Estimating the Effects of Friendship Networks on Health Behaviors of Adolescents

This paper estimates the effects of friends' health behaviors, smoking and drinking, on own health behaviors for adolescents while controlling for the effects of correlated unobservables between those friends. Specifically, the effect of friends' health behaviors is identified by comparing similar individuals who have the same friendship opportunities because they attend the same school and make similar friendship choices, under the assumption that the friendship choice reveals information about an individual's unobservables. We combine this identification strategy with a cross-cohort, within school design so that the model is identified based on across grade differences in the clustering of health behaviors within specific friendship patterns. Finally, we use the estimated information on correlated unobservables to examine longitudinal data on the on-set of health behaviors, where the opportunity for reverse causality should be minimal. Our estimates for both behavior and on-set are very robust to bias from correlated unobservables