The Gender Wage Gap
-‐ In Swedish Municipalities
Bachelor’s thesis within Economics Authors: Josefine Göthberg
Jonna Rickardsson Tutors: Charlotta Mellander
Jenny Ljungström Grek
Department: Economics, Finance and Statistics Jönköping Spring 2015
Bachelor’s thesis within Economics
Title: The Gender Wage Gap in Swedish Municipalities
Author: Josefine Göthberg, josefine.gothberg@gmail.com
Jonna Rickardsson, rickardssonjonna@gmail.com
Tutors: Charlotta Mellander
Jenny Ljungström Grek
Date: May 2015
Keywords: Gender wage gap, Wage gap, Regional economics, Gender inequality, Gender segregation, Swedish labor market JEL Classification: J16, J31, R10, J7
Abstract
Though successively decreasing over time, gender wage gaps are still large in all western countries. When gender wage gaps exist, there is an unequal distribution of economic power between men and women. This paper examines variables that significantly relate to the differences in the size of the gender wage gap across Swedish municipalities. With data gathered from Statistics Sweden and the Swedish Social Insurance Agency for the year 2011, a series of OLS regressions are performed. By examining what variables are statistically related to variations in the gender wage gap over municipalities, for example, average wage, human capital, gender segregation and work absence, the aim is to further contribute to the field of gender economics. The results in this paper show that the gender wage gap exists in all 290 Swedish municipalities. It varies greatly with women earning only 56 percent of men’s wages in Danderyd to women earning 87 percent of men’s wages in Haparanda. In municipalities where average wages are high the gender wage gap is large. Long-term illness and lowered capacity to work are strong factors negatively affecting the gender wage gap. In municipalities where women are more affected by long-term illness and lowered capacity to work than men the gender wage gap is larger. Furthermore, there is a significant relation between the gender wage gap and human capital. The gender wage gap is larger in municipalities where a large share of the population has a higher education.
TABLE OF CONTENTS 1. INTRODUCTION ... 1 1.1PURPOSE ... 2 1.2DELIMITATIONS ... 3 1.3DISPOSITION ... 3 2. LITERATURE REVIEW ... 4 2.1HUMAN CAPITAL ... 4
2.2GENDER SEGREGATED LABOR MARKETS ... 5
2.3WORKING TIME ... 6
Parental leave ... 6
Sick leave and work capacity ... 7
Part time work ... 7
2.4LOCATION AND THE GENDER WAGE GAP ... 8
3. HYPOTHESES AND EXPECTED RESULTS ... 9
4. METHODOLOGY ... 11 4.1EMPIRICAL MODEL ... 11 4.2VARIABLES ... 11 Dependent Variable ... 11 Independent variables ... 12 4.3ECONOMETRIC METHOD ... 14 5. EMPIRICAL RESULTS ... 15 5.1DESCRIPTIVE STATISTICS ... 15 5.2CORRELATION ANALYSIS ... 15 5.3REGRESSION ANALYSIS ... 16 6. ANALYSIS OF RESULTS ... 19 6.1IMPLICATIONS ... 22 7. CONCLUSION ... 23 8. BIBLIOGRAPHY ... 25 9. APPENDIX ... 28 9.1GENDER SEGREGATION ... 28
9.2 REGRESSION EQUATIONS WITH STANDARDIZED BETA VALUES ... 28
9.3CORRELATION MATRIX ... 29
FIGURES AND TABLES
Table 1. Variables, definitions and expected signs ... 9
Table 2. Municipalities with largest and smallest gender wage gap ... 11
Table 3. Descriptive statistics ... 15
Table 4. Wage gap correlations ... 16
Table 5. Regression equations ... 17
Table a: Gender segregation variables ... 28
Table b. Standardized beta values ... 28
Table c. Correlation matrix ... 29
Figure a. Standardized residual distribution ... 30
Figure b. Plot of Standardized residuals ... 30
1. INTRODUCTION
Around the world, issues of gender equality are very hot topics and gain a lot of media attention. In western countries gender quotas are being implemented and discussed, EU has directives regarding equal pay for equal work (European Commission, 2014) and family policies, such as parental pay, are restructured to promote gender equality (Statistics Sweden, 2014). However, though successively decreasing over time, gender wage gaps are still large. When gender wage gaps exist, the result is an unequal distribution of economic power between men and women, which in this paper refers to the situation when men or women have more economic resources at their immediate disposal. In essentially all developed countries most positions of authority are held by men. Men are highly over represented in CEO positions of larger companies as well as among other top positions, contributing to the unequal distribution of economic power in our society (Bihagen et al., 2013).
According to a report made by the UN, Sweden is among the highest ranked in the world when it comes to gender equality (measuring reproductive health, empowerment and labor market participation) (Human Development Report, 2014). But at the same time, Sweden has among the highest gender wage gap among European Union member countries (European Commission, 2014). The size of the gender wage gap, however, differs greatly between Swedish municipalities. The aim of this thesis is to examine what variables that significantly relate to the variations in the gender wage gap between Swedish municipalities.
Several factors that influence the gender wage gap have been identified, including working time, differences in human capital, gender segregation on the labor market, women's representation in trade unions and many more (Mahy et al., 2006). One reason for the large gender wage gap in Sweden is that, in relation to men, women tend to work fewer hours per week and fewer weeks per year (Blau & Kahn, 2000). For instance, in Sweden women work part-time to a greater extent than men (Statistics Sweden, 2014). Furthermore, even though an equality bonus for splitting parental leave more equally was implemented in Sweden in 2008 (Swedish Social Insurance Agency, 2014), the majority of parental leave is still utilized by women. In 2013, women stood for 75 percent of the parental leave, greatly affecting the hour’s women work and thus the size of the wage gap (Statistics Sweden, 2014). Other explanations for the differences in working time include that women spend more time at home with sick children and are affected by illness and long-term lowered capacity to work to a greater degree than men.
Differences in human capital are an interesting aspect of the gender wage gap. More women than men have a higher education, which is not reflected in the sizes of their wages. This contradicts the human capital theory, which suggests that investments in human capital result in higher earnings (Becker, 1964). Wage differences among educated women and men can however, to some degree be explained by gender segregated labor markets, which is the separation of women and men into different sectors and occupations. Throughout history, there has been a tendency for female-dominated sectors to pay lower wages (Blau & Kahn, 2000). According to Statistics Sweden, Sweden has had among the most gender segregated labor markets in the European Union and shows minimal changes in this area over time. In 1985, merely six percent of women and five percent of men worked in labor force sectors with
equal gender distribution. In 2014 these numbers were 14 percent for female workers and 13 percent for male workers (Statistics Sweden, 2014).
