This paper proposes a model-based empirical method for a policymaker to effectively target adolescents with a prevention program to reduce risky behavior such as marijuana consumption. We assume that peer pressure is the mechanism that determines how social interactions influence these behaviors. To identify the most influential individuals, we estimate social marginal effects using observational data on multiple social connections with a generalized method of moments (GMM) framework. Our empirical strategy uses the observed characteristics of distant individuals in a multilayered network space as instruments to address the endogeneity of the networks that arise from homophily. We use the Add Health data to find positive peer effects, for friends and classmates, on both cigarette smoking and marijuana use, with the friends’ effect having a greater impact. Our findings suggest that policymakers can benefit from using social marginal effects to target high-influence individuals in their prevention programs.