Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
8 The Urban Economy Transformed Much of development economics is concerned with fostering economic growth, alleviating poverty, and improving the general standard of living of people in low- and middle-income countries. Employment and earnings from labor are central concerns because for most households, labor earnings represent the largest portion of total income, and poverty is often the result of insufficient access to adequate employment. As urban population projections make clear, the urban economies of many developing countries are facing an unprecedented challenge to generate a sufficient number of new jobs at reasonable wages to absorb a growing and better- educated labor force. The purpose of this chapter is to examine several of the critical elements in the nature of urban labor supply and employment. Of course, no single chapter can do justice to the complexities of urban economic relations, and the attempt made here is far from comprehensive. From the great variety of issues associated with urban economies, we focus on a selected few that have the greatest demographic resonance. In many ways, the urban economy shapes urban demography. Urban eco- nomies generate the resources upon which governments draw to extend public services, whether in family planning, education, or health, and to undertake in- vestments in the public health infrastructure. The scale of the urban economy can enable specialization in economic activities and foster the growth of private mar- kets, with the markets in health services being of particular significance. At the level of families, income from labor and economic enterprises provides a means of purchasing health, education, and other services. The incomes of women, when considered in relation to those of men, have much to do with the returns to marriage. The economic returns to schooling have been a central theme in this report, and we have repeatedly emphasized their importance to family strategies involving lower fertility but greater schooling per child. Urban incomes and edu- cational returns also affect migration, whether from rural areas or from other urban 300
THE URBAN ECONOMY TRANSFORMED 301 areas. For these reasons, no study of urban demography can neglect its economic . . unc erpmmngs. To set the discussion in context, we begin with an introductory section that sketches the main sectoral and spatial features affecting labor markets and urban economies. This section highlights the analogies between the social externalities that arise from urban diversity and proximity and the economic externalities that arise from the same general sources. As will be seen, there is an economic role for social networks, one of the mechanisms highlighted in previous chapters. Our treatment of sectoral and spatial factors must be brief, but several superb literature reviews can be consulted for more detail (see, for example, Williamson, 1988; Eberts and McMillen, 1999; Becker and Morrison, 1999~. Unemployment and poverty in urban labor markets in developing countries are often attributed to la- bor market failure, but our review highlights how the differences between labor markets in high- and low-income countries have probably been overemphasized. Economists are now reexamining the structure of wages and earnings within urban labor markets in low-income countries to assess more carefully the role of human capital and institutional factors in determining earnings. Many of the issues now being raised have long been themes in research on industrially more-developed economies. In the following section, attention turns to urban growth and the economic re- turns to schooling. The potential contribution of education to improving economic productivity, lowering income inequality, and promoting growth and development is well known, and almost all governments allocate a considerable share of their budgets to public expenditures on education. For many capital- and natural- resource-starved countries, human resources may represent the only route to suc- cessful development. As noted above, the economic returns to schooling are also important determinants of demographic behavior. They influence parental assess- ments of the payoffs for investment in children's education and fertility decline. Also, to the extent that schooling returns are higher in urban than in rural areas or differ across cities, they exert influence on both the rate and direction of migration. Yet when the urban supply of educated labor is increasing rapidly, as it appears to be in most developing countries, one wonders how high returns to schooling can be maintained. This section explores the issues by way of case studies. Migration, another central element in the size and nature of the urban labor force, is discussed in the next section, where we reassess the conceptual frame- works that so dominated thinking in the 1970s and 1980s the Todaro (1969) and Harris and Todaro (1970) migration models. Given the changing nature of urban systems, to what extent are the assumptions made by Todaro and Harris- about the attractions of urban "formal-sector" jobs and dualistic conceptions of the urban economy still valid and useful? In considering the issues, we examine migration as it is influenced by urban economic structure, review what is known of urban economic mobility and income dynamics, and then explore (in the next sec- tion) dualistic assumptions about segmented urban labor markets. Here we look
302 CITIES TRANSFORMED critically at the proposition that sectoral barriers and rigidities are a root cause of urban economic inequalities, and examine the evidence for increasing hetero- geneity in the urban informal sector. We next review several case studies showing how labor demand and supply factors have come together to generate inequality in urban Brazil, Taiwan, and China, countries that can offer rich longitudinal data on urban earnings. The penultimate section of the chapter examines international facets of urban economies, factors that doubtless will increasingly condition urban labor markets in developing countries in the future and affect their labor absorption capacity. We present evidence on the urban implications of foreign direct investment, discuss the role of globalized networks of high-end business services, and examine how heightened exposure to world markets with their attendant benefits, costs, and uncertainties can affect the lives of urban residents disproportionately. The final section of the chapter presents conclusions and recommendations. SECTOR AND SPACE At the outset, we must distinguish sectoral from spatial influences. To see why, consider Engel's law that the income elasticity of demand for manufactured goods and services is greater than that for primary products. When demand conforms to Engel's law, an increase in a country's income per capita shifts the composition of its demand toward manufacturing and services and away from agriculture, mining, and related sectors. Shifts in the demand for labor (and other inputs) in each sector accompany these changes in final demand. Although it is tempting to think of manufacturing and services as urban activi- ties and to conclude that national income growth leads inexorably to urbanization, nothing in Engel's law actually implies this. Engel effects are aspatial. Addi- tional argument is needed to show why manufacturing and services tend to be located in cities, that is, in spatially concentrated clusters of labor and production. Spatial concentration offers some cost and productivity advantages, but because these come bundled together with disadvantages (e.g., congestion, negative exter- nalities), the spatial side of the story has its own logic. Engel effects that steadily shift the composition of demand can at first be expressed in spatial concentration and higher levels of urbanization, but later be expressed in the spatial deconcen- tration of the urban population, if not in lower levels of urbanization as such. In this section, we first examine sectoral effects, aiming mainly to show how urbanization can be a natural consequence of national economic growth or, more precisely, a consequence of the factors that produce national growth. We then take up spatial effects by considering both the benefits and costs of urban spatial con- centration. A review of the productivity advantages associated with spatial clus- tering suggests that urbanization may well be a cause of national economic growth in addition to being a consequence. The mechanisms involved are the spatial
THE URBAN ECONOMY TRANSFORMED 303 counterparts to the forces highlighted by the "new growth theory" in economics (e.g., Lucas, 1988~. In sorting through the many linkages between national and urban growth, we really cannot hope to disentangle cause from consequence. Even so, it may be useful to review the strong associations that have been documented repeatedly in the literature. We note four points. First, where national statistical systems permit such a decomposition, studies have shown that the urban share of the na- tional economy can be far greater than its population share alone would suggest. In the case of Mexico, a careful analysis of 1998 economic census data at the municipality level by Sobrino (2002) shows that the five largest metropolitan ar- eas accounted for 53 percent of national value added in industry, commerce, and services while housing only 28 percent of Mexico's population (United Nations, 2001~. Mexico's 10 largest cities, with only one-third of the country's population, generate 62 percent of national value added. Figures such as these underscore the importance of city economies to national levels of income. Even from an ac- countant's perspective, it is difficult to conceive of national economic activity and urban activity as somehow occurring in separate spheres. Second, because the size of the urban informal sector tends to be poorly mea- sured in national income accounts we discuss the informal sector later in this chapter there is reason to think that the urban share of the national economy may often be underestimated. One study of West Africa reconstitutes the production and trade flows of 1960 by applying social accounting matrix techniques (sour, 1994~. These revisions suggest that national gross domestic product (GDP) was understated by as much as 30 to 50 percent in the national income accounts. The revised methods reveal that West African cities provided two-thirds of real GDP, rather than the mere 38 percent indicated in the conventional accounts. Third, to get a sense of the urban contribution to national growth, one can look at cross-country data on growth in value added (World Bank, 2000a). Of the total growth in national value added among low-income countries from 1980 to 1998, fully 86 percent is attributable to growth in industry and services.! The urban contribution must be somewhat less than this because some services and industries are located in rural areas. Nevertheless, the figure suggests the extent to which national economic growth derives from urban growth, and it conveys something of the force of Engel's law. Fourth, a number of statistical studies have found that higher levels of na- tional income and faster national growth spur urbanization. Among the many studies that have examined such correlations, we single out the following few. Moomaw and Shatter (1996), using data from 90 countries for 1960, 1970, and 1980, show that the urban share of the population increases with national income per capita; the share is also increased by industrialization, trade orientation, and ~ Excluding India and China, calculations by the panel put the industry and services share of growth at 83 percent.
304 CITIES TRANSFORMED foreign capital inflows. Brockerhoff (1999) examines the population growth rates of a large number of developing-country cities. He finds that growth in gross na- tional product (GNP) per capita encourages city growth. In earlier work, Becker and Morrison (1988) reach a similar conclusion for the countries of sub-Saharan Africa. Gaviria and Stein (2000) also find a strong positive effect of growth in GDP per capita on the population growth rates of large cities.2 Although studies such as these do not firmly establish the direction of causation, they do show that the association between urban and national economic growth is robust and worthy of attention. Sectoral Effects Kelley and Williamson (1984) give a highly instructive account of the sectoral factors involved in urban growth. They formulate what is termed a "computable general equilibrium" model and use simulations to illustrate their main analytic points.3 These authors consider the case of a hypothetical country that is a price taker in international markets, which is to say that the prices of its agricultural and manufactured goods are taken to be externally determined. Changes in the relative price of manufactured goods affect the relative demand for labor in man- ufacturing. If manufacturing is indeed an urban activity, then an increase in the relative price of its products leads to faster urban growth and, in time, to a higher urban population share. In this way, changes in the international terms of trade between manufactured and agricultural goods can have an impact on the national level and pace of urbanization. Two types of technological progress are considered in the Kelley-Williamson framework. Neutral or "balanced" technical change which enhances the pro- ductivity of both agricultural and manufacturing activities to the same extent acts much like an increase in average national income and exerts an influence on relative demands for labor through Engel effects. Unbalanced technical change, in which faster progress is made in manufacturing than in agriculture, is expressed in labor markets by faster growth in the demand for manufacturing jobs. This fo- cused impact on manufacturing is augmented by the Engel effects that generally accompany technological progress. In the Kelley-Williamson simulations, unbal- anced technological change is shown to exert a powerful influence on the pace of urban growth. Technological change is less potent, however, when prices are determined in domestic rather than world markets, especially when those domestic markets are 2The Gaviria and Stein estimate retains its strength in the presence of country fixed effects. A1- though urban population growth and national income growth are correlated because both depend on country-specific fixed factors, there is clearly more to the story than this. 3 The model contains a number of sectors, but for convenience we refer here only to manufacturing and agriculture. Other sectors play key mediating roles. In particular, the prices of nontradeable goods and services, such as urban housing, are bid up as the model's population shifts to urban areas. These changes in the urban cost of living keep the extent and pace of urbanization in check.
THE URBAN ECONOMY TRANSFORMED 305 competitive. In such cases, technological progress favoring manufacturing will reduce marginal costs in that sector; under competitive pressures, the result will be lower prices for manufactured goods. Price declines, in turn, slow growth in the derived demand for manufacturing jobs. The differences in the effects of technological progress can be seen by comparing the price-taking, open-economy model of Kelley and Williamson with a similar model developed for India (Becker, Williamson, and Mills, 1992) in which prices are determined largely in domestic markets. In the latter case, technological progress continues to exert some influ- ence on urban population growth, but this influence is mild by comparison with the price-taking open-economy case. Just how technological change takes place in poor countries has been much debated. A number of studies suggest that when these countries are able to export manufactured goods, supplier-customer linkages are established that facilitate the adoption of new technologies and spread commercial know-how (Pack and Page, 1994; Tybout, 2000~. The recent literature is ably summarized by Rauch (2001), who describes how exports can give firms and entrepreneurs an entree to interna- tional business networks. Technological progress is generally thought to be skill- biased, in the sense that adoption of new technology raises the relative demand for educated, skilled labor (Johnson, 1997~. This may explain in part why cities possess (and attract) better-educated workers. But as Rodrik (1997) points out, it is very difficult to separate the effects of technological change from those of phys- ical capital accumulation, which is also thought to be strongly complementary to skills (Fallon and Layard,1975~. The process of capital accumulation is acknowledged to be central to eco- nomic growth some authors (e.g., Young, 1995) assign it the leading role and of course a large portion of any country's investment in plant, equipment, and in- frastructure will be found in its cities. But the size of a country's capital stock is one matter and its spatial concentration another. The agglomeration economies that urban sites offer to private firms (discussed in the next section) provide a par- tial explanation for the urban concentration of capital. However, these private ad- vantages depend crucially on governments' complementary capital investments in urban public services and infrastructure (Eberts and McMillen, l999~. Productive cities must be highly capital-intensive. An urban bias in government expenditures may further concentrate capital in cities, as discussed below. But does urbanization itself help mobilize the savings from which such in- vestments are made? It is certainly reasonable to think so.4 As documented in Chapter 4, city populations are more concentrated in the working and saving ages than are rural populations, and they contain relatively fewer young dependents. 4 Two gaps between domestic savings and investment are worth mentioning here. First, urban savings are not necessarily invested domestically; capital flight is evident in sub-Saharan Africa, where, as Collier and Gunning (1999b) show, about 39 percent of all domestic savings is invested outside the region. Second, as discussed later in this chapter, foreign direct investment is an increasingly important source of capital in the cities of some developing countries.
306 CITIES TRANSFORMED BOX 8.1 Industrialization and Social Capital in Indonesia There is much interest in the linkage between social capital and industrialization, but as yet very little by way of data with which to explore this linkage. Cross-sectional studies have suggested that the two are positively associated, but such studies leave it unclear whether social capital supports economic development, economic development gives rise to social capital, or the two are linked in a system of mutual causation. Exploiting unusual panel data on Indonesian districts in the period 1985 to 1997, Miguel, Gertler, and Levine (2001) find that growth in a district's manufacturing employ- ment appears to spur the development of local social capital, as measured by participation in arts and recreational groups and some types of credit cooperatives. However, the spa- tial impact is uneven. Because some districts become sites of flourishing manufacturing growth while others do not, social capital also grows unevenly. Outmigration occurs as mi- grants seek manufacturing jobs in nearby districts, and migration, in turn, erodes individual incentives to invest further in social capital in the origin districts. The loss of social capital is particularly evident for informal credit and mutual assistance groups, which depend for their survival on an assumption of stable and reciprocal exchange. Upon examining the other direction of causation, the authors find no compelling ev- idence that social capital either attracts or fosters manufacturing growth. (As they ac- knowledge, establishing causation is difficult even with longitudinal data.) A complicating factor is that the local groups studied in this research are often chapters or affiliates of na- tional groups, and are overseen by their national boards and by government. Hence, local groups probably cannot be viewed as groups that fade or flourish according to purely local circumstances. In national populations, at least, lower dependency ratios have been shown to fos- ter higher average savings rates (Higgins, 1997; Deaton and Paxson, 2000; Lee, Mason, and Miller, 2000~. Urban incomes are higher than rural on average, and this must also encourage savings (Loayza, Schmidt-Hebbel, and Serven, 2000b). In addition, cities house the financial institutions that should allow savings to be monetized rather than held in the form of illiquid physical assets. In formulating their general equilibrium model, Kelley and Williamson (1984: 59-60, 218-20) thought it safe to assume that urban households have higher savings rates than rural households. All this notwithstanding, however, one finds surprisingly little research in which urban savings rates are compared directly with rural rates.5 Indeed, the recent evidence suggests that the savings rates of rural households may well be higher than urban rates. In analyses based on cross-national datasets, Loayza, Schmidt-Hebbel, and Serven (2000a,b), show that with other things held constant, the level of urbanization is negatively associated with average savings rates. 5Severe conceptual and measurement difficulties the valuation of flows from consumer durables such as urban housing (Gersovitz, 1988), the measurement of business savings (retained profits), and the absence of spatially disaggregated data on government savings render such comparisons uncommon.
