I am interested in Arcidiacono et al (2007) which deals with whether forward-looking ‘rational addiction’ models best explain patterns of smoking among late middle-aged men. The suggestion is that one cannot observe whether decisions by young people to smoke or drink are consistent with rationally thinking through the health consequences of smoking or not.
As a general rule relatively few adverse heath effects occur in the first half of life. Young males who smoke, according to Hodgson (1992), have the same cumulative probability of surviving to age 35 as non-smokers. At age 45 (age 65, age 85 respectively) the corresponding ratio of the cumulative probabilities is 1.02 (1.18, 2.11 respectively) so that a smoker faces a 2% higher cumulative probability of being dead than a non-smoker at age 65, 18% at age 65 and 211% higher at age 85.
This evidence surprises me. The impression I had gained from other studies was that the smoking risks over middle ages were larger than these. For example, Doll et al (2004) examine smoking behavior of 34,439 male British doctors over a 50-year period and found that long-term smokers died 10 years younger than non-smokers – that is possibly consistent with the Hodgson data. But among men born around 1920, Doll et al found that prolonged smoking from early adult life tripled mortality to 43 per cent among smokers compared to 15 per cent among non-smokers between ages 35-69 – this seems much larger than the increased cumulative probabilities cited above. Or am I missing something? Moreover, according to Doll et al cessation at age 50 halved the mortality hazard and cessation at age 30 avoided most of it. There are huge public health gains from getting people to quit cigarettes.
According to the Hodgson claims, the disease consequences of smoking only really bite in the second-half of a male’s life. This means that a young male may choose to smoke rationally with complete account taken of future harmful consequences (without relying on hypotheses of above-average discount rates or impulsiveness levels) simply because, even with moderate discount rates, the cost of a premature death in the second half of their life has low present value.
In fact, examining data on young males won’t tell you if they are rationally choosing to smoke with a reasonable discount rate or behaving irrationally and myopically by enjoying the future now without regard to future consequences. In short you cannot test the ‘rational addiction’ hypothesis of Becker and Murphy (1988) using data on young males.
The Arcidiacono et al (2007) paper itself examines whether late middle-aged males – who do face significant extra mortality risks from smoking (and drinking). Do these males update their consumption behaviour in the face of adverse health shocks? Do they display rational forward-looking behaviour or is their behaviour myopic? They find that older males are rational addicts with moderately high discount rates who fully anticipate the risks associated with heavy smoking and drinking even though smoking, while young, can make sense at moderate discount rates. Myopic models in this setting imply much higher levels of drinking and smoking than do occur. Rational addiction models imply lower levels of these consumptions because individuals account for their future health costs. This is hardly surprising given that the conventional wisdom supposes that older people are more mature and have lower discount rates than adolescents. They will also be observing the health consequences of smoking.
Smoking when young does pose particular neurobiological costs as I have suggested before that are real even if they are not necessarily fatal. Also those who begin smoking or drinking when they are young are more likely to remain addicted to cigarettes or to become alcoholics than those who start later in life. It is thus costly to smoke when you are young even though, according to Arcidiacno (2007) it may be rational to smoke if you don’t mind smelling like a compost heap throughout your life, and don’t worry too much about dying a decent way down the track because you discount the future at a plausible rate.
I think this is an interesting viewpoint. In the past I have argued that young people behave irrationally because they have high discount rates – they are impulsive and so on. The argument constructed by Arcidiacono (2007) suggests that this issue is unlikely to be resolved by appealing to evidence and that observationally it doesn’t matter much because of the life-profile of health risks.