Closing the gender gap in education: What is the state of gaps in labour force participation for women, wives and mothers

The educational gender gap has closed or reversed in many countries. But what of gendered labour market inequalities? Using micro-level census data for some 40 countries, the authors examine the labour force participation gap between men and women, the “marriage gap” between married and single women’s participation, and the “motherhood gap” between mothers’ and nonmothers’ participation. They find significant heterogeneity among countries in terms of the size of these gaps, the speed at which they are changing, and the relationships between them and the educational gap. But counterfactual regression analysis shows that the labour force participation gap remains largely unexplained by the other gaps. E differences in education between men and women has been a priority of development organizations and the international community for many years. Pursued by institutions like the United Nations and the World Bank, the third Millennium Development Goal (MDG) aims to “eliminate gender disparity in primary and secondary education, preferably by 2005, and in all levels of education no later than 2015”.1 Equality of educational opportunities between men and women has also been acknowledged in the Universal Declaration of Human Rights of 1948 and in the 1979 United * SITE, Stockholm School of Economics and Harvard University, email: ina.ganguli@ hhs.se. ** Harvard University, email: [email protected]. *** The Graduate Institute, Geneva, and Harvard University, email: [email protected]. The authors gratefully acknowledge the support received from the Women and Public Policy Program and the Women’s Leadership Board at the Harvard Kennedy School, and from the Center for International Development at Harvard University. They would also like to thank Iris Bohnet, Claudia Goldin, Marcela Escobari, Victor Lavy and Lant Pritchett for helpful discussions. Responsibility for opinions expressed in signed articles rests solely with their authors, and publication does not constitute an endorsement by the ILO. 1 For a list of the Millennium Development Goals, see: http://www.unmillenniumproject.org/ goals/gti.htm [accessed 26 May 2014]. International Labour Review 174 Nations Convention on the Elimination of All Forms of Discrimination against Women.2 A key rationale for the above MDG is to contribute to the attainment of other development goals, such as economic growth, world poverty reduction, the eradication of disease and epidemics, and the improvement of other well-being indicators. According to the World Bank (2012), “there is no investment more effective for achieving development goals than educating girls”. The differences in education between men and women have now disappeared, or even been “reversed” in women’s favour, in almost all developed countries and in many developing countries as well, including among cohorts born over 60 years ago in several countries. The question, then, is whether these countries have also made progress in eliminating other inequalities facing women in the labour market, not only in terms of broader gender equality and women’s empowerment for economic participation, but also between married women and single women, and between mothers’ and non-mothers’ labour force participation.3 Also, how have developed countries fared relative to developing countries in this respect? Based on micro-level census data from some 40 countries, this article shows how these inequalities or “gaps” have changed over time. First, we document the closing of the gender gap in education and rank countries by the year in which this gap closed or reversed. We then turn to analysis of the gender gap in labour force participation, the “marriage gap” and the “motherhood gap”. These gaps still exist in many countries, although there is significant heterogeneity in terms of their size and the speed at which they are changing. Finally, we investigate the relationships between the gaps. While our cross-country research design does not allow us to identify causality between the gaps or the impact of economic development on them,4 it does provide a useful descriptive analysis of how they relate to one another. Despite the international community’s significant focus on reducing the gender gap in education and the remarkable reversal of the gap in so many countries, this is, to the best of our knowledge, the first study to use micro-level data for so many countries to study the state of these gaps and how they have changed over time. 2 Article 26 of the Universal Declaration of Human Rights stipulates that: “Everyone has the right to education ... and higher education shall be equally accessible to all on the basis of merit ... Education shall be directed to the full development of the human personality and to the strengthening of respect for human rights.” 3 We acknowledge the importance of other dimensions of women’s empowerment, such as political participation, earnings and property rights protection. Eliminating gender disparity in these areas is clearly necessary on grounds of human rights and, according to a large body of literature, to promote development (Bandiera and Natraj, 2011). In this article, we focus on labour force participation, which has often been regarded by international organizations as a first step towards improving women’s living standards, and as a mechanism for increasing income and economic growth. 