Occupational sex composition and wages in the Faroe Islands

(1)

Department of Sociology

Master thesis in Sociology, 30 ects Spring 2017

Supervisor: Magnus Nermo

Occupational sex composition and wages in the Faroe Islands

Maria Hansen

(2)

Abstract

This study explores the relationship between wages, gender and occupational sex composition. The study is based on censored 2011 census data on the employed labour force in ages 25-69 in the Faroe Islands; a small Nordic country. The aim of this study is to contribute to the consisting overall knowledge of the relationship, as well as to the specific Faroese context and its lacking labour market studies. The analysis is divided in to two parts.

Firstly, multiple regression analysis is applied to the whole labour market examining: 1) if a gender wage gap is present, and 2) the shape of the relationship between wages and occupational sex composition. Secondly, similar multiple regression analyses are applied to the genders separately to examine: 3) to what extent the associations between wage, sectors and occupational sex composition vary by gender. This is done in the overall labour force as well as in part-time only. Main findings demonstrate that some of the overall associations are consistent with previous findings in other contexts, such as the presence of the gender wage gap, and for women, wages have a partially decreasing correlation with an increasing proportion of women. However, unlike previous findings, the male wages are highest in the least sex-integrated compositions and they have no association with sector. Sub analyses show that for women, public sector has a positive association with wages, especially for part-time workers where it also is positive for men. The paper concludes on a small discussion on future research and contextual factors.

Key words

gender pay gap, occupational gender segregation, occupational sex composition, women's wages

(3)

Introduction

It is well established that the gender wage gap in most European labour markets has decreased since the 1960s, but still exists (Blau and Kahn, 2000; Eurostat, 2015). An interesting feature is that the gender segregation by occupation in the labour market is recognized as being related to the gender wage gap (e.g. Kilbourne, England and Beron, 1994; Cohen and Huffman, 2003). The gender segregation and difference in wages has most often been attributed to; gender essentialism (Charles and Grusky, 2014), human capital theory (Hill and Killingsworth, 1989; Mincer and Polacheck, 1974; Tam, 1997), devaluation theory (England, 1992; Karlin, England and Richardson, 2002; Kilbourne et al, 1994) or discrimination (Yip and Wong, 2014). Such theories imply linearity between wages and gender composition, but more recent research has added more complexity to the association.

Cotter, Hermansen and Vanneman (2004) found in the US that wages are highest for sex integrated occupations, and similar results have been found in both Denmark (Albæk and Thomsen, 2011) and Sweden (Magnusson, 2013). The relationships detected are nonlinear, but the wages in female dominated occupations are still found to be lower than in male dominated occupations. A further argument regarding Nordic countries is that their welfare states have to be taken into consideration, as the large public sector plays a great role in the occupational gender segregation (Hansen, 1997). The public sector has generated more part- time employment and more flexibility for caring responsibilities, which is both created by and creates occupational and sectorial gender segregation (Drange and Egeland, 2014).

The relationship between wages and gender by occupational sex composition is not fully established and the field contains many disagreements of both the patterns and the reasons behind them, both in Europe and in the Nordic countries. The main aim of this study is therefore to contribute to the existing knowledge by studying a new context, i.e. the Faroe Islands, which is a self-governing region of Denmark. Not many social studies are conducted on the Faroe Islands in general, and the latest study investigating the gender pay gap is from 2008, and was based on 60 per cent of the public sector with a sample of 3663 people and did not incorporate occupational sex composition (see Joensen, 2009). The second and more specific aim of this study is therefore also to contribute to the limited existing knowledge on the Faroese labour market.

(5)

2 Using 2011 census data from the Faroese national statistical authority, this study investigates the relationship between gender, wages and occupational sex composition, with added interests of employment sector and household characteristics. The Faroe Islands have high employment rates, high part-time rates and high fertility, as well as a large fishing industry; therefore, it will be interesting to compare the relationship between gender and wage in the Faroe Islands differs to other countries. The study is based on the complete employed population between the ages of 25 to 69. In order to get permission to use individual register data, the information had to be censored into several smaller datasets that cannot be merged. Only two datasets were chosen for this study, due to the dependent variable and main controls were present in both; such as working hours and education. In this way the datasets contain exactly the same people and are relatable, but do not risk the confidentiality of individual respondents. Dataset 1 is used for main analyses and Dataset 2 complements the discussion and findings with household characteristics. Despite the limitation of censorship, the datasets are believed to be sufficient for answering the research questions, which are:

1. What is the wage difference between genders in the Faroe Islands?

2. How is the relationship between occupational sex composition and wages in the Faroe Islands shaped? Do the wages decrease by the proportion of women in an occupation or is the relationship nonlinear?

3. To what extent do the associations between wage, sector and occupational sex composition vary by gender?

The paper is divided into five main sections. The first is an introduction to the context of the Faroe Islands. The second introduces previous theory and findings divided into: occupational sex composition and wages; non-linearity and its relationship with wages; and the special role of the Nordic welfare model. The third section communicates the data and methods used for the study, and the fourth the results. The final part summarises and discusses the findings in relation to the literature and provides concluding remarks.

