• No results found

Gender equality and the economic potential of women

N/A
N/A
Protected

Academic year: 2021

Share "Gender equality and the economic potential of women"

Copied!
80
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Gender equality and the economic potential of women

Abstract:

This thesis deals with the effect of gender equality on economic growth. Due to large gender inequalities, the economic potential of qualified women is not optimally used. A panel data study is performed for three time periods. The focus will be on 20 OECD countries and the research will be conducted on a country level basis. The results show that the evidence on the relevance of gender equalities in explaining observed differences in per capita growth rates is

mixed. Gender inequality in education and a gender pay gap hamper economic growth. Remarkably, sectoral gender segregation promotes economic growth. So it is suggested that

gender specialization goes along with a productivity gain. Moreover, the number of female legislators hampers economic growth and, on the other hand, the percentage of women in

parliament is positively related to economic growth.

Carolien D.L. Calkhoven University of Groningen Master’s Thesis Student number: 1705504 Email: caroliencalkhoven@hotmail.com Date: 30-07-2012 JEL-code: O49, J16, C23

Keywords: Gender equality, economic growth

(2)

2 1. Introduction

Women and men are legally equal, but clearly, they are not economically equal. Equality between women and men is one of the European Union’s main objectives and tasks (European Commission 2011). However, only one in three managers is a woman, and women work predominantly in sectors or occupations where wages are the lowest (European Commission 2006). The average pay differential for men and women in full-time jobs in OECD countries is more than 18 percent. In Japan and Korea, women earn at least one-third less than men. In Germany, Switzerland, Canada, and the United States, women earn over 20 percent less than men (OECD 2008). In 2007 the unweighted mean for the pay gap was 17 percent within the EU-countries. There was however considerable dispersion between member states, from 4.4 percent in Italy to 30.4 percent in Estonia (Löfström 2009). This study tries to find the effect of gender equality on GDP growth. Or, to put it differently, what is the economic potential of women?

According to the European Commission (2012) the gender inequality results in lost resources which are bad for economic growth. Or, as OECD (2008) states it, the economic potential of qualified women is not optimally used. Despite making up nearly half of the workforce and accounting for 60 percent of new university graduates in the EU, women continue to be under-represented in economic decision-making positions. Women’s untapped talent which could benefit businesses and society as a whole, represents a wasted investment in human capital (European Commission 2011).

There are gender inequalities in labour participation, work duration, education, and wages. Moreover, there is strong sectoral gender segregation, there are few women in management functions, and few women in political functions.

(3)

self-3 employment. Calculations of these gains shows that there is a potential for increased GDP of between 15 and 45 percent in the EU member states. This suggests that there are major benefits to be gained from enhancing gender equality.

With respect to women in management functions, Valerio (2009) states that women leaders are good for business. As United States businesses expand into new markets, cultures, and workforces across the United States and around the world, the companies that integrate gender diversity into their business strategy prove to be more successful. Catalyst (2004) has shown that the Fortune 500 companies with the highest percentages of women corporate officers, experiences, on average, a 35.1 percent higher return on equity and 34 percent higher total return to shareholders than did those with the lowest percentage of women corporate officers. The European Commission (2011) adds to this that gender diversity brings a number of vital benefits to boardrooms, such as higher returns, better overall performance, better risk management and greater employment of female talents.

According to the European Commission (2011), in 2010, only three governments from EU member states were led by women, while the average number of female members of national parliaments was 24 percent. The percentage is above 40 percent in the Netherlands and Sweden and below 10 percent in Malta and Hungary. With respect to senior ministers of national governments the share of women increased from 22 percent in 2005 to 27 percent in 2010. Multiple studies show that when women are well-represented in decision-making bodies, the overall quality of governance rises and levels of corruption decrease (OECD 2008).

Other research focuses its analysis of gender inequality and economic growth on a specific part of gender inequality. There is not yet a comprehensive view of the effects of gender inequality on economic growth. This study is the first to include multiple factors of gender inequality in one model, to research their effect on economic growth.

(4)

4 and gives some recommendations. There is a Data Appendix which gives a detailed description of the variables and their sources and a complete list of the countries included in the analysis.

2. Theoretical background and hypothesis development

First multiple gender inequalities are discussed before turning to standard regressors of economic growth.

2.1 Gender inequality

2.1.1 Labour Participation

Over the last years, there is a striking increase in the female labour participation rate. In the Netherlands, the participation rate of women rose steadily from 59 percent in 1995 to 74 percent in 2009 (OECD). According to Boeri et al. (2005), there was a strong increase in the female labour participation rate in the whole European Union and in North America, between 1960 and 2000. Furthermore, the gender employment gap, defined as the difference in employment rates between men and women, is falling in all EU-15 countries. On average, the gender gap almost halved from 30 percent in 1980 to 16.7 percent in 2000 (Boeri et al. 2005). Between 2009 and 2010 the gender employment gap narrowed by 0.4 percentage points in the EU from 13.3 to 12.9 percent. However, significant differences exist throughout the EU. Malta, Italy, and Greece have the lowest rates for women while Lithuania, Estonia, and Latvia have the lowest rates for men (European Commission 2011).

(5)

5 Figure 1: Female labour force participation rate 1996 – 2009 (Source: OECD)

One important motivation for the desire to increase the participation rate, is to reinforce the sustainability of social protection systems (Taskforce Part-time Plus 2010, European Commission 2011). However, according to Nelen et al. (2011), the sustainability of the welfare state can only be improved if labour participation in terms of the number of hours worked improves. So there should not only be a focus on the participation rate, but also on the number of hours worked.

Hochberg and Schmid (2005), based on a panel of 16 European countries and Japan for the period between 1993 and 2003, estimate the effect of the increasing participation rate on GDP growth to be an average of 0.4 percent per annum. The effect is greater in countries experiencing strong growth, such as Spain and Ireland, and is more limited in countries experiencing weaker growth, such as Germany and Italy. WWC (2006) states that the total potential benefits of reducing the gender segregation of jobs and increasing women’s participation could be worth between £15 billion and £23 billion or 1.3 to 2.0 percent of GDP

40 45 50 55 60 65 70 75 80 85 90 96 97 98 99 00 01 02 03 04 05 06 07 08 09 Fem al e lab o u r fo rc e p ar tici p ation r ate (% to tal) Austria Belgium Denmark Finland France Germany Greece Hungary Iceland Ireland Italy Luxembourg Netherlands Norway

Portugal Slovak Republic

Spain Sweden

(6)

6 in Great Britain. Löfström (2009) interprets gender labour market equality as women and men working to the same extent in paid jobs, having an equal share of part-time work and self-employment. Calculations of these gains shows that there is a potential for increased GDP of between 15 and 45 percent in the EU member states.

