The large sibling correlations in risky behaviour between siblings raise the possibility that adolescents may directly influence the actions of their brothers or sisters. We assess the extent to which correlations in substance use and selling drugs are causal. Our identification strategy relies on panel data, the fact that the future does not cause the past, and the assumption that the direction of influence is from older siblings to younger siblings. Under this assumption along with strong restrictions on dynamics, one can identify the causal effect from a regression of the behaviour of the younger sibling on the past behaviour and future behaviour of the older sibling. We also estimate a joint dynamic model of the behaviour of older and younger siblings that allows for family specific effects, individual specific heterogeneity, and state dependence. We use the model to simulate the dynamic response of substance use to the behaviour of the older sibling. We find that smoking, drinking and marijuana use are affected by the example of older siblings, but only a small faction of the linke between siblings is causal.