DISCUSSION PAPER NO. 1504 98253 Entering the City Emerging Evidence and Practices with Safety Nets in Urban Areas Ugo Gentilini July 2015 Entering the City: Emerging Evidence and Practices with Safety Nets in Urban Areas July 2015 Ugo Gentilini Senior Economist Social Protection and Labor Global Practice Ugentilini@worldbank.org 1   Abstract. Most safety net programs in low and middle-income countries have hitherto been conceived for rural areas. Yet as the global urban population increases and poverty urbanizes, it becomes of utmost importance to understand how to make safety nets work in urban settings. This paper discusses the process of urbanization, the peculiar features of urban poverty, and emerging experiences with urban safety net programs in dozens of countries. It does so by reviewing multidisciplinary literature, examining household survey data, and presenting a compilation of case studies from a ‘first generation’ of programs. The paper finds that urban areas pose fundamentally different sets of opportunities and challenges for social protection, and that safety net programs are at the very beginning of a process of urban adaptation. The mixed-performance and preliminary nature of the experiences suggest to put a premium on learning and evidence-generation. This might include revisiting some key design choices and better connecting safety nets to spatial, economic and social services agendas compelling to urban areas. JEL Classification: O1, O2, O18, I38 Keywords: urban areas, urban poverty, social protection systems, safety nets, social assistance, cash transfers, slums, migration, mobility, social inclusion, urban development, jobs, structural transformation, urbanization, cities. 2   Table of Contents Acknowledgements ............................................................................................................................. 7 Executive summary ............................................................................................................................ 8 Overview ........................................................................................................................................... 11 Section I. Understanding urbanization .......................................................................................... 15 1.1 Urban dynamics beyond the tipping point ................................................................................... 15 1.2 The economics of structural transformation................................................................................. 18 1.3 Urbanization, with and without development .............................................................................. 21 1.4 Quality of urbanization................................................................................................................. 24 Section II. The challenge of urban poverty .................................................................................... 26 2.1 The urbanization of poverty ......................................................................................................... 26 2.2 Poor areas or poor people? ........................................................................................................... 28 2.3 Is urban poverty different? ........................................................................................................... 32 2.4 Social assistance in urban areas: preliminary estimates from survey data ................................... 40 Section III. Key issues in design and implementation of urban safety nets ................................ 45 3.1 Identifying the poor ...................................................................................................................... 46 Assessments ................................................................................................................................ 46 Targeting ..................................................................................................................................... 51 3.2 Outreach and enrollment .............................................................................................................. 60 3.3 Mobility and portability ............................................................................................................... 65 Mobile populations ..................................................................................................................... 65 Residency .................................................................................................................................... 70 Managing labor incentives .......................................................................................................... 75 3.4 Governance arrangements ............................................................................................................ 78 Quasi-formal and community-based organizations .................................................................... 86 Social intermediation services .................................................................................................... 87 3.5 Complementary interventions ...................................................................................................... 88 Section IV. Lessons from country case studies .............................................................................. 99 4.1 The Urban Dibao Program in China .......................................................................................... 102 4.2 Urban Prospera Program in Mexico ........................................................................................... 109 4.3 Familias en Acción in Urban Colombia ..................................................................................... 116 4.4 Safety Net Programs in Nairobi’s Slums ................................................................................... 123 4.5 The Programa de Apoyo Temporal al Ingreso (PATI) in urban El Salvador............................. 130 4.6 Family Rewards: A U.S. Experimental Conditional Cash Transfer Program ............................ 137 4.7 Urban Voucher Program in Gaza ............................................................................................... 143 4.8 Social Pensions in Urban Delhi .................................................................................................. 150 4.9 Philippines: Modifying a Conditional Cash Transfer for Urban Areas...................................... 157 4.10 Program Keluarga Harapan in urban Indonesia ....................................................................... 165 3   References ....................................................................................................................................... 172 Figures Figure 1. Population trends and projections, 1950-2050 ................................................................... 15 Figure 2. Sources of urban population growth in India, 1963-2011 (million people) ....................... 16 Figure 3. Distribution of population by size of urban areas, 1990-2030............................................ 17 Figure 4. Structural transformation in 74 countries, 1990-2005 ........................................................ 19 Figure 5. Number of supermarkets in Dhaka (left) and their share in Delhi’s retail sector (right) .... 19 Figure 6. Urbanization and per-capita income ................................................................................... 21 Figure 7. Changes in urban population and GDP per capita, 1985-2010........................................... 22 Figure 8. Level and proportion of urban populations living in slums, 1990-2012 ............................ 23 Figure 9. Urbanization and poverty.................................................................................................... 27 Figure 10. Urban poverty numbers and share ($1.25/day line), 1990-2008 ...................................... 28 Figure 11. Poverty rates in urban and rural areas, latest available data ............................................. 29 Figure 12. Ratio of urban to rural poverty numbers by urbanization level, latest available data ...... 30 Figure 13. Perception of poverty and urbanization in Latin America and the Caribbean.................. 31 Figure 14. City size and inequality..................................................................................................... 31 Figure 15. Labor market activity in urban and rural areas ................................................................. 32 Figure 16. City size and job types in Ethiopia, 2012 ......................................................................... 34 Figure 17. Percentage of formal and informal employment in Delhi, 2009-10 ................................. 35 Figure 18. Mapping the urban and rural safety net in India ............................................................... 39 Figure 19. Coverage of social protection programs, poorest quintile ................................................ 41 Figure 20. Coverage of safety net programs, poorest quintile by country income group .................. 42 Figure 21. Coverage of social protection programs, poorest quintile by region ................................ 42 Figure 22. Safety net programs in urban (left) and rural (right) areas in lower-middle income ........ 43 Figure 23. Poverty map in Vietnam based on rates (left) and numbers (right) .................................. 45 Figure 24. Poverty map in Slovenia based on rates (left) and numbers (right).................................. 46 Figure 25. Urban diamonds for Nairobi (left) and Dakar (right) ....................................................... 41 Figure 26. Map of areas by level of precariousness in San Salvador................................................. 42 Figure 27. Poverty prevalence in Addis Ababa, 2012........................................................................ 50 Figure 28. Results of enrollment process in Bogota, 2011 ................................................................ 60 Figure 29. Expansion of Cazuca (Soacha) over 1996-2007 (left) and its current administrative boundary with Bogota (right) ............................................................................................................. 82 Figure 30. Mapping urban development, economic and social protection activities ......................... 89 Figure 31. Brazil’s PBSM: differentiated productive inclusion in urban and rural areas .................. 89 Figure 32. Housing continuum model ................................................................................................ 96 Figure 33. Typology of case studies by program pathways and urbanization levels ......................... 96 Figure 34. Number of Dibao beneficiaries in urban areas (1996-2013 ............................................ 103 Figure 35. Trends in urban Dibao thresholds (2003-2013) .............................................................. 106 Figure 36. Number of Oportunidades beneficiaries in urban areas (1997-2013)............................. 110 Figure 37. Clustering of AUPs by level of precariousness .............................................................. 131 Figure 38. Trainings demanded by PATI participants, 2013 .......................................................... 133 Figure 39. PATI’s impact on labor participation ............................................................................. 135 Figure 40. Matching targeting measures, poverty lines, and interventions...................................... 146 Figure 41. Main difficulties in application process for pension schemes (percent) ......................... 153 Figure 42. Distribution of poor households by type of toilet facility, urban and rural Areas .......... 159 4   Figure 43. Distribution of poor households by type of materials of wall (%), urban and rural Areas................................................................................................................................................. 160 Figure 44. Share of poor households as identified by the Listahanan (%)....................................... 163 Figure 45. Take up rate for the CCT program (%)........................................................................... 163 Figure 46. Coverage of KPH and other national safety nets ............................................................ 158 Figure 47. Beneficiary incidence of KPH and other national safety nets ........................................ 167   Tables Table 1. Urban poverty and spatial planning in Surakarta and Makassar cities, Indonesia ............... 25 Table 2. Poverty rates by area ($1.25/day line), 1990-2008 .............................................................. 27 Table 3. Employment composition in urban and rural areas by region ............................................. 33 Table 4. Selected large-scale crises impacting urban areas, 2003-2014 ............................................ 38 Table 5. Average coverage of social protection programs by poverty status .................................... 41 Table 6. Global coverage of social protection programs by quintiles ................................................ 42 Table 7. Global composition of safety net portfolios in urban and rural areas .................................. 44 Table 8. Comparative merits of targeting methods ............................................................................ 53 Table 9. Urban and rural PMT models in Honduras .......................................................................... 57 Table 10. Pilot approaches in urban SNAP by state, strategy and beneficiary profile ...................... 62 Table 11. Contrasting China’s hukou and Vietnam’s ho-khau systems ............................................ 72 Table 12. Work-related incentives in selected countries in Eastern Europe and Central Asia .......... 77 Table 13. Roles of local governments in U.S. anti-poverty and social protection programs ............. 79 Table 14. Contrasting Bogota, Ciudad Bolivar and Commune 4, select social indicators ................ 83 Table 15. Examples of small-scale self-employment programs in urban areas ................................. 90 Table 16. Summary of features in selected case studies .................................................................. 101 Table 17. Social protection programs in urban China ...................................................................... 103 Table 18. The three phases of Familias, 2000-present ..................................................................... 116 Table 19. SISBEN III scores ............................................................................................................ 118 Table 20. Benefit structure, 2000–2006 ........................................................................................... 120 Table 21. Benefit structure, 2007–2011 ........................................................................................... 120 Table 22. Benefit structure, 2012–present ....................................................................................... 120 Table 23. Benefit structure in Gaza’s voucher program .................................................................. 147 Table 24. Features of social pensions programs in Delhi ................................................................ 152 Table 25. Key features of Pantawid CCT and MCCT ..................................................................... 158 Table 26. Number of households assessed and identified as poor ................................................... 159 Boxes Box 1. Defining an urban area ........................................................................................................... 17 Box 2. Evolution of thinking about spatial economic models ........................................................... 20 Box 3. Urbanization Reviews............................................................................................................. 24 Box 4. Aspirations and livelihoods among the unemployed youth in Addis Ababa ......................... 34 Box 5. The cognitive tax among the urban and rural poor: insights from behavioral economics ..... 36 Box 6. Supporting Syrian refugees in urban Lebanon with electronic vouchers ............................... 39 Box 7. Urban activism and social media ............................................................................................ 40 Box 8. Combining census and qualitative data: a typology of marginalized urban areas in Romania ............................................................................................................................................................ 48 Box 9. Devising urban-sensitive PMTs.............................................................................................. 55 Box 10. High food prices and targeting in urban Ouagadougou and Bobo-Dioulasso, Burkina Faso ............................................................................................................................................................ 59 5   Box 11. Enrolling the urban elderly and working poor: evidence and practices from the U.S.......... 63 Box 12. Benefit-induced migration? A summary of the evidence ..................................................... 65 Box 13. Brazil’s Bolsa Familia: an urban perspective ....................................................................... 66 Box 14. Piloting portability of safety nets in urban Chhattisgarh, India ............................................ 69 Box 15. Portability and interoperability: lessons from the U.S. ........................................................ 69 Box 16. Experimenting with institutional reforms in China .............................................................. 73 Box 17. Sudden exit from an unconditional cash transfer program in Bulgaria ................................ 76 Box 18. Urban safety nets and labor informality: the Universal Child Allowance program in Argentina ............................................................................................................................................ 78 Box 19. India’s welfare funds for informal construction workers ..................................................... 78 Box 20. Institutional coordination and urban public works: insights from the Trabajar program ..... 84 Box 21. The crisis-response public works program in urban Latvia.................................................. 84 Box 22. Combining formal and informal provision of safety nets: food banks in urban areas ......... 86 Box 23. Institutions for local social intermediation: the case of urban Chile Solidario..................... 87 Box 24. Connecting safety nets and jobs: the Sustainable Livelihood Program in the Philippines... 91 Box 25. Productive inclusion in Brazil’s favelas ............................................................................... 93 Box 26. The Rental Support Cash Grant program in Haiti ................................................................ 96 Box 27. Mobile Water Payment Innovations in Urban Kenya .......................................................... 97 Box 28. Violence, confinement and stigma in urban Colombia ........................................................ 99 Box 29. SISBEN and related targeting processes ............................................................................ 117 Box 30. Opportunity costs of enrolment .......................................................................................... 119 Box 31. Gaza’s refugee population .................................................................................................. 144 Box 32. Assessing urban poverty: identifying pockets and hotspots ............................................... 161 Box 33. The poor student allowance program: a case study from Jakarta ....................................... 168 6   Acknowledgements This paper presents preliminary findings from ongoing work conducted in collaboration with various colleagues, particularly Ruslan Yemtsov, Maria Concepcion Steta Gandara and Ulrich Hoerning. Special thanks to Arup Banerji and Anush Bezhanyan for their strategic guidance, as well as to the paper’s peer-reviewers, including Linda Gibbs, Sumila Gulyani, John Blomquist, and Qaiser Khan. Several colleagues have contributed with inputs and case studies, including Elena Glinskaya, Dewen Wang, Pablo Acosta, Claudia Rodriguez Alas, Francesca Lamanna, Yoonyoung Cho, Laura Dávila, María Claudia Vásquez, Shrayana Bhattacharya, Changqing Sun, Avantika Prabhakar, Luisa Fernandez, Eric Zapatero, Rubén Villanueva, Yuko Okamura, Shanna Rogan, Mohamed Ihsan Ajwad, Pantaleo Creti, Aline Coudouel, James Riccio, Issa Sanogo and Paul Skoczylas. Maddalena Honorati, Claudia Rodriguez-Alas and Ana Sofia Martinez Cordova have been instrumental in providing ASPIRE survey data on urban social protection. The paper also benefited from the advice and support of Philip O’Keefe, Margaret Grosh, Jehan Arulpgramasam, Edmundo Murrugarra, Alexandra Posarac, Friederike Rother, Manuel Salazar, Gail Nayowith, Robert Palacios, Giuseppe Zampaglione, Carlo Del Ninno, Phillippe Leite, Syed Hashemi, Mattias Lundberg, Omar Arias, Karla McEvoy, Benedicte de la Briere, Ignacio Raul Apella, Maria Laura Sanchez Puerta, Micheal Woolcock, Marcela Ines Salvador, Jamele Rigolini, Leonardo Lucchetti, Eduardo Moreno, Somik Lall, Judy Baker, Marianne Fay, Margaret Arnold, Ruth Hill, Shaohua Chen, Richard Verbeck, Emanuela Galasso, Gaetano Vivo, Alys Willman, Ellen Hamilton, Hideki Mori, Leslie Elder, Harold Alderman, John Hoddinott, Dan Gilligan, Naureen Karachiwalla, and Meghan Olivier. Excellent research assistance was provided by Alies Van Geldermalsen and Jorge Miranda. Yet all views expressed in the paper only reflect those of the author. 7   Executive summary Currently, 3.9 billion people, or 54 percent of the world population, live in urban areas. Those settings are expected to host, by 2050, an additional 2.4 billion people, with about half of such increase occurring in five middle-income countries. Cities present a range of opportunities for upward socio- economic mobility. For instance, in 2014 they generate about 65 percent of global GDP, with the world’s 300 largest metropolitan economies accounting for nearly half of global output. However, more urbanization does not always mean more development. For example, when demand for housing, jobs, and services outstrips the capacity of cities to providing them, then urban areas can produce congestion and present complex challenges. For instance, the number of people in developing countries living in slums is about 863 million, or 32.7 percent of the urban population. More broadly, poverty is rapidly urbanizing: while the number of the urban poor remained stable at around 285 million people over 1990-2008, the share of urban poverty out of total poverty increased from 17 percent to 24 percent over the same period. Urban poverty exhibits a range of peculiar features, including being characterized by income volatility and informality; physical mobility and migration; constrained commuting to jobs; poor and risky housing conditions; low quality and highly congested social services; possible looser social networks and community bonds; and socio-economic marginalization, including because of violence and various forms of discrimination. Are safety nets effectively reaching the urban poor? Those programs, also known as ‘social assistance’ and including non-contributory transfers in cash or in-kind, tend to have relatively limited coverage of the urban poor. Household survey data shows that 16.6 percent of the urban household in the poorest quintile are covered by some form of safety net program, as opposed to 23.4 percent of the poorest rural quintile – a gap of about 7 percentage points. Such difference soars to nearly 24 percentage points in some middle income countries. Why such differences in urban versus rural coverage? This may be due to various factors, including a mainstream perception that safety nets may not be needed or appropriate in urban areas, chiefly because of more vibrant labor markets. Yet while the urban poor are covered by social insurance and labor market interventions for a rate double that of rural areas, those programs only reach a fraction of the urban poor (i.e., 3-4 percent for them). Also, poverty may be underestimated in urban areas, especially since the relatively higher urban cost of living is often not fully factored into the construction of poverty lines. From another perspective, a sizable share of the portfolio of interventions for urban poverty tend to fall under the remit of urban development and area-based upgrading programs. These provide the critical supply-side of urban infrastructure (e.g., drainage, water supply, public sanitation facilities, etc.) and, arguably, are less geared toward the ‘people’ or demand-side of the poverty equation. The political economy dynamics may also differ, as the middle class is larger in urban areas, and local insider-outsider issues may play an important role in shaping safety net regimes and preferences. In some contexts, large-scale untargeted subsidy schemes have been considered a de-facto urban safety net, although these tend to be particularly regressive. Finally, technical bottlenecks in design and implementation in urban safety nets have often stifled their performance, including in terms of assessments, targeting, communication, enrollment, benefit structure, and institutional arrangements. Such urban ‘nuts and bolts’ issues have been largely underexplored and are the central focus of this paper. Poverty assessments in urban areas show that the scope and focus of interventions can change quite remarkably pending on whether poverty is expressed in prevalence or absolute terms, i.e., areas where poverty rates are highest (generally rural areas) or areas with the highest number of poor people (often 8   urban areas): these two approaches can lead to fundamentally different conclusions on ‘where the poor live’, including with implications for budget allocations in those areas. However, limiting the assessment to consumption and income measures may not capture key multifaceted issues that tend to define poverty in urban contexts, such as demonstrated in Kenya and Senegal; indeed, several countries, e.g., Mexico and Romania, are adopting multidimensional approaches to urban poverty. Generally, urban programs use multiple targeting methods to select and prioritize potential beneficiaries. A recurrent question among social protection practitioners is whether and how to adapt methods such as proxy means testing (PMT) to urban contexts. Country case studies show that formula for rural populations may not accurately target urban poor and vice versa. Another frequently- raised issue revolves around reaching households with characteristics that are seldom captured in PMT models – e.g., not what kind of kind of materials their house is made of, but whether they have a house at all, i.e. the homeless or street children. Some programs, e.g., Philippines, were designed to specifically reach those populations. Traditional community-based targeting may be less compelling to urban areas, although countries’ experiences underscore the importance of mobilizing and ensuring participation of neighborhoods and building on informal networks. Beyond targeting, a number of steps should be put in place to ensure program ‘uptake’ by perspective beneficiaries. A key emerging lesson is the need for extensive communications and outreach tailored to urban settings, like those set out in Mexico and Colombia. The experience of the United States, for example, shows that enrollment may require customization for particular categories of urban dwellers, such as the elderly, the working poor, ethnic minorities, and newly-arrived migrants. Relatedly, the issue of mobility, whether daily or seasonal, calls for greater portability of benefits. Technology can represent both an opportunity as well as a barrier to portability, depending on the interoperability of systems deployed. A significant proportion of urban dwellers may not reside within the ‘formal’ and ‘legal’ system, hence raising challenges on how to support people lacking documentation and residency status. To unbundle such complex issue, countries like India and Kenya are considering various pathways for gradually formalizing those settlements while providing public assistance. The governance arrangements of social protection across countries are a product of legislative and political processes reflecting fundamental cultural preferences, historical initial conditions and technical considerations. Where central governments devolve responsibilities for financing and administration of social assistance to local governments, this can create both opportunities and challenges. For example, the physical proximity of municipalities with different levels of institutional and financing capacities can generate spatial inequities among neighboring areas. However, given the universe of different urban actors and operators, local governments can help innovate and harness the wealth of local organizations, such as in Brazil. These often play a key role in supplementing and integrating state-level capacities, especially in times of distress. Social intermediation services play an important role, facilitating the connection between demand and supply of social programs. Indeed, those services address the ‘choice overload’ problem that prevents poor people from effectively tapping the social protection system, including because of challenges – frequent in urban areas – such as limited awareness on existing interventions, high opportunity costs to accessing them, distrust or lack of familiarity with formal bureaucracies, etc. Urban social safety nets would need to be better integrated with interventions in the spatial, economic and social realms. It is the combination of those domains that may, ultimately, enhance the prospects for upward mobility of the urban poor. With the comparative advantage of being in ‘direct contact’ with beneficiaries, safety nets can play an important role in connecting the poorest to urban opportunities and sevices. This may include, for example, rethinking how to link safety nets to housing policy, devise ways to raise the employability of the poor in self or wage employment, and 9   incentivize the uptake of social services relevant to urban areas, such as daycare, or investing in the prevention of violence or health risks. Examples from both middle and low-income countries provide some initial evidence on practices to strengthen those sectorial linkages in urban settings. Yet relatively little is known about how these dimensions interact, how to identify and prioritize interventions, and what are the possible trade-offs. A collection of summary case studies illustrates pathways for introduction of urban safety nets and emerging lessons at diverse stages in the urbanization process, including China, Mexico, Colombia, Kenya, El Salvador, United States, Palestinian Territories, India, Philippines and Indonesia. Their experience shows that, in many ways, we are at the very beginning of a key agenda for an ever- urbanizing world. 10   Overview Cities are magnets of opportunity and have been playing such role for centuries. Indeed, the process of development is often closely intertwined with that of urbanization, that is, the spatial concentration of people, activities and infrastructure. Since 2007, most to the world’s population live in urban areas, including for pursuing upward mobility, accessing better services, and enhaning quality of life. Indeed, evidence shows that urban areas tend, on average, to offer higher standards of living than rural areas. Looking beyond averages, however, there is ample variation across and within countries. More urbanization doesn’t automatically translate into more development, and not all urban dwellers can reap the potential benefits of urbanization. So what shapes the balance of opportunities and challenges in urban areas? In principle, the form and functionality of cities are the result of two competing forces: on one hand, ‘agglomeration’ economies generate positive externalities from the process of urbanization (e.g., skills, innovation, economies of scale and productivity). Such effects have been the underlying tenet of the economic explosion of urban areas: for instance, by 2025, the global urban economy will have grown over 20 times its level of 1950, or representing 75 percent of the world’s economy and accounting for US$63.7 trillion. On the other hand, urbanization can produce negative externalities and congestion (e.g., disease, violence, crime, and socio-economic exclusion). In particular, when the demand for housing, jobs, transport and services outstrips the capacity of city governments to provide them, then urban areas can nurture, perpetuate and amplify various forms of poverty and marginalization. Indeed, an estimated 863 million people currently live in precarious settlements or ‘slums’ lacking access to basic water and sanitation services: in Nairobi’s slums, for example, 1 out of 500 people has access to toilets; in Dhaka, only 9 percent of the poorest quintile has access to a sewage line. Overall, the share of urban poverty out of total poverty increased from 17 percent to 24 percent over 1990-2008, which clearly indicates that poverty is urbanizing and it is doing so rapidly. Therefore, the outlook of cities around the world often depend on the management of those positive and negative forces and trade-offs. Against this backdrop, countries are increasingly recognizing the need for making urbanization more ‘inclusive’. As systems to address urban poverty are reimagined, there is growing interest in the role that social protection in general, and safety nets in particular, can play in urban areas and during the urbanization process. By ‘safety nets’ we refer to non-contributory transfers in cash or in-kind, including programs such as unconditional and conditional cash transfers (CCTs), school feeding, and public works targeted to poor people. Those programs are part of larger social protection systems and have been traditionally implemented in rural areas with more limited applications to urban contexts. This paper is one of the first attempts to review issues, evidence and practices on safety nets in urban contexts. In particular, the paper is geared to better understand the role of safety nets in urban areas, document current experience and practices, and identify emerging lessons and challenges. It does so by drawing from multidisciplinary published literature, household survey data from 112 countries in World Bank’s ASPIRE database, and a compilation of case studies. While its focus is on low and middle-income countries, examples from high-income countries are also provided to highlight and deepen the discussion on specific themes and features. So are safety nets reaching the poor in urban areas? Evidence shows that 16.6 percent of the urban household in the poorest quintile are covered by some form of safety net program, as opposed to 23.4 11   percent of the poorest rural quintile – a gap of about 7 percentage points. Such difference soars to nearly 24 percentage points in some middle income countries. Why such contrasting coverage performance between urban and rural areas? This could stem from multiple reasons. For example, there might be a mainstream perception that safety nets may not be needed or appropriate in urban areas, chiefly because of more vibrant labor markets. Yet, the poorest individuals tend to have precarious, low-pay and informal jobs. Survey data also shows that a minor fraction of the urban poor (i.e., 3-4 percent) are covered by social insurance and labor market programs. Relatedly, active labor market programs, including skills trainings for self and wage employment, have often found it challenging to match labor market needs with the aspirations, capabilities and profiles of the poorest households. In the majority of developing countries, poverty is still more highly concentrated in rural areas. Therefore, rural settings may have been prioritized in allocation of scarce resources, especially on the basis of poverty prevalence in those areas. Poverty assessments from Vietnam and Slovenia, for example, show that the scope and focus of interventions can change quite remarkably pending on whether poverty is expressed in prevalence or absolute terms, i.e., areas where poverty rates are highest (generally rural areas) or areas with the highest number of poor people (often urban areas): these two approaches can lead to fundamentally different conclusions on locus of poverty or ‘where the poor live’. Overall, poverty may be underestimated in urban areas, with the relatively higher urban cost of living being not systematically factored into the construction of poverty lines. From a different perspective, political economy dynamics matter: the middle-class is larger in urban areas, and local insider-outsider issues may play an important role in molding safety net regimes and preferences. Also, a sizable share of the portfolio of interventions for urban poverty tends to fall under the remit of urban planning and development, such as slums upgrading programs. Those essential and critical interventions have focused on the engeneering and supply-side of urban infrastructure (e.g., drainage, water supply, public sanitation facilities, etc.), and less on the demand or ‘people’ side of the poverty equation. Generalized subsidy schemes have been often been regarded as a de-facto urban safety net: for example, in Liberia and Georgia various subsidy programs represent 86.2 percent and 63.6 percent of the safety net portfolios in urban areas, respectively. Yet the distributional impacts of such measures is considerably regressive, with limited benefits accruing to the poorest segments of the urban populace. In terms of operational ‘nuts and bolts’, the successful experience with rural programs, compounded with the notion that cities present more conducive operational environments, have bolstered expectations on the performance of safety nets in urban settings. As countries roll-out such first generation of programs, the initial performance, however, seems to be lower than predicted. For instance, a range of technical hurdles have stifled an effective identification of the urban poor, or targeting them within the fluid expansion and contraction of urban informal settlements over time; it can be challenging to reach and communicate with prospective beneficiaries about available programs, including because of high mobility or characteristics that make people ‘hard to reach’ (e.g., being homeless); and even when people are reached, programs may not be attractive enough to offset relatively high urban opportunity costs or addressing particular bottlenecks (e.g., for elderly and seasonal migrants), including resulting in limited actual program uptake and enrollment. The early stage of implementation in countries like, for example, Mexico, India and United States provide a vivid illustration of those challenges, as well as ways in which countries have gradually adapted and refined approaches to complex urban contexts. 12   Cities often wear multiple institutional hats. Some of the largest metropolitan areas can at the same time be an administrative region, province and city. Where central governments devolve responsibilities for financing and administration of social assistance to local governments, this can create both opportunities and challenges. For example, the physical proximity of municipalities with different levels of institutional and financing capacities can generate spatial inequities among neighboring areas. However, given the universe of different urban actors and operators, local governments can help innovate and harness the wealth of local organizations, such as in Brazil. These often play a key role in supplementing and integrating state-level capacities, especially in times of distress. Also, in a range of countries poor people may not live within the ‘formal’ and ‘legal’ system, hence raising challenges on how to support people lacking documentation and residency. To unbundle such complex issue, countries like Kenya and India considering options to formalize those settlements while providing public social assistance. The review shows a great variety in the trajectory of introduction and expansion of urban safety nets (whether starting first in urban or rural areas), the level of urbanization at which countries have introduced those programs, and possible design adjustments to urban contexts. For instance, some countries have gradually built on mature rural safety net programs and transitioned them to cities, such as Mexico. Other countries have followed an opposite pattern, commencing programs in urban contexts and expanding them to rural areas, like China. In other cases, like Gaza and New York City, selected programs were conceived only for urban areas. There are also countries where safety nets covered both urban and rural areas from the start, without envisioning major design adjustments across space. However, their application in urban areas had been accompanied by a different set of linkages to complementary interventions (e.g., the urban- specific ‘productive inclusion’ measures in Brazil), ignite positive interaction between space and people (e.g., connecting safety nets and rental support measures in Haiti), and offer the playfield to test innovations (e.g., the delivery of digitalized food entitlements in urban Raipur, India). In most low-income countries, however, safety nets tend to be prevalently rural with limited urban interventions. Yet, as high food prices in 2007-08 were rapidly transmitted to urban consumers, programs were launched in urban areas, such as the vouchers-based program in urban Burkina Faso. Other countries, such as the Democratic Republic of Congo, Ethiopia, Mali and Tanzania are currently planning or beginning to implement urban safety nets interventions. Taken together, these considerations suggest that there is an overall ‘first-generation’ of urban safety nets programs bulging in different contexts. In other words, we are at the beginning of a journey where interest, practices and know-how are growing, but where the role of safety nets in urban areas – and in the urbanization process more widely – remains a complex, dynamic and largely pristine domain. Therefore, given the early stages in the learning process, defining the ultimate role of safety nets in urban areas is no easy task. With the comparative advantage of being in ‘direct contact’ with beneficiaries, evidence suggests that safety nets can play an important role in helping connect the poorest to urban opportunties and services. This may include, for example, rethinking how to link safety nets to housing policy, devise ways to raise the employability of the poor in self or wage employment, and incentivize the uptake of social services relevant to urban areas, such as daycare, or investing in the prevention of violence or health risks. It is the combination of those domains that may, ultimately, enhance the prospects for upward mobility of the urban poor. Examples from both middle and low-income countries provide some initial evidence on practices to strengthen those sectorial linkages in urban settings. Yet relatively little is known about how these dimensions interact, how to identify and prioritize interventions, and what are the possible trade-offs. 13   The reminder of the paper is organized around four sections. Section 1 briefly lays out key trends, concepts and analytics underpinning the urbanization process and its role in development. Section 2 examines urban poverty quantitatively and qualitatively, and presents household survey estimates on coverage of urban safety nets in different contexts. Section 3 identifies a range of issues that are critical for the design and implementation of urban safety net programs. Such section is structured around six sub-sections, including unveiling experiences with urban poverty assessments and targeting methods, outreach activities and enrollment practices of perspective beneficiaries, program design parameters (e.g., structure of benefits), mobility and portability of programs, considering complementary interventions, and devising institutional arrangements. These practices are based on case studies as well as a range of practices from others contexts as published in the literature and guidance materials. A summary of key design and implementation issues from ten case studies is offered in section 4, including presenting each country case with a consistent format to ease identification of key practices and lessons. In particular, the China case study sets out the remarkable evolution of the urban Dibao program. An unconditional transfer reaching over 21 million people, Dibao is the result of multi-year practice with local-level experimentation and innovation. A second case study takes a fresh look at the initial challenges encountered by the Prospera program as it expanded onto Mexican cities, and the measures that were progressively adopted to manage the process. In Colombia, the paper reviews the conception and performance of a conditional cash transfer in a large metropolitan area like Bogota. The experience of two unconditional cash transfer programs in Nairobi’s slums, Mukuru and Korogocho, is reviewed. There, the programs underscored the importance of combining formal and non-governmental arrangements to operate in such complex environments. The experience of El Salvador’s Programa de Apoyo Temporal al Ingreso (PATI) shows how to provide a combination of public works and skills to select households in the poorest and most violent urban neighborhoods. Key findings from multi-annual research on the Family Rewards program, a conditional cash transfer implemented in New York City, identifies lessons relevant for developing countries. A discussion of an urban safety net in protracted crises underpins the subsection on Gaza’s voucher program, including being recently leveraged to deliver assistance from multiple programs with different objectives (e.g. food, shelter, education materials). The India case study unveils the challenges of providing social pensions in large-scale slums in Delhi. These include unveiling the complex web of dynamics that underpin slum’s socioeconomic texture, and how that affects access to formal public safety nets. The experience of the Philippines and shows how the national Pantawid program was customized to reach specific beneficiary profiles such as the homeless in urban Manila. Finally, the paper discusses the experience of a safety net program like Program Keluarga Harapan (PKH) reaching beneficiaries in both Jakarta and other urban areas in Indonesia. In many ways, the discussion presented in the paper is the tip of the iceberg. We are at the beginning of a long-term agenda, with approaches to urban safety nets epitomizing the ‘science of delivery’, that is, an iterative process of discovery, learning and organic adaptation in approaches. It is against such backdrop that the paper should be contextualized and lessons interpreted, including with a view of identifying emerging practices, learning needs and help advancing practical know-how in an ever- urbanizing world. 14   Section I. Understanding urbanization Section summary. Currently, 3.9 billion people, or 54 percent of the global population, live in urban areas. Out of the ten fastest-urbanizing countries, nine are in Sub-Saharan Africa. By 2050, an additional 2.4 billion people are expected to live in urban areas, with about half of such increase occurring in six countries. Presently, cities generate about 65 percent of global GDP, with the world’s 300 largest metropolitan economies accounting for nearly half of global output in 2014. However, more urbanization does not always mean more economic growth. For instance, while the structural transformation process (which urbanization is part of) followed a more traditional and steady trajectory in some contexts (e.g., East Asia), pathways in others have been more irregular (e.g., Sub-Saharan Africa). For instance, the number of people in developing countries living in slums is about 863 million, or 32.7 percent of their urban population. This implies that the quality and governance of urbanization matters, including the importance of aligning the pace of urbanization with local capacities to manage it. As part of that process, urban development interventions revolve around the engineering and supply of infrastructure, housing and services in given areas; instead, social protection, and safety nets in particular, center on demand-side of the equation by reaching people more directly. Both approaches are part of holistic frameworks for inclusive cities. 1.1 Urban dynamics beyond the tipping point “By the year 2000, over half the world's population is likely to be living in urban areas”. The prediction of the 1979 World Development Report was reasonably accurate1: in 2007, the world reached the tipping-point whereby the global urban population outstripped that of rural areas (figure 1). Currently, 3.9 billion people, or 54 percent of the global population, live in urban areas2. Such rate may increase to 66 percent in 2050, when an additional 2.4 billion people are expected to live in urban areas. Nearly 90 percent of them may be concentrated in Asia and Africa, with about half of such increase occurring in six countries (i.e., India, China, Nigeria, Indonesia, Pakistan and the Democratic Republic of Congo) (UNDESA 2014). The unprecedented pace at which urbanization is unfolding has led to define it as a global ‘force of disruption’ (Dobbs et al. 2015). Figure 1. Population trends and projections, 1950-2050 Source: UNDESA (2014)                                                              1 The 1979 WDR was the second edition of the World Bank’s flagship report and featured “urbanization patterns and policies” as a key issue (World Bank 1979, p.72). 2 Just a few countries are home to half of the world’s urban population. China has the largest urban population (758 million), followed by India (410 million). These two countries account for 30 percent of the world’s urban population and, with another five countries, the United States of America (263 million), Brazil (173 million), Indonesia (134 million), Japan (118 million) and the Russian Federation (105 million), account for more than half of the world’s urban population. 15   In 2014, high levels of urbanization at or above 80 percent characterized Latin America and the Caribbean region. Africa and Asia, in contrast, remain mostly rural, with 40 per cent and 48 per cent of their respective populations living in urban areas. Over the coming decades, the level of urbanization is expected to increase in all regions, with Africa and Asia urbanizing faster (by 1.5 and 1.1 per cent per annum, respectively) than the rest (less than 0.4 percent annually). Out of the ten fastest-urbanizing countries, nine are in Sub-Saharan Africa (with Burkina Faso being the fastest- growing3). In terms of economic groups, upper-middle-income countries have experienced the fastest pace of urbanization since 1950. At the time, only 20 per cent of their population lived in urban areas, while they are now 63 percent urban. This percentage is expected to rise to 79 percent urban by 2050 (e.g., Brazil, China, Iran and Mexico). The current proportion of the population living in urban areas is 39 percent in lower-middle-income countries and 30 percent in low-income countries. By 2050, these countries are expected to be, on average, 57 and 48 percent urban, respectively. The composition of factors affecting urban population growth can play out differently in various countries, including in India and China. In the former, the major source of population growth is still natural growth, including accounting for approximately 40 million of the urban increase in 2001-2011 (figure 2). Yet demographic dynamics, with dropping birth rates, has led to a decline in natural population growth share in cities from 59 percent in 1991-2001 to 44 percent over the last decade. At the same time, the net migration share in urban growth is up from 21 percent over the last decade to about 24 percent, or 22 million people, over 2001-11. The remaining 32 percent of urban growth is due to reclassification of towns and expansion of urban agglomerations (IIHS 2013). In contrast, China’s urban population growth over 2000-2010 (nearly 100 million people) can be attributed to a less extent to natural increase (15 percent), and more to net migration (43 percent) and urban reclassification (42 percent) (World Bank 2014). Figure 2. Sources of urban population growth in India, 1963-2011 (million people) Source: IIHS (2013) More than half of the urban population in developing countries now lives in cities of less than 0.5 million people, and about 60 percent in settings of less than 1 million (figure 3). Nearly one-in-ten live in the 28 mega-cities of 10 million inhabitants or more. The number of mega-cities has nearly                                                              3 See http://citypopulation.de/.   16   tripled since 1990; by 2030, 41 urban agglomerations are projected to house at least 10 million inhabitants each. The fastest growing urban agglomerations are medium-sized cities and cities with less than 1 million inhabitants located in Asia and Africa (Christiaensen and Todo 2013; Ferre et al. 2012). Figure 3. Distribution of population by size of urban areas, 1990-2030 Source: UNDESA (2014a) Just like there is significant diversity around the types of urban areas, i.e., from megacities to small towns, there is also variance on how countries define an urban area from the outset. Indeed governments often use different classification metrics and techniques to classify an area as ‘urban’. Those areas are typically defined through one or multiple criteria, including administrative boundaries, economic indicators (e.g., workers engaged in nonagricultural employment), population dynamics (size and density), or physical characteristics (e.g. paved streets, water supply systems, sewerage systems or electric lighting) (box 1)4. Box 1. Defining an urban area In producing the 2014 update of the annual Urbanization Prospects report, UNDESA noted that 125 of the 233 observed countries used administrative criteria to distinguish between urban and rural areas. Out of these, 65 countries used administrative designations as their sole criterion. In 121 cases, the criteria used to define urban areas include population size or population density, and in 49 cases such demographic characteristics were the only criterion. However, the lower limit above which a settlement is considered urban varies considerably, ranging between 200 and 50,000 inhabitants. In the case of cities, population statistics are often reported in terms of the territory delimited by administrative boundaries; yet these may not necessarily coincide with the extent of the urbanized territory as delimited by other standards. Thus, the ‘city proper’ as delimited by administrative boundaries may not include suburban areas where a sheer portion of the population working or studying in the city resides. Furthermore, two or more adjacent cities may be separately administered, although they might jointly form a single urbanized region. In some case, administrative boundaries of some cities may cover large shares of land primarily devoted to agriculture. Therefore, two auxiliary concepts have often been used to define a ‘city’: the first is the concept of an urban agglomeration, which refers to the population contained within the contours of contiguous territory inhabited at urban levels of residential density. In 2014, city data for 79 of the 232 countries were based on the concept of urban agglomeration. The second concept is that of metropolitan region, which expands the definition of urban agglomeration and includes additional surrounding areas of lower settlement density that                                                              4 A number of procedural and political economy challenges from classifying (or not) an urban areas. These may include, for example, processes to redrawing municipal boundaries as cities and towns expand, including through local notifications on a no-objection basis. Also, once designated a statutory town, local governments may lose preferential treatment in intra-government transfers and public resources. Finally, official urban population statistics may not be consistent between states. 