• No results found

The effect of gender inequality in education on economic growth in South-East Asia and the Pacific

N/A
N/A
Protected

Academic year: 2021

Share "The effect of gender inequality in education on economic growth in South-East Asia and the Pacific"

Copied!
34
0
0

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

Hele tekst

(1)

BSc Thesis Economics and Business

The Effect of Gender Inequality in Education

on Economic Growth

in South-East Asia and the Pacific

Name: Student ID: Specialization: Faculty: Supervisor: Date:

Charlotte M. L. van Leeuwen 11078200

Economics

Economics and Business Ms. N. J. Leefmans 31/01/2018

(2)

2 Statement of Originality

This document is written by Student Charlotte van Leeuwen who declares to take full responsibility for the contents of this document.

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

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

(3)

3 Abstract

Lately, gender inequality in a variety of areas has been an issue in many countries in this world. One of the areas that should be addressed in developing countries is the gender bias in education. This paper empirically investigates the effect of gender inequality in education on economic growth using yearly panel data between 1960 and 2010 for a set of 20 countries within South-East Asia and the Pacific. Results based on Ordinary Least Squares regressions conclude that the gender inequality in education has a significant positive direct effect and a significant negative indirect effect through the investment ratio, the life expectancy at birth and the labour force share on economic growth. However, the negative indirect effects outweigh the positive direct effect. As follows an increase the female-male ratio of average years of schooling of 1 percentage point will lead to an increase of 0.0033 percentage point in the gross domestic product per capita. This implies that decreasing gender inequality will result in an increase in economic growth.

(4)

4 Table of contents

page

1. Introduction

2. Literature review

2.1 Determinants of economic growth

2.2 Gender inequality as determinant of economic growth

3. Data and research method

3.1 The model

3.2 Selection of variables

4. Descriptive statistics

5. Empirical results and analysis

5.1 Results for the basic model

5.2 Results for the model including time, country and region dummies

6. Conclusion References Appendix 5 9 9 13 14 17 19 21 22 24 26

(5)

5 1. Introduction

In the year 2000, eight Millennium Development Goals (MDGs) were established by the United Nations with the overall purpose of reducing poverty (United Nations, 2015). The 191 member states of the UN and several international organizations committed to help achieve these goals by 2015. The third MDG was the promotion of gender equality and empowerment of women (United Nations, 2015). This goal is among other things the result of what Sen called the “missing women”, referring to the excess mortality of women due to gender inequality (Duflo, 2012). Since closing the gender inequality gap was on many states’ and organizations’ agenda, gender inequality did decline over the last decades (United Nations, 2015). However, the year 2005 was the earliest date to achieve the third Millennium Development Goal with its main target of ending gender disparities in primary and secondary education. This target has been missed in more than 60% of countries for which data was available (Unterhalter, 2006). Moreover, The Millennium Development Goals Report 2015 states that despite many successes, gender inequality persists. Women are still being discriminated and more likely to live in poverty compared to men.

Gender inequality should change in different areas (Kabeer, 2005) such as with respect to employment and politics. This paper will focus on the inequality in education. Various studies (Kabeer, 2005) conclude that more educated women are more likely to use contraception and antenatal care. It also increases the participation of women in a wider range of decisions, their access to resources and their role in economic decision-making. Furthermore, they seem less likely to suffer from domestic violence since they appear to be better at dealing with violent partners. Overall, access to education increases women’s well-being and capability to exercise control over their lives.

Besides closing gender gaps in education being a goal in itself, it can also be seen as an instrument to achieve other objectives such as other development goals. It has been argued that inequalities in education are significant concerns for human development and well-being (D Hicks, 1997). Cooray and Potrafke (2011) state that the increase in education for girls has a positive effect on children’s health and education through their educated mothers. Gender inequality in education also affects the economic development and growth of a country (Klasen, 2002; Cooray & Potrafke, 2011). Female education increases productivity (Knowles, Lorgelly & Owen, 2002) and human capital (Cooray & Potrafke, 2011), through which it affects economic growth.

(6)

6 Stotsky, Shibuya and

Kebaj (2016) examined the trends in gender equality and women’s advancement for the International Monetary Fund (IMF). One of their findings was that over a period from 1980 to 2014 the female to male ratio of gross secondary enrollment increased in all regions. As can be seen in figure 1, gender equality in education improved for Africa,

Middle East & Central Asia, and Asia & the Pacific, where for the latter the improvement was particularly large. In 2014, the ratio for Asia and the Pacific even exceeded Europe’s ratio. Governments in South Asia are very engaged and successful in advancing basic education (Unterhalter, 2006). If we split Asia and the Pacific up into South Asia (SA) and East Asia and the Pacific (EAP) we still encounter an increase in gender equality in education for both regions as can be seen in figure 2.

Graph constructed by author

Note: Graph based on the average of the available Gender Parity Indices (GPI) for that year of the countries by region Source: Gender Statistics, the World Bank (www.databank.worldbank.org/data/reports)

Graph conducted by Stotsky, Shibuya and Kebaj (2016) Note: female to male ratio for gross secondary enrollment using 5 year intervals from 1980 to 2014 Source: Trends in Gender Equality and Women’s Advancement; IMF Working Paper, February 2016

0 0,2 0,4 0,6 0,8 1 1,2 G e nde r P ari ty Inde x (G P I) Year

Figure 2. Secondary School Enrollment GPI for East Asia &

Pacific and South Asia

East Asia and Pacific South Asia

(7)

7 Besides the strong increase in gender equality in education, SA and EAP also experienced strong economic growth over the past years with an annual average of 3.69% for SA and 3.47% for EAP in the period 1990-2015. Figure 3 shows the time series of the GDP per capita for both regions.