The size of the gender wage gap in Sweden fluctuates with level of education. The wage gap increases with higher educational levels implying that female workers in higher-educated jobs are exposed to a larger gender wage gap than less educated women (Albrech et al., 2003). The wages for men and women are more equal as they enter the workforce. Although, as workers gain experience and wages increase, women's wages fall behind. Furthermore, Swedish companies are no exception to the trend of male power dominance. Studies show that most CEOs of larger Swedish companies are men and that women’s advancement to higher positions and higher wages is slow (Bihagen et al., 2013). A phenomenon called the glass ceiling, which according to Albrecht et al. (2003) is highly present in the Swedish labor market. The wage gap, and thus the difference in economic power between the genders, differs in size depending on where in Sweden you live. According to research done by Cotter et al. (1996), industries are more gender segregated (female or male dominated) in less populated areas compared to urban areas. Furthermore a larger fraction of the population is employed within industries that typically are highly gender segregated in rural areas (Cotter et al., 1996). These factors indicate that the gender wage gap should be larger in less populated and more rural areas. However, in urban areas the educational level is much higher, which is linked to a larger gender wage gap (Statistics Sweden, 2014; Albrecht et al., 2003). Therefore, these two theories are contradictory and suggest different outcomes when comparing differences in regional gender wage gaps.
The main results from this paper show that the gender wage gap varies greatly between municipalities and is largest in the municipalities where average wages, human capital levels and population density are high. In these areas, women are more affected by lowered capacity to work than men, which is another strong factor determining the size of the gender wage gap. The statistically strongest variable in explaining the variation in the gender wage gap is average wage.
1.1 Purpose
The aim of this paper is to examine variables that significantly relate to the differences in the size of the gender wage gap across Swedish municipalities, and thus the difference in economic power between men and women.
There is not much research done comparing the gender wage gap between Swedish municipalities. Statistics Sweden provides countless information about wage differences between men and women at a national level, however not much information is available at a regional level. Thus, this paper contributes to the field of gender economics by examining what variables have a strong impact on the variations of the municipal gender wage gap for example, average wage, human capital, gender segregation and work absence.
This paper examines the municipal gender wage gap by looking at female wages as a percentage of male wages. With data gathered from the Statistics Sweden and the Swedish social insurance agencyfor the year 2011, a regression analysis is executed.
1.2 Delimitations
The limitations of this paper come from the lack of availability of some data. Data is not available on working time thus this is not directly included in the model in this paper. Although, several variables that do concern working time are included such as parental leave, sick leave, and lowered capacity to work. However, data is not available on type of employment. Variations in working time due to type of employment certainly affects the gender wage gap since women work part time to a greater extent than men (Statistics Sweden, 2014). This may lead to the model in this paper explaining a somewhat smaller part of the variations in the gender wage gap. Another limitation is that when computing gender segregation variables, a selection of eight gender segregated industries are examined and not all industries.
1.3 Disposition
This paper is organized as follows. Section 2 reviews the theoretical background and previous findings regarding determinants of the gender wage gap. Section 3 states the hypotheses tested in this paper. Section 4 presents the data, variables and empirical models and methods. The empirical results are presented in Section 5 and analyzed and discussed in section 6. Section 7 concludes the paper and provides suggestions for future research regarding the gender wage gap.
2. LITERATURE REVIEW
The European Union has directives regarding the principle of equal pay for equal work (European Commission, 2014). Although these directives exist, significant gender wage gaps are still evident in all member states (Mahy et al, 2006). Research shows that the gender wage gaps are decreasing in most European Union member countries, however only slowly (European Commission, 2014). Mahy et al., (2006) argue that, despite EU legislation, there is no “natural” trend towards equal pay. In Sweden, the gender wage gap has virtually remained unchanged since 1994 (Statistics Sweden, 2014). There are several factors influencing the gender wage gap. For instance, differences in human capital, gender segregated labor markets and working time (Mahy et al., 2006). This theory section will review existing theories regarding the gender wage gap.
2.1 Human capital
One of the most influential theories in explaining wage differences is the human capital theory (Becker, 1964; Mincer, 1958). The human capital theory by Mincer (1958) and Becker (1964), emphasizes that productivity is determined by the human capital of the worker. Earnings are equivalent to a worker’s return on human capital. Thus, people who invest more in themselves through education and labor market experience will earn higher wages (Becker, 1964). Mincer wage theory introduced one of the earlier equations1 and explanations on wage differences. The Mincer model explains differences in income by differences in experience and schooling, where it is assumed that all individuals have the same abilities and opportunities to enter any occupation. The model shows that as time goes on, and more skills and experience are attained, income increases. Different occupations require different training, and training reduces years of earnings (Mincer, 1958). The human capital theory emphasizes that experience and on-the-job training seem to have similar effects on earnings as education and other training (Becker, 1964). Women’s higher labor market absence due to for instance childbearing and more schooling, results in women having lower labor market experience than men (Mincer, 1958). However, according to De la Rica et al., (2008) women with higher education will participate in the labor force to a higher extent than women with lower education because of the higher investment in human capital that they have undertaken. Hence, the Mincer wage equation would suggest a larger gender wage gap between women and men with lower education levels. Some researchers, however, argue that this theory fails to explain the differences in wages between men and women satisfactorily (Schuld et al., 1994). The human capital theory and the Mincer equation do not consider what specific type of job workers hold, which is a major factor affecting the wage rate. In contrast, Thurow’s job-competition theory (Thurow, 1975) states that wages are based on the marginal productivity of the job (Thurow, 1975). This allows for differences in wages depending exclusively on the type of job a worker hold without taking into account the worker’s stock of human capital (Thurow, 1975). Hence, the
job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- job- 1 𝐿𝑛𝑦 = 𝑙𝑛𝑦
!+ 𝑟𝑆 + 𝐵!𝑋 + 𝐵!𝑋!