According to the European Commission (2011) women have a higher level of tertiary educational attainment than men in the EU. Therefore, gender gaps in participation could result in lower average labour force productivity than in the absence of such gender inequalities (Klasen 1997). Nowadays the professional careers of women do not fully reflect their skill levels, which is a waste of human resources and competences, while human capital is the key to competitiveness in the global economy (European Commission 2011). This leads to the following hypothesis.

H.1 Reducing the gender inequality in labour participation has a positive effect on economic growth.

2.1.2 Work duration

Even though the participation gap becomes smaller in all EU-countries, there are still a lot of differences between the female labour force and the male labour force. Although the incidence of part-time working women is highest for the Netherlands, in all EU countries women work much more on a part-time basis than men (Boeri et al. 2005, Evers et al. 2005). In 2009 31.4 percent of European women and just 8.1 percent of men worked part-time. Noteworthy is the fact that the countries with the highest female labour participation rates, also have among the highest part-time rates (European Commission 2011). Besides the gender inequality in part-time employment, the European foundation for the improvement of living and working conditions (2007) state that the part-time employment is also not equally distributed among age groups, nor among countries, sectors or occupations.

(7)

7 least 65 percent (Finland), but on average it is 78 percent (Boeri et al. 2005). In addition, there is a net pay gap between part-time and full-time labourers, which via the theory of efficiency wages has an effect on per hour labour productivity (CPB 2007). This pay gap is also confirmed by the OECD (1999) which states that, on average, part-timers receive lower levels of earnings per hour worked.

Evers et al. (2005) state that the determinants of labour productivity in relationship to the amount of working hours a week are, amongst other things, the balance between working- and private life, coordination- and tuning-costs, motivation, the amount of productive hours per day, and the nature of the function (teamwork or stand-alone). Thurik and Van der Wijst (1982) state that part-time employees may have other motives to work than full-time employees, which could cause a difference in per hour labour productivity. The amount of education, training, and experience also differs between part-time and full-time labourers, this is because, rationally, part-time labourers would invest less in human capital than full-time labourers (Evers et al. 2005).

(8)

8 Figure 2: Link of amount of hours a week to per hour productivity (Source: Evers et al. 2005)

Nelen and De Grip (2009) state that on the basis of the theory of human capital, it is expected that part-time workers are less productive than full-time workers anyway, because there is less investment in the development of part-time workers’ human capital. Evers et al. (2005) agree that there are differences in education and training levels between part-time and full-time labourers, and add that there is also an experience effect. According to CPB (2005), education, training, and experience all have a positive effect on the per hour labour productivity. The OECD (1999) highlights the fact that, on average, part-timers tend to receive lower levels of training compared to full-timers. A positive effect of part-time work is that it is good for employment. Part-time work means that, ceteris paribus, more workers can fulfil one full-time job (in terms of FTE) (CPB 1998).

This leads to the following hypothesis.

H.2 Reducing the gender gap in work duration has a positive effect on economic growth.

2.1.3 Education

The female to male ratio of the youth literacy rates in all regions of the world has risen in the period from 1990 to 2002. This indicates that there is an improvement in gender equality in education. However, South Asia, Sub-Saharan Africa and the Middle East, and North Africa consistently lack behind the global average. In 2002 the ratios were 79 percent, 89 percent

Amount of hours a week (X)

(9)

9 and 87 percent respectively, compared to a 92 percent global average (Chen 2004). Another indicator for gender equality in education is the female to male students ratio in primary and secondary schools. All regions show an increase in this ratio, however again South Asia, Sub-Saharan Africa and Middle East, and North Africa are lacking behind on the other regions (Chen 2004).

Various studies have proven a negative connection between gender inequality and economic growth. Klasen (1999) finds that countries in South Asia, Sub-Saharan Africa and the Middle East, and North Africa could have had an additional 0.5 to 0.9 percentage point per year income per capita growth if they had achieved gender equality in schooling during the period 1960 to 1992. Moreover, Dollar and Gatti (1999) find that larger female secondary education attainment tend to lead to higher growth rates, while male secondary achievement tends to lead to smaller growth rates, using data for over 100 countries over five-year intervals and two-stage least squares estimation. Hill and King (1993) find that gender inequality in education has effects on the level of aggregate output. They use panel regressions for 152 countries during the period 1960 to 1985, and find that low female to male primary and secondary school enrolment ratio is associated with a lower level of GNP, even after controlling for the effects of female education on GNP.

(10)

10 economic growth’. Finally, there is the demographic transition effect. High female education is one of the significant causal factors of fertility decline. This lower fertility rate initiates a process known as the demographic transition, which may give rise to a transitory increase in the rate of economic growth. First reduced fertility tends to lower the youth dependency burden1 and the total dependency burden2, which tends to increase the supply of aggregate savings within an economy. Moreover, lower fertility rates tend to, over time, result in a larger working-age population share because of previously high population growth. This large labour force will boost economic growth and investment demand. This higher investment demand and the, as already mentioned, higher domestic savings, will result in an increase in the investment rate, which should again increase economic growth (Bloom and Williamson 1997).

However, as Figures 3 and 4 show, nowadays in most European countries the gender gap in education has closed or even reversed. Therefore, closing the gender gap could also mean an increased participation in education of men.

All in all, this results in the following hypothesis.

H.3 Closing the gender gap in education has a positive effect on economic growth.

1 The youth dependency burden is defined as the ratio of the number of youths to the number of people in the working-age population.

(11)

11 Figure 3: Female to male students in primary education (Source: WDI)

(12)

12 Figure 4: Female to male students in secondary education (Source: WDI)

2.1.4 Gender segregation

Men and women have very different occupations and often work in different sectors. In the health and social work sector alone, women make up 80 percent of all workers. And, as women bear the burden of unpaid work and childcare they tend to work shorter hours. As a result, they generally work in sectors and occupations where jobs are compatible with their family responsibilities. Consequently, women are more likely to work part-time, be employed in low-paid jobs and not take on management positions (European Commission 2011a). Many studies acknowledge a very strong occupational segregation between men and women (Boeri et al. 2005, Evers et al. 2005, European Commission 2006, UN 2010, OECD 2008).

85 90 95 100 105 110 115 120 125 85 90 95 00 05 Fem al e to m al e st u d e n ts in se co n d ar y e d u cation (% ) Austria Finland France Ireland Netherlands Norway

(13)

13 According to Bettio et al. (2008) for the European Union as a whole, segregation as measured by the IP index3 is still relatively high, reaching 25.3 percent level for occupational segregation and 18.3 percent for sectoral segregation. However, there are wide country variations, with a gap of about 10 percentage points between the most and least segregated country. For both occupational and sectoral segregation the high-segregated countries are Estonia, Slovakia, Latvia, and Finland, and the four low-segregated countries are Greece, Romania, Malta, and Italy. It can be concluded that the well-known opposition of the 1990s between highly-segregated Nordic countries and low-segregated Mediterranean countries has given way to a similar opposition between (part of) the East and (part of) the Mediterranean. There is no significant change between 1992 and 2000 in the index of occupational segregation for either EU-27 or EU-15. However, for the current decade there is a slight upward trend which is more pronounced for sectoral segregation. Modest change at the aggregate level hides contrasting patterns at country level between 1997 and 2007. Austria, the Czech Republic, Denmark, Norway, Sweden, and the UK experienced relatively fast de-segregation, with decreases in the IP index between 1.5 and 2.8 percentage points. However, in Bulgaria, Ireland, Italy, Latvia, Romania, and Spain segregation increased.

