Effects Of Female Education On Economic Growth: A Cross .

Oztunc, Chi Oo, Serin / Effects of Female Education on Economic Growth: A Cross Country Empirical Studyhas been long concluded that education of women hasa positive effect on economic growth in all societiesespecially in developing countries, and called for moreattention on women’s education. She also explainedthat the net returns on education and training ofwomen are greater than that for men. Schultz (1961)who was the first economist, studied the difference inworker productivity by gender, based on their workexperience and schooling. Becker and Thomes (1994)also indicated that different wages among workerswas in line with the difference in their education level,training, and work experience. Bourguignon andMorrison (2002) examined how higher educationlevels lead to lower fertility rates by affecting percapita income growth and decreasing mortality rateper child. Barro (1991) revealed that more educatedhouseholds are likely to have higher productivity thanmore children; this shows that education and fertilityrate have a negative effect on each other. Barro (1991)also said that the economic growth rate is related tochanges in human capital and schooling years. Badenand Green (1994) however argued that women’seducation is not the main factor for improved childhealth and welfare and reduced fertility, and that it ismore effective to spend directly on child health andfamily planning to reduce fertility and child mortalityinstead of investing in educating women. Dollar andGatti (1999) also studied the relationship betweengender inequality in education and economic growth.They explained the positive relationship betweenthe education of women and economic growth byusing a five-year growth interval and by controllingfor possible endogeneity among education andeconomic growth with the help of instrumentalvariable estimation. Klasen and Lamanna (2008)also investigated the impact of gender inequality ineducation and employment on economic growth indeveloping countries through a 41-year period (1960–2000). He revealed that gender inequality in educationand employment can reduce economic growth, andalso stated that reducing gender inequality would leadto economic development because the education ofwomen has a huge impact on fertility and the creationof human capital for the next generation.Dollar and Gatti (1999) found a negative link betweenthe gender gap in education and economic growth, incontrast to Barro’s (1999) view that there is a positiverelationship between the two. This prompted Klasen(2008) to avoid using the same methods as the previousresearchers, to discover why they found differingrelationships between gender gap in educationand economic growth. Klasen (2008) showed veryclose links between gender inequality in labor forceparticipation and employment, and also found thatdecreasing marginal returns on education meansthat the education of women is limited to the lowerlevel. An increase in male education levels when themarginal return on the education of women is higherthan that for males, will affect economic performance.Many theorists believe that one of the most effectiveways to reduce the fertility rate per woman and thechild mortality rate—which will positively impacteconomic growth—is by promoting the educationof women as this would lead to education for thenext generation. Bloom and Williamson (1998)also examined education and economic growth byfocusing on “demographic gift”; how falling fertilityrates lead to a favorable demographic constellationafter 20 years. They also found that a woman’sincreased employment earnings provide her withgreater bargaining power within her household. Anincrease in women’s earnings also leads to greatersavings and economic growth. Bloom and Williamson(1998) came to this conclusion by using the shortterm growth model, unlike the other researchers whoused the long-term growth model. Therefore, Bloomand Williamson’s (1998) conclusions may differ fromthose of other researchers.Human Capital Theory expounds the view thateducation leads to higher skills among workersfor better productivity, and that the rate of returnon education investment surpasses that of otherinvestments. Human capital theorists use variousmethods to corroborate these views. First, theyexamine employee wages by comparing workers’education levels. Using the “normal” assumptionsof competitive labor and goods markets, humancapital theorists conclude that education levelsnot only help workers acquire skills for betterproductivity, but also encourage them to improvetheir abilities and attain greater earning power.Correspondence Theory also provides numerousimplications for poverty reduction policies as itis an effective anti-poverty strategy that should beincluded in education and skills training especiallyamong poor households. Studying data from ruralChina, Brown (2006) concluded that the educationof mothers—compared to that of fathers—has ahigher effect on the investment on education fortheir children. Using household survey data fromBrazil and Ghana, Thomas (1994) also found thatthe education of daughters is largely affected by theeducation of mothers, while the education of fathershas a significant effect on the education of sons.Controversially, Quisumbing and Maluccio (2003)had opposite results with their study in Ethiopia,which found that the assets brought to the marriageby the mother are more than those brought by the351

