Friends and peers are key to whether, when, and how much alcohol adolescents drink and are therefore central to epidemiological theories and perspectives on adolescent alcohol use. Because the consequences of early drinking can extend across the life course through many mechanisms, understanding the nature and extent of friend influences is of critical public health importance and central to understanding the societal disease burden that accumulates from individual behavioral decisions. Many alcohol use interventions are designed around friend influence, the idea that adolescents with friends who drink are likely to learn and adopt similar drinking behaviors themselves. The nature of friend influence - a social learning process - on individual drinking, however, is generally confounded with the joint friend selection process by which peers become intimates in the first place. Disentangling these reciprocal pathways requires new statistical methodologies developed to analyze complicated research designs capturing who is friends with whom, how and why those friendships dissolve and new ones form, and how and why individual drinking changes coincidentally. To address the issues of friend selection and friend influence on alcohol use complete longitudinal social network data recording all adolescent friendships in nine schools over three waves will be analyzed using an innovative new class of co-evolutionary longitudinal network-behavior models. This novel approach allow friendships, alcohol use, and other behavioral covariates to mutually influence each other, thereby controlling one for the other, so that friend selection and influence effects over time can be isolated and compared.
The specific Aims are (1) to simultaneously estimate and compare the importance of friend selection and social influence on alcohol use initiation, (2) changes in levels of use, and to assess alternative alcohol use mediational (3) friend selection (4) and friend influence pathways. Alternative mediational pathways are motivated from interactional theory and include social and individual processes such as commitment to school, relationships with parents, and traditional beliefs/attitudes. The analyses will be based on seven small (N=798) and two large schools from the National Study of Adolescent Health, one rural and mostly white (approx. N=800), the other urban and racially mixed (approx. N=1,700). These data are preferable to those used in most prior friend influence studies because complete social networks are measured, friend measures are drawn from their own reports, and the data is longitudinal. Friendship data was collected first in school, and then subsequently at two additional in-home interviews. This study will contribute to alcohol use epidemiology by using innovative new methodologies that embed adolescents into dynamic environments that they influence and are influenced by. The co-evolutionary network-behavior approach thus holds great promise for using existing epidemiological data to learn about alcohol use in adolescence in ways that are richer descriptively and inferentially than existing methodologies while more accurately reflecting the complexities of the social worlds within which adolescents are embedded.
The public health relevance stems from the fact that alcohol use is a widespread social problem in adolescence and many interventions are developed around social learning, the idea that adolescents with friends who drink are likely to learn and adopt similar drinking behaviors. Accordingly, there is a critical need for research accurately estimating how friends influence each other while accounting for the role alcohol use plays in who becomes friends with whom in the first place. Disentangling these friend influence and selection processes is key for determining accurate expectations for interventions, developing more effective interventions incorporating friends and peers in realistic ways, and for assessing the effectiveness of those interventions.
