Urban Segregation and Governance in the Americas
This page intentionally left blank
Urban Segregation and Governanc...
60 downloads
1145 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Urban Segregation and Governance in the Americas
This page intentionally left blank
Urban Segregation and Governance in the Americas Edited by Bryan R. Roberts and Robert H. Wilson
URBAN SEGREGATION AND GOVERNANCE IN THE AMERICAS
Copyright © Bryan R. Roberts and Robert H. Wilson, 2009. All rights reserved. First published in 2009 by PALGRAVE MACMILLAN® in the United States—a division of St. Martin’s Press LLC, 175 Fifth Avenue, New York, NY 10010. Where this book is distributed in the UK, Europe and the rest of the world, this is by Palgrave Macmillan, a division of Macmillan Publishers Limited, registered in England, company number 785998, of Houndmills, Basingstoke, Hampshire RG21 6XS. Palgrave Macmillan is the global academic imprint of the above companies and has companies and representatives throughout the world. Palgrave® and Macmillan® are registered trademarks in the United States, the United Kingdom, Europe and other countries. ISBN-13: 978–0–230–60960–0 ISBN-10: 0–230–60960–0 Library of Congress Cataloging-in-Publication Data Urban segregation and governance in the Americas / edited by Bryan R. Roberts and Robert H. Wilson. p. cm. ISBN 0–230–60960–0 1. Marginality, Social—Latin America. 2. Segregation—Latin America. 3. Discrimination in housing—Latin America. 4. Political stability—Latin America. 5. Marginality, Social—Texas—Austin. 6. Segregation—Texas—Austin. 7. Discrimination in housing—Texas— Austin. I. Roberts, Bryan R., 1939– II. Wilson, Robert Hines. HN110.5.Z9M26 2009 363.5′1—dc22
2008041659
A catalogue record of the book is available from the British Library. Design by Newgen Imaging Systems (P) Ltd., Chennai, India. First edition: May 2009 10 9 8 7 6 5 4 3 2 1 Printed in the United States of America.
Contents
List of Maps, Figures, and Graph
vii
List of Tables
ix
1
2
Residential Segregation and Governance in the Americas: An Overview Bryan R. Roberts and Robert H. Wilson Advances in Research Methods for the Study of Urban Segregation Carolina Flores
Part I
Residential Segregation in Greater Buenos Aires Fernando Groisman and Ana Lourdes Suárez
4
Urban Governance and Intra-Urban Population Differentials in Latin America: A Case Study of Metropolitan Lima, Peru Paul A. Peters
6
7
8
21
The Metropolis
3
5
1
39
55
Residential Segregation in the Mexico City Metropolitan Area, 1990–2000 Andrés Villarreal and Erin R. Hamilton
73
Residential Segregation in Montevideo: Challenges to Educational Equality Ruben Kaztman and Alejandro Retamoso
97
Residential Segregation in Santiago: Scale-Related Effects and Trends, 1992–2002 Francisco Sabatini, Guillermo Wormald, Carlos Sierralta, and Paul A. Peters Residential Segregation in São Paulo: Consequences for Urban Policies Haroldo da Gama Torres and Renata Mirandola Bichir
121
145
vi
CONTENTS
Part II Rapidly Growing, Mid-Size Cities 9
The Process of Cumulative Disadvantage: Concentration of Poverty and the Quality of Public Education in the Metropolitan Region of Campinas José Marcos Pinto da Cunha and Maren Andrea Jiménez
10
Changing Patterns of Residential Segregation in Austin Carolina Flores and Robert H. Wilson
11
Spatial Differentiation, Inequality, and Urban Policy: The Findings Bryan R. Roberts and Robert H. Wilson
169 187
205
Notes on the Contributors
223
Index
227
Maps, Figures, and Graph
Maps 3.1
4.1
5.1a
5.1b
5.2
6.1
7.1 8.1
9.1
9.2
Indicators of spatial autocorrelation: Global and Local Moran’s I without health insurance coverage, 1991 and 2001 Local measures of spatial auto correlation for the Northern Cone of Lima and spatial congruence for Metropolitan Lima, SES and absolute education, 1992 Clusters of census tracts with low (light grey) and high (dark) proportions of adults with middle school education using Local Moran’s I in 1990 Clusters of census tracts with low (light) and high (dark) proportions of adults with middle school education using Local Moran’s I in 2000 Spatial distribution of large manufacturing firms in Mexico City based on the percentage of total workers employed, 1994 Percent of people 25 to 59 years old with years of education below the department mean (2002–2004) and percentage of the population in irregular settlements by Census tracts (2004), Montevideo Geographically weighted regressions at two spatial scales, Santiago, 2002 Local Moran for average years of schooling of the head of the household and presence of shantytowns and illegal settlements. City of São Paulo, 2000 Results for Local Moran’s I for mean years of schooling of household head. Metropolitan region of Campinas, 1991 Results for Local Moran’s I for mean years of schooling of household head. Metropolitan region of Campinas, 2000
51
66
85
86
90
105 138
151
178
179
viii 10.1 10.2
MAPS, FIGURES, AND GRAPH
Local Moran’s I: Adult population with high school or less, 1990 Local Moran’s I: Adult population with high school or less, 2000
197 197
Figures
Residential segregation in Santiago, 2002 (Dimension 2), by socioeconomic group, at different scales (Isolation Index) 8.1 CHAID model for the urban infrastructure indicator
7.1
133 155
Graph 6.1
Percent of children who repeated first grade in 2001, by age of preschool attendance and household risk level
111
Tables
Education variables for segregation indices Alternative spatial definitions of data 3.1 Dissimilarity and isolation indices of segregation 3.2 Determinants of per capita family income (OLS regression) 4.1 Metropolitan-level segregation, 1993 4.2 Regional-level segregation by education and SES levels, 1993 5.1 Indexes of dissimilarity, isolation, clustering, and centralization for the Mexico City metropolitan areas computed at the census–tract Level 5.2 Dissimilarity indexes for two different definitions of the Mexico City metropolitan area computed at the city–block, census–tract, and municipality levels 6.1 Indices of segregation, 62 neighborhoods of Montevideo, 1986–2004 6.2 Percentage of public school enrollment for children ages 4 and 5 years old by educational level of the neighborhood, Montevideo, 1995 and 2004 7.1 Segregation of education measures (0.000) and their changes at multiple levels, 1992 and 2002 7.2 Simple correlations between social problems and segregation, 1992 and 2002 8.1 Residential segregation index (dissimilarity and isolation) by education. Scale of survey areas of São Paulo’s urbanized region, 1991 and 2000 8.2 Coverage of some public services among the poorest people and urban infrastructure access levels according to types of areas (%), São Paulo, 2004 2.1
2.2
31 32 46 52 64 64
82
88 104
111 135 135
149
153
x
TABLES
App 8.1 Summary of urban services access indicator 9.1 Index of dissimilarity and Global Moran’s I by education* and poverty indicators, metropolitan region of Campinas and municipality of Campinas, 1991 and 2000 9.2 Spatial distribution of school infrastructure by Local Moran’s I score, Municipality of Campinas, around 2000 10.1 Measures of segregation, 1990–2000 10.2 Index of average distance to the closest fire station from segregated and non-segregated areas; 1990 and 2000 (Average distance in 1990 100) 11.1 Segregation measures based on educational attainment of adults for eight cities
159
176
181 194
199 208
1
Residential Segregation and Governance in the Americas: An Overview Bryan R. Roberts and Robert H. Wilson
I
n this volume, we present studies of the emerging pattern of socioeconomic segregation in seven major cities of Latin America and, for interregional contrast, one in the United States. Our aim is twofold: One is to contribute to the understanding of contemporary urbanization in Latin America by highlighting contemporary processes of socioeconomic segregation in major Latin American cities through comparable data and measures; the other is to explore the challenges that these emergent patterns of segregation entail for urban government, for equity in access to services, and for the life chances of the urban population. Of the regions of the developing world, Latin America was the first to urbanize rapidly and currently has the highest levels of urbanization at 75.5 percent in 2000. Though the growth of the largest Latin American cities has abated in recent years, these cities still represent some 25 percent of the region’s urban population, and most of that urban population lives in cities of over 100,000. The size and spatial complexity of these cities make socio-spatial segregation an important social policy issue. It was, of course, an important issue in the urban research of the 1960s and 1970s, but three new developments make segregation analysis especially relevant and timely. The first development is the change that occurred in the macroeconomic structure of the region following the end of import-substituting industrialization (ISI) in the 1980s, in general, moving the region’s economies in a neoliberal direction with greater freedom of trade, privatization,
2
ROBERTS AND WILSON
labor market deregulation, and a reduced role of the state in the economy. For cities and their spatial organization, the important implications are a freeing of land markets and greater freedom for the private sector to provide housing and transport and communications infrastructure and to develop large-scale commercial enterprises, such as megastores and shopping malls. The market is now organizing Latin American cities to an extent that has not occurred before. As in the United States, marketdriven housing and commercial development have become key factors in socio-spatial segregation. One consequence, as we will see, is the emergence of new patterns of spatial segregation, such as gated communities. Second is the influence of the political changes of the end of the 1980s and 1990s that have brought both democratic and decentralized forms of government to the countries of Latin America and their cities. Economic and social policy reform in the region has resulted in considerable decentralization of central government functions to lower levels of government, with most basic educational and health services being administered by municipalities. Furthermore, the ideology of reform, heavily promoted by international agencies such as the World Bank, emphasizes greater citizen participation in government through oversight councils in health and education. These reforms make locality important to the quality of social service delivery since much now depends on the competence of local administrators and the material and human resources of their districts. The current emphasis in most Latin American countries on targeted social policies also has the effect, indirectly, of emphasizing locality and its characteristics. Many of these programs, such as the antipoverty program, Oportunidades, in Mexico, have a clear community component, and beneficiaries were initially identified using GIS data employed in segregation analysis. The decentralization emphasis on locality is particularly problematic in terms of the large Latin American cities, whose metropolitan areas are divided into many municipalities and health and school districts. Santiago, for example, has more than 40 municipalities, but they are without an effective overall metropolitan authority. In this situation, socioeconomic differences between municipal jurisdictions can lead to an overall inequity in service provision due to differences both in municipal revenues and in demand resulting from the socioeconomic characteristics of the local population. The fragmentation of many Latin American cities into a multiplicity of jurisdictions is a major challenge for urban planning, and it requires a detailed analysis of socioeconomic segregation and its consequences at the neighborhood and municipal level. Finally, there are the advances in methodology that permit a more detailed and systematic analysis of spatial segregation patterns. The
RESIDENTIAL SEGREGATION AND GOVERNANCE
3
increasing availability of geo-coded censuses enables us for the first time to analyze socioeconomic segregation at a disaggregated level. Previously, analysis of segregation in Latin America relied on neighborhood surveys or on identifying clearly visible markers of segregation, such as squatter settlements or the mansions of the rich.1 In contrast, the analysis of disaggregated census data, as will be done in the chapters that follow, gives us a more accurate and representative view of overall patterns of segregation. Why is residential segregation important? Ultimately this comes down to why locality matters or the issue of neighborhood effects (Sampson, Morenoff, and Gannon-Rowley, 2002). The theoretical assumption is that a range of physical and social characteristics of neighborhoods and the nature of the amenities they contain create advantages or disadvantages for individuals or families that are additional to those arising from individual socioeconomic characteristics. Studies show that the social dimensions of segregation, particularly in terms of the degree of residential stability and social cohesion, affect outcomes such as health, crime, and education. Moreover, there are residential spillover effects with residents benefiting from or being disadvantaged by proximity to advantaged or disadvantaged neighborhoods (Sampson and Raudenbush, 1999). In the sections that follow, we outline the theories that shape the empirical analysis of residential segregation. We will also introduce the case of the United States to highlight the specificity of urban residential segregation in Latin America, which mainly has been based on class in contrast to race and ethnicity in the United States. Considering the U.S. case also provides a preliminary assessment of whether contemporary patterns of urban residential segregation are now converging in the Americas under the impact of global economic forces.
Residential Segregation and the Modern City Contemporary theories of urban spatial differentiation arose in the context of the emerging industrial cities of the United States, in which land was relatively easily available for expansion and city centers were rail and road hubs, which concentrated administration and the most specialized services and commerce. Although these conditions did not apply to established “pre-industrial” cities and needed to be modified by topography and the nature of transport routes, theorists used industrial cities to model spatial organization when market forces had relatively free play. In these early approaches, a key postulate was that competition between businesses and social groups over space led to those with similar needs and resources grouping together. Competition was mediated by social
4
ROBERTS AND WILSON
and economic factors. One important factor was land value, determined by relative distance from the center. Central location, as the most widely accessible location, was estimated to be the most valuable, with the land price gradient declining with increasing distance from the center. A second factor was the relative demand that particular business, government, or residential uses had for central location. Price and need were essentially trade-offs, with preparedness to pay high prices depending both on the relative advantages of central location for different users and on the relative needs that a user had for space. For example, businesses, such as financial or professional services that had a need of centrality because of the specialized services they offered and that occupied relatively little space, were able to pay high prices for central location. More space-intensive uses, such as warehouses, which also needed to be centrally located, would locate on the margins of the central business district. Manufacturing industries with substantial need of space would locate further out, though in the days of rail transport, they, too, would seek a central location. A similar logic was seen as determining residential segregation. The rings of working-class housing around the center represented various trade-offs. The cost of land was still relatively high, but concentrating population in tenements or row housing offset those costs. For the working class, the lack of space and the noise and pollution of central locations were offset by the savings of time and transport in their journeys to work. The middle and upper classes made different types of trade-offs, preparing to pay for space and less pollution and tolerating the time and transport costs of the journey to work. These considerations were formally expressed in Ernest Burgess’s (1925) theory of concentric zones. Subsequent critiques of concentriczone theories modified their application to take account of transport routes, their effect on land value gradients, and of sectoral and multicentered growth patterns (Harris and Ullman, 1945; Hawley, 1950). But the underlying logic of the concentric-zone model remained influential, suggesting that the operation of market forces would segregate cities spatially in terms of the characteristics of localities, such as their housing and land costs, transport, and social identity (Hawley, 1950, pp. 264–75). In terms of residential segregation, the assumptions of these early theories were that land markets operated efficiently and that developers catered to specific residential markets, differentiating the types of housing made available to different income classes. This differentiation depended upon a distribution of income that was uneven, but not as concentrated at the top as to prevent the middle classes aspiring to better housing conditions than the working classes. Thus, an implicit axis of residential segregation was social class as measured by income, although
RESIDENTIAL SEGREGATION AND GOVERNANCE
5
early human ecology analyses paid little attention to the cultural and class factors that differentiated space. These entered the analysis as studies showing a marked urban segregation emerging in U.S. cities based on race and ethnicity. U.S. cities in the early twentieth century were expanding rapidly through immigration. Immigrant groups came in successive waves, each one occupying a part of the city and particular low-end occupational niches, often displacing the previous ethnic groups whose second generation had become economically mobile and sought better accommodation. The Irish, the Italians, and the Central Europeans each occupied distinctive areas of the city. With the mass migrations of African Americans from the South, they, too, occupied distinctive areas of the city. Competition again provided the theoretical dynamic of these patterns of residential segregation. The ecological concepts of invasion and succession posited that like groups would seek to colonize the same urban space and, as they did so, prior groups would move out to colonize new residential spaces. Grouping together of people of similar class or ethnicity resulted from the operation of networks of social support that recruited fellow ethnics to the city, finding them both jobs and housing. It also resulted from the similar cultural and religious needs that fellow ethnics had, which were met by businesses or churches catering to specific ethnic groups. The early ecologists labeled this process symbiosis and commensalism—certain uses flourishing by providing a service that the majority population wanted in common. Gerald Suttles (1968) described the cultural distinctions and interactions that segregated the south side of Chicago into various ethnic communities, whose inhabitants could distinguish the streets and street markers that divided one community from another. Although the preferences and networks of the ethnic groups influenced their pattern of settlement, it was aided and abetted by the massmarketing strategies of real estate interests. Levittown in suburban New York represented the beginning of this phenomenon with some 17,000 mass-produced houses catering to veteran demand after the Second World War. The styles were homogeneous and the first rental contracts excluded non-whites. “White only” suburbs emerged in most U.S. cities in the mid-1950s with minority populations, particularly black ones, formally and informally kept out of new and old suburbs. When a black middle class moved out of the center, it was usually to a homogeneously black suburb. Theories of spatial differentiation, whether economic, cultural, or social, posit that residential segregation in the capitalist world is a “natural” outcome of urban growth in cities. Market factors, the interests of
6
ROBERTS AND WILSON
developers, and the preferences and prejudices of urban populations lead to patterns of spatial segregation that differ between cities in their intensity or ethnic and class contours but emerge everywhere. Moreover, the concentration of wealth or poverty intensifies individual advantage or disadvantage. Spatial segregation creates neighborhoods that differ in the quality of life and the quality of services, such as education and health. From this perspective, neighborhood becomes an important factor in inequality and its reproduction. This natural outcome is affected by state policies, but the action of the state does not modify urban spatial segregation in a consistent direction. State policies can reinforce segregation by building homogeneous blocks of low-income housing or diminish it by providing multi-income housing. While these considerations suggest that large cities are likely to be residentially segregated, the nature and intensity of that segregation will vary according to the middle-range variables that affect residential segregation: topography; the nature of the land market; transport infrastructure; the organization of the construction industry; income distribution within the city; ethnic cleavages; and the extent of state intervention.
U.S.-Latin American Cities Compared Global changes in the economic geography of cities and their industries are affecting patterns of segregation in both Latin American and the United States. The creation of free-trade areas and the generalized reduction of tariffs have eroded the significance of national borders for ordering urban growth and have exposed industries to increasing competition. The advantages previously held by the largest cities have been eroded under the new economic climate in both Latin America and the United States. In the 1970s and 1980s, the U.S. national economy suffered dramatic structural change, with very important regional as well as urban impacts (Wilson, 1993). Large cities lost their comparative advantage in manufacturing, as in the cases of Pittsburgh and Detroit, but many created new specializations in critical service industries such as producer services. By the 1990s in Latin America, the growth of the largest cities had slowed. In most countries, intermediate-size cities grew faster than the very large cities (Portes and Roberts, 2005). This erosion resulted from the increasing diseconomies of scale of the largest cities. These are now consolidated cities in which self-construction is less of an option for the poor and utilities must be paid for. The economies of the large Latin American cities survive and compete internationally, but in terms of highly productive branches of production and services. The largest cities
RESIDENTIAL SEGREGATION AND GOVERNANCE
7
became centers of increasingly complex urban mega-regions or constellations made up of one or more large metropolises, a number of other more specialized industrial, service, or commercial cities and a hinterland of villages and small towns. Examples of this type of urban development are the Mexico City basin (Garza, 1999) and the mega-region surrounding São Paulo, Brazil, that includes Campinas and São Jose dos Campos. The historical differences in the constitution of Latin American and U.S. cities entail, however, that the urban impact of the global economic forces differs between the two regions (Gilbert, 1994; Morse, 1971; Roberts, 1978). Thus, in Latin American cities, land markets have been more constrained by unclear title, communal, or state ownership than they have been in the United States. As we will see, it is only recently with regularization of land titles and privatization that land markets in Latin American cities have begun to operate effectively to differentiate the residential and business uses of space. Even so, land markets in Latin America are still limited by the lack of purchasing power of the mass of the urban population and the absence of home financing (Gilbert, 2002; Ward, de Souza, and Giusti, 2004). The nature of the transport infrastructure is a further variable that contrasts the U.S. and Latin American city. Poor road and rail networks have inhibited middle-class movement to the suburbs in Latin America, leaving them to the poor who traded the low cost and availability of peripheral land and against time-consuming journeys to work. In contrast to the United States, the Latin American pattern has been that of the centralization of higher-income groups and the peripheralization of poverty (Schnore, 1965). The building of new superhighways and rapid transit systems is, however, making peripheral location more attractive for the middle classes in many Latin American cities. Latin American cities have also lacked the large-scale real estate and construction enterprises that built much of the U.S. cities. Latin American cities were built either through self-construction or the customized construction of individual houses. Neither practice led to the rise of the tracts of relatively homogeneous housing that have populated the inner and outer suburbs of U.S. cities. In addition, the income distribution of Latin American cities has been so unequal that there are not the gradients in purchasing power that would lead to a fine differentiation in the quality of housing occupied by different segments of the population. Also, the search for livelihood of much of the urban population has involved a combination of informal settlement and informal economic activity, which in the absence of regulation has reached even into middle-class areas of the cities. The operation of these variables means that urban neighborhoods in Latin America are, in general, likely to be more
8
ROBERTS AND WILSON
heterogeneous in their social composition than ones in the United States (Sabatini, 2003). Elites have increasingly moved out from city centers, and in what Sabatini (2003) describes as a wedge-shaped pattern of relatively continuous settlement. Ethnicity is present in many Latin American cities as a potential basis for segregation, but it is a less powerful discriminator than the ethnic variables in the United States. This is the case even in the large Brazilian cities where there is a relatively clear residential segregation between whites and blacks/nonwhites (Telles, 2004, Table 8.1). This happens in part because the low incomes of the mass of the Brazilian urban population do not give them much control over where and with whom they live (Telles, 2004, p. 208). The growth in the U.S. Hispanic population during the1970s became a new and significant demographic trend affecting segregation. By 2000, the Hispanic population represented 13 percent of the U.S. population. In addition, in the 1990s, the number of foreign-born residents grew by just less than 60 percent (Singer, 2004; Frey, 2006). The spatial distribution of both the Hispanic and foreign-born populations was unlike the traditional pattern, and today significant representation of both groups can be found in most areas of the country, at least in terms of secondtier cities. The national dissimilarity index of Hispanic-white segregation and isolation are somewhat lower than that for blacks, but the index for Hispanics is highest in metropolitan areas with large Hispanic populations (Logan, 2003, pp. 239, 247). In the 1990s, changes in the racial/ethnic composition in the larger metropolitan areas were quite significant (Fasenfest, Booza, and Metzger, 2004). In most large cities and even a few metropolitan areas—including New York, Los Angeles, and Houston—the white population holds a minority share of total population. In 2000, the number of predominately white neighborhoods in the ten largest metropolitan areas decreased by 30 percent and nine of these ten areas witnessed increases in mixed-race neighborhoods, including in some suburban areas (Fasenfest, Booza, Metzger, 2004; Logan, 2003). In the United States, poverty rates vary among racial and ethnic groups (Mishel, Bernstein, Allegretto, 2007). These differences are manifested spatially in cities due to residential segregation. In metropolitan areas, a majority of the poor are white, but poor whites are not concentrated in highpoverty areas as are African Americans and Hispanics (Jargowsky, 2003). Poor whites are not likely to live in neighborhoods with only other poor whites, but rather are relatively dispersed throughout the metropolitan and rural areas. Poor blacks and, to a lesser extent, poor Hispanics, are likely to live with individuals in similar conditions. As the positive effects of antidiscrimination measures allowed movement out of inner-city neighborhoods
RESIDENTIAL SEGREGATION AND GOVERNANCE
9
in the 1980s and 1990s for some, the minority poor remained confined to these neighborhoods (Wilson, 1987). In the 1980s, poverty among minorities became increasingly concentrated and by 1990, 3417 census tracts had poverty rates exceeding 40 percent and contained a total population of 8.4 million (Jargowsky, 2003). In the 1990s, a remarkable reversal occurred, and the number of people living in high-poverty neighborhoods declined by 25 percent (Jargowsky, 2003, p. 2). The steepest declines occurred in the Midwest and South but were broadly observed among all groups, especially in the African American population (Jargowsky, 2003, p. 5). During this period, some revival of inner-city neighborhoods occurred as the result of gentrification, the return of upwardly mobile young adults and middle-class families, and significant in-migration of foreign-born populations (Singer, 2004). Accompanying the relatively slow decline in residential segregation defined by race and ethnicity described above is the fact that large U.S. cities are now much more diverse than 20 years ago (Frey, 2006). The demographic landscape of U.S. cities has changed significantly. The state has been less a factor in shaping the city in Latin America than it has been in the United States, where federal and local governments did not interfere in the operation of land markets. Instead, through zoning and other regulations, they formalized the uses of urban space over time, contributing to the homogeneity of United States urban business and residential neighborhoods. In the Latin American city, the state has been mainly absent from the regulation of space, resulting in the heterogeneous uses of space with commercial, industrial, and residential uses locating in the same neighborhoods. The state also has contributed little to the provision of housing for low-income populations, concentrating instead on working with and legalizing irregular settlements (see Ward 1998, pp. 168–86 for Mexico City; Santos (1981) for Rio; and Portes (1989) for Bogotá). Those government housing programs that did exist tended to favor middle-income rather than low-income groups (Gilbert and Ward, 1985). On both the regulation and the provision of housing, Latin American city and national governments are now becoming more active. These changes in the nature of the metropolis make research on spatial differentiation and its consequences for welfare a priority. This research has a long tradition in the United States and is now beginning in Latin America. Since different areas of a city differ considerably in the socioeconomic composition of their population and, because of decentralization, in the quality of the services they provide, demographic and public policy analysis need to be combined. In the 1990s, access in Latin America to representative urban household surveys and census data disaggregated
10
ROBERTS AND WILSON
to city neighborhood or tract level has facilitated the systematic exploration of urban spatial differentiation and its implications for welfare. An example of this approach was the team research in Uruguay led by Ruben Kaztman (1996; 1999), which used household surveys to identify neighborhood effects on levels of unemployment and other social indicators. Kaztman was able to show that people with similar age and educational characteristics differed in their levels of unemployment, depending on whether they lived in homogeneously poor or more socially heterogeneous neighborhoods. Thus, despite the particularities of Latin America’s urbanization, the bottom line is the same in Latin America as it is in the United States: a need exists to focus on concentrated urban poverty and its deleterious effects.
Decentralization and Urban Policy The spatial structure of a nation provides a unique lens for analyzing social and economic processes. The spatial structures of the large metropolitan areas in the Americas often accentuate the social and demographic disparities found within individual countries. These large metropolises with their distinctive spatial organization have presented new policy challenges, and although innovative responses to these challenges have emerged, the patterns are far from regular. The mapping of governmental structures onto national space creates opportunities as well as limitations for governmental action. But one should expect that as spatial structure changes, due, for example, to urban growth, the nature of demand for public policy and services would evolve, and the existing governmental structure might prove inadequate. In the United States, for example, new forms of urban government emerged during the country’s urbanization process (Burns, 1994). The public sector in many countries in the Americas has become more decentralized in recent decades. Tim Campbell (2003) has described decentralization in Latin America as the quiet revolution of the 1990s. The process of decentralization, usually conceived as the devolution of resources and decision-making authority to the subnational and local tiers of governments, weakens the traditional hierarchical and often rigid structure of power in which, at least in Latin America, the executive branch exercises authority over subordinate levels. Governmental centralization evolved into something that was accepted and even desired in most countries, at least until the middle of the twentieth century. While the specific reasons for and the dynamics of decentralization are unique to individual countries (Gibson, 2004), several structural factors developed toward the second half of the twentieth century that led to
RESIDENTIAL SEGREGATION AND GOVERNANCE
11
the reconsideration of apparent advantages of centralization. Rapid urbanization and, especially, the growth of “secondary” cities contributed to the problems of providing basic infrastructure, full-employment opportunities, and adequate services—all removed from the central (often primate) city. In addition, many countries began to show large regional inequalities because the existing governmental structures were unable to take effective advantage of the country’s resources. In this context, excessive centralization that made decision-making and bureaucratic procedures slower and more rigid was viewed as incompatible with existing social challenges. Decentralization began to appear increasingly as a developmental consensus, something that made sense within an increasingly democratic world order (Smith, 1988). By the early 1980s, decentralization had become associated with economic reform and structural adjustment models (Schuurman, 1998). National governments were perceived as bloated, top heavy, and in need of being “downsized” or “rolled back” (Wilson et al., 2008). Public services needed to be decentralized either to the market and/or nongovernmental (NGO) providers (outward) or to the local level (downward) in order to gain efficiencies and better quality services. In many Latin American countries, the Social and Emergency Investment Funds’ organizations and structures were often adopted as alternatives to governmental delivery systems (Nickson, 1995). At the local level, NGOs become important service providers in their own right or under contract with governments and thus became important actors in social policy. Despite these high expectations, the difficulties of decentralization have become apparent. Issues of inadequate capacity in subnational governments and the difficulties of intergovernmental relations have appeared (Wilson et al., 2008). In many Latin American countries, local governments, lacking adequate financial, human, and institutional resources, have frequently been unprepared to assume the new responsibilities required by decentralization (for Brazil and Mexico, see Wilson et al., 2008). The degree of local government readiness varies substantially from country to country and even within various regions of the same country. But the capacity of local government is a recurring problem, and efforts for administrative reform and the training of individuals to assume positions in local government are critical. Effective local government action does not depend exclusively on its administrative capacity. The intergovernmental dimension of government action has become a prominent issue, especially in the metropolitan areas that encompass multiple local governments. Local governments are linked to higher levels of government in several different ways, including through constitutional and statutory frameworks, fiscal relations, joint responsibilities of program implementation, and politically. In the United
12
ROBERTS AND WILSON
States, for example, even though the 1970s and 1980s witnessed significant decentralization of power from the federal government to state governments, federal regulatory control over states increased significantly (Advisory Commission on Intergovernmental Relations, 1984; Conlan, 1988). The particular circumstances of coordinating governmental action in metropolitan areas with a multitude of municipal governments present great and unresolved challenges. In broad measure, contemporary issues of metropolitan governance have been addressed in an ad hoc fashion. But appropriate structures—a fourth tier of government, confederations of municipal governments, or expanded use of interorganization collaboration—have yet to be broadly adopted (Wilson et al., 2008). The tension between spatial organization of metropolitan areas and the political representation embedded in current jurisdictions suggests that broad changes in metropolitan governance will be difficult. The promise of decentralization, in which relatively rigid centralized systems would devolve powers and cooperate with lower levels of government, has proven difficult to fulfill in practice, particularly in the large metropolises in the Americas. Decentralization holds the potential for improved local government action, but in metropolitan areas local governments operate within a set of complex intergovernmental relations that can constrain, if not impede, their actions. Among these challenges, this study devotes particular attention to those policies that affect urban form. A broad range of cultural and economic forces determines patterns of housing settlements within cities, as discussed above. But in most countries a variety of public policies may also affect settlement patterns and urban form. Public infrastructure investments and land-use policies will obviously affect urban form, but other policies, such as urban services provision and regulation of financial markets related to housing investment, may also have spatial impacts. But settlement patterns, especially the spatial distribution by socioeconomic status or other class cleavages, clearly affect the demand for urban services. For example, in cities where poverty populations are spatially segregated, the demand for public services will be equally affected by space. But the provision of public services may, in fact, reinforce patterns of segregation. Perhaps purposively, the spatial provision of services may reinforce spatial manifestation of socioeconomic disparities. This assertion of mutual determination of urban form and the spatial dimension of urban policy raises a number of important empirical questions. First, how might spatially segregated urban areas affect the provision of services? Residents in a city generate a demand for services, such as education, health and other social services, as well as housing, transportation, police, and so on. In fact, most urban services will not be provided
RESIDENTIAL SEGREGATION AND GOVERNANCE
13
ubiquitously across urban space. Even though some policy problems may be largely citywide in nature, for example, non-point-source pollution (such as automobile emissions), most services must take location into account when designing service delivery systems. Although the mobility of residents means that not all services need be provided in all areas of the city, nevertheless service provision will need to take into account the location of residents (Austin, 1974; Bigman and ReVelle, 1978; DeVerteuil, 2000; Morrill and Symons, 1977; Smith, 1977; White, 1979). For example, lack of coverage in all basic infrastructures, such as water, wastewater, electricity, and so on, has a specific spatial dimension. We also know that concentrations of populations without adequate water and wastewater can create serious public health challenges. Populations dependent on public transportation represent a particular spatial demand that must be met. Similarly, populations with high shares of children create higher demand for education services than populations found elsewhere in the city. To the extent that communities are organized and act politically, the effective demand and supply of services can be affected. Second, how do urban policy and services affect urban settlement patterns? The action of the state with respect to land use regulation and infrastructure provision has a fundamental effect on urban form. These public investments affect property values and are fundamental to urban expansion. As is often the case, the lack of state action in such matters as the provisions for low-income housing can lead to unplanned and seemingly chaotic patterns of growth. The universal presence of squatters’ settlements in Latin American cities represents failures of state action and creates inefficient and inequitable levels of urban services. The spatial analysis of the provision of public services is bounded by at least two features: the type of service itself and its delivery characteristics. The type of good/service provided largely determines the spatial distribution of a public facility. Services that must be provided locally, for instance, are qualitatively different—from the spatial perspective— from those services that can be provided from a distance (Bradshaw and Muller, 2004). Whether to provide locally or at a distance will be mediated by the mobility of the population. Service provision to the homeless represents a particular challenge (Lee, 1993). Teitz (1968) provides some guidelines for distinguishing types of public facilities in terms of their spatial characteristics. First, Teitz adopts the classic differentiation between those goods that can be used collectively with zero short-run marginal costs, such as an air pollution monitoring system, a non-congested highway, a library, and so on, and those goods that actually show congestion. Second, the nature of the service area can be used to classify areas. Thus, a good/service showing a point
14
ROBERTS AND WILSON
pattern (when the final phase of distribution is flexible and intermittent such as with medical centers, post offices, police, and libraries) should be distinguished from those goods and services showing network patterns (services that call for continuous connections in space, such as water, sewer, electric power, gas, telephone, highways). Other important differences relate to the direction of flow—distributors or collectors; relationship with the city—one to one, one to many, or many to many; the hierarchical properties—relationship with other points or other nodes in the network; and whether the individual has the ability to choose, as with libraries and education, or not, as with fire stations and police. In addition one should consider the existence of alternatives—non-public sources such as NGOs, quasi-public, or private entities—and whether these alternatives are free of charge and perfect substitutes. Decision making on the location of public facilities may be entirely in public hands or delegated to the voluntary or private sectors. In each case, the factors determining the optimal spatial distribution of the facilities are different; thus, the effect of the socioeconomic distribution of the population on location decisions will vary from case to case. In particular, while the government and the voluntary sector may want to target the needy population, private providers may focus on serving the population with a higher power of demand. Political issues arise when the public facility is considered noxious. If this is the case, communities tend to oppose the construction of the facility (such as mental health care facilities, landfills, and so on). In addition, the structure of local government can also have an effect on the efficiency of services delivery in metropolitan areas, a point of great relevance to the highly fragmented system of local government in the United States.
Research Questions The literature on residential segregation and urban policy in Latin America, as described above, reflects the evolving nature of cities in terms of their economy, demography, and spatial form. But these changes occur in the context of intervention of the state, whether in terms of neoliberal development policies or actions of local governments. This study attempts to assess changes in socioeconomic segregation and the implications for urban governance. The specific questions to be examined are: Question 1: How has the spatial pattern of socioeconomic segregation changed in metropolitan areas at the end of the twentieth century?
RESIDENTIAL SEGREGATION AND GOVERNANCE
15
Even though population growth, at least in the largest cities, has slowed, we are interested in the spatial organization of the residential population. As discussed above, spatial organization is subject to change despite little demographic growth. This question will be addressed at two levels: (1) assessing the overall level of segregation in the city, and; (2) identifying notable changes in specific areas of the city. The analysis of spatial form has advanced significantly both in terms of the methods and the data availability, as will be elaborated in chapter 2. Calculations of segregation measures will be made at two points in time, through traditional measures of segregation as well as the recently developed global and local measures of segregation. An array of factors, such as housing markets, structural economic change, demographic dynamics, creation of new urban subcenters, and infrastructure investments, will be considered in interpreting the evolution of the spatial form of the metropolitan areas. Question 2: Does the evolving pattern of spatial segregation affect social inequality? A central feature of the Latin American city in the twentieth century was high levels of poverty, as described above. Aggregate measures of poverty and other socioeconomic variables have, indeed, improved in recent decades, but does this improvement result in less social inequality? This question will be examined through the lens of spatial form. In particular, do patterns of segregation reduce or increase inequality? This project investigates this question with a range of methodological approaches using different socioeconomic variables. The flexible research methodology with multiple variables reveals the extraordinary complexity of spatial form and social inequality. Question 3: How is the provision and demand of urban services affected by the changes in urban form and social inequality? As discussed above, urban form can affect the demand for urban services and action of the state can affect urban form. The question to be addressed here is, given the evolving spatial form in metropolitan areas, are new issues for urban policymaking and provision of services arising? This research question will be answered through an interpretative narrative relating the spatial pattern of poverty and the policy issues this pattern generates. In some cases, this narrative can be supplemented by a more formal analysis of demand or outcomes for a service. A range of urban services is considered, acknowledging the fact that delivery systems of different services vary.
16
ROBERTS AND WILSON
Research Design Understanding socioeconomic segregation and issues of urban governance in the large Latin American metropolises is complicated by the demographic, economic, and ecological differences among them. These differences are partly historical, reflecting the period at which Latin American countries began to urbanize rapidly. Difference between early urbanizers, such as Argentina and Uruguay, intermediate urbanizers, such as Chile, then Brazil and Mexico, and later urbanizers, such as Peru, are potentially important. In the more recent period, the very rapid growth of secondtier cities provides yet another factor that can explain variation in spatial form. Furthermore, institutions based on the differences between highly centralized unitary states, such as Chile or Peru, relatively centralized federations, such as Argentina and Mexico, and relatively decentralized federations, such as Brazil, may be important. To capture this variety, case studies were sought in the major metropolises of six countries, Buenos Aires, São Paulo, Santiago, Mexico City, Lima, and Montevideo. The population range of these larger cities is from Greater Montevideo’s 3-million population to Mexico City’s 19 million. Four of the cities, Buenos Aires, Montevideo, Lima, and Santiago, are considered primate, in that each city dominates the urban landscape of the country. The other two cities, São Paulo and Mexico City, are indeed large and powerful cities, but do not dominate the urban network of each country. As noted in an earlier section, all six cities have displayed relatively slow growth in recent decades. To allow for greater differences in urban size and recent population growth, the second-tier cities of Campinas, in Brazil, and Austin, in the United States, are included for study. Thus, the potential effects of differences in economic structure, topography, and demographic change as well as other factors of local context, especially local governance systems, are key variables in explaining potential variation across cities. The project was initiated in the “Workshop on Urban Governance and Intra-urban Population Differentials in Latin American Metropolitan Areas” held at the University of Texas at Austin in November 2002 and funded by a grant from the Mellon Foundation. At the workshop, information on data availability in different country contexts was discussed and a consensus around methods of analysis and key issues for future research identified. In 2004, a team of interdisciplinary researchers, including demographers, sociologists, and planners with expertise in these metropolitan areas was formed, and papers were commissioned, following common guidelines. The group met again at the conference “Spatial Differentiation and Governance in the Americas” in November 2005, at which time papers of individual cities were presented. After another round
RESIDENTIAL SEGREGATION AND GOVERNANCE
17
of revisions based on refinement of variables and methods, the papers of this volume were consolidated. Even though researchers adopted a common set of variables and methods, in the spirit of innovating in spatial analysis, they were encouraged to use additional measures and techniques. Note 1. The “classic” studies available in English of Latin American urbanization in the 1970s and 1980s are mainly studies of the megacities: Rio by Anthony Leeds (1969, 1974) and Janice Perlman (1976); Sao Paulo by Lucio Kowarick (1977; 1980); Mexico City by Susan Eckstein (1977), Peter Ward (1990), Wayne Cornelius (1975), and Larissa Lomnitz (1977); Bogotá by Alan Gilbert and Peter Ward (1985); Lima by Henry Dietz (1980) and David Collier (1976); Santiago by Manuel Castells (1983); Buenos Aires by James Scobie (1971).
References Advisory Commission on Intergovernmental Relations. (1984). Regulatory federalism: Policy, process, impact and reform. Washington, DC: Advisory Commission on Intergovernmental Relations. Austin, C. M. (1974). The evaluation of urban public facility location: An alternative to cost-benefit análisis. Geographical Análisis, 6, 135–45. Bigman, D. and ReVelle, C. (1978). The theory of welfare considerations in public facility location. Geographical Analysis 10, 229–40. Bradshaw, T. and Muller, B. (2004). Shaping policy decisions with spatial analysis. In M. Goodchild and D. Janelle (eds.), Spatially integrated social science, 300–322. Oxford: Oxford University Press. Brent Hall, G. and Peters, P. (2003). Global ideas and local practicalities in education policies and planning in Lima, Peru. Habitat International, 27, 629–51. Burgess, E. W. (1925). The growth of the city. In Robert E. Park, E. Burgess, and R. D. McKenzie (eds.), The city, 47–62. Chicago: Chicago University Press. Castells, M. (1983). The city and the grassroots: A cross-cultural theory of urban social movements. London: Edward Arnold. Conlan, T. (1988). New federalism: Intergovernmental reform from Nixon to Reagan. Washington, DC: Brookings Institution. Cornelius, W. (1975). Politics and the migrant poor in Mexico City. Stanford, CA: Stanford University Press. DeVerteuil, G. (2000). Reconsidering the legacy of urban public facility location theory in human geography. Progress in Human Geography, 24(1), 47–69. Dietz, H. A. (1980). Poverty and problem-solving under military rule: The urban poor in Lima, Peru. Austin: University of Texas Press. Eckstein, S. (1977). The poverty of revolution: The state and the urban poor in Mexico. Princeton, NJ: Princeton University Press.
18
ROBERTS AND WILSON
Eckstein, S. (1990). Urbanization revisited: Inner-city slum of hope and squatter settlement of despair. World Development, 18(2), 168–69. Fasenfest, D., Booza, J., and Metzger, K. (2004). Living together: A new look at racial and ethnic integration in metropolitan neighborhoods. Washington, DC: Center on Urban Metropolitan Policy, Brookings Institution. Frey, W. (2006). Diversity spreads out. Washington, DC: Center on Urban Metropolitan Policy, Brookings Institution. Garza, G. (1999). Global economy, metropolitan dynamics and urban policies in Mexico. Cities, 16(3), 149–70. Gilbert, A. (1994). The Latin American city. New York: Monthly Review Press. ——— (2002). On the mystery of capital and the myths of Hernando de Soto: What difference does legal title make? International Development Planning Review, 24, 1–20. Gilbert, A. and Ward, W. (1985). Housing, the state, and the poor. Cambridge: Cambridge University Press. Harris, C. D. and Ullman, E. L. (1945). The nature of cities. The Annals, 242, 7–17. Hawley, A. (1950). Human ecology: A theory of community structure. New York: Ronald Press. Jargowsky, P. A. (1997). Poverty and place: Ghettos, barrios, and the American city. New York: Russell Sage Foundation. ——— (2003). Stunning progress, hidden problems: the dramatic decline of concentrated poverty in the 1990s. Washington, DC: Center on Urban Metropolitan Policy, Brookings Institution. Kaztman, R. (1997). Marginalidad e integracion social en Uruguay. Revista de la CEPAL, 62, 91–116. ——— (ed.) (1999). Activos y estructuras de oportunidades. Montevideo: CEPAL/ PNUD. Kowarick, L. (1977). The logic of disorder: Capitalist expansion in the metropolitan area of Greater São Paulo. Brighton, UK: Institute of Development Studies. ——— (1980). A espoliação urbana. Rio de Janeiro: Paz e Terra. Lee, J. (1993). Creating effective human service delivery systems for the homeless. Ph.D. Dissertation, University of Southern California. Leeds, A. (1969). The significant variables determining the character of squatter settlements. América Latina, 12, 44–86. ——— (1974). Housing settlement types, arrangements for living, proletarianization and the social structure of the city. In W. Cornelius and F. Trueblood (eds.), Latin American Urban Research 4, 67–99. Beverly Hills, CA and London: Sage. Lewis, O. (1961). The children of Sanchez: Autobiography of a Mexican family. New York: Random House. Logan, J. R. (2003). Ethnic diversity grows, neighborhood integration lags. In B. Katz and R. E. Long (eds.), Redefining urban and suburban America: Evidence from census 2000, 235–56. Washington, DC: Brookings Institution. Lomnitz, L. L. A. (1977). Networks and marginality: Life in a Mexican shantytown. New York: Academic Press.
RESIDENTIAL SEGREGATION AND GOVERNANCE
19
Mishel, L., Bernstein, J., and Allegretto, S. (2005). The state of working America: 2004/2005. Ithaca, NY: Cornell University Press. Morrill, R. and Symons, J. (1977). Efficiency and equito: Aspects of optimum location. Geographical Análisis, 9, 215–25.Morrill, R. L. (1991). On the measure of geographical segregation. Geography Research Forum, 11, 25–36. Morse, R. (1971). Trends and issues in Latin American urban research, 1965–1970. Latin American Research Review, 6(1), 30–52. Morse, R. M. and Hardoy, J. (eds.) (1993). Rethinking the Latin American city. Baltimore, MD: Johns Hopkins University Press. Nickson, R. A. (1995). Local government in Latin America. Boulder, CO: Lynne Rienner. Perlman, J. (1976). The myth of marginality. Berkeley: University of California Press. Portes, A. (1989). Latin American urbanization in the years of the crisis. Latin American Research Review, 24(3), 7–24. Portes, A. and Johns, M. (1989). Class structure and spatial polarization: An assessment of recent urban trends in Latin America. In W. L. Canak (ed.), Lost promises: Debt, austerity, and development in Latin America, 111–37. Boulder, CO: Westview Press. Portes, A. and Roberts, B. (2005). The free market city: Latin American urbanization in the years of the neoliberal experiment. Studies in Comparative National Development, 40(1), 43–82. Roberts, B. (1974). Organizing strangers. Austin: University of Texas Press. ——— (1978). Cities of peasants: The political economy of urbanization in the third world. London: Edward Arnold. ——— (1991). Household coping strategies and urban poverty in a comparative perspective. In M. Gottdiener and C. Pickvance (eds.), Urban life in transition, 135–68. Newbury Park, CA: Sage. Sabatini, F. (2003). La segregación social del espacio en las ciudades de América Latina. Serie Azul, 35. Instituto de Estudios Urbanos y Territoriales, Facultad de Arquitectura, Diseño y Estudios Urbanos, Universidad Pontificia Católica de Chile. Sampson, R. J. and Raudenbush S. W. (1999). Systematic social observation of public spaces: a new look at disorder in urban neighborhoods. American Journal of Sociology, 105(3), 603–51. Sampson, R. J., Morenoff, J., and Gannon-Rowley, T. (2002). Assessing “neighborhood effects”: Social processes and new directions in research. Annual Review of Sociology, 28, 443–78. Santos, C. N. F. dos. (1981). Movimentos urbanos no Rio de Janeiro. Rio de Janeiro: Zahuar. Schnore, L. F. (1965). On the spatial structure of cities in the two Americas. In P. M. Hauser and L. F. Schnore (eds.), The Study of urbanization, 347–98. New York: John Wiley & Sons. Schuurman, Frans J. (1998). The decentralisation discourse: Post-Fordist paradigm or neo-liberal cul-de-sac? In C. Kay (ed.), Globalisation, competitiveness, and human security, 150–66. London: Frank Cass.
20
ROBERTS AND WILSON
Scobie, J. R. (1971). Argentina: A city and a nation. New York: Oxford University Press. Singer, A. (2004). The rise of new immigrant gateways. Washington, DC: Center on Urban and Metropolitan Policy, Brookings Institution. Smith, D. M. (1977). Human geography: A welfare approach. New York: St. Martin’s Press. Suttles, G. (1968). The social order of the slum. Chicago: University of Chicago Press. Szelényi, I. (1983). Urban inequalities under state socialism. New York: Oxford University Press. Teitz, M. B. (1968). Toward a theory of public facility location. Papers of the Regional Science Association, 21, 35–52. Telles, E. E. (2004). Race in another America: The significance of skin color in Brazil. Princeton: Princeton University Press. Ward, P. (1990). Mexico City. Boston, MA: G.K. Hall. ——— (1998). Mexico City. New York: Wiley. Ward, P., de Souza, F., and Giusti, C. (2004, December). Colonia land and housing market performance and the impact of lot title regularization in Texas. Urban Studies, 41(13), 2621–46. White, A. (1979). Accessibility and public facility location. Economic Geography 55, 18–35. Wilson, R. H. (1993). States and the economy: Policymaking and decentralization. New York: Praeger. Wilson, R. H., Ward, P., Spink, P., and Rodríguez, V. (2008). Governance in the Americas: Decentralization, democracy and subnational government in Brazil, Mexico and the United States. Notre Dame, IN: University of Notre Dame Press. Wilson, W. J. (1987). The truly disadvantaged: The inner city, the underclass, and public policy. Chicago: University of Chicago Press.
2
Advances in Research Methods for the Study of Urban Segregation Carolina Flores
T
he analysis of spatial segregation involves several methodological issues. First, there are multiple dimensions of socioeconomic residential segregation and different ways of measuring each. Second, social processes in space will not necessarily correspond to the geographical definitions used in comprehensive data collection, such as national population censuses, thus creating problems of measurement. But the field of spatial statistical analysis and geocoded census data creates new opportunities for addressing these issues and advancing the state of the art. Finally, of particular relevance to this volume, is that population censuses have yet to be standardized across countries in the Americas, as has been the case with the European Union, and, thus, formal, quantitative cross-country comparisons are not yet feasible. In this chapter the methodological frame used in the case studies that follow is presented. This chapter first defines the multiple dimensions of segregation and their measurement. The types of measurement errors common to this type of analysis, including the problem of areal definition and spatial autocorrelation, are explored. Finally, the types of variable and areal definitions adopted in the case studies are presented. Defining Residential Segregation Residential Segregation is defined as “the degree to which two or more groups live separately from one another, in different parts of the urban
22
FLORES
environment” (Massey and Denton, 1988, p. 282). Therefore, socioeconomic segregation can refer to the separation between the residences of the lower class and the residences of the middle and upper classes or to the degree to which different ethnic groups are located separately from one another in different neighborhoods of the city. Although the concept of residential segregation seems fairly straightforward, in fact, the separation of groups in urban space manifests itself in several forms, and each corresponds to a different aspect of the phenomenon of segregation. Massey and Denton (1988) have identified five of these dimensions: evenness, exposure, clustering, centralization, and concentration. For purposes of comparison, each study utilizes the first three. In this section, these measures are defined and their strengths and weaknesses discussed. Evenness The dimension of “evenness” refers to the “differential distribution of social groups among areal units in a city” (Massey and Denton, 1988, p. 283) or how “even” the distribution of the different groups of the population across spatial units within the city is. In terms of this dimension, residential segregation is at its minimum when social groups are evenly distributed across neighborhoods or when all spatial units of the city have an equal proportion of the population belonging to the different social groups. In contrast, evenness is minimized— and residential segregation is maximized—when any pair of individuals from two different social groups does not inhabit the same spatial unit in the city. The most widely used measure of the dimension of evenness is the dissimilarity index (Duncan and Duncan, 1955; Teauber and Teauber, 1965). For instance, if the city’s total population is divided between the poor (x) and non-poor (y), the dissimilarity index is computed as follows: D=
1 n xi yi ∑ − 2 i =1 x y
(1)
where i 5 1 . . . n are the spatial units or neighborhoods within the city; xi and yi represent the number of poor and non-poor populations in the ith zone respectively, and x and y are the total number of poor and nonpoor populations within the entire city. One of the advantages of the dissimilarity index is that it offers a straightforward interpretation. A result that ranges from 0 to 1 can be
RESEARCH METHODS FOR URBAN SEGREGATION
23
interpreted as the proportion of the poor population (x) that needs to be moved from one neighborhood to another in order to make the distribution of the population “even” across spatial units within the city. Therefore, a high dissimilarity index implies that the poor population is located in certain spatial units, whereas the non-poor population is located in other spatial units, which means that the level of segregation in the city is high. Measuring socioeconomic segregation using the dissimilarity index as expressed in equation (1) presents at least two problems. The first problem has to do with the fact that the dissimilarity index was originally conceived to compute the “evenness” in the distribution of two groups comprising the total population. Therefore, spatial segregation between socioeconomic groups depends upon the definition of two socioeconomic groups and thus on the definition of poverty or of whatever group characteristic is being used to measure segregation. While the definition of racial groups employed in the measurement of racial segregation is rather straightforward, socioeconomic groups are less clearly defined. Besides, it can also be argued that there are more than two relevant socioeconomic groups. In this case, the dissimilarity index is not able to compute the “evenness” in the spatial distribution of three or more groups. The dissimilarity index has been modified for the analysis of the spatial distribution of more than two groups through the multi-group dissimilarity index (Reardon and Firebaugh, 2002). This index however, does not have a straightforward interpretation, as does its original counterpart. The second problem with using the dissimilarity index as a measure of socioeconomic residential segregation relates to the fact that it evaluates segregation as the departure from an “equal” rather than a “random” distribution of the population, which seems rather arbitrary (Cortese et al., 1976; Cohen et al., 1978). Another measure that gauges the dimension of evenness is the entropy index. This index is more appropriate if socioeconomic residential integration—as opposed to segregation—is likened to a random distribution of the population. The entropy index “measures departure from evenness by assessing each unit’s departure from the reference group ‘entropy’ of the whole city.” (Massey and Denton, 1988, p. 285). This index, also known as the information index (H), analyzes the distribution of groups within spatial units in relation to the randomness of the distribution in the area. However, since its value depends on the level of entropy that is linked to the proportion of the minority population in the area, this index does not fulfill the requirement of compositional invariance.1 In fact, the level of entropy changes as the proportion of the minority changes. Thus the entropy index changes throughout time and from
24
FLORES
place to place even if the spatial distribution of the population remains unchanged, which makes it difficult to compare the index between different cities and/or time periods. Another commonly used indicator of evenness is the Gini index, defined as the “mean absolute difference between minority proportions weighted across all pairs of areal units, expressed as a proportion of the maximum weighted mean difference, which occurs when minority and majority members share no area in common” (Massey and Denton, 1988, p. 285). This index, however, is redundant in that it captures the same information as the dissimilarity index. Graphically, the latter is the maximum vertical distance between a Lawrence curve and the diagonal (Teauber and Teauber, 1965) while the Gini Index reflects the difference between the two curves. Exposure The dimension of exposure refers to the degree of potential contact allowed by the fact of sharing a physical residential area. In contrast to the dimension of evenness, the dimension of exposure does not depart from an ideal “even” distribution; instead “it attempts to measure the experience of segregation as experienced by the average majority or minority member” (Massey and Denton, 1988, p. 287). Socioeconomic residential segregation—in terms of isolation or lack of exposure—is captured by the isolation index (xPx): n ⎡x xi ⎤ xPx = ∑ ⎢ i × ⎥ ( xi + yi ) ⎦ i =1 ⎣ x
(2)
where xi and yi represent the poor—minority—and non-poor—majority— population in each spatial unit “i” and x represents the number of poor individuals in the city. The isolation index “measures the extent to which minority members are exposed only to one another rather than to majority members” (Massey and Denton, 1988, p. 288). It ranges from 0 to 1 and can be interpreted as the probability that a poor individual shares the spatial unit where he/ she lives with other poor individuals. The converse of the isolation index is the interaction index (xPy) that measures the degree to which the poor population in spatial unit i is exposed to non-poor individuals inside the boundaries of the neighborhood (Lieberson, 1981). Similar to the isolation index, the exposure index ranges from 0 to 1 and can be interpreted as the probability that a poor individual shares the spatial unit where he/ she lives with other non-poor individuals.