THE URBAN ECONOMY TRANSFORMED 307 Loayza and Shankar (2000) explore the case of India in detail, using time- series data for 1960-1995. They find private savings rates to be positively asso- ciated with the share of agriculture in GDP. From such aggregate data, it would appear that rural Indians, who lack other means to diversify the risks they con- front, engage in more precautionary savings than do urban residents. Weather and related sources of risk introduce a substantial element of uncertainty in rural areas, and rural economies often lack the protective institutions found in cities. In China, replacement of the commune system with a system of household responsibility reforms that spurred income growth but also removed some insti- tutional protections against risk evidently led to sharp increases in rural savings rates (Kraay,2000~. By contrast, the protections extended in urban areas through work units and state subsidies left these populations without the same incentives to save, at least until recently. According to Kraay, urban household savings rates remained well below rural rates into the mid-199Os. Unfortunately, the data avail- able to Kraay did not include savings among urban temporary migrants, who un- doubtedly save higher fractions of their income than do permanent residents. If a household's educational expenditures were considered alongside its mon- etized savings, might this broader view show that urban savings rates rival rural rates? Tsai, Chu, and Chung (2000) explore the possibility for Taiwan, and dis- cover that metropolitan households engage in less monetized saving but make greater investments in their children's schooling. Schooling investments can sub- stitute in several ways for savings, especially given the obligation for educated Taiwanese to repay their parents by providing them with support in old age. Still, the result is surprising because Taipei and the country's other large cities contain many financial institutions that should facilitate savings.6 Spatial Effects: Agglomeration Economies To examine how the sectoral effects sketched above are expressed in spatial terms, we turn to the theory of agglomeration economies, which addresses the productiv- ity and cost advantages derived from the spatial concentration of production and labor. The main elements of this theory date to the pioneering work of Marshall (1890), Weber (1909), Christaller (1933), and Losch (1954) in economic geog- raphy. The outlines of the theory were set forth in Chapter 2, and here we need only mention the extensions and specific features that appear to be of particular salience for developing countries. In a world with transportation and communication costs, agglomerating ten- dencies arise from several sources. Scale economies in production can lead firms to concentrate their activities instead of spreading production across spatially dis- persed small plants. If production is broadly conceived to include the recruit- ment and training of labor, the maintenance of inventories, and the formulation of 6See Gersovitz (1988) for a review of other attempts to include educational expenditures with monetized savings.
308 CITIES TRANSFORMED contractual relations with other firms, and if transport and communication costs affect each of these elements, firms can derive many productive advantages from locating in large, diverse economic clusters. As discussed in Chapter 2, the ad- vantages of spatial concentration can be grouped into two categories: localization economies, which pertain to the benefits of clustering with other firms in a given industry, and urbanization economies, which have to do with the benefits obtained from the size and diversity of the cluster. Localization economies arise when spa- tial concentration gives firms easier access to industry-specific inputs, such as specialized labor or services. Urbanization economies stem from the need for inputs used across industries and sectors, such as generalized legal and financial services and well-educated labor. Some of these productivity benefits are market- mediated, termed pecuniary externalities, whereas others are pure externalities, which often take the form of informational spillovers. Agglomerating forces can be either strengthened or weakened by the natural heterogeneities of space. For instance, rivers and natural harbors often form the kernel of urban economies. Yet when valuable mineral deposits are geographi- cally isolated, firms engaged in their extraction may find it necessary to establish operations in locations remote from cities. Although the spatial outcomes differ, in each case firms choose their locations by balancing scale economies against transport costs. Because sites with natural advantages tend to attract more firms and people, it can be very difficult to separate natural advantages from localization and urbanization effects. Even where longitudinal data are available, they seldom provide sufficient over-time variation in urbanization and localization measures for their effects to be clearly distinguished. One study attempting to isolate the role of natural advantage, that of Ellison and Glaeser (1999) for the United States, finds that as much as one-fifth of the geographic concentration of production, and quite possibly more, can be explained by such natural, location-specific features. Still, even this estimate leaves a large portion of concentration to be explained by localization and urbanization economies, as well as by other factors. A point much stressed in reviews of the literature (notably that of Eberts and McMillen, 1999) is that the advantages of clustering may or may not materialize, depending on the availability and quality of public services. Public provision and regulation of transport is one example of critical infrastructure; another is the re- liable delivery of such inputs as electricity and water (World Bank, 1994: 30-31~. As discussed in Chapter 5, the provision of basic public services is far from being complete in the cities of most poor countries, especially when account is taken of the reliability and adequacy of supply, and it is likely that there are significant clustering advantages still to be realized in most cities. Other publicly provided services such as health and education services may have significant comple- mentary effects on the costs of production. For instance, highly educated work- ers may much prefer to live where their children can be assured of good-quality schools. To live elsewhere, they might insist on a wage premium that would raise costs for firms. In short, when public services are unavailable or unreliable, the
THE URBAN ECONOMY TRANSFORMED 309 advantages of spatial concentration are likely to remain largely theoretical and to be threatened by congestion and the other diseconomies of agglomeration. Evidence for developing countries What is the special relevance to poor countries of the general points made above? In high-income countries, transportation costs have greatly declined as a share of total costs, and this has weakened one rationale for spatial concentration (Quigley, 1998~. In most poor countries, however, transportation and communication costs surely remain high. Are economies of scale in production then sufficiently great to induce spatial concentration? On this point the empirical evidence suggests caution. In reviewing a number of empirical estimates for developing-country manufacturing, Tybout (2000) finds evidence that scale economies are either slight or altogether absent.7 Without stronger evidence of scale effects, a theory built on transport costs and conventional scale economies alone appears inadequate. At least for the moment, the more promising findings are contained in the literature on cross-firm and cross-industry spillovers. Several empirical studies of localization and urbanization effects have been conducted in Brazil and India, although studies are regrettably scarce in the remainder of the developing world (Henderson, 1982, 1988; Shukla, 1996; Mitra, l999~. In his analysis of Brazilian cities, Henderson (1982,1988) detects little evidence of urbanization economies, but finds substantial evidence of within-industry localization economies. For the industries he studies, the elasticity of a firm's output with respect to local employ- ment in its industry (his preferred measure of localization) generally exceeds one- tenth. With reference to India, Shukla (1996) examines the elasticity of a firm's output with the size of the surrounding urban population, this being a measure of urbanization economies. She also finds significant positive elasticities averaging about one-tenth and ranging as high as one-fifth in the case of the basic metals and alloys industries. These are effects of considerable substantive importance, as can be appreciated by comparing the productivity of a firm located in a town with that of an otherwise identical firm located in a large city. In tests comparing urbanization with localization measures, Shukla finds that the urbanization effects are larger in 11 of the 13 industries she examines. Mi- tra (1999) also examines the case of India, estimating production functions for electrical machinery and cotton and cotton textiles. His estimates suggest that economies of urbanization are greater in cities of intermediate size than in either smaller or larger cities. For instance, electrical machinery firms derive the great- est productivity benefits from locating in cities of 1.5-2.5 million; for firms in the cotton-related industries, the maximum benefit is obtained from cities of 2.5- 5.0 million. Although Mitra and Shukla stress urbanization economies in their 7Likewise, in high-income settings, scale economies in manufacturing are perhaps less important than they once were, owing to declining fixed costs in computing and the use of techniques permitting smaller, more specialized production runs.
310 CITIES TRANSFORMED work, Henderson is not alone in emphasizing localization economies. A study of Mexican manufacturing by Grether (1999) shows that within-industry spatial concentration generates substantial productivity benefits. Eberts and McMillen (1999) call attention to a curious gap in research on de- veloped countries: agglomeration effects are examined in some studies and the effects of public infrastructure in others, while very few studies test for public- private complementarities. The situation is hardly different in poor countries. Us- ing an array of measures of provision of infrastructure and social services, Shukla (1996) examines the implications for industry productivity in Maharashtra. In this Indian state, public-sector capital inputs are shown to enhance the productivity of private firms. Water supply, roads, electricity, and health and educational facilities are evidently gross complements in their direct impact on private output; they ap- pear to be substitutes for private industries and public enterprises. Jimenez (1995) and the World Bank (1994: 15) review similar findings from a range of developing and developed countries. What is lacking, however, is direct evidence on whether public capital itself enhances the returns to spatial clustering. Spatial concentration and growth: Dynamic elements Over the past decade, the dynamic effects arising from agglomeration have at- tracted an increasing amount of research attention. The surge of interest is of- ten dated to the emphasis on scale and externalities in new theories of economic growth (Lucas, 1988), but economists have long been intrigued by the idea that proximity and diversity can combine in forms of social learning that generate innovation. Alfred Marshall (1890) anticipated the current interest by more than a century, writing on both the static and dynamic aspects: . . . great are the advantages which people following the same skilled trade get from near neighborhood to one another. The mysteries of the trade become no mysteries: but are as it were in the air.... Good work is rightly appreciated, inventions and improvements in machin- ery, in processes and the general organization of the business have their merits promptly discussed: if one man starts a new idea, it is taken up by others and combined with suggestions of their own; and thus it becomes the source of further new ideas. Modern theoretical treatments have drawn out the implications of such social in- teraction for innovation (e.g., Jovanovic and Robb, 1989~. As noted in Chapter 2, one important theme is that spatial proximity and urban diversity are especially important in the design and testing phases of product development and become less important once the production process has been standardized (Duranton and Puga, 20014. Empirical tests of these ideas are not yet common, in part because
THE URBAN ECONOMY TRANSFORMED 311 convincing tests require longitudinal data on the social and economic interactions highlighted in the theory.8 One important strand in this literature addresses the question of whether social interactions can affect the economic returns to schooling. As was seen in Chapter 5, developing-country cities differ from rural areas not only in the average level of adult schooling, which is of course higher in cities, but also in the diversity of educational attainment. Cities are locations where well- and poorly educated workers are more likely to come into contact. Might this proximity improve the prospects for informational spillovers and other beneficial externalities? Some evidence to this effect has emerged for the United States. Rauch (1993) examines the earnings of individual workers in U.S. metropolitan areas. He finds that with individual education and experience held constant, individual earnings in these areas tend to increase with average level of education. The effect is robust to the inclusion of a variety of individual and metropolitan covariates. This is intrigu- ing but hardly conclusive proof of beneficial spillovers: the marginal productivity of any one input (say, unskilled labor) is apt to increase with the levels of other productive inputs (say, skilled labor) in the local economy, even in the absence of spillovers.9 Stronger evidence of spillovers is provided by Moretti (2000), who finds that the earnings of individual well-educated workers increase with the pro- portion of similarly well-educated workers in a metropolitan area, again holding individual characteristics constant. This positive effect is inconsistent with con- ventional theories of marginal productivity, and may therefore represent a true beneficial spillover. Moretti's estimate is strong enough to withstand a battery of tests for selection effects and other statistical biases. If well-educated labor does indeed generate positive spillovers, an increase in the proportion of well-educated labor in the local economy will raise the earnings of the less educated, as is seen in the earlier work by Rauch (1993~. These are U.S. examples, to be sure, and we are not aware of comparable research in the cities of poor countries. However, there is no compelling reason to rule out similar effects in these cities. If social interaction has the effect of increasing productivity and spawning in- novation, what is the mechanism by which this occurs? Nothing appears to be known about this for workers, but where investment and innovation in manufac- turing are concerned, some researchers see a key role for the social networks of owners and entrepreneurs. Collier and Gunning (1999b) give a general account fin developing countries, of course, even conventional longitudinal studies of firms and industries remain rare. In a follow-on to the Indian research described above, Mitra (2000) examines productivity growth using panel data on a number of industries from the late 1970s to the early 1990s. Only a third of the industries studied show significant urbanization effects (as measured by the urban population of the state). These are weaker results than those obtained by Mitra (1999). 9Consider a Cobb-Douglas production function Y = Ua!Sl-a!, with U being unskilled labor and S skilled labor. The marginal product of unskilled labor is orUa!-lSl-a!, which increases with the level of S. Without spillover effects, the marginal product of unskilled labor U will fall as the level of U increases. Hence, tests for spillovers focus on the empirical evidence against diminishing marginal productivity.
312 CITIES TRANSFORMED of the role of social networks in sub-Saharan Africa, and Murphy (2002) explores their specific role in Mwanza, a small Tanzanian town. In keeping with the long tradition of network analysis in sociology, these au- thors stress two roles for social networks. First, networks are viewed as impor- tant mechanisms for the transmission of new information, especially when they contain individuals who can serve as "bridges" between otherwise unconnected groups. Networks can thus present their members with new economic options and offer examples of the consequences of innovative decisions. (This idea is well expressed in the above quotation from Marshall.) Second, from the view- point of a given network member, his or her contacts constitute a set of personal relations differing in reliability and trustworthiness, ranging from persons with whom there are strong ties to others with whom ties are infrequent or weak. Collier and Gunning (1999b) and Murphy (2002) apply these ideas to deci- sions about business expansion in Africa. As they describe the situation, the trusted network members of African entrepreneurs are often others within the same family or ethnic group. The dilemma is that in these trusted circles, new economic information will tend to appear less frequently than in wider networks. But the wider networks include contacts who are not trusted to the same degree, and an entrepreneur hearing of a profit opportunity through such contacts may feel that the risk is simply too great to proceed. In some settings, an entrepreneur might turn to a third party say, a firm that conducts creditworthiness checks, audits, and the like to assess business risks and ensure the reliability of a potential partner. Also, where governments maintain well-functioning and transparent commercial courts, an entrepreneur can expect to recoup some of the losses from a failed partnership and thus limit the downside risks of business expansion. In sub-Saharan Africa, however, the key business service sectors have yet to develop in many cities, and few small-scale entrepreneurs can count on the effectiveness of commercial courts and the impar- tiality of local officials. In the absence of these institutions, African entrepreneurs must fall back on their trusted network partners and abandon profit opportunities not conveyed through their most trusted local circles. The net effect is a contrac- tion of investment and slower economic growth than might occur in the presence of complementary business services and a coherent legal system. To sum up our review of spatial concentration and innovation, we find many reasons to believe that localization and urbanization effects can raise productivity and foster economic growth. As our review shows, theories suggesting such ef- fects have well outpaced any supporting empirical research. Hence no strong con- clusions can yet be rendered. But we see enough in the combination of compelling theory and (limited) empirical findings to conclude that urban spatial concentra- tion must be one cause of national economic growth. it The size of its contribution is not yet known but is surely of considerable significance. None active area of research on spillover effects is the performance of industrial clusters in developing countries (Schmitz and Nadvi, 1999; Bell and Albu, 1999).
THE URBAN ECONOMY TRANSFORMED Spatial Deconcentration 313 We have been describing the benefits of the spatial concentration of resources, but these advantages run up against limits imposed by congestion and the rising prices of land and other location-specific nontradeables. Data on urban systems suggest that with economic development comes a tendency toward urban decen- tralization (Rosen and Resnick, 19804. This tendency is not always evident, and given the structural changes involved, its meaning is debateable. Many factors can influence both development and decentralization, perhaps the most prominent among them being national transportation policies and investments. But mono- causal theories of decentralization are seldom seriously advanced in the literature. Most researchers acknowledge the powerful role played by economic forces, but stress that even these forces are not necessarily dominant. Political and historical factors are also given their due, with particular reference to the spatial aspects of political power and access to bureaucrats. The complexities of decentralization can be seen in two recent empirical anal- yses using cross-national comparative data. Ades and Glaeser (1995) aim to ex- plain differences in the population size of each country's largest urban agglom- eration.~i Other things being equal, these cities tend to be relatively larger in countries with higher levels of national income per capita and higher ratios of imports and exports to national income. The main city's population is smaller, however, in countries with higher national road densities and greater government expenditures on transport. Taken together, these findings imply that higher levels of income per capita will yield decentralized city systems only if accompanied by deeper investments in transport. Three political factors are found to matter: main cities are larger when they are national capitals, and both political insta- bility and dictatorship forms of government are associated with larger city size. Political effects are also seen in the earlier analysis of Henderson (1982), who shows that decentralized city systems are associated with decentralized political systems. Gaviria and Stein (2000) revisit some of these issues with panel data on the countries in the Ades-Glaeser sample; they focus on the population growth rate of the largest city rather than its size. Some of the earlier findings survive. For instance, where Ades and Glaeser show that higher levels of GDP per capita are associated with larger main cities, Gaviria and Stein show that faster GDP per capita growth rates are associated with faster city population growth. But Gaviria iiThey include country size as a control variable, which has the effect of expressing all results in relative terms. i2Again a control variable the growth rate of the country's total urban population is introduced to convert the results to relative terms. See Chapter 4 on the United Nations procedures for estimating city-specific population growth rates and the possibility of measurement errors in these data. Gaviria and Stein find that, with other things held constant, larger cities tend to grow more slowly than smaller ones. This tendency has been well documented by the United Nations (2001) and has long been embedded in its city projection methods.