4 See Bandiera and Natraj (2011) for a summary of existing studies on gender inequality and development, and their limitations. These authors suggest that most of the existing literature does not adequately address issues related to reverse causality and omitted variables. The gender gaps in education and labour force participation 175 Research background and rationale It has been suggested that educating girls and achieving the MDG on gender equality in education will lead to a range of improved economic and social outcomes for developing countries (Schultz, 2002). Evidence from recent empirical work, focusing primarily on developing countries, shows that there are indeed positive economic consequences and social externalities from improving women’s education. For example, increases in women’s education have been associated with reductions in fertility (e.g. Osili and Long, 2008), decreases in infant mortality and increases in life expectancy (see Dancer, Rammohan and Smith, 2008; Behrman and Deolalikar, 1988). Overall, it appears that educating women has beneficial effects on children’s health (see Glewwe, 2000), schooling, and adult productivity (Lam and Duryea, 1999; Strauss and Thomas, 1995). The beneficial effects are greater than those produced by the same level of education for fathers (see Schultz, 2002). Moreover, studies have shown that the benefits gained from expanding female education are far greater than the benefits gained from other public interventions, such as improving family planning service provision or increasing the number of physicians in the population (Kingdon, 2002). Empirical evidence also shows that gender equality in education leads to higher economic growth (e.g. Abu-Ghaida and Klasen, 2004), while low investment in women’s education leads to slower economic growth and reduced income levels (Dollar and Gatti, 1999; Klasen, 1999). Why would increases in women’s education produce improvement in other indicators of gender equality and socio-economic development? In an attempt to answer this question, we now discuss the theoretical reasoning behind the potential effect of narrowing the gender gap in education on women’s labour force participation generally, on the participation of married versus unmarried women, and on the participation of those with children versus those without children. Education and labour force participation The theory of human capital investment relates differences in earnings to differences in schooling, training, and other assets (Mincer, 1974; Becker, 1991). Specifically, it relates expected lifetime labour force participation to one’s incentive to acquire education and training. According to this theory, education increases the productivity and, thus, the earnings of individuals. And existing studies have indeed established a strong causal relationship between education and income at the individual level.5 Human capital theory explains why women have traditionally had fewer incentives to invest in education and training given their shorter expected labour force participation (Becker, 1992). Many factors have been put forward 5 At the macro level, however, the cross-country evidence linking education to growth is inconclusive (Pritchett, 2001 and 2006). The factors that may explain the discrepancy between macro and micro analyses include measurement error in macro regressions and issues related to the quality of education. International Labour Review 176 to explain why this has changed over time and why the gender gap in education has closed and even reversed in many countries. In the case of the United States, for example, the factors include the technological change that led to the use of mechanical power rather than human energy in workplaces (Galor and Weil, 2000), the rapid expansion of the service sector (Becker, 1992), the alleviation of household chores through the use of piped water, electricity and appliances (Killingsworth and Heckman, 1986; Greenwood, Seshadri and Yorukoglu, 2005; Goldin, 2006), greater ability to control pregnancies through access to birth control technology, and lower “effort costs” of college preparation and attendance for girls than for boys (Goldin, Katz and Kuziemko, 2006). The increase in divorce rates and the decline in family size have also been suggested as correlates and possible determinants of the reversal of the gender gap in education. Higher educational attainment has allowed women to increase their expected lifetime labour force participation, shifting their time horizon. Accordingly, their expected benefits from an increased investment in education have led them to plan careers instead of “having jobs” and being the “secondary worker” within the household (Goldin, 2006). Indeed, research shows that returns to women’s education are in many countries either equal to or higher than those accruing to men. Consequently, the opportunity cost of staying at home increases with women’s level of educational attainment. It is important to acknowledge the fact that cultural, social and political factors may also affect women’s labour force participation, beyond economic forces. Specifically, social barriers (idem, 1990), cultural factors (e.g. I . lkkaracan, 2012; Beblo and Ortlieb, 2012), particular views on the role of women in society (e.g. Folbre, 1994) and within the household (Badgett and Folbre, 1999), or formal restrictions (e.g. Boserup, 1970) may limit women’s labour supply, even in situations where labour market outcomes and returns to education are expected to be high. Marriage and labour force participation Married women’s labour force participation has been the subject of much debate in the literature. Mincer’s (1962) model was the first that tried to explain the apparent puzzle of the increase in married women’s labour force participation in spite of steady increases in men’s real earnings in the United States. According to this pioneering work, household income and wages have two effects on female employment status. On the one hand, the income effect suggests that wages are negatively related to labour supply: as wages increase and income goes up individuals may decide to spend less time working in the labour market and devote more time to leisure. On the other hand, the substitution effect suggests that as wages increase, the opportunity cost of not working also increases, thereby prompting individuals to increase their labour supply. Mincer found the positive substitution effect of general wage increases on women’s employment to be greater than the negative income effect through the husband’s wage, with the difference being large enough for increases in women’s wages to explain much of the increase in married women’s labour force participation. The gender gaps in education and labour force participation 177 Goldin (1990) shows how changes in income and wage elasticities can affect women’s labour force participation, while Blau and Kahn (2007) examine trends in married women’s own-wage and cross-wage elasticities. They find that in the United States from 1980 to 2000, married women’s own-wage elasticity fell by about half and that their responsiveness to their husbands’ wages