'May You Be the Mother of a Hundred Sons': Gender Inequality in Child Mortality in India and its Proximate Determinants

Half a million girls a year are sex-selectively aborted in India (Jha et al., 2006); many others never reach their fifth birthday due to infanticide or sex-selective neglect. The Indian government has tried to curb discrimination against girls by adopting different policies, such as banning sex-selective abortions. However, based on the 2001 census, the ratio of girls to boys aged less than six years old keeps decreasing. Even putting moral considerations to one side, the importance of curbing this trend is considerable. Social unrest, violence, kidnapping of women and the like have been documented to occur whenever the sex ratio becomes too skewed among young adults (Hudson and Den Boer, 2002). It should, therefore, be a priority for the Indian government to halt these social practices leading to discrimination against and to the elimination of female foetuses and female toddlers. In this thesis, it is hypothesised that parents perform a cost-benefit analysis while deciding the allocation of vital resources among their children. This cost-benefit analysis is based on many factors. The economic system, the cultural system including, notably, religion, and household characteristics are believed to have an impact on parents’ willingness to discriminate against some of their daughters. Parents, however, have limited resources and, therefore, cannot implement perfectly their desire to keep, or not, all their children. The focus of this thesis is on the impact of religion on gender inequality in child mortality and its proximate determinants, namely, height-for-age z-score, weight-for-age z-score, vaccination, treatment/advice for diarrhoea and access to oral rehydration salts. Chapters two and three discuss the factors which lead parents to discriminate against some of their children; chapter two is devoted to religion while chapter three summarizes other key arguments from the literature. After discussing the data—mainly from the National Family and Health Survey (NFHS)—and the model specification in chapter four, chapter five uses a model-based approach to determine empirically what leads some parents to discriminate against their daughters. Using different techniques, that is, OLS, logit, Poisson, the Heckman correction for sample selection and the Blinder-Oaxaca decomposition, this chapter reaches some interesting conclusions. First, Indian girls are clearly more at risk of dying during infancy, relative to boys, than would be expected based only on biological differences. The difference in mortality between boys and girls seems to be caused mainly by differences in access to health care. We also conclude that the Sikhs are the religious group with the highest level of gender inequality among their children. One other interesting conclusion of this chapter is that even if we find evidence of selection bias in the data in the case of vaccination, correcting for this bias does not change markedly our qualitative results. In other words, even after accounting for the fact that some girls may have been eliminated before we can collect data on them, our qualitative results do not change. Chapter six also tries to explain gender inequality in child mortality and its proximate determinants but, this time, using a design-based approach, that is, by taking into account the complex survey design used by the NFHS. We conclude that a design-based approach is more suitable for studying gender inequality in child mortality and its proximate determinants than the more widely used model-based approach. The reason is simply that it is impossible, in the context of survey data, to control for all factors affecting children‘s health and access to health care. We also demonstrate that the inclusion of state level and individual level variables in the same regression can lead to misleading results. In addition to illustrating the misspecification problem potentially widespread in the literature, and the hierarchical structure of the research question, this chapter also reports some interesting results concerning gender inequality among children and religion. The data show that the impact of being Muslims or Christians is the same for boys and girls in terms of survival, nutritional outcomes and access to health care, ceteris paribus. In the case of Sikhs, gender inequality appears in terms of children being underweight, with girls achieving poorer nutritional outcomes than boys, ceteris paribus. With new technologies allowing parents to know the gender of their unborn child early during pregnancy, parents do not need to rely any more on post-birth solutions to eliminate any undesired child. Stated differently, it has never been easier for parents to implement their preference for sons. Chapter seven, therefore, looks at the stated son preference of both men and women. We conclude that economic motives significantly influence the preference of women who have never married but this is not the case for never married men. This chapter focuses mainly on never married men and women due to a potential endogeneity problem, that is, a two way relationship between the sex ratio of children and stated son preference. Using instrumental variable techniques, we conclude that this concern is not warranted. However, it is also clear that the determinants of stated son preference vary with marital status. Our results show, notably, that economic motives become important for men once married, while belief on the acceptability of domestic violence become important for women once married. In terms of religion, we conclude that Christians have the lowest son preference, followed by Hindus and Muslims. The results also show that the Sikhs are the religious group with the highest level of son preference in India among currently married men. This thesis shows that religion is a factor in predicting whether a child lives or dies. A way forward would be to involve religious leaders in fighting the problem. The different initiatives taken by the Sikh community, for example, by publishing the Sikh Reht Maryada, are steps in the right direction