(6)

3

The Faroe Islands

The Faroe Islands are a small archipelago in the middle of the North Atlantic located halfway between Iceland and Norway, which has since 1948 been a self-governing region of the Danish kingdom with own parliament, flag and language. The country is made up of 18 islands and had in January 2017 a population of 49,884, of which 25,525 were employed in November 2016 (Hagstova.fo, 2017). The Faroe Islands participate in the Nordic Cooperation established by the Helsinki treaty and thus share the Nordic welfare model and have similar social and economic structures as other Nordic countries, including common labour market establishments (Government.fo, 2016; Nordic Co-operation, 2017). The biggest national income comes from the fishing industry, which accounts for over 97% of the expert volume, and the second largest industry is tourism (Sansir, 2015). Faroese statistics show that; the country together with Iceland has the highest labour force participation rate in Europe with 76 per cent employed women and 79 per cent employed men between the ages of 15 to 75 with the largest single occupation being care work, which also is the most female dominated occupation in the labour market (Hagstova.fo, 2014a). The Faroese women have the highest part-time ratio in the Nordic countries, with over half (51 per cent) working less than 35 hours a week (Hagstova.fo, 2014b). However, the country also has the highest fertility rate and is the only one above replacement level in the Nordic region with 2,5 children on average per woman (Norden, 2015). Despite the need and interest for statistical data and research equal to other countries, not much social research is conducted in the Faroe Islands due to the limited statistical resources and the small size of the country. When locating previous research on gender in relation to labour market and wages, not much is found, even though data is collected and available.

The latest Faroese study investigating the gender pay gap was conducted by Joensen (2009), who used data from 2008 on 60 per cent of the Faroese public sector employees comprised of 3663 people of which 63 per cent were women. His study indicates that women in the public sector are employed in less qualified occupations, have lower levels of education and work fewer hours compared to men. The raw gender wage gap was found to be that men on average earned 16,3 per cent more than women. After including controls for age, working hours, education high/low qualifications of the job and occupation, there still existed an unexplained wage gap to women's disadvantage of 2,9 per cent on average. The occupations were divided hierarchically into two groups of lower and higher qualified

(7)

4 occupations in part of the analysis. In the former 58 per cent of the included women were positioned and 38 per cent of the men with an unexplained wage gap between 1.4 - 2 per cent, and the latter was comprised of the remaining 42 per cent women and 62 per cent men with an unexplained gender wage gap of 2-5 per cent. Thus, the study found that a larger proportion of men compared to women work in higher qualified occupations in the public sector in the Faroe Islands and these higher qualified jobs also have larger unexplained gender wage gaps to women's disadvantage. Joensen's (2009) study is limited by only analysing a part of the public sector, and therefore cannot be generalised to the remaining work force. By studying the whole labour market, both private and public sector, the knowledge of the Faroese labour market would be substantially enhanced. It cannot be said whether the results will differ much in this study compared to that of Joensen's, but at least it accounts for more factors and all workers. It has in other countries, for example, been found that earnings in sectors differ (e.g. Hansen, 1997; Schøne, 2015), and thus this could affect the wage differences between the genders. A part from studying the whole labour market, this study further includes the before unstudied relationship between wages and the sex composition of occupations in the Faroe Islands. As the controls differ, the studies cannot be directly compared, but nevertheless they both enlighten the field with different aspects.

Theory and previous research

Overall the gender gap in earnings strongly decreased in the second half of the 20th century as more women entered the labour market and some of the former traditional male occupations; however, the wage gap still exists in most countries (Blau and Kahn, 2000). In 2013 women on average earned 16.4 per cent less than men in the European Union (EU) ranging from 5 to over 20 per cent in differences (Eurostat, 2015). These gender pay gaps are calculated from the average gross hourly earnings and are unadjusted for individual characteristics that partially explain the difference, such as education, gender and experience.

The Eurostat release claims that a gender pay gap is the "consequence of various inequalities (structural differences) in the labour market such as different working patterns, differences in institutional mechanisms and systems of wage setting" and is linked to legal, social and economic factors (2015, p. 6).

(8)

5 Occupational sex composition and wages

The gender pay gap is frequently linked to occupational segregation by gender in the labour market (e.g. Kilbourne, England and Beron, 1994; Melkas and Anker, 1997; Ellingsæter, 2013). According to Ellingsæter gender segregation is defined as "the unequal distribution of women and men across the occupational structure, including occupations, workplaces, industries and sectors" (2013: 501). The concept thus encompasses both vertical segregation of different hierarchical positions as well as horizontal segregation into different occupations.

Women, compared to men, are generally more likely to fill positions of lower-pay and lower- status (Melkas and Anker, 1997) and research has also found that occupations with a higher share of women are less rewarded than those with higher shares of men (Kilbourne, England and Beron, 1994).

Occupational gender segregation is argued to have negative effects on both the society and the individual. On the societal level the segregation can lead to economic inefficiency if not all resources are used sufficiently. On the individual level the segregation reinforces the prevailing gender differences regarding wages, work hours and career opportunities (Larsen, Holt and Larsen, 2016).

Several different theories seek to explain the origins and the effects of occupational gender segregation, and these can be divided into supply, demand and institutional explanations (Larsen et al., 2016). The supply side regards the employees' decisions and actions; here education is important, as it affects individuals' placement on the labour market.

It has been argued that men and women self-select into different areas of education with women choosing areas of less economic capital leading to lower wages (Oschenfeld, 2014).

Charles and Grusky (2014) argue that gender segregation in the labour market is underpinned by gender essentialism and male primacy, which contrast with gender egalitarian values and the egalitarian education and employment rates. They found that the horizontal segregation is based on the belief that women and men have different traits and thus suit different task requirements: women as more caring and men as more manual. The vertical segregation was found based on the belief that men are better suited in positions of authority and domination.

