Population genetics and economic growth


Jason Collins


February 13, 2012

The title of this post comes from a 2002 paper by Paul Zak and Kwang Woo Park. The title is mildly deceptive, as the paper has many elements and ideas crammed into it beyond population genetics. The model described by the authors includes working, consumption, saving, marriage, genetic diversity, sexual selection, intelligence, beauty, education, the Flynn effect, family size effects and more. While many of these elements deserve consideration, this is ultimately the paper’s weakness. Even though most of the assumptions are reasonable and well supported in the literature, the resulting mix is hard to disentangle, with a few factors dominating the results.

Zak and Park built an age-structured model in which agents’ cognitive ability and beauty are genetically determined, with human capital a function of cognitive ability and education. The model agents are paired with potential partners over a series of rounds, and decide whether they will marry their potential partner based on their beauty and human capital (together called “pizzazz”). In each round, they weigh the potential benefits to marriage, including increased income if their partner has higher human capital, the joy of marriage due to their partner’s pizzazz and the potential partners in future matching rounds. If they agree to marry, they then decide family size, trading off consumption (in both their young adulthood and old-age) and children. These interactions drive the population dynamics and economic output in the economy.

The agents do not seek to maximise biological fitness directly. As I have posted before, this is a workable choice if you can offer a link between the factor that the agent seeks to maximise and their fitness. However, where agents gain utility from a basket of outcomes that are linked to fitness to varying degrees, this creates a fitness trade-off between the activities. In the case of Zak and Park’s model, the agents trade-off marriage, children and consumption of goods. As a result, an agent with lower preference for consumption of goods relative to children would have a biological advantage and could invade the population.

Zak and Park run their agents through a number of simulated scenarios. In the baseline scenario we see a typical evolutionary biology result - a preference for pizzazz increases the accumulated human capital and beauty of the population, reducing its variance. Low pizzazz people are rejected, causing them to disappear from the gene pool. Beauty increases towards its upper bound, while human capital can continue to accumulate. It is the increase in human capital that drives long-term growth. The base-line simulation generated a one per cent growth in human capital per generation over 40 generations, which the authors suggest is a reasonable approximation of the last 800 years.

In an increased inequality scenario, higher variance in traits for half of the population results in low growth as there are low marriage rates and reproduction. The increased inequality results in less pairings where each agent meets the others’ criteria. The population shrinks, with output bottoming out and eventually picking up as agents become less choosy and start to reproduce. Total output is largely a function of population size. A similar effect is seen when there is a bimodal distribution in beauty. Initially there are low marriage and reproduction rates as high pizzazz individuals reject those with low pizzazz. This causes lower population, with associated reduced total output, which recovers once people become less choosy. Greater genetic diversity also decreases marriage rates, population and output.

In each of these scenarios, output is largely tied to population size and the degree to which agents are willing to mate with whomever they are paired. As a result, the authors’ discussion of economic output is effectively a discussion of population size. Only in the pandemic scenario, where different segments of the population are eliminated, is any commentary on per capita income made. In that case, the authors note that there is little effect of the pandemic on the output of survivors.

The last scenario explored by the authors is love. Love increases the probability that agents agree to marry despite more dissimilar levels of pizzazz. This scenario makes for massively increased output with increased population. The authors see this as illustrating the benefits of the right balance  between diversity an assortive mating. Assortive mating drives an increase in average pizzazz, but if too strict, population plummets as no one is willing to mate with the agent they are randomly paired. They call this balance the “The Goldilocks Principle”.

This conclusion is interesting, but I am not sure that is the world we are in.  Many women may be increasing their standards as their income increases, but per capita income is also increasing despite lower marriage rates. The demographic revolution was not primarily due to lower pairing, but due to the timing of pairing and lower fertility within pairs. Assortive mating is relatively easy in our assorted world. A university educated person is likely surrounded by other people of similar quality.

Apart from a near identical paper that Zak and Park released in 2006 (with an extra section on inequality), there has been little further work in the footsteps of this model. I suspect this is because the paper has too many elements, and that there is more reward in using simpler models to explore each element in-depth.