314 CITIES TRANSFORMED and Stein fail to confirm the Ades-Glaeser result on dictatorships. They note that the effects of openness to trade depend on geography: the population growth rates of landlocked cities tend to fall with an increase in the ratio of trade to GDP, while no effect of trade can be detected in the growth rates of port cities. It appears that a set of empirically replicable "stylized facts" about urban Reconcentration has yet to be established. As suggested by the analysis of Ades and Glaeser (1995) and by many coun- try case studies, investments in transportation infrastructure can play an important role in decentralization. As improvements are made in transport and commu- nication, firms are freed to seek the locations that offer them the best combina- tion of productivity advantage, rents, and wages. Henderson (1988) describes this process for Korea. In the region surrounding Seoul, manufacturing began to Reconcentrate in the 1970s, moving from the inner to the outer rings of the re- gion. Alternative sites in the south of Korea (near Busan) also began to grow more rapidly than Seoul. Firms in the machinery and metal sectors, whose pro- duction had become routinized, followed the logic of the product development cycle and sought cheaper land and labor outside the capital. Large firms some- times maintained head offices in Seoul but decentralized their production plants; other firms relocated all their operations but stayed in contact with Seoul bureau- crats by remaining within a half-hour's drive. In the Korean case, better transport links within and outside the capital region greatly facilitated decentralization; so did the increasing availability of good-quality local public services, which allevi- ated the concerns of high-skilled workers about relocating. Although the Korean government offered firms a number of targeted incentives for relocation, including tax breaks and preferential access to credit, in Henderson's account these policies had little effect. As the Korean example makes clear, the process of decentralization combines economic with political elements in a way that limits the explanatory power of narrowly focused economic models. Still, some insights into the process can be gleaned from formal analyses. The most fully articulated of the economic theo- ries is that of Henderson (1982,1985,1988), in which the organizing concepts are those of localization and urbanization economies. As discussed above, firms ben- efiting from localization economies can reap productivity advantages from spa- tial clustering with other firms in their industries. A system of specialized cities can then emerge, with the smallest cities of the system being dominated by in- dustries whose localization economies are exhausted at low levels of clustering (e.g., in industry employment). The largest cities then house the industries with gains from localization that persist even in large clusters 33 Specialization of this type establishes the spine of the city system, to which industries drawing benefits from urbanization economies city size and diversity then attach themselves. i3Such localization effects are most powerful when industry prices are held fixed, as in the case of production for export.
THE URBAN ECONOMY TRANSFORMED 315 In all cities, the effects of clustering on congestion costs, rents, and wages provide counterpressures that keep the agglomerating tendencies in check. Although Henderson makes only modest claims for the explanatory power of this elegant theory he offers it as a set of regularities that might be glimpsed beneath the embroidery of other economic relations the theory has drawn at- tention because of the ease with which it can be linked to international markets. Cities specializing in goods traded on world markets will be directly affected by changes in world prices. An increase in any given world price will be expressed in a national city system by an increase in the number of cities that specialize in the production of the good in question. Thus, when world prices change for goods in which medium-sized cities specialize, the spatial implications should be evident in this portion of the city hierarchy. National policies that subsidize or tax par- ticular goods will have effects akin to those of international prices. In principle, at least, these are empirically testable propositions, although proper tests would require unusually detailed, city-specific data. Fujita, Krugman, and Venables (1999) have fashioned an alternative city sys- tem theory one that can also be linked to international markets from rather different materials. Perhaps the most striking feature of their theory is that it is wholly free of reference to localization and urbanization externalities. Rather, the concepts at the core of the model are those of diversity, increasing returns to scale, and transport costs. As used here, "diversity" refers to the preferences of consumers for a wide range of manufactured goods. When consumers have only weak preferences for diversity, the ensuing pattern of demand encourages firms to produce a narrow range of goods with high levels of output for each, thereby exploiting the available economies of scale. When preferences for diversity are stronger, a greater range of goods is produced but at lower output levels for each. The cost of conveying finished goods to consumers is a main determinant of the spatial clustering of firms. As firms and their labor forces gather around any one location, the consumer market associated with that location expands, and this, in turn, raises its attractiveness for other firms. Such agglomerating tenden- cies are countered by the need for firms to exploit profit opportunities in loca- tions that are not so heavily populated by their competitors. After much complex argument see Krugman (1991) for the seminal paper in this line of research, Fujita, Krugman, and Mori (1999) for some of the detail, and Neary (2001) for a penetrating critique this approach can be shown to yield an urban hierarchy with a dominant central city and subsidiary cities occupied by industries with high transportation costs. Interestingly, the decentralizing tendencies can be hidden in small populations but are exposed as the national population grows. In one early expression of these ideas, Livas Elizondo and Krugman (1992) ex- plain how openness to international trade can shape the roles of larger and smaller cities. (The authors have Mexico City and Monterrey, California, in mind, with the United States playing the role of trading partner.) They imagine a scenario in which the population of Monterrey initially equals that of Mexico City. As
316 CITIES TRANSFORMED protectionism toward the United States increases, the allocation of population be- tween these cities becomes unstable, and a tendency emerges for labor to concen- trate in one large city. In this way, market size effects cause closed economies to produce dominant megacities. When the economy is opened to trade, this lessens the need for firms seeking large markets to locate in cities and permits a more dispersed city system. In these theories, small cities are viewed mainly as small concentrations of industry. But small and intermediate-sized cities are often linked to the sur- rounding rural settlements by complex two-way interactions; these include trade; employment; and the provision of services, such as health care and secondary education (Hardoy and Satterthwaite, 1986b; Kamete, 1998~. One of the most important linkages is the stimulus that prosperous agriculture can provide for urban development. Reviews of urban change often highlight how many of the most rapidly grow- ing urban centers are found in areas where the value of agricultural production is increasing most rapidly (Manzanal and Vapnarsky, 1986; Hardoy and Satterth- waite, 1988; Blitzer, Davila, Hardoy, and Satterthwaite, 1988; UNCHS, 1996~. Among the most important factors are the value per hectare of the crops (the higher the value, generally the more local urban development occurs); the po- tential for local value-added activities (and the scale of forward and backward multiplier linkages between agricultural production and local urban enterprises); and the structure of landownership (the greatest stimulus to local urban develop- ment generally being the existence of a large number of prosperous, relatively small farms growing high-value crops) (Hardoy and Satterthwaite,1988~. Examples of the strong stimuli for urban development provided by agriculture include fruit production in the Upper Valley of Rio Negro in Argentina (Man- zanal and Vapnarsky, 1986) and in Bangalore, India, the multiple interconnec- tions between rural silk cocoon producers and urban enterprises involved in col- lecting and processing the cocoons, making silk, and making and selling saris (Bhooshan, 1986; Benjamin, 2000~. In some cases, however, prosperous agricul- ture gives little stimulus to nearby urban development, especially where landown- ership is highly concentrated and more extensive crops are grown (Hardoy and Satterthwaite,1988~. Where markets are dominated by large local merchants who control access to transport and marketplaces and, in many instances, access to cap- ital, credit, and information, much of the value of agricultural production can be steered out of the region (Tacoli,1998~. There is little to suggest that governments recognize the potential for pros- perous agriculture to support urban development beyond the local level. Many agricultural towns and cities have been starved of the support they require from higher levels of government to aid economic expansion and meet the needs of rapidly growing populations. In some instances, agricultural policies have pre- vented or discouraged rural producers from diversifying production and trapped them in low-profit crops with few forward and backward linkages. Examples can
THE URBAN ECONOMY TRANSFORMED 317 be found in many Asian nations with policies designed to ensure adequate rice production to feed urban populations (Douglass, 1998, 1999~. Comprehensive urban/rural development frameworks and regional spatial planning in the 1970s and 1980s generally focused on trying to expand industrial production in smaller urban centers and often failed to identify and support the potential comparative advantages of each locality. However, this failure to support a prosperous agri- culture that could in turn underpin urban development also relates to obvious political constraints, such as inequitable landowning structures and pricing and marketing structures that held down rural incomes. While many of these are national-level constraints, the internationalization of trade and production is an increasingly important dimension that affects local economies through the rise of international agro-industry and the resulting marginalization of small farmers (Bryceson, l999~. Where statistics are available on occupational structures within smaller urban centers, they often show a significant proportion of the labor force working in agriculture, livestock, forestry, or fishing. The panel found no recent studies on occupational structures in smaller urban centers. However, case studies of cities with under 100,000 inhabitants in predominantly rural regions of low-income na- tions, undertaken during the early 1980s, revealed that it was common for 10-30 percent of the workforce to be employed in agriculture, with many more having occupations that serve demand generated by agricultural production or agricultural incomes (Hardoy and Satterthwaite, 1986b). Agriculture may also provide the livelihoods of a significant proportion of the population in larger cities although the portion of any city's workforce in agriculture is greatly influenced by how much the urban boundaries extend beyond the built-up areas. The fact that large proportions of the economically active populations in some of the largest cities of China and Bangladesh work in agriculture reflects the setting of urban bound- aries that encompass large areas of agricultural land. Statistics on urban occupa- tional structures are also unlikely to include most of those who engage in urban agriculture for instance, those who do such work on a part-time basis or outside normal hours or who are not registered as working in agriculture (Smit, Ratta, and Nastr,1996~. Urban Bias Revisited Few would disagree with the statement that the policies of developing-country governments have often favored large cities and the urban elites (Lipton, 1976~. Exchange rate, tariff, and tax policies appear to have penalized rural dwellers more than urban; by distorting economic signals, they have also reduced national out- put. In recent years, however, such price distortions have been reduced in many developing countries, often through reforms undertaken in programs of structural adjustment (Collier and Gunning, l999a,b). Many price-related biases no doubt remain, and in rural areas they are compounded by poor transport systems and
318 CITIES TRANSFORMED (in some cases) restrictions on competition among agricultural wholesalers and retailers. Lipton (1993) argues that price distortions are no longer the main expression of urban bias, and that expenditure biases have become more important. In his view one that is widely shared public investments in infrastructure and social spending are often undertaken in cities without sufficient attention to the rates of return that could be secured from alternative rural investments.~4 In many countries in sub-Saharan Africa, public employment appears to have expanded in capital and other cities well beyond what can be justified on efficiency grounds (Becker, Hamer, and Morrison, 1994~. The urban bias argument is in part an argument about missed opportunities to improve economic efficiency through rural investment, but inequities are perhaps its dominant theme.~5 The demographic data presented in earlier chapters of this report for instance, the data on public service delivery discussed in Chapter 5 document the many disadvantages suffered by rural populations by comparison with their urban counterparts. Although we have also drawn attention to the urban poor, who are themselves disadvantaged, these data serve as a reminder of the gross inequalities between urban and rural residents in many developing countries. Although no one would dismiss the possibility of an urban bias in expendi- tures, the element of bias is more difficult to isolate than is commonly recognized. The spatial concentration of urban populations allows economies of scale and scope to be exploited in urban infrastructural investments, and these economies are not readily captured in dispersed rural settings (Becker and Morrison, 1999: 1722; Montgomery, 1988~. Investing in such rural infrastructure as irrigation, trans- portation, and agricultural extension can improve rural productivity, reduce food costs, and allow labor to be released for urban occupations. But when capital is scarce and the unit costs of rural investments are high, an argument can be made for an initial focus on urban investments (Williamson, 1965~. In principle, one can imagine a fair-minded social planner choosing to schedule investments in urban and then rural phases. But these are speculative arguments. Acknowl- edging the many difficulties involved in assessing price and expenditure biases, Williamson (1988) points to the curious absence of empirical studies linking such i4For an analysis of the formidable difficulties involved in assessing rates of return to rural road investments, see van de Walle (2002). i5A genre of political-economic models explores these issues, noting that urban interests can some- times be served by rural investments. In an extreme case (such as the Soviet Union in its industrializa- tion phase), urban elites interested only in industrial development can agree to subsidize rural infras- tructure to increase food supplies and keep urban wages low (Dixit, 1969). As Hansen (1979) argues, neither the rural nor urban elites (large landowners and industrialists) can be expected to have much interest in diminishing the pool of subsistence labor in their sectors. While industrialists might see benefits in giving urban workers schooling and other social infrastructure, measures to raise produc- tivity in traditional agriculture may not work to benefit rural landlords and can be disadvantageous for urban industrialists. In this sense, an urban "bias" in public expenditures is to be expected, assuming that government is controlled by wealthy elites.
THE URBAN ECONOMY TRANSFORMED 319 policy biases even to urbanization. The efforts of Malpezzi (1990) to link urban growth to measures of price distortion yield almost no correlation, and the earlier studies of Montgomery (1987) and Becker and Morrison (1988) are no more suc- cessful in isolating an effect. In short, although there is every reason to think that a measure of urban bias continues to plague pricing and expenditure policies in many countries, the magnitude of this bias remains unknown. ECONOMIC RETURNS TO SCHOOLING Earlier chapters have described the ways in which economic returns to schooling can affect demographic behavior. That returns to schooling have a spatial di- mension is well recognized. As suggested in the preceding section, gaps in school- ing returns can be expected to emerge across the national space as economies urbanize and decentralize their city systems. The greater payoff to schooling in urban than rural areas is a main factor perhaps the dominant one in higher rates of migration among the educated. Rates of return to schooling are also potentially important determinants of fertility. When the returns to investing in education are high, this can encour- age families to adopt productive and reproductive strategies in which their fertil- ity is reduced and investments in their children's education are increased. The "quantity-quality transition" is a central mechanism in economic development: it ultimately produces slower labor force growth and higher levels of human capital per worker. The logic of supply and demand prompts doubt, however, about whether high returns to schooling can be sustained (Fallon and Layard, 1975; Montgomery, Arends-Kuenning, and Mete, 2000~. In recent years, the supply of educated labor has been increasing rapidly in many developing countries, and although figures specific to urban areas are not widely available, we believe the urban increases have been particularly large. If other things were held constant, such supply shifts would depress the economic returns to schooling and mute the demographic re- sponses described above. Of course, countervailing forces can be at work on the demand side that maintain or even increase high returns to schooling. The accu- mulation of physical capital can raise the relative demand for better-educated labor if skilled labor is complementary to capital (Fallon and Layard, 1975~. Exoge- nous technological change perhaps linked to manufactured exports can also increase demand for better-educated labor. Recent economic theories of endoge- nous growth and the small empirical literature described above (Rauch, 1993; Moretti, 2000) suggest that urban concentrations of highly educated labor may generate positive spillovers that raise the returns to lower levels of schooling. i6See Krueger and Lindahl (2001) for a review of the concept and measurement of returns to schooling.
320 CITIES TRANSFORMED Such demand-side forces could well keep schooling returns high even in the face of rapid shifts in supply. Psacharopoulos (1985, 1994) summarizes a large number of empirical esti- mates of the rate of return to primary, secondary, and tertiary schooling. (Many of these estimates are based on urban earnings regression equations.) In the de- veloping world, the private returns to primary education range from 17 to over 40 percent, with the high estimates being found in early studies (some using ques- tionable data) of sub-Saharan Africa. At the secondary-school level, estimates of returns vary from 16 to 27 percent, and at higher levels of schooling they lie in the 19-28 percent range.~7 Although private returns are unreliable guides to the social value of investments in schooling for that purpose, they would need to be adjusted for educational subsidies, credentialism, and other factors that impart upward bias they approximate the measures that families themselves consider in making demographic decisions. Are shifts in the supply of educated labor generally reducing the returns to schooling? There is some evidence to this effect for Korea (Kim and Topel, 1995), where the expansion of university schooling drove down its private rate of re- turn in the 1980s. Elsewhere in Asia there is little evidence of any systematic decline; see Montgomery, Arends-Kuenning, and Mete (2000) for a review. In sub-Saharan Africa, however, unmistakable evidence has emerged of falling rates of return in those countries whose economies have been stagnant or worse. In such settings, the demand-side forces may simply be too weak to maintain the returns to schooling when rapid expansion is underway on the supply side. The case of Taiwan, a country that has experienced rapid economic growth, is instructive. Lee and Mason (2001) document a substantial increase in the sup- ply of better-educated urban adults. By 1998, about 25 percent of urban men in Taiwan had acquired a junior college or higher degree, and roughly 33 percent had graduated from high school, leaving 42 percent with a junior high school education or less. Urban Taiwanese women had achieved very similar levels of schooling. Yet as the supply of better-educated labor was expanding from 1978 i7Although some high estimates are found in sub-Saharan African, the range is wide even here. For example, Cohen and House (1994) find that in the Sudan, the returns to primary school are lower than those to secondary, which in turn are lower than the tertiary returns a reversal of the pattern usually assumed to hold and that the returns are lower at all three levels of schooling than is typically reported in the literature. Bigsten, Collier, Dercon, Fafchamps, Gauthier, Gunning, Isaksson, Oduro, Oostendorp, Pattillo, Soderbom, Teal, and Zeufack (1998), having access to rich longitudinal data on workers in manufacturing firms in five countries (Ghana, Cameroon, Kenya, Zambia, and Zimbabwe), also show that in manufacturing, the returns to schooling are lowest for those with primary education, greater at the secondary level, and still greater for university graduates. i8See Knight and Sabot (1990) for an unusually penetrating discussion of the biases in rate-of-return estimates, based on two unusual surveys of wage earners in Nairobi and Dar es Salaam. In addition to years of schooling attained, these data include independent measures of cognitive and reasoning abilities. Earnings were found to be better predicted by cognitive scores which arguably measure the amount of learning that takes place in school than by years of schooling as such. The findings raise questions about the meaning of conventional rate-of-return calculations using years of schooling.
