97929 Poverty and Shared Prosperity in Brazil’s Metropolitan Regions: Taking Stock and Identifying Priorities “The catalogue of forms is endless: until every shape has found its city, new cities will continue to be born. When the forms exhaust their variety and come apart, the end of cities begins.” Italo Calvino, Invisible Cities (1972) Acknowledgements This report was prepared by a team led by Aude-Sophie Rodella (Economist, GP Poverty) under the guid- ance of Louise Cord (Sector Manager, LCSPE), Emmanuel Skoufias (Lead Economist, GP Poverty), Magnus Lindelow (Sector Leader, Brazil), and Roland Clarke (Sector Leader, Brazil). The team included: Ali Sharman (JPA, GP Poverty), Renan Pieri (Consultant, GP Poverty), Martha Viveros (Consultant, GP Poverty), Ana Luiza Machado (Consultant, GP Poverty), Adam Ratzlaff (Consultant, GP Pov- erty), Yevgeniya Svachenko (Consultant, SARCE) and John Burgess (Consultant, GP Poverty). The team substantially benefitted from discussions with Luis-Felipe Lopez-Calva (Lead Economist, GP Pov- erty), Philippe Leite (Senior Economist, GP Social Protection), Anna Fruttero (Senior Economist, GP Pover- ty), Shomik Mehndiratta (Lead Urban Transport Specialist), Steven Farji Weiss (Consultant, GP Transport and ICT) and Alessandra Campanaro (Senior Economist, GP Urban, Rural and Social Development). Over the course of this work the peer reviewers were: Joao Pedro Azevedo (Senior Economist, GP Poverty), Anna Fruttero (Senior Economist, GP Poverty), Philippe Leite (Senior Economic, GP Social Protection and La- bor), Claudia Baddini (Senior Social Protection Specialist, GP Social Protection and Labor), Somik Lall (Lead urban Economist, GP Urban, Rural and Social Development), Pete Lanjouw (Research Manager, DECPI). Contents Executive Summary 8 Introduction 11 I. Economic growth and redistribution have generated significant poverty reduction in Brazil’s RMs 14 A. GDP has consistently grown in RMs since 2004 but slightly slower than at the national level 14 B. Economic growth has most benefitted the poor, resulting in declining inequality across and within RMs 15 C. Monetary poverty has decreased considerably, especially in the North and Northeast, but vulnerability remains a challenge 16 D. Upward economic mobility has been strong in Brazil’s RMs 17 E. The poor and vulnerable face specific challenges to their upward mobility 19 II. Labor markets, demographic changes, and transfers have been key to poverty and inequality reduction 21 A. Labor income drove poverty and inequality reduction in Brazil’s RMs 21 B. Improved labor market outcomes and policies increased labor incomes 25 C. Gender differences persist in labor outcomes across RMs 25 III. Increased access to a range of services has almost eliminated multi-dimensional poverty 28 A. Access to basic goods and services is approaching universal coverage in RMs 28 B. Equitable access to quality education remains a problem 29 IV. Spatial inequities are pronounced within RMs – heightened by constrained mobility 33 A. Core municipalities have lower levels of poverty and higher access to services than peripheries but are also more unequal 33 B. The effect of urban mobility on these spatial differences warrants further analysis 34 Conclusion 38 References 40 Annexes 43 Annex 1. What is a Metropolitan Region? 44 Annex 2. Economic growth in Brazil’s RMs and Brazil as a whole 46 Annex 3. Shared prosperity in Brazil’s RMs and Brazil as a whole 48 Annex 4. The Gini Index in Brazil’s RMs and Brazil as a whole 50 Annex 5. Monetary poverty in Brazil’s RMs with and without adjusting for cost of living 52 Annex 6. Poverty lines in Brazil 56 Annex 7. Intra-generational mobility through synthetic panels in Brazil’s RMs 57 Annex 8. General aspects by income group across Brazil’s RMs 58 Annex 9. Multi-dimensional poverty in Brazil’s RMs and Brazil as a whole 60 Annex 10. The Human Opportunities Index in Brazil’s RMs 61 Annex 11. Differences between cores and inner peripheries in the RMs of Brazil’s Northeast and Southeast, 2010 63 Annex 12. Various indicators in Brazil’s Northeast and Southeast RMs 64 Annex 13. Evolution of commuting time in Brazil’s RMs 69 Annex 14. Oaxaca-Blinder Recentered Regression (RIF) for core/inner periphery and core/outer periphery of the RMs of Brazil’s North/Northeast and Southeast, 2010 70 Annex 15. Labor regressions by year and income group for working age adults living in RMs of Brazil 71 Annex 16. Measuring shared prosperity at the sub-national level 73 Annex 17. From Favelas to “Areas of Special Social Interest” (AEIS) 74 Boxes Box 1. States, municipalities and metropolitan regions in federal Brazil 13 Box 2. Poverty and Vulnerability through an international measurement lens 18 Box 3. Social Programs in Brazil’s RMs 23 Box 4. Tax incentives and business registration simplification for reducing informality 24 Box 5. Minimum wages in Brazil and its RMs 26 Box 6. Measuring non-income-based levels of social welfare in metropolitan Brazil 30 Box 7. Quality of education indicators have improved in RMs 31 Box 8. Crime in metropolitan Brazil 35 Box 9. Towards more inclusive urban transport: The RMs São Paulo and Rio 36 Figures Figure 1. Inside and outside the metropolitan region 12 Figure 2. In recent years real GDP growth in the RMs has been slowing down relative to Brazil as a whole. 15 Figure 3. GDP growth grew the most in the traditionally poorer Fortaleza and Recife 15 Figure 4. Income growth from 2004 to 2012 in RMs was highest for the poorest deciles 15 Figure 5. Inequality remains higher in RMs than in urban and rural areas 15 Figure 6. Inequality has fallen in all RMs but remains high and heterogeneous across RMs 16 Figure 7. Growth reduced poverty more than redistribution, on average, in Brazil’s RMs. 16 Figure 8. Most people living in poverty in Brazil are located in non-metropolitan urban areas 17 Figure 9. Vulnerability levels have declined but remain significant 17 Figure 10. National, regional and international lines show a consistent reduction in poverty and vulnerability but levels vary substantially across thresholds 18 Figure 11. Average annualized real growth rate of income is high for the national bottom 40 living in RMs 19 Figure 12. The middle class makes up almost half bottom 40 living in Brazil’s RMs 19 Figure 13. Certain demographic groups are over-represented among the poor 20 Figure 14. More extreme poor have access to services than the moderate poor but worse labor outcomes 20 Figure 15. Labor income contributed most to decreasing poverty and inequality in RMs 22 Figure 16. The contribution of labor income to poverty reduction was the largest in metropolitan setting 22 Figure 17. Recipients of Bolsa Família are primarily the extreme poor. 23 Figure 18. Many of the moderate poor do not receive benefits. 23 Figure 19. The real minimum wage increased substantially between 2004 and 2011 25 Figure 20. Women’s labor income is an important contributor to reduction of poverty and inequality in Brazil RMs 27 Figure 21. The share of female and male workers by occupational category has remained relatively stable since 2004 27 Figure 22. Access to services is higher in the RMs but Brazil as a whole is catching up. 28 Figure 23. The Northeast remains the poorest region, but rates have fallen significantly 29 Figure 24. The HOI is higher in RMs than in the nation but still low in school quality, sanitation, and home Internet 30 Figure 25. Educational attainment and education inequality improved across RMs 31 Figure 26. Grade repetition and dropout rates declined across RMs 31 Figure 27. More primary school children are enrolled in private school across income group in RMs (%) (2013) 32 Figure 28. Differences between cores and peripheries are notable across RMs with the exception of some indicators in São Paulo 34 Figure 29. Inner peripheries are growing the fastest across RMs 34 Figure 30. Homicide rates have soared in Fortaleza and declined in Recife in recent years 35 Figure 31. Public transportation in Recife appears plagued by crime and is limited outside the core 36 Figure 32. 58 RMs and RIDEs 45 Figure 33. 12 Metropoles according to criteria of spatial/economic integration (REGIC/IBGE, 2007) 45 Figure 34. Poverty and vulnerability headcounts with and without adjusting income for cost of living 55 Figure 35. Comparison of poverty lines in Brazil 56 Figure 36. Stayers, sliders, climbers from 2004 to 2012 in Brazil RMs 57 Figure 37. Matrix of multidimensional and income poverty in Brazil 60 Tables Table 1. Poverty reduction and shared prosperity have both advanced in Brazil’s RMs since 2004 9 Table 2: Households’ monthly expenses on public and private transportation per income decile (POF 2009) 37 Table 3. GDP (in 2000 R$ millions) and real annual GDP growth (%) 46 Table 4. GDP per capita (in 2000 R$) and real GDP per capita growth (%) 47 Table 5. Mean income (R$2012) and income growth (%) of the total population and bottom 40% of the national income distribution 48 Table 6. Mean income growth (%) of the population and bottom 40% of each state living in the RM 49 Table 7. Gini Index across individuals using household per capita income 50 Table 8. Gini Index across Brazil RM households by gender, race, age and education of household head 51 Table 9. Poverty headcounts with and without adjusting for cost of living 52 Table 10. Percent difference in headcounts when adjusting income for cost of living 54 Table 11. Comparison of headcounts using different poverty lines in Brazil 56 Table 12. Characteristics by income group living in Brazil RMs 2012 58 Table 13. Probability of being moderate poor vs. extreme poor, vulnerable vs. moderate poor, or middle class vs. vulnerable (logit regressions) for individuals living in Brazil’s RMs 59 Table 14. Human Opportunities Index 2004 and 2012 62 Table 15. Characteristics of core and inner periphery by income group 2010 63 Table 16. Breakdown of workers’ commuting time house-to-work by RM 69 Table 17. Hedonic wage regression by year and income group across Brazil’s RMs (18 to 64 yr olds) 71 Table 18. Probability of being employed by year and income group across Brazil’s RMs (18 to 64 yr olds) 72 Table 19. Ten Largest Favelas (2010 census, IBGE) 74 List of acronyms ALMP Active Labor Market Policies AUs Aglomerações Urbanas (Urban Agglomerations) BF/BFP Bolsa Família Program BPC Benefício de Prestação Continuada (Brazil’s non-contributory pensions program) CadUnico Cadastro Único (Bolsa Família Single Registry) CEDLAS Center for Distributive, Labor and Social Studies—University of La Plata, Arg. CPF Cadastro de Pessoas Físicas (Personal Identification) FGV Foundation Getulio Vargas FPIC Funções Públicas de Interesse Comum (Public Actions of Common Interest) GFTS General Transit Feed Specification GIC Growth Incidence Curve GDP Gross Domestic Product HOI Human Opportunity Index IBEU Urban Well-Being Index IBGE Brazilian Institute of Geographic and Statistical Research IMF International Monetary Fund IPEA Institution of Applied Economic Research Ipeadata IPEA’s demographic, macroeconomic, regional, and social databases LAC Latin America and the Caribbean MDS Ministry of Social Development MPI Multidimensional Poverty Index NE Northeast PNAD Pesquisa Nacional de Amostra por Domicílios (National Household Sample Survey) PND Plano Nacional de Deselvolvimento (National Development Plan) OB Oaxaca-Blinder REGIC Regiões de Influência das Cidades (Regions Influenced by Cities) RIDE Região Integrada de Desenvolvimento (Development Integrated Region) RM Região Metropolitana, Metropolitan Region RMBH Metropolitan Region of Belo Horizonte RMRJ Metropolitan Region of Rio de Janeiro RMSP Metropolitan Region of São Paulo SAE Secretaria de Assuntos Estratégicos (Secretary of Strategical Matters) SAGI Secretary of Evaluation and Information Management SEDLAC Socioeconomic Database for Latin America and the Caribbean Executive Summary I n the 20th Century, Brazil rapidly urbanized and is now not only an urban nation but a metropolitan one. Brazil’s sprawling regiões metropolitanas (metro- politan regions, or RMs, which are municipal clusters) are now home to almost 50 million people and much of the country’s economic vitality. The RM spatial level and its supporting governmental institutions have thus become critical to Brazil’s future development. While challenges remain for tackling deprivation in rural areas, poverty in Brazil is now predominantly urban. More than six in 10 Brazilians in extreme poverty were living in urban settings as of 2012. Of these, over a fourth was concentrated in the 10 largest RMs. Brazil 10 largest metropolitan regions (RMs) 8 Brazil’s largest RMs achieved laudable progress both service delivery and safety nets (including the Bolsa in poverty reduction and shared prosperity between Família cash transfer program). 2004 and 2012. Whether measured by income or multi-di- mensional indicators, poverty in metropolitan regions Vulnerability remains stubbornly large and pervasive plummeted during the eight-year study period. Progress in Brazil’s metropolitan regions and will be a key chal- was primarily driven by increases in labor incomes, but lenge to sustaining and deepening shared prosperity also the expansion of public services and transfers. Ex- gains achieved over the past decade. Close to 6.5 mil- treme poverty reached an unprecedented low in metro- lion people, or 11.3% of the metropolitan population, are politan regions of 2.3% compared to 3.6% at the national not poor but have incomes insufficient to make them mid- level, and poverty dropped to 4.6%, just over half of the dle class. The vulnerable face a high risk of falling into pov- national level of 9%. Behind poverty reduction was both erty in the event of economic shocks, given the predom- redistribution and economic growth, which resulted in inant role of labor income in their household finances. high income growth among the bottom 40% and a falling Additionally challenges include the high share of the vul- income inequality. Those dynamics were accompanied by nerable working informally or unable to find employment an expansion of access to services reaching almost the en- at all. Furthermore, the vulnerable group in RM settings tire metropolitan population for a number of key services. in Brazil is likely to be even larger than what is captured by the national vulnerability line of R$291. Using regional However, those remaining poor are hard to reach vulnerability lines defined by the World Bank to compare through growth alone and the moderate poor, in countries of the region (US$10) enlarges this group to 21.8 particular, may face constraints on two sides: access million, a full 38% of the population. to labor markets to support their upward mobility and the same attention from social programs as the Poverty and shared prosperity are converging across extreme poor to meet their basic necessities, partic- the RMs—with the historically lagging North and ularly in terms of education and sanitation. There are Northeast RMs having improved greatly but still fac- also issues of equity in terms of those remaining behind, ing pronounced challenges. While the North and North- with households headed by afro-descendants, females, east RMs have historically been the poorest, and remain and young adults all over-represented among the poor. so, the rate at which these RMs have reduced poverty and In general, the poor and vulnerable have limited partic- grown incomes of the bottom 40 percent is causing them ipation in labor markets, particularly the extreme poor. to converge towards the richer RMs of the South. In par- However, the moderate poor suffer more than the ex- ticular, two RMs of the Northeast—Fortaleza and Recife— treme poor in access to basic services which may reflect consistently stand out as lead performers, with the high- Brazil’s focus to date on eliminating extreme poverty. est rates of GDP growth, income growth of the bottom This study highlights some of these remaining gaps in 40%, poverty reduction, and income inequality reduction. Table 1. Poverty reduction and shared prosperity have both advanced in Brazil’s RMs since 2004 Extreme Extreme Poverty HC Poverty HC Bottom Mean inco- poverty HC poverty HC 2004 (%) 2012 (%) 40 income me change 2004 (%) 2012 (%) growth (%) (%) Brazil All 7.6 3.6 22.4 9.0 7.4 5.0 Brazil RMs 4.7 2.3 14.7 4.6 7.7 4.7 Brazil Rural 18.4 9.4 45.4 24.6 6.8 5.9 Brazil Non RM- Urban 5.9 2.8 19.7 7.2 7.3 4.9 Source: World Bank calculations using PNAD 2004, 2012. Note: Rural, urban, and metropolitan are exclusive groups. Poverty lines corre- spond to real values of R$70 (extreme) and R$140 (moderate) based on June 2011 Reais. The bottom 40 refers to the bottom 40 percent of the national income distribution living in the area. Growth refers to growth in the real value of income. 9 Poverty and equity outcomes vary significantly be- Poverty and Vulnerability lines in Brazil tween the cores of RMs and their peripheries. The core municipalities (i.e. the capitals of the given states) offer better access to services and formal job markets than their peripheries (the surrounding municipalities). However, in- equality remains higher in the core, reflecting the more Vulnerability: R$141-291 diverse income structure in the center of metropolitan ar- Moderate Poverty: R$71-140 eas. Concentric dynamics can be observed: the further a Poverty line R$140 Extreme Poverty: R$0-70 municipality is located from the core, the worse its indica- tors. This underlines the importance of better integrating the outer peripheries into the RMs so as to make labor and economic opportunities offered by RMs more inclusive. A * lines based on monthly per capita income. key goal will be improving mobility across the metropoli- tan sphere to enhance citizens’ access to services and jobs. In a less favorable economic context, the health of metropolitan regions, which comprise over 70% of Brazil’s GDP, will help determine growth and shared prosperity across Brazil. History has shaped a complex institutional structure in urban Brazil that has strong bearing on the design and implementation of policies, programs, and projects requiring coordination (Melo and Pereira 2013). This is particularly visible in the RMs, where acute challenges exist in getting municipalities, RM agen- cies, states and the federal government to work together toward common social and economic goals. While met- ropolitan governance is beyond the scope of this study, this level of governance directly impacts service delivery efficiencies and therefore strongly relates to policies re- garding poverty, vulnerability, and shared prosperity in urban settings. 10 Introduction R eflecting deep-rooted historical dynamics, region, with more than 84% of its population living in ur- Brazil’s population is not only heavily urban1 ban areas, primarily along the coastline and in the south. but is also increasingly metropolitan. Between The change happened very quickly, as noted by Ricardo 1940 and 2010, Brazil’s population increased almost five- Neves, “no other country of comparable size managed, fold. In 1970, Brazilian census takers found for the first in only two generations, to go from a rural to an urban time that their country’s urban population surpassed its one” (Perlman, 2008).2 However, given its speed, the ur- rural counterpart. Today, Brazil is among the most ur- ban centers developed often in an ad hoc manner with banized countries in the Latin America and Caribbean limited planning, bringing together the haves and have- nots, creating an “enduring framework of inequality,” in 1  This research uses the classification developed by Instituto Bra- the words of urban analysts George Martine and Gordon sileiro de Geografia e Estatística (IBGE), the Brazilian Statistical Office. McGranahan (2010). Despite some efforts at urban plan- The criteria that subdivide territory into urban and rural areas are based on the laws of each Brazilian municipality and updated with ning, the absence of vigorous implementation had con- each census. Urban households are those located inside the urban sequences on the development of those areas.3 The result perimeter of a city or village. Urban areas are classified as “urbanized areas” (characterized by buildings, streets and intense human occu- was often concentrated poverty, health hazards, environ- pation), “not urbanized areas” (legally defined as urban but charac- mental blight and a negative perception of further urban- terized by rural occupation) and “urban isolated areas” (also legally defined as urban but separated from a municipality by rural area ization. Consequently, instead of using urban planning to or by another legal limit). On the other hand, rural households are those located in rural areas external to urban perimeters and in ru- ral agglomerations (legally defined rural areas formed by adjacent 2  While the spatial distribution of the urban population in the buildings separated by no more than 50m). Rural agglomerations are Southeast and South regions has remained relatively stable, the classified as “of urban extensions” (located outside the legal urban Northeast has declined from 35% to just under 28% of the total. The perimeter but developed as a result of urban expansion), “villages” Center-West and the North have increased their shares although (rural agglomerations of no business or private character, charac- each now accounts for only 7-8% of the national urban population terized by minimum number of services and equipment), “nuclei” (“Urbanization Review for Brazil” – draft, September 2013, LCSDU). (rural agglomerations liked to a single owner or company) and “oth- 3  As noted by Martine and MacGranaham (2010), while “master er agglomerations”. This research primarily refers to metropolitan, plans” have been part of urban planners’ discourses for a long time, non-metropolitan urban, and rural. the reality is that urban growth has outstripped land use planning. 