According to Löfström (2009) and Bettio et al. (2008) the traditional gender roles in the home are not only an obstacle to women’s participation in the labour market, they also make flexibility poor since sex-typed occupations and careers restrict the mobility in the labour market. In addition, Löfström (2009) regards gender equality as an economic application of Le Chantelier’s principle, which essentially states that the fewer restrictions one has to consider, the easier it is to achieve a specific goal. So with gender equality in the labour market, work in society is distributed rationally between the sexes, which means that a given occupation is allotted to the person most suitable and not due to prejudices or discriminating rules or practices. When all normal cases are concerned, this leads to a better economic outcome than in alternative cases. In other words, the lack of mobility in the labour market could have a negative effect on economic growth. Bettio et al. (2008) add to this that it is indicated that skill and labour shortages are likely to affect mixed occupations less than male or female dominated occupations in the medium run.

(14)

14 On a more microeconomic level, it is argued that workplace diversity is good for firm performance. Biologists, behavioural economists and psychologists have suggested, during the recent global economic downturn, that more diverse teams make better informed decisions, leading to less risk taking and more successful outcomes for companies (Hausmann et al. 2010). Robinson and Dechant (1997) state five ways in which corporate diversity promotes corporate growth. First, corporate diversity enables a better understanding of the marketplace. Demographic projections indicate that the marketplace is becoming more diverse, and matching the diversity of a company to the diversity of the marketplace increases the ability to penetrate markets. Secondly, diversity increases creativity and innovation. More precise: ‘attitudes, cognitive functioning, and beliefs are not randomly distributed in the population, but tend to vary systematically with demographic variables such as age, race, and gender’ (Robinson and Dechant 1997). Third, diversity gives more effective problem-solving. At first heterogeneity may induce more conflict in the decision making process, but in the end the variety of perspectives that emerges cause decision makers to evaluate more alternatives and more carefully explore the consequences of these alternatives. Fourthly, diversity enhances the effectiveness of corporate leadership. This topic will be elaborated further on. Lastly, diversity promotes more effective global relationships. Cultural sensitivity is essential in an international environment and ethno-cultural diversity makes corporate leaders more sensitive to cultures not in North America. This final argument is not applicable to gender diversity.

Figure 5 gives an overview of the gender distribution between and within different sectors for the EU-27 in 2005. It can be seen that certain sectors are male dominated and others female dominated.

WWC (2006) states that the total potential benefits of reducing the gender segregation of jobs and increasing women’s employment could be worth between £15 billion and £23 billion or 1.3 to 2.0 per cent of GDP in Great Brittan.

All in all, this results in the following hypothesis.

(15)

15 Figure 5: Distribution of employment, by sector, sex and part-time/ full-time status EU-27 in 2005 (%) (Source: Parent-Thirion 2007)

2.1.5 Wages

According to United Nations (1990) and CPB (2007), there is a net pay differential for women. Sectors where women are the majority have lower wages than those dominated by men (European Commission 2011a). More than 50 years after the signing of the Treaty of Rome, which affirmed the principle of equal pay for men and women for the same work or work of equal value, women across the EU, in 2008, earn 17.5 percent less on average than men and there has been no reduction of the gender pay gap4 in the last few years. Again, there

(16)

16 are significant country variations. It varies from nearly 31 percent in Estonia to below 5 percent in Italy (European Commission 2011). These figures are illustrated in Figure 6. The European Commission’s ‘Strategy for equality between women and men (2010-2015)’ has closing the gender pay gap through legislative and non-legislative measures as a core objective. The Strategy sets out actions in five areas: the economy and the labour market, equal pay, equality in senior positions, tackling gender violence, and promoting equality beyond the EU (European Commission 2011a).

When considering the female labour force participation and the gender pay gap, the result might be counter intuitive. The greater the gap between male and female labour force participation, the smaller the average wage gap. Nevertheless, this is easily explained. In countries with low female labour force participation, only women with a relatively high education work professionally. Their pay is relatively high when compared to the male collective, due not least to the fact that their level of education is higher than the average for gainfully employed men. When the female labour force participation rises, the differences between the sexes in terms of education levels will (probably) be reduced. The gender pay gap will increase as a result, which means segregation and discrimination will become more relevant as explanations (Löfström 2009).

Figure 6: Pay gap between women and men in unadjusted form in EU member states in 2008 (Source: European Commission 2011)

(17)

17 According to Löfström (2009), a gender pay gap may mirror not only differences in productivity and labour market gender segregation, but also discrimination and the fact that women bear greater responsibility for home and family, entailing longer periods of absence from the labour market and more part-time work. These lower earnings of women in female-dominated sectors and occupations cannot be explained by productivity differences between sectors and occupations. Löfström (2009), also questions this. Are wages low because these occupations or sectors are low-productive, or because these jobs are female dominated? When women ‘take over’ what was previously a male dominated occupation, the wage development seems to slow down. The wage gaps are highest in management positions where the educational background and work experience of women and men are very similar (OECD 2008). This suggests that pay differentials are partially due to discrimination, women may not be paid as much, in relation to their productivity, as men are (Löfström 2009).

As is argued by Stiglitz (1986), the net productivity of workers is a function of the wage rate they receive, which is also known as the theory of efficiency wages. The theory of performance related pay (PRP) is confirmed by Gielen et al. (2006), who find that PRP increases labour productivity at the firm level by about 9 percent. However, the theory of efficiency wages is not supported by the CPB (2004). They state that the effect is short term, temporary, and above all, inefficient.

The theory of efficiency wages might imply that due to the gender pay gap women are less productive than men. Moreover, according to Löfström (2009) removing all forms of pay discrimination and bringing about a less segregated labour market should result in a boost of women’s pay and so an increase in the labour supply. Chen (2004) also states that it is expected that women increase their labour supply due to an increase in opportunity costs of leisure.

This leads to the following hypothesis.

H.5 Removing the gender pay gap has a positive effect on economic growth.

2.1.6 Managerial functions

(18)

18 Greece, and Italy. Nevertheless, although representing a large share of management and professional positions in the United States, women still made up only 2 percent of Chief Executive Officers in the Fortune 500 Companies. For the whole OECD area, it is estimated that only 7 percent of the directors of leading companies are women. Women are also badly represented on the boards of major companies. Over 46 percent of OECD large firms even have no women at all on the board, while only 23 percent have more than one woman on the board. The number of female directors is highest in the Scandinavian countries of Norway and Sweden, and is lowest in Italy, Portugal, and Japan. North American companies are among the global leaders in terms of the percentage women on their boards, although women still represent less than 13 percent of board members in the United States and only 11 percent in Canada (OECD 2008).