RESEARCH METHODS FOR URBAN SEGREGATION n ⎡x xi ⎤ xPy = ∑ ⎢ i × ⎥ ( xi + yi ) ⎦ i =1 ⎣ x
25
(3)
The dimension of exposure assumes that spatial proximity facilitates interaction between groups. It assumes that as segregation increases, the level of interaction—or exposure—decreases. When comparing residential segregation—in its dimension of exposure— between two cities or two periods of time, one must keep in mind that the level of exposure depends upon the size of the minority group (Blau, 1977, p. 23).2 Thus, like the entropy index, as the city-wide level of the variable, for example poverty or minority status, increases, residential segregation measured as exposure will necessarily decrease. The eta-squared or correlation ratio is often used to correct for this compositional bias.3 Clustering The dimension of clustering refers to the extent to which areal units inhabited by a particular group adjoin one another in space, thereby forming a cluster. Massey and Denton (1988) identify several measures of clustering such as the Absolute Clustering Index (ACL), the Spatial Proximity Index (SP), and the Relative Clustering Index (RCL). But for this study, the spatial Moran’s I (see equation 4), a measure of spatial autocorrelation, or the degree to which “things” in one place resemble “things” in adjacent locations, is used. For example, if the group of interest is the poverty population, the spatial Moran shows the degree to which the poverty rate in one neighborhood is similar to the poverty rate in adjacent neighborhoods. I=
n∑ ∑ wij (xi − x )(x j − x ) i
j
S0 ∑ (xi − x )2
(4)
i
where n is the number of areal units in the encompassing area, xi is the value of the variable in neighborhood i, and xj is its value in neighborhood “ j.” When measuring the spatial clustering of the poor, the spatial Moran corresponds to the weighted average of the deviations from the average variable value in a particular neighborhood and the deviations from the average value in the neighborhoods nearby. The weights (wij) define what can and cannot be considered as “nearby” and are expressed in W, the weight matrix, in which wij 5 1 if i and j are contiguous, and 0 otherwise.4 Since certain spatial units have a larger number of contiguous spatial units than others, weights are standardized by S0 = Σ Σ wij , the i j number of shared boundaries in the encompassing area.
26
FLORES
When there is no evidence of spatial autocorrelation, the expected Spatial Moran’s I (EI) tends toward zero since its expected value is estimated as EI = −1 (n − 1) that converges to zero as the number of geographical subunits (n) increases. In other words, the Moran’s I refers to the relationship between the value of a variable in a particular neighborhood or spatial unit and the average of that variable in the spatial units in the vicinity—those considered by the spatial weight matrix W. The latter is known as the spatial lag of the former.5 The correlation can be visualized in the Moran’s scatter-plot, which gives an idea of the relationship between the value of the variable in neighborhood i and its value in the vicinity—its spatial lag. This not only provides an efficient way to measure clustering at the local level but also to visualize areas where poverty concentrates.6 Spatial Moran’s I is the slope of the simple regression between the standardized values of these two elements. Anselin (1995) defines the local indicators of spatial association (LISAs) as “any statistic that satisfies the following two requirements: (a) the LISA for each observation gives an indication of the extent of significant spatial clustering of similar values around that observation; (b) the sum of LISAs for all observations is proportional to a global indicator of spatial association” (Anselin, 1995, p. 94). The most commonly used LISA is the Local Moran’s I that identifies the areas where clustering is statistically significant. These areas are called “hot spots”—areas with a high proportion of, say, minorities surrounded by areas with similarly high proportions of the same group. Local Moran’s I also identifies “cold spots”—areas with a high proportion of the majority surrounded by similar areas. Mathematically, local Moran’s I is defined as I i = (xi − x )∑ wij (x j − x )
(5)
j
This LISA indicates how similar one neighborhood is to surrounding neighborhoods. For normally distributed variables, local Moran’s I is asymptotically normal; thus, the index can be statistically tested to identify areas where clustering is statistically significant.7 This way, highpoverty areas significantly surrounded by similarly deprived areas are called hot spots, whereas areas of low poverty significantly surrounded by other neighborhoods of low poverty are called cold spots.
Types of Measures The dissimilarity index and the isolation index are simple and straightforward measures that account for the dimensions of evenness and exposure,
RESEARCH METHODS FOR URBAN SEGREGATION
27
both relevant in the case of Latin American cities. However, these measures have an important shortcoming: they do not identify which particular neighborhoods are segregated and which are not. In fact, in order to analyze the consequences of socioeconomic residential segregation, one needs a local measure that allows comparing the experiences of people residing in a particular place to the experiences of similar people located anywhere else. The Local Moran’s I solves this problem. As noted above, the decomposition of the global measure of spatial autocorrelation indicates which neighborhoods belong to a cluster of poverty, which neighborhoods are segregated in terms of clustering, and which neighborhoods are not segregated. The measure of segregation available through the Local Moran’s I differs from the dissimilarity and the isolation indexes in that the Local Moran’s I provides a local measure of segregation, whereas the latter two provide a global measure of segregation. These indexes also differ in that the former takes into account the spatial dimension of segregation. In fact, the definition of segregation in terms of clustering only makes sense when one neighborhood is related to other neighborhoods in the vicinity. This is a spatial relationship: neighborhoods are related because they are contiguous—they share boundaries—or because their centers are located at a certain distance from one another. The spatial aspect of segregation is what makes the difference in the analysis of residential segregation. A local measure of evenness or exposure would be proportional to the poverty rate in the neighborhood regardless of the poverty rate in the surrounding areas, which is not different from the analysis of the effects of neighborhood poverty on individual outcomes. The spatial dimension of segregation implies the idea of “embedded poverty” rather than poverty alone. Thus, in the dimension of clustering, the analysis of residential segregation means analyzing the effects of concentrated poverty in an area that is larger than the neighborhood itself, but spreads to other neighborhoods in the area.
Measurement Errors: Defining Areal Units and the Problem of Scale One of the most important problems in the measurement of segregation is the definition of areal units—neighborhoods—within the urban space. Conceptually, a neighborhood is difficult to define, since its boundaries are constantly being redefined by its inhabitants. Methodologically, the problem of the selection of boundaries suggests that segregation is, in a sense, a problem that varies with the scale in which it is measured.
28
FLORES
Segregation can be measured using areal units of different sizes such as census blocks, census tracts, or county subdivisions. The grid problem, also called the modifiable areal unit problem (Massey and Denton, 1998; Sabatini, 2004; White, 1983), refers to the fact that the same distribution of the population leads to different levels of segregation, depending on the size of the selected spatial unit. Since large areal units are more likely to be more heterogeneous than small areas, the division of the city into a few large areal units results in a smaller dissimilarity index than when small areal units are considered. Thus, the selection of areal units is critical when the aim is to compare residential segregation between two cities or between two points in time. On the other hand, segregation might have different implications on different scales. Large clusters of poverty are arguably more consequential than small clusters of poverty, since the former are more isolating than the latter. In sum, the definition of the boundaries of the neighborhood is relevant to the analysis of the consequences of segregation, because it affects the level of segregation itself and because segregation measured on different scales has different consequences. Thus, it is likely that they are each a different phenomenon. Moreover, the definition of the boundaries of the neighborhood depends upon the nature of the problem being analyzed. For instance, the locality affecting economic activity may well be much more extended than the neighborhood affecting kinship. Some activities such as work can be carried out miles away, whereas others such as mutual support are more likely to occur closer to home (Lupton, 2003). Thus, the contextual effects have to be measured on different scales depending on the outcome under study (Gephart, 1997). In this sense, physical spaces are measured in order to capture social spaces. Neighborhoods cannot be regarded as containers in which social interactions take place; neighborhoods are sets of social networks whose boundaries depend upon the nature of the dimension being examined (Massey, 1994). This suggests that the boundaries of the neighborhood are internally defined by the scope of social interactions that take place in space. Glennester et al. (1999) define neighborhoods as entities that are made up of layers of interactions, defined by travel areas, physical characteristics, and boundaries drawn by service providers. Thus, neighborhoods are socially and physically defined. The size and shape of areal units will vary from place to place according to the characteristics of the natural and constructed environment and the forms of local, political, economic, and social interaction. Kearns and Parkinson (2001) argue that a neighborhood exists at three levels: the home area that fosters the psychological goals of attachment, belonging, and values; the locality that confers social status; and the urban district that provides a wider
RESEARCH METHODS FOR URBAN SEGREGATION
29
landscape of social and economic opportunities. In practice, the selection of the areal units is often arbitrary, since it depends upon the availability of data. Nonetheless, researchers should consider that areal units of different sizes might lead to different conclusions. The most appropriate size will depend on the outcome under analysis and other considerations, such as the nature of the educational system.
Dependence, Autocorrelation, and Spillovers in Space Spatial dependence refers to the tendency for observations close in space to be more highly correlated than those further apart. According to the first law of geography, “everything is related to everything else but near things are more related than distant things” (Tobler, 1970, p. 234), implying that things in one place depend upon things in places nearby. The underlying assumption is that spatial propinquity—and the interaction between spatial units—generates a spatial process by which things that are close to each other affect one another. Spatial dependence refers to the idea that spatial proximity affects behavior, due to exposure and diffusion. Behavior in and antecedent conditions of spatial unit i have consequences on behavior in adjacent spatial unit j because the latter is exposed to the former (Morenoff et al., 2001). Spatial dependence exists because exposure leads to a process of diffusion of behavior that spreads from one unit of analysis to its vicinity. Spatial dependence is also a result of measurement error. In fact, the boundaries for collecting information—delineation of space into the spatial units we call neighborhoods—are arbitrary and may well miscalculate the ecological nature of the spatial unit in which the processes we are trying to measure occur. Either way, the first law of geography states that spatial dependence is the rule rather than the exception (Anselin and Bera, 1998); as such, accounting for the correlation of observations closely located in space is as important as dealing with other common data-related problems, such as time autocorrelation in panel data and heteroskedasticity in cross-section data. Technically, spatial dependence is a property of joint density functions and as such, it is virtually impossible to be verified in practice. Spatial autocorrelation—as a moment of the joint distribution—emerges as a more manageable approach—it can be estimated and tested—to tackle the problem of spatial dependence (Anselin and Bera, 1998). One way to test for autocorrelation is through the spatial Moran’s I explained above. Spatial spillovers can be defined as the benefits—or costs—that trickle down from a source to elements in the surroundings. In a sense, spillovers
30
FLORES
are a direct consequence of the phenomenon of spatial dependence, inasmuch as they can be understood as spatial externalities enabled by a spatial autocorrelation process. Thus, it is likely that high levels of segregation— defined as the spatial separation of groups—hinder the degree to which different social groups interact and affect one another. In other words, residential segregation inhibits positive externalities, or spillovers, from high-performing to low-performing schools, or from integrated families to excluded families. Although spatial autocorrelation indicates the strength of spatial spillovers, it needs to be corrected, generally by means of spatial models. When a cross-sectional data set shows spatial dependence, traditional econometric techniques are no longer useful for estimation and testing. Failure to account for spatial dependence in a cross-section analysis has consequences similar to failure to correct for autocorrelation in a time-series analysis, by which things at one point in time are correlated to things in later periods. Despite the similarities between spatial analysis and time-series analysis, one of the main problems in spatial dependence models is that the notion of spatial shift is much less clear than the notion of time shift: while a time unit (t) has only one unit ahead (t 1 1) and one unit behind (t 2 1), in space, neighboring units can be multiple. Thus, spatial models use a spatial lag operator that refers to the weighted average of random variables in neighboring units. Thus, the definition of a neighborhood set for each location is of critical importance. The spatial lag for the dependent variable y in the spatial unit i can be expressed as N
⎡⎣Wy ⎤⎦ = ∑ wij y j i
(6)
j =1
Value wij corresponds to the spatial weight applied to the dependent variable in unit j, and depends on the spatial relationship between spatial unit i and spatial unit j. In general, when wij takes value 1 it indicates adjacency between spatial units i and j. In this case, the assumption is that dependence is restricted to adjacent units. Assuming that diffusion is a more continuous process that does not require adjacency, spatial weights can also be calculated as the inverse of the distance between spatial units i and j (Cliff and Ord, 1981). This way, closer units will have a bigger impact on unit i than more distant spatial units. Thus, the first step that is necessary for modeling spatial dependence is to build a (N3N) matrix that captures the spatial relationship between all spatial units in the city. The matrix is known as the spatial weights
RESEARCH METHODS FOR URBAN SEGREGATION
31
matrix (W) and its construction implies making assumptions about the nature of the process of diffusion. The elements in this matrix are used to calculate each spatial unit’s spatial lag. By convention, the elements in the diagonal—the spatial relationship between spatial units with themselves or wii—are set to 0. Also, the elements of the spatial weights matrix are typically row-standardized such that the sum of all wij is 1. As an example of the spatial lag resulting from an adjacency matrix, when a spatial unit has three neighborhoods—each representing onethird of its “neighborhood”—with dependent variables of values 3, 5, and 1 respectively, the spatial lag of unit i will be 3 = (1 3) ∗ 3 + (1 3) ∗ 5 + (1 3) ∗ 1. In this sense, the spatial lag represents the weighted average of the value of y in neighboring units, also called a spatial smoother (Anselin, 1988).
Variables and Spatial Units Utilized in the Case Studies Spatial segregation can be defined along a wide range of demographic variables, such as income, education, poverty, and race/ethnicity. To enhance comparability across cities in this project, a variable common to case studies was sought. An inventory of data sources from the various countries was developed, and it was determined that the only socioeconomic variable available for a geographically defined unit in all cities was level of education. Therefore all case studies present segregation according to education levels. Education can be formulated in real and relative terms, where the former is the number of adults with or without a primary education and the latter captures the share of individuals below the mean national urban education level (see table 2.1). Most case studies incorporate other socioeconomic variables, such as income and poverty Table 2.1
Education variables for segregation indices
Variable
Minority Population (pi)
Majority Population (12pi)
1. Relative Measure of Education
Heads of households 25 and older with below the mean education level (calculated at two points in time). Heads of households 25 and older with primary* education or less.
Heads of households 25 and older with above the mean education level (calculated at two points in time). Heads of households 25 and older with more than primary* education.
2. Absolute Measure of Education
Note: * The degree attainment used should be appropriate to the urban area at question. For example, in the case of Austin it would be more appropriate to use adults with less than a secondary (less than high school) education.
32
FLORES
status, in order to develop a more complete understanding of segregation patterns. All authors adopt a two-group classification and some utilized multi-group segregation measures. Due to both data collection and reporting procedures and the boundaries of governmental jurisdictions, spatial segregation can be measured at multiple levels. First, administrative boundaries can be used to define aggregate units of analysis for the selected segregation measures. For example, in order to measure the spatial segregation in a given metropolitan area, the administrative boundaries for municipal districts, census tracts, and city blocks are available (see table 2.2 for availability of spatial data in the United States. Most countries considered here follow a similar framework). In this study, metropolitan areas are examined at two levels of analysis: global and local. Global-level segregation measures are calculated using indicators constructed at the district- or the census-tract levels. Measures of local segregation are calculated for lower-level spatial units. The spatial level affects the empirically determined segregation outcomes, as discussed above. In the Santiago and Mexico case studies, formal comparisons are made of indices calculated with different spatial levels. As anticipated, lower values on the segregation index are found at the higher level of spatial definition. At higher spatial definitions the likelihood of heterogeneous population, that is, low segregation, is more likely than at a local level. In the following case studies, authors incorporate the lowest possible spatial definition, but some studies include higher-level definitions as well. Although this strategy enhances comparability, the authors of the Santiago study discuss the possibility that alternative spatial definitions are capturing distinct social processes, and, therefore, the lowest level of spatial definition will not necessarily be the best choice for all types of analysis. Table 2.2
Alternative spatial definitions of data
Data Layer
Level of Analysis
Description
Census Block Group
Local
Census Tract
Local
Municipal District
Global
Metropolitan Region
Global
In the United States, an optimal Census block group gathers 1,500 people; however it varies from 300 to 3,000 people. In the United States, an optimal census tract gathers 4,000 people; however it varies from 1,000 to 8,000 people. This is a subregional layer that may provide a different view of segregation. The metropolitan region layer can be used to provide a “baseline” regional measure of segregation to compare local levels.
RESEARCH METHODS FOR URBAN SEGREGATION
33
Linking Public Policy to Spatial Segregation An examination of the relationship between residential segregation and urban services is one of the research questions in this study (see chapter 1). To pursue this objective, authors were given flexibility on the choice of policy areas considered, recognizing that the data available and policy priorities could vary significantly across cities and countries. Three of the studies explore the impact of segregation on access to employment. Three studies assess access to different types of infrastructure for areas defined by segregation, and one study considered access to health insurance. Since all the studies in this volume use measures of educational achievement for assessing segregation, many of the studies examined educational system policies. Furthermore, authors were given flexibility in the choice of methods for exploring the nature of the relationship between segregation and urban services, thereby producing an eclectic set of analyses that offer useful perspectives on future lines of research.
Notes 1. There are a number of requirements that a measure of segregation has to fulfill (Teauber and Teauber, 1965). One is the composition invariance that states that if the number of persons of group “m” in each unit increases by a constant factor “p,” and the number and distribution of persons of all other groups is unchanged, segregation is unchanged. Although not all scholars agree that composition invariance is a desirable property of segregation measures. it is important if segregation needs to be compared between units or between two points in time. Other requisites/requirements are: Organizational equivalence (if a spatial unit is divided into “k” units, each having the same group proportions as the original unit, segregation remains unchanged. Likewise, if “k” units with identical group proportions are combined into a single unit, segregation is unchanged); Size invariance (if the number of persons of each group in each spatial unit “k” is multiplied by a constant factor “p,” segregation is unchanged); Transfer principle (if an individual of group “m” is moved from spatial unit “i” to unit “j” where the proportion of persons of group “m” is greater in unit “i” than in “j,” then segregation is reduced); Exchange (if an individual of group “m” in unit ‘′’ is exchanged with an individual of group “n” in unit “j,” where the proportion of persons of group “m” is greater in unit “i” than in unit “j,” and the proportion of persons of group “n” is greater in unit “j” than in unit “i,” segregation is reduced); Additive organizational decomposability (if J spatial units are clustered in K clusters, then a segregation measure should be decomposable into K clusters, that is, a sum independent within—and between—components); and Additive Group Decomposability (if
34
FLORES
M groups are clustered in N supergroups, then a segregation measure should be decomposable into a sum independent within—and between—supergroup components). 2. Exposure and isolation indexes are sensitive to population composition. As the minority gets smaller, the majority increases in number, which increases the minority’s probability of being exposed to the majority. Thus, per construction, segregation is smaller in places where the minority is relatively more represented than in places where minority groups are relatively small. 3. Eta squared is calculated as follows: V = ( xPx − P ) (1 − P ), where xPx corresponds to the isolation index and P = x ( x + y ) . 4. This is known as the contiguity matrix. In this case, neighborhoods in the surroundings are exclusively those that share a boundary. There are other ways of defining the weight matrix that considers, for instance, distance as a parameter in the definition of the surroundings. In this case, the matrix is known as the distance matrix and the value of wij is the inverse of the distance between the center of neighborhood i and the center of neighborhood j. 5. This way, analyzing spatial autocorrelation is similar to time analysis where the dependent variable in time t has its correlate in the past (time t21). The difference is that the spatial autocorrelation vicinity goes in multiple directions— north, south, east, west, and the coordinates in between—whereas vicinity in time analysis is usually restricted to periods in the past. This is why spatial lag needs to be simplified, calculating the average in the vicinity. 6. Following this logic, segregated neighborhoods, defined for a particular variable such as poverty, would be those would be located in the upper-right corner of the upper-right quadrant of the Moran’s I scatter plot. 7. The moments of the Local Moran’s I can be derived using the principles outlined by Cliff and Ord (1981, pp. 42–46). See Anselin, 1995, p. 99.
References Anselin, L. (1988). spatial econometrics: Methods and models. Dordrecht/ Boston/London: Kluwer Academic Publishers. ——— (1995). Local Indicators of Spatial Autocorrelation- LISA. Geographical Analysis. 27(2), (April). ——— (1999). Interactive techniques and exploratory spatial data analysis. In P. Longley, M. Goodchild, and D. Rhind (eds.), Geographical information systems 2nd edition. New York: John Wiley & Sons. Anselin, L. and A. Bera (1998). Spatial dependence in linear regression models with an introduction to spatial econometrics. Chapter 7 in A. Ullah and D. E. A. Giles (eds.), Handbook of applied economic statistics. Statistics, Textbooks and Monographs, p. 155. New York: Marcel Dekker. Blau, P. (1977). Inequality and heterogeneity. London: Collier Macmillan Publishers. Cliff, A. and Ord, J. K. (1981). Spatial processes: Models and applications. London: Pion.
RESEARCH METHODS FOR URBAN SEGREGATION
35
Cohen, J. K., Frank Falk, R., and Cortese, C. F. (1978). Understanding the standardized index of dissimilarity. American Sociological Review, 43, 590–92. Cortese, C. F., Frank Falk, R., and Cohen, J. K. (1976). Further considerations on the methodological analysis of segregation indices. American Sociological Review, 41, 630–37. Duncan, O. and Duncan, B. (1955). A methodological analysis of segregation indices. American Sociological Review, 20, 210–217. Gephart, M. (1997). Neighborhoods and communities as context for development. In J. Brooks-Gunn, G. Duncan, and L. Aber (eds.), Neighborhood poverty. Volume I: Context and consequences for children, pp. 1–43. New York: Russell Sage Foundation. Kearns, A. and Parkinson, M. (2001). The significance of neighborhood. Urban Studies, 38, 2103–2110. Lieberson, L. (1981). An asymmetrical approach to segregation. In C. Peach, V. Robinson, and S. Smith (eds.), Ethnic segregation in cities, pp. 61–82. London: Croom-Helm. Lupton, R. and Power, A. (2003). Social exclusion and neighborhoods. In J. Hills, J. L. Grand, and D. Piachaud (eds.), Understanding social exclusion, pp. 118–140. Oxford: Oxford University Press. Massey, D. and Denton, N. A. (1988). The Dimensions of residential segregation. Social Forces, 67(2), 281–315. Reardon, S. and Firebaugh, G. (2002). Measures of multigroup segregation. Sociological Methodology, 32(1), 33–67. Sabatini, F. (2004). Medición de la segregación residencial: Reflexiones metodológicas desde la ciudad latinoamericana. In F. Sabatini and G. Caceres (eds.), Barrios cerrados en Santiago de Chile: Entre la exclusión y la integración residencial. Cambridge, MA: Lincoln Institute of Land Policy. Teauber, K. and Teauber, A. (1965). Negroes in cities: Residential segregation and neighborhood change. Chicago: Aldine. Tobler, W. (1970). A computer movie simulating urban growth in the Detroit region. Economic Geography, 46, 234–240. White, M. (1983). The measurement of spatial segregation. The American Journal of Sociology, 88, 1008–1018.
This page intentionally left blank
Part I
The Metropolis
This page intentionally left blank
3
Residential Segregation in Greater Buenos Aires Fernando Groisman and Ana Lourdes Suárez
Introduction Social research has recently demonstrated renewed interest in the study of residential segregation in Latin America. In large part, this is the result of the increase in the adverse social effects of spatial segregation during the 1990s. There is a certain consensus regarding the role of the region’s neoliberal policies in influencing this trend. Furthermore, although it is controversial, some studies have indicated that levels of segregation are on the rise (Arriagada and Rodríguez, 2003). Residential segregation and its association with differential opportunities for well being is a topic long debated in sociological research. However, there exist few studies of the magnitude of the phenomenon in Latin America, its evolution, and the mechanisms through which it operates to affect the life chances of the population. The lack of appropriate empirical evidence for many countries in the region in part influences this fact. In Argentina the topic has been partially explored, with a focus on the study of the changes associated with the increase of gated communities that spatially concentrate those of the highest incomes (Torres, 2001; Thuillier, 2005). Usually residential segregation is defined as the separation of two or more groups in urban space, or in other words, as the extent to which two or more groups live separated from each other in different parts of the urban territory (Massey and Denton, 1988, p. 282). The work of Wilson (1987) gave more attention to residential segregation by race and ethnicity. For Argentina, and particularly for the large cities, the most pertinent
40
GROISMAN AND SUÁREZ
divisor is class structure or social strata. For that reason, this work concentrates on residential segregation by socioeconomic status. This study, in large part exploratory, concentrates on the patterns of residential segregation in Greater Buenos Aires via two approaches. Its evolution during the nineties is described through various variables of stratification that reflect the socioeconomic level of households, as well as others that tap into aspects of urban infrastructure and access to services. The method employed consists of secondary data analysis to quantify the levels of and change in residential segregation. This study uses population and household census data from 1991 and 2001 as well as household surveys available from the National Institute of Statistics (INE). The paper is structured as follows: first, we describe Greater Buenos Aires and the demographic changes that have taken place between 1991 and 2001; second we explain our research method; and, finally, we present our results and conclusions.
Tendencies in the Spatial Configuration of Greater Buenos Aires Greater Buenos Aires forms part of the Metropolitan Region of Buenos Aires (MRBA). This is the country’s largest urban area. Within its boundaries lie 18,000 square kilometers (0.7 percent of the Argentinean territory) and more than one third of the country’s population (38.4 percent), as well as a large part of the gross domestic product and industrial employment. Greater Buenos Aires is composed of the city of Buenos Aires and its Conurbation (Conurbano Bonaerense)—consisting of 24 partidos, or districts. The urban agglomeration constitutes an “urban entity” as much from the functional point of view—being the field of daily movements of the population, particularly the commute between home and work—as from a physical point of view—forming an urban sprawl without any significant breaks in continuity. The context of recent spatial tendencies is the great urban explosion that took place since the last half of the past century. The extraordinary urban growth of the area was associated with a process of suburbanization, primarily characterized by the extension of the city in the direction of its periphery. The growth in urbanization consolidated a first ring, and later the second and the third rings. During the period of import-substituting industrialization that began during the 1930s, the dynamic of appropriation of space in the metropolitan region of Buenos Aires was stimulated by redistributive policies that promoted access to housing for the middle and working classes (Torres, 1993). Those of lower socioeconomic status
RESIDENTIAL SEGREGATION IN BUENOS AIRES
41
moved toward the suburbs, motivated by the offer of affordable lots that permitted access to land ownership. This population movement followed the trail of the railways: the fares for its users, due to subsidized public service companies, favored the movement of these sectors to the periphery. In this way the growth of the suburbs proceeded at very low densities and with a slow development of urban infrastructure. As a result, a disordered and poorly planned urban space formed with deficiencies in basic infrastructure, urban services, and transportation. In the decades of 1960s and 1970s, metropolitan growth slowed with a consistent reduction in the relative participation of migration. The pattern of urban space occupation described above was reinforced and intensified. The last military government initiated the eradication of the villas miseras, or slums, of the city of Buenos Aires. The displaced population situated themselves in the Conurbation. During this period, the government also introduced changes in urban policies to limit the occupation of parcels of land lacking infrastructure and services. The effect of both policies was a dislocation of the urban poor toward the periphery, a densification of poor neighborhoods, and an incipient process of illegal land occupation. After 1980, the growth of Greater Buenos Aires was characterized by a low population growth rate, and this tendency persisted through the nineties. The partidos, located in the second ring (where 70 percent of the total growth of the agglomeration occurred) have grown the most; whereas those located closest to the federal capital have lower demographic growth. The main tendency in population changes between 1991 and 2001 was residential polarization, as much inside the city of Buenos Aires as in its surrounding areas. Between 1991 and 2001, the population of Buenos Aires decreased, from 2,965,403 to 2,776,138. In contrast, the population of the surrounding Conurbation increased from 7,952,624 to 8,684,437 in the same decade. In this period the most significant actors in the definition of social spaces were the upper classes (Cerrutti and Grimson, 2004). Between 1991 and 1999, the number of luxury houses multiplied fourfold (Torres, 2001), and the number of working- and middle-class houses decreased more than 10 percent. The process of suburbanization of the elites intensified, paralleling the increase in the growth of gated communities. All measures of the magnitude of the expansion of gated communities indicate an extraordinary growth during the nineties. Between 1990 and 1991, 91 different communities were registered, whereas in 2001, there were 461 (Torres, 2001). This type of urbanization, which includes various types—country clubs, gated communities, clubes de chacra, and
42
GROISMAN AND SUÁREZ
the megaemprendientos—marks a new pattern in the appropriation of space in Buenos Aires. At this same time, at the other extreme of residential segmentation, the city witnessed an increase in precarious settlements, as well as a general deterioration in the neighborhoods traditionally home to the lower and middle classes. According to estimates from the Info-Habitat team at the Universidad Nacional de General Sarmiento, approximately 1 million people are distributed between 700 to 1,000 villas and settlements within the Conurbation. In the city of Buenos Aires today there are approximately 100,000 people residing in precarious settlements. To these numbers of precarious settlements must be added the “nearly one million people in irregular conditions that are difficult to register: a family that occupies one lot; squatters in houses in the city; and, clandestine lots, especially in the third ring of the Conurbation. In sum, the total comes close to 2 million people,” (Fernández Wagner, 2005).
Methodology Urban segregation is a multidimensional phenomenon; as a result, a great variety of segregation indicators exists. In this work we rely on the three most frequently used indicators: the Duncan dissimilarity index, the Bell isolation/exposure index, and the spatial Moran’s I. The dissimilarity index belongs to the family of evenness indicators, which permits the comparison of the distribution of two groups, one of which is the segregated or minority group. The index is based on the difference between the proportion of individuals in the minority group and the proportion of the rest of the population in each spatial unit, calculating the proportion of the minority group which would have to change residence in order to reach a distribution equal to that of the total area in question. The Bell isolation/exposure index captures the exposure or potential contact between the two groups. It takes into account the representativeness of the groups in the total population and measures the degree of potential contact between members of the same group in each one of the areas. Finally, Moran’s I measures the degree of spatial clustering according to a given characteristic of a territory. The estimations were produced with the microdata from the two most recent National Censuses of Population and Housing (Censo Nacional de Población y Vivienda—CNPV) conducted in 1991 and 2001. The variables used to construct the groups were the following: educational level of the household head, as a proxy for social stratification;
RESIDENTIAL SEGREGATION IN BUENOS AIRES
43
incidence of structural poverty, measured by the dichotomous indicator of Unmet Basic Needs (Necesidades Básicas Insatisfechas—NBIs); the proportion of households with deficient infrastructure or classified as “Type B”; migration status; and, health insurance coverage for the household head. The analysis employed data for partidos and fracciones censales (units similar to census tracts) for the case of the Conurbano, departamentos, and neighborhoods within the city of Buenos Aires. The difference between a partido and fracción censal lies not only in the size of the spatial unit but also in the boundaries and government body in charge. In effect, the partido coincides with the municipal jurisdiction at the Conurbano while the fracción is a censual unit that subdivides the area of each partido and departamento. It must be kept in mind also that there is a great disparity in size between these censual units. The city of Buenos Aires is divided into 21 departamentos with an average of 48,800 households each and 285 fracciones with an average of 3,500 households. Similarly, the Conurbano is made up of 24 partidos with an average of around 99,000 households and 369 fracciones with 6,400 households each. Our research hypothesis is that as a result of the processes of social exclusion and increasing poverty that took place during the decade of the nineties, segregation by socio economic status has become more pronounced. To test this, we turn to the stratification of households by health insurance coverage and educational level of the household heads. Head of Household Educational Level. We compute indices for different social strata according to the highest level of education completed by household heads. Head of Household Health Insurance Coverage. We dichotomize whether or not the household head had access to health care through a public system, a prepaga or a mutual. Since the middle of the last decade access to health care in Argentina has been intrinsically tied to formal employment. The increasing precarious nature of employment during recent decades thus affected the way in which people access health care. By looking at health care coverage then, we can measure the processes of segregation according to the labor market characteristics of the household head. Head of Household Migratory Status. The minority group is composed of household heads born outside of Argentina—the pattern of international migratory movements into the country during the last decades is that of citizens from bordering countries coming to Argentina. Household Poverty as Measured by the NBI. The segregated group is composed of those households that posses at least one of the components
44
GROISMAN AND SUÁREZ
for the composite index of NBI: (a) more than three persons per room (that is, overcrowding); (b) housing precariousness; (c) sanitary deficiencies (absence of a bathroom connected to a sewage system); (d) children between the ages of 6 and 12 who are not enrolled in school, and; (e) four or more household members for each employed member and head of household with low educational attainment. Type of Housing. The minority group is composed of those households that reside in deficient housing classed by the Census as Type B housing and casillas. This refers to a censual classification that groups all houses that meet one of the following criteria: dirt floors, floors with loose bricks, or floors made of another material (i.e., that do not have ceramic, tile, mosaic, marble, wood, or carpeted floors); or, absence of running water inside the house or bathrooms without running water. This indicator allows us to approximate a broader group of infrastructure deficient households than those captured by solely utilizing the NBI. The variation in type of housing reflects, although in an indirect form, the public housing policy during the decade under study. The prevailing zoning code since the middle of the seventies imposes stringent standards for housing. The 1977 Provincial Law 8912 of Zoning of the Province of Buenos Aires regulates the organization of land in the province, defining the use, subdivision, occupation, and infrastructure provision. It obliges the municipalities to adhere to these definitions of land use, and demands basic infrastructure works, defining dimensions and equipment to be furnished. This law makes real estate developers responsible for the public works in their developments (Catenazzi and Fernández Wagner, 2004). This technical regulation should have improved housing conditions, decreasing contamination and risk of flooding and favoring the provision of collective infrastructure. However, it produced the opposite effect for low-income families— infrastructure provision increased land values, reducing the offer of lots accessible to those sectors (Garay, 2004).
Results We proceed to analyze the changes in the composition of households according to the selected variables over the course of the decade. It is possible that the changes in the spatial distribution of households are not reflected in changes in the composition of households by socioeconomic groups used in this analysis of segregation and vice versa. However, the analysis of both processes permits a more complete approximation of the social transformation that proceeded in the area under study.
RESIDENTIAL SEGREGATION IN BUENOS AIRES
45
Between 1991 and 2001—as much in the city of Buenos Aires as in the Conurbation—the educational levels of household heads increased: the number of those that had not finished secondary school decreased from 50.3 percent to 39 percent of the population of the city and from 80 percent to 72 percent of the Conurbation. The percentage of those without health insurance increased from 15.2 percent to 21.1 percent in the city and from 32.1 percent to 44.5 percent in the Conurbation indicative of the poor labor market during the period. The number of households headed by immigrants from neighboring countries increased from 4.4 percent to 5.6 percent in the city and from 5.4 percent to 6.7 percent in the Conurbation, as has been documented elsewhere (Cortés and Groisman, 2004; 2005). In the case of the Conurbation, we also include the percentage of households with at least one NBI living under precarious housing conditions. In both instances the changes are small. Those households with one or more NBIs declined from 16.6 percent to 15 percent and those living under precarious housing conditions increased from 14.7 percent to 16 percent. Table 3.1 shows the dissimilarity and isolation indices of segregation for 1991 and 2001. One result worth mentioning is that both in the Conurbano Bonaerense and the city of Buenos Aires there were no major changes in segregation as measured by the level of education of the heads of households. This suggests a lack of spatial displacement during the period at the level of the unit of analysis: partidos and fracciones in the Conurbation and departamentos and neighborhoods in the city of Buenos Aires. Given the strong increase in the educational levels of household heads between 1991 and 2001, the small decrease in the indices of educational segregation are indicative of the persistence of segregation. In this sense the City of Buenos Aires stands out, in that the increases in the educational levels of heads of household were large while educational segregation indices for the city decreased more moderately in comparison to the Conurbation, particularly for those with secondary education or higher. An increase in segregation as measured by both indices is noticeable only when they are computed using the health insurance coverage of heads of households. In effect, as much for the city of Buenos Aires as for the Conurbation—and for the different types of territorial units selected—residential segregation according to this attribute increased, and this evolution was more evident among the residents of the city of Buenos Aires. Although the percentage of household heads without health insurance had increased between 1991 and 2001, this increase was not homogenously distributed across the area of study. The fact that the index of isolation increased the most indicates the generalized extension of this deficit.
46
GROISMAN AND SUÁREZ
Table 3.1
Dissimilarity and isolation indices of segregation
City of Buenos Aires Below primary By departamento education By neighborhood Up to primary By departamento education By neighborhood Secondary By departamento education By neighborhood College By departamento By neighborhood Health By departamento insurance By neighborhood Migrant By departamento By neighborhood Conurbation Below primary education Up to primary education Secondary education College Health insurance Migrant NBI Type B housing
By partido By fracciones By partido By fracciones By partido By fracciones By partido By fracciones By partido By fracciones By partido By fracciones By partido By fracciones By partido By fracciones
Dissimilarity Index
Isolation Index
1991
2001
1991
2001
16.9% 17.0% 20.6% 21.4% 23.0% 23.4% 26.0% 27.5% 10.3% 12.0% 13.7% 15.6%
19.8% 19.5% 20.9% 21.0% 22.5% 22.4% 25.7% 26.8% 18.8% 19.7% 24.7% 27.0%
10.0% 10.2% 39.3% 39.9% 53.9% 54.5% 16.2% 17.1% 16.0% 16.3% 5.1% 5.1%
5.9% 6.1% 27.8% 28.3% 41.7% 42.3% 20.9% 21.8% 24.2% 24.6% 9.4% 9.2%
10.7% 21.3% 14.0% 26.8% 17.8% 34.0% 27.8% 49.1% 13.0% 24.2% 15.1% 28.2% 15.7% 32.9% 30.7% 47.8%
11.5% 22.5% 13.6% 27.0% 13.9% 27.5% 24.9% 47.4% 14.8% 26.2% 9.1% 13.8% 17.3% 31.5% 28.3% 47.3%
25.3% 27.9% 66.2% 68.5% 23.5% 29.7% 5.4% 9.0% 32.6% 36.1% 5.9% 8.1% 16.9% 22.2% 20.2% 28.0%
21.1% 23.7% 35.5% 49.9% 35.0% 49.3% 6.9% 12.1% 46.3% 49.7% 12.0% 12.9% 16.4% 20.2% 28.2% 38.3%
Source: Authors’ calculations using the National Census of Population and Housing of 1991 and 2001.
The explanation of this evolution can be found in what took place in the labor market during this period. Households’ access to health insurance coverage is assured through social welfare systems—for the employed, retired persons, and pensioners—and the access to this protection is intimately tied to the insertion of household heads in formal employment and the continuation of this status until retirement. During the decade 1991 to 2001, the formal employment of household heads decreased, precarious employment increased, and evidence exists that occupational instability associated with the loss of formal jobs increased.
RESIDENTIAL SEGREGATION IN BUENOS AIRES
47
Higher segregation is an indicator of the differential impact of deteriorating labor conditions according to place of residence. The influence of the surrounding urban environment is reflected in people’s employment trajectories through a variety of mechanisms, for example, stigmatization because of neighborhood of residence, high costs of transportation, difficulties in entering and leaving jobs, and so on. It is also possible that those who lost their formal jobs moved to cheaper neighborhoods. On the other hand, over the course of the decade there were no changes in the distribution of households according to the type of housing where they reside. However, the isolation index of those in precarious housing conditions increases, implying a relative increase of this group. In other words, this reflects the extension of housing precariousness in areas of high population concentration. Finally, with regards to the migratory condition of the heads of the households, a different pattern presented itself in the city of Buenos Aires and in the Conurbation. In the former, the indices of dissimilarity and isolation increased, while in the latter, dissimilarity decreased, although isolation did not. If segregation behaves differently depending on the variable by which it is measured, an implicit assumption is that each should reflect different dimensions of social inequality, and, accordingly, we could expect that the relative values for each dimension coincide spatially. Estimating the correlation coefficients of the spatial distributions of the variables utilized to measure segregation confirmed the high association between them. The spatial analysis confirms the above results. A high degree of spatial correlation exists in the metropolitan area of Buenos Aires when it is evaluated separately in terms of educational attainment as well as health insurance coverage of heads of households. Map 3.1 illustrates the results for health insurance coverage. As educational attainment is a determinant of the quality of labor market insertion, it can be expected that the spatial correlation of both these indicators would be similar. The estimations obtained confirm this assumption. The percent of households with heads with little education in each territorial unit—in this case censual fracciones—is a good predictor of the values of this same indicator in the contiguous surrounding space. This is illustrated with the value of the global Moran’s I. In other words, the metropolitan area reflects a pattern of spatial distribution of households characterized by the existence of territories with similar social composition. The characteristic most relevant to this territorial configuration is the existence of residential clusters. The households with highly educated heads concentrate in the northeast zone of the region that borders the north of the city of Buenos Aires and the
48
GROISMAN AND SUÁREZ
partidos of the north of the Conurbation. In a complementary manner, the periphery of the region—in a ring shape—concentrates household heads with little education. Between 1991 and 2001 this organizational territorial pattern persisted. However, the grade of spatial association reduced slightly—the global Moran’s I decreased from 0.85 to 0.78—which is a reflection of the increase in the educational level of subsequent generations. This reduction is marginal in relation to the percentage change in the educational attainment of household heads during the period (approximately 10 percent). This also reflects the slight changes in the territorial clusters: the area where highly educated households concentrate increased and the area that contains households with low education marginally decreased. We can examine whether these values, aggregated at the metropolitanregion level, reproduce themselves in a similar form in the two geographic spaces that make up the region: the city of Buenos Aires and the Conurbation. In the first, the fringes of high and low education in the north and south of the city respectively are clearly formed, while the Conurbation shows several different clusters both for the highly educated and for the poorly educated. This difference also reflects the greater demographic and geographic heterogeneity, as well as the larger size of the Conurbation with respect to the city of Buenos Aires. Over the decade, the increase in the educational level of the heads of households was much greater in the city of Buenos Aires. This is reflected in a marked decrease in the global Moran’s I in the city from 0.842 to 0.648, as opposed to the Conurbation where it decreased from 0.634 to 0.607. An increase in the territorial rift between the northern and southern zones of the city materialized. With respect to the changes observed in the clusters in the Conurbation the volume of concentration of those highly educated households increased slightly. This augmentation reflects the increase in the gated communities located in three focal points: north, west, and south. Paralleling this, it appears that three spaces which concentrate households of low resources consolidated themselves: in the northeast, in the west, and in the south. Within the Conurbation, small-scale segregation is increasing, which may affect the degree of clustering of homogeneously poor neighborhoods through the presence of enclaves of wealth. By way of conclusion, it is clear that the metropolitan area of Buenos Aires exhibits a pattern of high residential segregation according to the social composition of its households—approximated by the educational attainment of their heads—that manifests itself in high spatial
RESIDENTIAL SEGREGATION IN BUENOS AIRES
49
correlation and the existence of geographic clusters of households of similar resources. This pattern was not significantly modified during the decade of the nineties, notwithstanding the increase in the educational level of new cohorts. The second dimension of analysis corresponds to the health insurance coverage of heads of households. The spatial analysis of households according to this characteristic allows us to approximate the existence of residential segregation associated to a householder’s insertion in the labor market. These can be observed in map 3.1 that presents the local Moran’s I—the dark gray and light gray zones indicate, respectively, low and high concentration of households without health coverage. In the metropolitan region we have a nucleus—in the northeast—composed of households with high proportions of health insurance coverage, whereas in the periphery the opposite predominates. For the Conurbation and the city, the analysis is consistent with that focused on educational attainment with the global Moran’s I for health insurance coverage declining from 0.334 to 0.215 for the city, but increasing from 0.559 to 0.589 for the Conurbation. The spatial analysis appears to confirm, then, that the high residential segregation that the metropolitan region exhibited in 1991 persisted throughout the decade. Furthermore, it appears that clusters based on the socioeconomic characteristics of the households were further intensified (see map 3.1). The persistence of high levels of segregation between 1991 and 2001 begs the question of what has happened since then, especially with regards to the relationship between the labor market and residential segregation in the context of the economic expansion in the years following 2001. We rely on the information from the Permanent Survey of Households (Encuesta Permanente de Hogares—EPH) conducted by the INE to test the incidence of the characteristics of the urban surroundings of households by the performance of people in the labor market in a period of economic growth. Variables reflecting characteristics that are not direct attributes of households but are a sign of the conditions of the habitat location may affect labor outcomes. This influence involves a variety of mechanisms that reinforce one another: residence-labor mismatches, stigmatization, information problems, and so on. As of 2003 it is possible to identify households surveyed that share the same urban surroundings. Two variables were constructed, which measure, respectively, the presence of a certain percentage of households with piped gas and households with residential ownership but not ownership of the land on which it is located—usually a proxy for illegally occupied land and/or villas. The
Moran’s I=0.6632 W_PRO_JEFNOCU
W_PRO_JEFNOCU
Moran’s I=0.6877
PRO_JEFNOCU
PRO_JEFNOCU
1991
Map 3.1 Indicators of spatial autocorrelation: Global and Local Moran’s I without health insurance coverage, 1991 and 2001
RESIDENTIAL SEGREGATION IN BUENOS AIRES
51
2001
Map 3.1 Indicators of spatial autocorrelation: Global and Local Moran’s I without health insurance coverage, 1991 and 2001
thresholds utilized in each case were 20 percent and 35 percent for the Capital and the Conurbano. The two dummy variables equal 1 if the household pertains to a surrounding area that meets the criterion and 0 if not. Probability models (logit) and least squares regressions (OLS) were run on the variables that reflect the insertion in the labor market of individuals and on income both individuals and families. The models included socio-demographic attributes usually associated to differential results on these variables: age, sex, and education. The period analyzed corresponds to the first trimesters of the years 2004 and 2005. The results (not shown here) confirm that the probability of being employed among the economically active population is significantly lower for members of households that reside in precarious housing conditions. It also confirms that for employed members of these households, the odds of being employed in unstable occupations are higher. The analysis on per-capita family income included variables on household
52
GROISMAN AND SUÁREZ
composition—that is, educational level of the household head, presence and number of children, and number of employed members (see table 3.2). Educational level and numbers of those employed are positive for income, but number of children is negative. The results indicate that for both the city of Buenos Aires and for the partidos of the Conurbation, residence in urban surroundings with low piped-gas coverage indicates lower household incomes. Because Argentina experienced impressive economic growth between 2004 and 2005, we sought to test if these households in particular benefited or were harmed during this brief period. To this end we incorporated interaction variables between the lack of gas provision and precarious housing conditions and a dummy variable for the year 2005 (table 3.2). The results were insignificant for the city of Buenos Aires. In the Conurbation, there has been a significant effect on household income. The influence of the interaction term—gas deficient in the year 2005—is negative, which reflects lower incomes for the group and, as such, differential access to the benefits of economic growth. The interaction coefficient for households living under precarious housing conditions has a positive sign, indicating in this case a greater benefit. These results permit us to establish the incidence of these variables in time, albeit without a clear pattern. Table 3.2
Determinants of per capita family income (OLS regression)
Coefficients b
2004
2005
City of Buenos Aires
Constant Education of head Number of children Number of employed Gas deficient Housing deficient
6.432*** 0.240*** 20.483*** 0.315*** 20.112*** 20.016
6.406*** 0.239*** 20.459*** 0.343*** 20.176*** 20.066
Conurbation
Constant Education of head Number of children Number of employed Gas deficient Housing deficient
5.765*** 0.246*** 20.302*** 0.204*** 20.218*** 20.033
5.946*** 0.257*** 20.335*** 0.261*** 20.317*** 0.036**
City of Buenos Aires Conurbation
Pooled data 2004–2005 2005 * DEF GAS 20.121 2005 * DEF HOUSING 20.284 2005 * DEF GAS 2005 * DEF HOUSING
Note: *** Significant at 0.01. ** Significant at 0.05. Source: Authors’ calculations using data from EPH-INDEC.
20.239*** 0.400***
RESIDENTIAL SEGREGATION IN BUENOS AIRES
53
Conclusions What is the pattern of residential segregation in Greater Buenos Aires? At the end of the nineties, segregation, measured both by the dissimilarity and isolation indices, increased only when estimated according to the degree of health insurance coverage. This implies that changes originated in the labor market differentially impacted households according to their location. It should be kept in mind that there was a significant deterioration in the labor market during the nineties. Unemployment rose from 6 percent in 1991 to 12 percent in 1994 and 18 percent in 2001. This phenomenon had its greatest effect on wages, employment possibilities and job quality, particularly for the lower skill levels. On the other hand, residential segregation according to socioeconomic status—measured by the educational attainment of the head of household— was stable during the period, even though during the extremes of the period, an overall increase in the educational level of household heads occurred. The spatial analysis appears to confirm, then, that the elevated residential segregation that the metropolitan region exhibited in 1991 persisted during the nineties. Furthermore, the configuration of clusters based on the socioeconomic characteristics of households has increased. These results highlight the necessity of deepening the study of residential segregation in Argentina, a theme underrepresented in social research. The growing—or persistence of high levels of—residential segregation in Argentina was a result of causes associated with a continuing trend in the labor market toward the exclusion of certain individuals. A substantial rise in educational levels did little to mitigate segregation. Moreover, results obtained for recent years confirm the level of impact that households’ urban surroundings have on labor outcomes in a period of economic growth. This creates a challenge for policy makers to improve social outcomes, especially with respect to urban services, but also with respect to education (Groisman and Suárez, 2008) and health. References Arriagada, C. and Rodríguez, J. (2003). Segregación residencial en áreas metropolitanas de América Latina: Magnitud, características, evolución e implicaciones de política. Serie Población y Desarrollo 47, CEPAL. Catenazzi, A. and Fernández Wagner, R. (2004). Gestión local del hábitat, en curso de posgrado Desarrollo local en áreas metropolitanas—Módulo 5. Universidad Nacional de General Sarmiento. Cerrutti, M. and Grimson, A. (2004). Buenos Aires, neoliberalismo y después. Cambios socioeconómicos y respuestas populares. The Center for Migration
54
GROISMAN AND SUÁREZ
and Development. Working Papers Series. Princeton University (CMD 04–04d-July). Cortés, R. and Groisman, F. (2004). Migraciones, mercado de trabajo y pobreza. El área metropolitana en los noventa. Revista de la CEPAL, 82, 173–91. ——— (2005). Determinantes de la participación de migrantes en el mercado de trabajo del Gran Buenos Aires (1994–2004). Revista Latinoamericana de Estudios del Trabajo, 17, 55–74. Fernández Wagner, R. (2005). Interview conducted by the newspaper Clarin. October 5. Garay, A. (2004). Dimensión territorial de lo local. En curso de posgrado Desarrollo local en áreas metropolitanas—Módulo 2. Universidad Nacional de General Sarmiento. Groisman, F. and Suarez, A. L. (2008). Segregação residencial e conquistas educacionais na Argentina In: Luiz Cesar de Queiroz Ribeiro and Ruben Kaztman (orgs.) A cidade contra a escola?: Segregação urbana e desigualdades educacionais em grandes cidades da América Latina. Rio de Janeiro: Letra Capital, 33–58. Massey, D. and Denton, N. (1988). The Dimensions of Residential Segregation. Social Forces 67(2), 281–315. Thuillier, G. (2005). El impacto socio-espacial de las urbanizaciones cerradas: el caso de la Región Metropolitana de Buenos Aires. EURE—Revista Latinoamericana de Estudios Unbano Regionales, 31(95), August, 5–20. Torres, H. (1993). El mapa social de Buenos Aires (1940–1990). Buenos Aires Universidad de Buenos Aires: Dirección de Investigaciones. Secretaría de Investigación y Postgrado. Facultad de Arquitectura, Diseño y Urbanismo, Buenos Aires. ——— (2001). Cambios socioterritoriales en Buenos Aires durante la década de 1990. EURE (Santiago), 27(80), 33–56. Wilson, W. J. (1987). The truly disadvantaged. The inner city, the underclass, and public policy. Chicago and London: University of Chicago Press.
