A Companion to Urban Economics - Arnott and McMillen - Chapter 4

A Companion to Urban Economics Edited by Richard J. Arnott, Daniel P. McMillen Copyright © 2006 by Blackwell Publishing Ltd C H A P T E R F O U R Cross-Country Patterns of Urban Development Stephen Malpezzi 4.1 INTRODUCTION Urban economists love to suggest the following thought experiment to students. Consider a world without cities. More specifically, suppose that the world’s popu-lation was randomly distributed across space. Omitting some totally uninhabit-able places such as Antarctica, there would be, very roughly, about 2 ha of land for each person in the world, some more arable than others. Suppose, in turn, that the capital stock was randomly distributed; there is, very roughly, something like $20,000 worth of tangible capital (computers, tools, buildings, furniture, trucks – all the things we use to make other things) per capita. If you were lucky, you might find yourself on somewhat fertile land, well irrigated, in a mild climate, with some tools that might help you gather or grow enough food to stave off starvation. If I were unlucky, I might be in a less favorable location with cap-ital that is now mostly useless. World GDP would fall from something like $40 trillion steeply toward zero. The survivors of this thought experiment would quickly be on the move, banding together and beginning to cluster in more favorable locations. The world’s economy would begin to regroup and recover, although it would probably take many decades to rebuild a world in any way recognizable. The thought experiment seems bizarre if not silly, but it does in fact clearly illustrate the importance of location. The thought experiment sounds so weird mainly because it collapses many millennia of migration, capital formation, trading, and technical change – the processes that underlie our long history of economic development – into a very 56 S. MALPEZZI short time frame. Now, anyone who reads newspapers or even watches CNN is aware that the level of economic development varies tremendously between countries such as the United States, France, or Japan, and countries such as India, Ghana, or Mexico. We will discuss some of these differences more rigorously below. But for now we note that these differences across countries provide a sort of natural experiment, or laboratory, for examining some of the relationships between urbanization and economic development. Other contributions to this volume, especially the other chapters in this part (by Rosenthal and Strange, Duranton, and O’Sullivan) discuss theoretical under-pinnings for the relationship between urbanization and development – including a focus on trade, transaction costs, and agglomeration, as well as the “usual” textbook economies of scale. The other contributions also discuss empirical evi-dence in passing, often from firm or small area-level data. In this chapter we take another empirical, bird’s-eye view, largely from the perspective of cross-country comparisons. Other than a motivating paragraph or two, we leave theory to other chapters, and we are not presenting state-of-the-art formal econometric tests; rather, we are presenting basic facts about economic development and urbanization. The chapter relies heavily on simple analyses of individual country data, most from one of two sources. The World Bank’s 2004 World Development Indicators database is the source for most of our basic data on population, incomes, and quality of life measures. The United Nations is the source for much of the basic data on the population of individual cities. Other sources are noted as we go. Elaboration of many results summarized in this chapter, including discussion of data issues, can be found in a longer companion publication, Malpezzi (2004). This chapter and a spreadsheet containing the basic data can be found through the author’s website, located at www.bus.wisc.edu/realestate. 4.2 PATTERNS OF ECONOMIC DEVELOPMENT AND PATTERNS OF URBANIZATION Alert readers are already suspicious. “Development” is a multidimensional, not to say slippery, concept. And the common practice of lumping the world’s countries into two or three groups, such as “developing” (broadly, low-income countries), “transition” (largely formerly communist countries in Eastern or Central Europe, and the former Soviet Union), and “developed” is crude at best. It is nearly impossible to come up with a simple taxonomy that seems satisfactory. Let us begin with the consideration of national income, or the closely related concept of gross domestic product (the total value of final goods and services produced within a country’s borders in a year). Rich countries, such as the USA, Canada, most of Western Europe, and Japan, have per capita GDPs of the order of $30,000. The poorest countries, such as Angola, Laos, or Burma, have per capita GDPs of around $300. While everyone knows that poor countries contain some very rich individuals and rich countries some poor ones, the fact that national averages differ by about two orders of magnitude is astounding and CROSS-COUNTRY PATTERNS OF URBAN DEVELOPMENT 57 begs for explanation: see Easterly (2001), Landes (1998), and Maddison (2001), for example. It is also important to note that GDP per capita is a very useful and widely used measure of the complex and fuzzy notion we have of “development,” but like all specific measures it is imperfect. For example, GDP is well known to undervalue nonmarket production, such as housework, and to fail to account fully for many environmental costs. On the other hand, a great many develop-ment outcomes – such as literacy, life expectancy, and the cleanliness of the environment – are correlated with income or GDP per capita, albeit imperfectly. Probably the most important development indicator that is largely uncorrelated with GDP per capita is the distribution of national income among households. But in the end if we have to pick a single measure for initial analysis, GDP per capita is clearly the place to start. To keep our analysis manageable, in this chapter we will focus on relationships between aspects of urbanization and GDP per capita, with a few notes about other measures; Malpezzi (2004) presents some additional detail. 