Authors: Luisa Corrado, University of Rome Tor Vergata and University of Cambridge; Roberta Distante, FEEM

This paper proposes a novel approach to address identification of social endogenous effects by generalizing the methods commonly used in standard dynamic panel data analysis to the peer effects setting. Our methodology shows how one can estimate peer effects free of the reflection problem in a dynamic context where individual- and group-specific unobservable effects are controlled for. We apply a dynamic linear-in-means model for analyzing the importance of social ties for eating behavior of US youth. We show that the main drivers of eating behavior are habituation and imitation effects. Imitation effects explain most of the variation in the Body Mass Index of individuals who were normal-weight and overweight during adolescence. Obese adolescents, instead, become future obese adults through wrong habits enforced by imitative behavior.