The economics of cities.

Glaeser, Edward L.

The fundamental questions of urban economics are: Why do cities exist~ How does density - or agglomeration - affect people and firms? Why do some cities flourish and others decay? Why are social pathologies often more extreme in cities?

These questions address how spillovers actually operate. If the effects of agglomeration and local spillovers lie behind phenomena as important as economic growth, business cycles, and the formation of human capital (as many researchers now suspect), then urban economics has a special role to play in helping us to understand how these spillovers work in their rawest form. My work tries to use the evidence from cities both to understand urban density itself and to shed light on other topics that are hopefully of interest to the broader economics community.

The Causes and Extent of Agglomeration Economies

Cities now exist for three primary reasons: 1) they reduce transport costs for goods; 2) they eliminate the space between people; and 3) cities facilitate a faster flow of ideas. More precisely, dense agglomeration reduces the transport costs for goods, people, and ideas. These three different sources of "agglomeration economies" have equivalents in other literatures. For example, the role of reduced transport costs for goods in the formation of cities is similar to the idea that the comovement of output over the business cycle occurs because a productivity shock to one firm increases demand for other firms.

Albert Ades and I(1) measure causes of urban agglomeration using cross-county evidence by looking at the extent to which countries' populations are concentrated in a single city. We find that population is more spread out in countries where transport costs for physical goods are lower (measured by development of internal transport networks), and when external trade is smaller (which Paul Krugman and Raul Livas(2) argue is a further implication of transport cost models of urban agglomeration). While transport costs do matter, our results suggest that political factors, for example dictatorship and instability, are far more important in explaining which countries have concentration in a single city. For example, dictatorships have 50 percent more of the population in their largest cities than do stable democracies. When political systems are not stable and democratic, politicians respond to the rent-seeking activities of people who live in their cities by transferring rents to those cities, and population flows follow these rents.

To understand why people in the United States may be more productive in cities, David Mare and I(3) examine why workers are paid substantially higher wages in cities. The urban wage premium persists even after we control for a full battery of individual factors (education, race, age), job-related factors (industry and occupation dummies), and for differential selection into cities. Unless workers in cities were more productive, firms would leave. Thus, even though real wages seem to be constant over space (as evidence on real prices suggest that they are), we believe that there is a productivity premium in cities.

Surprisingly, and counter to many theories of agglomeration, the urban wage premium does not immediately accrue to workers who come to the city, and it does not disappear immediately (or at all) for workers who leave the city. Instead, there appears to be a slow but steady increase in the rate of wage growth for workers in cities relative to workers outside cities (the urban wage premium is also higher among older workers). One possible interpretation is that the urban wage premium works through faster skill accumulation in cities which accrues over time, and stays with workers when they leave cities. I(4) provide a theoretical analysis of this view and some suggestive evidence that shows that the individuals who choose to live in big cities are drawn disproportionately from groups who would presumably value the skill accumulation role of cities (that is, young, college educated persons).

Hedi Kallal, Jose Scheinkman, Andrei Shleifer, and I(5) examine the connection between local area characteristics and the growth of particular "city-industries" (for example steel in Detroit or retail trade in New York). We find that employment growth is faster in city industries that are highly competitive (where competition is the number of firms relative to total employment), not concentrated (where concentration is measured by the share of the city-industry in the total city's employment), and in diverse cities (where diversity is measured by employment concentration of the city in its largest industries). We interpret these findings as a test of growth theories; they may imply that diversity and competition, not industrial concentration, inculcate growth, presumably through the generation of new ideas.

Scheinkman, Shleifer, and I(6) extend this work and examine the connection between city-level characteristics, population growth, and income growth. Population growth and income growth have moved together at the city level over the past 40 years, and all of our results hold for either of these variables. Initial levels of schooling predict later growth. The connection between schooling and growth has increased since 1970, perhaps because the rise in returns to skill has made the intellectual spillovers that are available in high human capital cities more important. Cities with high unemployment levels that were concentrated in manufacturing have seen a substantial decline in both manufacturing and non-manufacturing employment. Few attributes of the public bundle (that is, the level of taxation, or types of spending) seem to influence the growth of particular cities.

