Dynamic Social Interactions and Health Risk Behavior
We study risky behavior of adolescents. Concentrating on smoking and alcohol use, we structurally estimate a dynamic social interaction model in the context of students' school networks included in the National Longitudinal Study of Adolescent Health (Add Health). The model allows for forward-looking behavior of agents, addiction effects, and social interactions in the form of preferences for conformity in the social network. We find strong evidence for forward looking dynamics and addiction effects. We also find that social interactions in the estimated dynamic model are quantitatively large. A misspecified static model would fit data substantially worse, while producing a much smaller estimate of the social interaction effect. With the estimated dynamic model, a temporary shock to students' preferences in the 10th grade has effects on their behavior in grades 10, 11, 12, with estimated social multipliers 1:53, 1:03, and 0:76, respectively. The multiplier effect of a permanent shock is much larger, up to 3:7 in grade 12. Moreover (semi-) elasticities of a permanent change in the availability of alcohol or cigarettes at home on child risky behavior implied by the dynamic equilibrium are 25%, 63%, and 79%, in grades 10, 11, 12, respectively.
We are grateful to Steven Durlauf, Matt Jackson, David Levine for their extensive comments and suggestions at different stages. We also thank Larry Blume, Pierre-André Chiappori, Aureo De Paula, Marcel Fafchamps, Bryan Graham, Bob Lucas, Jen Overbeck, Tom Sargent, José Scheinkman, Giorgio Topa, and seminar participants at Yale, UPenn, NYU, Chicago, Cornell, Queens, University of Toronto, University of Queensland, Melbourne Business School, Université Laval, Australian National University, University of Western Australia, Workshop on Social Identity and Social Interactions in Economics in Quebec City, and the Australasian Economic Theory Workshop in Canberra. Özgür thanks the Economics Department at Sciences Po in Paris, where a big part of this research was done, for its hospitality. Özgür is also grateful for financial support to Melbourne Business School through MBS Competitive Grant schemes in 2013-2017, to “La Chaire du Canada en Économie des Politiques Sociales et des Ressources Humaines” at Université Laval, and to SSHRC, FQRSC, and IFM2. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.