11 Figure 1. Inside and outside the metropolitan region Other state Non-metropolitan regions: other state + outer periphery Outer • The outer periphery is composed of the municipalities directly bordering the RM. periphery • Urban and rural rates in this report refer to locations outside of the RM. Inner periphery Metropolitan region (RM): core + inner periphery Core • The core is the capital municipality and has the same name as the RM. (Capital) • The inner periphery is composed of the other municipalities within the RM. improve the evolution of the city, policy makers sought trict, Brasília, remain Brazil’s primary metropolitan regions. to limit the permanent settlement of new rural migrants As of 2012, these ten RMs accounted for approximately and the poor, which often led to the creation of informal 31% of the Brazilian population. urban settlements known as “favelas.” Migration to metro- politan regions did eventually slowdown, due mostly to The study consists of eight papers. The present paper pro- economic growth in traditionally lagging regions and the vides an overview of the evolution of poverty, inequality country’s transition to a more advanced stage of urban- and shared prosperity across Brazil’s large metropolitan ization. However, urban poverty accounts for 61% of the regions from 2004 to 2012. Seven separate notes focus extreme poor and 59% of the moderate poor and urban analysis on individual metropolitan regions of the less-de- areas contain large concentrations of households which veloped North (Belém) and Northeast (Recife, Salvador, are highly exposed to risks. and Fortaleza) and the wealthier and largest RMs, which are located in the Southeast (Rio de Janeiro, São Paulo, The objective of this study is to provide an overview of and Belo Horizonte).5 Across key poverty and shared pros- the evolution and status of poverty and shared pros- perity indicators, the RMs are benchmarked to each other perity in Brazil’s metropolitan regions (regiões met- and compared to other RMs in Brazil as well as to the na- ropolitanas–RMs). To the authors’ knowledge, this is the tion and the respective state. Within each RM, the analysis first study of poverty and shared prosperity diagnostics at distinguishes between the core of the RM (capitals of the the metropolitan region level. It aims to inform the World given states) and other municipalities within the RM (the Bank’s engagements in metropolitan areas and large mu- inner periphery) and in the municipalities directly border- nicipalities to address the twin goals of poverty reduction ing the RM (the outer periphery) (Figure 1). This spatial and shared prosperity in Brazil. The RMs considered in this lens is important to understand the highly interrelated study consist of the nine formed between 1972 and 1974 economic, social and governance challenges facing RMs under the Complementary Law 14 of June 8, 1973 which and their peripheries. allowed for multiple municipalities to incorporate into one região metropolitana (RM). These include: Belo Hori- While the ten metropolitan regions analyzed in this zonte, Belém, Curitiba, Fortaleza, Porto Alegre, Recife, Rio study are all large urban agglomerations, there are de Janeiro, Salvador, and São Paulo.4 While additional RMs important differences in their population sizes and so- have formed since, these nine RMs and the Federal Dis- cio-economic compositions. The RMs vary substantially in their population size. As of the 2010 Census, 19.7 mil- 4  IBGE representatively samples nine metropolitan regions plus the federal district for the PNAD. This sampling allows for an in-depth analysis of the evolution of poverty and equity at the metropolitan 5  The RM of Curitiba, Porto Alegre and Brasilia (DF) are integrat- level. Thus, while there are other RMs, for the purpose of this note, ed into the overall analysis but only the seven RMs of Belém, Recife, “RMs” only refers to the PNAD’s nine RMs of Belo Horizonte, Belém, Salvador, Fortaleza, Rio de Janeiro, Sao Paulo and Belo Horizonte are Curitiba, Fortaleza, Porto Alegre, Recife, Rio de Janeiro, Salvador, and further investigated into separates notes, reflecting the World Bank São Paulo, plus Brasília (Distrito Federal). more specific engagement with this municipalities and states. 12 Box 1. States, municipalities and metropolitan regions in federal Brazil Article 25 of Brazil’s constitution, enacted in 1988, gives states the right to create metropolitan structures. This was a shift from the 1967 constitution, which had placed metropolitan regions under federal authority. While giving states the authority to create such bodies, the current constitution does not set any further requirements regarding their structure or funding. Further complicating the situation is that the 1988 constitution also established municipalities as full federation members. That is to say, the more than 5,000 municipalities that are scattered across Brazil today are not subordi- nated to states or to any structure created by states. They enjoy the same autonomy and sovereignty. While municipalities within an RM have a vested interest in working together given their proximity, any action undertaken by a metropolitan agency must be specifically agreed to by all municipalities involved. Incentives to coordinate vary considerably. The political economies of municipalities and metropolitan regions and their capacity to coordinate are thus determinant factors of success. Adapted from Metropolitan Governance and Finance in São Paulo, in “Financing Metropolitan Governments in Devel- oping Countries,” edited by Bahl, R., Linn, J. and Wetzel, D. (2013). lion people lived in the RM São Paulo, more than the entire erty reduction, access to services, and shared prosperity population of Chile and over twice that of New York City. achieved so far, will require assertive joint action by mul- In contrast, only 2.1 million people lived in the RM Belém. tiple levels of government (See. Box 1). Including poverty The portion of the respective state population living in and vulnerability in the larger agenda of metropolitan re- each RM also varies, ranging from 25% in Belo Horizonte gions could provide a much-needed integrated focus. to 74% in Rio de Janeiro. Furthermore, there are large dif- ferences in the racial composition of the RMs. While the citizens of the RMs in the North and Northeast of Brazil predominantly self-identify as pardos (mixed origin), the RMs located in the South and Southeast have larger pop- ulations that call themselves brancos (white). Across RMs, the share of the population that self-identifies as pretos (black) is relatively low, ranging from 3% to 12%, with the exception of Salvador, where the figure is 27%. Addressing challenges will require coordinated ac- tions at the federal, state, metropolitan, and municipal levels. Due to the federal nature of the country, Brazilian states and municipalities have strong autonomy and sov- ereignty. This in turn has major bearing on the implemen- tation of policies that require all levels of government to work together. In spite of the complexity of coordination, metropolitan regions offer potential for poverty reduc- tion and shared prosperity in Brazil. Indeed, most of the challenges, particularly those related to services and ur- ban mobility and sustaining the substantial gains in pov- 13 Economic growth and redistribution have generated significant poverty reduction in Brazil’s RMs A. GDP has consistently grown in RMs since the top 10 RMs in Brazil representing 44% of the national 2004 but slightly slower than at the national GDP in 2011. level Economic growth in the RMs has been slightly slower L arge shares of Brazil’s GDP are concentrated in than the national rate over the last decade, over all, metropolitan regions, making their inclusive and in the last two years of the data (2010 and 2011). growth vital to the country’s overall economic From 2004 to 2011, annual GDP growth was just under health. As countries develop, economic activity gener- the national 4 percent average in the RMs, mainly reflect- ally becomes more concentrated in certain areas, with ing the greater susceptibility to the 2002-03 crisis and agglomeration effects increasing returns in those areas lower growth in 2011. While the RMs were slightly less (Duranton, 2013; WDR, 2009). Brazil is no exception, with affected by the 2009 global crisis than the rest of Brazil, a large share of its GDP concentrated in its metropolitan their growth rates have been trending lower than the na- regions. For instance, in the Northeast the three largest tional average since then (Figure 2). Still, there was con- metropolitan regions: of Salvador (Bahia), Recife (Pernam- siderable variation in growth rates across RMs over the buco), and Fortaleza (Ceará) concentrate 20% of the re- 2004-2011 period, with two of the poorest RMs in 2004, gion’s population, and contribute one half to two thirds of Fortaleza and Recife, growing the fastest while the other all economic activity in their respective states (2010). RMs two poorest RMs in 2004, Belem and Salvador, grew at a function as centers of growth for Brazil, in terms of GDP pace in the bottom half of RMs and in the latter case, the and other measures captured by the REGIC survey,6 with slowest of all. 6  The REGIC 2007 survey identified the core municipalities of the Statistical Office classifies as important for trade, banking, services, 10 RMs as the top poles of attraction among areas that the National federal public management, and corporations. 14 Figure 2. In recent years real GDP growth in the RMs Figure 3. GDP growth grew the most in the has been slowing down relative to Brazil as a whole7 traditionally poorer Fortaleza and Recife 8 7 7 Avg. annualized growth 2004/11 (%) 6 6 5 5 GDP growth (%) 4 4 3 2 3 1 2 0 -1 1 -2 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Belém Fortaleza Recife Distrito Federal Belo Horizonte Curitiba São Paulo Rio de Janeiro Porto Alegre Salvador Brazil RMs Brazil Source: World Bank calculations using Ipeadata/IBGE. Source: World Bank calculations using Ipeadata/IBGE. Figure 4. Income growth from 2004 to 2012 in RMs Figure 5. Inequality remains higher in RMs was highest for the poorest deciles than in urban and rural areas 0.58 0.57 10 0.56 Annualized income growth 8 0.54 rate (%) 2004/2012 0.52 Gini Index 0.54 6 0.50 0.48 4 0.46 2 0.44 10 20 30 40 50 60 70 80 90 100 2004 2005 2006 2007 2008 2009 2011 2012 GIC (RM) Growth rate in mean (RM) BR All BR RM BR Urban BR Rural Source: World Bank calculations using PNAD 2004/2012. Source: World Bank calculations using PNAD 2004, 2012. B. Economic growth has most benefitted the areas experiencing the smallest declines (Figure 5). More- poor, resulting in declining inequality across over, inequality remained highest in metropolitan set- and within RMs tings than in the rest of Brazil. This speaks to the particular profile of people who live in RMs—namely, the 10 RMs The growth incidence curve (Figure 4) indicates that encompass large and diverse populations, with high vari- growth in urban and metropolitan areas was positive ation among incomes. When viewed across RMs (Figure 6), for the entire income distribution, but disproportionate- inequality declined in all RMs, with the greatest declines ly benefited the lower deciles underscoring an inclusive in Recife and Fortaleza, the two RMs that initially had the growth pattern.7Not surprisingly then, income inequal- highest levels of inequality. ity steadily declined between 2004 and 2011 before it started to flatten in 2012. The sharpest declines were in On average, growth played a larger role than income urban areas, followed by metropolitan regions, with rural redistribution in reducing poverty in Brazil’s RMs from 2004 to 2012—with the notable exception of the RMs Re- 7  Graph to be updated to 2012. cife and Fortaleza, which were the two of the poorest RMs 15 Figure 6. Inequality has fallen in all RMs Figure 7. Growth reduced poverty more but remains high and heterogeneous across RMs than redistribution, on average, in Brazil’s RMs 0.65 RM rate 2004 RM rate 2012 25 23.0 22.5 Growth Redistribution Total Reduction in Moderate Poverty (%) 19.2 0.60 20 10.6 4.5 14.5 Gini Coe cient 13.7 13.5 0.55 15 11.4 2.4 9.2 4.6 0.50 10 3.8 7.6 6.2 5.8 3.8 4.8 0.45 5 3.3 3.2 3.1 1.7 0.40 0 Distrito Federal Belo Horizonte Salvador Rio de Janeiro Brazil RMs São Paulo Belém Fortaleza Recife Porto Alegre Curitiba Brazil Belem Fortaleza Salvador Belo Horizonte Distrito Federal Sao Paulo Curitiba Porto Alegre Rio de Janeiro Recife Source: World Bank calculations using PNAD 2004, 2012. Source: World Bank calculations using PNAD 2004/2012. in 2004, and Curitiba (where growth and inequality each incidence of 25% of the population, compared to only 7% contributed half of the gains in poverty). The Datt-Ravail- in non-metropolitan urban areas and nearly 5% in metro- lon decomposition8 shows that growth contributed two politan areas. However, given that the majority of Brazil- thirds of the poverty reduction with income redistribution ians live in urban or metropolitan areas, the majority of providing the remaining one third during the study peri- the poor live in urban and metropolitan areas. In partic- od (Figure 7). Looking at growth elasticity of poverty in ular, the 10 largest RMs concentrate 15% of Brazil’s poor the respective RMs also confirms the critical role played and nearly 20% of the extreme poor. by growth in RMs, with metropolitan poverty more elastic than at the national level. RMs presenting high incidence While poverty rates are low in metropolitan settings, of inequality, such as the RM Salvador, display the expect- the sheer size of the RMs means that they still have ed lower elasticities. The decrease in inequality seen in the very large numbers of poor people. Since 2004, the 10 RM Recife also translates into greater elasticity of poverty largest RMs in Brazil had reduced poverty by close to 70% reduction to mean income growth. Whereas a one per- and extreme poverty by over 50%. As of 2012, 4.6% of cent increase in mean income growth over the 2004-11 the metropolitan population lives in poverty and 2.3% period resulted in a reduction in poverty of -2.27 percent live in extreme poverty. By comparison, Brazil’s overall in the case of the RM Recife, a corresponding increase only poverty rate in 2012 was 9% and the extreme poverty translated in reduction of -1.06 percent for the RM Salva- rate was 3.6%. Though lower than the national rate, given dor, the lowest of all RMs. that over 2.6 million people living RMs are still poor, it is important to continue providing support to ensure that C. Monetary poverty has decreased these people can rise out of deprivation and that the vul- considerably, especially in the North and nerable do not fall back into poverty (See Figure 8). Northeast, but vulnerability remains a challenge Poverty rates vary across RMs but have fallen consid- erably since 2004 in all of them, especially in those lo- The majority of the poor in Brazil live in urban a reas. cated in the poorest states. In 2012, all RMs had lower In 2012, poverty was more common in rural areas with an poverty rates than the national average of 9%, as those in the traditionally poorer Northeast and North had rates 8  The Datt-Ravallion methodology decomposes the evolution of poverty changes into two components: (1) the income growth com- closer to 8%, while those in the South and Southeast had ponent, i.e., the change in poverty due to a change in the mean in- rates closer to 3%. In the last decade, the states with the come in the absence of changes in income distribution, and (2) the strongest poverty reduction have been in the North and redistribution component, i.e., the change in poverty due to changes in the Lorenz curve while keeping the mean income constant. Northeast, though their states continue to exhibit signifi- 16 Figure 8. Most people living in poverty in Brazil Figure 9. Vulnerability levels have declined are located in non-metropolitan urban areas but remain significant 45 50 35 2012 2004 Vulnerability (R$140-290) Headcount (%) 40 30 28.9 29.4 Poverty (R$140) headcount (%) Pop. living in poverty (millions) 35 40 26.0 24.4 25 30 21.3 21.0 30 19.1 18.9 18.0 20.1 19.3 25 20 18.0 18.3 15.6 15.5 20 15 20 11.3 11.0 15 9.5 9.1 10 8.4 7.9 10 6.8 10 5 5 0 0 0 2004 2005 2006 2007 2008 2009 2011 2012 Brazil RMs Belém(N) Fortaleza(NE) Salvador(NE) Rio de Janeiro(SE) Distrito Federal São Paulo(SE) Porto Alegre(S) Curitiba(S) Recife(NE) Belo Horizonte(SE) RM poor Rural poor Urban poor RM rate Rural rate Urban rate Source: World Bank calculations using PNAD 2001/2012. Note: Rural, urban, and metropolitan are exclusive groups. Source: Work Bank calculations using PNAD 2004, 2012. cantly higher rates of poverty and may need to look to the RM, bottom 40 income growth exceeded that of the mean strategies deployed in the RMs to spur poverty reduction. population income growth, showing that growth has not only benefitted the bottom 40 but benefitted them more Vulnerability is an important challenge for Brazil’s than the top six quintiles. The Northeast RMs performed RMs, particularly in the Northeast. Along with strong particularly well in terms of shared prosperity, with its RMs poverty reduction, vulnerability rates have also fallen occupying three out of the four top positions. Notably, across RMs and in 2012 stood at 11% on average (See Fig- while the RM Salvador had the lowest GDP growth rate ure 9.) However, while RMs in the South and Southeast cut over the period, average income growth was among the vulnerability rates by half, RMs in the Northeast reduced highest for the bottom 40. At the other end of the spec- vulnerability by closer to one quarter. Given the large trum, the RMs Belém and Rio de Janeiro performed the reduction in poverty rates in the North and Northeast, a worst in terms of shared prosperity (See Figure 11.) Due to relatively decline in vulnerability not surprising, because poverty reduction, many of the bottom 40% are vulnera- a large portion of the people graduating out of poverty ble and close to half are middle class. Thus, promoting the are likely to have joined the ranks of the vulnerable, rather income growth of the non-poor has become key to sus- than making it through to the middle class. When using taining and deepening shared prosperity (See Figure 12.) international lines, rates of vulnerability are even higher and significantly so, indicating that the problem is likely The expansion of shared prosperity has contributed to even larger than that shown here (see Box 2). a large portion of the population of Brazil’s RMs climb- ing the socio-economic ladder over the past decade. D. Upward economic mobility has been By 2012, more than three in four Brazilians had middle strong in Brazil’s RMs class incomes while five in six metropolitan Brazilians did. In the absence of panel data, assessing intra-generational The RMs grew incomes of the bottom 40% by more mobility in the metropolitan setting is complicated. How- than the average for their respective states between ever, using synthetic panels (Annex 7) to drill down on 2004 and 2012. Close to a fourth of the Brazilians in the bottom 40% of the income distribution live in the ten larg- growth of the bottom 40% compared to the mean. We present here est metropolitan regions. The growth of their income was this measure at the national level (bottom 40% of Brazil residing in the given RM) and at the state level (bottom 40% of the state residing robust at close to 8%, on average.9 Furthermore, in every in the corresponding RM) to provide a finer lens on the reading of shared prosperity in a large federal country such as Brazil. The bench- 9  As one of the twin goals of the World Bank, shared prosperity mark for shared prosperity in Brazil remains at the national level. See is measured by the growth of the income per capita of the bottom Annex 16 on the measurement of shared prosperity at the sub-na- 40% of the income distribution; pro-poor growth is measured by the tional level. 17 Box 2. Poverty and Vulnerability through an international measurement lens Using international lines to measure poverty in Brazil’s RMs shows significant poverty reduction but rescale pov- 50% metropolitan residents are either poor or vulnerable erty and vulnerability to more salient challenges. Since in 2012 2004, extreme poverty using the global line ($1.25 PPP per day) fell by over half, to reach 2.5% in 2012. Similarly 15.6% are poor strong results are seen with regional lines but with inci- 38 % are vulnerable dence levels classifying a larger share of the metropoli- tan population as poor. Indeed, using the regional lines for extreme ($2.5 PPP) and moderate ($4 PPP) poverty, poverty is seen falling by two thirds and half, respectively, to reach 6.3% and 15.6% in 2012. While a similar reduction is seen with vulnerability, close to 40% of metropolitan Brazilians continue to live in a state of vulnerability to poverty as measured with the $4-$10 PPP lines. In recent years, R$70 and R$140 per capita per month, administrative poverty lines defined for the Bolsa Família and Brasil Sem Miséria programs, have been increasingly used in place of official poverty lines. Thus, the monetary poverty measures considered in the rest of this study refer to individuals, unless noted otherwise, with per capita household incomes between the following thresholds: poverty–below R$140, extreme poverty and the extreme poor–below R$70, the moderate poor–between R$70 and R$140, vulnerable–between R$140 and $$291, and mid- dle-class–R$291+.10 Cost of living is not taken into account in order to be consistent with national methodologies of measuring poverty in Brazil. As such, poverty comparisons between Brazil RMs and Brazil are likely more favorable for the former given that the cost of living in metropolitan areas is generally higher than average. The same applies for regional differences, particularly the South/Southeast and North/Northeast divide. See Annex 6 for more infor- mation about poverty lines in Brazil and see Annex 5 for a comparison of results adjusting and not-adjusting for cost of living. Figure 10. National, regional and international lines show a consistent reduction in poverty and vulnerability but levels vary substantially across thresholds. 50 Vulnerable Rate (R$140-R$291/month) 38.1 39.4 39.0 38.1 36.5 37.1 37.2 40 34.9 Moderate (R$140/month) 30.4 Extreme (R$70/month) Headcount (%) 30 34.4 26.9 26.1 23.0 20.8 Regional Vulnerable ($4 PPP-$10 PPP/day) 20 17.3 15.6 Regional Moderate ($4 PPP/day) 10 6.2 4.6 Regional Extreme ($2.50 PPP/day) 3.9 4.0 3.4 3.2 2.6 2.5 Global Extreme ($1.25 PPP/day) 0 2004 2005 2006 2007 2008 2009 2011 2012 Source: WB calculations using PNAD 2004 to 2012, R$ lines refer to the real value based on June 2011. 10 10  SAE recently updated its definition of the middle class in Brazil, raising the threshold income to R$291 per capita. It also defined three types of “middle class:” lower-middle class, R$291-441; “middle” middle class, R$441-R$641; and higher middle class, R$641-$1,019 (SAE, 2013). Vulnerability is defined as income per capita between R$140 and R$291. http://www.sae.gov.br/site/?p=17351#ixzz2h0ho0ALf. 18 Figure 11. Average annualized real growth rate of Figure 12. The middle class makes up almost income is high for the national bottom 40 living in RMs half bottom 40 living in Brazil’s RMs Average annualized growth income 2004/12 (%) 9 8.4 8.3 8.3 100 8.0 7.9 7.7 8 7.5 7.3 7.3 Share of the bottom 40 (%) 2012 6.8 6.6 80 7 6 60 5 4 40 3 2 20 1 0 0 Brazil RMs Brazil RMs Brazil Belém Belo Horizonte Salvador Fortaleza São Paulo Distrito Federal Curitiba Porto Alegre Rio de Janeiro Rio de Janeiro Recife Curitiba São Paulo Distrito Federal Salvador Belém Fortaleza Porto Alegre Belo Horizonte Recife B40 RM All RM Extreme poor Moderate poor Vulnerable Middle Class Source: World Bank calculations using PNAD 2004, 2012. Source: World Bank calculations using PNAD 2004, 2012. economic mobility between 2004 and 2012 indicates that the economic ladder.13 While gender differences in the approximately 20% of the people living in RMs saw an im- probability of moving from extreme to moderate poverty provement in their socio-economic class.11 These results decreased between 2004 and 2012, being a woman still are similar to those found at the national level.12 meant having a lower likelihood of moving out of vulner- ability and into the middle class. White households were But gender and race continue to influence socioeco- more likely than afro-descendant households to move nomic mobility in the RMs. Analysis of the transition from extreme to moderate poverty, but afro-descendants between income groups shows that females across all were more likely than whites to move from moderate pov- income groups are less likely than males to move up erty to the vulnerable group. 