Philippine Escudié
- Lieu
-
Îlot Bernard du Bois
- Salle 24
AMU - AMSE
5-9 boulevard Maurice Bourdet
13001 Marseille - Date(s)
-
Vendredi 27 novembre 2026
12:00 à 12:45 - Contact(s)
-
Garance Desrousseaux - garance.desrousseaux[at]univ-amu.fr
Martin Dhaussy - martin.dhaussy[at]univ-amu.fr
Lilie Timricht - lilie.timricht[at]univ-amu.fr
Résumé
Adolescents who use substances do not occupy random positions in their friendship networks. But it is hard to tell whether network position shapes consumption, or whether consumption shapes who becomes friends with whom. This paper proposes a method to separate the two. Using two waves of the Add Health saturated sample, I estimate a stochastic actor-oriented model of how friendships and alcohol, tobacco and cannabis use evolve together. Both channels are at work: adolescents choose friends whose consumption is close to their own, and their consumption is formed to match their friend’s consumption. I then use the estimated model to simulate counterfactual networks in which friendship choices no longer respond to consumption, and compare the link between centrality and consumption in the observed and counterfactual networks. Selection explains a large part of what we observe. Cannabis users are at the margin of the network on every centrality measure, but these associations lose significance in the counterfactual. The positive link between drinking and popularity, and the link between smoking and having central friends, also largely disappear. By contrast, the link between drinking and the reach of one’s connections does not change, which is consistent with network position itself affecting consumption. Which centrality measure we use therefore matters, and centrality measured on the observed network can be a misleading predictor of consumption.