Peer effects in adolescents' delinquent behaviors: Evidence from a binary choice network model

• Contributes to the social interaction literature using a binary choice network model;
• Shows evidence for endogenous and contextual effects in all activities considered;
• Results are robust to alternative weighting matrices based on friend heterogeneity;
• Results are sensitive to the friendship reciprocity assumption.

This paper studies peer influences in adolescents' deviant behaviors, including drinking alcohol, doing dangerous things, skipping school and physical fighting, by a binary choice network model with heterogenous rational expectations proposed in Lee et al. (in press). For a wide range of parameter values for social interactions, this model has a unique equilibrium and is not subject to multiple equilibria and other model coherency related issues for discrete choice models. I specify peer reference group based on actual friendship networks. The data are from the National Longitudinal Study of Adolescent Health (Add Health). A school fixed effect strategy is employed to control for the confounding effects. Two ways of calculating marginal effects are evaluated. Robustness analysis is performed with regard to several alternative spatial weights matrices. I find that both endogenous and contextual effects exist in all the activities considered, even after controlling for school fixed effects. Similar to the case of continuous variables, the results are robust to several alternative specifications of the weighting matrix based on friend heterogeneity, providing justification for the common practice of assigning equal weight among friends. The changes in the results under the friendship reciprocity assumption indicate the relevance of the real friendship nomination network.