Abstract
Governments world-wide increasingly rely on gambling revenues, increasing the importance of understanding who gambles and why. Previous literature used Tobit and Heckman models to statistically analyse participation in gambling. These models make strong assumptions about the nature of gambling participation. We examine the double hurdle model as an alternative to other statistical approaches to modelling gambling participation and spending for lotteries in the province of Alberta, Canada. Our results for lotteries, based on data from a 2002 survey of gambling prevalence in Alberta, clearly prefer the double hurdle model, which yields different results than the commonly used Tobit model. This has important implications for governments who rely on revenues from lottery to fund many different programs in their jurisdictions.
Acknowledgements
The authors thank the editors at IGS and three anonymous referees for their helpful comments on earlier drafts of this paper. Portions of this research were supported by funding from the Alberta Gaming Research Institute.
Notes
1. This discussion draws on the application of Tobit and double hurdle models to cigarette smoking by Garcia and Labeaga (Citation1996).
2. For detailed information on the 2002 Alberta survey, see Smith and Wynne (Citation2002). We do not use alternative data sources like the Canadian Survey of Household Spending (SHS) because the public use micro data in the SHS contains only aggregated household spending on all types of gambling, not detailed information on household spending in specific types of gambling like lottery tickets.