DECOMPOSING THE PEER EFFECT ON ADOLESCENT SUBSTANCE USE: MEDIATION, NONLINEARITY, AND DIFFERENTIAL NONLINEARITY*

Abstract

Although the correlation between peer delinquency and delinquency is one of the most consistently demonstrated findings in delinquency research, researchers have focused primarily on the direct, linear, and additive effects of peers in statistical models, rather than on empirically modeling mediating, nonlinear, and moderating processes that are specified by theory. To address these issues, we measure respondent delinquency and peer delinquency with illegal substance use and then decompose the effect of peer substance use on self-reported substance use. Logistic hierarchical models on a sample of adolescents from the Project on Human Development in Chicago Neighborhoods (PHDCN) indicate that the effect of peer substance use on self-reported substance use is partially mediated by perceptions of the health risks of substance use. In addition, the direct statistical effect of peers is nonlinear: On average, the peer effect decreases at higher values of peer substance use, which is consistent with a “saturation” effect. We also find that the functional form of the peer substance use/substance use relationship is dependent on the neighborhood context. In neighborhoods with more opportunities for crime, the peer effect is initially strong but decreases as peer substance use increases, which is consistent with a saturation effect. Conversely, in neighborhoods with fewer opportunities for crime, the effect of peers is initially small, but as delinquent peer associations increase, the peer effect increases multiplicatively.