1

Growth, Girls’ Education, and Female Labor:

A Longitudinal Analysis

Accepted for Publication in

The Journal of Developing Areas

November 4, 2005

Jane Arnold Lincove

LBJSchool of Public Affairs

University of Texas at Austin

(512) 232-2651

DRAFT

THANK YOU FOR YOUR COMMENTS.

Growth, Girls’ Education, and Female Labor:

A Longitudinal Analysis

Abstract

Economic growth is expected to lead to investments in human capital that expand labor markets. For women, the link between economic growth and labor market participation is more complex. Investments in female education are expected to increase women’s productivity at home, but the relationship to labor force participation is less clear. Research has identified a U-shaped relationship where women leave the labor market at early stages of economic development to focus on home production and return when a white-collar sector develops (Sinha, 1967; Durand, 1975; Pampel & Tanaka, 1986; Goldin, 1995; Horton, 1996; Mammen & Paxson, 2000; Juhn & Ureta, 2003). Previous studies do not account for the massive growth in female education that has occurred during the past thirty years. This study replicates previous models using time-series analysis and including consideration of increases in female schooling. While cross-sectional evidence suggests that both education and growth lead to initial reductions in female participation, fixed-effects estimations indicate that girls’ schooling has only positive effects on female participation over time. The results suggest that investments in female education can overcome potential reductions in female participation due to increases in wealth, and policies to invest in girls’ education appear to have benefits for labor markets, as well as family production.

Introduction

Economic growth is expected to increase human capital in developing countries through investments in education and expanded labor markets. Investments in female education are particularly important for transforming economic growth into positive development outcomes as female literacy is linked to fertility, child health, and expanded economic and political opportunities for women (Summers, 1994). While economic growth and female education have contributed to reduced fertility and improved child health, they have not resultedin universal expansion of labor market opportunities for women.

International comparisons illustrate that female labor force participation is high in low-income countries and highly-developed countries, and relatively low in middle-income countries, creating a U-shaped relationship between national income and female participation. Theorists have attributed this relationship to changes in labor market structure (Sinha, 1967; Durand, 1975; Juhn & Ureta, 2003), social norms regarding the nature of women’s work (Goldin, 1995), and cultural factors such as religion, social mobility, and family structure (Youssef, 1974; Semyonov, 1980; Horton, 1996). This study focuses on the important intervening effect of female education in the relationship between economic growth and female labor force participation. Theoretically, female education can increase productivity of women at home (Becker, 1975; 1985) or in the workplace (T.W. Schultz, 1960). In the past thirty years,developing countries have invested significant resources to expand female schooling (Hill & King, 1993; Herz & Sperling, 2004). Since girls’ education is a preferred policy strategy to promote development (Summers, 1994; T.P. Schultz, 2002; Herz & Sperling, 2004), it is important to examine how these investments influence labor markets. If female education results in greater productivity at home, investments may accelerate the decline in female participation. Alternatively, if female education increases labor market opportunities for women, investments may counteract the negative effects of initial economic growth. Previous cross-sectional studies have controlled for female education, but longitudinal analysis has not given adequate consideration tothe effect of changes in female access to schooling over time.

The relationship between education, women’s work, and economic growthhas profound implications for development policy. If the U-shaped curve is a typical development path, periods of economic growth may lead unexpectedly to reductions in female labor force participation. More interestingly, investments in girls’ education may lead to even more dramatic reductions in the size of the labor force if the benefits of education accrue mostly in the home. It is still unclear how households and labor markets respond to girls’ education. Only longitudinal analysis can determine if reductions in female labor force participation are a typical response to increases in female schooling.

This study adds to understanding of the role of women in developing countriesover time by integrating theories of education, labor, and growth in longitudinal analysis. First, previous cross-sectional and longitudinal models are replicated with added consideration of the time-lagged relationship between female education and workforce participation. Next, fixed-effects models are tested to determine if the relationshipsidentified in cross-sectional results reflect a development path over time. The results support previous conclusions about theU-shaped relationship between wealth and female labor force participation. However, controlling for country-specific fixedeffects, the negative effect of initial economic growth is small, and female schooling is only positively related to female labor force participation over time. If expansion of female education accompanies economic growth, developing countries will most likely see little or no decline in female labor force participation as growth progresses. Simulations suggest that low-income countries in particularly are likely to see increases in female participation when economic growth is accompanied by investments in girls’ education.

