Background Behavioural homophily is frequent in communities: individuals are more likely to befriend others with similar substance use, such as drinking, illegal drug use and smoking. This homophily results from a mix of influence and selection effects and contributes to intergroup inequalities in health. Yet, the underlying mechanism driving this homophily remains opaque. Most studies have assumed that homophily results from simple contagion. However, this is unlikely when the behaviour in question is deemed illegitimate. Instead, the theory of complex contagion claims norms are important drivers of the dissemination of behaviour in a network. Yet, so far, the role of norms in explaining behavioural homophily has received little attention in community health research. This paper tests this theory and investigates whether and how normative expectations affect the homophily of smoking in adolescents at school.

Methods We used the second wave of the SILNE study, a whole social network study of adolescents, carried out in 2016 in seven European countries. Adolescents were asked to nominate up to five friends (also referred to as ‘alters’) with a name-generator taken from the Add Health design. Homophily was investigated for having tried smoking and for regular smoking. Three groups of norms were investigated: (1) a school-level policy score (STP) describing the comprehensiveness and enforcement of school anti-tobacco policy; (2) injunctive norms (friends’, parents’, and teachers’ disapproval of smoking, and (3) stigmatisation of smokers. We used Bayesian Dyadic Regressions with Markov Chain Monte Carlo Sampling, to model heterophilous vs homophilous ego-alter dyads regarding smoking status. Random components were included at ego, alter and school levels, and controlled for socio-demographics ego-alter homophily on sex, age, socioeconomic status and school achievement. We analyzed 33,400 dyads from 7,420 adolescents, attending 55 schools. The analysis was performed with the BGLIMM SAS procedure.

Results Heterophily of smoking status decreased strongly with the joint smoking disapproval of the ego-alter dyad (Beta=-0.76; 95%HPD:-0.83, -0.68) and increased with ego-alter difference in disapproval of smoking (Beta=0.35, 95%HPD: 0.26,0.44). The same was found for stigmatization of smokers: ego-alter stigmatization decreased heterophily (Beta=-0.15; 95%HPD: -0.21,-0.10) and ego-alter difference in stigmatization increased heterophily (Beta=0.11; 95%HPD:0.04,0.17). STP was not associated with heterophily on smoking status. In addition, this heterophily decreased with ego-alter heterophily on school achievement, on sex and on parental smoking but not on the adolescent own socio-economic status.

Conclusion Norms and stigmatization strengthen smoking status homophily in adolescents. The cross-sectional design is however a limitation. Homophily in smoking norms is instrumental in preventing the social dissemination of a risky behavior.