With respect to the managerial positions of women in the European Union, the European Commission (2012) states that, in general, women have fewer opportunities than men to advance in their careers and that women’s skills are not being used to their full potential. Women account for 45 percent of the employed people across the European Union. Also, women accounted for around 56 percent of the people in tertiary education, and, for many years, account for a majority in tertiary level graduates. So with respect to education, women enter the labour market better equipped than men, but their level of representation declines in senior positions. In January 2012 women held on average just 13.7 percent of board seats of the largest publicly listed companies in the EU Member States. Moreover, in January 2012 only 3.2 percent of chairpersons were female, compared to 3.4 percent in 2010 (European Commission 2011a, European Commission 2012).

(19)

19 board member and the percentage of women as business leader for a selection of European countries in 2008.

Figure 7: Female board members in 2008 (% total) (Source: EC DG Justice)

Figure 8: Female business leaders in 2008 (% total) (Source: EC DG Justice)

Diversity at the top enhances the effectiveness of corporate leadership since diversity results in a better understanding of the complexities of the environment and more ingenious decisions. Homogeneity is believed to result in a narrow perspective while diverse top managers take a broader view (Robinson and Decant 1997, WWC 2006, OECD 2008, Hausmann et al. 2010). The European Commission (2011) adds to this that gender diversity brings a number of vital benefits to boardrooms, such as higher returns, better overall

(20)

20 performance, better risk management and greater employment of female talents. The European Commission (2012) also mentions some arguments in favour of diversity. According to recent estimates women control about 70 percent of global consumer spending. More women in management positions could help mirroring the market and so lead to market share gains. Moreover, diversity boosts creativity and innovation. A more diverse board results in better performance because decisions are based on evaluating more alternatives compared to homogenous boards. Lastly, employers that focus on diversity will be in a better position to tap into an increasingly educated and skilled segment of the talent pool (Catalyst 2004).

Valerio (2009) states that women leaders are good for business. As United States businesses expand into new markets, cultures, and workforces across the United States and around the world, the companies that integrate gender diversity into their business strategy prove to be more successful. Catalyst (2004) has shown that the Fortune 500 companies with the highest percentages of women corporate officers, experience, on average, a 35.1 percent higher return on equity and 34 percent higher total return to shareholders than did those with the lowest percentage of women corporate officers. The European Commission (2010) however state that 30 percent of women in top positions is the key threshold for improved performance. The European Commission (2012) states that studies from various countries show that companies with a higher share of women at top levels deliver strong organizational and financial performance. Amongst these studies is research from McKinsey & Company which shows that companies with the most gender-diverse management teams had 17 percentage-point higher stock price growth between 2005 and 2007 compared to the industry average and their average operating profit was almost double. In addition, in the absence of strong governance, female directors can exercise strong oversight and have a ‘positive, value-relevant impact’ on the company. The European Commission (2010) adds that more women in senior positions is the key to economic stability and growth.

All in all, this results in the following hypothesis.

(21)

21 2.1.7 Political functions

In political decision-making, women are outnumbered by men in all countries. In 2005 less than 25 percent of all parliamentary seats are occupied by women in the OECD area. In just nine countries women hold at least one-third of parliamentary positions, with Sweden having the highest rate of female representation at 45 percent. However, most OECD countries have a female representation rate of less than a quarter of seats and in the United States only 15 percent of all legislators are women. Worldwide it is estimated that only 16 percent of all legislative positions are occupied by women, and in many countries there are no female representatives at all (OECD 2008).

(22)

22 Figure 9: Women in parliament in 2006 (% total) (Source: WDI)

Figure 10: Female legislators in 2006 (% total) (Source: WDI)

There is a greater chance that the needs of all citizens are more closely reflected by policies when women do participate in governance. Just like in corporate leadership, women and men bring different perspectives to decision-making, and the effectiveness of state and its policies is limited when women are underrepresented. Moreover, its representational quality is diminished. Multiple studies show that when women are well-represented in decision-making bodies, the overall quality of governance rises and levels of corruption decrease

(23)

23 (OECD 2008). The European Commission (2010) adds to this that more women in decision-making would advance the implementation of equality legislation and policies that would in turn allow more women to work and on an equal basis with men. Moreover, in an editorial in the Economist, headed ‘Forget China, India and the internet: Economic growth is driven by women’ it was stated that: ‘(…) More women in government could also boost economic growth: studies show that women are more likely to spend money on improving health, education, infrastructure and poverty and less likely to waste it on tanks and bombs.’ (from Löfström 2009). The United Nations produces an index measuring the relative voice of women and men in public life called Gender Empowerment Measure (GEM)5. A positive and significant correlation between gender equality in terms of power and influence and GDP per capita in each respective EU country is found. So countries with more women in decision-making have a higher level of GDP per capita. This however, does not determine the causal relations. A higher level of GDP per capita might mean that more women choose to become politically active, or male resistance to women in politics declines as GDP grows. If the latter is true, it could also work vice versa if the ‘pie shrinks’. Causality could also run the other way around. More women in politics could help bring about reforms that enhance gender equality, which in turn impacts positively on GDP (Löfström 2009). The correlation between GEM and GDP per capita is illustrated in Figure 11 below.

All in all, this leads to the following hypothesis.

H.7 More women in political functions has a positive effect on economic growth.

Table 1 below summarizes all gender inequalities found in literature that potentially affect economic growth.

(24)

24 Figure 11: Gender empowerment and GDP per capita in EU member states in 2007 (Source: Löfström 2009)

Table 1: Gender inequalities potentially affecting economic growth

Gender Inequality: Source:

Labour participation Klasen (1997); Hochberg and Schmid (2005); WWC (2006); Löfström (2009)

Work duration OECD (1999); Evers et al. (2005); Nelen and De Grip (2009)

Education Hill and King (1993); Bloom and Williamson (1997); Klasen (1997); Dollar and Gatti (1999); Chen (2004)

Sectoral gender segregation Robinson and Dechant (1997); WWC (2006); Bettio et al. (2008); Löfström (2009); Hausman et al. (2010)

Wages Stiglitz (1986); Chen (2004); Gielen et al. (2006); Löfström (2009)

Managerial functions Robinson and Dechant (1997); Catalyst (2004); WWC (2006); OECD (2008); Valerio (2009); European Commission (2010); Hausmann et al. (2010); European Commission (2011); European Commission (2012)

(25)

25 2.2 General aspects of economic growth

In this section control variables that are suggested in the literature to also have an impact on economic growth are discussed6.