Examining the relation between both variables will be the main goal of this paper. In particular, the research question is: What has been the effect of gender inequality in education on economic growth for South-East Asia and the Pacific in the period 1960-2010. If a negative effect can be shown, countries which still have high gender inequality in education such as the Middle East & Central Asia and Africa (see figure 1) could be politically incentivized to invest more in women’s education.

Several studies already examined the determinants of economic growth and the effect of education on economic growth. However, relatively few papers have examined the possible effect that gender inequality in education has on economic growth. The articles published on this matter are conducted based on different assumptions, data sets and methodologies and their results are rather contradictive. As most of these papers suggest, further studies on his topic should be done in order to get a better insight on what effect the gender inequality in general has and specifically the effect in education on economic growth. The most recent papers on this topic are based on data ranging from 1960 to 2000 for a wide variety of countries worldwide with the dependent variable being the economic growth rate. Our paper’s value added is that we will look at the effect of gender inequality on the differences in levels of GDP per capita on a yearly basis instead of on the growth rates based

0 10000 20000 30000 40000 Av g G D P pe r ca pit a Year

Figure 3. Average GDP per capita, PPP

(constant 2011 international $)

East Asia and Pacific South Asia

Graph constructed by author

Note: Graph based on the average of the available for that year GDP per capita, PPP (constant 2011 international $) of the countries by region

(8)

8 on 5 or 10 year intervals. We will also limit our study for the region South-East Asia and the Pacific over a longer period (1960-2010). By combining methodologies from past papers, previous findings will be investigated using Ordinary Least Squares (OLS) regressions.

The thesis is structured as follows. Previous literature and findings on the determinants of economic growth and on gender inequality in education and its relation to economic growth will be reviewed in section 2. In section 3 the research method and variables used in the regressions will be specified. The descriptive statistics will be outlined in section 4 and section 5 will analyze the results. In section 6 a sensitivity analysis will be conducted to check the robustness of our findings and section 7 will end the paper with a conclusion.

(9)

9 2. Literature review

Various papers have studied the relation and effects between gender inequality in education and economic growth. However, their results do not always seem consistent with each other. This might be due to differences in control variables, gender inequality measures, sample data and/or methodologies. An overview of previous studies and their findings will be discussed in this section.

2.1 Determinants of economic growth

The variables that affect economic growth have been widely studied. Often the models to determine the effects on economic growth are either based on the neoclassical growth theory (Romer, 2011) or on the endogenous growth theory (Barro, 2001).

The determinants of output per worker (growth) in the Solow growth model (based on the neoclassical growth model) are the level of physical capital, human capital and the level of technology. In the short run economic growth is also determined by conditional convergence, meaning that countries with a low initial real income tend to grow faster economically compared to rich countries. However, large variations in growth in the long run – both across countries and over time – only occur from an increase in change in technology. The latter is taken as exogenous in the model and could correspond to knowledge or to the education levels and skills of the labor force. In other words, differences in education could be the main driver of differences in economic growth in the long run according to the Solow growth model.

Barro’s (2001) endogenous growth models include various variables that theoretically can be linked to growth. In one of his studies he examined the role of human capital on economic growth. His research was based on cross-country panel data for 100 countries divided over 3 periods from 1965 to 1995 with initial level of GDP, government consumption as a percentage of GDP, rule of law, international openness, inflation rate, fertility rate, investment ratio, terms of trade and education being the exogenous variables.

2.2 Gender inequality in education as determinants for economic growth

Examining the effect of gender inequality in education as a determinant of economic growth has been studied to a lesser extent. Barro (2001) however did examine the latter as

(10)

10 well by differentiating between male and female school-attainment. He came to the conclusion that variation in education levels for men and women affect the economic growth differently. While secondary and higher levels of education have a significant positive effect on economic growth for men, there is no significant effect between education levels for women. He states that this is due to women not being well utilized in the labor markets in many countries. However, at primary level, male schooling is insignificantly related to growth whereas female primary education is significantly and positively related to growth. A previous cross-country study of Barro and Lee (1994) with data ranging from 1965-1985 for 133 countries reported that in regressions that include male and female years of schooling together, secondary female education is negatively related to economic growth whereas secondary male education is positively related.

These findings were criticized by Dollar and Gatti (1999), Knowles et al. (2002) and Klasen (2002). They find that these findings disappear once regional and time dummy variables are included in the regression. Moreover, Barro’s and Lee’s (1994) findings could be related to the problem of multicollinearity between male and female years of schooling. The correlation was measured to be above 0.9 between both variables.

Klasen (2002) avoids these problems in his research by taking the ratio of female average years of education over males average years of education as a measure for gender inequality. On top of that he includes region and 10 year interval dummies. His data set included 109 industrial and developing countries classified in 7 different regions and ranges from 1960 to 1992. In his conclusion he stressed the finding that gender inequality in education directly affects economic growth because of a decrease in average human capital. If the number of educated women increases, the productivity and the size of the labor force will boost and hence economic growth increases. There is an additional indirect effect through the impact that gender inequality has on population growth and investment. Higher human capital leads to increasing returns to physical investment which increases the possibility to invest more. More investment increases economic growth. Also, more gender equality in education reduces fertility rates which reduces population growth. This increases GDP per capita. Better promotion and a more balanced gender equality in education for the regions Sub-Saharan Africa, South Asia, and the Middle East and North Africa over 1960 to 1992, could have increased their annual economic growth rates by 0.9 percentage points faster. In Sub-Saharan Africa the payoff of promoting gender equality would have had the strongest effect on their economic growth compared to the other regions investigated.

(11)

11 Klasen and Lamanna (2009) updated previous studies of gender gaps in education on economic growth with more recent data and extended them to gender bias in employment. Their methodology follows Klasen (2002) but now they use more recent data (from 1960 to 2000). They concluded that not only the women who are discriminated towards education are harmed, but the entire society is (in developing countries). Their results again indicated a negative causal effect of gender bias in education on economic growth. Because of a sharp decrease in the gender gaps in education from 1980 to 2000 in South Asia and the Middle East & North Africa, the effect of gender inequality on growth became less harmful.