4.2.1 The broad context of urbanization DENSITIES ACROSS COUNTRIES The simplest way of thinking about urbaniza-tion is the existence of above-average density. Densities vary both across countries, and within them. We have noted that there are about 2 ha of land for every person in the world. The USA has an above-average endowment of raw land, by world standards: about 3.25 ha per person (or about 0.3 persons per hectare, or pph). But there are some other countries with even larger areas relative to their population. Canada has about 30 ha of raw land per person, Australia 40, Russia a dozen. At the other extreme, examples of higher densities include China’s 0.75 ha per person (or about 1.35 pph), while India and Japan have about 3.5 pph, South Korea and the Netherlands about 4.75, Bangladesh 10, and Singapore and Hong Kong about 65 pph. DENSITIES WITHIN COUNTRIES However, the figures presented in the previous paragraph are extremely gross; within countries, densities vary even more within than across countries. Within the USA, most of the country’s population lives within a few hundred miles of the major coasts (including the Great Lakes); with a few exceptions, such as Denver and Salt Lake City, most of the country is fairly empty from, say, Minneapolis, until one reaches a hundred miles or so of the Pacific Ocean (Rappaport & Sachs 2003). This pattern is not atypical; many countries have some fairly dense areas and some (often large) “empty quarters.” For example, almost 90 percent of Canada’s population lives within 200 miles of the US border; most of China’s population lives within 100 miles of the coast; and very few Australians live very far inland. Figure 4.1, from the Center for Interna-tional Earth Science Information Network (www.ciesin.org), maps population density around the world. The average densities of US states range from about 4 pph in New Jersey (denser than India or Japan), 3 pph in Rhode Island (denser than Germany), and over 2 pph in Connecticut and Massachusetts to less than 58 S. MALPEZZI Figure 4.1 World population density. Source: The Center for International Earth Science Information Network (CIESIN), Colombia University 0.05 pph in Nevada and New Mexico, and less than 0.2 pph in Alaska. To give some idea of the statewise differences, if the entire USA excluding Alaska was settled at New Jersey’s density, the country would contain well over 3 billion people. DENSITIES WITHIN CITIES Within cities, densities vary even more remarkably. Figure 4.2 presents simple density patterns for half a dozen illustrative locations; these are taken from Bertaud and Malpezzi (2003), which presents data for a larger sample of about 50 world cities. Among large cities in the USA, New York has a central density approaching 200 pph, falling off rapidly to 50 pph or less about 20 km from midtown Manhattan. But because the central area of the city comprises a relatively small area, the average density of the New York metro-politan area is only about 40 pph. Chicago and Los Angeles (not shown here) have central densities of 50–70 pph, and average densities of around 20; the central density of Chicago is higher, but many readers will be surprised to find that the average density of Los Angeles, 22 pph, is greater than that of Chicago, 16 pph. At the other extreme, the central density of Atlanta is only 25 pph, although it exhibits an even faster drop-off with distance from the center, from a lower base, and an average density of 6 pph. It turns out that this pattern, of a high central density, followed by a rapid initial drop-off, that slows as we move out from the center, is a consequence of a qualitatively similar pattern in land rents and real estate prices. These, in turn, are derived from the value of access to a central location; patterns of rents, property values, and population density are determined by the trade-off between rent and transport costs between different locations. A very simple analysis of the process can be found in Alonso (1960); for more details, see Daniel McMillen’s chapter in this volume. This density pattern at least roughly corresponds to reality not only in most US cities, but in fact in most cities in market-oriented econom-ies. It is sometimes referred to as a “negative exponential” pattern, because in a very idealized form it can be summarized by CROSS-COUNTRY PATTERNS OF URBAN DEVELOPMENT 59 300 200 200 100 100 0 0 0 10 20 Distance 0 10 20 30 40 Distance 150 300 100 200 50 100 0 0 0 10 20 30 40 50 0 10 20 Distance Distance 1000 500 0 110 100 90 80 70 60 50 40 0 2 4 6 8 10 12 14 0 10 20 Distance Distance Figure 4.2 Average persons per hectare, 7 km annuli, six world cities. Source: Bertaud and Malpezzi (2003) y(x) = y0e−bx, where x is the distance from the center, y(x) is the density at any distance x, y is the density at the center, b is a density parameter (often called the “gradient,” because it is the rate of change of density as x increases), and “e” is the base of natural logarithms. ... - tailieumienphi.vn