Jess Gaspar and I(7) ask whether information technology will make cities obsolete. Improvements in telecommunications will shift time away from face-to-face contacts towards electronic contacts within a relationship, but these improvements also increase the overall number of relationships, and many of these new relationships also will involve face-to-face contact. The theoretical effect of telecommunications on cities is ambiguous. Empirically, we find that across countries and across areas within countries, urbanization and telephone usage go together, even after controlling for costs and income. Telephone contact is higher between regions that are close spatially. Business travel (another form of face-to-face contact) has soared as information technology has improved. Silicon Valley, which should have the best access to the newest telecommunications technology, is the textbook example of geographic concentration of industry. Still, there doesn't seem to be persuasive evidence that the internet will destroy the city.(8)

Glenn Ellison and I(9) attempt to measure agglomeration economies by developing an index of geographic concentration that measures the degree of concentration of particular industries. This index takes the overall concentration of manufacturing employment as given and corrects for the fact that the lumpiness of manufacturing will suggest geographic concentration whenever returns to scale dictate that production take place in a few large plants. We find that the famed geographic clusters of particular industries are exceptions rather than rules. While the famous examples are all quite observable in the data, the mass of industries display a statistically significant level of industrial concentration that appears (to us at least) to be relatively mild economically. The fact that industrial clusters appear to be relatively rare also can be interpreted as suggesting that many of the most important agglomeration economies occur across, rather than within, industries.

Guy Dumais, Ellison, and I(10) extend this work and use the Longitudinal Research Database (which provides almost complete information on every manufacturing industry in the United States) to examine the dynamic components of geographic concentration. We find a large amount of movement of particular industries across space (even among the most concentrated industries) which casts doubt on the general applicability of anecdotes used by theorists to suggest that geographic location often is determined by decades-old happenstance. The longitudinal research database also suggests that the geographic concentration of manufacturing is a delicate balance of the creation of new manufacturing plants, the expansion and contraction of pre-existing plants, and plant closures. Plant openings occur away from pre-existing industrial centers and act to lower geographic concentration. Plant closures act to reinforce concentration, because they are much more likely away from pre-existing industrial centers (even after controlling for the fact that these peripheral areas are likelier to have newer plants). We also find that while industries that share input-output relationships locate together, these effects are small. The dominant force in determining which manufacturing industries locate together is that they hire the same kinds of workers.

Urban Pathologies

American cities are not just technological wonderlands. Many urban centers face extreme poverty, crime, and other social problems. Fundamentally, the social problems of cities appear to be the result of many of the same agglomeration economies that fuel urban productivity. The same intellectual spillovers that make Silicon Valley also may lead to the transmission of ideas (or norms) about crime within the inner city.

Bruce Sacerdote and I(11) ask "Why is there more crime in cities?" The connection between city size and crime rates is pervasive and longstanding. We find that only 15 percent of the connection between city size and crime appears to come from lower probability of arrest in cities, and 25 percent appears related to higher financial returns from criminal activity in cities. Almost 50 percent of the urban crime premium is related to the fact that the disadvantaged (particularly female heads of household) end up in cities.

The connection between growing up with one parent and crime is not nearly as strong as the connection between living in areas with large numbers of single parents and the level of crime. While this difference could be explained by omitted area-level variables that are correlated with single-parent families, it may also come from spillovers within those areas. Having one parent may not induce a child to become a criminal, but growing up in an area where all one's peers all have missing parents may create peer effects that lead to crime. Sacerdote, Scheinkman, and I(12) create a methodology for measuring the extent of social interaction based on a simple local interactions model. We find large amounts of social interaction in criminal behavior, and we believe that these social interactions explain the wide diversity of crime rates across space.

To further examine the role of local spillover effects, David Cutler and I ask whether ghettos are good or bad.(13) We document that young African-Americans growing up in segregated areas are more likely to be unemployed or idle (that is, neither at work nor in school), or to have a child out of wedlock, and are less likely to graduate from high school, than young African-Americans who grow up in less segregated areas. This is true even after controlling for city-specific characteristics and looking at the difference between white and African-American outcomes within cities. Furthermore, we use a variety of instruments (including number of governments and topographical barriers) in order to avoid the problem that individuals choose their neighborhoods. All of our results essentially confirm the striking negative effect of segregation on African-American outcomes in 1990. We suspect that the mechanism linking racial segregation with poor outcomes for non-whites is not physical access to jobs, but rather the intellectual and social isolation of disadvantaged communities which then works against the acquisition of human capital among youth.

Cutler, Jakob Vigdor, and I examine the roots of segregation and its current history.(14) We find that segregation rose during every decade between 1890 and 1960, but that it has declined quite substantially in the twenty years since 1970. Segregation cannot be explained by many city-level characteristics except for city size: bigger cities are generally more segregated. We use housing prices and survey evidence to determine the extent to which segregation occurs: because of a greater willingness among whites to pay to live in white neighborhoods (in which case, whites should be paying more in more segregated cities) or; because of "collective action racism," where whites collude to force blacks to live in particular neighborhoods (in which case, we would expect African-Americans to be paying more to live in more segregated cities). We find that in 1940, segregation was driven primarily by collective action racism (supported by legal devises including restrictive covenants), but that by 1990, the remaining segregation is driven primarily by white tastes for white neighborhoods.