Where y is earnings, 𝑦0 is earnings of someone with no education and no experience, S is years of schooling and X is years of potential labor market experience.
competition theory takes the existence of gender segregated labor markets into account when explaining the wage gap. However, whether male-dominated jobs have higher marginal productivity per worker than female-dominated is not certain.
According to De la Rica et al. (2008), the size of the gender wage gap increases with higher levels of education and thus with higher wages. The gender wage gap is smaller for lower levels of education, and as the level of education increases, the wage gap between male and female workers also increases. This phenomenon varies between countries and sectors. In Sweden the wage gap between men and women at the top of the wage distribution is shown to be remarkably large compared to the wage gap for lower wages (Albrecht et al., 2003). Furthermore, gender wage differences are smaller when entering the workforce and in the early years of workers’ careers. As experience and wages increase, women's wages fall behind. This phenomenon is called the glass ceiling, and refers to the situation where women's wages increases up to a point where there is a limit on their wage prospects (De la Rica et al., 2008). Educational choices of women also affect the gender wage gap. It is argued that Swedish women make educational choices characterized by a low probability of reaching top positions at larger companies. Moreover, female-dominated occupations’ have a tendency to offer fewer possibilities for professional advancement, a phenomenon called sticky-floors (Bihagen et al., 2013).
Educational levels differ significantly between different locations. Women are overall more educated than men in the Swedish municipalities (Statistics Sweden, 2014). In 2012 rural municipalities had the largest difference in education between men and women, with women being much more educated than men (Statistics Sweden, 2013). This suggests that, since there is a wage gap in rural areas even though women are substantially more educated, it may at least be less prominent compared to urban areas where the genders are more equally educated. Furthermore, a larger share of the population in cities and urban areas has a higher education. Higher education results in a larger wage gap, therefore in municipalities where a higher share of the population has a higher education the gender wage gap is larger (Cotter et al., 1996; McLaughlin & Perman, 1991; Baum-snow & Pavan, 2011).
There are contradictory theories regarding human capital and the gender wage gap. The human capital theory suggests that a worker’s earning is represented by the stock of human capital. While women are more educated, men have more labor market experience and training because of female’s higher absence from the labor market. Additionally, since women spend more time in school - they “lose” years of labor market experience, which could be one reason for their higher education yielding lower returns.
2.2 Gender segregated labor markets
Women and men are not evenly represented in different sectors and industries. The segmented labor market theory divides the labor market into two different sectors, the primary sector and the secondary sector, with no crossover capability. The primary sector, which is a male dominated sector, includes the higher-status and better-paid jobs. The secondary sector, which is considered a female dominated sector, is characterized by low-skilled jobs that require little training. The jobs in this sector are considered unattractive, wages are low and job turnover is high (Doeringer & Piore,
1971). This division of the labor market can lead to different groups, like men and women, receiving different wages.
Gender segregated labor markets are pointed out as a major cause of the gender wage gap. Throughout history, one of the most persisting characteristics of women's status in the labor market has been their tendency to work in lower paying, female dominated jobs, greatly affecting their wages and the gender wage gap (Blau et al., 2000; Cotter et al., 1996). There are several explanations for labor markets being highly segregated. It is argued that gender segregation on the labor market could, in some ways, be explained by the fact that women choose occupations where the cost of career interruptions is lower (De la Rica et al., 2008). A study conducted by Bygren and Kumlin (2005) shows that the different educational choices of men and women significantly affect the occupational gender segregation in Sweden. Another explanation for the existence of gender segregated labor markets is that organizations that already employ a high share of a certain gender will continue to hire persons of the same gender to a greater extent, leading to further labor market segregation and an increase of the gender wage gap (Bygren & Kumlin, 2005).
There are several aspects of gender segregated labor markets. For instance, gender segregation within industries i.e. how female or male dominated industries are and, to what extent the population is employed within these gender segregated industries. A higher level of gender segregation within industries, meaning a more extreme gender composition where women are overrepresented in female-dominated sectors and men overrepresented in male-dominated sectors, increases the gender wage gap. An occupational structure, where a higher share of the population is working in industries that are generally more gender segregated, for example industrial industries, transportation and farming, also increases the gender wage gap (Cotter et al., 1996). Therefore, where labor markets are more gender segregated in either one or two of the aspects, the gender wage gap is larger. According to Cotter et al. (1996), the gender composition within the different industries in the US accounts for 35 percent and the difference in occupational structure (whether the local labor market consists to a high extent of gender segregated industries) accounts for 65 percent of the gender wage gap differences between urban and rural areas (Cotter et al., 1996).
2.3 Working time
Women tend to work fewer hours per week and fewer weeks per year than men (Blau et al., 2000). Selmi (1999) suggests that this disproportionate distribution of working time between men and women explains a significant portion of the gender wage gap. Therefore, policies creating strong incentives for employers and families to change their behavior are essential for reaching more equal work patterns between men and women. Long absences from the labor market due to parental leave or long-term sick leave, and working part time reduce women’s earnings size directly and indirectly by lowering their work experience and employment stability (Mandel & Semyonov, 2005).
Parental leave
Women’s absence from the labor market when having children affects their wages. The expectation by employers that women will be absent due to childbearing also affects women's wages (Selmi, 1999). According to Selmi (1999) women are exposed
to a statistical discrimination, because as a group, they exit or are absent from the labor market to a higher extent than men. This results in all women being discriminated, even those who do not follow this anticipated female work-path. Thus, women without children or those who have limited interruptions from the labor market when having children are still not able to compete for jobs and higher wages on the same conditions as men.
Family policies are structured to facilitate women's employment and protect their rights in the labor market. The wage gap is somewhat smaller when these policies are more developed (Mandet & Semyonov, 2005). Swedish policies regarding the interaction between family and work, for example the parental leave policy and the daycare system, give Swedish women strong incentives to participate in the labor force. This should thus decrease the gender wage gap (Albrecht et al., 2003). Other policies in Sweden that are implemented to increase equality are for example the equality bonus, where a bonus is paid out to parents when the parent with the least utilized parental leave days is on parental leave (The Swedish Social Insurance Agency, 2014). However, the majority of parental leave days in Sweden are still utilized by women. In 2013, women stood for 75 per cent of the parental leave, greatly affecting the number of hour’s women work and thus the wage gap and economic power between the genders (Statistics Sweden, 2014).