Just like Klasen (2002) and Klasen & Lamana (2009), also Knowles, Lorgelly and Owen (2002) used cross-country panel regressions to examine the relation in the long term. The only difference is that Klasen (2002) used growth levels of GDP per capita as the dependent variable and Knowles et al. (2002) used levels of GDP per capita. Knowles et al. (2002) examined the effect of gender inequality on economic development by including health and male and female human capital as separate explanatory variables in the neoclassical Solow growth model. Their measurement of gender inequality in education was based on average years of schooling of the population of 15 years and older from 1960 to 1990. This data originates from Barro and Lee’s data base from 1996 for 73 countries worldwide. Based on ordinary least squares regressions, they concluded that there is a significant positive long run effect of gender inequality in education on output per worker.

Dollar & Gatti (1999) studied the effects of gender inequality in education, health, society and marriage, and politics on economic growth. Based on cross-country panel data they found that gender inequality in education reduces economic growth. In line with Duflo’s (2012) findings, they empirically found that an increase in per capita income lead to improvements in different measures of gender equality and also vice versa. So all studies that examine the effect of gender inequality on economic development should check their outcomes for simultaneous causality and should (where needed) correct for this problem.

Seguino (2000) finds the opposite result in her cross-country panel regressions. Her sample contains data from the period 1975-1995 consisting of 20 semi-industrialized, lower- and middle-income countries in East Asia that rely on exporting sectors in which women provide the bulk of labor. She uses a derivation from a neoclassical production function with output as a function of capital stock and skill-adjusted female and male labor supply measured as human capital and technological progress. Human capital was measured by the average years of total and secondary education per female and male. Seguino (2000) came to the conclusion that gender wage inequality and growth are positively related. Through gender

(12)

12 inequality in education, cheaper female workers can be employed which boosts investments and in turn boosts economic growth.

While various studies have examined the effect of gender inequality in education, their results are contradictive in the sense that most studies based on worldwide cross-country data find a negative effect on economic growth while one study focused on 20 semi-industrialized countries finds a positive effect. This paper will differentiate itself from previous studies by focusing on a subset of countries in the region South-East Asia and the Pacific instead of looking at countries worldwide and will include yearly data instead of 5 or 10 year intervals. Also, while previous studies’ datasets ranged until at most 2000, this paper will extend the data with an additional 10 years, using data for the period 1960-2010.

(13)

13 3. Data and research methods

3.1 The model

This paper will examine both the direct and indirect effect of gender inequality in education on economic development for the regions South-East Asia and the Pacific. The direct effect will be measured by:

𝑙𝑛𝑔𝑑𝑝𝑐𝑎𝑝𝑖𝑡 = 𝛼1+ 𝛽1𝑙𝑛𝑔𝑑𝑝𝑐𝑎𝑝60𝑖+ 𝛽2𝐼𝑛𝑣𝑖𝑡−1+ 𝛽3𝑂𝑝𝑒𝑛𝑖𝑡+ 𝛽4𝐿𝐹𝑆𝑖𝑡+ 𝛽5𝐿𝐸 𝑖𝑡+ 𝛽6𝐸𝐷𝑖𝑡+ 𝛽7𝑅𝐸𝐷𝑖𝑡 + 𝛽8𝑋 + 𝜀 (1)

To test the indirect effects some additional tests and regressions will be run to measure the effect of gender inequality on life expectancy, the investment ratio and the labor force share respectively:

𝐿𝐸𝑖𝑡 = 𝛼2+ 𝛽9𝑅𝐸𝐷𝑖𝑡 + 𝜑 (2)

𝐼𝑛𝑣𝑖𝑡−1= 𝛼3+ 𝛽10𝑂𝑝𝑒𝑛𝑖𝑡+ 𝛽11𝐿𝐹𝑆𝑖𝑡+ 𝛽12𝐸𝐷𝑖𝑡+ 𝛽13𝑅𝐸𝐷𝑖𝑡+ 𝛾 (3) 𝐿𝐹𝑆𝑖𝑡 = 𝛼4+ 𝛽14𝐸𝐷𝑖𝑡+ 𝛽15𝑅𝐸𝐷𝑖𝑡 + 𝛿 (4)

In these equations the dependent variable lngdpcap represents the natural logarithm of the GDP per capita. The 𝛼’s are the constants of the regressions. The control variables in the regressions are the following: Inv is the ratio of investment over GDP per capita; Open is the measure for the openness to trade; LFS is the labor force share; LE measures the life expectancy at birth; ED represents the average years of education and X is a measure for regional, country and year dummy variable. Note that the dummy variable will only be used in the sensitivity analysis in section 6. RED is the variable of interest which measures the gender equality in education by the female to male ratio on average years of schooling. 𝜀, 𝜑, 𝛾 and 𝛿 are the error terms (Klasen, 2002). The second part of this chapter will provide a more detailed description about these variables. The subscripts i and t represent the country and the year respectively. Panel data for 20 countries from the regions South Asia (SA), East Asia (EA) and the Pacific (PAC) will be taken on yearly basis from 1960 to 2010. A list of the countries included can be found in Appendix 1. Based on panel data regressions and using

(14)

14 the statistical software STATA, OLS regressions will be conducted to test the effects and significance of regressions 1-4.