Theoretical Background

The U-Shaped Female Participation Curve

The U-shaped female participation curve was first suggested by Sinha (1967) who observed that very poor countries have high female participation, which tends to fall during early stages of development. Middle-income countries have relatively low female participation. At the other extreme, highly-developed countries also have high levels of female participation. Sinha attributed the downward portion of the U-shaped curve to a decrease in female agricultural work during shifts to urbanization and non-agricultural production. The upward portion of the U-shaped curve occurs when women return to the labor force at advanced stages of development to fill service sector jobs. Durand (1975) also attributes international differences in female participation to the gendered structure of labor markets, arguing that female participation can rise or fall during development depending on whether industries that employ women expand or contract. Citing Boserup’s (1970) theory of women in development, Pampel and Tanaka (1986) argue that women are excluded from early industrial jobs due to physical limitations, gender discrimination, and the domestic demands of large families. Early stages of mechanized farming and industrial expansion cause women to leave the formal economic sector to focus on domestic production until white collar jobs emerge that reward education. Importantly, women’s access to these jobs depends on their access to education during the intervening period of development (Boserup, 1970).

Goldin (1995)applied the U-shaped curve to family production, arguing that a social stigma excludes women from participating in industrial jobs. In Goldin’s model, industrialization causes a husband’s wages to increase. This income effect enables the wife to substitute home production for labor market participation. At later stages of development, the wife can access white-collar jobs that do not require heavy laborand offer wages sufficient to surpass both the stigma against female work and the benefits of her productivity at home. A shortcoming of this approach is the assumption that the social stigma is a global norm transcending cultural differences. Galor & Weil (1996) offer an alternative explanation, suggesting that women are excluded from industrial labor because of inferior physical endowments rather than social norms.

Although the U-shaped curve was originally observed in cross-sectional data from multiple countries, theorists speculate that this is a development path traveled by individual countries over time (Sinha, 1967; Pampel & Tanaka, 1986; Goldin, 1995). The U-shape consistently appears in multi-countrydata from the 1950s and 1960s (Sinha, 1967; Durand, 1975; Pampel & Tanaka, 1986) through the 1990s (Goldin, 1995; Mammen & Paxson, 2000; Juhn & Ureta, 2003). This indicates a persistent pattern, but does not prove that countries travel through the U-shape over time. Country-level longitudinal studies have observed several individual countries experiencing declines or increases in female participation in relation to labor market changes (Sinha, 1967; Durand, 1975; Horton, 1996; Juhn & Ureta, 2003). The most convincing evidence of a longitudinal path comes from Mammen and Paxson (2000), who use fixed-effects time-series modeling to illustrate that the U-shaped curve describes female participation over time in multi-country analysis, although the effect size does diminish compared to OLS regression models due to country-specific effects that are stable across time.

The Role of Female Education

Theories of the U-shaped curve rely implicitly or explicitly on economic growth being correlated withincreases in female access to education. Literacy is a prerequisite for the upward slope of the U-shaped curve where women gain entry into modern, white-collar jobs that reward education. Boserup (1970) argues that access to schooling is crucial to women’s capacity to benefit from modernization and gain access to rewarding jobs. Thus, the connection between education and labor market participation is central to whether development contributes to economic freedom for women.

Education has not been the focus of previous studies of female participation. Pampel and Tanaka (1986) and Goldin (1995) both control for female education in regression models. However, these authors use current female enrollment levels as a proxy for the education level of women in the labor market. If female enrollments are growing quickly, current enrollment rates should vary significantly across generations. Mammen and Paxson (2000) are able to correct for this problem using cross-sectional household-level data. In India and Thailand, they find that a woman’s educationalattainment is associated with the probability of white collar employment, suggesting that education may play a role in the upward portion of the U-shaped curve. Longitudinal and multi-country studies have yet to deal adequately with the effect of growth in female schooling over time.

The effects of female schooling are particularly important for policy-makers because promoting girls’ education is a central development strategy (T.P. Schultz, 2002). Unlike male education, female education is promoted as a policy to increase household productivity by reducing fertility and improving child health, as well as a strategy to build the labor force (Summers, 1994). For many countries,female education has changed dramatically during periods of economic growth (Hill & King, 1993). Female primary school enrollment rates in developing countries have increased from 52 percent thirty years ago, to 88 percent today.[1] Growth in individual countries from all regions has been very impressive. To cite a few examples, Algeria increased girls’ secondary enrollment from 6 percent in 1970, to 73 percent in 2000; Brazil increased from 26 percent in 1970, to full enrollment in 2000; South Korea increased from 32 percent in 1970, to 94 percent in 2000.[2] Importantly, increases such as these are often the result of explicit policies to increase girls’ education. For example, Mexico’s PROGRESA program, which provides stipends for school attendance, is credited with increasing girls’ primary completion by 15 percent; a scholarship program for girls in Bangladesh is credited with almost doubling female enrollment; and Indonesia has reached 90 percent enrollment for girls through a program with a focus on building new schools that meet the specific educational needs of girls (Herz & Sperling, 2004). Countries that have not invested in girls’ education have experienced limited growth (Herz & Sperling, 2004). Girls’ secondary enrollment remains below 20 percent in Pakistan, Cambodia, and many sub-Saharan African countries.[3]

