Abstract

This paper investigates the leave-out strategy of instruments by using the leave-out community ratio of household access to in-yard water sources and community water infrastructure as instruments for hours in fetching water time, and the data on disease symptoms. The results show that community-level access to clean water is significantly associated with both water-relevant and irrelevant disease symptoms, which suggests that the correlation between community-level access to clean water and child health is at least partially due to endogenous project placement potentially with respect to unobserved community wealth. The paper concludes that the OLS estimates have a potential endogeneity bias problem and that IV estimates under this strategy is subject to endogenous project placement and is not valid. A policy implication of this study is that careful attention should be paid to both self-selection and endogenous project placement in studying the effect of water accessibility on child health.