Sick leave and work capacity
Being absent due to own illness does not only affect the income while being ill, but could also signal less commitment to the job and therefore affect wages and job opportunities more permanently (Hansen, 2000). Women are on average more absent from work than men and will therefore accumulate less work experience and obtain lower wages (Albrecht et al., 1999). In Sweden women are overrepresented in sick leave exceeding 14 days in all sectors. Approximately two thirds of all sick leave days are accounted for by women. Nine percent of women’s sick leave is due to pregnancy related causes, however, correcting for these, women are still overrepresented, leading to lower participation and fewer hours worked and thus a larger gender wage gap (Statistics Sweden, 2011).
Part time work
Differences in working time are highly affected by the type of employment. For example, in Sweden women work part-time to a greater extent than men (Statistics Sweden, 2014). In 1987, 45 percent of Swedish women worked part-time but only six percent of the men. In 2013 these numbers were 30 percent of the women and 11 percent of men showing that women are still the majority in working part-time (Statistics Sweden, 2014). The unequal proportions of male and female part-time workers could be explained by several different factors. For example, women are being overrepresented in sectors where part-time employment is more common because of limited full time positions available. Furthermore women might choose to reduce working hours when having children, they might also experience expectations to reduce working hours. This affects their working time and therefore their wages (Statistics Sweden, 2011).
2.4 Location and the gender wage gap
Research confirms that workers earn more in larger cities and urban areas than in rural areas, and that is the case for both women and men (Andersson et a., 2014; Cotter et al., 1996; McLaughlin & Perman, 1991; Baum-snow & Pavan, 2011). There are several possible explanations for this difference in wages but differences in human capital levels are one major cause of the city size wage gap. Higher-wage social analytical skills are concentrated in more urban areas, these regions have been found to have distinct advantages in attracting high-skilled people. In contrast, lower-wage physical skills are concentrated in rural areas (Florida & Mellander, 2014). Firms in urban areas could therefore be more productive and high skilled. Since firm productivity is related to wages, and workers and firms are more productive and efficient in larger cities and urban areas, wages are higher in these areas (Andersson et a., 2014; Baum-Snow & Pavan, 2011).
The higher share of educated people and top positions in urban areas results in a larger gender wage gap in larger cities and urban areas. Men and women are more equally educated in urban areas compared to non-urban areas, where men are less educated than women and also less educated compared to men in urban areas. According to De La Rica et al. (2008), women with higher education are less absent from the labor market, and due to the proportion of higher educated women being larger in urban areas, this may suggest that urban-women and men share the family-related burden more equally and therefore positively affect the gender wage gap. Furthermore, gender segregation on the labor market also results in a larger gender wage gap. Labor markets are significantly more gender segregated in rural areas compared to urban areas. Industries in less populated areas have a more extreme gender composition where women are overrepresented in female-dominated sectors and men overrepresented in male-dominated sectors. This is the situation for both the educated and non-educated population (Cotter et al., 1996; Glauber & Smith, 2013). There are also different occupational structures in rural areas compared to larger cities and urban areas. A higher share of the rural population is working in industries that are generally more gender segregated, for example industrial industries, transportation and health care. In comparison, labor markets and occupations in urban areas are more gender integrated and more industries are represented compared to rural areas (Glauber & Smith, 2013). Since gender segregated labor markets increases the gender wage gap, the gender wage gap should be large in rural areas (Cotter et al., 1996).
3. HYPOTHESES AND EXPECTED RESULTS
The theories presented in this theory section provide somewhat different predictions of where the gender wage gap is largest. Based on the theoretical framework the hypotheses are formulated to examine what factors are strongest in affecting the wage gap and thus where the wage gap is largest. Table 1 summarizes the expected signs on variables based on theory.
Table 1. Variables, definitions and expected signs
Variable Definition Expected sign:
women wage men wage
Average Wage Average wage for FA-regions -
Population density Log population density in the FA-region - Human capital
population share
The share of municipal population with three years or more of higher education
- Human capital gender
ratio The proportion, women divided by men, with three years or more of higher education + Gender segregation
within industries
The degree of gender segregation within the eight chosen sectors in the municipalities (higher value - more segregation)
- Gender segregation
share of labor market
The share of the municipality population that is employed within the eight chosen most gender segregated industries
-
Parental leave Women’s share of parental pay -
Sick leave, gender ratio
The proportion, women divided by men, of sick leave. - Sick leave, population
share
The total mean value of sick leave for the municipalities. - Work capacity gender
ratio
The proportion, women divided by men, with lowered capacity to work. -
According to theory, the gender wage gap increases with level of education. This phenomenon suggests the gender wage gap to be larger where education levels are higher (ceteris paribus) or where a larger proportion of the municipal population has a higher education of three years or more. For simplicity, from here on education is equated with human capital.
𝐻!: The gender wage gap is larger in municipalities with higher shares of human capital.
According to theory, the gender wage gap is less significant in lower-paid jobs and sectors and in the early years of workers’ careers. But as experience, education and wages increase, women's wages fall behind. The presence of the glass-ceiling phenomenon further increases the gender wage gap at the top of the wage distribution. This suggests the gender wage gap to be larger in municipalities where average wage is higher (ceteris paribus).
𝐻!: The gender wage gap is larger in municipalities where the average wage is higher.
The gender segmented labor market, where men and women work in different jobs and sectors, is according to theory a major reason behind the gender wage gap. One of
the most persisting characteristics of women's status in the labor force has been their tendency to work in lower paying, female dominated jobs. This suggests the gender wage gap to be larger in municipalities where the labor market is more gender segregated (ceteris paribus).
𝐻!: The gender wage gap is larger in municipalities where labor markets are more gender segregated.
Differences in working time are discussed as a main cause of the gender wage gap. The tendency of women as a group to work fewer paid hours per week and fewer paid weeks per year than men, reduces women’s direct and future earnings, lowering their work experience and their employment stability. This suggests the gender wage gap to be larger in municipalities where women are more absent from the labor market in relation to men (ceteris paribus).
𝐻!: The gender wage gap is larger in municipalities where women in comparison to men are more absent from the labor market.
4. METHODOLOGY
In this section, the empirical model, variables, and econometric method are presented.
