The Limits of Smithian Growth

Pardon the delay for this post. Today, we will continue from our last post our examination of Smithian economic growth. Named for Adam Smith, arguably the most important founder of economics, Smithian growth is economic growth that is driven by market size. Under a Smithian growth model, markets, and thus the size of the economy (in per-capita terms as well as absolute terms) can be expanded through trade integration and population growth. We reviewed last time Ortman and Lobo (2020), which demonstrates how Smithian growth applied to Pueblo civilization in modern-day New Mexico before the Spanish were established.

However, Smithian growth is not the only game in town. We briefly reviewed several growth models in the last post. Most notable among them is the Schumpeterian model. Named for the economist Joseph Schumpeter, the driving mechanism behind Schumpterian growth is technological innovation and creative destruction. Under this view, politicians interested in economic growth would pursue technological advancement, rather than trade and population growth, as the first policy.

Today, we’ll examine a few more historical cases of (suspected) Smithian growth. Then we will consider why Smithian growth has fallen out of fashion in recent years. As usual, this work is supported by a Living Literature Review grant from Coefficient Giving, though the work represents my own conclusion and not those of the funder.

Many researchers have identified Smithian growth phenomena in the preindustrial world, and we can only look at a few of them today. One of them, as noted, is Ortman and Lobo’s (2020) review of pre-Spanish Pueblos.

Kelly (1997) provides a Smithian analysis of growth in the Sung (or Song) Dynasty (AD 960-1279). At the dawn of the Sung Dynasty, China was in a subsistence state with very little industry or large-scale trade. By the end of the 11th century, China built a prosperous economy that was not surpassed anywhere in the world until the eve of the Industrial Revolution. Large-scale industries developed, particularly ironworking. At the heart of it, according to Kelly, is the extensive canal system and the trade integration that it enabled.

The paper has two main components. In the first part, Kelly (1997) builds a theoretical model of how interconnection fosters growth. The model has an interesting and counterintuitive feature: growth shows a threshold pattern, whereby the size of the economy, as a function of the extent of the transportation network, is relatively flat until the size reaches a critical value. After the critical value, the economy grows rapidly in the extent of the network until the entire region is integrated into a unified trading network. Beyond that point, there is little benefit to further development of the transportation network. This is counterintuitive, at least, if your intuition is that economic growth should occur gradually as a function of the extent of the network.

In the second component, Kelly (1997) applies this model to Sung China. Reliable figures for GDP were not available, and so he uses the real value of currency in circulation as a proxy. Kelly finds that per-capita GDP was relatively flat until the 11th century, grew rapidly in the 11th century as the canal system was fully built out, and returned to stagnation from the 12th century, all in accordance with the predictions of the model. According to the model, once the canal system was complete, no further gains from integration should be possible, and in the absence of Schumpeterian technology-driven growth, stagnation should be expected.

The use of money in circulation as a proxy for GDP can be questioned. Kelly himself acknowledges that the Wicksell effect may apply, in that his reconstruction of the growth of the money supply might be explained by the higher velocity of money that is typical in a more commercial society, as opposed to actual economic growth. But to my knowledge, the basic picture of rapid growth in the 11 century, followed by stagnation, is not in serious dispute. More recently, Broadberry, Guan, and Li (2017) agree with the notion of an efflorescence in the Sung Dynasty, with GDP per capita fluctuating at a high level into the Ming Dynasty before entering into decline.

As to the interpretation that Smithian dynamics resulting from the canal system explain this efflorescence, subsequent authors have largely agreed with this interpretation but added some nuance. Pingsheng and Jinfang (2024) identify canal-driven Smithian growth as one of three factors—the other two are a shifting focus on commerce and overseas trade—that contributed to economic growth in the 11th century. Ho (1956) discusses the importance of the introduction of early-ripening Champa rice in 11th century China, which was among the technological innovations of Sung China briefly mentioned by Kelly (1997), and it is hard to imagine that this didn’t make at least some contribution to growth. These explanations are complementary, not necessarily in competition with each other, and I find it implausible to suppose that the canal system did not make at least a major contribution toward Sung economic growth.

Subsequent work to Kelly (1997) have generally confirmed his model that transportation infrastructure triggers economic growth in a threshold manner, as opposed to a gradual manner, though it adds nuance. Trew (2020) finds that the development of transportation infrastructure in England and Wales from 1710 to 1881 led to a “spatial takeoff” of growth. Compared to Kelly’s work, Trew (2020) benefits from a much more complete data set. It broadly confirms Kelly’s conclusion, though it emphasizes the spatial aspect of transportation infrastructure; simply building more roads and railways helps little if they do not form the right connections.

