Social distance and homophily in adolescent smoking initiation.

BACKGROUND: Studies often demonstrate homophily in adolescent smoking behavior, but rarely investigate the extent to which this is due to the peer network processes of selection versus influence. Applying the concept of social distance, this study examines these two processes for smoking initiation.
METHODS: We analyzed socio-centric network data from the National Longitudinal Study of Adolescent Health (N=2065; grades=7th-12th). Social distance (degrees of separation), combined with stability and change in friendship networks, was used to derive indicators of peer selection and influence on initiation. Multilevel modeling was used to predict initiation from these indicators, and propensity score modeling was used to determine whether these associations remained after adjusting for pre-existing differences between initiators and consistent non-smokers.
RESULTS: We found that both peer influence and selection effects increased the likelihood of initiation even after adjusting with propensity score weights and demographic controls. While the effect size for peer influence depended on the overall proportion of smokers at the school, the selection effect was independent of the school environment. De-selection and indirect influence effects were not significant after controlling for school norm interactions.
CONCLUSIONS: The association between peer smoking and adolescent smoking initiation appears to be due to both peer selection and direct influence. However, "friends of friends" effects are likely to be confounded with contextual factors. Given that smoking initiation is primarily associated with close personal interactions between the adolescent and his/her friends, prevention efforts should focus on the role of smoking in fostering personal relationships among adolescents.