THE URBAN ECONOMY TRANSFORMED 321 to 1998, the rates of return to schooling were actually increasing for those with secondary schooling and above, the largest gains being recorded by men with a college or junior college education. Meanwhile, men with less than a junior high school education saw their rates of return to schooling fall. For urban women, rates of return have also been rising among those with a junior college or higher education, although the trends at the low end of the educational distribution are less clear. Evidently, then, over a period of rapid increase in the supply of work- ers with secondary and tertiary schooling, other forces perhaps having to do with technological change and capital accumulation have somehow conspired to enhance the rates of return to that schooling. The case of Brazil, studied by Lam and Dunn (2001), may exemplify the sit- uation of countries experiencing slower but still positive economic growth. This research focuses on the nine largest Brazilian cities, whose populations together grew by 25 million between 1970 and 2000. The private rate of return to schooling has fallen in these cities, but the decline has been so small as to barely register. For men, the rate of return fell from 13 percent in 1977-1979 to 12 percent in 1997-1999 in the nine cities combined. The largest city, Sao Paulo, exhibited the lowest rates of return (12.6 and 11.2 percent at the beginning and end of the pe- riod, respectively), but the decline here was about the same as elsewhere. There is no evidence that urban rates of return were depressed by labor force growth the declining rate of return is almost entirely a phenomenon of the 1990s, by which time labor force growth rates were themselves on the decline (Lam and Dunn, 2001: 17~. Although the Brazilian data testify to the surprising durability of high private returns to schooling, they do not reveal what demand-side factors could be responsible for that persistence. In urban Kenya, an expansion in education comparable to that of Taiwan and Brazil has not been accompanied by economic expansion; indeed, urban levels of consumption per capita have fallen steadily, declining by one-quarter from 1978 to 1995 (Appleton, Bigsten, and Manda, 1999~. Over this period, the rate of return to secondary schooling deteriorated badly, falling by nearly two-thirds (Apple- ton, Bigsten, and Manda, 1999: Table 7~. For those with primary schooling, the wage gains relative to having no schooling also declined.~9 At the tertiary level, however, it appears that the rates of return to schooling have either been constant or increased with time. Although declines in the quality of secondary schooling may have undermined its rate of return, Appleton, Bigsten, and Manda (1999) are skeptical that school quality can be the main explanation. In their view, the high returns enjoyed by early Kenyan cohorts were probably due to their ability to secure employment in well-protected niches of the urban economy. The experience of Kenya shows that where macroeconomic growth has halted and deterioration set in, continued expansion of the supply of schooling can drive Nonetheless, the social rates of return to primary schooling held steady because there were reduc- tions in the costs of schooling teacher salaries fell along with the general decline in wages.
322 CITIES TRANSFORMED down its returns, except perhaps for the relatively small group of students who are able to make their way to university. It is not known whether the Kenyan experience applies to other sub-Saharan countries, although it is surely not atyp- ical of countries in this region and elsewhere suffering from negative economic growth. Knight and Sabot (1990) show that in Tanzania, for example, recent co- horts of primary-school leavers have seen schooling returns of only 12 percent, well short of the 17 percent returns enjoyed by earlier cohorts. In urban Ghana, wage rates for the unskilled in manufacturing firms fell steadily during 1991- 1995, as they had been doing for well over a decade, but the wages of skilled workers remained constant over the period, and the returns to skill thus increased (Teal, 2000~. Matched worker-firm data assembled by Teal (2000) indicate that in Ghanaian manufacturing, skilled workers are more complementary to capital than are unskilled workers. A modest amount of capital accumulation in Ghanaian manufacturing may well have helped stabilize the wages of the skilled workers. The lesson to be drawn from this collection of case studies is that educational expansion need not result in declines in urban returns to schooling, although sup- ply shifts can certainly exert downward pressure when not opposed by forces on the demand side. The literature has not yet isolated the most important of the demand-side influences, although there is much speculation about the effects of physical capital accumulation, export orientation, and skill-biased technological change (Montgomery, Arends-Kuenning, and Mete, 2000~. This is an area in urgent need of research attention. MIGRATION AND ECONOMIC MOBILITY Migration is generally viewed as a behavioral response to spatial differences in standards of living.20 Much of the literature represents migration in terms of in- dividual decisions involving comparisons of real wages or earnings. But in recent years this literature has substantially broadened, and it now accommodates a va- riety of assumptions about the relevant set of decision makers and the economic outcomes they may consider. One now sees family models of decision making that stress the distinct opportunities and costs facing women and men (Riley and Gardner, 1993; Yang and Guo, 1999; De Jong, 2000), household production per- spectives in which some family members are deployed to urban areas and others to rural so as to maximize expected incomes (Zhao, 1999), and portfolio models in which migration is one tactic in the family's larger strategy for managing risk (Stark, 1991~. Although much of the attraction of urban areas for migrants lies in the prospects for upward economic mobility after arrival, very few studies have examined urban income dynamics with the appropriate longitudinal data. Hence, 20Thorough reviews of the migration literature in economics include Yap (1977), Williamson (1988), Lucas (1997), and Becker and Morrison (1999). We stress the volitional aspects of migration here, recognizing that in some countries and regions, substantial migratory flows result from violence, political repression, and natural disaster.
THE URBAN ECONOMY TRANSFORMED 323 it is difficult to say whether migrants' expectations generally coincide with the urban realities. This section closes with a review of what little is known about urban economic mobility. We take the early Todaro (1969) and Harris and Todaro (1970) models of mi- gration as points of departure for a broader discussion. These models are notable for clearly articulating how rural-to-urban migration is linked to the structure of the urban economy, as expressed in its labor market segments and wage rigidities. (The models do not address urban-to-urban migration.) In developing their mod- els, Todaro and Harris placed special emphasis on the urban formal sector, which at the time was conceived to be a segment of the labor market offering relatively high wages and security of tenure. In what follows we review briefly the core elements of these models and describe how the discussion has advanced in the decades since they were formulated. Revisiting the Todaro and Harr~s-Todaro Models Textbook discussions of rural-to-urban migration often depict it as a response to the higher levels of real earnings available in urban areas. If rural and urban earn- ings are determined by market forces, a gap between them signifies a departure from the textbook ideal of an economy-wide equilibrium in which the value of the marginal product of labor is the same across locations. As rural-to-urban mi- gration takes place, the shift of labor drives down urban real earnings and allows rural earnings to rise, leading the economy toward that equilibrium ideal. Mi- gration thus enhances the efficiency with which the economy's full resources are deployed (Adelman and Robinson, 1978; Morrison and Guo,1998~. This benign view of labor markets and migration is challenged by Todaro (1969) and Harris and Todaro (1970), who argue that in an important segment of the urban labor market the so-called "formal sector" wage rates are institution- ally determined and set at levels too high to clear the market. Todaro and Harris further assume that rural residents will migrate or not depending on the prospects for formal-sector employment. Such jobs can be secured, however, only after a period of open unemployment and job search that will commence upon the mi- grant's arrival. Hence in deciding whether to move, potential migrants will take into account not only the formal-sector urban wage rate wit, but also the proba- bility p of securing such a job.2i If the urban expected wage wit p exceeds the rural wage wr, a rural resident will judge migration to be economically rational, provided that moving costs are negligible.22 2iThe time horizons envisioned in these models differ: Todaro (1969) imagines an infinite horizon with ptt) representing the probability that a migrant has obtained a formal-sector job by t periods of residence in the city; in Harris and Todaro (1970), the migrant's time horizon is the length of a single period, and p is the chance of securing a formal-sector job in that period. 22To simplify his presentation, Todaro (1969) assumes that urban informal-sector earnings can be ignored; in effect, these earnings are treated as being fixed at zero. The probability p of securing a formal-sector job is specified by Todaro (1969) as the ratio of the number of new jobs created in a
324 CITIES TRANSFORMED In this framework, an incentive to migrate persists until wit p = wr, that is, un- til urban expected wages come to equal rural wages. Because urban formal-sector wages are fixed at wit, additional migration to cities can serve only to increase ur- ban unemployment and reduce p, the probability of employment. Eventually, un- employment will be driven high enough to equate urban expected earnings wit p to rural earnings. This is an equilibrium of sorts, but, as Todaro emphasizes, reach- ing it may require high rates of urban unemployment (Lucas, 1997: 734~. The models also imply that attempts to stimulate the demand for labor in urban areas are apt to be counterproductive: with urban wages fixed, such demand-stimulating policies can only induce further rural-to-urban migration, raising the total amount of urban unemployment and possibly increasing the urban unemployment rate.23 Hence, development effort is better expended on rural areas; if wr is increased, the spillover benefits will be evident in higher urban employment probabilities. In the early 1970s, this was seen as a compelling way to frame the issues. The Todaro and Harris-Todaro models drew together several strands of the liter- ature that were just coming into prominence at the time notably the notion of a dualistic, highly segmented urban labor market and the idea of equilibrium in expectations and they provided further justification, if any was needed, for de- velopment efforts directed to rural populations. Supportive empirical evidence soon emerged in aggregate and some individual-level studies, notably those of Annable (1972), Barnum and Sabot (1977), and Cole and Sanders (1985~. But as the Harris-Todaro framework came under increasing scrutiny, several key as- sumptions that had initially appeared sensible proved difficult to justify (Berry and Sabot, 1978; Kannappan, 1983, 1984, 1985,1988~. Have urban formal-sector wage rates wit indeed been rigid and fixed at levels high enough to preclude market clearing? As Watanabe (1976) observes, it must have appeared reasonable to assume so in the immediate postindependence era, when trade union pressures and desires to establish "livable" wage levels resulted in high declared minimum wages in some developing countries. In Watanabe's account, trade unions were particularly vigorous in East Africa, the region that supplied much of the inspiration for the Todaro and Harris-Todaro models. By the late 1970s, however, substantial evidence of wage flexibility had emerged in Kenya (Collier and Bigsten, 1981), and from 1978 to 1995, urban real minimum wage levels fell there by half (Appleton, Bigsten, and Manda, l999~. Even in the period to the number of urban unemployed who would be competing for such jobs; Harris and Todaro (1970) specify p to be a function of the stock of urban employed relative to the full urban labor force. The Todaro approach assumes that once a formal-sector position is assumed, it can be held indefinitely; the Harris-Todaro approach, by contrast, depicts the urban labor market as being a kind of job lottery with wholesale turnover in each period. Whether the urban unemployment rate would necessarily increase is unclear. In the Harris-Todaro model, rural wages wr rise with rural outmigration; both wr and the urban employment probability p can adjust to establish equilibrium. In simple versions of this model, rightward shifts of urban demand reduce the equilibrium urban unemployment rate, although the total numbers unemployed can Increase.
THE URBAN ECONOMY TRANSFORMED 325 late 1960s and early 1970s, according to Watanabe (1976: Table 2), the general tendency across developing countries was to allow the real value of minimum wages to erode with inflation, and there is reason to doubt whether the legal mini- mums were ever systematically enforced in very many countries (Teal, 2000~. The scope for high, rigid urban wage setting is now believed to be narrow and is perhaps confined mainly to the public sector.24 To be sure, in some countries government is the dominant formal-sector employer according to Collier and Gunning (1999a), it accounted for half of formal employment in Kenya as recently as 1990 and wages appear to be set above market-clearing levels in some cases (Knight and Sabot,1988~. There are instances in Egypt, for example in which government hiring rules appear to exert a disproportionate influence on urban la- bor markets.25 But in many countries, prolonged weak economic performance has created pressures on government budgets, making it increasingly difficult to finance their large public-sector wage bill. Consequently, public workforces have been maintained only by allowing salaries to decline. In the Sudan, for instance, real public-sector wages in the late 1980s had fallen by 30 percent from the levels that prevailed only a decade earlier (Cohen and House, 1996~. Urban open unemployment rates which also figure prominently in the Todaro and Harris-Todaro models have likewise proved difficult to interpret. Although exceptions can be found, the evidence has generally failed to show persistently high rates of open unemployment across the spectrum of developing- country cities (Kahnert, 1987; Kannappan, 1988~. Open unemployment is sim- ply not a sustainable job-search strategy for many urban job seekers. Fallon (1983), for instance, finds that in urban India, it is mainly secondary school-leavers who are openly unemployed; vigorous demand for university graduates has kept down their unemployment rates, and those who lack secondary schooling have few means to sustain themselves without work of some kind. For Venezuela, Schultz (1982) undertook a careful examination of the role played by unemploy- ment rates in guiding migrants among alternative destinations. Recognizing that 24We have not discussed "efficiency wage" models that allow firms to set wages for some of their employees above the levels prevailing in the external labor market, either to reduce turnover and train- ing costs or to encourage greater employee effort. These models are fundamentally different from the Harris-Todaro perspective in that the internal wage premium is not rigid, but is determined by the returns to firm-specific human capital, the costs of training and monitoring employees, and external market conditions. See Lucas (1997) and Becker and Morrison (1999) for reviews. 25 According to Assaad (1997), the Egyptian government has officially guaranteed a public-sector position to all university graduates and graduates of technical secondary schools. The promise of em- ployment has not been honored in recent years the last cohorts to receive the promised public-sector jobs graduated in the early 1980s but the guarantee remains on the books and has had a perverse effect on recorded rates of open unemployment. Those who want to maintain a place in the queue for public employment may register themselves as unemployed even as they work (unrecorded) in the private sector. The wage gap between public- and private-sector jobs has fallen in Egypt, according to Assaad, although discrimination against women in the private sector increases the relative attrac- tiveness of public-sector employment for them, as do the nonwage benefits attached to government jobs.
326 CITIES TRANSFORMED open unemployment might not be a viable job-search strategy for poor poten- tial migrants, he carried out separate analyses of migration by level of education. Schultz uncovered clear evidence that rates of unemployment mattered most for the better-educated potential migrants. Among those with primary schooling or less, unemployment rates exerted no discernible influence on locational decisions. Wage rates, however, mattered for all education groups. The view that urban labor markets are separable into formal and informal sec- tors, with the formal sector offering high wages and long tenure and the informal sector low wages and insecure employment, is now recognized as simplistic and in need of substantial reappraisal. We return to the issues later in this chapter. The point to note here is that the distributions of earnings in formal and informal jobs often overlap; many studies have failed to detect any pronounced formal-sector wage advantage for comparable workers. Even where a formal-sector wage ad- vantage once existed, it has tended to erode with time, with the deterioration being especially marked in Africa. As discussed below, a notable development of the past 20 years is the increasing informalization of urban labor and product markets, which has further weakened the case for focusing solely on the formal sector. In short, the assumption that rural migrants are motivated mainly by the prospect of formal-sector employment places more emphasis on this one segment of the urban labor market than is warranted. Of all the assumptions that motivate the Todaro and Harris-Todaro models, perhaps the most understudied is that migrants must rely on a "move first, then search" strategy with regard to urban employment. Although this assumption cannot be categorically rejected, several case studies suggest that it, too, is open to question. An analysis by Bigsten (1996) of rural-to-urban migration in Kenya shows that many migrants have jobs arranged before they move, with circular mi- gration providing a mechanism for transmitting job information. In a detailed study of migrants to Delhi, Banerjee (1981,1983,1986) finds that over half of mi- grants either had their urban jobs lined up before arrival or had received strong as- surances that a position would be available. These Indian migrants tended to take positions similar to those held by their urban relatives and contacts. In Manila (Aratme, 1992) and Bangkok (Phongpaichit, 1993), there is clear evidence of pre- arranged employment, especially for informal-sector jobs. If these experiences from Kenya, India, the Philippines, and Thailand are indications of what is true more generally, open unemployment upon urban arrival may be less common than envisioned in the early theory.26 26Fields (1975) develops an interesting variant on the Harris-Todaro model in which urban residents can choose either to search full time for a formal-sector position (while being openly unemployed) or to search part time (while holding an informal-sector job) at reduced search efficiency. He retains the assumption of rigid formal-sector wages but allows informal-sector wage rates to be flexible. In the model's equilibrium, the rural wage wr equals the expected urban wage from full-time search, which in turn equals the expected urban wage derived from informal-sector employment combined with part- time search. Because Fields assumes that informal-sector urban wages are flexible, his model predicts that in equilibrium, urban informal wage rates will lie below the levels of rural wages. A comparison
THE URBAN ECONOMY TRANSFORMED 327 It is not really surprising that migrants would draw on information from their families, social networks, and other contacts before deciding to relocate. Informa- tion chains and social support networks have long been recognized as important to international migration (Palloni, Massey, Ceballos, Espinosa, and Spittel, 2001; Fortes, 1995~. Social networks have not received the same attention in studies of internal migration, but the literature shows that here, too, the networks of potential migrants can provide information and supportive ties that are helpful in job search. A study of Khartoum (Cohen and House, 1996) finds that urban employers often recruit new workers by asking their current employees to recommend family mem- bers or friends; employment information thus tends to circulate along family and ethnic lines. Kannappan (1988) also stresses these informal channels. A number of authors describe how rural-to-urban migran ts take care to maintain their rural connections (Naved, Newby, and Amin, 2001; Townsend and Garey, 1994; Hoops and Whiteford,1983~. In addition to the studies cited above, Hugo (1981), Fuller, Lightfoot, and Kamnuansilpa (1985), Menon (1987,1988), and Root and DeJong (1991) stress the importance of both family ties and social networks to migration decisions. Moreover, as discussed in Chapter 6, social networks are important in facilitating migrants' adjustment to urban life (Lomnitz, 1997; Pick and Cooper, 1997; Seller, 1997; Srinivasan,1997~. The Composition of Migrant Streams Although the point is not much emphasized in economic theories of migration (see Stark, 1991, for an exception), it is obvious that the demographic and eco- nomic implications of migration differ depending on whether moves are perma- nent or temporary in nature. The case of China is of particular interest because in this country, temporary migration was a response to new economic opportuni- ties emerging just as powerful state restrictions on movement were being eased. The Chinese case also illustrates how migration can present men and women with different risks and opportunities. Yang and Guo (1999) describe temporary migration in China, relying on an unusually detailed study of Hubei province in central China. Here as elsewhere in the country, the household registration system (huLou, described in Chapter 4) assigns each citizen either an agricultural or a nonagricultural registration status. In the proreform era, a rural migrant could not gain access to urban employment, education, or social services without first obtaining a nonagricultural registration from the local urban government. When economic reforms began to stimulate new economic activities in Chinese cities, resulting in an increasingly diverse employ- ment structure only a part of which fell under government control, the possibility of wage rates alone would indicate that urban informal workers are poorer than rural residents. Once expectations of securing formal-sector employment are factored in, however, the levels of economic well-being of the two groups are seen to be the same. Expectations of upward mobility just compensate for the low level of current earnings among those currently employed.
