Natural selection and saving


Jason Collins


November 29, 2013

In the academic literature at the intersection of economics and evolutionary biology, evolution of time preference (patience) is one area that has received much attention. This makes some sense, as most economic models that consider decisions over time include time preference. Time preference is normally included in the model through a discount rate of a fixed value, so an evolutionary analysis might help in determining what that value is.

One of the first articles to ask what rate of time preference might be expected to evolve was by Ingemar Hansson and Charles Stuart, who examined the intergenerational rate of time preference. Intergenerational time preference is reflected in the rate of saving of one generation for the benefit of following generations. The first generation sacrifices their own consumption for consumption of their descendants.

This paper is another of those that, at least for the headline result, should strike one as obvious. Time preference would evolve such that when the person seeks to maximise utility (in an economic sense), they will also be maximising fitness. In this case, they would follow the “golden rule” of saving, which is the rate of saving that maximises the steady-state level of consumption across generations.

This outcome contrasts with empirical evidence that the golden rule tends not to be followed, with savings rate in developed, Western countries well below the golden rule. Savings rates are somewhat closer to (and sometimes argued to be higher) than the golden rule level in Asian countries.  Why there is this departure requires consideration of other factors, such as aggregate risk (as Robson and Samuelson consider in this paper).

Hansson and Stuart determined that, if people followed the golden rule of saving, the discount rate would reflect the long-term population growth rate. This requires that no generation be weighted less than any other generation, which indicates a strong concern about future generations.

Based on population growth in a number of European countries since 1500, Hansson and Stuart suggested that the discount rate implied by their model would lie between zero and a few per cent per year, or between zero and one per generation. While they selected population growth rates since 1500 largely due to the lack of earlier demographic data, their upper bound estimate implicitly assumes that evolution of time preference can be relatively rapid to reflect recent population growth rates. In this case, it would also suggest that discount rates are increasing. As population growth increases, one should cut saving and focus on consumption today.

Moving beyond the headline result, an interesting part of the paper is when Hansson and Stuart ask what are the consequences of their model for economic growth. A harsh environment may increase the marginal benefit of consumption (the benefit from each additional unit) and decrease population density, which will in turn increase the level of labour that each person produces. As labour supply increases, this in turn increases the productivity of and the level of capital.

Following this logic, the authors suggest that harsh natural environments select for genotypes that have a stronger preference for saving, leading to an equilibrium with low population density and high per-capita capital. Selected traits include a preference for work and accumulation of physical capital. Hansson and Stuart suggest that this might explain why humans left the Malthusian state first in regions with harsh winters.

This results seems intuitive, although it is an interesting contrast with a Malthusian model of the economy. In a Malthusian model, high levels of technology and productivity are reflected in high population densities. To reconcile Hansson and Stuart’s thinly populated, harsh environment with this Malthusian picture, it might be necessary to imply that preferences were shaped when the first populations entered the harsh regions, and that preferences have not significantly changed despite these previously unpopulated environments now having much higher population densities.

One interesting comment in the paper refers to a methodological argument by Stigler and Becker, who stated that it is worth assuming that preferences do not differ importantly between people. Hansson and Stuart suggest that this is the case where the population to be modelled consisted of a homogeneous people from a given environment, but that this would not be as applicable in a “melting pot” such as the United States.