4
Urban Governance and Intra-Urban Population Differentials in Latin America: A Case Study of Metropolitan Lima, Peru Paul A. Peters
Introduction Latin American megacities have developed within a specific set of cultural, social, and economic conditions that continue to exert tremendous influence over the daily lives of individuals and families within them (Gilbert, 1996). The combination of inequitable economic opportunities, unequal social divisions, formal and informal urban development patterns, and weak land-use planning create divergent and highly localized patterns of spatial segregation within the major cities of Latin American (Sabatini, 2003). Other physical geographic influences (such as abrupt mountain features and coastal boundaries) also exert considerable influence on the shape of the urban environment in Latin America, with many of the poorest neighborhoods located in environmentally degraded and marginal areas. As a result of these and other urban development factors, large divisions between the elite and poorer classes have emerged in Latin American megacities. Despite these important considerations and the wealth of theoretical literature on the region, little empirical evidence has attempted to systematically capture the extent and magnitude of socio-spatial segregation in Latin American megacities. While detailed ethnographies and community studies document the ways in which the poor are funneled to certain
56
PETERS
neighborhoods, while the rich separate themselves in gated communities, few broad-scale urban spatial analyses exist to confirm the scope of this segregation in the Latin American context. Importantly, the recent emergence of detailed, geo-coded census datasets, along with the development of segregation measures that address the inherent spatial nature of segregation allow researchers a potentially fruitful opportunity for articulating the principal influences of urban structure in the processes of socio-spatial segregation. The segregation of urban residential space has important implications for the provision and administration of public services such as housing, education, and health care. Many of these services rely on the ability of local communities to contribute a sufficient local tax base and donate the necessary funds for program operations that are increasingly decentralized to municipal governments. At the same time, the ability of local areas to efficiently manage and implement programs regulates how social services are distributed across metropolitan regions. In marginal areas with low levels of socioeconomic well being, the equitable distribution of services can be challenging. Thus, in analyzing urban development patterns across a metropolitan region, linking the differentiation of social segregation measures to measures of social service programs such as education can provide insight to how local and regional governments are able to respond to the needs of the population in key areas. Education has a powerful impact on the possibilities that children and youth have to determine and enhance their individual futures, those of their families, as well as those of the communities in which they live and work (Hall and Peters, 2003). In particular, basic education levels give people the essential learning tools that empower them to survive, fully develop their potential, work and live with dignity, and make informed decisions. Most importantly, a quality basic education generally can foster a desire to continue to learn at further stages in the education system. However, the provision of a quality education is related not only to the quality of schooling itself, but is intimately linked to household well being and the quality of life in the broader community. Planning to meet the educational demands in contexts such as Metropolitan Lima requires new educational curricula, new forms of education, new techniques of needs assessment, data collection and financing, and new approaches to institutional methods of analysis (Peters and Hall, 2004; United Nations, 1992). This chapter links the measurement of population differentials and socio-spatial segregation in Metropolitan Lima with levels of school quality in the public primary school system. Previous research has examined school quality within specific areas of Lima and local patterns of segregation within the metropolitan region, although this topic remains
RESIDENTIAL SEGREGATION IN LIMA
57
understudied within Peru and Lima in particular (Hall and Peters, 2003; Peters and Hall, 2004; Peters and Skop, 2007). The analytic framework developed here addresses the relationship between urban governance, spatial segregation, and education levels within Latin America in general and Metropolitan Lima in particular. First, the process of urban development and education service provision in Metropolitan Lima is discussed. The nature of these processes is linked via the role of metropolitan and national governance structures. Second, a framework for analysis is presented along with available data and relevant measurement tools. Third, a detailed analysis of social segregation patterns is undertaken in conjunction with the quality of schools service provision. Urbanization, Governance, and Segregation Rapid urbanization and population growth of megacities in Latin America has placed intense pressure on the ability of governments to provide basic infrastructure and services and to ensure adequate employment for much of the working-age population. In concert with this dramatic growth, decentralized and weakened local governments combined with far-reaching changes in national political and economic structures have led to uncertain conditions for much of the population. The longterm economic instability of the region as a whole has contributed not only to difficulties in providing sufficient employment and housing for residents, but also to a fragmented and weak bureaucratic system of governance. This fragmentation has extended into the provision of education services in Peru, where differentials in education provision and quality differ greatly across the country and within the metropolitan region. This section outlines the links between processes of urbanization, the nature of urban governance, and spatial segregation patterns. Urbanization of Metropolitan Lima The rapid growth of Metropolitan Lima is marked by the difficulties of integrating a rapidly urbanizing migrant population into a weak national and metropolitan economic structure. The import-substitution industrialization (ISI) policies of the 1950s and 1960s promoted macro-level agrarian reforms, universal education, and increased labor benefits that together led to rapid increases in rural-urban migration and urban population growth. Within the metropolitan area, the centralized national government was challenged by the largely unplanned growth of housing in the periphery of the city and a highly segmented labor market with a
58
PETERS
high incidence of informality. These processes and policies strengthened the concentration of urban poverty and intensified social and physical fragmentation of different social groups. The growth patterns and reactionary processes of the 1960s and 1970s had a major impact on the nature of urban poverty and inequality by the early 1990s. With the implementation of massive structural adjustment policies in the 1990s by the Fujimori government, fragmentation and exclusion of select social groups became even more pronounced via political arrangements and governance structures. The role of municipal actors within the metropolitan area was reinforced, leading to a fragmented pattern of regulation and governance within the metropolitan region. Implementation of national-level policies was downloaded onto local governments and regional councils, many of which were unprepared for the additional financial burden and technical responsibilities. These processes of change strongly influenced both the nature of urban development within Metropolitan Lima, which was already highly segregated, and the nature of social service provision for urban residents. While wealthier residents have established exclusive “citadels” as a means of protecting and enhancing their positions of power, wealth, and status, those in poverty have often been left to reside in degraded informal settlements, squatter housing, and urban tenements in both the historic center and the outer periphery. The settlement of the poor is complicated by an unpredictable and inequitable land market that discourages formal regularization and promotes a continuance of informal market processes. Though many of Lima’s squatter settlements established in the past half-century have gradually consolidated to become improved secondgeneration housing, a large proportion of settlements are composed of quasi-legal or illegal housing (Calderón, 2004). In general, the poor have been left to fill in the “gaps” in the urban environment, often in the most precarious and squalid of physical environments. The physical expansion of Metropolitan Lima, largely via informal means, has transformed the urban structure from one modelling European compact-cities to one intensely fragmented along multiple social and physical dimensions (Joseph, 2005). These tendencies and economic liberalism have exacerbated the polarization tendencies of the ciudad rica and the ciudad pobre, with segregation rising at the local level while not drastically changing at the macro-level (Janoschka, 2002; Peters and Skop, 2007; Sabatini, 2006). The interaction between processes of social change at the local level and economic processes at the macro level are contributing to an increasingly fragmented and privatized urban space. This restructuring is highlighted spatially by an increasing presence of gated communities, shopping malls, country clubs, and industrial parks in concert with
RESIDENTIAL SEGREGATION IN LIMA
59
an increasingly dynamic and mobile population (Borsdorf, 2003; Dear and Flusty, 1998; Sabatini and Cáceres, 2004). In addition to the fragmentation of physical space, social-service provision is also becoming highly differentiated between and within urban areas (Peters and Hall, 2004). Access to high-quality basic education depends not only on the provision of infrastructure and resources by the central Ministry, but also on the socioeconomic well being of local communities and the resources of local municipalities. The evident inequity in education service provision, sociospatial inequality, and processes of urban governance further illustrate the issues for broader social development schemes.
Urban Governance and the Provision of Education Primary school completion rates in Peru between 1985 and 2003 have improved vastly for youths between the ages of 11 and 13 and between 14 and 16 years of age from 53.9 percent to 72.5 percent and from 64.5 percent to 91.2 percent respectively (Peters and Hall, 2004). Females showed a particularly sharp rise in their completion rates by the time they were 13, moving from 51.8 percent in 1985 to 73.2 percent in 2003. The proportion of the population considered extremely poor also showed a marked improvement in both categories of youths in this time period, moving from 27.5 percent to 53.7 percent by age 13 and from 35.6 percent to 78.4 percent by age 16. However, strong differences are visible between children classified into the three different socioeconomic categories. Despite the higher overall completion rates in 2003, only 53.7 percent of those in extreme poverty complete primary school by the time they are 13 years of age, and 78.4 percent by the time they are 16 years of age. Despite these differences, the majority of young Peruvians have completed primary school, thus suggesting that this particular measure of education indicates relative homogeneity among Lima’s residents. The central Ministry of Education is the primary agency responsible for education provision in Peru. Education policy, core curriculum, and technical support are all provided centrally with subregional directorates that report directly to the national office. While education funding has increased gradually over time, it is generally unstable and insufficient to provide the necessary funds to support education at a local level. As such, private schooling has been used to fill that gap with the number of approved private schools steadily increasing through the 1990s. The number of private schools is strongly correlated to the wealth of an area, with areas of higher socioeconomic status having a higher proportion of children attending private schools.
60
PETERS
Beyond the visible physical separation, the segregation of social class in Metropolitan Lima is manifested across a variety of economic, social, and political dimensions. In particular, the quality and availability of education services in Lima mirrors the general patterns found by the physical separation of different social groups. Those households with the means to provide a quality education for their children often do so in the growing private school system. In response to widespread demand, many private schools, particularly for early childhood and primary education, have opened in poorer areas of the city, taking advantage of the often mistaken perception that a private education is of higher quality. Parents use their often-limited financial resources to school their children outside of the public system. As such, the public system is dominated by those families without the means for a private education, largely living within the poorer areas of the city. It is generally understood that even in terms of public education; those schools located in better-off neighborhoods perform better across a variety of measures (Carlson, 2000). Differences include the level of funding by the Ministry and the respective abilities of neighborhoods to provide additional services to individual schools. Primary education, although mostly funded by the state, is not entirely free in Peru, and parents are often asked to contribute money for uniforms, classroom materials, infrastructure upgrades, and even teacher salaries. Within communities where income levels are lower and the combined means of parents and their families are reduced, these additional and unaccounted means of school survival become difficult (Peters, 2002). In areas where the poor are highly segregated, these factors are only further exacerbated. As such, the contextual effects of poverty on educational attainment may become even more exacerbated. The ability to discern separate and joint effects of the relationship between social processes and spatial location is difficult. In relating educational attainment and spatial segregation patterns, the difficulty lies in attributing specific factors to outcomes in educational attainment and local context. Although it is well known that pupils from deprived areas have lower mean educational attainment levels than pupils from more advantaged areas, it is unclear what the direct effect of living in a deprived area is over other factors such as individual ability, family circumstances, and schooling characteristics (Flowerdew and Pearce, 2001; Garner and Raudenbush, 1991). The extent to which neighborhood contextual factors either compound or contribute to a child’s educational development is unclear (Peters, 2002). The following section advances a methodology for analyzing the relationship between socio-spatial segregation and patterns of educational attainment.
RESIDENTIAL SEGREGATION IN LIMA
61
Framework for Linking Segregation and Educational Attainment Although most of the poor tend to be concentrated in specific areas of the Latin American megacity, and this is certainly the case for Lima, the continued fragmentation of urban space and increasing inequality in public services complicate neat generalizations and sweeping conclusions. In other words, local realities and external influences confound conventional explanations of the dynamics of socio-spatial segregation. Over time it has become apparent that urban development patterns and levels of sociospatial fragmentation are not occurring in the same manner between and within different urban settings. Unfortunately, the empirical story regarding the geography of urban spatial segregation and urban governance in the Latin American megacity is incomplete. Given this, a combination of segregation indicators, education measures, and empirical techniques are used to provide a broader insight into urban development patterns at multiple levels with regards to school quality at the local level. There are a multitude of spatial methods and measures available, either spatial extensions of traditional measures, explicit spatial formulas not based on common segregation indices, or spatial techniques employed to analyze spatial segregation and differentiation (Wong, 2002). However, despite the development of many spatial measures of segregation, few are utilized in specific segregation studies, thus, there is little evidence to confirm their importance (Reardon and O’Sullivan, 2004; Wong, 2003a). To address this shortcoming, this analysis combines several aspatial and spatial multigroup and multilevel measures of segregation across several dimensions. Socio-spatial segregation is measured by using a combination of education variables and a composite indicator of socioeconomic status (SES). First, segregation is calculated using both relative and absolute education levels. Relative levels of education are calculated by the number of individuals attaining levels of schooling above or below the national average of seven years, one year above primary school completion. Due to data limitations, primary school completion is used here. Absolute levels of education are the number of adults above or below a secondary education level. For multigroup measures of educational segregation, five classes are used: less than primary, primary, complete primary, complete secondary, some post-secondary and completed university. Second, segregation is calculated using the national index of poverty as calculated by the Instituto Nacional de Estadísticas y Informática (INEI). This index calculates poverty based on several social and economic dimensions, although the specific calculation is not made public (Escobal, Máximo, and Carmen, 2001). For multigroup measures,
62
PETERS
poverty is calculated using the available five classes for low, medium-low, medium, medium-high, and high. For two-group measures, poverty is calculated as those individuals in the lowest three categories as compared to those in the highest two.
Available Data Census data for this paper are from the 1993 Peruvian Census of Population and Housing (IX Censo de Población y IV de Vivienda), which is the most recent complete census for Peru conducted at the household and individual levels (INEI, 1993). The complete dataset is available aggregated to the city-block level and calculations of spatial segregation measures at the micro level are possible. A partial release of census micro-data variables and complete spatial data were obtained on a CD provided by the INEI in 2003. The 2005 partial census, made available publicly from the INEI, is also used for the calculation of macro-level segregation at the district level and comparison to more current figures. Given the length of time since the release of the last complete census available at the local level, the 1993 data are supplemented in this chapter through the addition of the 2005 partial census, the 2002 ENAHO (Encuesta Nacional de Hogares) survey at the district and regional levels (INEI, 2002), and education-related data from the Ministry of Education. While detailed indicators of segregation cannot be calculated using these supplemental surveys, analysis of distributions and differentials in education levels are useful in measuring the change of educational attainment in the last decade. School quality is calculated using a combined indicator from aspects of the available data and based on the broader institutional goals identified by the Ministry of Education in their country-specific documents, Education for All (Ministerio de Educación del Perú, 2000; Peters and Hall, 2004). Ten school characteristics relating to the dimensions of education systems (inputs, processes, outputs, outcomes, and context) are identified and calculated for all public primary schools in Metropolitan Lima (Peters and Hall, 2004). Each of the indicators is calculated individually and then combined using equal weighting to create an index ranging between 0 and 1, where 0 is the lowest possible quality, and 1 is the highest possible. Each indicator is provided a target value, or 100-percent weight; a minimal value, or 0-percent weight; and, an acceptable range of values. The target values represent minimal values that are considered acceptable by local education planners and overall Ministry of Education goals (Hall and Peters, 2003). The indicators are the number of computers
RESIDENTIAL SEGREGATION IN LIMA
63
available for students to use; the presence of a library; access to reading materials; student-to-teacher ratios; teacher-to-classroom ratios; the gender ratio for students; infrastructure quality in terms of deficiencies in public water connections, electricity, sewage, or building construction; the promotion rate of students; and, finally, the percentage of teachers with post-secondary teaching certificates. The lowest geographic level of analysis for measuring segregation via the 1993 census is at the census-zone level with census micro-data aggregated by city block. Most measures of segregation rely on measuring the population differentials between aggregate groups within given geographic boundaries. Thus, whereas data are available at the city-block level, these data are grouped within census zones, which are subsets of administratively defined municipal districts set forth by the national government. In general, census-zone boundaries do not change greatly from census to census; however, given the rapid urbanization of Lima in the last decades of the twentieth century, this has not always been the case. In an effort to remain consistent across time, spatial data identifiers from the 2000 census update and spatial data files were recalculated with 1993 identifiers using an available equivalency file. Segregation measures were calculated using a tightly coupled analysis program within the GIS software package ArcView 3.3 and ArcGIS 9.2 (Peters, 2005). This software allows the linking of population data drawn from census files to spatial data layers within the GIS user interface. Multiple analyses can be performed on the same base layer, using different levels of geographic aggregation to calculate segregation. For example, using the city-block layer, segregation can be calculated at the census zone, district, region, or metropolitan levels. Thus, in order to account for changing values of segregation indicators at the different spatial levels, results can be compared side by side and the most reliable outputs selected (Wong, 2003a). Additionally, outputs of each analysis can be easily mapped using the GIS software package and visually compared to expected results.
Socio-spatial Segregation Patterns The patterns of spatial segregation in Metropolitan Lima, as in other Latin American cities, vary widely depending on which aspect of segregation is being measured and at which spatial scale the analysis is performed (Peters and Skop, 2007; Sabatini, 2003). The nature of segregation in Lima is highly fragmented, with vast, expansive homogeneous areas with similar population characteristics and small fragmented areas with marked differences.
64
PETERS
Given the overall size of the urban population and the large geographic area, most measures of segregation fail to adequately capture the extreme levels of differentiation between social groups. The four conal regions of Lima (north, city center, east, and south) show relatively small differences in their overall measures of educational attainment. The percentages of those with the two extreme educational levels—less than primary schooling and more than secondary schooling—are 22.7 and 9.4 for the north, 21.3 and 5 for the city center, 26.6 and 8.5 for the east, 17.6 and 12.1 for the south. The segregation values presented in tables 4.1 and 4.2 display the calculated levels of segregation for relative and absolute levels of education in Metropolitan Lima and in each of the five regions. As is expected, the segregation values for these indicators are generally low, although the exposure index (xPy) is high in the case of absolute education. This indicates that in general, the population differentials at the metropolitan level are relatively homogeneous. The high values at the metropolitan level for the exposure index with absolute education support this, where individuals without a secondary school diploma generally live close to those without a secondary diploma. Still, the overall residential dissimilarity between social classes in Lima is readily apparent. One reason for this is finding is that the absolute level of education in Lima is high, with a large proportion of the population completing primary school and obtaining a secondary school diploma. However, education outcomes are not translated into higher levels of SES for all sectors of the population. It is the finding Table 4.1
Metropolitan-level segregation, 1993
Indicator
Education (absolute)
Education (relative)
Education (categories)
SES
0.33 0.70 — —
0.19 0.37 — —
— — 0.25 0.41
0.52 0.48 0.41 0.79
D xPy
SD(m) S
Table 4.2
Regional-level segregation by education and SES levels, 1993
Region
Southern Cone City Center Northern Cone Eastern Cone Callao
Relative
Absolute
Multi-group
SES
D
xPy
D
xPy
SD(m)
S
SD(m)
S
0.13 0.18 0.14 0.16 0.16
0.42 0.25 0.40 0.41 0.35
0.26 0.24 0.26 0.31 0.26
0.79 0.57 0.77 0.75 0.73
0.25 0.16 0.23 0.28 0.18
0.40 0.47 0.47 0.51 0.56
0.35 0.31 0.34 0.40 0.27
0.88 0.82 0.73 0.98 0.91
RESIDENTIAL SEGREGATION IN LIMA
65
that supports the weak results for education differentials through standard segregation measures. Differences become more apparent when segregation is analyzed at lower geographic levels. In order to discern the degree of segregation at different levels, indicators of segregation must be calculated at different levels of aggregation, using available census data and comparable measures. Whereas the results above were aggregated at the metropolitan and regional levels (using census zones as the unit of analysis), segregation measures can also be aggregated at the municipal district, census zone, or school catchment area levels. At the regional level, the segregation results are more mixed. Values for most measures and indicators remain low. The exception for education is the exposure index for absolute education levels. These values continue to suggest that for most regions, with the exception of the city center, those without a primary education do not live close to those above this level. Expanding on these results, the SES indicator is measured for the same regions using multigroup spatial measures. For the SD(m) measure (a spatial dissimilarity calculation that accounts for the nearest neighbor) segregation values continue to be low. However, the S-index, which is an explicitly spatial measure, displays high levels of segregation in all metropolitan regions. This difference suggests that the segregation of SES is largely spatial in Metropolitan Lima, and the analysis via tabular methods may mask important spatial processes. The above indicators combine to suggest that the levels of educational attainment within the metropolitan region are both highly fragmented and relatively unbroken at the same time. In other words, educational attainment levels remain diverse across much of the metropolitan area. The city center and surrounding areas include large proportions of the population with similar education levels. In contrast, the periphery of the city, as exhibited in the Northern Cone, has higher levels of segregation and isolation with lower levels of educational diversity. Given the historic development patterns by the different classes within the metropolitan area, this patterning makes sense. The lower portion of map 4.1 displays the S-index of spatial congruence. This measure produces standard deviational ellipses that correspond to the spread and direction of population groups. In this case, the spatial distribution of the five population groups for SES and educational attainment are displayed by region within Metropolitan Lima. For each region, segregation can be interpreted by analyzing the spread and direction of five ellipses. For regions with high levels of segregation, there is little spatial overlap of the ellipses, whereas for regions with lower segregation there is more spatial overlap. The ellipses displayed here correspond to the values
66
PETERS
Map 4.1 Local measures of spatial auto correlation for the Northern Cone of Lima and spatial congruence for Metropolitan Lima, SES and absolute education, 1992
of the S-index reported in table 4.2. The most segregated region appears to the eastern cone, with the northern cone appearing less segregated. In comparing SES and education, it is apparent that the spatial congruence between the five SES groups is much less than for the five education groups. This is supported by the much lower values of the S-index. To further the analysis, the local Moran’s I was calculated for relative levels of educational attainment and SES. This measure calculates the degree of clustering that would be expected for levels of education either above or below the global spatial distribution. In this case, the relative indicator used is for the percentage of the population without a secondary school diploma, with the inverse being those with a secondary diploma. In the top portion of map 4.1, the clustering of this variable is easily apparent. Using the Moran’s I index values displayed by standard
RESIDENTIAL SEGREGATION IN LIMA
67
deviation, the areas of highest clustering are along the boundary between the city center and the Northern Cone, in the city center near the Eastern Cone, and in the south-central part of the city center. The darker-shaded areas correspond to those areas with higher than expected index values, or a greater clustering of individuals without a secondary school diploma. These values are to be expected, as this area is a strong division between better-off areas in the city center and poorer areas in the northern cone. Likewise, the slight clustering of lower than expected values in the city centre corresponds to middle-to upper-class neighborhoods where higher education levels would be expected. Overall, the Moran values for the metropolitan area are quite high, with little probability of this distribution being random.
Patterns of Education Quality and Segregation As with the patterns of spatial segregation described above, there are strong differences in school quality across the metropolitan region and within local areas. Many of these differences are related to the abilities of local governments and communities to provide adequate funding and support for schools, but aspects are also related to the contextual environment in which the school is located. This section focuses in greater detail on the northern cone of Lima, using similar techniques as those presented above. According to the index of school quality, overall variation between school districts is quite high, with some districts having mean quality levels below 5.0, while others have mean levels above 8.1. In contrast to the regional-level distribution of educational attainment where the Southern Cone had the most favorable distribution in terms of the smallest percentage with less than primary and the highest percentage with post-secondary, this cone appears to have the lowest mean levels of school quality. This may relate to the high levels of poverty found in this region, where students attending public primary schools most likely will be from poorer families and neighborhoods. Children from better-off families and neighborhoods would thus be likely to attend private schools and would not be reflected in these statistics. The greatest variation in mean education quality between districts appears to be in the city center, which contains some of the highest and lowest values. These results suggest that while some municipal districts are performing well in providing high-quality education for students, others are lagging behind in key areas. Relating these results to the analysis of segregation, some of the highest levels of diversity in educational
68
PETERS
attainment levels are also found in this area, meaning that there are large numbers of individuals from all education backgrounds. Thus, it is likely that within these areas there are strong differences in where families can send their children to school, with some very poor neighborhoods limited in options for high-quality schools, with better-off neighborhoods having higher-quality schools as alternatives. In addition, those families with the financial means will often choose to send their children to private schools, presumably of higher quality than the local public schools. Perhaps most telling of the state of public schools across the region is the percentage of students attending private primary schools. Across the metropolitan region the percentages vary between less than 5 percent and more than 50 percent or students outside the public system, having broad implications for the effective and efficient provision of a quality education for all students. Generally, the lowest percentages of students in private school are in the northern cone and southern cone, where the vast majority of students would thus be within the public system. However, within the city center, some districts have more than 50 percent of primary school students attending a private institution, presumably leaving the remainder of those to attend the nearest public school. The analysis (not included here) of the distribution of public schools and educational quality shows that the districts in the city center with lower mean education quality levels correspond to those with higher proportions in private school. Where private school attendance is high, public schools may not receive the community support required to ensure a high quality education. Conversely, for those areas where the majority of students attend public school, there is a strong community incentive to demand quality education. In planning for public education services then, it is necessary to not only evaluate the quality of schools within the public system, but also to address the differences between public and private provision. The description above furthers the earlier assertion in this paper that although education levels across Metropolitan Lima appear to be relatively uniform, key variables relating to the quality and efficiency of the education system display large differences at both regional and local levels.
Conclusions The primary goal of this paper was to examine population differentials and socio-spatial segregation in metropolitan Lima, relating these patterns to educational attainment. The nature of segregation in Lima is fragmented, with expansive homogeneous areas with similar population characteristics at the metropolitan level and more fragmented areas of marked difference
RESIDENTIAL SEGREGATION IN LIMA
69
appearing at subregional levels. Combining spatial statistical measures of residential segregation with analysis of educational attainment, it is shown that strong differences are indeed apparent across the metropolitan region, having implications for how local and regional governments are able to respond to the needs of the population in key areas. The geographic patterns of social segregation in Metropolitan Lima are mixed, displaying differing results depending on the scale of analysis and the variables selected for differentiation. In general, segregation values at the metropolitan level are low compared to cities in North America or Europe; however, the values correspond to similar results found in other Latin American cities (Sabatini, 2003). Across most of the variables selected, segregation at the metropolitan level was relatively low, with values for education levels and employment types not exceeding 0.2 for most measures. The combined index of SES does suggest higher levels of segregation, with values near 0.5, although this is still not considered extreme. In relation to education, attainment levels are uniform across much of the metropolitan area, with the majority of adults obtaining a primary education, and few completing secondary school. The population in the northern cone and periphery of other areas display lower attainment levels than the population in the city centre. However, high levels of spatial segregation become apparent at regional and local levels when examining specific aspects of education service provision. While educational attainment levels from the census do not appear to have high levels of segregation at either the metropolitan or local levels, education quality and service provision are highly segregated. Using a combined index of education quality, differences between municipal districts was measured, with the results suggesting that in some areas, schools are performing far better on average than others. Using individual school points for the northern cone, these assertions are supported by strong visible differences in the locations of high-performing schools and those with lower measured school quality. In particular, the differences in school quality appear to be related to the proportion of students attending private school, where districts with higher private-school attendance have lower average school quality. This result is of importance in how local governments respond to the educational needs of those students within the public school system. Public education at the primary level relies heavily on an active local population of parents with the ability to contribute to the education of their children, often through informal means. In marginal areas with low levels of socioeconomic well being, or in better-off areas where only students from the poorest families attend public schools, the provision of education is challenged by a general lack of resources within the community.
70
PETERS
Thus, in analyzing the urban development patterns across a metropolitan region such as Lima, linking spatial segregation patterns to social service outcomes can provide important insight into how local and regional governments are able to respond to the needs of the population. References Borsdorf, A. (2003). Cómo modelar el desarrollo y la dinámica de la ciudad latinoamericana. Eure-Revista Latinoamericana De Estudios Urbano Regionales, 29(86), 37–49. Calderón, J. (2004). Accesso por los pobres al suelo urbano y mercados inmobiliarios en Lima Metropolitana. Cambridge, MA: Lincoln Institute of Land Policy. ——— (2004). The formalisation of property in Peru 2001–2002: The case of Lima. Habitat International, 28(2), 289–300. Carlson, B. A. (2000). Achieving educational quality: What schools teach us, learning from Chile’s P900 primary schools. Desarrollo Productivo, 64 (diciembre) Santiago: ECLAC. Dear, M. and Flusty, S. (1998). Postmodern urbanism. Annals of the Association of American Geographers, 88(1), 50–72. Escobal, J., Máximo, T., and Carmen, P. (2001). Focalización geogáfica del gasto social: Mapas de pobreza. Economía y Sociedad, 43, 11–17. Flowerdew, R. and Pearce, J. (2001). Linking point and area data to model primary school performance indicators. Geographical and Environmental Modelling, 5(1), 23–41. Garner, C. L. and Raudenbush, S. W. (1991). Neighbourhood effects on educational attainment: A multilevel analysis. Sociology of Education, 64, 251–62. Gilbert, A. (1996). The megacity in Latin America. Tokyo: United Nations University Press. Hall, G. B. and Peters, P. A. (2003). Global ideals and local practicalities in education policies and planning in Lima, Peru. Habitat International, 27(4), 629–51. Instituto Nacional de Estadísticas y Informática (INEI). IX Censo de Población y IV de Vivienda. (1993). Lima, Peru: Instituto Nacional de Estadísticas y Informática. ——— (2002). Encuesta Nacional de Hogares (ENAHO). Lima, Peru: Instituto Nacional de Estadísticas y Informática. Janoschka, M. (2002). The new model of the Latin American city: Fragmentation and privatization. EURE—Revista Latinoamericana de Estudios Urbano Regionales, 27(85), 11–29. Joseph, J. A. (2005). La ciudad, la crisis y las salidas: Democracia y desarollo en espacios urbanos meso. Lima, Peru: Universidad Nacional Mayor de San Marcos. Ministerio de Educación del Perú. (2000). Educación para todos 2000 Perú: Informe nacional de educación. Lima, Peru: Ministerio de Educación del Perú.
RESIDENTIAL SEGREGATION IN LIMA
71
Peters, P. A. (2002). Assessing education quality in Peru: The role of information in education planning and management. Master of Environmental Studies’ thesis, School of Planning, Faculty of Environmental Studies, University of Waterloo, Waterloo. ——— (2005). Fragmentation of urban space in Latin America: A GIS approach to the analysis of segregation in Lima. In R. Garcia Alvarado, K. Abarca Cárdenas, I. López Meza, and R. Biere Arenas (eds.), Libro de ponencias: 2ndo Congreso Internacional Ciudad y Territorio Virtual, pp. 92–96. Concepción, Chile: Ediciones Universidad Bio. Peters, P. A. and Hall, G. B. (2004). Evaluation of education quality and neighbourhood well-being: A case study of Independencia, Peru. International Journal of Educational Development, 24(1), 85–102. Peters, P. A. and Skop, E. H. (2007). The geography of poverty and segregation in Metropolitan Lima, Peru. Journal of Latin American Geography, 6(1), 149–71. Reardon, S. F. and O’Sullivan, D. (2004). Measures of spatial segregation. Sociological Methodology, 34(1), 121–62. Sabatini, F. (2003). La segregación social del espacio en las ciudades de América Latina. In Serie Azul. Santiago, Chile: Instituto de Estudios Urbanos y Territoriales, Universidad Católica de Chile. ——— (2006). The social spatial segregation in the cities of Latin America. Washington, DC: Inter-American Development Bank (IADB), Sustainable Development Department, Social Programs Division. Sabatini, F. and Cáceres, G. (eds.). (2004). Barrios cerrados: Entre la exclución y la integración residencial. Santiago, Chile: Instituto Geografia, Pontifica Universidad Católica de Chile. Sabatini, F., Wormald, G., and Contreras, Y. (2003). Informe de Avance. La guerre de la basura de Santiago: Del derecho al suelo y la vivienda al derecho a la ciudad. Santiago, Chile: Pontifica Universidad Católica de Chile. United Nations. (1992). Agenda 21: Programme of action for sustainable development. New York: United Nations. Wong, D. W. S. (2002). Modelling local segregation: A spatial interaction approach. Geographical & Environmental Modelling, 6(1), 81–97. ——— (2003a). A comparison of traditional and spatial measures of segregation: Some empirical findings. In J. W. Frazier and F. M. Margai (eds.), Muticultural geographies: The changing racial/ethnic patterns of the United States, pp. 247–62. Binghampton, NY: Global Academic. ——— (2003b). Spatial decomposition of segregation indices: A framework toward measuring segregation at multiple levels. Geographical Analysis, 35(3), 179–94.
This page intentionally left blank
5
Residential Segregation in the Mexico City Metropolitan Area, 1990–2000 Andrés Villarreal and Erin R. Hamilton
T
he concentration of poverty in large U.S. cities is an issue of great public and academic concern. Understood as the outcome of various processes, including the restructuring of the economy and loss of employment among low-skilled workers (Wilson, 1997), the movement of manufacturing firms away from the city centers (Kain, 1968), and discriminatory housing policies restricting the location of low-income minority groups, concentrated poverty in inner-cities is considered an important social problem facing many large urban centers in the United States. Neighborhood poverty has been associated with a broad range of negative social outcomes such as high school dropout rates and teenage pregnancy (Harding, 2003), adult depression (Ross, 2000), and domestic violence (Cunradi et al., 2000), among others. Compared to the vast literature on segregation and neighborhood effects in the United States, relatively little is known about the spatial distribution of poverty in Latin American urban areas and its consequences for individual well-being. In this chapter we examine recent trends in socioeconomic segregation in Mexico City, one of the largest urban centers in the Western hemisphere. We use census-tract and city-blocklevel data to measure segregation along various dimensions identified by Massey and Denton (1988). In the final section of the chapter, we also consider how spatial segregation may affect the employment opportunities of residents of disadvantaged neighborhoods in the city. Specifically, we assess the spatial mismatch hypothesis, according to which the distance
74
VILLARREAL AND HAMILTON
between disadvantaged neighborhoods and high-paying manufacturing firms restricts the employment opportunities of the urban poor.
Population Growth, Physical Expansion and Socioeconomic Segregation in the Context of Mexico City Mexico City experienced rapid growth through the middle decades of the twentieth century. In 1950, the population of Mexico City was slightly more three million, with an annual growth rate of over 5 percent (Garza and Schteingart, 1978). At this rate, which was more or less sustained through 1970, the population of the city almost doubled every ten years between 1940 and 1970. In 1970, more than 9.2 million people lived in Mexico City, but by then the annual growth rate had declined to an average of about 2.3 percent between 1970 and 1990, when the population reached 14.7 million (Ward, 1998). Mexico City’s population growth led to an outward expansion of the city, particularly towards the north and the southeast, as volcanic mountain ranges block expansion directly west, east, and south. The fastest growth in Mexico City in the 1950s and 1960s occurred in the state of Mexico; during this time, the population of the Federal District grew at an annual rate of just over 5 percent, while the population of state of Mexico increased by over 20 percent annually (Garza and Schteingart, 1978). Indeed, between 1950 and 1970, Mexico City’s geographic area doubled from 29,000 hectares to 60,000 hectares (Delgado, 1990). Similarly, Ward (1998, table 2.2) observes a direct relationship between growth rates and distance from the city center across time. While natural increase composed the majority of population growth in Mexico City in the second half of the twentieth century, Mexico City’s outward expansion was made up mostly of rural migrants. For example, while more than half of the population growth in the Federal District in the 1950s and 1960s was due to natural increase, a full three-fourths of the growth in the state of Mexico during the same time was due to the irregular settlement of low-income migrants in and around the Texcoco lakebed (Garza and Schteingart, 1978). To some extent, center-city residents also migrated out towards the periphery as rental units were transformed into commercial and service enterprises, but this migratory pattern made up a very small part of the outward expansion (Coulomb, 1984, as cited in Delgado, 1990). The contemporary spatial distribution of Mexico City’s poorest residents has been conditioned by the pattern of settlement of rural migrants during the second half of the twentieth century (Rubalcava
RESIDENTIAL SEGREGATION IN MEXICO CITY
75
and Schteingart, 1987; Schteingart, 1989; Delgado, 1990; Ward, 1998). Contrary to the traditional pattern of settlement of migrants in U.S. cities described by Chicago School theorists, according to which migrants settle in the central parts of the city and move outwards as their economic condition improved (Park and Burgess, 1925), successive waves of rural migrants to Mexico City historically have tended to settle in tenements, often in illegally occupied lands in the outer rings of the city, despite the fact that their peripheral location often results in a physical separation from sources of employment. This divergent pattern results for several reasons. John Turner (1968) developed a theoretical model of urban expansion suggesting that in late transitional cities in Latin America, such as Mexico City in the second half of the last century, poor migrants have settled in the periphery because the inner-city and its immediately surrounding areas have developed to such a point as to economically prohibit low-income settlement. Ward (1998) confirms this process in Mexico City, adding that government rent controls in inner-city housing in Mexico City and social networks connecting migrants to housing on the periphery interrupt the classic model of urban migrant settlement as well. Indeed, land and housing markets heavily influence the settlement of migrants in Mexico City as well as the overall socioeconomic distribution of the city’s population. Public housing programs, namely the Institute of the National Fund for Worker Housing (INFONAVIT) and the Public Employees Housing Fund (FOVISSSTE), were largely ineffective at reshaping the basic pattern of low-income peripheral settlement, as they served formal, blue-collar workers and middle-income state employees, and did not address the housing needs of the most poor (Ward, 1998). The National Popular Housing Fund (FONHAPO) was developed in the early 1980s to address this oversight, but its services have only reinforced the peripheral settlement of the poor through the development of affordable housing on the outskirts of the city and the facilitation of self-help housing already located there. Public housing programs thus left most housing needs largely unmet in Mexico City, and low-income settlers resorted to land invasions, which were essentially ignored through 1970, or illegal purchase, which the government overlooked (Ward, 1998). A ban on the development of low-income subdivisions within the Federal District, beginning in the 1950s, encouraged low-income settlers to purchase land and construct their own housing on affordable but undesirable plots in poor locations in the outer parts of the city. These transactions were largely illegal, either because subdivision developers did not meet basic servicing requirements, or full land titles were not given. In many cases, settlers illegally purchased land from agrarian collectives known as
76
VILLARREAL AND HAMILTON
ejidos. Until the reform of the agrarian law in 1992, ejido members were forbidden from selling their land to private investors. The historical experience of growth and expansion in Mexico City has been described using the term contornos, or surrounding areas, in order to differentiate the process from the idea of concentric rings, developed to describe U.S. city growth (Schteingart, 1987). As the city grew and expanded, peripheral areas became incorporated into the concentrated inner area, reaching a certain level of urban consolidation and stability at the same time that new peripheral areas were filled by in-migrants. Prior studies of segregation in Mexico City have generally used descriptive maps to show the distribution of socioeconomic inequality across the city, and these studies have shown that the outer contornos of Mexico City are indeed populated by the poorest sectors of the population, resulting in a center-periphery pattern of segregation: the wealthiest sectors residing just southwest of the city’s center in areas marked by greater socioeconomic heterogeneity, while the poorest concentrate on the outer edges of the city in homogenously poor areas (Rubalcava and Schteingart, 1987; 2000). This review of Mexico City’s population growth, physical expansion, and corresponding segregation sets the stage for our analysis of segregation in Mexico City between 1990 and 2000. We build on this prior work by analyzing the evenness, isolation, centralization, and clustering of Mexico City using indexes developed to provide comparative measures of these conceptually different dimensions of segregation for U.S. cities. However, before proceeding, it is important to briefly review demographic and economic trends during the 1990s in Mexico City. In the 1990s, Mexico City grew at a rate of 1.58 percent, reaching a population of just over 18 million by 2000 (Bush and Gomez, 2003) with over 50 percent of its inhabitants living outside the Federal District (Graizbord et al., 2003). In terms of net migration, the Federal District lost, on average, about 100,000 people each year, while the state of Mexico gained about the same number (CONAPO, 2000). In the past two decades, Mexico City’s economy has undergone a shift from an industrial to a service base, resulting in a polarization between a very large section of the population providing informal and small-firm consumer services, on the one hand, and a very small section of the population employed in professional, high-end producer services (Ariza and Solis, 2005). Given this background, we expect to uncover several important trends and patterns in segregation in Mexico City between 1990 and 2000. Most importantly, we expect to find a decentralization of the poor and a centralization of the wealthy, an opposite pattern to that found in the United States.
RESIDENTIAL SEGREGATION IN MEXICO CITY
77
Moreover, the settlement of rural migrants in the periphery of Mexico City also leads us to expect an uneven distribution of the poor, but an even more sharply uneven distribution of the wealthy, with similar patterns of isolation. In other words, we do not expect many census tracts to be shared by both residents of high- and low-socioeconomic status. In terms of clustering, we expect higher-income strata to be clustered together towards the immediate south and west of the city, whereas lower income strata will be clustered around the city’s periphery. Because the city center has lost population while the outer area has continued to gain population, and because economic trends suggest an increasing polarization between occupational sectors, we expect that segregation will have increased slightly in Mexico City between 1990 and 2000.
Data Our analysis of income segregation in the Mexico City metropolitan area is based on data from the 1990 and 2000 Mexican population censuses carried out by the National Institute of Statistics, Geography, and Informatics (INEGI). The censuses contain detailed information on the income and education levels of all adults, as well as various measures of the quality of housing and access to services, such as water and electricity, which may be used to estimate the level of poverty at the community level. INEGI provides tabulated data for all census tracts (known as Areas Geoestadísticas Básicas, or AGEBs) located in urban areas throughout the country. These tabular data are sufficient to compute the various measures of income segregation described below. A standard set of 70 variables is available for the 1990 census, whereas 170 variables are available for 2000. Additional variables at the census-tract and city-block levels were also obtained from the Ministry of Social Development (SEDESOL) in Mexico. These additional variables allow us to estimate the number of households living below the recently established poverty lines for the various areal units. Our first challenge is to define the boundaries of the Mexico City metropolitan area to be used in our analysis. Until recently there has been no widely recognized definition of statistical metropolitan areas in Mexico similar to those available from the Census Bureau in the United States (Federal Register, 2000). The exact definition of Mexico City has generally been up to individual researchers’ discretion. Fortunately, the Mexican National Population Council (CONAPO) has recently undertaken the task of defining a standard set of metropolitan areas in 2000 based on the
78
VILLARREAL AND HAMILTON
aggregation of municipalities, the smallest recognized political boundaries in Mexico (CONAPO, 2004). We therefore employ CONAPO’s definition of the Mexico City metropolitan area for 2000, which consists of the 16 districts, or delegaciones that make up the Federal District, plus 59 municipalities in the neighboring states of Mexico and Hidalgo.1 Due to data limitations we are forced to define the Mexico City metropolitan area in a more limited way for 1990, using only the 16 delegaciones of the Federal District and 7 municipalities belonging to the state of Mexico.
Measurements The task of obtaining appropriate measures of disadvantage among Mexico City residents is a difficult one. Simple measures of poverty based on thresholds of monetary income per person or household may not be adequate in a society in which the poor are regularly involved in nonmonetary forms of wealth-generation such as bartering and self-construction of housing. Income thresholds also make comparisons with other Latin American countries more difficult. For these reasons we decided to test several alternative proxies for low socioeconomic status based on the educational attainment of adults (those over 15 years of age), and to compare them against the income earned by workers 12 years of age or older. Previous work by the CONAPO has found a composite index of marginality to also be useful in distinguishing the most vulnerable segments of Mexican society. The marginality index is created using a principal component analysis with nine variables that measure various individual-level characteristics such as income and education, as well the characteristics of dwellings such as the percentage of houses with dirt floors, without electricity, and without running water and sewage (CONAPO, 1993).2 In order to further validate our measures of low socioeconomic status at the census-tract level, we also computed a reduced form of CONAPO’s marginality index using five variables available in the tabulations at the census-tract level for both the 1990 and 2000 censuses. The five variables included are the percentage of workers earning twice the minimum wage or less; the percentage of residents 15 years or older who are illiterate; the percentage of residents 15 years or older with incomplete primary education or less; the percentage of dwellings with no sewage; the percentage or dwellings with no running water; the percentage of dwellings with only one room; and, the percentage of dwellings with dirt floors.3 Three educational categories were used as proxies for low socioeconomic status based on the percentage of residents 15 years of age or older who had (1) no formal education; (2) primary education or less,
RESIDENTIAL SEGREGATION IN MEXICO CITY
79
and; (3) middle-school education (secundaria) or less. Since available tabulations for workers’ incomes are based on multiples of the minimum wage, we used the two lowest groups, namely, those earning one minimum wage or less and two minimum wages or less as indicators of low socioeconomic status. Because tabulations for the census are based on multiples of the current value of the minimum wage, they reflect very different standards across time. The minimum wage lost 30.6 percent of its value in Mexico City during the decade of the 1990s. We therefore used linear interpolation to estimate the number of workers in each census tract in 2000 earning the equivalent of one and two times the 1990 minimum wage. We also used two different measures as proxies for high socioeconomic status (affluence): the percentage of residents 18 years of age or older with college education or more and the percentage of workers earning more than 5 times the minimum wage. A final set of measures of poverty was derived from information provided by SEDESOL. In order to target its poverty alleviation programs to the poorest neighborhoods in urban areas, SEDESOL has estimated the number of households and the total number of residents living below three different poverty lines down to the city-block level using discriminant analysis with 19 variables from the 2000 census. The variables measure characteristics such as the gender, age, level of education of household heads, the presence of children and elderly, the quality of housing, and ownership of various domestic appliances.4 The three poverty lines are meant to distinguish residents with insufficient income to (1) satisfy basic nutritional requirements (pobreza alimentaria); (2) satisfy basic nutritional, health, and educational expenses (pobreza de capacidades); and (3) satisfy the basic nutritional, health, and educational expenses, plus those costs associated with clothing, footwear, housing, and transportation (pobreza de patrimonio) (SEDESOL, 2002; n.d.). SEDESOL poverty estimates are used below to further validate our proxies for poverty at the census-tract level, as well as to estimate segregation at various levels of aggregation, including the city-block level. An examination of the measures of education at the census-tract level reveals an overall increase in educational attainment of Mexico City residents between 1990 and 2000. For example, the percentage of residents with no education in the average census tract decreased from 5.7 percent to 4.0 percent during the decade, while the percentage of residents with college education or more increased from 17.6 percent to 20.4 percent. By contrast, changes in the income measures reveal a decrease in the real wages of Mexico City workers, especially those in the lower end of the income distribution. The percentage of workers earning less than one minimum wage (in constant 1990 terms) in the
80
VILLARREAL AND HAMILTON
average census tract increased from 19.5 percent to 25.0 percent over the decade, while the percentage earning less than two minimum wages decreased from 63.4 percent to 53.1 percent. Finally, an analysis of the correlation between our various measures generally validated their use as proxies for poverty and affluence. For example, the marginality index is positively correlated with the percentage of residents with low educational attainment at the census-tract level and negatively related to the percentage with college education. The index is also positively related with the percentage of workers earning less than one, and, especially, less than two minimum wages and negatively related with the percentage earning more than five minimum wages. The percentages of residents below the household and individual-based poverty lines estimated by SEDESOL are also strongly associated with our measures in the expected directions. Measures of Segregation In the analysis presented below we examine four different dimensions of spatial segregation, namely evenness, isolation, clustering, and centralization, for each of our measures of poverty and affluence. Since the indexes of segregation used to measure evenness, isolation, and clustering have been defined elsewhere in this volume, in this section we only discuss the measure of centralization. Members of a minority group may be spatially concentrated in the center of a city or in its periphery. To measure the extent of a group’s centralization, we use the relative centralization index (RCE) first proposed by Duncan and Duncan (1955). The index may be defined as follows: ⎛ n ⎞ ⎛ n ⎞ RCE = ⎜ ∑ Xi −1Yi ⎟ − ⎜ ∑ XiYi −1 ⎟ ⎝ i =1 ⎠ ⎝ i =1 ⎠
(1)
where the n areal units are first sorted by increasing distance from the city center—in our case the central plaza or zocalo. The figures Xi and Yi are the respective cumulative proportions of groups X and Y’s population up to and including residential unit i, while Xi21 and Yi21 are the cumulative proportions up to the preceding unit (Massey and Denton, 1988).5 The RCE represents the share of group X who would have to move to equalize the two groups’ distribution around the city center, and it varies between 21 and 1, with 0 being equal distribution of two groups around the city center, 1 being the complete centralization of minority group X, and 21 being the complete decentralization of minority group X. If our hypothesis regarding the settlement of the urban poor in the outer rings
RESIDENTIAL SEGREGATION IN MEXICO CITY
81
of Mexico City is correct we should expect a negative centralization index for our measures of poverty and a positive one for our measures of affluence, contrary to values commonly found in the United States.
Results Evenness The first panel in table 5.1 shows the indexes of dissimilarity for our various measures of disadvantage and affluence calculated at the censustract level. In order to allow comparisons across both years, the indexes are computed only for our first (more restrictive) definition of the Mexico City metropolitan area.6 Perhaps the most interesting finding is that the highest educational and income groups—those with college education or more and those with incomes greater than five minimum wages—are more segregated than the lowest groups. In other words, it appears that the most affluent residents of Mexico City are less evenly distributed than the poorest residents. Moreover, with the exception of the lowest educational category consisting of adults with no formal education, the indexes of dissimilarity indicate a greater level of segregation for higher-educationalstatus groups in general. The dissimilarity indexes, computed using categories of income among employed residents 12 years of age or older, show a similar pattern: residents earning five times the minimum wage or less were more segregated than those earning two minimum wages or less, and the latter were in turn more segregated than those earning only one minimum wage in 1990. Residents in the higher of the two income categories available in 2000 were also more segregated.7 Of our various measures of disadvantage, it seems that the highest thresholds (middle education or less and less than two minimum wages) capture a more meaningful distinction with regard to the spatial distribution of Mexico City residents. Comparisons of segregation levels across countries and categories should generally be approached with caution. Nevertheless, because the index of dissimilarity is less sensitive to the overall proportion constituted by a particular group within a metropolitan area, it allows us to place the segregation of disadvantaged groups in Mexico City in comparative perspective. The indexes of dissimilarity shown in table 5.1 suggest a considerably lower level of segregation of the most disadvantaged residents of Mexico City compared to racial minorities in the United States. Logan, Stults, and Farley (2004) find an average index of dissimilarity of 65.2 for black-white segregation, 51.6 for Hispanic-white segregation, and 42.2 for Asian-white segregation in U.S. metropolitan areas for 2000. Of course, one important
Table 5.1 Indexes of dissimilarity, isolation, clustering, and centralization for the Mexico City metropolitan areas computed at the Census–tract Level Isolation Dissimilarity Variable Education No education Primary education or less Middle school education or less College education or more Wages (base 1990) Less than 1 min. wage Less than 2 min. wages More than 5 min. wages
Clustering Eta2
Isolation Index
Moran’s I
Centralization z–value
1990
2000
1990
2000
1990
2000
1990
2000
1990
2000
1990
2000
22.06 16.99 24.81 35.57
23.09 14.80 28.64 36.44
7.29 27.23 57.68 25.57
5.23 24.93 61.85 29.63
1.61 3.31 9.12 11.04
1.26 2.46 12.14 13.21
0.425 0.590 0.653 0.701
0.452 0.670 0.705 0.734
36.1 50.9 54.0 59.9
42.2 61.5 65.3 66.1
–11.23 –7.28 –12.02 13.53
–10.24 –6.52 –13.17 14.41
16.39 23.14 39.94
16.94 21.72
21.27 66.79 19.30
27.88 57.29
2.66 7.78 10.51
3.44 7.30
0.445 0.548 0.651
0.592 0.651
37.4 46.0 54.8
55.3 60.3
–4.69 –8.37 6.38
–7.89 –9.72
Notes: The dissimilarity index is multiplied times 100. The census tract containing the zocalo is used as the geographic center for the index of centralization.