Cantoni and Yuchtman (2014) provide a similar analysis on distance to universities. The first universities in Germany were founded in 1386 amidst the Great Western Schism (1378-1417). Universities trained students in Roman and canon law, and they proceeded to take positions where they applied that knowledge, reducing uncertainty in trade and increasing trade volumes. This is an interesting result that links transportation access (in this case, access to universities), institutions, and economic growth.

Donaldson and Hornbeck (2016) provide a result that contrasts somewhat with Kelly (1997). They examine the impact of railroads in the 19th century on the market value of agricultural land in the United States. It is worthwhile to appreciate just how important railroads were. They estimate that by 1890, 64% of the value of agricultural land was dependent on railroads, and if rail had disappeared, at most 20% of the lost value could be made up with canals and country roads. This estimate is made by comparing the value of land in counties that did have railroad access and those that did not. In contrast to Kelly, though, Donaldson and Hornbeck do not show any clear threshold effect; even after the agricultural market is connected, farmers continue to benefit from additional railroad access. The difference may be that Donaldson and Hornbeck’s analysis begins after the U.S. agricultural market was already integrated and thus past the point where a threshold effect would have applied. However, this is speculative, and it remains unclear just how real the threshold effect is.

It would be wrong to say that the concept of Smithian growth no longer applies in the modern world, but within both academia and policy, Schumpeterian models based on technology have clearly become more dominant.

Peretto (2015) builds a model showing two phases of growth: Smithian and Schumpeterian. The latter, in turn, is based on two forms: in-house innovation by firms and new firms that are created around new products. Under Peretto’s model, Smithian growth phases tend to follow a sigmoid shape: a period of rapid growth followed by stagnation, when the market expansion that drives growth is played out. However, Smithian growth can bring a society to a point where sustaining Schumpeterian mechanisms kick in, which Peretto argues happened in the Industrial Revolution.

The final paper for today is Chilosi, Lecce, and Wallis (2025). They examine economic growth in Britain from 1500-1800 and find that Smithian mechanisms were the main drivers of growth during that period. They do so by building a measure of specialization through a data set of job titles, and they find a correlation between specialization and economic growth.

However, the connection between specialization and industrialization is more tenuous. There is no real relationship between specialization and industrial takeoff, but the paper’s abstract leaves out this important statement which changes the interpretation of the result. From the paper’s conclusion,

However, when we look for connections between our measures of specialisation and industrialisation, we find that a robust relationship between patterns of specialisation and subsequent industrialisation only emerged during the middle of the eighteenth century. The division of labour was an engine of long-run economic growth, but not - by itself - a cause of the Industrial Revolution.

In other words, specialization does indeed still matter going into the industrial era.

Today, we’ve reviewed a handful of papers related to historical Smithian economic growth. I take away a handful of tentative conclusions from these papers.

First, it does look like there is robust evidence for a threshold effect for transportation infrastructure, and not a simple gradual relationship (i.e. increasing transportation infrastructure by X% increases GDP by Y%). Furthermore, Trew (2020) emphasizes the spatial nature of infrastructure; simply building more will not have much benefit if it does not make strategic connections.

I think about this chart a lot.

Notice the slowdown in road construction after 1980, following the bulk of the construction of the Interstate Highway System. It would be tempting to attribute this to environmental regulation and public opposition to road construction, and I imagine that these are important factors at the margins. But I also suspect that the more significant factor is that, analogous to the models of Kelly (1997) and other researchers, the American economy became fully integrated to the point where further road construction would have had sharply declining marginal value. I furthermore suspect that if the economic rationale for further road construction was sufficiently compelling, it would happen regardless of opposition.

A major question for policymakers is, under what circumstances today is there a strong case for public investment in transportation infrastructure, in light of its potential to stimulate Smithian economic growth? That is a question that I hope to explore in a later post.

Second, I think it is wrong to conclude that Smithian growth no longer applies in the industrial world. Rather, I think the issue is that Schumpeterian growth has been so strong since the early 19th century that Smithian mechanisms have been obscured. Kelly (1997) makes roughly this point when he considers several episodes of transportation expansion in Britain in late 18th century. Thus phenomena such as shrinking population and protectionism are as much headwinds to growth today as they were in the preindustrial era.