11  In the absence of panel data, analysis of the evolution of in- come over the period is complemented by a zoom-in on income E. The poor and vulnerable face specific mobility using synthetic panels, based on the approach recently de- veloped by Dang, Lanjouw, Luoto and McKenzie (2011). The analysis challenges to their upward mobility identifies people who left and people who stayed in poverty, and the potential hypotheses that lie behind the trends. The main advantage As of 2012, over 9 million people living in the 10 larg- of this approach is that it does not need to impose much structure on the individual income-generating process. Instead, it allows us to est RMs in Brazil were either poor or vulnerable to calculate lower and upper bounds on the movements in and out of poverty, with households headed by young adults, af- poverty, depending on the assumption regarding the individual-spe- cific error term. Synthetic panels are built using two cross-section ro-descendants and females over-represented among datasets, 2004 and 2012. The methodology and results are further these groups. About a third of the metropolitan poor and detailed in Annex 7. Caution should be applied in the interpretation of those results. Indeed, in zooming in on metropolitan regions, the vulnerable live in the Northeast, a half in the Southeast, significant reduction of the sample size presents a first limitation— while the North, South, and Distrito Federal each have the RMs are pooled to overcome this limitation, extracting the re- less than a tenth. Comparatively, less than a fifth of the spective RMs from this pooled sample. This solution only partially remedies the sampling limitation. In addition, the very nature of a entire metropolitan population lives in the Northeast and metropolitan setting signifies that a higher degree of migration is more than three fifths lives in the Southeast. Children and to be expected. This weakens some of the assumptions used when building synthetic panels—namely that there are good reasons to households headed by young adults, afro-descendants, believe that the composition of the group observed with the first and single mothers are significantly over-represented cross-section dataset will be similar if not identical to the composi- tion of the group captured by the second cross-section dataset. For among both the poor and vulnerable (See Figure 13). those reasons, the results are presented in Annex 7 and are not fur- ther expanded upon in the main analysis. 13  The analysis was done using a simple logit model to capture 12  According to Fruttero, Castaneda, Lopez-Calva & Lugo (2012), factors at play for income groups to transition into the next upper over the period 2003 to 2011 and using a vulnerability line of R$250, group (e.g., from extreme to moderate poverty, from moderate pov- 19.8% of the Brazilian population rose in socio-economic status. erty to vulnerability). Results are presented in Annex 8. 19 Figure 13. Certain demographic groups Figure 14. More extreme poor have access to services are over-represented among the poor than the moderate poor but worse labor outcomes 80 100 Populatino share of income group (%) 2012 70 Percent of income group (%) 2012 80 60 50 60 40 30 40 20 20 10 0 0 15 to 25 Afro-desc Female Single Children Sanitation Secondary+ Informality Labor Unemployment year HH HH HH Mother HH (0-15 yrs) participation R$70 R$70-R$140 R$140-R$291 R$291+ R$70 R$70-R$140 R$140-R$291 R$291+ Source: World Bank calculations using PNAD 2012. Source: World Bank calculations using PNAD 2012. This is especially true for young- adult and single-mother Labor market indicators such as hourly wage, infor- households, which account for more than twice as many mality rates, labor force participation, and unemploy- households in poverty or vulnerability as in the general ment strongly differentiate the moderate poor from population. Female-headed households are also over-rep- the extreme poor. This hints at the key role played by resented among the moderate and extreme poor but labor markets in the incidence and depth of poverty. have similar shares of the vulnerable group as they do in While the unemployment rate among the extreme poor the general population. in RMs is 84%, it falls to 27% and 21% for the moderate poor and vulnerable, respectively, and to around 5% for The poor and vulnerable continue to have lower edu- the middle class (See Figure 14). If they do work, the ex- cation and to lack access to some basic services such treme poor are more likely to be self-employed, informal, as sanitation. While access to services is approaching or work part-time compared to the moderate poor. On the complete coverage in RMs, there are still households that other hand, the moderate poor resemble the vulnerable lack adequate sanitation and whose children who are in terms of labor force participation and unemployment not enrolled in school. Both of these services impact the rates, though they have higher levels of informality.15 entire life cycle of a person due to the long-term effects of sanitation on health and education on characteristics ranging from labor earnings as an adult, smoking, drink- ing and relationships.14 In RM settings, this situation is, surprisingly, most commonly found among the moderate poor (See Figure 14). For example, 18% of the RM moder- ate poor live in households with school-age children who do not attend school compared to 12-13% of both the ex- treme poor and vulnerable. Adult educational attainment follows the same pattern, with the extreme poor and vul- nerable having higher rates than the moderate poor. Like- wise, 26% of moderate poor individuals live in households lacking sanitation, compared to 18% for extreme and vul- 15  While informality offers an opportunity for firms to operate un- nerable households. der lower regulatory and wage costs, it often means low insurance for workers, under-saving for retirement, unfair competition, and 14  See Becker and Tomes (1979), Feinstein et al. (2006), Heckman noncompliance with tax collection and the rule of law. It ultimately et al. (2006), and World Bank (2010b). creates a drag on productivity and growth (World Bank, 2007). 20 Labor markets, demographic changes, and transfers have been key to poverty and inequality reduction A. Labor income drove poverty and dependency ratio) reduced poverty. Transfers played a inequality reduction in Brazil’s RMs positive role in every RM and, after labor income, were the second- or third-biggest factors most commonly found to I ncreased labor income contributed the most have reduced poverty. to poverty reduction in each of the RMs. Using a Shapley decomposition to separate the factors be- Transfers and falling dependency ratio also contribut- hind poverty changes shows that labor income contrib- ed to poverty and extreme poverty reduction. Changes uted close to 80% of the change in poverty and over 95% in transfers16 and demographics17 contributed to poverty of the change in extreme poverty between 2004 and and extreme poverty reduction while changes in contrib- 2012 across Brazil’s RMs. (See Figure 15). This surpassed 16  The PNAD defines income from capital and social programs the national rate by close to 20 and 50 percentage points, using the same variable. In order to separate these two sources of income, potential income from Bolsa Família and Benefício da respectively, indicating that labor markets are especially Prestação Continuada is imputed based on the number of eligible critical in those settings, including when compared to children in the family for Bolsa Família and the value of the variable (i.e., if it is the same as the minimum wage, the variable is assumed non-metropolitan urban settings (see Figure 16). Within to be income from BPC). The potential income from social programs RMs, labor income consistently played the leading role is then added to the income from abono de permanência and remit- tances to define “transfers,” while the remaining income is defined as in poverty and extreme poverty reduction, contributing “capital gains.” See Firpo, Pieri, Pedroso, and Souza (2013) and Barros over 60% in each one. The smallest contribution of la- et al. (2006). bor income to poverty reduction occurred in the three 17  The share of adults means the number of adults in a household. It is a demographic measure particularly useful in appreciating the Northeastern metropolitan regions, whereas the largest evolution of dependency ratios and/or the effect of “youth bulges”— was seen in the RM Rio de Janeiro. Across RMs, changes notably in combination with the share of occupied, i.e., the number of adults per household in the workforce. (See Azevedo, Inchauste in transfers and the share of adults in households (falling and Sanfelice, 2012.) 21 Figure 15. Labor income contributed most Figure 16. The contribution of labor income to poverty to decreasing poverty and inequality in RMs reduction was the largest in metropolitan setting Brazil All Brazil RM Brazil Rural Brazil Urban 20 Poverty R$70 Poverty R$140 Gini 0 Contribution to poverty reduction R$140 (%) 4.2 -12.3 -11.3 -14.4 -12.5 0 -20 -9.1 -9.8 -5.7 Share of change (%) -20 -11.3 -39.6 -24.6 -40 -61.9 -78.9 -62.1 -40 -60 -60 -69.6 -80 -45.9 -80 -25.8 -25.4 -78.9 -9.8 -100 -95.1 -100 Share of adults Labor income Non-labor income Share of adults Labor income Non-labor income Source: World Bank calculations using PNAD 2004/2012; Note: The share of adults (occupied) corresponds to the number of adults (occupied adults) in the household. Non-Labor income includes transfers, pensions, capital and other non-labor incomes. Poverty lines used are R$70 (extreme poverty) and R$140 (poverty). utory pensions and changes in employment had neg- bor income. In Recife and Fortaleza, the RMs with the two ative effects on extreme poverty reduction. Across RMs, largest reductions in Gini over the period, labor incomes changes in transfers and the share of adults in households dominated, with the second-highest factor contributing (falling dependency ratio) reduced poverty, with the latter less than 12% in each. In the RM Salvador, labor income particularly influential in Porto Alegre and Curitiba. Trans- accounted for over 50% of the reduction in Gini, but fers played a positive role in every RM and, after labor in- transfers were close behind and mattered more in the RM come, were generally the second- or third-biggest factor Salvador than in any other RM. Across RMs, transfers, de- in reducing poverty (trading off with the share of adults in mographics, and employment (apart from Rio de Janeiro the household). However, transfers played a smaller role and Belém) contributed to income inequality reduction, than at the national level, likely due to the overpowering while capital generally had the opposite effect, especially aspect of labor income in RMs. With a small, negative in- in São Paulo. In the case of the RMs Salvador and Belém, fluence on poverty, an increasing share of adults occupied contributory pensions also exerted a drag on inequality in the work force actually acted to raise extreme poverty, reduction exceeding 25%. highlighting some of the employment challenges of the extreme poor. This is consistent with the national trend, In RMs, changes in household demographics and la- though the magnitude is doubled in RMs and every RM bor market participation accounted for more of the except for Distrito Federal displayed this trend. Changes in change in income inequality than at the national level, contributory pensions also contributed to higher extreme while transfers accounted for less. Labor income had a poverty in every RM (and more negligibly, also to higher relatively similar influence at both levels. But other fac- poverty), suggesting that a smaller number of extremely tors had differing impacts. The effect of share of changes poor RM households received contributory pensions. This in the household share of adults was almost 25% in the may indicate a shift towards working aged households in RMs while it was 13% at the national level. The impact of metropolitan areas. the share of employed differs similarly, with a contribution towards the reduction in the Gini of almost 20% in RMs Changes in labor incomes contributed most (69%) to compared with only 3% at the national level. Capital had income inequality reduction across RMs but was not a much larger negative impact on income inequality re- the dominant factor within every RM. For example, duction in the RMs, while transfers and contributory pen- changes in the share of employed adults in the house- sions accounted for half of the importance in RMs as at the hold (labor participation) in the RM São Paulo and share national level. The marked contribution of demographics of adults (dependency ratio) in São Paulo and Porto is particularly important for a country in the advanced Alegre contributed more to reduction in the Gini than la- stages of its demographic transition. A sizeable share of 22 Box 3. Social Programs in Brazil’s RMs The Brazilian Government is implementing multiple programs to fight multidimensional poverty. These programs include Bolsa Família and Segurança Alimentar. Many are run by the Ministry of Social Development (MDS, Ministério do Desenvolvimento Social e Combate à Fome) as part of the Brasil Sem Miséria plan, which began in 2010. The flagship program against poverty, the conditional cash transfer (CCT) scheme Bolsa Família, dates to 2004. It provides R$70/ month money to households in extreme poverty (based on an extreme poverty line of R$70 per month), with an additional R$30-32 for each child in the family. This can be higher based on conditions fulfilled by the family. The purpose of Bolsa Família and other CCT programs is to provide additional income to families that are below a given income threshold, provided that they fulfil certain conditions that help future human capital accumulation and over- all development. In the case of Bolsa Família, these conditions include school attendance for children between the ages of 6 and 17, participation in maternal and child health courses provided by the Health Ministry, and remaining up to date on vaccinations. Furthermore, households must have a child under the age of 17 or be expecting a child. A major portion of the extreme poor in Brazil’s RMs receive benefits from Bolsa Família, but the moderate poor are largely left out in some RMs. Percentages of RM populations receiving benefits from Bolsa Família range from 10% (Curitiba) to 31% (Fortaleza). This represents over 40% of registrants in the CadUnico in each RM. In seven out of the 10 RMs, the extreme poor are the majority of recipients of BF. This means that a large portion of the neediest is receiving social support. However, in RMs in the North and Northeast, which have some of the highest poverty rates, there may be little room left for the moderate poor to also benefit from social assistance. In these RMs, the share of the extreme poor registered in CadUnico but not receiving Bolsa Família benefits is lowest, but the gap between the moderate poor and extreme poor not receiving benefits is largest. In greater or lesser form, this pattern exists across RMs. This may be what is reflected in the reduced access to services and educational attainment of the moderate poor, discussed in section C.D of this report.18 Figure 17. Recipients of Bolsa Família Figure 18. Many of the moderate poor are primarily the extreme poor do not receive benefits 100 60 Share of poor not receiving BF (%) 80 50 Percent of BF recipients (%) 40 60 30 40 20 20 10 0 0 Belem Fortaleza Rio de Janeiro Porto Alegre Salvador Sao Paulo Recife Belo Horizonte Distrito Federal Curitiba Belem Salvador Fortaleza Recife Rio de Janeiro Porto Alegre Belo Horizonte Sao Paulo Curitiba Distrito Federal Extreme Poor Moderate Poor Non-Poor %CadUnico Receiving BF Extreme poor not receiving BF Moderate poor not receiving BF Source: World Bank calculations using Cadastro Único (MDS SAGI online database). Source: World Bank calculations using Cadastro Único (MDS SAGI online database). 18 18  The income aggregate used by the Cadastro Único program varies from that used in the computation of the poverty numbers provided at the beginning of this note. 23 Box 4. Tax incentives and business registration simplification for reducing informality Micro and small enterprises (MSEs) play an important role in Brazil’s economy, employing 67% of the labor force and contributing 20% to the GDP, according to IBGE.19 MSEs are also critical for RMs’ development, employing about 60% of the workers in the largest RMs (RAIS and CAGED, 2013). Studies show that simplification of business registration procedures and streamlining of tax systems encourages MSEs to formalize, while formalization in turn leads to job creation and increases firm productivity, which could have positive implications for poverty reduction and shared prosperity. This pattern is especially important for RMs where informality continues to be a main issue, particularly for the poor and vulnerable. Prior to the mid-1990s, the complexity of the Brazilian tax system presented a serious constraint to MSEs’ growth and development and helped create high informality. To address this problem, the Brazilian Government imple- mented a new simplified tax system for micro (annual gross revenues up to R$120,000) and small firms (up to R$720,000) in 1996.20 Known as the SIMPLES, its goal was to reduce cost of doing business, facilitate tax payments and broaden the tax base. This hope was that this would facilitate growth of small, labor-intensive firms and enable them to compete with large firms. For that reason, SIMPLES explicitly excluded from program eligibility all activities that by law require the employment of professionals from regulated occupations. SIMPLES established a new progressive tax rate system that combined federal, state and municipal taxes into one. Businesses using SIMPLES were able to substitute up to eight taxes and social security contributions for a single tax rate on their annual gross revenues, with rates varying from 3% to 5% for micro-enterprises, and from 5.4% to 7% for small firms (Fan et al. 2008).21 Moreover, while the state and municipal taxes—the Imposto Sobre Circulação de Mer- cadorias e Prestação de Serviços (ICMS) and the Imposto Sobre Serviços (ISS)—were initially not included in SIMPLES, states and municipalities could enter into agreements with the Federal Government to transfer to it the collection of the corresponding taxes through an increase in the SIMPLES rates (Fajnzyblber et al. 2011). Several evaluations of this reform showed a significant positive impact on firm formalization leading to job creation and increased firm productivity (Monteiro and Assuncao, 2006; Fajnzyblber et al. 2011). Monteiro and Assuncao used the 1997 Urban and Informal Economy Survey (ECINF) which covered 40,000 firms in the Brazilian state capitals and metropolitan areas and applied difference-in-difference estimation method. They found that the SIMPLES reform increased formality by 13 percentage points, which in turn stimulated investment and switching of the firms from short-term to long-term projects. Fajnzylber et al., using the 1997 and 2003 ECINF and implementing a regression discontinuity analysis, also found that SIMPLES led to significant increase in formalization of enterprises. Moreover, formalization was associated with job creation, increase in capital intensity, and productivity. Other countries in LAC have implemented innovative reforms to encourage formalization of businesses. For exam- ple, Mexico in 2002 enacted the Rapid Business Opening (SARE) reform, which reduced the average number of days, procedures and office visits required to register a business. Evaluation of that reform using discriminatory analysis 19 20 21 showed that it significantly increased the number of formal businesses and generated wage employment (Bruhn, 2013). 19  http://thebrazilbusiness.com/article/brazilian-tax-simples-nacional. 20  The definition of micro and small enterprises was amended in 2006: a firm was considered micro if its annual gross revenues were equal or less than R$240,000, and it was considered small if its annual gross revenues were between R$240,001 and R$2,400,000. In 2011, the pres- ident of Brazil proposed to increase the ceiling from R$240,000 to R$360,000 for micro firms, and from R$2.4 million to R$3.6 million for small firms. http://riotimesonline.com/brazil-news/rio-business/tax-changes-for-small-businesses-in-brazil/# 21  The taxes and contributions covered by SIMPLES included (1) corporate income tax (Imposto de Renda das Pessoas Jurídicas–IRPJ); (2) tax on industrialized products (Imposto sobre Produtos Insutrializados–IPI ); (3) social security contribution to fund unemployment insurance and other social programs (Contribução para o PIS/PASEP); (4) social security contribution on net profits (Contribução Social sobre o Lucro Liqui- do–CSLL); (5) social security contribution (Contribução para o Financiamiento da Seguridade Social–COFINS); and (6) employers’ social security contributions (Contribução para a Seguridade Social a cargo de Pessoa Jurídica). 24 Figure 19. The real minimum wage increased have supported the transition to formality by lowering substantially between 2004 and 2011 bureaucracy and the tax burden for small enterprises to declare formal employees. (See Box 4.) Still, while all RMs 20 improved in labor market indicators, large variations be- 15 tween RMs persist. As with other dimensions of the anal- Change in minimum wage (%) ysis, the Northeast RMs continue to lag behind the South- 10 east. In 2012, the highest performer in the Northeast was 5 still below the lowest performer in the Southeast in terms 0 of hourly wage, informality, unemployment, and average -5 labor income. -10 Brazil increased the minimum wage, raising it in real 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 terms by more than 70% from 2002 to 2013, directly Source: World Bank calculations using Ipeadata. Note: Change is the percentage change in real terms from January of each year. benefiting workers in the formal sectors in RMs but also those working in the informal sectors. Not only did its population is entering retirement. Yet there remains a this directly benefit the more than 10 million workers paid young population (particular in the North and Northeast) at minimum wage levels, it also helped workers across the whose productive participation in labor markets will de- labor income distribution by raising the wage floor.22 Even fine how shared prosperity is sustained, especially in met- though the minimum wage primarily concerns workers of ropolitan settings which attract young migrant workers. the formal sector, in dynamic labor markets, it also serves as a reference point for informal work and for many of the B. Improved labor market outcomes self-employed. According to IBGE, the change in mini- and policies increased labor incomes mum wage raised the share of wages in the GDP from 46% to 58% between 2003 and 2012.23 Labor incomes increased by more than one third over- all between 2004 and 2012 and increased the most in C. Gender differences persist in labor the lowest deciles. While the average level of labor in- outcomes across RMs come across RMs in Brazil increased by 36%, workers in the lowest deciles of the household per capita income Female labor market income contributed significantly distribution saw their labor incomes increased by much poverty and inequality reduction in RMs. (See Figure more—73% for the first decile and 93% for the second. 