RESIDENTIAL SEGREGATION IN MEXICO CITY
83
limitation with cross-country comparisons is that the dissimilarity index is sensitive to the population size of the census tracts used to compute it. The dissimilarity index is generally higher for smaller units. The average size of a census tract in Mexico City is 4,588, whereas it is 4,000 in the typical census tract in the United States.8 With respect to the observed trends in segregation, the results shown in table 5.1 indicate relatively small changes in the evenness of the distribution of the poor in Mexico City between 1990 and 2000. This finding is not surprising given the generally slow pace at which segregation changes occur. Nevertheless, some interesting patterns are detected. The index of dissimilarity increased for every educational category examined except for adults with primary education or less. The increase was highest for those with middle-school education or less (15.4 percent), while the decrease for those with primary education or less was 12.9 percent. Taken together, the changes in the dissimilarity indexes for the different educational categories suggest that it was the moderately low and the very highest socioeconomic-status groups that became more segregated during the 1990s. The changes in the dissimilarity index for the two income categories suggest a slightly different pattern. Whereas the dissimilarity index for employed adults earning less than one minimum wage remained essentially unchanged during the decade, those earning less than two minimum wages became more segregated. It should be noted, however, that the results for the two income categories in 2000 may be affected by the linear interpolation used to adjust for changes in the real value of the minimum wage between 1990 and 2000, and are therefore less reliable.
Exposure The second panel in table 5.1 shows the isolation indexes for our four educational groups and three income categories. In order to allow for comparisons of the degree of contact the various groups have with non-group members as well as their relative isolation over time, the correlation ratio or Eta2 was also computed. In contrast to the isolation index, the correlation ratio is not dependent on the overall composition of the metropolitan area and is also symmetrical in that its value for a group is equal to that for the rest of the population (Massey and Denton, 1988). The cross-sectional results for the correlation ratio are much more consistent than those just described for the dissimilarity index: the correlation ratio increases consistently for groups with higher levels of education. The most isolated are those with college education or more. And it is precisely this group that became more isolated over time. According to the raw isolation index,
84
VILLARREAL AND HAMILTON
the probability that a randomly selected adult with college education or more shares a census tract with another such adult is 0.30 in the year 2000. The correlation ratio is also higher for higher income categories in both years, although we are unable to determine the change in the correlation ratio for the highest income category between 1990 and 2000. Clustering One of the clearest patterns in the spatial distribution of residents belonging to different socioeconomic groups in Mexico City during the decade of the 1990s has to do with the level of clustering. According to the Moran’s I values and their associated z-scores shown in the third panel in table 5.1, every educational and income group was significantly clustered (at the .001 level) in both years. In other words, the distribution of high and low poverty groups are significantly more clustered than would be expected by mere chance. Although the z-scores are not strictly comparable, since their underlying distributions are not the same and cannot generally be assumed to be normal, the increasing values nevertheless suggest both a slightly increasing level of clustering for higher-status groups and an increase over time. If true, this would suggest that Mexico City may be experiencing a greater separation of the poor and non-poor at the multicensus tract level. In order to obtain a clearer picture of whether the clustering is due to the spatial grouping of low- or high-poverty tracts, as well as where the clusters are located within the city, we produced maps displaying the values of the local Moran’s I for residents with middle-school education or less in 1990 and 2000. Maps 5.1a and 5.1b show census tracts with high levels of disadvantage that are significantly located next to other highlevel tracts; those with low disadvantage next to others with low disadvantage. As expected, the clusters of high disadvantage are located in the outer areas of the city, but especially in the northern and eastern sections, whereas the low-disadvantage clusters are located in the southern and western sections of the city. A comparison of the maps for 1990 and 2000 suggests that the increase in clustering during the decade may have occurred not only by an increase in the high disadvantage clusters, but also to a large extent by a greater consolidation of contiguous tracts with low levels of disadvantage in the southwestern part of Mexico City. However, comparisons of the levels of clustering should be made with caution, as they may be sensitive to changes in the underlying distributions caused by differences in the overall proportions of the disadvantaged tracts. An overall increase in educational levels may be responsible for the greater clustering.
RESIDENTIAL SEGREGATION IN MEXICO CITY
85
Map 5.1a Clusters of census tracts with low (light grey) and high (dark) proportions of adults with middle school education using Local Moran’s I in 1990
Centralization Finally, the RCEs shown in the last panel in table 5.1 are consistent with our hypothesis that Latin American urban centers, such as Mexico City, are characterized by the opposite pattern observed in the United States, namely, the decentralization of disadvantaged groups. The lowest educational attainment and those with the lowest incomes have negative
86
VILLARREAL AND HAMILTON
Map 5.1b Clusters of census tracts with low (light) and high (dark) proportions of adults with middle school education using Local Moran’s I in 2000
indexes of centralization in table 5.1. By contrast, residents with the highest educational attainment levels (those with college education or more) and those in the highest income category (more than five times the minimum wage) have positive indexes of centralization. We were unable to find studies reporting the level of centralization in U.S. metropolitan areas for low-income groups to be used as benchmarks. However, Ellen (2000) finds highly positive centralization indexes for
RESIDENTIAL SEGREGATION IN MEXICO CITY
87
African Americans in a sample of 220 metropolitan areas in the United States. The average centralization index in her sample is 0.358.
Segregation along Official Poverty Lines at Different Levels of Aggregation In the preceding section we used the census-tract-level variables to examine four dimensions of residential segregation. Additional tabulations available at the city-block level for 2000 allow us to further analyze how the level of aggregation affects our estimates of segregation. These tabulations also allow us to examine segregation levels along the three poverty lines identified by SEDESOL (pobreza alimentaria, pobreza de capacidades, and pobreza de patrimonio). Unfortunately, these data are only available for 2000, which prevents us from examining changes over time. Table 5.2 shows the dissimilarity indexes at the city-block, census-tract, and municipal levels for the estimated number of persons and households considered to fall below the three poverty lines as well as the number of household heads with primary education or less (the latter measure is closer to the variables used in our previous analysis). Also, in order to determine how sensitive our previous estimates of segregation are to the specific boundaries used for Mexico City, we computed the dissimilarity indexes for both the CONAPO definition of the Mexico City metropolitan area made up of 75 municipalities and delegaciones, as well as for our more restrictive definition including only 23 municipalities and delegaciones. The dissimilarity indexes for the six measures of poverty at the census-tract level are substantially higher than those obtained with our proxies based on low educational attainment and low-income levels in the previous section of the chapter. The dissimilarity indexes range from 0.36 to 0.46 for the CONAPO definition of the metropolitan area using the three household poverty levels and from 0.33 to 0.43 for the more restrictive definition of the Mexico City metropolitan area. These values are high in absolute terms, and begin to approach those found for racial segregation in the United States. Predictably, the dissimilarity indexes are higher when households are aggregated into smaller units of analysis. In other words, the dissimilarity indexes are higher when computed at the city-block level than they are at the census-tract level, and higher at the census-tract level than they are at the municipality level for every variable used. Finally, with regard to the effect of the boundaries of the Mexico City metropolitan area used, the results are generally quite similar at
Table 5.2 Dissimilarity indexes for two different definitions of the Mexico City metropolitan area computed at the city–block, census–tract, and municipality levels Dissimilarity Index (in %)
Education Heads with primary education or less Household Poverty Households below poverty line 1 (patrimonio) Households below poverty line 2 (capacidades) Households below poverty line 3 (alimentaria) Individual Poverty Persons below poverty line 1 (patrimonio) Persons below poverty line 2 (capacidades) Persons below poverty line 3 (alimentaria) Notes: The dissimilarity indexes are multiplied times 100.
CONAPO Definition
Other Definition
Block
Census-tract
Municipality
Block
Census-tract
Municipality
31.87
26.12
12.60
31.86
26.03
10.47
41.64 51.56 60.51
35.59 41.56 46.19
20.90 25.94 30.26
40.06 50.36 60.55
33.27 38.49 42.72
15.68 19.39 21.73
40.41 50.02 57.17
34.41 40.40 44.33
19.90 25.50 29.25
38.59 48.61 56.52
31.84 37.07 40.40
14.61 17.97 19.69
RESIDENTIAL SEGREGATION IN MEXICO CITY
89
the census-tract level, thereby justifying our use of the more restrictive boundaries for our comparison between 1990 and 2000. For example, the dissimilarity index for household heads with primary education or less (the closest indicator to the ones used in the previous section) is almost identical for both definitions of the metropolitan area (0.2612 for the CONAPO definition compared to 0.2603 in the more restrictive definition). The values obtained for the six measures of poverty are somewhat higher for the CONAPO definition. This is not surprising given that the 75 municipalities and delegaciones included in the more expansive definition include areas that are less urban and therefore more likely to include tracts with higher concentrations of poor households. For the sake of brevity, our calculations of the remaining indexes of segregation at the city-block level and for the official poverty measures are not presented, but were generally found to be consistent with the patterns described in the previous section.
The Spatial Mismatch Hypothesis: How Segregation May Affect Employment The spatial mismatch hypothesis was first proposed by John Kain (1968), who argued that the movement of manufacturing firms out of inner-cities into the suburbs in search of lower land costs, coupled with the inability of poor black residents of U.S. inner-cities to relocate near these sources of employment (due to discriminatory and restrictive housing policies), creates a spatial mismatch between employers and low-skilled African American workers. This spatial mismatch was thought to explain the growing unemployment of inner-city black residents who could not easily find jobs in firms located far away through informal networks, and who may not have been able to afford the transportation costs of commuting to the suburbs. Kain’s spatial mismatch hypothesis has received mixed empirical support in the United States. Most studies looking at residential segregation and/or residential suburbanization and black employment do not support the theory, whereas studies that look at employment suburbanization and black employment do (see Holzer, 1991). The debate regarding these findings centers on the extent to which measures of segregation reflect a spatial mismatch. That is, Kain argues that it is not the evenness of the distribution of blacks and whites across cities but “whether the housing market discrimination confines them to a narrow and spatially concentrated segment of the metropolitan-area housing market” (1992, p. 380). Thus the problem is not one of evenness but of clustering and centralization.
90
VILLARREAL AND HAMILTON
Percentage Workers in Large Manufacturing Firms 0.00% 0.00% – 0.25% 0.25% – 0.50% 0.50% – 2.54%
Map 5.2 Spatial distribution of large manufacturing firms in Mexico City based on the percentage of total workers employed, 1994
In Mexico City we have seen that disadvantaged social groups are spatially clustered on the periphery of the city, an opposite pattern to the spatial concentration of blacks in U.S. inner-cities. However, to the extent that employers are located closer to the city center in Mexico City, poor Mexico City residents may also experience a spatial mismatch that influences their overall employment opportunities. In order to examine this possibility, we used data from the 1994 economic censuses at the census-
RESIDENTIAL SEGREGATION IN MEXICO CITY
91
tract level to identify the location of manufacturing and service firms and assess their distance from the previously identified clusters of low-income census tracts. Map 5.2 shows the location of large manufacturing firms in the Mexico City metropolitan area in 1994. The specific variable used is the total number of firms with more than 100 workers located in a particular census tract divided by the total surface area of the census tract. We divide the number of firms by the surface area because we want to identify the areas with the highest concentration of firms. Interestingly, the highest concentration of manufacturing firms is located in the central part of the city and slightly towards the north. A comparison of map 5.2 with the distribution of residents with low levels of education in maps 5.1a and 5.1b reveals a clear mismatch between the location of residents with low educational levels and the source of blue-collar manufacturing jobs. Additional maps generated using the various measures of poverty defined by SEDESOL revealed a similar mismatch.9 The relatively great distance between poor residents of Mexico City and manufacturing jobs may indeed be aggravating their socioeconomic conditions. Further analysis at the individual level is required to determine whether the distance from the neighborhood of residence to the high-employment centers is in fact associated with lower employment rates among Mexico City’s urban poor.
Conclusion In this chapter we have uncovered important trends in socioeconomic segregation that reflect Mexico City’s historic and contemporary patterns of demographic, physical, and economic growth. First, we found that the high socioeconomic groups—those with highest levels of education and wages—are more segregated in terms of the evenness of their distribution and their isolation within census tracts than are the low socioeconomic groups. On the whole, these groups became slightly more segregated between 1990 and 2000, a trend that supports theories of greater polarization between socioeconomic groups in cities undergoing economic transformation due to globalization (Graizbord et al., 2003; Sassen, 1991). There were some interesting exceptions to this rule, particularly, groups with education and income just above the lowest levels, which deserve greater attention. The fact that high-income groups are more unevenly distributed and isolated than low-income groups suggests that their daily experience may be more socially homogenous than that of low-income groups, a finding in opposition to that of Rubalcava and Schteingart (2000). It is possible that this increasing segregation of the wealthiest sectors of Mexico City’s
92
VILLARREAL AND HAMILTON
population reflects recent patterns towards increasing isolation and physical separation of the wealthy detected in other global cities (Marcuse and van Kempen, 2000). Our analysis of clustering and centralization shows a pattern of segregation consistent with the historical pattern of migrant settlement in Mexico City. In the 1990s, the poorest sectors of the population were still located on the periphery of the city, especially towards the north and east, whereas the better-off sectors were clustered just south and west of the city center. Our measures of centralization confirm these patterns: disadvantaged groups are decentralized whereas wealthier groups are more centralized. This pattern has important consequences for access to employment. Through mapping large manufacturing firms across the city, we uncovered a spatial mismatch between the location of these sources of employment and the location of low-income groups. The location of commercial and service-sector firms not shown in the maps presented above also reflect a mismatch. To the extent that disadvantaged groups are unable to overcome this physical separation (through transportation, for example), their employment opportunities may be limited. Future work should further explore this important relationship between the spatial locations of sources of employment and low-income groups and the employment levels of those groups. Ultimately, this spatial mismatch may contribute to the continuation of socioeconomic inequality in Mexican urban areas. Our work was hindered by a number of data and methodological obstacles. Problems that we could not completely overcome involved differences between 1990 and 2000 in terms of availability of data, definition of city boundaries, and changes in the national minimum wage. It is possible that our findings misrepresent true changes in educational and wage segregation because we were unable to control for the changing age structure of the population between 1990 and 2000. That is, because educational levels tend to be lower among older Mexico City residents, changes in the aggregate level of education at the census-tract level may not only reflect changes in the socioeconomic level of residents but may also be a result of changes in the age structure of the population. Future work should strive to overcome these measurement problems. Notes Research was supported in part by a grant to Andrés Villarreal from the National Institute of Child Health and Human Development (grant number 1R03HD051673). Research was also supported by the Andrew W. Mellon Foundation Center Grant on Urbanization and Internal Migration to the
RESIDENTIAL SEGREGATION IN MEXICO CITY
93
Population Research Center at the University of Texas at Austin. The authors would like to thank the staff at the Ministry of Social Development (SEDESOL) in Mexico, and Javier Suárez Morales in particular for sharing their data with us and for various helpful suggestions. 1. In 2000, the inclusion of questions about municipality of work on the national census allowed for the development of an official Metropolitan Zone of Mexico City (ZMCM), which has similar guidelines for inclusion to the U.S. Statistical Metropolitan Area (see Federal Register, 2000). The new Mexican guidelines define metropolitan zones in Mexico as places with a central core of at least 50,000 inhabitants and surrounding municipal districts that contribute at least 15 percent of their working population as commuters to the central core, or, conversely, where at least 10 percent of the central core workers commute from that particular surrounding municipal district. Additional surrounding municipal districts are included in the zone if they also meet certain standards regarding contiguity, population integration, non-agricultural employment base, population density, and physical proximity to the central core (CONAPO, 2004). 2. The variables included in the computation of the marginality index are (1) the illiteracy rate among individuals 15 years of age or older; (2) the percentage of individuals 15 years of age or older with incomplete primary education; (3) the percentage of residents in dwelling without sewage or toilet; (4) the percentage of residents in dwellings without electricity; (5) the percentage of residents in dwellings without running water; (6) the percentage of dwellings with more than two occupants per bedroom; (7) the percentage of residents in dwelling with dirt floors; (8) the percentage of residents living in towns of less than 5000; and (9) the percentage of the economically active population earning less than two minimum wages (CONAPO 1993). 3. In order to make this marginality index comparable for both years the same factor loadings obtained in 1990 were applied to the 2000 variables. 4. The 19 variables are (1) rural locality (population less than 2500); (2) the type of material used for floor of dwelling; (3) the presence of bathroom or toilet in dwelling; (4) the presence of a household head with no formal education; (5) the presence of a household head with incomplete primary education; (6) the presence of a household head with incomplete middle school education; (7) the age of household head; (8) gender of the household head; (9) the demographic dependence ratio; (10) the number of residents per bedroom; (11) ownership of a video cassette recorder; (12) ownership of a gas range; (13) ownership of a refrigerator; (14) ownership of a washing machine; (15) ownership of a car; (16) whether at least one resident of the household has social security benefits; (17) children ages 5–15 years not attending school; (18) children ages 12–15 who work; and (19) children less than 12 years of age. 5. In order to sort the census tracts in Mexico City by their distance to the Zocalo we first estimated the centroid for each census tract and then computed the distance from each centroid to the centroid of the tract containing the Zocalo. 6. As described in the methods section above, the indexes presented are for a single group using the remaining population as reference.
94
VILLARREAL AND HAMILTON
7. Unfortunately, the interpolation method used does not allow us to obtain a reliable estimate for the number of individuals earning five times the minimum wage or less in 2000. 8. See http://www.census.gov/geo/www/tiger/glossry2.pdf, last accessed on November 24, 2008. 9. One important caveat is in order: due to data limitations, map 5.2 is based on the more restrictive definition of Mexico City. It is possible that the further outlying areas of the city excluded from this restrictive definition contain a high concentration of manufacturing and service sector firms.
References Anselin, L. (1995). Local indicators of spatial association—LISA. Geographical Analysis, 27, 93–115. Ariza, M. and Solis, P. (2005). Dinámica de la desigualidad social y la segregación especial en tres areas metropolitanas de México. Paper presented at the IUSSP XXV Conference on Population. July 18–23. Bush, V. P. and Gomez, C. A. (2003). Escenarios demográficos y urbanos de la zona metropolitana del valle de México. Mexico City: CONAPO. Consejo Nacional de Población (CONAPO). (1993). Indicadores socioeconómicos e índice de marginación municipal, 1990. Mexico City: CONAPO. ——— (2000). Published statistics. http://www.conapo.gob.mx/00cifras/ 00indicadores.htm (accessed November 24, 2008). ——— (2004). Delimitacion de las zonas metropolitanas de México. Mexico City: CONAPO. Cunradi, C. B., Caetano, R., Clark, C., and Scafer, J. (2000). Neighborhood poverty as a predictor of intimate partner violence among White, Black, and Hispanic couples in the United States: A multilevel analysis. Annals of Epidemiology, 10(5), 297–308. Delgado, J. (1990). De los anillos de la segregación: La ciudad de México 1950– 1987. Estudios Demográficos y Urbanos, 5, 237–74. Duncan, O. D. and Duncan, B. (1955). Residential distribution and occupational stratification. American Journal of Sociology, 60, 493–503. Ellen, I. G. (2000). Is segregation bad for your health? The case of low birth weight. Brookings-Wharton Papers on Urban Affairs, 1, 203–29. Federal Register. (2000). 65(294). Washington, DC: U.S. Government Printing Office. Garza, G. and Schteingart, M. (1978). Mexico City: The emerging megalopolis. In W. Cornelius and R. V. Kemper (eds.), Metropolitan Latin America: The challenge and response. Beverly Hills, CA: Sage. Graizbord, B., Rowland, A., and Aguilar, A. G. (2003). Mexico City as a peripheral global player: The two sides of the coin. The Annals of Regional Science, 37, 501–18. Harding, D. J. (2003). Counterfactual models of neighborhood effects: The effect of neighborhood poverty on dropping out and teenage pregnancy. American Journal of Sociology, 109, 676–719.
RESIDENTIAL SEGREGATION IN MEXICO CITY
95
Holzer, H. (1991). The spatial mismatch hypothesis: What has the evidence shown? Urban Studies, 28(1), 105–22. Kain, J. (1968). Residential segregation, Negro employment, and metropolitan decentralization. Quarterly Journal of Economics, 82, 175–97. Lewis Mumford Center. Ethnic diversity grows, neighborhood integration is at a standstill, 2001. http://www.albany.edu/mumford/census (accessed November 24, 2008). Logan, J., Stults, B. J., and Farley, R. (2004). Segregation of minorities in the metropolis: Two decades of change. Demography, 41, 1–22. Lomnitz, L. A. (1977). Networks and marginality: Life in a Mexican shantytown. New York: Academic Press. Marcuse, P. and van Kempen, R. (2000). Introduction. In P. Marcuse and R. van Kempen (eds.), Globalizing cities. Oxford: Blackwell. Massey, D. and Denton, N. (1988). The dimensions of residential segregation. Social Forces, 67(2), 281–315. ——— (1993). American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard University Press. Park, R. E. and Burgess. E. W. (1925). The city. Chicago: University of Chicago Press. Pezzoli, K. (1988) Human settlements and planning for ecological sustainability: The case of Mexico City. Cambridge, MA: MIT Press. Ross, C. (2000). Neighborhood disadvantage and adult depression. Journal of Health and Social Behavior, 41, 177–87. Rubalcava, M. and Schteingart, M. (1987). Estructura urbana y diferenciación socioespacial en la zona metropolitana de la ciudad de México (1970–1980). In G. Garza (ed.), Atlas de la Ciudad de Mexico, pp. 108–14. Mexico City: Colegio de México. ——— (2000). Segregación socioespacial. In G. Garza (ed.), La ciudad de México en el fin del Segundo milenio, pp. 287–96. Mexico City: El Colegio de México. Sassen, S. (1991). The global city: New York, London, Tokyo. Princeton, NJ: Princeton University Press. Schteingart, M. (1989). Dinámica poblacional, estructura urbana y producción del espacio habitacional en la zona metropolitana de la ciudad de México. Estúdios Demográficos Urbanos, 12, 521–48. Secretaría de Desarrollo Social (SEDESOL). (2002). Medición de la pobreza: Variantes metodológicas y estimación preliminar. Mexico City: SEDESOL. ——— (n.d.) Nota técnica para la medición de la pobreza con base en los resultados de le Encuesta Nacional de Ingreso y Gastos de los Hogares, 2002. Mexico City: SEDESOL. Telles, E. (1995). Structural sources of socioeconomic segregation in Brazilian metropolitan areas. American Journal of Sociology, 100, 5, 1199–1223. Turner, J. (1968). Housing priorities, settlement patterns, and urban development in modernizing countries. Journal of the American Institute of Planners, 34, 354–63. Ward, P. (1998). Mexico City. West Sussex, UK: John Wiley & Son. Wilson, W. J. (1997). When work disappears: The world of the new urban poor. New York: Knopf.
This page intentionally left blank
6
Residential Segregation in Montevideo: Challenges to Educational Equality Ruben Kaztman and Alejandro Retamoso
Introduction Until the 1960s, Montevideo could have been conceived as a single territorial unit—compact and consolidated, with neighborhoods and zones defined by the centrality of a labor market structured around government employment and industrial production (Kaztman, Filgueira, and Errandonea, 2005). The characteristics of this spatial profile sharpened during the economic hegemony of the import-substitution model. The collapse of this model resulted in profound changes in both the labor force and the social morphology of the city. The situation of the labor force was deeply affected by the decline of industrial and state employment and the weakening of workers’ ties with the labor market. Unable to cover the rents in the center of the city, unemployed and precarious workers were forced to move to the periphery, most of them finding new homes in irregular settlements. Thus, the main feature of recent processes of residential segregation in Montevideo is the high level of homogeneity in the social composition of poor neighborhoods. These transformations presented unexpected challenges to the governance of the city. This chapter has two main purposes. The first is to characterize the nature of the aforementioned social transformations. The second is to examine the ways in which those responsible for educational policy have confronted the threats posed by these transformations for the achievement of an egalitarian society. We do this through the analysis of two phenomena: the recent
98
KAZTMAN AND RETAMOSO
trends in the distribution of the socio-educational groups in Montevideo, and the relationship between the social composition of neighborhoods and the provision of educational services. But before embarking upon this analysis, a brief characterization of the socio-demographic profile of the city is in order. Changes in the Social Morphology of the City during the Past Decades Compared with other large Latin American cities, Montevideo’s process of urbanization appeared early and has been a gradual phenomenon. By the middle of the last century, a little more than 20 percent of the country’s population resided in rural areas, whereas almost half of the urban population of Uruguay was concentrated in Montevideo—thus accounting for 57 percent of the urban population. The primacy of the capital weakened in the last decades, even though it was ameliorated by the growth of the metropolitan population residing in departments adjacent to Montevideo. Uruguay is divided into 19 administrative units referred to as departments. The department of Montevideo, which includes the capital, is physically the smallest and is bordered by two other departments—Canelones, and San José. It is also important to note that a small part (4 percent) of the department of Montevideo continues to be considered rural,1 and that a significant part of the urban expansion of the city was at the expense of these rural areas, particularly in the form of irregular settlements. As a result, in 2004 one out of every eight individuals (12.4 percent) in irregular settlements resided in rural areas. Population Movements in Montevideo Between 1985 and 1996 Montevideo grew at a rate of 2.3 per 1,000 while the corresponding rates in the neighboring departments were of 18.5 and 6.9 per 1,000, respectively. These differences increased during the period 1996 to 2004. While Montevideo experienced a negative growth rate of 1.5 per 1,000, Canelones and San José grew at 11.5 and 8.0 per 1,000, respectively. Thus, the slight variations in the size of the Montevideo population between 1996 and 2004 hide two important movements: one within the department and the other toward adjacent departments. A large part of the population growth of these two departments can be attributed to emigration from Montevideo,2 which in the period from 1991 to 1996 displayed a cumulative net migration rate of 211.74 percent.3
RESIDENTIAL SEGREGATION IN MONTEVIDEO
99
Montevideo’s low growth between 1985 and 1996 and its negative growth between 1996 and 2004 were then counteracted by the phenomenon of metropolitization spreading out from the aforementioned departments. 4 Through the last four decades, the population in the periphery of the capital residing in those departments went from 10 to approximately 20 percent of the total of Greater Montevideo. In the case of Canelones, at least, the available beach resort infrastructure operated as an attraction pole for the middle and lower-middle classes. Comparisons made by the authors of the 2004 population census and Encuesta Continua de Hogares (ECH) data from the National Institute of Statistics (INE) show that the population expansion in the periphery of Montevideo seems to have been driven by the incorporation of households with socioeconomic profiles slightly different from those of the lower classes that reside in the city. Although it shares with the lower-class neighborhoods of Montevideo a high proportion of households at the beginning stages of their family life cycle, in the periphery there is a lower proportion of infants and children, higher educational achievement among the population ages 25 to 59, higher per capita income, and, consequently, lower indices of poverty and indigence. The process of suburbanization within the department of Montevideo was very different, particularly in terms of infrastructure and land occupations. In essence, a large part of this population movement went in the direction of new urban parceling in rural areas or urban zones with low population density. Upon these lands, occupied at the margin of legality, the so-called irregular settlements expanded rapidly. Taken as a whole, from the movements described above it is clear that the apparent “quietness” of the inter-census population growth of Greater Montevideo hides important transformations in the manner in which social strata use urban space.
Characteristics of Neighborhoods Losing or Gaining Population “New” neighborhoods growing during the last decade were characterized by poor households and irregular settlements, a relatively high proportion of children and adolescents, and a relatively low mean level of educational achievement among its economically active residents. Everything points to an increase in the territorial concentration of young households with scarce human resources, as well as problems with housing and poverty. Cities can suffer significant transformations in the social composition of their neighborhoods due to social mobility, migration movements, or differential fertility rates. Unlike other cities in Latin America (and other periods in the history of this city), the preceding trends do not support
100
KAZTMAN AND RETAMOSO
the hypothesis that processes of upward mobility were relevant to explain recent changes in Montevideo’s social morphology. Instead, the massive displacement of poor households toward the periphery and the natural increase of those same households appear as the main determinants of these changes. Land occupations and the formation of irregular settlements, central features in these transformations, were concentrated in the capital of the country. Of every ten Uruguayans living in irregular settlements through the country as a whole, nearly eight lived in Montevideo. Characteristics of Neighborhoods with Irregular Settlements One classic characteristic of neighborhoods with a high concentration of irregular settlements is the relatively high proportion of families in the first stages of their life cycle. Examining the 62 neighborhoods in Montevideo, the Pearson’s r correlation between the percentage of the population in irregular settlements and the percentage of children and adolescents is positive and statistically significant at 1 percent (0.767). In turn, the correlation between the percentage of the population in irregular settlements and the percentage of persons 60 years and older is also significant, but negative (20.792). Generational Imbalances in the Spatial Distribution of the Population in Montevideo Within Latin America, Uruguay stands out for its older age structure and a strong imbalance in the welfare of distinct generations. These differences are reflected in the profile of the Montevideo neighborhoods. An indicator of the relative weight of older adults in the age structure is the percentage of people ages 65 and older as a total of the population. In Uruguay that figure was 13 percent in the year 2000, twice as much as the corresponding figure for the whole region. This feature is even more accentuated in Montevideo, with about 15 percent of its population in this cohort. The aging of the city’s population was clear during the last decade. It is explained by a long-term low fertility rate, an increase in international migration (which affects primarily the young population), and by the age structure of the population that immigrated to Montevideo from rural areas and small and large towns.5 As we have seen, the poorest and youngest families moved to the peripheral neighborhoods, thus reinforcing the relevance of the life cycle in the spatial distribution of the population in Montevideo. The differences
RESIDENTIAL SEGREGATION IN MONTEVIDEO
101
between growing and declining neighborhoods with regard to educational levels of the adult population, indices of poverty, and presence of irregular settlements support the idea that the generational imbalance is one of the primary axes ordering the social geography of the city. The Uruguayan history of the past 50 or 60 years is congruent with that vision. Older households were the main beneficiaries of the inertia of the strong state’s welfare institutions which were at their height in the middle of the last century. Today, they are in a clear decline, lacking the capacity to attend the new structures of risk faced by young families with relatively low skills.6 Some studies have already pointed out this national peculiarity. For instance, they show that, in the regional scenario, Uruguay has the largest imbalance between the general poverty index and the quotient between childhood poverty and general poverty (Kaztman and Filgueira, 2001). Recent data from Montevideo indicates an increase in the city generational imbalance. In effect, from 1986 to 1988 the poverty index among children and adolescents between the ages of 0 to 17 years rose to 51.5 percent, while the overall poverty index was 34.5 percent, and the ratio between them 1.49. From 2002 to 2004, the corresponding figures were 48.2 and 28.1 respectively, and the ratio between the two rose to 1.71. In the period under consideration, overall poverty reduced about 6 percentage points. However, while childhood poverty decreased only 3 points, elderly poverty decreased by 13 points. Residential Segregation in Montevideo The transformations just described are part of important processes of residential segregation in the past 20 years, which changed the social morphology of the city. This section analyzes the characteristics of those processes. Geographic Scale This work concentrates on the department of Montevideo and its 62 neighborhoods, though in some cases we also work with census tracts.7 As the meaning of different levels of territorial aggregation is intimately related to the phenomenon under study, it is useless to debate the level of aggregation most appropriate to a study without having clarified its object of analysis. For instance, the significance of territorial limits are different for children, adolescents, or adults, which indicates the need to consider stages in the life cycle as one of the criteria to define the adequate level of geographic aggregation.
102
KAZTMAN AND RETAMOSO
Another problem related to the appropriate level of scale is that, for many analytical purposes, the relevance of community context for the formation of habits, expectations, and behaviors is mediated by the level of susceptibility of actors to surrounding influences. Reasonably, this susceptibility will be directly proportional to the importance of a person’s surroundings as a source of social capital. Along this line of reasoning, you would expect people with stable employment and/or participation in civic organizations such as unions, churches, political parties, sports clubs, and so on, to be more receptive to the normal and valorized patterns of these institutions than to those predominating in their neighborhoods. Conversely, those with unstable ties to the labor force and with scarce or no participation in civic organizations will be more susceptible to the influence of forces in the area surrounding their places of residence. When our interest centers on the level of well being of residents, we will be interested in geographic boundaries that establish the relevant parameters of the life conditions that people share. A flood zone, for example, defines a neighborhood where a calamity affects all of its residents. The same can be said of borders that circumscribe shared problems of neighborhood infrastructure, of local employment opportunities, of levels of insecurity, or of insufficient transportation or services. In the case of Montevideo, the first point to underscore in relation to scale is that both neighborhoods and census tracts are relevant to understand the behavior of their residents. Many studies corroborate the importance of these geographical units to predict social situations and risk behaviors associated with enduring poverty and the mechanisms of intergenerational transmission of poverty (Kaztman, 1999; Kaztman and Retamoso, 2005; Cervini and Gallo, 2001; Macadar et al., 2002).8
Indicators Used to Analyze Spatial Segregation Residential segregation will be examined on the basis of educational indicators. This choice has an important substantive support in the growing significance of knowledge for the new forms of material production and, therefore, the growing significance of educational attainments as determinants of differences in workers salaries and well-being (ECLAC, 2001–2002; Kaztman, 2002) Two educational indicators were chosen to analyze changes in spatial segregation. The first is an absolute measure—the percentage of people ages 25 to 59 years old with only an elementary education.9 This percentage declined from 42.3 percent from 1986 to 1988, to 28.1 percent
RESIDENTIAL SEGREGATION IN MONTEVIDEO
103
from 1995 to 1997, and to 22.4 percent from 2002 to 2004. The second is a relative measure, the percentage of adults in this same age group whose education is below the department mean. For the corresponding periods, these percentages were 51.8, 49.0 percent and 50.9 percent. We chose the group 25 to 59 years old because the large majority of people in these ages are economically active and emancipated from their families of origin. Having formed their own families, most of them will be household heads or their spouses. Thus, information on their spatial distribution will give us clues not only on the physical isolation of adults in different educational categories, but also on the conditions which favor the isolation of children and adolescents in those households.10
How Has Spatial Segregation in Montevideo Evolved in the Past Two Decades? The Measures of Segregation The literature has produced a multitude of indices to measure residential segregation. Each emphasizes different aspects of the phenomenon, such as the uniformity in how groups are distributed across territories (e.g., the dissimilarity index, analyses of variance or entropy, the Gini index, etc.); the level of potential exposure to other groups in the same territorial unit (the exposure or isolation index); the level of concentration of a group in particular parts of the city; or, the level of proximity between territorial units where populations with similar characteristics reside (for example, absolute or relative measures of clustering, the index of spatial proximity, and indicators of spatial autocorrelation such as the Moran’s I). The appropriateness of each should be evaluated by two criteria: the characteristics of the social categories whose spatial segregation you are interested in capturing, and the analytical aims orienting your inquiry.11 With that in mind, we selected four segregation indices: the residential segregation index based on variance (when dealing with interval level variables), the Duncan’s index of dissimilarity, the exposure index, and the global Moran’s I. The total variance of years of schooling is decomposed into two parts: between neighborhoods and within neighborhoods. The indicator illustrates the percentage of the total variance for years of schooling for the population residing in the 62 neighborhoods of Montevideo that is explained by differences in the mean education levels of the neighborhoods. The larger the proportion of total variance explained by the
104
KAZTMAN AND RETAMOSO
variance between subunits, the higher the homogeneity within, and the heterogeneity between, neighborhoods. According to this data, in the last two decades the homogeneity in the socio-educational composition within neighborhoods increased, as did the heterogeneity between neighborhoods. Between 1986 and 1988, 17.5 percent of the total variance was explained by the variance between neighborhoods. By 1995 to 1997, this percentage was 22.6 and by 2002 to 2004, it was 26.7 Thus, there is a trend for lower- and higher-educated populations to reside in different areas of the city. The Duncan index of dissimilarity varies between 0 and 100, with values closer to 0 indicating that the distribution of the population with a determined attribute in the subunits of a territory is similar to its distribution in the entire territory. Values closer to 100 indicate situations of maximum segregation. Table 6.1 shows that between 1986 and 2004 there was an increasing dissimilarity in the distribution of the population of Montevideo calculated on either an absolute or a relative measure of educational attainments. As such, in order to reach an egalitarian distribution of the population with low educational attainment, in the period 2002–2004 it would have been necessary to redistribute about 41 percent of the population in the department. Map 6.1 illustrates that in the last of these three periods, the 62 neighborhoods in the department were distributed according to three highly differentiated areas: one located in the southeast coast of the city with high levels of education; another with low levels of education in the
Table 6.1
Indices of segregation, 62 neighborhoods of Montevideo, 1986–2004 1986–1988
1995–1997
2002–2004
Variation 1986–1988 to 2002–2004
Percentage of 25 to 59 years old with up to an elementary education Duncan Index of 30.5 34.1 37.3 22.1 Dissimilarity Exposure Index Global Moran’s I by 0.78 0.79 0.79 Neighborhoods Percentage of 29 to 59 years old with educational attainment below the department mean Duncan Index of 31.7 37.0 41.5 30.9 Dissimilarity Exposure Index 210.5% 0.42 0.42 0.38 Global Moran’s I by 0.76 0.78 0.80 Neighborhoods
RESIDENTIAL SEGREGATION IN MONTEVIDEO
105
% Pop. Irregular Settlements 0–19.9 20–49.9 50–100
% Year of Education Below Mean 0–30 30.1–50 50.1–70 70 1–100
N E
W S
Map 6.1 Percent of people 25 to 59 years old with years of education below the department mean (2002–2004) and percentage of the population in irregular settlements by Census tracts (2004), Montevideo
peripheral areas of the department (in particular in the neighborhoods located in the northeast and west); and, a third area with intermediate levels of education in the central zone of the city. This picture coincides with many characteristics of the dynamics of its urban transformation, with growth in the periphery clearly led by low-income populations and a concentration of the most affluent population at the east of the department. Map 6.1 also shows the spatial location of the irregular settlements in Montevideo. The exposure index measures the probability that an individual shares their neighborhood with a different individual. Its interpretation indicates, for example, that, if in an area where a member of group X resides its value is 0.2, on average 2 of every 10 residents will be of group Y. Consequently, the value of this index will be very small in cases of high segregation (Martori and Hoberg, 2004). As changes in the values of the index depend on changes in the relative size of the social category under consideration, we seek to minimize these effects by presenting in table 6.1 only the percentage of the population ages 25 to 59 with educational attainment below the department mean, which, as we saw above, does not significantly vary between the two three-year periods used in this study.
106
KAZTMAN AND RETAMOSO
The decrease in the interaction between people with low and high qualifications in the neighborhoods of Montevideo is consistent with the increase in the homogeneity in their social composition previously documented. These figures attract attention because the weak decline in the index doesn’t appear to reflect, as do the previous indicators, either the severity of the processes of residential segregation or the rhythm in which the occupation of land and the establishment of irregular settlements grew during the period considered. A possible interpretation is that the educational threshold required for stable and protected jobs grew faster than the mean educational levels.12 Thus, the opportunities for interaction between people below and above the educational attainment mean may not reflect the opportunities for interaction between the poor and non-poor. This happens particularly within neighborhoods receiving mostly young families who, by virtue of the age of their adult members, may show average academic achievements significantly higher than those of previous generations at similar socioeconomic statuses. With respect to the indexes of spatial contiguity of geographic units with similar values on a given variable, our analysis shows that in the period from 1986 to 2004 the territorial area covering neighborhoods with similar socio-educational composition grew in size. It appears reasonable to assume that if significant effects of the homogeneity of the social composition of territorial units on the material situation and mental contents of residents is supported by empirical evidence, these effects would be more accentuated the larger the territorial area with the same social composition. To measure the relative frequency to which neighborhoods are or are not adjacent to others with the same social composition we use the global Moran’s I. A positive correlation reveals the existence of spatially contiguous units with similar values. A negative correlation indicates that high (low) values in a subunit neighbor low (high) values in another. It is important to remember that the Moran’s I characterizes a city globally, and as such allows for comparison with other cities or with the same city in different periods. As can be observed in table 6.1, there is no indication of significant changes in the isolation of people who reside in these territorial units with respect to those residing in neighborhoods with higher levels of education. Comparing the data with those from the 1996 census we find that, for neighborhoods, the results from the census coincide with those from survey data for 1995 to 1997 (0.78 and 0.79, respectively), whereas for census tracts the autocorrelation is higher (0.87).13
RESIDENTIAL SEGREGATION IN MONTEVIDEO
107
Residential Segregation and Education at the Elementary School Level in Montevideo. Challenges The preceding sections illustrated the alterations in the patterns and levels of residential segregation in the department. It has been shown that the most relevant characteristics of receiving neighborhoods were a rapid escalation in the number of irregular settlements; a predominance of families in the first stages of their life cycle; a relatively low level of educational attainment among adult residents; a low proportion of people with structured labor market trajectories and identities based on work experiences;14 and, unlike previous migratory processes, a majority of residents coming from central areas of the city rather than from other regions of the country. The new components of the urban scenario posed important challenges to the educational system, in particular to elementary education. Among other things, the available information clearly indicated that many children moved from areas where schools were relatively well equipped to areas with substantial deficit in school equipment and infrastructure. This recognition led to concentrating the educational efforts on problems of infrastructure. However, it was also clear that the most important challenge was not the lack of physical infrastructure but the new conditions in which children were socialized. The alterations in the social profile of school children caused by the transformations of neighborhoods and families demanded changes in the design of the curricula, in the training of teachers, and especially in the organization of the classroom activities now populated by students with a profile markedly different from those who attended public schools in previous years. Although it is difficult to disentangle the intricacy of the problems that the spatial concentration of new forms of poverty poses to the educational authorities, it should be recognized that the quality of the relationships that schools maintain with families and neighborhoods is a strong determinant of the efficacy of their role in the learning process. That is, a large part of the success of institutionalized education depends on the more or less harmonious way in which the efforts and influences of these three spheres interact. From this perspective, the impediments facing families concentrated in disadvantaged areas to complement the school efforts become more relevant: too many children, scarce human resources, weak ties to the
108
KAZTMAN AND RETAMOSO
labor market, low expectations for social mobility, and poor housing infrastructure. Neighborhood effects on the socialization of school children also become relevant: the tone of neighbors’ sociability; the quality of communal institutions; or the scarcity of peers and adults in the vicinity who can act as role models with regard to habits, behaviors, and expectations for school achievements. In sum, when families and neighborhoods fail to provide adequate support, it is more difficult for the educational system to develop its key role in the social integration process, namely, its ability to dissociate the educational achievements of poor children from their conditions of origin. We now turn to the reactions of the educational system to the new social reality defined by the concentration of poor families in disadvantaged neighborhoods. Policies Designed to Improve School Performance of Poor Children At the elementary school level, the Uruguayan educational system has responded to the worsening social situation of the school-age population through a series of measures whose design and application were consolidated as of 1995. In this section, we address three of those measures: special schools, the increase in preschool enrollment, and the provision of free lunches in schools. Special Schools As previously mentioned, the provision of elementary public education expanded in the nineties. Beginning with a traditional model, one which only distinguished between urban and rural schools, a move was made to a model based on a diversity of options. The diversity sought to address, through alternative schooling formats and compensatory schemes, the growing heterogeneity in childhood experiences. The design of educational policies focused on the most vulnerable sectors through two special programs: the Contexto Sociocultural Crítico (Critical Sociocultural Context, from now on CSCC) schools and the Tiempo Completo (Complete time, from now on TC) schools. On the basis of the transformation of public schools already existing in those areas, the TC schools began operating in the poorest geographic areas at the beginning of the nineties. In their first stages, their main contribution consisted solely of the extension of the school day. It is only in 1995 that these schools began to implement teaching programs specific to the institutions. After 1995, the rhythm in the construction of new
RESIDENTIAL SEGREGATION IN MONTEVIDEO
109
classrooms and in the transformation of ordinary public schools to TC was increased, and the TC teachers received a differential bonus salary.15 With respect to the CSCC schools, although they too are located in poor areas, they retain normal school hours and do not use a specific teaching program. To attract the most experienced teachers they rely on basic compensatory schemes. Between 1995 and 2004, enrollment in CSCC schools increased from 21.4 percent to 22.3 percent of total enrollment and enrollment in TC schools increased from 2.1 percent to 6.6 percent of total enrollment, whereas common public schools declined from 76.5 percent to 71.4 percent of total enrollment. These figures indicate a slight increase in the proportion of CSCC schools and a much more significant increase, although from small initial figures, in the proportion of TC schools. Similar to 2004, the distribution of special schools in 1995 was in line with the objective of the reform with most (47 percent) concentrated in neighborhoods of low socioeconomic composition. On the one hand, the majority of institutions specifically designed to increase educational opportunities among disadvantaged children were located in places with unfavorable school and neighborhood contexts. On the other hand, throughout the period, the proportion of special schools (CSCC more than TC) grew in neighborhoods of low and medium socioeconomic composition. There was also a strong increase of CSCC and TC schools in affluent neighborhoods perhaps because in 1995 the number of public schools in these neighborhoods was significantly smaller than in poorer neighborhoods. In addition, we cannot reject the hypothesis that some affluent neighborhoods contain small homogeneously poor areas and that, in those cases, public schools, which most of the time recruit in areas smaller than a neighborhood, enroll only children from those small areas. Furthermore, first- to sixth-grade enrollment in special schools is concentrated in areas with very unfavorable sociocultural profiles. As could be expected from previous analytical results, the authors also found that the enrollment in those schools was higher in receiving neighborhoods (44.9 percent) than in sending neighborhoods (15.7 percent). Preschool Education The expansion of public preschool educational services—strengthened as of 1995—was an important initiative against the mechanisms behind the intergenerational reproduction of poverty and social exclusion.16 Evidence recently accumulated on preschool services in Uruguay indicates at least three regularities. First, it clearly indicates an expansion in the education of children ages four and five. Second, it shows that preschool attendance reduces the chances of repeating grades in elementary school. Third, it also shows that the expansion of coverage mainly
110
KAZTMAN AND RETAMOSO
benefited children from the poorest social strata. As a whole, these three tendencies indicate clear advances in social equity as a result of transformations in the educational system.17 As a result of advances in preschool education, in 2004 more than 9 out of every 10 children 5 years old, and more than 7 of every 10 children 4 years old, attended either public or private preschool education. The public sector exerted a clear leading role in the expansion of coverage with a clear trend of increasing coverage from the early 1990s. The results of studies conducted by the Gerencia de Investigación y Evaluación de ANEP corroborate the importance of advances in preschool education for further school achievements.18 Graph 6.1 reveals the association between the length of preschool attendance and repetition of first grade for children who come from homes with differing levels of risk (as measured by mother’s educational level). The results show that the levels of repetition for children in risk situations with no preschool attendance are double those of children who attended preschool since age 4, as well to those of children who attended preschool from age 5 but do not come from disadvantaged families. Finally, with respect to the impact of these advances on equality, table 6.2 shows the growing presence of the public sector in providing preschool education in poor neighborhoods, as well as the consequent reduction in the gaps in preschool attendance among children residing in neighborhoods with different levels of disadvantage. Moreover, when controlling by the socioeconomic level of children’s households, the analysis shows that children coming from the most disadvantaged households practically doubled their enrollment during the period.19 Three factors appear to be the most relevant in explaining these advancements: first, the notable expansion in public preschool education; second, an increasing awareness among parents about the importance of early schooling for their children’s later academic achievement; and third, the undeniable functions of preschools as places where children are protected, cared for, and fed as a response to the deeply felt necessities of the poorest urban families. Free Lunch at Schools With the new characteristics of poverty in Montevideo, the need to feed children at schools became increasingly important. Evidence of malnutrition produced quick responses from educational authorities, driven not only by humanitarian concerns, but also by recognition of the impact of nutrition on children’s learning capacity. Existing programs of school meals were reinforced and new models introduced free lunches and the Copa de Leche system. These initiatives were part of the design of the new
Percent
RESIDENTIAL SEGREGATION IN MONTEVIDEO 45 40 35 30 25 20 15 10 5 0
111 42,1
34,7 25,0
21,6
20,7
6,2
From 4 years old
From 5 years old No Risk
No preschool attendan
Risk
Graph 6.1 Percent of children who repeated first grade in 2001, by age of preschool attendance and household risk level Source: ANEP-CODICEN, 2004.
Table 6.2 Percentage of public school enrollment for children ages 4 and 5 years old by educational level of the neighborhood, Montevideo, 1995 and 2004 Neighborhood educational level (terciles of the average years of study for the population 25 to 59, between 2002–04) Low Enrollment in public preschool 1995 40.9 2004 69.6 % of variation 1995–2004 70.2 Enrollment in public and private preschool 1995 56.8 2004 80.5
Medium
High
Total
41.8 66.3 58.8
32.6 44.0 34.9
38.4 62.3 62.3
74.3 88.4
83.0 93.8
70.3 90.2
Source: Authors’ own calculation based on ECH data from the INE.
public protection nets aimed at counteracting the effects of a worsening social situation. The nutritional policies implemented by the educational system receive strong financial backing from the impuesto de primaria (elementary education tax). Approved in 1986, but applied since 1990, this tax goes directly to finance a variety of areas (transportation, equipment, repairs, etc.), but the majority of the collected funds go to school nutritional programs.20 Suffice to say that according to a survey conducted by the INE in 2004, the school nutritional programs constitute the most extensive network of
112
KAZTMAN AND RETAMOSO
free meals in the country. Free lunches are most often served in schools located in the most disadvantaged neighborhoods. Furthermore, the proportion of students who eat lunch at school is directly proportional to the level of disadvantage of the school and its neighborhood, with over 74 percent of students in the most disadvantaged neighborhoods using the free-lunch system between half and all the time, compared to 25 percent of students in the most advantaged neighborhoods. Likewise in the schools with unfavorable sociocultural contexts, 74 percent of students use the free lunch system between half and all the time compared to 10 percent in schools with favorable sociocultural contexts. In sum, by assuming an important role in an area traditionally reserved for families, the widely extended free-lunch school network operated as another pillar of the public education system, particularly in neighborhoods of low socioeconomic status.
Final Considerations: Choosing Measures to Neutralize the Impact of Residential Segregation on the Academic Achievement of Poor Children In the past two decades, Montevideo experienced an important acceleration in its socioeconomic residential segregation. Several factors were responsible for this dramatic increase in residential segregation, the main one being a continuous deterioration of income and social protection levels among low-qualified urban workers. Thus, many young people, with little human capital and few stable and protected job opportunities, were expelled from central urban areas to the periphery of the city. Higher socioeconomic homogeneity in residency caused homogeneity in the use of local services, especially in those services, such as elementary schools, health care centers, transportation, and leisure, which usually recruit their users from the vicinities. In this way, to their weak links to the labor market, workers in the poorest neighborhoods in Montevideo added a growing isolation from the societal mainstream. These changes affected the profiles of households, neighborhoods, and schools—the three most significant contexts for the socialization of children. More households found it difficult to design strategies to satisfy their members´ basic needs. Maybe affected by a combination of weak ties with the labor market, declining expectations for social mobility, and growing isolation from the middle classes, poor urban household showed a trend toward unstable family arrangements and single parenthood units, which implied a growing percentage of children cohabiting with only one of their biological parents. Under these circumstances, the pool of family resources for socialization—especially time availability—was
RESIDENTIAL SEGREGATION IN MONTEVIDEO
113
affected, and parents found more difficulty in controlling their children, supporting their daily academic activities, and complementing the school efforts in the learning process. Schools, in turn, found it increasingly more difficult to foster learning abilities in groups of homogeneously poor students. Many teachers and school authorities working in areas of highly concentrated poverty found themselves overwhelmed by the level of unsatisfied basic needs of their students, by the weak support they received from their families, and by a notorious deficit in the cognitive development and in the basic social skills that children bring to the classroom. Beyond their own families and schools, neighborhoods are the next relevant human aggregate, providing the context within which poor children interact and the behaviors to which they are exposed. Neighbors may influence children’s attitudes and expectations through several paths, and likewise, peer groups in the vicinity influence children through mechanisms similar to those observed in schools, as mentioned further on. There are different potential routes of influence that adults may take: One is as role models, as successful examples of the use of certain means to achieve certain ends; another is as actors who help to define and support the normative patterns that regulate the relations between neighbors and that establish the tone and level of sociability; and, a third is as promoters of the positive or negative image that the neighborhood presents to the rest of the city, an image that may influence the formation of children’s identities and their feelings of self-esteem. A final form is as builders and/ or supporters of local institutions. In this sense, some of the new characteristics of poor urban neighborhoods tend to impair, rather than to support, their socializing effect. The instability of the patterns of communal life, the scarcity and low quality of neighborhood institutions, and the shortage of adults who could function as role models and who could exert efficient informal social controls on child behavior contribute to undermining the socializing force of the neighborhood or, at least, to detour it from the conventional paradigm that defines education as the principal path to social mobility and individual development. Putting all this together, the type of changes affecting each one of these important socializing agents—family, schools, and neighborhoods—have started a vicious cycle in no way beneficial to children’s educational performances. During Montevideo’s integrated past, the more harmonious interaction between the three spheres—each one complementing and reinforcing the others—facilitated the functioning of the educational system. The new characteristics of urban poverty have modified this relationship. As families and neighborhoods failed to provide their expected complementary roles, schools found formidable barriers to developing
114
KAZTMAN AND RETAMOSO
key roles in the processes of social integration on the basis of equity, that is, their unique capacity to dissociate children’s academic achievements for their socioeconomic backgrounds. A study developed by the UN Economic commission for Latin America and the Caribbean (ECLAC) office in Uruguay helped educational authorities to start thinking about how to strengthen the capacity of schools to overcome the difficulties imposed by the socioeconomic background of neighborhoods and families. The study addressed most of the problems discussed here and provided a detailed description of the state of teaching and learning in the country. It also offered critical information for the development of a proposal for the reform of the educational system oriented to disassociating children’s educational achievement from their socioeconomic backgrounds. As described in the body of this article, the reform of the Uruguayan educational system paid special attention to the location of facilities. It placed new centers for initial education (kindergartens)—special schools, and free-lunch school programs—in neighborhoods with the highest concentrations of poverty. Simply stated, the strategy consisted of focusing educational efforts in places where the risks for children to stay behind were higher.21 However, in the design of the interventions there is no sign that those risk situations stemming from the social composition of the schools or the social composition of the neighborhoods themselves were duly attended to. To what types of risk do we refer? The social composition of schools defines the peer groups with which a child will have opportunities for daily interactions. Studies of academic achievement show that at least four factors affect differential school performance. First, a child’s companions at school mold his or her expectations for academic achievement. Second, the more heterogeneous school peers are, the greater the variety of experiences and of problem-solving practices to which a child is exposed, and the greater the opportunities to develop cognitive and social skills useful for both their school performance and their future insertion in the labor market (Betts, Zau, and Rice, 2003). Third, similar positive effects on academic and labor attainments may be fostered by social networks based on heterogeneously constituted peer groups. Finally, among the most disadvantaged children, daily contact with their peers from other classes may create early feelings of sharing problems and destinies. Those feelings are important components of their general sense of belonging to a unique, single society, beyond present disparities in the material conditions of their families. One may assume that, for poor children, the relative importance of those early experiences of citizenship will be higher the more the microcosms of the classroom resemble the social composition of the whole society.
