Estimating the causal effect of maternal education on infant mortality with DHS data for Iran

S. Farrokh Mostafavi, Institute for Management and Planning Studies

In this paper I try to estimate the counterfactual causal effect of maternal education on infant mortality with DHS data for Iran. In order to obtain estimates of the causal effect, we need to control for certain variables such that the condition of ignorability holds. I try to identify and control these variables, which include biological and psychological characteristics of mother, her values, beliefs, and preferences, socioeconomic status, and her access to health services. Given that these variables are not measured, specific proxies are employed to control some of them. Mother’s education is defined as a dichotomous variable in order to minimize problems arising from unobserved heterogeneity. Logistic regression results indicate that the introduction of controls attenuate the coefficients of maternal education considerably, though they never become insignificant. Socioeconomic and community characteristics of mothers contribute to the observed strong statistical association. Problems in modeling and estimation are identified and discussed.