This paper develops and estimates a model with multiple schooling choices that identifies the causal effect of different levels of schooling on health, health-related behaviors, and labor market outcomes. We develop an approach that is a halfway house between a reduced form treatment effect model and a fully formulated dynamic discrete choice model. It is computationally tractable and identifies the causal effects of educational choices at different margins. We estimate distributions of responses to education and find evidence for substantial heterogeneity in unobserved variables on which agents make choices. The estimated treatment effects of education are decomposed into the direct benefits of attaining a given level of schooling and indirect benefits from the option to continue on to further schooling. Continuation values are an important component of our estimated treatment effects. While the estimated causal effects of education are substantial for most outcomes, we also estimate a quantitatively important effect of unobservables on outcomes. Both cognitive and socioemotional factors contribute to shaping educational choices and labor market and health outcomes. We improve on LATE by identifying the groups affected by variations in the instruments. We find benefits of cognition on most outcomes apart from its effect on schooling attainment. The benefits of socioemotional skills on outcomes beyond their effects on schooling attainment are less precisely estimated.
Education, Health and Wages
C32: Multiple or Simultaneous Equation Models: Time-Series Models; Dynamic Quantile Regressions; Dynamic Treatment Effect Models
C38: Multiple or Simultaneous Equation Models: Classification Methods; Cluster Analysis; Factor Models
HCEO Working Groups: