Population and Economic Growth

I am glad to be back after a hiatus in the Philippines two weeks ago. I’ll have much more to say about this trip later, but I do dislike being away from the blog. I very much hope to get back to a weekly writing schedule, but for now my work schedule does not permit that.

For as much as I have written about population and fertility trends, I have neglected to discuss the basics, and that is something that I hope to correct over the next few weeks. This work is part of my Scaling in Human Societies project, which is generously supported by OpenPhilanthropy’s Living Literature Review program. This post is a less developed version of the material that I intend to go on the website later. I also intend to further discuss energy and economic growth, as I did for the previous two posts, but I confess that this project has so far not been as fruitful as I had hoped.

Population and Economic Growth: Theory

There are many dimensions to the question of how population growth affects average well-being, but today I’ll stick with the question of the size of the economy, and gross domestic product per capita in particular. GDP is a fairly straightforward metric, and while it does not fully encapsulate human wellbeing, it is a reasonably good proxy for most aspects of wellbeing that we care about. As explained by Sibe, Chiatchoua, and Megna (2016), there are two main hypotheses on how this relationship works.

First is the Malthusian hypothesis, named for the granddaddy of pessimistic analysis of population growth, Thomas Malthus. In his original essay, Malthus (1798), Malthus posits that population tends to grow in a geometric (exponential) fashion, while food supply grows in an arithmetic (linear) fashion, and so inevitably, at some point, food supply must become inadequate and the population is pushed down to a subsistence level. In modern usage, the term “Malthusian” has become a byword for any viewpoint that increased population, in some way, depresses the average standard of living.

The second hypothesis can be called the Kremerian hypothesis, named for Michael Kremer’s work, Kremer (1993), which I will discuss in detail below. Taken broadly, a Kremerian analysis is one in which an increased population increases the average standard of living. The main mechanism for this, according to Kremer (1993), stems from the nonrivalry of ideas. This means that if I use an idea to improve my life somehow, this does not diminish the ability of anyone else to use the same idea, and so in a larger population, there is more ultimate value to creating an idea. Conversely, in a larger population, technological progress tends to be faster because there are more people who have the potential to come up with ideas.

Darrat and Al-Yousif (1999) use the label “revisionist” to for what Sibe, Chiatchoua, and Megna (2016) call “Kremerian”, contrasting this to the Malthusian view that is (was?) “orthodox” in development economics. They also articulate the “transition” view, which is that population size does not have a great effect on economic performance, but a poor economy tends to foster greater population growth.

Kremer’s Long-Term Growth

Let us now take a closer look at Kremer (1993), which appears to me to be the most important paper establishing a positive effect of population growth on economic growth. The author, Michael Kremer, has done much notable work in development economics and won the 2019 Nobel Prize in Economics, along with Esther Duflo and Abhijit Banerjee, for work with randomized control trials for poverty alleviation (work about which I imagine I would write less favorably if I were to address the topic in detail). Kremer was also a coauthor of Market Commitment Working Group (2005), which developed advance market commitments for vaccines and was a major intellectual force behind Operation Warp Speed to quickly develop COVID-19 vaccines.

Kremer (1993) notes that much work in development economics treats technological progress as exogenous, which means that it happens at its own pace and is independent of the dynamics of the economy. He argues that it should be regarded as endogenous instead, and in particular, technological progress is faster when the population size is higher. Technological progress is, in turn, the most important driver of long-term economic growth.

The main mechanism that Kremer (1993) posits for this is, as noted above, the nonrivalry of ideas. He also discusses several mechanisms which will look familiar to anyone knowledgable of agglomeration economies, which explain why, all else equal, larger cities tend to have greater per-capita economic output than smaller cities. It is little surprise that the same factors are at work on the level of a full economy. More populous regions offer a greater potential for specialization, which is more efficient; intellectual fruitful contact; and larger market sizes. I would add to this Boserup’s (1985) argument that a greater population density allows for more efficient infrastructure such as roads and rail.