4.1 Empirical Model
The aim of this paper is to examine what factors and variables that significantly relate to the variation of the size of the gender wage gap between municipalities. Based on the theoretical framework and previous research, as well as inspired by the Mincer equation, the following model is formulated:
Female average wage/ Male average wage =𝐵!+ 𝐵! (Average wage) + 𝐵!𝐿𝑛
(Population density) + 𝐵! (Human capital) + 𝐵! (Gender segregation) + 𝐵! (Parental leave) + 𝐵! (Sick leave) + 𝐵! (Work capacity)+ 𝜀
4.2 Variables
Dependent Variable
The dependent variable Gender wage gap is expressed as women’s earnings as a proportion of men's earnings in the municipality where they live. The variable is calculated by dividing women’s total average wage over men’s total average wage for each Swedish municipality. The data is gathered from Statistics Sweden (2015) and covers the entire Swedish population that had a wage income in 2011. The variable varies greatly between the municipalities, the extremes showed in table 2. A lower value indicates that women’s wages as a share of men’s wages is lower and that the gender wage gap is larger.
Table 2. Municipalities with largest and smallest gender wage gap
Municipality Female average wage* Male average wage* Womens share of mens wages Municipality Female average wage* Male average wage* Womens share of mens wages Danderyd 29591 53117 0.56 Haparanda 16395 18881 0.87 Lidingö 25456 42367 0.6 Bräcke 16280 18844 0.86 Vellinge 20666 32196 0.64 Berg 15344 18045 0.85 Lomma 22951 35832 0.64 Överkalix 16414 19419 0.85 Täby 24954 38077 0.66 Övertorneå 15767 18551 0.85 Vaxholm 22688 33990 0.67 Gotland 16723 19915 0.84 Kungsbacka 20295 30298 0.67 Strömsund 15894 18970 0.84 Kävlinge 19310 28232 0.68 Borgholm 15757 18891 0.83 Perstorp 16007 23555 0.68 Åre 16321 19572 0.83 Öckerö 18530 27324 0.68 Sorsele 15231 18415 0.83 * Monthly wages in SEK
N=290
Source: Statistics Sweden 2015
Independent variables
Based on what previous research found to affect the gender wage gap, this model includes the following independent variables. All data is for year 2011.
Average wage
Calculated as total wage income for the municipality divided by the municipal population with a wage income. Therefore, this variable illustrates the average wage per worker in the municipality. The variable is expressed in 1000s SEK.
Population density
The data on population density is logged and measures population per square kilometer. The data is gathered from Statistics Sweden and describes population density for Functional analysis regions2. FA regions are chosen instead of
municipalities since the municipal population often works outside the municipality where they live.The inclusion of a population density variable is to measure whether the municipality is an urban area or rural area. Theory suggests that there are large differences in wages (and the gender wage gap) between urban and more rural areas.
Human capital variables
The human capital variables are based on data on educational levels for women and men. The data is gathered from Statistics Sweden (2015).
Human capital population share
This variable is calculated by dividing the municipal population age 25 to 64 with a higher education by the total municipal population age 25 to 64. Therefore, this variable shows the share of the municipal population with three years or more of higher education, where a higher value implies that a larger proportion of the municipality population has a higher education.
Human capital gender ratio
This variable is the share of women with higher education divided by the share of men with higher education per municipality. The higher the value, the more educated are women in relation to men in the municipality.
Gender segregation variables
To be able to measure the degree of gender segregation in the municipalities, two new variables are computed using data on employment from Statistics Sweden (2015). The data divides the Swedish labor market into 50 industries. The four most female dominated - and the four most male dominated industries, employing at least one per cent of the total national labor force are included (Table a, Appendix). The one percent threshold level is to eliminate the effect of very small industries, which are often location bound. These variables are calculated on both FA-regional level and municipality level. Measuring for FA-regions might be more in accordance with theory to reflect that people do not always work in the municipality where they live.
2 FA-regions, functional analysis regions is a regional division based on commuting statistics from
2003 and trends in commuting patterns, capturing workers movements. Each FA-region includes one or more municipalities.
Gender segregation, within industries
This variable is computed by calculating the mean value of the proportions of men in the male-dominated industries and the proportions of women in the female-dominated industries per municipality and FA-region. This variable explains to what extent the eight chosen industries are gender segregated (female or male dominated). A higher value indicates that industries are more female or male dominated and thus more gender segregated.
Gender segregation share of labor market
This variable is calculated by dividing the share of the municipal or FA population employed within the eight chosen industries by the total workforce in the municipality or the FA-region. A higher value of this variable indicates that a larger share of the population is employed within these highly female- and male-dominated sectors, suggesting that the municipal or FA labor market is more gender segregated.
Work absence variables
Four different variables measuring municipal variations in absence from the labor market due to short term sick leave or long term lowered capacity to work are included in the model. To measure variations between municipalities in transitory sick leave we introduced two sick leave variables, which are both based on the Swedish measure of sickness and rehabilitation days in the municipalities (Swe: “nya sjukpenning talet”)3. The other two variables aim to show variations between the municipalities in work capacity levels. The variables are based on a measure of the amount of people with a permanent or temporary lowered capacity to work in each month and municipality (Swe: “SA beståndet”)4. The data is gathered from the Swedish Social Insurance Agency (2015) for all variables.
Sick leave gender ratio
This variable is calculated by dividing women’s value of the sick leave measure by men’s value. This variable explains how affected by short term illness women are in comparison to men in the municipality. The higher the value, the more affected by illness are women, in comparison to men in the municipality.
Sick leave population share
This variable is calculated by dividing the share of the population that was sick by the municipal population. It explains how affected by illness the municipality is in general, and is the value of the sick leave measure for men and women. The higher the value, the larger the proportion of the municipality population has suffered from illness.
Work capacity gender ratio
This variable is calculated from the work capacity measure by dividing the share of women with lowered capacity to work by the share of men with lowered capacity to work in the municipality. The higher the value, the more affected by lowered capacity to work are women in comparison to men.
3 The measure of Sickness and rehabilitation days is defined as the number of days of sick leave and
rehabilitation days in a month, divided by the number of insured minus the number of people with no capacity to work. Only people age 16 to 64 are included in the measure.
4 Explains the number of people with a permanent or temporary lowered capacity to work. The
Work capacity population share
This variable is calculated by dividing the total amount of people with lowered capacity to work by the working age population in each municipality. It explains how affected the municipality is in by lowered capacity work. A higher value indicates that a larger proportion of the municipal population has a lowered capacity to work.
