Is Addiction "Rational"? Theory and Evidence

Abstract

A standard model of addictive process is Becker and Murphy's "rational addiction" model, which has the key empirical prediction that the cunent consumption of addictive goods should respond to future prices, and the key normative prediction that the optimal government regulation of addictive goods should depend only on their interpersonal externalities. While a variety of previous studies have supported this empirical contention, we demonstrate that these results are very fragile. We propose a new empirical test for the case of cigarettes, using state excise tax increases that have been legislatively enacted but are not yet effective, and monthly data on consumption. We find strong evidence that consumption drops when there are announced future tax increases, providing more robust support for the key empirical prediction of the Becker and Murphy model. But we also propose a new formulation of this model that makes only one change, albeit a major one: the incorporation of the inconsistent preferences which are likely to provide a much better platform for understanding the smoking decision. We find that with these preferences the model continues to yield the predictions for forward-looking behavior that have been tested by others and by ourselves. But it has strikingly different normative implications, as with these preferences optimal government policy should depend as well on the "internalities" imposed by smokers on themselves. We estimate that the optimal tax per pack of cigarettes should be at least one dollar higher under our formulation than in the rational addiction case.

The value of risk‐free cigarettes – do smokers underestimate the risk?

“…Gruber (2001) and Gruber and Köszegi (2000), we will adopt the admittedly crude, recent estimate by Shaw et al (2000) Table 9. Value of life year ranges from USD 2,300 to 6,700 when we do not use a positive discount rate, and from 3,600 to 18,600 with a 5% discount rate.…”

“…In a similar way, but instead using value of life figures from Viscusi (1993) along with figures of the time lost from smoking in Manning et al (1991), Gruber (2001) and Gruber and Köszegi (2000) calculate internal costs of smoking to be roughly USD 30 per pack of cigarettes, which corresponds to a value of life year of USD 115,000. However, these value of life year estimates are calculated from value of life estimates, without appropriate correction for a lower quality of life at a higher age.…”

Optimal harvesting of fish stocks under a time-varying discount rate

“…In particular, although hyperbolic discounting may place greater weight on the interests of future generations, there are problems with incorporating hyperbolic discounting of utility into social cost-benefit analysis (Pearce et al, 2003;Groom et al, 2005). Because hyperbolic discounting can explain a range of perplexing human behaviours, including drug addiction (Gruber & Koszegi, 2001), sub-optimally low savings rates (Laibson, 1997;Laibson et al, 1998;Harris & Laibson, 2001), procrastination (ODonoghue & Rabin, 1999a,b;Benabou & Tirole, 2004) and various others (Akerlof, 1991), this paper does not make the normative recommendation that hyperbolic preferences be incorporated into optimal policy making. 4 On the other hand, it is important to understand the descriptive implications of what might happen if a planner does use hyperbolic discounting in making policy, if only so that we can be aware of potential problems before they arise.…”

Rational Addiction with Optimal Inventories: Theory and Evidence from Cigarette Purchases in Japan

“…There are 1 In Angrist et al (2000), moving averages of the three days' wind speeds and wave heights preceding the trading day were measured as the dummy instrument, stormy. 2 Here, I do not analyze the issues of self-control and time-inconsistent preferences presented by Winston (1980) and Gruber and Köszegi (2001). In addition, I do not analyze the influence of chemical dependency, which was discussed by Cameron (2000).…”

“…Such a typical endogeneity problem is caused by the omission from the model of a variable correlated with the explanatory variable. Moreover, this type of endogeneity bias caused the coefficient of next year's tax rate to be positive in the analysis of Gruber and Köszegi (2000), and the coefficient on lagged consumption to be negative in the study of Keeler et al (1993).…”