The total effect of gender inequality in education on economic development will be computed using path analysis (Klasen, 2002), i.e. taking the sum of the significant direct and indirect effects. The direct effect will be measured by coefficient 𝛽6. The indirect effect will

be measured by the direct effect of a control variable on the economic growth times the direct effect of the gender equality ratio, RED, on that control variable. The total effect of gender inequality in education on economic development will thus be measured as follows in case all the coefficients will have a significant effect:

𝛽7+ (𝛽9∗ 𝛽5)+ (𝛽13∗ 𝛽2) + (𝛽15∗ 𝛽4)

The variable of interest in this paper will be gender inequality in education, as measured by the female-male ratio of average total years of schooling of the adult population (RED). An increase in the variable RED thus measures a decrease in inequality as the female to male ratio increases. This paper expects RED to have positive direct and indirect effects on economic growth. As a result, the total effect is also expected to be positive, implying gender inequality to negatively affects the economic growth. This prediction is based on the results of previous studies on this topic (Klasen, 2000; Dollar and Gatti, 1999; Knowles et al., 2002). Data on growth, investment and openness are drawn from the Penn World Tables Mark 7.1. Labor Force Share (LFS) is computed by the labor force (population aged 15 and 65 years) over the total population. This data is obtained from the World Bank and the Penn World Tables. Data on life expectancy has been obtained by the World Bank as well. All data on education are drawn from the database constructed by Barro and Lee. Their database contains human capital data at five year intervals. For this paper’s analysis, an estimation on the yearly education levels has been conducted based on linear interpolation. A complete overview about the sources of the variables included in the regressions can be found Appendix 2.

3.2 Selection of variables

In our model, the dependent variable economic growth will be measured by the level of real GDP per capita corrected for purchasing power parity, using the chain index at 2005 constant prices in international dollars (I$). This measure or the growth rate of this measure

(15)

15 has also been used by Klasen (2000), Dollar and Gatti (1999), Seguino (2000), Knowles et al. (2002) and Barro (2001). Due to the large standard errors of GDP per capita compared to the standard errors of the other included variables, the natural logarithm of GDP per capita is taken. For more details on the reasoning behind taking the natural logarithm, see discussion in section 4 on the descriptive statistics.

According to the Solow Growth model (Romer, 2012) the economic growth of a country is positively related to the level of human capital, physical capital and productivity. The model also indicates that economic growth in the short run is determined by conditional convergence. Other studies who based their research on the endogenous growth theory found that besides human capital and physical capital, also determinants like the openness to trade significantly affect economic growth (Barro, 2001). Based on these findings, the some control variables will be added in the regression. A description is provided below.

The natural logarithm of GDP per capita in 1960 (lngdpcap60) will be included to test for conditional convergence. According to the neoclassical growth model, countries with lower initial GDP levels tend to growth at a higher rate compared to countries with higher initial levels of GDP.

The physical capital in a country will be represented by the average annual rate of investment as a percentage of GDP (Inv) of a country. Barro (2001) concluded that the investment ratio positively and significantly affects growth. To avoid the problem of simultaneous causality, lagged values (1 years prior) of the investment ratio will be used in our analysis.

The control variables labor force share (LFS), life expectancy (LE) and average years of education (ED) are included to represent human capital. LFS is calculated by taking the ratio of the labor force over the population of a country. The labor force (growth) has been included in several studies on economic growth as a control variable for levels of human capital such as Knowles et al. (2002), Klasen (2002) and Klasen & Lamanna (2009). An increase in the labor force share is expected to increase the economic growth. Dollar and Gatti (1999) include the average life expectancy as variable in their regression on economic growth. Life expectancy can be seen as a measure for the overall health in a country. Health in turn is expected to be positively correlated with human capital. A healthier workforce will be more productive and efficient which will lead to more economic growth. In our analysis, health will be measured as the years of life expectancy at birth. ED is measured by the average total years of schooling for total population. Education is seen as an important proxy

(16)

16 for human capital and is expected to have a positive relation with economic growth. This variable was also used in Klasen (2002), Klasen & Lamanna (2009) and (Seguino, 2000).

The openness to international trade of a country will be included in the regression and will be measured as the ratio of imports plus exports to GDP. International trade will lead to specialization in the product the country has a comparative advantage in (Krugman , Obstfeld and Melitz, 2015). So the more specialization, the higher its productivity, the higher its economic growth. This positive effect has been proven by Barro (2002). Hence, he concluded that the effect diminishes as a country gets richer up to a point that additional “openness” doesn’t contribute to the economic growth anymore. The degree of openness of a country is also included in several comparable studies (Klasen, 2002; Klasen & Lamanna, 2009).

Our variable of interest will be the female-male ratio of average total years of schooling of the adult population (15 years or older). The closer this ratio to 1, the more gender equality in education. The ratio rather than separate variables for male and female years of schooling is taken due to the multicollinearity problem that arises when using the latter variant (Klasen, 2002). To find the effect of gender inequality in education has on the economic growth be the core of this research.

(17)

17 4. Descriptive statistics

The tables in this section give a detailed overview about the variables used and the data availability for these variables. As can be seen in table 1, data on GDP per capita have very large standard deviations. By taking the natural logarithm, the standard deviations are reduced significantly. For this reason, instead of working with the change in GDP levels over time, the effect on the percentage change of the GDP per capita will be estimated.

Table 1. Summary Statistics gdpcap, lngdpcap

Table 2 shows the summary statistics of all variables used in regression 1. As can be observed, some variables are missing data (up to 40 out of 1020 observations). For Macao, data on growth, investment, openness and labor force share was missing for 1960-1970. The countries Fiji, Hong Kong, Indonesia, Macao, Nepal, Papua New Guinea and Singapore, lacked data on the investment ratio’s for 1959. For Taiwan, data on labor force share was missing from 1960 until 1977 and on life expectancy from 1960 until 1999.

Table 2. Summary Statistics lngpdcap, lngdpcap60, Inv, Open, LFS, LE, ED, RED

Source: Author’s calculations.

Table 3 shows the correlations between the different variables. A high correlation between the ratio of gender inequality in education (RED) and the average years of education

(18)

18 (ED) and between RED and the life expectancy at birth (LE) can be observed. So one should bear in mind that this problem of multicollinearity could bias the results obtained in this study.