17   are under the direct influence of the city (e.g., through established transport networks, road linkages or commuting patterns). In 2014, for 28 countries data for the capital city were reported in terms of urban agglomeration. Source: UNDESA (2014) 1.2 The economics of structural transformation Nowadays cities generate about 65 percent of global GDP, with the world’s 300 largest metropolitan economies accounting for nearly half of global output in 2014 (Parilla 2015; Dobbs et al. 2012). As such, it is often argued that the development process is closely intertwined with a gradual structural transformation in the sectoral composition of the economy (Kuztets 1966; Lewis 1954). Such process, far from being linear, has prompted longstanding and competing views around the factors, sequence and weight of factors spurring such evolution. Yet cross-country evidence shows a number of empirical regularities accompanying the process. For example, the rise in agricultural productivity at the early stages of development often represents the primary driver of growth and poverty reduction. As average incomes grow, other sectors tend to take-off causing a decline in the share of agriculture in GDP and employment (Timmer 2009)5. For instance, it is estimated that between 1981 and 2001 about three-quarters of China’s poverty reduction performance took place in rural areas (Montalvo and Ravallion 2010). This has lead China to move from an agriculture-based to a transforming economy as defined by World Bank (2008). Figure 4 shows such transformation by examining the dwindling share of agriculture in GDP as the proportion of rural poverty rates declines (which, as we’ll discuss in section 1.3, is a close proxy for urbanization and income growth)6.                                                                5 Country experiences can indeed diverge significantly from these patterns. First, a country may fail to generate economic growth, in which case the pattern might still hold, but the transformation fails to take place. Second, a country might experience an extremely rapid transformation—with a falling share of agriculture in GDP and employment--but not experience much economic growth, so the pattern fails to hold. Third, a country might experience extremely rapid economic growth, but fail to have an equally rapid structural transformation, in which case both the pattern and the commensurate transformation fail to hold. Understandably, the policy implications in each case are radically different (Timmer 2007). 6  In the graph, the share of agriculture in GDP in agricultural, transforming and urbanized economies is 29 percent, 13 percent and 6 percent, respectively; GDP per-capita is $379, $1,068 and $3,489, respectively.    18   Figure 4. Structural transformation in 74 countries, 1990-2005 Source: World Bank (2008a) The transformation process is unfolding rapidly and in very tangible ways in both rural and urban areas. For instance, agrifood systems, which by definition represents a sizable share of the economy in agriculture-based countires, are undergoing a profound transformation making rural and urban areas more connected and integrated (Reardon and Timmer 2014). A vivid illustration of the forces at play is the emergence of supermarkets chains that organize and concentrate production, intermediation and distribution activities. Between 2001 and 2009, the volume of activity of supermarkets in India rose from $0.2 billion to $5.1 billion; in the Philippines, it soared from $1.9 billion to 6.8 billion; and in China it increased from $13.1 billion to $91.5 billion (Reardon et al. 2012). At the same time, food consumption patterns in urban areas have evolved towards higher intakes of fats, processed products, and prepared foods (Popkin 2006). Taken together, such transformations “… have led to a remarkable ‘quite revolution’ in food supply chains within rural areas and from there to urban consumers” (Masters et al. 2014). Figure 5 illustrates the rise of supermarkets in Dhaka and Delhi. Figure 5. Number of supermarkets in Dhaka (left) and their share in Delhi’s retail sector (right) Source: Reardon et al. (2012) 19   While these developments somewhat pertain to a sector or even sub-sector, they convey a broader point about transformation – that is, urbanization is only a component and, to some extent, an outcome of a range of dynamic interactions of multiple factors along the rural-urban spectrum. In other words, the process of urbanization should not be interpreted as the mere generation of cities; instead it needs to be understood within the broader framework of how economic systems change, evolve and develop over time and space. The conceptual underpinnings of urbanization have been matter of ample research and empirical attention across disciplines. Box 2 summarizes some basic urbanization models. Arguably, one of the most relevant explanations is the notion of agglomeration economies – that is, the positive externalities that are generated when activities, firms and people locate near one another (Glaeser 2010). There are three core channels through which agglomeration economies materialize, namely sharing, matching and learning. Agglomeration through sharing occurs when large numbers of firms or workers benefit by drawing on a common pool of resources. These can include public goods and infrastructure, such as transport facilities and education infrastructure. Firms can also share a variety of intermediate inputs, as well as engage in indirect risk sharing and pooling. For example, when firms experience a positive productivity shock, they may expand employment and vice-versa. Agglomerations sparked by matching are generally framed in terms of frictions in labor markets. One way of thinking about the labor market is that it matches different types of workers and firms: the better the match, the higher the benefits to both. Larger cities facilitate the connection between different profiles of workers and firms. Finally, large cities provide more opportunities for people and firms to learn from each other and from the environment around them. Workers may also find it easier to switch jobs, taking valuable knowledge with them, while firms may be able to learn more easily from their suppliers and customers. In some ways, this more fluid exchange of knowledge is a form of sharing. However, learning is distinct from sharing in that both the generation of knowledge and its diffusion benefit from these interactions. A similar argument can be made for the accumulation of skills through learning. For example, young, unskilled workers may become skilled as a result of face-to-face interactions with more skilled and experienced workers as present in large cities7. In short, physical proximity may facilitate sharing, matching or learning and thus lead to agglomeration economies, which spurs productivity and innovation. Yet those linkages are not automatic, as next section will argue. Box 2. Evolution of thinking about spatial economic models Research on the spatial distribution of economic systems spans a wide variety of disciplines, including economics, geography, urban planning, and regional science. This box simply draws out a few of the more important conceptual threads that are interwoven in that body of research. Perhaps the oldest spatial economic model was developed by von Thiinen in the early 1800s. In his work, the spatial land market equilibrium features a land rent gradient that declines monotonically with distance from the town. Given this rent gradient, it follows that production will be organized spatially so that crops are produced within concentric rings: crops with the highest value per unit of land (net of transport costs) will be grown closest to the town, with ever-lower-valued crops produced farther from the town. The von Thiinen model has had a profound influence on the way economists think about the spatial location of productive activities, which extends well beyond agriculture. Another model features ‘central place theory’ has been quite influential in the thinking of geographers, regional scientists, and urban planners. These were developed by German economists Losch and Christaller in the early part of the twentieth century. Central place theory abandons the von Thiinen assumption of constant returns to scale; instead, it explicitly recognizes the existence of firm-level scale economies in the production of nonagricultural goods (typically due to indivisibilities in production technologies). These scale economies combined with transportation costs lead producers to                                                              7 At a global level, such matching process has also been a source of inequality, including as argued by Maskin (2010). 20   cluster together in urbanized central places out of which are satisfied the product demands of consumers dwelling within both the urbanized area itself and the surrounding agricultural hinterland. However, central place theory has been vulnerable to criticism on the grounds that it is not grounded in market fundamentals. In particular, it has been deemed suitable for analyzing optimal rules for collective decisions of a central-planning type, but are quite unsuitable for analyzing the outcome of decentralized decision making. Perhaps for this reason, central place theory generally has been a more popular analytical tool among geographers and planners than among economists. Although already mentioned in the text, ubiquitous in much of modern urban and regional economics is the notion that increasing returns to scale exogenous to firms give rise to ‘agglomeration economies’ that explain the concentration of economic activity in cities. Positive external economies have been attributed to a variety of sources, including complementarities in labor supply and production across firms both within and across industries; smoothing of seasonal fluctuations in demand and input supply; and technological spillovers that reinforce the creation and diffusion of innovations among entrepreneurs. Two related analytical traditions in spatial economics follow directly from consideration of agglomeration economies. The first is the concept of urban centers as growth poles, or nodes of productive activity that essentially generate their own economic gravitational fields. The traditional view is that urban growth poles cause a spread of benefits of agglomeration economies from more productive (i.e., efficient) urban economies to rural areas. The second important concept related to agglomeration economies is the market potential function. The idea here is that firms tend to locate nearest to the maximum number of potential customers. Since the potential customer base includes urban households and firms (in addition to those lying in the hinterland), the possibility for self-reinforcing urban or regional growth is a logical follow-on to this conceptualization of firm location decisions. Finally, recent contributions in spatial economic analysis are being provided under the so-called ‘new economic geography’, including several papers by Krugman. In those models, as in central place theory, the degree to which economic activity is concentrated within an urban space follows from the interaction between transport costs, firm-level economies of scale in the production of manufactured goods, and factor mobility. What differentiates the new economic geography from its predecessors is its ability to formalize the intuitively appealing concepts central to earlier work in this area – central place theory, cumulative causation, and location theory – into a unified framework. In such context, agglomeration economies emerge as the endogenous outcome of the interaction of a small number of economically meaningful parameters, including scale economies, the spatial extent of the market, and the cost of distance. Source: Renkow (2007) 1.3 Urbanization, with and without development In general, cross-country data show that the rate of urbanization and levels of economic prosperity are positively correlated over time. In fact, almost no country has reached income levels of more than $10,000 per-capita before reaching an urban population of about 60 percent (figure 6). Figure 6. Urbanization and per-capita income 21   Source: World Bank (2014a) While urbanization is inextricably linked to economic transformation, more urbanization does not always mean more economic growth. For instance, while the urbanization process followed a more traditional and steady trajectory in East Asia and the Pacific, pathways in some Sub-Saharan Africa countries have been more irregular and pulled in different directions (figure 7). Figure 7. Changes in urban population and GDP per capita, 1985-2010 Source: World Bank (2013a) While an in-depth discussion of the causes of such trends go beyond the scope of this paper, it is important to emphasize that, as mentioned for structural transformation, the relationship between income and urbanization is not linear. For instance, one manifestation of such premise is the widespread phenomenon of slums. For example, cross-country simulations for Africa show that an increase of 1 percent in urbanization will yield a 2.3 percent increase in slums: as put by Arimah (2010, p.16), “… the pernicious effect of urbanization on the incidence of slums is indicative of the process of urbanization without development”. According to the Expert Group Meeting on slum indicators8, “a slum is a contiguous settlement where the inhabitants are characterized as having inadequate housing and basic services. A slum is often not recognized and addressed by the public authorities as an integral or equal part of the city”. Therefore, slums also include squatter settlements. These are created by the illegal occupation of land and are in contravention of official building regulations. Relatedly, slum households were defined as “a group of individuals living together under the same roof and lacking one or more of the following conditions: access to improved water; access to improved sanitation; structural quality/durability of dwelling; sufficient living space that is not overcrowded; and security of tenure”. Urban growth often occurs so quickly that basic information about slum populations are not available. Sometimes slum populations are intentionally excluded from household surveys because informal settlements do not have legal recognition (Save the Children 2015). Latest available estimates show that, as illustrated in figure 9, the number of people in developing countries living in slums is about                                                              8 Hosted by the UN in 2002, the Group assembled 35 international professionals as well as staff members of the Urban Secretariat and UN-HABITAT with the objective to contribute to the development of indicators for the “Cities without Slums” (Target 11) of the Millennium Development Goal 7, “Ensure Environmental Sustainability”. 22   863 million, or 32.7 percent of their urban population9 (figure 8) (UNHABITAT 2013). In 2012, slum prevalence – or the proportion of people living in slum conditions in urban areas – was highest in sub- Saharan Africa. In that region about 62 percent of the urban population lives in a slum. In South Asia, slum prevalence is about 35 percent, while in Latin America and the Caribbean, 24 percent of the urban population was classified as living in slum conditions. The lowest slum prevalence is observed in North Africa, with a level of 13 per cent. Figure 8. Level and proportion of urban populations living in slums, 1990-2012 Source: UNHABITAT (2013) Slum settlements may have differing degrees of marginalization depending on the recognition – or level of recognition – by public administrations10. For example, in India the 2011 Census differentiated between ‘notified’, ‘recognized’, and ‘identified’ slums: the first category represents those legally notified under relevant ‘Slum Acts’ in different states; the second are those that an authority of the state recognizes formally as a ‘slum’ in some form, either within a policy or through a survey; finally, the third category is those that the census counts as slums but that lack either notification or even a ‘recognition’ by a state authority. The latter represents the largest category, including 37 percent of identified slums (Bhan et al. 2014). But here it is important to mention that slums may often not be legally recognized and, as such, generating institutional quandaries on whether and how to support people in need living there. Given the overall limited public support to those environments, slum dwellers may tend to pay higher prices for, among other services, privately- supplied water. We’ll come back to these issues in section 2. While an extensive discussion on slums goes beyond the scope of this report, it is here important to underscore that slums can elicit differing views. The debate is sparked by the wide recognition of slums’ grim living conditions, but also by the fact that slums may offer relatively low-cost housing and potential proximity to work (World Bank 2013b). For example, evidence from Pune, India, shows that poor households prefer to live in centrally located slums with closer proximity to jobs, although these are located on riverbanks that are prone to flooding (Lall and Deichmann 2009). Yet in other contexts such as Kenya, puzzles emerge around low-quality but high-cost of slum shelter. One study, for instance, estimated that while 92 percent of Nairobi’s slum residents are tenants and crowded in 2.6 persons per room, rents are significant: in 2004, slum households paid ‘structure owners’ an                                                              9 Urban slum population was 760 million in 2000 and 650 million in 1990. 10 See Hehta and Dastur (2008) for a compilation of slums approaches (adaptive and proactive) and case studies. 23   average rent of about US$11 per month, a sum that amounts to US$31 million for the year (Gulyani and Talukdar 2008). Whether and how those dynamics are helping or hurting the poor is a matter of debate, with views ranging from seeing the Earth as a “planet of slums” (Davis 2006) to deliberate positions “in praise of slums” (Kenny 2013). 1.4 Quality of urbanization The discussion around economic and spatial transformation implies that the quality and governance of urbanization matters for development. Indeed, evidence and practice underscore the importance of synchronizing the pace of urbanization with local capacities to manage it, including in terms of supply and regulation of housing and services (Lall et al. 2007). In this regard, it is important to identify key factors and bottlenecks shaping urbanization in a given context, including through robust diagnostic tools. The identification of those constraints lies at the heart of approaches like the Urbanization Reviews (box 3). Those tools examine three main dimensions of urban development: first, the planning of urban areas is central, including charting a course for cities by setting the terms of urbanization, including policies for urban land, basic infrastructure and public services. A second key function is connecting, or making a city’s markets (labor, goods, and services) accessible to other cities, markets and to neighborhoods within the city. Finally, the financing of urban areas includes funding for large capital outlays needed to provide infrastructure and services as cities grow and urbanization gathers speed. For the framework of planning, connecting, and financing to work, a solid governance structure is a prerequisite. Box 3. Urbanization Reviews Applying the 2009 World Development Report policy framework, the World Bank’s Urbanization Reviews offer diagnostic tools to identify policy distortions and analyze investment priorities. Each review starts by assessing a country’s or region’s spatial transformation: how the urban economy is evolving, how demand for the city is changing with economic development, the pace of new arrivals, and how these new arrivals into the city are finding places to live and commuting to their jobs. It then compares the city’s observed patterns with benchmarks in other places or with past conditions. Such comparisons help reveal how policy distortions constrain urbanization and how investment shortfalls restrict the benefits from it. Once the review has identified the possible constraints and shortfalls, it proposes policy options. It aims to show how a city can harness economic and social benefits not just today, but in the future, as economies grow, technologies change, and institutions are strengthened. To test the relevance of the tools and policy framework in different development circumstances, the World Bank has piloted the Urbanization Reviews in more than 10 countries at varying stages of urbanization Source: World Bank (2013b) While the urban development agenda does have a direct bearing on poverty, these are often regarded as distinct and sectoral issues. For example, a review of urban planning documents in the Indonesia’s cities of Surakarta and Makassar found that “… most stakeholders, particularly the local government work units, continue to regard spatial planning and poverty reduction as two unrelated subjects. In addition to this, most [of local government officials] have a sectoral view of poverty reduction and therefore regard these matters as the sole concern of other agencies who are responsible for the social sector and activities within it” (Sambodho et al. 2013). The relationship between urban development and poverty becomes particularly clear when we move across the various gradients and sizes of urban areas. As shown in the case of Indonesia, the poor who reside in inner city areas tend to experience increased levels of welfare compared to the poor in other locations in the city such as periurban areas. Indeed, the significance of the positive influence of the inner city spatial context lies in the relatively sound infrastructure conditions, a reduced disaster risk, and greater access to the city’s economic 24   resources such as markets, factories, or employment opportunities. Conditions like these increase the ability of the poor to protect and develop their livelihood assets and exemplify the relationship between poverty and urban planning (table 1). Table 1. Urban poverty and spatial planning in Surakarta and Makassar cities, Indonesia Spatial aspects Inner-city areas The Suburbs: river Peri-urban and new Implications for of urban banks of coastal/marine development areas planning poverty Arrangement and -Arrangements of -Arrangements of -Provision of decent -Revitalization and provision of settlements settlements in areas close housing for poor renovation of settlements -Risk of fire in to rivers or in coastal newcomers infrastructure in slum densely populated areas (formal/informal areas areas -Arrangement of workers) -Development - Issues with magersan settlements in flood - Arrangements of planning and settlers (Surakarta) prone (Surakarta) and settlement for flood settlement control in -The lack of coastal inundation victims (Surakarta) illegal and social/public space (Makassar) areas -The growth of illegal uninhabited areas due to high population settlement in unsuitable - Disaster migration density areas (swamps and plan for flood and landfill sites) fire disasters The provision of - Limited access to - Limited s to clean - The poor condition of - Provision of clean clean water and clean water and water and sanitation in clean water and water and sanitation sanitation sanitation in slum slum areas sanitation infrastructure in slum areas including waste areas - Pollution and - No integrated waste - Control of waste management - Issues of accumulation of garbage management system in pollution from environmental in rivers illegal settlement areas households and home sanitation; - Accumulation of industries slaughterhouses garbage in settlement - Waste management located in the areas along river banks and disposal systems residential areas in slum areas (Makassar) - Storage of garbage and waste management is not maximized Land tenure - Land ownership Land ownership status - The development of - Legalization and status status (private/family, (settlements on the sea) illegal settlements in certification of land tenant) - The process of uninhabitable areas for settlement in - Possibility of compensation as part of (swamps and landfill public or private eviction due to a relocation program sites) areas occupying from the flood plains - Control of illegal private/government settlements in areas land that are uninhabitable (swamps, over rivers, at landfill sites) Economic - Continued limited - Continued limited - Continued limited - Revitalization of integration of the access by the poor to access by the poor to access by the poor to coastal natural poor formal sector formal sector formal employment resources or changing employment employment - Job security and social the livelihood of opportunities -Degradation of natural protection for formal fishers as part of a resources (SDA) and a workers (Makassar) transition into the decline in the economic urban economy potential of fisheries - Program (Makassar) improvement and certification of 25   human resources to allow them to enter into formal employment Source: Sambodho et al. (2013) In a stylized way, urban development revolves around the engineering and supply of infrastructure, housing and services in given areas; instead, social protection, and safety nets in particular, center on demand-side of the equation by reaching people more directly. Yet as we’ll see for a number of cases studies in section 4, e.g. El Salvador and India, there are several pressure points where the agendas converge, such as in measuring urban poverty or land tenure issues. In this regard, several frameworks have been set out for conceptualizing ‘inclusive’ urban development approaches. For example, building on Blomquist (2014), incorporates the three pillars of urbanization reviews mentioned in box 3 (i.e., planning, connectivity and financing) and supplement it with a fourth pillar on inclusiveness. The latter might encompass broader issues at the intersection of enhancing the economic productivity of the poor (e.g., skills) and social equity, but it also include a range of social protection interventions relevant for urban contexts. Pending on how these four pillars interact and the capacity to manage them, they may generate positive or negative externalities associated with urbanization. The level and balance of such effects may in turn affect overall levels of urban poverty, prosperity and quality of life more broadly. During the course of the next sections, other complementary frameworks will be laid out, including on synergies between social protection and other dimensions, as well as on the specific role of safety nets in urban contexts. Section II. The challenge of urban poverty Section summary. Poverty is rapidly urbanizing: while the number of the urban poor remained stable at around 285 million people over 1990-2008, the share of urban poverty out of total poverty increased from 16 percent to 23 percent over the same period. Urban poverty exhibits some peculiar characteristics in terms of context, livelihoods and sources of vulnerability and poverty profiles. These are often characterized by high income and physical mobility, including seasonal migration; constrained commuting to jobs; poor housing conditions, increasingly located in high-risk prone areas; access to low quality and congested social services; and socio-economic marginalization, including perception of employment discrimination because of living in fragile/crime areas. The coverage of safety nets among the urban poor is systematically lower in urban areas relative to rural settings: about 16.6 percent of the urban household in the poorest quintile are covered by some form of safety net program, as opposed to 23.4 percent of the poorest rural quintile. The difference in safety nets coverage of the poorest quintile grows to up to 24 percentage points in upper-middle income countries. The coverage of urban social insurance and labor markets is double that of the rural poo, although it is of very low level (3-4 percent). In terms of program composition, public works seem to be overwhelmingly implemented in rural areas, the coverage of conditional cash transfers is slightly higher in urban areas, and the relevance of unconditional cash transfers, targeted subsidies and ‘other social assistance programs’ is considerably higher in the urban portfolio relative to the rural areas. 2.1 The urbanization of poverty Consistently with the discussion in section 1.2, as urbanization unfolds poverty tends to decline and “urbanize”. Indeed, the urbanization process often proceeds in conjunction with two other trends: first, there is a reduction in global poverty. Such reduction occurrs in both urban and rural areas, although the prevalence of poverty remains significantly higher in the latter. In 2008 the average poverty rate in rural areas was 29.4 percent, while in urban areas was less then half of it, or 11.6 percent (table 2). In both contexts, poverty rates nearly halved between 1990 and 2008. 26   Table 2. Poverty rates by area ($1.25/day line), 1990-2008 1990 1996 2002 2008 Rural Urban Rural Urban Rural Urban Rural Urban East Asia & Pacific 67.5 24.4 45.9 13.0 39.2 6.9 20.4 4.3 Europe and Central Asia 2.2 0.9 6.3 2.8 4.4 1.1 1.2 0.2 Latin America & Caribbean 21 7.4 20.3 6.3 20.3 8.3 13.2 3.1 Middle East & North Africa 9.1 1.9 5.6 0.9 7.5 1.2 4.1 0.8 South Asia 50.5 40.1 46.1 35.2 45.1 35.2 38.0 29.7 Sub-Saharan Africa 55 41.5 56.8 40.6 52.3 41.4 47.1 33.6 Total 52.5 20.5 43.0 17.0 39.5 15.1 29.4 11.6 Source: World Bank (2013c) Second, the other accompanying trend involves a growing share of urban poverty as a percentage of total poverty, primarily due to slower reduction given the lower initial conditions and higher population growth in those contexts (Spence et al. 2009; Ravallion et al. 2007). Indeed, the share of urban poverty in global poverty increased from 17.9 percent to 24.4 percent over the 1990-2008 period (World Bank 2013c)11. Figure 9 illustrates these trends. Figure 9. Urbanization and poverty Source: World Bank (2013c), Ravallion et al. (2007) In terms of distribution across urban settings, most global poverty is concentrated in medium and small towns (World Bank 2013c). Recent cross-country analysis examined the distribution of poverty across the urban spectrum. In general, it shows that only a fraction of the poor live in megacities, while the majority is located in medium and small towns. For example, in Brazil 72 percent of the poor live in urban areas but, surprisingly, only 9 percent of them reside in megacities. The bulk of poverty is concentrated in medium (17 percent) and very small towns (39 percent). Of course, there are exceptions: in Mexico, 16 percent of the poor live in larger cities (more than 1 million inhabitants);                                                              11 For example, in India the share of population living in urban areas increased from 19.9 percent to 31.2 percent over the past three decades. Almost simultaneously, poverty decreased by 6.2 percent. Yet, the number of urban poor increased by 34.4 percent while the rural poor decreased by 15.5 percent, which equals to a net increase in the share of urban poverty from 18.7 to 26.8 percent (GoI 2011). 27   also, in the least urbanized regions such as Sub-Saharan Africa, 40 percent of the urban poor live in the largest urban area (e.g., Niger, Senegal, Sierra Leone, and Swaziland). 2.2 Poor areas or poor people? Such urbanization of poverty is likewise evident when absolute numbers are considered. Yet some interesting and somewhat different results emerge. Indeed, while the number of rural poor declined from 1.42 billion in 1990 to 896 million in 2008, the number of urban poor r (figure 10). At one point, in 2002, in Latin America and the Caribbean the number of urban poor (32.2 million) even outpaced that of rural poor (25.5 million). Based on simple projections, the global number of urban poor may exceed the number of rural poor before the year 2030, including between 2023 (model based on short-term trends) and 2032 (model based long-term trends). In relative terms, the share of urban poverty out of total poverty increased form 17 percent in 1990 to 24 percent in 2008. Figure 10. Urban poverty numbers and share ($1.25/day line), 1990-2008   Source: author’s calculations based on World Bank (2013c) and population figures in WDI online.   Similarly, when national poverty lines are considered, 86 out of 91 developing countries for which data is available show higher poverty rates in rural areas (points above the dotted line) as opposed to urban areas (countries below the dotted line) (figure 11).   28   Figure 11. Poverty rates in urban and rural areas, latest available data 80 Zambia Congo, D.R. 70 60 Rural poverty (%) Mexico Honduras 50 40 Ghana Azerbaijan 30 20 India 10 Chile 0 0 10 20 30 40 50 60 70 80 Urban poverty (%) Source: author’s calculations based on national poverty lines and population figures in WDI online. We then examine the absolute number of poor people. In particular, figure 12 plots the ratio between the number of urban poor to rural poor against the level of urbanization (numbers above 1 indicate higher numbers of urban poor people compared to rural). The size of the bubble indicates the total population in the country. Three interesting observations emerge. First, in about one-quarter of the sample (i.e., 22 countries or 23 percent) there are more poor people in urban areas than in rural areas. Second, the analysis confirms what we observed when considering poverty rates – i.e., a strong correlation between urbanization and urban poverty levels (R2=0.6). Third, such a tendency seems to accelerate as countries become predominantly urban, i.e., they surpass the 50 percent urbanization threshold: before that point, in only one case (Swaziland) the number of urban poor are higher than the rural ones. Conversely, after an urbanization rate of 75 percent, in no country the number of rural poor is higher than the urban poor.   29   Figure 12. Ratio of urban to rural poverty numbers by urbanization level, latest available data 4.00 Jordan Ratio of number of urban poor/number of  3.50 3.00 2.50 Mexico rural poor 2.00 1.50 Armenia Indonesia Colombia 1.00 El Salvador India 0.50 Nigeria S. Africa 0.00 10 20 30 400 50 60 70 80 90 Ethiopia ‐0.50 Turkey Burundi Urbanization (%) Morocco Source: author’s calculations. The graph doesn’t include Chile and Uruguay where the urban/rural ratio is 75 and 11, respectively. These findings are also consistent with studies investigating the dynamics of chronic poverty over time. For example, Lucchetti et al. (2014) propose a definition of chronic poverty whereby a household is defined so if its income was remained below US$4/day (2005PPP) in 2004 and 201212. The study finds that, when considering poverty rates, living in rural areas is associated with higher poverty persistence over the 8-year period. However, when considering the absolute number of people, chronic poverty becomes more pervasive in urban areas. In Brazil, for example, for every one chronic poor in rural areas, there are two that live in urban areas. Some of these findings for the region also find echo in qualitative data. Based on surveys included in the Latinobarometro online database13, figure 13 shows the correlation between urbanization and the share of respondents who perceive ‘poverty and social inequality’ to be the ‘most important problem in the country’.                                                                12 The study did not control for possible spells of in-and-out poverty that have been documented in the literature, e.g., the influential papers by Yaqub (2000) and Baulch and Hoddinott (2000). 13 http://www.latinobarometro.org/latOnline.jsp.   30   Figure 13. Perception of poverty and urbanization in Latin America and the Caribbean 10 8 Share of respondents (%) 6 4 2 0 5 10 15 20 25 30 35 40 45 Urbanization (%) Source: author’s calculations based on data from Latinobarometro and WDI online. At its core, the issue of poverty prevalence and numbers suggests that further thinking is needed on how best to allocate resources aimed at poverty reduction – that is, whether to target poor areas or poor people14. While each approach tells its own story, the growing interest in both perspectives provides more nuanced information on intra-country variation in poverty than was previously available. This might be an important development to shape the agenda of social protection programs in urban areas for the years to come. The issue of urban inequality is, arguably, less researched than urban poverty. However, a growing number of studies are investigating inequality from different perspectives. For instance, new estimates show that average incomes of the highest income quintiles are magnified by city size, so that income inequality as measured by Gini index is increasing with urban size (Behrens and Nicoud 2013). In other words, larger urban areas generate more wealth and are at the same time more unequal than smaller cities (figure 14). Figure 14. City size and inequality Source: Behrens and Nicoud (2013)                                                              14 There is here an interesting parallel with the debate on international assistance in middle-income countries (Kanbur and Sumner 2011). 31   Similarly, new evidence from 65 developing countries suggests that urban-rural gap accounts for about 40 percent of mean country inequality and much of its cross-country variation (Young 2013). For example, recent analysis on a sample of 37 countries shows that almost 80 percent of them have higher Gini values in urban areas than rural (UNHABITAT 2008). For instance, differences can be as high as 0.1 for countries such as Bangladesh, Cambodia, Nepal, Uganda, Mozambique and Botswana. Differences are also noteworthy in terms of human capital. For example, in some countries like Tajikistan and Venezuela, the education gap between the richest and the poorest quintile is greater in urban than in rural areas. The gap is widest in Venezuela, where pupils from the richest urban families have, on average, almost 8 years more schooling than those from the poorest ones, compared with a gap of 5 years between the wealthy and poor in rural areas (UNICEF 2012). 2.3 Is urban poverty different? Urban poverty exhibits some peculiar characteristics in terms of context, livelihoods and sources of vulnerability and poverty profiles. For instance, being poor in an urban setting is often associated with higher physical and income mobility, even among the poorest households. The wider availability of short-term employment entails that poor urban laborers tend to seize opportunities scattered over multiple activities, locations and within short timeframes. Populations, therefore, tend to be highly mobile and dynamic. As a result of mobility, migration and competition, in some instances poverty tends to be more dynamic and transitory than in rural areas. For example, in urban Mexico only 7 percent of extreme poor households in 2002 still had the same status in 2007 (Rascon and Rubalcava 2008); in Indonesia about 20 percent of households being initially surveyed in urban areas cannot be found in the same residence within a period of 6 months; similarly, recent data shows that about a third of urban Indonesian residents moved in or out of poverty in less than a decade (World Bank 2011). Given those features, it is important to closely examine labor market dynamics. Preliminary analysis of survey data from 61 developing countries shows that labor force participation rates in urban areas are lower than in rural settings. On average, about 40.8 percent of the urban workforce is inactive compared to 34.2 percent in rural areas. As shown in figure 15, such differences attenuate as countries’ income per capita grows. Figure 15. Labor market activity in urban and rural areas 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% li_urb li_rur lm_urb lm_rur um_urb um_rur Employed Unemployed Inactive Source: ASPIRE data. Note: li=LIC; lm=LMIC; um=UMIC 32   The composition of employment in urban and rural areas is also different. Despite some regional variations, wage and salaried workers account for 60 percent of urban employment, own account worker (or self-employed) for about 30 percent, business owners who employ other workers (“employers”) for 4.3 percent, and unpaid workers for 6.4 percent. In contrast, self-employment accounts for 44 percent of rural employment, while unpaid work constitutes over 20 percent of rural jobs. As income per capita increases, the percentage of wage workers or employers soars, while the share of own account workers and non-paid employees tend to fall. In low income countries, almost 60 percent of urban workers are own account or non-paid employees (table 3). Table 3. Employment composition in urban and rural areas by region Wage and Own account Employers Unpaid employee salaried employee urban rural urban rural urban rural urban rural Regions 59.0 35.2 30.3 43.6 4.3 3.3 6.4 20.9 East Asia & Pacific (13) 62.0 25.3 28.5 58.3 4.5 6.1 6.0 13.4 Europe & Central Asia (13) 82.3 61.6 8.9 16.7 3.6 2.4 1.7 20.9 Latin America & Caribbean (19) 62.6 42.0 28.8 39.9 5.1 5.1 3.7 12.9 Middle East & North Africa (8) 66.1 54.8 19.4 22.8 7.4 5.4 2.8 21.8 South Asia (8) 58.5 35.7 30.0 36.3 2.6 0.8 10.1 30.1 Sub-Saharan Africa (33) 43.5 18.6 43.4 56.5 3.6 1.7 9.7 26.1 Country income groups Low income countries 39.4 15.7 45.4 52.2 3.9 1.9 13.2 33.2 Lower middle income countries 59.2 33.8 30.7 46.6 4.5 4.1 6.3 21.2 Upper middle income countries 73.7 51.2 20.2 34.7 4.3 3.1 2.6 12.3 Source: ASPIRE data Labor activities can also vary by city size. For example, for youth the prevalence of self-employment, family work and agricultural activities falls with city size and private sector waged employment increases with city size (figure 16). In small towns up to 83 percent of working youths are self- employed or in unpaid family work, but this share falls to 24 percent in Addis Ababa (World Bank 2015a). Among the youth, survey data on a sample of respondents show that almost no unemployed youth said they were planning on making a living in self-employed activities (just 2 percent). However, four times this number end up in self-employment just four months later. Box 4 discusses some specific features around youth unemployment in Addis Ababa.   33   Figure 16. City size and job types in Ethiopia, 2012 Addis Ababa Other big towns Small towns Proportion of active labor foe 70 60 50 40 30 20 10 0 Government or Private Private Paid domestic Self-employed Unpaid work for Agriculture NGO (temporary) (permanent) work family Source: World Bank (2015a) Box 4. Aspirations and livelihoods among the unemployed youth in Addis Ababa In 2012, unemployment rates among males aged 15 to 24 in Addis Ababa were 21 percent, and 36 percent among those who have just graduated high school. What are their aspirations for employment opportunities? What do they do and how do they survive while they are without work? How do the youth find jobs? A recent poverty assessment conducted by the World Bank in Ethiopia investigates those questions. Survey data identifies two different types of unemployed youth: lower educated youth who are native to Addis Ababa and higher educated recent migrants. The first type of unemployed youth was primarily sampled in slum areas in non-central sub-cities of Addis Ababa and the second type was primarily sampled around the job vacancy boards. Findings show that the unemployed youth of Addis Ababa rely heavily on money from their parents, particularly those who have just graduated or moved to Addis Ababa. Those with degrees, and recent migrants, are getting three times the financial support from their parents than someone who was born in Addis, or had no education. Families provide savings to youth on graduating or on moving to the capital in order to support themselves while they look for work. Recent migrants are far less likely to have savings, formal or informal than those who are unemployed and native to Addis. On average, they have only enough money to survive a few weeks on their own savings, at their regular rates of expenditure. More than half of those who were unemployed and remained after 4 months, had engaged in temporary work during this time. Half of casual/daily jobs for men are in construction. Only one fifth (22%) of unemployed youth did not work at all during 4 months. Well-educated individuals were no less likely to have taken work over the 16 weeks during which they were tracked. As a result many well-educated are engage in temporary jobs (such as those in the construction sector) for which they are over qualified. Unemployment contains considerable boredom on a daily basis for unemployed youth. The unemployed spend on average two thirds (16 hours) of their time, or half of all time not sleeping, in their own homes or yards. Of those waking hours, remarkably, respondents report spending at least 3 hours per day on average “doing nothing”, even after having been asked about 20 different activity categories, and asking about any other time spent that had not been accounted for. This time spent doing nothing does not include all other leisure activities reported (on average 3.4 hours a day), nor does it include time spent socializing with friends (1.7 hours per day on average). Visiting vacancy boards is the most common form of job-search method and the one that yields the highest probability of finding a permanent job. Job seekers usually try a range of different routes into work, including asking their social networks and going door to door to ask businesses for vacancies. However, vacancy boards and newspaper, particularly vacancy boards, are the most common forms of job search. They are used by 44% of the urban unemployed, compared to 22% who ask friends or relatives for a job. They are particularly used for finding permanent jobs: although 38% of jobs found by unemployed youth in Addis Ababa were found at job boards, 69% of permanent jobs were found at job boards (compared to 63% and 31% found through networks respectively). Searching at vacancy boards can be time consuming and expensive involving many visits to the central vacancy boards, each of which costs more than the median daily expenditure of unemployed youth. Those with lower levels of education that do not visit vacancy boards state it is because they will not find work there (60%) reflecting the fact it is more often skilled jobs that are posted on the boards. The majority of those with higher levels of education that do visit more but the costs of transport were prohibitively high (82%). Of the sample of unemployed youth sampled at vacancy boards, 83% 34   had stopped visiting these boards after 4 months because it was too expensive to travel to the board. The average cost of a trip to the town center to look for work is estimated to be 15 Birr which is higher than the average median expenditure of 14 Birr per day among the two samples of unemployed youth. One in four educated, active job-seekers secured permanent employment in four months of search, but rates of success are much lower for those who are less educated and less actively looking for work. After 16 weeks 21% of the type 2 unemployed, had found permanent jobs compared to only 6% of the type 1 unemployed. This means that type 2 unemployed will stay in unemployment and poverty for a longer period of time. Among those well-educated, actively seeking work, one third (32.8%) had been unemployed for 6-12 months and almost one fifth (18.9%) for longer than this. Among those native to Addis, 35.3% had been unemployed for longer than one year. Rates of discouragement are much higher among the type 1 unemployed. Source: World Bank (2014c) The informal sector in cities is often extensive and, according to a recent estimate, it accounts for more than half of total jobs (Ghani and Kanbur 2013). In India, for example, almost 60 percent of total urban employed are wage workers, and 67 percent of this category are informal wage workers (IIHS 2013). For example, as noted by the case study for India (section 3), over half of Delhi’s workforce is informal, most of which is living in slums (figure 17). Figure 17. Percentage of formal and informal employment in Delhi, 2009-10 Formal Sector Informal Sector Household/Other 59.91 56.82 35.82 32.28 31.91 32.8 32.42 10.38 7.67 Total Male Female Source: Bhattacharya et al. (2014) The wider availability of short-term employment entails that the opportunity and transaction costs for poor households tend to be significant, including in terms of time, forgone income and transportation costs. For example, in the context school enrollment in Mexico it was argued that “…the opportunity cost of switching from employment to schooling is much higher in urban than in rural areas, where the [Oportunidades] scholarships were between one half and two thirds of children's full time wages” (Attanasio et al. 2008, p.6). In Lima and Rio de Janeiro most of the poor live 30-40 kilometers from employment hubs, entailing an average commuting time of 3 hours per day. In Montevideo, residents living in slums outside the city cite the lack of access to public transport as a major constraint to accessing jobs. In Kampala, many motorized transport options are unaffordable for the poor, with transport fares accruing to 41 percent of incomes for the poorest 20 percent of Kampala‘s population (in fact, 70 percent of urban workers walk to work) (World Bank 2015b). There could also be other types of indirect costs. For instance, working conditions and lifestyles in urban areas may affect the poor in ways that pose a considerable ‘cognitive tax’ – that is, people may have limited ‘bandwidth’ due to perpetual concerns on daily transport, weekly rent, monthly pay, and limited family or community support. These may contrast with the more seasonal, peaks-and-troughs 35   patterns in rural livelihoods (box 5). While specific behavioral research on urban poverty is limited, it might be plausible that the urban poor experience a cognitive tax of different nature than the rural poor’s. Box 5. The cognitive tax among the urban and rural poor: insights from behavioral economics New experiments in the field of behavioral economics are investigating whether and how monetary poverty can directly impede people’s cognitive functions. For example, one recent study in an urban context surveyed poor and rich shoppers at a New Jersey mall. By proposing hypothetical scenarios, researchers induced thoughts about finances, e.g., “your car is having some trouble and requires $X to be fixed. You can pay in full, take a loan, or take a chance and forego the service at the moment... How would you go about making this decision?” These scenarios, by touching on monetary issues, were meant to trigger thoughts of the participant’s own finances. After viewing each scenario, and while thinking about how they might go about solving the problem, participants performed tasks used to measure cognitive function, such as the Raven’s test (i.e., a common component in IQ tests that measures the capacity to think logically and solve problems independent of acquired knowledge). The results found that thoughts about finances reduced cognitive performance among poor participants, but not in well-off ones. A second study was fielded in a rural context and examined the cognitive function of farmers over the planting cycle. It was found that the same farmer showed diminished cognitive performance before harvest, when poor, as compared with after harvest, when better-off. This cannot be explained by differences in time available, nutrition, or work effort, and stress. Instead, it appears that poverty itself reduced cognitive capacity – that is, poverty-related concerns consumed mental resources, leaving less for other tasks. Taken together, these studies suggest that, as the authors put it, “… being poor means coping not just with a shortfall of money, but also with a concurrent shortfall of cognitive resources. The poor, in this view, are less capable not because of inherent traits, but because the very context of poverty imposes load and impedes cognitive capacity. The findings, in other words, are not about poor people, but about any people who find themselves poor”. Such perspective may have important implications for urban safety net programs, particularly in terms of weighing, on one hand, conditionalities or co-responsibilities (in case envisaged) and, on the other hand, adding cognitive taxes on the poor. Source: Mani et al. (2012) Transaction costs can also be amplified by specific phenomena such as violence. In Mexico, in some instances street violence makes it necessary for children and youth to take public transport to school (Latapi and de la Rocha 2004). For similar reasons, residents in low-income areas of Dar es Salaam were found to spend between 10-30 percent of their income on transport (Dudwick et al. 2011). More broadly, insecurity, crime, gender-based domestic violence and intergenerational conflict tend to loom large in urban settings and generate social and economic costs. It is estimated that 30 percent of hospital admissions in Latin America are the result of urban violence, while the associated health costs account for up to 5 percent of GDP in Colombia. Relatedly, living in informal settlements is reported as a key source of anxiety of the poor, including daily fears of violence and abuse. The overcrowding of poor-quality housing in marginal areas often further perpetuates marginalization, discrimination and neighborhood stigma. For example, a unique 30-year longitudinal study from Brazil shows that favelas residency is the most widely perceived stigmatizing factor, including among 96 percent of the interviewed households. Comparing within and across generations, the study shows that, although there have been notable improvements in a range of dimensions, the stigma of favela life is reflected in a lack of return on educational investment and earning differentials between favelados and non-favelados in the same areas of the city (Perlman 2010). As a favela dweller in Sao Paulo put it in the Voices of the Poor report, “… one day a company called me for a job, but when they realized I lived in Bode [a favela] they changed their minds, thinking that I was one of those marginais they couldn’t trust”. 36   Cost of living in urban areas, including that of food, housing and other expenditures, can be considerably higher than rural settings. Urban economies are complex, market-based and integrated. Poor urban and rural households generally devote a similar share of income on food (i.e., 50-70 percent of household budget). However, urban households’ access food almost entirely through market-based transactions, a feature also referred to as “commoditization”. For example, in urban Peru the share of households’ income spent on market-purchased food is three times higher than that of rural households. While urban dwellers are more likely to be shielded from seasonal fluctuations in domestic commodity prices, they tend to be more exposed to global food price volatility15. Moreover, the nutrition transition towards higher calories, fats and pre-prepared foods introduces new forms of health and nutrition risks, including obesity and cardiovascular diseases. Expenditures on housing – and rents in particular – represent another relatively larger cost item for the urban poor. For example, in urban Nepal dwellers spend about 28 percent of their income on housing while rural households about four times less than such level, or about 7 percent (Garrett and Ersado 2003). Also, the poorest are more likely to rent than the better off. For example, in Kampala 78 percent of households in the poorest quintile rent their homes, compared with 63 percent for the richest quintile (Lall 2010). Access to basic services such as water, electricity and sanitation tends to improve as city size increases. In Vietnam, the difference between the share of population with access to piped water in the largest versus the smallest towns is around 25 percentage points; in India and Brazil, such difference is 20 and 10 percentage points, respectively. There is also uneven access within cities of similar size. For example, in Dhaka only 9 percent of households in the poorest quintile benefit from a sewage line, and less than a third of them access piped water16 (World Bank 2013, 2009). In terms of provision of urban social services, major challenges revolve around their low quality, saturation and households’ inability to afford them. In rural areas, the availability of services is often limited, hence reducing the options for people to access a specific service. In urban areas, the density that cities offer makes it cheaper to expand services – evidence from 78 countries shows that it cost significantly less to provide piped water in urban areas as opposed to sparsely populated settings. But density also entails limited scope for spatial expansions of new infrastructure. As a result, the competition for those services is high. Studies suggest that prospects for better social services is, alongside higher urban wages, a key determinant of migration, hence contributing to service congestion As it was observed, “… having access to a pit latrine is not the same in a rural setting where it is used by one family and can be sited to avoid contaminating water sources, and urban settings where 50 households share it” (Haddad 2012). In some contexts, it is even worse. For example, a recent report showed that in Monrovia’s West Point shantytown, which is home to more than 40,000 people, only five public toilets are available (Save the Children 2015). During the rainy season, which includes rainfalls for over 20 inches (50cm) per month, the water flows through the streets, mixing with feces and contaminating the wells most people rely on for drinking water. As discussed in section 1 on the quality of urbanization, when the supply of land, housing and services does not keep up with the rising demand from growing populations, low-income households often                                                              15 Exposure to food price volatility has been a factor that contributed in igniting urban riots and instability in a number of countries with pre-existing socio-political discontent (Barrett 2013). 