We estimate the effect of education and student status on the propensity to smoke. Our estimation strategy accounts for the endogeneity of education by "differencing out" the impact of unobserved characteristics correlated with educational attainment. This is accomplished by exploiting education...