A number of authors have drawn attention to the importance of human capital for the theory of economic growth (Uzawa 1965, Romer 1986, Lucas 1988, Lucas 1990, Barro 1996, Easterly and Levine 1997, Helpman 2004, Heijdra 2009). They argue that human capital is the engine of economic growth. Economic theory suggests that the initial level of human capital matters for economic growth.

Moreover, the initial level of output matters for economic growth. Several authors discuss the effects of initial GDP per capita on economic growth (Barro 1991, Sachs and Warner 1995, Harrison 1996, Barro 1997, Easterly and Levine 1997). All authors find that the initial level of GDP per capita has a negative effect on economic growth.

The average rate of population growth is found to have a negative effect on per capita GDP growth (Kormendi and Meguire 1985, Mankiw et al. 1992, Kelley and Schmidt 1995, Bloom and Sachs 1998).

As a fourth control variable, several studies make notice of the effect of economic and political institutions on economic growth (Barro 1996, Acemoglu et al. 2001, Easterly and Levine 2001, Alfaro 2003, Helpman 2004, Rodrik et al. 2004). Economic and political institutions affect the incentives to innovate and accumulate, and they also affect the ability of countries to accommodate change. Barro (1991) and Sachs and Warner (1995) add to this the effect of political instability.

The next control variable that is mentioned by several authors is fiscal policy (Barro 1991, Sachs and Warner 1995, Barro 1997, Alfaro 2003). Barro (1991) uses government consumption to GDP as a measure for fiscal policy. Barro (1996) states that the growth rate is enhanced by lower government consumption.

6 For a more thorough overview of growth determinants see Durlauf, S., Johnson, P. 2005, ‘Growth Econometrics’, in Aghion, P., Durlauf, S., Handbook of economic growth, Elsevier, pp. 570-754.

(26)

26 The investment share of GDP is found to have a positive effect on economic growth (Barro 1991, Sachs and Warner 1995, Barro 1997). However, the relation between investment and economic growth is more complex, as we might have reverse causation.

As a final control variable, several studies have drawn attention to the effect of trade on GDP growth (Barro 1991, Easterly and Levine 1997, Dollar and Kraay 2003, Alcala and Ciccone 2004, Helpman 2004). Exports plus imports over GDP is used as a measure for this. Helpman (2004) states that there is no simple relationship between trade policies and growth. The effects depend on an economy’s characteristics. Barro (1996) states that the growth rate is enhanced by improvements in the terms of trade, which is measured as the ratio of export to import prices.

Nevertheless, it must be noted that the robustness of the relationship between some of these variables and GDP growth has been questioned by Levine and Renelt (1992).

3. Research methodology and statistical design

In this section the methodology used to test the hypotheses stated in Section 2 is presented. Due to data availability constraints, three time periods are used to test the hypotheses. Period 1 runs from 1993 until 2008, and covers 9 countries. Period 2 runs from 2003 until 2008 to include indicators on managerial functions, political functions, and institutions. Period 2 covers 19 countries. Period 3 runs from 2003 until 2006 to include indicators on wages and investment, and covers 18 countries. A complete list of the 20 OECD countries included is provided in Table A.2 in the appendix. As dependent variable GDP growth per capita is used. Potential difficulties are serial correlation and heteroskedasticity in the residuals (Hill et al. 2008). The standard errors are corrected using White’s diagonal coefficient variance method to end up with robust standard errors. A remedy for serial correlation is to use first difference models7. Another reason to use first difference models is to solve a problem with nonstationarity8. A drawback of first difference models is that the number of observations is

7 In the tables in Section 6 Empirical Findings, all models are tested for serial correlation in the residuals. I test for it using a method suggested by Wooldridge (2002, p. 283). It is a method for testing for serial correlation in first difference models. First the standard regression is ran. Then the residuals are saved and regressed on the lagged residuals. A positive and significant coefficient indicates serial correlation. It can be concluded that none of the regressions has problems with serial correlation, therefore the regressions are only corrected for heteroskedasticity.

(27)

27 reduced. In differencing there are T – 1 time periods for each i, rather than T (Wooldridge 2002).

The main model is given by:

Where y denotes the average growth rate of real per capita GDP.

The vector Xi,t consists of explanatory variables which indicate gender inequality and

consists of 7 sets of factors which affect per capita GDP growth. Table 2 below gives an overview of these sets and the variables which are used as indicator. x denotes the vector of

regression parameters that are to be estimated. i,t is a disturbance term. The goal of this

research is to find the sign and significance of the coefficients x after controlling for Zi,t, t,

and i.

Zi,t is the vector consisting of control variables that are suggested by literature to also

have an effect on per capita GDP growth. z denotes the vector of coefficients. The number of

control variables varies for each time period due to data availability constraints.

t captures the period fixed or random effects and i captures country fixed or random

effects9. The period effect parameter is included to control for country invariant time specific effects that might have an influence on per capita GDP growth. The country effects parameter is included to control for country specific time invariant effects10.

spurious regressions due to nonstationarity (Verbeek 2000, Baltagi 2008, Hill et al. 2008). The variables for Period 2 and Period 3 could not be tested for a Unit Root since the time periods are too short. Based on the results for Period 1 and the line graphs of the variables for Period 2 and 3 it was decided to also use first difference models for these time periods.

9 A Hausman test is employed to see which model is most appropriate.

(28)

28 Table 2: Indicators for factors affecting per capita GDP growth

Factors Indicators/ Variables

Explanatory

Labour participation (Set A) - Labour force participation female value over male value

Work duration (Set B) - Weekly hours worked female value over male value

- Percentage female part-time employment

Education (Set C) - Female to male students in primary education

- Female to male students in secondary education

- Lifelong learning female value over male value

Sectoral gender segregation (Set D) - IP-index

Wages (Set E) - Gender pay gap

Managerial functions (Set F) - Percentage female chairpersons - Percentage female board members - Percentage female business leaders Political functions (Set G) - Percentage female legislators

- Percentage female senior ministers - Percentage women in parliament Control

Population growth - Population growth rate

Institutions - Sophistication

Fiscal policy - Government consumption to GDP

Investment - Investment share of GDP

(29)

29 4. Data

In this section the data and variables are described. Table A.1 in the Data Appendix gives a complete description of the dependent, explanatory, and control variables, their abbreviation, and their sources. Table A.2 contains a complete list of the 20 countries included in the analysis.

The country and year coverage of the used datasets are the basis for the country and year selection for this study, which consists of the EU-15 and 5 other OECD countries. Having only 20 OECD countries in the sample could be a drawback. This is because there is a lot more variation in the data world-wide, than between these OECD countries.

The relevant selection of indicators is chosen based on data availability, which is known as a deductive approach. Appropriate indicators for all gender inequalities are identified, however, not every indicator is available for a long time span. Therefore, three time periods are tested. It could be a drawback that some indicators are not available for a long time span because it reduces the number of observations.

