Association for Public Policy Analysis & Management

Numerous studies have linked poor neonatal health, typically measured by low birth weight (LBW), with delayed cognitive outcomes in childhood. However, these studies have typically employed correlational methods that compare unrelated children who differ in birth weight (BW). This approach leaves open the possibility that it is not LBW that causes poorer cognitive outcomes, but rather some other factor—such as family disadvantage or maternal stress—that causes both LBW and poorer child outcomes.

Further, previous studies have not examined whether these associations between BW and child outcomes differ by socioeconomic status (SES). This is important because if the negative influence of LBW on child outcomes is stronger for low-SES children, this puts these children—who are already at greater risk for LBW and poorer developmental outcomes—at even greater disadvantage.

In this study, we are able to take a more causal approach to the question of whether neonatal health predicts poorer cognitive outcomes by comparing twins whose birth weights differ. Comparing outcomes among twins is a particularly strong test of effects because it not only eliminates shared family characteristics, but also eliminates the influence of important aspects of the prenatal environment shown to influence neonatal health and child outcomes. Data are drawn from the Early Childhood Longitudinal Study-Birth Cohort (ECLS-B), the only nationally representative sample that contains rich data on neonatal health and child outcomes among twins. We use these twin pairs in sibling fixed-effects models that control for unobserved family and prenatal characteristics that do not vary between siblings and that typically bias the association between BW and child outcomes in correlational OLS models, as well as child gender, which could differ among twin pairs. Further, we ran separate models using additional indicators of neonatal health, such as fetal growth rate (FGR), that may more accurately capture neonatal health, and thus its relationship with cognitive outcomes, than BW alone. Child outcomes included math and reading scores, measured when children were approximately 4-years-old.

Results indicate a positive within family effect of BW (in grams) on reading (B=0.001, p=.014) and math (B=0.002, p=.006) and of FGR on reading (B=0.053, p=.013) and math (B=.069, p=.003), indicating that even among a sample of twins that share family and prenatal environments, more robust neonatal health predicts better cognitive outcomes. Findings also demonstrated evidence of moderation in effects by maternal education, with stronger effects of BW, log BW, and FGR on most outcomes for children of less educated mothers. For example, a 10% increase in birth weight was associated with a .25 increase in math scores at kindergarten for children of mothers with less than a college degree, but only a .10 increase for those with college educated mothers, suggesting that children with poor neonatal health born into low-SES families may be doubly disadvantaged.