An attempt to reclaim some ground for the rational choice model, in the spirit of Gary Becker (R.I.P.)
Macroeconomic models with micro-foundations commonly have a parameter for the subjective discount rate. Goods in the future are, naturally, valued less than present goods. But attempts to ground the discount rate in the standard rational choice model have ranged from ad-hoc (e.g. Irving Fisher, who adduced a number of psychological factors) to uninteresting truisms (e.g. Ludwig Von Mises, who argued that time preference must be positive because the opposite is unthinkable). Time preference is not, however, a sui generis component of the rational choice model. In fact, the preference for present goods over future goods masks two quite different phenomena: what I will call the opportunity cost effect, and the uncertainty effect. Separating these out will not only clarify the microeconomics of time preference, but will also go a long way in explaining what appear to be empirical irregularities under the assumption of a monolithic time preference parameter.
The uncertainty effect is the more familiar of the two, though the economics discipline is not in the habit of identifying it with uncertainty. Because the future is perceived dimly, the passage of time casts a shroud of doubt around all expected events. Accordingly, we are willing to pay some positive price both to bring uncertain future goods into the certain present, and to push certain and present bads into the uncertain future.
This uncertainty is, of course, subjective. The psychological factors predisposing someone to regard the future as substantially uncertain, or to count on enough certainty in the world to make long-range plans, are still for economics a black box. Nevertheless, by casting it in terms of expectations rather than “personality”, we can shrink the the size of that black box, even if we cannot entirely eliminate it, and derive fruitful empirical predictions from the ground gained.
There is some resistance to the identification of time preference with uncertainty. It is commonly assumed that time preference would be just as strong in the absence of uncertainty. Doubtless, it does not explain the whole of it. But the remainder can be accounted for by the opportunity cost effect. In short, because time is scarce and one’s lifespan uncertain, time spent waiting is time wasted.
Like the uncertainty effect, the opportunity cost effect results in a preference for present goods over future goods. The more rapidly I consume goods now, the more total goods I expect to be able to consume over the course of my lifespan. However, unlike the uncertainty effect, the opportunity cost effect also results in a preference for present bads over future bads. This effect will dominate as the occurrence of the bad becomes more certain. When going to the doctor for a shot, we want to “get it over with” quickly, rather than postponing it. The fact – often cited by behavioral economists – that anticipation is unpleasant and cognitively burdensome, is thus easily incorporated into the rational choice framework, and without bringing in some ad hoc disutility of anticipation.
The opportunity cost effect is, in the end, also subjective; but again, fruitful conclusions can be drawn from the ground gained. It will tend to dominate, for example, where one’s own mortality weights heavily on his mind. Though the uncertainty effect may make the prospect of pushing a bad past the end of one’s life enticing, we can observe it most purely in medical situations, where death is the alternative to some unpleasance, rather than an escape from it. Where there is a will to live, then, a sooner operation will be preferred to a later, even if the odds of death are not reduced by the earlier operation. Wherever one expects to hear “Let’s get it over with”, there the opportunity cost effect dominates.
Still, in most cases, the uncertainty effect will outweigh the opportunity cost effect. On net, I will prefer to bring goods into the present, and push bads into the future. However, if these two effects combine to form what we observe as a pattern of time preference, then the marginal propensity to bring goods into the present will be greater than the marginal propensity to push bads into the future. Quite simply, the effects compound for goods, and cancel for bads. This prediction is quite different from the symmetry implied by a monolithic time preference, and one that appears to be borne out by experience.
Formulating time preference this way highlights the usefulness – and the limits – of economic models featuring infinitely lived households. The usual justification – that any PDV of a good 80 years in the future is sufficiently indistinguishable from zero at normal discount rates – begs the question if those discount rates are a function of expected lifespan. An infinitely lived individual (or household, assuming bequests), having no opportunity cost effect, might plausibly have a discount rate sufficiently low to make benefits 80 years removed seem quite valuable.
Such models will be useful, on the other hand, where we need to abstract away from the opportunity cost effect and to isolate the uncertainty effect. Any model, then, that features infinitely lived households, perfect certainty, and a positive discount rate, cannot reasonably claim to have micro-foundations. An infinitely-lived agent facing a world with perfect certainty would be, all else equal, completely indifferent between an apple today and an apple next week. In other words, his discount rate would be zero. If (per Mises) this result is unthinkable, then it is because that world is unthinkable, not because it does not follow from the postulates of human action.