Impact of the Prosper Intervention On Peer Selection and Influence Processes

Universal substance use prevention programs typically include a focus on peer relationship processes that may be related to substance use initiation. Logic models typically emphasize building refusal skills and increasing positive peer influences. Our goal is to conceptualize the potential impact of these programs in terms of contemporary views of peer selection and influence processes in dynamic social networks. In the present paper we test whether the PROSPER intervention program had an impact on three such processes across 6th to 9th grades: (1) the tendency to establish friendships with deviant peers; (2) the tendency to establish friendships with peers similar to oneself on levels of deviance; and (3) the tendency for one's deviance to become more similar to that of one's friends. We examine these issues with respect to three forms of deviance: alcohol use, cigarette use, and delinquency.

We analyzed data from the PROSPER community-level randomized control trial, which was conducted in two consecutive grade-cohorts in 28 rural and semi-rural communities in Iowa and Pennsylvania. Half of these communities were randomly assigned to receive the PROSPER intervention, which included both a family-based program (6th grade) and a school-based program (7th grade). Students in each community provided reports of up to seven school-based friends prior to the intervention (fall 6th) and each spring from 6th to 9th grade. Students also responded to questions about substance use and delinquency at each assessment. We analyzed the interplay between changing patterns of deviant behavior and changing friendship patterns by using stochastic actor-oriented modeling as implemented in Snijders' SIENA program. For each type of deviant behavior, SIENA models that controlled for a wide range of endogenous social network dynamics produced separate model coefficients and standard errors for the three key peer selection and influence parameters. These parameter estimates and standard errors were entered into a random effects meta-analysis to determine whether they differed significantly for intervention and control communities. Results indicated that the PROSPER intervention was associated with a diminished tendency for youth to report friendships with delinquent peers (p < .05); a stronger tendency for youth to select friends similar to themselves on levels of smoking (p < .01); and diminished influence from friends' alcohol use (p < .10).  Overall, results suggests that the PROSPER intervention altered peer processes in complex ways that require more complex conceptual models of the interdependence of peer selection and influence dynamics.