What Drives Economic Growth?

Economic growth is one of the defining features of the modern world, so much that most of us assume that from one year to the next, we should expect greater incomes, more technological advancement, and growth in our investments. But economic growth on a scale that is noticeable on a human lifetime has not been the normal human experience, and it is not guaranteed that growth will continue into the future. Indeed, barring major revisions to our understanding of physics, it is guaranteed that it will eventually cease.

Today and over the next few posts, my goal is to understand how economic growth occurs. This is not merely a matter of academic curiosity. I fear that policymakers and the general public have become too complacent about growth, treating it as an automatic and inevitable process. My hope is that understanding the mechanisms that drive growth will help steer policymakers to constructive policies, and that understanding that growth is not automatic will insure that they treat growth as a priority.

To begin this series, I want to introduce a concept known as Smithian growth, named for Adam Smith and expounded in his foundational work, The Wealth of Nations (Smith 1776). As usual, this work is conducted as part of a Living Literature Review grant from Coefficient Giving, but the contents are strictly my own views.

Adam Smith does not present the concept that would be named after him in a single package, but rather, he develops the principles throughout The Wealth of Nations. The core mechanism of Smithian growth is division of labor. When market sizes are larger, there is greater potential for individuals to specialize. Market size is, in turn, determined by two factors: overall population size and the degree of market integration.

Specialization, according to Smith, is important for two major reasons. First, by repeating a narrow task, a worker can improve their skill via learning-by-doing. Second, by focusing on a smaller task set, there is less of a cost in the form of task-switching.

Smithian growth does not rely on technological advancement, yet it carries the potential for a self-reinforcing process of growth, albeit with limits as division of labor faces diminishing returns. Specialization creates more wealth, which allows an expansion of market sizes, which create more opportunity for specialization, which creates more wealth.

Smithian growth is far from the only game in town, and so before we proceed more deeply with that topic, let us briefly survey the other major growth mechanisms. As described by Aghion, Akcigit, and Howitt (2015), Schumpeterian growth, named for the economist Joseph Schumpeter, has technological innovation as the driving mechanism of growth. The basic tenets of the Schumpeterian model are as follows:

(a) Long-run growth results from innovations; (b) innovations result from entrepreneurial investments that are themselves motivated by the prospects of monopoly rents; and (c) new innovations replace old technologies. In other words, growth involves creative destruction.

The Swan-Solow growth model (or simply the Solow model) is named for a pair of papers: Solow (1956) and Swan (1956), which independently developed the model. In its most basic form, the Solow model presents the size of the economy at time t, Y(t), as a function of labor L and capital K as follows:

Here, α and β are positive parameters that add to 1 and are determined empirically. R is the residual. The Solow model in its most basic form is something of a black box, since we can simply define R = Y/(LK), and the equation is true tautologically. Trying to understand the residual has been the subject of much work since 1956 and a topic that I plan to discuss later on, and in fact that I did discuss last year in the context of Ayres and Warr (2005), who argue that the residual can be expressed as energy consumption.

An important contrast between the Solow model and both the Smithian and Schumpeterian models is that, unlike the latter two, the basic Solow model treats both market size and technology as exogenous to the model, meaning that, while market size and technology might affect economic growth through the residual, economic growth does not affect market size and technology.

Why Nations Fail (Acemoglu and Robinson 2012) argues that inclusive political and economic institutions are central to growth. Considering growth in historic and current societies, the authors find that inclusivity of institutions is the decisive factor that distinguishes fast-growing from stagnant and failing societies. The authors won the 2024 Nobel Prize in Economics for their work. The thesis of Why Nations Fail is agnostic to the direct mechanism for growth, and it is compatible with Smithian, Schumpeterian, or Solovian models.

Lucas (1988) proposes human capital as the central explanation for growth. Human capital is built through education, skills, and health. Again, human capital is compatible with, but not reducible to, the Smithian mechanism for growth. Gallup, Sachs, and Mellinger (1999) propose geography as a key determinant of an economy’s growth potential. Many researchers, such as Bloom, Canning, and Sevilla (2001), have argued that a demographic transition, or a transition to low birth rates, is a driver of growth by temporarily reducing the dependency ratio of a society.

These explanations are generally complementary, and each may have value in its own context. For the remainder of today’s post, we will focus on Smithian growth.

When economic growth in a society is observed over time, there may have been both population and trade growth as well as technological advancement. A major challenge is to disentangle how much those two factors contributed to growth. To test Smithian growth empirically, we many need to find a society that saw significant change in population and/or trade integration, but not in technology level, and observe the growth rates of that society.

This is exactly what Ortman and Lobo (2020) do with Pre-Hispanic Northern Rio Grande Pueblos, in modern-day New Mexico, from the years 1250 to 1650. Over that time, Puebloan society in New Mexico achieved a fourfold increase in economic activity, or an average annual growth rate of 0.8%. Over this time, both technology levels and regional population (though not population of individual settlements) were relatively constant in Puebloan society, isolating trade integration as the decisive factor behind growth.

