This study is the first to examine the determination of birth outcomes employing individual data at national level in the U.S. The data source is from the 1987 U.S. linked birth/infant death certificates. Our analytical framework is to estimate the infant health production function from a behavioral model in which health inputs are themselves choices. We place major emphasis on instrumental variables method of estimation to correct for the endogeneity bias. The effects of endogenous inputs, such as prenatal care, maternal age, and fertility, on birth weight are investigated. We specially focus on the analysis of the effect of prenatal care on birth weight controlling for endogeneity of prenatal care, maternal age, and fertility. We find that OLS underestimates the gains to prenatal care by a factor of 9.4 for blacks, about 3.1 for whites, and 4.8 for pooled blacks and whites sample. Black mothers who seek prenatal care one month earlier give birth to babies 124 grams more in birth weight while for white mothers the corresponding number is about 99 grams. The policy implications for this study are straightforward. The benefits for public prenatal care programs appear substantial, especially for blacks. Moreover, cost-effective approaches to improving birth outcomes through the expanded utilization of prenatal care will motivate women, who otherwise have received late or no care, to begin care in the early trimester.