Having the wrong friends? Peer effects in adolescent substance use

Swedish cross-sectional survey data on young individuals aged 12–18-year-old was used to analyse school–class based peer effects in binge drinking, smoking and illicit-drug use. Significant and positive peer effects were found for all three activities. By introducing school/grade fixed effects, the estimated peer effects were identified by variation in peer behaviour across school–classes within schools and grades, implying that estimates were not biased due to endogenous sorting of students across schools. Further, endogeneity bias due to bi-directionality of peer effects was found for binge drinking and smoking. Controlling for this source of endogeneity resulted in even stronger peer effects.