View article

In this paper, we estimate a tobacco consumption rational addiction model with panel data. We apply Becker and Murphy's (1988, Journal of Political Economy 96, 675–700) proposal to individual rather than aggregate data. Given the censoring nature of the dependent variable, we use a double-hurdle specification after testing that the factors governing the consumption equation are not the same as those that govern the participation equation. We use non-linear panel data methods to consistently retrieve the structural form parameters, as usual linear panel data methods applied to the consumption equation conditional on participation do not provide consistent estimates. The empirical results reject the myopic model and give support to the rational addiction one. Finally, although long-run elasticities are bigger than short-run elasticities, the individual's response to prices is very small.