declined by about 40 per cent. Heim (2007) also examines married women’s labour supply elasticities in the United States, over the period 1979–2002, finding that their wage and income elasticities for both participation and hours decreased dramatically in absolute terms over the period. Juhn and Murphy (1997) examine women’s employment and earnings at different points of their husbands’ wage distribution from 1959 to 1989 and find an increasingly weak relationship between women’s labour force participation and their husbands’ earnings. They also confirm a positive relationship between women’s wages and labour force participation, casting doubts on the idea that married women’s labour force participation increased to compensate for slow growth in their husbands’ earnings. In addition to economic factors, Fernandez (2007) shows how changes in culture and social norms, as part of a rational intergenerational learning process, may have contributed to the increase in married women’s labour force participation. Motherhood and labour force participation In the neoclassical model of the family, women tended to specialize in domestic production, including child rearing and traditional household activities, whereas men were the primary earners and specialized in formal production through labour market activities (Mincer, 1962; Becker, 1981 and 1991). Moreover, even among women in the labour force, the early years of child rearing were associated with a decrease in working experience and wages. However, the neoclassical model was found to be inadequate in explaining the rise in married women’s employment, so subsequent models focused on other determinants of mothers’ labour force participation (e.g. Bowen and Finegan, 1969; Smith and Ward, 1985). First, the expansion of education appears to be related to a decline in fertility rates (Caldwell, 1982; Galor and Weil, 1996 and 1999) and postponement of motherhood (see Strauss and Thomas, 1995, for a review of the literature). According to Caldwell (1982), education influences fertility by increasing the cost of children and the perceived importance of investing in children’s human capital, leading women’s optimal fertility choices toward “higher quality” (Becker and Lewis, 1973; Willis, 1973). Becker, Murphy and Tamura (1990) show how higher levels of human capital lead to an increase in the returns on human capital, as well as fewer children. Galor and Weil (1996) relate increases in women’s wages to the decline in fertility, which, in their theoretical framework, is a consequence of the increase in the cost of children more than the increase in household income. Several recent contributions to the literature have tried to establish a causal relationship between education and fertility, addressing issues of reverse International Labour Review 178 causality and possible omitted variables (e.g. Breierova and Duflo, 2004; McCrary and Royer, 2011; Duflo, Dupas and Kremer, 2010). Second, advances in household production technology have reduced the time required for domestic production. In addition, advances in medical technology have reduced the time budget associated with childbirth and parenting, while effective birth control allows women to determine the size and timing of their families and to respond rationally to market signals (Barker and Feiner, 2004). Third, the shift of economic activity towards the service sector, the consequent introduction of scheduled part-time employment, which is more “female friendly”, and the increase in returns to education have led to growth in the labour force participation of married women and mothers (Goldin, 2006; Goldin, Katz and Kuziemko, 2006). This too has increased educated women’s opportunity cost of staying out of the labour force (Attanasio, Low and Sánchez-Marcos, 2009). Expected relationship between education and labour force participation gaps Based on the literature reviewed above, we would expect an increase in women’s educational attainment to be associated with increased labour force participation – both among women generally and among wives and mothers6 – and with lower fertility rates and smaller wage gaps between men and women. In other words, empowering women with more education should increase their economic participation and give them greater access to economic resources, leading to improvements in many other spheres. Following previous empirical analyses (mainly country-specific studies or studies focusing on a limited number of countries), we expect to observe a U-shaped relationship between women’s participation and development, as documented by Goldin (1994) and Mammen and Paxson (2000). When women have low levels of education, they are employed in low-skilled, often manual activities for which a social stigma exists, but when they are educated, they find employment in more socially acceptable labour market activities. There has so far been no large cross-country analysis of the reversal of the educational gender gap and the inequalities attaching to women’s labour force participation. Accordingly, the remainder of this article provides evidence on the dynamics of these gaps, relating them to factors such as levels of GDP, urbanization and labour market rigidity. 6 We acknowledge that there may be social norms which prevent women from working outside the household, in which case an investment in women’s education could not increase their labour force participation. Moreover, gender inequalities can be multidimensional. For example, Sundaram and Vanneman (2008) find that in Indian districts featuring a higher share of women in employment, girls’ literacy lags further behind boys’. Their suggested explanation is that in this context, where more mothers work outside the household, more girls have to work to help the family. It would also be important to examine women’s participation by sector with a focus on employment status categories. Due to data limitations, we are unable to do this here (for an interesting analysis on India, however, see Huisman and Smits, 2009). The gender gaps in education and labour force participation 179 Data The primary source we used to construct our data set is the International Integrated Public Use Microdata Series, IPUMS-International (Minnesota Population Center, 2009). This is the largest publicly available compilation of