Evolution of time preference by natural selection

One of the few areas where there is active research on the link between evolutionary biology and economics is the evolution of economic preferences (some papers in this area are in my economics and evolutionary biology reading list). Economic preferences are the way an actor will rank a set of choices based on characteristics such as the amount received, the probability of certain outcomes, and the timing with which the outcomes are received.

Preferences about the timing of an outcome are known as time preference (often measured as a discount rate). Someone who values goods now much more than goods received later has a high rate of time preference, while someone who gives goods in the future a relatively higher value has a low rate of time preference. We would call someone with a low rate of time preference patient.

So, what rate of time preference would have evolved under the forces of natural selection? Discounting the future makes sense in an evolutionary context as future outcomes might not be realised, population growth might make benefits received in the future worth less, and there is the chance of events such as death.

One of the seminal papers in the analysis of the evolution of time preference is Alan Rogers’s Evolution of Time Preference by Natural Selection (ungated version here). Time preference had been touched on before by R.A. Fisher, and in an earlier paper by Hansson and Stuart who examined intergenerational time preference. But Rogers’s paper was the first to look at this question on an intragenerational basis, which is the context in which economists usually consider it.

Rogers examined the optimal same-age transfer  that would be made from a mother to her daughter (e.g. from the 20-year old mother to the 20-year old daughter). In making such decisions, the mother would need to consider the remaining reproductive life of her daughter, that her daughter is only 50 per cent related to the mother, and the rate of population growth. As the transfer is same-age, the mother and daughter have the same remaining reproductive life. If there is no population growth (which was effectively the case for most of human history), only the degree of relatedness would matter and the discount factor is effectively one half per generation. Under these conditions, Rogers argued that the long-term real interest rate should be about 2 per cent per year.

As for the analysis by Hansson and Stuart, this rate appears low against measured discount rates, which are typically above 10 per cent per year. A simple static analysis of this nature is missing something. Some other papers grapple with this issue, and I will post about them soon.

Subsequently, Robson and Szentes argued that there are “serious problems with Rogers’ model” (ungated version here and extended version here), and that a particular rate of time preference should not flow from the analysis. They argued that unrealistic assumptions drive the results, including the assumption of identical offspring, which is not the case when offspring vary with age, and the assumption of a single same-age transfer where there are many possible same-age transfers (e.g. 20-year old mother to 20-year-old offspring, 30-year old mother to 30-year old offspring). Robson and Szentes also showed that the rate of time preference would depend upon the nature of the survival function faced at each age (i.e. the probability of death).

Regardless, Rogers’s paper is an important one, and despite a couple of earlier pieces of work on the evolution of time preference by other authors, Rogers’s paper is often seen as the seminal paper that kick started the evolutionary analysis of preferences. That is not a bad legacy to have.

Update: Alan Rogers responds in the comments.

4 thoughts on “Evolution of time preference by natural selection

  1. I would argue that my results are not flawed, and neither are those of Robson and Szentes. They simply ask different questions. In my view, evolution has shaped the genetic bases of preferences only slowly, whereas market forces equilibrate quickly. For this reason, I ignored markets when modeling the evolution of preferences, and I took preferences as fixed when modeling markets. Robson and Szentes, on the other hand, built a model in which both forces (markets and evolution) operate on similar time scales, and they solved for the simultaneous equilibrium of both forces. It is a brilliant piece of mathematics, but I don’t see it as relevant to humans. It is likely that human time preferences are far from evolutionary equilibrium, because credit markets are evolutionarily novel.

    1. Thanks for your reply Alan. My choice of the word “flawed” was probably not the right one, so I have replaced “shown to be flawed” with the precise claim from Robson and Szentes’s article that there are “serious problems”. I’ve also added a pointer to your response.

      I’m going to have to put some more thought into your argument about whether their approach is relevant to humans before I can offer any intelligent comment on your reply.

  2. I’d also like to clarify another point, concerning the assumption of same-age transfers. This critique is about section II of my 1994 paper. It ignores section III, which relaxes the assumption of same-age transfers. In my view, section III has the important results; section II was an effort to make those results intuitively accessible by paring the model down to bare essentials. It was never intended to be realistic.

  3. Not sure if this is relevant, but I’ve spent a good deal of time thinking about the early development of tools in paleolithic societies. A tool, by definition, saves more time and effort than were required to make it. OTH, if one lives too close to the absolute subsistence level it may not be possible to take time out to fashion a stone tool without starving to death in the meantime. One lives literally from hand to mouth.

    The conclusion I am inclined to draw is that one’s future time preference may depend on how much time one can spare from subsistence food production. Generalizing, when you are flush you can afford longer time horizons. Does this ring true?

Comments welcome

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s