A strong association between the risk behavior of individuals and the risk behavior of their friends has been documented in many studies (see reviews: Ivaniushina & Titkova, 2021; Montgomery et al., 2020). There are two processes that contribute to this association: social influence, in which a person changes their behavior to align more closely with the behavior of their friends, and social selection, in which people tend to befriend those who engage in similar behaviors. It is generally understood that these processes often occur simultaneously (Montgomery et al., 2020). Therefore, the estimation of influence effects must consider selection effects, and without knowledge of the latter, the effect size of the former cannot be estimated correctly (Manski, 1993). The use of longitudinal data and recently developed methods of statistical modeling (RSiena, discussed in the Methods section) allows for the accurate disentanglement of these intertwined processes.

Different behaviors, however, might exhibit the effects of influence and selection differently. Our meta-analysis shows that both the selection effect and the influence effect are significant in alcohol consumption, which indicates that adolescents prefer to select friends who are similar to them regarding alcohol consumption and that adolescents adjust their drinking behavior to match their friends’ behavior (Ivaniushina & Titkova, 2021). Similar results are obtained for smoking (Schaefer et al., 2012).

Unlike smoking, which is highly visible, and alcohol consumption, which is relatively visible among friends, sexual behavior is highly private. Still, studies show that peers might influence one’s sexual debut (Ali & Dwyer, 2011), risky sexual behavior (Ajilore, 2015), or contraceptive use (Ali et al., 2011). The main disadvantage of such studies is that authors do not differentiate peer selection and peer influence effects, which makes the results less reliable. Sexual (il)literacy might be another type of content passing through the friendship networks. Different actors might be the source of information about sexual relationships for adolescents. At the earliest developmental stages, it is parents who educate their children on sexual health (De Graaf et al., 2011). However, during adolescence, the reference group for acquiring this type of knowledge changes, and peers become an important source of information about sex (Nogueira Avelar E Silva et al., 2020). There is a lack of studies focusing on the spread of sexual (il)literacy in adolescent networks. However, intervention studies indicate the effectiveness of peer-led interventions in increasing one’s sexual literacy (Mitchell et al., 2021).

In our research on adolescent risk behavior, we conducted a longitudinal survey in Russian vocational schools focusing on drinking, smoking, sexual behavior, and sexual literacy. The survey had a social network component, thus allowing us to study peer influence and peer selection on many covariates. In Russia, students choosing vocational education come to their vocational schools from other schools at the age of fifteen and form new friendship ties in these new settings, staying there for three to four years of training. They study there between the ages of 15 and 18, a time when they eagerly try to develop new patterns of behavior, seek new friendships, and are most vulnerable to peer influence. This makes such schools a perfect ground for studying the co-evolution of peer relations and adolescent risk behavior.

While there are many network-based studies on smoking and drinking, to our knowledge, no study has ever used dynamic social network analysis to model the diffusion of sexual behavior and (il)literacy among adolescents and compare it with other risk behaviors.

We use our unique longitudinal dataset collected between 2016 and 2020 from 13 vocational colleges (65 groups, 1800 students) in St. Petersburg, Russia. Students in vocational schools often come from low SES families, and risk behaviors are more prevalent there than in academic schools, making this environment advantageous for our study. Furthermore, our longitudinal design enables us to observe the de novo formation of social networks as we track students from their first year of enrollment onwards. The survey includes data on smoking (assessed by both occurrence and frequency), drinking (measured by frequency and intensity of drinking), and sexual experience (a binary indicator of sexual activity) across four waves, along with measures of sexual literacy obtained in two waves. The sexual literacy index is constructed based on students' knowledge of various contraception methods and their effectiveness in preventing unwanted pregnancies and sexually transmitted infections. Students were asked to provide names of people they communicate with most often in their group. Directed friendship networks are built using these nominations. Groups with a sufficient number of responses to these questions (80% as a rule of thumb) in all waves were included in the analysis. For each group, a stochastic actor-oriented model (SAOM) was built using the RSiena package. SAOM models simultaneously and explicitly model the co-evolution of social networks and actors’ behavior (Steglich et al., 2010). Actors create, maintain, or disrupt ties; they may also change their behavior. The unobserved process of change takes place in micro-steps, with the alteration of one tie at a time or the movement of one step at a time on a behavioral scale. There are separate parts of model specification: one for network change and another for behavioral change. This allows separate conclusions to be drawn regarding the selection and influence processes that occur in networks. The results of separate SAOMs for each group are summarized using meta-analysis.

