A longitudinal investigation of peer influence, peer selection and adolescent smoking in UK schools

Smoking in adolescence is strongly associated with friends‘ smoking behaviour. This relationship is not straightforward to understand, and there is inconsistent evidence from social network studies regarding the relative importance of peer influence (effect of friends on behaviour) and peer selection (effect of behaviour on friendship). The inconsistencies are in part due to difficulties in the statistical analysis because of the inherent dependencies between individuals in social networks. A promising advancement in the effort to disentangle the mechanisms that lead to peer group behavioural homogeneity in friendship networks is the stochastic actor oriented model, as implemented in SIENA. To date, few studies have used SIENA to examine influence and selection processes in adolescent smoking. In this presentation, we apply the SIENA methodology to data collected through A Stop Smoking in Schools Trial (ASSIST) to assess the extent to which influence and selection contribute to homogeneity of smoking behaviour in peer groups. ASSIST collected three waves of behavioural data (including own smoking behaviour, and other behavioural attributes) and social network data (students named up to six friends) on a yearly basis from over 10,000 students in 59 schools and will thus also allow investigation of variation in network-behaviour relationships across schools. This paper uses data from over six hundred 12-13 year old students in three schools. The findings of these analyses will be presented and their implications for the future of smoking prevention interventions discusse