Abstract

Most smokers have begun the habit before the age of 18 and understanding the dynamics of smoking in adolescence and the influence of social networks may help us design better interventions.¹ In 2001, Queens University Belfast began a longitudinal study of adolescents (the BYDS study) over annual five waves, on individual attributes, behavioural traits and school-friendship networks.

We have systematically reviewed the prospective studies on peer network effects on adolescent smoking behaviour and disparate results, quality and methodological perspectives are evident. A common feature is an emphasis on topology and metrics² but with little account taken for nodal characteristics and broader school or contextual factors.

To illustrate their importance, we use data from 1000 adolescents in eight BYDS schools first to describe how peer network topologies vary between boys and girls and across schools.

A relatively new method of analysis is employed based on stochastic actor-orientated models (implemented in SIENA³). We will present data from multiple waves to illustrate how this can accommodate both network topology and individual characteristics to trace the effects of peers on the evolution of smoking behaviour within in a continuous time stochastic process and place our findings (on gender differences and school level factors) in the context of our systematic review.