The paper’s first finding sheds light on the idea that living downtown has increased in popularity relative to life in the suburbs. One reason for the increase in center-city popularity is that amenities in downtown areas have improved (better nightlife, safer streets, and so on) Another reason for the increased popularity of city living is that working in centers of cities makes employees more productive to a greater extent than in the past. Broadly speaking, the finding that people have slowed in their movements to the suburbs, while employment continues to suburbanize in some types of cities, argues for the amenity-based explanation. If, instead, productivity was the main driving force, the suburbanization of employment would be more likely to cease everywhere. Yet employment continues to suburbanize outside of the largest US cities.

The result on the decentralization of manufacturing employment confirms and extends some new thinking on how declines in manufacturing employment affect local areas. As a share of total employment, manufacturing has been falling since the end of World War II. But manufacturing declines in the past two decades have appeared to devastate local communities to a much greater extent than in previous decades. The paper supports one explanation for this difference—in the 1970s, and 1980s, manufacturing losses were much larger in the centers of cities than in outlying areas (such as rural America). An example big center-city job losses in the earlier period is provided by the garment industry of New York City, which lost thousands of workers in the 1970s. Downtown areas, however, have a better chance of recovering from manufacturing losses given the high education levels of the working population living near city centers, which makes it easier for those areas to diversify their employment bases. Since 2000, however, geographic patterns of manufacturing job losses have changed. Rural areas, with lower education levels, are now suffering big job losses as well. In particular, the big increase in Chinese imports in the 21st century has significantly reduced employment in US manufacturing plants that had previously migrated to areas with a lower cost of living and lower education levels. Unfortunately, these areas have less chances for economic renewal than did US cities in the 1970s and 1980s. The link between the location of manufacturing activity, the education levels present in those areas, and the ability for high- and low-education areas to recover from manufacturing losses is consistent with other research presented at the conference that examines the “China shock” in more detail.

The results on disparities in earnings based on geographic location provide some additional evidence on the phenomenon of so-called superstar cities. Economically speaking, cities have been doing better than rural areas during the past few decades. But some cities have been doing especially well, and this shows up as a widening of the city-level income distribution during the past few years, even after accounting for city-level density. In other words, average incomes in cities at the same density level are now more widely dispersed than before. This is consistent with the idea that some cities are especially productive. Figuring out why these cities are performing so well—and whether their outsized success will continue—is a key challenge for policymakers and researchers in the 21st century.

This paper uses a comprehensive source of yearly data to study private-sector labor demand across US counties during the past five decades. Our focus is on how employment levels and earnings relate to population density—that is, how labor markets in rural areas, suburbs, and cities have fared relative to one another. Three broad lessons emerge. First, the longstanding suburbanization of employment and population in cities with very dense urban cores essentially stopped in the first decade of the 21st century. For cities with less dense cores, however, the decentralization of employment continues, even as population patterns mimic those of denser areas. Second, a dataset that begins in 1964 clearly shows the decentralization of manufacturing employment away from inner cities, a trend that has long been a focus of the urban sociological literature. Starting in the 1990s, however, manufacturing employment fell sharply, not just in cities but also in rural areas, which had experienced less-intense deindustrialization before then. Finally, average earnings dispersion across counties with similar density levels fell during most of our sample period. But after the 1990s, this dispersion rose, probably because of an increase in earnings dispersion among very dense counties (“superstar cities”). We also note that our results are consistent with explanations of rising individual-level earnings inequality within cities that rest on fundamental changes in how basic job tasks are performed, rather than where particular jobs are located.