Direct Empirical Tests Of The Alternative Frameworks
Chaloupka (1991) used data on individuals from the second National Health and Nutrition Examination Survey to test the rational addiction model. He regressed daily cigarette consumption on cigarette prices, past, present, and future, and lagged and future cigarette consumption. The price parameter estimates were not statistically significant, providing weak support for the rational addiction model.
Becker et al. (1994) used cross sectional data on US states to test the rational addiction model. A key implication of the model is that people are forward-looking in making decisions about consumption of an addictive good. They found that, as cigarette prices rise, current consumption of the good falls, which is what the rational addiction model predicts.
Auld and Grootendorst (2004), using annual aggregate data on milk and the rational addiction model as the underlying conceptual framework, found that milk is the most addictive of all commodities evaluated, including cigarettes! The literature contains other critiques of the models and tests as is, perhaps, to be expected given the relative youth of the topic and its difficult ‘fit’ with standard economics.
Imperfectly Rational Model
Even if people are not rational and forward-looking, they may be simply forward-looking. Gruber and Koszegi (2001) tested whether cigarette consumption was negatively related to announced cigarette excise tax increases. They found that it was, which they interpreted as evidence for the notion that people are forward-looking but not necessarily time-consistent in their preferences.
There is some empirical evidence on hyperbolic discounting by smokers in particular. Indicators of hyperbolic discounting include: the use of commitment devices, discount rates that vary according to the time horizon with short-terms exceeding longer term ones, and actual behavior not matching stated plans. Odum et al. (2002) evaluated discounting on the part of small samples of current, former, and never smokers using two nonlinear decay models, one for exponential and the other for hyperbolic discounting. They found the hyperbolic model provided a better fit between the two. Current smokers discount health gains and losses more than never smokers do using both exponential and hyperbolic functional forms. Several other studies, however, have produced conflicting results, and strong empirical support for a key role of hyperbolic discounting in the smoking decision is not yet available.
There is a large literature on the relationship between advertising and smoking. Two reviews are in the Further Reading section. However, the vast majority of studies have used aggregate data, which do permit a direct test of the effect of cues on smoking. There is some evidence that teen exposure to TV advertising has a positive and significant impact on the probability of smoking. However, most direct testing of the effects of cues has been conducted by noneconomists.