20). Though less than the contribution of male labor in- This was not unique to RMs, as labor income also grew come, female labor income still contributed almost one fastest for the lower deciles at the national level. third of the decline in extreme poverty and over 15% of the decline in both inequality and poverty. Similarly, in- A number of factors contributed to raising labor in- creased female labor force participation helped drive comes in the RMs since 2004, chief among them: lower down poverty, extreme poverty and inequality, while unemployment and informality rates as well as high- declining participation rates for men drove up extreme er hourly wages. Efforts to increase access to education poverty. translated into an increased availability of skilled workers in 2012 compared to 2004, as evidenced by an increase of almost one third in the rate of adults with a secondary or 22  The substantial increase of the minimum wage in Brazil also has had direct impact on pensioners, a growing portion of Brazil’s pop- higher level of education. While labor force participation ulation, because there is an automatic link between the minimum remained relatively stagnant over this period, unemploy- pension and the minimum wage. At present all pension recipients— ment in the RMs dropped by almost half to 7%. Echoing which include all people aged 65 and above—receive at least the minimum wage, which is almost 10 times the extreme poverty line this improving labor market environment was the de- (R$70). crease in informality from 31% in 2004 to 22% in 2012. 23  SAE, (2013), Vozes da Nova Classe Média - Caderno #4 : Classe Média e Emprego Assalariado, Brasilia: Secretaria deAssuntosE- Implementation of special tax and registration regimes stratégicos(SAE). 25 Box 5. Minimum wages in Brazil and its RMs Brazil’s first minimum wage was enacted in 1936 by President Getúlio Vargas. Following Courseuil and Servo (2002), Brazilian research on the topic can be divided into three groups: (1) papers that estimate the effect of minimum wage on wage distribution, (2) papers that investigate the relation with income distribution, and (3) papers that analyze other effects on the labor market. One possible challenge for empirically studying the effects of changes in the minimum wage is to determine a coun- terfactual or a control group that was not affected due to the simultaneous nature of changes for all individuals. For this, the literature has tended to concentrate on wage distribution analysis focusing on employed workers or, when analyzing issues related to poverty, doing these using aggregated measures. Barros et al. (2001) use PME data to identify who is affected by the minimum wage. Following individuals from 1995 to 1998, they decompose variation in pover- ty level and show that components of the minimum wage not related to labor markets significantly reduced poverty. Firpo and Reis (2007) show that a continuous increase of the minimum wage above inflation between 2001 and 2005 caused a reduction in wage inequality from 30% to 60%, depending on the inequality measure used. Such effect on inequality is also found in Jales (2012) for the period between 2001 and 2009. In this paper, the author proposes a methodology for estimating the latent distribution of wages in the absence of the minimum wage policy. Using this framework, the author finds that the reduction in wage inequality was accompanied by an increase in unemployment and a reduction of the formal sector of the economy that consequently reduced labor tax revenues. Another topic in the literature begins with the creation of regional wage floors in 2000.24 Five states enacted this policy: Rio de Janeiro, São Paulo, Santa Catarina, Rio Grande do Sul, and Paraná. Moura and Neri (2008) present ev- idence of the impact for the first years of regional wage floors in Rio de Janeiro and Rio Grande do Sul, finding that the floor is not binding for most of the firms paying lower salaries. Courseuil, Foguel and Hecksher (2012b) estimate the impact of wage floors on Parana and São Paulo using PNAD data and the methodology developed by Abadie, Diamond and Hainmueller (2010) known as synthetic control. They find positive effects on income and employment but only for some occupations. The establishment of regional floors also presents opportunity to use additional econometric tools to study poverty and unemployment. It further underlines the necessity of bringing a geograph- ical lens to the analysis, because regional floors can change wage distribution distinctly by state and occupation. Looking at RMs to assess how changing the minimum wage could affect wage distribution for the present analysis, the authors implement a regression discontinuity design (RD) for each of the RMs using the PNAD and following Lee (2008) and Lee and Lemieux (2010). The results show that for all except Rio and São Paulo, there is a peak around the minimum wage value. This shows that a large share of workers have their incomes concentrated at the minimum wage level and therefore changes in its level can substantially affect the labor market for the employed. Given the bunching of workers around the minimum wage line across RMs, a second step in the analysis involves looking at the possible effects this policy could have around the cut-off of wages distribution. The inference is that a worker receiving the minimum wage would expect his or her own wage to be exactly or barely above the minimum wage value. Thus, the authors of the present report look at the cut-off generated by the minimum wage value (wage demeaned equals wage less minimum wage values) to estimate the effect of the minimum wage on formality, schooling years, the probability of the head of household being a man, and access to assets. The results show a strong discontinuity when formality is used as a dependent variable, echoing the results of Jales (2012). This means that the implementation of a minimum wage policy against low wages has the adverse effect of reducing the proportion of formal contract vacancies in the labor market, particularly in RMs where wages are concentrated around the minimum wage or below it. However, no impact is found on other outputs when restricting the analysis to the occupied population. 24 24  Regional floors vary but can’t be lower than the federal mandatory statutory minimum wage. Moura and Neti (2008) document how some firms do not pay the wage floor in spite of the law. 26 Figure 20. Women’s labor income is an important Figure 21. The share of female and male workers contributor to reduction of poverty and inequality by occupational category has remained relatively in Brazil RMs stable since 2004 20 25 Percent of total workers (%) 15yrs+in RMs 10 20 Share of change (%) in RMs 0 -10 15 -20 10 -30 5 -40 -50 R$70 R$140 Gini 0 Senior o cial Agriculture Artisan Clerk Industrial Armed forces Professional Repair/mainten. Services/vendor Technician Women share of occupied Women labor income Men share of occupied Men labor income Other non-labor income Pensions Transfer female 2004 male 2004 female 2012 male 2012 Source: World Bank calculations using PNAD 2004, 2012. Note: “Transfers” are based on imputed values of social program payments. Poverty lines used are R$70 (extreme) Source: World Bank calculations using PNAD 2004, 2012. Note: workers 15 years and and R$140 (moderate). See for details on poverty lines in Brazil. Brazil RMs only. older. Brazil RMs only. Females continue to have worse labor outcomes than males across indicators such as hourly wage, infor- mality, labor force participation, unemployment, and part-time employment. But, since 2004, the differences in rates between males and females of all of these indica- tors except for hourly wage have lessened. However, the fact that the difference in wages slightly increased over the same time period indicates that progress towards gender equity in the labor market is not extending to fi- nancial outcomes. When controlling for factors that may affect wages, such as migration status, union affiliation, weekly hours worked, informality, age, and education, the wages that women living in RMs earned was significantly less than men’s. In fact, being a woman was the largest determinate of the variables considered, and, alarmingly, this negative effect increased, especially for poor women, from 2004 to 2012. (See Annex 15.) Continued gender segmentation in labor markets may also contribute to these differential outcomes, which show a higher share of women in services and a significantly larger share of men in industrial employment (Figure 21). 27 Increased access to a range of services has almost eliminated multi-dimensional poverty Access to basic goods and services is stove. Over 90% of individuals lived in households that approaching universal coverage in RMs had basic sanitation, children enrolled in school, and at least one household member with eight years of educa- C ompared to the national rates, access to basic tion. While those coverage rates are encouraging, they goods and services in the RMs is higher, but also signal the complexity of reducing the remaining the gap between the nation and the RMs is gaps. More targeted and integrated approaches will be closing.25 Rates of child school enrollment, access to elec- required for this task. tricity, and possession of basic assets are similarly high at the RM and the national levels, but RMs are perform- Figure 22. Access to services is higher in the RMs ing slightly better than the nation in terms of safe water, but Brazil as a whole is catching up shelter, and educational attainment (See Figure 22). RMs have markedly greater sanitation coverage than national 100 figures, with rates close to 15 percentage points above the 80 Percent with access (%) national level. Compared to 2004, the access gap between RMs and Brazil has narrowed across indicators, shrinking 60 the most in terms of assets and safe water. 40 As of 2012, over 90% of the population in RMs had ac- 20 cess to basic goods or services. Over 97% of individuals 0 lived in households that had quality shelter, safe water, Education Assets Electricity Sanitation Shelter Water Enrollment electricity, and possessed two out three of the following three goods: refrigerator, telephone, and clean cooking All 2012 RM 2012 All 2004 RM 2004 25  See Annex 9 on the calculation of the MPI. Source: World Bank calculations using PNAD 2004, 2012. 28 Figure 23. The Northeast remains the poorest region, but rates have fallen significantly 35 2004 35 2012 30 30 Share of Households (%) 2012 Share of Households (%) 2004 25 25 20 20 15 15 10 10 5 5 0 0 Brazil RMs Belém Fortaleza Salvador Rio de Janeiro Curitiba Porto Alegre Distrito Federal São Paulo Belo Horizonte Brazil RMs Recife Belém Fortaleza Recife Salvador Distrito Federal Belo Horizonte São Paulo Porto Alegre Curitiba Rio de Janeiro Transient Not poor but deprived Chronic Transient Not poor but deprived Chronic Source: World Bank calculations using Cadastro Único (MDS SAGI online database). Source: World Bank calculations using Cadastro Único (MDS SAGI online database). While monetary poverty has fallen significantly in Bra- tional level. While some of the RMs have had low levels of zil RMs, the reduction is more striking in multi-dimen- multi-dimensional poverty since 2004, such as São Paulo, sional poverty. Chronic poverty—which combines others such as Fortaleza and Belém achieved this low level these two forms of deprivation—has been virtually of chronic poverty through more significant reductions of eliminated and the remaining monetary poverty now around 7 percentage points (See Figure 23). Since mone- constitutes the main challenge, particularly in the tary poverty in RMs is also low, the largest group of peo- Northeast. While monetary poverty remains the main in- ple who are poor only in monetary terms, called the tran- dicator for poverty reduction, many countries, including sient poor—makes up only 3.8% of the population. Still, Brazil,26 recognize the importance of looking at poverty not only do those percentages represent large numbers from a multi-dimensional perspective, which measures of people due to Brazil’s size, but the deprivation criteria access to different types of goods and services. (See Box (threshold cut-offs) may need to be revised to better ac- 6 and Annex 9 for detailed descriptions of this approach.) count for the specificities of vulnerability in a metropoli- The indicators and deprivations criteria used in the pres- tan setting.28 ent analysis are based on previous work done on chronic poverty in Brazil (World Bank, 2013).27 B. Equitable access to quality education remains a problem MPI analysis for the RMs shows that rising incomes and a concerted effort by the Brazilian government have Children in Brazil’s RMs have better opportunities nearly eradicated chronic poverty within the large than in other settings, but equitable access to quality metropolitan regions. Its incidence fell below 0.5% in all education and sanitation is still not universal. Using ten of the RMs considered, and was also low at the na- 28  A number of Latin America countries such as Mexico and Co- 26  Brazil’s Ministry of Social Development (MDS), which is charged lombia have designed their own MPI indicators to monitor poverty. with the Brasil sem Miséria plan and the Bolsa Família program, clear- This may provide Brazil with relevant experience to follow suit in ly states that poverty is a multi-dimensional phenomenon—even the future. (Santos, M., 2013: Measuring Multi-Dimensional Poverty though the ministry has primarily used administrative income pov- in Latin America: Previous Experiences and the Way Forward, OPHI erty lines (R$70 and R$140) to set and monitor its poverty objectives. Working Paper N.66.) The Santos paper analyzes methods most used See MDS (2013) : Programa Bolsa Família : uma década de inclusão e ci- by researchers to select poverty dimensions; existing data or con- dadania; MDS/ SAGI (2014) : Pobreza multidimensional: subsídios para vention; theory, namely the selection of dimensions based on im- discussão à luz do MPI/OPHI. plicit or explicit assumptions of what people value or should value; 27  See Annex 9 on the Multi-Dimensional Poverty Index (MPI) public “consensus,” ongoing deliberative participatory processes, and methodology. empirical evidence regarding people’s values. 29 Figure 24. The HOI is higher in RMs than in the nation but still low in school quality, sanitation, and home Internet 100 80 60 HOI 40 20 0 Attendance Grade Sanitation Water Electricity Cellphone Internet 2012 RM 2012 All 2004 RM 2004 All Source: World Bank calculations using PNAD 2004, 2012. Box 6. Measuring non-income-based levels of social welfare in metropolitan Brazil Though its philosophical roots reach back to Aristotle, non-monetary approaches to welfare only became prevalent within economic literature in the late 20th century, following Amartya Sen’s seminal Tanner Lectures on basic ca- pability equality delivered at Stanford University in 1979. Since then, various approaches to measuring well-being through non-monetary indicators have been applied to assess poverty and inequality in developing countries. One such indicator is the Multidimensional Poverty Index (MPI), developed by Sabina Alkire and James Foster. The MPI examines the intensity of deprivation across indicators that are believed to be closely linked with well-being, such as education, housing conditions, and the accumulation of assets. When a household lacks access to a certain number of these basic services or goods, it is considered to be “multi-dimensionally poor.” (See Annex 9.) Alkire et al. (2014) argue that the MPI can directly compare well-being outcomes between urban and rural locations. However, evidence suggests that for a more precise understanding of well-being, the MPI should be tailored to the location. Generally, income-based poverty lines correspond to different levels of well-being in urban and rural settings due to different market structures in those areas and generally underestimate urban levels of poverty. This makes the MPI an important component to compare settings.29 However, to characterize urban poverty beyond its comparability with rural settings, the MPI should be tailored to factors relevant to life in cities. Having “access” to a particular service is not enough in an urban setting, where population density may impinge on quality of such services as sewage, water and trash, and where lapses in service are more problematic. To enhance the urban MPI, indicators of quality of life in urban settings that are not as applicable in rural settings, such as home eviction rates, lack of transportation, and exposure to crime should be considered. However, to date there is no good repository of such information. In order to tailor the MPI to urban settings in Brazil, more and better data is needed. The national household survey (PNAD) is limited in sample size and scope of questions asked during the interview, while the Census faces a similar issue of question scope and comes out only every 10 years. Data being gathered by municipal and state govern- ments and service providers are therefore key to better tracking and understanding basic well-being in urban Brazil. Policy makers need to adopt systems to easily share and connect these data so as to create a more comprehensive picture of well-being at a more disaggregated geographic level. Alternatively, specific surveys can be undertaken in order to directly collect this type of data. Since 2008, the state of Minas Gerais has been doing this in its Travessia 29 30 program, which uses the MPI to target areas most in need of funding and technical assistance to provide services and access to housing.30 29  See extract of DANIDA. 2002. “Improving the Urban Environment and Reducing Poverty.” Workshop paper. Copenhagen, Denmark from http://web.mit.edu/urbanupgrading/urbanenvironment/issues/how-much-poverty.html. 30  Tribunal de Contas do Estado de Minas Gerais. 2011. “Relatorio de Auditoria Operacional: Programa Travessia”. Belo Horizonte. http:// www.tce.mg.gov.br/IMG/Travessia.pdf. 30 Box 7. Quality of education indicators have improved in RMs All RMs increased average years of education and reduced educational inequality between 2004 and 2012. Across metropolitan Brazil, average years of education increased by 1.1 years. In 2012, the Distrito Federal and the largest RMs, São Paulo and Rio de Janeiro, achieved the highest average educational attainment rates (more than 9 years), while Recife (8.2 years) and Fortaleza (7.6 years) had the lowest. All RMs performed better than their respective states in both 2004 and 2012, with the rates differing most for the middle performers. In general, the RMs with high- er levels of educational attainment had lower levels of educational inequality.31 (See Figure 25.) This indicates the importance of increasing educational attainment, particularly in lagging RMs such as Recife and Fortaleza, where the education Ginis were 0.32 and 0.36, respectively. Grade repetition and drop-out rates have improved across RMs but there is still variation among RMs. From 2005 to 2012, every RM lowered its rates of dropouts and grade failures for elementary education. On average, RMs cut the dropout rate by half to reach just 3% and decreased the grade failure rate by 4 percentage points to 10% in 2012. (See Figure 26.) These rates are similar to those seen across all of Brazil. In 2012, RMs with the highest grade failure rate also had the highest dropout rates. This was notably true for Salvador, which had a grade failure rate more than 60% higher than the overall RM rate and, despite improving the most over the time period, retained one of the highest dropout rates across RMs. The reverse applied in the best-performing RMs. For instance, São Paulo’s dropout rate was under 1% and its grade failure rate was half that of all Brazil RMs. The reason why some RMs (notably São Paulo) have lower failure rates could be related to a policy called Progressão Continuada (Continued Progressions) or Aprendizagem em Ciclos (Learning in Cycles). In São Paulo, this policy allows students to fail only at the end of a school cycle (fifth grade or ninth grade). The program has been effective in cutting drop-out rates, but its impact on learning is controversial. Looking at an index of school infrastructure and teacher training rates shows that, on average, schools located in RMs improved and performed slightly better than schools overall. The main reason is that the RMs’ municipal schools perform better than municipal schools overall. Differences between the performances of state schools are more limited. Between 2007 and 2012, school infrastructure and teachers’ training rates improved across RMs. Especially impacting training was a substantial increase in teachers’ access to higher education. As of 2012, São Paulo and Belo Horizonte performed well in both infrastructure and teacher training. Rio de Janeiro performed better than the Brazil RM rate in infrastructure but below the Brazil RM rate in training. Meanwhile, RMs located in the Northeast all performed below the Brazil RM average in both. Figure 25. Educational attainment and education Figure 26. Grade repetition and dropout rates inequality improved across RMs declined across RMs 12 0.45 25 0.40 Average years education (25yrs+) 10 20 0.35 Failure/dropout rate (%) 8 0.30 15 0.25 6 0.20 10 4 0.15 0.10 5 2 0.05 0 0.00 0 Brazil RMs Brazil RMs Brazil Belem Distrito Federal Sao Paulo Rio de Janeiro Curitiba Salvador Porto Alegre Belo Horizonte Recife Fortaleza Salvador Rio de Janeiro Belem Recife Belo Horizonte Fortaleza Sao Paulo RM 2012 RM 2004 Gini2012 Gini2004 Failure 2012 Failure 2005 Dropout 2012 Dropout 2005 Source: World Bank calculations using PNAD 2004, 2012. Source: Ministry of Education. Note: Elementary school rates. Note: Adults age 25 yearsold and up used in calculations. 31 31  Educational inequality is measured by the Gini index where 1 is total inequality and 0 is total equality of years of education attained by the adult (25 years+) population. 