Comments

in 1899 adna weber began his masterpiece on urban growth in the nineteenth century by writing, “The most remarkable social phenomenon of the present century is the concentration of population in cities.”¹ One hundred years later, the evolution of cities is still among the most interesting and important of social phenomena. America is in the midst of an urban transformation so profound that it is changing completely the spatial organization of economic activity. In 1900 urban America lived and worked in high-density communities that allowed most people to travel with their own feet. By the middle of the century, the typical city still had a dense urban core where people worked, but the majority of urban residents lived in suburban communities and commuted by cars. At the dawn of the twenty-first century, jobs have followed workers, and America is increasingly a nation of moderate-density edge cities.²

The data document a truly massive change in the American city. In 1940 only one of the ten largest cities in the United States (Los Angeles) had a population density below 10,000 people per square mile. In 1990 population density levels are below 7,500 people per square mile in seven out of the ten largest cities. Cities of the past were built at higher densities because, after all, in 1920 cars were rare.³ Even in 1960, only 61 percent of commutes were by private vehicle; by 1990 the figure had risen to 83 percent.⁴ In 1940 the overwhelming number of urban jobs appear to have been close to the city center. In 1996 only 24 percent of jobs in metropolitan areas are within three miles of the central business district. The dense walking city of the nineteenth century has been replaced by the medium-density driving city of today.

This decentralization of the American city proceeded in two waves. First, people moved their residences to the suburbs. This movement of urban workers to the suburbs started in the late nineteenth century with the first commuter trains and streetcars. The introduction of the mass-produced automobile democratized the suburbs and led to a massive reshaping of the urban landscape. But as late as 1960, 63 percent of jobs were in the central city, and 51 percent of metropolitan area population lived in the suburbs.⁵ People lived at low densities, but they worked at high densities.

In 2000 people both live and work in the suburbs. This paper is focused on the second wave of urban development: the decentralization of employment. First, we document basic facts about the decentralization of employment. In 1996 we find that across the United States, the average metropolitan area is remarkably decentralized.⁶ The median employee works seven miles from the city center, and the median resident lives eight miles from the city center. New York City is apparently the only major metropolitan area that really has a monocentric employment structure (the median employee in New York works only 3.8 miles from the city center). Chicago, for example, is almost as decentralized as Los Angeles. Across cities, we show that the correlation between city age and decentralization is surprisingly weak, and that regional patterns are less extreme than is often thought. The Northeast and the Midwest have many areas that are just as decentralized as the South and the West.

We use cross-industry variation to try to understand what drives the sub-urbanization of industry.⁷ We examine four reasons why firms choose particular locations: land costs, access to ideas, access to workers, and transport cost savings for inputs and outputs. This framework emphasizes the fact that cities exist to save transport costs for goods, people, and ideas.⁸ We use a number of (admittedly weak) proxies for land use to test whether industries that use more land tend to suburbanize. Industries with higher electricity usage (a possible proxy for land use) are more likely to decentralize. Manufacturing industries, which are generally more land intensive, are also more likely to decentralize.

One piece of evidence suggesting that...