4.1 Dependent variable

The GDP growth data is obtained from the World Development Indicators from the World Bank. The annual percentage growth rate of GDP per capita is based on constant local currency. GDP per capita is based on gross domestic product divided by midyear population. Table 3 below gives the descriptive statistics of the dependent variable over the three time periods.

Table 3: Descriptive statistics dependent variable

Variables GDP 1 GDP 2 GDP 3

Mean 2,268 2,240 2,338

Minimum -4,479 -4,993 -1,601

Ireland '97 Estonia '08 Portugal '03

Maximum 10,385 10,772 8,257

Ireland '08 Estonia '06 Slovakia '06

Standard deviation 2,181 2,612 1,743

Coefficient of variation 0,962 1,166 0,745

(30)

30 4.2 Explanatory variables

Below the explanatory variables and their sources are described. Table A.3 in the appendix gives the descriptive statistics of the explanatory variables.

4.2.1 Labour Participation

The gender participation gap is measured as the female labour force participation rate over the male value and is obtained from the World Development Indicators. The labour force participation rate is the proportion of the population aged 15 and over that is economically active. This variable will be referred to as the LP variable. The variable is used to test H.1.

4.2.2 Work duration

To measure work duration, several variables are employed. The first variable is the female weekly hours worked over the male value and is obtained from the OECD iLibrary. It is measured as the average usual weekly hours worked in the main job. This variable will be referred to as the WHW variable.

The second variable is female part-time employment as a percentage of total part-time employment. It is measured as regular employment in which working time is substantially less than normal. However, definitions of part-time differ by country. The data is obtained from the World Development Indicators. This variable will be referred to as the PE variable.

Both indicators are used to test hypothesis H.2. 4.2.3 Education

To find the effect of gender inequality in education on economic growth three variables are used. The first variable measures the ratio of female to male students in primary education. It is defined as the percentage of girls to boys enrolled at primary level in public and private schools. The data is obtained from the World Development Indicators. This variable will be referred to as the EDUP variable.

(31)

31 The final variable on education measures the female percentage that participates in lifelong learning over the male value. The absolute value of the distance from 1 is used as indicator for gender inequality in lifelong learning. The data is gathered from Eurostat. It is defined as the percentage of population aged between 25 and 74 that participates in education or training. This variable will be referred to as the LLL variable.

The above mentioned variables are used to test H.3. 4.2.4 Gender segregation

The level of sectoral gender segregation is measured with the help of the IP index11. It can be interpreted as the proportion of the workforce which would need to change sectors in order to remove segregation, considering the female and male shares of sectors. It ranges from 0 to 0.5, were 0 represents complete equality and 0.5 complete inequality. To define sectors the NACE12 index is used. NACE OC0 and NRP are omitted from the analysis due to data

availability concerns13. The data is obtained from Eurostat. This variable will be referred to as the SEG variable. H.4 is tested using this variable.

4.2.5 Wages

The data on the gender pay gap is obtained from Eurostat, and is defined as the gender pay gap in unadjusted form in percentages. This variable will be referred to as the WAGE variable. This variable is used to test H.5.

4.2.6 Managerial functions

To infer the relationship between the position of women in management functions and economic growth multiple variables are used. The first variable is the percentage female

11 The IP index is defined as:

Where N represents the total number in employment, M the total number of males in employment, Mi the

number of males in sector i, and Fi the total number of females in sector i.

12 Available at http://circa.europa.eu/irc/dsis/employment/info/data/eu_lfs/lfs_main/lfs/lfs_statistical_ classifications.htm.

(32)

32 chairpersons. This variable will be referred to as the CP variable. The data is obtained from the European Commission database on women and men in decision making and is defined as chairman of the board of directors (or supervisory board in case of separated supervisory and executive functions).

The second variable is the percentage female members of board. The data is collected from the European Commission database on women and men in decision making and is defined as members of the board of directors (or supervisory board in case of separated supervisory and executive functions) and the count includes the chairmen. This variable will be referred to as the BM variable.

The final variable on managerial functions is the percentage female business leaders. This variable will be referred to as the MAN variable. It is defined as the number of persons whose occupation is recorded as being in category 121 (Directors and chief executives) or 13 (Managers of small enterprises) of the ISCO (see footnote 11) classification. This data is also obtained from the European Commission database on women and men in decision making.

The above mentioned variables are used to test H.6. 4.2.7 Political functions

The representation of women in political functions and its effect on economic growth is analyzed using three variables. The first variable is female legislators, senior officials or managers as a percentage of total, and the data is obtained from the World Development Indicators. This variable will be referred to as the FL variable.

The second variable is the percentage of women as senior minister. It is defined as members of the government who have a seat on the cabinet or council of ministers. The data is again obtained from the European Commission database on women and men in decision making. This variable will be referred to as the SEN variable.

The final variable on the political position of women is the proportion of seats held by women in national parliament and is defined as the percentage of parliamentary seats in a single or lower chamber held by women. The data is obtained from the World Development Indicators and the variable will be referred to as the PAR variable.

(33)

33 5. A first look at the data

In this section the data will be analyzed to get some feeling with the data. First a graph of the dependent variable from 1985 till 2010 will be analyzed. Second, the tentative relations of the explanatory variables with the dependent variables are illustrated in scatterplots. Finally, a potential problem with multicollinearity is analyzed with the help of correlation matrices of the explanatory variables.

5.1 GDP growth data

(34)

34 Figure 12: GDP growth per capita in constant local currency (Source: WDI)

(35)

35 Figure 12: continued -20 -15 -10 -5 0 5 10 15 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 GDP gr o wt h p e r cap ita (an n u al % ) Ireland Italy Luxembourg Netherlands Norway Portugal

Slovak Republic Spain

(36)

36 5.2 Scatterplots

To examine whether the hypotheses suggested in Section 2 at least show the intuitive relations, scatterplots of the explanatory variables with the dependent variable are presented. However, these scatterplots do not account for possible joint determination and cannot be used to infer a direction of causality. The number of scatterplots per variable differs due to the fact that not all variables are included in each time period.

5.2.1 Labour Participation

The relation between labour participation equality and economic growth is tested using H.1. Figure 13 shows the scatterplots for H.1. It is illustrated that H.1 is only partly confirmed. In Period 1 LP is negatively related to GDP, which is in contrast to H.1, however Period 2 and 3 show a weak positive relationship.

Figure 13: Scatterplots LP versus GDP in Period 1, 2, and 3

5.2.2 Work duration

H.2 is used to test the relation between gender equality in work duration and GDP growth per capita. Figure 14 shows the scatterplots. Again, the hypothesis is only partly confirmed. Period 2 and 3 show that WHW is positively related to GDP, which is in line with expectations. However Period 1 shows a negative relationship.