Parental leave
The Parental leave variable is the share of the parental pay that is utilized by women in the municipality. The data is gathered from the Swedish Social Insurance Agency (2015). The higher the value, the larger is the proportion of the parental leave that is utilized by the mother.
Urban variable
Using a principal component analysis the variables Population density and Average
wage are combined into one variable. This to correct for multicollinearity problems.
Population density and average wages are most often higher in urban areas, hence we call this combined variable for Urban variable since it captures if the municipality is located in an urban or area or not.
4.3 Econometric Method
The gender wage gap is examined over Swedish municipalities for the year 2011, using a regression analysis. When plotting the data, all variables show a linear relationship with the dependent variable (the variable Population density is logged). Therefore, an ordinary least square regression analysis is suitable.
Multicollinearity in the model results in biased results. In this model, some of the independent variables are correlated with each other and thus show higher VIF values (Table 5) and Pearson correlation coefficients (Table c, Appendix). Correlations are discussed further in the next section and a principal component analysis is used to work out parts of the problems. Furthermore, several regressions are run to detect and avoid the existing multicollinearity.
A residual analysis is conducted to examine whether the residuals in the model will affect the validity of the results. No problems with heteroscedasticity are found. The residuals are normally distributed with a mean value of zero. Furthermore, the correlations between all independent variables and residuals are zero. When examining all independent variables the regression suits the data well, since only eleven municipalities have standardized residuals with a value above two (Table d, Appendix).
5. EMPIRICAL RESULTS
In this section, the findings of this paper will be presented. First, descriptive statistics and correlations are discussed, than the outcomes of the regressions are presented.
5.1 Descriptive statistics
The gender wage gap exists in all 290 Swedish municipalities. However, it is not evenly distributed among the municipalities, but varies greatly with women earning only 56 per cent of men’s wages in Danderyd to women earning 87 percent of men’s wages in Haparanda. These differences can to some degree be explained by variations in the independent variables. Table 3 provides descriptive statistics for the dependent and independent variables.
Table 3. Descriptive statistics
Variables Minimum Maximum Mean SD
Wage gap 0.56 0.87 0.74 0.04
Average wage 198.65 494.25 246.22 32.05
Population density 0.20 677.58 139.18 217.59
Human capital population share 0.083 0.547 0.17 0.72
Human capital gender ratio 0.90 2.96 1.73 0.38
Gender segregation share of labor market 0.32 0.58 0.43 0.43 Gender segregation within industries 0.78 0.91 0.86 0.02
Parental leave 0.68 0.87 0.77 0.03
Sick leave gender ratio 0.96 2.51 1.62 0.24
Sick leave population share 3.76 11.63 7.08 1.17
Work capacity gender ratio 0.85 2.16 1.48 0.20
Work capacity population share 0.04 0.17 0.10 0.02
Urban variable* -2.37 5.35 0.00 1.00
* Principal component variable of average wage and population density. N=290
Source: Statistics Sweden 2015, Swedish Social Insurance Agency 2015.
5.2 Correlation Analysis
In Table 4 the bivariate correlations between the dependent variable and the independent variables are illustrated. All wage gap correlations are significant at a 5 percent level (Table 4). All wage gap correlations except for the Gender segregation
within industries5 variable are significant at a 1 percent level. The bivariate
correlations between the Gender wage gap and Gender segregation share of labor
market6 and Parental leave show conflicting signs based on theory. The correlation
between Average wage and the Gender wage gap is -0.627, which is the highest of all bivariate correlations with the dependent variable. The correlation between Work
capacity gender ratio and Gender wage gap (-0.543) is the second highest wage gap
correlation.
5 The variable Gender segregation within industries is based on municipalities instead of FA-regions
due to better correlation values.
6 The Gender segregation share of labor market variable is based on FA-regions instead of
Table 4. Wage gap correlations7 Variable Wage gap: Women wage Men wage Average wage -0.627** Population density -0.414**
Human capital population share -0.314**
Human capital gender ratio 0.303**
Gender segregation share of labor market 0.355**a Gender segregation within industries -0.104*
Parental leave 0.248**a
Sick leave gender ratio -0.257**
Sick leave population share 0.271**
Work capacity gender ratio -0.543**
Work capacity population share 0.433**
Urban variable -0.587**
** Significant at 1% level. * Significant at 5% level. a, variable show conflicting sign. N=290
Source: Statistics Sweden 2015, Swedish Social Insurance Agency 2015.
Several of the independent variables are correlated, for example Population density is higher in cities and urban areas, in these areas education levels (Human capital
population share) are higher and Average wages are higher. Also, in urban areas,
work absence (Sick leave population share, Work capacity population share etc.) is lower and labor markets are less gender segregated (Table c, Appendix). The Human
capital population share variable is one of the variables that are most correlated with
others, for example; Average wage (0.771), Work capacity population share (-0.693),
Parental leave (-0.507), Sick leave population share (-0.496), Gender segregation within industries (-0.492) and Population density (0.448). The variable Average wage
is also correlated with many of the independent variables (Table c, Appendix). Moreover, the variable Human capital population share is highly correlated with the
Human capital gender ratio (-0.738). The two Human capital variables are equally
correlated with the Gender wage gap, but their correlations are contrariwise.
Other variables were tested and found not to be significant or did not show strong statistical relationship with the gender wage gap and were therefore excluded from the model. Municipal population is an example, the variable did not show a distinct relationship with the dependent variable. Also the proportion of female versus male residents in the municipalities showed no direct relationship with the gender wage gap. Neither did the size of the proportion of the municipal population in working age (25 to 65 years) to the whole population have any significant correlation with the dependent variable.
5.3 Regression Analysis
To examine to what extent the variables are related to the municipal variations in the gender wage gap OLS regressions are used. Based on the F-statistics, all regression equations are significant at a 1 percent level. The results from the regressions are summarized in Table 5.