16 Differences between and within cities can be significant. For example, in Argentina the share of population with “unsatisfied basic needs” was less than 10 percent in the federal capital, while the average in 25 smaller urban areas was almost 17 percent. Yet a wide diversity emerged within Buenos Aires itself, where between 8 and 26 percent of the population lacked access to basic services. 37   resort to slums. For example, Kariuki and Schwartz (2005) analyzed data from 47 countries (93 locations) and found that the average water prices charged by private vendors compared with the public network were 4.5 times higher in peri-urban or unplanned settlements with unclear tenure. In Accra’s slums, the cost of water from private vendors can amount to 10 percent of households’ monthly income (Dudwick et al. 2011). In illegal settlements in Turkey, dwellers faced prices from 9 to 30 times higher than publicly-provided services (Leitmann and Baharoglu 1999). The rapid expansion of cities, as well as the management of densities within them, requires institutions that manage land effectively and strategically. These include a transparent system to convert land use, a clear definition of property rights, a robust mechanism of land and property valuation, and a strong judicial system. However, the capacity to formulate, oversee and enforce standards is generally limited in developing countries, leading to haphazard and unplanned urbanization trajectories. As such, increases in settlements can often occur near risky locations, e.g., railways such as in Pakistan, as well as disaster-prone locations like riverbanks. In Dakar, for example, the fastest population growth over the last 20 years occurred in peri-urban areas, and 40 percent of that population growth happened on high-risk lands. Indeed, the issue of disaster risk looms large in urban areas (table 4). This process, with low-income households occupying the most hazardous urban land, is not static. Data from Cali, Colombia, shows that localized hotspots of small scale disaster events change as inner-city neighborhoods gentrify, governments improve hazard management, and new informal settlements form at the urban periphery (World Bank 2013b). Table 4. Selected large-scale crises impacting urban areas, 2003-2014 Country Shock Year Main cities Number of affected Total damage affected people (Mill) (US$ bill) Lebanon Conflict/Syrian refugees 2014 Multiple 1.16 n.a. Japan Earthquake 2011 Multiple 0.27 210 Haiti Earthquake 2010 Port-au-Prince 3.4 n.a. China Sichuan earthquake 2008 Multiple 45.9 85 Myanmar Cyclone Nargis 2008 Yangon 2.4 4 Indonesia Java earthquake 2006 Yogyakarta 3.1 3.1 India Floods 2005 Mumbai 20 3.3 Pakistan Kasmir earthquake 2005 Muzaffarabad 5.1 5.2 United States Hurricane Katrina 2005 New Orleans 0.5 125 Iran Earthquake 2003 Bam 0.2 0.5 Source: ED-DAT database; IFRC (2010) Overall, the urban population exposed to cyclones is estimated to rise from 310 to 680 million between 2000 and 2050; similarly, the number of people facing earthquakes risk may increase from 370 million to 870 million over the same time period17 (IFRC 2010). At the same time, urban protracted crises are also increasingly frequent, with a growing share of conflict and conflict-related crises occurring in urban areas (see box 6). In other words, the issue of ‘fragile cities’ is becoming a central issue in the humanitarian system.                                                                17 For other urban-specific crisis data, see also the “Urban Social Disturbance in Africa and Asia”. The dataset covers 55 major cities, 23 in Sub-Saharan Africa and 32 in Central- and East Asia, in 49 different countries for the 1960-2006 period. A total of 3,375 events have been recorded for these cities over 1960-2006, of which 1,378 involved lethal casualties (http://www.prio.org/Data/Economic-and-Socio-Demographic/Urban-Social-Disturbance-in-Africa-and-Asia/). 38   Box 6. Supporting Syrian refugees in urban Lebanon with electronic vouchers In some cases, the scale of refugee operations can reach a magnitude that could be considered a de-facto urban settlement. The current humanitarian operation supporting Syrian refugees in Lebanon is a case in point. Part of the assistance package includes the provision of electronic food vouchers by WFP, reaching about 882,850 individuals as October 2014 (or 75 percent of the entire refugee population in the country). The program delivers about US$30 per person, for a total of US$345 million disbursed in 2014. Through its network of 416 shops, the program is spurring significant know-on effects in the local economy, including the generation of nearly 1,300 jobs. Participating supermarkets and small retailers have made capital investments for meeting the additional demand, such as enhancing storage and space capacity, for about US$3 million. Overall, the program is expected to generate US$517 million in the economy. Vouchers were delivered in partnerships with the private sector (i.e., Mastercard®), with online monitoring and cost reconciliation systems that would ensure payments of merchants within 48h from transaction. The range of technological innovations introduced with the program is also helping to inform the Emergency National Poverty Targeting Programme (E-NPTP) in the host country of Lebanon, hence charting an interesting pathway for sharing practices in a complex urban environment. Source: Sanogo (2014) Municipal governments have a range of responsibilities for basic infrastructure service delivery, land use, economic development planning, and provision of social services. As a result, they tend to vary significantly in accounting practices and their capacity to deliver services. For example, in Colombia about 1,100 municipal governments have responsibilities for basic infrastructure service delivery, land use, economic development planning, and provision of social services. While municipal tax collection has risen with decentralization and administrative reforms across all categories of cities, small and medium towns sometimes may have not kept pace with larger cities in their ability to increase local revenues. As a result, municipalities tend to vary significantly in accounting practices and their capacity to deliver services. An outcome of such diversity is the uneven capacity to coordinate interventions and assign clear responsibilities among government levels in highly- populated areas (Samad et al. 2013). For example, a mapping of safety net programs in urban and rural India revealed the multiple objectives, target groups, sub-targets, interventions, and the intricate web of interactions between them (figure 18). Figure 18. Mapping the urban and rural safety net in India Source: IIHS (2013) 39   The issue of fragmentation is central to the system-building agenda of social protection (World Bank 2012); but a further, explicit recognition of the rural-urban dimensions may introduce an additional layer of complexity, or make such complexities more visible. As noted by Bhan (2013, p.3) “… the emergent urban moment is not without its own agenda and seeing where entitlements-based social safety nets fit in within this agenda is no simple task”. As the next section shows, while the rural arm of the net has been subject to much investment, there is a basis for urban safety nets to build upon and enhance performance in those contexts. Differently from rural areas, cities tend to generate more limited community arrangements, social networks and support mechanisms. Especially in new settlements and slums, barriers such as different languages, lack of family connections and the dynamic in-and-out-flow of temporary migrants, for example, can create substantial pockets of social exclusion and marginalization (see India case study in section 4). At the same time, the type of ties are forged in urban contexts that seem at first glance anecdotal may actually be the result in a deliberate strategy, that is, with migrants coalescing and establishing closer bonds with populations with similar ethnic, religious and political background (Saunders 2011). Therefore, urban informal networks often exist, but may take different shapes. In Ethiopia, for the example, the ‘idir’ system is widely present in urban contexts. Looking ahead, various social media are generating new forms of urban connections that significantly depart from more traditional mechanisms of social connection and interactions. This has been particularly evident in recent social upheavals, although may find broader applications and diffusion in daily urban dynamics (Box 7). Box 7. Urban activism and social media The emergence of Tahrir Square in Cairo as a beacon for recent revolts is a testament to how place and history come together in unexpected ways. As of today, there has been little in-depth work exploring the cyclical and reciprocal relationship between social media, traditional media, and the urban spaces in which uprisings come into being. For instance, in the case of Egypt’s wave of the Arab Spring, the first spark that mobilized the Egyptian uprising on January 25, 2011 was the popular Facebook page “We are all Khaled Said”, launched in June 2010. That page offered Egyptians an interactive platform for documenting human rights violations, making anti-government claims, and mobilizing support. As the subscribers grew in numbers and interacted, the Facebook page transformed into an organizing hub for the Egyptian uprising, disseminating the call to protest (on January 25th, 2011). In addition, one other main organizational online hub was Twitter, where activists used the hash tag #Jan25 to invite others to join the conversation as well as organize amongst themselves. On Twitter, many activists discussed and planned the day of January 25th by using the @ reply function. In fact, it was observed that many of these tweeters considered themselves “citizen journalists” and made it their mission to get the word out with a flow of videos and pictures, which created a new type of urban activism mediated by means of internet and mass communication. As the tweeters navigated between virtual and physical space, continual updates from protesters in Egypt sustained the uprising in Egypt. As noted by Alsayyad (2012, p.63) “… what happened in Tahrir Square during the uprising shows that even in the 21st century urban space remains the most important arena for dissent and social change. The new spatial arrangements articulated in the cities of the Arab world during the recent uprisings have brought back the ‘urban question’ underscoring the need to study the new ‘urban qualities’ that influence the grievances, organizational forms, and consciousness of insurgent citizens”. Source: Alsayyad (2012) 2.4 Social assistance in urban areas: preliminary estimates from survey data Based on household survey data from 112 countries included in the World Bank ASPIRE database, this section provides a brief snapshot of the comparative urban-rural coverage of social protection programs. The analysis will center of safety nets (or social assistance) and is disaggregated by country income groups, regions, household poverty status, and type of safety net programs. 40   Overall, the coverage of safety programs is lower in urban areas compared to rural settings. On average, 21.3 percent of the urban population in low and middle income countries is covered by some form of safety net; in rural areas, that rates is 27.7 percent. In terms of reaching household in the poorest quintile, safety nets cover 16.6 percent of the urban bottom 20 percent, which is considerably lower than the 23.4 percent of the poorest rural quintile (figure 19). Figure 19. Coverage of social protection programs, poorest quintile 30 25 23.4 20 16.6 (%) 15 Urban 10 Rural 4.3 3.3 5 2.3 1.5 0 Safety nets (social assistance) Social insurance Labor market programs Source: World Bank ASPIRE dataset (April 2015) Similar estimates emerge when absolute measures of poverty are adopted, such as the $1.25/day poverty line: in such cases, only 15.4 percent of the urban poor are covered relative to 25.1 percent of the rural poor, a difference of almost 10 percentage points18 (table 5). Table 5. Average coverage of social protection programs by poverty status   Poor Non-poor Total Social assistance Urban 15.4 22.3 21.3 Rural 25.1 29.2 27.7 Social insurance Urban 4.3 18.1 16.1 Rural 1.9 7.5 5.4 Labor market programs Urban 3.2 3.4 3.3 Rural 1.6 2.4 2.1 Notes: data refers to average coverage in low and middle income countries. Estimates for coverage are post-transfer. Poverty refers to $1.25PPP poverty line. Source: World Bank ASPIRE dataset (April 2015) Data shows that, across quintiles, the coverage of safety nets in urban settings is lower than in rural areas for all cases except Q4 (table 6). They also show that the coverage of social insurance is consistently higher in urban relative to rural areas; social insurance seems regressive, with coverage                                                              18 There are minor differences in coverage between the poor as defined by the ‘bottom 20 percent’ or those living below the absolute poverty line ($1.25/day) for Low Income Countries and Lower Middle Income Countries; yet those differences are sizable for Upper Middle Income Countries and High Income Countries. 41   rates increasing from 2-4 percent in Q1 to 18-34 percent in Q5. Labor market programs, instead, are found to be generally of low coverage, including being below 4 percent throughout income groups in both urban and rural contexts. Table 6. Global coverage of social protection programs by quintiles Q1 Q2 Q3 Q4 Q5 Total (poorest) (richest) Social assistance Urban 16.6 20.3 24.2 27.2 18.0 21.3 Rural 23.4 27.1 26.4 26.5 34.9 27.7 Social insurance Urban 4.3 5.5 11.9 24.2 34.4 16.1 Rural 2.3 1.4 1.9 3.4 18.0 5.4 Labor market programs Urban 3.3 4.0 3.9 3.1 2.4 3.3 Rural 1.5 1.7 2.2 2.3 3.0 2.1 Source: World Bank ASPIRE dataset (April 2015) Importantly, there are significant income and regional variations in coverage. For example, the difference in safety nets coverage of the poorest quintile grows from about 9 percentage points in LICs to about 24 percentage points in UMICs (figure 20). Figure 20. Coverage of safety net programs, poorest quintile by country income group 90 80.2 80 70 63.3 60 56.8 52.1 50 (%) 40 30 22 21.7 20 13.2 12.4 10 0 Low income countries Lower middle income Upper middle income High income countries countries countries Urban Rural Source: World Bank ASPIRE dataset (April 2015) In terms of regional coverage, however, the scenario seems more mixed (figure 21). For instance, in some cases, like in the East Asia and Pacific and Sub-Saharan Africa regions, the safety net coverage of the poorest quintile in urban and rural regions is similar, with the former displaying a slightly higher urban coverage and the latter a higher rural one. In other contexts, differences are significantly marked, including the Latin America and Caribbean and South Asia region where, respectively, urban coverage is 14.5 and 26.8 percentage points lower. An opposite trend is observed in the Middle East and North Africa, where urban coverage of the poorest is 20 percentage points higher in urban areas, 42   including because of generous subsidy schemes such as the Baladi Bread and Ration Card programs in Egypt. Although decreasing over time, evidence points to “… an urban bias in the distribution of food subsidy consumer benefits in Egypt, (…) [with] per capita benefits from food subsidies were about 10 percent higher in urban than in rural areas (LE 197 a person a year versus LE 178 a person a year)” (World Bank 2010, p.15). Figure 21. Coverage of social protection programs, poorest quintile by region 70 63.9 57.2 58.2 60 54.2 53.6 49.4 50 46.6 38.5 40 27.5 30 20 8.6 10.1 10 0.7 0 Latin America East Asia and Middle East and Sub‐Saharan Eastern Europe South Asia and Caribbean Pacific North Africa Africa and Central Asia Urban Rural Source: World Bank ASPIRE dataset (April 2015) In terms of individual classes of safety net programs, survey data allowed for disaggregated analysis for 8 interventions. These include unconditional cash transfers, social pensions, conditional cash transfers, unconditional in-kind transfers (including food), school feeding, public works, targeted subsidies, and other social assistance programs19. When analyzing the urban-rural differences in program composition, some basic patterns emerge (table 7): first, public works seem to be overwhelmingly implemented in rural areas. There is very limited survey evidence of urban public work programs, which explains an almost 10 percentage point difference between urban and rural coverage rates. Second, the coverage of conditional cash transfers is slightly higher in urban areas: although the coverage is only a 2 percentage point difference, this is almost double that of rural areas (where coverage is 2.5 percent). Third, when looking at the share of programs’ coverage within the total coverage for that areas, some further findings surface: for example, the relevance of unconditional cash transfers, targeted subsidies and ‘other social assistance programs’ is considerably higher in the urban portfolio relative to the rural areas.                                                              19 Unconditional transfers except non-contributory social pensions; public works include cash and food for work; general subsidies are excluded from the analysis, which only includes targeted subsidies for food, utilities and transport, as well as fee waivers; finally, ‘other social assistance programs’ include scholarships and other unspecified non-contributory programs. 43   Table 7. Global composition of safety net portfolios in urban and rural areas Urban Rural Coverage % of total urban Coverage % of total rural coverage coverage Unconditional cash transfers 5.5 22 4.1 13 Conditional cash transfer programs 4.5 18 2.5 8 Social pensions 2.1 8 2.8 9 Unconditional in-kind transfers 2.9 11 2.2 7 School feeding 2.2 9 3.4 10 Public works 0.1 0 9.9 31 Targeted subsidies 4.9 19 4.7 15 Other social assistance programs 3.2 13 2.7 8 Source: World Bank ASPIRE dataset (April 2015) Those patterns are particularly marked in lower middle income countries. While low income countries and upper middle income countries have generally consistent compositions in urban-rural program portfolios, low-middle income countries feature, on one hand, a more sizable urban share of targeted subsidies and CCTs (i.e., accounting for a combined 70 percent of the regional portfolio); in rural areas, on the other hand, more than half (55 percent) of the program composition features public works, including large-scale programs such as India’s National Rural Employment Guarantee Scheme (figure 22). Figure 22. Safety net programs in urban (left) and rural (right) areas in lower-middle income UCT OSA UCT OSA CCT SP UIT TS SF CCTs TS SP UIT SF PW PW Source: World Bank ASPIRE dataset (April 2015). Note: PW= public works; UCT = unconditional cash transfers; CCT= conditional cash transfers; OSA = other social assistance; TS = targeted subsidies; SF= school feeding; UIT= unconditonal in-kind transfers Against this background, the next section will explore key issues in design and implementation of safety nets in urban areas. 44   Section III. Key issues in design and implementation of urban safety nets Section summary. Poverty assessments increasingly point to the importance of considering the issue of ‘density’. For example, the scope and focus of interventions can change quite remarkably pending on whether consumption-based poverty is expressed as prevalence in a given area or in terms of number (or density) of people: these two are not the same – areas which are very poor may also be sparsely populated. Multidimensional approaches have been often used to measure and identify taxonomies of poor urban areas, including in terms of space, services and employment. Generally, urban programs use multiple targeting methods to select and prioritize potential beneficiaries. A recurrent question is whether and how to adapt proxy means tests (PMT) approaches to urban contexts. Country case studies show that formula for rural populations may not accurately target urban poor and vice versa. Another frequently-raised issue concerns reaching households with characteristics that are seldom captured in PMT models – e.g., not what kind of kind of materials their house is made of, but whether they have a house at all, i.e. the homeless or street children. Some programs, e.g., Philippines, were designed to specifically reach those populations. Beyond targeting, a number of steps should be put in place to ensure program ‘uptake’ by perspective beneficiaries. A key emerging lesson is the need for extensive communications and outreach tailored to each urban community, like those set out in Mexico and Colombia. The experience of the United States show that enrollment may be especially challenging for particular categories of urban dwellers, such as the elderly and working poor. Relatedly, a distinguishing feature of the urban poor is that they are often more mobile than poor people living in rural areas, including as a result of seasonal migration. This raises the issue of portability of benefits. Technology can represent both an opportunity as well as a barrier to portability, depending on the interoperability of systems deployed. A significant proportion of people do so not reside within the ‘formal’ and ‘legal’ system, hence raising challenges on how to support people lacking documentation of residence. To unbundle such complex issue, countries like India are considering a set of ‘notifications’ for gradually regularizing illegal settlements while providing public assistance. The governance arrangements of social protection across countries are a product of legislative and political processes reflecting fundamental cultural preferences, historical initial conditions, and technical considerations. When central governments devolve responsibilities for financing and administration of social assistance to local governments, including urban municipalities, this can create both opportunities and challenges. Relatedly, the physical proximity of municipalities with different levels of institutional and financing capacities can generate spatial inequities among neighboring areas. Given the universe of different actors and operators, local governments can help ensure coordination and coherence among the web of local organizations. These often play a key role in supplementing and integrating state-level capacities, especially in times of distress. Social intermediation services do not bring directly material benefits to the families; they instead stand right in-between the demand and supply of social services and facilitate access to programs. The services address the “choice overload” problem that prevents extreme poor from effectively using the social protection system, including because of challenges – frequent in urban areas – such as limited awareness on existing interventions, high opportunity costs to accessing them, distrust or lack of familiarity with formal bureaucracies, etc. Urban social safety nets would need to be better integrated with interventions in the spatial, economic and social realms. Examples from Brazil, Philippines, Kenya, Haiti, and Colombia provide some initial evidence on practices to strengthen those sectorial linkages in urban settings. This section discusses core practical issues that underpin urban safety net programs. In doing so, we draw from both case studies presented in section 4 and illustrations from other experiences. The first sub-section examines the key function of identifying the urban poor, including assessments, targeting and definition of program objectives. Subsequently, we turn the attention to outreach efforts and enrolment of perspective beneficiaries into the program. While recognizing that some of these functions can overlap, they may help to locate them within an program overall cycle. The following section lays out core issues around mobility and portability, followed by a discussion on institutional and governance arrangements, including the role of different government levels, Quasi-formal and community-based organizations, and social intermediation services. This section concludes with examples of possible complementary interventions in spatial, economic and social domains. 45   3.1 Identifying the poor Assessments Assessments of risks, vulnerabilities and needs provide the core diagnostics to inform design choices in safety nets. In this regard, there are various specificities to measuring urban poverty. For example, surveys are generally representative at the urban or possibly city level, while in some contexts the speed at which cities expand and contract could make census data rapidly outdated. Urban poverty is often of localized and territorial nature, which may require a granular approach in measurement as well as frequent ‘refreshers’. In some cases, like observed in the Philippines (see section 4), poverty assessments may take place in contexts where administrative maps and on-the-ground reality may not fully match. Households may live in settlements (or even in specific floors of multi-store buildings) not present on officials planning documents. Hence in a number of circumstances, data collection activities were carried out that are representative at a disaggregated level within cities, as well as representative for subgroups in the population. This may allow for spatial analysis, which is critical for urban planning and targeting of poor areas (World Bank 2011). Poverty assessments also point to the importance of considering the issue of ‘density’. This pose an additional choice to be made in terms of not only selecting metrics and indicators, but also how they are expressed. For example, the scope and focus of interventions can change quite remarkably pending on whether consumption-based poverty is expressed as prevalence in a given area or in terms of number (or density) of people. In fact, such choice has deep implications for planning and allocation of resources for poverty reduction. Although the mainstream approach is to consider rates of poverty, country studies clearly illustrate the difficult choices faced by policymakers: they may have an interest both in areas where poverty is high, but also in areas which have the most poor (Murthi 2014). These two are not the same – areas which are very poor may also be sparsely populated. For example, in Vietnam the district and commune poverty maps based on headcounts or prevalence support the argument that assistance should be targeted on less densely populated, largely forested, ethnic minority areas. On the other hand, density-based analysis shows that poverty is highest where the population density is highest, that is, in the two delta areas of Vietnam and along the coastal lowland areas. Figure 23 below presents the results from mapping such approaches. Figure 23. Poverty map in Vietnam based on rates (left) and numbers (right) Source: Swinkels and Turk (2007) 46   Similar contrasting scenarios have been found in a number of other countries, like Estonia and Slovenia. Take the latter for example, a map of which is proposed in figure 24. Based on poverty rates (left map below), under EU rules the eastern half of the country is considered less developed and the western half more developed. Yet when we consider poverty in absolute terms, there is a visible concentration of poor people in urban areas, including the capital city of Ljubljana (the dark red patch in the middle of the country, right map below). This concentration is typical of many capital cities which may be better off on average but still home to large numbers of poor people. Under EU rules, Ljubljana, like many capital cities in Central and Eastern Europe, has limited access to cohesion funds. As put by Murdhi (2014), “… this may suggest the need to devise more refined approaches of allocating resources for poverty reduction, with capital cities and urban areas more widely given greater eligibility as many of the poor live in better-off areas”. Figure 24. Poverty map in Slovenia based on rates (left) and numbers (right)   Source: Murthi (2014) Another issue in urban poverty assessment is the choice of indicators. Traditional appraisals often revolve around a specific aspect of urban poverty, such as consumption and expenditure measures. As shown in reviews of practices in South Africa, Nigeria, Jamaica and Vietnam, the focus on a specific aspect of urban poverty stems from the complexity and costs of the assessments (Baker and Schuler 2004). More recently, a number of assessments defined poverty as a multidimensional phenomenon and thereby rely on multiple indicators for its analysis. Although methods to select and aggregate indicators vary, multidimensional approaches have been used to identify taxonomies of poor urban areas, such as in Romania (box 8). The issue of multidimensionality is also linked to our ‘framing’ discussion in the introductory section of this paper (figure 25), and it is relevant for other issues discussed later in this section. Indeed, urban poverty tends to constrain people in terms of space, services and employment, although the way these dimensions interact can vary over time and location. For example, figure 25 shows diamonds to track performance over different urban poverty dimensions in slums of Nairobi and Dakar20. The analysis shows that while poverty incidence, unemployment, low education and housing quality affect slum dwellers, important variations exists, with slum residents in Nairobi being better educated and those in Dakar enjoy better spatial conditions (Gulyani et al. 2010, p.4).                                                              20 The analysis by Gulyani et al. (2010) also includes more disaggregated information for housing living conditions and other related variables examined with the use of polygons. 47   Figure 25. Urban diamonds for Nairobi (left) and Dakar (right) Source: Gulyani et al. (2010) These considerations have important conceptual and operational implications. Indeed, they confirm that the focus on welfare poverty measures alone, which are often the basis for safety net programs, are a necessary but insufficient conditions to achieve broad-based urban poverty reduction. They also show that challenges may vary from city to city (and even from neighborhood to neighborhood), and that a context-specific approach is required to assess the multiple dimensions of urban poverty. Box 8. Combining census and qualitative data: a typology of marginalized urban areas in Romania A range of urban poverty diagnostic tools were recently completed in Romania. These suggested three main criteria for defining and analyzing poor urban areas, namely human capital (and population), employment and housing conditions. Specifically, the tools used both the terms ‘disadvantaged’ and ‘marginalized’ urban areas: the former are areas which meet one or two of the abovementioned criteria. In contrast, urban marginalized areas refer to areas where all three criteria. A total of seven indicators were used for those criteria. Using data of the 2011 Population and Housing Census, those indicators were calculated for each urban census sector (the Census contained 50,299 census sectors in urban areas, with an average of 216 inhabitants in each). The threshold for indicators was set at the 80th percentile (table below). Dimension/criteria Indicators Urban threshold (80th percentile) Human capital Proportion of population in the census sector between 15 and 64 22.1 years that completed only 8 grades of school or less Proportion of persons with disabilities, chronic diseases or other 8.0 health conditions that make their daily activities difficult Proportion of children (0‐17 years) in total population 20.5 Employment Proportion of persons aged 15‐64 years neither in formal 22.2 employment nor in education Housing Proportion of dwellings not connected to electricity 0.0 Proportion of overcrowded dwellings (<15.33 square meters per 54.7 person) Insecure tenure: proportion of households that do not own the 12.3 dwelling 48   A census sector is considered disadvantaged on human capital if it has a relatively high concentration of at least two of the following groups: working‐age population with poor education, children, and people with disabilities, chronic diseases or other health conditions. A “high concentration” means that the share of the respective group in the total relevant population of the census sector is among the highest 20% values of all urban census sectors in the country. A census sector is defined as having low formal employment if its share of the working age population that is without formal employment and not in education is among the highest 20% of all urban census sectors in the country. Finally, a census sector is defined as having poor housing if any two of the three housing indicators have values above their thresholds. That is, if it has a relatively high concentration of at least two of the following groups: (i) people living in dwellings not connected to electricity (ii) people living in overcrowded dwellings, and (iii) households with a low degree of security of plot tenure. "Concentrate" here again means that the share of the respective group in the census sector is one of the highest 20% values of all urban census sectors in the country. In urban areas, 1,139 census sectors met the criteria of being an urban marginalized area. These are located in 264 cities and in the capital Bucharest. Some 342,933 people live in these areas. Through qualitative research, these areas included four subtypes. The first is ‘ghetto’ areas of low‐quality blocks of flats or in former workers colonies. This includes low quality housing facilities built before 1990 for the workers of the former socialist large enterprises. Most often these are small to medium sized buildings (housing 150‐500 inhabitants) concentrated in one or two low quality blocks. Usually, apartments in ‘ghettos’ include only one room of 9 to 15 square meters9, overcrowded with furniture and numerous families with many children. Usually these small rooms are used as a bedroom, living room, office and kitchen, for large families of up to 11 members. In spite of this, many of them are clean and tidy on the inside. In addition to poor living conditions, ghetto residents face three major problems. These include paying for utilities, fear of being evicted from the room due to overdue debts (for tenants of social housing), and weak support networks and bad reputation of the area (low social capital). The second subtype of urban marginalized communities concern slum areas of houses and/or improvised shelters. These are located in old neighborhoods at the outskirts of towns and cities with very poor communities that include Roma and non‐Roma. These often have extended in size since 1990. In addition to low‐quality houses made of adobe, many additional improvised shelters were put together over time, either within the courtyards of the old houses or on public areas. These shelters are often made of plastic and paperboard with some wooden frames. Houses and shelters are very small, between 6 and 40 square meters, but accommodate large families with many children. The houses are situated randomly, one next to the other, with very little space between them. These types of areas are usually located next to a river or train tracks. The main problems for these areas, in addition to extreme poverty and miserable housing conditions, are the lack of identity papers and property documents. The third subtype of urban marginalized communities includes modernized social housing. These are well endowed with infrastructure and utility services (sometimes better than the rest of the urban areas) but accommodate poor people in difficult social situations that are eligible for these houses. Yet they are isolated. If the location of a social housing area is torn from the vital tissue of the city, away from income earning opportunities, and if it is inhabited by a single socio‐ economic or ethnic group hen– in spite of modern housing conditions ‐ the area is segregated and has little development potential. The fourth subtype includes social housing in buildings in “historical city centers”. These are old neighborhoods, where inhabitants have lived for more than 30 to 35 years. Except for the location in the city central areas, the living situation of these communities to a large extent resembles slum areas. Some people were allowed to stay in some ruined buildings but were not given identity papers as tenants living at that address, given that the building was administratively registered as ‘destroyed’. This implies that that person cannot get a job, has no right to medical care or social benefits, and so on. Unlike the other types of urban marginalized areas, historical city areas are not inhabited by communities with strong ties, intense daily interaction or leaders. Instead, small nuclei of neighbors live in these areas who need to be treated on an individually basis. Source: World Bank (2013r) In El Salvador, employment and education dimensions were integrated with security and crime- related variables. These provided the backbone for the PATI program examined in section. In particular, the Urban Poverty and Social Exclusion Map represents a rigorous statistical and 49   geospatial effort led by the Ministry of Economy. Using information from the 2007 VI Population Census and the V Housing Census, the Map produces geo-referenced census data allowing analysis down to urban block-level (figure 26). Figure 26. Map of areas by level of precariousness in San Salvador Classification of AUPs in the municipality of San Salvador Zoomed-in view of the AUPs in zip code 01, San Salvador Source: Rodriguez-Alas et al. (2014) The identification of ‘precarious urban settlements’ (AUPs) is based on five steps. First, precarious households are identified. Precarious households are identified using the Unsatisfied Basic Needs (UBN) method based on housing indicators. This includes four main dimensions, namely physical space relative to the size of the family (overcrowding), inadequate access to potable water, inadequate sanitary infrastructure, and precarious housing building materials. To classify a household as precarious it needs to fulfil a minimum of two UBNs out of four. On average, there are 20.9 households per block, with 59.5 percent of them classified as precarious. Second, precarious blocks are identified. Blocks are the smallest territorial unit or micro area from the census. In order to control for population size in different blocks, a method was devised whereby a block with 5-9 households is ‘precarious’ when more than half of its households are themselves precarious (as identified in step 1); for blocks of 10-29 households, these are precarious when 30- 49.9 percent of households are so; and for more populous blocks of over 30 households, these are precarious when 10-29.9 percent are so. Third, AUPs are identified. Once individual precarious blocks have been selected, the next action is the identification and delineation of agglomerations of such blocks. Once these conglomerates are identified, they are delimited by or ‘assigned’ to municipalities; within such clusters, AUPs are identified by grouping neighboring precarious blocks with a minimum of 50 precarious households 50   (on average, AUPs include 279.7 precarious households). At that point, an 8-digit unique identifier is assigned to each AUP. Fourth, AUPs are divided in four clusters. Once the AUPs have been identified, they are ranked based on the index of residential marginality (IMARES) and the index of social exclusion (IEXCS). The former reflects the relative housing deprivation faced by AUP residents and it includes 5 housing indicators (with values ranging from 0 to 100, or the highest level of precariousness). The latter index, the IEXCS, measures household-level social exclusion in terms 18 employment, human capital and access to social services-related categories. For this index, a 3-points scale (with 3 being the highest exclusion level) is used to score each category and calculate average score. At this point, k-means method for cluster analysis is conducted to generate conglomerates of AUPs. The results are four levels of AUP clusters precariousness (extreme, high, moderate and low). As a result of this process, some 2,508 AUPs were identified in urban El Salvador, including 495,981 households with a population of more than 2 million; this represents more than half of the urban population and the equivalent of the total rural population. About 19 percent of AUPs show a level of extreme precariousness and 32 percent a high level of it. Fifth, and finally, AUPs are ranked within each cluster. This is achieved by creating an index of social-economic stratification (IESOCE) which is the average of two indicators, one measuring household-level assets (TV, phones, refrigerators, etc.) and another one for educational attainments of the head of the household. Targeting The choice of targeting methods depends heavily on the specific objectives of the social protection intervention. Program designers must have a clear vision on who are the intended beneficiaries before they start assessing alternative methods. For example, does the proposed program seek to reduce destitution among elderly widows? Does it aim at increasing children’s human capital while providing income support to families? Or assist unemployed youth? A clear statement of program goals is a necessary precursor to discussions regarding targeting methods and implementation. In urban localities, the choice of methods can be shaped by several contextual factors. One is the geography of urban poverty. In some towns and cities, the urban poor are concentrated in select areas. In others, they are widely dispersed and in still others, they move frequently from place-to-place. For example, in Addis Ababa poverty is not concentrated in specific locations (kebeles), but instead is spread out throughout the city. Very few kebeles have poverty rates higher than 50 percent or lower than 10 percent (Figure 27). A proxy means test (PMT) model was constructed to assess whether poverty status could be accurately predicted in Addis Ababa using a few easily observed characteristics of a household (Olinto and Sherpa 2014). Eligibility defined through a PMT system seemed to work quite well. Simulations show that nearly all beneficiaries in a program of 500,000 would be in the bottom 50 percent and three-quarters would be below the poverty line where PMT targeting was used. 51   Figure 27. Poverty prevalence in Addis Ababa, 2012 Source: World Bank (2014c) Where poverty is concentrated, active beneficiary selection – for example, conducting a census of all households to obtain information needed to determine eligibility – might be feasible and desirable if budget and logistics would allow so. Indeed, a number of current urban safety nets adopt a door-to- door (saturation) approach in selected urban areas, such as the ‘pockets of poverty’ and the ‘poverty hotspots’ in the Philippines’ Pantawid program (see section 4). But if the poor are dispersed or mobile, active selection can be considerably more costly to implement, as the Mexican experience illustrates. Another issue is whether targeting should be conducted on the basis of households or individuals characteristics. In rural areas, the registered beneficiary of social protection programmes is often the household head; the assumption being that (s)he will share benefits with other household members who are usually family members. Households in urban localities are more demographically diverse, ranging from single-person households to nuclear families to collections of loosely-related individuals. How this should be approached depends in part on the specific objectives of the intervention. For example, the Colombian and Mexican conditional cash transfer programmes described in section 4 seek to improve children’s human capital and, as such, target the households in which children live. By contrast, El Salvador’s PATI program which seeks to assist unemployed youth targets individuals. Generally, urban programs use multiple targeting methods to select and prioritize potential beneficiaries (see table 8 for a summary of the comparative advantages of different targeting methods). In the case of PATI, the intervention provides temporary income support ($100/month for 6 months) to vulnerable urban populations in exchange for their participation in skills training programs and community projects. The existence of detailed poverty maps allowed urban neighborhoods to be targeted at block-level. Within selected localities, interested individuals pre- register to the program. The requirement for individuals to do so, along with a program wage rate below market rates for unskilled work, acts as a screening device; in other words, individuals self- select into the program. Program officials visit these individuals at home to verify residency and conduct a simple means test and prospective beneficiaries were then prioritized using predefined criteria including gender and age. Both government and nongovernment actors take part in the final stage, a verification process to generate a prioritized list of participants. About 72 percent of PATI beneficiaries belong to the two poorest income quintiles, 45 percent lived in extreme poverty and 63 percent were women. 52   Table 8. Comparative merits of targeting methods Targeting Appropriate Advantages Limitations Urban method circumstances considerations Means testing - where declared income is - in the best of cases, - requires high levels of Generally excludes verifiable very accurate literacy and informal sector - where administrative documentation of workers and, thus, capacity is high economic transactions, has been noted to - where benefits to preferably of income discourage transition recipients are large enough - administratively from informal to to justify costs of demanding where there formal sector. This administering means test are meaningful attempts does not necessarily at verification mean less labor or - Targeting programs to productivity, but less the monetary poor may visibility. (e.g., not ultimately address Bolsa Família, different dimensions of Brazil; Child poverty Support Grant and Old Age Grant, South Africa) Proxy-means - reasonably high - verifiable, may allay - may seem arbitrary to As housing is often testing (PMT) administrative capacity concerns over some a main component - programs meant to address politicization or - requires large body of of PMT the chronic poverty in stable randomness of benefit literate and probably approach is less situations assignment computer-trained staff, accurate for renters, - where applicable to a large - uses readily moderate-to-high levels particular those who program or to several observable household of information and shift residence programs so as to maximize characteristics technology frequently. return for fixed overhead - is less likely than - inherent inaccuracies at - welfare and -good availability of data means test to affect household level, although income can change (recent census and work effort good on average rapidly in urban household survey) - insensitive to quick settings and the changes in welfare, as in a PMT captures crisis or in some indicators that are transition countries somewhat static - Targeting programs to - recertification in the monetary poor may urban settings needs not ultimately address to be done more different dimensions of frequently than in poverty rural setting and data - availability of data in collection costs can developing countries is be high (e.g., frequently limited (census Familias en Acción, or household surveys) Colombia) Community - where local communities - takes advantage of - may lower authority or - The approach is targeting are clearly defined and local information on cohesion of local actors often paired with cohesive individual - may continue or PMT. - for programs that propose circumstances exacerbate patterns of - communities tend to include a small portion of - allows for local social exclusion to be less well- the population definition of need and - if local definitions of defined - for temporary or low welfare welfare are used, - local officials and benefit programs that cannot - transfers costs of evaluation is more community leaders support an administrative identifying difficult and ambiguous may exert political structure of their own beneficiaries from - Combination with PMT pressure over intervention to is sometimes difficult as targeting 53   community (this can PMT targets poverty - increased mobility also be seen as a based on consumption leads to weak social limitation) (monetary poor) and cohesion and less communities tend to performing CBT consider other dimensions of poverty, resulting in inclusion and exclusion errors Demographic - where registration of vital - administratively - inaccurate where - Urban areas likely targeting statistics or other simple demographic have better records demographic characteristics - low stigma characteristics are weak of birth than rural is extensive - often politically correlates of poverty but the approach - where a low-cost targeting popular may not easily method is required accommodate temporary migrants - might be difficult to outreach to and enroll the elderly - programs targeting the urban youth are politically very popular as these minimize the risk of social unrest Geographical - where considerable - administratively - depends critically on the - May be targeting variations exist in living simple accuracy of information particularly suited standards across regions - no labor disincentive - Level of disaggregation for high- - where administrative - unlikely to create depends on availability of concentration of capacity is sufficiently stigma effects data. Usually these data poverty, like slum limited to preclude use of - easy to combine with are only disaggregated at neighborhoods individual/household other methods provincial or district level (wither for cash assessment - helps to estimate the and very rarely at a lower assistance or for - where delivery of national, provincial , level (neighborhood, public works). intervention will use a fixed district and community community, etc) - Low income site such as a school, clinic, targets (when data - performs poorly where families living in or ration shop available) poverty is not spatially wealthy - helps to know the concentrated neighborhoods number of households - can be politically (providing services to survey in each controversial and as domestic district if PMT or other help) may be methods are to be missed. applied - poverty maps are easily combined with other indicators that might be appropriate for urban settings (employment, education, housing, security, night lightning, etc.) Self-targeting - countries with very low - administrative costs of - cost on the recipient, - This is often a administrative capacity targeting likely to be which lowers the net basis for public - settings where individuals low value of the benefit works targeting (on are moving rapidly in and - unlikely to induce - stigma may be wages). out of poverty labor disincentives considerable - Demand for participation during 54   - where a wage or - may be difficult to find a economic crises is consumption patterns means of delivering a often higher than separates poor from large benefit capacity to meet it. nonpoor (e.g., consumption - On demand surveys are of inferior goods) often used to reduce costs but can result in higher exclusion errors as the poorest households are those with less access to communication, less capacity to cover the registration costs (transports, admin fees, etc) Source: adapted from Alderman et al. (2014) A recurrent question is whether and how to adapt PMT approaches to urban contexts. Background work for this review has shed some light on recent practices (Evans 2015). While many PMT-based programs start in rural areas, when counties expand programs to urban settings the initial PMT formulas need to be updated21. In the absence of sufficient income data and robust checks for means- testing, many countries utilize PMTs to estimate income or consumption and allocate fiscal resources to poor and vulnerable groups. PMT formulas use observable correlates to consumption or income that ideally are not prone to fraud, such as housing material that cannot be moved. Meta-evaluations have found PMT to be one of the highest performing mechanisms for targeted social protection22. Country case studies show that formula for rural populations may not accurately target urban poor and vice versa. In particular, it is interesting to examine specific countries that utilize PMT both in urban and rural areas and observe if and how those approaches differed. A non-exhaustive list of such examples include Cameroon, Ghana, Kenya, Mexico, Mozambique and Philippines (box 9). Box 9. Devising urban-sensitive PMTs A range of countries are adapting their PMT models to urban contexts. For example, Cameroon is piloting a layered (geographic, community, PMT) targeting approach that includes differing formulas for rural and urban areas. The rural formula contained fewer variables with unique questions, including religion of head and cart ownership. The formula for urban areas contains considerably more variables, mainly assets, such as a mobile or land line, CD or video player, ventilator or A/C, stove, and car. The reason for this difference is that rural poverty was 55% and only 12% in urban areas, and the characteristics of urban poverty tends to have higher variation. The model opts for quantile and not OLS regression, as the former avoids complications caused by outliers. Given high levels of poverty overall, the model will be refined to further reduce exclusion errors. In Ghana, the Livelihood Empowerment Against Poverty (LEAP) program includes a PMT component, in addition to categorical, geographic and community-based targeting. The formula is not publically available, though simulations show that a modified PMT with geographic differentiation would have lower inclusion errors, though more mixed impact on exclusion, with elderly and widows most adversely impacted, likely to economies of scale effects. In Kenya, the Orphans and Vulnerable Children (OVC) program uses multiple targeting approaches, including PMT. The program originally had                                                              21 While outside of the scope of the paper, PMT tends to perform more effectively when combined with other targeting methods and implementation for the social protection system remain critical for achieving poverty reduction outcomes (Leite 2014) 22 See for example Coady, Grosh, Hoddinott (2002) where among the 10 methods PMT only lags workfare and youth categorical targeting. An update of a similar analysis will help shed new light given the widespread use of PMT over the past decade.  