31 Box 7. Quality of education indicators have improved in RMs (cont.) Still, inequity in access to quality education – a key vector of upward mobility – is particularly concerning in RM set- ting. More than 60 percent of the upper class and over 15 percent of the middle class children go to private schools in the largest metropolitan regions of Brazil (Figure 27). Not surprisingly, these rates are even lower for the children residing vulnerable and poor households, hovering respectively at less than 10 and 5 percent. Across metropolitan areas, opting for private school is the most prevalent in the North and Northeast RMs, where close to 90 percent of the upper class send their children to private schools and the rates even increase for the poor and vulnerable households. These results suggest that school quality is a particularly prevalent issue in the RM of Recife, Fortaleza and Salvador, although private school enrollment rates are also significant in the RM Belém and Rio. Figure 27. More primary school children are enrolled in private school across income group in RMs (%) (2013) 100 Poor Vulnerable Middle Class Upper Class 80 Percent children enrolled in private school 60 40 20 0 National Urban RM RM RM RM RM RM Belo RM Rio de RM São RM RM Porto Distrito (Non-RM) Average Belém Fortaleza Recife Salvador Horizonte Janeiro Paulo Curitiba Alegre Federal Source: World Bank Calculation, PNAD 2013. the Human Opportunity Index32 to assess coverage and eral) the HOIs for grade progression on time and access equity in children’s access to basic services (opportunities) to sanitation are relatively high, while they are relatively shows that children in all RMs have more opportunities low in other RMs such as Fortaleza and Belém. The RMs than at the national level, ranging from 81 (out of 100) in of Salvador and Rio de Janeiro present a mixed picture, RM Fortaleza to 92 in RM São Paulo. These numbers com- having the two lowest HOIs for grade completion on time pare to an index of 71 in all of Brazil. High performance on yet relatively good (in comparison to other RMs) HOIs on the composite HOI is largely driven by the near universal sanitation and home Internet. coverage rate in RMs for access to safe water, electricity, and child school enrollment (See Figure 24). The worst-per- Inequality in coverage captured by low HOIs for on-time forming indicator across RMs was for grade progression on grade progression and sanitation highlights a need for time. The average rose by only about 2 points since 2004, more targeted interventions in RMs experiencing these compared to about 10 points for sanitation and 21 points low HOIs. Indeed, unequal distribution of coverage based for home Internet. This is particularly concerning because on circumstances (such as parental education or income) not only was grade progression on time the worst-per- drives the low results for grade progression on time and forming indicator, but it has not been improving. sanitation in the RMs in question. The penalty for inequal- ity of grade progression on time was the highest among There are differences between the RMs both in terms opportunities for the RMs and was the same as the nation- of overall HOI and particular lagging dimensions. al average (6.4 points). The analysis shows that the edu- While in RMs such as São Paulo, and Brasilia (Distrito Fed- cation of the household head and household per-capita income continued to be strong determinants in the in- 32  The HOI is a measure ranging from 0 to 100 of the coverage rate equality of distribution of basic opportunities among chil- of children’s access to basic opportunities adjusted for the equity of distribution based on uncontrollable circumstances such as gender, dren. It could therefore hinder intergenerational mobility race, and parents’ income. For a comprehensive explanation of how in RMs as in the rest of the country. HOI is calculated, see Annex 10. 32 Spatial inequities are pronounced within RMs - heightened by constrained mobility A. Core municipalities have lower levels of core: access to sanitation and adults’ rates of secondary poverty and higher access to services than education. But rates of school enrollment are general- peripheries but are also more unequal ly similar between cores and their inner peripheries. Access to sanitation is lower in every RM outside the core. I ncome inequality within core municipalities is This pattern is most marked in the Northeast, where the consistently higher than within inner peripher- core-periphery difference is over 20 percentage points. ies. Census data from 2010 shows a clear pattern of Similarly, adults (15 years and older) living in the core are higher levels of income inequality in the core municipal- more likely to have higher secondary education rates, ities of RMs (i.e., the state capitals) than in the rest of the with an over 10 percentage point difference between the municipalities of the RMs. (See Figure 28.) Among the six core and inner periphery in each RM with the exception RMs in the Northeast and Southeast, core municipalities of São Paulo, which differed by only half as much (4.5 per- have a Gini Index that is 10 points higher, on average, than centage points). By contrast, school enrollment does not in their inner peripheries, with the lowest (Salvador) and show large differences, and, as of 2010, was close to 90% highest differences (Fortaleza and Recife) in the North- across geographical areas and across RMs. With the inner east. That inequality is higher in the cores of these RMs is peripheries growing their populations at a faster rate than not surprising since density brings together people from the cores, improving these access rates will only become various backgrounds, but it is this density that makes low- more important (See Figure 29). ering inequality especially critical given the greater likeli- hood of physical interaction between people from vary- Workers earn less and are more likely to be informal ing economic levels. outside the cores of the RMs. In the inner peripheries, hourly wages are 40% lower than in the cores, as a sim- In terms of access to services and educational attain- ple average across the Northeast and Southeast RMs. In ment, two measures are systemically lower outside the addition, informality is always higher outside the cores, 33 Figure 28. Differences between cores and peripheries are notable across RMs with the exception of some indicators in São Paulo 100 Di erence of inner-periphery 50 from core (%) 2010 0 -50 -100 Fortaleza(NE) Recife(NE) Salvador(NE) Belo Horizonte(SE) Rio de Janeiro(SE) São Paulo(SE) Extreme poor HC Moderate (non-extreme) poor HC Vulnerable HC Gini Sanitation rate Secondary+ rate Informality rate Hourly wage(R$2012) Source: World Bank Calculation, Census 2010. Figure 29. Inner peripheries are growing the fastest across RMs 40 Core Inner-periphery Outer-periphery 35 30 between 2000 and 2010 Population growth (%) 25 20 15 10 5 0 Belem Fortaleza Recife Salvador Belo Rio de Sao Paulo Curitiba Porto Distrito Horizonte Janeiro Alegre Federal Source: World Bank calculations using IBGE. though in São Paulo the difference is small. The Oaxa- tom and the top of the distribution. This means that both ca-Blinder re-centered regression (RIF) run for seven RMs the vulnerable and better-off residents of peripheral mu- confirms those findings (See Annex 14). Median labor in- nicipalities are working in lower quality jobs and receiving come per capita in the inner peripheries of the RMs is low- lower wages, regardless of formality, than the core, so that er than in the metropolitan core, ranging from -30% in the improved mobility matters for the full spectrum of the in- RM São Paulo and Salvador to close to -50% in the RM For- come distribution. taleza.33 The gap for the bottom of the distribution (10th quantile) is particularly pronounced in the RMs Rio, For- B. The effect of urban mobility on these taleza, Salvador, and Belo Horizonte (all over 20%). When spatial differences warrants further analysis comparing an RM to its outer periphery, a stark contrast is observed between RMs experiencing small to null differ- Better urban mobility will be critical to connecting the ences (the RMs Rio, São Paulo, and Belo Horizonte,) and RMs’ most deprived inhabitants and to job opportuni- the rest experiencing large differences, for both the bot- ties and to overcome other spatial mismatches. As one of the key issues of the June 2013 protests, urban mobility undoubtedly constitutes a priority issue for Brazil policy 33  This measure excludes pensions and benefits received from so- cial programs like Bolsa Família and Benefício de Prestação Continuada. makers—and one directly related to shared prosperity. 34 Box 8. Crime in metropolitan Brazil RMs in the North and Northeast had the highest rates Figure 30. Homicide rates have soared in Fortaleza of violent crime of RMs in Brazil in 2012 (Figure 30). and declined in Recife in recent years Crime levels rose in four out of ten RMs, remained roughly stagnant in 3 RMs, and fell in another 3, with 90 RM 2008 RM 2012 Core 2012 strongest decline in Recife. The RM Fortaleza had the 80 Homicide Rate (per 100,000) highest rate of all, with 69 homicides per 100,000 resi- 70 dents per year, while the RM São Paulo had the lowest 60 50 at 17. Looking within the RMs, there is no clear pattern 40 between center and periphery crime rates. In general, 30 crime rates in the core of the RM were similar to the 20 overall RM, although the core had much higher rates 10 in Fortaleza, while in Belem and Rio de Janeiro, crime 0 Fortaleza (NE) Recife (NE) Salvador (NE) Belem (N) Curitiba (S) Distrito Federal Porto Alegre (S) Rio de Janeiro (SE) Sao Paulo (SE) Belo Horizonte (SE) affected the core slightly less than the overall RM. Fortaleza, once one of the safest RMs in terms of homi- cide, is now the most dangerous, while Recife, once the Source: World Bank calculations using CEBELA, 2014. “Mapa da Violencia”; Ipeadata/IBGE. most dangerous RM in Brazil, has become the safest. The murder rate doubled dramatically in the RM Fortaleza, while in the RM Recife, it fell by close to one third of the 2008 rate. Recife’s success in reducing homicides may reflect Pernambuco state’s Pacto Pela Vida, a public safety initiative that began in 2007 and has 138 different programs. The state’s large reductions in poverty and inequality (the largest across all RMs) may also have played a role, as various studies have linked poverty and inequality to crime rates. (See Fajnzylber, Lederman, and Loayza, 2002 and Poveda 2011.) Given the changing dynamics of crime and the relatively high homicide rates across the country, more research into security programs such as the Pacto Pela Vida or Rio de Janeiro’s Unidade de Polícia Pacificadora (UPP – Pacifying Police Unit) and their effect on social welfare would help policy makers tackle both issues of crime and of poverty and inequality facing Brazilians. While governments at the national and state level have started to invest in better monitoring and evaluation systems of crime and violence, more remains to be done to evaluate the cost effectiveness of the numerous citizen security programs developed over the past couple of years. To do so will require not only information pertaining to the incidence of crime and violence but also disaggregated information on the programs themselves such as geo-refer- enced funding amounts, police force capabilities, crime reporting, and the dates of implementation or scale-up of any actions to account and disentangle the effects of those interventions. The fact that educational levels and returns to education recent analysis on jobs displacement and joblessness du- are lower outside the core and lower still outside the RM, ration in the United States (Andersson et al. 2014) gives as shown in the OB analysis conducted for the stand-alone further credence to spatial mismatch. It found that better notes, suggests that people are constrained in their ability job access significantly decreases the duration of jobless- to commute to better labor opportunities in the centers of ness among lower-paid workers, particularly traditionally the RMs. A growing consensus exists that poor job accessi- more challenged groups (non-Hispanic blacks, women bility contributes to poor labor outcomes for lower-skilled and older workers). The example of the RM Recife (Figure ethnic minorities (Kain 1992, 2004; Ihlanfedlt and Sjoquist, 14) where the localization of homicides closely matches 1998; Gobillon et al, 2007). On the other hand, no consen- that of bus stops also suggests that non-physical barriers sus exists on the magnitude of the spatial mismatch and can constitute additional bottlenecks to the mobility of on which groups most are affected (Ihlanfedlt, 2006). More vulnerable groups (Figure 31). Analysts looking at pover- 35 ty, social protection, and labor market in Brazil stand to not necessarily mean that commuting time has no rela- gain a more refined strategic policy lens if they integrate tion to poverty or economic growth. Rather, concerning this dimension into their work. The expansion of big data the differences between the core of the RM and its inner analysis in the country could help in this task.34 periphery, this could indicate that lack of reliable trans- portation may keep people working closer to home at the While urban mobility is undeniably critical to well-func- cost of lost earnings from higher wages in the core. On the tioning labor markets in metropolitan settings, con- other hand, the categorical construction of the commut- straints on commuting time variables limit the analy- ing variables in both the census and the PNAD may also sis. This report highlights spatial features of poverty and not adequately capture household commuting patterns shared prosperity in metropolitan and peri-metropolitan and ways in which mobility could be a bottleneck to ac- settings. But commuting time variables captured by both cess to services and labor market opportunities (Anders- the PNAD and the census (long-form) do not provide sig- son et al. 2014). nificant results, either in differences in commuting time between residents of different areas of the metropolitan Figure 31. Public transportation in Recife appears regions or across time. (See Annex 13.) Still, previous stud- plagued by crime and is limited outside the core ies have shown that improved transportation – notably through better integration of different transportation modes (ex. train and buses) can increase access to job op- portunities (Pereira and Schwanen, 2009). Taking the case of the metropolitan region of Rio de Janeiro, in 2010 a res- ident of Rio’s periphery spent an average of 86 minutes commuting every day (close to a ¼ of its workday) and the RM Rio de Janeiro has the lowest proportion of workers taking less than half an hour to get to work (around 40 percent in 2009). This situation curtails the access of poor and vulnerable to better labor opportunity - 55 percent of the region’s jobs are located in the Municipality of Rio de Janeiro – notably by reducing the opportunity cost to commute. Incidentally, the lack of significant results does Source: Pernambuco Open Data—Grande Recife; PPV; IBGE. Box 9. Towards more inclusive urban transport: The RMs São Paulo and Rio Transport infrastructure figures prominently in Brazil’s development strategy. In 2013 the State of São Paulo allocat- ed almost two thirds of its fiscal budget to the sector, with over 22% alone earmarked for the expansion and oper- ation of the Metro in the São Paulo Metropolitan Region (RMSP). Road and rail investments in urban mobility rank among the top investment priorities in Brazil’s large cities. These programs have traditionally focused on improving efficiency, reducing emissions, and promoting the competitiveness of cities. But more recently, policies aimed at addressing the needs of the urban poor and other underprivileged groups are increasingly gaining a salient role in the transport agenda. Corseuil and Pereira (2012a) shows that the monthly transport costs of the lowest decile represented over 20% of their income v. 15% on average, and 13% for the highest decile in 2009. 34  In particular, the use of administrative registries such as the RAIS or the CadUnico in connection with the census and commuting data (using GTFS data of public transportation) offers much potential to refine the analysis in metropolitan and peri-metropolitan settings. Exploratory cross-sectoral discussions confirm the possibility of such analysis and its relevance for Brazil. (TBC) 36 Box 9. Towards more inclusive urban transport: The RMs São Paulo and Rio (cont.) Over the last 20 years, state and metropolitan governments have combined policy and infrastructure initiatives to provide the more disadvantaged with improved access to opportunities. On the policy side, municipal and state au- thorities subsidized peripheral residents’ using the integrated fare scheme known as “Bilhete Unico Integrado” (BIU). As a result, the share of household income in RMSP dedicated to transport fell by half for the moderate and extreme poor to 9% and 13%, respectively. The BUI policy and associated investments in public transportation have enabled low-income people to travel more often and access previously unreachable districts in the search for higher-paying and better-quality jobs and opportunities. The impact evaluation for the inter-municipal BUI in the Metropolitan Region of Rio de Janeiro (RMRJ) shows, for instance, that the policy increased the number of formally employed people in the municipalities with BUI, particularly in peripheral areas. On the infrastructure side, significant investments in the suburban rail/metro network increased physical accessi- bility to formal jobs, particularly in peripheral areas. In the RMSP, 150,000 low-income families gained access, on average, to an additional 2.5 potential jobs within a 45-minute radius as a direct effect of Metro Line 4 implemen- tation. In addition to large-scale transport projects, Brazilian cities have in the last few years experimented with models of mobility consisting of medium- and low-capacity systems aimed exclusively at integrating low-income no-go areas into the fabric of the city. Examples of such “pro-poor” transport infrastructure can be found in the RM Rio, such as the aerial cable car system (teleferico) in the informal settlements of Complexo do Alemão and Providên- cia, and the Rubem Braga elevator connecting the Metro with a favela in a central district. On a daily basis, each of these modes serves between 8,000 and 10,000 people residing in low-income neighborhoods. While more nuanced analysis is needed to assess the impact on job creation35, land value and uses, and other welfare outcomes, these systems have undoubtedly enhanced citywide accessibility and transport mode choices in formerly underserved communities. Perhaps more importantly, the new integration and connectivity have increased a sense of belonging among residents of the favelas. Table 2: Households’ monthly expenses on public and private transportation per income decile (POF 2009) Urban transport expenses (R$) Share of income (%) Household Household income per Public Private Total Income (R$) Public Private Total capita transport transport transport transport 1st decile 54.82 61.34 116.16 532.03 10.30 11.53 21.83 2nd decile 64.75 97.14 161.90 917.20 7.06 10.59 17.65 3rd decile 71.03 118.74 189.77 1,165.42 6.10 10.19 16.28 4th decile 83.82 164.72 248.54 1,490.95 5.62 11.05 16.67 5th decile 82.69 213.93 296.63 1,730.79 4.78 12.36 17.14 6th decile 88.07 262.23 350.30 2,102.56 4.19 12.47 16.66 7th decile 89.47 350.45 439.92 2,573.93 3.48 13.62 17.09 8th decile 86.57 454.56 541.14 3,257.67 2.67 14.04 16.71 9th decile 83.07 727.52 810.59 4,669.59 1.78 15.58 17.36 10th decile 76.66 1,426.78 1,503.45 10,872.28 0.71 13.12 13.83 Average 76.89 427.44 506.33 3,211.25 2.46 13.31 15.77 Source: IBGE (2010), IPEA (2012). 35 35  On urban transport expansion and employment see the case of the United States in Baum-Snow (2014). 37 Conclusion T he last decade brought significant welfare gains to metropolitan Brazil. Solid growth in these economic powerhouses helped drive extreme poverty down from 4.7% in 2004 to 2.3% in 2012 and pov- erty from 14.7% to 4.7%. The incomes of the bottom 40 grew by 7.7% a year, compared to 4.7% for the RM mean. Moreover, there were signs of income convergence, as the traditionally poorer Northeastern RMs experienced faster gains in both poverty reduction and shared prosperity than the other seven RMs. Overall, growth and higher la- bor incomes were the catalyzing force behind these gains, accounting for 95% of the fall in extreme poverty. Howev- er, in the three RMs in the Northeast (Fortaleza, Salvador and Recife), redistribution also played an important role. Still, while poverty has fallen to just 4.7% in metropol- itan Brazil, a significant number of people (2.6 million) remain poor and an even larger number of people (6.5 million) are vulnerable to sliding back into poverty. Furthermore, despite their impressive gains, the three northeastern RMs and northern RM Belém remain the poorest, with poverty rates exceeding 8%, almost double the average for metropolitan areas. These four RMs also have the lowest levels of access to sanitation and second- ary education, and the highest levels of private school en- rollment for middle and upper class children—suggest- ing significant quality problems in their schools. While these RMs made important progress in reducing income inequality, there are important spatial differences within them. Poverty remains higher in their inner peripheral municipalities compared to their core capital cities and la- bor market opportunities less attractive (lower wages and higher informality). The socioeconomic gaps between 38 these four RMs and nearby urban and rural areas are likely per classes on one side and the poor and vulnerable to continue acting as a strong pull factor, attracting more on the other, especially the moderate poor. While ac- poor and vulnerable households from the interior of the cess to basic services has greatly improved, the analysis state and the periphery, thus spurring further RM popula- confirms that challenges remain in key sectors such as tion growth. This is in contrast to the southern RMs, where sanitation and education, both in coverage and quality. there remain few significant welfare disparities between Understanding those challenges and the bottlenecks that each RM and the rest of its state. cause them is critical to sustaining the gains achieved and avoiding a permanent disconnect between those that can The first area relates to better understanding the chal- afford to opt out or compensate for quality shortcomings lenges that the poor face in accessing quality employ- in public systems and those who can’t. In addition, the ment even though they live in economic powerhouses. gaps between the extreme and moderate poor warrant The extreme poor are virtually excluded from the labor further investigation and could have important policy market and hence the benefits of growth. In RMs, over implications. As noted, those gaps may indicate that the 80% of them are unemployed, with the rates in the RMs moderate poor face specific challenges in access to ser- São Paulo and Rio de Janeiro exceeding 90%. Given Bra- vices due to a prioritized policy focus on extreme poverty zil’s unprecedented increase in formal sector employment and may also include some transient poor that have lost and historically low unemployment, this disconnect of the their employment. poor from labor markets and stubbornly high levels of in- formality among the vulnerable are of deep concern. In a The fourth area refers to the need for further explo- context of possible slower growth for Brazil in the future, ration of the spatial dynamics of RMs—and their re- it will be crucial to create better performing and more in- gional specificities, and the role played by mobility clusive labor markets to sustain the gains achieved so far. in effective access to services and labor opportunities for all metropolitan residents. This is especially true giv- The second area pertains to better understanding en that the inner-periphery municipalities are growing what caused the differences in inequality reduction in population faster than the cores across RMs, requiring rates observed across RMs—particularly in the North- increased coordination between metropolitan municipal- east. While two of the Northeastern RMs, Recife and For- ities. Indeed, inner periphery municipalities are symbiotic taleza, achieved the highest reductions in inequality over with their core municipalities and efficient urban mobility the time period (from 0.63 Gini points to 0.50 and from is key to development. The center of the RM consistently 0.60 to 0.51 respectively), the third (Salvador) is falling be- displays better outcomes—with the exception of inequal- hind in this dimension. Further investigation behind those ity—suggesting a need for better connectivity to allow differences could help shed light on how to sustain and different municipalities to contribute and share their deepen inequality reduction all over Brazil. growth potential. Constrained mobility of the metropol- itan poor and vulnerable lowers their capacity to leverage The third area concerns the persisting gaps in service the economic opportunities potentially offered by metro- delivery and quality—between the middle and up- politan regions. The diagnostic highlights four areas where more research is warranted to help further refine policies aimed at reducing poverty and vulnerability and enhancing shared prosperity 4. Understanding RMs ' 1. Understanding the 2. Understanding the 3. Understanding service spatial dynamics , their challenges faced by the poor di erences in inequality delivery gaps in both access regional speci cities and the in accessing quality reduction between RMs and quality role played by urban mobility employment in e ective access to services and labor opportunities 39 Blinder, A.S. 1973. “Wage Discrimination: Reduced Form References and Structural Variables.” Journal of Human Resources, 8, 436-455. Bruhn, M. 2013. “A tale of two species: Revisiting the ef- Abadie, A., A. Diamond, J. Hainmueller. 