This paper is concerned with the relationship between child mortality and the use of health care. The authors develop a simultaneous model of fetal and postnatal mortality (hazard processes) and input demand equations for prenatal medical care and institutional delivery (probit equations). This model is applied to retrospective data from Peninsular Malaysia covering 1950-88. The results show that prenatal medical care and institutional delivery have strong beneficial effects on child survival probabilities, and that these effects are substantially underestimated when adverse self-selection among users of health care is ignored. The effectiveness of prenatal health care in Malaysia improved until 1980, and then deteriorated. Fetal survival is not random with respect to unobservables, but ignoring it introduces only mild biases in infant and child mortality estimation. While mother-level heterogeneity plays an important role, the authors do not find evidence of effects common to her sisters. Higher infant and child mortality rates among young mothers are explained by their lower likelihood to purchase health care, and by their higher rate of premature deliveries. Finally, fetal mortality underreporting is greater for short pregnancies and for women with little education.