For alcohol consumption there are both significant peer influence and peer selection effects in Russian vocational schools, indicating that adolescents prefer to select friends who are similar to them with regard to alcohol consumption, and that adolescents adjust their drinking behavior to match their friends’ behavior. It is very much in line with the results of all previous studies (see meta-analysis in: Ivaniushina & Titkova, 2021). For smoking our modeling shows the existence of peer selection but no significant peer influence. Previous data on selection and influence effects on adolescent smoking behavior is ambiguous. While some authors demonstrate that friends' similarity in smoking behavior is driven mainly by social selection rather than peer influence (Kiuru et al., 2010; Mercken et al., 2010), other studies have revealed peer influence effects as well (Schaefer et al., 2012). It seems that in Russian context students that smoke at late adolescence acquire smoking habits before the age of fifteen, and in vocational schools ‘smokers’ are getting acquainted with each other through hanging out for a smoke during breaks outside of school buildings. For sexual behavior and sexual (il)literacy we are currently running RSiena models, and we expect weak effects (if any) of peer selection and peer influence on sexual behavior and relatively significant peer effects on sexual (il)literacy among adolescents in Russian vocational schools. Full results will be presented in the paper at ECER.

Ajilore, O. (2015). Identifying peer effects using spatial analysis: The role of peers on risky sexual behavior. Review of Economics of the Household, 13(3), 635–652. Ali, M. M., Amialchuk, A., & Dwyer, D. S. (2011). Social Network Effects in Contraceptive Behavior Among Adolescents. Journal of Developmental & Behavioral Pediatrics, 32(8), 563–571. Ali, M. M., & Dwyer, D. S. (2011). Estimating peer effects in sexual behavior among adolescents. Journal of Adolescence, 34(1), 183–190. De Graaf, H., Vanwesenbeeck, I., Woertman, L., & Meeus, W. (2011). Parenting and Adolescents’ Sexual Development in Western Societies: A Literature Review. European Psychologist, 16(1), 21–31. Ivaniushina, V., & Titkova, V. (2021). Peer influence in adolescent drinking behavior: A meta-analysis of stochastic actor-based modeling studies. PLOS ONE, 16(4), e0250169. Kiuru, N., Burk, W. J., Laursen, B., Salmela-Aro, K., & Nurmi, J. E. (2010). Pressure to drink but not to smoke: Disentangling selection and socialization in adolescent peer networks and peer groups. Journal of adolescence, 33(6), 801-812. Manski, C. F. (1993). Identification of Endogenous Social Effects: The Reflection Problem. The Review of Economic Studies, 60(3), 531. Mercken, L., Snijders, T. A., Steglich, C., Vartiainen, E., & De Vries, H. (2010). Dynamics of adolescent friendship networks and smoking behavior. Social networks, 32(1), 72-81. Mitchell, K. R., Purcell, C., Simpson, S. A., Broccatelli, C., Bailey, J. V., Barry, S. J. E., Elliott, L., Forsyth, R., Hunter, R., McCann, M., McDaid, L., Wetherall, K., & Moore, L. (2021). Feasibility study of peer-led and school-based social network Intervention (STASH) to promote adolescent sexual health. Pilot and Feasibility Studies, 7(1), 125. Montgomery, S. C., Donnelly, M., Bhatnagar, P., Carlin, A., Kee, F., & Hunter, R. F. (2020). Peer social network processes and adolescent health behaviors: A systematic review. Preventive Medicine, 130, 105900. Nogueira Avelar E Silva, R., Raat, H., Reitz, E., Plat, M., Deković, M., & Van De Bongardt, D. (2020). Longitudinal Associations Between Sexual Communication With Friends and Sexual Behaviors Through Perceived Sexual Peer Norms. The Journal of Sex Research, 57(9), 1156–1165. Schaefer, D. R., Haas, S. A., & Bishop, N. J. (2012). A dynamic model of US adolescents’ smoking and friendship networks. American journal of public health, 102(6), e12-e18. Steglich, C., Snijders, T. A. B., & Pearson, M. (2010). Dynamic Networks and Behavior: Separating Selection from Influence. Sociological Methodology, 40(1), 329–393.