(37)

37 Figure 14: Scatterplots WHW versus GDP in Period 1, 2, and 3

Figure 15 shows the scatterplots which are also used to test H.2. The expected relationship is observed for Period 2 and 3. If women represent the majority of part-time employment, GDP growth is lower.

Figure 15: Scatterplots PE versus GDP in Period 1, 2, and 3

5.2.3 Education

(38)

38 Figure 16: Scatterplots EDUP versus GDP in Period 1, 2, and 3

Figure 17 shows the relation between EDUS and GDP. From the scatterplots for Period 2 and 3 it can be concluded that gender inequality is related with lower economic growth. This is again in line with H.3. Nevertheless, for Period 1 it is shown that gender inequality in secondary education is related with higher economic growth. It will be interesting to see if this relationship is significant.

Figure 17: Scatterplots EDUS versus GDP in Period 1, 2, and 3

The relationship between training and other lifelong learning and GDP growth is presented in Figure 18. The observed relationship in Period 2 and 3 is as opposed to expected. The scatterplots show that gender inequality in lifelong learning is related to higher economic growth. Period 1 shows a weak opposite relationship.

(39)

39 Figure 18: Scatterplots LLL versus GDP in Period 1, 2, and 3

5.2.4 Gender segregation

The relationship between sectoral gender segregation and GDP growth is shown in Figure 19. The observed relationship is for all periods as opposed to expected in H.4. This means that according to Figure 19 stronger sectoral gender segregation is associated with higher GDP growth per capita. So it is suggested that gender specialization goes along with a productivity gain. It will be interesting to see if this relation is also significant.

Figure 19: Scatterplots SEG versus GDP in Period 1, 2, and 3

5.2.5 Wages

Figure 20 is used to test H.5. The relation as shown by the scatterplot is as opposed to expected. The figure shows a positive relationship between a gender pay gap and GDP growth per capita. Again, it will be interesting to see if this relationship is also significant.

(40)

40 Figure 20: Scatterplot WAGE versus GDP in Period 3

5.2.6 Managerial functions

To illustrate the relationship between gender inequality in managerial functions and GDP growth Figure 21, 22, and 23 are used. All figures are used to test H.6. Figure 21 shows the relationship between the percentage of women as chairmen and GDP growth. The relationship is as expected for both periods.

Figure 21: Scatterplots CP versus GDP in Period 2 and 3

Figure 22 illustrates the relationship between the percentage of women as members of boards and GDP growth. Again, the relationship is as expected for both periods. A more equal gender distribution (that is, more women as board members) is positively related to GDP growth.

(41)

41 Figure 22: Scatterplots BM versus GDP in Period 2 and 3

Finally, Figure 23 shows the relationship between women as leaders of businesses and GDP growth. The scatterplots show no evident relationship between these two variables.

Figure 23: Scatterplots MAN versus GDP in Period 2 and 3

4.2.7 Political functions

Figure 24, 25, and 26 are used to test H.7. Figure 24 illustrates the relationship between the percentage of female legislators and GDP growth. Unfortunately, no conclusions can be drawn from the scatterplots.

(42)

42 Figure 24: Scatterplots FL versus GDP in Period 1, 2, and 3

Figure 25 is used to analyze the relationship between the percentage of women as senior minister and GDP growth. The scatterplot shows a negative relationship, which is as opposed to expected. It will be interesting to see whether the relationship is also significant.

Figure 25: Scatterplots SEN versus GDP in Period 3

Finally the relationship between the proportion of seats held by women in parliaments and GDP growth is analyzed using Figure 26. The result is as opposed to expected. According to the scatterplots PAR is negatively related to GDP.

Figure 26: Scatterplots PAR versus GDP in Period 2 and 3

(43)

43 From the above analysis it can be concluded that there is some evidence in support of H.3 and H.6. However, H.1 and H.2 are only confirmed for Period 2 and 3. For H.4, H.5, and H.7 the evidence is even as opposed to expected.

5.3 Contemporaneous correlations

In this section the correlation coefficients between the dependent variable and the explanatory variables are computed. The correlations are computed for the same time period for the explanatory variables and the dependent variable. Therefore, Table A.4, Table A.5, and Table A.6 in the appendix present the correlation matrices for, respectively Period 1, Period 2, and Period 3. Again, as with the scatterplots, these contemporaneous correlations do not account for possible joint determination and cannot be used to infer a direction of causality. However, these correlations do give an indication of the potential of the explanatory variables to explain patterns of economic growth. Moreover, the correlation matrix can be used to analyze multicollinearity. High levels of correlation could indicate a multicollinearity problem. When using strongly correlated variables, the coefficient estimates may change erratically in response to small changes in the model or the data. However, multicollinearity does not reduce the predictive power of the model as a whole, but coefficient estimates become unreliable. Therefore, the correlation matrix will be analyzed extensively14. Correlations above 0.6 will be interpreted as strongly correlated, and correlations below 0.6 as weakly correlated.

After analyzing the correlation matrices it can be concluded that multicollinearity will not impose serious problems for the models in Period 1 and 2 since the explanatory variables are not strongly correlated15. For Period 3, PAR – LP, BM – LP, and PAR –SEN are strongly correlated. This could impose a potential multicollinearity problem. For the specification of the regression models it will be seen if there are qualitative changes in the results when either variable is removed.

14 Another method of dealing with multicollinearity is performing a Principal Component Analysis (PCA). However, by using PCA, it is very hard to find which variables are significant. Therefore I choose to analyze the correlation matrix, and use this analysis for dealing with multicollinearity.

(44)

44 6. Empirical findings

6.1 Endogeneity

In growth regressions, reverse causation or regressor endogeneity is an important problem. If GDP per capita is regressed on some of the control variables over the same period, the resulting relationship could simply reflect an association between contemporaneous shocks to both. A considerable problem is that it does not reflect evidence of a causal relationship. Moreover, when these (potentially) endogenous regressors are included in the regression, the resulting estimator is inconsistent. That means that it does not converge to its true value even as the sample size becomes infinitely large. Therefore, all control variables are replaced by a suitable instrument. The instruments are the average of the preceding two years16 (data permitting). These instruments are considered both relevant and predetermined, as it is expected that they are correlated with its own value in the subsequent year, but not jointly determined with the subsequent GDP growth per capita.

6.2 Regression results

As already mentioned, appropriate indicators for all gender inequalities are identified, however, not every indicator is available for a long time span. Table 4 below gives an overview of which factors are covered per period.

Table 4: Factors covered per period

Factors Period 1 Period 2 Period 3

Labour participation   

Work duration   

Education   

Sectoral gender segregation   

Wages 

Managerial functions  

Political functions   

(45)

45 6.2.1 Period 1

The benchmark model includes all three control variables for this period. Different specifications of this model will be tested. Table 5 gives an overview of the regression results of the benchmark model for Period 1. The first column gives the OLS results, the second the Random Effects results, and the final column the Fixed Effects results.