2010. Synthetic con- fect of registration reform on informal business owners in trol methods for comparative case studies: estimating the Mexico.” Journal of Development Economics 103: 275 effect of California’s tobacco control program. Journal of the American Statistical Association, v. 105, n. 490, June 2010. Chao, L. and J. Lyons. 2013. “Bill Comes Due for Brazil’s Middle Class.” The Wall Street Journal. October 8, 2013. Abou-Alsamh, R., 2012. “Brazil Shows How Too Much http://online.wsj.com/news/articles/SB100014240527023 Consumer Debt Can Threaten an Economy.” International 04795804579097412611960306. Business Times. July 14, 2012. http://www.ibtimes.com/ brazil-shows-how-too-much-consumer-debt-can-threat- Centro Brasileiro de Estudos Latino-Americanos (CEBELA). en-economy-723428. 2014. “Mapa da Violencia.” Ipeadata/IBGE. Abramo, P., 2003. Quatro notas preliminares sobre a teoria Corseuil, C., Pereira, R., 2012a. Gastos das familias brasilei- econômica da favela: localização residencial e mercado ras com transporte urbano público e privado no Brasil: uma imobiliário, in. P. Abramo, ed. A Cidade e a Informalidade : análise da POF 2003 e 2009, IPEA Research Paper N.1803 O Desafio Das Cidades Latino-Americanas. Rio de Janeiro, Brazil. Corseuil, C., Foguel, M., and Hecksher, M. 2012b. Efeitos dos pisos salariais estaduais sobre o mercado de trabalho: uma Aguilar, J. and T. Yepes. 2013. “Location of the Poor: Neigh- nova abordagem empírica. borhood versus Household Characteristics: the Case of Bogotá.” GDP-IDB Project “Urbanization and Develop- Corseuil, C. H., L. Servo, L. M. S. 2002. Salário mínimo e ment: Delving Deeper Into the Nexus.” bem-estar social no Brasil: uma resenha da literatura. Texto para Discussão, n. 880, Ipea, Brasília. Azevedo, J.P., S. Franco, E. Rubiano, and A. Hoyos. 2010. “HOI Stata Module to compute Human Opportunity In- Da Mata, D. Deichmann, J., Henderson, V., Lall, S. 2005. dex.” Statistical Software Components S457191, Boston Determinants of city growth, Brazil World Bank Policy Re- College Department of Economics. http://ideas.repec. search Working Paper. org/c/boc/bocode/s457191.html. This version 1.7. Dang, H.A., P. Lanjouw, J. Luoto and D. McKenzie. 2011. Azevedo, J.P., V. Sanfelice, and M. C. Nguyen. 2012. “Shap- “Using repeated cross-sections to explore movements in ley Decomposition by components of a Welfare Measure.” and out of poverty.” Policy Research Working Paper Series World Bank. (mimeo). 5550, World Bank. Baum-Snow, N., 2014. Urban Transport Expansions, Em- Duranton. G. 2014. “Growing through Cities in Developing ployment Decentralization, and the Spatial Scope of Ag- Countries.” Policy Research Working Paper Series 6818, glomeration Economies, NBER working paper, February World Bank. Becker, G. S. and N. Tomes. 1979. “An equilibrium theory of Fajnzylber, P., W. Maloney, and G. Montes-Rojas. the distribution of income and intergenerational mobili- “Does formality improve micro-firm performance? ty.” Journal of Political Economy 87(6), 1153–89. Evidence from the Brazilian SIMPLES program.”  Jour- nal of Development Economics  94.2 (2011): 262-276. Barros, R.P.. de, Corseuil, C. H., Foguel, M., Leite, P. Uma avaliação dos impactos do salário mínimo sobre o nível de Fajnzylber, P. (2001). “Minimum Wage Effects Throughout pobreza metropolitana no Brasil. Economia, v. 2, n. 1, p. 47- the Wage Distribution: Evidence from Brazil’s Formal and 72, 2001. Informal Sectors.” Anais do XXIX Encontro Nacional de Economia. Bianchi-Santarossa, R., A. Castaneda, A. Fruttero, S. Lach, L.F. Lopez-Calva, M.A. Lugo, and J. Solomon. 2012. “Pov- Fan, Q., J. Reis, M. Jarvis, A. Beath, and K. Frauscher. erty Dynamics in Brazil: Patterns, Associated Factors and 2008. “The Investment Climate in Brazil, India, and South Policy Challenge.” World Bank. Africa: A Comparison of Approaches for Sustaining Eco- 40 nomic Growth in Emerging Economies.” Washington, DC: signs in economics. Journal of Economic Literature 48 (2), World Bank 281-355. Feinstein, L., R.Sabates, T. Anderson, A.Sorhaindo and C. Mallén, P. 2013. “Brazilians Are Accumulating Debt by Hammond. 2006. “What are the effects of education on Not Understanding Credit Card Conditions.” International health?” Measuring the effects of education on health and Business Times. December 6, 2013. http://www.ibtimes. civic engagement: proceedings of the Copenhagen sym- com/brazilians-are-accumulating-debt-not-understand- posium. OECD, 171-354. ing-credit-card-conditions-1499238. Ferré, C., F. Ferreira, and P. Lanjouw. 2012. “Is There a Met- Melo, M. and A. Pereira. 2013. Making Brazil Work: Check- ropolitan Bias? The relationship between poverty and city ing the President in a Multiparty System, Palgrave MacMil- size in a selection of developing countries.” World Bank lan SOA, August Economic Review, 26:3, 351-382. Meyer, B., W. Mok, and J. Sullivan. 2009. The under-Report- Ferreira, F., J. Messina, J. Rigolini, L. López-Calva, M. Lugo, ing of Transfers in Household Surveys: Its Nature and Conse- and R. Vakis. 2013. “Economic Mobility and the Rise of the quences. National Bureau of Economic Research. http:// Latin American Middle Class.” World Bank. www.nber.org/papers/w15181. Firpo, S. and M. Reis. 2007. “O salário mínimo e a queda Monteiro, J. and J. Assunção. 2006. “Outgoing the shad- recente da desigualdade no Brasil”, in R. Paes de Barros, ows: estimating the impact of bureaucracy simplifica- M. Foguel and G. Ulyssea (eds): Desigualdade de renda no tion and tax cut on formality and investment.”  Pontifícia Brasil: uma análise da queda recente (Brasilia, IPEA), Vol. 2, Universidáde Católica, Department of Economics, Rio de pp. 499–506 Janeiro. Fontes, G., R. Simoes, and A. Hermeto Camilo de Oliveira. Motte-Baumvol, B and C. Nassi, 2012 “Immobility in Rio de 2010. “Urban Attributes and Wage Disparities in Brazil: A Janeiro, beyond Poverty.” Journal of Transport Geography, Multilevel Hierarchical Model.” Regional Studies, 44:5, pp 24, 67-76 595–607. Moura, R. and Neri, M. 2008. Efetividade do “salário míni- Fortin, N., T. Lemieux, and S. Firpo. 2010. “Decomposition mo estadual”: uma análise via regressões quantílicas para in Economics.” NBER Paper N.16045. dados longitudinais. Economia Aplicada, São Paulo, v. 12, n. 2, p.239-273 Heckman, J., J. Stixrud, and S. Urzua. 2006. “The Effects of Cognitive and Noncognitive Abilities on Labor Market Neusa, Q. and M. Correa, 2011. “Making Brazilians Safer: Outcomes and Social Behavior.” Journal of Labor Econom- Analyzing the Dynamics of Violent Crime.” World Bank. ics. University of Chicago Press 24(3): 411-482, July. Oaxaca, R. 1973. “Male-Female Wage Differentials in Urban Intergovernmental Panel on Climate Change. 2007. “IPCC Labor Markets.” International Economic Review, 14 (Octo- Fourth Assessment Report: Climate Change.” Geneva, ber), 693-709. Switzerland: IPCC. Observatório das Metrópoles - IPPUR/UFRJ. 2013. “Ibeu: International Monetary Fund. 2013. “Brazil: Technical Note índice de bem-estar urbano.” organização Luiz Cesar de on Consumer Credit Growth and Household Financial Queiroz Ribeiro and Marcelo Gomes Ribeiro. Rio de Janei- Stress.” ro: Letra Capital. Jales, H. 2012. Estimating the effects of minimum wage in Bra- Onder, H. 2013. “Reaching the Bottom 40 Percent: Do Na- zil: A Density Discontinuity Design Approach. Available at: tional Goals Translate Sub-nationally?” World Bank, PREM www.eesp.fgv.br/sites/eesp.fgv.br/files/file/Hugo_Jales.pdf Economic Premise, November http://siteresources.world- bank.org/EXTPREMNET/Resources/EP130.pdf Lee, D. 2008. Randomized experiments from non-random selection in U.SI house elections. Journal of Econometrics Oliveira, L., D. de Souza, L. dos Santos, M. Antunes, N. 142 (2), 675-697. Brendoline, and V. Quintaes, 2013, “Construction of a Consumption Aggregate Based on Information from the Lee, D. and T. Lemieux. 2010. Regression discontinuity de- Brazilian Consumer Expenditure Survey 2008-2009 and 41 its use in the Measurement of Welfare, Poverty, Inequality Waiselfisz, J. 2014. “Mapa da Violência 2014: Os Jovens do and Vulnerability of Families.” IBGE. Brasil.” Centro Basileiro de Estudos Latino-Americanos (CE- BELA) and FLACSO Brasil. Pereira, R., and T. Schwanen, “Tempo de deslocamento ca- sa-trabalho no Brasil (1992-2009): diferenças entre regiões Weinstein, B. 1983. The Amazon Rubber Boom: 1850-1920, metropolitanas, níveis de renda e sexo. Texto para Discussão Stanford University Press No. 1813”. Instituto de Pesquisa Econômica Aplicada (IPEA): Brasília/Rio de Janeiro, 2013 Wilson, J. and N. Svajlenka. 2012. “Metro Brazil: An Over- view of the Nation’s Largest Metropolitan Economies.” Ravallion, M. and Q. Wodon. 1999. “Poor Areas, or Only Washington, DC: Brookings. Poor People?” Journal of Regional Science, 39:4, 689-711. World Bank. 2000. World Development Report 2000/2001: Ravallion, M. and S. Chen. 2003. “Measuring pro-poor Attacking Poverty. World Bank. growth.” Economics Letters, 78:93-99. World Bank. 2007. “Economic performance in Latin Amer- Rocha, S. 2003. Pobreza no Brasil. Afinal, de que se trata? Rio ica and the Caribbean: A Microeconomic Perspective.” de Janeiro: Editora FGV. World Bank. Secretaria de Assuntos Estratégicos (SAE). 2013. “Governo World Bank. 2010a. Handbook on Poverty and Inequality, define que a classe média tem renda entre R$ 291 e R$ 1.019 ed. J. Haughton and S.R. Khandker. World Bank. (Cidade Verde, em 24.07.2013).” http://www.sae.gov.br/ site/?p=17351#ixzz35UobUtKL World Bank. 2010b. Human Opportunity for Children in Brazil: An Assessment with the Human Opportunity Index. Secretaria de Avaliação e Gestão da Informação do World Bank. Ministério de Desenvolvimento Social e Combate à Fome. “Tabulador de Informações do CadÚnico.” Available at http:// World Bank. 2013a. “Shifting gears to accelerate shared aplicacoes.mds.gov.br/sagi/cecad/tabulador_tabcad.php prosperity in Latin America and the Caribbean.” World Bank. Sen, A. 1999. Development as Freedom. Oxford: Oxford Uni- versity Press. World Bank. 2013b. “The World Bank Group Goals: End Extreme Poverty and Promote Shared Prosperity.” World Shorrocks, A. 1999. “Decomposition Procedures for Dis- Bank, Washington, DC. http://www.worldbank.org/con- tributional Analysis: A Unified Framework Based on the tent/dam/Worldbank/document/WB-goals2013.pdf. Shapley Value.” University of Essex. (mimeo). World Bank. 2014. “Poverty and Shared Prosperity. The Skoufias, E., P. Leite, and R. Narita. 2013. “Expanding Mi- World Bank’s Twin Goals and Their Measurement Chal- crofinance in Brazil: Credit Utilisation and Performance lenges”, Policy Research Report, World Bank Development of Small Firms.” The Journal of Development Studies, Research Group (to be published in October 2014) 49(9):1256-1269. Skoufias, E. and G. Lopez-Acevedo. 2010. “Sources of Welfare Disparities Within and Between Regions in Latin America and the Caribbean.” World Bank. The Economist. 2012. “Tick, Tock, Brazil’s Pension System”, March 24. Available online at: http://www.economist. com/node/21551093. Urban Age. 2009. “Cities and Social Equity: Inequality, ter- ritory, and urban form.” London School of Economics. Waiselfisz, J. 2013. “Mapa Da Violência 2013: Homicídios E Juventude No Brasil.” Centro Basileiro de Estudos Lati- no-Americanos (CEBELA) and FLACSO Brasil. 42 Annexes Annex 1. What is a Metropolitan Region? It is first and foremost an instrument for planning and together make up an integrated socio-spatial process. management for municipalities that have public func- Included are territorial planning; soil and land use; envi- tions of common interest (Funções Públicas de Interesse ronment and water resources; transport and commuting; Comum—FPIC). The functions are services and structures sanitation (water, sewage, draining); and health and edu- and activities that, shared across municipal boundaries, cation. Metropolitan governance in Brazil—a timeline Today there are: ¨ Creation of the Metropolitan Regions (RMs) in the 1970s • 55 Metropolitan Regions (Regiões Metropolitanas – RMs) established by the states using different criteria. - They become “anchors” in the regional and national development of the country (linked the second PND—Plano • 3 Integrated Regions of Development (Regiões Integradas Nacional de Desenvolvimento II- PND 1975-1979) de Desenvolvimento—RIDEs) established by the Federal government (Teresina, Petrolina/Juazeiro and Brasília). - Existing metropolitan areas prioritized for institutionalization. • 12 Metropoles (Regiões de Influência das Cidades— REGICs) : - Creation of the initial management model - 1 Large National Metropolis: São Paulo - Creation of different metropolitan institutions - 2 National Metropoles: Rio de Janeiro and Brasília ¨ Federal Constitution of 1988: - 9 Regional Metropoles: Belo Horizonte, Porto Alegre, Curitiba, - Metropolitan question is brought to the state level Fortaleza, Salvador, Recife, Belém, Manaus and Goiânia. (Estadualização) - The new PNAD-Continua will be representative for 20 - Conflicts are initiated at the federal level that are not yet metropolitan regions (Manaus, Belém, Macapá, São Luís, resolved more than 25 years later Fortaleza, Natal, João Pessoa, Recife, Maceió, Aracaju, Salvador, Belo Horizonte, Vitória, Rio de Janeiro, São Paulo, - Lack of national criteria and references Curitiba, Florianópolis, Porto Alegre, Vale do Rio Cuiabá, and Goiânia) as well as the RIDE of Teresine (Grande Teresina). - Institutional process of metropolização - Metropolitan management is fragmented and made more vulnerable 44 Typology of the management of public functions of common interest (Funções Públicas de Interesse Comum—FPIC) Public Functions Main Public Main Cooperation Types Functions Characteristics Difficulties Cooperative Public Less structured Transport -Federal laws and programs -Different institutional Functions sector in terms Sanitation guiding States and development levels in of Federal municipalities states and municipalities Government but -Main financing by Union more structured -State protagonist on RM, sector in local and especially in transport system state level (National -Experiences on consortiums Agencies) involving municipalities and states; -Facilities to access federal resources for infrastructure in the RM considered by IBGE Strong/ Highly Health -Strong national system that -Different territorial bases structured sector- Education guides, controls and finances for management of these shared competence the policies in states and functions and/ or national municipalities; -Different institutional system - Facilities to access federal levels in states and resources for infrastructure in municipalities the RM considered by IBGE “Non (?) Cooperative” Housing -Sectors controlled by -Difficulties in controlling Public Functions Land Use Control municipalities using different land market criteria; -Plan limited to -Federal financing directly to municipalities territories municipalities; -Interest conflicts in -State financing directly to municipalities’ border municipalities areas Source: IPEA (2013). Figure 32. 58 RMs and RIDEs Figure 33. 12 Metropoles according to criteria of spatial/economic integration (REGIC/IBGE, 2007) Total population: 93.8 million (IBGE, 2010) Total population: 63.2 million (IBGE, 2010) PIB 2010: R$ 3.2 trillion (IBGE, 2010) PIB 2010: R$2.0 trillion (IBGE, 2010) PIB per capita: R$34.246 (IBGE, 2010) PIB per capita: R$32.966 (IBGE, 2010) 45 Annex 2. Economic growth in Brazil’s RMs and Brazil as a whole Table 3. GDP (in 2000 R$ millions) and real annual GDP growth (%) Area 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Brazil 11,308 11,795 11,950 12,267 12,408 13,117 13,531 14,066 14,924 15,695 15,643 16,822 17,282 4.3 1.3 2.7 1.1 5.7 3.2 4.0 6.1 5.2 -0.3 7.5 2.7 Brazil RMs 5,550 5,644 5,709 5,662 5,582 5,867 6,149 6,335 6,761 6,982 7,038 7,487 7,536 1.7 1.1 -0.8 -1.4 5.1 4.8 3.0 6.7 3.3 0.8 6.4 0.7 Belém 69 77 78 80 80 87 89 94 100 102 102 104 106 10.9 2.0 1.8 0.1 9.1 2.7 5.3 6.2 1.8 0.7 1.5 1.9 Fortaleza 144 148 146 157 153 160 168 177 184 202 207 228 241 2.3 -1.2 7.9 -2.6 4.6 4.8 5.4 3.7 9.9 2.7 10.0 5.7 Recife 173 177 179 192 187 194 206 214 229 237 245 273 280 2.3 1.5 7.2 -2.8 4.1 6.0 4.1 7.0 3.2 3.4 11.7 2.6 Salvador 240 250 250 258 250 272 299 295 305 308 328 336 301 4.1 0.2 3.2 -3.1 8.8 9.8 -1.4 3.3 1.1 6.4 2.6 -10.4 Belo Horizonte 340 364 382 390 392 433 441 479 522 560 533 603 599 7.2 4.9 2.2 0.5 10.5 1.8 8.7 8.9 7.3 -4.7 13.0 -0.7 Rio de Janeiro 1,060 1,069 1,068 1,069 1,007 1,097 1,097 1,110 1,174 1,156 1,181 1,228 1,265 0.8 0.0 0.0 -5.8 9.0 0.0 1.2 5.8 -1.5 2.1 4.1 2.9 São Paulo 2,383 2,432 2,462 2,365 2,356 2,412 2,591 2,676 2,874 2,962 2,960 3,131 3,170 2.1 1.2 -3.9 -0.4 2.4 7.4 3.3 7.4 3.1 -0.1 5.8 1.2 Curitiba 265 280 282 286 304 317 320 331 368 388 388 414 419 5.6 0.8 1.6 6.2 4.1 1.1 3.4 11.0 5.5 0.0 6.9 1.2 Porto Alegre 361 385 389 399 393 417 430 426 445 459 459 500 468 6.6 1.0 2.7 -1.4 6.0 3.2 -0.8 4.4 3.0 0.0 8.9 -6.4 Distrito Federal 516 465 473 466 461 478 507 532 560 609 635 669 686 -10.0 1.7 -1.4 -1.2 3.7 6.2 4.9 5.3 8.6 4.3 5.3 2.6 Source: World Bank calculations using Ipeadata/IBGE. 46 Table 4. GDP per capita (in 2000 R$) and real GDP per capita growth (%) Area 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Brazil 6,897 6,946 6,932 7,025 7,015 7,224 7,346 7,531 8,111 8,277 8,170 8,819 8,983 0.7 -0.2 1.3 -0.1 3.0 1.7 2.5 7.7 2.0 -1.3 7.9 1.9 Brazil RMs 10,796 10,558 10,508 10,269 9,982 10,191 10,515 10,669 11,500 11,618 11,590 12,557 12,540 -2.2 -0.5 -2.3 -2.8 2.1 3.2 1.5 7.8 1.0 -0.2 8.3 -0.1 Belém 3,983 4,173 4,147 4,142 4,061 4,239 4,272 4,402 4,768 4,770 4,739 4,944 4,988 4.8 -0.6 -0.1 -2.0 4.4 0.8 3.0 8.3 0.0 -0.7 4.3 0.9 Fortaleza 4,891 4,831 4,677 4,946 4,732 4,771 4,904 5,069 5,226 5,605 5,671 6,305 6,585 -1.2 -3.2 5.8 -4.3 0.8 2.8 3.4 3.1 7.3 1.2 11.2 4.4 Recife 5,396 5,294 5,294 5,610 5,386 5,474 5,725 5,881 6,270 6,345 6,496 7,407 7,542 -1.9 0.0 6.0 -4.0 1.6 4.6 2.7 6.6 1.2 2.4 14.0 1.8 Salvador 8,151 8,002 7,864 7,990 7,619 8,024 8,653 8,390 8,216 8,103 8,474 9,403 8,345 -1.8 -1.7 1.6 -4.6 5.3 7.8 -3.0 -2.1 -1.4 4.6 11.0 -11.3 Belo Horizonte 7,403 7,550 7,752 7,783 7,684 8,186 8,172 8,717 9,561 10,039 9,442 11,134 10,969 2.0 2.7 0.4 -1.3 6.5 -0.2 6.7 9.7 5.0 -5.9 17.9 -1.5 Rio de Janeiro 10,172 9,834 9,733 9,631 8,987 9,605 9,505 9,516 10,170 9,815 9,951 10,379 10,619 -3.3 -1.0 -1.1 -6.7 6.9 -1.0 0.1 6.9 -3.5 1.4 4.3 2.3 São Paulo 13,753 13,602 13,580 12,860 12,645 12,610 13,353 13,597 14,952 15,101 14,969 15,908 15,994 -1.1 -0.2 -5.3 -1.7 -0.3 5.9 1.8 10.0 1.0 -0.9 6.3 0.5 Curitiba 9,845 10,097 9,888 9,842 10,222 10,176 10,047 10,153 11,589 11,891 11,716 13,049 13,083 2.6 -2.1 -0.5 3.9 -0.4 -1.3 1.1 14.1 2.6 -1.5 11.4 0.3 Porto Alegre 10,066 10,345 10,285 10,428 10,147 10,464 10,637 10,398 11,244 11,370 11,289 12,619 11,762 2.8 -0.6 1.4 -2.7 3.1 1.7 -2.2 8.1 1.1 -0.7 11.8 -6.8 Distrito Federal 26,206 22,658 22,544 21,716 21,034 20,938 21,750 22,320 22,820 23,798 24,357 26,025 26,287 -13.5 -0.5 -3.7 -3.1 -0.5 3.9 2.6 2.2 4.3 2.3 6.8 1.0 Source: World Bank calculations using Ipeadata/IBGE. 47 Annex 3. Shared prosperity in Brazil’s RMs and Brazil as a whole Table 5. Mean income (R$2012) and income growth (%) of the total population and bottom 40% of the national income distribution Area 2004 2005 2006 2007 2008 2009 2011 2012 Brazil All (All) 584.09 620.77 678.00 695.27 729.69 748.51 796.16 860.92 6.3 9.2 2.5 5.0 2.6 6.4 8.1 Brazil All (B40) 134.05 145.16 163.74 169.80 186.68 197.75 215.00 236.59 8.3 12.8 3.7 9.9 5.9 8.7 10.0 Brazil RMs (All) 773.37 838.47 899.28 919.88 953.55 973.55 1,039.59 1,115.66 8.4 7.3 2.3 3.7 2.1 6.8 7.3 Brazil RMs (B40) 141.76 154.88 176.32 182.94 200.80 214.46 235.38 256.93 9.3 13.8 3.8 9.8 6.8 9.8 9.2 Belém (All) 484.79 502.34 551.65 644.05 623.13 592.39 689.93 717.34 3.6 9.8 16.7 -3.2 -4.9 16.5 4.0 Belém (B40) 145.69 152.67 171.60 185.71 202.74 206.69 229.28 242.38 4.8 12.4 8.2 9.2 1.9 10.9 5.7 Fortaleza (All) 479.15 503.27 523.87 533.61 596.71 632.53 656.84 682.89 5.0 4.1 1.9 11.8 6.0 3.8 4.0 Fortaleza (B40) 134.25 146.64 165.57 173.78 187.01 197.64 222.19 248.48 9.2 12.9 5.0 7.6 5.7 12.4 11.8 Recife (All) 524.60 550.02 588.08 555.15 623.91 654.47 654.67 697.18 4.8 6.9 -5.6 12.4 4.9 0.0 6.5 Recife (B40) 128.30 139.00 160.34 163.78 178.25 194.08 212.73 245.50 8.3 15.4 2.1 8.8 8.9 9.6 15.4 Salvador (All) 528.02 586.28 658.32 702.25 753.52 796.73 841.26 886.12 11.0 12.3 6.7 7.3 5.7 5.6 5.3 Salvador (B40) 130.35 145.28 167.24 172.27 189.56 203.50 225.23 247.11 11.5 15.1 3.0 10.0 7.4 10.7 9.7 Belo Horizonte (All) 710.10 768.02 869.49 881.41 922.27 975.97 1,066.26 1,181.01 8.2 13.2 1.4 4.6 5.8 9.3 10.8 Belo Horizonte (B40) 146.65 163.41 180.68 191.21 205.78 222.93 249.45 278.41 11.4 10.6 5.8 7.6 8.3 11.9 11.6 48 Table 5. Mean income (R$2012) and income growth (%) of the total population and bottom 40% of the national income distribution (cont.) Area 2004 2005 2006 2007 2008 2009 2011 2012 Rio de Janeiro (All) 869.38 886.57 995.24 970.95 1,030.26 1,073.22 1,055.57 1,108.19 2.0 12.3 -2.4 6.1 4.2 -1.6 5.0 Rio de Janeiro (B40) 149.43 163.89 183.20 184.07 208.15 219.96 238.70 253.31 9.7 11.8 0.5 13.1 5.7 8.5 6.1 São Paulo (All) 821.48 946.89 986.79 1,011.21 1,010.48 1,008.53 1,119.36 1,237.15 15.3 4.2 2.5 -0.1 -0.2 11.0 10.5 São Paulo (B40) 145.01 158.17 181.75 190.91 211.99 222.98 243.86 266.09 9.1 14.9 5.0 11.0 5.2 9.4 9.1 Curitiba (All) 920.04 908.70 913.46 1,079.39 1,081.77 1,105.26 1,137.93 1,260.62 -1.2 0.5 18.2 0.2 2.2 3.0 10.8 Curitiba (B40) 151.06 162.13 193.71 203.37 201.08 221.69 254.15 265.95 7.3 19.5 5.0 -1.1 10.3 14.6 4.6 Porto Alegre (All) 885.22 941.71 984.43 957.45 1,027.11 1,017.45 1,124.77 1,168.76 6.4 4.5 -2.7 7.3 -0.9 10.5 3.9 Porto Alegre (B40) 150.23 161.82 182.10 188.32 208.15 225.59 245.39 263.18 7.7 12.5 3.4 10.5 8.4 8.8 7.3 Distrito Federal (All) 1,114.78 1,214.89 1,348.12 1,488.19 1,527.06 1,570.09 1,659.77 1,655.29 9.0 11.0 10.4 2.6 2.8 5.7 -0.3 Distrito Federal (B40) 142.42 156.50 178.25 186.05 199.58 225.53 231.67 254.48 9.9 13.9 4.4 7.3 13.0 2.7 9.8 Source: World Bank calculations using PNAD 2004/2012 Table 6. Mean income growth (%) of the population and bottom 40% of each state living in the RM Area All income B40 (of UF) All income B40 (of UF) All annualized B40 (of UF) annuali- 2004 income 2004 2012 income 2012 growth (%) zed growth (%) Belém 484.79 110.36 717.34 164.83 5.0 5.1 Fortaleza 479.15 87.13 682.89 170.56 4.5 8.8 Recife 524.60 86.06 697.18 178.07 3.6 9.5 Salvador 528.02 84.64 886.12 162.32 6.7 8.5 Belo Horizonte 710.10 153.88 1,181.01 301.09 6.6 8.8 Rio de Janeiro 869.38 207.80 1,108.19 295.07 3.1 4.5 São Paulo 821.48 214.11 1,237.15 370.69 5.3 7.1 Curitiba 920.04 193.46 1,260.62 357.46 4.0 8.0 Porto Alegre 885.22 216.09 1,168.76 355.17 3.5 6.4 Distrito Federal 1,114.78 182.36 1,655.29 345.29 5.1 8.3 Source: World Bank calculations using PNAD 2004, 2012. 