Estimating Peer Effects on Health in Social Networks

Abstract

We recently showed that obesity can spread socially from person to person in adults (Christakis and Fowler 2007). A natural question to ask is whether or not these results generalize to a population of adolescents. Three separate teams of researchers have analyzed the National Longitudinal Study of Adolescent Health (Add Health) and shown evidence of person-to-peron spread of obesity, but they use different methods and disagree on the interpretation of their results. Here, we conduct our own analysis of the Add Health data, provide additional evidence from the Framingham Heart Study on the social spread of obesity, and use Monte Carlo simulations to test the econometric methods we use to model peer effects. The results show that the existence of peer effects in body mass is robust to several specifications in both adults and in adolescents.