55   high inclusion and exclusion errors with 43% of the poorest not being reached. There is consideration of updating the PMT formula with geographic consideration of drought-prone areas, and such changes could reduce both errors. In Morocco, following a pledge for universal health coverage in 2002 the government expanded its contributory program, and initiated its non-contributory health scheme RAMED as a pilot in 2008 and was expanded nationally in 2012. RAMED utilizes both community and PMT targeting. The program has distinct formulas for rural and urban areas, most notably using a self-reported income measure in the former and asset score in the latter, which includes agricultural land ownership, equipment such as a tractor, and livestock with animals such as a horse or camel receiving higher asset scores than a goat or sheep. Both use measures of socio-economic conditions, sharing some variables such as personal transport, sanitation and phone, though urban scores also include electricity, people per room, and type of water access. While self- reported income is not ideal due to possible gaming, the country’s PMT provides a good example of how PMT scoring criteria vary between rural and urban areas due to geographic robustness of income data due to generally more in-kind versus monetary transactions in rural regions. Also Mozambique provides an interesting example. There a public works program useful approach that combines geographic, community-based, and PMT with urban/rural models. Geographic targeting includes the poorest districts, communities then present a list of potential beneficiaries, and finally a PMT is used on the list to determine eligibility. Such a layered targeting approach tends to result in greater allocation to beneficiary group. While the PMT models share many variables, the rural model includes land size and livestock, for example, and urban electricity and computer. Similar to other countries, the weights often vary substantially, with the dependency ratio having almost double the value in urban areas, likely due to higher living costs, and higher education having a weight of 56 as compared to 35 in rural areas. Asset ownership also has more weight in rural areas with TV and radio more than double in weight than urban areas. Finally, in the Philippines, the National Household Targeting System for Poverty Reduction (NHTS-PR) provides a useful case of a PMT model used by multiple programs and that distinguishes between both urban and rural areas. The PMT was developed in 2007 and is used for enrolment in the CCT program Pantawid Pamilyang Pilipino, as well as determining eligibility for subsidized health care. Both for expediency of roll-out and improved identification, the NHTS-PR began with separate models in rural and urban areas, and has the objectives of minimizing exclusion errors and quickly enrolling as many of the poor population as possible. The PMT utilizes data from the country’s main living standards survey (FIES) and labor force survey (LFS), which use the sample is representative of 17 regions, as well as rural and urban strata. The models predicts income with some geographic differences in explanatory variables, their coefficients, and the cut-off scores. There are 87 different spatial cut-offs, 6 for cities and districts and 81 for the provinces. Rural and urban differences exist for the 81 provinces, though these are not publically available. NHTS-PR could further improve performance, which is less accurate due to more fluid labor market conditions in urban areas that affect the scores, as well as streamlining the implementation of two separate models. Source: Evans (2015), Del Ninno and Mills (2015), Tesliuc et al (2014), GoH (2014) The experience of Mexico, for example, provides an illustration of the urban and rural PMT formula differences, including the variables used their coefficients, and weighting. As part of the Opportunidades expansion in urban areas, the program updated it PMT in 2002 to include rural/urban and regional distinctions, which resulted in 19 regional models. The PMT was again updated in a pilot in 2009 due to undercoverage of urban poor, and refined again the following year. The new PMT decreased exclusion errors in urban areas by nearly 15 percentage points. The differing spatial PMT formulas and cut-off lines are due to more monetary income in the form of work income and transfers, such as pensions in urban areas, and better labor market conditions in terms of earnings and opportunities than rural areas. For instance, remittances have twice the weight in the rural formula and not having a refrigerator three times. In urban areas, by contrast, renting a living space has twice the weight as rural areas, likely reflecting the higher share of expenditure for housing in urban areas. There are variations in housing conditions, with the floor condition having nearly twice the weight in rural areas. There are a number of additional variables in urban areas, such as not having a computer, which are strong indicators of welfare, though less prevalent in rural areas and weaker predictors of welfare. 56   Similarly, Honduras introduced a model to estimate individual income in 2011 (which was updated in 2013) and that included differing urban/rural formulas for its PMT (see table 9). For example urban has fewer variables for trash disposal, and lacks a variable for number of dependents, whereas rural has no variables for floor or ceiling material, a different water source variable (river or creek vs pipe in urban areas), no public lighting variable, fewer housing tenure and asset questions (the latter category lacks cable TV and air conditioning) and fewer education questions. The differences may be explained by being statistically significance and economic theory, such as differing demand and supply of goods and services, and education being a weaker determinant of income in rural areas, where wages are generally lower and fewer high income jobs exist. Second the coefficients vary in magnitude, with electricity, car and motorcycle having more predictive power in rural areas, likely as these are more common in cities. In other words, country experiences show that, when PMT is used for targeting selection or validation purposes, national public polices using PMT require formulas that differ in the variables and weighting to predict the poor and non-poor. Such a crucial exercise greatly increases the targeting efficacy of programs. Table 9. Urban and rural PMT models in Honduras Variable Urban Coefficient Rural Coefficient Main floor material Cement brick ‐0.1163166 Concrete slab ‐0.118942 Other ‐0.9322116 Main wall material Mud, Stick/pole, or ‐0.2093934 Adobe ‐ Reed 0.0885795 Main ceiling material Straw or palm 0.4022802 Energy for cooking Firewood ‐0.0660531 Propane 0.2165538 Electricity 0.0688944 Electricity 0.3179408 Water supply Public service pipeline 0.325467 Private service pipeline 0.3186978 River, creek, spring ‐0.200248 Lighting Public service 0.3805126 Candle 0.5395925 Candle or gas lamp 0.5032669 Candle or gas lamp ‐ 0.1501492 Torch ‐ 0.2857441 Use for fertilizer ‐0.3890326 Trash disposal Public residential 0.1318495 collection Anywhere ‐ 0.1126941 Other 0.4221727 Leased 0.1304556 Received for services 0.3720387 worked Housing tenure Owner and is paying ‐0.1540316 Owner and fully paid ‐0.1859029 Legal ownership ‐0.1652357 Given unpaid -0.2522964 Asset ownership Stove 0.1340853 Stove 0.15752 Refrigerator 0.1231428 Refrigerator 0.1458223 Equipment 0.1346149 Equipment 0.1216635 Car 0.2145574 Car 0.1690345 57   Motorcycle 0.086188 Motorcycle 0.184862 Cable tv 0.1268079 Air conditioning 0.2331725 Characteristics of head and Age of head squared ‐0.0000273 other members Log number people per ‐0.0803866 Log number people per ‐ bedroom bedroom 0.0753729 Log total people in -0.3120473 Log total people in ‐ household household 0.2044808 Number of dependents - 0.0583459 Ratio men and women 0.0617549 Ratio men and women 0.0563343 age 15+ age 15+ Number women 0.0455317 Number women 0.0688455 reproductive age reproductive age Average age of 0.0058992 Average age of 0.0060577 household household Average education of 0.0471609 Average education of 0.0660389 household household Years schooling of 0.0153325 Years schooling of 0.0125137 household head household head Years schooling of 0.0059764 spouse Number children school ‐ age 0.0666262 Average permanent 1.033205 contract in province Number children ‐0.0615585 school age attend Max years schooling 0.0238685 other members HH member living 0.0676005 HH member living 0.1660616 outside country outside country Agricultural province ‐0.6771227 Constant 6.486824 Constant 6.490012 Source: Evans (2015) Another frequently-raised issue concerns reaching households with characteristics that are seldom captured in PMT models – e.g., not what kind of kind of materials their house is made of, but whether they have a house at all, i.e. the homeless or street children. In this regard, the Philippines’ conditional cash transfer program (the Pantawid Pamilyang Pilipino Program) was modified to address the specific profile of homeless families (Mendoza et al. 2013). While section 4 reviews such experience more in detail, we here mention that the program began by using a PMT to identify prospective beneficiaries. However, the PMT was predicated on the assumption that beneficiaries had an address and lived in a physical dwelling. As such, it omitted homeless households, nearly all of whom had children less than 18 years of age. A ‘modified’ CCT (MCCT) was introduced to respond to those needs and using the national CCT as a platform to reach those ‘invisible’ profiles. Protocols for targeting processes were modified, and so were the benefit structure, including assistance to education, health, housing and security, and psychosocial support23.                                                              23 Relatedly, street children represent another target group that requires a unique set of comprehensive services, potentially including temporary shelter, counseling, and reunification with family or fostering, such as observed in Kampala with street children from the Karamoja region (Alderman et al. 2014). 58   These experiences contrast with efforts to establish a social safety net program in the slums of Nairobi. There, the absence of recent urban poverty data, together with significant movement across and within slums, meant that efforts to use geographic targeting had to be abandoned. Further, resources for communicating information were limited – this made it difficult to ensure that potential beneficiaries were aware of the program, a problem exacerbated by high mobility and weak social connections. This led to a high rate of exclusion errors. Eventually in one slum, targeting had to be redone using a census that was both costly and also delayed implementation. Somehow similarly to Kenya, the experience of urban Burkina Faso shows the opportunities and limitations of combining local urban knowledge enshrined by NGOs with rigorous PMT analysis (box 10). Box 10. High food prices and targeting in urban Ouagadougou and Bobo-Dioulasso, Burkina Faso The 2008 food price shocks that hit most developing countries revealed the high vulnerability to food insecurity of cities. In Burkina Faso, for example, between December 2007 and February 2008 the price of a food basket increased by 23 percent. By June 2008, food expenditures in Ouagadougou represented 75 percent of total household spending, compared to 50 percent in the previous year. Such increases were accompanied by violent demonstrations in the country’s main cities. As a response, the Ministries of Social Affairs and Health launched a food voucher program in Ouagadougou and Bobo-Dioulasso. The targeting process included three main steps. The first included the pre-selection of poor areas, drawing mostly on quantitative information of the Red Cross based on its long-standing experience in the two cities. Once areas were identified, a significant data collection exercise was carried out covering 142,000 households from preselected poor areas in the two cities. Households were approached according to the quality of the dwelling and whether or not they had durable items. Households’ interviews were based on a two-step questionnaire: the first aiming to assess whether or not the household is potentially vulnerable, and the second to be filled only if the first step was passed. The number of households surveyed represented almost half of the whole population in the two cities. To calculate the PMT score, a set of 31 variables were computed. The score was based on characteristics and equipment of the house, household assets ownership, social and demographic characteristics of the head of household (age, gender, marital status and education), number of children and adults, sources of food, and access to basic social services. Based on results from the PMT, 31,500 most vulnerable households – 200,000 individuals – were identified as ‘very poor’ and selected for the voucher programme24. The group largely included female-headed households that rent a 8-10 sq meter housing without latrine, make a living out of unskilled irregular daily labour, have very poor access to health, and children are out of school. Of these households, 21,300 were in Ouagadougou and 10,225 in Bobo-Dioulasso. Once the targeting was completed, a beneficiaries zoning classification was conducted. Households were clustered according to neighbourhoods, which helped to determine the number of shops to be selected for the programme in each location/district, as well as inform the design of the distribution strategy, including staffing requirements, logistics, monitoring and budget. The capacity of a total of 250 shops was assessed, out of which 100 shops selected in Ouagadougou and 50 in Bobo-Dioulasso. The experience and knowledge of the NGOs of the poorest areas in Ouagadougou and Bobo Dioulasso was an important ingredient in geographic targeting. However, such model might be challenging to replicate for a substantial scale-up of the programme. The very similar levels of food insecurity among the surveyed households limited the performance of the PMT. Moreover, the use of the PMT to identify food insecure households was somewhat limited by the fact that it is based on structural poverty variables, like assets, housing and income. Although food security and poverty are highly correlated, a selection based only on structural poverty variables could miss the short-term spells and cyclical aspects of food insecurity. Also, when the programme started, 10 percent of the pre-selected households could not be found. The main reasons for that were the high mobility of the poor urban population and the time span between the survey and the actual registration. Source: Creti (2014c) Finally, it is important to also underscore the importance of the political economy of urban targeting. This stems from the longstanding debates around the role of the middle class and median voters in redistribution preferences (Sumner 2012; Birdsall 2010; Pritchett 2005). More recently, the issue of ‘urban-biased’ targeting was fostered by the combination of urban youth unemployment and riots –                                                              24 An additional 70,000 households identified as ‘poor’ are received Plumpy Doz or locally fortified flour if they have children aged 6–24 months. 59   factors that we discussed in section 2. Under those circumstances, there is some evidence in support of those urban targeting dynamics, as for example provided by Mozambique’s experience in 2008 and 201025. As chronicled by Zapatero and Villanueva (2014), the combination of high cost of living in urban areas and cuts in social spending led to a waves of riots that, over those two years, led to 20 deaths and 700 injured people. Part of the government response included both broad-based investments in social protection programs, as well as specific urban programs such as subsidies for urban transport and the introduction of ‘Cesta Basica’, an urban food voucher program26. 3.2 Outreach and enrollment After targeted populations have been identified, a number of conditions must be met to ensure program ‘uptake’ by perspective beneficiaries. A key emerging lesson is the need for extensive communications and outreach tailored to each urban community. Television and radio spots are useful – but not if the poor have no access to these. For instance, Brazil included a range of informal mechanisms, such as local associations, loudspeakers and churches (Villarosa 2014). Also, registration sites must be accessible; if these cannot be reached easily because of distance, poor public transport, safety or other reasons – again prospective beneficiaries will not be able to apply. Furthermore, where illiteracy rate are high, or some physical disabilities might be prevalent, individuals who are unable to complete forms may simply not apply for benefits they are entitled to. For example, South Africa, which uses passive selection for its Old Age and Child Support Grants, has addressed this concern through the use of local community committees who assist prospective beneficiaries – such as elderly widows – with the application process. We’ll further discuss how the U.S. is addressing those issues in urban safety net programs. In rural areas, social protection programs often do censuses from which beneficiaries are selected. Such an approach is appropriate when the target population is concentrated within a given locality, when households are relatively static and when household members are likely to be at home when program staff visit. But in urban areas, these conditions are less likely to be met, which makes the census approach both more costly and less accurate. For example, in rural settings the Mexico Oportunidades program conducted a census of all households in targeted localities and a proxy means                                                              25 The country’s institutionalized unconditional cash transfer program, i.e., the Programa Subsidio de Alimentos (PSA) reaching about 287,400 beneficiaries at the time of the riots in 2008, was itself restricted to urban areas during the phase of economic reforms in the 1990s (Vera Soares et al. 2010). 26 As chronicled in Zapatero and Villanueva (2014), at the beginning of 2008, the combination of the food, fuel and financial crises and the partial dismantling of fuel subsidies, lead to an important raise in the cost of living for the Mozambican population, especially in urban areas. On February 5, 2008, riots broke out in Mozambique in the capital city of Maputo over rising minibuses fares and bread prices. These riots started a series of violent clashes in the country, resulting in six death and more than 100 injured. The Government response was to cut the price of diesel for minibuses and to freeze those over the next two years. After two years, despite the Government’s release of the National Strategy for Basic Social Protection 2010-2014 in May 2010, coverage of social protection in rural and urban settings was still limited. This, together with announcements of price increases for bread by 25 percent, and water and electricity tariffs by 11 and 13.5 percent respectively, sparked a new wave of urban riots in September 2010. These led to 13 deaths and more than 600 injured, with cities being looted and paralyzed for 3 to 5 days. The government eventually activated a number of urban measures, such as a subsidy to urban public transport users and an urban poverty-targeted voucher program for basic food commodities (Cesta Basica). From 2011 to 2013, the number of beneficiaries supported by social protection programs increased by 29 percent, with an increase by 20 percent in urban settings. Among these, the government (National Institute of Social Action, or INAS) started the implementation of labor-intensive public works in urban settings for the first time in 2012, benefiting over 1,500 households in the municipalities of Maputo and Manjacaze. In addition to increasing the number of beneficiaries, INAS managed to improve the generosity of the PSSB by raising the minimum benefit from 130 Metical in 2012 to 280 Metical in 2014, which represents about 34 percent of the poverty line for an urban poor household in Maputo. 60   test score was calculated for all households. When the program expanded onto urban areas, the percentage of eligible households was expected to be substantially smaller than in rural areas, hence the application of the rural approach to urban areas was deemed too costly. A preliminary self- selection stage was introduced to save on administrative resources that would be otherwise allocated to screening out higher income households as well as the resource costs (time and money) associated with the census. Within selected urban communities, an extensive information campaign preceded the registration process. It used a variety of dissemination methods including TV and radio advertisements; the distribution of flyers; placing posters in churches, schools, health clinics, and marketplaces; and through vehicles that make loudspeaker announcements. This was followed by the establishment, for a short while, of a program office in each locality where individual households could register for the program. Households that were deemed eligible were subsequently visited at home to verify this information, after which households themselves were asked to visit the program’s office for checking their application results. Despite these efforts, only 51 percent of eligible urban households enrolled in the program in the initial phase (see section 4 for a case study). Consequently, program officials rethought how to reach potential participants, such as the full-day poor working mothers and those living in highly-populated areas. Given beneficiaries’ higher mobility and opportunity costs, processes were put in place to reduce the time needed for program enrollment (e.g., using short pre-screening interview). A social intermediation service (Modelo de Atención Personalizada de Oportunidades) was introduced to directly reach out to potential beneficiaries, providing personalized service to navigate the social protection system, and establishing a relationship of trust and support27. Colombia’s Familias en Acción conditional cash transfer program provides an interesting contrast to Mexico’s approach. The government of Colombia maintains an indicator of economic well-being called SISBEN, which ranks households into six categories. Aggregating this information to the locality level meant that the Colombia government could identify urban neighborhoods with high levels of poverty. A census-style survey was administered in localities that fall into the two lowest socioeconomic strata to identify prospective beneficiaries. This was supplemented by extending the survey to households who request it. Despite this, in Bogota, enrollment levels were lower than expected. About two-thirds of surveyed households didn’t apply. Reasons for such outcome included that prospective beneficiaries were unfamiliar with program benefits; they did not know that they could register; and they did not have time to register because of holding multiple jobs. Among those that attempted to enroll, about half didn’t manage to do so because of insufficient knowledge of eligibility criteria (figure 28). To address low take-up rates, part of the strategy included a month- long, large-scale registration process was launched with new locations established where individuals could learn more about the program and register for it (see section 4).                                                                27 See also section 3.4 on social intermediation. 61   Figure 28. Results of enrollment process in urban Bogota, 2011 Yes: 53.3% Yes: Managed Why? (main) 33.7% to enroll? Not in SISBEN I: 30.6% No: Didn’t meet other criteria: 15.6% Went 46.6% Process overlength: 7.3% enrolling? No: 66.2% Why? (main) Not aware of program benefits: 35.8% Thought wasn’t entitled: 29.2% Had no time: 6.7% Source: MacMaster et al. (2013a, 2013b) Urban enrollment may be especially challenging for particular categories of poor dwellers. In this regard, he Supplemental Nutrition Assistance Program (SNAP) – the U.S. premier safety net program providing unconditional food vouchers – lends some interesting experience and practices. About 91 percent of SNAP’s 46.6 million beneficiaries live in urban and periurban areas, hence making the program a key backbone in those contexts (CSM 2014; Oliveira 2013). On average, about 72-75 percent of eligible households actually enroll in the program: in particular, evidence shows that two categories among eligible low-income beneficiaries – the elderly and working poor – are especially prone to low enrolment. Among eligible elderly, enrolment rates are up to 20 percentage points lower than average participation, while the working poor exhibit enrolment rates 10 percentage points lower than average (CSM 2014; Lefin 2011). Key reasons for limited participation among those ‘underserved’ profiles include lack of information about eligibility and/or application processes, perceived or real burdens of applying, low benefit amount, and stigma (Mabli et al. 2011; Burstein et al. 2009). In order to address the specific constrains faced by the elderly and working poor, a series of pilot models were recently tested to facilitating their enrollment in urban areas in six states. The pilots included strategies to raise awareness on the program, provide application assistance, and simplify the application process (table 10). These were generally successful in enhancing the participation among eligible low-income elderly, but not among the working poor. Box 11 discusses more in detail the findings and lessons from the pilot process. Table 10. Pilot approaches in urban SNAP by state, strategy and beneficiary profile Source: Kauff et al. (2014). Note on states abbreviations: MA=Massachusetts; WA=Washington; WI=Wisconsin; MI=Michigan; OH=Ohio; PA=Pennsylvania. 62   Box 11. Enrolling the urban elderly and working poor: evidence and practices from the U.S. Among potential SNAP beneficiaries, the elderly and working poor tend to exhibit enrolment rates between 10 and 20 percentage points lower than average. Key reasons for limited participation among those ‘underserved’ profiles include the following: - Lack of information about eligibility and/or application processes. Although eligibility is determined through income and asset tests and is not restricted to families, many elderly individuals believe they are ineligible because they have assets or they do not have dependent children living with them. Working poor individuals often believe they are ineligible because of their earnings or because of the value of their vehicles. In the case of legal immigrants, informal communication within communities about applications has in some cases generated concerns of losing work permits. Some people, especially seniors, do not know how to apply, or even how to find out how to apply, for benefits. - Perceived or real burdens of applying. Seniors may find it difficult to get to the SNAP office because of lack of transportation, health issues, or physical limitations. While most states have tried to address this issue by waiving the face-to-face interview requirement at initial certification and allowing telephonic signature, seniors may not be aware of the option to conduct the eligibility interview and ‘sign’ the application over the telephone. Working people may find it difficult to get time off from work to go to a SNAP office. The required documentation of earnings and assets may seem burdensome and an invasion of privacy. Research has documented that seniors, in particular, often perceive interactions with SNAP office personnel as unpleasant, and application requirements may be difficult to understand. - Low benefit amount. In 2012, the minimum SNAP benefit for one- or two-person households was $16; households with three or more members could receive less. Benefits for workers may be low because of their earnings. Benefits for seniors may be low because many live alone but have Social Security or Supplemental Security Income that brings them to, or close to, the poverty level. For some, the costs of applying for SNAP (particularly in terms of the time required to complete the paperwork) may be high relative to its benefits. - Stigma. Embarrassment, feelings of failure, hurt pride, dislike of government assistance, and loss of independence are all reasons cited by elderly and working persons for not participating in SNAP. Research has documented that these groups may feel they should not need SNAP benefits and that others are needier. In order to address the specific constrains faced by the elderly and working poor, a series of pilot models were recently tested to facilitating their enrollment, including in urban areas of Chelsea and Worcester cities (Massachusetts); Clark County (Washington); Lucas County (Ohio); Hillsdale, Jackson and Lenawee Counties (Michigan); Rock County (Wisconsin); and Philadelphia County (Pennsylvania)28. The pilots included strategies to raise awareness on the program, provide application assistance, and simplify the application process. In terms of awareness, States conducted one or more of the following: developing and testing messages that educate about SNAP; identifying and targeting efforts to participants in other assistance programs that make them likely eligible for SNAP (that is, list strategies); marketing SNAP and demonstration program services through print materials and media advertisements; and collaborating with community organizations and employers to share information about SNAP and demonstration services. For application assistance, all states but Washington, the state’s key subcontractor hired staff to provide application assistance directly to elderly and working poor clients; in Michigan and Massachusetts, the subcontractor also collaborated with other community organizations to provide application assistance, and in Washington, the state contracted directly with community organizations to provide assistance. Finally, Michigan and Pennsylvania simplified the application process through waivers and administrative changes. In Pennsylvania, another waiver allowed elderly applicants to self-declare medical expenses rather than provide verification. Administrative changes allowed state staff to use self-declared shelter expenses and data the state had verified within the past six months for other programs instead of requiring income, residency, and citizenship documentation from SNAP applicants. A number of interesting results emerged. In terms of enhancing participation among the elderly, the program produced statistically significant impacts in Pennsylvania and Michigan, but not in Ohio. Both successful states granted waivers for enabling seniors to apply without visiting a SNAP office, including phone-based interviews and home visits. Both pilots worked with state agencies to obtain lists of seniors that indicated likely eligibility for SNAP, including drawing from other government programs (e.g., Medicaid) and community-based organizations. The latter, for instance, proved very                                                              28 Pilots also included rural areas, such as Dane and Green Counties in Wisconsin. These, for instance, were severely affected by economic downturns, including as a result of the loss of manufacturing jobs in neighboring Rock County and the closure of General Motors’ plant in Janesville. 63   cost-effective: in Pennsylvania, referrals costs from community organizations totaled less than $30,000, but accounted for 36 percent of households that enrolled in SNAP through the pilot. Existing technology infrastructure—namely, web-based SNAP application system and proper MIS—was key for the pilot success. However, the generation of the desired volume of calls for telephone-based application assistance entailed activities such as community presentations and information distribution. It was reported that seniors were frustrated with the multi-step process and points of contact. Indeed, they may have had to engage in several calls before meeting a person face to face – that is, first when they called the hotline to inquire about SNAP and schedule application assistance, second for their application assistance appointment, and third when a community partner called to schedule an appointment to review documents. This process typically took three weeks or more. Overall, in Pennsylvania and Michigan about 85 percent and 82 percent of the applications submitted as part of the pilot were awarded, respectively. Effects were strongest for the oldest senior households (which likely have more mobility challenges than younger ones), suggesting that eliminating the need to visit a SNAP office may be a key driver of the effects. In Ohio, the pilot didn’t yield to statistically significant increase in enrollment. Part of the reasons included the application submission by paper, which made it challenging to track those pertaining to the pilot29. There were also specific issues with the community organization executing the demonstration, such as the limited follow up with applicants assisted and therefore no data existed on the approval rate among them, or reasons for denial. However, the Ohio pilot underscored some of the seniors’ unique challenges (beyond mobility and transportation) that individualized application assistance can address. First, meeting seniors in the community at a location of the senior’s choice permitted pilot staff to reach seniors where they felt most comfortable. Second, to address the frequent issues among seniors of impaired vision or hearing, pilot staff used large print materials and offered to sit with clients as they completed telephone interviews or to meet them at local government premises. Third, staff highlighted the importance of working patiently and slowly with those who may need longer to process complex eligibility information, or who may want to have longer conversations because they desire companionship. Fourth, the pilot developed a four-step process to reduce the number of no-shows by scheduling appointments close to seniors’ homes, confirming the date and time of the appointment, placing a reminder call the day prior to the appointment, and talking to a senior who missed an appointment. In terms of increasing the participation among the working poor, the pilots in Massachusetts, Washington and Wisconsin didn’t spur any statistically significant increase. Various reasons may have contributed to such outcome: for example, it is possible that pilot program services were not different enough from other pre-existing outreach activities. In some cases, it was challenging to identify new locations for engagement and application assistance that offered both confidentiality and a sufficient client flow (health centers were prime application assistance sites because they typically met both criteria). Also, the personality and approach of staff conducting engagement and application assistance is critical, including both interpersonal and marketing skills. Yet also in those cases a number of valuable lessons emerged. For example, in pilots envisioning collaboration with employers to raise awareness, staff eventually learned that, to gain employer cooperation, the message must be framed in a way that allows them to see the benefits participation could bring to the company and does not reflect negatively on the company’s wage rates30. In other cases, potential applicants did not understand that assistance for application from social workers was just the first step of a process. This caused some confusion when the SNAP office contacted them to confirm information or ask for additional documentation. To address such confusion, pilot staff developed a one-page information sheet describing the organization, the role of the person who had assisted them with their application, and instructions about what to do next and how to access information about their case. This proved to be a helpful tool when explaining to perspective participants what to expect after their application was submitted. Source: Kauff et al. (2014), CSM 2014; Lefin (2011)                                                              29 Staff had laptop computers equipped with wireless internet cards, which they planned to use to help clients submit applications online. However, staff reported that seniors preferred to see their paper application completed so they could monitor what was being submitted, and often distrusted electronic submission of their personal information. While it is possible that submitting paper (rather than electronic) applications may require more work for government staff (e.g., for data-entry), there was no perceived difference in processing time for general versus pilot-related SNAP applications. 30 For instance, explaining that employees are likely to be more productive if they receive proper nutrition seemed to be an effective strategy. In addition, explaining that the employer’s part-time workers may not know they qualify for such an important benefit takes the focus off the specific company and how much it may pay their workers. 64   3.3 Mobility and portability Mobile populations As noted in the previous sections, the most effective targeting strategies for identifying, screening, and enrolling potential beneficiaries for social protection programs depend on the objectives of the program and the related characteristics of targeted beneficiaries. A distinguishing feature of the urban poor is that they are often more mobile than poor people living in rural areas. This section addresses issues involved in finding and tracking mobile populations, improving portability of benefits, and linking urban safety nets to complementary programs. We identify three types of mobile urban populations, including (national) urban migrants31, the working poor populations spending substantial time travelling to and from work, and the homeless. Migrants frequently move to urban areas in search of employment, but other motivations include fleeing conflict and insecurity, reuniting with family, or seeking health and other services. People migrating for employment sometimes include temporary or seasonal migrants and may include individuals or entire families. These individuals are vulnerable to falling into poverty if they do not find employment or the employment is terminated. These various motivations for moving to an urban area may provide an indication that the individual or family is fleeing chronic poverty or may be newly poor in response to an income or security shock. Thus, it can be useful to include migrant status in the targeting criteria for urban programs, and active targeting of neighborhoods with large migrant populations can be an effective strategy to identify newly eligible beneficiaries. This raises a concern that provision of social protection to urban migrants may induce greater migration. This is found to be the case in a minor set of cases, including about 10-15 percent of migrants (see box 12). Such concerns can be ameliorated by having similar targeting criteria and benefits for programs in rural and urban areas. This suggests migrant status may be an effective criterion in the first stage of targeting, but should not be used as a criterion for program eligibility. Box 12. Benefit-induced migration? A summary of the evidence The migration from prevalently rural, lagging regions to leading (mostly urban) areas has been an empirical regularity in the economic transformation and development process – that is, such dynamic lies at heart of the Lewis and Harris- Todaro models. But to what extent do public service differentials between rural and urban areas, instead of economic opportunity, induce migration? The questions is framed in terms of public services since the evidence base tend to revolve around that aspect of public policy alongside broader welfare benefits. While the issue seems not to have been examined with a specific safety net lens, services and welfare provide a convincing proxy for understanding whether and to what extent public benefits, including social assistance, may play a ‘magnet’ function. Internationally, the evidence of welfare benefit-induced spatial mobility is mixed. For example, in the EU studies show that within-union immigrants are either as likely or less likely to be receiving support, and no strong link is being detected between welfare generosity and immigration. Furthermore, the issue needs to be interpreted over larger time horizons, including weighing the costs and benefits from migration over time32. Also in developing countries, the disparity in public provision seems not to be the key driver for the large majority of migrants, although it appears to be so for 10-15 percent of them. Based on cross-country household survey data, evidence shows that in Bolivia, 13.3 percent of migrants reported to have moved to access better schools. In Romania, it was 10 percent. In Paraguay and Guatemala, over 15 percent of migrants moved due to poor living conditions. In Bulgaria, 15                                                              31 We do not include international migrants here, since they raise a different set of legislative and policy issues. 32  For example, in 1967 one of every five U.S. urban residents over age 14 had migrated from a rural area. Although such migration imposed short-term fiscal costs, in the long run, migrants paid more into the system than they had received from it (World Bank 2014).  65   percent of migrants sought better schools and 13 percent wanted better living conditions. In Brazil, analysis of census data shows the importance of public service differentials in influencing long-run migration decisions. In particular, poor migrants are in fact willing to accept lower wages to get access to better services. A full-time minimum wage worker earning Rs$7 per hour was willing to pay Rs$390 per year in compensating wage differentials to have access to better health services, Rs$84 for better access to sewage services, and Rs$42 for better access to electricity. In such contexts, it is likely that measures focused on improving rural public services would result in more of the people who choose to migrate doing so for reasons of economic opportunity (i.e., wage differentials). This could have beneficial spillover effects (i.e., adding to agglomeration economies in leading areas), while simultaneously easing pressure on local governments to accommodate large numbers of migrants. Studies on Mexico, for instance, seem to corroborate this point, although in terms of international migration. In that case, increasing household income through publicly provided safety nets, including the Oportunidades conditional cash transfer program, reduced migration. Indeed, households participating in Oportunidades seem less likely to migrate compared to nonparticipating households with similar characteristics. The transfer, in other words, is perceived as an opportunity cost for households, preventing migration for such individuals. Yet such deterrent-effect seem to come into play after a certain income threshold (8,287 Mexican pesos, of which 3,267 is provided by Oportunidades. Source: Ishikawa (2014), Dustmann et al. (2010), Lall et al. (2008), Stecklov et al. (2005) In some cases, separate targeting strategies may be needed to add newcomers to existing beneficiary lists. Initial targeting is often based on lists of beneficiaries for existing or recent programs. For example, initial beneficiary lists for Brazil’s Bolsa Família program were created using the Cadastro Unico, a registry that identified beneficiaries of the predecessor programs of Bolsa Escola and Bolsa Alimentacao (see box 13). Adding new beneficiaries, including recent migrants, can be costly and difficult to implement. Although Cadastro regulations in Brazil listed home visits as the preferred method of new registrations, in practice the program used a combination of home visits, public service delivery points (like schools), and office-based registrations with at-home spot checks on a subsample of registrants. A beneficiary survey conducted in 2005 indicated that 15 percent of beneficiaries were registered at home, 46 percent were registered at a school, 22 percent were registered in a municipal office, and the remainder were registered at health posts or community events. Relying excessively on community structures may carry risks of omitting selected groups. For instance, newly eligible households often have less political clout than established poor populations. For example, over time, slum dwellers in urban Kenya were able to attract the attention of politicians to try to improve their access to safety nets. Such gains may come at the expense of other poor constituencies that are harder to identify. Box 13. Brazil’s Bolsa Familia: an urban perspective Bolsa Família, Brazil’s national Conditional Cash Transfer (CCT) program, was created in 2003. The programme has three main components: income transfers, conditionalities, and complementary programmes. As of 2015, it provided transfers to 70 million individuals (direct and indirect beneficiaries) or 35 percent of the population. Extremely poor families (defined as earning R$77 (USD 31) per month or less) receive a base benefit of R$77 regardless of family composition. Both extremely poor and moderately poor families (defined as earning R$154 (USD 62) per month or less) receive a variable benefit according to the number of children in the family (capped at three), as well as whether the mother is pregnant or breast-feeding. Families receive R$25 for each child aged 14 or less, and R$42 for each teenager aged 15 or 16. The transfers are conditional on both health and education conditionalities. Household data is collected at the local level through a standardized survey, and is entered into a registry of low-income families called the Cadastro Único. Of the total population registered in CadÚnico, 48.1% are located in urban areas. In São Paulo and Rio de Janeiro, less than 10% of households are enrolled in Bolsa Familia despite the high incidence of poverty. The amount of the subsidy may be too small of an incentive for urban families to keep children in school. Other state add-ons, such as as Bolsa Carioca (in Rio de Janeiro), provide additional subsidies to account for cost of living in urban areas. However, most recent data shows there were only five in existence, and they were on a small scale. 66   Registration is meant to be conducted at home, but central registration points are also permitted in difficult to reach areas, and each municipality must have a permanent site for ongoing registration and updating of information. Outreach is conducted through the municipality, family social assistance centres, schools, health centres, churches and NGOs, as well as through the media. In difficult to reach areas, further outreach such as intensive media campaigns and bimonthly mobile team active recruitment is conducted. These procedures help minimize poor families being missed. Payments are made to a woman in each family. The complementary programs to Bolsa Familia in urban areas involve support to small and micro enterprise owners, and include training (particularly vocational training), equipment, micro lending, and formalization. These programs use a mix of active recruitment as well as self-identification. There is a strong focus on sensitization using local residents’ associations, community based organizations and churches, loudspeakers, and social networks (both informal and formal, such as Facebook), and also the police (particularly in pacified favelas). Much use is also made of local institutions and stakeholders in order to find and register participants, and the physical presence of the program (be it an office or field staff) is deemed important for outreach. In peripheral urban areas in the state of Bahia, a geographic mapping of areas to cover is undertaken first. Subsequently, there is a registration drive that includes both active search and individual requests. A survey is conducted collecting data on income, housing conditions, etc., just as for Bolsa Familia. Finally, a study of economic viability of the micro enterprise is conducted. Source: Villarosa (2014), Chioda et al. (2012), CGAP (2010), Soares (2010), Lindert et al. (2007) A related approach to targeting newcomers, similar to that used by Bolsa Família, is to screen potential beneficiaries at other service delivery points, like schools and health clinics. These service delivery points may then direct applicants to services for the poor or may serve to provide some of the services directly, such as targeted school meals. One concern is that pushing these additional responsibilities on schools and health clinics creates an added burden on service systems that are often already overstretched. A distinct issue concerns how to design urban safety nets to track beneficiaries of safety net programs as they move from rural to urban areas or across urban centers. When rural and urban programs have similar eligibility criteria, such tracking is needed so that chronically poor households do not fear moving from rural areas in search of better job prospects in urban areas out of concern for losing their benefits. Many national programs have offices in major urban centers for this reason. Migrants can register in their new urban location so that their benefits move when they move. Still, some poor urban migrants will not be aware of the presence of program offices or may not find them easy to locate. This provides an additional impetus to include active targeting in urban neighborhoods hosting large numbers of migrants, including slums, for example. Recent evidence has shown that transfer programs can be more effective at improving outcomes for poor households if there is a better understanding of the motivation for migration, and if programs are allowed to support some migration as a way to improve outcomes for the poor. Bryan, Chowdhury, and Mobarak (2014) report the results of an experiment providing an $11.5 incentive (covering transport costs to urban areas, plus a bonus when contacting program designers when in those urban centers) to households in Bangladesh to out-migrate during the lean season in search of temporary employment. Their experiment was conducted in Rangpur region of Northwestern Bangladesh in which the lean season is a period of substantial food insecurity. Roughly one-third of households in their study area send a family member to a nearby urban center to seek temporary employment. Successful migrants are able to remit meaningful sums to their families to help smooth consumption during the lean season, but the risk of failing to find employment places their families at even greater risk. The authors show that the offer of a one-time migration incentive increases the share of households sending a migrant by 22 percent and leads to a 30-percent increase in consumption at the origin (550-700 cals/person/day) for family members of induced migrants. These results show that sophisticated transfer programs that provide an inexpensive form of insurance for temporary rural-to- 67   urban migration can lead to substantial improvements in welfare for the poor. Such a program may have substantial added benefits by reducing the demand from these families to migrate permanently to urban areas, reducing pressure on urban services and sustaining rural economies and communities. Mothers working low-wage jobs in major cities sometimes must travel long distances to work each day, making them unavailable for at-home registration for programs and unavailable to visit distant registration centers. Identifying such workers for program registration may require visits to large employers of such workers or at-night or mobile registration initiatives. Although slum dwellers may be considered to have a fixed location, they are often substantially more mobile than rural populations because of the extreme vulnerability of their income, dwelling, and security situation. Establishing service centers or conducting additional registration in slums may help, but individuals also need to be informed that their benefits can follow them if they move, and should be told how to update their registrations. Another approach would be to use hotlines and call centers to allow preliminary screening of applicants, who could then be directed to a registration center or visiting by roaming registration teams. At-risk youth present an even more difficult challenge because they are not typically in school and may irregularly reside with their families or any stable guardians. Such populations require specially-designed programs needing comprehensive services that may include counseling, drug treatment, job training, and mentorship in addition to social protection. In the presence of highly urban mobile populations, policymakers designing safety net programs must consider a variety of factors that affect the portability of benefits—that is, the ability of a program to continue to provide benefits to beneficiaries as they move spatially. In this regard, it is important to distinguish between, on one hand, the portability of participation in a given program and, on the other hand, the portability of accessing transfers under a given program. In the former case, in the absence of a central registration system, programs need a strategy to facilitate beneficiary tracking. Programs can place the responsibility to inform the program on the beneficiary, but such initiatives should include information campaigns so that beneficiaries are informed and can undertake the steps to remain in the program. For example, in the Philippines, a process is in place for beneficiaries to declare any change in residency and keep participating in the Pantawid conditional cash transfer program. This process requires that such notification is made six months in advance of a move, as for example shown in urban areas of San Jose, Batangas. The increasing use of technology makes the portability of accessing transfers somewhat less of a challenge, especially when they include cash transfers as opposed to vouchers and in-kind transfers. For example, under Bolsa Família, cash transfers are provided through individual beneficiary cards that can be brought to outlets at many urban centers to redeem benefits. In Ecuador, a similar card system was used for the urban voucher program, but with mobile vendors that visit beneficiary communities monthly with equipment that allows beneficiaries to verify the amount of money available on their card and receive benefits. This can be an effective and secure system, but it may be costly to implement in programs at scale. Currently, the low cost and wide reach of mobile money programs offered by mobile phone providers makes this a potentially attractive method of delivering transfers (Aker 2012; Vincent and Cull 2011). In some settings, like voucher-based programs in the Gaza strip, portability of benefits is somewhat restricted in non-crisis times, and delivered in times of emergencies (Creti, 2014b). The limited portability of benefits is the result of compromising between local implementation capacity and operational complexity. Indeed, technology can represent both an opportunity as well as a barrier to portability, depending on whether tech-based transfers are made portable and the interoperability of 68   systems deployed. In India, for example, the possibility to make benefits portable helped to significantly reduce inefficiencies in the Public Food Distribution (PDS) program in urban areas of Chhattisgarh state. Yet even in high-income countries like the United States, the process toward ensuring full portability of flagship safety net programs encountered major bottlenecks as different states adopted differing operating systems. Boxes 14 and 15 discusses the India and U.S. experience, respectively. Box 14. Piloting portability of safety nets in urban Chhattisgarh, India The Public Food Distribution (PDS) program in India has been extensively examined in the literature. Until recently, most of the evidence pointed to a number of operational challenges that significantly affected the PDS efficacy. Among those, a key concern revolved around ‘leakages’ of food through the supply chain before getting to distribution points (i.e., ‘fair price shops’). Even at that point, there are numerous accounts of commodities being diverted, food getting underweighted, beneficiaries being overcharged, shops being closed, or food being falsely declared out of stock. These factors have resulted in considerable transaction costs to beneficiaries (e.g., queuing for food for half a day) and underprovision of entitlements. Until recently, one of the states where such challenges were pervasive was Chhattisgarh. Yet, survey-based empirical evidence documented an impressive improvement in the state’s PDS performance between 2004/5 and 2009/10. For instance, the program was able to cut the share of people that “reported no PDS purchase” from 75 percent to 32 percent, hence expanding coverage among eligible beneficiaries33. Also, the diversion of PDS grains was reduced from 51 percent to 10.4 percent. Compounded with political commitment and other factors, a key ingredient behind Chhattisgarh’s progress was the introduction of an automated system (COREPDS) in each participating fair price shop. This included the equipment of PoS devises with GPRS connectivity, biometric authentication scanner, and smart card slot. The piloting of such system commenced in 2007, including 151 shops and 170,000 beneficiaries in Raipur city. Differently from the previous models, it allowed beneficiaries to choose the shop where to access the benefits. In other words, benefits were made portable. This introduced a strong element of competition among shopkeepers and, as early evidence shows, a number of the above-mentioned challenges with underprovision were eliminated. While comprehensive evaluations are underway, the experience suggests that technology alone can improve but not fully address issues of transparency. The pilot instead shows that the empowerment of participants through choice (as provided by portability) was a key determinant in elevating people ‘from beneficiaries to customers’, hence letting market mechanisms and competition to largely address previous inefficiencies. Source: GoI (2013); Dreze and Sen (2011); Khera (2011); World Bank (2011) Box 15. Portability and interoperability: lessons from the U.S. Formally introduced in 1964 after pilots in the late 1930s, SNAP supports 46.6 million people (or 14 percent of the population) with monthly vouchers of US$133.4. Until 1988, SNAP’s vouchers were paper-based. Under that system, recipients could use their benefits at any authorized voucher retailer in any state. A resident of northern Indiana, for example, could cross the border and use food stamps in Michigan, just as New York residents could shop in New Jersey. Over 1988-1993, legislations allowed and encouraged to pilot test of electronic benefit transfers (EBT) delivery systems. The EBT is an electronic system that allows a recipient to authorize transfer of their government benefits from a Federal account to a retailer account to pay for products received. Once eligibility and level of benefits have been determined, an account is established in the participant's name, and benefits are deposited electronically in the account each month. A plastic card, similar to a bank card, is issued and a personal identification number (PIN) is assigned or chosen by the recipient to give access to the account. In 1996, the Personal Responsibility and Work Opportunity Reconciliation Act mandated that States implement EBT systems by late 2002, unless a State faced unusual barriers to implementation. In this process of transitioning to an EBT system, its ‘interoperability’ – or the ability of EBT systems in different states to communicate with each other – was a key concern. Indeed, States were pursuing stand-alone procurements and there existed no guarantees that EBT systems in any given state would be compatible with others. Indeed, unlike under the paper-voucher regime, with EBT States could require recipients to use their benefits only in the issuing state, thereby precluding from shopping in other states and depriving merchants of interstate business. The result was a series of difficulties created by the proliferation of incompatible, online and offline, systems which deprived recipients of the ability to use their benefits anywhere in the country. Also, this was particularly burdensome for merchants serving market areas transcending state borders, as well as for larger chains that would have had to purchase different EBT equipment                                                              33 The national average for the indicator is 55 percent. 