49 Annex 4. The Gini Index in Brazil’s RMs and Brazil as a whole Table 7. Gini Index across individuals using household per capita income Area 2004 2005 2006 2007 2008 2009 2011 2012 Δ 2004/12 Brazil All 0.57 0.57 0.56 0.55 0.54 0.54 0.53 0.53 -0.04 Brazil RMs 0.57 0.57 0.56 0.56 0.55 0.55 0.54 0.54 -0.04 Belém 0.54 0.54 0.54 0.56 0.52 0.51 0.52 0.52 -0.03 Fortaleza 0.60 0.58 0.56 0.55 0.56 0.56 0.52 0.51 -0.09 Recife 0.63 0.61 0.60 0.58 0.59 0.57 0.54 0.50 -0.12 Salvador 0.59 0.59 0.58 0.59 0.58 0.58 0.56 0.57 -0.02 Belo Horizonte 0.56 0.55 0.55 0.55 0.53 0.53 0.52 0.52 -0.04 Rio de Janeiro 0.56 0.56 0.56 0.56 0.55 0.56 0.54 0.54 -0.02 São Paulo 0.54 0.55 0.54 0.52 0.51 0.51 0.51 0.52 -0.02 Curitiba 0.56 0.54 0.52 0.52 0.50 0.51 0.48 0.49 -0.07 Porto Alegre 0.54 0.54 0.54 0.52 0.53 0.51 0.51 0.50 -0.03 Distrito Federal 0.63 0.60 0.60 0.61 0.62 0.62 0.60 0.58 -0.04 Source: World Bank calculations using PNAD 2004/2012. 50 Table 8. Gini Index across Brazil RM households by gender, race, age and education of household head Area 2004 2005 2006 2007 2008 2009 2011 2012 Δ 2004/12 All HH 0.58 0.58 0.57 0.57 0.56 0.56 0.55 0.55 -0.03 Gender HH Female 0.59 0.58 0.57 0.57 0.57 0.56 0.56 0.54 -0.05 Male 0.58 0.58 0.57 0.56 0.56 0.56 0.54 0.55 -0.03 Race HH Afro-descendant 0.53 0.52 0.51 0.50 0.49 0.50 0.49 0.48 -0.05 White 0.57 0.58 0.56 0.56 0.56 0.56 0.56 0.56 -0.02 Education HH Less than primary 0.46 0.45 0.43 0.43 0.44 0.42 0.40 0.39 -0.07 Primary complete 0.48 0.50 0.45 0.46 0.46 0.47 0.42 0.44 -0.05 Some secondary 0.48 0.48 0.46 0.47 0.46 0.45 0.44 0.51 0.03 Secondary complete 0.50 0.48 0.48 0.47 0.47 0.47 0.46 0.46 -0.04 Some tertiary 0.44 0.47 0.46 0.43 0.49 0.44 0.45 0.46 0.03 Tertiary complete 0.46 0.46 0.46 0.46 0.46 0.48 0.48 0.48 0.02 Age HH 15 to 19 0.58 0.55 0.52 0.46 0.48 0.55 0.48 0.45 -0.13 20 to 24 0.55 0.52 0.50 0.50 0.50 0.49 0.47 0.48 -0.07 25 to 29 0.56 0.58 0.56 0.55 0.59 0.57 0.59 0.56 0.00 30 to 34 0.62 0.62 0.58 0.58 0.56 0.59 0.57 0.56 -0.06 35 to 39 0.60 0.60 0.58 0.56 0.57 0.57 0.58 0.56 -0.03 40 to 44 0.57 0.60 0.58 0.56 0.54 0.55 0.54 0.55 -0.03 45 to 49 0.57 0.58 0.57 0.56 0.55 0.55 0.53 0.53 -0.04 50 to 54 0.55 0.56 0.56 0.56 0.54 0.53 0.54 0.54 -0.01 55 to 59 0.56 0.56 0.57 0.58 0.55 0.55 0.53 0.54 -0.02 60 to 64 0.58 0.57 0.58 0.56 0.57 0.57 0.55 0.54 -0.04 65+ 0.57 0.56 0.52 0.54 0.54 0.53 0.53 0.54 -0.03 Source: World Bank calculations using PNAD 2004/2012. 51 Annex 5. Monetary poverty in Brazil’s RMs with and without adjusting for cost of living Table 9. Poverty headcounts with and without adjusting for cost of living Headcounts with no adjustment for cost of living R$70 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 7.6 7.0 5.8 5.7 4.8 4.6 4.4 3.6 Brazil RMs 4.7 4.1 3.2 3.4 2.9 2.8 2.4 2.3 Belém 6.0 5.9 4.9 3.7 3.5 3.7 3.2 3.4 Fortaleza 8.5 8.5 5.9 6.2 5.6 5.0 3.5 3.3 Recife 10.8 10.0 7.0 8.3 6.9 6.1 6.4 4.3 Salvador 9.6 7.6 5.5 5.9 4.6 4.3 3.6 3.3 Belo Horizonte 3.8 2.7 2.3 2.4 2.2 2.0 1.5 1.1 Rio de Janeiro 3.1 2.4 2.5 3.1 2.2 2.3 2.4 2.8 São Paulo 3.7 3.3 2.5 2.5 2.1 2.4 1.7 1.9 Curitiba 2.3 2.3 1.3 1.6 2.2 1.6 1.4 1.7 Porto Alegre 2.8 2.4 2.1 2.6 1.9 2.0 1.4 1.4 Distrito Federal 4.8 3.8 2.4 2.1 2.8 1.9 2.2 1.9 R$70-140 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 14.8 13.9 11.5 10.4 9.3 8.7 6.7 5.3 Brazil RMs 10.0 9.3 7.2 6.3 5.6 5.2 3.4 2.3 Belém 16.6 16.7 12.9 10.4 9.7 9.6 7.9 5.5 Fortaleza 22.3 18.2 15.5 14.3 12.2 10.6 8.0 5.0 Recife 21.1 18.9 16.4 13.9 14.2 10.5 7.5 4.7 Salvador 17.5 16.6 12.3 11.2 9.7 8.9 6.5 4.6 Belo Horizonte 9.9 8.5 7.3 6.3 4.9 4.1 2.2 1.2 Rio de Janeiro 6.8 7.4 5.0 5.1 4.1 4.7 3.3 2.3 São Paulo 6.7 6.1 4.4 3.7 3.5 3.5 1.9 1.0 Curitiba 6.6 6.6 4.8 2.4 3.3 2.9 1.3 1.0 Porto Alegre 6.5 6.5 5.6 4.5 3.9 3.6 2.3 2.0 Distrito Federal 8.8 8.1 6.1 5.5 4.5 4.7 2.5 2.6 R$140-R$291 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 22.6 20.7 19.8 19.9 18.6 18.2 15.5 15.2 Brazil RMs 20.1 17.4 16.6 16.5 14.7 14.5 11.7 11.3 Belém 29.4 26.4 25.8 24.4 22.8 25.0 18.9 19.1 Fortaleza 28.9 26.7 27.7 27.6 24.9 25.1 21.3 21.3 Recife 26.0 23.8 25.5 25.8 23.5 23.7 20.2 18.9 Salvador 24.4 23.5 23.0 23.0 21.5 20.1 16.8 18.0 Belo Horizonte 21.0 19.0 17.6 16.6 15.1 12.7 10.6 9.1 Rio de Janeiro 18.0 15.6 14.6 14.3 13.9 12.0 11.3 11.0 São Paulo 18.3 14.4 13.1 13.6 11.2 11.9 8.9 8.4 Curitiba 15.5 13.5 15.0 11.9 9.3 10.4 7.6 6.8 Porto Alegre 15.6 14.5 13.0 12.8 11.6 11.7 8.6 7.9 Distrito Federal 19.3 15.6 13.8 14.0 13.5 12.6 10.0 9.5 Source: World Bank calculations using PNAD 2004/2012. Note: Where applicable, per capita household incomes are adjusted for cost of living based on Oliveira, et al. (2013). 52 Table 9. Poverty headcounts with and without adjusting for cost of living (cont.) Headcounts adjusting income for cost of living R$70 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 6.7 5.9 4.9 5.0 4.1 4.1 4.0 3.3 Brazil RMs 4.5 3.7 2.9 3.2 2.7 2.7 2.3 2.3 Belém 5.5 4.2 3.9 3.2 3.0 3.4 2.9 3.3 Fortaleza 7.2 6.9 4.7 5.5 4.4 4.5 3.2 3.1 Recife 9.0 8.4 5.5 7.4 6.2 5.5 6.2 3.9 Salvador 9.5 7.4 5.3 5.7 4.3 4.2 3.6 3.2 Belo Horizonte 4.0 2.9 2.5 2.4 2.2 2.0 1.5 1.1 Rio de Janeiro 3.0 2.3 2.4 2.9 2.2 2.2 2.2 2.8 São Paulo 3.7 3.1 2.5 2.5 2.1 2.4 1.7 1.9 Curitiba 2.0 2.0 1.2 1.4 2.2 1.6 1.4 1.6 Porto Alegre 2.6 2.3 2.0 2.5 1.9 2.0 1.4 1.4 Distrito Federal 5.5 3.8 2.5 2.1 2.8 2.0 2.3 1.9 R$70-140 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 13.9 12.4 10.3 9.5 8.3 7.4 5.5 4.7 Brazil RMs 9.6 8.7 6.7 5.9 5.2 4.6 2.9 2.1 Belém 15.3 14.4 11.0 9.6 8.6 7.3 5.5 4.7 Fortaleza 20.0 15.2 13.7 12.5 10.7 7.7 5.9 3.8 Recife 19.5 14.8 13.8 12.1 11.9 7.5 4.5 3.9 Salvador 17.0 16.4 11.8 10.7 9.4 8.6 6.1 4.4 Belo Horizonte 10.7 9.0 7.6 6.5 5.1 4.4 2.2 1.2 Rio de Janeiro 6.5 7.5 4.8 4.7 3.8 4.1 2.7 2.1 São Paulo 6.6 6.1 4.3 3.7 3.4 3.4 1.9 1.0 Curitiba 6.0 4.5 4.4 2.0 2.7 2.0 0.8 1.0 Porto Alegre 6.5 6.4 5.4 4.5 3.7 3.4 2.2 2.0 Distrito Federal 9.6 8.3 6.2 5.6 4.5 4.8 2.5 2.7 R$140-R$291 2004 2005 2006 2007 2008 2009 2011 2012 Brazil 22.0 21.1 19.9 18.8 17.7 16.6 15.2 13.9 Brazil RMs 19.5 17.2 16.5 15.7 14.1 13.5 11.6 10.5 Belém 26.9 26.7 26.2 21.8 21.7 22.5 18.4 17.4 Fortaleza 27.7 25.9 26.0 22.7 22.1 22.6 20.1 17.8 Recife 24.3 25.3 25.1 21.5 20.3 21.3 19.8 14.6 Salvador 24.5 22.9 23.0 23.0 21.3 18.1 16.5 17.6 Belo Horizonte 21.2 19.0 18.6 16.9 15.6 13.4 11.2 9.5 Rio de Janeiro 17.1 14.6 14.2 14.4 13.1 10.0 11.5 10.0 São Paulo 18.2 14.3 13.1 13.5 11.2 11.8 8.9 8.4 Curitiba 14.3 14.2 13.6 9.4 8.2 8.9 6.1 5.5 Porto Alegre 15.3 14.4 13.0 12.5 11.4 11.5 8.5 7.8 Distrito Federal 18.4 15.8 15.6 14.3 14.0 12.7 10.5 9.9 Source: World Bank calculations using PNAD 2004/2012. Note: Where applicable, per capita household incomes are adjusted for cost of living based on Oliveira, et al. (2013). 53 Table 10. Percent difference in headcounts when adjusting income for cost of living R$70 2004 2005 2006 2007 2008 2009 2011 2012 Brazil -11.6 -16.8 -14.5 -12.1 -14.4 -11.4 -10.0 -10.0 Brazil RMs -4.4 -8.5 -7.1 -5.4 -6.3 -3.4 -3.7 -1.5 Belém -7.5 -27.8 -20.9 -12.5 -16.3 -8.9 -9.3 -3.0 Fortaleza -15.4 -18.9 -20.9 -11.7 -20.8 -10.7 -10.9 -4.5 Recife -16.6 -16.1 -22.0 -10.5 -10.7 -9.8 -3.5 -8.3 Salvador -0.9 -1.4 -2.7 -3.2 -6.3 -2.7 -1.0 -0.8 Belo Horizonte 7.5 4.2 7.4 1.0 -0.8 2.1 0.0 4.5 Rio de Janeiro -3.2 -4.0 -2.1 -4.5 -3.0 -2.1 -8.9 0.0 São Paulo 0.0 -4.4 0.0 -1.6 -1.0 0.0 0.0 0.0 Curitiba -10.1 -13.7 -11.5 -8.9 0.0 0.0 0.0 -5.1 Porto Alegre -6.0 -5.2 -1.3 -5.5 -2.2 0.0 0.0 -0.4 Distrito Federal 14.6 0.7 1.9 0.0 0.0 2.9 2.7 2.4 R$70-140 2004 2005 2006 2007 2008 2009 2011 2012 Brazil -6.2 -10.7 -10.8 -9.3 -10.4 -14.2 -17.7 -11.8 Brazil RMs -3.4 -6.4 -6.4 -6.1 -7.4 -12.3 -16.4 -8.7 Belém -7.5 -13.7 -14.7 -7.7 -11.2 -23.3 -29.7 -14.6 Fortaleza -10.3 -16.4 -11.8 -12.7 -12.1 -27.4 -26.8 -25.6 Recife -7.4 -21.9 -15.4 -12.9 -16.4 -28.2 -39.8 -16.1 Salvador -2.9 -1.5 -4.2 -4.1 -2.5 -3.6 -5.1 -4.1 Belo Horizonte 9.1 6.6 5.3 2.6 4.0 5.9 2.5 2.0 Rio de Janeiro -4.9 0.2 -5.1 -8.5 -7.9 -12.3 -19.3 -8.0 São Paulo -1.3 1.0 -2.1 -0.3 -2.1 -2.3 0.0 0.0 Curitiba -9.0 -32.2 -9.1 -16.0 -19.8 -31.2 -40.0 1.4 Porto Alegre -0.3 -1.7 -4.7 -1.5 -5.4 -5.4 -1.8 -2.0 Distrito Federal 8.7 1.7 1.0 1.7 0.0 2.4 1.5 4.3 R$140-R$291 2004 2005 2006 2007 2008 2009 2011 2012 Brazil -3.0 2.3 0.4 -5.7 -4.8 -8.8 -1.9 -8.7 Brazil RMs -2.9 -1.0 -0.7 -4.9 -4.1 -6.7 -1.1 -7.0 Belém -8.6 1.1 1.7 -10.5 -5.1 -9.9 -2.7 -8.9 Fortaleza -4.2 -3.0 -5.9 -17.8 -11.1 -10.0 -5.8 -16.5 Recife -6.6 6.3 -1.4 -16.7 -13.6 -10.1 -1.9 -22.9 Salvador 0.3 -2.6 -0.1 -0.1 -1.4 -9.8 -1.9 -2.3 Belo Horizonte 1.1 -0.3 5.6 2.0 3.8 5.5 5.3 4.6 Rio de Janeiro -4.6 -6.7 -2.6 1.0 -5.4 -16.7 1.8 -8.6 São Paulo -0.6 -0.5 -0.2 -0.6 -0.3 -0.7 -0.3 -0.2 Curitiba -8.3 4.9 -9.7 -21.1 -11.4 -14.0 -19.1 -20.0 Porto Alegre -1.8 -0.8 -0.3 -2.0 -2.4 -1.8 -0.5 -1.6 Distrito Federal -4.7 1.1 12.8 2.1 3.5 0.2 5.3 4.0 Source: World Bank calculations using PNAD 2004/2012. Note: Where applicable, per capita household incomes are adjusted for cost of liv- ing based on Oliveira et al. (2013). Yellow highlights represent percentage differences less than 10 when adjusting income for cost of living versus not adjusting; blue highlights indicate percentage differences greater than 10. 54 Poverty headcount (%) Poverty headcount (%) Poverty headcount (%) 0 5 10 15 20 25 30 35 0 5 10 15 20 25 0 2 4 6 8 10 12 Brazil Brazil Brazil Brazil RMs Brazil RMs Brazil RMs Belém Belém Belém Fortaleza Fortaleza Fortaleza Recife Recife Recife Salvador Salvador Salvador Belo Horizonte Belo Horizonte Belo Horizonte R$70 2004 Rio de Janeiro Rio de Janeiro Rio de Janeiro R$70-140 2004 R$140-291 2004 São Paulo São Paulo São Paulo Curitiba Curitiba Curitiba Porto Alegre Porto Alegre Porto Alegre Distrito Federal Distrito Federal Distrito Federal Brazil Brazil Brazil Brazil RMs Brazil RMs Brazil RMs Belém Belém Belém Fortaleza Fortaleza Fortaleza Recife Recife Recife Salvador Salvador Salvador Belo Horizonte Belo Horizonte Belo Horizonte Rio de Janeiro Rio de Janeiro Rio de Janeiro R$70 2012 R$70-140 2012 São Paulo São Paulo São Paulo R$140-291 2012 Curitiba Curitiba Curitiba Porto Alegre Porto Alegre Porto Alegre Distrito Federal Distrito Federal Distrito Federal Figure 34. Poverty and vulnerability headcounts with and without adjusting income for cost of living COL COL COL No COL No COL No COL Source: World Bank calculations using PNAD 2004/2012. Note: Where applicable, per capita household incomes are adjusted for cost of living based on Oliveira, et al. (2013). 55 Annex 6. Poverty lines in Brazil Table 11. Comparison of headcounts using different poverty lines in Brazil 2001 2002 2003 2004 2005 2006 2007 2008 2009 2011 2012 Global Extreme Poverty 11.8 10.4 11.6 9.8 8.2 7.1 6.9 5.6 5.5 5.1 4.2 (US$ 1.25 PPP/day) National Extreme Poverty Line 9.9 7.9 9.0 7.6 7.0 5.8 5.7 4.8 4.6 4.4 3.6 (R$70/month) National Moderate Poverty 24.7 23.3 24.9 22.4 21.0 17.3 16.1 14.1 13.3 11.1 9.0 (R$140/month) Source: World Bank calculations using LAC Equity Lab - SEDLAC data (CEDLAS and World Bank). Brazil does not have an official poverty line. Most pover- the country. The international US$1.25 extreme poverty ty measurements compare to an absolute poverty line, line is also used on occasion by Brazil, notably in relation constructed using monthly household income. Several to the Millennium Development Goals (MDG). Comple- unofficial lines exist. They include lines constructed as a mentary to those lines, the lines used by the World Bank fraction of official minimum income (one quarter and one (US$1.25, US$2.50 or US$4 with purchase power pari- half ) as well as regionalized monetary lines that reflect ty—PPP) serve to harmonize poverty measurement and variable costs of living in different regions or areas of the compare the evolution of poverty across countries. The country, or a food basket price index based on minimum choice of a line may reflect the objectives of the analysis, calorie-intake recommendations by WHO and FAO.36 The for instance, an international comparison or the defini- lines produced by IPEA were long considered de facto tion of a public policy. poverty lines for Brazil, and have been used as such by the World Development Indicators (WDI) database of the As a result of methodological differences in the computa- World Bank. tion of lines and income aggregates, there are sometimes small differences between government and World Bank In recent years, R$70 and R$140 per capita per month, estimates. However, trends in Brazil are broadly consistent administrative poverty lines defined for the Bolsa Família across methodologies. and Brasil Sem Miséria programs, have been increasingly used in place of official poverty lines. It has become cru- Figure 35. Comparison of poverty lines in Brazil cial to monitor poverty rates using these administrative lines, particularly in studies of the evolution of poverty in 30 25 36  Based on consumption baskets established for each of the nine metropolitan areas and Brasilia, values are derived for 15 urban and 20 Headcount (%) rural areas in different regions, defining a total of 25 extreme (in- digência) and moderate (pobreza) poverty lines. These amounts are 15 adjusted to the reference date each year with varying prices for each 10 product, based on IBGE National Consumer Price Index. Concerning regional poverty lines, see Rocha, S. (2006) : Pobreza no Brasil. Afinal, 5 de que se trata?, Editora FGV, Rio de Janeiro, 2006.See also www.ip- eadata.gov.br. In December 2013, IPEA updated its extreme poverty 0 and poverty numbers for the period ranging from 2009 to 2012, but 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 no updated data on the regional lines used is available. For 2012, Global Extreme Poverty National Extreme Poverty National Moderate IPEA puts extreme poverty at 5.3% while all poverty is at 15.9%. (US$ 1.25 PPP/day) Line (R$70/month) Poverty (R$140/month) 56 Annex 7. Intra-generational mobility through synthetic panels in Brazil’s RMs In the absence of panel data, the analysis of the evolution Note: Different assumptions about the residuals yield the of income over the period is complemented by a zoom-in lower and upper income bounds—the assumption that on income mobility using synthetic panels, based on the error terms between periods are perfectly correlated sys- approach recently developed by Dang, Lanjouw, Luoto tematically underestimates the mobility between periods. and McKenzie (2011). The analysis identifies those who This functional form is thus used to estimate the lower left and those who stayed in poverty. The main advantage bound on mobility. Conversely, assuming that the error of this approach is that it does not need to impose much terms are perfectly uncorrelated overstates the amount of structure of the individual income generating process. In- mobility; this equation is thus employed to calculate the stead, it allows us to calculate lower and upper bounds on upper bound on mobility. the movements in and out of poverty, depending on the assumption regarding the individual-specific error term. Figure 36. Stayers, sliders, climbers Synthetic panels are built using two cross-section datasets from 2004 to 2012 in Brazil RMs (2004 and 2012), by estimating the relationship between income and two sets of variables: i. Time-invariant vari- 100 ables at the household level: traits such as gender, year of birth and parental education that do not change through- 80 Share population (%) out a lifetime—for each year; ii. Time-invariant variables at 60 the metropolitan area level such as unemployment rate, population of working age, and displacement rate. 40 Using the second round (2012), the relationship between 20 income and invariant household characteristics is mod- 0 eled using OLS and the income is estimated for round Belém Fortaleza Salvador Recife Belo Horizonte Distrito Federal Brazil RMs Sao Paulo Curitiba Rio de Janeiro Porto Alegre 1(2004) with the resulting coefficients as follows: , where perfect positive cor- Stayers Climbers Sliders relation of the error term is assumed to establish the lower bound. Source: World Bank calculations using PNAD 2004, 2012. Note: Income groups based on poverty lines R$70 (extreme) and R$140 (moderate) and middle class thresholds R$291 (lower) and R$1,019 (upper). This yields incomes for the two periods, one real, and the other estimated, to track households’ movements in and out of poverty, e.g., the probability of a household that was poor in period 1 to escape poverty by period 2 where p is the poverty line: This approach has been validated in a recent paper by Cruces et al. (2011) in the context of Latin America. 57 Annex 8. General aspects by income group across Brazil’s RMs Table 12. Characteristics by income group living in Brazil RMs 2012 All R$70 R$70-R$140 R$140-R$291 R$291+ Individuals based on HH access (%) Assets 99.7 97.6 98.9 99.5 99.8 Sanitation 90.9 81.8 73.8 81.7 92.6 Shelter 97.0 95.4 92.9 95.4 97.2 Water 98.3 93.6 94.4 96.2 98.7 8 years of education (for at least 1 90.1 75.8 82.9 87.9 90.5 member) Enrollment of children 90.8 86.3 82.4 87.6 92.6 Adults (15yrs+) Education less than primary (%) 30.3 46.1 54.0 45.3 28.6 Education secondary + (%) 36.5 26.4 17.2 26.1 37.9 Hourly wage (R$2012) 14.53 1.96 3.73 5.12 15.25 Informality (%) 22.2 96.8 66.9 37.5 20.5 Labor force participation (%) 65.8 32.0 51.9 55.6 67.6 Unemployment (%) 7.2 84.0 26.8 20.5 5.2 Population Share (%) 15 to 25 year household head 9.8 22.5 22.4 24.5 8.1 Afro-descendant household head 49.6 58.3 73.3 68.7 47.6 Female household head 42.3 56.8 57.6 46.7 41.1 Single mother household head 10.0 26.4 37.5 23.6 7.9 Afro-descendant individuals 49.6 60.8 70.9 67.4 46.5 0 to 15 years old 20.9 34.1 47.5 38.7 17.8 Source: World Bank calculations using PNAD 2012. Note: Afro-descendants include individuals self-identifying as “preto” (black) or “pardo” (mixed origin). 58 Table 13. Probability of being moderate poor vs. extreme poor, vulnerable vs. moderate poor, or middle class vs. vulnerable (logit regressions) for individuals living in Brazil’s RMs 2004 2012 Characteristics Moderate Vulnerable Middle class Moderate Vulnerable Middle class Age -0.008*** -0.009*** -0.046*** 0.018*** -0.009*** -0.049*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Age^2 0.000*** 0.000*** 0.001*** -0.001*** 0.000*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female -0.036*** -0.078*** -0.122*** 0.045*** -0.163*** -0.157*** (0.002) (0.001) (0.001) (0.003) (0.002) (0.001) Afro-descendant (vs. white) 0.042*** -0.238*** -0.724*** 0.384*** -0.132*** -0.680*** (0.002) (0.001) (0.001) (0.003) (0.002) (0.001) Dwelling has sanitation -0.074*** 0.427*** 0.778*** -0.400*** 0.368*** 0.677*** (0.002) (0.001) (0.001) (0.003) (0.002) (0.001) Dwelling walls masonry materials 0.144*** 0.140*** 0.273*** -0.417*** 0.316*** 0.231*** (0.003) (0.002) (0.002) (0.006) (0.004) (0.002) Household has 2 of 3 key assets 0.708*** 0.595*** 1.048*** 1.029*** 0.595*** 0.402*** (refrigerator, phone, clean stove) (0.002) (0.002) (0.003) (0.011) (0.011) (0.007) Dwelling has piped water 0.305*** 0.250*** 0.761*** 0.406*** 0.180*** 0.586*** (0.003) (0.002) (0.003) (0.006) (0.005) (0.003) Dwelling with electricity 0.012*** 0.061*** 0.159*** -0.062*** 0.055*** 0.144*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Average years of education 0.110*** 0.060*** -0.043*** -0.008*** -0.015*** 0.102*** (0.002) (0.001) (0.001) (0.003) (0.002) (0.001) Migrant -0.246*** -0.657*** -1.811*** -0.670*** 0.199*** -0.099*** (0.004) (0.003) (0.004) (0.013) (0.012) (0.008) Constant -0.008*** -0.009*** -0.046*** 0.018*** -0.009*** -0.049*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Observation 7,812,334 16,005,305 45,330,264 2,566,405 7,668,301 53,952,417 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 Source: World Bank calculations using PNAD 2004, 2012. Note: Afro-descendant includes individuals self-identifying as “preto” (black) or “pardo” (mixed origin). Note: Income groups based on poverty lines R$70 (extreme) and R$140 (moderate) and middle class lower threshold R$291. 59 Annex 9. Multi-dimensional poverty in Brazil’s RMs and Brazil as a whole Poverty has been defined as the “pronounced deprivation dwelling, (5) no flush toilet or pit latrine in the dwelling, (6) in wellbeing” (Houghton et al.). This is a difficult condition no electricity, and (7) lack of assets. Individuals are consid- to measure statistically and is thus typically expressed as ered to have a lack of assets when they do not own either a the amount of income required to attain a certain level refrigerator/freezer or a telephone/mobile. When a house- of well-being. However, there are other elements that are hold is deprived of three or more of these opportunities, it non-monetary that effect the level of individuals’ well-be- is considered to be “multi-dimensionally poor.” ing, their ability to compete in the labor market, and their resilience to economic shocks. In order to capture some of Utilizing a combination of multi-dimensional poverty and these dimensions, a different measure or indicator must monetary poverty can provide a clearer portrait of pov- be utilized. erty and those who face it. When these two indicators of poverty are combined, four separate groups emerge; (1) Multi-dimensional poverty examines the non-monetary the chronic poor (those poor both in income and multi-di- measures of poverty evident in a society. Similar to an mensional terms), (2) the transient poor (those poor only absolute poverty line, it is based on the total number of in income terms), (3) the “not poor but deprived” (those non-monetary deprivations that an individual faces. Rath- poor only in multi-dimensional terms), which could be er than simply measuring the level of income of an indi- considered as a form of vulnerability, and (4) the better vidual or household, multi-dimensional poverty examines off (those not poor in either term). The chronic poor may the deprivation of indicators that are believed to be close- be divided into those who fall below the moderate and ly linked with income levels and well-being. These dimen- extreme poverty lines. These groups make up the differ- sions are: (1) a school-aged child (7 to 15) is out of school, ent elements of poverty and problems faced by different (2) none of the household members has completed at least groups within society, providing stakeholders and policy eight years of schooling, (3) the dwelling is constructed makers with a tool to better understand the problem and with precarious wall materials, (4) no access to tap water in act on it. Figure 37. Matrix of multidimensional and income poverty in Brazil 2004 2012 Not poor but deprived Better o Not poor but deprived Better o RM: 1.3% RM: 86.5% RM: 0.6% RM: 95.4% All: 6.6% All: 75.7% All: 3.4% All: 89.2% $140 $140 Income poor Income poor Chronic poor Transient poor Chronic poor Transient poor RM: 1.3% RM: 10.9% RM: 0.1% RM: 3.8% All: 6.7% All: 11% All: 1.6% All: 5.8% 7 6 5 4 3 2 1 0 7 6 5 4 3 2 1 0 Multi-dimensionally poor Multi-dimensionally poor Source: World Bank calculations using PNAD 2004/2012. 