individual-level census data, consisting of decennial records of individuals and households. Our analysis is thus based on micro-level data for a selection of 40 countries in Africa, Asia, Europe and Latin America, as available, drawing upon the two most recent waves of the census for each country.7 In most cases, the interval between the two waves of the census is ten years. Appendix table A1 lists the countries in our analytical samples and the census years for each country. Our analysis focuses on the cohort of women and men aged 35–44 in order to concentrate on individuals who have finished their studies and made their decisions on marriage, fertility and labour market participation. Using the census data for each country, we measure several gender gaps, namely: • the education gap, defined as the difference in average years of education between men and women; • the labour force participation gap, defined as the share of men who are in the labour force minus the share of women who are in the labour force – i.e. employed or unemployed and job-seeking – or, where expressly noted for the purposes of our analysis, as the ratio of those shares;8 • the marriage gap, defined as the difference in employment rates between women who are married (or living in cohabitation) and single women or, where expressly noted for the purposes of our analysis, as the ratio of those rates; • the motherhood gap, defined as the difference in employment rates between women without children and women with three or more children.9 7 We acknowledge that our sample is biased towards developing countries. However, we were unable to add more developed countries by relying on other data sources because aggregate statistics consistent with the variables and demographic groups we use are not available. 8 We use the employment status variable (EMPSTAT) from the IPUMS data set. Although the variable has been harmonized to make work status comparable across countries and over time, some differences remain in regard to terms of classification and reference group. A detailed description for each country and census wave is available at: https://international.ipums.org/internationalaction/variables/EMPSTAT#comparability_tab [accessed 26 May 2014]. 9 We also calculate an alternative measure of the motherhood gap defined as the difference in employment between women with children and women without children. Other studies have examined the “family gap” by looking at the difference in earnings between women with children and women without children, but all of them have focused on developed countries, where the average number of children is lower (e.g. Sigle-Rushton and Waldfogel, 2007; Anderson, Binder and Krause, 2002; Waldfogel, 1998; Harkness and Waldfogel, 1999). Our measure based on three or more children is intended to reflect the higher number of children born to mothers in developing countries, which make up the majority of our sample, so that we have a more comparable share of women in each group. However, robustness checks showed that the basic results do not change depending on which definition of the motherhood gap we use in terms of the number of children. International Labour Review 180 We supplement the measures constructed from the IPUMS census data with country-level variables from the World Development Indicators, including measures of GDP per capita and urbanization rates. We also include measures of labour market rigidity extracted from the World Bank’s (2011a) Doing Business data set. The education gap Figure 1 shows the year of birth of the first cohort for which the education gap was closed according to the most recent available wave of the census. In most countries, this wave was administered in the early 2000s (see Appendix table A1). Of the 40 countries in our sample, 27 have closed the gap, while men still averaged more years of schooling than women in the remaining 13 countries at the time of the census. The first country where the gap was closed is Belarus, for the cohort born in 1945, and the last two are South Africa and Romania, for the cohort born in 1975. In the United States, the gap was closed for the cohort born in 1956. Interestingly, ten developing countries closed the gender gap in education earlier than the United States.10 In countries where the gap was closed in the 1950s, two generations have already experienced gender equality in years of schooling, while in those countries where the gap was closed in the 1970s, only one generation has experienced such equality. Most developed countries and some developing countries have not only closed the gender gap in education, but also significantly reversed it in favour of women (see Hausmann, Ganguli and Viarengo, 2009). Except for Austria, all of the countries where the gap remains in men’s favour are developing countries. As table 1 shows, however, the gap at age 25 is still very wide in 11 of the 13 countries that have yet to close it, ranging from 0.6 to 2.1 years of schooling, but much smaller in Austria and Mexico, at 0.1 years. The gap has closed in Viet Nam, and it has reversed significantly in Mongolia, Portugal, the Bolivarian Republic of Venezuela, Argentina, Brazil and Greece, where 25-year-old women have between 0.8 and 1.2 more years of schooling than men. In Latin America, the gap has been reversed in most countries (eight out of ten). Only Mexico and Bolivia exhibit a persistent gap disfavouring women, still significant in the case of Bolivia. Figure 2 relates women’s average years of education to the current size of the gender gap in years of education. Interestingly, several of the countries where the gap has reversed are not among those where women have the highest average levels of education (e.g. Argentina and Brazil). In Rwanda and Guinea, where women have the lowest average levels of education, the gender gap is between 1.2 and 2.3 years of schooling. In countries where the 10 Admittedly, the United States has had a more complicated history in this regard: between 1900 and 1930, women had roughly the same level of education as men (Goldin, Katz and Kuziemko, 2006), then the gap widened and subsequently closed again for the cohorts born in the 1950s. None of the other countries in our sample displays such a reversal over time. The gender gaps in education and labour force participation 181 197