Kremer (1993) builds a model that combines the aforementioned observations with the Malthusian hypothesis that population is limited by the ability to satisfy a population’s needs. The result is that population grows superexponentially over the long run, in that the growth rate also increases. Excepting some disruptive occasions in history, such as the Mongol conquests and the Black Death, an accelerating growth rate is generally observed from 1,000,000 BC—to my knowledge the earliest date used seriously in an economics paper—until the middle of the 20th century. Since then, though, it is clear that population growth is slowing and may reverse in the 21st century, and clearly without a pressing Malthusian constraint on population size.

Kremer refines the model with several income effects. First, he considers how rising income increases the opportunity cost of childbearing and thus deceases the average number of children born per couple, leading his model to predict that rising income will eventually cause population growth to reverse for non-Malthusian reasons. Second, he considers various mechanisms by which a rising average income may increase (or possibly decrease) research productivity, so that population size is not the only factor that influences the rate of technological progress.

Kremer (1993) notes a natural experiment to validate the model. Around 10,000 BC, rising sea levels disconnected some formerly connected regions of the world, which would not be reunited until the Columbian exchange around AD 1500. In accordance with his model, by 1500, larger regions that could support larger populations tended to be more advanced technologically and economically. For example, Afro-Eurasia was more advanced than the Americans, which in turn was more advanced than Australia, which in turn was more advanced than Tasmania. On Flinders Island, a small island off Tasmania, the human population went extinct entirely about 2500 BC.

There are a few reasons to be cautious about Kremer's (1993) results. On the subject of research productivity, or the amount of technological progress that results per researcher, Kremer notes the positive agglomeration effects of a higher population as discussed above, and he also notes in passing the negative effect of duplication of research effort. I wonder if the negative effects are greater than he lets on. I have discussed many times the famous Bloom et al. (2020), which documents sharply decreasing research productivity across a wide set of domains. Decreasing productivity is balanced by a sharp increase in the total volume of research, thus maintaining a steady rate of progress. If it can be shown that decreasing productivity is because of increased effort, then it would seriously undermine the case for faster progress with a larger population.

I am also uncomfortable with the way in which “technology” is represented numerically. It is not at all clear that technology should advance linearly in the amount of research effort expended, nor it is clear that carrying capacity should grow linearly in technology. To be fair, Kremer does address these concerns, but not in a way that makes me confident in the mathematics. To borrow a phrase from cosmology, Kremer’s model also suffers from a fine-tuning problem, in that so many parameters are chosen to fit observed data that the model is nearly unfalsifiable.

The above criticisms notwithstanding, Kremer (1993) is a valuable and very plausible account of how population contributes to economic growth.

Population and Modern Growth

Now let us consider a few studies that focus on the recent past. I’ll be referring to Granger causality a few times here, which I discussed last month. Briefly, Granger causality is a statistical test to determine, given two times series, if one “causes” the other by testing for correlation between lagged values of the first series and the second series. As noted last month, the statistical test does not really prove causation in an epistemological sense, and one needs to have a plausible causal mechanism in addition to the data.

First is London, Cayssials, and González (2022). They consider panel data of 111 countries consisting of GDP and population from 1960 to 2019. They find that the countries can be naturally clustered into three groups (with a handful of countries that don’t neatly fit into any group), which they label as mature, young, and transitional countries. In the young and transitional countries, bidirectional causality is observed between GDP and population, meaning that each one causes the other. In mature countries, however, population is not observed to cause GDP. Just as we observed before with energy and GDP, the relationship between population and GDP might break down for high levels of wealth. However, I’ll comment below on why this might be more of a statistical artifact than a real phenomenon.

Sibe, Chiatchoua, and Megna (2016) perform a similar causality test on panel data of 30 countries, covering 78% of the world’s population as of 2013, on GDP and population from 1960 to 2013. The result is also bidirectional Granger causality between the two variables.