RESIDENTIAL SEGREGATION IN MONTEVIDEO
115
Unlike the type of geographically targeted interventions implemented in Montevideo, a policy design based on the above findings and hypotheses should have focused more on the interaction between different social groups. This could be done, for instance, by encouraging classroom heterogeneity through the localization of educational establishments on the boundaries between neighborhoods of differing social classes; making some minimum social heterogeneity mandatory for schools and providing free transportation to facilitate student attendance to socially mixed educational centers; or by promoting housing policies and housing subsidies favoring mixed neighborhoods. The consideration of all these factors leads to policy interventions in urban planning which go far beyond interventions limited to the educational system. This way of thinking about public policy seems to have influenced some European cities where initiatives to integrate foreign workers were implemented by defining the location of public housing construction or by giving economic incentives to members of the middle classes to reside in previously homogenously poor areas, thus assuring a certain level of social mixing in new neighborhoods (Musterd and Ostendorf, 1998). Having said this, we should consider that social-mixing urban policies are not necessarily easy to implement, nor do they guarantee direct reductions in social distances for all societies and all times.22 However, they are the policies that better address the research results about educational achievements and residency. In this sense, they are more faithful to research results than the well known and commonly implemented territorially targeted educational policies. Moreover, it seems that the educational system will not be able to elude social-mixing policies without risking its key role in the construction of societies integrated on the basis of equity. To advance toward this goal, the medium- and long-term fragmentations associated to the rising trend in both, educational segmentation and residential segregation of the new urban poor cannot be ignored. Notes A preliminary version of this document was presented at the “Urban Governance and Intraurban Population Differentials in Latin American Metropolitan Areas” workshop at the University of Texas at Austin, November 17–19, 2005. 1. The definition of a territory as urban or rural depends on the attributes of the department, as defined in the Law of Populated Centers. Generally the definitions are based on aspects associated to the existence (or lack thereof) of parceling of lands, of street routes, and of public services.
116
KAZTMAN AND RETAMOSO
2. In 1996, 14 percent of the population of Canelones was not residing in the department five years before the date of the census. Among them, 73 percent came from the department of Montevideo (Retamoso, 1999). 3. More recent estimates do not exist. 4. We use “metropolitan area” and “Greater Montevideo” interchangeably. 5. Between 1996 and 2004, 122,000 people emigrated from the country, a figure similar to that registered during the dictatorship period (1975–1985). Fiftyseven percent of these people were male, and 66 percent were between the ages of 20 and 39. 6. The 1989 plebiscite, when retirement benefits and pensions were adjusted for inflation, is perhaps the clearest indicator of the success of the older generations’ efforts to retain the benefits of the old regime. In addition, solidly established corporate arrangements allowed some senior adults to retain benefits acquired in the labor market. Their actions contributed to balkanize the market, preventing the access of new generations to protected and stable jobs. 7. The census tract is only one of the relevant geostatistical divisions. The hierarchical order is department, census section, census tract, and zone (in urban areas this coincides with the manzana). On average, a census tract in an urban area contains approximately 10 manzanas. 8. The studies cited above use data from censuses and household surveys. Whereas the first instrument allows using data at both the neighborhood and census-tract level, neighborhoods are the lowest statistically meaningful level of aggregation in surveys. Although at this level of aggregation surveys are weaker than censuses in terms of statistical representativeness, the reverse is true in terms of analytical potential and timely information. 9. In Uruguay, elementary schooling consists of six years of schooling (first through sixth grades) and begins at age six. 10. We excluded people 60 years and older because their inclusion might have complicated the interpretation of our results for two reasons: first, because this group has significantly more weight than other age groups in the central areas of the city, and second, because of the noteworthy expansion of education in the last decades implied an important widening of the educational gap between elder and younger adults. 11. For example, if we want to test the hypothesis that certain ethnic minorities exhibit a larger or smaller propensity to aggregate than others, we would want to compare the dissimilarity index for each of these groups in the city. On the other hand if we are interested in exploring whether disadvantaged urban groups are spatially isolated from people who participate in the main social and economic spheres of society, we would apply an index that reveals the degree of physical proximity between the two categories, assuming that physical proximity is a good proxy for potential interaction. 12. This is reflected, among other things, in the high levels of formal training (high school education or more) required for performing relatively simple tasks, such as those demanded for jobs in supermarkets or at gas stations (Kaztman, 1997).
RESIDENTIAL SEGREGATION IN MONTEVIDEO
117
13. The high values for the Moran’s I indicate that the mean education in a neighborhood is strongly correlated with the mean education in neighboring neighborhoods. The significance of these results was not lost on the authors, seeding doubt as to the idealness of the educational indicators employed to measure these types of groupings in cities that, as in Montevideo, demonstrate a strong concentration of mean educational strata. Using low income as an indicator, Carolina Flores finds for Santiago, Chik—a city that traditionally has been much more segmented than Montevideo—a Moran’s I far below that of Montevideo, 0.34. The counter intuitive character of is crude comparison points out to the convenience benefit of broadening the analysis by using the same indicators for the two cities (Flores, 2005). 14. On the relationship between the process of residential segregation and the characteristics of insertion in the labor market, see a previous work written by the authors (Kaztman and Retamoso, 2005). 15. There were also clear advances in the definition of the criteria by which ordinary schools were selected to become TC schools, especially with regard to the identification of schools recruiting poor children and the selection of areas with large increases in their infant population. A more detailed description of the types of schools selected can be found in ANEP–CODICEN, (2004) and Clavijo, Francia, and Retamoso (2005). 16. Initiated in November 1998, Law 17.015 established preschool education as obligatory for all children at age five. This law stipulated a maximum term of four years for the Administración Nacional de Educación Pública (ANEP) to provide the conditions necessary to enact the law. The measurement attempted, among other things, to foment the cognitive abilities and skills appropriate for a child’s age, to promote motor-sensory development and socialization, and to help in the prevention of the negative effects that biological, nutritional, familial deficiencies, or other risk situations, have on a child’s mental and physical development. 17. Sufficient evidence exists to support these claims. For a good synthesis of the findings with regards to the relationship between preschool attendance and achievements in elementary school, see ANEP–CODICEN (2002). 18. See ANEP (2005), “Panorama de la Educación en el Uruguay: Una década de transformaciones 1992–2004.” 19. See ANEP/MECAEP/UNIVERSIDAD CATOLICA (2003) “Determinantes de la regularidad de la asistencia y de la deserción en la educación inicial uruguaya.” 20. Ley No. 15.809 (Art. 636), Ley de Presupuesto Nacional de Recursos y Gastos de abril de 1986. It taxes urban, suburban and rural properties throughout the country on the basis of their declared values in the Dirección de Catastro Nacional. 21. Paradoxically, the spatial concentration of the poor facilitated the implementation of focalized strategies of educational improvement. 22. This is especially true for highly stratified societies such as the Latin American ones. In considering the relative costs of these types of policies, it should be taken into account that, in most societies, the present forms of
118
KAZTMAN AND RETAMOSO
class relationships are the final product of long processes with deep historical roots of domination, negotiation, and conflict resolution. This means that social engineering aimed at modifying existing patterns of class relationships and social distances may be one of the hardest areas for social policies. In the case of education, the resistance to interventions of the types reviewed above arise from all sides and are most fierce when the classes that circulate within the main social and economic city circles carry stereotypical images of those who have remained outside those circles; in such cases, the higher classes may perceive the promotion of social integration through schools as a threat to both their expectations with regard to their own children’s educational advancements and the maintenance of their traditional social prerogatives.
References ANEP-CODICEN (2002). Educación inicial: Logros, desafíos y alternativas estratégicas para la toma de decisiones. Gerencia General de Planeamiento y Gestión Educativa. ——— (2004). Monitor educativo de educación primaria. Tipos de escuela, contexto sociocultural escolar y resultados educativos. Segunda comunicación de resultados. Montevideo. ——— (2005). El Gasto Educativo en Cifras, Serie Estadísticas Educativas No.5. Montevideo, Uruguay. ANEP/MECAEP/UNIVERSIDAD CATOLICA (2003). Determinantes de la regularidad de la asistencia y de la deserción en la educación inicial Uruguaya. Montevideo. Betts, J., Zau, A., and Rice, L. (2003). Determinants of student achievement: New evidence from San Diego. San Diego, CA: Public Policy Institute of California. Cervini M. and Gallo, M. (2001). Un análisis de la exclusión social: La segregación residencial en los barrios de Montevideo, 1986–1998. Tesis de grado de la FCE. Montevideo, Uruguay. Clavijo, M., Francia, M., and Retamoso, A. (2007). Las escuelas de Tiempo Completo: Una manera de entender la enseñanza y el aprendizaje. Proyecto Hemisférico Revistade la Cepal, 91: 133–52. ECLAC (Comisión Económica para América Latina y el Caribe) (2001–2002). The social panorama of Latin America. Santiago, Chile: CEPAL. Flores, C. (2005). Residential segregation and the geography of opportunitties: A spatial analysis of heterogeneity and spillovers in education. Unpublished manuscript. INE (Instituto Nacional de Estadística) (1990). Las necesidades básicas en Uruguay, Montevideo. Kaztman, R. (1997). Marginalidad e integración social en Uruguay. Revista de la CEPAL, 62 (August).
RESIDENTIAL SEGREGATION IN MONTEVIDEO
119
———(1999). El vecindario importa. In R. Kaztman (coord.), Activos y estructuras de oportunidades (Capítulo 4). Estudios sobre las raíces de la vulnerabilidad social en Uruguay. Montevideo: CEPAL. ——— (2002). Convergencias y divergencias: exploración sobre los efectos de las nuevas modalidades de crecimiento sobre la estructura social de cuatro áreas metropolitanas. Kaztman, R. and Filgueira, F. (2001). Panorama de la infancia y la familia en Uruguay. Montevideo: IPES—Universidad Católica del Uruguay—IN. Kaztman, R., Filgueira, F., and Errandonea, F. (2005). La ciudad fragmentada: Respuesta de los sectores populares urbanos a las transformaciones del mercado y del territorio en Montevideo. In B. Roberts, A. Portes and A. Grimson (eds.), Ciudades Latinoamericanas: Un análisis comparativo en el umbral del nuevo siglo. Buenos Aires, Argentina: Prometeo. Kaztman, R. and Retamoso, A. (2005). Segregación residencial, empleo y pobreza en Montevideo. Revista de la CEPAL, 85 (April). Macadar, D., Calvo, J. J., Pellegrino, A., and Vigorito, A. (2002). Proyecto segregación residencial en Montevideo: Un fenómeno creciente? Montevideo: CSIC. Martori, J. C. and Hoberg, K. (2004). Indicadores cuantitativos de segregación residencial. El caso de la población inmigrante en Barcelona. Geo Crítica (Universidad de Barcelona), 8(169). http://www.ub.es/geocrit/sn/sn-169.htm. Musterd, S. and Ostendorf, W. (1998). Segregation and social participation in a welfare state. In S. Musterd and W. Ostendorf (eds.), Urban segregation and the welfare state: Inequality and exclusion in western cities. London: Routledge. PNUD (1999). Desarollo Humano en Uruguay. Programa de Naciones Unida s para el Desarollo. Monterideo, Uruguay.
This page intentionally left blank
7
Residential Segregation in Santiago: Scale-Related Effects and Trends, 1992–2002 Francisco Sabatini, Guillermo Wormald, Carlos Sierralta, and Paul A. Peters
Introduction It is frequently argued that social inequality and residential segregation are directly associated. Given that social inequality has increased in recent times or has stabilized at high levels while residential segregation has remained high, there is a factual basis for making such an association. Underlying this association is an assumption about the symmetry between social differences and spatial forms. The theoretical and epistemological correlation of this assumption is that society reflects itself in space: an unequal society is reflected in segregated cities. Therefore, society can be understood without any reference to its spatial dimensions. At its extreme, this perspective holds that society would have no relation to space, which is absurd. A similar perspective is found in the new field of “global cities” research where researchers focus their attention on the spatial “effects” of economic globalization. In particular, the post-Fordist economy seems to exist ex ante its effects on the city and the space. This understanding of the relationship between space and society is not limited to urban studies and social sciences in Latin America. The social sciences have always had problems explaining the relationship between space and society. Spatial features have been reduced to empirical manifestations of unobservable realities or externalities, which are considered the true causes. Viewing cities and their segregation levels as observable
122
SABATINI, WORMALD, SIERRALTA, AND PETERS
evidence would allow us to better understand social and economic processes that are unobservable. In a sort of intellectual apostasy, urban scholars reduce the study of the object that defines their discipline, the city, to a mere epiphenomenon or the late-coming manifestation of a chain of underlying social facts, most of them unobservable. The chain of facts ends in observable phenomena, namely, spatial facts and their material basis. This idea of symmetry between social and spatial structures, which can be described as a “mirror thesis,” is a simple way to solve the theoretical and methodological difficulties that the relationship between space and society poses. Because of the simplicity of this thesis, it is also popular. However, the intellectual sacrifice it involves creates serious reductionism: researchers look for facts that fit a specific conclusion, namely, that segregation has necessarily increased because inequality has increased and new forms of social exclusion have emerged. To what extent does the study of segregation that is popular among urban contemporary scholars in Latin America suffer from this type of reductionism? Segregation trends in Santiago during the last decades, nonetheless, contradict the mirror hypothesis. Segregation is decreasing according to measures of different social dimensions and geographic scales. At the same time, inequality and new forms of social exclusion are consolidating, especially in labor markets and the political arena. This is the topic of the first two sections following the introduction of this chapter. The first section describes the characteristics of existing social inequality in Santiago. The second section discusses the various factors that underlie the results. We argue that these factors are also present in other cities in Latin America, given their relationship to neo liberal economic reforms. Even though residential segregation is declining in Santiago, its social effects have become more severe among the poor compared to the period prior to the economic reform at the end of the 1970s. This second paradoxical result is presented in the section, Trends in Residential Segregation. We relate this outcome to increases in labor instability and the weakening of links between poorer sectors of the population on the one hand, and political parties and the political system on the other. The section ends with a presentation of statistical results for Santiago. We also apply other statistical methods to measure the effects of segregation among the urban poor. Finally, we present our conclusions. Dynamics of Social Inequality: Socioeconomic Trends during the Past Two Decades As the World Bank has noted, inequality is not a new problem in Latin America (World Bank, 2004). Inequality in the Latin American context
RESIDENTIAL SEGREGATION IN SANTIAGO
123
is important both for the causes of the phenomenon and the forms that it acquires, resulting from inequality processes that are currently developing. According to the World Bank, there are two main causes for the reproduction of inequality in Latin America: First, key markets (notably those for capital, labor, land and education) tend to fail and thus exclude disadvantaged groups despite their productive potential; secondly, the unequal distribution of political influences means that those in stronger positions are systematically favored in the allocation of resources through market and government institutions alike. (World Bank, 2004, p. 361)
Therefore, inequality in Latin America can be explained, on the one hand, by the improper functioning of markets that are fundamental for welfare provision and, on the other hand, by the unequal distribution of political influences, as well as the absence of effective government institutions able to correct these market failures. In this vein, the new trends of social inequality in Chile are related to the replacement of the economic model, from an import-substitutingindustrialization model to an open-market (neoliberal) model. This neoliberal model is oriented toward exports and based on the principles of economic liberalism. This model, which has been considerably modified during the last two decades, has had a profound impact on the country’s socioeconomic structure. Two of the most important consequences of the new economic model are (a) the differentiation and segmentation of the structure of opportunities in labor, education, health, and social security dimensions (supply segmentation) and (b) the emergence of new forms of accessing opportunities in these fields, according to one’s socioeconomic status (demand segmentation). These two forces combine to generate high inequality in the access to opportunities for integration, both in terms of social welfare and social citizenship (Wormald et al., 2002). One of the characteristics that distinguish Chile within Latin America is the continual decline of poverty since the 1990s. According to data from the La Encuesta de Caracterización Socioeconómica (CASEN, a nationally representative household survey), the percentage of individuals and households in poverty decreased by more than half between 1990 and 2003.1 The decline is primarily related to economic growth and the active participation of the state in favoring the poorest groups. However, the decline of poverty has been accompanied by the persistence of income inequality and a new environment of social vulnerability that affects low-income households, both poor and non-poor, as will be discussed later.
124
SABATINI, WORMALD, SIERRALTA, AND PETERS
Between 1990 and 2003, the proportion of income of the lowest income quintile did not change substantially, remaining near 4.4 percent. However, 56 percent of the national income was concentrated in households from the upper quintile (CASEN, 1990; 2003). Hence, even though the total “pie” increased, yielding improvements for all socioeconomic groups, income inequality remained constant. This reinforces Chile as one of the most unequal countries in Latin America at the end of the twentieth century (World Bank, 2004, p. 2). The decline of poverty has been offset by the unbalanced integration of opportunities for improved social wellbeing, even for those households that have surpassed the official poverty line. Thus, a “grey zone” of social vulnerability has emerged. Using CASEN panel data from the 1996 to 2001 period, a recent study found that social mobility increased for the population in the first nine income deciles.2 However, upward mobility has not been the only kind of mobility experienced. On the contrary, the study shows that nearly 50 percent of the urban population is likely to fall below the poverty line. This is partially because households in the two poorest income quintiles are particularly vulnerable to unemployment or illness of the household head. In these cases, households are unable to retain their socioeconomic status (Contreras et al., 2004). Additionally, the vulnerability of the poor is also related to their unsteady insertion in the labor market, especially given the generally low level of skills. In this context, several aspects are worth highlighting. First, the proportion of informal workers in the urban economy is high. Using the International Labour Organization’s (ILO) definition of informal employment, more than a third of the country’s working population was employed in the informal sector in 2000 (35.4 percent). In the metropolitan region, 34 percent of the labor force was informal (Wormald and Salinas, 2003). Among hourly wage workers, the percentage lacking a written contract in the metropolitan region increased from 11.8 percent in 1990 to 15.8 percent in 1998. This shift suggests more uncertain access to health care and social security provided by formal contracts (Wormald et al., 2002). In 2003, the proportion of hourly-wage workers lacking a written contract in the country was 22.3 percent, and 68.6 percent of them belonged to the lowest two income quintiles (MIDEPLAN, 2004). As such, the greatest proportion of workers without the proper social security and health benefits is concentrated in poor households. Second, unemployment during this period affected the poorest households more strongly, both in periods of economic expansion and contraction (Sabatini and Wormald, 2004). For example, in 2004, a year in which the economy was growing, 66.9 percent of the unemployed population (9.9 percent of the labor force) belonged to the lowest
RESIDENTIAL SEGREGATION IN SANTIAGO
125
two income quintiles, whereas just 5.9 percent belonged to the highest income quintile. Additionally, the entry of women into the labor force is much lower among the poor. In 2005, for households in which both members of a couple were present, 58.8 percent were in the highest income quintile when both partners were working, whereas both partners worked in only 8.6 percent of the lowest quintile (MIDEPLAN, 2004). Therefore, among the poor, female labor force participation is low and unemployment high. When employed, the poor are more likely to be working in precarious jobs and mobilizing fewer resources in the labor market. Third, there are important differences in access to educational opportunities for households from different socioeconomic groups (Wormald et al., 2002). Supply and access to high-quality education are strongly segmented within Chile and especially within Santiago. The educational system in Chile is a mix of public and private schools with public schools having lower educational outcomes than their private counterparts. Children from low-income families generally attend public schools, given a lack of household economic resources to afford private schooling. As a consequence, the equality of opportunities underlying development models has been eroded due to the segmentation of social opportunities according to socioeconomic status. Finally, despite advances toward equity, Chile does not appear to be exempt from the cooption of the political system by the elite. The state has attempted to reduce the inequality generated by the current economic model through the implementation of diverse social programs. In 2003, the mean income of the poorest quintile households increased by 11.7 percent due to different subsidy programs (CASEN, 2003). However, the power structures in the economic, political, and social functioning of the country have often been denounced. The president of the Chilean Federation of Industry (SOFOFA), which includes large and mediumsized manufacturing firms, publicly stated that “Chile will not change if the elites don’t stop sucking the teat, and I don’t think the economic and political elites will decide to let it go.” (Lamarca, 2005) Labor has lost importance as a basis for social identity and action with Chilean workers having a diminished ability to organize and to exert political influence. At the same time, the legitimacy of political parties has been reduced, further weakening the ability of labor unions to pressure the state. Even though free-market policies encourage processes of individual integration and mobility, they reduce the opportunities for the emergence of collective actors able to impact the political system. Politics has emerged as a professional activity that is limited to power games between the state and the private sector. In Chile, as in other
126
SABATINI, WORMALD, SIERRALTA, AND PETERS
countries, a strategic alliance between private interests and the state’s supporting a liberal and global economy has been established. Political apathy is expected considering the indifference toward political parties, unwillingness to register to vote, and distrust in politicians. In this sense, apathy is more than a simple lack of interest in public issues or in politics in general. Political apathy is also a consequence of the way formal politics is currently practiced. Increasingly, politics appears as a marketing activity, relying on the use of mass media. Citizens have shifted their interest to local problems that affect a reduced number of people, numbers much lower than the mass media cares about. This increases the political exclusion of citizens, especially the poor. The infrastructure and service problems that affect poor neighborhoods or the environmental impact of projects that hinder quality of life, among others issues, only interest politicians to the extent they can be expressed in terms of more general, non-targeted, actions or proposals. Building from these observations, territory and place of residence become increasingly important in the social organization of the urban poor. There is some evidence that these groups are able to organize themselves in order to reclaim their rights for a better quality of life in the city (Sabatini and Wormald, 2004). This ability for action could be the source of new forms of social organization and empowerment that would allow them to fight against the negative effects associated with vulnerability and unequal access to social citizenship. However, these forms of actual or potential movements have a tenuous relationship with the formal political system. The weakening and discrediting of political parties are worrying issues when considering the long-term stability of the democratic system. These thoughts beg a more general question about the relationship between the socioeconomic aspects of inequality and current trends in residential segregation, especially in metropolitan areas. How are these two phenomena related? How does spatial segregation contribute to the “geography of opportunities” (Galster and Killen, 1995) that the poor face when trying to improve their well being?
Trends in Residential Segregation Although social inequality remains high and has adopted new forms in recent years, quantitative levels of residential segregation in Santiago decreased between 1992 and 2002. The decline was concentrated in highincome groups (the wealthiest 10 percent of households). This decline is largely explained by the gradual dispersion of these groups from the group
RESIDENTIAL SEGREGATION IN SANTIAGO
127
of municipalities that form the Barrio Alto.3 Upper-income groups have moved into other municipalities at the urban periphery, including municipalities where the poor typically lived. Residential segregation has also declined among the poor (the poorest 10 percent of households). Before presenting a quantitative description of these trends and discussing the underlying factors, we explain the methodology used in this work.
Methodological Approach Segregation is defined in this chapter along two objective dimensions: evenness and isolation. Further in this chapter we expand on a third, subjective dimension, which is important to explain the ghettoization that poor Santiago neighborhoods are currently experiencing. The first dimension of segregation is most often captured by index of dissimilarity discussed in chapter 2. We measure this dimension for each of the groups we will consider relative to the other groups as a whole. The second dimension measures the degree of social homogeneity within a given spatial unit in the city. This is captured by the isolation index, which measures the probability of interaction between families of the same socioeconomic status, as discussed in chapter 2. Whereas some authors have shown that these two indices are highly correlated for metropolitan areas in the United States, this is not necessarily the case in Latin America (Glaeser, 2001). For example, in Santiago high-income groups score high on segregation according to Dimension 1, but low according to Dimension 2. At the same time, the suburbanization of Santiago during the twentieth century concentrated the elite in a higher-income cone. However, unlike in many other world regions, high-income areas in Latin America can be among the most socially diverse areas within the city.4 While the elites practically removed themselves from other areas in Santiago, they shared the Barrio Alto with other social groups from the middle and lower-middle class. This feature corresponds to a variety of cultural patterns and is important to keep in mind when thinking about the future of segregation and the public policies influencing urban growth patterns. We will return to this topic later in the chapter. Households have been classified by socioeconomic group, following the standard used in social research and marketing studies in Chile. This scheme distinguishes five groups: ABC1 or the elite (10 percent); C2 or upper-middle (20 percent); C3 or middle (25 percent); D or lower-middle (35 percent); and E or low (10 percent). The groups were calculated using the measurements of education of the household head and ownership
128
SABATINI, WORMALD, SIERRALTA, AND PETERS
of a particular set of household consumer goods. Since each group had different rates of change between 1992 and 2002, with groups D and C3 increasing their share of households by three percentage points mainly at the expense of group E, the size of each group was adjusted in order to reflect inflections of the curves that produce these indices. As previously stated in the section Dynamics of Social Inequality: Socioeconomic Trends during the Past Two Decades, the number of households below the official poverty line has declined in Santiago, reaching 11 percent. This relatively low percentage contrasts with the fact that the lowest socioeconomic groups (D and E) represent 45 percent of the population. As stated earlier, low socioeconomic groups are not characterized as accurately by their poverty status as by their vulnerability to economic crises, health, or familial problems. This vulnerability has transformed residential segregation into a negative condition. The two selected indices were measured at different spatial scales: municipalities, census districts, census tracts, and city blocks.5 We will examine the evolution of segregation at these different scales. We do not believe the measurement of residential segregation at a lower scale is more precise than its measurement at a higher scale. Rather, we think each represents different phenomena—different scales embody realities that are qualitatively different and nest different, even contradictory phenomena (Wagensberg, 2004). In this vein, it is remarkable how studies of residential segregation have been dominated by the philosophically atomist idea that one can represent residential segregation by a single number that describes the lowest spatial scale.6 A better approach sees social systems as more than the simple addition of social aggregates, as is intuitively clear when looking at urban ghettos of poverty and discrimination. Within urban ghettos, important phenomena emerge, phenomena that are irreducible to the individuals or activities that compose these neighborhoods (Holland, 1998). The issue of scale is most often introduced in quantitative studies as a new dimension of segregation that is measured separately. In this vein, Massey and Denton (1993) included clustering as one of the five elements of their well-known multidimensional typology. It is possible, however, that clustering and residential segregation are mere synonyms, the first being just an attempt at capturing segregation at a higher spatial scale. We will limit this study to the two formerly mentioned objective dimensions of segregation and assume that clustering is an issue related to the scale at which segregation is calculated.7 A last methodological issue is the spatial definition of Santiago as a city. Here we recognize the relatively compact urban development of Santiago comprises 34 contiguous municipalities. Nonetheless, as with
RESIDENTIAL SEGREGATION IN SANTIAGO
129
many other cities in the world, Santiago’s density has declined in recent decades, and its shape, or urban-rural borders, have started to fade. To some extent, this is due to the proliferation of leisure homes on half an hectare of land or more in the rural-border areas that have become an alternative to the main residence or a weekend house for elite families and the middle class. This process is part of the spatial dispersion of the elite. Strictly, it would be necessary to study the evolution of segregation indices for a second definition of Santiago, one that also considers municipalities nearby the urban sprawl or compact city, where most of the leisure homes are located.
Elite Dispersion and the Decrease of Residential Segregation As in many other cities of the world, in Santiago the distribution of spatial concentration of different social groups, measured by the dissimilarity index, adopts the shape of a J curve with high-income groups being the most segregated and mid-income groups the least. Low-income groups are in an intermediate position. Varying the geographic scale of measure does not yield differences in the distribution of the dissimilarity index. We calculate the dissimilarity index by different scales of measurement, and include a diagonal line that serves as a methodological correction. According to this calculation, segregation decreases as the geographic scale increases. The higher the scale, the lower the index. The index approaches zero when the scale approaches the city level. The lower the scale, the higher the index, which approaches one at the minimum size available, the household level. The literature on segregation has referred to the variability of the dissimilarity index at different geographic scales as a grid problem. It is usually considered a methodological problem, and it is recommended that one use the same scale when different cities are studied, or when the same city is studied at different points in time. Recalling these last considerations, it may be argued that the grid problem does not reflect imperfections or distortions of the index but an expectable outcome of the change of scale. Changes in the level of spatial concentration for different socioeconomic groups are more pronounced at higher scales, namely at the municipal level, than at lower scales where the changes are less pronounced. In addition, the spatial segregation of the elites diminished between 1992 and 2002, and the decrease was higher precisely at the municipality level, where segregation was stronger than would be expected at that level. The area above the diagonal and below the curve of the ABC1 line (the elites)
130
SABATINI, WORMALD, SIERRALTA, AND PETERS
diminished between 1992 and 2002, whereas the area under the diagonal and above the ABC1 line increased (see figure 7.1). In the same vein, the level of spatial concentration of the E socioeconomic group, that is, the poorest households also decreased at the municipality scale. We have argued that since the start of the neoliberal reform, a transformation of the patterns of segregation of Latin American cities has initiated. Because Chile was the forerunner of these economic reforms, Santiago showed the spatial changes. Two characteristics of the transformation of segregation patterns are remarkable. On the one hand, there is the spatial dispersion of the elites, and, on the other hand, the physical proximity of the poor to the elites and to areas of higher urban dynamism (subcenters). We consider both to be contemporary tendencies and potential structural changes rather than quantitatively categorical facts.8 Regarding the dispersion of the elites, our argument relies on a series of factors and processes related to the emergence and growth of a different real estate economy, which has the ability to transform the urban structure. These factors are the liberalization of the economy and, particularly, of the urban land markets; the concentration and internationalization of capital and, especially, the development of real estate investment capital; and, the higher investment in urban infrastructure networks at the regional scale (highways, pipelines, metropolitan transportation systems, telephone and telecommunication systems, and fiber optics networks). The major private real estate projects, whether residential (mainly gated communities), commercial (shopping centers and big supermarkets), or office complex projects represent an ability to transform the city that private projects did not have prior to the economic reform when the real estate sector’s supply was fragmented and there were no private mega-projects. The recent exit of elite families—principally young families—from the high-income area where the elites virtually confined themselves until the circa 1980 economic reform is one outcome of these new real estate projects.9 These projects are attractive for private firms because they create a new way of maximizing the profits from land: the firms buy land, often in the popular suburbs where residents are employed in low-wage sectors and lower-level land uses result in low land prices. After developing a project, developers sell the land to better-off socioeconomic groups who employ it in higher-order uses, thus increasing the price. The development of infrastructure networks, promoted by the public authorities in order to gain competitiveness for their cities and countries, favors the spatial dispersion of the elites. The utilization of the distant periphery in the areas that surround the urban sprawl follows the same economic logic and lowers density as the urban-rural border fades. The design of gated communities and the large scale of projects are factors that contribute to
RESIDENTIAL SEGREGATION IN SANTIAGO
131
the allocation of affluent families in the middle of low-socioeconomicstatus neighborhoods. The spatial concentration of the elite in the traditional high-income cone is still quite marked, as is also the concentration of poor groups in the periphery. The elite families that move out of the cone create economic assets for themselves by colonizing new areas in the poor periphery. This is how new socially prestigious neighborhoods have emerged along with new opportunities for these affluent families in terms of increasing the value of their properties. In more recent years, the gentrification process has reached the whole socioeconomic spectrum, extending to different areas in the city including residential projects as well as commercial projects (shopping centers and big supermarkets) and office complexes. Applying the concept of gentrification to non-residential phenomena is non-orthodox, but it is a way to emphasize this generalized process in Latin American big cities. Whereas spatial segregation in the United States seems to play an important role in the configuration of new social identities emerging from massive processes of social mobility, in the unequal, hierarchical, and less dynamic social structures of Latin America, this does not happen, or, at least, it does not happen to the same extent. Rather, spatial segregation in Latin American cities is one strategy for increasing real estate value in the long run, a strategy that families from all socioeconomic groups try to copy. Inflationary and poor economies that yield uncertainty enhance opportunities for investment in what traditionally has been considered one of the few sure assets: urban properties in cities that grow at accelerated demographic and geographic rates. This may also apply to segregation in the United States, but we argue that this process is comparatively more important in Latin American cities. The second trend—the tendency of poor groups to seek out areas where better-off groups live or areas of higher urban dynamism—is related to the increase of employment instability and, probably, to the new forms of political exclusion we discussed earlier. In the context of an economy that dramatically increases uncertainty, especially among the poor, and of a political system in which the parties have separated from their social bases, spatial segregation appears as another factor that deepens social exclusion. The geography of opportunities is now more relevant than in former times. In the past, the potential for more organic and stable labor or political insertion made spatial segregation less negative than in the current context. These two tendencies of change in the patterns of segregation seem to combine differently in different countries. In Brazil, the tendency of the poor to approach the elites and more dynamic areas seems to be more
132
SABATINI, WORMALD, SIERRALTA, AND PETERS
patent than in Chile. Small-size favelas have multiplied in the metropolitan areas of São Paulo during the last decades (Marques and Torres, 2005). In Rio de Janeiro, where the favelas are larger and have a longer history, the densest favelas, on average, are near the beach and affluent neighborhoods. In Santiago, in contrast, partially due the liberalization of the land markets and the dramatic change in the real estate sector and partially due to their relatively small size of poor informal groups and massive public housing programs, the dispersion of the elites, is more evident. The most spatially isolated socioeconomic group in Santiago is group D, representing approximately 35 percent of the households and including the beneficiaries of public housing policies (see figure 7.1). Conversely, the least isolated is group E, the poorest representing about 10 percent of households. This group does not have comparable access to housing policies to the same extent as group D, which would explain the differences in the levels of segregation between the poor groups. More than group D, Group E is forced to relocate close to affluent areas with the goal of improving urban opportunities.10 In general terms, figure 7.1 shows that the level of spatial isolation of different socioeconomic groups is moderate, particularly for the elites, 50.00 45.00 40.00 35.00
%
30.00 25.00 20.00 15.00 10.00 5.00 0.00 E
D
Municipality
C3 Group District
C2
Tract
ABC1
Block
Figure 7.1 Residential segregation in Santiago, 2002 (Dimension 2), by socioeconomic group, at different scales (Isolation Index)
RESIDENTIAL SEGREGATION IN SANTIAGO
133
as we stressed before. As we will see later, segregation according to Dimension 2 is especially harmful among low-income groups, precisely where it is higher. On the contrary, for high-income groups the configuration of socially homogeneous areas, or Dimension 2 of segregation, does not offer significant advantages; whereas, spatial concentration (Dimension 1) does. This result is consistent with the two most classic notorious characteristics of the traditional segregation patterns in Latin American cities: the spatial concentration of the elites and the configuration of vast homogenously poor areas. In line with this analysis, between 1992 and 2002 the evolution of segregation according to Dimension 2 also involved declines in segregation for the elites in Santiago and, to a lower extent, for groups E (the poorest) and C3 (mid-income groups), as well as a slight increase in the spatial isolation of the most segregated households—those belonging to group D (poor) and C2 (upper-middle-income groups).
Residential Segregation in Latin American Cities Important theoretical and practical consequences derive from the contrasting findings regarding Dimension 1 and 2 of segregation for the elites and the areas where they live. We hypothesize that the catholic ethos of Latin American cities, vis-à-vis the protestant ethos that was in the origin of Anglo-American cities, or “city of suburbs” (Fishman, 1987), implies a concept of “the Other” that makes spatial segregation by socioeconomic status in Latin America less marked than in American cities. The difference may also be related to a type of urban adolescence that has dominated American cities since the Second World Word, due to massive and dynamic processes of upward social mobility. In their attempt to build and defend new social identities, socioeconomic groups turn to segregation, especially to the social homogeneity of their neighborhoods (segregation according to Dimension 2), as an additional resource they can use to differentiate themselves from the lower socioeconomic groups from which they are emerging (Sennett, 1971).11 Typically, in Latin American cities the elites represent just a third of the population of high-income areas. Even at the block level, the isolation index of the elites in Santiago did not exceed 40 percent in 2002 (figure 7.1). American suburbs, on the contrary, are more socially homogeneous both in racial and income composition, and this homogeneity is more stable through time. In Latin American cities, the neighborhoods that emerge as exclusive or elite are gradually penetrated by families of lower socioeconomic status,
134
SABATINI, WORMALD, SIERRALTA, AND PETERS
especially through multiunit, multistory building. Multistory buildings are explicitly prohibited and resisted in the suburbs of American cities, because they are seen as a means by which lower socioeconomic status groups can locate in the area. The higher-density, multistory buildings allow individual residents to afford the more expensive prices of the land; and ensure profits for real estate brokers. The fact that Latin American elites do not defend their neighborhoods in ways found in the United States is a peculiar cultural characteristic and may reveal a relatively greater openness to social mixture in shared space. Latin American researchers have not sufficiently studied this phenomenon, due to the oversimplification found in the mirror thesis and to the tacit acceptance of the idea that Latin America imitates the urban patterns and processes of the main capitalist country, the United States.
Trends in Residential Segregation According to Comparative Measures In this section we present and interpret the empirical measurement of residential segregation in Santiago (see table 7.1). Following the earlier analysis, the new analysis highlights the evidence that segregation in Santiago has not increased during the last inter-census period, in spite of the persistence of high levels of social inequality and of the development of the private real estate sector, which relies on increases in social segregation. In the presence of economic growth, a constant income distribution, as the one we described in the section Dynamics of Social Inequality: Socioeconomic Trends during the Past Two Decades previously, translates into increasing social distances among groups. With a few exceptions this has been the case in Chile since 1984. The dissimilarity and isolation indices utilize an absolute and a relative stratification of households according to the educational level of the head of the household (see table 7.2). The first classification distinguishes between heads of household that have and have not finished high school. The relative stratification distinguishes between those who are above and below the mean years of schooling of the city’s heads of the household. The spatial segregation, defined by the absolute stratification, shows that clusters appear to be evolving toward higher dispersion; this is to say, toward a decline of segregation according to Dimension 1. The relative stratification clusters indicate an increase in spatial concentration. However, whereas the absolute stratification clusters approximate real socioeconomic groups, the relative stratification clusters are actually taxonomic groups or social abstractions. Therefore, the more convincing change is the
135
RESIDENTIAL SEGREGATION IN SANTIAGO
Table 7.1 Segregation of education measures (0.000) and their changes at multiple levels, 1992 and 2002 Dissimilarity Index Absolute 1992 2002 Census Block Census Tract Census District Municipality
Table 7.2 and 2002
Exposure Index
Relative
D
1992 2002
Absolute D
1992 2002
Relative D
1992 2002
D
467
441 22.59 417
442
2.56 267
410 14.29 419
441 20.83
401
380 22.07 349
381
3.18 292 445 15.31 455
447 20.83
362
349 21.28 316
389
3.35 302 460 15.76 469
462 20.66
319
276 24.24 275
272 20.25 318 488 17.06 489
488 20.05
Simple correlations between social problems and segregation, 1992 1992
Youth Inactivity* Youth Unemployment* Unemployment* Teenage Fertility**
2002
Municipal
District
Tract
Municipal
District
Tract
0.911 0.703 0.792 0.214
0.831 0.589 0.674 0.131
0.734 0.374 0.555 0.095
0.902 0.570 0.790 0.017
0.722 0.284 0.847 0.033
0.567 0.177 0.626 20.011
* Significant at 0.01 level. ** Not significant.
first one. In addition, the reduction of segregation at the larger geographic scale corresponding to municipalities—or “decline of the scale of segregation,” according to the terminology we have introduced—verifies for both dimensions. Segregation according to Dimension 2, or the social homogeneity of space, is the dimension most clearly associated to negative social and urban effects. When we analyze this data, the reversal of residential segregation in Santiago is evident. The increase in exposure (to the interaction with other groups), that is, the reduction of social isolation, is again verified for the absolute stratification, whereas there are no changes for the relative stratification. The distribution of social isolation among different socioeconomicstatus households could be attributed to the spatial dispersion of mid- and high-income groups, especially the elites, to areas composed mainly of groups of lower socioeconomic status. The dispersion can be explained by young families’ departing the high-income cone, or, complementarily,
136
SABATINI, WORMALD, SIERRALTA, AND PETERS
by socially ascending households that did not previously live in that cone. We have argued that because of the slight relative increase of elite households in the city between 1992 and 2002, the main reason underlying the reduction in social isolation of the elites is likely to be their departure from the high-income cone. Finally, the proliferation of gated-community-type projects in the urban periphery directed to high- and mid-income groups, supports our argument as well. The calculations for spatial clustering (Moran’s I) shows that from 1992 to 2002 the index of clustering for the different socioeconomic groups combined diminishes for the absolute measure of education from 0.7276 to 0.6964, whereas it slightly increases on the relative measure from 0.6861 to 0.6907. Mapping the LISA clusters for absolute minority group proportions for 1992 and 2002 clearly shows that between the two years new residential clusters of better-off socioeconomic groups emerged, dispersed throughout the city, whereas the lower socioeconomic status clusters did not change significantly (map not shown).
Social Effects of Residential Segregation Despite the reduction of residential segregation in Santiago between 1992 and 2002, there were nevertheless social disintegration effects on the poor (groups D and E, about 45 percent of the population). Effects on unemployment and youth inactivity are considered here. The analysis of social effects of segregation rely on the relationship between the levels of spatial isolation (Dimension 2) of the poor (groups D and E) and the existence of social problems in those households. Census data allows the study of the following variables: unemployment of the head of household; youth unemployment (15–24 years old); and, teenage fertility (15–19 years old). Within a household, youth inactivity refers to the existence of young people who do not study or work, and teenage fertility to the presence of women aged 15–19 years who are mothers, regardless of their marital status. The paradox of declining segregation and social disintegration effects is attributed, partially, to the advancement of social exclusion in the three areas we formerly discussed: (1) employment instability; (2) segmentation in the access to services and social security; and, (3) disconnection between the social bases and formal politics. Currently, residential segregation is lower, but it has a more deleterious effect. Furthermore, a more detailed spatial analysis could show that since the recovery of democracy in 1990, segregation has become more pernicious due to the location of the social
RESIDENTIAL SEGREGATION IN SANTIAGO
137
housing projects. Since 1990, democratic governments have conducted massive social housing programs, reducing the so-called “housing-deficit” in Santiago and the country as a whole. But these (mainly high-density) projects are highly segregated in location. Table 7.2 shows how harmful spatial isolation (or segregation according to Dimension 2) is for the poor in Santiago. As spatial isolation increases, so does unemployment for both the heads of households and young. The geography of opportunities is clearly affected by residential segregation. On the one hand, there is a strong correlation between the households’ spatial isolation and the proportion of young people who neither work nor study. This is a worrisome finding because, as the police know and the studies show, youth inactivity is usually associated with crime and particularly with drug dealing in poor neighborhoods. This could signal a rise in the ghetto phenomenon in Santiago. On the other hand, according to these results, segregation is not statistically associated with teenage fertility, which is an encouraging outcome. The ghettoization of poor neighborhoods is a relatively recent process in Santiago, as well as in other Chilean and Latin American cities, whereas teenage fertility could be considered an advanced stage of what Massey and Denton call the “culture of segregation” (1993). Social effects of segregation are not the same at different spatial scales (Sabatini and Sierralta, 2006), as table 7.2 shows. To address this question, we use the Geographically weighted regression technique (GWR) to conduct a linear regression analysis of segregation and spatial diversity. GWR allows us to estimate not only the effects of segregation on social problems, but also how those effects vary across spatial areas of the city at different spatial scales. We focus on only one dependent variable, unemployed household heads from the lowest socioeconomic status (SES) groups, so as to determine the effect of segregation on this important variable. Using this technique, for each census tract, a regression was computed using the values of both variables (percentage of D and E households in the tract and percentage of unemployed heads of D and E households) for all the census tracts within a spatially weighted radius. Unemployment and segregation variables are measured as percentages for each census tract. Map 7.1 displays the results of the geographically weighted regressions at two spatial scales, the first representing a defined radius of 1000 meters, and the second a radius of 2000 meters. The effects are as expected for the municipalities in the periphery of Santiago, with the exception of municipalities of the high-income cone and some districts that have received
138
SABATINI, WORMALD, SIERRALTA, AND PETERS
500m Neighbourhood
250m Neighbourhood
Entropy (SES) 0.00 – 0.60 0.61 – 0.70 0.71 – 0.80 0.81 – 0.90 0.91 – 1.00 Metro Lines 2000m Sample
1000m Sample
GWR estimates < –1.50 –1.49 – –0.50 –0.49 – 0.50 0.51 - 1.50 1.51 + Metro Lines 0
5
10
Kilometers 20
Map 7.1 Geographically weighted regressions at two spatial scales, Santiago, 2002
new mid-and high-income households. The greater the degree of segregation, the higher the unemployment. The more ubiquitous and uniform poverty in an area, the higher unemployment is in that area (represented in map as light gray–colored areas). On the contrary, and this is a surprising finding, the relationship is the opposite in the consolidated, central zones of the metropolitan region as well as in the high-income cone and other areas of the gentrified periphery. The higher the spatial segregation of poor households, the lower the percentage of heads of poor households that are not working (represented by gray-colored areas in the lower portion of map). Comparing the direct (expected) effect with the indirect (unexpected) effect, the former is stronger. The negative effects of the development of homogenously poor residential areas on unemployment are more pronounced than the positive effects that occur in large areas of
RESIDENTIAL SEGREGATION IN SANTIAGO
139
the city in which the poor are a minority, and where (partially because of being a minority) the geography of opportunities is greater. In fact, the estimated values for the GWR (map 7.1) indicate that the direct relationship is stronger (up to 4.69), while the inverse relationship is weaker (down to 22.76). The first result is consistent with the argument we have been making thus far: segregation according to Dimension 2 contributes to the isolation of households and inhibits their interaction with people from other socioeconomic groups, and, in general, from activities other than those related to popular housing. The second result moves us to make an interesting territorial specification for the influence of segregation on unemployment among the lower socioeconomic groups. In cases where small “pockets of poverty” can be found within larger areas of higher socioeconomic status (central areas in Santiago, consolidated areas around accessibility axes and the high-income cone), spatial segregation of the poor (and its concentration) helps improve employment opportunities. We suggest that, for those poor neighborhoods segregated at a low scale, physical proximity among poor households facilitates the circulation of information about labor opportunities.
Conclusions The most important conclusion of this study is the recognition of the need to question some of the most common interpretations about residential segregation in Latin American cities. In particular, we need question the existence of a simple and direct relationship between social inequalities and spatial segregation, and, more generally, the underestimation of residential segregation as a phenomenon by itself, which is rather reduced to a spatial indicator of inequalities. Residential segregation to some extent is natural given that inequalities are, according to Tilly (1999), persistent in society. It follows that trying to control or to modify residential segregation would be both ingenuous and insignificant. On the contrary, we think that segregation is a continual process and that it has a complex relationship to social inequalities and social differences. On the one hand, the phenomenon is less determined and more open than what is usually assumed; and, on the other hand, it is a determining phenomenon for the life chances of people and households, especially among the poor (Dimension 2), but also among higher income groups (Dimension 1). We have confirmed that residential segregation in Santiago has decreased at the same time that social inequality has remained high, and
140
SABATINI, WORMALD, SIERRALTA, AND PETERS
new forms of social exclusion of the poor have been strengthened. This is our first paradoxical result. Segregation has become more pernicious in that it has declined in magnitude, but its negative effects have increased. This is our second paradoxical result. We have shown, additionally, the importance of scale in the analysis of the social effects of residential segregation. Segregation could help increasing or reducing unemployment of heads of low-income households depending on the characteristics of the broader urban context.
Notes A preliminary version of this study was presented at the Spatial Differentiation in the Americas conference, organized by the University of Texas at Austin (November 2005). The study is part of collaborative research between the Chilean National Institute of Statistics (INE) and the Universidad Católica de Chile, started in 2005. This research seeks to study the evolution of residential segregation at the household level by socioeconomic status, as well as its social effects. This study builds upon former work, especially the FONDECY project #1020877, Valparaiso, Santiago, and Concepcion: Creation of HighIncome Neighborhoods (2002–2004) and Spatial Segregation and Uses of Urban Land in Latin America, funded by the Lincoln Institute of Land Policy (1999–2000). 1. For the country as a whole, the proportion of poor individuals went from 38.6 percent to 18.8 percent, while the proportion of poor households went from 33.3 percent to 15.4 percent. For the metropolitan region, the decline was even steeper. At the individual level, poverty went form 33 percent to 13.5 percent, and at the household level, it went form 28.5 percent to 10.8 percent. 2. The panel survey was run by MIDEPLAN, using CASEN 1996 as the baseline. It is representative of the III, VII, VIII and metropolitan regions. It identifies a set of households interviewed in 1996, and follows them in 2001. The four regions included cover about 60 percent of the Chilean population. 3. Literally, Barrio Alto means “high neighborhood.” The term is used in Santiago to designate an affluent area of the city that is located close to the mountains. Both income and altitude are high in this area. 4. See Rubalcava and Schteingart (2000), as an example for Mexico City, and Preteceille and Ribeiro (1999) and Riberiro (2000) for Rio de Janeiro, whose work support this diversity. Villaça (1998) emphasizes this as one of the characteristics of areas where high-income groups concentrate in Brazil. 5. According to census data for 2002, the mean population sizes in these different geographic units were for Santiago: municipality, 152,115; census district, 13,866; census tract, 3897; and block, 125.
RESIDENTIAL SEGREGATION IN SANTIAGO
141
6. A recent article by David Wong (2005) is appropriately entitled “Formulating a General Spatial Segregation Measure.” For a more detailed discussion of atomism in the measurement of segregation, see Sabatini and Sierralta (2006). 7. Two other dimensions of segregation that are included in the multidimensional index developed by Massey and Denton, concentration and centralization, are also important to consider, both for methodological and theoretical reasons, as Sabatini and Sierralta explain (2006). 8. Sabatini and Caceres (2004) discuss four components of the transformation of segregation patterns. 9. The out-migration of new families formed by “the children of the Barrio Alto” explains the dispersion of the elite, and, secondarily, the demographic expansion of group ABC1 to the extent of including former C2 households. In fact, the number of ABC1 households increased by 17.1 percent between 1992 and 2002, which is just above the growth of the total of households in Santiago (14.6 percent). 10. The value of the isolation index clearly depends on the demographic weight of each group, a topic we will discuss bellow. 11. Sennett (1970) discusses the influence of these collective forms of adolescence in the preference of the inhabitants of American cities for living in socially homogeneous suburban neighborhoods.
References Abramo, P. (1998). La ville kaléidoscopique; Coordination spatiale et convention urbaine. Paris: Editions L’Harmattan. Anselin, L. (1995). Local indicators of spatial autocorrelation—LISA. Geographical Analysis, 27, 93–115. CASEN (1990). Encuesta de caracterización socioeconómica de hogares. Santiago, Chile: MIDEPLAN. ——— (2003). Encuesta de caracterización socioeconómica de hogares. Santiago, Chile: MIDEPLAN. Contreras, D., Cooper, R., Herman, J., and Neilson, C. (2004). Dinámica de la pobreza y movilidad social: Chile 1996–2001. Departamento de Economía, Universidad de Chile. August. Santiago, Chile. Fishman, R. (1987). Bourgeois utopias: The rise and fall of suburbia. New York: Basic Books. Galster, G. and Killen, S. P. (1995). The geography of opportunity: A reconnaissance and conceptual framework. Housing Policy Debate, 6(1), 7–43. Holland, J. (1998). Emergence: From chaos to order. New York: Basic Books. Lamarca, F. (2005). El Mercurio (Santiago) October 28. Marcuse, P. (2001, July). Enclaves yes, ghettos no: Segregation and the state. Talk given at the international seminar Segregation in the City, organized by the Lincoln Institute of Land Policy. Cambridge, MA.