These ideas are reinforced in e.g. popular culture, social interaction and gender specific socialization, resulting in everyone internalizing these as natural gender traits (Charles and Grusky, 2014). While such gender essentialist choices might lead to segregation, segregation can also shape norms that lead to gender essentialist choices (Melkas and Anker, 1997). This partially relates to the supply side of the human capital model, as Becker (1985) argues that

(9)

6 women are less inclined to invest in human capital such as education and on the job training as their careers are shorter and more likely to be interrupted by family obligations and domestic work. Thus, many supply theories argue that the investments made, or the lack of them, result in lower female wages relative to that of men.

The demand side in explaining gender segregation regards the employers' demand for a specific workforce, which can shape the conditions to fit a certain population, be it for example women or men (Larsen et al., 2016). This relates to Bergmann's (1974) 'overcrowding model', which states that an exclusion of women from 'male' labour can result in a segregated labour market with an excess of workers in the 'female' labour market. This excess results in lower wages and thus also a gender pay gap for equally productive work.

Becker (1985) also argues that segregation might be partially caused by the avoidance of hiring women, and such explanations can be associated with discrimination. Another discrimination interpretation of the relationship between decreasing wages with increasing occupational share of women is the belief that the female gender is being devalued. The advocates of the devaluation theory claim that society has institutionalized sex biased cultural values that result in decreased wages in female dominated occupations for both sexes (England, 1992). Yip and Wong (2014) argue that in Hong Kong female dominated occupations have lowest wages due to statistical discrimination based on aggregated fertility levels. They hold that companies take future losses into account when determining present wages based on the aggregate fertility and thus the higher chances of workers taking paid maternity leaves. Thus, many theories of demand argue that the gender wage gap partially results from statistical discrimination of females and female work. While many of the theories and studies mentioned in this paper relate the gender pay gap to occupational sex composition, it is important to note that gender segregation in the labour market does not in itself equals inequality, as long as occupations across the labour market are equal in terms such as wages, promotion opportunities and status (Melkas and Anker, 1997; Ellingsæter, 2013). However, the segregation does become a source of the gender gap when it leads to lower wages and fewer career opportunities for female workers (Melkas and Anker, 1997).

Nonlinear association between occupational sex composition and wages Most of the theoretical literature - including the crowding model, devaluation theory and partially the human capital model - imply for various reasons that there is a linear relationship between increasing shares of women and decreasing wages. Devaluation theory in particular

(10)

7 has gained much support in the past. (e.g. England, 1992; Killbourne, England and Beron, 1994). However, findings and arguments exist which contradict the theory. Hakim (1998) questions the devaluation theory as she argues that in Britain the highest wages are in occupations with mixed sex composition. This is supported by the work of Cotter et al.

(2004) who find similar patterns for the US with regard to sex-integrated occupations.

Further, England, Allison and Wu (2007) using longitudinal data found no causal effect of changes in wage on change in sex-composition (as would be predicted by the human capital model) and they found very little support for the devaluation theory in how changes in sex- composition affect the wages. It is, however, notable that most studies also find that occupations with high proportion of women mostly have lower earnings than the ones with high proportion of males. Magnusson's (2013) Swedish studies also show inconsistency with the linear assumption of the above theories, as the highest earnings for both men and women were found in sex integrated occupations, and thus she argues that wages are not a decreasing function of the share of females in the occupation, but rather the relationship is nonlinear.

Magnusson (2013), furthermore, using longitudinal studies concludes that the wage growth is highest for both genders for those who move from strongly male or female-dominated occupations to occupations with a more mixed sex composition.

The inconsistency with the devaluation theory that Magnusson finds is partly contradicted by a Danish study by Albæk and Thomsen (2011). They report that the relationship between share of females and wages in the overall Danish labour market is nonlinear, as the highest wages are in occupations with around 30 per cent female employees.

After the peak the association steeply decreases resulting in female-dominated occupations to have lower wages than male-dominated ones, and this pattern is also found when the labour market is divided into private and public sector. But, when occupational groups are analysed separately, the wages decrease by increasing proportion of women. The authors thus conclude that the reason for the nonlinearity is that sex integrated occupations have highest wages, but within the occupational groups of more similar wages, the results show consistency with devaluation theory (Albæk and Thomsen, 2011, pp. 8-9).

While theories on a linear association between wages and sex composition are numerous, Cotter et al. (2004) point out that not much literature is to be found regarding a nonlinear relationship between the share of females and wages. This was met by this study, as hardly any theories on why sex integrated occupations sometimes have highest earnings were located. The only answer is provided by Albæk and Thomsen, who found that in Denmark

(11)

8 the most integrated occupations have overall higher knowledge that leads to higher wages, but as mentioned, when divided by occupational groups (having the same knowledge level) the wages showed sings of devaluation (2011: p. 9). The relationship between wages and sex composition is therefore not clear-cut but rather has several dimensions that are not fully explained by any of the theories. But, while sex integrated occupations show highest wages in several studies, female dominated occupations appear to have lower wages than the male dominated ones in most studies. Therefore, the mechanisms of occupational segregation discussed can possibly affect the gender pay gap to some extent, but they do not seem to be fully explanatory. It has in regards to this issue been argued that in order for the theories on devaluation, human capital model and overcrowding to be maintained they need to be specified otherwise in order to account for the nonlinearity, which they fail to do now (Grönlund and Magnusson, 2013).

