Social influence from friends and other peers on Northern Irish adolescents' use of cannabis and other illicit drugs"

For those working with young people, it is invaluable to know to what extent any individual's decisions around drug use reflect the behaviour which their peers have modelled for them. Peer influence is, however, notoriously difficult to estimate, often because of the inadequacies of the available data. This thesis describes secondary statistical analysis of the Belfast Youth Development Study, a dataset with several attractive characteristics. Peer behaviour could be measured at t he level of the school grade, the friendship cluster and various related subgroups. Furthermore, data on network composition and drug use behaviours were available over multiple consecutive years, which is rare internationally. Using a suite of regression models, including standard linear, instrumental variables and fixed effects models, I demonstrate the importance of friends' behaviour and of the relationships between friends in determining individual• level drug use. Specifically, I find strong positive causal coefficients for friends' average use of cannabis and other drugs on ego's use. By contrast, the mean use of other grademates is found to have zero effect on that of ego. In dealing with the problems of identification, I employ a sound instrumental variable strategy exploiting network structure and capitalise on the longitudinal dimension of BYDS to produce an individual fixed effects model controlling for initial preferences and select ion of new friends. This is the first such model of adolescent drug use. Both of these strategies converge on a peer influence coefficient of 0.4, indicating that a 10% increase in friends' illegal drug use results increases ego's likelihood of use by 4%. The effect is much stronger for friends who reciprocate ego's friendship. There is also evidence that drug using groups are more tightly knit at their core but also more open around their margins than other groups. These results are discussed in terms of their applicability to educating young people, both on how their peers' behaviour might compromise their long-term health and personal autonomy, and, reciprocally, on the potential effects of their behaviour on their closest friends.EThOS - Electronic Theses Online ServiceGBUnited Kingdo