What determines the location of cities? Sometimes, we can clearly identify instances when city locations were chosen to achieve specific development or political goals, in remote or sparsely populated areas. For example, the site of Canberra, Australia's capital city, was selected in the early 20th century as a compromise between rival cities Sydney and Melbourne. For many older cities, we can make only educated guesses about their origins. In general, economists believe that people choose to concentrate at sites that have some productive or amenity value. A river, a harbor, or some other natural resource nearby might encourage settlement. There is also the role of local institutions--for example, well-defined property rights--that might make some places more attractive. If these kinds of local features aren't available everywhere, economic activity will be attracted to locations that are superior in resources and institutions.
Another factor that may determine the location of cities is the benefits derived from density itself--so-called agglomeration economies. Living or working in close proximity to businesses or other people can make workers more productive. For example, similar businesses might cluster together in order to have access to cheaper specialized inputs. Jerry Carlino's 2001 and 2009 Business Review articles and my own from 2011 discuss several potential sources of these agglomeration economies. (Of course, the effect of agglomeration economies on the location of cities does not preclude the influence of natural amenities.)
These complementary explanations both have something useful to say about the locations and relative sizes of cities. Of course, great agglomerations today are located near rivers, oceans, or other prominent features of the natural landscape. And many people who live in densely populated areas experience clear benefits from proximity to customers, employers, and producers.
What is perhaps less clear is how to judge the contributions of locational "fundamentals" and agglomeration economies--or more generally, economies of density--independently. Note that both natural fundamentals and economies of density have important limitations as stories for understanding the geographic distribution of economic activity. While natural features seem important, it is difficult to point to one or even several natural features that are valuable enough to explain a very large metropolitan area. For example, in Philadelphia, is proximity to the Delaware and Schuylkill rivers alone really so valuable as to encourage millions of people to crowd together on their banks? Similarly, on their own, stories featuring economies of density are also limited. If there are large economies of density, people will want to locate near existing concentrations of population, but these stories are silent on how a city comes to be in a particular location in the first place. Why is the greatest agglomeration in the Third Federal Reserve District' near the confluence of the Delaware and Schuylkill rivers and not, say, further upstream on the Schuylkill or closer to the Atlantic Ocean?
Furthermore, if there really are large economies of density--that is, density itself provides incentive for people to concentrate, in a virtuous circle--it's possible that any location could be the potential site for a city. All that is required for a large agglomeration is a smaller agglomeration or, in a sense, a city "seed." Intuitively, if you were to rewind history and replay the settlement of some large expanse of land, perhaps cities in this alternative history would be of different sizes and locations. Economists sometimes call this "path dependence" or "history dependence"--that is, present-day or long-run outcomes can depend on a series of historical events or shocks--and it is a potentially important, and unique, theoretical implication of models featuring economies of density.
EVIDENCE ON GEOGRAPHY FROM WAR AND DISEASE
In two papers, economists Donald Davis and David Weinstein reported a historical example paralleling this thought experiment. …