Theories of Female Education and Female Labor

From a human capital perspective, gains in female schooling should be related to labor market opportunities for women, because employment opportunities create an incentive for parents to invest in daughters (T.W. Schultz, 1960). Actualeconomic growth appears to be associated with expanded educational opportunities for girls, regardless of female labor force participation. Despite gains in access to schooling, gender segmentation and wage discrimination continue to limit women’s opportunitiesin many countries (Celle de Bowman, 2000; Tzannotos, 2003). If female schooling increases during economic growth but female labor force participation decreases or remains very low, an alternative to the human capital perspective is needed. Theories of education and development offer three explanations that are independent of labor market opportunities. First, education for both sons and daughters may be a normal good, with investments increasing as wealth increases, regardless of future returns to schooling (Goldin, 1995; Galor & Weil, 1996). Second, countries may invest broadly in education for girls and boys to maintain political stability and support democratic institutions (T. W. Schultz, 1960; Wise & Darling-Hammond, 1983). Third, home production, rather than labor market productivity, may be the target of female schooling in some countries.

Theory and empirical work support the third hypothesis. Case studies reveal that school curricula and instructional techniques often reinforce women’s domestic role and inferiority to men (Smock, 1981; Rose & Tembon, 1999; Wynd, 1999). At the same time, schooling promotes literacy, analytical thinking, knowledge of hygiene, and other basic skills that have been shown to improve home productivity(Jejeebhoy, 1995). Lam and Duryea (1999) argue that with initial levels of female schooling, families get greater benefits from the production of high-quality children than from women’s wages. With secondary schooling, women’s market wages exceed the value of home production, and women enter the labor market. This theory implies that the relationship between female schooling and labor force participation is also U-shaped. With initial increases in primaryschooling, women in less-developed countries leave the labor market to specialize in home production. After a generation of girls accesses secondary schooling, women reenter the labor market when wages reward educational investments. Additional support for this approach comes from demographic research, which demonstrates that primary education for mothers can improve child health, but only secondary education creates labor market opportunities (Gille, 1985; Mason, 1993; Jejeebhoy, 1995). This theoretical approach differs from theories of the U-shaped curve by attributing decreases in participation to changes in women’s home productivity rather than structural changes in the labor market.

Lam and Duryea (1999) find support for this hypothesis using household-level data in Brazil, where women with primary schooling have dramatically lower birthrates than uneducated women, but do not participate in labor markets at high rates, despite increasing wages. With a secondary education, Brazilian women display both low fertility and high labor force participation. Behrman et al.(1999)found that education in India increased for both males and females during the Green Revolution, a time when labor market returns to human capital increased significantly for men but not at all for women. These authors speculate that women’s home productivity in educating sons was valued to such an extent that female education becamean important component of economic growth, despite women’s exclusion from labor markets. A similar study of human capital returns in Pakistan also found that women’s education can accrue as benefits to male relatives, even if women are excluded from the labor force (Fafchamps & Quisumbing, 1999).

If the primary function of female schooling is to increase the home productivity of women, expansion of female education may contribute to the downward portion of the U-shaped curve. Alternatively, schooling may also counteract the downward effects of economic growth if educated women reenter the labor market. As educational opportunities expand, the effects of female schooling may also be growing over time. Thus a complete model of the U-shaped curve integrates the effects of both schooling and growth.

Other Theoretical Influences on Female Participation

A final group of theories argues that social, cultural, and political factors influence female participation as much as economic factors. These theories focus on non-economic obstacles to female participation, which may vary with each country’s culturaland political institutions (Youssef, 1974; Semyonov, 1980; Horton, 1996). Both Islam and Catholicism are associated in theory with lower female labor force participation or delegation of women to specific employment sectors (Youssef, 1974; Smock, 1981; Nassar, 2003). Women’s economic freedom may also be associated with political rights. Political freedom is linked to economic rights and the ability to freely participate in labor markets (Sen, 1996). These controls are also included in the models in this study.

Methodology

Data & Variables

Based on the theories outlined above, an integrated model of female participation should include the effects of economic growth, education, labor market structure, and cultural variables. Data were compiled from the World Bank’s World Development Indicators (WDI) (World Bank, 2002), United Nations Women’s Indicator Statistics (WISTAT) (UN, 1999), and the International Labour Organization’s (ILO) employment statistics (ILO, 2000). Full data are available for 141 developed and developing countries from 1970 to 2000. Because of the intervals at which the ILO published employment data, longitudinal observations are available for the years 1990, 1995, and 2000. This is not a sufficient time-frame to see an individual country transition from low- to high-income, but individual countries at different stages of development should experience the decreases and increases in female labor force participation described by theory.

Female labor force participation is measured using the ILO’s definition of economic participation. An individual is considered economically active if she is employed or seeking employment in the formal economy. This measure can be particularly problematic for women because they may not be counted as employed if they participate in family agricultural production or small family enterprises (Pampel & Tanaka, 1986; Heward, 1999). The shortcomings of ILO data are balanced by the benefit of the survey’s impressive international coverage. This study uses ILO measures of female economic participation with the caveat that women’s labor force participation may be underreported, particularly in countries with a large informal sector.