Table 4. Correlations between variables

(19)

19 5. Empirical results and analysis

5.1 Results for the basic model

This section will present the results for both the direct and indirect effects that the gender inequality in education has on the economic growth based on regression 1.1, 2, 3, and 4. Further this section the overall effect will be computed. Appendix 3 and 4 display the results of the panel regressions. The dummy variables will be included in section 5.2 to check the robustness of the results of the basic model. Because the natural logarithm of GDP per capita is taken as the dependent variable in regression 1.1, the coefficients should be interpreted by the following way: a unit change in the independent variable z will change the GDP per capita with 𝛽𝑧%. So an increase of 1 in the average annual rate of investment as a

percentage of GDP would increase the GDP per capita with 0.00964 percentage points. The results of regression 1.1 are presented in table 5. As can be observed, most control variables have a significant positive impact on the GDP per capita. These results are consistent with the expectations discussed before and confirm the findings of previous studies. The outcome for the conditional convergence variable however is contradicting with previous studies and with the Solow Growth Model. For the countries from South-East Asia and the Pacific, it appears that low initial levels of GDP per capita lead to lower future economic growth.

Also our variable of interest RED is inconsistent with our expectations since it has a negative direct effect on GDP per capita with a p-value smaller than 0.001. The model predicts an increase of 1 in the female-male ratio of average total years of schooling of the adult population to lead to a decrease of 0.728 percentage points in the GDP per capita. This implies that gender inequality has a direct positive effect on economic growth.

Regressions 2 until 4 in Appendix 4 explain the indirect effects of gender inequality in education on economic growth. Regression 2 regresses RED on the average life expectancy at birth, LE. The variable RED seems to have a very significant positive effect on the life expectancy. Life expectancy on its turn positively affects the GDP per capita. Based on path analysis, this finding leads to an indirect positive effect of RED on GDP per capita of (𝛽9∗

𝛽5) = 44.95 * 0.0146 or 0.65627. This implies a negative indirect effect of gender inequality in education on economic growth.

(20)

20 Table 5. Results of direct effect of gender inequality in education on economic growth

(1.1) lngdpcap lngdpcap60 0.468*** (0.0793) Inv 0.00964*** (0.00143) Open 0.00175*** (0.000239) LFS 0.00991*** (0.00177) LE 0.0146*** (0.00320) ED 0.265*** (0.0120) RED -0.728*** (0.146) Constant 2.307*** (0.609) N 963

Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 Table constructed by author

The same reasoning applies to the indirect effect of gender inequality in education on economic growth through investment. Investment might also depend on the openness and the labor force share of a country, so these variables were included as control variables. Regression 3 shows that the direct effect of RED on the investment rate is highly significant. Through its effect on the investment rate, the indirect effect of gender equality on economic growth would be of (𝛽13∗ 𝛽2) = 31.79 * 0.00964 or 0.3064556.

Regression 4 shows the direct effect of ED and RED on the labor force share. As can be seen, the effect of RED is positive and very significant. The indirect effect through the labor force share is equal to (𝛽15∗ 4) = 9.918 * 0.0.00991 or 0.09828738.

Based on path analysis, all indirect effect of gender inequality in education on economic growth are significantly negative. By taking the sum of the direct effect and the

(21)

21 indirect effect that RED has on GDP per capita, an overall effect of -0.728 + 0.65627 + 0.3064556 + 0.09828738 = 0.33301298 can be concluded. The positive indirect effect of RED through the investment rate, the life expectancy at birth and the labor force share seems to outweigh the negative direct effect of RED on LNGDPCAP. This indicates that 1 percentage point increase in the female-male ratio on average years of schooling, will increase the economic growth by 0.0033 percentage points. Put differently, decreasing the gender inequality causes the economic growth to increase.

5.2 Results for the model including time, country and region dummies

To check the robustness of the results on the positive direct effect of gender inequality in education on economic growth in section 5.1, some dummy variables will be added or combined to regression 1.1. The East Asian financial crisis that hit in 1997 could have a regional negative effect on economic growth for those countries (Radelet and Sachs, 1998). That is why the 20 countries included in the analysis have been divided into 3 different regions: East Asia (EA), South Asia (SA) and the Pacific (PAC). The dummy is constructed in the following way: EA = 1 if country is situated in the region East Asia, EA = 0 otherwise; SA = 1 if country is situated in the region South Asia, SA = 0 otherwise. So these dummy variables explain the differences between the Pacific and the other regions.

Besides region dummies, we also look at the effect of country dummy variables. The crisis did not hit every country equally hard, so it could be that country dummies will lead to more significant results than the regional dummy variables. They explain the differences between Australia and the other countries.

Also yearly dummy variables will be taken into account because they also could potentially explain fluctuations in the economic growth due to the East Asian financial crisis and so lead to more accurate estimates for our variable of interest RED. The yearly dummy variables will explain the differences between the year 1960 and the other years. Data ranges from 1960 until 2010. The year 1960 is represented by y1, year 1961 by y2, year 1962 by y3 etc. until year 2010 is represented by y51.

These variables will be regressed individually and combined. The results can be found in Appendix 3 in the regressions 1.2-1.6. In regression 1.2 regional dummy variables are added to regression 1.1, country dummies in regression 1.3, year dummies in regression 1.4, region and year dummies combined in regression 1.5 and country and year dummies in regression 1.6.

(22)

22 As can be observed in Appendix 3, the country dummy variables have a highly significant effect on the economic growth whereas the region and yearly dummy variables contain very little significance. More important is to note that in all regressions, the gender inequality in education ratio has a very high significance with coefficients ranging from -0.739 to -0.696. We can conclude that the results about the negative direct effect of RED on LNGDPCAP of -0.728 from regression 1.1 in section 5 is robust.

