Association for Public Policy Analysis & Management

I estimate peer-effects in bullying behavior using a nationally representative cross-section of students from sixth to tenth-grade from the Health Behavior in School-Aged Children (HBSC) 2009-2010 U.S. study. Bullying at school is a pervasive problem in the U.S. that can be detrimental for mental health and human capital accumulation (Brown and Taylor, 2008). This has motivated a number of studies that have tried to understand the determinants of bullying behavior. However, previous empirical works have overlooked the role of social interactions as a determinant of the individual propensity to become a bully. Neglecting social interactions might induce a serious omitted variable bias impairing our understanding of bullying behavior. Moreover, understanding peer effects is crucial as several education policies change the composition of students across schools and classrooms.

I quantify school-level peer effects in bullying behavior estimating a structural model of social interactions proposed by Soetevent and Kooreman (2007). The model accounts for self-selection into schools and unobserved school heterogeneity and allows for intra gender differentials in social interactions. Allowing for school-level random effects and controlling for a number of individual and school characteristics as well as average exogenous characteristics of one’s peers, I find strong evidence of peer effects, showing that one has stronger preferences for bullying another student when the prevalent behavior in his school tend to be violent. Specifically, the introduction of a bully in a representative school increases by almost 3% the expected percentage of bullies in the school. Reciprocally, the introduction of a non-bully in the school reduces the expected percentage of bullies by 1%. The estimated effect can be considered as a lower bound for the actual effect of social interactions. In particular, social interactions will have a larger impact at the classroom level, where social ties are stronger. The analysis suggests that any policy aiming at increasing the degree of heterogeneity in the composition of schools is going to be successful at reducing the prevalence of bullying behavior. Moreover, the impact of such a policy will be amplified by the multiplicative effect of social interactions.