The Ambiguous Role of Transportation Costs

Over time, transportation technology tends to improve and costs go down. Does this cause economic activity to become more concentrated or less concentrated? We’ll review three papers today that explore this question and give us an ambiguous answer but one that leans toward “more concentrated”.

As usual, today’s post is part of a Living Literature Review grant from Coefficient Giving, though all conclusions are my own. A companion piece on the main Scaling in Human Societies site is in progress.

Over the last few weeks, I have been looking over some research on the role of transportation and telecommunication technology on the concentration of economic activity. Although not exactly the same, there is a lot of overlap between the dynamics of cheap transportation and effective telecommunications.

What I have been calling the “death of distance” hypothesis, named after Cairncross (1997) but a much older idea, is that better transportation and telecommunications technology should make urban agglomeration less important and disperse economic opportunity. This has long been the dream of remote work, and the idea is quite intuitive, at least to me. Most major cities are becoming more congested, and at best, a ramp-up of investment in roads and public transportation will alleviate congestion only slowly and marginally. Meanwhile, the tools for remote work are improving rapidly: fast and reliable Internet connections and software for collaboration, for instance. If there really is some magic to in-person presence that justifies commuting and concentration in cities with congestion and high costs of living, we can now get pretty darn close to simulating in-person presence with virtual reality.

And yet, the literature, not to mention facts on the ground, have not been kind to the death of distance hypothesis. In the last post, I tried to argue that, while telecommunications do not negate urban agglomeration economies, they do make urban scaling a bit less steep. Unfortunately, that post was heavily caveated because the evidence offers, at best, tentative support to this form of the hypothesis. But the death of distance hypothesis keeps coming back up because it feels like something that “should” be true.

One complication is that “distance” is a bit vague. Angel et al. (2010) establish that urban densities have been generally declining all over the world in the 20th century, though Hanberry (2023) documents that this trend has reversed in the 21st century. This seems clearly to be a response to the prevalence of mass transit, and then personal automobiles, and so distance within cities is becoming less important.

As we’ll see below, there are two opposing forces at work that make the impact of transportation improvements between cities more ambiguous. On the one hand, as discussed last time, the fact that individuals can travel more freely between cities via highways and airplanes means that it becomes less important for those individuals to live in a particular cities.

On the other hand, and this may be the more important factor, cheaper transportation means that manufacturing can serve a greater geographical range. Then manufacturing tends to concentrate in small areas because the benefits of economies of scale outweigh the costs from transportation. We thus have a pattern of dispersal within cities and concentration between cities. Nonphysical goods, such as what is primary produced in Silicon Valley, have virtually no transportation costs, and it is with software that we see extreme levels of concentration. Companies are willing to endure punishing costs of living and severe governmental mismanagement to gain the advantages of concentration, despite the theoretical possibility of fully remote work. Personally, I still don’t really understand why this is the case, but it is, and we will continue to explore why that may be.

As a prelude to today’s main papers, the German economist Walter Christaller’s 1933 Central Place Theory (discussing in King 1985) is one of the foundational works in economic geography.

As relayed by King 1985, Christaller’s central place model envisions a hierarchy of city sizes. The model assumes that an agrarian population is spread evenly over an unbounded, featureless plain. The cost of travel is proportional to the distance traveled. Both consumers and businesspeople try to maximize profits, which are the sum of travel costs and production costs.

The economy has several kinds of goods, each of which have different prices and demand profiles. Some goods have high demand and low production costs, and for them, buyers will be more sensitive to reducing travel costs to purchase those goods. Sellers of those goods tends to be more numerous and located in a large number of small towns. Some goods have low demand and high production costs. For them, a seller needs to reach a large population for their business to be viable, and due to high costs, a buyer is less sensitive to having to make a long trip to procure them. It follows that sellers of these goods tend to congregate in a smaller number of large cities.

It follows from Christaller’s model that settlements tend to emerge in a hexagonal lattice as shown below. Larger cities, which sell more specialized goods in addition to the common goods, occur infrequently compared to small towns, which sell only common goods.

