This research aims to analyze how smoking among young Brazilians is influenced or not by the effects of peers at the level of school classes (colleagues), weighted by individual characteristics, contextual to individuals and institutions in the school households surveyed. The contribution of this study is to fill an existing gap in the Brazilian literature, where the existence of drug use pairs in general is still little explored by empirical academic research in the area of health and social economy. The analytical and econometric approach taken here aggregates issues related to non-linear models of consumption choices and lagged social interaction effects, where the issue of causality is resolved with such empirical modeling. Using individual-level data from the 2012 and 2015 of Pesquisa Nacional de Saúde do Escolar, estimates were performed using a multinomial logit model and an additional series of models were estimated by inputting control variables, adolescent family background, effects of year and region to mitigate the biases of the social interaction effects of school classes, in addition to the temporal effect for the year 2012. The results found showed that increasing the proportion of peers in a smoking room increases the relative risks of be an occasional and regular user, for the model with all variables (RRR=2,501; 2,064); while for the lagged pair effect, the relative risks of being an occasional smoker decrease in the 2015 period (RRR=0.614; 0.493). The estimation of the various econometric models, where the other regressors are added gradually, showed a reduction in the relative risks of smoking occasionally and regularly, albeit in a discrete way and still being statistically significant to fully explain smoking. Keywords: Peer effects. Smoking. Health adolescents. Pseudo-panel. Social econometrics