This paper estimates a simultaneous model of moderate and problem drinking, smoking, and wages using a random sample of employed Canadian men. The results indicate that sample selection into alcohol and tobacco use is not negligible. With all else in the system held constant, moderate and heavy...