A Multivariate Sample-Selection Model : Estimating Cigarette and Alcohol Demands with Zero Observations

Heckman's sample-selection model is generalized to the multivariate case and used to estimate demands for cigarettes, beer, and wine by individuals in the United States. The specification allows correlations between the error terms of multiple selection and level equations, and nests the popular bivariate sample-selection model and two-part model. Empirical results suggest that the proposed model performs better than the restricted specifications. Differentiated effects of variables on probabilities and levels of consumption also suggest rejection of the Tobit system. Gender differences are present, and demographic variables are more important than income in determining consumption of cigarettes, beer, and wine