This essay is based on the working paper “Transport Corridors” by Roy Ellis, Stephen Haber, and Jordan Horrillo.
Was Adam Smith right about the role of geography in economic development? According to The Wealth of Nations, published in 1776, economic growth is the product of productivity gains that come from specialization; the degree of specialization is influenced by the extent of the market; and the extent of the market is bounded by geographic factors as well as by policies. Smith focused in particular on access to navigable water, which determined the size of the market, and the suitability of soils for agriculture, which determined the depth of the market.
Smith could not, of course, operationalize his insights in a manner that would satisfy journal referees today. The rapid decline in the cost of computing over the past two decades, however, has permitted scholars to operationalize hypotheses about geographic factors and levels of economic development—and Smith’s insights about access to navigable water and soil quality have taken center stage.
A major challenge to that emerging literature on geography and economic development is the measurement of the extent of the market. In a first phase of research, scholars proxied the extent of the market by using the boundaries of present-day nation-states. A subsequent phase of research improved upon nation-state boundaries by employing latitude-longitude grid cells as the unit of analysis.
The problem is that a market is neither a political jurisdiction nor a grid cell. It is an economic-geographic unit in which prices integrate because factors of production and products move. Market extents are generated by contiguous geographic features, conditional on human alterations to those features at a point in time, and given transportation technologies available at a point in time.
We therefore estimate the extent of markets in Smith’s time, conditional on the transport technologies available in Smith’s time, the navigability of rivers and lakeshores in Smith’s time, sea ice density as close as we can measure it to Smith’s time (1750), time invariant terrain slopes, time invariant tsetse fly endemicity, time invariant naturally occurring harbors, and soil qualities conditional on pre-Green Revolution production technologies. That is, we develop a tool—which we refer to as a “transport corridor”—that measures the extent of markets in the eighteenth century.
We then measure the level of economic development in Smith’s time (as well as before and after his time) at the level of the market by constructing a geocoded panel dataset of cities and towns with at least 20,000 inhabitants for the years 1500, 1600, 1700, 1800, 1850, 1900, 1950, and 2000.
We detail below how we build the tool to measure both the extent of markets and their levels of development. To give readers a sense of the difference it makes to measure markets using our tool, rather than grid cells, we present a map of Western Europe centered on Paris.
If a researcher measured the extent of the market of Paris using the boundaries of present-day nation states, then the entire area of France would be counted as the market—even those regions geographically closer to Genoa or Geneva.
If a researcher measured the extent of the Paris market using a one-degree by one-degree grid cell (as is standard in the literature), then the Paris market would be the region covered in purple. It would be standard in the literature to code that purple grid cell as containing navigable water (measured as a dummy variable) and flat land (measured as the gross change in elevation within the grid cell). It would also be standard in the literature to code that grid cell for the distance (as the crow flies) to the coast and the distance (as the crow flies) to the nearest natural harbor. Nevertheless, nothing outside of the purple region would be included in the Paris market. Adjoining regions would be assigned to other grid cells. That is, the researcher would miss the fact that the Seine River, which bisects Paris, flows into the English Channel and was navigable in Smith’s time over that distance—effectively linking Paris, via the natural harbor of Le Havre, to London. She would also miss the fact that Paris is located on the Great European Plain—an expanse of flat land north of the Pyrenees, Alps, and Carpathians that runs from Southeast Britain to Moscow. Finally, she would miss the fact that almost the entirety of that plain—not just the purple region—is covered in fertile, loamy soils, well suited to agriculture.
A researcher measuring the extent of the Paris market using a transport corridor (shown in yellow) would, however, be able to take account of not only these facts about the Paris market, but others as well. For example, the Paris market did not just contain one natural harbor, it contained more than forty. In addition, she would see that the market of Paris did not contain one city—it contained thirteen, including London, Antwerp, and Rotterdam. That is, Paris was one of the wealthiest cities in the world in the eighteenth century because it was part of a network of wealthy cities with productive agricultural hinterlands that were linked to one another by some of the flattest terrain on the planet, slow-moving rivers, and natural harbors.