Garza-Rodriguez et al. (2016) look specifically at Mexico from 1960 to 2014, and they also find bidirectional causality between GDP and population.

Chang et al. (2017) find a decidedly mixed picture. Considering data from 1870 to 2013 and 21 countries, they find that GDP causes population growth in some, that population growth causes GDP in others, and both cause each other in still others, and in some, that there is no causal relationship in either direction. Overall, it is a minority of countries in which population growth Granger causes GDP growth.

Most studies that find a causal relationship between population and GDP find a positive, Kremerian relationship; an increase in population causes GDP per capita to increase. A minority of studies find a negative, Malthusian relationship, in that an increase in population causes GDP per capita to decrease. Yao, Kinusaga, and Hamori (2013) is one such study. Using a vector error correction model (some technical details explained here), this paper finds that population had a negative impact on GDP per capita in China from 1952 to 2007.

There are many more studies that I cannot cover for lack of time and space. See London, Cayssials, and González (2022) for a more thorough list. A scan through their chart confirms that most studies find either a positive effect or no statistically significant effect of population on GDP per capita, and a small number find a negative effect.

Interpreting the Numbers

All of the studies in the above section look for statistical causal relationships between population and GDP per capita for various countries. A few words of caution are in order in interpreting the numbers.

First, it cannot be emphasized enough that correlation is not causation, even though we call the technique Granger causation. As shown in Kremer (1993), there are plausible mechanisms for causation, but it is nevertheless still conceivable that the observed relationships are not causal.

Second, when searching for Granger causation in any context, time frame matters. In particular, common sense suggests, as argued for example in Simon (1998), that childbearing exerts a net cost on society for the first 18 years or so of the child’s life, as the child requires resources in the form of food, clothing, living space, parental attention, education, and so forth. After that time, the child hopefully enters the workforce and becomes a net positive for society. We should therefore expect an increase in birth rates to have a negative effect on GDP per capita for the first 18 years and a positive effect thereafter, though the effect may turn negative again after the person retires.

Third, it is not clear that results at the level of specific countries are meaningful. If we accept Kremer’s argument that nonrivalry of ideas is the main mechanism by which population contributes to wealth, we note that in the modern world, ideas freely cross national borders, and so the population of a specific country may not be very important. This may also create a free-rider problem, whereby national governments benefit from knowledge spillovers generated by a higher world population but enact policies to avoid the perceived domestic costs of higher birth rates.

There is much more I want to review, and the above represent only a first pass at the topic. From this, though, I would conclude that the balance on evidence indicates that the Kremerian hypothesis, as opposed to the Malthusian hypothesis, holds more weight, and we should expect that population growth to contribute to economic growth. Conversely, the expected coming population stabilization and decline will be a headwind to per-capita wealth. However, the evidence is not decisive, and if the effect was very large, the evidence would probably be more decisive. I expect that under reasonable forecasts of population decline in the late 21st century and beyond, society will be OK.

Why It Matters

To conclude for today, I want to step back and consider an important implication of this discussion, and not the obvious question of whether countries should adopt more pronatalist policies such as a child tax credit.

One of the central moral teachings of the Catholic Church, and one that I take very seriously, is of the sanctity of life. People tend to understand this phrase as being synonymous with opposition to abortion, but it is broader than that and includes opposition to killing an innocent human in any context, including euthanasia (Clarke (2022)), capital punishment, and dehumanization of immigrants. It is a straightforward teaching, yet few people, including few devout Catholics, believe in it in its entirely.

As much as I would like to believe that the sanctity of life is rooted in an understanding of the intrinsic value of a person as a creature of God, the blunt reality is that most people value others in instrumental terms. If the relationship between population and per-capita GDP is in fact breaking down, then an unhappy consequence is that citizens of wealthy countries are increasingly seeing new neighbors across the street, immigrants, or children as, at best, to be grudgingly tolerated.

In the 1972 final report for the Commission on Population Growth and the American Future, chairman John D. Rockefeller 3rd had the following to say.