According to Dmitriev (2022), the connection between central place theory and Zipf’s law is superficial. Both principles model how cities form a hierarchy with a small number of large cities and a large number of small cities. However, Zipf’s law shows how the rank-size distribution emerges from a stochastic process, while central place theory is a more static model.

Central place theory is an interesting model for how transportation costs influence the geography of cities, but it is a static model that does not paint a clear picture of those this structure emerges. Now, let us turn to three papers that present simple, dynamic models of how transportation costs relate to urban concentration.

Our first paper is Krugman (1991), a foundational paper in New Economic Geography. This paper builds a simplified model, in which there are two regions, and the model seeks to determine the dynamics of how manufacturing and population are distributed between the two regions. If transportation costs are high, then manufacturers and hence population are more even distributed between the two regions, as less trade between them is feasible. The model also finds that greater returns to scale and a greater share of the economy in the manufacturing sector also strengthen the concentration effect. He furthermore finds that with two regions, the existence of two cities may be an unstable equilibrium, depending on the model parameters. The larger region may gain more from economies of scale, strengthening its manufacturing sector and attracting population, creating a circular causal loop that ends with the entire population in one region. For other values of parameters, though, higher wages in the more populous region halt the movement of population to that region.

Our second paper is Puga (1999), which considers the role of transportation costs on agglomeration and find that there is not necessarily a monotonic relationship between them. If transportation costs are high, he finds that industry tends to be more dispersed, since in that case the cost of trade outweighs economies of scale. For intermediate transportation costs, industry tends to become more concentrated to take advantage of economies of scale. If transportation costs are low, then the effect of concentration depends on labor mobility. If labor can move freely between the regions, then low transportation costs are a further concentrating force for industry. If labor movement is restricted, then in the presence of low transportation costs, concentration is halted and reversed due to labor cost differentials.

As with Krugman (1991), the model of Puga (1999) is based on an economy with two regions and two sectors: industry and agriculture. The agricultural sector is considered to be fixed. When labor mobility is restricted, then concentration of industry in one region will raise wages due to competition for labor with agriculture, which in turns halts further concentration.

Helpman (1998) also builds a simplified two region model. In his model, there is a homogenous sector and and a differentiated sector. His model is similar to that of Krugman (1991), but with a few important differences. In Krugman’s model, there is a homogeneous sector, which he labels agriculture, that can be freely traded across the regions. In Helpman’s model, it cannot be traded freely. Krugman also assumes that the income from homogeneous product is spent entirely in the region from which it is derived, while Helpman assumes that the product generates demand in each region in proportion to the number of people that live there.

Helpman (1998) concludes that when economies of scale are strong or when housing costs are low relative to the cost of tradable goods, people prefer to concentrate in a single region, and the only equilibrium states are when one region or the other contains the entire population. However, if economies of scale are weak and housing costs are high, the equilibrium state is one in which both regions are populated.

Today, we focused on three papers that present very simple models for how transportation costs and the strength of economies of scale influence agglomeration. In some cases, the lure of accelerating returns to scale and the desire to avoid high transportation costs may induce people to concentrate in a single region. When returns to scale are weak and transportation costs are low, people may instead be motivated to spread out across regions.

Simple models such as in Krugman (1991), Puga (1999), and Helpman (1998) obviously suppress much of the complexity that is present in the real world. But they also help us understand core principles by removing unnecessary detail. Together, they are enough to illustrate the basic principle: low transportation costs tend to be a concentrating force, but this is not always the case.

In the course of researching this post and the last post, I have concluded that the consensus in the literature—that falling transportation and telecommunication costs are not leading to a death of distance—is probably correct, at least when we look between cities rather than within cities. Although I made at attempt to salvage the hypothesis in the last post, doing so requires a lot of hedging and caveating.

I find this conclusion unsatisfying. Some of the major critiques of the modern economic system is that it is a “winner takes all” system, in which certain superstar cities experience great prosperity while other decline and in which access to high status professions is increasingly out of reach for most people. Ever-growing public opposition to trade, labor migration, and new technology is motivated in part by the perception that these things foster economic concentration that creates wealth that is shared among a minority. However, I don’t see a plausible model of economic growth that does embrace free trade, free migration, and openness to new technology.