Nordic welfare states and women's work

The last explanation for the occupational gender segregation is the institutional one. The Nordic welfare state is argued to be essential to account for when studying occupational gender segregation in the Nordic countries (Hansen, 1997). Hansen (1997) argues that the occupational gender segregation is comparatively high in the Nordics compared to other countries, which is due to differing welfare regimes. Therefore, she argues, the general explanations of gender segregation are insufficient for analyses of the Nordics. Whereas most of the western and southern European welfare states are defined around male breadwinners and female homemakers, the Nordic welfare state has normalized women's integration to the labour market. The integration was made possible by converting much of previous unpaid housework, especially care work, to being paid labour together with providing supportive extensive services (Melkas and Anker, 1997). But the expansion also likely caused greater gender segregation with the high demand for female labour power in the aforementioned caring occupations (Ellingsæter, 2013; Hansen, 1997). From the expansion of the Nordic welfare state a very large public sector was created, which subsequently plays a great part in the Nordic labour markets - both as a work employer and a social benefits provider (Hansen, 1997).

Women and men are usually unequally distributed into sectors with more women in the public and more men in the private. This distribution is argued to account for a part of the overall gender pay gap as the sectors have different earnings (Hansen, 1997; Schøne, 2015).

(12)

9 The main explanation for the gender distribution into different sectors seems to be, apart from the demand of female labour due to the expansion of the welfare state, the role of motherhood. The welfare state has made it possible for Nordic women to more easily combine work and family lives with policies of parental leave and public childcare (Geist, 2005; Larsen et al., 2016). There are, however, some arguments that claim that women's use of parental leave and other family friendly policies leads to absences that can have harmful effects toward their career, both from the employers perspective (e.g. Evertsson and Duvander, 2011; Bihagen and Ohls, 2006) and through women's personal shift in work commitment (Halrynjo and Lyng, 2009). The public sector is seen as more family friendly than the private sector, and therefore is seen as more attractive for mothers (Schøne, 2015).

Hansen (1997) argues that it is usually assumed that employees have to make trade-offs between family flexibilities, which are better in the public sector, and wages, which are better in the private sector. She found that in Norway the male-dominated private sector has the best earning possibilities for both genders, but when women take on caring responsibilities, the public sector is much more attractive, as there will be no punishment for such caretaking. For men the private sector always has higher earnings no matter the female percentage, but for women the public employees has lower annual earnings in the male dominated occupations, but higher in the female dominated occupations. In this way for women, there did not necessarily have to be a trade-off.

Another issue reflected in the 'family friendliness' of the public sector is the high part-time employment rate (Drange and Egeland, 2014). The Nordic countries have generally very high proportion of the female employed population working part-time, however, the levels differ across countries. With estimates from 2015 in the ages 25-64 years, the highest prevalence was in Norway (around 35 per cent) then Sweden (33 per cent), Iceland (30 per cent), Denmark (27 per cent) with the much lowest being in Finland (14 per cent) (Norden Statbank, 2015. Own computations). The Faroe Islands are not included in the statistics, and it is not clear if they are included in the Danish estimates, but as mentioned earlier the level of part-time workers in the Faroe Islands is even higher, with a level around 51 per cent (Hagstova.fo, 2014b). The main reasons for women's high part-time work in the Nordic countries are argued to be "family responsibilities, health problems and workplaces/sectors not offering full-time positions to all employees" (Drange and Egeland, 2014, p. 23).

Rosenfeld and Birkelund (1995) argue that while the reasons behind working part-time might be close to identical across the Nordic countries, the levels vary due to the interconnectedness

(13)

10 of the family, market and state depending on the contextual economy, politics, ideology and history. For example, Finland has always had a "full-time culture" whereas Norway has not, which is also reflected in more modern times as over twice as many women in Norway (21 per cent) compared to Finland (9 per cent) give "childcare" as main reason for working part- time (studies mentioned in Drange and Egeland, 2014, p. 37). This could potentially also be the case for the Faroe Islands, which as mentioned earlier, has higher fertility rates and higher female part-time rates compared to the other Nordic countries. The choice of many women working part-time can, however, also negatively affect others in the occupation. Drange and Egeland mention the concept "part-time culture" which relates to the interconnectedness of the reinforcing work organization and supply (2014, p. 45). In occupations where there are good opportunities for part-time and where its consequences are smaller, more women choose to work part-time, especially after childbirth. However, these high shares of part-time female employment are also increasing the levels of involuntary part-time employment, as some women become restricted by the opportunities of the occupations, even though wanting to work full-time.

Overall the Nordic welfare state seems to have some isolated issues compared to other welfare states, and this has to be kept in mind when analysing the countries, as it might further cause complexity to the already complex relationship between occupational sex segregation, gender and wages.

Based on the literature, it seems safe to claim that the relationship studied between wages, gender and occupational sex composition still needs to be established, as many different and sometimes contradicting findings exist. The main aim of this study is to add to the general knowledge of the relationship by including the Faroese labour market in the debate; thereby also contributing to the very lacking knowledge of the particular Faroese labour market.

The expected findings of the research questions based on the literature are; 1) that there exists a gender wage gap to women's disadvantage; 2) that the relationship between wages and occupational sex composition is nonlinear with gender integrated occupations having higher earnings; and 3) that the association between wage and sector differ by gender, with private sectors having higher earnings for men, but not as much for women. Sub analyses of the last research question are, further, expected to show gender differences in the association between wage and household characteristics, as well as differences between full- time and part-time workers in both sector and household characteristics. The expected

(14)

11 findings are based on the Faroe Islands sharing the Nordic welfare model, however, due to being a small and perhaps distinctive country, the findings may as well be distinctive too, but no previous findings are to support such an expectation.