328 CITIES TRANSFORMED emerged for rural migrants to take urban jobs while continuing to be registered as de jure agricultural residents. These migrants whose urban location differs from their sector of registration are "temporary migrants" in the sense of the phrase used in China. "Permanent migrants" are those whose registration status has been reconciled with their location, as happens with official job transfers and with better-educated migrants who are more likely to persuade local officials of the need to convert their status.27 The consequences of these developments are evident in very large-scale mi- gration flows in China (Yeung, 2001), with credible estimates of the urban "float- ing population" (liudong renkou) of temporary migrants ranging as high as 50 million (Yang and Guo, 1999~. The jobs held by temporary migrants offer long hours; poor working conditions; low and unstable pay; and little access to the in-kind benefits available to longer-term urban residents, such as medical insur- ance, pensions, subsidized transport, child care, and housing (Wang and Zuo, 1999~. Construction work is a main source of employment, along with manufac- turing, nursemaiding, marketing and services, garment work, and scrap collecting (Yeung, 2001, citing several sources). Temporary migrants are somewhat hand- icapped in securing better jobs because they lack the social networks that help longer-term residents gain access to information and obtain recommendations. Temporary migrants in China are predominantly men about three-quarters of the total, to judge from the Hubei survey used by Yang and Guo probably because of the risks and stresses of their urban lives and the nature of the avail- able employment. Rural married women are especially unlikely to become tem- porary migrants, whereas marriage is evidently not a strong disincentive for rural men. Temporary migrants tend to have an intermediate level of educational attain- ment, roughly equivalent to middle or junior secondary school. Those with less schooling can find it difficult to negotiate the complexities of urban environments, whereas the better-educated are often able to convert to nonagricultural status, thereby becoming permanent migrants. Young (2001) notes that permanent mi- grants are concentrated mainly in the professional and technical occupations and dominate urban-to-urban migration flows, whereas temporary migrants are much more likely to be engaged in rural-to-urban moves. The sex differences that are so prominent in Chinese temporary migration are not seen across Asia.28 In Thailand, for instance, the pronounced seasonal- ity of agriculture has given rural women the opportunity for off-season migra- tion to Bangkok, where many take short-term jobs in the service sector. Among 27Yeung (2001) gives a detailed account of the recent efforts of local governments to establish in- termediate statuses, lying between the agricultural and nonagricultural categories of the huLou system, that would give rural migrants a degree of legitimacy and access to some urban entitlements, if not the full range. 28We do not attempt here a general review of male-female differences in migration, which vary greatly by country, region, and time period. The interested reader might consult Bilsborrow (1993), who provides a wide-ranging survey of female migration in developing countries with particular atten- tion to conceptual and measurement issues, and Singelmann (1993), who offers a statistical portrait.
THE URBAN ECONOMY TRANSFORMED 329 recent migrants to Bangkok, women outnumber men by the ratio of three to two (Phongpaichit,1993~. Some argue that the export orientation of Thailand, its rela- tively large urban service and tourist sectors, and employer perceptions that young women will work reliably for low wages have much enhanced the demand for female labor (Lim,1993~. Phongpaichit (1993: Table 50) shows that social networks of friends and rel- atives are particularly important sources of employment information for Thai fe- male migrants. Although male migrants also depend on social networks, they are more likely than women to conduct job searches for themselves. Some re- search suggests a gender-specific pattern in establishing migrant social networks: men are said to migrate first, with their contacts helping to establish networks and reduce risks, thereby facilitating subsequent rounds of migration on the part of women (Guest, 1993~. Urban Economic Mobility Surprisingly little research has explored urban economic mobility using data with a longitudinal dimension. In part this is because longitudinal data are generally uncommon. Even where such data have been gathered, however, rural income dynamics and rural risks appear to have attracted more interest. The dearth of urban studies also reflects the heterogeneity of urban employment and the diffi- culties involved in accurately measuring net earnings and household consumption expenditures in any setting. The few urban studies available tend to divide household consumption lev- els into quintiles or similar classes and record the transitions made among these classes over time, with a particular focus on transitions out of poverty (usually the lowest consumption quintile). These longitudinal studies challenge the view that upward mobility is more likely in urban than in rural areas. In both sectors, it appears, there can be considerable flux, and the estimates of transition rates for both are probably contaminated by measurement errors.29 Grootaert and Kanbur (1995) examine annual panel data on urban and rural households in Cote d'Ivoire from 1985 to 1988, a period in which per capita 29In addition to the literature reviewed below, see Gibson (2001) for Papua New Guinea. In Gibson's analysis, urban households exhibit less short-term variability in consumption than do rural households. Urban households may be better positioned to smooth consumption, at least over the short term, thanks to easier access to savings and loan accounts and the availability of medical or life insurance. This interpretation is consistent with the literature on savings rates in rural and urban areas reviewed earlier in this chapter, which suggests higher precautionary savings in rural areas. A rather different approach to measuring mobility, focusing on intergenerational mobility in the acquisition of schooling, is that of Andersen (2001) for Latin America. This analysis, using 18 surveys fielded in 1995-1998, examines the education attained by young adults as a function of the educational attainment of their parents. The percentage left unexplained by parental schooling is taken to be a measure of intergenerational mobility. Using the full national samples, Andersen finds that more-urbanized countries exhibit greater intergenerational mobility in schooling. However, separate regressions applied to urban and rural samples provide no clear evidence of greater intergenerational mobility in the urban areas.
330 CITIES TRANSFORMED consumption was falling for the country as a whole. Among the urban house- holds that were initially poor (but not in extreme poverty), the proportion exiting poverty ranged from 15 to 55 percent, depending on the region. Rural households exhibited very similar out-of-poverty transition rates. Carter (1999) investigates economic mobility during 1993-1998 in the South African province of KwaZulu- Natal. He also finds substantial economic mobility in both the urban and rural parts of the province, but uncovers some evidence that upward mobility is more common in the urban areas (Carter, 1999: Table 2,16~. In a detailed study of Lima, Glewwe and Hall (1998) examine per capita con- sumption expenditures from 1985 to 1990, this being an era of macroeconomic collapse when average incomes fell by 30 percent across Peru. Their study reveals substantial rates of mobility even for households in the lowest consumption quin- tile in 1985. Of these initially poor households, 40 percent had risen to the second or higher quintile by 1990. Conversely, among households initially in the second consumption quintile in 1985 those that might be described as near-poor some 24 percent had descended into the lowest quintile by 1990. Better-educated Peruvian households exhibited higher initial levels of con- sumption (in 1985) and were apparently more likely to sustain improvements in consumption even through the ensuing macroeconomic turmoil.30 Rural migrant households were found to be more vulnerable to consumption declines, as were households with children. Glewwe and Hall show that social network ties among local kin could not provide much protection against consumption declines because all internal transfers dwindled as the Peruvian economy shrank. The small minor- ity of households with international networks of kin and friends tended to be better protected. Two studies of Malaysia by Randolph and Trzcinski (1989) and Trzcinski and Randolph (1991) take a longer-term perspective on mobility, examining in- dividual transitions in relative earnings over the period 1967 to 1976. Using retro- spective rather than panel data and restricting their analyses to men, these authors find that rural-to-urban migration combined with occupational change (presum- ably the common experience among migrants) increased the likelihood of upward mobility in relative terms.3i Among the urban residents, men with greater edu- cation were found to be less likely to experience relative income decline over the period; as in the case of urban Peru, education appears to provide some protection against downward mobility. 30The latter is a bivariate result, which weakens when multivariate controls are introduced. It is disputed by Schady (2002), who was unable to confirm the result in a reanalysis using additional Peruvian datasets. 3iThe focus of this research is squarely on relative mobility. The earnings distribution in 1967 is divided into deciles, as is the 1976 distribution. A man who was in the lowest decline in both years is recorded as having experienced no mobility, even though his absolute income level might have changed considerably. These studies do not analyze urban and rural residents separately, and use retrospective reports of earnings that are of uncertain reliability for the self-employed and farmers.
THE URBAN ECONOMY TRANSFORMED 331 This smattering of longitudinal studies, in which countries undergoing crises and wholesale structural change are overrepresented, may not be a reliable guide to urban economic mobility in less turbulent times. For the broader picture, it is necessary to consider cross-sectional studies in which migrant earnings can be compared with those of urban natives.32 Vijverberg and Zeager (1994), who focus on Tanzania in the early 1970s, provide a particularly interesting example of this body of research. They find that rural-to-urban migrants attract lower wage offers than do urban natives in their first decade of city residence, but are able to secure higher earnings thereafter. One explanation for this crossover focuses on the mi- grants who arrive with the aim of accumulating capital to be used upon their return to the village. Such "target" migrants may well exert greater effort than urban na- tives in their first few years of city residence, much as international migrants are thought to do in their high-income host countries. Once they pass through an ini- tial period of adjustment to city life, the greater work effort expended by migrants is rewarded by higher earnings. Something of this pattern is also seen in a study of the Sudan by Cohen and House (1996), which shows increasing earnings for migrants by length of stay in Khartoum, possibly attributable in this case to the accumulation of firm-specific experience.33 In urban Bolivia, Chiswick, Patrinos, and Hurst (2000) find that rural migrants who have stayed in the city for more than 5 years have signifi- cantly higher earnings than migrants who have arrived more recently, although rural migrants evidently never surpass the earnings of urban natives. (By contrast, urban-to-urban migrants in Bolivia achieve higher earnings than city natives.) Fe- male migrants who cannot speak Spanish mainly older, rural-born women and recent arrivals face earnings penalties, whereas monolingual Spanish speakers (whether male or female) can command an earnings premium. These earnings paths testify to positive selectivity on average, migrants must be an unusually energetic and resourceful group and to the adjustment difficulties faced by many migrants upon arrival. THE INFORMALIZATION OF URBAN LABOR MARKETS As we have seen with regard to migration, the proposition that the urban economy can be divided into formal and informal sectors has attracted an enormous amount of research attention over the past 30 years. Doubt has been cast on a number of 32It has long been recognized that cross-sectional studies of migrant earnings by duration of stay are vulnerable to selectivity bias in that those migrants who have been successful are probably more likely to remain. Although some upward bias is probable, the size of the bias is unknown. 33In bivariate analyses, migrants with 6 or more years of residence in Khartoum were found to earn more than native residents in several sectors: public-sector employment, small but protected private employment, and unprotected private employment. Migrants earned somewhat less than urban natives when employed in large private firms and self-employed. These migrant-nonmigrant comparisons were not pursued in detail in multivariate analyses.
332 CITIES TRANSFORMED assumptions about these sectors, raising the question of whether it is still useful to organize thinking about the urban economy under such broad headings.34 The categories of formal and informal reflect a general belief that urban markets are separable into segments and that barriers of various kinds hinder movement from one segment to the other. If this is so, then urban economic diversity and in- equality can be explained, at least in part, by reference to the barriers that prevent earnings from being equalized.35 A thorough review of the issues is beyond the scope of this chapter, but in what follows, we touch on some of the arguments that link the formal-informal distinction to urban inequalities and speculate about the implications for urban income growth. As can well be imagined, urban jobs and economic enterprises are too var- ied, and the data about them too crude, for two categories of employment to be adequate. In addition, many city dwellers hold multiple jobs, such as when a wage earner in a large firm also operates a small business on the side. To define employment as formal or informal, the International Labour Organization (ILO) has highlighted only three of the many possible dimensions: whether a worker is self-employed; the size of the firm; and coverage by government systems of em- ployment registration, taxation, or regulation. In the ILO definition, the informal sector comprises the self-employed, those who are working in firms with fewer than five employees, workers with no registration, owners of a family business with fewer than five employees, and family members working in a family busi- ness without a specified wage.36 This classification assigns to the formal sector all government employees and both the owners and employees of large firms. 34Initially inspired by Hart's 1973 work in Ghana, the concept of an informal sector gained pop- ularity in the 1970s, thanks to the efforts of the International Labour Office's World Employment Programme and affiliated researchers (for example, Bromley and Gerry, 1979). Turnham, Salome, and Schwartz (1990) and Mead and Morrison (1996) provide critical reviews. 35Not all conceptions of the informal sector emphasize the barriers to mobility. De Soto (1987) views the informal sector as a response to the excessive power of the state and its overly restrictive regulations. The autonomy enjoyed by small firms and entrepreneurs is risked if by growing larger, they expose themselves to taxation and regulation (Tybout, 2000). Some observers view informal- sector growth as being beneficial because they believe it fosters labor market flexibility and adaptability more generally. An alternative view, held by some neo-Marxist functionalists, is that small-scale production provides a reservoir of labor willing to work at near-subsistence levels (see Deble and Hugon, 1982), keeping wage levels low (and profits high) in formal-sector enterprises. Informal-sector growth can also reflect efforts by large enterprises to shed costs by outsourcing and subcontracting some of their operations to unprotected workers. For Lautier (1994), the informal economy is not an economic sector in any conventional sense, but rather constitutes a parallel economy in which virtually any activity might be represented, from urban agriculture to transportation and financial services. Even Hart (1995) now proposes to abandon the "informal economy" label. 36In 1997, the ILO clarified its 1993 definition (based on size and capitalization of firms), defining the informal sector in terms of characteristics of the production units (enterprises) in which the activity takes place, rather than characteristics of the persons involved or their jobs. Accordingly, the popu- lation employed in the informal sector was defined as comprising all persons who are employed in at least one production unit of the informal sector, irrespective of their employment status or whether this employment is their main or a secondary job.