(23)

23 6. Conclusion

Gender inequality in education is an issue that should be addressed in order to increase the overall well-being of women. This paper tried to give additional motives on why gender inequality should be reduced by finding out whether reducing inequality has a positive on the economic growth for countries in South-East Asia and the Pacific. If so, politicians would have additional incentives to increase schooling for girls in the Middle East, Central Asia and Africa since enrollment ratios in these regions are still very gender biased.

Based on the significant increase in gender equality in education along with proper growth rates in South-East Asia and the Pacific, the study will focus on 20 countries from these regions. This research is based on yearly panel data between 1960 and 2010 and finds a positive direct effect of gender inequality in education on economic development and a negative indirect effect through the investment rate, the life expectancy at birth and the labor force share. By taking the sum of both effects, an overall effect of 0.333 has been conducted. This means that for countries in South-East Asia and the Pacific, increasing the gender inequality ratio in education by 1 percentage point will lead to an increase of GDP per capita by 0.003 percentage point.

The results of the total effect of gender inequality in education slowing down economic growth in this paper is in line with the results of previous studies on similar topics. Klasen (2002), Klasen and Lamanna (2009), Dollar & Gatti (2001) and Knowles et al. (2002) also found a negative effect of gender inequality in education on economic growth. However, where this paper found a positive direct effect, some studies (Klasen, 2002; Klasen and Lamanna, 2009) found a negative direct effect. These differences in results could be due to our sample being limited to countries in South-East Asia and the Pacific compared to worldwide cross-country selection by other studies. Seguino (2000) also limited her study but to a set of semi-industrialized (mostly Asian) countries and implied a positive effect of gender inequality in education on economic growth. Seguino (2000) however also concluded that gender inequality stimulates investment, where our paper concludes the opposite.

We can conclude that one should properly reflect on the methodologies and the sample selection of the studies done so far on this topic. There is a clear difference in results based on a variety of countries worldwide and results based on only a subset of countries. A proper study on why these results are significantly different should be conducted.

Recommendations for further research would be to study the indirect effect of gender inequality on economic growth through the labor force participation rate. Due to a lack of

(24)

24 data, this was not performed in this paper. Researching this issue would be interesting to see whether higher education levels for girls will be translated in to a higher participation rate of women in society. This on its turn could have a significant effect on the growth of a country. Also similar research with a sample that includes different subsets of countries would be recommended to check whether that leads to similar results compared to previous studies on a wide variety of countries worldwide. Also one could update this paper using the lagged values for openness and RED to limit simultaneous causality problems.

Results suggest that one should continue promoting gender equality in education. Especially in Africa, the Middle East and Central Asia where gender inequality in education still persists. Besides being incentivized solely to boost economic growth, one should have more incentives to reduce gender inequality in education for the sake of women’s well-being and for the other consequences of more gender equality in education such as its positive impact on health, child education and mortality rates.

(25)

25 References

Aliber, M. (2003). Chronic Poverty in South Africa: Incidence, Causes and Policies. World Development, 31(3), 473-490. doi: 10.1016/S0305-750X(02)00219-X

Barro, R. (2001). Human Capital: Growth, History, and Policy. American Economic Association, 91(2), 12-17.

Barro, R. J., & Lee, J. W. (1994, June). Sources of economic growth. In Carnegie-Rochester conference series on public policy (Vol. 40, pp. 1-46). North-Holland.

Cooray, A. & Potrafke, N. (2011). Gender inequality in education: Political institutions or culture and religion? European Journal op Political Economy, 27(1), 268-280. doi:10.1016/j.ejpoleco.2010.08.004

Duflo, E. (2012). Women Empowerment and Economic Development. Journal of Economic Literature, 50(4), 1051–1079. doi: 10.1257/jel.50.4.1051

Dollar, D. & Gatti, R. (1999). Gender Inquality, Income, and Growth: Are Good Times Good for Women? World Bank Working Paper

Hicks, D. (1997). The Inequality-Adjusted Human Development Index: A Constructive Proposal. World Development, 25(8), 1283-1298.

Jacobs, J. (1996). Gender Inequality and Higher Education. Annual Reviews Sociology, 22(1), pp. 153-185.

Kabeer, N. (2005). Gender Equality and Women's Empowerment: A Critical Analysis of the Third Millennium Development Goal. Gender and Development, 13(1), 13-24.

Klasen, S. (2002). Low Schooling for Girls, Slower Growth for All? Cross-Country Evidence on the Effect of Gender Inequality in Education on Economic Development. The World Bank Economic Review, 16(3), 345-373.

(26)

26 Klasen, S., Lamanna, F. (2009). The Impact of Gender Inequality in Education and

Employment on Economic Growth: New Evidence for a Panel of Countries. Feminist Economics, 15(3), 91-132. doi: 10.1080/13545700902893106

Knowles, S., Lorgelly, P. & Owen, P. (2002). Are educational gender gaps a brake on economic development? Some cross-country empirical evidence. Oxford Economic Papers, 54 (1), 118-149.

Krugman, P.R. , Obstfeld, M. and Melitz, M.J. (2015), International Economics. Theory & Policy, 10th edition, Pearson Global Edition, ISBN-10: 1-292-01955-7, ISBN-13: 978-1-292-01955-0, Chapters 1-12.

Radelet, S., & Sachs, J. (1998). The onset of the East Asian financial crisis (No. w6680).

National bureau of economic research. Romer, D. (2011), Advanced Macroeconomics, fourth edition, McGraw-Hill, New York.

Seguino, S. (2000). Gender Inequality and Economic Growth: A Cross-Country Analysis. World Development, 28 (7), 1211-1230.

Sen, A. (1990). More Than 100 Million Women Are Missing. The New York Review of Books, 37(20)

Stotsky, J., Shibuya, S., Kolovich, L. and Kebhaj, S. (2016). Trends in Gender Equality and Women’s Advancement. IMF Working Paper

United Nations (2015). The Millennium Development Goals Report 2015.