Data and methods

Data

The data used comes from the Faroese national census data from 2011 acquired from the Faroese statistical agency (hagstova.fo), which is seen as fit for studying the relationship between wages, gender and occupational sex composition, as all those variables and additional ones are included. Furthermore, access to the whole labour market provides a very strong sample. Unfortunately, due to the preservation of individual anonimity, I cannot as a Master student acquire a full dataset, but rather the agency tailored smaller datasets to fit the needs of my paper with as many variables as possible, of which two were chosen for the final analyses.

The two full datasets were comprised of the same 37,965 people, i.e. the whole population over the age of 15. The population, however, has been narrowed down. Firstly, the study only considers the employed population, and therefore only includes those working at least 1 hour peer week, dropping everyone with less. Secondly, everyone in ages under 25 and over 69 were dropped, in order to get the most representative sample of the labour market. This is, further, based on that the Faroese statistical agency, have found that part-time hours is most common in older and younger ages, due to retirement or studies(Hagstova.fo, 2014b), and therefore it is not held that they are fully representative of the Faroese labour market as a whole and might affect the results. The resulting population in both datasets are 21,563 observations.

In both datasets the dependent variable is the logarithm of the median annual wage.

The original wage variable was coded in smaller intervals, and in order to use OLS regressions, I recoded the interval to be the mid point of the interval; for example those earning between 190,000-200,000 are set to earn 195,000 DKK/year. While not being as great as the exact wage, it is believed that the results will still not be too far from the actual wages and therefore not interfere too much with the overall results. Further, the data has

(15)

12 truncated the wages at the top resulting in everyone earning 1,000,000 or more are set to earn 1,000,000. It was decided to keep them in the population, as it will more likely play down the gender gap estimations rather than overstate them, and analyses have been made both with and without showing any particular difference. The wages were, moreover, logarithmically transformed as the distribution was a bit skewed and to the benefit of predicting per cent difference of wages rather than absolute differences.

The main independent variables of Dataset 1 is a female dummy, depicting the female gender in reference to the male, and the occupational female per cent representing the sex composition of occupations that the observation works in. The variable of sex composition is gained directly from the agency in divisions of five 20 per cent categories: 0-20%, 20-40%, 40-60%, 60-80% and 80-100% depicting the share of females.

The control variables included are: age, which is divided into age groups of 25-39 years, 40-54 years and 55-69 years; educational level, which is classified into 1) compulsory, 2) upper-secondary, 3) undergraduate and 4) postgraduate; and weekly hours worked that is divided into <15 hours, 15-34 hours, 35-47 hours and 48+ hours. Up until now both datasets contain the same variables, the ones that differ between the two are: in Dataset 1, which is the main dataset used, a private dummy is added describing the sector worked in and the ISCO International Standard Classifications of Occupations (ILO, 2012), for occupational group worked in. The more detailed occupations themselves are not included as it became to detailed for me to acquire, but this variable is controlled for to account for vertical segregation that most likely explains a part of wages earned. Dataset 2 on the other hand, has the variable household describing age of youngest child and number of adults in the household, which is recoded in to two variables: age of youngest child divided into the categories; no child under 18, child aged 0-5 and child aged 6-17; and number of adults in household is divided in to; single adult, 2 adults, and 3+ adults. For Dataset 2 these household characteristics, except 3+ adults, are the main independent variables.

(16)

13

Table 1 Descriptive statistics of included variables and datasets

Variable

Men Women Dataset

% min max mean st. d. % min max mean st. d. 1 2

Annual Wages (1000 DKK) 5 1000 255 156 5 1000 170 62 √ √

Private sector 75 36 √ -

Sex composition

0-20% fem. 53 3 √ √

20-40% fem. 14 6 √ √

40-60% fem. 22 22 √ √

60-80% fem. 8 23 √ √

80-100% fem. 3 46 √ √

Age groups

25-39 32 32 √ √

40-54 40 42 √ √

55-69 27 26 √ √

Education

Compulsory 18 23 √ √

Upper secondary 50 42 √ √

Undergraduate 23 28 √ √

Postgraduate 9 57 √ √

Weekly hours

<15 3 7 √ √

16-34 8 45 √ √

35-47 50 43 √ √

48+ 39 5 √ √

ISCO groups

1: Managers 10 4 √ -

2: Professionals 14 27 √ -

3: Technicians and associate

professionals 26 10 √ -

4: Clerical support workers 3 13 √ -

5: Services and sales workers 8 30 √ -

6: Skilled agricultural,

forestry and fishery workers 8 0.4 √ -

7: Crafts and related trades

workers 15 0.7 √ -

8: Plant and machine

operators and assemblers 9 5 √ -

9: Elementary occupations 6 10 √ -

0: Armed forces occupations 0.1 0 √ -

Household

No child <18 52 44 - √

Youngest child 0-5 years 22 25 - √

Youngest child 6-17 years 26 31 - √

Single adult 13 10 - √

Two adults 50 55 - √

Three+ adults 37 35 - √

Note: In Dataset 1 the number of men is 11,086 and females 9,433. In Dataset 2 the number of men is 11,675 and females 9,809.

However, the percentages in the categories in both datasets are the same. The only difference between the datasets is a small difference in the mean wage varying with 3000 DKK for men and 1000 DKK for women. The wage from Dataset 1 is present in the table.

Limitations of the data

The data used has some limitations, of which one is the concept of time. By only using cross- sectional data the study cannot incorporate changes over time, and therefore the study cannot study if and why changes occur as only patterns and correlations at one point in time can be established. Other limitations in this study originate from grouping of the variables, which is legally required to preserve anonymity for the individuals. This is to eliminate specific

(17)

14 identifiers that can risk the privacy of individuals that might be more easily targeted in a small geographical place like the Faroe Islands. In order to gain access to all the variables wanted as controls, such groupings had to be done before my access, and the benefits of controlling for all the variables is very important. The grouped variables include the age, hours worked, education, wage, occupational sex composition, and household characteristics.

