The joint selection and purchase of alcohol and tobacco in Spain is modeled using bivariate probit and endogenous switching regression analysis to account for the interaction between drinking and smoking. Different expenditure structures for tobacco and alcohol arise when one or both are consumed. Endogenous switching among structures exists because households are not randomly assigned. The results show that the correlation term of the selection equations and most correction terms for selfselectivity bias are statistically significant. The results also show that income and household demographic variables are important determinants of both selection and expenditures on tobacco and alcohol.