| My dissertation consists of three essays that use different types of variation across the life cycle to study health outcomes and behaviors in children and young adults. The first essay uses co-variances in two correlated child-age profiles that are observed prospectively: the child-health profile and the parental-investment profile. It estimates the child health formation function's return-to-investment profile, and argues that early childhood is a critical period for health formation with high returns to investment.;The second essay traces the racial difference in health profiles to the intrauterine environment as measured at birth. It uses the variances in birthweight resemblance in natural occurring twin pairs. The estimation rests upon two orthogonality relationships between parental investment and birthweight variations. It documents racial differences in birthweight variance components. It also raises questions about variance decomposition estimates in general by reporting sensitivity to slight perturbations in twin sample composition (MZ-DZ, gender) and twin mortality risks.;The third essay probes a recently developed method that uses genetic markers as instrumental variables, which is built on the premise that a child's inheritance of alleles from parents can be considered a randomization mechanism at conception, and hence a 'clean' source of variation to identify the causal effects of behavior. It reports sexual differentiated patterns of association between genetic markers and behaviors, in what could be considered first-stage regression. Coupled with the hypothesis of functioning sexual dimorphism in genetic expression, it suggests that the genetic marker as instrumental variable approach needs an unambiguous understanding of the first stage estimates, and needs to consider the potential violation of monotonicity conditions. |
