Modeling and Measuring the Genetic Determinants of Child Development

The longstanding debate over whether human capabilities and skills are shaped more by "nature" or "nurture" has been revitalized by recent advances in genetics, particularly in the use of polygenic scores (PGSs) to proxy for genetic endowments. Yet, we argue that PGSs embed not only direct genetic effects but also indirect environmental influences, raising questions about their validity for causal analysis. We show that these conflated measures can mislead studies of gene-environment interactions, especially when parental behavior responds to children's genetic risk. To address this issue, we construct a new latent measure of genetic risk that integrates individual genotypes with diagnostic symptoms, using data from the National Longitudinal Study of Adolescent to Adult Health linked to restricted individual SNP-level genotypes from dbGaP. Exploiting multiple sources of variation--including the Mendelian within-family genetic randomization among siblings--we find consistent evidence that parents compensate by investing more in children with higher genetic risk for ADHD. Strikingly, these compensatory responses disappear when genetic risk is proxied by the conventional ADHD PGS, which also yields weaker--and in some cases reversed--predictions for long-run outcomes. Finally, we embed our latent measure of genetic endowments into a standard dynamic structural model of child development. The model shows that both parental investments and latent genetic risk jointly shape children's cognitive and mental health development, underscoring the importance of modeling the dynamic interplay between genes and environments in the formation of human capital