THE URBAN ECONOMY TRANSFORMED 333 A few examples using definitions similar if not identical to that of the ILO- may suffice to indicate the extent of the informal sector.37 Arnaud (1998), review- ing evidence from a number of West African cities, finds it reasonable to classify fully 70 percent of employment as informal. In Kazakhstan, only 27 percent of the country's labor force pays social taxes; even if all of these workers were urban residents, the formal sector thus defined would constitute only 56 percent of the urban labor force (Anderson and Becker, 2001~. For Kyrgyzstan, data on social contributions indicate that just 64 percent of the labor force works in the formal sector (Becker and Paltsev,2001~.38 In urban India, the self-employed accounted for 16 percent of employment in 1987-1988, but by 1993-1994 their share had risen to 34 percent. In Mumbai alone, informal jobs increased from 44 percent of the total in 1971 to 55 percent in 1981 (De skip ande and De shpande, 1991 ~ . In S. ao Paulo, the formal sector's share of employment fell from 54 percent in 1988 to 43 percent in 1998 (Buechler,2000~. Boxes 8.2 and 8.3 provide detailed analysis of the informal sector in Brazil, and Sao Paulo in particular. To many observers, ex- amples such as these indicate that the informal sector's share of urban employment is large and may be on the rise. garners to Mobility? It is one thing to say that the informal sector is large and growing, but quite another to assert that those who work there are separated from the formal sector by cross- sectoral barriers. Surprisingly few studies have explored this distinction with data in which individual workers are followed over time. Maloney (1999), who uses longitudinal data from urban Mexico in the early l990s, uncovers little in his data to support dualistic views. For Mexico, the evidence on employment transitions is more consistent with the operation of "a well-integrated market in which workers search for job opportunities across sectors than with one in which informal work- ers seek permanent status in the formal sector and stay until they retire" (Maloney, 1999: 291~. The data in question were collected over a 15-month period from 1991 to early 1992, a relatively prosperous period for Mexico characterized by very low (mea- sured) unemployment. Selected transition rates for those working at the beginning and end of this period are shown in Table 8- 1. Even over this short time span, high rates of individual mobility are evident, with considerable movement to and from formal-sector salaried employment. (The formal sector comprises all firms with 37No comprehensive assessment of levels and trends in informal employment can be made here. Sethuraman (1981) provides estimates of the share of the urban labor force working in the informal sector for a number of cities in the late 1970s. 38These are upper-bound estimates for Kyrgyzstan because some informal-sector participants main- tain nominal formal-sector job status to retain eligibility for benefits, and many outside the official labor force have informal-sector occupations.
334 CITIES TRANSFORMED BOX 8.2 Rising Informal Employment in Sao Paulo Between 1989 and 1999, according to Buechler (2000), public-sector employment in Sao Paulo shrank from 635,000 to 609,000, and private registered salaried employment also declined, from 3.4 to 2.9 million. The remainder one measure of Sao Paulo's informal labor force grew from 2.4 to 3.7 million, an annual growth rate of over 4 percent. (See Lam and Dunn, 2001, for other measures of informal employment.) In 1988 industry employed 29 percent of workers in the municipality, but by 1998 this share had fallen to only 18 percent. Schiffer (2000) argues that much of the city's manufacturing activity has relocated to other cities, and that Sao Paulo has become much more service oriented. There has been a noticeable "casualization" of work, with an in- crease in small businesses operating at the fringes of the formal economy, engaged in self-employment, part-time work, home work, construction, sweatshop manufacturing, street vending, and other retail trade. Buechler (2000) argues that these sectors have grown as a result of the increasing competition that accompanies economic openness and liberalization in effect, that globalization has led to informalization. While manufacturing has declined, there has been considerable growth in high-end, skill-intensive services. With increases in outsourcing, some skilled and professional jobs formerly listed under manufacturing are now classified under services. Sao Paulo's infor- mal sector has strong links to the formal sector, as is evident in the number of unregistered workers working in large firms and the self-employed who are associated with a single such large firm. As Buechler notes, formal-sector employees may choose to depart for the informal sector in their search for jobs offering greater independence and flexibility (in her qualitative interviews, quits were recorded more often than layoffs). Also, informal-sector growth presumably reflects efforts by employers to evade cost-increasing regulations, taxes, and social payments. six or more workers.) The percentages on the diagonal from upper left to lower right show that formal salaried employment is the most persistent of the statuses, but even these job spells can be short, with some 22 percent of formal job-holders leaving in little more than a year. The likelihood of exiting formal salaried em- ployment declines with education (not shown), suggesting that the better-educated BOX 8.3 Does Urban Population Growth Swell the Informal Sector? Examining evidence from the nine largest Brazilian cities, Lam and Dunn (2001: Figure 6, Table 2) find no correlation between the percentage of urban workers who are unregistered and the growth rate of either the urban labor force or the urban population. Nor do they find a correlation between unemployment rates and rates of city population or labor force growth. Population and labor force growth rates have been falling since the 1980s, while unemployment rates have been rising. The rise in urban unemployment in Brazil must have been due primarily to other factors perhaps macroeconomic shocks and policies. Difficulties in accommodating la- bor force growth are unlikely to have been a primary cause.
THE URBAN ECONOMY TRANSFORMED TABLE 8-1 Employment Transitions in Urban Mexico, 1991-1992 (percentages of workers by employment status in 1992, given status in 1991) 335 1992 Status Self- Informal Contract Formal 1991 Status Unpaid employed Salaried Work Salaried Unpaid 27 16 18 6 13 Self-Employed 1 69 9 5 10 Informal Salaried 3 16 41 7 29 Contract Work 0 15 10 45 26 Formal Salaried 0 6 7 5 78 NOTE: The rows do not sum to 100 percent because we have omitted several statuses: out of the labor force, in school, unemployed, and employed in "other" work than listed above. SOURCE: Maloney (1999: Table 5~. may perceive brighter prospects for advancement within the formal sector than do workers with less education.39 Maloney finds that by their own accounts, about two-thirds of those leaving formal salaried positions for self-employment did so voluntarily. Transitions to self-employment are more likely for prime-age workers, probably because some start-up capital is needed to begin a small business. In Mexico, at least, self- employment is not an entry-level sector for the young.40 These findings echo the views of Gregory (1986), who expresses doubt about dualistic characterizations of urban labor markets in Mexico. Tybout (2000), assembling evidence on rates of formal-sector entry and exit in Chile, Colombia, Morocco, Korea, and Taiwan, also documents high transition rates. Lacking longitudinal data, other researchers have attempted to glean evidence on formal-informal barriers from data on wage rates or earnings by sector of em- ployment. Unfortunately, evidence of persistent wage differences does not nec- essarily constitute evidence of barriers to mobility. Any job is associated with a bundle of nonwage characteristics, and the earnings from a particular type of job would be expected to vary systematically with such characteristics.4i Persistent 39A study of Cote d'Ivoire also shows that formal education is an important determinant of formal-sector employment, with a suggestion of credentialism on the part of formal-sector employers (Grootaert, 1992). 40Similar conclusions were reached by Blau (1986), who studied transitions from employee to self- employed status in Malaysia. He found that among men in their prime working ages, transitions from employee to self-employed are much more likely than transitions from self-employed to employee. Among migrants to urban areas, the likelihood of being self-employed rises with time since migration. It appears that in Malaysia, as elsewhere, a significant number of the self-employed are small-scale entrepreneurs whose operations require a base of start-up capital that must be pieced together over time. 4iApart from average wages or earnings, the relevant job characteristics include other benefits apart from earnings, such as subsidized health care or social security; the variance of earnings; the future
336 CITIES TRANSFORMED wage differences (among comparable workers) could be taken as evidence of barriers to mobility, but might instead reflect sorting according to nonwage char- acteristics. Small-scale entrepreneurs, for instance, could prefer their work to better-paying employment in a large firm because they value the autonomy and independence that comes with being one's own boss. All this renders the inter- pretation of wage differences problematic. Interestingly, in the study of Mexico mentioned above, Maloney (1999) finds that transitions from self-employment and contract work to formal salaried em- ployment bring significant declines in earnings, whereas those leaving formal salaried employment for either self-employment or contract work realize signifi- cant earnings gains. But it is difficult to be sure that these are real losses and gains, because formal-sector jobs in urban Mexico often provide health benefits as part of the compensation package. Transitions to self-employment from other infor- mal work which probably do not entail a loss of benefits also bring substantial increases in earnings.42 Maloney shows that informal salaried-worker earnings are the lowest among the urban groups he studies, although this finding may be attributable in part to implicit training costs and other unmeasured factors. Similarly mixed results appear in other research. In a study of urban Brazil, Telles (1993) finds that self-employed men who are protected by labor legisla- tion do as well as or better than their formal-sector counterparts, but unprotected workers fare worse. For urban Peru, Yamada (1996) compares earnings in the informal sector with those in public and private formal-sector jobs. Informal- sector work provides an income premium relative to the private formal sector, perhaps because it gives a greater return to entrepreneurial abilities. In 1985 the Peruvian public sector still offered an earnings premium, but by 1990 this pre- mium had vanished in the wake of repeated macroeconomic shocks and hyper- inflation. In Indonesia, however, Thomas, Beegle, and Frankenberg (2000) show that in 1997, urban public-sector employment provided a substantial wage pre- mium relative to private-sector employees and that this premium increased when the Asian financial crisis reached Indonesia in 1998. The urban self-employed returns to skills and on-thejob training, which may not be fully reflected in current earnings; the risks and stresses of working conditions; the degree of autonomy and independence that can be exercised by the job-holder; and the likelihood of involuntary job loss. As Maloney (1999) argues, to interpret evidence of earning differentials in terms of market barriers and rigidities, one must have a sense of what these differentials would have been in the absence of such rigidities. In a hypothetical world in which individuals have identical preferences, labor markets are in equilibrium, and earnings are flexible, jobs would provide levels of earnings that would just compen- sate for their nonwage characteristics (Rosen, 1974). Although they would generally offer different wage rates, in equilibrium all types of jobs would provide the same level of utility. This hypothetical situation can serve as a benchmark against which empirical evidence of earnings differentials can be measured. 42Another interpretive difficulty, noted by Lucas (1997: 737), is that data on the earnings of the self-employed may not adequately distinguish net earnings from sales, and may overstate the premium associated with self-employment.
THE URBAN ECONOMY TRANSFORMED 337 had no precrisis earnings advantage over private-sector employees in Indonesia, but made a substantial gain (in relative terms) during the crisis. In summary, the recent evidence on wage differences by sector is far from uniform, revealing much country-to-country variation and offering few general lessons. Most of the studies reviewed by the panel find that the informal sec- tor contains some groups with decidedly lower earnings than those available in formal-sector jobs. But the informal sector also contains what might be termed an "upper tier" of individuals and enterprises whose net earnings appear to exceed those of formal-sector occupations. The formal sector is itself heterogenous and difficult to characterize, with public-sector employment offering a wage premium in some settings but not in others (Lucas, 1997: 762~. Taken as a group, the studies reviewed here lend little support to the notion that persistent earnings differences are the result of barriers to mobility that separate formal from informal work. We do not see convincing evidence that such barriers are insurmountable or that they cause urban earnings to be more unequal than they might otherwise be. To be sure, the literature on this issue remains thin, but the burden of proof appears to have shifted to those who contend that mobility constraints are significant. Definitive conclusions must await longitudinal data on employment transitions, and for the moment such data remain scarce. Increasing Heterogeneity If the developments of the past two decades could be summed up in a sketch, it might read as follows. In the 1980s and l990s, macroeconomic crises and restruc- turing had the effect of curtailing wage-earning employment in many public and private enterprises. International assistance was often extended on terms that re- quired reductions in government expenditures, a commitment to privatization, and the imposition of cost containment measures in state-owned enterprises. Where such programs were implemented, parastatal enterprises and many private firms cut employment; in many cases, recruitment to the civil service all but ceased.43 Some formerly protected, subsidized, traditional industries shrank as governments cut tariff protections and exposed their economies to the world markets. Buenos Aires, for example, lost a third of its industrial jobs from 1974 to 1985, and its unemployment rate rose from 6 percent in 1991 to 18 percent in 1995 (Prevot- Shapira, 2000~. Those who lost wage employment in the formal sector often had little recourse but to take up informal employment. At the same time, however, informal-sector entrepreneurs moved aggressively into fields that had been largely abandoned by the public sector, such as trans- portation. Large industrial enterprises developed subcontracting links to informal units that specialized in business services (Sassen, 1994b), such as small main- tenance and repair workshops and computer services. In China, the towns and small cities of the Pearl River delta became home to numerous small industrial 43McCulloch, Baulch, and Cherel-Robson (2000) give a detailed account for the case of Zambia.
338 CITIES TRANSFORMED units connected by subcontracts to large enterprises in Hong Kong. Where new urban elites emerged, consumer service firms also sprang up to provide them with caterers, couriers, dog walkers, security guards, and home deliverymen, much as in the cities of high-income countries. In short, while the "lower tier" of the urban informal sector may have been the recipient of substantial inflows from the formal sector and elsewhere, with earn- ings thereby deteriorating, other segments of the informal sector may well have been revitalized and linked profitably to domestic and even international partners. If in an earlier era the urban economy could be depicted as divided along the lines of formal and informal sectors, today there is an increasing appreciation of the polarization taking place within the informal sector itself (see Crankshaw and Parnell 2003, for the case of Johannesburg) . These developments are exceedingly difficult to quantify, and no doubt vary greatly among countries and among cities in any given country; nevertheless, they merit discussion. In the view of many observers, the changes sketched above have had the effect of making employment more precarious, even for those who hold formal-sector positions. The stratum of supervisors and skilled employees, which might once have appeared reachable by means of steady advancement within a firm, has be- come unsettled as a result of growing job flexibility and economic reversals. Pres- sures on government payrolls and formal-sector salaries have presented employees with few options but to seek outside supplements to their incomes.44 Patterns of work and residence may be becoming increasingly fluid, with many workers al- ternating between formal and informal activities. Even for skilled workers, fixed- term contracts and the vagaries of international demand have transformed working life from a stable, dependable routine to a ceaseless search for the next contract and connection. In a way, then, even the conditions of employment in the formal economy have become informalized, at least to a degree (Vandemoortele, 1991; Roberts, 1991~. It is an open question whether developments such as these can provide a firm foundation for future urban income growth. Is it really likely that the informal sector, thus roiled and reconstituted, will become sufficiently dynamic to generate economic growth? Except where migration responds to dramatic growth in demand (as in the case of China described earlier), the opportunities available in the lower tier of the informal sector will probably remain quite limited. Urban agriculture often practiced by street food vendors who grow their food locally may have some potential for growth through closer integration with rural areas. The lower tier will probably continue to function as an entry point for some migrants and as a safety net for others, an economic niche where financial considerations are balanced against traditional social obligations (Ellis and Faure, 1995~. The role of crisis 44 In Buenos Aires, the contratos basuras are growing in number, and some 20 percent of job seekers are actually currently employed workers looking for a second job because of a drop in their salaries (Prev8t-Shapira, 2000).
THE URBAN ECONOMY TRANSFORMED 339 regulator and cushion, however, has its limits. It is difficult to imagine that large enterprises are likely to develop or significant skills accumulate within this tier. One recent study of the informal sector in sub-Saharan Africa reports that the most successful informal firms are neither the smallest nor the largest, but those of intermediate size (Faure and Labazee, 2000~. These enterprises rarely evolve from smaller units, except in the cases of civil engineering and transportation. They are typically formed by the transfer of commercial capital and are created by investors with vocational training, such as skilled workers made redundant in the formal sector. These units often have the capacity to grow in their technical sophistication (auto parts and other small urban metallurgy enterprises are developing in Nigeria, as are maniac and cereal milling firms) and rarely employ recent migrants, perhaps because success requires both skills and urban social connections. Even in this upper tier of the informal sector, however, business profits are not necessarily reinvested in the accumulation of capital. Rather than expand the scale of their operations, many successful small entrepreneurs evidently prefer to diversify their activities and reduce risks by investing in trade, transportation, and construction (Collier and Gunning, l999a,b). The continued vitality of the informal sector will depend on access to reliable public services and infrastructure (water, electricity, communications, road net- works), on the quality of urban management, and on the development of comple- mentary business services.45 Infrastructural constraints can be especially frustrat- ing for informal-sector businesses, which typically lack the capital to substitute for poorly delivered public services (for example, by purchasing their own generators to protect against electricity outages). Informal-sector development is also hin- dered by the absence of complementary business services and by the weaknesses of legal systems and governmental oversight (Collier and Gunning, l999a,b). It is not surprising, then, that informal-sector profits tend to be used for purposes of diversification rather than capital accumulation. In summary, our impression is that the informal sector encompasses an in- creasingly diverse set of activities; one can speak of high tiers and low tiers, of protected and unprotected workers, and yet further distinctions among informal enterprises are doubtless warranted. The formal sector is also more heteroge- neous than was recognized in the early theoretical conceptions. In addition, the links between the informal and formal sectors appear to be growing increasingly complex and subtle. According to our reading of the evidence, no insurmountable barriers appear to divide these two sectors. If this view is correct, then the terms "formal" and "informal" are now useful mainly as evocative labels that suggest general char- acteristics, rather than being descriptive of deep-seated market rigidities and du- alisms. The diversity of both formal and informal sectors, as well as the frequency 45But for a cautionary view, see Davis, Kang, and Vincent (2001), who find that poor water supplies do not greatly constrain micro enterprises in two Ugandan towns. These small units have water needs not unlike those of households in volume.