142
SABATINI, WORMALD, SIERRALTA, AND PETERS
Marques, E. and Torres, H. (2005). São Paulo: Segregacao, pobreza e desigualdades sociais. São Paulo, Brazil: Editora Senac. Massey, D. (1994). Space, place and gender. Cambridge: Polity Press. Massey, D. and Denton, N. (1988). The dimensions of residential segregation. Social Forces, 67(2), 329–57. Massey, D. and Denton, N. (1993). American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard University Press. MIDEPLAN. (2004). Principales resultados de empleo de la encuesta CASEN. Santiago: MIDEPLAN. Preteceille, E. and Ribeiro, L. C. Q. (1999). Tendências da segregação social em metrópoles globais e desiguais: Paris e Rio de Janeiro nos anos 80. Revista EURE, 26(76), 79–102. Ribeiro, L. C. Q. (2000). Cidade desigual ou cidade partida? Tendencias da metrópole do Río de Janeiro. In L. C. Q. Ribeiro (ed.), Futuro das metrópoles: Desigualdades e gobernabilidade. Río de Janeiro, Brazil: UFRJ-FASE. Rubalcava, R. M. and Schteingart, M. (2000). Segregación urbana en el Área Metropolitana de la Ciudad de México. In Gustavo Garza (ed.), La Ciudad de México en el Fin del Segundo Milenio, pp. 287–96. México: El Colegio de México-Gobierno del Distrito Federal. Sabatini, F. and Cáceres, G. (2004). Los barrios cerrados y la ruptura del patrón tradicional de segregación en las ciudades latinoamericanas: El caso de Santiago de Chile. In Gonzalo Cáceres and Francisco Sabatini (eds.), Los barrios cerrados en Santiago de Chile: Entre la exclusión y la integración social, pp. 9–44. Santiago, Chile: Instituto de Geografía, PUC Chile. Sabatini, F., Cáceres, G., and Cerda, J. (2001). Segregación residencial en las principales ciudades chilenas: Tendencias de las tres últimas décadas y posibles cursos de acción. Revista EURE, 27(82). Sabatini, F. and Sierralta, C. (2006). Medição da segregação residencial: Meandros teóricos e metodológicos e especificidade latino-americana. In José Marcos Pinto da Cunha (ed.), Novas metrópoles paulistas: População, vulnerabilidade e segregação. Campinas, Brazil: Nepo-Unicamp. Sabatini, F. and Wormald, G. (2004). La guerra de la basura de Santiago: Desde el derecho a la vivienda al derecho a la ciudad. Revista EURE, 30(91), 67–86. Sennett, R. (1970). Families against the city: Middle class homes of industrial Chicago, 1872–1890. Cambridge, MA: Harvard University Press. ——— (1971). The uses of disorder: Personal identity and city life. New York: W. W. Norton. Tilly, C. (1999). La desigualdad persistente. Buenos Aires, Argentina: Manantial. Villaça, F. (1998). Espaço intra-urbano no Brasil. São Paulo, Brazil: Studio Nobel. Wagensberg, J. (2004). La rebelión de las formas. Barcelona: Tusquets. Wong, D. (2005). Formulating a general spatial segregation measure. The Professional Geographer, 57(2), 285–94. World Bank (2004). Inequality in Latin America. Washington, DC: World Bank.
RESIDENTIAL SEGREGATION IN SANTIAGO
143
Wormald, G., Cereceda, L. E., and Ugalde, P. (2002). Estructura de oportunidades y vulnerabilidad social: Los grupos pobres en la Región Metropolitana de Santiago de Chile en los años 90. In G. Wormald and R. Kaztman (eds.), Trabajo y ciudadanía. Montevideo, Uruguay: CEBRA Editores. Wormald, G. and Salinas, V. (2003). Informalidad en Chile durante la década de los años 90s. Working paper for the Latin American Urbanization in the Late Twentieth Century seminar. February, Buenos Aires, Argentina.
This page intentionally left blank
8
Residential Segregation in São Paulo: Consequences for Urban Policies Haroldo da Gama Torres and Renata Mirandola Bichir
Introduction Brazilian cities are highly segregated in spatial terms. Besides the prevalence of shantytowns, these cities generally present a radial-concentric urban structure, with the rich population concentrated in the center of large cities while the poor are located in peripheral, more distant areas (Santos and Bronstein, 1978; Villaça, 2000; Caldeira, 2000). In spite of its intensity, residential segregation is not a major theme in the Brazilian social debate. Whereas academic discussions on social issues focus on poverty and income inequalities, segregation is rarely used as a basis for public policy–making. This situation may be partially explained by conceptual difficulties. Brazilian urban sociology originally mistook “segregation” for the concepts of poverty and inequality, as well as the lack of access to basic public services. Since the 1970s, the debate over the living conditions of the population in peripheral areas and shantytowns—usually composed of low-income migrants who moved to the metropolis in search of job offers and who had limited access to those benefits made available by the state (Camargo, 2005; Kowarick, 1979). In this context, the concept of segregation mixed notions of poverty, inequality, and lack of access to the urban infrastructure and the services generally provided by the state. According to Sabatini (2001), this same confusion often occurs in other Latin American countries.
146
DA GAMA TORRES AND MIRANDOLA BICHIR
The first tradition of Brazilian urban studies was developed during the 1960s and 1970s in essays written as part of the debate on the theory of marginality.1 This mainly Marxist literature viewed urban dynamics as a consequence of larger economic processes located in space and especially highlighted the so-called production of peripheral areas as a process required by the reproduction of the working class at low cost (Bonduki and Rolnik, 1982; Kowarick, 1979; Maricato, 1982). Such peripheral areas would be far from the urban centers and characterized by lack of infrastructure, precarious housing solutions (that is, shantytowns, illegal settlements, and self-built homes), and illegal land ownership. With state complacency, real estate brokers and the like would speculate to develop those illegal settlements (Bonduki and Rolnik, 1982; Maricato, 1982; Santos, 1979; Ribeiro and Lago, 1991; Vetter and Massena, 1981). Their precarious nature would indicate a set of so-called extortions—a situation of urban plunder, according to Kowarick (1979). More recently, discussions on socio-spatial segregation have been structured around a different perspective, although conceptual difficulties remain (Taschner and Bogus, 2000; Villaça, 2001). Following major international studies, issues such as “urban restructuring” and globalization emerged within the debate to shed new light on the broader tradition of using the urban arena as a locus for treating “macro issues” and leaving the urban studies per se at a secondary level. In this sense, segregation loosely refers to the residential separation between different income groups; however, unlike other debates—especially in the United States—it does not consider racial segregation (Jargowski, 1997; Massey and Denton, 1993; Mingione, 1999; Wilson, 1990). 2 Residential segregation is rarely measured in empirical terms, with few studies discussing the problem from a public-policy perspective. As a result, the consequences of segregation for families and for public policies are seldom considered (Torres, 2004). We believe that an adequate empirical approach for understanding segregation should be based on a more detailed analysis of the urban socioeconomic landscape and encompass at least four relevant dimensions. First, it is important to define the concept of segregation in such a way that allows for appropriate measurement. Second, criteria must be selected to identify the most segregated areas in a given metropolis. Third, it is important to understand the processes that lead to segregation—for example, land and real estate dynamics and state policies, among others. 3 Finally, such an approach requires that we consider the socioeconomic and public-policy consequences segregation has for the residents of those areas. In this paper, we try to address some of these issues by assessing
RESIDENTIAL SEGREGATION IN SÃO PAULO
147
the impact of segregation on the access to urban infrastructure services in the city of São Paulo.
Segregation: Concepts and Metrics We are interested here in two major dimensions of segregation: the spatial concentration patterns of given social groups in specific areas and the degree of social homogeneity within such areas (Sabatini, 2001; Torres, 2004).4 A key issue related to the definition of “segregation” refers to the degree of homogeneity required to identify an area as being segregated5—an issue also associated with the scale of segregation, being any measurement in this field influenced by the size and number of units of analysis (census tracts, districts, etc.). Thus, segregation should be understood as the gathering of a particular group in a given area6—either in terms of race/ethnicity, income, or other cleavages relevant to that specific social context. In this more general sense, the population concentration in both shantytowns and gated, high-income condominiums, for instance, may be regarded as different aspects of the segregation process. Usually referred to as “residential segregation,” this concept is strongly present in North American urban studies and allows for the development of synthetic segregation indicators that may be used to compare data across regions and over time (Massey and Denton, 1993). The dissimilarity index is likely to be the indicator most employed in this field despite its limitations (Sabatini, 2004). As discussed in chapter 2, it measures the share of the population in a given social group that would have to move in order to create a distribution of social groups within an area similar to the distribution existing in the city as a whole. Other metrics—such as “isolation indexes”—are also used to get to such results (Massey and Denton, 1993; Sabatini, 2001). Besides the dissimilarity index, we also use the Moran’s index, which estimates the degree of homogeneity or heterogeneity between neighboring urban areas (Bichir, Torres, and Ferreira, 2005). In this paper, we work mainly with the geographic unit of analysis of “survey areas” employed by the Brazilian Demographic Census of 2000,6 corresponding to an intermediate area between census tracts and districts. However, we will be attentive to other geographic scales to consider the problems regarding micro- and macro-segregation.7 Survey areas refer to spatial divisions based on groups of census tracts employed to run the longer, more detailed survey for the Brazilian Demographic Census. Data from the 1991 census were also built following the same approach and
148
DA GAMA TORRES AND MIRANDOLA BICHIR
unit of analysis.8 The results obtained through the dissimilarity index and the Global Moran index are presented subsequently.
Poverty and Segregation in São Paulo When discussing the evolution of socioeconomic indicators for the heads of household in the urbanized region of São Paulo between 1991 and 2000, one can notice a significant improvement in their social conditions.9 For instance, the share of heads of household with incomplete primary education (less than eight year of schooling) decreased, going from 61.3 percent in 1991 to 52.1 percent in 2000, according to census data. Likewise, census data indicate that the share of heads of household with less than three years of schooling decreased from 24.5 percent to 19.3 percent during the 1990s. The share of heads of household with 11+ years of schooling (secondary school) went up from 24.7 percent to 31 percent, even though the share of those with university degrees remained virtually stable—from 10.6 percent in 1991 to 11.3 percent in 2000. Therefore, during that decade, the low-schooling population seems to have decreased in relative terms in the urban region of São Paulo. Such data is consistent with more general educational and income indicators available from other sources such as the Monthly Employment Survey by the Brazilian Statistics Bureau. In the case of São Paulo, a reduced share of low-educated people also implies a decrease in segregation levels—even though this is not always true for other regions and/or situations. Massey and Denton (1993) have pointed out the paradox of having increased social segregation in parallel with substantial improvements in social conditions: between 1910 and 1940, when residential segregation rose sharply in the United States, despite much better living conditions and higher income levels. In São Paulo, it is possible to observe that segregation remains stable when considering education dissimilarity indexes10 between heads of household with 0–3 and 15+ years of schooling (scale of survey areas), while decreasing significantly when we compare the less educated (0–3 years) with those with secondary education (11+ years). It also decreased for all other comparisons considered (table 8.1). Table 8.1 shows that the general level of segregation in terms of the educational background is high in São Paulo, but it has been decreasing lately. We also considered an alternative measure of segregation, the isolation index, which indicates the extent to which low-educated families are isolated from more educated ones in their residential locations. This alternative measure indicates that isolation has also been decreasing for the group with less than 8 years of schooling.
RESIDENTIAL SEGREGATION IN SÃO PAULO
149
Table 8.1 Residential segregation index (dissimilarity and isolation) by education. Scale of survey areas of São Paulo’s urbanized region, 1991 and 2000 Segregation indicators Dissimilarity index Dissimilarity between heads of household with 0 to 3 and 151 years of schooling Dissimilarity between heads of household with 0 to 3 and 111 years of schooling Dissimilarity between heads of household with 0 to 7 and 81 years of schooling Dissimilarity between heads of household with 0 to 7 and 111 years of schooling Dissimilarity between heads of household with 0 to 7 and 151 years of schooling Isolation index Isolation of the group with 0 to 7 years of schooling
1991
2000
0.619
0.616
0.487
0.423
0.276
0.276
0.514
0.381
0.641
0.584
0.681
0.577
Source: IBGE, Demographic Censuses of 1991 and 2000.
However, it is interesting to notice that dissimilarity is not decreasing in terms of income—having otherwise increased from 0.63 percent to 0.67 percent when comparing heads of household with 0–3 and 201 minimum wages (Torres, 2004). In other words, the decline in educational segregation does not reflect a broader dynamic of reduced segregation, but rather an expansion in secondary education coverage that is now reaching poor areas. Contrary to the situation of Santiago (Sabatini, 2001), where macrosegregation fell while micro-segregation increased, segregation in terms of income has increased in São Paulo for the scales of both census tracts and survey areas (Torres, 2004). Such indicators suggest a growing social isolation between different income groups in the comparison of large areas and even locally within the survey area. Results also indicate the extension and depth of this phenomenon: poorer families live spatially isolated from others groups—especially in urban peripheries—and their degree of isolation has increased in the past 10 years. The data for São Paulo also shows a low level of segregation between blacks and whites in the scale of survey areas for 2000: a dissimilarity index of 23.7 percent, which is well below that of American cities. Generally speaking, such a low level of segregation in terms of race occurs because poor areas are ethnically mixed, whereas rich areas are mainly white. In summary, these results indicate there is an important residential segregation in São Paulo, determined above all by a socioeconomic component. This conclusion is also consistent with the existing literature on
150
DA GAMA TORRES AND MIRANDOLA BICHIR
the subject. The key new factor in this respect refers to the decline in educational segregation during the 1990s, while the income-based segregation was on the rise. Other Segregation Measures: The Moran Index The Global Moran index (I) can also be employed to measure segregation through spatial self-correlation (see discussion of this in chapter 2). Such an indicator is more often used as a summary of the spatial distribution of a variable, acting as an alternative metric for segregation. When compared to those indicators more commonly used in segregation studies—the dissimilarity index, for example—Moran’s I incorporates a rather innovative dimension: testing the degree to which the level of a variable for a given area is similar to that of neighboring areas. In areas without segregation, for instance, poor people (or any other variable of interest, such as the concentration of blacks) will be evenly distributed in a given area and its neighboring ones, for all areas of the city, with a Global Moran near 0. In segregated areas, those with a high concentration of poor people will be close to others of the same kind (likewise, areas with a high concentration of non-poor individuals will have neighboring non-poor areas) with a Global Moran of near 1. Using the Moran index applied to schooling data for the city of São Paulo11 following the scale of survey areas, the results indicate high levels of residential segregation for 2000, as in the case of the dissimilarity index. The Global Moran index shows high spatial correlation among survey areas—0.74 percent for the variable of “average years of schooling” of the head of household. Overall, areas of low schooling are located close to other similar areas, thus suggesting a high degree of social homogeneity in schooling levels within the region. However, the results capture mainly a macro-segregation dimension not capable of showing what is happening in shantytowns and smaller gated communities. These elements are next presented in detail. Strategies to Identify Segregated Areas in the City of São Paulo Besides defining strategies to measure segregation, such as those suggested by the dissimilarity and Global Moran indexes, it is also important to identify those places in which segregation occurs more strongly. To that effect, we have employed the so-called Local Moran index, which identifies spatial clusters according to a given variable of interest. The Local Moran provides an excellent strategy for identifying macrosegregated areas. As mentioned before, micro-segregated areas such as
RESIDENTIAL SEGREGATION IN SÃO PAULO
151
shantytowns cannot be precisely identified by such a technique when they are located in a non-poor region.12 Map 8.1 shows the results obtained. Here, the Local Moran results indicate the existence of segregated areas in terms of schooling, especially in peripheral parts of the city: the extreme east and south and some specific areas of the north—a distribution well-known and coherent to the literature on the subject for the city of São Paulo. Such areas are identified in gray on map 8.1. The black areas, along the southeast vector of the city, indicate spatial concentration of high levels of schooling—these are also the richest areas of São Paulo. White identifies transition areas, with intermediary levels of schooling.
Legend ~
Sao Paulo Illegal settlements Local Moran High-high Intermediary Low-low
Map 8.1 Local Moran for average years of schooling of the head of the household and presence of shantytowns and illegal settlements. City of São Paulo, 2000
152
DA GAMA TORRES AND MIRANDOLA BICHIR
In general, it is possible to conclude that the gray areas, located in the peripheries of the city, are segregated in terms of schooling, concentrating a continuous share of the population with low-education levels. Map 8.1 also indicates that areas with low education levels (low-low) concentrate the highest share of shantytowns and illegal settlements. In the next section, we investigate to what extent living in such areas influences the access of the population to urban infrastructure policies. Segregation and Public Policies: Access to Urban Infrastructure Having identified the most segregated areas in the city of São Paulo (macrosegregation), we can now discuss the consequences of living in these areas in terms of poor people’s access to urban infrastructure. This section is based on results from a survey performed in November 2004 by the Center for Metropolitan Studies (CEM-CEBRAP) The survey (1,500 questionnaires) focused on the poorest 40 percent of the city of São Paulo and aimed at verifying several aspects of their living conditions, including access to public policies—i.e., education, health, urban infrastructure, and income transference programs (Figueiredo et al., 2005). We discuss here the key determinants of urban infrastructure policies chosen due to the strong association, usually present in the literature, between lack of urban services—that is, water, sewage, energy, and garbage collection—and segregated areas. As discussed in the introduction of this chapter, this association, often leading to the idea that lack of infrastructure (especially in the periphery) is the same as segregation, is quite frequent in the Brazilian urban studies literature, (Lago, 2002; Marques, 2003; Ribeiro, 2002; Vetter, Massena, and Rodrigues, 1979; Vetter et al., 1981; and Vetter and Massena, 1981, among others). However, other social dimensions must also be considered. Taking into account different variables from the survey’s questionnaire, we have developed an indicator of access to urban services: water supply—including frequency of supply so as to verify intermittent service, since it is virtually universal within the city13—sewage, garbage collection, public lights, road pavement, green and leisure areas, and public transport conditions. As we can see in table 8.2, the coverage of these services is generally high, even among the poorest people of the city of São Paulo. Green areas, public lights, and sewage system are the only exceptions, being less frequent in the low-low peripheral areas of São Paulo. Considering such a high coverage, we have built a more restrictive indicator by aggregating all these dimensions. This should allow us to verify potential differentiations of access in areas apparently well served.
RESIDENTIAL SEGREGATION IN SÃO PAULO
153
Table 8.2 Coverage of some public services among the poorest people and urban infrastructure access levels according to types of areas (%), São Paulo, 2004 Indicator
Water Daily water supply Energy supply Sewage system Garbage collection Road pavement Public lights Public transport Green and leisure areas Synthetic indicator Inadequate services (0 to 0.88) Insufficient services (more than 0.88 to 0.89) Adequate services (more than 0.89 to 1.0)
Type of areas (Local Moran)
Total Coverage
Low-Low
Intermediary
High-High
96.1 88.8 96.6 71.1 91.8 80.0 71.4 89.1 39.6
98.0 98.4 98.6 80.6 94.3 90.4 84.1 90.0 56.9
96.8 95.7 99.2 87.9 94.6 90.8 85.2 92.7 71.0
96.7 92.0 97.4 75.0 92.7 83.7 76.0 89.6 46.8
27.0
33.6
44.3
38.4
29.6
33.6
33.4
33.1
43.3
32.7
22.3
28.5
Source: CEM-Cebrap/Ibope. Survey of access to public services by the poorest population in São Paulo. November 2004.
It should be noted that in statistical testing using factor analyses, such variables are highly correlated, thus indicating that their provision is not an isolated occurrence, but rather part of a more general dynamic of urban services. These services contribute to the so-called production of the “built environment.” They have to be delivered at the door of the population, and sometimes even within their homes, as opposed to health and education equipment, which imply population deployment. These dimensions significantly impact the interpretations on conditions of access and the consequences of residential segregation. Two situations were considered for each of the nine variables comprising the indicator of access to urban services: adequate and inadequate, the former valuing 1 and the latter 0. The final indicator corresponds to the average of such variables, varying along a scale from zero (0) to one (1), where zero is the worst situation and one is the best. This indicator was then aggregated in three different groups to specify the access levels measured by the lowest-income population of the city. The results are shown in table 8.2 above, where we can see that a significant share of the poorest population of the city (38.4 percent) still have—according to such a restrictive indicator—an inadequate urban
154
DA GAMA TORRES AND MIRANDOLA BICHIR
infrastructure. Generally speaking, the less-served households have no sewage collection, are distant from green areas, and/or without public lights and daily water supply. However, almost all are supplied with water and energy and have garbage collection. When de-aggregating this information by type of area, it is possible to observe significant differentials: in the central area (high-high), 27.7 percent of poor households have inadequate access, whereas the intermediary and peripheral (low-low) areas get 36.6 percent and 44.3 percent, respectively. In other words, the access to urban infrastructure in more segregated areas of the city remains a significant problem. However, such results cannot prove per se the relation between residential segregation and inadequate access to urban services. There could be other intervening factors, such as the lack of social capital in suburban areas or the legal conditions of access to the land, just to mention two additional factors that may explain the difference in access to public services (Putnam, 2000; World Bank, 1999). Therefore, it is also critical to assess whether the access to urban services is a consequence of other variables included in different models— for example, family income, education, political preferences, and so on. This argument is further explored by using a chi-square automatic detector (CHAID), which identifies population segments with more or less access to urban services according to different independent or predicting variables.14 This model is useful for exploratory analyses and allows for the creation of groups according to the types of interactions observed. Additionally, CHAID provides a means to test the influence of different types of variables. A substantial range of variables was tested as drivers of access to urban infrastructure policies consistent with the different hypotheses on access to urban services—from the legal status of land (institutional dimension), residential segregation, and social participation to political behavior and individual characteristics. The CHAID model is represented in figure 8.1, summarizing the variables included and their order of importance. Those variables presented first are more likely to explain the situation verified. Individual cells present the levels of access to urban infrastructure measured by each specific group, as well as the size of the group vis-à-vis the total population and the value of chi-square for each association. The model’s significance level is 5 percent. This model first indicates that the relevance the legal status of land—that is, households located in shantytowns or illegal settlements—plays as a differential for the access to urban services. Whereas only 30.8 percent of households not located in such areas have inadequate access to urban infrastructure, 44.7 percent of those located in shantytowns and/or illegal settlements have inadequate access to these services. This result probably
Access Inadequate: Insufficient: Adequate:
38.4% 33.1% 28.5%
Shantytowns and illegal settlements No
Yes Access
Access Inadequate: Insufficient: Adequate:
Inadequate: Insufficient: Adequate:
30.8% 34.4% 34.9%
Local Moran (education level of head of household)
Time in the neighborhood (months)
Intermediary
High-high
44.7% 33.0% 23.3%
More than 277
Low-low
Access
Access
Access
Access
Inadequate: 23.7% Insufficient: 29.3% Adequate: 47.0%
Inadequate: 22.0% Insufficient: 39.9% Adequate: 38.1%
Inadequate: 44.0% Insufficient: 30.7% Adequate: 25.3%
130 to 276
Up to 129 Access
Access
Inadequate: Insufficient: Adequate:
25.2% 38.3% 36.5%
Inadequate: Insufficient: Adequate:
Time in the neighborhood (months) More than 277 Access Inadequate: Insufficient: Adequate:
48 to 276
Inadequate: Insufficient: Adequate:
Up to 0.5
Access 34.1% 35.6% 30.4%
Access
Inadequate: 14.4% Insufficient: 44.7% Adequate: 38.9%
Inadequate: Insufficient: Adequate:
Annual participation in non religious association
No
Yes
Access Inadequate: Insufficient: Adequate:
Figure 8.1
55.6% 29.5% 14.7%
Family income per person (minimum wage)
Up to 47
Access 12.8% 41.9% 45.4%
Inadequate: Insufficient: Adequate:
41.4% 31.6% 27.0%
Access 53.2% 23.0% 23.8%
Inadequate: Insufficient: Adequate:
CHAID model for the urban infrastructure indicator
36.7% 36.8% 26.5%
0.5 and more
Access 67.7% 21.7% 10.6%
Inadequate: Insufficient: Adequate:
44.2% 37.1% 18.5%
156
DA GAMA TORRES AND MIRANDOLA BICHIR
indicates the difficulties faced by the state to expand services to illegal areas that do not comply with urban regulations. Many of these areas are not included in the city’s official maps, nor are they present in the major information systems used by town hall officers, the real estate register upon which urban taxes are based and collected. Furthermore, urban service providers do not have legal backing to operate in such areas (Maricato, 1996; World Bank, 1999). The state avoids making investments for fear of the land being repossessed by its legal owners or for fear of being sued by the Public Prosecution Office. This leads to low-quality services or none at all. In view of that, it is important to consider such micro-segregated areas as different from other macro-segregated ones in terms of public policies: there seems to be several barriers—including institutional ones—to public investments in those places. It is also possible to clearly observe in the map 8.1 that areas with low education levels (low-low) are concentrated in the highest share of shantytowns and illegal settlements. In other words, the distribution of the urban services considered here is not necessarily explained by the distribution of more or less segregated areas; alternatively, it may be explained by the presence of shantytowns and illegal settlements. Second, the model indicates that among those households located in shantytowns and illegal settlements, the length of time they live in the neighborhood is also a key element to determine access levels. Those residing longer in the neighborhood present better access levels, which seems to indicate a continuous access to urban improvements. Such a result is consistent with the ethnographic literature (Almeida and D’Andrea, 2004). For those who have not been around for long, family income plays an important role in ensuring better access. For those households not located in shantytowns and illegal settlements, macro-segregation—defined by the Local Moran according to years of schooling as shown in the section Strategies to Identify Segregated Areas in the City of São Paulo—seems to be the element that best explains the conditions of access to urban infrastructure. This result points out the impact of segregation itself on the access to public policies, since access levels are far worse in more segregated regions. Some studies in the Brazilian literature support this. Marques (2000; 2003), for instance, highlights the relation between state bureaucracies and city officials in decisions about public investment—in certain cases restricting the money spent on poor areas to prioritize richer ones. From this point of view, state investment patterns may be explained by both politicians’ political priorities and the “hierarchical selectivity” of bureaucratic officials.15
RESIDENTIAL SEGREGATION IN SÃO PAULO
157
Thirdly, the model also shows that among those households not located in shantytowns and illegal settlements, but in areas with a high proportion of low-educated heads of household (low-low), the key variable to explain access differentials is annual participation in non-religious associations. Those who participate have better access levels, which seem to indicate a kind of “network effect”: participation in associations provides different contacts and information, influencing the access to public services (Gurza Lavalle and Castello, 2004). Finally, among those households not located in shantytowns and illegal settlements, but located in “intermediary” areas in terms of education, those who have been living longer in the area have better access levels. As mentioned, this result seems to reveal a history of urban consolidation in these places. Appendix 8.1 synthesizes the information from the CHAID diagram, highlighting the share of individuals from each group with inadequate access levels to urban infrastructure, and its representation vis-à-vis the city’s total population. In brief, the results discussed so far lead to a number of general conclusions: a. Households located in shantytowns and illegal settlements present the most significant variations in urban access levels. Such places characterize micro-segregation situations that were not observed through the Local Moran model, since it is structured in a different unit of analysis. Following a broader institutional literature, we believe that this result cannot be explained solely by residential segregation hypotheses and that they also reflect institutional dimensions—especially that of land use regulation and its impact on service provision (World Bank, 1999). b. The macro-segregation variable (provided by the Local Moran) is important for those households that are not located in shantytowns and illegal settlements. Such individuals have less access to urban infrastructure despite living in “legal” areas, that is, those included in the Town Hall’s information systems, among people who pay land taxes. Our major hypothesis here is that political decisions and internal bureaucracy from both the town hall and utilities processes have created a “discriminatory” dynamic over time for those low-income areas, as suggested by Marques (2003). c. The significance of the income variable in defining the access to urban infrastructure of households is surprisingly limited. Most likely, this result is due to the fact that the survey focused on the poorest 40 percent of the population, thus significantly reducing the variability of this variable. However, it must be noted that the
158
DA GAMA TORRES AND MIRANDOLA BICHIR
income levels observed are still heterogeneous and could theoretically differentiate the levels of access.16 d. The variable related to time of residence in the neighborhood also plays an important role, even though it is less stressed in the literature on the subject. Ethnographic descriptions show increasing improvements in migrant households (Almeida and D’Andrea, 2004). Over time, low-income urban areas such as the shantytowns considered in this study gradually improve their organization capabilities and access levels to government representatives, ensuring better services (Perlman, 2002). e. Finally, the secondary relevance of variables related to association and political participation comes as a surprise. The literature has long stressed the fundamental significance of the so-called “social capital” to guarantee better access to goods and public services (Gurza et al., 2004; Putnam, 2000). Similarly, pork barrel politics have also been used to explain differences in services access (Ames, 1995). However, such relevance is likely to occur with policies other than those analyzed in this study. Although results from the model employed here to determine urban services are in line with our expectations for given areas, they also show that overall the interpretation of low-income individuals’ access to public services is rather complex, involving variables derived from different analytical perspectives. Similar analytical exercises for other metropolitan regions are likely to indicate key elements that will allow us to develop a more general model for the city as a whole. Discussion The arguments presented in this paper were structured based on three key elements. The first refers to the definition of segregation and the development of indicators for residential segregation (dissimilarity and Global Moran indexes) so as to determine the levels of residential segregation found in the city of São Paulo. We have shown that segregation is high and has increased for income, while remaining stable for education. The second element regards the identification of those areas in which segregation is stronger, such as locations with clusters of individuals of high and low education, as defined by the Local Moran. The most segregated areas are located in the periphery of the city, consistent with the literature. Due to the unit of analysis adopted, such results did not capture the micro-segregation dimension.
Appendix 8.1 Summary of urban services access indicator Group
Resident in shantytowns or illegal settlements, recently in the neighborhood (up to 129 months), and with family income per capita of up to 0.5 min. wage Non resident in shantytowns or illegal settlements, resident in areas defined as low-low in Local Moran, and without participation in non-religious association. Resident in shantytowns or illegal settlements, recently in the neighborhood (up to 129 months), and with family income per capita of more than 0.5 min. wage Resident in shantytowns or illegal settlements and in the neighborhood between 130 and 276 months. Non resident in shantytowns or illegal settlements, resident in areas defined as low-low in Local Moran, and with participation in non-religious association. Non-resident in shantytowns or illegal settlements, resident in areas defined as intermediary in Local Moran, and in the neighborhood between 48 and 276 months. Resident in shantytowns or illegal settlements and in the neighborhood for more than 277 months. Non-resident in shantytowns or illegal settlements and resident in areas defined as high-high in Local Moran. Non-resident in shantytowns or illegal settlements, resident in areas defined as intermediary in Local Moran, and in the neighborhood for up to 47 months. Non-resident in shantytowns or illegal settlements, resident in areas defined as intermediary in Local Moran, and in the neighborhood more than 277 months. Total
Percent of inadequate access
Participation of each group in the analyzed population (percent)
67.7
11.9
53.2
7.6
44.2
12.6
41.4
20.3
36.7
9.6
34.1
7.8
25.2
10.2
23.7
8.3
16.4
3.6
12.8
8.1
38.4
100.0
Source: Center for Metropolitan Studies. Survey of public service access of the poorest population in São Paulo. November, 2004.
160
DA GAMA TORRES AND MIRANDOLA BICHIR
Third, we have organized the data from the survey of access to public services for the poorest population of São Paulo around the geographical divisions indicated by the Local Moran to assess the effect of residential segregation on the population access to public services. Such analysis allowed us to observe that the segregation dimension is related to the less access to public services presented by some groups, regardless of family income. Although this association is widely discussed in the literature, it is often addressed differently from the perspective presented here. Both segregated families and low-income ones can have some access to urban infrastructure, especially to those services of universal coverage in the city such as water and energy supply. In most cases, the type of place of residence and the number of years living in the neighborhood are the factors that most strongly determine the chances a given family has of accessing urban services such as street lights, sewage collection, and nearby green areas. On the one hand, the illegal status of the land (shantytowns and illegal settlements) seems to produce a significant impact on service access, although this microsegregation dimension appears to be related to the institutional conditions of service provision and not only to the segregation dimension. On the other hand, macro-segregation as defined by the Moran approach also implies a lower level of access in “legal” areas. In view of these arguments, it would make sense for public policies to consider a spatial or territorial perspective—in other words, to consider shantytowns, illegal settlements, and other segregated areas as special interest places for public policies. To this effect, a number of operational initiatives would be required: ●
●
●
Give municipal administrations access to information systems that systematically capture and register the existence of informal and illegal areas. Information systems are important both to substantiate policy decisions and to bring the issue to the public debate (Torres, 2005). Reduce institutional barriers to public investment in irregular areas, including changes in land regulation.17 Identify and fight bureaucratic practices and policies that discriminate segregated areas within the government. Bureaucrats should be particularly trained to ensure that low-educated migrant populations receive adequate treatment.
Although the above propositions may seem rather straightforward and almost simplistic, their implementation implies overcoming substantial problems. Peripheral areas are often ignored by decision makers. Also,
RESIDENTIAL SEGREGATION IN SÃO PAULO
161
the demand for urban investments involves powerful interests from different stakeholders: high-income families, the media, construction, and real estate companies compete for public resources to be invested in their priority areas (Marques, 2003). Handling such pressures and providing a territorially homogeneous urban policy is the major challenge currently faced by urban administrators in Brazil. Notes 1. In the 1960s and 1970s, the term marginality was a central concept in Latin America, specifically in reference to the issue of exclusion (or marginal integration) of the lower-income share of the population from the so-called dependent and peripheral capitalist system then being developed in Brazil (see Perlman, 1977). 2. Telles (2003) is one important exception. 3. These specific dimensions, which will not be addressed in this paper, are more extensively discussed in the Brazilian literature (see Santos, 1979; Vetter and Massena, 1981; Smolka, 1983, 1987; Ribeiro and Lago, 1999; Villaça, 2000; Marques, 2003). 4. Massey and Denton (1993) also mention other dimensions, such as concentration (measuring the density of poverty) and centralization (measuring its location vis-à-vis the downtown area). We consider such elements to be less relevant for the Brazilian debate. 5. For Villaça (2004), the “high concentration” of high-income population in certain areas is the key element that defines it. 6. The demographic census is based on two questionnaires: the shorter one covers all the population, whereas the longer one focuses on a sample corresponding to 10 percent of the population. 7. Macro segregation refers to a wider spatial scale, such as neighborhoods or districts. Micro segregation is related to a detailed geographic scale, such as blocks or census tracts as unit of analysis. Important distinction may be observed between the two phenomena in empirical terms: macro segregation may be reduced in a given urban area while the micro one increases, as it was recently the case in Santiago, Chile (Sabatini, 2001). 8. Data from the 1991 census have been reconstructed for the 2000 census survey areas through GIS overlays. 9. Due to data availability, we have taken into consideration the 21 cities that make up the urbanized region of São Paulo and account for more than 91.4 percent of the total population of the metropolitan area. In general, we have excluded small cities of a more rural nature (that is, Santa Isabel, Juquitiba, and Salesópolis). All large cities have otherwise been considered (São Paulo, Osasco, Guarulhos, Santo André, São Bernardo, and so on). 10. This indicator is rather limited and may be criticized for a number of reasons. First, it does not capture segregation within the regions used as units of analysis. Second, this indicator varies according to the size of the unit of
162
11.
12.
13.
14.
15.
16.
17.
DA GAMA TORRES AND MIRANDOLA BICHIR
analysis, causing the so-called grid problem (Sabatini, 2001) and increasing the complexity of inter-area comparisons. Third, this is a non-spatial indicator; it does not identify whether all segregated individuals are concentrated in specific places within the city, or rather spread across it, a problem known as the “checkerboard” dilemma (Sabatini, 2004). In spite of such difficulties, this index has been widely employed, indicating the high level of residential segregation between blacks and whites in all large American cities, for instance, with its highest peaks found in Chicago, Detroit, and Kansas City (Massey and Denton, 1993). Contrary to the discussion for the dissimilarity index, we will work here with data only for the city of São Paulo. This option was chosen due to the data available for the survey, used in the next section. In order to run the Local Moran through a LISA map, we have used a second order queen contiguity matrix. This kind of spatial contiguity was chosen for operational reasons: its distribution fits best to the data presented in the survey with low-income families, employed to analyze urban infrastructure policies. Geographical coding identified the survey’s 1500 cases according to the areas defined by the Local Moran with a second order contiguity matrix: 516 high-high cases (high schooling survey area, similar to its neighbors); 492 intermediary ones; and 445 low-low cases (low schooling survey area, similar to its neighbors). Data from the Pesquisa Nacional por Amostra de Domicílios (PNAD) indicate that water coverage for the Metropolitan Area of São Paulo accounted for 99.2 percent in 2003. The CHAID (Chi-squared Automatic Interaction Detector) technique, based on the chi-square association, allows for the hierarchical classification of individuals through a log-linear model. It is used to study the relation between a dependent variable and a series of predicting variables that interact among themselves. From a double-entry table between the predicting and the dependent variables, the model is able to test all possible partition categories for the predicting variable, looking for the one that presents the highest statistical chisquared value. From the resulting partition, it is possible to group data accordingly and conduct a new analysis within each subgroup, repeating the whole procedure for the dependent variable and the remaining predicting ones. It is understood as a set of ideas, beliefs and visions of the society, either explicitly or implicitly, shared by members of the policy professional community (Marques, 2003). Survey data indicate that in 39 percent of cases family income is between 0 and 0.5 minimum wages; in 35.9 percent of cases it varies from 0.5 to 1 minimum wage; in 19.5 percent the range is between 1 and 2 minimum wage; in 5.1 percent it goes from 2 to 5 minimum wage. “Only well functioning land markets can provide an adequate supply of housing, and maintaining these markets is another task that deserves attention from the public sector. Providing universal registration and establishing clear property rights to all urban land will require strengthening existing institutions. Ill-defined land rights render land useless and discourage the
RESIDENTIAL SEGREGATION IN SÃO PAULO
163
redevelopment of entire portions of a city. But simply providing security of tenure creates incentives to improve housing and infrastructure dramatically. To avoid adding to the backlog of problem housing and neighborhoods, new development must meet basic—but not excessive—compliance standards” (World Bank, 1999, p. 146).
References Assembléia Legislativa do Estado de São Paulo (ALESP) (2000). Cadernos do Fórum São Paulo Século XXI. São Paulo, Brazil: Diário Oficial. Almeida, R. and D’Andrea, T. (2004). Pobreza e redes sociais em uma favela paulistana. Novos Estudos CEBRAP, 68 (March), 94–106. Ames, B. (1995). Electoral strategy under open-list proportional representation. American Journal of Political Science, 39(2), 406–33. Bichir, R., Torres, H., and Ferreira, M.P. (2005). Jovens no município de São Paulo. Explorando os efeitos das relações de vizinhança. Estudos Urbanos e Regionais, 6(2), 53–69. Bonduki, N. and Rolnik, R. (1982). Periferia da Grande São Paulo: Reprodução do espaço como expediente de reprodução da força de trabalho. In E. Maricato (ed.), A produção capitalista da casa (e da cidade) do Brasil industrial, pp. 117–54. São Paulo, Brazil: Alfa-Ômega. Caldeira, T. (2000). City of walls: Crime, segregation, and citizenship in São Paulo. Los Angeles: University of California Press. Camargo, C. (2005). São Paulo, 1975—Crescimento e pobreza. São Paulo, Brazil: Ed. Loyola, 1976. Figueiredo, A., Torres, H. G., Arretche, M., and Bichir, R. M. (2005). Determinantes do acesso a serviços públicos. Brasília: IPEA, Projeto BRA/04/052, Rede de Pesquisa e Desenvolvimento de Políticas Públicas REDE-IPEA II. Mimeograph. Gurza Lavalle, A. E., and Castello, G. (2004). As benesses deste mundo associativismo religioso e inclusão socioeconômica. Novos Estudos CEBRAP, 68 (March), 73–93. Jargowski, P. (1997). Poverty and place: Ghettos, barrios and the American city. New York: Russell Sage Foundation. Kowarick, L. (1979). A espoliação urbana. Rio de Janeiro, Brazil: Paz e Terra. Lago, L. (2002). A lógica segregadora na metrópole brasileira: novas teses sobre antigos processos. Cadernos IPPUR, 15/16, número especial. Maricato, E. (1982). Autocostrução, a arquitetura possível. In E. Maricato (ed.), A produção capitalista da casa (e da cidade) no Brasil industrial, pp. 71–93. São Paulo, Brazil: Alfa-Ômega. ———. (1996). Metrópole na periferia do capitalismo: Ilegalidade, desigualdade e violência. São Paulo, Brazil: Hucitec. Marques, E. (2000). Estado e redes sociais: Permeabilidade e coesão nas políticas urbanas no Rio de Janeiro. Rio de Janeiro, Brazil: Revan/Fapesp. ———. (2003). Redes sociais, instituições e atores políticos no governo da cidade de São Paulo. São Paulo, Brazil: Anna Blume/Fapesp.
164
DA GAMA TORRES AND MIRANDOLA BICHIR
Massey, D. S. and Denton, N. A. (1993). American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard University Press. Mingione, E. (1999). Urban poverty in the advanced industrial world: Concepts, analysis and debates. In Mingione, Enzo (ed.), Urban poverty and the underclass, pp. 30–40. New York: Blackwell. Perlman, J. (1977). O mito da marginalidade—Favelas e política no Rio de Janeiro. Rio de Janeiro, Brazil: Paz e Terra. ———. (2002). The metamorphoses of marginality: Favelas of Rio de Janeiro: 1969–2002. World Bank Conference. Trinity College. Putnam, R.D. (2000). Bowling alone. New York: Touchstone. Ribeiro, L. C. Q. and Lago, L. (1994). Reestruturação nas grandes cidades brasileiras—O modelo centro/periferia em questão. Rio de Janeiro. Presented at XV Encontro Anual da ANPOCS, Caxambu. Ribeiro, R. (2002). “Segregação, acumulação urbana e poder: Classes desigualdades na metrópole do Rio de Janeiro. In Cadernos IPPUR, year 15/16, special number: Planejamento e Território: Ensaios sobre a desigualdade. Sabatini, F. (2001). Transformação urbana e dialética entre integração e exclusão social: Reflexões sobre as cidades latino-americanas e o caso de Santiago do Chile. In M. Oliveira (ed.), Demografia da exclusão social, pp. 165–90. Campinas, Brazil: Unicamp. ———. (2004). Medición de la segregación residencial: Reflexiones metodológicas desde la ciudad latinoamericana. In G. Cáceres and F. Sabatini (eds.), Barrios cerrados en Santiago de Chile: Entre la exclusión y la integración residencial, pp. 277–307. Santiago, Chile: Lincoln Institute of Land Policy and PUC-Chile. Santos, C. (1979). Velhas novidades nos modos de urbanização brasileiros. In L. do Prado Valladares (ed.), Habitação em questão, pp. 17–47. Rio de Janeiro, Brazil: Zahar. Santos, C. and Bronstein, O. (1978). Meta-urbanização: O caso do Rio de Janeiro. Revista de Administração Municipal, 25(149), 6–34. Smolka, M. (1983). Segregação social no espaço: Definição do objeto de análise. In M. Smolka (ed.), Estruturas intra-urbanas e segregação social no espaço: Elementos para uma discussão da cidade na teoria econômica, pp. 38–75. Rio de Janeiro, Brazil: IPEA/Anpec. ———. (1987). Para uma reflexão sobre o processo de estruturação interna das cidades brasileiras: O caso do Rio de Janeiro. Espaço e Debates, 21, 17–28. Torres, H. G. (2004). Segregação residencial e políticas públicas: São Paulo na década de 1990. Revista Brasileira de Ciências Sociais, 19(54), 41–56. ———. (2005). Informação demográfica e políticas públicas na escala regional e local. Conference paper presented at Reunión de expertos sobre población y desarrollo local. Santiago, Chile: CELADE. October. Taschner, S. and Bógus, L. (2000). A cidade dos anéis: São Paulo. In L.C. Queiroz (ed.), O futuro das metrópoles: Desigualdades e governabilidade, pp. 247–84. Rio de Janeiro, Brazil: Revan/IPPUR/Fase. Vetter, D. and Massena, R. (1981). Quem se apropria dos benefícios líquidos dos investimentos do estado em infra-estrutura? Uma teoria de causação circular.
RESIDENTIAL SEGREGATION IN SÃO PAULO
165
In L. Machado da Silva (ed.), Solo urbano: tópicos sobre o uso da terra, pp. 49–78. Rio de Janeiro, Brazil: Zahar. Vetter, D., Massena, R., and Rodrigues, E. (1979). Espaço, valor da terra e equidade dos investimentos em infra-estrutura no Município do Rio de Janeiro. Revista Brasileira de Geografia, 41(1/2), 32–71. Vetter, D., Pinto, D., Friedrich, O., and Massena, R. (1981). A apropriação dos benefícios das ações do estado em áreas urbanas: seus determinantes e análise através da ecologia fatorial. Espaço e Debates, 1(4), 5–37. Villaça, F. (2000). Espaço intra-urbano no Brasil. São Paulo, Brazil: Nobel. ———. (2004). A pesquisa sobre segregação: conceitos, métodos e medições. Espaço & Debates, 24(45), (January/July) 87–109. Wilson, W. (1990). The truly disadvantaged: The inner city, the underclass and public policy. Chicago: University of Chicago Press. World Bank. (1999). Entering the 21st century: World development report 1999/2000. New York: Oxford University Press.
This page intentionally left blank
Part II
Rapidly Growing, Mid-Size Cities
This page intentionally left blank
9
The Process of Cumulative Disadvantage: Concentration of Poverty and the Quality of Public Education in the Metropolitan Region of Campinas José Marcos Pinto da Cunha and Maren Andrea Jiménez
Introduction The rapid process of urbanization in Latin America over the past 50 years has resulted in large mega-cities characterized by high-income inequality, poor housing conditions, and reduced access to public services, particularly among the urban poor. Indeed, rising levels of urban poverty has proved to be a general characteristic of Brazilian urbanization. However, in recent years decreasing urban primacy in Latin American countries has led to high growth rates within smaller urban agglomerations. Whereas some research has documented the extent of social segregation in the mega-cities of Latin America, less attention has been given to that of its secondary cities. Practically no research has been done on the relationship between segregation and access to public services in secondary cities. In recent years the Metropolitan Region of Campinas (MRC), Brazil, located about 100 km west of the city of São Paulo, has had one of the highest annual growth rates of the state of São Paulo. In 2000, 6.32 percent of the state’s population lived within the region. In fact, during the 1970s the São Paulo state government began moving economic production away from the state capital of São Paulo towards the interior of the state, thus
170
PINTO DA CUNHA AND JIMÉNEZ
fueling rapid economic and population growth in Campinas. As a result, Campinas can be classified as an “emergent metropolis,” in that the majority of its growth has occurred within the last 30 years. As such, the processes and consequences of urban expansion within the region have been largely unexplored (NEPO/NESUR, 2004). The region’s pattern of urban expansion resulted in a complex territory that reflects the contradictory nature of economic growth. This growth led to a process of urbanization that favored the increase of gated communities for middle- and high-income households. It simultaneously led to the concentration of the poor in periphery neighborhoods characterized by precarious urban infrastructure as well as the creation of favelas throughout the region (NEPO/NESUR, 2004). Accordingly, as is characteristic of many Latin American cities, the spatial distribution of Campinas is one where the affluent concentrate in the socially heterogeneous center of the region, while the poor tend to be homogenously concentrated in a large area occupying the southwest of the region (Cunha, et al., 2006). In Campinas, rapid urbanization and population growth has often outpaced the capacity of the local government to provide basic infrastructure and public services to its residents. With this in mind, our objectives are twofold. We first document the evolution of segregation in the metropolitan region from 1991 to 2000, the most recent period for which census data is available. We use two measures of segregation, the index of dissimilarity and Moran’s I, to evaluate changes in the patterns of residential segregation over the decade. Second, we examine the relationship between the structural quality and average test scores of public schools in the municipality of Campinas with the socioeconomic characteristics of the areas in which they are located, in order to explore variations in the relationship between neighborhood characteristics and the quality of public education within the municipality.1
Overview of the Metropolitan Region of Campinas Demographic and Economic Characteristics The Metropolitan Region of Campinas, consisting of 19 municipalities and containing almost 2.2 million residents, is without a doubt one of the most important regions in Brazil, not only because of its economic production but also because of its prominence in Brazilian technological production (polo tecnológico). As such, at the same time that the MRC expanded and assumed national importance it also accumulated—and
RESIDENTIAL SEGREGATION IN CAMPINAS
171
continues to accumulate—undesirable consequences, the majority of which have manifested in other Brazilian metropolises. These include a high concentration of poverty, unemployment, violence, increasing inequality and unequal economic development, and overall, a strong tendency towards social segregation. As shown previously (Cunha et al., 2006), from a demographic standpoint the creation and expansion of the MRC reveals similarities with other metropolises in the country. That is, expansion occurred as a function of high population growth, particularly in the peripheral areas of the region, although there are clear indicators that diverse processes, such as the growth of suburbs and municipal seats other than Campinas, also had an impact on growth in the metropolitan region. In the 1950s, Campinas became one of the most notable cities in the interior of the state of São Paulo both because of its dynamic economy and its population density. In a predictable manner, urbanization in the region accompanied economic growth; from 1946 to 1954 the MRC tripled its total urban area (Zimmermann, 1988). Principally in the 1970s, Campinas received large government investments from the state of São Paulo, turning it into one of the major axes of industrial expansion into the interior of the state. As a result, the major push to move industrial production from the Metropolitan Region of São Paulo to the interior of the state rapidly increased population growth in the municipality of Campinas, as well as the metropolitan region as a whole. Population growth in the last decade has progressed along several axes of expansion in the region, the majority of which follow the main highways. Whereas land occupation in three directions (west, southwest, and north) has been driven by the offer of low-income housing, there is a large concentration of housing attractive to those of higher incomes also in the north and in the southeast. These types of housing include gated communities, nature preserves, and even a complex for high technology production. Although many directions of expansion and population concentration exist, residential segregation within the region is most clearly defined by the Anhangüera Highway, which runs from the northwest to the southeast of the MRC (Cunha et al., 2006). From an economic standpoint, the MRC has progressively increased its share of industrial production in São Paulo state. In 2004, production in the metropolitan region was responsible for 7.4 percent of the state’s GDP. In fact, 7.8 percent of value-added production of São Paulo state is due to activities in the region (Fundação SEADE, 2005). Although there are differences between municipalities, particularly those farthest from the metropolitan center, the economy of the MRC remains principally an urban, industrial one.
172
PINTO DA CUNHA AND JIMÉNEZ
Social-spatial Heterogeneity As in other metropolitan regions in Brazil, Campinas has a significant level of segregation, although in some zones it is possible to observe the coexistence of residents of various socioeconomic strata, such as areas where favelas or irregular occupations are juxtaposed with middle- and high-income housing. Our previous work demonstrated spatial differences in the region according to housing infrastructure using such indicators as the connection to a wastewater system and the number of bathrooms in a household.2 Between 1991 and 2000, housing quality significantly improved throughout the region, while at the same time the concentration of precarious housing conditions increased in the most peripheral areas of the municipality of Campinas, particularly the southwest. Similarly, there is a clear differentiation between the zones delimitated by the Anhangüera Highway. There is a “corridor of affluence” to the east of the region, whereas in the southwest and west the lowerincome population tends to concentrate in what we label the “corridor of poverty”3 (Cunha et al., 2006). Whereas this spatial distribution is very distinct from that of the concentric-circles model present in other metropolitan regions such as São Paulo, it does express a center-periphery dichotomy in a different form, with the Anhangüera Highway as the dividing line.
Concepts, Measures, and Data Although studies of residential segregation in the United States most often focus on spatial differences based on race, populations also can be geographically concentrated according to socioeconomic status, lifecycle position, and/or ethnicity (Frisbie and Kasarda, 1988). In the case of Brazil, socioeconomic status is a far stronger predictor of residence than race (Telles, 1992; 1995). Indeed, measures of residential segregation by race in Brazilian cities are moderate when compared to the same measures calculated for U.S. cities. In all cases, residential segregation is meant to refer to the phenomenon where two or more social groups reside in physically distant areas of the urban fabric (Massey and Denton, 1988; 1989). However, it is important to note that residential segregation (for example, physical distance) does not necessarily equate social exclusion (that is, social distance), although it is a possible indicator of it (see Park, 1967). Regardless, researchers have embraced the idea that residential segregation is a complicated phenomenon that can exist in varying dimensions.
