摘要：Objectives. We investigated the associations between smoking and friend selection in the social networks of US adolescents. Methods. We used a stochastic actor-based model to simultaneously test the effects of friendship networks on smoking and several ways that smoking can affect the friend selection process. Data are from 509 US high school students in the National Longitudinal Study of Adolescent Health, 1994–1996 (46.6% female, mean age at outset = 15.4 years). Results. Over time, adolescents’ smoking became more similar to their friends. Smoking also affected who adolescents selected as friends; adolescents were more likely to select friends whose smoking level was similar to their own, and smoking enhanced popularity such that smokers were more likely to be named as friends than were nonsmokers, after controlling for other friend selection processes. Conclusions. Both friend selection and peer influence are associated with smoking frequency. Interventions to reduce adolescent smoking would benefit by focusing on selection and influence mechanisms. Adolescent smoking has declined over the past 2 decades 1 yet remains a significant determinant of current and future health outcomes. 2 Given the importance of peers during adolescence, it is unsurprising that numerous studies have documented a strong association between friendships and smoking. 3–7 The processes linking smoking and friendships, however, are quite complex, 8 and several questions remain about the nature of this association. Of particular interest is the lingering issue of endogeneity: how much peer networks affect smoking behavior versus how much smoking affects friendship choices. 9–11 Addressing endogeneity is essential to prevention and cessation efforts, as distinct causal processes behind selection and influence imply different policy and intervention prescriptions to combat the negative effects of smoking and other risky health behaviors. We have focused on explaining 2 empirical patterns: why friends have similar smoking behavior and why popular students are sometimes more likely to smoke than are less popular students. We tested whether similarity in smoking among friends is owing to peer influence or to selecting similar peers as friends. We also tested whether popularity leads to smoking or whether smoking increases students’ popularity. We used longitudinal data and a recently developed statistical model to simultaneously model the effect of the friendship network on smoking and the effect of smoking on friend selection. 12,13 Rather than creating a summary of each individual's network position (e.g., centrality), in this approach we incorporated the complete friendship network. In particular, the model predicted changes in the friendship network (because of smoking and other factors) and adolescents’ smoking behavior (because of friendship network factors). This model has been applied successfully to smoking among European adolescents 5,6,13–15 and other health outcomes, such as depression. 16–18 To date, however, to our knowledge these methods have not been used to examine smoking among US adolescents.