Table 5. Regression equations
Variables Eq. (1) VIF Eq. (2) VIF Eq. (3) VIF Eq. (4) VIF
Constant 1.416** 1.294** 1.464** 1.218** (14.40) (14.76) (20.91) (15.22) Average wage -0.001** 3.539 - -0.001** 2.678 - (-10.93) (-14.05) Population density -0.001 2.276 -0.005** 1.408 - - (-0.39) (-3.94)
Human capital population share 0.149**a 4.812 -0.088* 2.212 - 0.044a 2.812
(3.34) (-2.42) (1.16)
Human capital gender ratio -0.003a 3.013 - - - (-0.37)
Gender segregation share of labor market -0.019 1.719 - - - (-0.43)
Gender segregation within industries -0.376** 1.485 -0.475** 1.378 -0.516** 1.379 -0.455** 1.375 (-4.37) (-4.76) (-6.69) (-5.00)
Parental leave -0.051 1.564 - - -
(-0.80)
Sick leave gender ratio -0.008 1.334 -0.019* 1.279 - -0.017* 1.267
(-1.17) (-2.28) (-2.29)
Sick leave population share 0.002 1.586 0.004* 1.405 - 0.003 1.415
(1.21) (2.31) (1.88)
Work capacity gender ratio -0.064** 1.403 -0.073** 1.251 -0.065** 1.160 -0.059** 1.266 (-7.35) (-7.37) (-7.97) (-6.48)
Work capacity population share 0.34 2.726 - - - (0.30)
Urban variable - - - -0.021** 2.183
(-8.76)
R2 0.613 0.428 0.585 0.525
adj R2 0.598 0.416 0.581 0.515
t-values within brackets
** Significant at 1% level. * Significant at 5% level. a variable show conflicting sign
N=290 for all variables
Source: Statistics Sweden 2015, Swedish Social Insurance Agency 2015.
Equation 1 examines the Gender wage gap including all independent variables. When all variables are included, the model has an adjusted R2 of 0.598. However, as can be observed in Table 5, in this equation several variables are insignificant. The only variables that are significant at a 1 percent level are Average wage, Human capital
population share, Work capacity population share and Gender segregation within
industries. The variables Average wage, Human capital population share, Human
capital gender ratio and Population density all show higher VIF values. Thus,
multicollinearity seems to be present, which affects the significance of variables in the model and may lead to biased results. In municipalities where the Gender wage gap is higher, there is a larger proportion of educated people, and a smaller ratio of educated women to men. When including both Human capital variables, their contrary effects8
result in at least one of the variables becoming insignificant. Although Human capital
population share is significant at a 1 percent level (Table 5, equation 1), the
conflicting sign makes it hard to interpret. Based on the standardized beta-values (Table b, Appendix), Average wage is the strongest variable in the regression followed by Work capacity gender ratio. Indicating that in municipalities where average wages are higher the gender wage gap is larger and when women are more affected by lowered capacity to work than men, the wage gap increases.
8 When combining the two Human capital variables in a principal component analysis, the component
Equation 2 excludes some of the insignificant variables from the previous equation9.
Average wage, which is significant in equation 1 is also excluded since it shows a
high VIF value and might be the reason for other variables becoming insignificant or showing conflicting signs in equation 1. In this equation all variables are significant at a 5 percent level and four variables are significant at a 1 percent level. All variables show the expected signs. The adjusted R2 is 0.416, which is lower than in the previous equation. It is the lowest adjusted R2 among the equations due to the exclusion of Average wage. The VIF values are lower compared to equation 1. No VIF values are above 2.5 and most are under 1.5 thus there are no problems with multicollinearity in this equation compared to the previous equation. The sick leave variables are significant at a 5 percent level however they are the weakest variables in the equations (Table b, Appendix). Indicating that when women are more affected by short-term illness the gender wage gap only increases slightly. Furthermore when the municipality is more affected by short-term illness the gender wage gap only increases marginally. Work capacity gender ratio is the strongest variable in the equation, followed by Gender segregation within industries and Population density. Therefore, in municipalities where industries are more male or female dominated the wage gap increases, and that the gender wage gap is larger in more densely populated FA-regions.
Equation 3 is a linear stepwise regression and includes three of the independent variables. The equation has an adjusted R2 of 0.581. All variables are significant at a 1 percent level and show the expected signs. The VIF values are low so no multicollinearity problems are present in the model. As in equation 1, Average wage is the by far strongest variable in the equation (Table b, Appendix). This is also the variable that is most correlated with the wage gap (Table 4).
Equation 4 includes the principal component variable, Urban variable. This variable combines the Average wage and Population density to avoid multicollinearity in the equation since these variables are highly correlated, (0.571) (Table c, Appendix). The model has an adjusted R2 of 0.521. The variables Human capital population share is insignificant and show conflicting sign. The VIF values are all under 3, the highest being 2.812 for Human capital population share. Furthermore Sick leave population
share is insignificant. The strongest variable in the equation is the Urban variable
capturing whether the municipality is located in an urban area or rural area and indicating that the gender wage gap is larger in urban areas.
The variables Parental leave, Gender segregation share of labor market and Work
capacity population share are not significant in any regression.
9 The excluded variables are those which repeatedly, when tested in many different regressions, shown
to be insignificant.
6. ANALYSIS OF RESULTS
In this section, the hypotheses are answered and the findings of this paper are analyzed and discussed.
Human capital
The first hypothesis states that the gender wage gap is larger in municipalities where human capital levels are higher. The results in this paper are consistent with theory which states that the wage gap increases significantly with level of education (De la Rica et al, 2008). Based on the regressions a significant relationship between human capital levels and the gender wage gap is evident; the gender wage gap is largest in municipalities with higher shares of educated people (Human capital population
share). The municipalities Danderyd, Lomma, Lidingö and Täby all have among the
highest gender wage gap (Table 2) and the highest share of educated people. Other municipalities with high share of educated people are Lund, Umeå and Uppsala. These municipalities have major student cities, however, the gender wage gap in these municipalities are not significantly higher in proportion to other municipalities. In the regressions, the Human capital population share variable is only significant and shows the expected sign in equation two. This could be because the variable is highly correlated with Average wage (0.771) (Table c, Appendix), which has a stronger relation to the Gender wage gap. The Human capital gender ratio variable is above one in all but two municipalities, indicating that more women than men have a higher education in almost all 290 Swedish municipalities. In municipalities where women are much more educated in comparison to men, the Gender wage gap is smaller. The statistical indications of the two human capital variables are contradictory. In municipalities with a higher share of people with higher education, women and men are more equally educated. Several of the municipalities with the highest proportion of educated people, are also among the municipalities with lowest Human capital
gender ratio (Danderyd, Lidingö, Täby, Solna, Lund, Stockholm, Nacka). This
implies that in the most highly educated municipalities, men are more educated compared to other municipalities, and the proportions of women and men that are educated are more equal. Therefore, another explanation for why the gender wage gap is larger in more “educated” municipalities is the larger proportion of men with a higher education. With higher education comes higher wages, thus with a larger proportion of men earning higher wages compared to men elsewhere, the result is an increasing wage gap. Although the gender wage gap exists in municipalities where almost three times as many women than men have a higher education, it is more prominent where the proportions of the genders are more equal. For instance, Danderyd is the municipality with the lowest value of the Human capital gender ratio at 90 percent, hence, men are more educated than women. Furthermore, Danderyd has the largest Human capital population share with 55 percent of the population having a higher education. Additionally, Danderyd has the largest gender wage gap with women earning only 56 percent of men's wages (Table 2).