We generate transport corridors for 50,603 initial points, placed at roughly seventy-five kilometer intervals, covering the terrestrial globe. To do this we measure river flows and lake extents as of the late eighteenth century by restoring to modern-day river maps the sandbars, rapids, and waterfalls that impeded river traffic prior to the invention of dynamite and earth-moving equipment and by removing from those maps the canals human beings dug to get around those impediments. We then model transport costs using eighteenth-century transport technologies, expressed in energy units (megajoules), taking into account the differential energy requirements of the boats used by Lewis and Clark to move across navigable water, a horse-drawn Conestoga wagon to move over land, and a human porter to move overland in the regions of Africa with endemic tsetse flies, which killed horses, mules, and oxen. Each of the 50,603 points gets the same energy budget (which is based on historical wagon draws covering fifty miles on flat ground). Shippers employed the most efficient combination of the three technologies to overcome friction and gravity to move one metric ton of goods as far as they could, given terrain slopes, navigable water, and the presence of tsetse flies.
Our tool first estimates the region accessible from each of the 50,603 points, which we refer to as a hinterland. It then estimates the region composed over overlapping hinterlands, which we refer to as a transport corridor. The following image compares hinterlands and transport corridors for three regions of the world with dramatically different physical markets.
The first frame on the map, at left, represents Nsheng, the capital of the Kuba Kingdom, which was the largest city in the eighteenth century in the present-day Democratic Republic of the Congo. Because of the prevalence of tsetse flies it was not possible to utilize draft animals to pull wagons. To prevent flooding, Nsheng was not located on a river, but on a mesa between three rivers. Thus, the size of its market was the region that could be reached on foot. Because no neighboring markets could be reached, Nsheng’s hinterland and transport corridor are the same size.
The center frame represents Santiago, Chile, which despite the ability to use draft animals, still wrestled with steep terrain and lack of access to navigable water. The transport corridor of Santiago is eighty times larger than that of Nsheng, but is twenty times smaller than that of Philadelphia, shown in the third and final frame. The transport corridor of Philadelphia was not only immense, it also included 185 natural harbors. It also covered some of the most productive farmland in the world.
If Adam Smith’s insights were right, we should find the majority of eighteenth-century urban development in the Old World occurred in transport corridors that were large, included natural harbors, and had highly productive land. As a preliminary test of this hypothesis, this draft of our paper focuses on those transport corridors that fall into the top quartile of both size and agricultural productivity distributions. We map this group of 4,195 large and productive corridors. We show the regions covered by large, agriculturally productive transport corridors with natural harbors in teal, and those covered by large, agriculturally productive transport corridors that did not include natural harbors in brown.
We find that urban places and populations in the Old World concentrated in large transport corridors with high soil quality that had at least one natural harbor. Those transport corridors account for only 10 percent of Earth’s surface area, yet they contained 76 percent of all cities and 81 percent of total urban population in 1700.
We also discover that the pattern of development holds until the early twentieth century. While not shown here, 70 percent of all urban population in 1900 was located inside this same set of transport corridors. We think this is remarkable, given the dramatic change in transport technologies during the nineteenth century. Our findings begin to break down by the modern day, but this region covering 10 percent of the world still explains 50 percent of all urban population in the year 2000.
We also discover that the patterns of the Old World are replicated in the New World, but with a lag because the technologies of the Old World—draft animals, wagons, steel tools, oats, wheat, rye, barley, and rice—diffused across the New World with a lag. While not shown here, circa 1900, the urban development in the New World had taken place in regions with similar natural endowments as the Old World.
The implications of our findings are that geographic factors did not just play a major role in economic development in Smith’s time, they continued to do so even after fossil fuel technologies made it possible to move goods vast distances at low cost.
Read the full working paper here.
Jordan Horrillo is a Data Research Analyst at the Hoover Institution. He is also a Ph.D. candidate at Stanford University with an interest in comparative politics and American politics.
Research briefings highlight the findings of research featured in the Long-Run Prosperity Working Paper Series and broaden our understanding of what drives long-run economic growth.
