Modeling Consumers' Participation in Gambling Markets and Frequency of Gambling

Survey data on participation in gambling typically contain many zeros. The presence of many zeros presents methodological problems for the analysis of participation in gambling markets and gambling expenditure. The most common techniques for handling zeros in gambling data have been the Tobit estimator and the Heckman selectivity estimator. Recent research indicates that hurdle models (Jones 1989, 2000) and the Cragg (1971) model, are better suited to analyze participation in gambling. We apply these models to gambling participation in Canada and find that the double hurdle model is preferred in two of the three forms of gambling examined.