Unterhalter, E. (2006). Measuring Gender Inequality in Education in South Asia. United Nations Children’s Fund

(27)

27 Appendix:

Appendix 1: List of countries used for analysis by region

East Asia (EA) South Asia (SA) The Pacific (PAC)

- China (CHN) - Fiji (FJI) - Hong Kong (HKG) - Indonesia (IDN) - Japan (JPN) - Korea (KOR)

- Macao, China (MAC) - Malaysia (MYS) - Philippines (PHL) - Singapore (SGP) - Taiwan (TWN) - Thailand (THA) - Bangladesh (BGD) - India (IND) - Nepal (NPL) - Pakistan (PAK) - Sri Lanka (LKA)

- Australia (AUS) - New Zealand (NZL) - Papua New Guinea

(28)

28 Appendix 2: Definition and sources variables

Variable Definition Data Source

lngdpcap The natural logarithm of per capita income over a

period between 1960-2010, in chain series and corrected for PPP, at 2005 constant prices (in I$)

Penn World Tables 7.1

lngdpcap60 The natural logarithm of initial GDP per capita in 1960

Penn World Tables 7.1

Inv Average annual rate of investment as a percentage of GDP over period between 1959-2009

Penn World Tables 7.1

LFS The labor share ratio measured by the ratio of the total labor force (population between the age of 15 and 64 years) over the population of a country over period between 1960-2010

World Bank (labor force)

Penn World Tables 7.1 (Population)

OPEN Average Openness as a ratio of GDP over period

between 1960-2010

Penn World Tables 7.1

LE Years of life expectancy at birth over period between 1960-2010

World Bank

ED Average total years of schooling for total population (15 years or older) over period between 1960-2010

Barro and Lee

RED Female-Male ratio of average total years of

schooling of the adult population (15 years or older) over period between 1960-2010

Barro and Lee

X Dummy variables for regions SA, EA and PAC,

countries and years

Note:

The Penn World Tables 7.1 were taken from Heston, Summers and Aten (2012) accessed on the website of the Rijksuniversteit Groningen (www.rug.nl/ggdc/productivity/pwt/pwt-releases/pwt-7.1)

Data on Labor Force were accessed from the database of the World Bank for all countries but Taiwan, for which the data was accessed at Statistics from Statistical Bureau (SSB) of the Republic of China (Taiwan) website (https://eng.stat.gov.tw) Barro and Lee data were taken from their website (www.barrolee.com/)

Data not available at World Bank for Taiwan, data found through different source for 1978-2010 (SSB Taiwan) Data not available at World Bank for Taiwan, so found on www.indexmundi.com/g/g.aspx?c=tw&v=30

(29)

29 Appendix 3: Regressions on economic growth

(1.1) (1.2) (1.3) (1.4) (1.5) (1.6)

lngdpcap lngdpcap lngdpcap lngdpcap lngdpcap lngdpcap

lngdpcap60 0.468*** 0.482*** 0.0148 0.590*** 0.621*** 0.0417 (0.0793) (0.0945) (0.0381) (0.0611) (0.0737) (0.0427) Inv 0.00964*** 0.00964*** 0.00931*** 0.0111*** 0.0112*** 0.00988*** (0.00143) (0.00144) (0.00141) (0.00159) (0.00159) (0.00154) Open 0.00175*** 0.00181*** 0.00143*** 0.00199*** 0.00199*** 0.00146*** (0.000239) (0.000239) (0.000240) (0.000248) (0.000248) (0.000252) LFS 0.00991*** 0.00827*** 0.0176*** 0.00496** 0.00354* 0.0173*** (0.00177) (0.00171) (0.00210) (0.00155) (0.00149) (0.00216) LE 0.0146*** 0.0155*** 0.0105** 0.00896* 0.00786 0.00587 (0.00320) (0.00321) (0.00319) (0.00423) (0.00428) (0.00419) ED 0.265*** 0.261*** 0.287*** 0.222*** 0.214*** 0.271*** (0.0120) (0.0120) (0.0122) (0.0145) (0.0143) (0.0158) RED -0.728*** -0.738*** -0.712*** -0.717*** -0.739*** -0.696*** (0.146) (0.147) (0.145) (0.151) (0.152) (0.150) SA 0.0528 0.0668 (0.325) (0.232) EA 0.186 0.319 (0.246) (0.177) BGD -1.160*** -1.280*** (0.108) (0.129) CHN -1.815*** -1.846*** (0.139) (0.143) FJI -3.257*** -3.265*** (0.264) (0.271) HKG -0.566*** -0.563*** (0.0778) (0.0807) IDN -0.800*** -0.886*** (0.115) (0.127) IND -0.856*** -0.977*** (0.112) (0.132) JPN -0.0619 -0.0472 (0.0518) (0.0540) KOR -0.742*** -0.747*** (0.0832) (0.0855) LKA -1.960*** -1.955*** (0.115) (0.119)

(30)

30 MAC 0.894*** 0.826*** (0.0840) (0.0939) MYS -0.729*** -0.764*** (0.0885) (0.0925) NPL -1.115*** -1.265*** (0.124) (0.151) NZL -0.601*** -0.582*** (0.0577) (0.0603) PAK -0.815*** -0.926*** (0.119) (0.135) PHL -1.034*** -1.094*** (0.0970) (0.105) PNG -0.165 -0.317* (0.0972) (0.130) SGP 0.221** 0.168 (0.0851) (0.0913) THA -0.421*** -0.497*** (0.108) (0.118) TWN 0 0 (.) (.) y2 0.000902 0.00389 -0.0296 (0.0919) (0.0922) (0.0880) y3 0.00954 0.0136 -0.0218 (0.0919) (0.0923) (0.0880) y4 0.0477 0.0529 0.0150 (0.0920) (0.0924) (0.0881) y5 0.0532 0.0596 0.0208 (0.0922) (0.0925) (0.0883) y6 0.0538 0.0614 0.0212 (0.0923) (0.0926) (0.0885) y7 0.0544 0.0635 0.0187 (0.0926) (0.0929) (0.0888) y8 0.0324 0.0431 -0.00647 (0.0929) (0.0932) (0.0892) y9 0.0471 0.0592 0.00487 (0.0933) (0.0936) (0.0896) y10 0.0600 0.0737 0.0143 (0.0936) (0.0940) (0.0901) y11 0.0844 0.0993 0.0362 (0.0941) (0.0944) (0.0906)