340 CITIES TRANSFORMED of passage between them, means that the root causes of urban inequalities must be sought in the opportunities and constraints that can be found in both sectors. EARNINGS INEQUALITY: CASE STUDIES Ideally, the developments described in the three preceding sections changes in the returns to schooling, the economic mobility experienced by migrants, and the increasing differentiation of both formal and informal urban sectors could be drawn together to form a comprehensive picture of urban economic inequalities. At the moment, that goal lies beyond reach, but the broad outlines of a portrait of inequality can be glimpsed in selected country case studies. We examine the case of Brazil first, and then consider Taiwan and China.46 Brazil It is plausible to suppose that most countries have rich and poor cities and that much of urban inequality has to do with differences in income across cities. In- deed, earlier chapters of this report have documented substantial socioeconomic differences by city size, as seen in levels of public service provision, children's schooling, and fertility, among other indicators. But the demographic surveys we have relied upon are not of sufficient size to allow within-city inequalities to be clearly exposed. To examine both inter- and intracity inequality, Lam and Dunn (2001) ex- ploit a 20-year series of large labor force surveys for Brazil, covering the period 1977 to 1999. Restricting attention to the country's nine largest cities, Lam and Dunn (2001: 15) find that earnings inequality is "primarily generated within cities rather than across cities. Although differences in mean earnings between cities are as large as two to one, these differences make a relatively modest contribu- tion to overall inequality, given the much larger earnings differentials observed within every city." Taking the variance of the logarithm of earnings as a measure of inequality, Lam and Dunn show that over 90 percent of earnings variance is attributable to within-city earnings differences.47 A large component of earnings variance about 45 percent is due, in turn, to variation in the number of years of schooling attained by Brazilians and the 46Many factors other than those discussed here can affect earnings inequality. For instance, Ashen- felter and Oaxaca (1991) report findings from a set of studies examining labor market discrimination in developing countries. Large earnings differentials between men and women that are attributable to discrimination are found in Tanzania, Brazil, and Nicaragua; similarly large differences are found between Africans (blacks) and non-Africans (Asians) in Tanzania, and between scheduled and non- scheduled castes in Delhi. 47Lam and Dunn are unable to adjust nominal earnings for price differences among Brazilian cities, which they acknowledge may be considerable. Such price differences affect the variance of earnings across cities, and leave unclear whether the observed cross-city differentials are nominal or real.
THE URBAN ECONOMY TRANSFORMED 0.65 c) .m ~ 0.55- tn .' . _ c) 0.45- 0.35 - I ~ Northern Cities | ,~ \ | ~ Southern Cities | ~ 6` ON, \ ~ R ~ \ ~ fig , 1980 1985 1990 1995 Year 341 FIGURE 8-1 Residual earnings variance in Brazil's nine largest cities, 1977-1999. NOTE: The figure shows the s2 from earnings regressions that include age and educational attainment. SOURCE: Interpolated from Lam and Dunn (2001~. economic returns to schooling.48 Schooling in Brazil accounts for an unusually large share of earnings inequality by international standards taking about three times the share found in the United States. Interestingly, the variance of earnings in Sao Paulo is the lowest among the nine cities considered by Lam and Dunn, partly because the returns to schooling are slightly lower there. Although schooling can account for a great deal of earnings variance in Brazil, it leaves a considerable portion of that variance unexplained. Over the 20-year pe- riod examined by Lam and Dunn, the unexplained component of earnings variance increased substantially, as can be seen in Figure 8-1. The high inflation of the late 1980s played a role in the rising trend, but that trend must also reflect the increas- ing heterogeneity of Brazil's urban labor markets discussed earlier (see Box 8.2 for Sao Paulo). 48Consider the earnings equation Y = AS + X'~y + c, where Y is the logarithm of earnings, S is schooling attainment in years, X represents other explanatory factors, and ~ is a disturbance term uncorrelated with the S and X variables. The variance of Y can be expressed as Vy=92Vs+2 pCsXy+~'Vxy+~62' where Vy, Vs. and Vx are variances, and Cs,x is a covariance matrix. Lam and Dunn (2001) focus on the two main roles of schooling evident in this expression: the return to schooling coefficient ,B and the variance of schooling Vs. Their regressions are sex-specific, with age being the only X variable included.
342 0.35 - 0.30 - a) ~ . _ ~ 0.25- u) . _ ~ 0.20- u' a) 0.15 0.10 - CITIES TRANSFORMED p Urban Men ~ Urban Women 1980 1985 1990 1995 Year FIGURE 8-2 Residual earnings variance in urban Taiwan, 1978-1998. NOTE: The figure shows the s2 from earnings regressions that include age and educational attainment. SOURCE: Lee and Mason (2001~. T. alwan Using similar techniques, Lee and Mason (2001) examine changes over time in earnings inequality in Taiwan.49 Inequality (again measured by the variance in the logarithm of earnings) has been low there by world standards, and from 1978 to 1988 it declined further before rising in the mid-199Os. Within Taiwan's urban sector alone, earnings inequality showed little change from 1978 until the 1990s, when it began to rise. Lee and Mason attribute some of the increase in inequality to the entry of women into the urban labor force, which increases the spread of ur- ban earnings because women generally earn less than men. However, two related factors partly offset this effect: the variance of women's earnings is smaller than that of men, and the male-female earnings gap has begun to close. Higher returns to schooling in urban labor markets have also increased the variance of earnings, but higher schooling attainment has tended to decrease it. Yet much of the recent increase in urban inequality, at least for Taiwanese men, is due to unmeasured forces that generated markedly higher earnings vari- ances in the 1990s. Figure 8-2 shows the time path of the residual variance in urban earnings for men and women, with these variances estimated net of age and educational attainment for each sex. As can be seen, the residual variances fluctu- ated with a mild downward drift in the years 1979 to 1994, but then rose sharply. 49Lee and Mason are not able to include the self-employed in their analyses.
THE URBAN ECONOMY TRANSFORMED 343 The increase predates the Asian financial crisis years, although the crisis must have had a further unsettling effect. But while earnings variances were rising for men, decidedly less change is evident for women, and nothing akin to the urban increase is apparent in earnings variances in rural areas (not shown). It is tempting to speculate that in Taiwan, increasing urban variances are the product of global- ization and informalization, but no direct evidence confirms this notion as yet. China Khan, Griffin, and Riskin (1999) find evidence of increasing income inequality within China's cities over the period 1988 to 1995, when many of the work-unit and government subsidies previously enjoyed by urban households all but disap- peared. Although the composition of urban incomes changed considerably (cash incomes from employment rose from 44 to 61 percent of household income over this period, while housing and miscellaneous subsidies fell from 39 to 11 percent), inequality also increased within each category of income. The within-category effects were by far the dominant influence on overall urban inequality. Khan, Griffin, and Riskin (1999) also find that measured urban poverty rates increased somewhat over the period, and speculate that greater increases might well have been recorded if proper account could have been taken of urban price changes and the incomes of the enormous "floating population" of migrants, which are not covered in the available data. Using a sample of large Chinese cities that administer adjacent rural counties, Wei and Wu (2001) show how urban economic growth can spill over to bene- fit such rural areas, thereby reducing urban/rural inequalities. The main focus of this analysis is on the international trade orientation of Chinese cities in the period 1988-1993. To measure inequality between cities and their rural surround- ings, Wei and Wu construct ratios of urban and rural GDP per capita, using city- specific data and including only the rural areas that fall under a city's control. For cities that grew more open with respect to trade those for which the value of city exports relative to city GDP rose inequality between the city and its rural counties fell. The results are robust to several changes in empirical specification and are not limited to the coastal cities alone. They hint at the increasing impor- tance of international linkages to the character of internal, national-level economic relationships a topic explored in the next section. THE FUTURE OF URBAN LABOR MARKETS: GLOBAL LINKS AND LOCAL OUTCOMES The increasing speed and depth of international communication and economic exchange will surely have profound consequences for the cities of poor coun- tries. But these countries differ greatly in the extent to which their economies are exposed to such forces: a few are rapidly dismantling barriers to trade and in- vesting heavily in communications infrastructure, others are proceeding at a more
344 CITIES TRANSFORMED cautious pace, and some appear to be decidedly reluctant participants in global networks of exchange. Historical legacies, sunk costs, and policy stances vary so greatly among poor countries that no general pronouncements about globalization and its urban consequences are really possible. The spatial implications are especially complex. Even a partial list suggests just how many contingencies need to be taken into account in thinking about the local spatial expressions of international linkages. Reductions in trade barriers give local firms access to larger markets and enable them to exploit economies of scale. Yet when firms that were formerly protected make their international debuts, the less efficient among them will likely be driven back to a purely domes- tic role. When firms in high- and low-income countries are linked in customer- supplier relationships, this link can provide a conduit for the transfer of technology and economic learning, whether about production processes, markets, or commer- cial strategies. But technology transfer is probably less broadly beneficial when it involves local firms that extract and export raw materials as compared with firms that export manufactured goods. In some countries, manufactured exports are produced in enclaves with easy access to raw materials (or sited across national borders from high-income consumers), whereas the firms that were once shel- tered by import substitution policies are found in the large metropolises (Becker and Morrison, 1999: 1719~. The returns to having superior transport and com- munications infrastructure are likely to favor the large cities of most countries, at least until investments can be undertaken in their smaller cities. But if this obser- vation would appear to suggest a measured and orderly spatial progression, con- sider that in some cases notably in China's Pearl River delta massive programs of infrastructural investment as part of the development of the greater Hong Kong metropolitan region have been put in place with surprising speed, enabling smaller cities to exploit specialized niches and linkages in the international economy. In what follows, then, we attempt no general predictions. Rather, we high- light a few of the developments associated with globalization that appear likely to impinge on urban demography. We first consider how foreign direct invest- ment can shape local economic space. Next, we examine an emerging sector of high-end business services a notably dynamic element in some large cities with effects that ripple across their populations. Finally, we explore what is known about the risks and economic volatility to which urban populations are exposed as they engage more fully with the international economy. Foreign Direct Investment As Douglass (1997) argues, the influence of international linkages on city growth is often overlooked or summarized too glibly in indicators such as ratios of exports to GDP. He writes (p. 12~: Missing from the analysis have been a wide array of equally powerful international relations formed through direct investment in production
THE URBAN ECONOMY TRANSFORMED 345 BOX 8.4 Foreign Direct Investment: Shaping Indonesia's Cities In Indonesia, the period from the 1970s to the l990s saw foreign direct investment (FDI) emerge as a powerful force, which encouraged the growth of a network of large cities but generally neglected rural areas and smaller cities (Douglass, 19974. In the 1970s about three-quarters of total foreign investment went to two sectors: natural resource extraction (mining and forestry) and resource processing (the chemicals, minerals, and metals manufacturing sectors); and textiles, which accounted for about one- quarter of all foreign investment. Investments in manufacturing were concentrated mainly in Jakarta and the surrounding province of West Java. Key port cities along Java's northern coast also attracted investment, as did cities in Sumatra and Kalimantan from which oil, timber, and some cash crops were shipped. FDI was substantially more spatially concen- trated, according to Douglass, than were domestic firms. In particular, foreign investment neglected the densely settled rural areas of Java, Bali, and eastern Indonesia, which lacked significant natural resources. By 1985, no less than 76 percent of manufacturing employ- ment was found in Java, and of this, some 86 percent was urban based (Douglass, 1997: 1274. The consequences were seen in rapid outmigration from rural Java to Jakarta and the cities on Java's northern coast. The early 1980s saw steep declines in commodity prices, including oil, accompanied by a dwindling of FDI in the outer islands. Indonesia began to reconfigure its develop- ment strategies to emphasize export-oriented manufacturing, an activity in which its newly literate labor force might prove attractive to foreign investors. The country soon became a favored site for Japanese investment. By the latter part of the 1980s, Japan was pro- viding over a third of all foreign investment in Indonesia, and FDI was also flowing in even greater quantities, in total, from Taiwan, Korea, Hong Kong, and Singapore. But the export-oriented manufacturing and foreign investment projects were highly concentrated in spatial terms, being located mainly in Jakarta, Bandung, and Surabaya (Douglass, 1997: 135-6~. More recently, rural dwellers within range of these large cities have been increasingly involved in urban employment, whether as daily and weekly commuters or as short-term residents, thanks to improvements in transport and communication. Such urban connec- tions have clearly improved the incomes of rural families (Douglass, 1997: 1374. However, the costs of urban concentration have been high. It is doubtful that the concentration of In- donesian manufacturing in a few large urban centers can be sustained for long, given the stress it places on the urban natural environment and the need for enormous investments in public services to keep pace with expanding populations. and services, licensing agreements, concessionary contracts for the extraction of natural resources, and finance by transnational enter- prises, their home governments, and international donor agencies. All of these linkages have, in fact, been expanding much more rapidly than international trade, and when seen together they comprise an increasingly formidable set of parameters.... One important strand in this set of relations that of foreign direct investment (FDI) is explored in Box 8.4 for the case of Indonesia. In this instance, a pattern
346 CITIES TRANSFORMED of foreign investment has emerged with clear spatial delineation, unfolding in dis- tinct phases marked by an increase in the urban concentration of investment. We know of no broader treatment than this of the spatial implications of international linkages and capital flows, but as in Indonesia, it is likely that foreign investment often has an urban focus. As Kaminsky, Lyons, and Schmukler (2001: Figures 1 and 2) show, private capital flows to developing countries have become increasingly important and now account for some 80 percent of the total flows, including official aid. These private flows are a mix of FDI and other financial streams that wend through the capital markets, involving bank and trade-related lending together with portfolio flows, the latter of considerable importance to the economies of Latin America and the East Asia Pacific region. As discussed by the World Bank (2002a), private capital flows are cyclical and sensitive to the macroeconomic and financial conditions of high-income countries, as well as to local risk factors. FDI, although less volatile than portfolio flows, is also subject to pronounced cyclical swings. The peak in FDI to developing countries was reached in 1999, and just 2 years later, invest- ments had fallen by 8 percent from that peak (World Bank, 2002a: 37~. Recent projections indicate that strong FDI growth will probably resume in East Asia and in some countries of other regions. For the foreseeable future, however, only a few developing countries can expect to receive any substantial foreign investment. At present, FDI is highly concentrated in only a handful of very large countries: Brazil, China, and Mexico alone absorb about half of all FDI flows to developing countries (World Bank, 2002a: 39~. For the countries that can attract them, such investment flows provide important additional capital, but they also present additional risks of exposure to the economic volatilities transmitted through world financial networks (De Gregorio and Valdes,2001~. Globally Linked Business Services Many corporate firms in high-income countries (and some of their counterparts in developing countries) have penetrated multiple regional and international mar- kets, operating directly or through affiliates. International firms operating in local markets often require services that are tailored to these markets and purchased from local suppliers. The globalization of finance; the growth of transnational investment; the spatial dispersal of factories, service outlets, and offices; and the creation of facilitating information technology all have contributed to a demand for specialized business services (Roberts, 1994; Hampton, 1996; Bagchi-Sen and Sen, 1997; Bowe, 1998; McKee, Garner, and McKee, 2000~. As discussed in Chapter 2, high-end business services can be viewed as com- plex bundles of inputs. These bundles may involve contributions from several types of specialized professions, and where speed is of the essence in production, only the communications infrastructure and diverse economies of large cities are
THE URBAN ECONOMY TRANSFORMED 347 BOX 8.5 The Internationalization of the Informal Economy Information technologies, adopted by some migratory networks and diasporas, allow inter- nationalization of the informal economy. A striking example is provided by the networks of Mouride Senegalese migrants, who use Internet connections in Dakar to manage trading in money and goods among New York, Dakar, and Touba (Tall, 20004. The internationalization of informal sectors is also tied to growth in transnational or- ganized crime. Prostitution, child trafficking, smuggling, and arms dealing are now glob- alized phenomena. The production and distribution of illegal drugs have important con- sequences for urban economies through the creation of jobs and the consumer demands of those directly employed. Drug trafficking is probably the principal export of Colombia and Myanmar, where the resources thus generated are invested in real estate and large- scale industrial enterprises. In Colombia, a system of "Mafia-providence" (Le Bonniec and Sauloy, 1992) transforms crime and corruption into factors of economic moderniza- tion. In Lagos and Cape Town, trafficking in narcotics provides incomes far in excess of normal formal-sector salaries, and finances enterprises in the trade and transport sectors (Fottorino, 19914. likely to be able to provide the requisite input mix.50 Once the services have become routinized, however, some elements in their production can be relocated outside the large cities to take advantage of lower rents and congestion costs (Du- ranton and Puga, 2001~. It is the nonroutine, innovative, and customized aspect of business services that tends to keep them concentrated in the larger cities. In many major cities, then, a new economic core of corporate service activities is being formed, sometimes displacing older manufacturing and administrative ac- tivities. In some cases, the scale, power, and profitability of the new corporate core activities are so visible as to suggest the emergence of a new urban economy. To be sure, even in the most dynamic of cities, the corporate services sector is but one part of the larger city economy. Cities with a core of international business func- tions retain much of their former economic character and still have a great deal in common with cities lacking that core. But as international networks strengthen, the growth paths of internationally linked cities appear likely to diverge from those of other cities. (See Box 8.5.) Although less clearly apparent in the smaller cities, the emergence of an in- ternationally linked, modern services sector has become evident in those cities as well. Furthermore, regionally focused firms share many of the characteristics of firms with more widely dispersed operations and contacts. Although the regional firms need not negotiate all the complexities of international borders and country regulations, they, too, have regionally dispersed networks of operations that re- quire centralized control and servicing. Hence, regional firms have also fed the demand for business services in their corporate headquarters and regional offices. 50See Black and Henderson (1999a) for the United States.
