Son-Preference, Gender Differentials in ChildLabor and Schooling, and EfficiencyAlok Kumar∗August 2011AbstractThis paper studies the effects of son-preference by parents on child laborand schooling in a model with bilateral altruism between parents and children.Results suggest that son-preference has differential effects on child labor andschooling. It leads to gender differential in child labor with female childrenworking more than male children. But, it does not lead to gender differen-tial in schooling. Only when parents cannot give bequests, female childrenreceive less schooling than male children. Binding bequest constraint resultsin inefficiently high level of child labor and low level of schooling. Reversetransfers (transfers from children to parents) in the second period result ininefficiently high level of schooling and low level of child labor, a result whichis in contrast to models of Baland and Robinson (2000) and Horowitz andWang (2004). The empirical evidence from rural areas of Bangladesh showsthat son-preference is an important factor explaining the gender differentialin child labor.Keywords: gender bias, son-preference, schooling, child labor, efficiency,bequests, MICSJEL Code: J22, I20, D60Address: ∗ Department of Economics, University of Victoria, Victoria,British Columbia, Canada, V8W 2Y2, Email: [email protected] IntroductionEmpirical evidence suggests that son-preference (parental gender bias in fa-vor of sons) is wide-spread in many regions of the world, particularly inAsia and Middle-East (Williamson 1976, Boserup 1980, Behrman 1988). Inrecent years, particularly due to spread of sex-selection techniques, a largeliterature has emerged which studies the socio-economic determinants andconsequences this bias. Previous works on son-preference have studied itseffects on fertility and sex-ratio (Ben-Porath and Welch 1976, Bloom andGrenier 1983, Leung 1991, Clark 2000), excess mortality among female in-fants (Das Gupta 1987, Sen 1990), and differential access to health (Chenet. al. 1982, Pande, R. 2003), nutrition (Sen and Sengupta 1983, Behrman1988, Hazarika 2000) and education (Behrman, Pollak, and Taubman, 1982,1986, Davis and Zhang 1995, Alderman and King 1998, Orazem and King2007).Empirical evidence also shows that the incidence and the intensity ofchild labor is higher for female children than male children.1Edmonds andPavcnik (2005) using UNICEF MICS (Multiple Indicator Cluster Survey)data find that the incidence of child labor female children (72.1 percent) ismuch higher compared to male children (64.8 percent). They also find thatfemale children are more likely to work long hours than male children. Theircalculation shows that 22.1 percent of female children worked 20 hours ormore per week. The corresponding percentage for male children was 19.4percent. Allais (2009) using SIMPOC survey data for sixteen countries findssimilar evidence.In this paper, I develop a model to study the effects of son-preferenceby parents on child labor, schooling, bequests, and welfare. Using MICSdata of rural Bangladesh for 2005-06, I provide evidence that son-preferenceis an important factor in explaining the gender differential in child labor.While the main contribution of this paper is to study the implications ofson-preference, I also analyze the effects of gender differential in earningsfunctions on child labor and schooling. This issue has been addressed byother papers most notably by Horowitz and Wang (2004). However, I deriveimportant new results.In the model, there are two periods. A family consists of parents and two1Child labor includes market and domestic work. But if we take only market workthen the incidence of child labor is higher among male children than female children (seeEdmonds 2007 for a thorough discussion).1children – one male and one female. Both parents and children are altruistic.Parents’ utility depends not only on their own consumption, but also on theutility enjoyed by their children. Similarly, children’s utility depends notonly on their own consumption and leisure, but also on the utility enjoyedby their parents. Children are endowed with one unit of time in the firstperiod, which can be allocated among three activities: labor, schooling, andleisure. Children incur disutility from both child labor and schooling. Ahigher level of schooling in the first period leads to a higher level of humancapital (earnings opportunities) in the next period. While parents care aboutboth children, they may put more weight on the utility of their male children.Parents choose levels of child labor, schooling, bequests and savings in thefirst period to maximize their utility. Children can also give transfers toparents (reverse transfer) in the second period.I distinguish between two cases: pure son-preference case and pure earn-ings function bias towards male case. In the pure son-preference case, Iassume that parents put more weight on the utility of male children, butearnings functions are identical for both male and female. In the pure earn-ings function bias towards male case, parents care equally about both maleand female children, but males have superior earnings function. This case issimilar to one analyzed by Horowitz and Wang (2004).In the model, I derive following main results. Firstly, when parents cangive bequests, both male and female children receive equal amount of school-ing, but female children work more than male children in the son-preferencecase. In the case of earnings function bias, not only male children workless, but also receive more schooling than female children. Secondly, whenthe parents cannot give bequests, in the son-preference case, male childrenreceive more schooling than female children and work less. However, in thecase of earnings function bias male children can receive more or less schoolingthan female children and can also work more or less than female children.Thirdly, in the son-preference case time allocated to leisure for female chil-dren is lower than male children. Alternatively, total time allocated to laborand schooling to female children is higher than male children. On the otherhand, in the earnings function bias case, the time allocated to leisure or totaltime allocated to labor and schooling is same for both male and female chil-dren. This implication is robust to whether bequest constraint binds or not.Finally regarding efficiency, allocations are efficient when parents can givepositive bequests. In the