Most research on birth outcomes has found a direct relationship between appropriate prenatal care and increased birthweight. Researchers concede, however, that without a randomized design, which is clearly unethical, one cannot determine how much of the association is due to the medical intervention and how much is due to the characteristics of the women receiving the care. In short, the degree of selection bias is unknown and potentially substantial. In this paper we test for selection bias and estimate its direction and magnitude. We find that adjusted mean differences in birthweight between women who obtain intermediate as opposed to inadequate prenatal care substantially underestimate the effects of care that would be observed under random assignment. In particular, ordinary least squares estimates indicate that the gains to intermediate care are 113 grams for black infants, 76 grams for white infants and 92 grams for Hispanic infants. Under random assignment, black infants would experience gains of 130 grams, whites 234 grams, and Hispanics 183 grams. The gains for adequate as opposed to intermediate care are relatively minor. The results point to adverse selection in the demand for prenatal care.