A consequence from the grouping is that age squared, which often is included as an independent variable in similar research to account for its nonlinearity, cannot be added.

Further, the variable of age of the youngest child in the household, if any, does not mention how many children there are all in all and whether being a child of the observation; it could in some instances be perhaps a sibling of a person still living at home with their parents. It does neither differ between having had no children or having no children under the age of eighteen. Despite the variable being a bit vague, the number will mainly be analysed as being the child of the observations, since this is held to be the main relation. The variable of number of adults also provides a problem. Additional adults do not distinguish between the relations, e.g. being an adult living with your parents, a parent living with an older child, or living with a cohabitant. Therefore, these associations have to be analysed very cautious, but due to incorporating only people in ages 25 to 69, it is mainly held that being two adults is cohabitation, and this is what the interpretation will be. The category of being 3+ adults in the household is more troublesome, as it is not sure whether the observation is a parent living with children or a child living with parents. Due to the interpretation difficulties, but also the high number of observations (over 7,000) the estimate is accounted for in the regression models, but is not interpreted.

The last limitation of preserving anonimity is that with two datasets comes the inability to control for all the variables together; e.g. the fact that I now cannot account for household characteristics together with the occupational group and sector. Therefore I cannot read the association between having children and the sector, which is much discussed in the literature. Overall, Dataset 1 will be the main one and the addition of Dataset 2 with its household characteristics can shed light on a different side to be discussed in relation to the other dataset, as the population is the same.

As mentioned, the dependent variable is the measure of income by annual wage.

Whereas many studies use the hourly earnings as the indicator of income and the pay gap, this was not available for me. The limitation is that I cannot study the difference in pay for the hourly wages and therefore relate exactly to the official gender wage gap reports. On the

(18)

15 other hand, the study has the possibility to study reward structures. Hansen (1997) argues that hourly wage might underestimate the impact of family situations and responsibilities on the wage inequality, as it would not reflect, for example, the possibility of working overtime that might be greater for men than women. Finally, the study is limited by a lack of control variables, such as experience, number of children, area of residence and others.

Analytical strategy

The aim of the thesis is to add to the existing knowledge of the association between wages, gender and occupational sex composition by studying the Faroese labour market using cross- sectional data from 2011. The analyses are performed using multiple linear regressions studying the association between the dependent variable (logarithm of wages) and the independent variables. Due to having dummies for five occupational sex compositional categories, its differing association with wages can be accounted for similarly with establishing the pattern between the two. The analyses are divided into two, and the first part includes the whole labour market focusing on the first two research questions - 1) the gender wage gap and 2) the shape of the relationship between annual wages and occupational sex composition.

In exploring the gross gender wage gap I use annual wage and account for hours worked. This provides an estimate of the unadjusted gender wage gap in annual earnings. The evolvement of the gendered wage gap, if any, will be analysed by adding additional controls to subsequent models. Firstly, by controlling for age group, educational level, ISCO group and sector worked in, and secondly by accounting for occupational sex composition; both to see if the gender gap diminishes as well as for defining the pattern of the relationship between sex composition and wages. It is mainly Dataset 1 that is used due to containing sector and occupational group (ISCO) worked in, which are seen as very important in predicting wages. The analysis will, however, be supplemented by Dataset 2 looking at the association between wages and household characteristics. These characteristics are defined by dummies for having a younger or older child in the household as well as dummies for being 2 adults compared to none. It is unfortunately not possible to control for sector and ISCO group in Dataset 2. However, it is still interesting to get an insight into the association of the Faroese household factors and yearly wages for discussion.

In the second part of the analysis research question three is examined. Here women and men are analysed separately in order to study whether the overall relationship and

(19)

16 associations found are prevalent for both genders or if any associations differ between them.

The question is based on the literature that argues that there exists a gender difference in the association between wages and occupational sex composition in the Nordic countries (e.g.

Hansen, 1997). By exploring other parameters in the relationship between gender, wage and occupational sex composition more knowledge is to be gained of the interrelationship as well as the specific context. This gender difference analysis is divided into two sub parts, as it examines both all workers and, secondly, it examines the part-time workers, due to its high prevalence in the Faroe Islands. In both sub parts the analysis starts with descriptive statistics of the distribution of sexes by the variable studied, and secondly, the associations are studied using OLS regressions. The regressions use the same combinations as in the first part, and in that way, the results can be analysed in relation to each other; the labour market as a whole;

divided by genders; and the division by genders in part-time.

Results

The gender wage gap

The first step in the analysis focuses on the first two research questions and studies the overall gender wage gap and the shape of the relationship between annual wages and occupational sex composition. Model 1 in Table 2 below indicates that men have 12 per cent higher annual wages than women when only adjusting for hours worked. When adding controls for education, age group, occupational group and sector the gap increases to 15.9 per cent, suggesting that women, compared to men, earn even less when accounting for human capital and occupational group (See Model 2 in Table 2). Regressions not presented indicate that the gender gap seems to widen when including occupational group and sector, whereas education narrows it¹. In Model 3, occupational sex composition is added, which makes the gap decrease by 2,4 per cent from Model 2 as well R²increases. This suggests that the sex composition of the occupation worked in is associated with annual wages and the gender wage gap. Based on regressions not shown, the ISCO groups representing vertical segregation seem to explain a lot of the association found between occupational sex composition and wages, as when not included the associations are smaller and weaker. In Model 4 (based on Dataset 2) sector and ISCO controls are not available, which probably

1For example, the gender wage gap would be 17 per cent in Model 2 had education not been controlled for..