348 CITIES TRANSFORMED The transformation of urban real estate is one vivid indicator of the emerging services sector. Housing is being constructed and remodeled to the tastes of new high-income professionals; firms selling high-profit goods and services are out- bidding lower-profit firms for desirable commercial space. Neighborhood shops tailored to local needs are being replaced by upscale boutiques and restaurants catering to high-income urban elites. The political economy of these developments is no less important. A growing share of a city's total payroll and tax revenues can be accounted for by the new specialized service core. The high-end business service sector is an increasingly important factor in shaping urban economic diversity and inequality. Exposure to Risk from World Markets Because many large cities act as the interface between their national economies and the international markets, their populations are directly exposed to the vari- abilities of these markets. To be sure, rural populations can also be profoundly affected by international developments, such as changes in the world prices of primary products. Rural economies, being less diverse than urban economies, are arguably more vulnerable to certain types of shocks. Cities possess financial and some social welfare institutions that can mitigate risks and help smooth con- sumption (Gibson, 2001~. In general, however, the spatial concentration of city populations must enhance the local economic multipliers that transmit economic downturns (and upturns) from one subgroup to another. Urban economies are more monetized than rural economies, and rural strategies for weathering a crisis (e.g., growing food for own consumption) are not available to many of the urban poor. Also, the construction and real estate sectors notorious for their boom- and-bust cycles and sensitivities to the state of capital markets are far more im- portant to city than rural economies, as are the inventories held by manufacturing firms. City economies would appear to contain some especially volatile sectors that might raise risks for urban residents overall. As countries negotiate their way toward liberalization and a deeper engage- ment with world markets, they adopt adjustment policies that mediate the effects of international markets and cause these effects to impinge differently on rural and urban populations. A number of African countries, for instance, have eased marketing restrictions and lifted price ceilings on agricultural goods, thereby re- moving much of the bias against agriculture that was so prominent in the 1970s (Collier and Gunning, l999b). But even as these reforms work to the benefit of some rural residents, they can hurt urban residents who rely on subsidized food and other goods. In many of the recent economic crises seen across the develop- ing world, international market forces and domestic policies are mingled in a way that makes it very difficult to anticipate the urban consequences. The complexities are well illustrated in the case of Zambia(McCulloch, Baulch, and Cherel-Robson, 2000), where developments in international markets combined
THE URBAN ECONOMY TRANSFORMED 349 with a host of policy reforms to generate rising rates of urban poverty. The collapse of world copper prices long that country's largest and most valuable export in the late 1970s eventually led Zambia to adopt a series of structural adjustment programs beginning in the mid-1980s, which included the removal of ceilings on the prices of maize and other agricultural commodities and major re- forms that improved agricultural marketing. These long-overdue reforms clearly benefited rural residents. For urban consumers, however, the removal of subsidies on maize meal sharply increased the cost of living and sparked rioting in Lusaka and other major towns of the Copperbelt. On the international front, Zambia lifted many foreign exchange controls in the early 1990s, removed licensing and quan- tity restrictions on exporting and importing, and reduced and greatly simplified tariffs on imports. These new policies were intended to lay the foundation for future growth in exports and employment. But in the near term, as McCulloch, Baulch, and Cherel-Robson (2000: 11) write: . . . the collapse of the manufacturing sector has been dramatic. Com- panies operating behind high levels of protection have been unable to withstand the simultaneous shocks of trade liberalization and the re- moval of subsidized credit. Employment in formal manufacturing fell over 40 percent . . . the textile industry has almost collapsed. Urban poverty rates in Zambia (calculated from three surveys with data on house- hold consumption expenditures) rose from 47 percent in 1991 to 63 percent in 1998; meanwhile, rural rates of poverty fell from 88 to 77 percent (McCulloch, Baulch, and Cherel-Robson, 2000: Table 3b). The experience of Indonesia during the Asian financial crisis also shows how urban and rural populations can be affected differently by international develop- ments. The Indonesian phase of the crisis is usually dated to the collapse of the rupiah in January 1998, after which there ensued a period of rapid inflation, tight- ening credit, and government control over lending that paralyzed banking and nearly brought the construction sector to a halt. Levinsohn, Berry, and Friedman (1999) studied the price effects associated with this crisis and found that contrary to what is often thought, the very poor do not occupy niches in the urban economy that are somehow sheltered from international economic shocks. Indeed, increases in the cost of living during the crisis were greatest for the urban poor, somewhat smaller for the urban nonpoor, and smaller still for rural households. The impacts on employment, wages, and earnings are closely examined by Thomas, Beegle, and Frankenberg (2000) and Smith, Thomas, Frankenberg, Beegle, and Teruel (2000~. Over the decade leading up to the Asian financial cri- sis, substantial wage growth benefited Indonesian men and women. As the crisis unfolded through 1998, employment rates that were already high remained so. The impact of the crisis was not seen in employment as such, but rather in wage rates, which fell by some 40 percent in urban and rural areas. This spectacular
350 CITIES TRANSFORMED decline in wages all but erased the gains made by men over the preceding decade and nearly erased the gains for women as well. The urban declines in wages were slightly larger than the rural declines, and of course in cities, a much higher fraction of workers earn wages (Smith, Thomas, Frankenberg, Beegle, and Teruel, 2000~. Wages fell for almost all urban groups, although the percentage declines were evidently larger for low-skilled wage earners, especially among men without government jobs (Thomas, Beegle, and Frankenberg, 2000~. For the urban self-employed, hourly earnings fell about as much as wages, on average, although again upper and lower tiers of self- employment can be distinguished, with greater declines occurring in the latter (Thomas, Beegle, and Frankenberg,2000~. In rural Indonesia, the self-employed (mainly farmers) appear to have been left largely unaffected by the Asian financial crisis, and rural family incomes tended to fall less than wage rates, probably be- cause rural households could activate multiple coping strategies. Urban residents could not protect themselves to the same degree, although both Smith, Thomas, Frankenberg, Beegle, and Teruel (2000) and McGee and Firman (2000) detect an increase in urban agricultural work, which might be interpreted in terms of coping strategies.5i Frankenberg, Beegle, Thomas, and Suriastini (1999) ask whether Indonesian children were more likely to leave school in the peak year of the crisis (1998) as compared with the previous year. They find marked increases in school dropout rates in both the urban and rural areas of Indonesia, with the increases being rela- tively small at the primary level of schooling but quite substantial at the secondary level. Urban dropout rates rose more than did rural rates, and although increases were recorded across the economic spectrum, dropout rates rose much more for poor families. Curiously, however, the negative effects of the crisis on children's education were not seen in their health. Frankenberg, Beegle, Thomas, and Suri- astini (1999) show that although use of public and preventive health services de- clined in the crisis, no deterioration in children's health was apparent. Children's height for age and weight for height showed little change from 1997 to 1998, and measures of anemia actually exhibited some improvement. The possibility of such mixed responses to macroeconomic crisis is empha- sized by Schady (2002), who examines changes in urban children's school en- rollment in Peru from 1988 to 1992. Peru's crisis was as severe as that faced by Indonesia and might have been expected to produce the same sort of reac- tions. Yet Schady finds little change in enrollment rates in Lima and other Peru- vian cities. Perhaps, as he argues, macroeconomic crises lower the opportunity 5iSumarto, Wetterberg, and Pritchett (1999) present findings from a large-scale qualitative survey conducted in each of Indonesia's 4025 subdistricts (kecamatan). The data drawn from interviews with local officials in each subdistrict suggest that households in the middle range of income re- sponded to the crisis by working more, reducing consumption, drawing down savings, and selling assets. In the view of these local officials, the poor households in their districts had far fewer options available to them.
THE URBAN ECONOMY TRANSFORMED 351 BOX 8.6 Cross-Border Migration Flows Between Kazakhstan and Russia Employing unusually detailed time-series data from Kazal~stan, Musabek, Becker, Seiten- ova, and Urzhumova (2001) show how migration is influenced by economic conditions there and in neighboring Russia. Monthly data are available on aggregate flows from 1995 to 1999. The migration flows are substantial: on net, 13 percent of Kazal~stan's urban population (and 9 percent of its rural population) emigrated between 1990 and 1999, with by far the largest outflow going to Russia. Urban Kazal~stanis have higher propensities to migrate to Russia and are far more sensitive to wages and exchange rate movements than are their rural counterparts. This is especially true for young adults (aged 18-294. Migration responds to changes in relative Kazal~stani/Russian mean wages with a lag of about 7 months and an elasticity estimated to be near unity on average. Migration is also highly responsive to changes in relative capital investments per worker and to movements in exchange rates. costs of schooling by reducing the earnings from children's employment.52 The marginal benefits of schooling can also be affected if the crisis is expected to be long-lasting. Indeed, when earnings decline more for those with few skills and low education, it is conceivable that a crisis may even raise the perceived returns to schooling. Few studies have explored the spatial implications of economic crisis in any depth by asking whether certain types or sizes of cities appear more vulnerable. In the formerly socialist countries of West Asia, the prolonged crises associated with economic restructuring appear to have hit the secondary cities harder than the capitals. Bishkek, the capital of Kyrgyzstan, experienced severe difficulties in the first few years of the post-Soviet era, but as its economy steadied, it began to attract migrants from the secondary cities, where local industry had utterly collapsed with dim prospects for recovery. Between 1989 and 2000, Bishkek added about 130,000 people, while the secondary cities of Kyrgyzstan lost 55,000 residents (Government of the Kyrgyz Republic, 2000~. Levels of poverty in Bishkek have been low in comparison with rural rates and those of other cities, and its wage rates have been relatively high even when adjusted for differences in human capital (Anderson and Becker, 2001~. If anything, the gaps in living standards between Bishkek and the remainder of Kyrgyzstan may well have widened over the 1990s. In many ways, then, the economic crisis and adjustments in Kyrgyzstan appear to have reestablished the favored position of the capital city that was so characteristic of the Soviet era. In Kazakhstan, both urban and rural areas have lost population, and there are now substantial migratory flows both to and from Russia (see Box 8.6~. The total urban population declined from 9.8 million in 1992 to only 8.4 million in 1998 52This argument requires a narrow definition of opportunity costs: when adult earnings also decline, children's earnings that are lower in money terms can have a large effect in terms of household well- being.
352 CITIES TRANSFORMED (Government of Kazakhstan 1999, and unpublished data). But the capitals of Almaty and Astana continued to grow, at least during the second half of the 1990s, and again it was the secondary cities that lost population. Health recovery appears to have been stronger in the country's large cities (Becker and Urzhumova, 2001~. In both Kyrgyzstan and Kazakhstan, capital-city residents have generally fared better than those in secondary cities during a very difficult period of transition. CONCLUSIONS AND RECOMMENDATIONS Conclusions The urban demographic bonus is both a boon and a challenge for developing- country cities. As population projections make clear, the urban economies of low- and middle-income countries can expect a very large increase in the sup- ply of labor over the next several decades. In addition, as fertility declines, child dependency ratios will decline, and a larger proportion of the population will tem- porarily move into the working ages (Bongaarts, 2002~. In some Asian settings, declining fertility has also been credited as a major contributor to sustained eco- nomic growth (Mason, Merrick, and Shaw, 1999; Bloom end Williamson, 1998~. Slower growth in the number of school-age children has enabled greater educa- tional investment per child, and lower dependency ratios have produced higher national savings rates and reduced temporarily the need for certain types of pub- lic expenditures. In this sense, an extraordinarily large cohort of working-age adults produces a one-time demographic bonus. But for countries to realize their demographic bonus, their economies must be strong enough to absorb the grow- ing workforce without experiencing increased unemployment or depressed wages. Rapid economic and political changes increase the complexity of the challenge. Urban labor markets and returns to schooling in developing countries are dependent not only on labor supply, but also on technological change and capi- tal formation. This chapter has examined several key elements in the nature and magnitude of the structure of labor demand and supply in low- and middle-income countries. Urban labor markets determine individual and family incomes, which establish incentives for migration and profoundly influence decisions about in- vestments in children's schooling and family size. A popular view is that urban labor markets in developing countries are riddled with imperfections and that unemployment and poverty are the direct result of labor market failure. Yet our review has highlighted the considerable uncertainty that surrounds the functioning of urban labor markets in these settings. Urban economies have changed greatly since the formulation of the highly influential models of Todaro (1969) and Harris and Todaro (1970~. Although rapid growth in the supply of labor might appear to threaten the economic returns to school- ing, our analysis shows that where macroeconomic growth has been moderate or strong, the urban returns to schooling have generally been maintained, and in
THE URBAN ECONOMY TRANSFORMED 353 some cases returns to university education have increased. On the other hand, where macroeconomic growth has been weak, as in much of sub-Saharan Africa, rapid increases in the supply of better-educated urban labor have resulted in a marked decline in the returns to schooling. Evidently, accumulation of physical capital and technological progress can sustain educational returns even in the face of rapid shifts in labor supply, but if these are lacking, returns to schooling will be threatened. Although there is substantialfftuctuation in individual incomes in urban areas, rural-to-urban migrants are able to attain earnings comparable to those of native urbanites after an adjustment period. Surprisingly little attention has been paid to individual income dynamics within urban areas. The few longitudinal studies available do not clearly reveal higher rates of upward or downward mobility in cities as compared with rural areas; in both there is a great deal of flux. However, studies of migrants based mainly on cross-sectional surveys that may overrep- resent the more successful migrants generally show that rural migrants undergo a period of adjustment to city life during which their earnings are low, but sub- sequently achieve earnings levels that rival and sometimes exceed those of urban natives. Recommendations Given the partial nature of this review, it is impossible to formulate a list of specific research recommendations on the relationship between population and labor force growth and labor absorption. Space constraints have forced us to ignore many critical issues, such as the role of unions, discrimination, and occupational segre- gation in the demand for and supply of labor. We have also chosen to overlook issues relating to hours worked, retirement decisions, job attachment, work effort, and the nature of contracts. And finally, we have not discussed the technological advances that might raise the marginal productivity of labor in agriculture. Never- theless, our review has uncovered two central research themes that will doubtless play a large part of a future research agenda in this area. Cities and city-regions Our review has underscored the need to place urban de- mographic issues within an urban economic context. However, we have paid little attention to the economy and demography of the region immedi- ately surrounding cities and metropolitan regions what we might call the urban-regional economy. Greater attention to this subject in future demo- graphic research is important because these surrounding regions are inte- grally related to what is happening to cities and metropolitan regions. Globalization andinequality There has been much recent speculation about whether the globalization of economic relationships heightens urban in- equalities. Case studies of Brazil, Taiwan, and China provide some evi- dence of a growing dispersion of urban incomes in the 1990s. Although
354 CITIES TRANSFORMED the sources of the trends are not clearly identifiable, such empirical find- ings are consistent with the view that urban labor markets are increasingly heterogeneous and volatile, in part because of their exposure to world mar- kets. An examination of the effects on urban populations of international economic shocks and crises shows that city dwellers can be disproportion- ately affected; the evidence from Indonesia is especially clear on this point. The spatial effects are not always focused on cities, however, and urban residents also draw considerable benefits from exposure to world markets. Nevertheless, as globalization proceeds, more research will be needed on both the costs and benefits to urban residents of their increased exposure to world markets.