69   in each state in which they operated. There were two sets of responses to enhance interoperability. One early measure was for individual states to achieve interoperability by deploying EBT equipment on both sides of a border. For instance, Ohio allowed merchants on the Indiana side of the border to participate in Ohio’s EBT system in order to ensure that recipients living near the border would retain access to FSP retailers. At the same time, some states voluntarily addressed the problem by collaborating in the development of QUEST. Overseen by the National Automated Clearing House Association, QUEST is a series of evolving rules intended to create a “uniform operating environment for EBT.” The voluntary QUEST protocol enables benefit recipients resident in one QUEST state to access their food stamp and nonfood stamp benefits in all the other QUEST states. The QUEST logo was typically displayed at participating ATM and POS machines, informing consumers that their benefit cards are accepted at a particular machine. This protocol helped to some extent to preserving the portable nature of vouchers and forestalling the proliferation of incompatible EBT systems. Yet the voluntary nature of QUEST made it nonbinding for states to participate or comply. This led to problems in areas that span QUEST and non-QUEST states. For example, benefit recipients residing in the neighboring cities of Gallup, New Mexico and Window Rock, Arizona are unable to shop in the other city since Arizona is a QUEST state while New Mexico is not. While QUEST’s growth reduced the problem of interoperability among the 31 participating states, participation was voluntary and merchants remained concerned about the lack of a single national EBT standard. National merchants still had to invest in different kinds of POS equipment in each market. Realizing that a lack of interoperability could derail the full implementation of a national EBT system for vouchers, Congress passed the Electronic Benefit Transfer Interoperability and Portability Act of 2000. That law provided national standards and required states to develop interoperable EBT systems for SNAP vouchers, thereby ensuring the portability of benefits across state lines. To further assist merchants, the law prevented states from shifting the compliance costs to authorized food stamp retailers and Congress simultaneously agreed to pay 100 percent of the conversion costs, provided that the total amount spent on all states in a given year did not exceed $500,00034. In addition, the Act prevented states from placing limits on the geographic areas in which benefit recipients could use their benefits. The SNAP experience calls for caution in implementing technology for national programs in a decentralized manner. While the 1996 legislation required that states implement EBT systems for vouchers, the act provided states with the flexibility to develop systems as they saw fit. Yet, the SNAP is a national program that transcends state lines and benefits are intended to be portable. By allowing the implementation of EBT in a decentralized fashion, a situation was created whereby states were developing systems that were potentially technologically incompatible. The adoption of QUEST protocols and the passage of the 2000 legislation helped to set standards and make SNAP benefits portable and interoperable. Source: Oliveira (2014), Stegman et al. (2003) Residency A critical quandary in conceiving and implementing urban safety nets is the question of residence. The premise of the debate lies in the realities of how residents inhabit cities in developing countries. A significant proportion of people do so not reside within the ‘formal’ and ‘legal’ system, but through a range of practices ranging from squatting and occupation to violations of building and planning norms within individual structures, or to the conversion of rural land into urban residential developments. For example, in Delhi less than 25 percent of the city’s residents lived in ‘planned’ legal settlements, the rest in unplanned areas (GoD 2009). Against this backdrop, a recurrent question is whether urban social protection regimes can be built without explicitly delinking tenure and support (through law and policy) in a context where a significant portion of residents – and a majority of those in need – are deemed to inhabit the city ‘illegally’. Given the relevance for urban safety nets of the issue at hand, it is important to clarify some further concepts without diverting the discussion into purely legalistic spheres. In general, ‘legal’ settlements                                                              34  Overall, the EBT proved to be effective in a number of areas relative to the paper system, such as easing administration and reduce fraud. However, the initial assumption that technological solutions would lower costs would not always prove correct, and in some occasions EBT turned out to be more expensive than anticipated. When coupled with changes in EBT’s economic and market structure, the increased cost of EBT to the states was seemingly shifted to benefit recipients and merchants.  70   are those built on plots marked in the development area of the government master plan, in concordance with the residential use allocated to that plot (i.e., the ‘planned’ settlements mentioned above, therefore are legal and formal areas). In some cases, people may live in plots recognized in the administrative master plan, but their use is not for residential purposes. Yet home-buyers in these areas typically have formal document as proof of legally defensible transaction. In a way, titles of these homes are ‘formal’, but the use of such land for residency is illegal. In other cases, residents may have either occupied a land or paid someone who has done so before them. In such instances, settlements are both informal and illegal35. There are three basic ways in which spatial illegality can hamper access to safety nets. The first is by explicitly excluding residents living in illegal settlements from the onset; the second is by making it extremely difficult to comply with residential proof – the hurdle of ‘paper work’ is a key bottleneck that we’ll further discuss in section 3, for instance in the case of India. Finally, residence-based exclusions can occur through cycles of evictions and resettlement. In such cases, public authorities may be less likely to invest in programs for communities that are prone to be evicted. Furthermore, as people resettle post-eviction, documents that evidenced proof of address in the previous location (when available) becomes no longer valid36. The latter is also closely connected to the previous discussion on portability of benefits. A number of practices and proposals are emerging that may help unbundle such longstanding and complex issue. For example, countries like India are considering a set of ‘notifications’ for gradually regularizing illegal settlements while providing public assistance. Other approaches envision, for instance, access to programs through evidence of an ‘intention’ to reside in the city that includes residents at an early stage of this residence. Rather than asking residents to prove that they deserve to be included as urban residents by surviving for years in the city, it includes them from the very beginning (Bhan et al. 2014). As countries, however, consider policy options on whether and how to formally recognize areas and people, part of the social assistance function is provided by non-governmental organizations, such as in the Nairobi slums in Kenya (see section 3). Such interventions may often not have the capacity to provide support at scale, but they provide interesting insights on how social protection programs can integrate existing informal risk sharing and coping mechanisms of slum dwellers (Muiruri 2013). While illegal tenure is a central constraint in a number of contexts, the issue of residency – or lack thereof – can affect access to safety nets in a number of other ways. In particular, the discussion above is largely an institutional issue. But residency can also become a policy matter. For example, the systems of hukou in China and ko-khau in Vietnam have various provisions that limit access to safety nets by migrants. Table 11 contrasts basic features in those systems.                                                                35 However, illegal areas can be ‘regularized’, or be given formal and legal titles often decades after they are built. Such regularization or legitimization of settlements can take different forms, e.g., on-site upgrade. 36 An additional barrier is faced by those who rent rather than own housing. The predominance of urban lease cycles of less than 1 year (e.g., 6 or 11 months) imply that renters remain precarious, unacknowledged, and face difficulty in obtaining proof of sufficient length of residence to qualify for the service provision, especially within illegal neighborhoods where their status is further marginalized. 71   Table 11. Contrasting China’s hukou and Vietnam’s ho-khau systems China Vietnam Types of registration Historically, agricultural vs. non- Permanent and Temporary agricultural hukou types, but a single name was adopted in a number of provinces after an unified registration was introduced Parallel residence permit by Yes No cities/provinces Local pilot of reform Yes No Restriction on labor mobility No restriction on geographic No, but the law requires registration. mobility, but certain restrictions exist for sectoral and/or occupational mobility Effect on housing and land right Yes Migrants, with few exceptions, can purchase buy houses and land titles Effect on access to public education Historically, it was very limited, but Difficulty in access to public schools this has been significantly improved in recent years Effect on access to public health Similar to the effect on access to Health insurance cards can only be service public education received in the registered place of residence; those who use health insurance not in their place of residence/registration, face large (70%) co-payments. Access to other health programs (such as immunization) also based on residence. Some provinces/localities may open this temporary residence. Effect on electricity and water No Some reports that the unregistered access pay higher fees Eligibility for social protection Ineligible for rural migrant workers Temporary and unregistered programs and their families individuals not eligible Source: adapted from Demombynes (2014) As we’ll discuss in section 3, the China’s urban Dibao program represent the largest unconditional cash transfer program for urban areas in developing and emerging countries, including reaching over 20 million beneficiaries. Given the implications that the hukou system has on accessing urban Diabo by migrants, it might be interesting to further explore the hukou system and its evolution in some detail. The hukou system links the provision of benefits to the place of origin, not residence. This implies that the large volume of migrants from rural to urban areas (about 167 million people) may not access safety net programs even if ‘in need’. Over time, however, China has introduced a range of gradual policies that enable local experimentation with different levels of relaxation of the hukou system. Clearly, this has broad-based implications for Dibao eligibility. Box 10 sets out a compilation of pilot experiences with hukou reforms. A number of lessons emerge from the experimentation process. For instance, local reforms have been least complete in the large cities where rural migrants are most concentrated, at least for migrants from outside the municipal or provincial jurisdiction. At the same time, urban hukou in small and medium cities entitles the migrant to less generous social services and social protection, contributing to the limited success of the policy aimed at attracting migrants to smaller cities. Box 16 sets out a compilation of pilot experiences with hukou reforms. 72   Box 16. Experimenting with institutional reforms in China The original rationale of Hukou was that migration for work was temporary and that families of migrants would stay behind and access services in rural areas. Although this was true in the early stages of China’s economic transition, the situation has changed dramatically over the past couple decades. Rural-to-urban migration has become more permanent, with the majority of migrants having no aspirations to return to rural areas. In addition, a second generation of migrants, born and raised in cities, has no attachment to the rural areas from which their parents migrated. Starting in 1997 and culminating in a 2001 national policy, measures were gradually introduced to encourage selected rural migrants to apply for urban hukou in small cities and towns – i.e., the ‘small city free’ policy. In 2006, the State Council promulgated a milestone document that provided a comprehensive policy framework for the fair treatment of rural migrant workers in cities with respect to their entitlement to social services and other measures. All fees levied on rural migrants were removed, such as temporary residence fees and management fees, family planning fees, urban expansion fees, and management and service fees. More recently, the State Council formulated a national policy on hukou reform in 2011 and issued reform guidelines linked to the city’s administrative level. In towns and county-level cities, migrants can apply for permanent local hukou for themselves and family members (spouse, unmarried children, and parents) if they have legally stable employment and a residential apartment (including leased)37. In prefecture-level cities, migrants can apply for permanent local hukou for themselves and family members if they had legally stable employment for over three years, lived in a legally stable place of residence, and contributed to social insurance for a certain number of years. In municipalities directly under central management, vice-provincial-level cities, and other large cities, strict quota control policies continue. Local hukou experimentations include four types of models. These are hereafter discussed and provide a key source of learning for China’s institutional reforms towards a residence-based safety net system for Diabo. - A score system for hukou conversion. Guangdong is the largest migrant-receiving province, housing nearly 30 million migrant workers. It was the first province to replace the traditional hukou quota system with a point system for hukou conversion, gradually lowering the conversion criteria for migrants. The points are calculated based on education, vocational certificates and profession, years of social insurance contribution, charitable activities such as blood donation and volunteer work, and government awards. Between 2010 and 2011, about 696,000 migrants were converted through the point system (against a three-year target of 1.8 million conversions). In 2013 Tianjin promulgated its point system (to take effect in 2014), and Beijing announced that it will formulate its residence permit system in 2014. - Strict and fixed conversion criteria with rationing. Shanghai was the first city to introduce the residence permit system open to all, but the qualifying conditions are among the strictest. The Shanghai system prioritizes three categories: those with college degrees or special talents and those who work, do business, or invest in Shanghai (and their families); those who have stable employment and housing; and those reunited with family members with Shanghai hukou. The system features points calculated based on the resident’s age, years of experience, and social insurance contributions in Shanghai, as well as educational and technical qualifications (e.g., residents who make a significant investment in Shanghai earn 100 points; those providing false information lose 150 points.). A total of 120 points is required for a residence permit holder to be entitled to social benefits. Residence permit holders must make seven years of social insurance contributions before applying for hukou. In addition, Shanghai has a tight overall quota on hukou conversions, and the number of conversions has to date been very low. - Localized hukou conversion through exchange of rural and urban entitlements. Chongqing has encouraged family migration with hukou conversion but only for those who are rural residents of Chongqing. Hukou transfer to urban districts requires that migrants work or do business in the area for more than five years, purchase commercial property, or make significant investments or tax payments. This is lowered to three years for migration to a township within the municipal boundary. The key feature is the so-called “exchanging three rural clothes for five urban clothes” policy: the “rural clothes” being homestead land, farmland, and contracted forest land, while the “urban clothes” are pension, medical insurance, housing, employment, and education. Those converting from rural to urban hukou can keep their farm, homestead, and forest for three years but must give it up thereafter if they wish to retain their urban hukou. Chongqing has, however, been easing the exchange requirements in recent years. Using a lottery, authorities in Chongqing give                                                              37 As this review is being prepared, the Ministry of Public Security is formulating a roadmap for hukou reform, aiming for implementation by 2020. The December 2013 Urbanization Work Conference of the central government also called for an “orderly conversion” of rural migrants and proposed a numerical target of 100 million long-term rural migrants to be converted to urban hukou holders. 73   residence permit holders access to subsidized public housing rentals, with the subsidy covering about half of the market rental price. In the three rounds of lotteries to award subsidized public housing rentals, more than 100,000 people were granted subsidized rental units. These subsidized public rentals are also open to long-term residents who do not own residential property, allowing for the creation of mixed neighborhoods. - Hukou conversion of local residents without exchange of rural rights. Chengdu introduced a residence permit system with two types of permits: temporary and permanent. The residence permit and hukou conversion is only open to those who are already residents of rural areas of Chengdu prefecture. Local migrants apply for temporary permits if they stay between one month and one year and for a permanent permit if staying over a year. Local migrants will be issued residence permits if they have contracted jobs, register a business, purchase housing, or are dependents of residence permit holders. Residence permit holders enjoy more public services and welfare than temporary residence holders and are eligible for hukou conversion. A residence-based approach for access to social services will encourage mobility and give workers an incentive to move to places where they can earn the highest returns on their labor, which will improve allocative efficiency in the labor market and help enhance productivity. At the national level, removing all mobility restrictions will play a major role in narrowing rural-urban and regional income gaps. Furthermore, making social entitlements available to all workers and their families in their areas of their residence will help deepen the human capital base, promote a healthier workforce, and alleviate social tensions. The figure below sets out illustrative pathways of reform. Stylized pathways of hukou reform Source: World Bank (2013) These are based on possible sequence of hukou reforms that gradually expands the residence permit system to diminish the relevance of hukou, ultimately leading to a simple population registry system. It shows the progressive expansion (in terms of both coverage of population and extent of entitlements) of the registration permit system among nonlocal populations over time and the parallel reduction in full hukou conversion thresholds. It also notes key steps to reach the end goal of all local and nonlocal citizens merged into a single population registry system with common entitlements based on city of residence rather than original hukou status. Source: Wang and Glinskaya (2014); World Bank (2014, 2012); Umapathi et al. (2013); O’Keefe (2004). As China is experimenting with relaxing hukou requirements, it appears that local reforms have been least complete in the large cities where rural migrants are most concentrated, at least for migrants from outside the municipal or provincial jurisdiction. At the same time, urban hukou in small and medium cities entitled the migrant to less generous social services and social protection, contributing to the limited success of the policy aimed at attracting migrants to smaller cities38. More to the point,                                                              38 Reforms in larger cities have generally been oriented to better-skilled and richer migrants, significantly limiting the labor market impacts of the reforms and reducing their equity benefits. Migrants are excluded in a variety of ways, for example, through entry barriers on skills, investments, or income or through rationing by strict income, work, and 74   the lessons from piloting show a key quandary in reform: benefits are national in scope, but the costs are overwhelmingly local because of the intergovernmental financing arrangements. This is a classic externalities issue, making collective action a major challenge. Cities capture only some of the benefits of financing entitlement reform, and the localized returns on investments remain unclear with a mobile migrant population. While their localized choices not to fund or to underfund basic services for migrants may therefore be understandable, the resulting situation is suboptimal from a national perspective. As a result, the introduction of a modern residence system needs to be national and unified, accompanied by a change in intergovernmental fiscal responsibilities that would promote fiscal sharing arrangements for social service provision. These are among the issues discussed in the next section. Managing labor incentives The longstanding issue of labor disincentives receives considerable attention among policymakers. Because of the relatively stronger and vibrant labor markets in urban areas, the quandary of labor- compatibility (and of dependency on social assistance) is a recurrent tenet in urban safety net policies. As such, this section briefly unbundles key issues and country practices in managing labor disincentive risks. It also tries to connect such theme with that of labor informality, which represents another important policy issue in urban areas. In theory, a number of factors in safety net design affect work incentives. These may include the level or size of benefits, their structure, the profile of beneficiaries, and integration with labor market activation policies (Tesliuc et al. 2014; Grosh et al. 2008). In principle, the relevance of benefit levels stems from concerns that programs are large enough to allow beneficiaries to ‘live off the benefit’. A proxy for this factor is program’s generosity, or the share of transfers in beneficiaries’ consumption or income. In developing countries, survey data shows that the average benefits size is 23 percent of the poor’s income or consumption – a level not even sufficient, on the average, to close the poverty gap among beneficiaries (World Bank 2014). Given the low initial or pre-transfer conditions of beneficiaries (e.g., the poorest devote between 60-80 percent of their income on food), it’s unlikely that a transfer of one-quarter of their income would generate work disincentives. Nonetheless, countries are managing possible size-related disincentives in different ways. One option is to establish limits per recipient households. For example, in Albania the limit to safety net transfer levels was initially linked to unemployment benefits (the maximum was set at 2.5 times unemployment benefits); similarly, in Uzbekistan the maximum benefit level is equal to three times the minimum wage. Another practice includes the differentiated benefits per household. For example, in the Kyrgyz Republic the benefit from guaranteed minimum income program is paid only for the eligible people in the household (children, the elderly, people with disabilities), and thus the                                                              residence requirements. Uncertainties and potentially high opportunity costs with respect to rural landholdings constrain demand by migrant workers for urban hukou. In developed areas, rural land values are high, and rural hukou holders could lose the windfall from land conversion if they change from agricultural to nonagricultural hukou. Local urban residents have concerns about the potential effects on service quality in cities if their localities have to absorb the costs of service provision for migrant populations. One survey of local residents in Guangzhou shows that more than half of urban residents agree that rural migrants deserve the rights to enjoy health insurance, have compulsory education, join the labor union, and vote. However, about the same number are against migrants applying for unemployment compensation, Dibao, and low-rent housing (Liu 2008). Another study, found that residents from places with better public service provision and higher public service quality tended to be more reluctant to accept migrants. The study also found that urban residents with lower socioeconomic backgrounds were more reluctant to accept migrants, as were residents of cities with higher employment pressures (Wang 2010). Managing such perceptions may be as significant an element in hukou reform as the technical and policy issues. 75   difference between the actual family income per capita and the guaranteed minimum is multiplied by the number of eligible persons (instead for all household members, such as in Romania). Another important factor in labor disincentives is the issue of marginal tax rates, or how sharply transfers are reduced when beneficiary’s earnings increase. Indeed, when the provision of a transfer is based on income levels, then those benefits could introduce an implicit tax on earnings, or a marginal tax rate. For example, means-tested transfers aimed at ensuring a minimum income level may imply that program participants face a 100 percent marginal tax rates: in such case, a small increase in non-program income may result in sudden program exclusion or in an equal reduction in program benefit. One way in which countries have managed the risk is to ‘make work pay’ – that is, reducing the benefit for each dollar earned by something less than a dollar of benefit (e.g., Croatia). Other methods consist in imposing time limits to social assistance and offering flat-rate design. In the latter case, programs provide a basic benefit to eligible households topped by a variable supplement based on the level of households’ vulnerability (e.g. Armenia). Some programs that, by design, may impose high likelihood of labor disincentives could, in practice, not do so. For example, taken literally, the China Dibao program implies that participants face a 100 percent marginal tax rate – that is, as incomes exceed the Dibao thresholds, participants disqualify for the program. In practice, however, evidence shows that the benefit withdrawal rate (i.e., the average rate at which benefits respond to changes in household income from other sources) is only 12-15 percent per annum, i.e. a 100 Yuan increase in income for participants results in only an average of 12-15 Yuan drop in transfer receipts over one year39 (Ravallion and Chen 2013). In other programs like CCTs, there is little a priori reason to suppose that they would undermine the uptake of work opportunities (to the extent that they exist). Indeed, in most cases, once a household is qualified, program support is not linked to earnings. Such design avoids creating a high marginal tax rate at the threshold or qualifying income level. So, rather than aiming at harsh exit of beneficiaries from the program (see box 17), the main issue for many CCTs is to ensure that newly qualified households can join the program, including with effective beneficiary registries and recertification processes. Box 17. Sudden exit from an unconditional cash transfer program in Bulgaria In January 2008, the government of Bulgaria shortened time limits in a flagship cash transfer program (Guaranteed Minimum Income) and began enforcing them more strictly, resulting in a drop in participation from around 60,000 to just below 50,000 in only two months. The rationale for the measure was that the buoyant labor market created an opportunity to remove employment disincentives in the social assistance system and to move people off benefits and into work. A survey in April 2008 of the initial cohort of affected beneficiaries confirmed their highly vulnerable profile. For example, almost 60 percent of the affected beneficiaries had primary education (four grades), had less than four grades of education, or were illiterate. About 70 percent of those who lost their benefits because of the change in time limits remained unemployed three months after the stoppage of the guaranteed minimum income support, despite a situation of overall high labor demand. About a quarter (27 percent) of the affected beneficiaries stated that they had not looked for jobs after they stopped receiving safety net support. They pointed to several reasons, such as the lack of qualifications, low pay offered or a higher reservation wage, family reasons (child care), and long distance to travel between the offered job and the place of residence. Not looking for a job didn’t seem to be driven by the receipt of other forms of social benefits. Source: Tesliuc et al. (2014)                                                              39 This compares favorably to standards as, for example, set out by Kanbur et al. (1995). The authors argued that, under conventionally assumed and seemingly plausible assumptions for the relevant parameters, the optimal benefit withdrawal rate would be around 60-70 percent. 76   Another important issue to consider is whether the beneficiary household includes members of working age who do not work (as much as expected). A more formal definition of this group is individuals of working age who are not in employment, education, training, or disabled (the NEETD group). The relative size of such group will put some bounds on the size of expected work disincentives. In Eastern Europe and Central Asia, the share of NEETD group ranges from 36 to 64 percent of total beneficiaries. Conversely, in a number of cases program may explicitly target people unable to work, including disabled people, elderly, etc. Following Barrett (2006), such programs may somewhat encourage a ‘positive’ dependency among such cohort. Finally, a compelling feature to countervail possible disincentives is to design programs in ways that explicitly encourage or facilitate work effort. A number of countries combine the provision of social assistance with labor activation measures (although there is limited evidence on their effects on work disincentives and employability). The measures are targeted to adult beneficiaries able to work and are meant to link beneficiaries with employment services, hence encouraging them to remain active in the labor market40. In addition, some countries may include other income-generation conditions and assets (table 12). Table 12. Work-related incentives in selected countries in Eastern Europe and Central Asia Country Registration with Work Earnings Other income conditions employment services requirement disregards Albania Y Y N Applicants who refuse agricultural land are not eligible Armenia Y N N Bulgaria Y Y N Applicants who refuse agricultural land are not eligible Kyrgy Rep. Y N N - Lithuania Y Y N - Romania Y Y Y - Source: Tesliuc et al. (2014) The issue of compatibility with work incentives is also connected to the complex issue of work formalization (and formal contributions to social protection). Individuals transition in and out of the formal social protection system throughout their active lives (Packard et al. 2012; Ribe et al. 2012). Some of these transitions depend on factors outside their control, e.g., when firms restructure, or when new technologies change the composition of the labor force. Nevertheless, individuals can also make choices that influence those transitions. Because it may often be challenging to enforce the mandate to contribute to social protection, some individuals may opt for informal sector jobs and/or contribute for shorter periods of time during their careers based on expected costs and benefits. Factors for decision-making around informal and formal labor may include, among others, the anticipated net earnings in the formal and informal sectors; the expected value of the bundle of social protection benefits (such as unemployment and pension benefits); and the benefits accruing from remain outside the formal social protection system. While cross-country experience is limited, available evidence on safety nets for informal workers show that there might be an impact on keeping or widening the level of informality, while there appears to be no effects on incentives to leave the formal sector because                                                              40 A number of positive impacts on children’s wellbeing can be attributed to safety net programs. These may include investments in early childhood nutrition and education which are closely correlated with income earning potential in adulthood (Hoddinott 2014; Alderman 2011). Yet these are seldom accounted for in labor simulations and models, probably because of the timeframe for benefits to materialize (often up to a generation). 77   of the programs. Box 18 and 19 discusse, respectively, the case of Argentina’s Universal Child Allowance program and India’s welfare funds. Box 18. Urban safety nets and labor informality: the Universal Child Allowance program in Argentina Argentina’s Jefes program was established at the height of the 2002 economic crisis and enrolled 2 million beneficiaries. The rapid creation of the program led to criticism about selection of beneficiaries. A transition to better designed programs began in 2006, when economic prospects where more stable and conducive. In particular, beneficiaries were transferred to two programs: an employment and training program, Seguro de Capacitación y Empleo (or Seguro), and a CCT program named Familias. The transition was rolled out until late 2009, when the new CCT, the Universal Child Allowance program (Asignacion Universal por Hijo, or AUH), was introduced and beneficiaries from Jefes and Familias were absorbed into it. Currently, AUH is a key government instrument for providing social assistance to the unemployed and informal workers. In particular, the AUH program provides a monthly benefit similar to the transfer for children of formal workers (about 10 percent of the minimum wage per child). In 2012, the program envisaged a monthly transfer of $340 per child (up from the original level of $180) for a maximum of 5 children. About 80 percent of the amount is transferred directly, while the remaining 20 percent is provided upon verification of school and health conditionalities. The AUH accounts for about 0.8 percent of GDP and, since 2009, reached approximately 3.5 million children per year, equivalent to 15 percent of the country’s households (and 29 percent of youth under 18 years). Based on urban survey data conducted in 31 cities with more than 100,000 inhabitants, quasi-experimental evidence shows that AUH significantly reduced the likelihood of work formalization among (informal) beneficiaries by about 8.2 percentage points. However, at the same time there seems to be no evidence that the program encouraged informality among registered workers. Source: Apella (forth.); Garganta and Gasperini (2012); Ribe et al. (2012); Rofman and Oliveri (2012). Box 19. India’s welfare funds for informal construction workers Worker welfare funds are one model India has developed for providing social protection to workers in the informal economy. These funds have been set up by various State governments as well as by the Central government of India. These funds are targeted for informal workers in specific industries and, in most cases, are raised from a cess or tax on the production/output in specified industries, especially those in which there is no direct recognized employer-employee relationship, and typically there is no contribution from government or the workers. For example, The Building and Other Construction Workers Welfare Cess Act, passed in 1996, mandates all states to form a welfare board and collect one percent of the construction cost from employers and developers. State governments collect a cess on construction projects worth more than one million rupees and transfer the amount raised to the Welfare Fund for Construction Workers, which registers workers and brings them under cover of welfare schemes such as pensions, health insurance and scholarships for children. The performance and impact of such welfare funds has not been systematically investigated. However, large shares of funds collected remain unused. According to labor ministry data, Rs 11,127 crore has been collected by various state governments across the country as building and construction workers’ cess until 30 September 2013. Of these funds, 87% were unutilized. The Delhi construction workers welfare fund spent 8.53 percent of the Rs 1,196 crore raised, while Uttar Pradesh utilized 0.91 percent of Rs 739 crore and Haryana reported an expenditure of 2.17 percent of Rs 803 crore. In addition, entry costs into welfare funds and affiliated social protection programs for informal workers are very high. Cumbersome paperwork and application processes to prove eligibility risk excluding the poorest, who often do not have the ability to acquire the relevant documents and attestations. However, there is little information available on the extent of transaction costs incurred by workers when they apply to register for fund and SP services. Source: World Bank, various internal materials. 3.4 Governance arrangements Government levels The governance arrangements of social protection across countries are a product of legislative and political processes reflecting fundamental cultural preferences, historical initial conditions, and 78   technical considerations. While specific institutional models and financing regimes do affect the scope and capacity of local governments, there is a range of critical functions that these provide in the realm of social protection and antipoverty policy. For example, Weir (2012) proposed a taxonomy based on four functions for local urban governments in the United States, including those of connector, system builder, innovator, and advocate41. The description of such functions and related examples are set out in table 13. Table 13. Roles of local governments in U.S. anti-poverty and social protection programs Role Actions Examples Connector Connect low-income residents to existing Milwaukee and Chicago’s Campaigns to expand take-up state and federal assistance. of earned income tax credits. System builder Oversee new connections among local New York City’s EarlyLearn NYC which merged child organizations to promote more effective care, Head Start and universal pre-kindergarten funds to service programs; combine state and federal support system redesign. income streams in new ways. Chicago Regional Housing Choice Initiative, a consortium of housing authorities making housing vouchers portable in the region. Innovator Launch new programs to support security San Francisco’s health care access Program Healthy San and opportunity for low income residents. Francisco. New York City’s Social Impact Bonds. Advocate Support nonprofit efforts to secure federal San Antonio support for Choice and Promise and state funds, may require matching Neighborhoods program. grants. Lobby to secure additional state and federal Mobilization to increase Community Development funds for low-income residents or to change Block Grant Funds. federal and state programs to support more effective local action. Fairfax county, Virginia, anti-homelessness campaign Support coalitions of nonprofits advocating with NGOs, faith groups and corporate sponsors. for low-income residents. Source: Weir (2012) When central governments devolve responsibilities for financing and administration of social assistance to local governments, including urban municipalities, this can create both opportunities and challenges. Among urban communities in Albania, for example, evidence shows that officials’ use of local information unlikely to be obtained on the basis of a questionnaire or formula was associated with better targeting than those that relying on proxy indicators alone (Alderman 2002). On the other hand, decentralized management can create spatial inequity in levels of local capacity, management of thresholds and benefits due to discretion in eligibility rules. In a number of high-income contexts, social assistance programs are administered by local, regional or provincial tiers of governments, rather than by national-level departments. For example, in countries like Switzerland, Austria, Canada, Norway, Spain, and Italy the administration and setting of threshold rates are entirely the responsibility of the provinces or municipalities. In other high- income countries, there is a standard national social assistance threshold and benefit rate, which may                                                              41 In the case of Brazil, for example, the State of Rio de Janeiro made some important innovations and adaptations to the federal Bolsa Famila program. In Rio, this was achieved through the Renda Melhor program, or Rio Sem Miseria, which for example enhanced payment systems and increased transfer levels (SEASDH 2014). 79   only vary due to cost of living differences42. Similarly, there are variations in funding principles and procedures between central departments and local government. The main examples of large-scale social assistance schemes managed and funded nationally are those of Australia, New Zealand and UK. The share of central financing varies from 50 percent in Denmark and close to 100 percent in France and Australia43. Also in the United States, a relatively small proportion of cities’ budgets is devoted to redistribution, with the bulk of funding for social protection coming from the federal government44. In China, mixed financing responsibilities between enterprises and local governments were common in the early phases of the urban Dibao expansion. The situation changed over time, with the share of central transfer increasing from 29 percent in 1999 to 65 percent in 2012, including supporting the poorest provinces. Central transfer also varies significantly between provinces. The richer coastal provinces—the receiving places of rural migrant workers—receive no central budgetary allocations. In contrast, both central and western provinces receive central budgetary allocation. Within a province, the richer prefecture cities normally receive no or small budgetary allocations from central and provincial governments, while the central and/provincial governments played a more prominent financing role for cities in lagging areas. For example, in 2012 about 95 percent of funds for urban Dibao for Zhencheng city (Guangdong region) were received from local government. In contrast, in Heilongjiang province, 70 percent of funds were centrally-provided (with the remaining 30 percent equally split between provincial and local city government). The process of annual determination of provincial Dibao transfers hinges on a general provincial funding formula as well as a range of factors. Those include efficiency of resource use, the level of financial contribution of provinces, beneficiary numbers, overall program performance, and local fiscal capacity. Effective management of urban safety net programs must account for differences across a continuum of urban settings in terms of size of the urban area and the rate of urbanization. Poor populations can differ substantially in their income-earning opportunities, access to markets and coping mechanisms in response to shocks between small urbanized towns in largely rural areas, district or provincial capitals, and very large cities. Similarly, rapidly urbanizing areas will have more migrants and may have more dynamic economic opportunities with greater movement in and out of poverty. Here, we consider how to account for these issues in the design of urban safety nets.                                                              42 Among OECD countries, Germany and Finland have locally administered schemes but national guidelines for eligibility thresholds (with thresholds that vary locally according to differences in the cost-of-living); Japan has six geographical cost-of-living bands. In Sweden, thresholds can be set higher locally according to fiscal resources available to local governments. The most extreme case is of Switzerland where there are national guidelines but wide local variation in actual threshold since, in theory, all 3,000 cantons can set their own threshold. 43 In the Netherlands central funding share is 80 percent, in Austria it is about 60 percent, while in Japan and Bulgaria it is close to 75 percent. 44  Historically, large U.S. metropolitan areas had the fiscal space to support some level of redistribution. During the Great Depression of the 1930s, local governments – especially big cities – pressed Washington for assistance to the poor, advocating for work and relief programs. Banded together in the U.S. Conference of Mayors, large cities constituted a powerful intergovernmental lobby ensuring that poverty and opportunity remained on the national agenda. In this regard, Weir et al. (2005) set out a typology of urban coalitions, including those party-imposed, interest-based, governor- brokered, and intra-metropolitan interest-based. However, reductions in federal urban assistance and devolution have made cities increasingly reliant on their state governments at a time when cities may have lost political strength in state legislatures. Local governments, and even big cities, have become a more limited force in national politics, with possible implications for the visibility and prioritization of needs of low-income residents. 80   Along the continuum of size of urban areas, small urbanized towns may be more like their surrounding rural areas on factors that affect safety net design. These small towns may have better infrastructure and services than surrounding rural villages, but they are also remote and may not offer substantially different economic or employment opportunities. In terms of targeting the poor and choice of safety net program, safety net designs could be similar in small urbanized towns and rural areas. However, in urbanized towns it may be less costly to provide transfers linked to banking services or to include conditionalities linked to health services. Larger district and provincial capitals tend to have better infrastructure and access to services that can facilitate integration with safety net programs. It may also be less costly to target and administer programs for urban poor in these more densely populated centers because many more poor households can be reached in a geographically concentrated area. For these same reasons, these peri-urban areas often have served as settings for experimenting with program designs in order to learn what is most effective. Pilot safety net programs can benefit from better quality of infrastructure and services and lower cost of identifying and reaching beneficiaries. Although these differences with poorer, more remote areas must be kept in mind, these settings factors have made peri-urban areas popular for experimentation early on in programs. Despite these advantages, many problems associated with cities also begin to be more concentrated in these moderate-sized urban centers, including crime, presence of gangs, larger numbers of unemployed, and more intractable chronic urban poverty and slums. More generally, the institutional framework that underpins the different dimensions of urban poverty is generally fragmented. In principle, urban safety net policymakers tend to place a stronger emphasis on multi-sectoral service delivery, skills transfer and graduation-related interventions. These approaches may require a higher degree of integration among institutions, government levels, and public-private partnerships. In practice, however, in a number of lower income countries the roles and responsibilities are often unclear and spread across multiple actors and levels. In India, for example, it was noted that urban health spans across four ministries and nutrition across six ones, with multiple scales involved within each of those. In particular, local-level municipalities tend to vary significantly in accounting practices and their capacity to deliver services; cities, provinces and municipalities themselves may have their own programs which may not necessarily be consistent with national schemes. While this challenge may be compelling also for rural areas, the spatial proximity and concentration of those administrative entities in urban contexts amplifies the need for coordination among them. Such institutional quandaries are particularly compelling for large capital cities. There arrangements are complicated and unique due to the challenge of reconciling local, regional and national interests. The multiple identities and the way the capital city is governed as a ‘city state’, ‘city within a state’ or a ‘federal district’ significantly determines the intergovernmental (fiscal) relations between the city and the national government. For example, in Ethiopia cities fulfill both ‘state’ and ‘municipal’ functions (including a range of urban poverty-related issues): for the former, local governments generally receive transfers in the form of block allocations from the center or regions while, for the latter, they are expected to fully fund their municipal functions from municipal revenues. However, those revenues are often not sufficient to meet the growing municipal functions which is associated with the rapid urbanization in the country (e.g., Addis Ababa more than doubled its population in 23 years, from 1.4 million in 1984 to 3.1 million in 2012). As a result of such functions-funding mismatch, significant gaps emerge in access to social protection and basic services in cities (World Bank 2014; Demissie 2010). 