This article deals with home-made food. The consumption of such specific goods involves an important null consumption issue. A two-step procedure is used to estimate whether the decision to consume is distinct from the decision on the amount of consumption. Data from the '1991 French Food Consumption Survey' show the relevance of double hurdle models. Finally, the estimation of Cragg's model is improved with a new undersampling procedure. Numerous and sometimes opposite effects of household characteristics on the probability of consuming home-produced food and on the amount consumed are obtained.