At this point, long-time readers might notice similarity between Smithian economic growth and agglomeration economies in cities. Indeed, in a previous post, I cited Lobo et al. (2013), with the lead author the same Lobo as in Ortman and Lobo (2020), who performed a regression on (log of) economic output against (log of) population for metropolitan and micropolitan areas in the United States and found a slope of 1.146 with an impressive R² value of 0.97. Smithian growth for a full economy is, in a sense, little more than the principle of agglomeration economies extended beyond a single city, defined here as a unified labor market.

Getting back to Ortman and Lobo (2020), they spend the first third of the paper reviewing the principles of settlement scaling theory. They derive the 7/6 power rule, which is that socioeconomic quantities such as economic output tend to vary with the 7/6 power of population size, assuming a well-mixed population. Their derivation is similar to the derivation I discussed here, which was drawn from Bettencourt (2013).

They proceed to discuss specialization. The number of tasks performed in a settlement of N well-connected people is shown to be N^(1-δ), where δ is shown to be 1/6. The number of tasks performed by a given individual is thus N^(-δ). In other words, while a society as a whole becomes more versatile with growing population, individual members become more specialized. I find this to be intuitively sensible.

To identify population, the authors use a technique known as uniform probability density analysis, which estimates the population of a region based on the architectural footprint and the pottery assemblage. They find that the population of the region was generally stable over time, but the average size of settlements increased. This implies that the number of settlements decreased, and the population generally consolidated over the study period.

Three additional metrics are used to estimate growth dynamics.

• The ratio of potsherds derived from painted fine-ware serving vessels to potsherds derived from utility-ware cooking vessels, which can be derived from archaeological data, is a metric of economic output.

• The average floor area of rooms associated with a component is a proxy for the number of possessions that individuals possesses and thus is also a measure of economic output.

• The ratio of chipped-stone fragments to utility-ware potsherds in samples is a proxy for productive diversity, which is closely related to specialization. In a more specialized economy, a smaller portion of the population needs to make pots, and therefore there should be more pots per stone tool.

The ratio of fine to utility sherds generally increased until the 16th century, at which time it decreased. The ratio of chipped stone to utility sherd generally decreased, albeit unevenly, until increasing in the 16th century. The average (log of) site size generally increased through the whole study period. Mean room area decreased until the second half of the 14th century and then increased until a decrease in the second half of the 16th century.

The authors find that their measures for economic output and activity diversity fall within the margin of error of the theoretical predictions for a society without technological change. It is a very satisfying conclusion that provides a compelling case for the validity of Smithian growth in Pueblo society. But some aspects can be questioned.

Ortman and Cooper (2021) find that pottery sherds found in the Chifeng region in Northern China can be viewed as a socioeconomic rate, and they find that the theoretical superlinear scaling pattern holds for settlements in that region. They suggest that the volume of pottery sherds is a good proxy for overall consumption, but this is unproven, and it is plausible that pottery volume, indicating social or ritual activity, behaves differently from overall consumption.

Keuschnigg, Mutgan, and Hedström (2019) raise the fundamental objection of causation to urban scaling. Although it can be readily observed that larger cities tend to have higher economic output than smaller cities, it is less obvious whether size causes output to increase, or whether the reverse is true, in that more prosperous cities attract residents and thus grow larger. Here, the authors analyze data regarding cities in Sweden and find that at least 39% of the observed superlinear output scaling effect can be explained by population characteristics and selective migration. It is not clear whether such a mechanism would apply to a preindustrial region such at the Pueblo, but at least that offers another data point to increase the robustness of urban scaling theory.

Finally, we’ve noted before the practical challenges in delineating city boundaries for the purpose of performing a scaling analysis. The rules of thumb, that a “city” should correspond to a unified labor market and Marchetti’s Constant, that a city is a 30 minute commuting radius from a central business district, leave much room for interpretation. Cottineau et al. (2015) find that the observed scaling properties are sensitive to the manner in which city boundaries are demarcated. It is difficult enough to objectively demarcate cities in modern societies, and it is much more so to do so with ancient cities based on archaeological evidence.

These weaknesses notwithstanding, I find that Ortman and Lobo (2020) provide compelling evidence for Smithian economic growth among the Pueblo in New Mexico.

Today, we have briefly surveyed models of economic growth and have taken a close look at Ortman and Lobo (2020), which finds that Smithian growth was sustained for centuries among the Pueblo in New Mexico.

Nevertheless, modern understanding of economic growth remains fragmented, with several complementary approaches pursued by governments that view growth as a high priority. With an eye toward Schumpeterian mechanisms, governments pursue focused research and development. With an eye toward Smithian mechanisms, governments pursue infrastructure projects such as roads and airports, free trade, and welcome immigration. Skepticism toward growth, at least in part, motivates the inverse of these policies.

Regardless of one’s philosophical orientation toward economic growth, good policy requires an accurate understanding of how growth actually occurs. That Smithian growth has occurred historically and still occurs in the modern world is, in my view, beyond reasonable dispute. But the magnitude of that growth, and its importance relative to Schumpeterian mechanisms, is debatable. Next time, we will consider two papers—Chilosi, Lecce, and Wallis (2025) and Peretto (2015)—that cast some doubt on the importance of this mechanism.

This weekend, I posted an article on the main Scaling site about the emergence of more extensive human large-scale cooperation in the Neolithic. That post closely follows a blog post from last month. I still need to prepare the material in Part II of the series for the Scaling site.