81   Relatedly, the physical proximity of municipalities with different levels of institutional and financing capacities can generate spatial inequities among neighboring areas. The example of Soacha and Bogota in Colombia set out by Davila (2013) clearly illustrates the point. The municipality of Soacha is one of 116 municipalities in the province of Cundinamarca and it is the most populated municipality neighboring Bogotá. Currently, Soacha has the tenth largest population in the country, housing some of Bogotá’s population inflow of the last decades and part of the city’s recent migrant population, especially from low income households. The high population pressures to which Soacha has been subjected and its strategic location have led to chaotic urbanization. This has been compounded by institutional and governance difficulties the result of which are vast informal settlements lacking adequate basic infrastructure services, educational facilities, parks and road infrastructure. The area of Cazucá is an example of the rapid conurbation between Soacha and Bogotá. The urban fabric that defines the border between Soacha and Bogotá as the physical limit between the two territories has disappeared. Figures 31 shows how that process led to basically ‘one city’ in just over a decade. Soacha’s Commune 4, where the neighborhood of Cazucá is located, is a zone of extreme social vulnerability. Commune 4 constitutes an even poorer area than neighboring Ciudad Bolívar in Bogotá. The latter is one of the poorest areas of the capital, but it nonetheless belongs to Bogotá, and so it benefits from much more organized and better endowed local government action, such as infrastructure upgrading as part neighborhood regularization programmes, social services such as libraries and parks, and connection to the public transport system and, through this, access to many of the capital city’s services. As the right-hand of figure X shows, the border between Ciudad Bolívar and Commune 4 appears as completely blurred. Following Davila (2013, p.144), “… for any visitor (…), what one sees is an uninterrupted spatial continuum of simple dwellings. Residents of Cazucá, however, are very clear in their mind about which street separates them from Bogotá. Those who live on their side of this invisible line, for example, do not have access to reliable water services”. Figure 29. Expansion of Cazuca (Soacha) over 1996-2007 (left) and its current administrative boundary with Bogota (right) Source: Davila (2013) 82   Those difference clear stand out in data on basic services: for example, comparing populations living on each side of administrative boundary, Commune 4 shows a 68.4 percentage-points worse off in terms of share of population without access to piped water, 59.6 points in population without access to sewerage, and almost 50 points among populations without gas (table 14). Table 14. Contrasting Bogota, Ciudad Bolivar and Commune 4, select social indicators Share of Share of Share of Incidence of Share of population population population pregnancy among population over 12 without without without gas teenage girls (12- with no primary piped water sewerage 19 years) education Bogota (average) 1.4 2.1 20.2 6.8 10.2 Ciudad Bolivar (Bogota) 5.5 6.5 14.1 11.2 17.9 Commune 4 (Soacha) 73.9 66.1 63.5 13.1 26.2 Percentage points 68.4 59.6 49.4 1.9 8.3 difference between Ciudad Bolivar and Commune 4 Source: Davila (2013) In terms of specific experience with safety nets, the Familias en Acción conditional cash transfer program was present in Soacha. According to Davila (2013), the program was one of the few tangible signs of the presence of the state in a context mired with insecurity (see box X later in this section). In Bogota the governance of Familias differed from elsewhere in the country (see section 4). In part because Bogota had financial resources to implement its own complementary programs, Familias was not operated through the mayor’s office, as in other cities, but was operated instead by the federal government. In part because of this difference in governance, the program lagged behind other cities in coverage rates of poor households until a major expansion in 2012. In some cases, municipality governments are sufficiently advanced and have adequate resources to design and operate their own safety net programs, as in Bogota. These municipal governments often seek some political advantage in operating their own program, as an opportunity to expand largess to the electorate and sometimes to distinguish program features from national programs according to their own political philosophy. In other words, in decentralized contexts mayors, for example, can play an important role in priority- setting. In a compilation of interviews with four Colombia mayors, one of them, Jose Caicedo from Zipaquira, mentioned “… we are going to ‘build’ people first and later we can build things made of cement, surface roads, erect walls… my priority is given to three issues: education, health, and care for vulnerable people” (Davila 2009, p.15). However, city-specific programs should be designed to be sufficiently consistent with their corresponding national social protection or else they risk hindering policy coherence. Importantly, as mentioned in section 2, there might be complementaries as well as overlaps between the institutional coordination of urban development and social protection programs. This is particularly compelling when similar interventions are used in both domains. For example, depending on how they are designed, public works programs can be considered either an urban development 83   intervention or a safety net program45 (Subbarao et al. 2013). The case of Trabajar in Argentina illustrates these issues (box 20), while Latvia provides an example of a ‘pure’ safety net response in crises situations (box 21). Box 20. Institutional coordination and urban public works: insights from the Trabajar program In 1996, the government of Argentina established a workfare program known as Trabajar. Through the execution of small infrastructure facilities, Trabajar sought to improve the living standards of the communities in which subprojects were located and create opportunities for temporary employment for poor workers in both urban and rural areas. The program was managed and implemented by the Ministry of Labor and Social Security through staff at the national, regional, and provincial levels. Subprojects were proposed by municipalities, communities, national agencies, and civil society organizations. The subprojects were designed to be labor intensive and relatively small, with the average project costing less than $100,000 and employing an average of 20 workers. The types of subprojects eligible for financing included rehabilitation, expansion, and new construction of community or public infrastructure, such as sewerage, latrines, potable water, housing, roads, urban works, irrigation, schools, and health centers. The wage rate was set at the same low level for urban and rural areas (the idea being that if the rate was not attractive to urban dwellers, perhaps they were not so poor). Trabajar encountered a number of difficulties in larger municipalities (cities with more than 100,000 residents). First, larger municipalities found it difficult to insert Trabajar projects, the review and evaluation of which occurred on a monthly basis, into municipal and master plans that had already been formulated and that included mostly fairly complex works. Second, Trabajar projects were small, and larger municipalities found they did not fit well into the large-scale projects that made up their capital investment plan. Third, in larger municipalities, infrastructure projects fell under the purview of the public works agencies, not the social assistance agencies, as was the case for smaller municipalities, and the public works agencies often concluded that the benefits from a Trabajar project did not outweigh the cost of proposing and implementing one. Finally, larger municipalities found it easier to contract out the work rather than employ low- skilled workers in need of more supervision. Despite these difficulties, Trabajar was still popular in large urban areas. The Trabajar program staff developed a proposal to address these issues. That proposal involved changing the project cycle for larger municipalities so that they would have an opportunity to work with a projected financial envelope of Trabajar funds and integrating them into master plans. It also allowed financing of a series of small stand-alone projects that could be part of a larger infrastructure project. These changes were never implemented, however, because the new government replaced Trabajar with the Jefes de Hogares program. Source: Fay et al. (2005) Box 21. The crisis-response public works program in urban Latvia In 2009, in response to rising unemployment rates and a relatively weak social safety net, the government of Latvia introduced an emergency public works program known as the Workplaces with Stipends (WWS) program. The program’s goal was to strengthen Latvia’s social safety net in response to the unprecedented drop in economic activity and rapidly rising poverty. Specifically, the public works program created temporary labor-intensive employment for people who were unemployed but ineligible for unemployment insurance benefits.                                                              45 In general, the benefits of public works programs are two-fold: on one hand, they provide income and support to participants in the form of wages or similar compensations in return for their labor; on the other hand, they create or rehabilitate assets for communities and beyond. In practice, programs put a relative emphasis on one or the other approach. For instance, some programs may prioritize the provision of transfers for poor, food insecure and vulnerable populations relative to the assets generated (i.e., transfers represent between 50-80 percent of the budget). This is sometimes referred to as a ‘safety net approach’ in public works (e.g., the Ethiopia Productive Safety Net Program (PSNP) fall under this category). In other cases, programs may aim to create and maintain sustainable infrastructure while encompassing broader employment-related objectives, like South Africa’s Expanded Public Employment Program. In this case, such ‘public employment programs’ are often managed by Public Works ministries and may not have a specific poverty-reduction objective (i.e., they allocate about 25 percent of budget for transfers). 84   In contrast to many public works programs around the world, the Latvian emergency public works program was open to residents of urban and rural areas who were registered as unemployed but were not receiving unemployment insurance benefits. Program participants were eligible to participate for up to six months with a two-week minimum requirement; jobs were provided on a first-come, first-served basis. There was no limit to the number of times a worker could benefit from the public works program, but reentry into the program required signing up on the waiting list, which sometimes took months to clear. From 2009 to 2011, about 123,000 people benefited from the WWS program. The Latvian public works program was rationed through a self-targeting mechanism with two main components. First, a relatively low wage—by Latvian standards—was offered to public works participants: this included a stipend of about 80 percent of the binding net minimum monthly wage, or about $200. This monthly stipend was not subject to taxes or social contributions, and all program participants were automatically insured against work-related accidents. Second, the program required that the ratio of labor costs to material and tool costs be about 80 percent. Despite the low stipend and the labor-intensive work offered, the program proved very popular and was continually oversubscribed. The waiting list was always nearly twice the number of available positions. Overall, the WWS program was successful in targeting its intended beneficiaries. Almost 83 percent of those who enrolled in the program had incomes in the bottom 20 percent of Latvia’s income distribution; 96 percent were in the bottom 40 percent of the income distribution. When the unemployment rate began to decline, the government of Latvia put a plan in place to phase out the WWS program. The key to the phase out plan was to erode benefit levels so that incentives to participate decreased. In July 2011, the stipend was reduced to $128 per person per month from $200. As job opportunities in the labor market started to emerge, fewer people relied on the WWS program—as evidenced by the shorter waiting lists at employment offices. Funding to municipalities was based on a published formula that included the number of unemployed people in the municipality and the number of unemployed people in the municipality who were not receiving unemployment insurance benefits. Both of these numbers are maintained by the state employment affiliates, which are centrally funded. the central government provided municipalities with technical assistance to illustrate tasks eligible for program support. The government followed up with inspections to ensure compliance. However, the severity of the crisis, and the lure of central funding, led all municipalities to work hard to find projects that would qualify for funding from the WWS program. Currently, Latvia implements a very small public works program to complement its Guaranteed Minimum Income safety net. The stipend paid is relatively high—more than twice the WWS rate—and there are some training elements incorporated into the program as well. The purpose of this program is to train the long-term unemployed while providing them with a subsistence income. Source: Ajwad (2014) The stages of transition in rates of urbanization will also shape effective design of safety net programs. Rapidly urbanizing areas have more migrants and less embedded community structures to assist in targeting, for example. This requires more robust systems to identify recent migrants or newly eligible households. To facilitate management of this process, some urban programs develop block- or neighborhood-level administrative offices. These can take advantage of local knowledge of changing residents to assist in targeting and to monitor delivery of benefits and services. The objectives of safety net programs are also often different in rapidly urbanizing areas, where increasing employment and access to public services may be the highest priority for the poor. These households may not be as poor as chronically poor households in areas undergoing a slower process of urbanization. Those seeking employment may be younger individuals, for example, rather than families with older household heads. The risk of ignoring unemployed males in their 20s, for example, is that these individuals may be the ones most likely to resort to crime if they are unable to find work. Indeed, insecurity and crime are major concerns affecting the poor in many major cities. Safety programs that do not explicitly account for these problems often have difficulty reaching the poor with benefits or are not as effective because the poor are dealing with the pernicious effects of crime 85   and poverty. In response to such concerns, El Salvador developed strategies for working in highly insecure urban areas in order to improve the effectiveness of its PATI program. In order to counter high crime rates in many urban areas, the PATI program put in place pilot job entry projects that included job training, employment counseling, workplace linkages, seed capital, and mentoring for youth (de Sanfeliú and Acosta 2014). These programs were intended to improve opportunities for more sustainable income for this vulnerable population. Although initial evidence indicates that these programs had limited effects, it was recognized that improvements in their designs may ultimately lead to more effective strategies to improve employment prospects for youth and reduce crime. Quasi-formal and community-based organizations One important institutional issue emerging in urban programs is the partnership platform with non- state actors, particularly local nonprofit and community-based organizations. Given the universe of different actors and operators, local governments can help ensure coordination and coherence among the web of local organizations. These often play a key role in supplementing and integrating state- level capacities, especially in times of distress (see box 22). Weir (2012) reviewed some interesting institutional arrangements from the United States. For example, in 2006 the city of New York created a Center for Economic Opportunity to serve as an incubator for new poverty reduction initiatives, including overseeing the pilot CCT program Opportunity NYC. Other cities have innovated on a smaller scale, often with nonprofits taking the lead. Also, the city of Seattle worked with the Neighborhood Farmers Market Alliance to enhance the use of SNAP vouchers at farmers markets. Likewise, Philadelphia collaborated with a local nonprofit, The Food Trust, to provide supplemental funds for SNAP recipients who shop at farmers markets. Yet, in most cities such web of organizations has largely developed in an uncoordinated way and struggled to respond to growing needs. For example, a study revealed that 42 percent of nonprofits ran a deficit in 2009, with the smaller ones being hardest hit. A 2011 survey of nonprofits in Los Angeles County found that only half were financially stable, and that providing basic services for the poor failed at twice the rate of other services. Box 22. Combining formal and informal provision of safety nets: food banks in urban areas Food banks provide food to charities and other grassroots organizations, which in turn distribute them to vulnerable populations. As such, they tend to complement more institutionalized, state-provided safety net programs. Recent estimates show that nearly 60 million people turn annually to food banks in high-income nations, or about 7.2 percent of their population. Most EU countries support the Food Aid Programme for the Most Deprived Persons (MDP) of the European Commission, these are integrated with other networks (e.g., FEBA network, churches, etc.) that partially overlaps with the MDP. Also, countries have national food bank systems, such as in Germany where the Die Tafeln network reaches nearly 1.5 million people. In the United States, Feeding America covers about 80 percent of the American food banking system (which reaches almost 40 million people), many of which (about one-third also benefit from SNAP). While food bank initiatives were recently launched in Saudi Arabia, Taiwan, and Hong-Kong, these also have a significant presence in middle-income countries. For example, in South Africa food banks support about 378,000 per year, while India, Mexico and Turkey are also ramping-up their capacity in the realm. In Brazil, food banks reach almost 1.5 million people per day (Restaurante Popular): these are government-funded program where beneficiaries pay about $0.5 for two full meals a day (the government subsidizes about $2 per person). Overall, such scenario has a number of implications for urban social protection policy partnerships. For instance, the diversity of local models, activities and operations pose significant challenges for coordination. For example, in the United States over 40 percent of clients served by Feeding America had not applied for SNAP since they (erroneously) assumed they were ineligible. A full mapping of existing actors and their capacity would help establish more strategic and operational partnerships between formal and informal provisions of safety nets. Source: Gentilini (2013). 86     Social intermediation services Social intermediation services do not bring directly material benefits to the families; they instead stand right in-between the demand and supply of social services and facilitate access to programs. The services address the “choice overload” problem that prevents extreme poor from effectively using the social protection system, including because of challenges – frequent in urban areas – such as limited awareness on existing interventions, high opportunity costs to accessing them, distrust or lack of familiarity with formal bureaucracies, etc. As such, they must be well integrated within the social protection system, have inter-operable information systems to track the availability of programs and the needs of the population46, and employ qualified, informed and well-trained social workers. In line with these features, a range of countries such as Chile (Chile Solidario), Mexico (the mentioned Modelo de Atención Personalizada de Oportunidades), and Colombia (Red Unidos) are investing in social intermediation is a key backbone in their social protection, especially in urban areas. Box 23 briefly describes the Chilean experience in connecting institutions at local level. Box 23. Institutions for local social intermediation: the case of urban Chile Solidario Chile Solidario is not a specific program or social benefit, but a management model based on the articulation of institutional and local networks to provide social protection to the poorest families. It aims to coordinate mechanisms to organize the delivery of integrated services to support families in extreme poverty (though, where gaps have been identified, Chile Solidario led to the creation of additional programs), by guiding them to effectively use the social services network to overcome factors responsible for their poverty. Chile Solidario began its operation in 2002. In 2012, the effective cumulative program coverage amounted to 482,558 families (around 2 million people). The Family Support component and the preparation of a family-specific development plan (Programa Puente) are the backbone of Chile Solidario. (Since 2006, additional vulnerable target groups were incorporated into the system, such as homeless individuals, the elderly living alone, and children dependent of adults in prison, and the support component was adapted accordingly). The Family Support component is the entry point to the system for 2 years. The service basically responds to the need of providing families with a personalized, caseworker service to navigate the social protection system, and establishing a relationship of trust, aiming at developing greater self-confidence and psychosocial support to successfully face the challenges of access and permanence in the network of social protection programs. The family support service features professionally trained. (Twice a year, Family Counselors go through a binding performance evaluation process, and only those who achieved satisfactory scores can continue in the job. The performance evaluation assesses 3 dimensions: personal abilities, knowledge of social services and ability to connect with them, and productivity (coverage; number of families graduating; social empowerment of families); 95 percent of family counselors are university graduates). Family Counselors who regularly visit families at home. One Family Counselor works with between 60 and 100 families simultaneously, some of them in the initial intensive phase (weekly or biweekly) and others in the phase of monitoring and follow-up (monthly, bimonthly or quarterly visits). The frequency of contact with each family decreases over time. The Family Support Service is complemented by a family cash transfer (flat amount per family), called Bono de Protección Familiar, delivered on a monthly basis and intended to help finance the costs associated with access to services. The transfer value decreases every 6 months. In addition to guaranteed benefits, families covered by Chile Solidario have preferential access to a number of social programs. Such access is enabled through interagency agreements that provide institutional conditional transfers to providers of services. All families that conclude the Family Support stage automatically enter in a monitoring and tracking phase of life conditions. During a 3 year period, families keep guaranteed benefits and preferential access to social programs, and the Bono de Protección is replaced by a Bono de Egreso (Exit Cash Transfer) for 36 months operating as a “prize” to the completion of the first stage. Since Chile Solidario is a management model, it was necessary to develop a set of management tools that facilitated its operations, which comprised: (i) an Interagency Coordination System that aims at achieving both horizontal (between institutions) and                                                              46 In a context of rationed social services, for instance, social intermediation services may need to negotiate priority access for their beneficiaries. Social intermediation services may also need to advocate for tailoring the design of available programs to the needs of their beneficiaries; changing eligibility rules to include their beneficiaries; or for the implementation of new programs, to cover emerging sources of vulnerability, or as bridges for the extreme poor to be able to access other programs (think, for instance, at a literacy program for becoming eligible for a job training program). 87   vertical (between levels of government administration) coordination. The daily management of Chile Solidario occurs at the local level. However, a national coordination process focuses on the articulation and coordination of resources (programmatic, managerial and financial) necessary for an effective and efficient local implementation; (ii) an Integrated Social Information System (SIIS) containing data of families and their members. In addition to being the backbone of Chile Solidario’s targeting system (based, for the most part, on a proxy means approach), the SIIS is the device that allows calculating the demand for services and monitoring the available supply, family support service and changes in the living conditions of the beneficiary families; (iii) a sub-National Management component based on annual work plans, both at regional and local level, and a mechanism to transfer implementation resources to Municipalities; and (iv) a Financial Management Component based on a mechanism of conditional transfers to institutions that provide services and benefits to the population covered by Chile Solidario. Evaluations of Chile Solidario found that beneficiaries have better access to direct cash transfer programs, and also better knowledge of the supply of social programs. In urban areas, Chile Solidario is found to have positive effects on psychosocial dimensions and participation in job training programs. In particular, studies found a 20 percent increase in the employment rate of females only if they were not employed before 2002, and if their family enrolled in Chile Solidario after 2004, when the supply of training and employment programs was significantly scaled up. The demand for social programs induced by the Chile Solidario translated therefore into employment outcomes only for people who were previously out of the labor force, and when met by a corresponding increase in the availability of training programs side (Carneiro et al. 2014; Galasso and Carneiro 2008). Source: Camacho et al. (2014) 3.5 Complementary interventions This sub-section explores complementarities and possible linkages between social assistance and other key dimensions in urban areas. The discussion follows the framework laid out in the introduction of the paper (figure 1), including on the role of safety nets in helping the poor pursuing upward mobility. The following discussion is not meant to offer a comprehensive overview of the all complementary interventions, but to simply touch upon some emerging lessons and experiences. The dimensions discussed here are illustrated in figure 30 in terms of three broad circles, including encapsulating spatial, economic and social realms. Each of these have their own measures as well as common areas where they intersect. Examples are provided of issues that pertain to one or more circle, with the central area encompassing matters relevant for the three dimensions. So the next paragraphs will provide some considerations on each of those overlaying circles.   88   Figure 30. Mapping urban development, economic and social protection activities ‐ Transport to job poles  “Spatial” urban  “Economic” space  planning and  and job agenda  development  ‐ Violent/crime areas  “Social” space        ‐ Targeted programs for poor slum dwellers  (including safety  ‐ Labor‐intensive public works  ‐ Poor areas with unclear land tenure  nets)  ‐ Affordable housing for the poor  ‐ Integrated service delivery  ‐ Urban disaster risk management  ‐ ALMPs for subsistence entrepreneurs ‐ Unskilled urban migrants  ‐ Childcare services for poor working women  ‐ Unconditional transfers    ‐ Demand‐side transfers for social services (CCTs)  Source: adapted from World Bank (2014b) In terms of employment, there is growing interest around the safety net ‘plus’ agenda (or ‘productive inclusion’), including ways to harness the linkages between social assistance and labor-market interventions. Such agendas are in many ways nascent, especially in urban low-income countries. As we’ll explore in the next section of case studies, countries such as El Salvador and Latvia have experimented with connecting safety nets with employment opportunities, especially self- employment47. As discussed in section 2, self-employment and small-scale entrepreneurship can feature significantly in urban job profiling. Conceptually, the poorest and most vulnerable among the self-employed are referred to as subsistence entrepreneurs – or those who are self-employed out of necessity and who often lack skills and entrepreneurial traits – as opposed to vocational and transformational self- employed workers. For instance, it is often suggested that programs to promote entrepreneurship should target individuals with the highest growth potential, with sufficient cognitive skills and entrepreneurial aptitude. One recurrent finding from the literature, however, is the different profile and needs faced by subsistence self-employed, including peculiar individual constraints in terms of aptitude, skills and social capital. A recent review of 106 small-scale self-employment interventions underscored the importance of a                                                              47 In terms of linking safety nets to wage employment opportunities, one noted example in the literature is Argentina’s Jefe y Jefas program. This began as a non-contributory compensation to unemployed heads of households with poorly documented eligibility requirements (Galasso and Ravallion, 2004). As the fiscal costs for the program became unsustainable, an additional requirement of either 20 hours of public works organized by the municipalities (not the center) or else an equivalent commitment to education and training was added. In order to encourage new employment, the program also offered to pay wage subsidies for up to 6 months to new entries into formal sector jobs. Despite a favorable assessment in regards to poverty reduction, the program was less of a mean to address unemployment during a financial crisis, and more a vehicle to bring new entrants – generally women – temporarily into the labor market. Few beneficiaries, however, took up training or schooling alternatives or were able to take advantage of the wage subsidy option. 89   carefully-devised process of beneficiary profiling, selection of technical interventions, their implementation, and evaluation (Cho et al. 2014). Among those programs, there are only a handful that are specifically targeted to urban areas (table 15). Despite none of these programs have an explicit link to safety net transfers, they can elicit important information for their design and performance of safety net ‘plus’ approaches. Small scale self-employment programs could be an important complementary intervention to social safety nets, including through supports for skills development, financing, mentoring and networking (see box 24 for an example). Table 15. Examples of small-scale self-employment programs in urban areas Country Description Kenya adolescent girls in the urban -Microfranchise in a few business areas (e.g., food and hairdressing) slum in Nairobi (Girls Empowered by for adolescent girls in Nairobi slum areas Microfranchise) -Skills training, business infrastructure and products, access to finance provided to about 2,400 girls between the ages of 17 and 19. Lao PDR adolescent girls in three -Identified strong business ideas through business idea presentation provincial capitals. and support young entrepreneurs -Business skills training, mentorship, and seed grants provided. - Out of 100 entrepreneurs participated in training, 30 was selected to receive services. Liberia adolescent girls in -About 2,100 girls between age 16 and 27 who are out of school but communities in Greater Monrovia and literate identified through community mobilization. Kakata city (business development -Six months of skills training and business development, access to track) finance, life skills training and mentorship provided -In addition, savings opportunities and child care services provided Ethiopia Women Entrepreneurship -Female owned micro-and small enterprises (MSEs) that have a Development Project (WEDP) growth potential assessed by participating financial institutions -Access to finance, skills training and business development, technical assistance provided Bangladesh BRAC Ultra Poor (Urban) -Women from ultra-poor households in urban slums -Assets transfer, skills training for business development and lifeskills. -About 200 households in 13 slums were assisted by the project Venezuela Technoserve Business -Business plan competition & start-up advice Solutions to Poverty -About 44 entrepreneurs received business consulting Source: Cho (2015) Cho et al. (2014) noted that although some programs select poor and vulnerable beneficiaries (BRAC’s Ultra Poor program), others assess the success potential for businesses to select beneficiaries (e.g., Ethiopia WEDP, AGI Lao, Technoserve Venezuela). Some programs report that business plan review/competition serves as a tool to screen applicants when targeting entrepreneurs with limited skills and education. Given a potentially highly heterogeneous group of beneficiaries, an appropriate profiling system for beneficiaries is key, including to structure the package of services and avoid offering “off-the-shelf” benefits. For example, in light of the poor water and sanitation 90   systems in urban slum, BRAC’s UP urban programs provided, in addition to business development services, healthcare support and awareness. The reviewed urban programs provide combinations of services, although some focus more on training (e.g., AGI Liberia) and others give a greater emphasis on financing (Ethiopia WEDP). Business training is the most commonly offered service, with vocational training added (AGIs, microfranchising). Various financial services are included, but only a few utilized the traditional microcredit loan. AGI Liberia provided savings account to the participating girls to foster better resource management skills; IRC’s microfranchising and BRAC’s UP transfer in kind business products and assets. The scale of those urban projects varies between 44 individuals in Venezuela to 2,400 girls in the IRC’s microfranchsing project in Kenya. All programs are relatively small scale and at pilot stage. Early findings show that closely matching the design of the project to the needs of beneficiaries can in part explain the programs’ success. For example, the promising results in Liberia are partly sparked by the provision of child care, the fact that training took place in the beneficiaries’ residential areas, and the addition of saving account and life skills training to empower girls. While IRC’s microfranchising is ongoing, preliminary evidence suggests low uptakes and high drop out of the program due to the high mobility of beneficiaries. Also the BRAC’s urban program has not yet been evaluated, although anecdotal evidence suggests that the project faced a different set of challenges than in rural areas48. In general, emerging experience shows that the menu of business activities offered under small-scale self-employment programs need to further reflect urban contexts. For instance, business activities tend to be concentrated to a few occupations, such as tailors and hairdressers for women and electricians or carpenters for men. However, more occupations can be considered and explored. In this regard, it is important to conduct entrepreneurship awareness and sensitization, so that applicants see themselves as potential business people; outreach activities need to be considered, as subsistence urban workers with limited social networks but high mobility are the hardest group to reach; applications should be simple, given the target group’s lack of education and skills; potential beneficiaries would need support in preparing these applications in selecting types of activity and requested services; and there should be a local market analysis to assess whether chosen businesses/activities are economically viable and not subject to market saturation. Box 24. Connecting safety nets and jobs: the Sustainable Livelihood Program in the Philippines The Sustainable Livelihood Program (SLP) was commenced in January 2011 to help graduating poor Pilipino households (included in the national poverty registry) and CCT beneficiaries which form a subset of those. From 2011 to 2014, the SLP served 478,281 families (with 6 percent of these being urban) and out of which 87 percent (413,944 families) were Pantawid Pamilya CCT beneficiaries. Projects under the SPL umbrella are run in partnership with relevant ministry departments. For instance, through the Trabahong Lansangan ng Programang Pantawid Pamilyang Pilipino, the Department of Public Works helps to guarantee employment for Pantawid Pamilya beneficiaries. The Department of Labor and Employment (DOLE) plays a pivotal role to ensure that SLP Employment Facilitation track beneficiaries are gainfully employed. Employment fora and job fairs are conducted, convening employers whose job vacancies match with those of SLP participants. The DOLE also facilitates the ‘HELP ME Convergence Program,’ aiming to implement a sustainable and responsive method to address child labor. has two tracks: enabling micro-enterprises and employment facilitation through capacity development and partnership building. The job opportunities track has a primary urban focus and provided skills profiling, job matching, occupational guidance and job referrals. The Department of Welfare and                                                              48 See http://blog.brac.net/2013/03/bracs-ultra-poor-program-migrates-to-the-city/ 91   Social Development (DSWD)’s job matching services is provided through the assistance of City/Municipal Social Worker and the DSWD Project Development Officer (PDO). The field implementers are tasked to determine the training needs and job qualifications of the participants relative to existing employment needs. Their duty includes networking with existing job placement agencies of government and private sector. Continuous upgrading of technical skills is provided to the participants in order to equip them to independently search for employment opportunities thereafter. The SLP offers various modalities: for example, the Self-Employment Assistance–Kaunlaran (SEA-K) Capital Seed Fund provides up to P10,0000 per participant. Another program modality is the Pre-Employment Assistance Fund (PEAF), with cash assistance for SLP participants who have potential or guaranteed employers that need financial assistance to obtain the necessary pre-requisite requirements for the job. By December 2014, a total of 895 projects have been implemented. Monitoring at SLP is conducted monthly, quarterly and annually. Moving ahead, the SPL aims to strengthen its monitoring and evaluation processes by enhancing the information system, laying out financial and work plans, data visualization (through GPS based monitoring), and mobile phone-based data collection. Source: GoP (2015) Looking beyond programs for small scale self-employment, a number of interesting practices are emerging from Latin America. These present some similar challenges as those discussed in Cho (2015), although also present peculiarities and lend further qualitative evidence. One of the most interesting cases is Brazil’s Plano Brasil Sem Miseria (PBSM). The framework includes three components designed to (i) increase income among the poor (e.g., Bolsa Familia), (ii) provide access to services (e.g., for education, health, social assistance and food security), and (iii) foster productive inclusion. The latter component envisioned a clear distinction between urban and rural approaches, including in terms of objectives and specific programs (figure 31). Figure 31. Brazil’s PBMS differentiated productive inclusion activities in urban and rural areas Productive microcredit  Program to enhance  Seed distribution  productive rural activities  program  Pronatec  Mulheres Mil   Program Crescer  Program to support environment  Universal access  PBSM  conservation (Bolsa Verde)  to water program  National Program for Access to  Technical Education and Jobs   Food purchase  Technical assistance on  program  rural extension  Urban areas   Rural areas   Source: devised based on Peas-Sousa (2013) As noted by Paes-Sousa (2013), the set of interventions for rural areas offer technical assistance to extremely poor farmers, quality seeds, and resources for the acquisition of equipment and other inputs to increasing the quantity and quality of food production. The access to quality water and electric power is also fundamental to improve life and working conditions in rural areas, and PBSM therefore pays particular attention to these aspects. In urban areas, instead, is more geared toward promoting professional qualification and skills, including for those seeking wage employment as well as self- employment. Recent reviews have also shed light on key issues emerging from operating in selected favelas, such as those of Rio de Janeiro. For example, the analysis by Villarosa (2014) documented a vibrant set of 92   initiatives, including to connect participants listed in the CadUnico (i.e., Bolsa’s he management and information system for Bolsa, with various interventions and skills trainings (box 25). The evidence on income gains from those trainings seems mixed. Program participation doesn’t seem to place a significant opportunity cost to participants, whose transaction costs are generally covered (transport, fees, etc.); also, there is anecdotal evidence of savings from possible expenses as a result of new skills (e.g., domestic repairs). Yet a central contribution of the trainings seem to revolve around galvanizing motivations for planning activities and projects, many of which however are yet to materialize. At the same time, qualitative benefits such as self-esteem and social integration seem significant, especially in initiatives where favela and non-favela people participate in common training activities. Box 25. Productive inclusion in Brazil’s favelas There is a range of initiatives and organizations are present in the Alemão Pavão-Pavãozinho-Cantagalo, and Formiga areas of Rio. These include among other FIRJAN (the Federation of Industries of Rio de Janeiro), and particularly those by SESI (Serviço Social da Indústria – the social service of the Industries Federation), SEBRAE (Serviço Brasileiro de Apoio Às Micro e Pequenas Empresas – the national agency supporting small and medium enterprises); the Instituto Pereira Passos (IPP) of the Municipality of Rio de Janeiro; the State Secretariat of Social Assistance and Human Rights (SEASDH); the social components of PAC (Programa de Aceleração do Crescimento, a large public investment program including, among others, slums upgrading interventions). SESI is one of the organizations of the wider FIRJAN system and it is present in all the favelas with ‘UPP’, or pacified areas. The package that is offered by SESI includes measures on education, health, culture, knowledge, leisure, sport, specific activities for the elderly, and vocational training. While basic vocational training is offered by SESI itself, professional training is provided by SENAI (Serviço Nacional de Aprendizagem Industrial – national service for industrial learning), another organization of the FIRJAN system. Fees are usually paid in order to attend the courses of SENAI, but through SESI Cidadania these are provided for free to the residents of the pacified favelas; in addition, free transport to the SENAI centers is offered. One of the pillars of SESI Cidadania is the presence of the program in the field and its rooting in the community. One or two agents from the local community are selected to work full time in each UPP favela. In addition, a volunteer employee of FIRJAN is designated for each favela, including to help partner with other stakeholders. PRONATEC is mainly a vocational training program that is funded by the federal government, managed by state or municipal governments, and implemented through different partnerships. The supply of vocational courses through PRONATEC is provided by SENAI and, partly, by SENAC (Sistema Nacional de Aprendizagem do Comércio – national learning system for trade). People who are interested can have access to the list of courses that are available on a website, apply on-line or at a local CRAS (Centro de Referência da Assistência Social – referral center for social assistance), which is also the entry point for CadUnico and thereby Bolsa. Then, the list of candidates who have been registered in the CadÚnico goes to SESI. From this moment on, it is SESI and its partners who take care of the candidates from PRONATEC. Yet some issues exist with data management, as data systems underpinning CadÚnico/PRONATEC and SESI Cidadania seem not fully integrated. The Program for the Development of Entrepreneurship in Pacified Communities is run by SEBRAE in all the areas with UPP. It aims at supporting micro and small entrepreneurs (those whose monthly turn-over is respectively below USD 2,500 and below USD 150,000). As for SESI, SEBRAE has a long-standing presence (since 1996) in the favelas of Rio de Janeiro. However, the pacification of the favelas brought new challenges, together with the need for an integrated approach, focused on the specificities of the business environment of the favelas. The other entry point of the program is the recently approved law for microentrepreneurs, which aims at promoting the formalization of their businesses by simplifying the bureaucratic procedures – most steps can be made on-line, and the cost for keeping the registry up to date ranges between USD18 and 21 per month (even so, the default rate is high). All micro and small entrepreneurs are eligible for the program. The program does not have a constant presence in the favelas. However, SEBRAE field staff (which are mostly made of contracted consultants) visit large favelas twice a week, and the small ones once every two weeks. According to SEBRAE managers, the main result of the program so far has been the systematization of a methodology for poor areas, including the provision of 8 interventions. More than 14,000 people have been attended by the program so far, and more than 2,000 entrepreneurs formalized their business – half of which in 2012 only. 93   Empresa Bacana (“Cool” Firm) is a program that supports the formalization of microentrepreneurs. It is run by the IPP of the Municipality of Rio de Janeiro, and its activities in the field are closely integrated to those of SEBRAE. Empresa Bacana enters the favela after SEBRAE made the preparatory mapping and mobilization. Its presence in the field is not permanent – Empresa Bacana spends approximately one month in each favela, focusing on the articulation between different institutions and sectors who are supposed to participate in the process of formalization of micro-enterprises. These institutions gather in given dates in the favela to attend altogether the different steps that are necessary to formalize the firms to participate in the process of formalization of micro-enterprises. These institutions gather in given dates in the favela to attend altogether the different steps that are necessary to formalize the firms. Finally, ‘PAC Social’ is the social component of PAC projects of the State Government of Rio de Janeiro. These projects are among the largest under PAC in Brasil, and have been concentrated in the areas of Manguinhos, Alemão and Rocinha. Although PAC Social is not a proper productive inclusion project, some of its components showed to be relevant in this respect. In addition, employment and income generation is formally one of the social components of PAC according to the federal norms that regulate it. In this respect, the experience of PAC Social can provide useful insights about the articulation between urban upgrading and productive inclusion Source: Villarosa (2014) Similarly to the considerations discussed for safety nets, the challenge of locating, reaching and informing potential and past participants looms large as a key barrier for the productive inclusion programs’ uptake, monitoring and evaluation. Yet it is clear that program implementers in favelas have deployed a variety of ways to enhance awareness and mobilization, including creative solutions ranging from fairs to social networks. From a related standpoint, the high mobility and dynamic nature of slums livelihoods make it challenging to ensure proper documentation for program enrolment. For example, the provision of school certificates (grade 5 or 6) might prove difficult for migrants, hence preventing them from participating in trainings and other activities. At the same time, programs have shown considerable flexibility in adapting services based on customer needs, including in terms of customizing working-time schedules. Yet the prevailing approach envisage the provision of “packages” of predetermined interventions. Evidence also points to some trade-offs that emerge from program implementation. For example, the formalization of micro enterprises by SEBRAE may generate a number of positive spillovers, including enhanced access to credit and economies of scale in purchasing supplies; notwithstanding new legislative provisions on the matter, the costs of the bureaucracy entailed by the formalization process seems to sometimes outweigh its intended benefits, hence discouraging applications. These types of barriers also emerge for other program areas, such as the set-up of small business activities as part of PAC Social’s work with NGOs. The physical presence of actors on the ground seems to be a key ingredient for success for program implementation. Where institutions were physically present in the community – whether as part of the NGOs strengthening process through PAC Social or as SEBRAE officials supporting the formalization process – these were perceived as significantly more effective than institutions located remotely or adopting a time-bound approach, such as Empresa Bacana. Qualitative evidence also vividly describes the aspirations of urban slum dwellers, especially the youth. They appear to be strongly motivated and highly dynamic. Those insights were key to interpret the role of the different programs provided – that is, rather than being an end in themselves, they serve as components of wider and longer-term livelihood trajectories, including networking and diversifying skills with as many courses and experiences as possible. A nuanced understanding of motivations helps also interpret some potential divergence in attitude towards certain activities (and 94   related outcomes) – for example, a key driver for attending trainings by SEBRAE seems to be the access to formal pensions; this somewhat conflicts with the objective of generating “true- entrepreneurs” as aimed by the professionals involved in the supply of those trainings. A key challenge emerging from the case study is the limited linkages to labor markets. More precisely, “what happens after the training” is a core concern of most participants. Also, part of the problem seems to lie in the internal system of mere attendance-based targets in SESI and SEBRAE. Programs are currently strengthening particular (technical) aspects of the employability of participants, but these are tempered by gaps in other areas (e.g., verbal expression, etc.) and limited intermediation of the involved institutions with potential job opportunities. While some steps to address the issue are taken– e.g., coaching programs and guidance, databases for best pupils, etc. – ensuring a proper continuum of post-training interventions remains a core and complex area to be strengthened. Engagement in non-pacified areas remains a key bottleneck in Brazil for systematic engagement in slums areas. The challenges from operating in non-UPP contexts, including in terms of both calibrating public assistance and generating private sector incentives, would likely dwarf those of non-violent areas. Another key issue revolves around the demand for trainings and support. Villarosa (2013) reported that almost half of the micro-entrepreneur in Rio’s favelas “… considered that no training or technical assistance would be useful to run their businesses”. Also, it was noted that “… traditional approaches do not work with the young anymore: they are hardly interested in spending “4 hours in a class room and then becoming assistant bricklayers”. Arguably, a significant factor in brightening the prospects for success in the observed programs would hinge on increasing the attractiveness and customization of such trainings. An area that may deserve further analysis is the interactions between the complementary domains of social assistance and productive inclusion. This would include harnessing the built-in synergies between CRAS/Cadastro (including targeting) with the other programs such as provided by the FIRJAN arms. Finding ways to overcome the hurdles around compatibility in management and information systems might be part of a larger agenda on inter-sectoral coordination in slum areas. In terms of spatial linkages, the way social assistance and spatial policy intersect is another key and much underexplored domain. Previously in this section we touched upon issues related to land tenure, which is a central concern in spatial matters. So in this context, we’ll briefly discuss another related aspect, which is that of housing and the key role it plays in shaping opportunity (Chetty et al. 2015). Each type of formal tenure arrangement, such as freehold, delayed freehold, registered leasehold, public or private rental, shared equity and cooperative tenure, and informal arrangement, like customary ownership, religious and non-formal tenure systems, have their advantages and limitations (IDMC 2015). Multiple tenure systems are the result of the historical evolution of legal pluralism, under which statutory, customary and religious laws co-exist and overlap. As such, figure 32 shows how different forms of tenure can be represented and correlated along a housing continuum.   95   Figure 32. Housing continuum model Source: Gupta (2015) Those housing options have also been pursued in fragile and post-disaster context, including to facilitate the return to more normal and stable conditions by vulnerable groups. One such experience, for example, is Haiti’s rental support cash grant (box 26). While the initiative has been implemented as an urban and disaster risk-related program, it clearly enshrines the potential for being more fully connected to social protection programs, including in terms of targeting, supporting documentation, or payments. Box 26. The Rental Support Cash Grant program in Haiti During the peak of the Haitian displacement crisis, more than 1.5 million people were living in over 1,500 camps; in early 2013, these were reduced to 320,000 people living in 385 camps. The Rental Support Cash Grant aimed to help closing the cycle of displacement and putting families back to living conditions comparable to those pre-earthquake. Despite the huge scale of the displacement, the program enabled over 500,000 Haitians to leave unplanned displacement camps. In particular, it provided financial payments to displaced families/individuals for a fixed-term lease in accommodation rented from a private-sector landlord. Housing conditions were subject to rigorous assessment, including a=only allowing those classified as viable or ‘green’ to be part of the program. Cash grants included $500 per family, $25 for transport costs, and an unconditional $125 if the family was still in the rental property 6-8 weeks after the program. Families were however responsible for providing their own housing solution after the one year rental period. Implementation experience shows that a key challenge is to establish those families legitimately in need and eligible for assistance and being genuine camp residents, registering them, and establishing a beneficiary list. In Haiti, a decision was made early on to assess vulnerability and needs using camps as the unit of analysis. In other words, if a camp was judged to be a priority for assistance, all the families inside that camp would benefit from the program. After a process of registration and communication, implementing agency staff guided the beneficiary to a low cost rental property of choice. The agency member may choose to pay beneficiaries through banks, or through mobile phone money applications. The next step is the relocation, and the need to dismantle tents and closing the camp to avoid among other, security and health risks. Haitian protection teams were however mainstreamed into all steps of the process, starting with the first interaction with each family at registration when heads of households were asked if they had any members of their family in need of extra assistance and provision which ranged from livelihoods skills, sexual health education, and replacing lost identity documentation. Source: Gupta (2015), Fitzgerald et al. (2014) Similarly, initiatives being successfully roll-out as urban service delivery programs may show potential for wider applications, including with safety nets. For example, the Mobile Water Payment program in slums of Kenya offered the option to pay water bills through an app on mobile phones. Data shows that, at least in the Kiamumbi area, such payment method increased bill payments and enhanced cost recovery by water service providers, and enhanced prospects for better service delivery. Given the widespread us of mobile technology for cash transfer programs in Kenya, the use of service- oriented apps may also suggest scope for integrating platforms for safety net delivery for targeted beneficiaries (box 27).   