| Decades of criminological research have pointed to peers and peer groups as sources of and conduits for delinquency. Often, however, there has been a lack of clarity about the role of peers. This lack of clarity concerns whether the respondent delinquency-peer delinquency correlation results from peer influence or peer selection. Recently, the use of longitudinal social network analysis has provided valuable insights that have strengthened our understanding of peer selection and peer influence. While researchers often find evidence of both selection and influence generally, less inquiry has focused on the conditions in which peer influence is more or less likely, and whether differences exist across behaviors. The first question for this dissertation concerns whether students select friends based on their similarity in marijuana use, alcohol use, and bullying. Next, it assesses whether students are influenced by the behavior of their peers. Additionally, it examines whether potential peer or personal characteristics moderate these selection and influence effects. Using five waves the Teenage Networks in Schools (TEENS) dataset, a bi-weekly survey of a 9th grade cohort, this study found evidence of considerable short-term change in students' friendships, marijuana use, alcohol use, and bullying. The results show that students selected friends based on similarity in sex/gender and approval of delinquency, rather than similarity in marijuana use, alcohol use, or bullying. Additionally, there was no evidence of peer influence across any of the dependent variables. Prior school year behavior, however, was a significant predictor of behavior during this short-term study. Lastly, there was weak evidence to suggest popularity (negative effect) and unstructured socializing (positive effect) may be important moderators of marijuana use selection and bullying influence, respectively. Theoretical as well as practical implications of these findings are discussed, along with limitations of this study and future directions for research. |