(31)

31 y12 0.0615 0.0777 0.0116 (0.0936) (0.0940) (0.0903) y13 0.0841 0.102 0.0295 (0.0941) (0.0944) (0.0909) y14 0.136 0.155 0.0777 (0.0946) (0.0950) (0.0915) y15 0.131 0.151 0.0699 (0.0952) (0.0956) (0.0922) y16 0.113 0.134 0.0485 (0.0959) (0.0962) (0.0929) y17 0.134 0.156 0.0662 (0.0965) (0.0969) (0.0937) y18 0.130 0.154 0.0606 (0.0972) (0.0975) (0.0945) y19 0.162 0.187 0.0894 (0.0979) (0.0983) (0.0954) y20 0.143 0.170 0.0709 (0.0985) (0.0989) (0.0961) y21 0.143 0.171 0.0678 (0.0992) (0.0996) (0.0969) y22 0.145 0.174 0.0671 (0.1000) (0.100) (0.0978) y23 0.139 0.169 0.0570 (0.101) (0.101) (0.0987) y24 0.144 0.176 0.0586 (0.101) (0.102) (0.0996) y25 0.164 0.197 0.0751 (0.102) (0.103) (0.101) y26 0.185 0.219* 0.0888 (0.103) (0.104) (0.102) y27 0.194 0.229* 0.0947 (0.104) (0.105) (0.103) y28 0.207* 0.244* 0.105 (0.105) (0.105) (0.104) y29 0.221* 0.260* 0.117 (0.106) (0.106) (0.105) y30 0.223* 0.262* 0.116 (0.106) (0.107) (0.105) y31 0.241* 0.281** 0.131 (0.107) (0.108) (0.106)

(32)

32 y32 0.227* 0.269* 0.115 (0.108) (0.109) (0.107) y33 0.243* 0.286** 0.127 (0.109) (0.109) (0.108) y34 0.259* 0.303** 0.140 (0.110) (0.110) (0.109) y35 0.260* 0.306** 0.140 (0.111) (0.111) (0.110) y36 0.247* 0.294** 0.127 (0.112) (0.112) (0.111) y37 0.251* 0.299** 0.129 (0.112) (0.113) (0.112) y38 0.257* 0.306** 0.133 (0.113) (0.114) (0.113) y39 0.212 0.262* 0.0833 (0.114) (0.115) (0.115) y40 0.258* 0.310** 0.122 (0.116) (0.116) (0.116) y41 0.275* 0.328** 0.138 (0.116) (0.117) (0.117) y42 0.255* 0.309** 0.113 (0.117) (0.118) (0.118) y43 0.266* 0.321** 0.119 (0.119) (0.119) (0.120) y44 0.268* 0.324** 0.119 (0.120) (0.121) (0.121) y45 0.282* 0.340** 0.132 (0.121) (0.122) (0.122) y46 0.284* 0.343** 0.132 (0.122) (0.123) (0.123) y47 0.297* 0.357** 0.144 (0.123) (0.124) (0.124) y48 0.322** 0.383** 0.165 (0.124) (0.125) (0.125) y49 0.319* 0.381** 0.159 (0.125) (0.126) (0.126) y50 0.323* 0.386** 0.154 (0.126) (0.127) (0.128) y51 0.374** 0.438*** 0.203 (0.127) (0.128) (0.129)

(33)

33

Constant 2.307*** 2.062* 6.524*** 1.838*** 1.513* 6.672***

(0.609) (0.858) (0.382) (0.434) (0.623) (0.404)

N 963 963 963 963 963 963

Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

(34)

34 Appendix 4: Indirect effects of RED on economic growth

(2) (3) (4) LE Inv LFS RED 44.95*** 31.79*** 9.918*** (0.819) (2.885) (1.963) Open -0.0232*** (0.00539) LFS 0.0405 (0.0362) ED -0.881*** -0.996*** (0.223) (0.138) Constant 32.54*** 8.476*** 7.279 (1.275) (1.912) (6.287) N 980 985 992

Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001

Referenties

GERELATEERDE DOCUMENTEN

The model will be able to predict dune evolution for high flow up to upper stage plane bed conditions, low flow where the bed and bedforms become partly immobile and

This finding is supported by Tabak and Koprak (2007) who state that a nurse in a higher position of authority will tend to dominate less experienced and younger

These voltages, given by G &amp; C C , will be relayed back to the power supply (depending on the switching topology) source via an intrinsic body diode that is present inside

In the long run, it will be to the distinct advantage of the South African society as a whole if a culture of respect for fundamental rights and the constitutional process of

The model results reveal the existence of stable equilibrium states with more than one inlet open, and the number of inlets depends on the tidal range and basin width (section 3)..

Uit de door mij gegeven voorbeelden van biomedische toepassingen van ”pleisters plakken” zal het duidelijk geworden zijn dat het werk in mijn onderzoeksgroep zich

In hoeverre bestaat er een verband tussen de gecommuniceerde identiteit en de gemedieerde legitimiteit van organisaties op social media en in hoeverre spelen het gebruik van

There are several measures to compute the readability of a text, such as Flesch-Kincaid readability[10], Gunning Fog index[9], Dale-Chall readability[5], Coleman-Liau index[6],