Previous research has found that youth smoking choices are strongly influenced by peer smoking. However, these studies often fail to account for simultaneity and nonrandom peer selection. This article describes an equilibrium model of peer effects that incorporates both of these features, and estimates its parameters using data on California teenagers. Identification is aided by using the influence of observable variables on group selection as a proxy for the influence of unobservables. I find that the effect of peer smoking on the decision to smoke is much weaker than found in previous studies.