96   Box 27. Mobile Water Payment Innovations in Urban Kenya According to census data, Nairobi had 3.2 million inhabitants in 2009, with 30 percent of them living in slums. A particular innovation has emerged in this context, including the payment of water services through mobile phones apps. A number of urban contexts in Africa are trapped in a vicious cycle of low bill payment rates (unpaid water bills cost the urban water sector in Africa almost $500 million a year – equivalent to 0.07% of the continent’s GDP), limited cost recovery by providers, and poor service provision to consumers. Part of the latter problem is also the result of demand for services outstripping the supply. Also, high transaction costs incurred by customers and cumbersome paper-based billing processes for water service providers are further obstacles to efficient and secure revenue collection. In this context, given the spread of mobile phones in Kenya, mobile water payment solutions may offer a secure, low-cost and increasingly accessible mechanism to support the financial and operational sustainability of urban water services. The ability to pay remotely for water bills seems to offer customers both time and money savings. Also, utilities can boost their revenue collection and reduce the administrative burden of bill processing. With the near universal ownership of mobile phones amongst urban households, water bill payments can be done by linking a customer’s account to their mobile phone number, so the account and its payment functions are displayed via a simple on-screen menu. Kenya, is one of four African countries that has adopted water payment models. In 2009, it accounted for more than half of the nation’s urban piped connections. By 2011, through the mobile water payment application pioneer M-PESA, 2,250,607 people were served in Nairobi city; 2,922 in nearby Kiamumbi; 372,366 in Nakuru; 181,512 in Kisumu; and 220,198 in Eldoret. Kiamumbi presents an interesting case: in less than a year, water utility bills grew from 21 percent in December 2010 to 76 percent in September 2011; over the same period, the share of revenues of service providers collected through such model increased from 14 percent to 65 percent. Before December 2010, bills could only be paid there by bank deposit, with their nearest bank branch being 4 km away. High time and cost savings were the main motivations for mobile water payment adoption, with women benefiting most from time savings. Of households shifting to mobile water payments, 84 percent had formerly taken a public bus to pay at the bank (travel cost is $0.40). In comparison, the transaction fee that mobile network operators charge for each mobile water bill payment to customers paying their water bill with M-PESA is $0.20 cents – a direct saving of $0.20 per month. Respondents almost universally viewed time savings as a reason for transitioning to M-PESA payments. Still user’s fees may not always cater for the poorest households that tend to pay their bills in small but multiple instalments. Indeed, most stakeholders interviewed in the studies felt it was the wealthier and professional segments (as in Kiamumbi) that are most likely to use such payment services. While cost-benefit issues would hinge upon tariff structure, regulatory position, and competition amongst mobile money providers, it’d be interesting to explore how those innovations could be connected to the provision of other interventions, including safety nets. As discussed in section 4, Kenya has a range of safety net programs, some of which operate in slums areas. Connecting those agenda might be an interesting way of bridging issues around service provision that tend to be approached in parallel. Source: Hope et al. (2011) As discussed in previous sections, the issue of violence often defines significant parts of the urban space and affects their social and economic tissue. This is sometimes referred to as the ‘territorial’ dimension of urban areas, including in some cases the formation of microcosms of activities, norms and relations that can be deeply ingrained in marginalized neighborhoods and areas. Clearly, violence is a multidimensional issue, both in terms of its causes and its implications, and the role of safety nets in those contexts is an underexplored issue. However, some evidence on the broad impacts on violence is emerging49. For instance, an early randomized evaluation of the Oportunidades program in Mexico found that women receiving smaller transfers were less likely to experience violence than a control group, but a subset of women receiving larger transfers were more likely to be victimized, particularly when male partner education levels were lower (Angelucci 2008). Their findings suggest that when the income transfer is large, it threatens the male contribution to the household, and that the benefits men experience from the higher income are outweighed by the sense of ‘disempowerment’ they feel. A separate evaluation of the                                                              49 The review of the violence here presented draws from Willman (2015). 97   urban component of Oportunidades found no evidence that beneficiary women were at more risk of physical violence (Rivera et al. 2006). Over time, the risk of physical violence appeared to decrease: a mid-term evaluation of Oportunidades found that women in beneficiary households were less likely to experience physical violence than women in comparable, non-beneficiary households, but more likely to experience emotional violence50 two to six years after the program ended (Bobonis et al. 2006). Looking at the longer-term impacts of the program (five-to-nine years after implementation), Bobonis and Castro (2010) find that physical and emotional abuse did not vary significantly between beneficiary and control groups51. Overall, Bobonis et al. (2013), find that beneficiary women are 40 percent less likely to experience physical abuse, but are more likely to be threatened with physical abuse, than non-beneficiary women. That is, as time went on, men were more likely to threaten violence than actually use it, which could potentially reflect an equalizing of power relations. In some cases, CCTs have been associated with decreases in violence even in the short-term. In Peru, an evaluation of the Juntos CCT program suggests that districts where the program was implemented saw a nine percent decrease in physical violence, and an 11 percent decrease in emotional violence, compared to areas that were not part of the program (Willman and Corman 2013). A global review of CCT programs singled out Juntos as the only program impacting on gender relations (Holmes and Jones 2010). This was reasoned to be related not to the cash transfer, but to the linking of the transfers to other services and the willingness of Juntos staff to address the issue in community meetings. In Brazil, the expansion of the Bolsa Familia CCT program was found to be associated with a decrease in domestic violence, but these effects were strongest for women with higher levels of education. A 2010 study found an overall decrease in violence, using female homicide (among the 15-49 age group), as a proxy for domestic violence52. One study estimated that if the program were expanded to an additional 25 percent of women with an average of two children each, the incidence of domestic violence in the municipalities with violence rates equivalent to the national average would be reduced by 5 percent (Perova et al. 2013). However, it is noted that the observed decrease was concentrated among women with higher education levels; no reduction in homicide is found among women with primary education or less53. A study of an unconditional cash transfer program in Kenya found strong impacts on violence (Haushofer and Shapiro 2013). The study evaluated the GiveDirectly program, a direct, unconditional cash transfer program administered via mobile phone using Kenya’s M-PESA program. Households were selected via census data (they must have a thatched room and mud floors and walls) and got assigned to receive transfers via a closed lottery. Two indices were constructed – one to measure experience of violence (physical, sexual, emotional) and one to measure attitudes about violence (for example, whether it’s okay to discipline a wife, whether husbands should make decisions unilaterally. The evaluation found that transfers were associated with a 30-50 percent reduction in reports of physical violence by men against women, and a 50-60 percent decrease in women reporting rape within the marriage. The effects were estimated to be larger when women                                                              50 Emotional violence generally refers to intimidation and other forms of psychological abuse, which are often part of the pattern of domestic violence. 51 They also find higher rates of marital dissolution in the short-term, especially among younger and better-educated households. They offer the explanation that beneficiary women may experience abuse initially, whereupon those with the means of leaving the relationship do so, and achieve the economic means of moving out of the CCT program. Those beneficiaries who are still in the program nine years later may be those with fewer means to leave violent relationships. Another suggested explanation is the dissolution of partnerships with migration. 52 More specifically, an increase in the amount of the cash transfer per woman was associated with a drop in the female homicide rate. 53 The education analysis is done using number of homicides as the rate cannot be calculated; population data does not include an estimate of the number of individuals at each education level.  98   received the transfers, but still present when men received them. The study did not observe any reductions in emotional violence (similar to Prospera or ex-Oportunidades studies). Overall, it is hard to generalize across such diverse experiences. Safety nets can vary greatly in their objectives, underlying theories of change and operational models. Some programs may have economic empowerment of women as the overall objective, with reductions of violence as a secondary goal, or simply as an expected spillover effect. Some offer complementary interventions, while others may be unconditional and offer soft accompanying measures, which itself can vary greatly. Some work exclusively or primarily with women, while others actively engage male partners and explicitly try to change norms and attitudes around violence. In general, the evidence base for the impact of safety nets on violence remains thin, but some evaluations point to promising practices. The majority of evaluations on the matter in developing countries have been done in few countries and some could be subject to selection bias, because women that are attracted to these types of programs may already be more ‘empowered’ compared to others. Added to this, incidence of violence and degree of violence are notoriously difficult to measure, and are unlikely to change significantly within the short time periods generally covered by projects. This is particularly difficult in contexts where violence shapes aspects of everyday life, from physical mobility to neighborhood stigma (box 28). Box 28. Violence, confinement and stigma in urban Colombia Qualitative evidence from urban Colombia can provide insights into the socio-economic effects of violence. When asking people what are the most important problems in the Soacha Commune 4 municipality near Bogota, Davila (2013) reported that insecurity was the chief concern (alongside lack of legalization of their land, followed by lack of access to water). Insecurity and fear shaped people’s lives and mobility. As mentioned in focus groups discussions, “… those that leave before 6 am have to walk down in groups towards the Southern Motorway”; or “… after 8 pm this becomes critical”. At evening, armed gangs impose a curfew in the neighborhood. Stories of domestic violence were common, with women spending most of their time at home. The presence of local armed groups was a difficult issue to quantify and investigate, but it surfaced in interviews with residents and local officials. People associated these groups with certain areas within the neighborhood (e.g. “that hill belongs to him”) and of drug-trafficking in the area. However, the majority of people interviewed mentioned to be in receipt of the Familias en Acción program (see section 4). The public schools in the area constitute what is perhaps the most tangible state presence, although being equipped with low numbers of teachers, poor teacher training, limited material resources. This is compounded by labor market uncertainty faced by young people upon completing their schooling. In a small survey, the majority of people responded that if they had the opportunity, they would leave the area. Preference for staying was often justified on the basis of low living costs of land, utility services, housing, and rent (36.6 percent of households in Commune 4 were tenants), despite the stigma as perceived in the press or in the labor market (“once they learn where we live we won’t be hired, especially by large firms”). In other words, given the low probability of finding a better job and a higher income elsewhere, residents remain despite the negative aspects of the area. This is relevant since one might see Commune 4 as a transit area, one step on the way to settling in Bogotá. Source: Davila (2013) Section IV. Lessons from country case studies Summary. A typology of urban safety nets is proposed based on the pathways for introducing programs and countries’ stage in the urbanization process. Based on Wang and Glinskaya (2014), the China case study sets out the remarkable evolution of the urban Dibao program. An unconditional transfer that, as mentioned, reaches over 21 million people, Dibao is the result of multi-year practice with local-level experimentation and innovation. Drawing from Davila (2014), the Mexico case study takes a fresh look at the challenges encountered by the national CCT program Prospera as it expanded onto Mexican cities, and the measures that were progressively adopted to manage the process. The summary 99   subsection of Vásquez (2014) propels us to a third case study on Colombia. There, we review the conception and performance of a CCT in a large metropolitan area like Bogota. The section on Kenya reviews the experience of two unconditional cash transfer programs in Nairobi’s Mukuru and Korogocho slums (Creti 2014a); the programs were conceived out of the food prices-induced crisis in 2009 and underscores the importance of combining governmental and non-govrnmental arrangements to operate in such complex environments. The summary of Rodriguez-Alas et al. (2014) sets out the compelling experience of El Salvador’s Programa de Atencion Temporal al Ingreso (PATI) in targeting and providing a combination of public works and skills to select households in the poorest and most violent urban neighborhoods. Riccio presents key findings from multi-annual research on the Family Rewards program, a CCT implemented in New York City, and identify lessons relevant for developing countries. A discussion of an urban safety net in protracted crises underpins the subsection on Gaza’s voucher program (Creti 2014b), including being recently leveraged to deliver assistance from multiple programs with different objectives (e.g. food, shelter, education materials). The case study drawing from Bhattacharya, Sun and Prabhakar (2014) unveils the challenges of providing social pensions in large-scale slums in Delhi. Their contribution surfaces the complex web of dynamics that underpin slum’s socioeconomic texture, and how that affects access to formal public safety nets. The ensuing section summarizes the experience of the Modified CTT in urban Philippines. The chapter builds on Okamura et al. (2014) and shows how the nationwide Pantawid CCT program was customized to reach beneficiary profiles such as the homeless in urban Manila. Finally, the case study Fernandez (2014) shows the experience of a national CCT program like Program Keluarga Harapan (PKH) reaching beneficiaries in both Jakarta and other urban areas in Indonesia. This section introduces a compilation of brief case studies illustrating lessons and practices from safety net programs in urban areas. One key emerging finding is that country practices vary remarkably, including, for example, in the pathways for introducing and expanding urban safety nets (figure 33). Some countries follow a phased approach, gradually building on mature rural safety net programs and transitioning them into urban areas; this is the case in Mexico and Colombia. In Mexico, about 40 percent of the Prospera (formerly Oportunidades) conditional cash transfer program beneficiaries currently live in urban and periurban areas (2.4 million people), up from 7 percent in 1997–98. A similarly phased pathway was followed by Colombia, and the case study examines the conception and performance of the Familias en Acción program in a large metropolitan area. Figure 33. Typology of case studies by program pathways and urbanization levels Pathways of roll‐out  From rural to urban; China Mexico, Colombia from urban to rural Urban only Kenya El Salvador U.S. (New York City), Gaza Urban and rural at the India, Philippines Indonesia same time 0–50 51–65 66–100 Urbanization (%)  Other countries have followed an opposite pattern, starting their programs in urban contexts and expanding them to rural areas. In China, the urban coverage of the Dibao unconditional cash transfer program rose from 0.85 million in 1996 to 21.4 million in 2013. This growth was accompanied by an expansion of the program to rural areas in 2007. Dibao currently reaches about 75 million people nationally. Other programs have covered both urban and rural areas from their outset. This is the case for the Philippines Pantawid Pamilya program, which currently covers 4.2 million households, and which recently introduced specific design variants for urban areas. India’s social pension programs also cover both urban and rural settings, and the case study reveals the complex dynamics that underpin 100   the socioeconomic texture of slums, and how that affects access to formal public safety nets by over half a million people. Indonesia has followed a similar pathway: a case study shows the experience of a national conditional cash transfer program, Program Keluarga Harapan (PKH), in reaching beneficiaries in both Jakarta and other urban areas. Finally, some programs are launched in urban areas only. The Kenya case study reviews the experience of two unconditional cash transfer programs in Nairobi’s Mukuru and Korogocho slums. El Salvador’s Programa de Atencion Temporal al Ingreso (PATI) provided a combination of public works and skills to select households in the poorest and most violent urban neighborhoods. The case study of the Family Rewards program in New York City presents key findings from multiyear research and contains useful lessons for other countries. And the Gaza case study shows how a national urban food voucher program can operate in a complex environment, including being leveraged to deliver assistance for multiple objectives such as food, shelter, and education. Table 16 presents a summary of the 10 case studies highlighted here. These experiences represent a first generation of urban safety net programs burgeoning in a variety of contexts. While interest, practices, and know-how are growing, the role of safety nets in urban areas—and in the urbanization process more widely—remains largely uncharted. It is thus critical that countries have meaningful opportunities to exchange their experience, identify learning needs, and together frame an agenda that will be central in social protection in the years to come. Table 16. Summary features of selected case studies Location Program Type Coverage China Urban Dibao Unconditional cash transfer 21.4 million people (national) Mexico Urban Prospera Conditional cash transfer 5.8 million households (national) (2.4 million urban) Colombia Familias en Acción Conditional cash transfer 3 million households (Bogotá) (111,172 in Bogotá) Kenya Nairobi Urban Social Protection Unconditional cash transfers 4,739 households (two Nairobi slums) Programme Urban Livelihoods and Social Protection Programme El Salvador Urban Temporary Income Public works+ 30,473 people (national) Support Program (Programa de Protección Temporal al Ingreso, PATI) United States Family Rewards Conditional cash transfer 2,400 households (New York City) Palestinian Urban voucher program Unconditional voucher transfer 230,000 people Territories (Gaza Strip) 101   India National Old Age Pension Unconditional cash transfers 563,000 people (Delhi slums) Scheme; National Widow (social pensions) Pension Scheme; National Disability Pension Scheme Philippines Pantawid Pamilya Conditional Conditional cash transfer+ 4,200,000 households (Manila) Cash Transfer and Modified Conditional Cash Transfer Indonesia Program Keluarga Harapan Conditional cash transfer 3 million households (national) (PKH) (1.5 million urban) Note: ‘+’ indicates links to labor market interventions 4.1 The Urban Dibao Program in China54 Rationale Social assistance programs in urban China have played a key role in facilitating economic transition and providing support for the poor and vulnerable. Before the 1990s, urban poverty was quite rare55.1 However, since the late 1990s, China has undertaken comprehensive reforms, including the restructuring of state-owned enterprises. Urban poverty consequently soared due to massive layoffs and unemployment, affecting some 28.2 million workers during 1998–2003. It is in this context that the urban Dibao program was piloted and formally introduced, alongside other measures for the unemployed—the so-called “three guarantee lines” of Dibao, unemployment insurance, and reemployment centers to offer training and services56. Shanghai was the first city to pilot the urban Dibao program in 1993, followed by Daliang, Qingdao, Yantai, Fuzhou, Xiamen, and Guangzhou. Based on the lessons learned from local experimentation, the State Council issued a directive to establish an urban minimum living guarantee program in 1997, and promulgated the Regulation of Urban Minimum Living Guarantee System in 1999. By the end of that year, urban Dibao programs had expanded to cover all Chinese cities. Urban Dibao beneficiaries rose from 0.85 million in 1996 (0.3 percent of the urban population) to 4.08 million in 2000 (1.2 percent of the urban population). Subsequently, they increased sharply to 22.4 million in 2003 (6 percent of the urban population); this spike could be largely attributed to state- owned enterprise layoffs. Between 2003 and 2010, the number of urban Dibao beneficiaries remained relatively stable, but has declined in recent years (primarily because of reductions in inclusion errors due to enhanced eligibility management and exit controls). In 2013, the program covered about 21.4 million beneficiaries—or 4.6 percent of the urban registered population—and claimed about 0.13 percent of China’s gross domestic product (GDP) (figure 34).                                                                54 The discussion largely draws from Wang et al. (2014) produced for this review. 55 Data shows that the urban poverty rate was 6 percent in 1981 and then fell sharply, but rose again in 1989.  56  Since the unemployed were no longer covered by danwei welfare—a comprehensive welfare system based on work units—this was the group most at risk of poverty in cities.  102   Figure 34. Number of Dibao beneficiaries in urban areas (1996-2013) 25 20 Million beneficiaries 15 10 5 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013   As state-owned enterprise reform was rolled out, the scope and nature of the urban Dibao program evolved, and it became a formal poverty-oriented measure to support low-income urban working households and the elderly. And it has had a significant effect: for example, between 2007 and 2012, the proportion of informally employed workers increased from 15.1 percent to 21.4 percent, while the share of unemployed workers declined from 27.6 percent to 18.7 percent. The urban Dibao program targets only urban residents with local hukou, which lays the basis for the residence system, and excludes migrant workers and their families (approximately 167 million people)57. In addition to Dibao, China’s urban social protection system includes a number of other programs, most of which were introduced in recent years. These include the program for tekun (“three no”) people (no labor ability, no income, no legal guardian or supporter), and support in the realm of education, health, employment, housing, disaster relief, and temporary assistance58 (table 17). Table 17. Social protection programs in urban China Program Targeted Group Benefit Tekun “three no” in urban areas; “five guarantee” in rural income support, medical and social care, housing areas Dibao poor families with per capita income below unconditional cash transfer to top up the gap threshold Medical Tekun family members, Dibao families, other subsidies for medical insurance contributions or for Assistance families with special difficulties medical expenditures                                                              57  Urban migrants constitute 11 percent of the national population and more than 20 percent of urban residents. If migrants had been allowed to enroll in the program, conservative estimates show that at least 15 million (about 10 percent of all migrants) would have been eligible for participation. 58 In 2012, there were 99,000 urban Tekun beneficiaries, medical financial assistance provided support for 84.5 million people, and other temporary assistance programs (education assistance, housing and heating subsidies, etc.) covered 2.6 million urban and 3.8 million rural households.  103   Education students in high school or above from Tekun or reduction/waiver of fees and charges, grants, living Assistance Dibao families allowance, internships Employme unemployed workers from Dibao families discount interest loans, subsidies for social insurance nt contributions, training subsidies, fee waivers, public Assistance works Housing Tekun people, dibao families public housing rentals, housing subsidies, subsidies Assistance for redevelopment of poor rural housing Temporary families suffering abrupt and severe shocks; the cash transfer/temporary housing, medical care, other Assistance homeless support Social disaster victims and affected people food, drinking water, temporary shelter/resettlement, Relief clothing, heat, medical treatment, epidemic prevention, emergency rescue In 2014, the State Council issued the Interim Measures for Social Assistance, which define for the first time the objectives, targets, and approaches of the various social assistance programs at the national level. Of these programs, Dibao serves as the backbone of China’s social assistance system; most of the other programs are linked to it in terms of eligibility and enrollment. While a rural version of Dibao was introduced in 2007 and reached about 53.4 million beneficiaries in 2012, we here focus on the urban version of the program. Targeting As a strictly means-tested program, household eligibility verification for Dibao includes income and asset verification, as well as a residency requirement. The value of an eligible family’s total financial resources, including income and assets, must be below the local assistance line. Household income is measured as cash income from any source, including earnings, social benefits, and private transfers. Savings and stocks are counted as part of income. Many cities also take into account ownership of durable goods59. There appears to be a significant degree of discretion granted to neighborhood committees that make the first assessment of applicant household assets. Regulations make reference to “actual living conditions” of the household as a basis for including non-income criteria in determination eligibility. While some criteria may be quite specifically defined (for example, per capita living space) in some regions, even more specific provincial instructions leave considerable room for interpretation. Some regions establish additional requirements, which are discussed further in the benefit structure section. For example, Shanghai excludes those who receive unemployment benefits; Guangdong requires participation in birth control programs where relevant; several provinces exclude households whose economic circumstances arise from drug, alcohol, or gambling addiction. The screening phase also addresses residence status and family structure. Only individuals who have an official local urban residence status are eligible. The current application of the hukou policy when determining Dibao eligibility presents several challenges. For example, it may lead to inequity                                                              59  For example, Beijing has specified that families that own goods such as a vehicle, motorcycle, or mobile phone, or who have pets, are ineligible for social assistance benefits.  104   between official urban residents and long-term migrant families. The migrant worker status may exclude a family from rural Dibao, hence also the families of migrant workers may therefore be excluded from rural Dibao. This issue is particularly pronounced in small towns near rural areas where many families have rural hukou but reside in towns in the long run. Perspective beneficiaries' applications have to be made through local government offices (at the street office or rural/township social assistance management office). Local officials, who usually live in the same neighborhood as the applicants, investigate and make recommendations to the appropriate district government. Usually, the district government relies on the assessment reports and recommendations of the street office, based on the findings of the community-level officials. Discretion is often exercised, and it is not uncommon for applicants to put pressure on officials. Community monitoring is also a factor in the process as the list is publicly disclosed. Recent evidence shows that between about 2 and 14 percent of urban households were eligible for social assistance benefits. However, between 28 and 51 percent of eligible families were actual beneficiaries, while up to 43 percent of the households that received urban Dibao benefits were ineligible60. The results also indicate that exclusion or undercoverage of the urban poor appears to be a significant issue. This is closely related to the possible conflicting objectives of the program as perceived by local and central administrators, which is discussed later in this case study. Benefit structure In theory, determination of the local Dibao benefit should be based on the minimum livelihood cost for food, clothing, shelter, utilities, medical care, and tuition expenses. In practice, there is considerable flexibility granted to provinces (and cities) for determining the threshold, leading to wide variations in approach. The approach selected is largely driven by local economic conditions and fiscal capacities. Some apply standardized budget approaches to measure subsistence food and nonfood expenditures based on household survey data; this is the case in Beijing, Shanghai, Hebei, and Chongqing. Other provinces and cities refer to minimum wage levels (usually 30–40 percent of the average wage). For example, guidelines for Liaoning suggest that the Dibao benefit level should be lower than 65 percent of the local minimum wage. Other provinces adopt the proportion of food expenditures to total expenditures or per capita income, such as in Inner Mongolia, Anhui, and Shandong. As figure 35 shows, the urban Dibao benefit threshold increased from an average $24 per month to $61 per month—a nominal growth of 9.6 percent annually. However, the ratio of the urban Dibao threshold to per capita disposable income declined from 21.2 percent to 16.6 percent between 2003 and 2013, due to higher growth in per capita disposable income; similarly, the ratio of the urban Dibao threshold to the minimum wage declined from 45.0 percent to 33.3 percent during the same period. In terms of international standards, the urban Dibao thresholds are on average below $2 per day,                                                              60 The majority of these participating-but-ineligible families had incomes close to the Dibao line. About 11 percent of these families had incomes below the line but were classified as ineligible as a result of the asset test. The fact that they actually received Minimum Living Standard Assistance (MLSA) benefits suggests that the assets considered in the eligibility estimation—vehicles and motorcycles—are probably easy to hide and thus difficult detect. The remaining 30 percent of the ineligible participants had incomes more than double the MLSA line, suggesting serious exclusion errors. But many of those with annual incomes above the eligibility line may have had sufficiently low incomes during some months of the year to be eligible for the program. 105   meaning they are inadequate to provide income support for the urban poor61. Also, threshold levels tend to be higher in better-off areas or large cities such as Beijing, Shanghai, and Jiangsu, where the fiscal capacities of local governments are less constrained. But in medium-size and small cities or in poorer areas, threshold standards tend to be lower. For example, in 2011, the monthly urban Dibao threshold in Shanghai was $75, while it was $30 in Xinjiang. Figure 35. Trends in urban Dibao thresholds (2003-2013) 400 50 Thresholds % of Disposable Income % of Minimum Wage 300 40 Thresholds (RMB/month) Percentage (%) 200 30 100 20 0 10 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Benefits may be paid in cash or in kind, though there is a clear preference for cash. Some provinces (for example, Shanghai) also allow for material support for households whose measured income is slightly above the Dibao level. All provinces also provide festival season bonus payments and/or gifts. Interestingly, provinces have tried to maintain an equilibrium between the local minimum wage, unemployment benefits, and Dibao levels. For example, in 2003 Beijing calibrated unemployment insurance at 66–85 percent of the minimum wage, urban Dibao benefits at 59 percent of that level, and the lowest public pension at 95 percent of the minimum wage. While Dibao is an unconditional transfer program, in practice some conditions are established. These are often introduced to counter possible labor-disincentive effects and include four core practices62: for example, some provinces allow for continuation of Dibao benefits for some time after beneficiaries find jobs that would take the household above the Dibao threshold. For example, Shanghai allows full benefit payments for three to six months after beneficiaries find work, with a gradual reduction in transfer; Yunnan provides full benefits for up to three months after finding work,                                                              61 Empirical studies show that the Dibao program had limited impacts on poverty reduction in its early stage due to extensive undercoverage and exclusion errors, but improved over time. Some studies suggest that Dibao has had a larger impact on reducing the depth and severity of poverty than on poverty rates—again, largely due to undercoverage and exclusion errors. For instance, recent evidence shows that Dibao reduced the poverty gap and severity among all eligible families by 13 and 30 percent, respectively, despite having virtually no impact on the poverty rate. If full coverage and delivery had been provided, the effect would have been much larger: 3 percent for the poverty rate, 54 percent for the poverty gap, and 79 percent for poverty severity.  62  While in theory, Dibao imposes a 100 percent marginal tax rate on participants (that is, a small increase in nonprogram income will result in an equal reduction in program receipts), in practice the rate is much lower. Recent studies estimate a marginal tax rate of only about 12–14 percent a year. While such a low rate makes it unlikely that the program would provide a serious disincentive for earning extra income, it raises concerns about how well the program reaches the poorest and adapts to changes in household needs, including transient poverty  106   and Shenyang for two to three months; Beijing reduces the payment after three months. Also, provinces can provide bonuses, such as in Shanghai, where a bonus of $15 per month is provided to Dibao beneficiaries who hold jobs. In other instances, provinces may demand participation in small, low-skill community works as required by the neighborhood committee. Simultaneously, efforts are made to connect beneficiaries with the local labor bureau (including examples of common Dibao and Labor Bureau office sites in cities like Shanghai, which appear to be relatively more effective). A typical rule found among reviewed practices was that beneficiaries could only refuse such work or training two or three times before being removed from the program. Program administrators can encourage work and self-monitoring for beneficiaries, following a model first piloted in Dalian. In Dalian, able-bodied Dibao recipients were organized into a nonprofit community organization that would assist in finding jobs and volunteer work. This method appears to have had some effect, with a 20 percent reduction in Dibao rolls in Dalian the year the new system was introduced. By late 2004, this model had been taken up in different forms in most coastal provinces, including Beijing, Shanghai, Wuhan, Ningbo, Jiaozo, and Longjian. As an additional measure, some provinces leverage the tax and fee system to encourage work among Dibao beneficiaries. Examples of this approach include waivers on business registration fees, five- year waivers on land tax when used for employment, and a two-year tax freeze on self-employment earnings. These practices were found even in poorer provinces such as Yunnan. Institutional arrangements Although urban Dibao expenditures have increased from $2.6 billion in 2004 to $11.3 billion in 2013, the ratio of program expense to GDP has only increased slightly—from 0.11 percent to 0.13 percent— suggesting that urban Dibao expenditures have not kept pace with the rapid growth in GDP. Mixed financing responsibilities between enterprises and local governments were common in the early phases of program expansion. The situation has changed over time, with the share of central transfers increasing from 29 percent in 1999 to 65 percent in 2012, including supporting the poorest provinces. The central transfer varies significantly between provinces. The coastal provinces—the destination of most rural migrants—receive no central budgetary allocations. In contrast, both central and western provinces receive a central budgetary allocation. Within a province, the richer prefecture cities normally receive no or small budgetary allocations from central and provincial governments, while the central and provincial governments play a more prominent financing role for cities in lower- income areas. For example, in 2012 about 95 percent of funds for the urban Dibao program for Zhencheng city (Guangdong region) were received from the local government. In contrast, in Heilongjiang province, 70 percent of funds were centrally provided, with the remaining 30 percent split equally between the provincial and the local city governments. Annual determination of provincial Dibao transfers hinges on a general provincial funding formula as well as a range of other factors. Those factors include efficiency of resource use, level of financial contribution of the provinces, beneficiary numbers, overall program performance, and local fiscal capacity. The relative weight given to these factors varies over time. 107   Lessons learned A more coherent framework for setting thresholds is one of the key priorities for the Dibao program. Advanced economies often apply a unified formula to set the threshold level for social assistance programs. This does not mean they choose a specific absolute level nationwide, but instead allow for variations due to regional cost-of-living differences. China could consider a gradual convergence, which is already under way, from the bottom up: setting thresholds from county (city) to prefecture, from prefecture to province, and finally from province to the nation as a whole. This approach would ensure greater equity, first within prefectures and then provinces over time. Both national and sub- national governments can play an active role in equalizing thresholds. China’s Ministry of Civil Affairs fully appreciates this issue and has initiated a process to harmonize localized approaches. One area for development is the relative impact of the program on the poor and near poor. While the Dibao program has performed well in excluding the nonpoor, its design raises risks of poverty traps for households just above the eligibility threshold. Eligible households have their incomes topped up to the Dibao threshold; they also receive noncash benefits including exemptions or reductions in education fees, subsidized health insurance, public housing, and subsidized utilities. As a result, they may be better off than households just above the Dibao threshold which are not entitled to such noncash benefits but have only slightly higher incomes. This raises the risk that Dibao households face high effective marginal tax rates of graduation from the program (risk of welfare dependency) and that the near-poor will feel unfairly treated. In practice, labor disincentives do not appear to be a problem, protection from poverty is a challenge, including as epitomized by exclusion errors. In addition, evidence suggests that there can be constraints regarding the entry of new participants, due to the costs of collecting and verifying information. Some of the “stickiness” of the Dibao program can be explained by the different weights and interpretations of Dibao objectives accorded by local administrators (who may favor a promotion objective) and the central government (which seems to put a premium on protection functions). Relatedly, local hukou experimentation and reform pathways are part of the broader process under way to test and establish a residence-based social assistance system. Yet the inclusion of rural migrant families into the urban Dibao program raises a number of critical issues. Overall, it may likely increase the fiscal burden of city governments. Given the likely destinations for most migrants, the additional costs of including them in the urban Dibao program would have important consequences for local government budgets in areas that are currently not eligible for central subsidies. Some simulations show that costs would be relatively low63.9 This involves the delicate issue of financing a nationwide positive externality, including revisiting central-local financing arrangements. The administrative management process would also need to be carefully considered. Rural migrants are mobile, and it is challenging to verify their income and assets. Grounding Dibao eligibility in a residency-based approach may require clear rules on some minimum duration of residency, in addition to other province-specific requirements. The key issue, in other words, is not hukou reform per se, but rather the necessary preconditions of entitlement and welfare reform that will ultimately lead hukou to a simple population registry system, instead of being the foundation of social assistance eligibility.                                                              63  The extension of the urban Dibao program to rural migrant workers may cost between 0.02 and 0.04 percent of GDP.  108   4.2 Urban Prospera Program in Mexico64 Rationale In the mid-1990s, Mexico’s government ran 15 food-based safety net programs operated by 10 different ministries and agencies65. The Prospera program, initially named Progresa and then Oportunidades, was launched in 1997 and largely replaced such schemes. The reform was initially sparked by a major macroeconomic crisis that hit the country in 1995 (costing 6 percent of GDP in that year alone) and built on the momentum generated by social conflict and an incoming political administration. Unlike previous safety net programs, Prospera took a new approach to poverty reduction. In the course of various cabinet discussions over 1995 and 1996, it became clear that besides a shift in thinking about the relationship between food subsidy programs, income transfers, and the human capital of the poor, a new approach would require reallocation of the budget for poverty programs; reorganization of the administrative apparatus devoted to poverty alleviation and human capital; a new emphasis on measurement of program results, particularly through the use of robust impact evaluations; and a renewed political relationship between the federal government and citizens for accountable, transparent, and politically neutral program operations. With strong leadership and commitment from top-ranking government officials, the program began with an initial coverage of 300,000 families and a budget of $58.8 million. The initial target households reached by Prospera were located in rural areas66. After the pilot phase in 1997, the program was expanded to periurban localities, scaling up to about 1.6 million households in 1998, and to some 2,476,400 households in 1999. Early results of an evaluation of the program’s impact in 1999–2000 were encouraging: they showed that the components being provided as an integrated package seemed to be an effective means of interrupting the intergenerational transmission of poverty, and that what was then known as Progresa could be expanded to households with limited resources in urban areas. Feedback from the evaluation was used to guide adaptation of some of the program’s components to this new context. This combination of positive outcomes in rural areas and expanded coverage made it possible, beginning in 2001, to incorporate households in periurban and urban localities of up to 15,000                                                              64  The discussion largely draws from Davila (2014) produced for this review.  65  Several observations were made about those programs and the need to reform them. First, given the distribution of poor households, there was an imbalance in the distribution of budget funds between urban and rural areas: more than 75 percent of the total budget was channeled to urban areas, where less than 40 percent of the poor lived. Second, there was unevenness in the amounts of targeted and generalized subsidies, with almost two-thirds of all resources channeled to the latter. Third, extensive population dispersion made it difficult to deliver food subsidies in rural areas (in 1995, approximately 10.5 million people lived in 183,000 localities of fewer than 500 inhabitants); fourth, a significant share of the total budget was absorbed by the administrative expenses of the ministries and agencies in charge of the programs. Fifth, there was little coordination between agencies, leading to duplication of efforts and, in the case of targeted programs, difficulties in systematically identifying poor households because of the different methodologies used. Sixth, with the exception of a few small targeted subsidies, food subsidy programs and nutritional or health interventions were run independently of each other, and they did not focus adequately on the most vulnerable members of the family (generally, children under two years of age and pregnant or nursing women). Finally, program operations and impact were not subject to systematic evaluation; on the whole, ministries and agencies operated with great discretionary authority and little accountability  66 A rural area is defined as one with fewer than 2,500 inhabitants; areas with a population between 2,500 and 15,000 are considered periurban, while areas with populations above 15,000 are considered urban. 109   inhabitants, for a total coverage of about 3,116,000 households. In 2002, the roll of beneficiaries was increased once again, to a total of 4,240,000 households, this time including urban localities of up to 1 million inhabitants but still excluding the country’s four largest metropolitan areas: Mexico City, Monterrey, Guadalajara, and Puebla. Households from these metropolitan areas came up for consideration in 2004, for a new total of 5 million. While the number of beneficiaries remained generally stable over 2005–08, a new wave of expansion took place in 2009, which marks the start of an “innovation phase” discussed below. From that year, priority was given to incorporating urban families; some 200,000 households were added, for a total of 5.2 million. By the next year, the program was serving 5.8 million households, a level that has stabilized and is similar to its current coverage of nearly 5,922,000 as of the end of 2013. Of this total, 58.6 percent of the households are in rural localities, 18.9 percent in periurban areas, and 22.4 percent in urban areas. Figure 36 shows the growth in combined periurban and urban beneficiaries of the program. Figure 36. Number of Prospera beneficiaries in urban areas (1997-2013) 3.0 2.5 2.0 Million beneficiaries 1.5 1.0 0.5 0.0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Targeting In rural towns, national census data were used to select localities with a high density of poor households and ready access to a school and a health facility. Households were reached using a door- to-door census that evaluated their socioeconomic status (proxy means test). The assessment primarily utilized household assets, but also considered education and household composition. Prospera staff then returned to eligible households to invite them to enroll in the program. As the program expanded into urban areas, the targeting approach changed. In principle, the program was targeted to cities and neighborhoods within cities with a high density of poor households as identified by the national census. But because the poverty rates in urban areas were substantially lower than those in rural, it was considered too costly to conduct a door-to-door census to determine program eligibility. Mass media were used to advertise the program, inviting families to visit the program recruitment office in the community, and soliciting their economic evaluation. 110   In order to handle the large number of applicants, a procedure was introduced in 2001 involving a pretest, the Urban Summary Module (Cédula Resumen Urbana). The pretest served to screen applicants and determine if they should take the full-length survey; if so, this was conducted at the pretest site67. Officials next visited the households to verify the information on the questionnaire68. Applicants had to subsequently return to the program office to receive confirmation of eligibility and register. Because of the greater complexity and opportunity costs of this process in urban areas, evidence shows that only 51 percent of eligible households initially enrolled in urban areas, compared to 97 percent of eligible households in rural areas. Out of those eligible, 24 percent were not aware of the program at all. The 2002 program expansion took this experience into account, along with an increase in potential demand from cities with up to 1 million inhabitants, given strong interaction between these localities and their neighboring areas. Accordingly, a radius of influence was established for each locality based on the number of inhabitants. A total of 131 service areas were identified, with a service unit installed in each to collect the necessary socioeconomic information from the households and determine their eligibility. This process was supplemented with fliers and radio announcements. This flood of messages generated more demand than the services could handle during the first weeks of enrollment, resulting in long wait times. After this initial oversaturation, applicants became discouraged, and eventually the personnel staffing the units found themselves with a limited number of people to serve. Experience with the processes in 2001 and 2002, together with the results of a 2003 evaluation, led to the recommendation of a new approach for future urban expansion. In 2004, units would continue to be installed, but their work would be backed up with information from surveys conducted in preidentified areas with the highest concentrations of poverty. It was decided to establish 130 formal service and registration centers to provide ongoing support to beneficiary households and supplement services offered by the 32 existing state coordination offices. The reason for setting up these centers was to reach out to the households and help them reduce the time and resources it cost them to travel to the state capital. The program also introduced traveling units associated with some of these centers, which made scheduled trips and shared information through a previously established social network in sessions with groups of the population. This approach made it easier for Prospera to reach out to beneficiaries through scheduled meetings with small groups—a forerunner of the personalized service model (Modelo de Atención Personalizada de Prospera, MAPO) introduced in 2009 and discussed in the next section. This expansion phase also introduced mechanisms for updating rolls, with reevaluation of a household’s socioeconomic conditions (recertification) or in response to an inquiry from the general public (ongoing verification). Between 2005 and 2008, the program maintained its coverage of about 5 million households, introduced some new interventions—for example, Support for Older Adults (2006), Energy Support (2007), and the Vivir Mejor strategy (2008)—and redefined some of the criteria for remaining in the program. With its current broad coverage, the program can directly identify any of the country’s 192,247                                                              67 While the full survey has more than 100 questions, including detailed information on household members and housing characteristics, the pretest has about 50 questions. 68 Responses to the full questionnaire were evaluated using a nationwide scoring system to determine whether the household was eligible to receive benefits under the program.  111   localities and prioritize them using the National Council for the Evaluation of Social Development Policy’s (CONEVAL’s) social gap index (Índice de Rezago Social), the National Council of Population’s marginalization index, and available statistical information. Localities and blocks are identified, and the families are interviewed using mobile devices69. The Prospera national coordination unit sends the list of localities/blocks under consideration to officials in the education and health sectors at both the national and state levels (through their state delegations). The list includes the estimated number of households that would benefit, as well as the codes corresponding to the health units that provide care for the active families in these localities. This information helps in determining whether education and health centers are accessible and have the requisite capacity to provide the intended services (such centers are often saturated in urban contexts). After applying the pretest, the standardized questionnaire was used to estimate households’ per capita monthly income using a specific linear regression model for urban areas. The proxy income is then compared to the level defined by CONEVAL as the minimum threshold of well-being (LBM)70 and eligibility is determined. Benefit structure As a typical conditional cash transfer program, cash transfers under Prospera are provided upon achievement of activities in education (school attendance), health visits, and nutrition-related initiatives. The amount of cash transfers received by participants depends on a range of factors, including the number of household members under nine years of age, the number of members receiving scholarships, school grade, and the number of elderly in the household. There is a maximum amount of support a household can receive: the ceiling is about $128 if a household has children on scholarship for elementary or middle school; the maximum for children on a high school scholarship is nearly $208. These amounts were updated every six months between 1998 and 201271. Survey data suggest that Prospera transfers represent on average about 25 percent of the poorest households’ monthly income. Until 2009, grants were the same for both rural and urban areas, after which an urban-adjusted benefit was introduced on a pilot basis. In particular, benefits for elementary education were eliminated and reallocated for middle and high school education. Also, performance-based scholarships were introduced. Early results from evaluations are mixed, and sectoral dialogue and arrangements are ongoing to determine the specific modalities of the scholarship structure. To reduce travel and wait time for beneficiaries, cash payments are generally made through bank cards usable at automated teller machines (ATMs). However, while about 94 percent of the municipalities that are considered urban and metropolitan have at least one bank branch, 88 percent of the municipalities with fewer than 15,000 inhabitants do not have bank cards. Thus, for rural areas, benefits are delivered through prepaid cards distributed at defined outlets (for example, government-run Diconsa stores, telegraph offices, branches of the National Savings and Financial Services Bank, gas stations) or service points temporarily set up for the purpose on specific days72.                                                              