RESIDENTIAL SEGREGATION IN CAMPINAS
173
In most all measures of segregation, the definition of the area and scale at which segregation is measured greatly affects the results. The modifiable area unit problem (MAUP) frequently arises in spatial analysis because of the arbitrary nature in which spatial units, such as census tracts, are designed. Usually spatial units are designed for methodological ease and do not reflect the neighborhoods in which people reside. Related to MAUP is the issue of scale, in that the level of aggregation of data used will greatly affect the results of studies of segregation. Indeed, our previous work in this region illustrates that area and scale are very important when evaluating patterns of spatial segregation in Campinas. The index of dissimilarity for the MRC as a whole is very different from the indices of dissimilarity calculated using only census tracts southwest or northeast of the highway; the area of concentrated affluence (northeast) is more socially heterogeneous than that of the area of concentrated poverty in the southwest (Cunha et al., 2006). If we had used a division other than the Anhangüera highway to divide the territory of the MRC—municipal boundaries, for example—our results would have been completely different. Although we recognize the importance of these methodological issues to the study of segregation, we are limited by the nature of the data available to us. To accomplish our objectives we use data from the Brazilian census in 1991 and 2000 at the census tract (setor censitário) level, the smallest area at which Brazilian census data is available. For the Metropolitan Region of Campinas in 1991 there were 1,815 census tracts, 39 of which are omitted from our analysis due to missing information (these census tracts have fewer residents than the minimum necessary for Instituto Brasileiro de Geografia e Estatística [IBGE] to release the data). In 2000, there were 3,106 census tracts; once those tracts with missing data are excluded, we have 3,064. Here it is necessary to identify indicators that permit us to adequately characterize social-spatial segregation. As mentioned above, this phenomenon can be evaluated according to many dimensions (income, race, religion, migratory status, etc.); yet, in Brazil, there is no question that segregation principally manifests according to socioeconomic status. However, one indicator alone is not capable of revealing the great differences between people residing in a region such as Campinas. Furthermore, the socioeconomic indicators most frequently used, such as the poverty line, are problematic for several reasons. First of all, Brazil unlike the United States has no federally mandated poverty line. Furthermore, there is consensus neither on its conceptual definition nor on how to calculate the Brazilian poverty line (although generally some sort of calculation of the cost of a food basket is used). Regardless of the debate surrounding
174
PINTO DA CUNHA AND JIMÉNEZ
the meaning and measurement of poverty, we agree with other authors (see Torres et al., 2003) that poverty is multi-faceted, and income levels alone are inadequate measures of poverty. With this in consideration, in this paper we use a conjunction of indicators that reflect not only human capital characteristics of household heads, but also factors related to housing quality and household composition. We use two sets of indicators of socioeconomic status. We follow the work of Torres et al. (2003) and conduct a factor analysis to obtain summary indices of poverty.4 Using the varimax procedure, we arrive at four factors of poverty in 1991, which combined, explain 75.07 percent of the variability in the variables analyzed. When using the same procedure for 2000, we have three poverty indicators, which explain 61.71 percent of variability in the same variables. As such, we arrive at four indicators of different yet overlapping indicators of poverty for 1991—socioeconomic status, neighborhood characteristics, household structure, and family life cycle. The results of the factor analysis for 2000 are very similar, although there is no significant indicator for household structure. As our main objective in this paper is to evaluate changes in the patterns of residential segregation from 1991 to 2000, we omit this fourth indicator and concentrate on the results for the three factors significant at both points of time. Although several variables indicating the level of schooling of the household head are included in the various poverty indicators, education has been consistently shown to be a strong indicator of socioeconomic status, and, as such, we use it to calculate measures of segregation as well. Years of education of the head of household is dichotomized according to the mean for each year that D is calculated (1991 mean 5.92 years, 2000 mean 6.63 years). We also determine whether or not the head of household has a primary education (4 years or less versus more than 4 years of schooling).
Results The Changing Pattern of Social-spatial Segregation The combination of different measurements of segregation as well as different indicators with respect to poverty permits us to better clarify the magnitude and nature of segregation in the region. During the 1990s, levels of education and income increased within the MRC as did access to basic public services such as running water, garbage collection, and
RESIDENTIAL SEGREGATION IN CAMPINAS
175
the waste water system. Additionally, levels of illiteracy decreased, both among children age 7 to 14 and among heads of households. This indicates an overall improvement in the socioeconomic conditions within the metropolitan region. When calculating the index of dissimilarity we use the two dichotomized head-of-household education variables. While recognizing the difficulties of using the dissimilarity index to evaluate the phenomenon at hand, particularly considering its variability depending on the spatial unit studied (Rodríguez Vignoli, 2001; Préteceille, 2004), the index is still useful as a summary indicator to evaluate the changes in the region during the 1990s, a decade that represents a crucial time of consolidation of the metropolitization process. Table 9.1 illustrates that the indices of dissimilarity both for mean years of education and whether or not the householder had a primary education decreased slightly from 1991 to 2000. After 1991, the index of dissimilarity according to mean education decreased from 33.70 percent to 31.13 percent. This means that in 2000 a little more than 31 percent of those households whose heads had less than 7 years of schooling would have to change their census tracts of residence so that their distribution across census tracts would be equal to their presence in the metropolitan region as a whole. Similarly, in 2000 the index of dissimilarity decreased from nearly 30 to 25.57, indicating that about a quarter of those households where a head has less than a primary education would have to move residences so that their distribution across census tracts would be equal to that in the entire metropolitan region. Overall, these changes in the dissimilarity index most likely can be attributed to increases in education in the Brazilian population. According to IBGE the average years of schooling of the Brazilian population older than 10 years of age increased from almost 5 years in 1991 to 6.4 years in 2003. The changes in the index of dissimilarity must be interpreted with some caution due to the sensitivity to changes in the composition of the population. Overall however, the dissimilarity index indicates relatively low levels of segregation according to education in both 1991 and 2000. Yet, as noted above, this summary index of segregation masks differences in the urban fabric. Table 9.1 also presents the global Moran’s I values for 1991 and 2000 by education and poverty indicators. Since our final objective is to evaluate the relationship between household characteristics and quality of public schooling, we use a critical distance threshold of 2 kilometers in order to calculate the weight matrixes. We base our rationale on a state law that mandates that children should live a maximum of 2 kilometers
176
PINTO DA CUNHA AND JIMÉNEZ
Table 9.1 Index of dissimilarity and Global Moran’s I by education* and poverty indicators, metropolitan region of Campinas and municipality of Campinas, 1991 and 2000 Measure Metropolitan Region of Campinas Index of Dissimilarity (Education) Mean years of education** Primary education or less (4 years vs. 4 years) Global Moran’s I (Education) Mean years of education** Global Moran’s I (Poverty Indicators) Socioeconomic status Neighborhood characteristics Household structure*** Family Life Cycle Municipality of Campinas Global Moran’s I (Education) Mean years of education** Global Moran’s I (Poverty Indicators) Socioeconomic status Neighborhood characteristics Family Life Cycle
1991
33.70 29.98
2000
31.13 25.57
0.5293
0.5813
0.5370 0.3706 0.1600 0.2913
0.3776 0.2755 — 0.3259
0.4614
0.5047
0.5163 0.3005 0.2196
0.4062 0.2640 0.2596
Notes: Global Moran’s I results significant at the p .05 level. * Years of education of the household head. ** As noted in the text, in 1991 mean years of education is dichotomized into less than six years and six years or more of education. In 2000, mean years of education is dichotomized into less than seven years and seven years or more of education. *** Only an indicator in 1991. Source: FIBGE, Demographic Census, 1991 and 2000.
from a public school. Additionally, Jargowsky and Kim (2004) argue that by using a distance threshold to define neighbors (as opposed to relying on shared territorial boundaries) partially addresses the arbitrariness of the boundaries as defined by the census bureau. These results demonstrate that there is significant positive clustering by head-of-household education level and all poverty indicators; however, the former is stronger than the latter. These results differ somewhat from those obtained using the index of dissimilarity, although generally they indicate that segregation is lower in 2000 than it is in 1991. One interesting exception is that of mean years of education. Here, it was not necessary to dichotomize the mean years of schooling of the household head in order to arrive at the global Moran’s I. This is the only indicator of segregation by education where segregation increases, albeit slightly (from 0.53 in 1991 to 0.58 in 2000). This perhaps points to the effects of the loss of information in dichotomizing a continuous
RESIDENTIAL SEGREGATION IN CAMPINAS
177
variable. Additionally, the Moran’s I for the family-life-cycle indicator of poverty increases slightly between 1991 to 2000, from 0.29 to 0.33, indicating an increase in the clustering of neighborhoods with younger heads of households and higher dependency ratios, although here again, only slightly. When mapping local Moran’s I values, we are able to visualize where clustering of affluence and poverty occurs within the Metropolitan Region of Campinas. In our analysis, hot spots (dark gray) are significant clusters of census tracts with high values on a variable, and cold spots (gray) are significant clusters of census tracts with low values on a variable. As our previous work demonstrated (Cunha et al., 2006), there is a significant clustering of highly educated households in the center of Campinas— mainly to the northeast of the Anhangüera Highway—whereas there is a large cluster of less-educated households in the area spanning the southwest of the MRC (See maps 9.1 and 9.2). The areas surrounding the municipal seats of Americana and Paulínia in the northwest corner of the MRC prove to be exceptions to the spatial pattern described above. However, both cities are known for their industries that attract highly educated workers. Additionally, the significant cluster of highly educated heads of households in the far south of the region represents the municipal seat of Indaiatuba, which recently became popular for the development of middle- and high-income gated communities. However, when assessing changes in the pattern of hot spots and cold spots by householder education, we observe the corridor of poverty stretching further to the northwest of the region, while the corridor of affluence spreads both northwards and southwards along the Anhangüera Highway. Similarly, the maps for clustering by our measure of socioeconomic status (poverty factor 1) demonstrate the move from highly concentrated areas of affluence in the center of the metropolitan region in 1991 to a more dispersed area of high socioeconomic states to the northeast of the highway (maps not shown). The neighborhood characteristics indicator of poverty departs from this corridor pattern in distinct ways. In 1991 the neighborhood characteristics indicator is the only one that portrays a clear traditional centerperiphery dichotomy; in 2000 it continues to do so. This indicates that more recently settled areas with high concentrations of affluent households lack the same basic neighborhood infrastructure as those areas where the poor concentrate. Still, in this analysis it is the indicator with the weakest level of clustering in 1991 and remains so in 2000 (0.37 in 1991 and 0.28 in 2000). The neighborhood-characteristics poverty indicator also showed the most heterogeneity in terms of the mix between
178
PINTO DA CUNHA AND JIMÉNEZ
Municipal boundaries Local Moran’s I Results (LISA) Years of Education of Household Head 0
5
10
20 Kilometers
Not Significant High-High Low-Low Low-High High-Low
N
Map 9.1 Results for Local Moran’s I for mean years of schooling of household head. Metropolitan region of Campinas, 1991 Source: IBGE, Demographic Census 1991.
middle zones, which is confirmed by the relatively lower global Moran’s I scores. In this case, hot spots and cold spots were not the wide swaths that they are for other indicators of socioeconomic status and poverty. While there have been significant improvements in the overall level of access to better housing, these improvements are not uniform across the metropolitan region. As we shall see below, unequal access to public services is not only characteristic of the Metropolitan Region of Campinas but also a feature of the municipality of Campinas itself.
RESIDENTIAL SEGREGATION IN CAMPINAS
179
Local Moran’s I Results (LISA) Years of Education of the Household Head Not Significant High-High Low-Low Low-High High-Low
N
Map 9.2 Results for Local Moran’s I for mean years of schooling of household head. Metropolitan region of Campinas, 2000 Source: IBGE, Demographic Census 2000.
The Geography of Opportunities: Social Segregation and Public Education The intent to know, measure, and characterize existing residential segregation becomes even more important when taking into consideration that in cities, the areas where families live represent an important factor in the improvement or deterioration of their material conditions, what Galster and Killen (1995) and others (such as Sabatini, 2004) refer to as the “geography of metropolitan opportunities.” In order to demonstrate
180
PINTO DA CUNHA AND JIMÉNEZ
this hypothesis, this study uses the example of public education in two dimensions: the quality of schools with respect to their infrastructure and student achievement measured by tests administered to all students in the State of São Paulo Academic System. We intend to examine whether the areas that are most segregated, most precarious, and furthest from the region’s center are also where the majority of schools with fewer resources and worse student performance are located. In 1996 the Brazilian government amended the constitution to ensure universal enrollment of all children from first to eighth grades. This amendment also dictated that 15 percent of municipal and state tax revenue be spent on public schools, as well as guaranteeing a minimum amount of federal spending for each student in these grades (Torres, Ferreira, and Gomes, 2005). However, universal enrollment of Brazilian children has not ensured universal levels of educational quality. Since all public schools are funded by the state or municipal government, theoretically, all schools should be of equal quality. Instead, at least in Campinas, school infrastructure and student academic achievement vary significantly from school to school. We obtained information about public schools from two sources. First, the 2003 Brazilian school census provides information on the structural quality of schools using the following four dichotomous indicators: whether the school has a library; whether the school has a sports field; whether the school has a computer lab; and whether the school has a science lab (for all variables, yes 1). These were the four measures of infrastructure that demonstrated the most variability between schools. When these four services are summed we obtain a single measure of school infrastructure by school which ranges from 0 to 4. Second, the System of Evaluation of Academic Achievement in the State of São Paulo (SARESP) is a standardized test administered to students attending state schools. The main objective of this test is to monitor the quality of the state educational system. In 2000, the test was administered to students enrolled in state schools in the fifth and seventh grade. Test scores range from 0 to 100 percent; in 2000, the average score in the Metropolitan Region of Campinas was 45 percent. In this case, information is only available for state-administered schools (n 84 schools). To begin, segregation (as measured by clustering) in the municipality of Campinas exists at similar levels as it does for the MRC as a whole (see table 9.1). Next, table 9.2 illustrates the location of public schools that offer up to one service and all four services, respectively, compared to hot spots, cold spots, and mixed areas as defined by the mean years of education of the household head. Overall, 30 percent of public schools in the municipality
RESIDENTIAL SEGREGATION IN CAMPINAS
181
Table 9.2 Spatial distribution of school infrastructure by Local Moran’s I score, Municipality of Campinas, around 2000 Local Moran’s I score of the where the school is located (total number of schools in that type of neighborhood in parentheses) No significant clustering (52) Clustering of high-education sectors (cold spot of poverty) (34) Clustering of low-education sectors (hot spot of poverty) (73) Low-education sector surrounded by high-education sectors (13) High-education sector surrounded by low-education sector (12) Municipality of Campinas (184)
Number of schools Number of schools Average with zero or one with all SARESP score infrastructure four infrastructure (number of measure measures schools in parentheses) 16 8
12 10
44.87 (23) 50.15 (20)
25
8
41.81 (32)
4
1
47.75 (4)
4
3
45.00 (5)
55
34
45.00 (84)
Source: 2000 Demographic Census and 2003 Censo Escolar.
offered 0 (no) or 1 service, whereas slightly more than 18 percent had all four infrastructure measures present. Geographically, poorer quality schools (those that offer no or one service) primarily are distributed towards the periphery areas of the municipality, although there are some situated in the central areas of the city (map not shown). When comparing the location of lower quality schools in relation to the hot spots and cold spots of poverty identified in our previous analyses, about half of these schools fall within these hot spots of poverty, although some are also located in cold spots of poverty (as defined by head of household education). At the other extreme, the situation is somewhat different; those schools that offer all four services are mostly concentrated in the center of Campinas and its surroundings. These high-quality schools are located in hot spots of poverty as well, but according to our classification only about 11 percent of schools located in hot spots of poverty can be qualified as “high quality,” whereas the same could be said of almost 30 percent of schools located in cold spots of poverty. The distribution of schools according to their mean scores (less than or equal to the mean score of 45 percent versus above the mean) on the SARESP exam in 2000 is also presented in table 9.2. When examining mean SARESP scores by the Local Moran’s I of the setor censitário where they are located, the average SARESP school score is more than 8 percentage points higher
182
PINTO DA CUNHA AND JIMÉNEZ
in areas of concentrated highly educated households than in areas with significant concentrations of less-educated households. Meanwhile, average scores for schools located in mixed areas or areas where there was no significant clustering are roughly similar and fall between the scores for hot spots and cold spots of poverty. What is most notable, however, is the low average SARESP score of students in the region (45.1 percent, 5 percent). We should emphasize that in Brazil, as a rule, low-income children only attend public schools, a fact that exacerbates the concentration of poverty and its effects on schooling. It is also clear that when we are dealing with public schools in Brazil, many centrally located schools also serve students residing in the peripheries of the municipality. Luhr and Cunha (2004) illustrate this when analyzing the case of one of the oldest schools in Campinas, located in the center of the municipality in one of the highest priced areas of the city. Overall, this does not diminish the importance of the results here emphasized since by law children should study in the areas closest to their residences. Overall, few students attend schools outside of their immediate area, as verified using information from the origin/destination5 survey conducted in the region in 2003.
Conclusions Our LISA analysis demonstrates that there are significant and large clusters of low- and high-poverty areas in both 1991 and 2000, even though our summary measures of segregation indicate moderate levels of residential segregation by education and poverty for both these years. We do not consider these results to be contradictory; rather, they point to the many facets of residential segregation, and highlight the importance of using more than one measure of segregation when studying this phenomenon. If Campinas is to be taken as a case of one of the many smaller, faster growing metropolises of Latin America, this may bode well for the future of metropolises in the region. Indeed, in the United States it is the fastest growing cities in the south and the west of the country with the lowest levels of residential segregation according to race and ethnicity (Glaeser and Vigdor, 2001). Perhaps this means that the new metropolises of Brazil are not condemned to repeat the same pattern of segregation and social exclusions as the cities that have preceded it, such as São Paulo and Rio de Janeiro. Furthermore, as our data on neighborhood infrastructure and public school quality show, equal access to services does not ensure equal quality of services. Our analysis demonstrates that the lack of utilities and waste disposal is not as concentrated as poverty; indeed, clusters of areas
RESIDENTIAL SEGREGATION IN CAMPINAS
183
lacking these public services are located throughout the metropolitan region—both in clusters of high- and low-socioeconomic status households. Furthermore, low-quality public schools tend to be found in poorer areas, but schools lacking infrastructure are found in wealthy areas of the municipality as well. The difference, though, is that affluent households tend to have the financial recourses to purchase goods to substitute the lack of public provision of services (for example, private schooling), whereas poor households have little choice but to utilize whatever services are offered. Even though residential segregation is not as extensive as in other cities, the opportunities for upward mobility for poorer residents in the Metropolitan Region of Campinas are constricted by the lack of quality public schooling available to them.
Notes 1. At this time school quality information is only available for those schools located in the municipality of Campinas. However, it is important to note that approximately 50 percent of public schools in the metropolitan region are located in the municipality of Campinas. 2. The greater accessibility to basic sanitation services in São Paulo state, in addition to the poor quality of information about basic sanitation because of citizens’ difficulties in distinguishing what type of service they have, leads to our conclusion that the number of bathrooms in a household is a more powerful predictor of differences in the structural quality of households. 3. Further analysis reveals that these regions are also ones with higher proportions of children and school-age population. For more information, see NEPO/ NESUR 2004. 4. The following characteristics of census sectors were used in the factor analysis: Percentage of children 7–14 years old who are illiterate; percentage of illiterate heads of households; percentage of heads of households with an elementary school education or less; average years of schooling of the heads of households; average monthly head-of-household income; percentage of households without garbage collection; percentage of households without piped water; percentage of households without a bathroom; percentage of households not connected to the sewage system; percentage of rented households; percentage of household heads 10–29 years old; average age of heads of households; percentage of the population who are young or old-age dependents; and the percentage of household heads of who are female with an elementary school education or less. 5. The origin and destination survey was conducted for the first time in the Metropolitan Region of Campinas in 2003 by the Empresa Paulista de Planejamento Metropolitano SA (EMPLASA), part of the Secretaria de Estado de Economic e Planejamento, with the objective of measuring people’s daily trips from their households.
184
PINTO DA CUNHA AND JIMÉNEZ
References Cunha, J. M. P., Eichman, J., Jiménez, M. A., and Luhr, I. Trad. (2006). Expansão metropolitana, mobilidade espacial e segregação nos anos 90: O caso da Região Metropolitana de Campinas. In José Marcos Pinto da Cunha (ed.), Novas metrópoles paulistas: População, vulnerabilidade e segregação, pp. 337–63. Campinas, SP: NEPO/UNICAMP. Frisbie, W. P., and Kasarda, J. D. (1988). Spatial processes. In N. Smelser (ed.), Handbook of Sociology, pp. 629–66. Newbury Park, CA: Sage. Fundação Sistema Estadual de Análise de Dados—SEADE. (2008). Available online at http://www.seade.gov.br/produtos/pibmun/index.php. Last accessed December 12, 2008. Galster, G. C. and. Killen, S. P. (1995). The geography of metropolitan opportunity: A reconnaissance and conceptual framework. Housing Policy Debate, 6(1), 7–43. Glaeser, E. L. and Vigdor, J. L. (2001). Racial segregation in the 2000 census: Promising news. Washington, DC: Center on Urban and Metropolitan Policy Survey Series, Brookings Institute. Jargowsky, P. A. and Kim, J. (2004). A measure of spatial segregation: The generalized neighborhood sorting index. Political Economy Working Paper. School of Social Sciences, University of Texas at Dallas. Luhr, I. T. and Cunha, M. J. (2004). Acessibilidade aos equipaments da educação pública em um grande centro urbano: O caso de Campinas. Research Report. Campinas, SP: NEPO/UNICAMP. Massey, D. S. and Denton, N. A. (1988). The dimensions of racial segregation. Social Forces, 67, 281–315. ——— (1989). Hypersegregation in U.S. metropolitan areas: Black and Hispanic segregation along five dimensions. Demography, 26, 373–91. NEPO/NESUR. (2004). Atlas da Região Metropolitana de Campinas. CD-ROM. First ed. Campinas, SP: Universidade Estadual de Campinas. Park, R. E. 1967. The urban community as a spatial pattern and a moral order. In R. H. Robert (ed.), Robert Park on social control and collective behavior, pp. 55–68. Chicago: University of Chicago Press. Préteceille, Edmund. (2004). A construção social da segregação urbana. Espaços & Debates, 25(45), 11–23. Rodríguez Vignoli, J. (2001). Segregación residencial socioeconómica: ¿Qué es? ¿cómo se mide? ¿qué está pasando? ¿importa? Serie Población y Desarrollo, 16. Population Division of the Economic Commission for Latin America and the Caribbean Santiago de Chile (CELADE): ECLAC. Sabatini, F. (2004). Medición de la segregación residencial: Reflexiones metodológicas desde la ciudad latinoamericana. In F. Cáceres and F. Sabatini (eds.), Barrios cerrados en Santiago de Chile: Entre la exclusión y la integración residencial. Cambridge, MA: Lincoln Institute of Land Policy. Sabatini, F., Cáceres, G., and Cerda. J. (2001). Segregação residencial nas principais cidades chilenas. Revista EURE, 82(27), 21–42.
RESIDENTIAL SEGREGATION IN CAMPINAS
185
Telles, E. E. (1992). Residential segregation by skin color in Brazil. American Sociological Review, 57, 186–97. ——— (1995). Structural sources of socioeconomic segregation in Brazilian metropolitan areas. American Journal of Sociology, 100(5), 1199–1223. Torres, H. da Gama. (2004). “Segregação Residencial e Políticas Públicas: São Paulo na década de 1990,” Revista Brasileira de Ciências Socias 54. Torres, H. da Gama and Gomes, S. (2002). Desigualdade educacional e segregação social.” In Novos Estudos Cebrap, 64, 132–140. Dossiê Espaço, Política e Políticas na Metrópole Paulistana. Torres, H. da Gama, Ferreira, M. P., and Gomes, S. (2005). Educação e segregação social: Explorando o efeito das relações de vizinhança. In E. Marques and H. Torres (eds.), São Paulo: Segregação, pobreza e desigualdade, pp. 123–42. São Paulo: Editora do Senac. Torres, H. da Gama, Marques, E., Ferreira, M. P., and Bitar, S. (2003). Pobreza e Espaço: Padrões de Segregação em São Paulo. Estudos Avançados 17(47): 1–37. Zimmerman, G. (1988). O Município no sistema tributário: Os municípios paulistas e o caso de Campinas. In: Fundação SEADE (ed.), A interiorização do desenvolvimento econômico no Estado de São Paulo (1920–1980). São Paulo: Fundação SEADE.
This page intentionally left blank
10
Changing Patterns of Residential Segregation in Austin Carolina Flores and Robert H. Wilson
The Literature on Urban Form and Poverty in the United States Spatial Form of the Cities and Racial/Ethnic Segregation in U.S. Cities Rapid urban growth and suburbanization in the United States, dating from the early part of the twentieth century, coincided with increasing residential segregation by race and income (Massey and Denton, 1988; 1993). During the early phases, industrialization was a primary factor in shaping new patterns of urban form. In searching for more land-intensive forms of production and after the truck became an important mode of transportation, manufacturing activities moved to suburban areas. The rapid growth in the demand for labor attracted large-scale European migration and the African American population from the South, and residential segregation by race became pronounced. Segregation was achieved through legal and non-legal, often violent, means (Massey and Denton, 1988; 1993). In the 1960s, federal legislation prohibited discrimination based on race and/or ethnicity in housing markets. But the national patterns of residential segregation changed slowly. Between 1970 and 2000, modest improvement in segregation and isolation of the black population could be observed as the black-to-white dissimilarity index declined (Glaeser
188
FLORES AND WILSON
and Vigdor, 2003, pp. 218–19), but today it remains high, especially in Northern cities (Logan, 2003, p. 239). The rapid growth in the Hispanic population in the 1970s became a new and significant demographic trend. By 2000, the Hispanic population represented 13 percent of the U.S. population. The heavy concentration of Hispanics in particular areas of the country declined in the 1990s; today, significant Hispanic populations can be found in most areas of the country. The national dissimilarity indices of Hispanic-white segregation and isolation are somewhat lower than for blacks-white segregation, but the index for Hispanics is high in metropolitan areas with large Hispanic populations (Logan, 2003, pp. 239, 247). In the 1990s, changes in the racial/ethnic composition in the larger metropolitan areas were quite significant (Fasenfest, Booza, and Metzger, 2004). In most large cities and even in a few metropolitan areas— including New York, Los Angeles, and Houston—the white population held a minority share of total population. In 2000, the number of predominately white neighborhoods in the ten largest metropolitan areas decreased by 30 percent, and nine of these ten areas witnessed increases in mixed-race neighborhoods, including some neighborhoods in the suburban areas (Fasenfest et al., 2004). Patterns of residential segregation in urban America are undergoing significant change, but the patterns are complex and vary across regions and by size of the cities. Differential growth rates across racial/ethnic groups contribute to this complexity.
Patterns of Neighborhood Poverty Poverty rates in the United States are significantly higher for racial and ethnic minorities than for the white population (Mishel, Bernstein, and Allegretto, 2005, chapter 5). These differences are manifested spatially in cities due to residential segregation. In metropolitan areas, a majority of the poor are white, but unlike African Americans and Hispanics, poor whites are not concentrated in high-poverty areas (Jargowsky, 2003). Poor whites are not likely to live in neighborhoods with only other poor whites but rather are relatively dispersed throughout the metropolitan areas and in rural areas. Poor blacks and, to a lesser extent, poor Hispanics are likely to live with individuals in similar conditions. In the 1980s, poverty among minorities became increasingly concentrated, and by 1990, 3,417 census tracts, containing a total population of 8.4 million, had poverty rates exceeding 40 percent (Jargowksy, 2003, p. 2). In the 1990s, a remarkable reversal occurred, and the number of
RESIDENTIAL SEGREGATION IN AUSTIN
189
people living in high-poverty neighborhoods declined by 25 percent. The steepest declines occurred in the Midwest and South, occurring among most groups, but especially among the African American population (Jargowsky, 2003, p. 5).
Methods and Definitions In this paper we measure the level of socioeconomic residential segregation in the Austin Metropolitan Area (Austin MSA), with particular attention given to changes within the Austin city limits. In addition, we analyze the level and changes in socioeconomic residential segregation during the decade of the 1990s and relate these changes to educational achievement. Residential segregation refers to the way in which two or more population groups are distributed in space. A city is socioeconomically segregated when the group that is defined as being socioeconomically disadvantaged is spatially separated from those that are socioeconomically advantaged. In contrast, an integrated city facilitates spatial interaction between these two groups. Measures of segregation generally assess the spatial separation between two groups: the majority and the minority. In this paper we divide the population into minority and majority using four different classifications. In order to measure the level of residential segregation between socioeconomic groups, the population is classified as follows: (1) poor (minority) and non-poor1 (majority); and, (2) adults 25 years old and older with high school education or less (minority) and adults with more than high school education (majority). In order to measure the level of residential segregation between races and ethnicities we divide the population into black (minority) and non-black (majority) and into Hispanic (minority) and non-Hispanic (majority). In one way or another, all measures of residential segregation account for the way in which population composition in a particular spatial subunit— for example, a neighborhood within a city—relates either to the population composition in the overall area, (the city), or in the other regional subunits, (the neighborhoods). Thus, the issue of defining a regional subunit is critical for interpreting the measures of segregation. In practice, however, the selection of the spatial units is arbitrary. It responds to the availability of data rather than to theoretical considerations. This paper uses census block groups as a proxy for the neighborhood, but recognizes its limitations for defining neighborhoods. The Austin MSA is divided into 800 census block groups, with an average of 300 people per unit.
190
FLORES AND WILSON
Overview of Austin Metropolitan Area The Austin Metropolitan Area is formed by Travis, Williamson, Bastrop, Caldwell, and Hays counties in central Texas. The city of Austin is set in the middle of Travis County and is surrounded by smaller cities in bordering counties: Georgetown in Williamson County, Elgin and Bastrop in Bastrop County, Lockhart in Caldwell County, and San Marcos in Hays County. The city of Austin dominates the urban geography of the MSA.
Urbanization and Racial Geography The growth of Austin and the evolution of its spatial demography vary significantly from that of the northern industrial cities of the United States, as discussed above. First, Austin is a relatively young city. In 1860, its population was 3,500 and did not reach 100,000 until the 1940s. It did not develop a significant manufacturing base until the 1970s. The racial and ethnic demography of the city follows a distinct historical pattern. In 1860 there were around 1,000 slaves in Austin, accounting for almost 30 percent of the population (Humphrey and Crawford, 2001). Given slavery, there were no distinct African American settlements. After the Civil War, the number of African Americans in Austin quickly increased. Whereas the Anglo population grew at an estimated 12 percent between 1870 and 1900, the African American population grew by 60 percent, the result of the large-scale migration of newly emancipated African Americans seeking work in cities. At this time, distinct settlements in Austin, populated by African Americans seeking better living conditions and more affordable housing, appeared in what is now considered East Austin and to the north and west of the city (Henneberger and Huff, 1979; Manaster, 1986). Despite the development of these settlements, African American housing remained somewhat spatially integrated during this period, largely due to the fact that many African Americans continued to work for white families. Mexicans were present in Austin in the 1850s, but in very small numbers (Humphrey and Crawford, 2001). The first recorded Mexican settlement in the city occurred in the 1870s on the banks of the Colorado River and Shoal Creek, areas considered undesirable for housing by the white population (Hamilton, 1913; Manaster, 1986). Between 1900 and 1930, housing patterns in the city become more segregated (Humphrey and Crawford, 2001). The application of Jim Crow laws, the advent of the Ku Klux Klan, and the 1928 city of Austin zoning
RESIDENTIAL SEGREGATION IN AUSTIN
191
ordinance contributed to the increased segregation (Henneberger and Huff, 1979). Under the latter, public facilities were only made available to the black population residing in East Austin, a strategy intended to concentrate African Americans. Spurred by the Mexican Revolution in 1910, the Mexican population began to increase significantly after 1920, and by 1930 it represented 10 percent of the Austin population (Henneberger and Huff, 1979; Humphrey and Crawford, 2001). The new Mexican migrants located principally in East Austin to the south of the African American area. A Catholic church was built in East Austin to minister to the growing Mexican population. Mexican migration continued through the 1940s, but Mexicans started locating in South Austin, producing a somewhat more spatially dispersed pattern than that found in the African American population, which was almost entirely concentrated in East Austin by the 1960s (Henneberger and Huff, 1979; Humphrey and Crawford, 2001). The East Austin communities’ isolation from the city was reinforced in the 1960s by the location of the major interstate highway, IH-35, thus creating a barrier a few blocks to the east of downtown. In the 1970s and later, the residential patterns of these two groups changed (Henneberger and Huff, 1979). Housing discrimination by race and ethnicity was prohibited by federal law in the 1960s, allowing greater dispersal of the two groups in the following decades. The black population in East Austin started to decline in the 1970s through a process of out-migration, but it increased in other areas of the city, especially in the northeast. The Mexican, and, increasingly, the Mexican American population, grew substantially in both East and South Austin.
Population Total population in the Austin MSA increased by 47.6 percent during the 1990s, reaching roughly 1,250,000 in 2000 (U.S. Census, 2008a). This rate of increase substantially exceeds that of the population in the whole state of Texas, which grew by 28.2 percent. The bulk of the MSA population resides in Travis County (68 percent in 1990) and particularly within the city limits (57 percent in 1990). Nonetheless, population started to spread toward the north; between 1990 and 2000, Williamson County grew 6 percent per year, ending the decade with 20 percent of the MSA’s total population. Likewise, Georgetown’s population grew 75.8 percent between 1990 and 2000, in contrast to 32.8 percent for the city of Austin. Bastrop and Hays counties
192
FLORES AND WILSON
grew by about 4 percent a year during the decade and accounted for 5 and 8 percent of the MSA population in 2000, respectively. Caldwell, the smallest county, grew 2 percent per year, accounting for roughly 2.5 percent of the MSA population. Travis County grew at the relatively moderate rate of 3.5 percent per year during the decade and represented 65 percent of the MSA population in 2000. Between 1990 and 2000, individuals in the Austin MSA living below the poverty threshold decreased from 15.9 to 11.1 percent. Poverty was particularly problematic in Caldwell County in 1990, where 30.9 percent of the population lived in poverty, but the rate declined to 13.1 percent by 2000. Similarly, the poverty level in Williamson County noticeably decreased from 10.1 percent in 1990 to a meager 4.8 percent in 2000. The reduction of poverty was also significant in Bastrop and Hays counties: from 17.9 percent and 20.9 percent in 1990 to 11.6 percent and 14.3 percent in 2000, respectively. Travis County and the city of Austin present a somewhat slower decrease in the poverty level of 3.5 and 3 percentage points, respectively. But at the end of the decade, the poverty rate in Austin exceeded that of all the counties, unlike 1990 where higher rates were found in most of the outlying counties. The level of education of the adult population also increased during the 1990s. In 1990, 59.5 percent of adults 25 years old and older had attained more than the high school level, whereas in 2000, 64.9 percent had done so. Bastrop and Hays counties show an important increase in education levels of the adult population. Williamson County seems to attract highly educated adults: by the end of the decade, the county achieved levels of education roughly similar to those in Travis County and the city of Austin. These are the places where adults with high levels of education concentrate. In the Austin MSA, the Hispanic population has increased from onefifth in 1990 to one-fourth of total population in 2000. The Hispanic population is overrepresented in Caldwell and Hays counties and is becoming increasingly important for the city of Austin. In contrast, the African American population diminished from 9.4 to 8 percent in the MSA between 1990 and 2000. This trend is found in all counties except Hays and Williamson, where the African American population is small and steady. Hispanic and black adults have lower levels of education than the average: in 2000, 35.3 percent of Hispanic and 50.7 percent of black adults had more than a high school level as compared to 70.8 percent of white adults. Similarly, 18 percent of Hispanics and 18.1 percent of African Americans lived below the poverty line, as compared to just 8 percent of the white population.
RESIDENTIAL SEGREGATION IN AUSTIN
193
Changes in the Spatial Form of Austin: 1990–2000 This section analyzes changes in socioeconomic residential segregation during the 1990s along three dimensions: unevenness of the population distribution, clustering, and the isolation of the underprivileged population. The measures are calculated for three population categories (each measured with three values): (1) the poor; (2) those with low levels of education; and, (3) racial/ethnic minorities. Although we anticipate that the pattern of segregation is similar for these three categories, performing a separate analysis can reveal the complexities of the segregation patterns. The analysis focuses on the entire Austin MSA, but special attention is given to the population within the city limits. Spatial Distribution of the Population Between 1990 and 2000, the share of population living below the poverty line decreased in all census block groups, but the overall pattern indicates that poverty outside of Austin declined more quickly than in the city itself. Areas with higher-than-average poverty rates (15.9 percent in 1990 and 11.1 percent in 2000) spread toward the southeastern parts of the city during the decade. The overall level of education among the adult population has increased throughout the MSA, but the increase in adult education levels is particularly noticeable in areas more distant from Austin. The levels of education within Austin changed little during the decade. In terms of race and ethnicity, the Hispanic population is concentrated southeast of the Austin MSA, whereas the African American population is spatially concentrated northeast of the city, closer to the city center. Residential Segregation One of the most important dimensions of residential segregation is the degree to which groups are evenly distributed across the spatial subunits. According to this dimension, socioeconomic residential segregation is null if all census block groups are an exact replica of the Austin MSA with respect to the share of the poverty population. Any departure from this equilibrium is interpreted as segregation of the poor. The dissimilarity index tells us what proportion of the population should move from one spatial subunit to another in order to make all subunits alike in the whole area. Thus, a high dissimilarity index mirrors a high level of segregation.
194
FLORES AND WILSON
The poverty population is unevenly distributed across the Austin MSA (see table 10.1), and it became slightly more so in the 1990s. The proportion of the poverty population that would need to be moved from one neighborhood to another in order for each census block group to replicate the Austin MSA increased from 43.5 percent in 1990 to 44.3 percent in 2000. Similarly, whereas in 1990 38.3 percent of adults would need to move in order to achieve an even distribution in terms of level of education, in 2000 the number increased to 41.6 percent. Within the city of Austin, however, spatial segregation of the poor has slightly declined, but in striking contrast, the segregation of adults with a high school education or less has noticeably increased. The poverty population is becoming slightly more evenly represented across neighborhoods, whereas the adult population with low levels of education is becoming more unevenly distributed. This seems counterintuitive, since poverty might be considered a proxy of low education and vice versa. The explanation might be that Hispanic populations consistently have lower levels of education than the rest, as observed above. In 2000, 35.3 percent of Hispanic adults had more than a high school education, compared to 65 percent of all adults. Thus, the trend in the segregation of adults with low levels of education is actually capturing the trend in the segregation of the Hispanic population. Within the city of Austin, the proportion of Hispanics that would need to move from one neighborhood to another in order to make the distribution of the population even increases from 39.3 percent in 1990 to 44.5 percent in 2000 (see table 10.1). The change is similar to the level of Table 10.1
Measures of segregation, 1990–2000 Poverty Population
1990 2000 Dissimilarity Index (in percent) City of Austin 43.0 40.7 Austin MSA 43.5 44.3 Isolation Index (in percent) City of Austin 16.4 14.6 Austin MSA 15.8 14.4 Global Moran’s I* Austin MSA 0.344 0.198
Adults/Less than High School
Black Population
1990 2000
1990 2000
1990 2000
38.0 38.3
45.3 41.6
52.3 45.7 52.1 45.2
39.3 44.5 40.6 42.9
20.3 20.4
24.9 21.3
28.1 17.4 25.2 15.2
19.3 19.3
0.520 0.610
0.715 0.701
Note: * Weight Matrix: distance based/ nearest neighbor (2); p-value .001 on all values. Source: Calculated using census data.
Hispanic Population
23.4 20.8
0.718 0.759
RESIDENTIAL SEGREGATION IN AUSTIN
195
segregation of the adult population with low levels of education. In the Austin MSA as a whole, the proportion of Hispanics that would need to move from one neighborhood to another in order to reach an even distribution of the population increased less noticeably, from 40.6 percent in 1990 to 42.9 percent in 2000. Again, the change is similar to the change in segregation in terms of education at the MSA level. The decline of the segregation of the poor within Austin, juxtaposed to the slight increase in segregation for the entire MSA, deserves further exploration. Given that poverty rates in all counties declined substantially during the decade, the segregation pattern would be consistent with a concentration of the poor in unincorporated areas of the MSA, where housing standards are less onerous and housing is less expensive. An alternative hypothesis is that the city of Austin is more successful in its spatially targeted welfare policy than are other local governments, including county governments. These hypotheses cannot be fully tested with the data available in this study, but certainly warrant further investigation. Residential segregation can also be understood as the probability that a member from one group interacts with a member from another group. Neighborhood homogeneity decreases the probability of interaction between groups and increases isolation. At first glance, this dimension seems rather similar to the dimension of evenness in the population distribution. The main difference is that exposure is sensitive to the number of minority members that reside in a certain regional subunit. Thus, the contribution to the level of isolation of a small neighborhood will be small, even though this neighborhood is inhabited mainly by members of the minority. The level of isolation also depends on the proportion of the minority in the whole encompassing area. That is, if the overall poverty level is low, isolation will be also low, whereas isolation increases when the poverty level increases. In this analysis we correct the index of isolation by the proportion of minority in order to be able to compare the levels of isolation in 1990 and 2000. Once the correction is made, the poverty population is found to have become less segregated (isolated) in both the city of Austin and the Austin MSA (see table 10.1). In fact, in the Austin MSA the probability that a poor person shares the neighborhood with another poor person diminishes from 15.8 percent in 1990 to 14.4 percent in 2000. Similarly, the black population has become substantially less isolated. The results from the dissimilarity and the isolation indexes for the Austin MSA appear to be contradictory. While the distribution of the poverty population is becoming more uneven (that is, dissimilarity increases), the poverty population is less isolated. This pattern could be
196
FLORES AND WILSON
produced if the poverty population in the Austin MSA is becoming more unevenly distributed in less populated areas. This would lead to a more uneven distribution of the population without increasing the probability of interaction between the poor. Less-educated adults and the Hispanic population both were more isolated in 2000 than in 1990 in the city of Austin and in the Austin MSA (see table 10.1). The increase in isolation is more intense in the city than in the MSA. Again, segregation of less-educated adults reflects the spatial segregation of the Hispanic population. At the end of the period, the black population achieved levels of isolation substantially lower than for the Hispanic population. A third way of understanding segregation relates to the degree to which groups form homogeneous clusters in space. This is called the “clustering” dimension of segregation and is measured in two ways. We first use a global measure—Moran’s I—that measures the degree to which the socioeconomic characteristics in any given neighborhood are similar to those in the contiguous neighborhoods. Using such a measure, one can predict the degree to which the poverty level in a given neighborhood is a good predictor of the poverty level in the contiguous neighborhood. Second, we use a local measure. The Local Moran’s I index spatially locates the segregated areas. Segregated areas are also called “hot spots,” or areas in which certain characteristics tend to concentrate. Technically, the Local Moran’s I identifies those census block groups with certain characteristic, for example, size of poverty population, that are significantly surrounded by census blocks with similar characteristics. In other words, hot spots form the homogeneous clusters. When the Moran’s I index is significantly larger than zero, the spatial distribution of the attributes is spatially correlated. The distributions of the poverty population and adults with low levels of education are positively correlated in space (see table 10.1); that is, populations that are poor and have low levels of education are clustered. Black and Hispanic populations are even more spatially clustered. The local spatial Moran’s I can be used to identify these clusters.2 Segregated areas with low levels of education are located mainly east of IH-35. To the northeast, the band of segregation stretched from Georgetown to Taylor in 1990, but this band was pushed farther to the east and away from Georgetown during the decade. To the southeast, the area of segregation stretched from south of U.S. Highway 290 to Elgin in a broad swath to San Marcos, a pattern little changed from 1990 to 2000. The spatial concentrations of populations with relatively high levels of education (majorities), indicated on the maps 10.1 and 10.2 in light gray, are found in West Austin in 1990 and extended further to the west in 2000.
Legend Main Roads Urban Areas Not Significant Hot Spot Cold Spot Integrated Area
Map 10.1 Local Moran’s I: Adult population with high school or less, 1990 Austin MSA: Local Moran on Education, 1990 Source: US Census Bureau in both cases. The data was accessed in 2008.
Legend Main Roads Urban Areas Not Significant Hot Spot Cold Spot Integrated Area
Map 10.2 Local Moran’s I: Adult population with high school or less, 2000 Austin MSA: Local Moran on Education, 1990 Source: US Census Bureau in both cases. The data was accessed in 2008.
198
FLORES AND WILSON
The spatial clustering for the black and Hispanic population is substantial, as indicated by the Global Moran’s I values in table 10.1. The most impressive evidence is the broad swaths of light gray to the west, indicating areas with limited minority population. The black population is clustered to the east and northeast of Austin and the Hispanic to the east, southeast and south of Austin toward San Marcos.
Residential Segregation and Public Services: Education and Access to Fire Emergency Services The Provision of Public Education in Austin Despite the very rapid increases in the United States toward universal coverage of primary education in the early twentieth century and in secondary education by mid-century, the public education system is marked by remarkable degrees of racial/ethnic segregation. Significant disparities in performance are found today across schools and population groups. The racial/ethnic composition of schools and their performance remain central challenges to the U.S. system. As one of many school systems subject to the U.S. Supreme Court ruling that separate educational facilities based on race were unconstitutional, the Austin Independent School District had to desegregate in the 1960s.3 This action led to the expansion of several school districts in suburban areas of Austin as a result of “white flight.” After many years of delay, Austin adopted a busing solution, taking children from one neighborhood to another in order to achieve racial balance within schools. Nationally and locally, support for busing declined over time and other alternatives were devised. Magnet schools, for example, provided specialized and high-level courses in areas such as math and science, languages, or the arts. These schools were located in the east side of the city, thus inducing higher socioeconomic status (SES) families to voluntarily bus high-achieving children to those schools. Although within a school the composition of the magnet students differs from that of the local student body, individual schools were able to demonstrate racial integration. In terms of student achievement, defined by the share of students passing the third-grade standardized test, schools to the east trail significantly those of other areas. Relatively low achievement scores in schools on the east and higher achievement in schools to the west and north mirror the patterns of segregation by educational attainment in the general population. Given the broad consensus of the effect of parents’ education on student education, the spatial segregation of adults with low educational
RESIDENTIAL SEGREGATION IN AUSTIN
199
achievement, both in terms of effect of the family and the neighborhood, may have a deleterious effect on student achievement to the east and south. City form affects the spatial demand for education services and spatial disparities in outcomes, thereby affecting the efficacy of educational services themselves. The children currently being educated in East Austin will not necessarily remain there as adults and thereby might not contribute to the reproduction of the spatial patterns in the future. Residential mobility in the United States is, in fact, quite high and the history of the African American community moving out of East Austin, noted above, illustrates this mobility. But the more significant development in this analysis is found in the growing Hispanic population. As a result of relatively low educational achievement and spatial segregation, the schools in areas with high shares of Hispanic population face great challenges.
Access to Fire Emergency Services To illustrate another application of segregation analysis to the study of public policy, we examine the provision of fire emergency services. An index for average distance, corrected for the share of the population residing in each area, to the closest fire station from each of the segregated (hot spot)and affluent (cold spot) areas was calculated (see table 10.2). In 1990, the average distance to the nearest fire station is almost three times higher outside than inside the boundaries of the city of Austin. In general and as expected, the city is better served than the rural areas in the Austin MSA, Table 10.2 Index of average distance to the closest fire station from segregated and non-segregated areas; 1990 and 2000 (Average distance in 1990 5 100) 1990
2000
City
City
Total
City
City
Total
Hot Spot
48.2 (64)
147.4 (111)
111.1 (175)
59.3 (80)
131.6 (71)
93.3 (151)
Cold Spot
43.6 (103)
n/a (0)
43.6 (103)
49.5 (157)
169.4 (8)
55.3 (165)
Total
62.8 (544)
177.3 (262)
100.0 (806)
64.7 (523)
201.2 (242)
107.9 (765)
Source: Census Data. Index equals the distance (Mi)*(Population of census block/population of MSA), normalized to average distance in 1990 5 100. Results for mixed areas are not shown.
200
FLORES AND WILSON
a finding that holds for both segregated and affluent areas, although there has been a slight deterioration for the general population over the decade (around an 8-percent increase in the index).4 Affluent areas have greater accessibility to services than do segregated areas, although the difference declined over the decade. Despite some deterioration in access for all segregated areas during the decade, access for those areas within city limits improved. People in segregated areas, on average, were better served in 2000 than in the previous decade due to the fact that a larger share of the segregated population was located within the city limits, where service is better than in rural areas. In 1990, 24 percent of segregated areas and 37 percent of population living in segregated areas were located within the city limits, whereas in 2000, 52 percent of the segregated areas and 57 percent of the population were in the city. In contrast, the population residing in affluent areas is increasing more rapidly outside the city. In 2000, a few areas of concentrated affluence, gathering 6 percent of the population of all affluent areas in the MSA, appear outside the city limits. Located in the environmentally attractive northwestern part of the Austin MSA, these areas are quite distant from the segregated areas in the east. But interestingly, these areas have substantially inferior fire and emergency services compared to their counterparts within the city limits (index values in 2000 of 169.4 percent and 49.5 percent, respectively).
Conclusions To illustrate the usefulness of the analysis of spatial clustering for public policy, two issues—quality of education and accessibility to fire services— were explored. The urbanization of the segregated areas (the segregated population living within the city limits which increased from 37 percent to 57 percent of all segregated population over the decade) and the suburbanization of the affluent areas (affluent population living outside the city limits which increased from 0 percent to 6 percent of total affluent population) have important implications for urban policy. The increased segregation of lower-educated populations that occurred at both the city and MSA level is likely exacerbating the problem of access to high-quality education for children in these areas. In contrast, some cold spots, composed of clusters of populations of higher educational attainment, have appeared outside the city limits and these areas often have their own independent school districts. These two observations
RESIDENTIAL SEGREGATION IN AUSTIN
201
suggest that urban form and the new geographic patterns are accentuating socioeconomic disparities. The fragmented structure of local government in the region means that, at the local level, there is no governmental venue capable of addressing these disparities. In terms of accessibility to fire services, within the city’s jurisdiction, the increasing segregation of the socioeconomically disadvantaged population produces better access to fire service provision than it does in rural areas. This phenomenon suggests a complex relationship between residential location decisions made by low-SES population and the quality of public services. Fire services may be somewhat unique in that the demand for such services will be higher in densely populated areas, thus resulting in greater accessibility of the service in urban than rural areas. The increasing segregation of lower SES population to the east and south of the city is likely affected by more affordable housing and lower taxes found there, topics not fully explored in this study. But among the affluent, distinct patterns are found. The small school districts in affluent areas tend to have high-performance schools. But some affluent areas outside of the city have low accessibility to fire services, suggesting that this service may not be a high priority or that its supply can be obtained by non-public-sector mechanisms. The highly fragmented structure of local government in the metropolitan area may be a factor playing an indirect role in the emerging patterns of segregation. The expanded opportunities for the more affluent to choose residential location may indirectly affect the quality of public services available to lower-economic-status individuals. The enhanced ability to examine urban form through spatial analysis provides tools for reexamining the relationship between segregation and urban services. We also see that the structure of local government and of service provision, as well as local revenue generation among local governments in a metropolitan area, will also need to be integrated into future analyses of the relationship. Notes 1. Data on poverty is obtained from census data and measured using “a set of money income thresholds that vary by family size and composition to determine who is in poverty. If a family’s total income is less than the family’s threshold, then that family and every individual in it is considered in poverty. The official poverty thresholds do not vary geographically, but they are updated for inflation using the Consumer Price Index. The official poverty
202
FLORES AND WILSON
definition uses money income before taxes and does not include capital gains or non-cash benefits (such as public housing, Medicaid, and food stamps).” (U.S. Census Bureau, 2008b). 2. Dark gray spots are areas where minorities concentrate, whereas lighter gray areas are the ones where the majority concentrates. 3. See Brown v. Board of Education, 347 U.S. 483 (1954). 4. The average corrected distances for all areas within and outside the city limits are higher than the average distances for hot and cold spots in 2000. This difference is explained by the larger than average index value for mixed areas; numbers not shown in the table.
References Brewer, J. M. (ed.) (1940). An historical outline of the Negro in Travis County. Austin, TX: Samuel Houston College. Fasenfest, D., Booza, J., and Metzger, K. (2004, April). Living together: A new look at racial and ethnic integration in metropolitan neighborhoods. Washington, DC: Center on Urban Metropolitan Policy, Brookings Institution. Glaeser, E. L. and Vigdor, J. L. (2003). Racial segregation: Promising news. In B. Katz and R. E. Long (eds.), Redefining urban and suburban America: Evidence from Census 2000, pp. 211–34. Washington, DC: Brookings Institution Press Center on Urban Metropolitan Policy. Hamilton, W. (1913). A social survey of Austin. Bulletin of the University of Texas, 273. Henneberger, J. J. and Huff, E. C. (1979). Housing patterns study. Segregation and discrimination in Austin, Texas. Austin, TX: Austin Human Relations Committee. Humphrey, D. C. and Crawford, W. W., Jr. (2001). Austin: An illustrated history. Sun Valley, CA: American Historical Press. Jackson, R. E. (1979). East Austin: A socio-historical view of a segregated community. M.A. Thesis, University of Texas at Austin. Jargowsky, P. A. (2003, May). Stunning progress, hidden problems: The dramatic decline of concentrated poverty in the 1990s. Washington, DC: Center on Urban Metropolitan Policy, The Brookings Institution. Logan, J. R. (2003). Ethnic diversity grows, neighborhood integration lags. In Bruce Katz and Robert E. Long (eds.), Redefining urban and suburban America: Evidence from Census 2000, pp. 235–56. Washington, DC: Brookings Institution Press. Manaster, J. (1986). The ethnic geography of Austin, Texas, 1875–1910. M.A. Thesis, University of Texas at Austin. Massey, D. S. and Denton, N. A. (1988). The dimensions of residential segregation. Social Forces, 67, 281–315. ——— (1993). American apartheid: Segregation and the making of the underclass. Cambridge, MA: Harvard University Press.