Denise DiPasquale and I examine a final form of urban pathology: riots.(15) We find a strong connection between urbanization and rioting across countries, particularly among countries that are ethnically fragmented. Using a sample of race riots in the 1960s, we find regular evidence for basic economic hypotheses about rioting (higher unemployment increases rioting; police resources decrease rioting) and little evidence for more sociological hypotheses about relative poverty and deprivation. We do find, however, that ethnic diversity is deeply linked to conflict, both in the United States and around the world. The major role of ethnicity requires models that move beyond the cost of punishment and the opportunity cost of time into understanding the far-reaching effects of ethnic identity.

Some of my work has focused on policy responses to these problems. One paper published in 1996 suggests that property taxes provide better incentives for revenue maximizing local governments who would choose to maximize property values in an attempt to maximize total revenues.(16) In a forthcoming paper, I argue that indexing transfer payments to local price levels could lead to pernicious results because of migration effects.(17) In a work on my own(18) and with Erzo Luttmer,(19) I consider the effects of rent control on misallocating housing across consumers. Overall, though, my primary focus has been on the issues in the study of urban economies, not on the appropriate government response to these economies.

Conclusion

Cities are unique combinations of the best and worst features of modern society. Agglomeration effects appear to create faster learning and greater productivity, but they also seem to further crime, riots, segregation and ghetto poverty. Further research on cities can play a role in helping us to both understand how to improve urban areas and how agglomeration economies actually operate.

1 A.F. Ades and E.L. Glaeser, "Trade and Circuses: Explaining Urban Giants," Quarterly Journal of Economics 1995, pp. 195-228.

2 P. Krugman and R. Liras, "Trade Policy and the Third World Metropolis," NBER Working Paper No. 4238, December 1992.

3 E.L. Glaeser and D.C. Mare, "Cities and Skills," Hoover Institution Working Paper E-94-11, 1994.

4 E.L. Glaeser, '"Learning in Cities," NBER Working Paper No. 6271, November 1997.

5 E.L. Glaeser, H. Kallal, J. Scheinkman, and A. Shleifer, "Growth in Cities," Journal of Political Economy 100 (1992)pp. 1126-52.

6 E.L. Glaeser, J. Scheinkman, and A. Shleifer, "Economic Growth in a Cross-Section of Cities," Journal of Monetary Economics 36 (1995) pp. 117-43.

7 J. Gaspar and E.L. Glaeser, "Information Technology and the Future of Cities," NBER Working Paper No. 5562, May 1996, and Journal of Urban Economics, forthcoming.

8 E.L. Glaeser, "Are Cities Dying," Journal of Economic Perspectives, Winter 1998, forthcoming.

9 G. Ellison and E.L. Glaeser, "Geographic Concentration in U.S. Manufacturing Industries: A Dartboard Approach," Journal of Political Economy 105 (1997)pp. 889-927.

10 D. Dumais, G. Ellison, and E. L. Glaeser, "Geographic Concentration as a Dynamic Process," NBER Working Paper No. 6270, November 1997.

11 E.L. Glaeser and B. Sacerdote, "Why is There More Crime in Cities?" NBER Working Paper No. 5430, January 1996.

12 E.L. Glaeser, B. Sacerdote, and J. Scheinkman, "Crime and Social Interactions," Quarterly Journal of Economics CXI (1996) pp. 508-48.

13 D. Cutler and E.L. Glaeser, "Are Ghettos Good or Bad?" NBER Working Paper No. 5163, June 1995, and Quarterly Journal of Economics CXII (1997) pp. 827-72.

14 D. Cutler and E.L. Glaeser, "The Rise and Decline of the American Ghetto," NBER Working Paper No. 5881, January 1997.

15 D. DiPasquale and E.L. Glaeser, "The L.A. Riot and the Economics of Urban Unrest," NBER Working Paper No. 5456, February 1996, and Journal of Urban Economics, forthcoming.

16 E.L. Glaeser, "The Incentive Effects of Property Taxes on Local Governments," Public Choice 89: pp. 93-111.

17 E.L. Glaeser, "Should Transfer Payments be Indexed to Local Price Levels?" NBER Working Paper No. 5598, May 1996, and Regional Science and Urban Economics, forthcoming.

18 E.L. Glaeser, "The Social Costs of Rent Control Revisited," NBER Working Paper No. 5441, January 1996.

19 E.L. Glaeser and E. Luttmer, "The Misallocation of Housing Under Rent Control," NBER Working Paper in progress.

Edward L. Glaeser is a Faculty Research Fellow in the NBER's Program on Economic Fluctuations and a Associate Professor at Harvard. He is profiled in this issue.