60 Annex 10. The Human Opportunities Index in Brazil’s RMs The Human Opportunities Index is an equity-penalized tain circumstance group. When this is disaggregated, it measure of the coverage of services deemed necessary represents the amount of the penalty that is contributed for the development of human capital in children. These to each specific circumstance. services are electricity, sanitation, access to education (measured by school attendance) and completion of pri- The six steps of building the Human Opportunity Index mary education on time. While the coverage rate allows policy makers and stakeholders to see overall access, the 1. Estimate a separable logistic model on whether child HOI also examines how access is affected by circumstanc- i had access to a given basic good or service as a func- es outside of the child’s control (i.e., parents’ education, tion of his or her circumstances. For education, age is also ethnicity and location of birth). HOI penalizes the cover- used to predict the probability of completing each grade. age rate for the differences in access based on these cir- The specification is chosen according to the needs of each cumstances. circumstance: quadratic for years of education, logarith- mic for real income, and categorical for age and the other The penalty can be interpreted as the percentage of peo- dimensions. In all cases the functions are linear in the pa- ple whose access would have to be reassigned to people rameters. From the estimation of this logistic regression, in the groups with below-average coverage rates in order obtain coefficient estimates. to achieve equality of opportunity. If all groups had ex- actly the same coverage rate, the penalty would be zero, 2. Given these coefficient estimates, obtain for each child and no reassignment would be needed. As coverage ap- in the sample the predicted probability of access to the proaches universality for all groups, that reassignment basic good or service in consideration based on the becomes smaller. predicted relationship and a vector of their circum- stances : The Dissimilarity Index, or D-Index, measures the differ- ences in coverage as a result of being a member of a cer- Graphically, the HOI can be explained as follows 3. Compute the overall coverage rate C, 100 90 Vulnerables Non- vulnerables 80 70 Penalty 60 Coverage 50 40 30 where wi = 1/n or some sampling weights. HOI 20 10 4. Compute the Dissimilarity Index , 0 10 20 30 40 50 60 70 80 90 100 Percentile (Coverage Group) Note: Vulnerability is de ned here in terms of opportunities and is distinct from the vulnerability line (R$291) based on income used in the rest of the analysis. 61 5. Compute the Penalty, Or 6. Compute the HOI, Table 14. Human Opportunities Index 2004 and 2012 Area Attendance Grade Water Sanitation Electricity Internet Cellphone Progression 2004 Brazil All 94.6 47.0 74.8 48.7 91.7 62.8 32.0 Brazil RMs 96.0 58.1 94.3 76.0 99.6 64.7 50.0 Belém 93.5 34.2 77.1 76.7 98.7 50.8 41.7 Fortaleza 94.7 48.6 80.5 48.3 98.2 61.0 34.9 Recife 94.2 41.8 85.1 28.8 100.0 63.4 39.8 Salvador 95.2 37.7 95.4 77.8 99.9 69.8 40.7 Belo Horizonte 95.6 59.4 97.5 80.3 99.3 57.3 55.4 Rio de Janeiro 96.7 49.4 95.9 81.9 99.9 66.7 53.1 São Paulo 96.3 74.1 98.7 84.2 100.0 64.9 47.9 Curitiba 96.3 59.6 95.7 81.3 99.5 64.2 52.2 Porto Alegre 96.1 59.1 97.0 87.1 99.3 65.1 73.5 Distrito Federal 97.0 55.8 97.1 94.2 99.4 67.0 72.3 2012 Brazil All 97.0 56.9 86.7 62.3 98.7 79.5 86.7 Brazil RMs 97.3 59.9 97.1 85.5 99.9 85.7 96.1 Belém 97.9 50.8 96.2 59.8 100.0 76.4 95.3 Fortaleza 96.8 51.3 90.2 65.3 99.9 76.2 96.6 Recife 97.0 54.9 95.2 60.2 99.7 87.8 95.7 Salvador 96.7 43.4 96.6 89.6 99.8 89.1 97.3 Belo Horizonte 97.0 56.7 97.8 84.1 100.0 77.7 97.9 Rio de Janeiro 97.3 50.1 97.4 90.2 100.0 86.9 94.5 São Paulo 97.6 72.9 98.0 93.7 99.9 90.0 95.4 Curitiba 92.3 54.0 98.4 85.1 100.0 78.6 94.7 Porto Alegre 98.2 59.6 99.2 90.2 100.0 77.4 97.7 Distrito Federal 98.1 54.4 96.0 95.0 100.0 87.5 98.2 Source: World Bank calculations using PNAD 2004, 2012. Note: HOI is calculated based on access of children age 16 and younger to ser- vices, adjusting for equality of distribution among the following circumstance groups: education and gender of the household head, per capita household income, gender, race, urban/rural location, the number of children in the household, and whether two parents are in the household. 62 Annex 11. Differences between cores and inner peripheries in the RMs of Brazil’s Northeast and Southeast, 2010 Table 15. Characteristics of core and inner periphery by income group 2010 All R$70 R$70-R$140 R$140-R$291 periphery core periphery core periphery core periphery core RM Fortaleza Education less than primary 48.5 34.6 63.5 55.5 63.4 61.0 54.0 51.1 Hourly wage(R$, 2012) 5.7 11.6 1.4 1.9 3.0 3.6 3.8 4.1 Informality (%) 38.5 30.3 93.7 90.6 62.7 54.6 46.0 43.6 Labor force participation (%) 58.4 63.9 28.0 20.3 46.2 46.6 54.8 56.9 Unemployment (%) 9.2 7.6 45.6 62.1 17.0 19.9 11.6 13.4 Sanitation 43.2 74.2 33.5 70.4 34.7 64.5 39.2 66.5 RM Recife Education less than primary 43.3 34.1 62.1 55.7 63.7 60.9 53.8 51.7 Hourly wage (R$2012) 8.3 14.2 1.8 2.0 3.4 3.4 4.3 4.3 Informality (%) 26.8 23.9 91.0 87.1 52.3 51.8 36.5 38.8 Labor force participation (%) 58.5 61.7 29.9 29.3 47.8 47.5 55.1 56.9 Unemployment (%) 14.7 12.3 71.9 75.9 29.5 31.0 20.6 21.6 Sanitation 48.1 69.5 38.1 60.9 35.5 56.2 39.5 56.1 RM Salvador Education less than primary 43.7 31.5 61.6 50.1 64.4 56.5 55.1 48.5 Hourly wage (R$, 2012) 8.9 12.4 1.6 1.7 3.2 3.9 4.2 4.6 Informality (%) 28.0 22.3 91.0 91.9 57.9 51.9 39.3 38.6 Labor force participation (%) 65.2 67.3 38.8 32.5 53.0 54.2 60.3 62.9 Unemployment (%) 15.9 12.9 70.2 75.7 31.7 33.9 21.5 22.7 Sanitation 66.5 93.0 54.9 87.8 53.3 87.5 59.8 87.8 RM Belo Horizonte Education less than primary 44.8 30.1 59.2 40.2 65.9 59.0 59.3 53.0 Hourly wage (R$, 2012) 8.7 16.5 2.4 2.6 3.7 4.0 4.3 4.7 Informality (%) 19.3 16.2 85.4 77.3 37.6 34.1 28.5 25.1 Labor force participation (%) 67.4 68.3 22.7 17.5 47.3 47.2 57.3 57.7 Unemployment (%) 7.7 6.4 76.8 82.5 28.3 31.9 16.8 19.8 Sanitation 81.9 96.2 74.1 93.8 69.5 91.9 73.9 90.7 RM Rio de Janeiro Education less than primary 38.8 28.5 52.9 42.9 60.2 56.6 52.2 49.2 Hourly wage (R$, 2012) 10.4 17.9 2.3 2.7 4.2 4.5 5.2 5.3 Informality (%) 25.2 18.7 88.6 87.8 48.2 39.6 36.7 29.8 Labor force participation (%) 61.0 61.6 20.9 14.7 47.0 43.7 55.3 54.8 Unemployment (%) 10.0 7.2 81.3 82.9 28.0 25.6 18.7 17.7 Sanitation 82.8 94.4 77.0 91.7 71.8 87.4 75.4 88.8 RM São Paulo Education less than primary 35.8 32.2 50.0 45.6 57.4 57.8 50.4 51.9 Hourly wage (R$, 2012) 12.1 19.1 1.9 1.9 3.9 4.7 5.1 5.7 Informality (%) 19.6 19.3 87.1 90.4 46.5 45.5 31.3 32.1 Labor force participation (%) 66.8 66.8 19.5 14.5 47.9 45.5 57.8 56.2 Unemployment 9.2 7.3 83.1 85.1 31.7 30.9 22.0 20.0 Sanitation 86.9 92.7 81.1 89.2 73.1 80.1 77.4 83.1 Source: World Bank calculations using Census 2010. 63 Annex 12. Various indicators in Brazil’s Northeast and Southeast RMs RM Fortaleza Homicide Rate 2012 RM Recife Homicide Rate 2012 RM El Salvador Homicide Rate 2012 RM Belo Horizonte Homicide Rate 2012 RM Rio de Janeiro Homicide Rate 2012 RM Sao Paulo Homicide Rate 2012 64 RM Fortaleza Gini 2010 RM Recife Gini 2010 RM El Salvador Gini 2010 RM Belo Horizonte Gini 2010 RM Rio de Janeiro Gini 2010 RM Sao Paulo Gini 2010 65 RM Fortaleza BF % 2013 RM Recife BF % 2013 RM El Salvador BF % 2013 RM Belo Horizonte BF % 2013 RM Rio de Janeiro BF % 2013 RM Sao Paulo BF % 2013 66 RM Fortaleza IBEU 2010 RM Recife IBEU 2010 RM El Salvador IBEU 2010 RM Belo Horizonte IBEU 2010 RM Rio de Janeiro IBEU 2010 RM Sao Paulo IBEU 2010 67 RM Fortaleza Informal % 2010 RM Recife Informal % 2010 RM El Salvador Informal % 2010 RM Belo Horizonte Informal % 2010 RM Rio de Janeiro Informal % 2010 RM Sao Paulo Informal % 2010 68 Annex 13. Evolution of commuting time in Brazil’s RMs Table 16. Breakdown of workers’ commuting time house-to-work by RM RM Minutes from house to work More than an hour to go to work (%) 1992 2012 Change (min) 1992 2012 Change (ppt) Belém 24.30 32.80 8.50 3.3 10.1 6.80 Fortaleza 30.90 31.70 0.80 8.1 9.8 1.70 Recife 32.30 38.00 5.70 9.6 14.0 4.40 Salvador 31.20 39.70 8.50 8.3 17.3 9.00 Belo Horizonte 32.40 36.60 4.20 10.6 15.7 5.10 Rio de Janeiro 43.60 47.00 3.40 22.2 24.7 2.50 São Paulo 38.20 45.60 7.40 16.6 23.5 6.90 Curitiba 30.20 32.00 1.80 8.6 11.3 2.70 Porto Alegre 27.90 30.00 2.10 6.1 7.8 1.70 Distrito Federal 32.10 34.90 2.80 8.7 10.6 1.90 Source: IPEA, 2013/ PNAD 1992, 2012, IBGE. 69 Annex 14. Oaxaca-Blinder Recentered Regression (RIF) for core/inner periphery and core/outer periphery of the RMs of Brazil’s North/Northeast and Southeast, 2010 Let’s consider two groups: A and B. The overall change in the distributional statistic v of per capita income Y overtime can be defined as: Where F is the cumulative distribution. Following Firpo, Fortin and Lemieux (2007, 2009) we can add and subtract the counterfactual distribution statistic for obtaining the Oaxaca-Blinder distribution where is the structure effect or the difference due to changes in coefficients and is the composition effect or the difference explained by changes in charac- teristics between the two groups. The standard Oaxaca-Blinder method (Oaxaca, 1973; Blinder, 1973) is a particular case of the above equation where we decompose the difference in mean wages. Typically applied to analyze difference in wages (for instance, between men and women), it can be interpreted as follows: if the wage structure of the reference group was held constant, how much of the gap could be explained by the differences in characteristics? However, the traditional OB method presents limitations. Indeed, while decomposing the mean is fairly straightfor- ward, thanks to the statistical properties of the expected value, decomposing quantiles is not. In the former, thanks to the law of iterated expectations (an extension of the law of total expectations— LTE), the estimated coefficient of a simple OLS regression can be interpreted as the effect of a chance on the mean value of the covariate on the uncondi- tional mean value of the dependent variable. On the other hand, the coefficient in the conditional quantile regression can only be interpreted as the effect of a change in the mean value of the covariates on the T th conditional quantile of the dependent variable, as the law of iterated expectation does not hold. Instead, the two-step method introduced by Firpo, Fortin and Lemieux (2007, 2009) replaces the dependent variable Y by the recentered influence function RIF(y;v) of the statistic ν. The recentering consists of adding back the distributional statistic ν to the influence function IF(y;ν): RIF(y;ν) = ν + IF(y;ν)..37 Hence we can apply OLS to obtain regression coefficients from RIF transformed variables and go back to the standard Oaxaca-Blinder decompositions. This allows generating counterfactuals for any distributional statistic like quantiles and Gini. In the present analysis, we apply the two-step method to provide a more granular view on the metropolitan dynamics between the center of the metropolitan region (i.e., its capital), the inner periphery of the metropolitan region, and its immediate outer periphery. Previous works have used mean decomposition methods to explain disparities between urban and rural areas (Ravallion and Wodon, 1999), and between regions and within regions (López-Acevedo and Sk- oufias, 2010). A more recent paper explores its application to intra-city welfare taking the case of Bogotá (Aguilar and Yepes, 2013). To the best of our knowledge, the application of the RIF method in combination with the Oaxaca-Blinder decomposition has not been used to look at metropolitan spatial disparities. 37  For a more extensive description of the respective methods see N. Fortin, T. Lemieux, and S. Firpo (2010): Decomposition in Economics, NBER Paper N.16045 70 Annex 15. Labor regressions by year and income group for working age adults living in RMs of Brazil Table 17. Hedonic wage regression by year and income group across Brazil’s RMs (18 to 64 yr olds) Characteristics All Poor Vulnerable All Poor Vulnerable 2004 2004 2004 2012 2012 2012 Migrant 0.074*** 0.066*** 0.060*** 0.053*** 0.098* 0.064*** (0.006) (0.016) (0.009) (0.006) (0.056) (0.017) Weekly hours worked -0.020*** -0.020*** -0.020*** -0.032*** -0.024*** -0.030*** (0.000) (0.001) (0.000) (0.000) (0.002) (0.001) Unionized 0.178*** 0.018 0.088*** 0.148*** -0.044 -0.012 (0.007) (0.031) (0.014) (0.008) (0.118) (0.027) Female -0.304*** -0.166*** -0.238*** -0.324*** -0.308*** -0.219*** (0.006) (0.017) (0.010) (0.006) (0.060) (0.018) Informal -0.288*** -0.411*** -0.278*** -0.268*** -0.508*** -0.363*** (0.006) (0.018) (0.010) (0.007) (0.067) (0.019) Age 0.060*** 0.045*** 0.047*** 0.038*** 0.057*** 0.031*** (0.002) (0.005) (0.003) (0.002) (0.017) (0.005) Age^2 -0.001*** -0.001*** -0.001*** -0.000*** -0.001*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Average years eductation -0.046*** 0.020*** 0.022*** -0.081*** 0.026 0.016* (0.003) (0.007) (0.005) (0.003) (0.024) (0.008) Average years eductation^2 0.009*** -0.000 -0.000 0.010*** -0.001 -0.000 (0.000) (0.001) (0.000) (0.000) (0.002) (0.001) Constant 0.731*** 0.856*** 1.011*** 2.193*** 1.243*** 2.094*** (0.034) (0.098) (0.055) (0.035) (0.321) (0.103) Observations 43,624 3,424 7,547 45,718 448 3,193 R-squared 0.550 0.353 0.375 0.507 0.319 0.382 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 Source: World Bank calculations using PNAD 2004, 2012. Note: Afro-descendants include individuals self-identifying as “preta” (black) or “pardo” (mixed origin). Note: Income groups based on poverty lines R$70 (extreme) and R$140 (moderate) and middle class lower threshold R$291. 71 Table 18. Probability of being employed by year and income group across Brazil’s RMs (18 to 64 yr olds) Characteristics All Poor Vulnerable All Poor Vulnerable 2004 2004 2004 2012 2012 2012 Age 0.103*** 0.100*** 0.111*** 0.089*** 0.068*** 0.135*** (0.000) (0.000) (0.000) (0.000) (0.001) (0.001) Age^2 -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Female -0.294*** -0.290*** -0.354*** -0.301*** -0.005 -0.563*** (0.001) (0.002) (0.001) (0.001) (0.003) (0.002) Afro-descendant (vs. white) -0.138*** 0.039*** 0.041*** -0.128*** 0.112*** -0.001 (0.001) (0.002) (0.002) (0.001) (0.004) (0.002) Dwelling has sanitation 0.040*** -0.251*** -0.140*** 0.071*** -0.277*** -0.061*** (0.001) (0.002) (0.002) (0.001) (0.004) (0.003) Dwelling walls masonry -0.131*** -0.227*** -0.158*** -0.065*** -0.250*** -0.029*** materials (0.002) (0.003) (0.003) (0.002) (0.008) (0.006) Household has 2 of 3 key assets 0.081*** -0.167*** -0.025*** 0.212*** -0.254*** 0.165*** (refrig., teleph., clean stove) (0.002) (0.003) (0.004) (0.006) (0.013) (0.015) Dwelling has piped water 0.064*** 0.004 -0.099*** 0.087*** -0.112*** -0.242*** (0.002) (0.003) (0.004) (0.003) (0.007) (0.006) Dwelling with electricity -0.219*** -0.018 -0.698*** -0.352*** 0.352*** (0.009) (0.012) (0.022) (0.027) (0.038) Average years of education 0.029*** -0.046*** -0.037*** 0.021*** -0.084*** -0.038*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Migrant 0.095*** -0.000 0.086*** 0.065*** -0.012*** 0.046*** (0.001) (0.002) (0.002) (0.001) (0.004) (0.002) Constant -0.941*** -0.729*** 0.157*** -0.419*** -0.552*** -1.113*** (0.009) (0.015) (0.024) (0.027) (0.041) (0.019) Observation 26,716,992 2,485,089 4,334,240 30,121,432 580,293 2,058,796 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 Source: World Bank calculations using PNAD 2004, 2012. Note: Afro-desc. include individuals self-identifying as “preto” (black) or “pardo” (mixed origin). Note: Income groups based on poverty lines R$70 (extreme) and R$140 (moderate) and middle class lower threshold R$291. 72 Annex 16. Measuring shared prosperity at the sub-national level The World Bank has recently identified two strategic goals: under-report consumption and/or income, the measure ending extreme poverty and boosting shared prosperity. of shared prosperity will still be completely robust. (World The two goals should be achieved in a sustainable way. Bank 2014). Here, sustainability is meant in a broad sense, that is, the economic, social, and environmental dimensions must be In a large and federal country such as Brazil, measur- considered together. ing shared prosperity at the national level may not fully capture some of the important within variation at the Boosting shared prosperity has been defined as “expand- sub-national level that particularly matters from a policy ing the size of the pie continuously and sharing it in such prioritization perspective. While the indicator of shared a way that the welfare of those at the lower end of the in- prosperity monitored by the World Bank is measured by come distribution rises as quickly as possible” (World Bank the growth of the income per capita of the bottom 40 per- 2013). Income growth among the bottom 40 percent of cent of the income distribution, given the present focus the income distribution in the population (the “bottom of the study on metropolitan regions, measures at the 40”) has been chosen as the indicator to be used to mea- national level (bottom 40% of Brazil residing in the given sure shared prosperity. Complementary to this indicator, RM) and at the state level (bottom 40% of the state resid- pro-poor growth is measured by the growth of the bot- ing in the corresponding RM) are both used to provide a tom 40% compared to the mean. finer lens on the reading of shared prosperity in a large federal country such as Brazil. However, the use of Shared This indicator presents measurement advantages in that Prosperity Indicators (SPI) at the sub-national level is not it is largely unaffected by measurement problems asso- without potential problems. As noted by Onder (2013), ciated with anyone who is not in the bottom 40 percent countries with strong heterogeneity between regions of the distribution. This feature is particularly important could face a case of “Simpson’s Paradox”—a special case in the case of an upper middle-income country such as of ecological fallacy where correlation of aggregates dif- Brazil. Indeed, non-response is a problem more pro- fers from the correlation of components. Incidentally, as nounced in the case of richer countries, where refusal the measurement of shared prosperity continues to be rates increase with income levels (see Meyer et al. 2009 tested, the benchmark for shared prosperity in Brazil re- on the issue of nonresponse in the United States). This mains at the national level. means that measures of inequality that are derived from household survey data may not adequately capture the full magnitude of changes in inequality if top earners are not represented in the samples. However, since the shared prosperity measure places close to no weight on anyone above the 40th percentile, nonresponse of top earners has a relatively small effect on the measure. When the rich completely drop out of the survey, the “bottom” 40 percent will somewhat overstate the location of the 40th percentile, but the effect of this on the mean of the bottom 40 percent is much smaller than the change in the mean of the distribution and the measure of inequal- ity. However, if top earners participate in the survey, but 73 Annex 17. From Favelas to “Areas of Special Social Interest” (AEIS) Table 19. Ten Largest Favelas (2010 census, IBGE) Name State Population 1 Rocinha RJ 69,161 2 Sol Nascente DF 56,483 3 Rio das Pedras RJ 54,793 4 Coroadinho MA 53,945 5 Baixadas da Estrada Nova Jurunas PA 53,129 6 Casa Amarela PE 53,030 7 Pirambú CE 42,878 8 Paraisópolis SP 42,826 9 Cidade de Deus AM 42,476 10 Heliópolis SP 41,118 According to the 2010 census, over 11.4 million people in favelas can be considered part of the middle class (i.e., (about 6% of the population of Brazil and equal rough- having an income between R$1,000 and R$4,000).39 ly to the total population of Portugal) live in aglomera- dos subnormais, subnormal agglomerations. These are IBGE defines a favela as a “settlement of 51 housing units more commonly known as favelas, although this term or more located on public or private property and charac- covers various realities, including “areas of special social terized by disordered occupation without the benefit of interest.”38 The vast majority (88.6%) of those people are essential public services.” But as noted by Perlman (2009), located in 20 large metropolitan areas, notably the met- while ‘favelas’ and ‘slums’ are both territories of exclusion ropolitan regions of São Paulo (596,479 individuals), Rio in cities that increasingly criminalize poverty, they exist de Janeiro (520,260), Belém (291,771), Salvador (290,488) in very different contexts and serve different functions. and Recife (249,432). The word favela has taken such negative connotations that most people use morroi (hill), communidade popular If favelas concentrate poor and vulnerable people, they (popular community), or simply communidade. There are are not devoid of social mobility either. A recent IPEA other forms of informal housing which have traditionally study estimated that up to 65% of the population living accommodated the poor. Among them are Cortiços (old single-family houses that have been subdivided to ac- 38  In Brazil, federal law (City Statute 10,257/2001) permits the designation of certain areas as “special districts,” which allows for dif- commodate multiple families); Cabeças de porco (tene- ferent or more flexible zoning codes. The designation of an Area (or ments), and vilas (workers’ housing consisting of attached, Zone) of Special Social Interest (AEIS or ZEIS) allows cities in Brazil to re-zone and create targeted services for a specific, legally defined geographical area 39  http://www.sae.gov.br/site/?p=14901 74 one-room apartments running back from the street along both sides of a narrow passageway). The IBGE’s favela definition itself is problematic. There are many settlements “of 51 or more shacks” that have gone unnoticed by the authorities, either because they are ad- jacent to another favela, between two conjuntos or in a particularly remote area. “Located on public or private property” is meaningless, because all housing is on pub- lic or private property, and this definition doesn’t speci- fy whether the land is being occupied legally or illegally. “Characterized by disordered occupation” only applies to some favelas, because others consist of very ordered vil- las – classic workers housing on both sides of a narrow passageway with utilities and drainage running down the center of it, for example. As for “without the benefit of public services,” it is well-known that older favelas are fully serviced (for instance, in Rio), whether as a result of government upgrading or the hard work of its residents (Perlman, 2009). The prevalence of subnormal agglomerations is induced and compounded by Brazil’s large housing deficit. Indeed, like other countries in LAC, Brazil faces an acute problem of housing scarcity, evident in the 8 million- unit “housing deficit” in 2011—a climb from 6.4 million units in 2005. Of this, 90% relates to the lowest income bracket. For many Brazilians, especially the poor, high property prices have made housing unaffordable, and poor households may in fact pay a higher price in slum areas relative to formal dwellings due to the low supply elasticity of housing in those areas (Abramo, 2003). For instance, according to a recent study by the Inter-American Development Bank (IDB, 2012), around 62% of families in São Paulo found it too expensive to own a house. 75 76 77 The World Bank 1818 H Street, NW, Washington, DC 20433, USA. www.worldbank.org 78