RESIDENTIAL SEGREGATION IN AUSTIN
203
Mishel, L., Bernstein, J., and Allegretto, S. (2005). The state of working America: 2004/2005. Ithaca, NY: Cornell University Press. U.S. Census Bureau (2008a). Available online at http://www.census.gov/main/ www/cen2000.html (accessed December 4, 2008). U.S. Census Bureau (2008b). Available online at http://www.census.gov/hhes/ www/poverty/povdef.html (accessed December 3, 2008).
This page intentionally left blank
11
Spatial Differentiation, Inequality, and Urban Policy: The Findings Bryan R. Roberts and Robert H. Wilson
A
s the twentieth century closed, the growth of the metropolitan areas in Latin America slowed and the irregular settlements on the periphery consolidated, even though globalization continued to subject labor markets to great stress. Research on cities, at least on Latin American cities, has not placed a priority on questions about urban spatial differentiation, especially residential segregation, as has been the case in North America. Rather, Latin American scholars have tended to focus on survival strategies of the urban poor through such perspectives as capabilities, vulnerabilities, and assets, whether in terms of finding low-cost housing or gaining access to informal labor markets. The concern over racial segregation among scholars of U.S. cities was not an issue in Latin American cities. This volume places the question of spatial differentiation in Latin American cities center stage. Taking advantage of spatially coded data and new tools of spatial statistical analysis, the studies presented here suggest that the study of spatial differentiation holds promise for scholars and policy analysts interested in the alleviation of urban poverty and inequality. These studies also speak to the role of the government in contributing to spatial differentiation and provide a range of approaches for assessing the effectiveness of urban service provision in highly differentiated urban spaces. We are now in a position to answer the research questions with which we began this study of socioeconomic segregation in seven major cities of Latin America, with Austin as a U.S. counterpoint. First, we asked how
206
ROBERTS AND WILSON
the spatial pattern of socioeconomic segregation had changed by the end of the twentieth century in the metropolitan areas we studied in terms of the level and pattern of segregation. Second, we were interested in what could be learned about social inequalities from the analysis of spatial segregation. Last, we wanted to know the extent to which patterns of social segregation are affected by and, in turn affect, urban policy, including the provision of urban services. Spatial Methods Before summarizing the findings reported in the city studies, some reflection on the methods used in the project, described in more detail in chapter 2, would be useful. We measure segregation by spatial indices that allow us to map different aspects of the segregation experience. Three indices were generally used—the index of dissimilarity, measuring the uneven distribution in urban space of a population with certain socioeconomic characteristics; the index of isolation, which calculates the likelihood of people living in close proximity to others with shared characteristics; and, the Moran index (Global and Local), which measures the degree of clustering of people with shared characteristics in different areas of the city. All these measures are affected by the sizes of the unit of spatial analysis, which in our cases range from census block, census tract, district, and municipality to metropolitan area. All other things being equal, the smaller the unit of analysis, the more likely it is that it will be segregated in terms of the uneven distribution of a population with certain socioeconomic characteristics (index of dissimilarity) or in terms of proximity to others with similar socioeconomic characteristics (indices of isolation and clustering). Another potential source of difference in measuring segregation is that individual socioeconomic characteristics may differ from each other in their impact on segregation. Thus, the São Paulo study reports, for example, that educational segregation is less marked in that city than is income segregation. The recognition of the differential impacts of individual characteristics allows for a more complex and nuanced understanding of the process of segregation, a process that varies significantly among the cities examined here. The challenge of comparing segregation between the cities in our studies is a considerable one because of the need to use measures of segregation, size of spatial units, and the socioeconomic characteristics that are comparable between metropolitan areas. It is also aggravated by differences in the range of socioeconomic microdata made available in city and national censuses and in the level of disaggregation at which these data are reported. To overcome these problems, we have tried to use similar
THE FINDINGS
207
size units of analysis, to employ at least two of the indices of segregation (dissimilarity, isolation, and clustering) in each study, and to include education as a common measure of socioeconomic segregation. A goal of this project has been to explore improved data sets available in many countries and new methods of analysis. While the three measures of segregation have been adopted in all studies, several authors applied additional methods of spatial analysis. The study of São Paulo developed a CHAID model; in Santiago, a geographically weighted regression technique was applied; in Mexico City, an index of centralization was employed; and multicategorical socioeconomic variables were used as measures in Lima. The results from these applications suggest that, methodologically, the time is ripe for spatial analysis. Changes in Spatial Patterns The question of spatial differentiation in Latin American cities did not become salient to scholars until the late twentieth century. The traditional forces structuring urban form, such as the rapid growth of metropolitan cities and great expansion of squatter settlements on their periphery in the 1960s and 1970s dissipated; the period of slow growth and consolidation of settlements that followed brought new elements to the urban development process. Most of the cities studied here have experienced the emergence of gated communities for the upper classes outside traditional enclaves. Another phenomenon affecting segregation in a number of cities is the deterioration in labor market conditions in recent decades. No dramatic changes have been reported concerning spatial patterns of segregation of the seven Latin American metropolitan areas in the 1990s, whether in terms of increasing or decreasing segregation (see table 11.1). Two partial exceptions are Santiago and Montevideo. Poverty levels in Santiago have decreased substantially in recent years and in chapter 7, Sabatini et al. calculate that there has been a decline in socioeconomic segregation as elites and, to a lesser extent, the poor disperse throughout the metropolitan area. In Santiago, however, the massive building of spatially concentrated public housing for the poor is a factor offsetting their dispersion throughout the metropolitan area. In Montevideo (chapter 6), the contrast between center and periphery has sharpened, as young, poor households move to irregular settlements on the periphery, and the city center both loses population and concentrates the elderly. Segregation by educational level has increased. In the past, the center housed many of the poor in tenements. Montevideo, compared with other Latin American cities, previously had relatively few irregular settlements; but as the economy weakened and unemployment
208
ROBERTS AND WILSON
Table 11.1 Segregation measures based on educational attainment of adults for eight cities Dissimilarity
Buenos Aires Lima Mexico City Montevideo Santiago Sao Paulo Austin Campinas
Isolation
Global Moran’s I
Year 1
Year 2
Year 1
Year 2
Year 1
Year 2
0.268 0.330 0.170 0.305 0.467 0.276 0.383 0.300
0.270 — 0.148 0.373 0.411 0.276 0.416 0.256
0.685 0.700 0.272 0.420 0.267 0.681 0.204 —
0.499 — 0.249 0.380 0.410 0.577 0.213 —
0.851 0.556 0.590 0.780 0.728 — 0.520 0.593
0.775 — 0.670 0.790 0.686 0.740 0.610 0.581
and informal employment grew, irregular settlements proliferated on the periphery. In the other Latin American cities, changes in segregation have been gradual. In Mexico City, Villarreal and Hamilton (chapter 5) report an increase in the segregation of the more highly educated groups, particularly the highest educated who are concentrated in the southwest of the city and of people with moderately low levels of education. The lowest educational group in Mexico City, as in Santiago, does not show increased segregation over time, probably reflecting, as Sabatini et al. suggest, their embeddedness in informal economic arrangements that depend on proximity to the wealthy. In São Paulo (chapter 8), segregation by education has decreased with time, but segregation has increased by income. In all the cities, educational levels have increased among the general population, diminishing the educational gap between those with less education and those with high levels of education. Consequently, segregation by education, except at the very highest levels, is likely to diminish. The case of Buenos Aires (chapter 3) is an interesting one in this respect. Groisman and Suárez show that spatial segregation in terms of educational levels diminished between 1990 and 2000, with educational levels rising throughout the city, even in the outer rings of the metropolitan area where poverty concentrates. But the socioeconomic index of health coverage for the population shows a rise in patterns of segregation in this period. Health coverage is associated with formal employment, which traditionally was high in Buenos Aires. However, in the 1990s, a process of job informalization occurred, which together with rising levels of unemployment, deprived a substantial number of households of social security coverage. Lack of coverage concentrates in the peripheral rings of the city, whereas the covered population is disproportionately
THE FINDINGS
209
represented in the center of the metropolitan area, in the city of Buenos Aires, and its affluent suburbs to the northeast. Within the metropolitan area, Groisman and Suárez report that the city experienced increasing polarization between its north and south as the number of luxury houses multiplied four times in the 1990s. Groisman and Suárez also note that between 1990 and 2001, the number of gated communities in Buenos Aires grew from 91 to 461. Even though data limitations prevent the calculation of segregation indicators for two points in time in Lima, there is little evidence of substantial change in segregation patterns. Peters notes some increase in the homogeneity of the poor areas on the eastern periphery along the mountains and an extension of high-income areas from the center to the coast on the east; but Lima remains perhaps the most socially heterogeneous of the seven cities, fragmented into socially diverse districts containing a range of housing types and with a poor transport infrastructure. The most recent of the metropolitan areas, Campinas, also has the least defined spatial structure (chapter 9). Whereas the other cities show a segregation pattern defined by the contrast between center and peripheral rings, Marcos and Jiménez make the contrast between corridors of affluence and corridors of poverty. In this regard, Campinas and Austin share common patterns. These second-tier cities have shown remarkable growth driven by decentralization of traditional manufacturing from the larger, industrial cities, but also resulting from development in advanced technology sectors supported by large, major universities. In both Austin and Campinas, the dividing line between affluence and low socioeconomic status is a major road. In recent years, this pattern of segregation has accentuated, but has not substantially changed. Some measures of segregation also declined in Austin, particularly for the African American and the poverty population. In contrast, the rapidly growing Hispanic population became more segregated as did the population of lower levels of education. Despite declines in segregation in some measures, Austin clearly maintains the U.S. pattern of sharp segregation based on ethnicity. The chapters make clear that two linked but separate processes produce these changes in spatial patterns of segregation. The most familiar one is the movement of population whereby people of similar socioeconomic characteristics occupy an area and make it compatible with their own lifestyles and culture. The area may be a newly developed one or it may be a previously settled one, which the newcomers make their own through what the Chicago urban ecologists labeled as invasion and succession. The second process is a change in the socioeconomic characteristics of the people settled in an area. Rising levels of unemployment
210
ROBERTS AND WILSON
and economic insecurity can, for example, transform skilled workers or middle-class employees into the “new poor.” We can see both processes at work in the cities analyzed in this volume. There is considerable intra-metropolitan population movement in the large Latin American cities. The central cities such as Buenos Aires, Mexico City, and Montevideo lose population, while the peripheral parts of the metropolitan area grow. Intra-metropolitan migration thus becomes a principal source of differential metropolitan growth, replacing rural-urban migration as a major source of city growth. In Montevideo, as Kaztman and Retamoso show, the growth of the city’s periphery is based on the outmigration from the center of young households seeking cheap accommodation (chapter 6). As important as these movements, however, is the movement of the middle and elite classes to the periphery in gated communities and other types of enclosed settlements. The development of gated communities is noted in all seven of the metropolitan areas. This change in the spatial pattern of segregation results from the greater opening of the Latin American urban economies to trade and foreign investment. Since the degree of foreign investment varies considerably between cities, with the highest in Buenos Aires, São Paulo, Mexico City, and Santiago, it is more marked in some cities than in others. In all the cities, as the chapters document, it has resulted in gated housing complexes for the middle classes and commercial developments, such as malls, that change the retail geography of the city. In Santiago, Campinas, Buenos Aires, Mexico City, and São Paulo, investment in transport infrastructure permits the gated communities to spread to the periphery of the city, which are often close to poor areas, thus diminishing the indices of segregation if not its reality. The new commercial developments contribute to the de-centralization of the middle classes, who can meet their shopping needs securely by car or, in some cases in Santiago and Buenos Aires, within the walls of a gated town. Despite the importance of the gated community phenomenon, it does not represent a major change in the pattern of socioeconomic segregation in the Latin American city. The residential developments reported in this volume are piecemeal and of relatively small scale, with sizes that rarely exceed a thousand people. They do not represent a substantial suburbanization of the middle classes. The chapters in the volume show that the second process of segregation is also important. In Buenos Aires, Santiago, and São Paulo, moderate changes in levels of segregation and general improvement in educational levels have occurred even though socioeconomic conditions of the population that remains deteriorate in a number of ways. Even where poverty
THE FINDINGS
211
has declined, as in Santiago, inequality and job insecurity has remained the same or increased. In Montevideo, Kaztman and Retamoso describe a situation in which formal workers lose their jobs and social security protection. They become impoverished and their previously socially heterogeneous neighborhoods become more homogeneously poor. In this respect, the contemporary Latin American city differs from its predecessor of the 1960s and 1970s. In the Latin American cities of the earlier period, spatial segregation was an active process resulting from the movements of both low- and high-income populations to settle the city, filling in available spaces. In the growing economies of the import-substituting industrialization (ISI) period, income opportunities were readily available, but housing was not, leading people to take what shelter they could find and adapt it to their needs. The contemporary city is now more consolidated and regulated with even squatter settlements housing second and third generation settlers who are often trapped in poverty and job insecurity. Accompanying the common trends in the pattern of spatial segregation, there are differences between the seven cities in their levels of segregation as measured by the Global Moran. Montevideo is currently the most segregated, followed in order by Buenos Aires, São Paulo, Mexico, Santiago, Campinas, and Lima. The explanation for these differences in levels of segregation is partly the maturity of a city’s urban and industrial development. Montevideo and Buenos Aires were the first Latin American cities to urbanize and industrialize and have a relatively consolidated and differentiated urban infrastructure. In contrast, Campinas is the last of these cities to develop and is still in-filling its urban infrastructure. Santiago, though an early urbanizer, is also the city that has been most radically economically restructured in recent years. Despite much higher levels of income inequality in Latin America, these Latin American cities show levels of socioeconomic segregation, as measured by the dissimilarity index, that are similar to or less than U.S. cities. Thus Austin, a mid-size U.S. metropolitan area, is more segregated ethnically than Mexico City in terms of poverty. These findings are surprising in light of the persistence of substantially higher levels of income inequality in Latin America than occur in the United States. Several factors account for this apparent paradox. One is the pattern of settlement described for the six established metropolitan areas (Buenos Aires, Lima, Mexico City, Montevideo, Santiago, and São Paulo) in which elites concentrate near the center and the poorer population on the periphery. In all six metropolitan areas, however, the density of the city center population and the small numbers of the elite compared to the general population means that the areas where the elite concentrate are relatively socially
212
ROBERTS AND WILSON
heterogeneous. Sabatini et al. point out that high-income districts in Latin America are possibly the most socially diverse in the city. Their study of Santiago argues that the densification of population in the city center through high-rise apartments for middle-class and lower-middleclass populations is a new additional factor that dilutes elite concentrations in these areas. The settlement of the poor in tenements and informal settlements close to elite housing continues into the new millennium as the poor seek to be near the job opportunities created by the wealthy and is particularly marked in the Brazilian cities. It occurs on a lesser scale in Santiago and Buenos Aires despite the new middle-class gated communities and shopping malls of the periphery. Nevertheless, lower levels of segregation were more common in the Conurbano of Buenos Aires than in the city itself. None of these Latin American cities have been either planned or zoned. Also, the low-income levels of the non-elite urban population have discouraged the development of a real estate market for tract housing for the middle or working classes on the scale that occurred in the United States. Consequently, housing is heterogeneous in size and quality in most urban neighborhoods located in the inner rings of the seven cities. In several of these cities, urban real estate developers provided only lots with minimal infrastructure, leaving it to the purchasers to erect or contract their own housing, as in the case of the settlement of the “rings” of Buenos Aires described by Groisman and Suárez. Poverty and unregulated settlements concentrate on the periphery in all seven of the Latin American cities, although in Campinas, which is the smallest and most recent to develop, such settlements tend to locate to the west of the city. The city studies show that the construction of public housing for lowincome populations has not been a major factor in the spatial segregation of the social classes. In the 1970s and 1980s, the small amounts of public housing that were constructed through government programs catered mainly to middle-income populations, as described for Mexico City by Villarreal and Hamilton. Indeed, it is only recently that public housing for the poor has become a factor in segregation, as Sabatini et al. show in Santiago.
What Can Be Learned about Social Inequalities from the Analysis of Spatial Segregation? Social inequality has become as important a source of policy concern as poverty in Latin America, where high levels of inequality have persisted even when poverty has diminished. Thus, Chile reduced poverty substantially in the 1990s, but maintained very high levels of inequality. Unlike
THE FINDINGS
213
high levels of poverty, high levels of social inequality are not synonymous with large-scale material or social deprivation; but the disproportionate concentration of wealth in a few hands can have negative effects for the welfare of the majority. Thus, despite remarkable improvement in poverty rates and education levels, the studies in this volume show some of those negative effects of social inequality as they are manifested in urban space. For most urban populations, the experience of social inequality is a local one, based on the quality and accessibility of services available in their neighborhood or district—health, education, sanitation, shopping. There are two main ways in which spatial segregation produces inequalities in the quality of the services provided. One is through the disadvantages that accumulate and reinforce each other in low-income neighborhoods. Homogeneity of poverty limits resources of the poor. In addition, the social context in these neighborhoods likely place more demands on schools, infrastructure, and other services because neighbors lack the resources to support them materially or socially. These neighborhoods are also likely to have a greater prevalence of crime and vandalism that takes an extra toll on neighborhood services. The political clout of such neighborhoods is also likely less than that of more affluent ones, diminishing the support that they receive from municipal governments. The second major way in which spatial segregation reinforces the negative effects of social inequality is through the limitations imposed by physical space on poorer populations. For those without resources, distance and travel is a greater limitation on their activities than they are for those with more resources. The poor are more dependent on local services than are the rich. And homogeneously poor neighborhoods are more likely to have low-quality services and lack external contacts and information than are socially heterogeneous neighborhoods. The wealthy can buy out of public education if a neighborhood public school is of low quality, whereas the poor need to send their children to a nearby public school. Likewise, the journey to work is more likely to be a constraint on resources for the poor than for the wealthy, who can afford the time and money to commute to work. The accounts in this volume provide a rich set of examples of how space and segregation interact with inequality. In three of the cities— Campinas, Lima, and Montevideo—the studies focus on the impact of socioeconomic segregation on the quality of education. In two cities, Lima and Campinas, data were obtained on the quality of public schools. In general, lower-quality schools were found in homogeneously poor areas. In Campinas, poor neighborhoods had lower educational achievement and fewer educational services in schools compared with socially heterogeneous neighborhoods. This was also the finding in Lima (chapter 4),
214
ROBERTS AND WILSON
though the socially heterogeneous city neighborhoods presented both the highest and lowest levels of school quality. Peters attributes this to a lower level of neighborhood social capital, either as a result of the absence of highly educated parents in some city-center neighborhoods or their “escape” into private education. In Lima, as in the other Latin American cities examined in this study, private schools have become the choice of the middle and upper classes. The authors report that this abandonment of the public sector results in public schools of very low quality even in socially heterogeneous central city neighborhoods because of the prevalence of private schools. The importance of public education for social integration is the major theme of Kaztman and Retamoso’s analysis of educational services in Montevideo. The pattern of segregation in Montevideo, particularly through the expansion of the periphery of the metropolitan area, has increasingly segmented education, with schools on the periphery being poorly equipped and serving homogeneously poor areas. Even in the center, public schools are less likely than in the past to cater to a socially diverse population. As in Lima, the growth in private education has become an important factor in segmentation. Thus, two out of five middleincome households in Montevideo and three out of four high-income households send their children to private schools. In Buenos Aires’s metropolitan area, the impact of segregation is explored through analyzing the uneven distribution and clustering of population with unfavorable labor market positions. The loss of health insurance coverage is concentrated spatially in the periphery, but also in the poorer areas of the city of Buenos Aires. A similar association is found between living in areas of deficient housing and being unemployed or employed in insecure jobs. Groisman and Suárez argue that this concentration of unfavorable neighborhood conditions is likely to have cumulative effects on their inhabitants’ welfare. The Buenos Aires metropolitan area is very large and the costs of travel to seek work or services are high. If most others have insecure jobs or are unemployed in the neighborhood, those without work will get little help from their neighbors either in terms of information or material support. Since free medical services are mainly located in the centers of municipalities in Buenos Aires, they are likely to be inaccessible to those isolated in homogeneously poor peripheral neighborhoods. In São Paulo, there has been a general improvement in neighborhood infrastructure and services, but Torres and Bichir (chapter 8) demonstrate that homogenously poor neighborhoods mainly located at the periphery are significantly worse off in terms of the quality of their infrastructural services and access to those services. They attribute this outcome to legal
THE FINDINGS
215
status in terms of squatter settlements and other illegal settlements. In the case of legal settlements, the authors attribute inadequate services in part to what they call a “discriminatory” dynamic on the part of the town hall and the utilities providers. They also show that factors such as local social capital as measured by membership of organizations and length of stay in the neighborhood have a favorable impact on access to services. In São Paulo, the low levels of social capital found in segregated areas of low-income households lead to lower levels of political participation and thereby reduce the effectiveness of this mechanism for achieving better services. The focus of the Mexico City study is on the spatial mismatch between where the poor population resides and formal employment opportunities in manufacturing, commerce, and the services. This mismatch is quite dramatic in the case of Mexico City since the poor populations are concentrated in the periphery of the city and employment opportunities are located closer to the center of the city. As in the case of Buenos Aires, travel distances in Mexico City are considerable and for populations located at the periphery entail high costs in time and money. For many households at the periphery, this situation confines them to employment in the informal sector, in insecure jobs, and without social security coverage. In Austin, segregation also results in spatial disparities in education outcomes and exacts costs in terms of accessibility to services. The residential segregation of adults with low educational achievement negatively affects student achievement in segregated neighborhoods. The spatial form of the city is affecting the spatial demand for education services and is creating disparities in outcomes, thereby affecting the efficacy of educational services themselves. In the examination of access to fire service, Flores and Wilson (chapter 10) show that the clustering of populations of similar demographic characteristics has differential impacts in different parts of the metropolitan area. Within the city, low-income areas actually have relatively good access to fire services as a result of higher densities, but similarly, areas outside the city have very poor access compared to other segments of the population. But, interestingly, some clusters of populations of higher educational attainment outside the city even have worse access, suggesting less reliance on this particular public service than lower income populations.
Urban Policymaking and the Provision of Urban Services The principal purpose of the case studies has been to characterize spatial segregation in the cities in the Americas. Many of the cases acknowledge the role of governmental action, such as land-use control and provision
216
ROBERTS AND WILSON
of physical infrastructure, in shaping these patterns. Other factors not included in these studies, such as land markets and topography, also shape cities. Even though these studies are not able, from a methodological perspective, to assess rigorously the role of government actions in shaping urban form, a number of observations can be made on the subject. Following these observations, implications for urban policy of segregation will be discussed.
Residential Segregation and Social Integration: A Challenge to Public Policy Housing patterns are a critical element of urban form. Given that not all activities of residents occur in the place of residence—as discussed elsewhere, a neighborhood should not be considered a bowl in which social activities occur—the accessibility and the associated costs of a particular site to others will affect the well being and opportunities for residents. The capability of people to move can overcome physical distance. The results in Buenos Aires, Mexico City, and Santiago indicate that contemporary patterns of residential segregation impose significant transportation burdens on residents of lower socioeconomic status in terms of access to employment. Areas of the cities with suitable employment opportunities are quite distant from the residences of this social group. As argued in the Buenos Aires study, dwellers in lower-status segregated areas are found to benefit less from economic growth than residents elsewhere. The problem is somewhat diminished in areas where the concentrations of lower-income individuals are not clustered. Although not all areas of mixed- or higher-socioeconomic status are necessarily the areas of greater employment opportunities, in general, areas of clustering of low-socioeconomic-status individuals are more likely to encounter greater transportation challenges to reach sites of employment. An interesting counter example is found in Santiago, where some gated communities have arisen in areas dominated by lower-status individuals. The location of these higher-income enclaves provides employment opportunities for traditional residents of the area. In general, however, increased clustering of low-socioeconomic-status populations in large metropolitan areas will likely create burdens on transportation systems, as well as impose transportation expenses on lower-socioeconomic-status individuals. Several of the papers discuss explicitly the mismatch hypothesis developed by Kain (1968) in the United States, where inner-city, often minority, residents are only qualified for occupations that are growing most rapidly in suburban areas. The clustering of higher-status individuals close to
THE FINDINGS
217
city centers and the poor on the periphery in most Latin American cities produces urban form quite different than it does in the United States. Despite the differences, both patterns of urban growth create significant employment mismatches. The effects of homogeneity in neighborhoods are further explored in the discussion of school performance and neighborhood composition below.
How Government Actions Affect Segregation In several of the case studies, instances of government action, or inaction, are identified as factors contributing to residential patterns. Certainly the role of governments in the development of cities through the provision of infrastructure and public housing is well studied, and the observations in the case studies are, therefore, anticipated. Although the analysis of the specific impacts of infrastructure investments and land-use controls on residential patterns are beyond the scope of this project, several observations are possible. The process of urbanization in Latin America has placed the poor on the periphery of metropolitan areas due, in part, to lower housing costs found there and occasionally as a result of governmental policy, through entirely ineffective provision of public housing and rent-control policy for inner-city housing in Mexico City. Infrastructure investments can make an area more attractive for housing and thus lead to higher housing costs. The studies in this book suggest that the more egregious examples of inadequate infrastructure—for example, lack of water, wastewater, and electricity—are rare. These metropolitan areas experienced periods of rapid growth decades ago and are now best characterized as mature cities and sizeable areas, and basic infrastructure is generally available. Nevertheless, some unresolved issues concerning public investments and services remain, especially the quality of urban services provided. Local governments, in the exercise of land use controls and public health regulation, often identify areas of cities unsuitable for residential construction. Prohibitions against residential construction in flood plains or restrictions on impervious ground cover may be ignored, and housing in such areas will likely not be provided full public services. In the case of São Paulo, residential occupations in violation of local land-use rules will almost certainly lead to inadequate provision of services due to prohibitions against extension of public services to these areas. Unless such prohibitions are accompanied by relocation strategies and affordable housing efforts, the plight of the poor will remain unresolved.
218
ROBERTS AND WILSON
We also find instances where the lumpiness of public investments can create inadequate services. In Buenos Aires, the lack of natural gas distribution systems in some parts of the city leaves residents without access to the service. But the São Paulo study cautions us on interpreting the reasons for inadequate public services for poor household areas. The authors argue that traditional measurements of segregation prove inadequate for analyzing the effects of poverty on public service coverage. A pocket of poverty in an area with generally higher levels of income will have better services than in an area with widely clustered areas of poverty. In the former, there are likely to be networks of infrastructure to which an individual household can link if incomes are sufficient. But once income levels are taken into account, time in residence is found to have a positive effect on access to services. However, in areas where low-income people are clustered, as in São Paulo’s periphery, infrastructure networks may not exist. This was a conclusion drawn in Buenos Aires concerning access to natural gas systems. Spatial clustering tools and other tools, such as CHAID, illustrated in the São Paulo study, can help diagnose the sources of inadequate service provision. Another element of public services provision highlighted in several case studies is governmental fragmentation in metropolitan areas. In the Latin American cities, local governments on the periphery generally have weaker governmental capacity than do the central municipalities. Poor governmental performance can lead to inadequate services and thus affect decisions regarding where individuals will locate. To the extent that local municipal services are affected by local revenue collection, municipalities with large shares of low-income population will likely be constrained to low-expenditure levels for public services. Although this study did not conduct a formal assessment of local government capacity, including the fiscal capacity, in most of these metropolitan areas, availability and quality of services decline as one moves away from the center. Thus, fragmented governmental structures and residential segregation patterns in metropolitan areas are interrelated. In several studies, the presence of gated communities is noted. In Santiago, gated communities in areas previously dominated by low-status households are actually leading to a decline in segregation. Gated communities can substitute private provision of some services, such as security, for their public provision. As another sign of public sector fragmentation, higher-income families can, in effect, opt out of public services. The case of education is discussed further below. The public sector shapes urban form and land markets. The quality of infrastructure and services becomes incorporated in the price of land. In turn, land markets sort the population among neighborhoods. In most of
THE FINDINGS
219
the metropolitan areas examined in this study, we find that lower quality and availability of infrastructure, lower socioeconomic status, and poor labor market accessibility are mutually reinforcing. This dynamic can be illustrated in education services.
Education Attainment, Schools, and Space In an earlier section of this chapter, the patterns of segregation according to educational attainment found in metropolitan areas were reported. Segregation by educational level is relatively high in most metropolitan areas, but there is no consistent pattern of change because of the general rise in educational levels among metropolitan populations that offsets the clustering of the highly educated. The relationship between these patterns and schools is a complicated one, but one in which mutual determination is undoubtedly at work. The Montevideo case examines a more general impact of segregation, the loss of heterogeneous social interaction. It argues that segregation diminishes the opportunities for the interaction between different socioeconomic groups. Even though the empirical results suggest that isolation is declining in many cities, in Montevideo expansion of clustered segregation of lower-socioeconomic-status individuals means that, at least in those neighborhoods on the periphery, individuals of different classes are less likely to interact. The authors argue that this impedes social integration, particularly among lower-status individuals and families. This means that the clustering of low-income populations impedes the effectiveness of educational policies even when targeted to improve the education of low-income groups. Thus, Kaztman and Retamoso report on government attempts to remedy the inequalities in public education in Montevideo through special programs targeted to poor neighborhoods; but they also point out that these programs are limited by the depressant effects on children’s education of homogeneously poor neighborhoods, where parents lack the time and resources to help their children. The role of social integration that could be served by public education is lost. Although the studies in this volume do not explore the specific neighborhood effect on educational outcomes, this should be a priority for further research (Flores, 2008). There is considerable evidence that the quality of schools varies in space, as in Lima, Montevideo, Campinas, and Austin. In particular, school outcomes are highly correlated with socioeconomic status and segregation. Fragmentation in local school systems can accentuate these disparities. In affluent areas, private schools are more commonly used
220
ROBERTS AND WILSON
in Montevideo and Lima, another example of households opting out of public services. In Austin, affluent areas often have their own small public school districts and produce outcomes much higher than in the larger inner-city school districts. Thus, the socioeconomic status of an area can affect the composition of the schools as well as outcomes. Relocating students to schools outside the neighborhood of residence can be accomplished through busing, as was used in the United States in the 1950s and 1960s to overcome racial segregation, or with portable vouchers, as seen in Santiago. Those that consider school outcomes from a relative deprivation perspective suggest that interaction in schools for children from significantly different socioeconomic groups can actually have a deleterious effect on the lower-status children since the “veil of ignorance” concerning social standing is lifted (Flores, 2008, pp. 32–33). The authors of the Montevideo study offer an opposing view, based in part on the deleterious consequences of the loss of social interaction among heterogeneous groups. Even if one acknowledges the need for additional research on an appropriate balance of social heterogeneity in schools, in segregated areas socialization and neighborhood effects likely place lower-socioeconomic students at risk and reduce educational achievement. Given the common pattern of segregation by educational attainment found in most of the metropolitan areas studied here, concern about the opportunities for children in lower-status families in segregated areas is warranted. In a new analysis by an author contributing to this volume, educational outcomes were subject to neighborhood effects (Flores, 2008). One of the original propositions of this study states that public sector action affects urban form, and urban form, particularly segregation, creates unique demands for public service. Several interpretations of the research findings support the former, but the research design adopted here does not allow a rigorous testing. On the latter, the evidence is much clearer. The disparities in the spatial demand for services, driven in no small part by the residential location opportunities available for lowersocial-status families, make the efficient delivery of services more difficult. Leaving aside the implications for social justice of segregated cities, the disparities likely increase the cost of providing services. For example, with less segregation, the cost of an education system of good quality might well be reduced. Although the public sector has made strides in securing basic infrastructure for metropolitan dwellers and some very significant improvements in terms of educational attainment and reduction of poverty can be observed, the trends observed in the cities in this study suggest that the forces behind segregation and the challenge that segregation brings to the public sector will not soon be abated.
THE FINDINGS
221
Closing Thoughts Significant spatial differentiation exists in the Latin American metropolitan areas. But on a variety of measures, the poor populations in these cities have far better socioeconomic characteristics as compared to lowerincome groups during periods of rapid urbanizations in the 1960s and 1970s. One observes that the settlements have consolidated and minimal levels of infrastructure are almost universal. But, spatial differentiation reinforces social inequalities both in terms of outcomes of prevailing social processes and the provision of public services. The availability of spatially coded data and new methods for spatial analysis suggest that this will be an important area for further investigation. The studies also indicate the usefulness of comparative analysis, even though common patterns of changes in urban form among these cities are rare. Local context and specificity shape outcomes in individual cities, but broad clustering of lower-income families found in most cities appears to constrain opportunities in all. Furthermore, the assessment of the quality of urban services, especially education, or access to services, will benefit from a more explicit consideration of space. In particular, the governmental fragmentation observed in large Latin American cities makes addressing spatial disparities in urban services especially challenging. But the spatial analyses presented in this volume demonstrate analytical purchase in identifying the causes of differentiation in demand for and quality of services and thus provide the basis for improving governmental performance. The very large metropolitan areas examined here have been growing slowly, especially in contrast to the rapidly growing second-tier cities. One should expect the evolution of urban form in these second-tier cities to be significantly different than what we find in the large consolidation metropolitan areas. The results from the Campinas study and its contrast to Austin in the United States suggest further lines of inquiry. The process of globalization and its effects on the urban form of the large metropolitan areas seems to be playing itself out differently in the second-tier cities. The structuring elements of ISI in the larger cities are not present in the second-tier cities. We initiated this project expecting to discover that spatial differentiation in cities in Latin America was indeed being affected, if not reshaped, by globalization and the expansion of market economies. Despite a general pattern of moderate decline in some measures of spatial segregation, new issues of spatial inequality are observed. The geography of opportunity in these cities is changing, but the ability of lower-income groups sorted to large neighborhoods occupied by individuals and families of
222
ROBERTS AND WILSON
similar circumstances to realize the benefits of this geography reaffirms that place does matter. References Kain, J. (1968). Residential segregation, Negro employment, and metropolitan decentralization. Quarterly Journal of Economics, 82, 175–97. Flores, C. (2008). Residential segregation and the geography of opportunities: A spatial analysis of heterogeneity and spillovers in education. Ph.D. Dissertation, LBJ School of Public Affairs, University of Texas at Austin.
Notes on the Contributors
Renata Mirandola Bichir is a Researcher at the Center for Metropolitan Studies, (CEBRAP), in São Paulo, Brazil, where she is currently working on segregation and urban poverty. Bichir has published numerous articles and papers on issues of urban spaces, poverty, and segregation. Recent scholarship includes neighborhood effects on youth in São Paulo; this research received a FURS/Brasil Blackwell award. Bichir is finishing her Master’s in Political Sciences at the University of São Paulo. Carolina Flores is an Assistant Professor at the Department of Sociology, Pontificia Universidad Católica de Chile (PUCC) in Santiago, Chile. Flores is also a doctoral candidate at the Lyndon B. Johnson School of Public Affairs at the University of Texas at Austin. She holds a Master’s in Economics from the PUCC and a Master’s in Public Policy from the London School of Economics. A Sociologist by training, her research activities focus on the effects of spatial concentration of urban poverty on several individual outcomes, such as education and integration to the labor market. Her teaching interests lie in the areas of quantitative research methods, spatial analysis, social policies, and sociology of education. Fernando Groisman is a Researcher at the National Centre for Scientific and Technical Research (CONICET) and at University of General Sarmiento (UNGS) in Argentina. His research focuses on labor market dynamics, inequality, social exclusion, and residential segregation. He has published numerous papers on these issues. Currently, Groisman is the President of the Argentinean Association of Specialists in Labor Studies. He holds a Ph.D. in Social Sciences from the Latin American Faculty of Social Sciences (FLACSO) and a Master’s degree in Labor Studies from University of Buenos Aires (UBA). Erin R. Hamilton is a doctoral student in the Department of Sociology and a trainee at the Population Research Center at the University of Texas at Austin. She studies the social demography of migration with a focus on
224
CONTRIBUTORS
Mexico. In addition to her work on segregation in Mexico City, she studies the social, economic, and geographic origins of recent Mexican migration to the United States and the health consequences of migration. Maren Andrea Jiménez is a doctoral candidate in the Department of Sociology at the University of Texas at Austin, where she also received her Master’s degree in Sociology. Her research interests include the processes of segregation and its consequences in Brazilian cities and the role that family characteristics and household composition play in affecting the health and well being of children and the elderly. Ms. Jiménez is currently employed as an Associate Population Affairs Officer in the Population Division (CELADE) of the UN Economic Commission for Latin America and the Caribbean. Ruben Kaztman is the Director of the Research Program on Social Integration, Poverty, and Exclusion (IPES) at the Catholic University of Uruguay. He is the author and co-author of numerous books published in Chile, Mexico, and Uruguay, and has also published in several international journals and magazines. Kaztman holds a Master’s in Sociology from the University of California at Berkeley. He is also the Coordinator of the Group on Urban Segregation Studies (GESU). Paul A. Peters is a Technical Specialist at Statistics Canada in the Health Information and Research Division. He is also a doctoral candidate in the Population Research Center, Department of Sociology at the University of Texas at Austin. He holds a Master’s in Environmental Studies from the School of Urban Planning at the University of Waterloo, Canada. His research focuses on the use of spatial tools to analyze the role of planning and governance in shaping urban social and physical change. His most recent work involves the intersections between complex systems and spatial inequality in Latin American megacities. José Marcos Pinto da Cuhna a Professor of Demography in the Department of Demography at the Institute of Philosophy and Social Sciences and a Senior Research Associate in the Population Research Center (Núcleo de Estudos de População, NEPO), University of Campinas, in Brazil. Professor Pinto da Cuhna’s main research area is the link between immigration and urbanization, a topic in which he has written extensively regarding the internal migration and metropolitization process in Brazil. Alejandro Retamoso is an Associate Researcher at the Research Program on Social Integration, Poverty, and Exclusion (IPES) at the Catholic University of Uruguay. Retamoso pursued graduate studies in Demography at the Population Division (CELADE) of the UN Economic
CONTRIBUTORS
225
Commission for Latin America and the Caribbean and Universidad de Chile. His most recent research focuses on poverty, social exclusion, and education. Author of numerous articles in specialized magazines, journals, and books, he is also a Consultant for the United Nations Children’s Fund (UNICEF)—Montevideo and the Inter-American Drug Abuse Commission Control of the Organization of American States. Bryan R. Roberts is a Professor at the Department of Sociology and the current director of the Teresa Lozano Long Institute of Latin American Studies at the University of Texas at Austin. Roberts has published numerous articles and papers on issues of citizenship, social policy, irregular settlements, and internal and international migration including MexicoU.S. labor markets, informal economies, community development, and urbanization. His latest publications include Rethinking Development in Latin America and Ciudades latinoamericanas: Un análisis comparativo en el umbral del nuevo siglo. Dr. Roberts received a Ph.D. in Sociology from the University of Chicago. Francisco Sabatini is a Professor at the Catholic University of Chile in Santiago, Chile, where he lectures and conducts research on residential segregation, environmental conflicts, and planning and land policies. Sabatini has served as an Advisor to the Chilean Minister of Housing and Urban Affairs since the return to democracy in 1990. Sabatini combines his academic work with involvement in NGO-based research and action projects in low-income neighborhoods and villages. Sabatini is a Sociologist holding a Ph.D. in Urban Planning who has published extensively and taught in several countries, mainly in Latin America. Carlos Sierralta is a Geographer and holds a Master’s in Urban Development from the Pontificia Universidad Católica de Chile, where he is currently working as a Teacher and Research Assistant. His main interests are urban sociology and residential segregation. Ana Lourdes Suárez holds a Ph.D. in Sociology from the University of California, San Diego and has undertaken doctoral studies in Anthropology from the University of Buenos Aires. Suárez is an Academic Associate Researcher at UNDP Argentina and a Researcher at the National Center for Scientific and Technical Research (CONICET) of Argentina. She teaches at the Catholic University in Buenos Aires. Her most recent research includes residential segregation, urban poverty, and labor markets. Haroldo da Gama Torres is an Economist and a Demographer specializing on the spatial distribution of population. Da Gama’s research activities
226
CONTRIBUTORS
have focused primarily on the application of Geographic Information Systems (GIS) to the study of social policy. He has also worked in the development of different GIS-related projects in the areas of urban planning, education, environment, and small businesses in Latin America. His most recent research includes the connections between social policies and urban inequalities. Currently, da Gama is the Coordinator on Knowledge Transfer of the Center for Metropolitan Studies (CEM-CEBRAP), in São Paulo, Brazil. Andrés Villarreal is an Associate Professor in the Department of Sociology and Faculty Research Associate in the Population Research Center at the University of Texas at Austin. He is currently engaged in a research project funded by the National Institute of Child Health and Human Development examining the effects of economic liberalization on social inequality in Mexico. He also conducts research on crime and violence in Latin America. His work has appeared in the American Journal of Sociology, the American Sociological Review, and Social Forces, among others. Robert H. Wilson is Associate Dean of Academic Affairs and Research and Mike Hogg Professor of Urban Policy, LBJ School of Public Affairs, at the University of Texas at Austin. Wilson has directed the Brazil Center, the Center for Latin American Social Policy in the Institute of Latin American Studies, and the University of Texas at Austin’s Urban Issues Program. His publications include Governance in the Americas, Public Policy and Community: Activism and Governance in Texas, States and the Economy: Policymaking and Decentralization, The Political Economy of Brazil, along with six other books and numerous articles, chapters, and special reports. Professor Wilson received a Ph.D. in City and Regional Planning from the University of Pennsylvania. Guillermo Wormald is a Professor at the Institute of Sociology at the Pontificia Universidad Católica de Chile. Wormald’s main research areas include economic sociology, poverty, and social exclusion. Among his most recent publications is “Santiago de Chile under the New Economy,” in Ciudades latinoamericanas. Un análisis comparativo en el umbral del nuevo siglo, edited by Alejandro Portes, Bryan Roberts, and Alejandro Grimson. Wormald has also published, with Florencia Torche, “Stratification and Social Mobility in Chile during the Last Decades of the XX Century” (forthcoming), Estratificación y movilidad social en América Latina ¿Es posible avanzar hacia una mayor igualdad?, edited by Rolando Franco, Arturo León, and Raúl Atria. Wormald holds a Ph.D. from the University of Sussex, UK.
Index
aerial units, see census data agrarian reform, in Peru, 57 Barrio Alto, Santiago, 127 Brown. v. Board of Education, 202 Burgess, Ernest, 4 theory of concentric zones (1925), 4 see also concentric zone theory Censo Nacional de Poblacion y Vivienda (CNPV), 42 census data, 21, 40, 43, 206–7 in Austin, 189 in Brazil, 147–8 in Buenos Aires, 40, 42 defined, 27 and GIS, 3 in Lima, 62 in Mexico, 73, 77 in Santiago, 128 in São Paulo, 147 in U.S., 201–2 chi-square automatic detector (CHAID) model, 154 Chicago, urban ecological theory of invasion and succession, 209–10 cities, 6–7, 15, 41 “clubes de chacra,” 41 “clustering,” see segregation, measures of commensalism, 5 concentric zone theory, 4, 212 contornos (Mexico), 76 Conurbation (Conurbano Bonaerense), 40, 41, 45
decentralization, 2, 10–12 demographics, 8 trends in Austin, 192–3 dependence, spatial, 29 deregulation, 2 discrimination, 8 dissimilarity index, see segregation, measures of diversity, ethnic and racial in contemporary U.S., 9 economy, 7, 11 education, 13, 214, 219–20 access to, 125 in Austin, 192 in Campinas, 177 factors affecting school performance, 114–15 and government instability, in Lima, 57, 59 levels of attainment: in Buenos Aires, 48; in Montevideo, 107 in Lima, 59, 62, 67 and “magnet schools,” 198 in Mexico City, 79 in Montevideo, 102–3, 108–11 and spatial segregation, 13, 56 in U.S., 198 ejidos, 75–6 “embedded poverty,” 27 employment informal, in Santiago, 124–5 unemployment, 208–9 Encuesta Continua de Hogares - ECH (Uruguay), 99
228
INDEX
Encuesta Nacional de Hogares ENAHO (Peru), 62 “evenness,” see segregation, measures of favelas, 170, 171 “first law of geography” (Tobler), 29 free trade areas, 6 Fujimori, Alberto (1990–2000), 58 gated communities, see housing gentrification, 9 in Santiago, 131 geographical information systems (GIS), 2 and spatially coded data, 205 Geographically Weighted Regression technique (GWR), 137 ghettoization, 127, 137 “global cities” research, 121 globalization, 121, 221 government delegaciones (Mexico), 78 departamentos (Argentina), 43 fragmentation, 221 influence of, in Latin American, 9 instability, in Santiago, 123 and market failure, 123 partidos (Buenos Aires), 40 regulation, 9–13 and service provision, 13, 217–18 “grid problem,” 129 health insurance coverage, 45 as indicator of segregation in Buenos Aires, 45 see also segregation, measures of housing, 4 affordable, in Mexico City, 75 “clubes de chacra,” 41 decrease of working and middle class, in Buenos Aires, 41 gated communities, 39, 207, 210, 216; in Buenos Aires, 41, 48; in Campinas, 170, 171, 177; in Santiago, 130 and growth patterns, 13
home financing, 7 informal (or irregular) settlements, 3, 205–7, 212; in Lima, 58; in Mexico City, 74; in São Paulo, 146; in Uruguay, 97, 99–100, 105, 107 Levittown, N.Y., 5 nature preserves, 171 patterns, in Austin, 190–1 policy, analysis, 9; in Mexico, 73 production of, 5 quality in Campinas, 172 self-construction in Latin America self-help, 74 shantytowns, 145 squatter settlements, 3, 211 villas miseras (in Buenos Aires), 41 working-class, 4 see also ghettoization immigration, 5, 47, 210 in Buenos Aires, 45, 47 in Montevideo, 98 in U.S., 5 import-substituting industrialization (ISI), 1, 211 in Lima, 57 in Santiago, 123 in Uruguay, 97 income distribution in Latin America, 7 as measure of segregation, in Mexico, 79 inequality and spatial form, 15 infrastructure, 14 provision, 13; in Buenos Aires, 41; as measure of poverty in Mexico, 78; in São Paulo, 145 regulation of, 13 spatial dimension of, 13 Instituto Brasileiro de Geografia e Estatistica (IBGE), 173 irregular settlements, see housing isolation index, 24
INDEX
labor and residence mismatches in Buenos Aires, 49 land markets, 7 and government intervention, 9 in Mexico City, 75 in Peru, 58 land occupations, in Montevideo, 100 land use and infrastructure provision in Buenos Aires, 41 regulation of, 13, 41 Levittown, N.Y., 5 marginality, 146 market failure and inequality, 123 megacities, 55 mega-regions, 7 Metropolitan Region of Buenos Aires (MRBA), 40 Mexican Revolution (1910), 191 Ministerio de Educación del Perú, 59, 62 Ministry of Social Development SEDESOL (Mexico), 77 “mirror thesis” of social and spatial structures, 122 modifiable area unit problem (MAUP), 173 National Institute of Statistics (INE), 40, 99 National Institute of Statistics, Geography, and Informatics (INEGI), 77 National Population Council CONAPO (Mexico), 77 Necesidades Basicas Insatisfechas (NBIs), 43, 45 neighborhoods, 7–9 characteristics of, in Montevideo, 99 homogeneity of, 195 and schools, 107 neoliberalism, 123
229
neoliberal reform, in Santiago, 130 “new economic model,” see neoliberalism nongovernmental organizations (NGOs), 11–14 Oportunidades (Mexico), 2 population growth: in Austin, 191–2; in Campinas, 169; of Hispanic in U.S., 188; in Mexico City, 74; in Montevideo, 98 poverty, 8 in Austin, 192, 193–4 in Campinas, 177 and education, 194–5 education as indictor of, 174–6 “embedded poverty,” 27 exclusion of the poor, 126 and the female labor force, 125 indices of, in Campinas, 174 in Lima, 59, 61 measures of, 78, see also segregation, measures of in metro areas, 8 in Mexico City, 74, 77, 87 and minorities, in U.S., 9 in Santiago, 123 in São Paulo, 148–50 spatial distribution of: in Mexico City, 73; in Montevideo, 100 survival strategies of urban poor, 205 in U.S., 188–9 public policy spatial dimension of, 12 and urban form, 12 real estate, 7 in Santiago, 130 segregation, 22 in Chicago, 5, 209–10 “cold spots,” 200
230
INDEX
segregation—Continued constitutional and statutory frameworks, 11 economic factors affecting, 210–11 and education, 208–9; in Buenos Aires, 41, 45, 47; in Campinas, 180–2; in Lima, 56; in Mexico, 79 and employment, in Mexico, 73 and ethnic groups, 5 and ethnicity, 39; in Latin America, 8; in U.S., 8 evolution of, in Montevideo, 103 indicators of, 42, 79; in Montevideo, 102–3 and gated communities, 39 generational, in Montevideo, 101 global, 3 “hot spots,” 196 and income: in Mexico City, 79–80; in São Paulo, 149 in Lima, 61, 63 local, 32 measures of, 24–8, 76, 80, 134–6, 189, 195–6, 206–7; in Austin, 194–5; “centralization,” 76, 85–7; “clustering,”76, 84, 92, 128, 196, 200; dissimilarity index, 42, 87, 103, 147, 170; error, 27; “evenness,” 22–4, 76, 81–3, 127, 195; “exposure,” 24, 83–4, 103; health insurance coverage, 49; index, 22; isolation, 127, 147, 195–6; Local Moran’s I, 27, 103, 147, 150, 170, 196; residential ownership, 49; scale, 173 mitigating impacts of, in Montevideo, 112–15 and neoliberalism, 39 patterns of, 207–9, 211–12; in Mexico City, 92; in Santiago, 122, 126 and poverty, in Mexico City, 87 and public policy, 33, 216 and race, 39 and residential patterns, 3, 4, 6, 39; defined, 21–2, 39; in Lima, 69; in Santiago, 129
and service provision, 198 social effects of, 136–7 socioeconomic, 2, 22, 40, 172, 211– 12; in Austin, 193; measurements of, 23 sociospatial analysis of, 2 spatial, 2, 12; analysis of, 21; and census data, 21; and education levels, 31; in Lima, 60; measures of, 15; patterns, 147; and social inequality, 15; variables of, 31 spatial units, see aerial units; census data and teenage fertility, 137 and transportation, 12–13 in U.S., 5, 187 and U.S. Supreme Court, 198 variable affect of, 6 service provision, 12–15, 214–15 in Austin, 198–200 and education, 156 emergency services, 199–201 in illegal settlements, 156–8 and public policy, 215 in São Paulo, 152–6 settlements, see housing slavery in Austin, 190 Social and Emergency Investment Funds’ organizations, 11 social exclusion, 43 socioeconomic trends, in Santiago, 122–6 Spatial Differentiation and Governance in the Americas (2005), 16 spatial isolation, 137 “spatial mismatch hypothesis,” 73, 89, 215 spillover effects, 29 sprawl, see suburbanization statistical analysis, spatial, 205 suburbanization, 40 in Buenos Aires, 41 contornos (Mexico), 76 and development of rail, 41
INDEX
in Montevideo, 99–100 in Santiago, 130 sprawl, 130 in the U.S., 7, 187 “white flight,” 198 suburbs, 5 “white only,” 5 Suttles, Gerald, 5 symbiosis, 5 System of Evaluation of Academic Achievement in the State of São Paulo (SARESP), 180 time-series analysis, 30 Tobler, Waldo, see “first law of geography” topography, as factor in segregation, 6 transportation, 3–7, 12 unemployment, see employment urban form density, 129 development patterns, in Lima, 61 fragmentation of, in Lima, 59 and service provision, 15 and social inequality, 15
231
theories of, 3–4 and transportation, 41 urbanization in Austin, 190 in Buenos Aires, 40 in Campinas, 170 and centralization, 16 and decentralization, 16 and governance, in Lima, 42 and immigration, in Lima, 57 and infrastructure provision, in Lima, 57 and land value, 4 periods of in Latin America countries, 16 and public policy, 12 villas miseras, 41 welfare policy in Austin, 195 Workshop on Urban Governance and Intra-urban Population Differentials in Latin American Metropolitan Areas (2002), 16 zoning, 9, 212–15