Typically Unobserved Variables (TUVs) and Selection into Prenatal Inputs: Implications for Estimating Infant Health Production Functions

We use survey data, augmented with data collected from respondents? medical records, to explore selection into prenatal inputs among a group of urban, mostly unmarried mothers. We explore the extent to which several theoretically important but typically unobserved variables (representing wantedness, taste for risky behavior, and maternal health endowment) are likely to bias the estimated effects of prenatal inputs (illicit drug use, cigarette smoking, and prenatal care) on infant health outcomes (birth weight, low birth weight, and abnormal conditions). We also explore the consequences of including other non-standard covariates and of using selfreported inputs versus measures of inputs that incorporate information from medical records. We find that although the typically unobserved variables have strong associations with both inputs and outcomes with high explanatory power, excluding them from infant health production functions does not substantially bias the estimated effects of prenatal inputs. The bias from using self-reported measures of the inputs is much more substantial. The results suggest promising new directions for research on the production of infant health.