Abstract
Cigarette consumption by US women is investigated. The dependent variable is discrete and censored at zero. The Gaussian hurdle model and negative binomial hurdle model are used to accommodate zero observations in the sample. Elasticities derived from the two models are extremely close.