A growing economic literature has begun to focus on the effect of parental investments in child health in developed countries. However, this literature is not conclusive. Empirical work has concentrated on estimating the effect of a wide set of parental inputs comprising maternal health behaviours like smoking, alcohol consumption or diet. As most of these inputs are parental choices and we commonly do not observe all inputs relevant for the child production function, estimates on the effect of health inputs suffer from endogeneity bias. This paper explores the effect of smoking, alcohol consumption, exercise and diet during pregnancy on birth outcomes and considers the problem of identifying the causal effect of these endogenous maternal health behaviours. The analysis controls for a wide range of covariates and exploits sibling variation in the Danish National Birth Cohort. The paper compares different estimation strategies based on diverging identifying assumptions on the nature of the heterogeneity between families and parental response to child health outcomes. It acknowledges that prenatal resource allocation is a dynamic process, i.e. that parental preferences, perceptions about the ways in which child health is generated, and - for children of higher birth order - earlier children's outcomes will shape parental investments in child health.