A panel of Italian single households is used to test for rational addiction in alcohol consumption. These monthly consumption data raise problems of measurement errors and unobservable heterogeneity. To deal with the zeros in the dependent variable we adopt a specification based on infrequency of purchase. GMM estimators are used to deal with errors in variables and unobserved heterogeneity. There is evidence that alcohol consumers are actually forward-looking. Past consumption is also significant in explaining current consumption thus detecting the addictive nature of alcohol. Discount rates, the strength of addiction, short and long run price and income elasticities are significant and in line with the theoretical predictions. These results and the inclusion of a number of demographic and geographic characters deliver valuable information for public policy purposes and suggest significant differences driven by individual heterogeneity.