(20)

17 explains the lower gender gap of the model. The controls of having a child in the household and number of adults in the household are included, which can highlight some domestic characteristics in relation to wage. The results indicate that having children in the household is positive for annual wages earned. A similar regression model, not included, without the household characteristics had a 1.4 per cent lower gender wage gap and lower R² than Model 4. This indicates that the household is important for annual wages and wage difference between genders.

Table 2 Unstandardized coefficients from an OLS regression of log annual wages of

employed people in ages 25-69 in the Faroe Islands 2011.

Variables Model 1 Model 2 Model 3 Model 4

Female -0.120*** -0.159*** -0.135*** -0.119***

(0.00900) (0.00936) (0.0103) (0.0105)

Age 40-54 0.110*** 0.108*** 0.150***

(0.00855) (0.00850) (0.00972)

Age 55-69 0.0375*** 0.0390*** 0.155***

(0.00966) (0.00962) (0.0114)

Upper Secondary 0.0173 0.0172 0.0827***

(0.0106) (0.0106) (0.0101)

Undergraduate 0.179*** 0.163*** 0.326***

(0.0136) (0.0136) (0.0115)

Postgraduate 0.291*** 0.297*** 0.443***

(0.0186) (0.0185) (0.0169)

20-40% -0.215*** -0.0355**

(0.0214) (0.0142)

40-60% -0.153*** -0.105***

(0.0133) (0.0116)

60-80% -0.171*** -0.0992***

(0.0172) (0.0145)

80-100% -0.0664*** -0.0253*

(0.0164) (0.0141)

Private -0.0265*** -0.0112

(0.00886) (0.00912)

Child 0-5 0.161***

(0.0113)

Child 6-17 0.161***

(0.0101)

2 adults 0.0179

(0.0127)

Constant 11.72*** 11.58*** 11.62*** 11.50***

(0.0191) (0.0230) (0.0241) (0.0233)

Observations 20,519 20,519 20,519 21,484

R² 0.149 0.258 0.267 0.220

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Note: Models 1, 2 and 3 are retrieved from Dataset 1 and Model 4 is from Dataset 2, therefore the varying observations. Not in the table: all models control for categories of hours worked. Models 2 and 3 control for separate ISCO groups. Model 4 controls for being 3+ adults. For full table contact the author.

(21)

18 Occupational sex composition

Studying the relationship between annual wages and the occupational sex composition, Table 2 above demonstrates that the highest annual wages are found in male dominated occupations with 0-20 per cent females (reference category), as all the others have a negative effect. Both Model 3 and 4 include the occupational sex composition, but due to the higher R² and the controls of ISCO and sector, Model 3 is chosen for the analysis of the association of sex composition with wages. The results indicate that the second highest wages are, unexpectedly, earned in the most female dominated occupations, and overall the estimates indicate a nonlinear pattern, which, is neither highest in the integrated occupations nor lowest in the female dominated occupations as expected.

Comparing the two genders

Occupational sex composition and sector

The first sub part of research question three examines the association between wage and the sex composition separately for men and women, as well as the associations with sector and household characteristics. Table 3 is the distribution of the gendered population by sector and sex composition, which demonstrates that the distribution is gendered. Two thirds of the studied women work in the public sector, but only one fourth of the men are publicly employed. Hansen (1997) found that in Norway women were especially concentrated in the public female dominated occupations, and that for men the tendency was opposite with them clustered in the private male dominated occupations. These patterns are the same for the Faroese labour market with those two categories being the largest³.

Table 3 Distribution of genders by sector and sex composition in the Faroese labour market.

Men and women ages 25-69 in 2011.

Gender Sector 0-20% 20-40% 40-60% 60-80% 80-100% Total Proportion in sector

Male

Public 33.1 14.4 22.8 21.1 8.6 100

N=2877 24.8%

Private 59.9 13.6 21.2 3.4 1.9 100

N=8720 75.2%

Female

Public 1.83 5.5 14.5 21.9 56.3 100

N=6166 63.1%

Private 4.6 7.1 41.7 21.5 25.2 100

N=3607 36.9%

3The numbers are marked by bold letters in Table 3.

(22)

19 Exploring the relationships further, the two genders are analysed separately regarding the association between log annual wages and the independent variables (see Table 4).

Table 4 Unstandardized coefficients from an OLS regression of log annual wages of employed people in ages 25-69 in the Faroe Islands 2011 by gender.

Men Women

Variables 5a 6a 7a 5b 6b 7b

Age 40-54 0.125*** 0.125*** 0.149*** 0.0910*** 0.0889*** 0.136***

(0.0136) (0.0135) (0.0149) (0.00946) (0.00942) (0.0115)

Age 55-59 0.0313** 0.0358** 0.138*** 0.0562*** 0.0506*** 0.154***

(0.0151) (0.0151) (0.0169) (0.0110) (0.0110) (0.0141)

Upper secondary

-0.0148 -0.0129 0.0545*** 0.0565*** 0.0532*** 0.116***

(0.0176) (0.0175) (0.0163) (0.0114) (0.0114) (0.0110)

Undergraduate 0.194*** 0.182*** 0.366*** 0.113*** 0.107*** 0.266***

(0.0219) (0.0219) (0.0189) (0.0154) (0.0154) (0.0125)