Table 5: Comparison OLS, Random Effects, and Fixed Effects Period 1 Benchmark model Period 1

OLS RE FE Dependent Variable GDP GDP GDP POP -2,08 -0,782 -0,67 (-1,686) (1,199) (1,373) FP 0,781* 0,823** 0,789** (0,465) (0,371) (0,392) TRA -0,038 0,029 0,059* (0,035) (0,026) (0,032) Observations 135 135 135 R- squared 0,063 0,048 0,652 F - statistic 2,915 2,221 8,158 Prob (F - statistic) 0,037 0,089 0,000 Countries 9 9 9 *** p<0.01, ** p<0.05, * p<0.1 Standard errors in parentheses

A constant is included in each regression but not reported in the table

(46)

46 Table 6: Hausman test Period 1

Hausman test Period 1

Test summary Chi-Sq. Statistic Chi-Sq. d.f. Prob.

Cross- section random 1,365 3 0,714

Period random 1,561 3 0,668

Cross-section and period random 3,096 3 0,377

To test the hypotheses stated in Section 2 the benchmark model will be complemented one-by-one with the explanatory variables. Secondly, all explanatory variables will be included in a single regression to see whether the results are mutually exclusive. That is, to see whether the coefficients and significance of the explanatory variables change when all explanatory variables are included in a single regression. Table A.7 presents the regression results for Period 1. Column (1) presents the results for the benchmark model. It can be seen that the explanatory power for this model is quite low since only 0.05 percent of the variance in the dependent variable is explained by this model.

6.2.1.1 Labour participation

Column (2) in Table A.7 presents the results in relation to H.1. No evidence is found in favour of H.1 as there is no significant relation between LP and economic growth. However, the coefficient does have the expected sign. Column (10) in Table A.7 does show a weakly significant positive relation. Nevertheless, this final model is regarded as not robust, since the coefficients and their significance change drastically compared to the preceding specifications. Table 7 below compares the final model of all three time periods with the benchmark model for the same number of observations as the final model. It can be concluded that the drastic changes in the final model for Period 1 and Period 3 are not caused by the differing number of observations. Therefore, it does not alter the conclusions.

6.2.1.2 Work duration

(47)

47 Table 7: Benchmark model and final model equal observations Period 1, 2, and 3

Period 1 Period 1 Period 2 Period 2 Period 3 Period 3

Dependent Variable GDP GDP GDP GDP GDP GDP POP -0,446 -2,247** -1,326 -0,970 5,250 3,849 (1,040) (1,125) (1,995) (1,693) (3,925) (3,629) INS x x -0,679 -0,946 -4,330 -0,545 x x (5,940) (5,575) (9,767) (6,302) FP 0,204 0,326 0,332 0,293 0,927* -0,056 (0,448 (0,347) (0,683) (0,540) (0,526) (0,999) INV x x x x -0,462 -1,599*** x x x x (0,433) (0,423) TRA -0,004 -0,076*** 0,078 0,158 -0,153** -0,106 (0,045) (0,027) (0,110) (0,100) (0,057) (0,06) LP 0,246* 0,524* 0,726** (0,140) (0,304) (0,304) WHW -55,090*** 16,576 -4,043 (17,883) (20,244) (24,349) PE 0,111 0,210 -0,028 (0,116) (0,142) (0,147) EDUP -0,454 -0,298 -0,105 (0,282) (0,732) (0,625) EDUS 0,076 -0,006 0,070 (0,100) (0,144) (0,147) LLL 11,017*** -0,947 -12,811** (2,244) (3,558) (4,214) SEG -4,419 82,045*** -40,550 (19,316) (27,733) (27,048) WAGE -0,254*** (0,061) CP 0,071 0,012 (0,080) (0,086) BM 0,010 -0,170 (0,072) (0,113) MAN -0,033 -0,052 (0,065) (0,060) FL -0,141** -0,135* -0,597* (0,059) (0,074) (0,324) SEN 0,057** (0,023) PAR 0,022 0,165** (0,070) (0,071) Observations 112 112 92 92 50 50 R- squared 0,006 0,188 0,633 0,779 0,659 0,945 Countries 9 9 19 19 18 18

County effects None None Fixed Fixed Fixed Fixed

Period effects Random Random Fixed Fixed Fixed Fixed

*** p<0.01, ** p<0.05, * p<0.1 Robust standard errors in parentheses

(48)

48 6.2.1.3 Education

Column (5) till (7) in Table A.7 show the results for H.3. All coefficients on education are insignificant. However, the coefficient of EDUP does have the expected sign. More importantly, all coefficients are insignificant in these models. This could have been caused by the unbalanced data for these variables. Due to the unbalanced dataset it was not possible to include both random effects. Based on the Hausman test it was decided which random effect was included. The lack of either random effect could have worsened the specifications. This also holds for the final Column (10).

6.2.1.4 Gender segregation

The relation between sectoral gender segregation and economic growth is presented in Column (8) in Table A.7. A positive highly significant relation is found which opposes H.4. This result could have been expected from Figure 19 in Section 5.2.4. These scatterplots showed a positive relation between sectoral gender segregation and economic growth. This result suggests that, ceteris paribus, an increase in the sectoral gender segregation increases per capita GDP growth. Or, in other words, gender specialization goes along with a productivity gain.

6.2.1.5 Political functions

Column (9) in Table A.7 presents the result to test H.7. The result is insignificant and as opposed to expected.

6.2.1.6 Control variables

The results in Column (1) till (9) in Table A.7 confirm the findings by Levine and Renelt (1992) that population growth and trade are not robustly related to economic growth. However, in Column (10) both variables are significantly negatively related with economic growth. The coefficients are as expected, nevertheless, the results are considered unreliable because the coefficients and significance have changed drastically in this model compared to preceding models.

Referenties

GERELATEERDE DOCUMENTEN

FIGURE 2 | LPS stimulation of endothelial cells in vitro induces the formation of EC subpopulations based on E-selectin and VCAM-1 expression.. (A) Histograms of HUVEC as one

Given that the ICC was a major issue in last Kenyan presidential elections in 2013 and continued to be an emotive issue which precipitated the government/ opposition divide in view

De provincie Overijssel koos dus voor het stimuleren van burgerinitiatieven door middel van een wedstrijd om vervolgens de uitvoering van de meest kansrijke initiatieven

Two conditions required to apply option theory are that the uncertainty associated with the project is market risk (the value-in‡uencing factors are liquidly traded) and that

Er werd beoordeeld in week 38 en week 41 door 14 beoordelaars. De scheutvorming bij Sweet Chico + werd het hoogst gewaardeerd. De bladstand was bij alle rassen vlak. De bladkleur

In this case study, the functions of the animations were agnostic towards the number of elevators and floors, meaning they would only need to be created once for all

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded

In standard PWM strategy with the programmed switching frequency, the harmonics usually occur at fixed and well-defined frequencies and are thus named “discrete