After two years of concentrated effort, we have concluded that, in the long run, no substantial benefits will result from further growth of the Nation’s population … . We have looked for, and have not found, any convincing economic argument for continued population growth. The health of our country does not depend on it, nor does the vitality of business nor the welfare of the average person.

Rockefeller could confidently state this purely instrumental view of the human person because it was the common wisdom at the time.

The trend of population peaking and declining looks durable and likely to be the reality across the world regardless of policies that are adopted. In response, it is tempting to seek to decouple population from wealth by heavy investment in technologies such as artificial intelligence and robotics. This will probably happen no matter what, but it is important to be aware of the hazards. (Further) decoupling may erode our basic concepts of human rights, if humans are increasingly seen as consumers rather than creators in the world. Then the atrocities associated with population control in the 20th century may be only a taste of what is to come.

Quick Hits

I mentioned Hoff (2010) before, an engaging overview of the Nixon administration’s aforementioned population commission. The paper includes the following passage.

Also in 1965, the National Review printed a supplement called “The Population Explosion,” which yielded nothing in apocalyptic rhetoric to the liberal doomsday literature of the era. In the lead article, “The Avalanche,” as the editors summarized it, “A science fiction novelist [Theodore Sturgeon, who incidentally fathered seven children] takes a hard look at the earth’s skyrocketing birthrate and admits it portends horrors even he finds hard to imagine.” 91 The author identified as the true science fiction of the era the food optimism promulgated by the “nutritional technological cohorts— the algae-and-yeast boys, the ranch-the-oceans fellows, and the transmutation-of-petrochemicals-into-proteins enthusiasts.” 92

I have been looking for a copy of this article without having to travel to Kansas, and so far I have not been successful. Can any sleuths help me out? As I’ve argued before, population control was so ingrained in the culture of wealthy nations in the 1960s and 1970s that it had a great influence of many other areas, especially legalized abortion and immigrant restriction, which are still felt even as “population control” has become a taboo phrase. This quote makes me wonder if there is also a connection between population control and opposition to high-yield food technologies. If so, that would be another big piece of the puzzle put into place.

Also on the population subject, mark your calendars for Friday, November 13, 2026. This is doomsday, according to Forester, Mora, and Amiot (1960). These authors observe increasing world population rates at the time, project the increase to the future, and conclude that this date is the time at which population will reach infinite. This conclusion was no doubt meant in jest, but it nevertheless illustrates that blindly extrapolating trends into the future, without understanding the dynamics of those trends, is a good way to make bad predictions.

And again on the population subject, two weeks ago there was a bombing at a fertility clinic in Palm Springs, California. The bomber was clearly a disturbed person, though his antinatalist ideology, and its more extreme variant of promortalism, is gaining traction. As this article explains, the term “antinatalism” was coined by David Benatar in his work, Better Never to Have Been: The Harm of Coming into Existence. I wrote about Benatar and other issues around population ethics two years ago. Most antinatalist arguments are based either on Malthusian concerns or on the belief that fewer or no children is a manifestation of women’s autonomy. Benatar presents a third strand of antinatalism, in addition to the word itself, that life is inherently not worth living. He argues that people might suffer, or live lives not worth living, and while creating a life not worth living is evil, creating a life that is worth living is morally neutral. I still fail to understand the rationale for this asymmetry.

And on a different subject, midterm elections occurred in the Philippines while I was there. I attended a rally (as a curious bystander, not a supporter) for Rodrigo Duterte, the former president (2016-2022) and former mayor of Davao City (several previous terms going back to 1988). Duterte won that election and is scheduled to take office in June, but he is currently in the Hague for charges related to extrajudicial killings in the drug war. Duterte bears much in common with other right-leaning populist leaders of the present and recent past, including Donald Trump, Boris Johnson, Jair Bolsanaro, Viktor Orbán, and others, including a flamboyant public persona and a checkered relationship with the rule of law.