Postgraduate 0.270*** 0.276*** 0.424*** 0.319*** 0.317*** 0.468***

0.125*** 0.125*** 0.149*** 0.0910*** 0.0889*** 0.136***

20-40% -0.210*** -0.0396** -0.189*** 0.0604**

(0.0290) (0.0185) (0.0392) (0.0307)

40-60% -0.150*** -0.100*** -0.0902*** -0.0553**

(0.0192) (0.0153) (0.0265) (0.0265)

60-80% -0.128*** -0.0576** -0.126*** -0.0472*

(0.0287) (0.0239) (0.0267) (0.0269)

80-100% -0.0430 0.0239 -0.0305 0.0156

(0.0339) (0.0326) (0.0256) (0.0260)

Private 0.0176 0.0221 -0.0791*** -0.0540***

(0.0148) (0.0150) (0.00936) (0.00989)

Child 0-5 0.187*** 0.105***

(0.0177) (0.0132)

Child 6-17 0.176*** 0.124***

(0.0161) (0.0114)

2 adults 0.0771*** -0.0833***

(0.0201) (0.0144)

Constant 11.53*** 11.56*** 11.45*** 11.46*** 11.49*** 11.44***

(0.0428) (0.0431) (0.0409) (0.0207) (0.0306) (0.0334)

Observations 11,086 11,086 11,675 9,433 9,433 9,809

R² 0.201 0.208 0.174 0.255 0.262 0.212

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Note: Models 5a, 5b, 6a and 6b are retrieved from Dataset 1 and Models 7a and 7b is from Dataset 2, therefore the varying observations. Not in the table: all models control for categories of hours worked. Models 5a, 5b, 6a and 6b control for separate ISCO groups. Models 7a and 7b control for being 3+ adults. For full table contact the author.

The gender separate analyses demonstrate that only small differences of associations exist between the genders. Men have unexpectedly no association between sector and wage, but women have, on the contrary, higher earnings in the public sector (see Table 4 Models 6a and 6b). The association between wage and occupational sex composition does not differ as much as expected, and further, both associations are contrary to what was expected. For both genders wages are highest in male dominated occupations⁴. The only difference between the genders, except the wage gap, is that for men wages are lowest in occupations with 20-40%

4 In the male-dominated occupations only 279 women work, this low number might be a selected group of high earners.

5 The figure was also made based on real annual wages and not their logarithm to show a more relatable graph,

(23)

20 women and for women their wages are lowest in the most integrated occupations (see Table 4 Models 6a and 6b). The associations are visualised in Figure 1 below⁵.

Figure 1

The high wages for males in the male dominated occupations seem to be pulled upwards by the very large male (2,275) occupational group of "technicians and associate professionals"

with around half earning over 335,000 DKK a year⁸. In the male-dominated occupations only 279 women work, this low number might be a selected group of high earners. For both genders it can be seen that second highest wages are found in female dominated occupations, however, these numbers are not significant due to large variation in the category for both genders. For men only 400 work in the female dominated occupations, and therefore the estimate is vulnerable for extremes. An example is that 112 men are in the ISCO group of

"professionals", in which more than 50 per cent earn above 365,000 DKK a year and around 20 per cent earn from 575,000 to 1,000,000 a year⁶. These few high earners positively shape the overall association of the few members in that category and also create the insignificant variation. For women, despite being a very large group, the association between wages and working in the female dominated occupations is not significant either.

5 The figure was also made based on real annual wages and not their logarithm to show a more relatable graph, but the shape changed slightly and was therefore not included.

8 Calculated from Dataset 1

11.91212.112.212.3Predicted logarithm of wage

0-20 20-40 40-60 60-80 80-100

Female percent in occupation

Men Women

Separately predicted log of annual wages by gender (Table 4, 6a and 6b)

(24)

21 The estimates of having a child and additional adult in the household are present in Models 7a and 7b (see Table 4). The numbers indicate a positive association with wages for both genders to have a child in the household compared to having no child. For men the association is higher in having smaller children, which might be related to increased breadwinner role in conjunction with perhaps the mother taking a leave and/or decreasing her work hours. For women, it is also found that having a child is positively associated with wages both under and over the age of 5, but the latter being stronger. The former is noteworthy, as I would expect women to temporarily go down in annual wages or staying at the same level due to leave and less flexibility to work long hours. However, here it is important to note that the association might also have been different if the variable was having a child in ages less than e.g. 1 or 2 years or a continuous variable, compared to the 5 years this dataset has. Regarding the association of amount of adults in the household, it is positive for men's wages to have one other adult in the house compared to them being the only one (the reference group). But, for women, the association is negative.

Part-time and sector

The second sub part of research question three examines the association between wage and sector, as well as wage and household characteristics, by gendered part-time population. This analysis is based on the very large part-time working female population, and the interest to see how sector might be related to their annual wages. Such an association might explain some of the wage differences between genders in the Faroe Islands. In Table 5 the genders are distributed by sector and weekly hours worked. In both the public and private male categories the main proportion works full time (35-47 hours a week), with the latter being a larger population and the highest overall category in the table. For women, full time is highest in the private sector, but for the public sector, the highest prevalence is in a part-time group (16-34 hours a week); being the highest female category in the table. The part-time under 15 hours a week does not differ much between the genders and it may be a selected group that works this few hours.

Occupational sex composition and wages in the Faroe Islands

Occupational sex composition and wages in the Faroe Islands

Abstract

Table of Contents

Introduction

The Faroe Islands

Theory and previous research

Data and methods

Results