The findings in this paper confirm the first hypothesis and finds that the wage gap increases with level of education and foremost with the increase of men with a higher education.
Average wage
The second hypothesis states that the gender wage gap is larger in municipalities where the average wage is higher. The results from the regressions show that in municipalities where average wages are higher the gender wage gap is much larger. This is the situation in municipalities where both women and men have higher average wage incomes. Therefore, where women earn more in relation to other women, men earn even more in relation to other men, creating an even larger gender wage gap. This is consistent with our hypothesis and the theoretical framework. According to Albrecht et al. (2003), the gender wage gap at the top of the wage distribution is shown to be remarkably large compared to for lower wages. Average
wage is by far the strongest variable in the regressions where it is included (Table 5)
and it also has the strongest bivariate correlation with the Gender wage gap (-0.627), (Table 4). Several of the municipalities, which have among the highest average wages, also have among the largest gender wage gaps (Danderyd, Lindingö, Täby, Lomma, Vaxholm, Vellinge). Furthermore, well-paid top work positions are more common in larger, more educated regions and cities where average wages are higher, and according to theory these are to a higher extent held by men, also increasing the wage gap in these locations (De la Rica et al, 2008; Bihagen et al, 2013).
Another important occurrence to discuss is that wages are often earned in other municipalities than where the workers live. Residential segregation results in people with different wage income, although maybe earned in the same municipality, to locate in different municipalities. Commuting municipalities, which have high average wages and a larger share of population with a higher education, and are located nearby cities with high population density, have the largest gender wage gap. This is the case for the municipalities Kungsbacka, Danderyd, Lidingö, Täby, Vaxholm, which are high-income locations. In these municipalities, the wages are often earned in other municipalities or urban areas for example Stockholm or Göteborg. Other municipalities nearby densely populated areas are for instance, Botkyrka and Södertälje, in which residents are likely to work in Stockholm, however, these municipalities have lower average wages, a less educated population and have a less prominent gender wage gap.
The results from the regressions confirm the hypothesis and find that in municipalities where average wages are higher the gender wage gap is much larger.
Gender segregation on the labor market
The third hypothesis states that the gender wage gap is larger in municipalities where labor markets are more gender segregated. Theory suggests that labor markets are more gender segregated in non-urban areas compared to urban areas (Cotter et al., 1996; Glauber & Smith, 2013) and the gender wage gap should thus be large in non-urban areas. The findings regarding gender segregation are somewhat inconclusive. The variable Gender segregation share of labor market illustrates the share of the population in the FA-region that works within the eight most gender segregated industries. These industries employ about 45 percent of the Swedish population. The size of the FA population proportion, however, varies from 30 to 58 percent (Table 3). This variable is not significant in the regressions. There are several possible explanations for this; one being that other contradictory variables are stronger in affecting the wage gap. For example Human capital population share and Average
segregation on the labor markets. The municipalities with the largest share of the population working within these industries are among the municipalities with the lowest average wages. For example Årsele, Vilhelmina, Sorsele and Övertårneo are all among the municipalities with highest share of population employed within these eight industries as well as among the municipalities with lowest average wage. Municipalities with higher average wages have a larger wage gap, and since the gender segregation on the labor market variable is not as strong, the effects on the gender wage gap is not clear.
The variable Gender segregation within industries illustrates how female or male dominated these eight industries are in the municipalities. This variable is significant in all regressions at a 1 percent level. Indicating that in municipalities where the industries are more female or male dominated the wage gap is larger. A clear relation between municipalities where industries are highly gender segregated and wage gaps are large is evident when looking at smaller municipalities generally specialized in gender segregated industries. After municipalities with high average wage and high educational levels, the municipalities with gender segregated industries have among the highest wage gaps. For instance municipalities with industries such as mining (Gällivare and Kiruna), paper mills (Hammarö) and nuclear plants (Östhammar) all have among the largest gender wage gaps. These municipalities also have lower educational levels and the municipalities show a higher absence from work with a higher share of the population having a permanent lowered capacity to work.
Thus the findings confirm the hypothesis that the gender wage gap is larger, all else equal, where labor markets are more gender segregated, especially in small industrial municipalities. However, due to other factors that are stronger in explaining the wage gap, the findings are somewhat inconclusive.
Work absence
The fourth hypothesis states that the gender wage gap is larger in municipalities where women in comparison to men are more absent from the labor market. According to theory women's absence from the labor market is a strong factor in explaining the gender wage gap (Mandel & Semyonov, 2005). Thus when women are more absent from the labor market, the wage gap increases. Women are affected by temporary, short-term illness, to a higher extent than men in all but one municipality. They are also more affected by long-term illness and lowered capacity to work than men, in all municipalities. The magnitude of the differences in illness between the genders varies between municipalities (Table 3).
A higher value of sick leave gender ratio should according to theory indicate a larger wage gap since women are more absent than men. This is consistent with the findings in this paper where the variable is significant at a five percent level when included. Several of the municipalities with the smallest difference in absence between men and women are also among the municipalities with the smallest wage gap, for example Överkalix. Moreover, Lidingö and Vaxholm have among the highest sick leave
gender ratio and gender wage gap. The variable sick leave population share is
significant in equation 2 and 4 (Table 5) indicating that when municipalities are more affected by short term illness the wage gap decreases.
A higher value of the work capacity gender ratio variable indicates that women to a higher extent than men have a lower capacity to work, and should thus lead to a larger