Educational Sciences: Theory & Practicefather the educational outcome of the daughters islower compared with the sons. They also conducteda similar investigation in Indonesia, and foundthat the assets of the mother have a positiveand significant effect on sons’ education, but nosignificant effect on daughter’s education, while amore educated father positively effects the educationof daughters. Kızılgöl (2012) also explored theimpact of gender inequality in education on povertyin Turkey. This paper indicated that educated, maleheaded households in urban areas have a lowerprobability of poverty compared to uneducatedfemale-headed households in rural areas. Moreover,Kızılgöl (2012) noted that increasing female-maleratio for literacy, for education of 10 years andabove as well as for earnings would lead to a higherprobability of poverty eradication.Using time series models with the help of economicvariables, Ince (2011) also studied how women’seducation is important for Turkey’s development.She noted that education is a significant factor forsocial and economic progress, especially in Turkey,as it can increase the welfare of its society. She alsoconcluded that education can be used as a predictorof human capital. Oxaal (1997) also considered therole of women’s education in reducing poverty, notingthat females from developing countries have lessopportunity for schooling than males. She also foundthat poverty forces children, especially girls, frompoor families to receive less opportunity for schooling.The opportunity cost of sending girls to school isvery great for poor households as the labor of girls isused to replace that of their mothers, for example inproviding childcare for their younger siblings.A panel data set has multiple observations on thesame econometric units. With panel data, eachelement has two subscripts; the group identifieri, and the within-group index denoted by t ineconometrics, which usually refers to time. Themost general linear representation of panel data is:yit kk 1 xkit βkit εit, i 1,2, , N, t 1,2, ,T (1)where N is the number of individuals, T is thenumber of periods.One set of data estimators allow heterogeneityacross panel units (and possibly across time) butconfines that heterogeneity to the intercept terms ofrelationship. There are fixed-effects and RE modelswhich impose restrictions on the above model ofthereby allowing only a constant to differ over i.In particular, it restricts the slope coefficients to beconstant over both units and time, and allows for anintercept coefficient that varies by unit or time. Fora given observation, an intercept varying over unitsresults in the following structure:yit xkit βkit ztδ ui εit(2)where is a 1xk vector of variables that vary overindividual and time, is the kx1 vector of coefficientson , is a 1xp vector of time-invariant variables thatvary over individuals, is the px1 vector of coefficientson , is the individual-level effect and is the disturbanceterm. The is either correlated or uncorrelated withregressors in and. If the is uncorrelated with theregressors, this is known as random effects (RE).RE estimators use the assumptions that the isuncorrelated with the regressors to identify the andcoefficients (Maddala, 2006).As our interest is in the impact of women’s educationon economic growth for Asia Pacific countries, theaforementioned literature has helped to inform theselection of unique explanatory variables for ourempirical study.To implement RE formulation of (2), it is assumedthat both and are non-zero processes, uncorrelatedwith the regressors; they are each homoskedastic inthat there is no correlation over individuals or time.For the T observations belonging to the unit of thepanel, one can assume the composite error processas follows:A Panel-Data Model: Random Effects Modelvit ui εitFor this study, the random effects (RE) model wasused to analyze the data set included in the paneldata processes. The RE model, also known as thevariance components model, considers the paneldata structure between dependent and independentvariables (Balestra & Nerlova, 1966; Baltagi, 2005;Hsiao, 2003). Using panel data for estimationensures control for missing and unobservedvariables and for the relationship identification ofcountry–specific effects (Arellano & Bond, 1991;Matyas & Sevestre, 1996).This is known as the error components model withconditional variance352cov(vit, vis) σu2 σε2 fort s(3)(4)and conditional covariance within a unit ofcov(vit, vis) σu2 for t s(5)cov(vit, vis) 0 for all t, s if i j(6)As the errors are correlated, generalized least squares(GLS) has to be used to obtain efficient estimates(Maddala, 1971). It can be written as follows:

Oztunc, Chi Oo, Serin / Effects of Female Education on Economic Growth: A Cross Country Empirical StudyβGLS θ σε2Tσu2 σε2(7)Thus the ordinary least squares (OLS) and fixedeffects model estimators are special cases of the GLSestimator with and , respectively. According to (7), if Tis large or is large relative to will be very close to zero,and the GLS estimator is very close to the fixed-effectsmodel estimator. In actual practice, it is not knownand needs to be estimated based on preliminaryestimates. Several methods have been suggested inthe literature (Fuller & Battase, 1973; Hauthakker,Verleger, & Shechan, 1974; Nerlove, 1971).Breusch and Pagan (1980) have developed alagrange multiplier (LM) test for , which maycompute the RE estimation. It is possible to estimatethe parameters of the RE model with full maximumlikelihood. The application of maximum likelihoodestimation continues to assume that the regressorsand are uncorrelated, adding the assumption thatthe distributions and are normal. The estimator willproduce a likelihood ratio test of corresponding tothe Breusch-Pagan test available for GLS estimator.Hausman test is usually applied to test for fixedversus RE models. If the regressors are correlatedwith the fixed-effects estimator is consistent but theRE estimator is not. If the regressors are uncorrelatedwith, the fixed-effects estimator is still consistent,albeit insufficient, whereas the RE estimator isconsistent and efficient. There are two scenarios inthe Hausman test: (1) If both fixed effects and REmodels generate consistent point estimates of theslope parameters, they will not differ meaningfully.(2) If the orthogonality assumption is violated, theinconsistent RE estimates will significantly differfrom their fixed effects counterparts (Baum, 2006;Hausman & Taylor, 1981). The Hausman test usesas a -statistic with degrees of freedom k where k isdimensionality of and estimators.MethodA panel regression is used based on theoretical andempirical literature that explores the link betweenfemale education and economic growth. A balancedpanel of 231 observations from 11 randomlyselected Asia Pacific countries over the 1990–2010period (21 years) is used in this study. The sampleof the countries represents all major regions in theAsia Pacific area, which reflects the characteristicsof female education on economic growth. Specialattention is given here to theoretical and empiricalliterature on how education as human capital affectsliving standards and economic growth as a societyin Bangladesh, Cambodia, China, India, Indonesia,Lao PDR, Malaysia, Myanmar, Philippines,Thailand, and Vietnam, where female education hasan impact on female labor force participation rates.A panel regression is used based on the studiesdone in the literature. There are fixed andrandom effects in a panel regression; however weuse and determine

Philippines, Thailand, and Vietnam. In addition, it emphasizes the impact of female education on economic growth as measured by GDP, literacy, fertility, and the female labor force. Using panel regression analysis, it is found that the fertility rate, female labor force participation rate and female primary school enrollment are