69  These mobile devices are smart phones used to collect socioeconomic information. They are able to capture, process, and transmit the pertinent data wirelessly.  70  The CONEVAL minimum basic threshold is used to identify the population that, even if it spent its entire income on food, would not be able to acquire food indispensable for proper nutrition.  71 During 1998–2011, the benefit amounts were pegged to cumulative increases in the national price index for the basic shopping basket. Beginning in June 2011, the criterion has been average variation in the rural and urban LBM. 72 Household representatives are given scheduled appointments over 25 working days during the course of the month; some 300 household representatives are served each day at the temporary service points using a point-of-sale terminal. 112   In localities without any banking services, transfers are directly made in cash through the program’s customer service desks (Mesas de Atención de Programa)73. Although evaluations of the impact of Prospera in urban areas had shown positive results in terms of health, nutrition, and education indicators, the impact was less impressive than in rural areas. Thus, it was deemed necessary to develop an intervention that would better respond to the needs of urban families living in poverty—taking into account their mobility, limited time, and higher opportunity cost to comply with conditions (for example, relatively well-paid job opportunities for young people that may compete with schooling). It was also noted that the epidemiological and nutritional profile of urban inhabitants differs from that of people living in rural areas. Among urban inhabitants as compared to their rural counterparts, there is a significant increase in the prevalence of noncommunicable diseases; continued high prevalence of low height and anemia; increased tendency toward obesity and being overweight; insufficient duration of exclusive breastfeeding; inadequate use of food supplements; and low coverage of certain preventive measures such as breast examination, detection of prostate disease, and use of the Pap smear. Since 2009, a range of practical, urban-sensitive refinements were introduced to the program. These include innovations such as the Alternative Health Model (Modelo Alternativo de Salud, MAS), a revised benefit structure, and wider use of banks to transfer payments; and the development of a new model for providing orientation and personalized service, MAPO. That service provides information to beneficiary households with sessions offered every two months and attended by groups of 40 household representatives, selected on the basis of proximity to the service center74. With the transition of cash payments through bank cards, beneficiaries no longer have to be present to receive their payments. As a result, it was important to establish new points of contact with beneficiaries and to devise a different service model. To provide such personalized service, the program has a territorial structure based on operating zones and microzones comprised of one or more neighboring localities. Each operating zone has a regional service unit; microzones have a service representative responsible for various tasks, including providing personalized services for the beneficiary households, and have better control over operational execution. The MAS approach encourages greater emphasis on preventive health care while placing less focus on the curative aspect; it also aims to strengthen measures to improve nutrition. The MAS design includes a plan of action for health, promotion of better nutrition, and encouragement of self-care. With this model, the idea is to change health programming and services from an emphasis on a doctor visit to a commitment on the part of all household members to practice prevention measures specific for their age, sex, and needs. In addition to monitoring the usual nutritional status of beneficiaries, the program replaced its previous dietary supplements—Nutrisano (porridge for children under five years old with some level of malnutrition) and Nutrivida (a drink for pregnant or lactating women)—with Vita Niño (for children under five regardless of their nutritional diagnosis) and Nutrivida tablets; these still contain micronutrients but have a lower caloric content. The self-care strategy is being introduced to complement other strategies through workshops for                                                              73 Payments are still made through prepaid cards. At the customer service desks, staff from the state coordination units provide guidance for household members, clarify any doubts, and handle required procedures (reporting changes in household membership, personal data, address, means of receiving the benefit, etc.).  74  Relatedly, beneficiary representatives serve on a community promotion committee (Comité de Promoción Comunitaria) dedicated to enhancing communication between households and program staff.  113   beneficiaries tailored to address the specific urban dynamics. Topics addressed at the workshops include healthy eating, use of the supplements and physical activity during pregnancy, breast and cervical-uterine cancer, hypertension and diabetes, prevention of addiction, and intrafamily violence, among others. Taking into account the opportunity cost for urban beneficiaries, it has been decided to schedule the workshops every two months instead of once a month, and to open up the possibility for the family to appoint another family member (over 18 years of age) to represent the household. In addition, there is a list of practical health actions households can follow if they miss a health workshop75. Institutional arrangements The institutional home for the program is provided by the Secretariat of Social Development (SEDESOL), which coordinates with the Secretariats of Health and Public Education, as well as the Mexican Social Security Institute. It involves three levels of government, because education and health services are decentralized and therefore provided by state governments, while municipal governments provide security and physical space. To ensure coordination between the different entities, the program has a council and a technical committee at the federal level, as well as state technical committees. Interviews conducted for this review suggest coordination often hinges on the goodwill of the actors involved rather than on preestablished agreements. Since 2000, the program has periodically reevaluated the socioeconomic and demographic conditions of beneficiary households through a recertification process. Beginning in 2011, a schedule was adopted to re-collect socioeconomic data from 20 percent of the roll of beneficiaries every five years. All the households in a given locality are covered, starting with the localities with the lowest social gap index score. Currently, the recertification period for the localities has been extended to eight years. Household socioeconomic information is evaluated applying the same methodology that was used to identify the beneficiaries initially, but with two thresholds: the LBM plus the threshold for ongoing verification of socioeconomic conditions (Línea de Verificación Permanente de Condiciones Socioeconómicas, LVPCS). The latter, a higher threshold, corresponds to the monetary level at which a household has sufficient estimated income to cover food, education, and health needs. If the estimated household income is lower than the LBM, the household will continue to receive all the same support from the program as long as its socioeconomic situation remains unchanged. Conversely, if the estimated household income is higher than or equal to the LBM but lower than the LVPCS, it is eligible to transition to a differentiated support scheme (Esquema Diferenciado de Apoyos)76. The length of time households can be included in this scheme is determined by either the number of years it will take for the children in the household to reach the age of 12, the remaining reproductive years of the women in the household, or the remaining years for household members between ages 12 and 22 to complete high school or vocational special-needs education, respectively77. Finally, if the estimated household income is between the LBM and the LVPCS but does not meet                                                              75 These actions include guidance on recovery from malnutrition or anemia in a child under five years of age, maintenance of normal parameters for a person with diabetes or hypertension over a period of six months, long-term use of a family planning method for sexually active women, and vasectomy for sexually active men.  76 This scheme maintains co-responsibilities but excludes the food support component (including infant feeding) and elementary education scholarships in the education component. Households remain under this scheme as long as their condition of eligibility is unchanged (their estimated income is not higher than the LVPCS) and they meet the indicated demographic criteria. 77 If there are household members between 12 and 22 years of age who are not attending school under a scholarship, the household has up to 24 months to update their status or support will be discontinued. 114   the established demographic criteria, or if the estimated household income is higher than the LVPCS, household support will be discontinued. Lessons learned Over the years, the urban arm of Prospera has evolved substantially and has achieved the highest urban coverage among Mexico’s safety net programs. Through a process of learning and documentation, the experience of the program has stressed the importance of refining practices to meet urban contexts. These experiences have been particularly compelling in the realm of outreach and familiarizing beneficiaries with the program, as well as in dealing with possible excess demand. Implementation has also pointed to the importance of thinking carefully about whether conditionalities are, beyond political economy considerations, relevant as a technical feature, especially when designed to reach mobile and time-constrained populations. Relatedly, incentive-based programs such as conditional cash transfers work to the extent that the supply side of services are of adequate quantity and quality. In rural areas, physical access to such services is a key constraint; in urban areas, a core bottleneck relates to the overwhelmed capacity of services to meet additional demand. As a result, for example, beneficiaries often opt for private health care—this is the case for some 10.5 percent of Prospera participants. In this context, it becomes critical to design programs that are portable in all their components. Part of the constraint was relaxed through smart-card payments, but a truly portable conditional transfer should also include choice of different service locations and hours to attend services (especially workshops and trainings). Another important aspect relates to restructuring of benefits. While pilots are under way, the structure of education incentives should be better tailored to youth, including encouraging cognitive and life skills, as well as being more aligned with the needs of the labor market. An example would be to sustain Prospera high-performing students by granting access to scholarships for higher education or vocational training. This measure would require broadening the dialogue to engage other sectors and improve inter-institutional cooperation (labor, private sector, training institutions, etc.). Also, since a large percentage of participating households are headed by women, there might be room to connect social assistance and social care services (for example, Programa de Educación Inicial no Escolarizado, Programa Escuelas de Tiempo Completo). Closer linkages to social insurance would be desirable as well (for example, life insurance for heads of household), as well as integration with housing policy (for example, Vivienda Digna). There is a need to better coordinate actions and define roles with municipal governments, especially in large metropolitan areas. The current social protection landscape is not yet a “system” able to address the various facets of urban poverty. Only about 36 percent of the poorest quintile of the urban population is covered by some form of social assistance intervention (compared to 77 percent in rural areas). Even though progress has been made in harmonizing approaches, interventions tend to maintain their own assessment, planning, operational, management, and information systems78.14 Prospera components such as MAPO are well rooted in the social fabric and territory, providing an important foundation for integrating longer-term sectoral interventions designed to address the needs of both people and areas within a unified urban poverty framework.                                                              78 For example, based on current practices of data collection by individual programs, in a given year a household may be interviewed six times to collect both shared and specific information. If the basic information was only collected once and shared with all the programs, pre-eligible households could be identified through a basic questionnaire, with the relevant households contacted later to answer the remaining questions. 115   4.3 Familias en Acción in Urban Colombia79 Rationale Colombia’s flagship conditional cash transfer program, Familias en Acción (Families in Action), was launched in 2000 in response to the economic crisis of the late 1990s. Familias was part of a wider set of interventions managed by the Social Support Network; these interventions included Empleo en Acción, aimed at providing temporary jobs in cities; and Jóvenes en Acción, designed to improve access to the labor market and job placement conditions for youth. Familias was rolled out in three phases (table 18). The first phase (2000–06) targeted municipalities with fewer than 100,000 inhabitants80. In 2005, the program began to include families that had been forcibly displaced, many of whom lived in large cities. At that point, a series of studies was launched to explore how to adjust Familias in large cities; this included pilot initiatives in Altos de Cazucáin Soacha, Comunas 1 and 3 in Medellín, and El Pozónin, Cartagena. This urban expansion was driven by an effort to support displaced people, as mandated by the constitutional court; as well as a goal of covering 3 million families, which resulted in expanding the program to cities of more than 100,000 inhabitants. Subsequent pilot impact evaluations showed that the pilots had positive effects on beneficiaries, although somewhat smaller in magnitude than in municipalities with fewer than 100,000 inhabitants81. Table 18. The three phases of Familias, 2000–present Characteristics 2000–06 2007–11 2012–present Geographic rural—municipalities of rural and urban—nationwide rural and urban— coverage fewer than 100,000 coverage nationwide coverage inhabitants Target population families with children families with children under families with children under the age of 18 the age of 18 under the age of 18 Targeting sisben sisben ii sisben iii instruments rupd rupd rupd (and ruv) siunidos indigenous censuses Targeting criteria sisben level 1 sisben ii level 1 sisben iii scores displaced population displaced population indigenous population Registration scheduled events scheduled events scheduled events open to displaced open to displaced population open to displaced                                                              79 The discussion largely draws from Vasquez (2014) produced for this review.  80 Municipalities with fewer than 100,000 inhabitants were considered rural. The evaluation made a distinction between the rural and urban areas of those municipalities, allowing it to produce comparative results. 81 In urban areas, the results were found to be considerably weaker in terms of total consumption, food consumption, and nutritional status. However, school attendance among children between the ages of 12 and 17 rose by 5 percent in those areas. The most significant results in these municipalities have been (1) a 9 percent increase in family consumption; (2) a 420-gram increase in birth weights; (3) an improvement in the nutritional status of children, with a 9 percent reduction in chronic malnutrition among seven-year-olds; and (4) an increase in school attendance of nearly 2 percent among children aged 8–11 and over 7 percent among children aged 12–17. 116   population population Number of target: 340,000 target: 3,000,000 no target families final: 520,000 final: 2,800,000 current: 3,000,000 Transfer model same for all differentiated by type of differentiated by type of municipalities municipality municipality Note: RUPD = Master Registry of Displaced Population; RUV = Master Registry of Victims; SIUNIDOS = Information System for National Agency for Overcoming Extreme Poverty; SISBEN = System for the Identification of Potential Beneficiaries of Social Programs. Supported by those findings, in 2007, the president of Colombia made an executive decision to expand the program into large cities. This began the program’s second phase (2007–11), during which Familias’s objectives evolved from a response to an economic crisis to addressing poverty and inequality, building human capital, and improving living conditions among poor and vulnerable families. The third phase, which began in 2012, was marked by a redesign based on a new multidimensional poverty assessment. At present, 3,004,922 families are registered with the program, which represents 82 percent of those eligible. This case study reviews Colombia’s experience in rolling out the Familias program in the capital, Bogotá. Targeting Since inception, the information base for the targeting of Familias households has been the System for the Identification of Potential Beneficiaries of Social Programs (SISBEN). The target group has been the poorest households (Level 1 in SISBEN). As the system evolved and new versions were launched, Familias households were reassessed accordingly. The current version, SISBEN III, has been complemented by other databases, in particular the Information System for the National Agency for Overcoming Extreme Poverty (SIUNIDOS) and censuses of indigenous communities (box 29). Box 29. SISBEN and related targeting processes SISBEN. Since 1994, the Colombian government has maintained a database to identify potential beneficiaries for social programs among the poor population. This individual targeting database has three components: (1) SISBEN index, (2) socioeconomic classification records, and (3) scoring software. There have been three versions of SISBEN (I, II, III). In the first two versions, there was an indicator that assigned values between 0 and 100 based on household living conditions (economic resources in SISBEN I; standard of living in SISBEN II). These established cutoff scores to create six levels, with Level 1 corresponding to the poorest population. SISBEN III preserves the 0–100 scoring system but has also introduced three geographic categories—14 major cities, the rest of the urban sector, and the rural sector—and eliminated the cutoff scores for delineating SISBEN levels. The SISBEN survey was administered via two modalities: census and by request. The census-style survey was administered in two phases, first by surveying homes in socioeconomic strata 1 and 2, and second by surveying specific additional sectors. The survey on demand was administered to households that requested it. Master Registry of the Displaced Population (RUPD). The objective of this registry is to have up-to-date information on the population that has been involuntarily displaced. Any family with children under the age of 18 that is registered in the RUPD is eligible to participate in Familias. Registration in the RUPD is guided by certain rules, some of which are applicable to RUV registration. Master Registry of Victims (RUV). This registry, which includes victim statements, is a tool for guaranteeing that victims of violence receive fair treatment, ensuring that they are recognized and given access to a full range of services and assistance. 117   UNIDOS. The National Agency for Overcoming Extreme Poverty uses SISBEN to target its beneficiaries. UNIDOS has 10,800 social workers who provide diagnostics for households living in extreme poverty through direct visits. Indigenous censuses. These are lists of census data delivered by the indigenous authorities to the Ministry of the Interior. The proxy means test in SISBEN III differentiates scores between three types of geographic areas. To ensure continuity in the process of strengthening human capital, the program established a transition stage for registered families with scores above the Familias cutoff level, but below that corresponding to participation in the nation’s health plan. This transition stage allows participants two additional years of participation in the program (table 19). Table 19. SISBEN III scores Sisben III Score Geographic Disaggregation Selected In Transition area 1. 14 major cities without their metropolitan areas: bogotá, medellín, cali, barranquilla, cartagena, bucaramanga, cúcuta, ibagué, pereira, villavicencio, 0–30.56 30.57–54.86 pasto, montería, manizales, and santa marta area 2. rest of the urban sector, consisting of cities other than the 14 major 0–32.20 32.21–51.57 cities, and towns and rural areas adjacent to the 14 major cities area 3. rural sector, consisting of rural areas other than those adjacent to the 14 0–29.03 29.04–37.80 major cities Targeting begins with identifying those families with a low SISBEN score. Next, to identify families in extreme poverty that are not registered with SISBEN—due to a lack of documentation, not having been surveyed, etc.—the targeting mechanism makes use of SIUNIDOS, thus guaranteeing Familias’s coverage of the poorest families. Based on survey and administrative data, a 2011 impact evaluation provided an in-depth analysis of the quantitative, qualitative, and operational issues underpinning eligibility in urban Familias. The study found that, in 2007, only about 40 percent of the surveyed households enrolled in the program. Among the approximately 60 percent that did not do so, the main reasons included lack of awareness about their entitlement to participate (51 percent); unfamiliarity with program benefits (23 percent); and lack of time due to work, taking care of another person, or having to attend to household matters (12 percent). By 2011, the share of households that did not enroll soared to 66 percent, while those that did enroll declined to 34 percent. Among those that did not enroll, a lower share than in 2007 was unaware of their entitlement to participate (29 percent), but a higher percentage was unfamiliar with the program benefits (36 percent). In other words, the evaluation found increased awareness about accessing the program, although there was still confusion about its benefits82. Moreover, among                                                              82 For instance, the share of beneficiaries that, between 2007 and 2011, knew they had to comply with conditionalities increased from 68.4 percent to 84.9 percent for education, and from 66.7 to 75.7 percent for nutrition. The share of participants aware of the structure of benefits increased from 10.1 percent to 32.8 percent—which still constitutes a very low level of awareness. In 2011, 65.8 percent of beneficiaries were not aware of the final graduation grant offered by the program, and 88.6 percent did not know that program benefits and participation would not change if a beneficiary moved to a different municipality.  118   those households that attempted to enroll, only about half eventually managed to do so: 53 percent were successfully enrolled, while 47 percent were not. The main reasons for unsuccessful enrollment were that households were not in SISBEN I (30 percent) or did not fulfill other eligibility criteria (16 percent). Also, the process of urban enrollment entailed substantial opportunity costs for beneficiaries (box 30). Box 30. Opportunity costs of enrolment As emerged in interviews with participants, the key difficulties in the Familas enrollment process relate to time and travel. Participants cited having to spend more than 12 hours to travel to enrollment centers (including spending the night there) and long queues. On average, the enrollment procedure took 6.3 hours—almost a full day of work (or $10.50). Procuring an education certificate took three days on average; for about half the sample, it entailed a cost of $1.10, or about 10 percent of the education subsidy for children aged 6–8. The nutrition documentation also took three days, with a cost of $1.20 for 10 percent of the participants (about 15 percent of the nutrition subsidy for children aged 7–11). These costs exacerbated the considerable time span between registration and first program payment, which occurred after 4.3 months on average. Additionally, about 41 percent of beneficiaries incurred transportation costs to withdraw the cash transfer, with an average cost of $1.20; about 89 percent of beneficiaries withdrew cash from automated teller machines (ATMs), which are limited in areas with a high density of Familas beneficiaries. About 11 percent of beneficiaries incurred fees in withdrawing the cash ($1.30), as their cards only permitted a given number of free transactions. Some 41 percent of beneficiaries found the overall payment process to be unpredictable. In Bogotá, the first registration efforts resulted in a very low coverage rate of 35 percent83. As of this writing, Familias has registered 111,172 people in Bogotá, which represents 56 percent of those eligible (about 198,000). This enrollment rate is significantly lower than in the rest of the country’s municipalities, which have an average enrollment rate of 84 percent. Of those registered in Bogotá, only 68 percent have accessed benefits. Benefit structure Based on a pilot program in 2004, the Familias benefit structure was revised in 2007 to introduce a grant component. This included a sum of $50.70 for advancement from grade 9 to grade 10, and of $117 upon completion of grade 11. The primary school subsidy was replaced with a revised version of the nutrition subsidy for children between the ages of 7 and 11. This subsidy was eliminated because no significant impact was found in terms of its reducing the dropout rate in primary school (which was already quite low). Since elimination of the education subsidy could have negative effects on family food consumption, program officials increased the nutrition subsidy for that age group. In the program’s third phase, however, the subsidy schedule was abandoned due to limited effects on grade advancement or graduation, and the nutrition and health subsidies were consolidated. Tables 20–22 show the evolution in benefit structures across the program’s three phases.                                                              83 At least 10 percent of registered families did not receive benefits, mostly due to erroneous information, failure by families to present certifications and other required documents, and a high rate of relocation among the families. 119   Table 20. Benefit structure 2000–06 ($) Nutrition Subsidy Education Subsidy Children Aged 0–6 Grades 1–5 Grades 6–11 19.50 5.80 11.70 Table 21. Benefit structure 2007–11 ($) Nutrition Subsidy Education Subsidy Families Families families with with with Children Children children Aged 0– Aged 0–6 aged 7–11 Grades 9– Group 6 only and 7–11 only Grades 2–5 Grades 6–8 10 Grade 11 1 19.50 0 0 5.80 9.70 13.60 15.60 2 19.50 19.50 7.70 0 11.70 17.50 23.40 3 19.50 19.50 7.70 0 9.70 13.60 15.60 4 19.50 0 0 5.80 11.70 11.70 11.70 Note: Group 1 = Medellín, Cali, Soacha; Group 2 = Ibagué, Neiva, Popayán, Santa Marta, Sincelejo; Group 3 = Bogotá, Baranquilla, Bucaramanga, Montería, Pasto, Pereira, Villavicencio, Yopal; Group 4 = remaining municipalities (1,105). Table 22. Benefit structure 2012–present ($) Health Subsidy Education Subsidy Group Aged 0–7 Transition Grades 1–5 Grades 6–8 Grades 9–10 Grade 11 1 23.40 0 0 9.70 11.70 17.50 2 23.40 7.70 3.90 9.70 11.70 17.50 3 23.40 7.70 5.80 11.70 13.60 19.50 4 27.30 7.70 5.80 13.60 15.60 21.40 Note: Group 1 = Bogotá, Medellín, Cali, Soacha; Group 2 = Ibagué, Neiva, Popayán, Santa Marta, Sincelejo; Group 3 = Baranquilla, Bucaramanga, Montería, Pasto, Pereira, Villavicencio, Yopal; Group 4 = remaining municipalities (1,105). For the nutritional component, the transfer amount was estimated taking into account the per-child service cost incurred by community households, multiplied by the average number of children under age 7 in families at SISBEN Level 1. For the education component, the transfer amount was based on the direct cost of school attendance for families in the first quintile. Beginning in the program’s third phase, this subsidy—which can be provided to a maximum of three children per family—was adjusted, taking into account the opportunity costs for labor and school attendance. Payments are made every two months, which means that the amount covers two transfers. Among the mothers participating in Familias in Bogotá, 85 percent have a bank account into which the payment can be transferred. The other 15 percent of these beneficiaries schedules payment through a bank draft, which requires them to go to the bank to claim the funds. An estimated 60 percent of this latter group do not 120   claim the funds within the allotted 20-day period, obligating the bank to return the funds to the program. This situation merits further investigation, and might involve the need for a more proactive role for the Bogotá Regional Office (DRB) in terms of reviewing cases and taking steps to avoid the problem, possible issues around targeting, and opportunity costs to collect—in some cases— relatively small transfer amounts. The program in Bogotá (and Medellín) verifies school attendance (no less than 80 percent of scheduled classes attended during the period) via online reports by the schools. For children attending private school (10 percent of the cases), mothers must submit a certificate of school attendance at one of the five points of service. For the health component, the condition set by the program is that all registered children under the age of seven in a family must receive age-based growth and development checkups in accordance with the health ministry’s growth and development protocol. Verification of health conditionalities is slightly more challenging than for education. Health service provider institutions in Bogotá maintain an Individual Health Service Provider Registry. The DRB requests these records to verify the fulfillment of child conditions in 13 public hospitals. For children that visited private clinics, families must submit records directly at one of the five DRB points of service (submission can be delegated to “leader mothers,” as discussed below). Field data indicate that verification can be made for only about 24 percent of health beneficiaries in Bogotá. One complicating factor is that people frequently change between contributory, partially subsidized, and fully subsidized health systems—each of which involves a different service provider. In contrast, for the education component, 90 percent of the beneficiaries attend public schools, whose attendance records are contained in a centralized database maintained by the district education secretary. Institutional arrangements Since its inception, Familias has been led and executed by the Office of the President, through a unit that has operated under a different name and structure in each administration. In Bogotá, the program commenced in 2007: initially, the Bogotá government decided not to participate, but the national government pursued the program anyway in the capital. The Department for Social Prosperity created the coordinating unit for program operations in Bogotá, which is now known as the DRB. Bogotá’s institutional model differs from that of the rest of the country, where the program is coordinated by national and local governments under agreements signed between the Department for Social Prosperity and the respective municipal mayor’s office. According to interviews with officials, the lack of coordination with the mayor’s office is a key constraint of the program in Bogotá, and contributes to the program’s mixed performance in the city. Relatedly, low coverage in Bogotá could be the result of a competition between national and district- level social protection programs. For example, there are indications that community leaders may have conveyed the message that families participating in Familias were excluded from district programs, thereby forcing them to choose between one or the other. On the other hand, because Bogotá offers different interventions, Familias has not raised widespread expectations, and the general public does not seem to have recognized the national government for bringing the program to the city. This may have diminished the likelihood of political interference in the program. However, at the local level, 121   there have been instances of politically intended abuses which have been exposed online84. The DRB has five points of service located in areas with large numbers of beneficiary families. These locations provide guidance and services for beneficiaries, including verification of conditions, permanent registration of displaced persons, registration updates, and resolution of issues regarding the program. Service is available Monday to Friday from 7 a.m. to 7 p.m. These extended hours of operation were set to accommodate mothers whose time is limited and who are generally more available after 5 p.m. The 2011 evaluation findings and recommendations were taken into account in the most recent phase of the program in Bogotá, especially with regard to communication campaigns. For instance, in 2012 the DRB spearheaded a large-scale outreach process, including the mobilization of financial, administrative, and technical resources. Five registration sites were set up in venues for large events (for example, stadiums and parks), with police and health authorities ensuring security and registration, respectively. Program registration was open for 15 days for a potential pool of 198,000 families, with nearly 110,000 registering. The Familias model is used to promote the participation of mothers—especially through mothers’ groups, the election of mother leaders, and mother leader committees—and of other activities for community welfare, and networking groups. In Bogotá, these activities are offered on a smaller scale than in the rest of the country. For example, in 2013 some 108,000 families were invited to 35 meetings to elect 714 mother leaders. The attendance rate was 12 percent, and only 308 mother leaders were elected. This testifies to the low level of participation among mothers in Bogotá at community events. Lessons learned In Bogotá, the overall performance of Familias has been generally positive, although with milder impacts in the rest of the country and with specific challenges. Some of the reasons for these differences in effectiveness include possible competition with other social protection programs, and a program design not specifically tailored for the urban poor. A key challenge for Familias in Bogotá relates to the involvement of and coordination with the mayor’s office. Viewing Familias as a complement to district-level efforts would facilitate intersectoral synergies—for example, verification of conditionalities (especially for health)—as well as more effective deployment of district resources to programs that would complement Familias. The thorny issue of opportunity costs for urban beneficiaries needs to be more fully investigated and considered, including as a bottleneck for program performance in its current design and objectives (for example, low claims of direct cash transfers; low attendance of mothers at community welfare events). Interviews with policy makers and practitioners underscored the need for a more comprehensive diagnostic of the causes of urban poverty and how urban programs can be designed accordingly. In tandem with a renewed emphasis on understanding opportunity costs, such diagnostics could help identify more appropriate objectives and program features, including reexamining the relevance of current conditionalities in a large city like Bogotá.                                                              84 The program in Bogotá has a website (http://familiasBogotásur.jimdo.com/) where notices and warnings are posted and the community is invited to report misconduct.  122   4.4 Safety Net Programs in Nairobi’s Slums85 Rationale Nairobi has an estimated population of 3.5 million, which is nearly 40 percent of the country’s urban population. The city has been growing at a rate of about 3 percent a year, which represents the highest urban annual growth rate in Africa. As much as 75 percent of this growth has been absorbed by informal settlements or slums. There are over 200 slums in Nairobi, which occupy only 5 percent of the total residential land of the city but concentrate 60 percent of its population. The number of slum dwellers is expected to double over the next 15 years. Nairobi’s slums vary considerably in terms of size and area, but their population density can reach up to 2,309 persons per hectare. The slum population is highly mobile: nearly half (48 percent) emigrated directly from rural areas, while 43 percent comes from other settlements in Nairobi itself, and 7 percent from another urban location in Kenya. Only 1 percent of Nairobi’s slum dwellers were born in the settlement where they currently live. Some 92 percent of slum dwellers pay high prices for housing units that are mostly illegal, substandard in quality, and crowded. Barely 19 percent of slum households have access to piped water, while 68 percent rely on shared toilets. Less than 1 percent of households are served by a public waste collection system. The under-five mortality rate was 86 percent in the slums compared to 64 percent for Nairobi as a whole and 75 percent for urban Kenya on average in 2008. The unemployment rate among adult slum dwellers is 26 percent. Disaggregated by age and sex, this rate increases to 46 percent among youth (aged 15–24) and 49 percent among women. Unemployment among youth is considered a key factor behind the increasing levels of insecurity and violence in the slums: as much as 63 percent of slum households report not feeling safe in their own settlement, and 27 percent report having suffered a criminal episode over the previous 12 months. In 2009, the Kenya Food Security Steering Group estimated that of approximately 9.5 million food- insecure people in the country, 4.1 million live in urban areas. And 90 percent of the households surveyed in two Nairobi slums (Korogocho and Mukuru) had reduced their number of meals and diet diversity. The researchers also found that up to 30 percent of the children in these two slums had been taken out of school. Another study undertaken by Oxfam, Concern Worldwide, and Care recommended an immediate response be launched to address the food security crisis in the urban slums of Nairobi. Yet, the overall rural orientation of existing safety nets and the absence of an established urban analysis and response system with clear trigger indicators made it challenging to mount a systematic urban response86. In February 2009, the Kenyan cabinet recommended the creation of a cross-ministerial task force to design and pilot a food subsidy intervention for the urban food insecure. In early 2010, an agreement                                                              85 The discussion largely draws from Creti (2014a) produced for this review. 86 Safety nets in Kenya reached almost 13.7 percent of the population in 2010. The country’s two largest safety net programs are the Food Assistance Program and the School Feeding Program, which account for over half (53.2 percent) of the nation’s total expenditure on social assistance. In recent years, a range of other programs was introduced, such as the Older Persons Cash Transfer, Cash Transfers to Orphans and Vulnerable Children, Hunger Safety Net Programme, Urban Food Subsidy Cash Transfer, and Persons with Severe Disability Cash Transfer. These five cash transfer programs are being coordinated and harmonized around the framework created by the National Safety Net Programme launched in 2013. Other safety nets include health programs such as the HIV/AIDS nutrition feeding initiative, the Health Voucher Programme, and public work schemes implemented in rural and urban areas to support youth employment.  123   regarding implementation of the Urban Safety Net Programme in Kisumu, Mombasa, and Nairobi was signed. Within this framework, Oxfam and Concern Worldwide launched two programs in Nairobi’s slums: the Nairobi Urban Social Protection Programme in Mukuru and the Urban Livelihoods and Social Protection Programme in Korogocho. This case study reviews their experience. Beginning in November 2009, 2,781 households in Mukuru and 1,958 households in Korogocho received monthly cash transfers for over two years (until December 2011). In Korogocho, the first group of beneficiaries received cash transfers until the second quarter of 2012. The overall objective of the two programs was “to improve livelihood security of the most vulnerable in Nairobi’s informal settlements.” Their specific objective was “to increase access to food for the most vulnerable households in selected informal settlements and to develop longer-term food and income security initiatives.” The programs had three components: an unconditional cash transfer, livelihood support (skills training, cash for work, and business grants), and advocacy. The following analyzes the largest component of these two programs, their cash transfer component. Targeting Although the selected slums fell under the administrative districts with the highest concentration of poor people in Nairobi87, geographic targeting was not based on poverty prevalence and severity as provided by the latest census (1999) and the Kenya Integrated Household Budget Survey 2005–06. The decision to implement the safety net program in Mukuru and Korogocho was based on previous experiences of the implementing agencies in these slums. However, slum boundaries do not correspond to those of administrative districts, making it challenging to use survey-based data. Further, such data may rapidly become obsolete in light of the highly dynamic development of informal settlements and the high mobility of their populations. In Korogocho, Concern Worldwide intended to narrow the geographic targeting to the most vulnerable areas inside the slums. This effort would be based on criteria such as access to services, shelter, infrastructure, level of food security, and use of negative coping strategies. However, the community considered such an approach politically and socially unviable. As a consequence, the program targeted the entire slum. The initial selection of households in Korogocho and Mukuru was community based and followed a similar methodology. The process comprised five steps: (i) development of selection criteria in agreement with community stakeholders; (ii) identification of eligible households by a selection team (made up of community and local partner representatives) through household visits; (iii) review of preliminary lists of eligible beneficiaries in community meetings; (iv) verification visits and interviews with a random sample of 10 percent of households identified as eligible; and (v) Issuance of the final list of beneficiaries. The outreach process included public communication of the program and visits to households. Despite these efforts, it was difficult to ensure broad-based awareness about the programs across the slums, chiefly because of the high mobility and weak social connections among slum dwellers. This created some level of exclusion errors, as determined by the midterm review through focus group discussions with non-beneficiaries.                                                              87 The districts of Embakasi, Kasarani, and Makadara have the highest poverty concentrations in the city. Mukuru spreads across both Embakasi and Makadara, while Korogocho is within Kasarani. 124   The eligibility criteria that were selected and applied by the communities (Step 1 above) were different in the two slums. In Mukuru, the first eligibility criterion chosen by the community was low income (households earning less than $0.50 per day); in Korogocho, it was food insecurity (households eating one or fewer meals a day). In both Mukuru and Korogocho, households benefiting from other programs were not eligible. Eligible households also had to meet at least another vulnerability criterion such as being child or female headed, having a person living with HIV/AIDS, or taking care of more than three orphans/vulnerable children. There were a number of design issues with these criteria, especially the income component. The heavy reliance on the judgment of community and local partner representatives with minimal external verification posed significant risks for targeting effectiveness. In Korogocho, complaints about this process received through the project grievance mechanism88, led Concern Worldwide to rerun the targeting process using a census approach where teams were required to visit each house to collect data on 18 indicators (rather than select households)89. These indicators were combined using multiple correspondence analysis into a single index, with a cutoff of 59 out of 100 based on an initial assessment. This approach improved targeting outcomes, but was more expensive and took longer. Once selected, recipients went through a registration process. In line with most of the safety net programs, including those analyzed in the flagship Kenya Social Protection Sector Review, the program required beneficiaries to have a national identification (ID). This requirement posed a significant challenge because 5 percent of the preselected recipients were ineligible for a national ID, particularly child-headed households, refugees, and people from border areas who could not prove their Kenyan nationality at birth90. Benefit structure The value of the transfer was set at $12.50 per month per household, in alignment with other government cash transfer programs. The level was set to cover one-third of the household food basket (one full nutritious meal per day of ugali, maize, and kale). Yet in practice, the transfer size only met 20 percent of the average household’s basic food needs. On the one hand, cross-program alignment in transfer size ensured coherence, predictability, and transparency over the household entitlement. It also potentially facilitated the graduation of beneficiaries to other government programs. On the other hand, the nationally fixed value did not take into account the difference between the food poverty line in urban and rural areas: for instance, the Kenya Integrated Household Budget Survey 2005–06 calculated the food poverty line at $14.70 in urban areas and $9.90 in rural areas—a 32 percent higher line in urban areas. Also, larger households, which in Nairobi’s slums are more likely to be poor, received less benefits per capita because the transfer was adjusted to household size. In a context of rapidly rising prices, not adjusting the transfer value to inflation affected the most disadvantaged household, which devoted large shares of income for food and was highly dependent on markets for accessing it. In 2011, in response to rising food prices, the cash transfer value was increased to $20                                                              88 Concern Worldwide provided laminated cards with hotline numbers to beneficiaries so they could call with any complaints. The hotline numbers were also posted on notice boards so nonbeneficiaries could use them as well. 89 The indicators captured data on food consumption, assets, support from other organizations, household size, income and characteristics of head of household, shelter, access to health services, orphans, people with disabilities, and pregnant and lactating mothers. 90 Due to legal regulations that apply to financial institutions, people without identity cards cannot use the M-Pesa mobile money transfer system. To circumvent this requirement, Oxfam allowed recipients without ID cards, such as child-headed households, to nominate alternative recipients, and around 480 did so. Another 300 were unable to do so and were thus excluded. 125   in Korogocho. This adjustment reflected the changes in three other national safety net programs91. Benefits were paid through M-Pesa, the mobile money transfer system provided by Safaricom. In Mukuru, households registered their phone numbers with M-Pesa. About 40 percent of the households did not have phones, and were given SIM cards for use in phones held by community health workers. In Korogocho, enrolled households were issued SIM cards even if they already had their own phones. Those who did not have phones had to use borrowed phones. Processing of the transfer was made directly through a system developed by Safaricom. All transactions were done electronically, so human resource requirements were minimal and there were no security or fiduciary risks to the implementing organization and its staff. The technical support provided by Safaricom, either directly or through its network of agents, was critical to the success of the payment system. This support ranged from raising awareness on how to make transactions through the system, explaining security features within the system, and providing customer service. Transfer and withdrawal fees were 3.6 percent of the transfer value. These fees were covered entirely by the implementing agencies. Transfers via mobile phone technology proved to be efficient and cost-effective. From the provider’s side, there was good mobile phone connectivity and a huge network of agents from which recipients could easily access funds. From the beneficiary’s side, the mobile technology was familiar given the high visibility of M-Pesa in Kenyan urban areas. Payments made through M-Pesa were quick compared to other payment methods, resulting in reduced opportunity costs for beneficiaries to collect the transfers. Payment through M-Pesa was extremely well received by recipients. It was considered more discreet and secure than cash—a particular advantage given the insecure context of the slums. Recipients also appreciated the flexibility of mobile cash, which could be collected from a range of locations at different times, spent directly without cashing out, or sent directly to their family. The system also allowed savings, but almost no recipients seem to have taken advantage of this option, arguing that money is too scarce to save any. Studies on the use of cash sent to urban dwellers through remittances or safety net programs in Kenya reveal a tendency to immediately cash out the entire value. Challenges with the payment mechanism also existed. First, recipients had to have a national ID card, which excluded numerous households, as previously discussed. Second, households that had to borrow phones could not receive messages regarding the arrival of the money or any other message from program staff. There were anecdotes of extortion from phone owners. The midterm evaluation in Korogocho argued that buying phones for those households that do not own one would not be a significant cost to the program and would reap significant benefits. Studies in other urban African contexts found that mobile cash transfers are more cost-efficient than other traditional payment methods even when beneficiaries are provided with mobile phones, because the initial fixed costs are amortized after 8–10 months. Institutional arrangements Both programs in Mukuru and Korogocho included a first phase of cash transfer assistance followed by a second phase of livelihood support, including cash for work, vocational training, and business grants. Public work schemes and vocational training were considered a graduation strategy from the unconditional cash transfer program. Program and government officials regarded this progression as appropriate, particularly in light of the high level of unemployment among youth in the slums and the dynamic labor market. However, the process and criteria for beneficiary graduation from cash transfer                                                              91 Cash Transfers to Orphans and Vulnerable Children, Home-Grown School Meals, and the Hunger Safety Net Programme increased their transfer values to $20 in response to the 2011 price increase.  126   to livelihood activities were not well documented. An evaluation of the Mukuru program highlighted that the local implementation partner did not receive guidance on either graduation or exit from the program. As a consequence, decisions were left to the discretion of the partners and communities. In addition, the exclusion of 1,047 households in October 2010, based on an external assessment, was not well communicated and was considered arbitrary. Eventually, Concern Worldwide did not implement the cash-for-work component in Korogocho because it would have overlapped with existing government programs, specifically the Slum Upgrading Programme and Kazi Kwa Vijana. The agency also did not plan a graduation strategy toward these government programs. Oxfam, on the other hand, implemented cash-for-work activities in Mukuru, but program evaluations documented problems related to self-targeting as an entry mechanism. These problems were mostly related to the high value of the benefit (up to $3.00 per day as compared to the $2.50 per day given in Kazi Kwa Vijana). The implementing agencies worked closely with the Ministry of Gender, Children and Social Development and the Prime Minister’s Office, both of which served on the task force created in February 2009. This coordination was meant to ensure consistency with existing national programs and to serve as a demonstration pilot for future government cash transfer approaches in slums. Due to delays in government funding, however, the agencies developed their own programs in the respective slums. Implementation involved a wide range of stakeholders. Partnerships established with existing civil society organizations facilitated the entry of the agencies into the slums and increased accountability. Oxfam partnered with a local nongovernmental organization, Mukuru Slum Development Projects; Concern Worldwide partnered with the Redeemed Gospel Church Development Programme. Local authorities and community members were actively involved in the identification and registration of beneficiaries. The government of Kenya and the World Bank participated in the program’s technical and advocacy aspects. The agencies liaised with both the United Nations (UN) Office for the Coordination of Humanitarian Affairs and UN-Habitat on advocacy work, particularly through the Urban Vulnerability Forum led by these UN organizations92. Monitoring and evaluation Concern Worldwide collected a wide range of livelihood data, such as household area of origin, duration of stay in the slums, reasons for migrating, food frequency and diversity by age group, household expenditures, savings and assets, experiences with the M-Pesa delivery mechanism, and how the received cash transfer was used93. Oxfam interviewed a small sample of recipients each month asking questions about food security,                                                              92 The forum works with the government, the UN, and nongovernmental and community-based organizations in informal settlements to develop assessment tools and response strategies appropriate for humanitarian needs in urban informal settlements in Kenya. From October 2009 to December 2011, the forum facilitated three major consultations with over 50 agencies working to establish an appropriate multistakeholder partnership for urban emergencies. It has also supported the development of monitoring and assessment tools for emergencies in urban settings and enhanced preparedness among national and local authorities in urban areas. 93 Concern Worldwide conducted a quantitative baseline survey in October 2009, and a follow-up in October 2010, to a panel sample of 156 beneficiaries. In parallel, qualitative fieldwork consisting of focus group discussions and in-depth interviews was conducted in October 2010. Further quantitative and qualitative data were collected for the midterm review in 2012. 127   water, soap, school fees, rent, clothes, medicines, transport, gifts, and loan repayments. The perspectives of different community members—including beneficiaries and nonbeneficiaries, village elders, social workers, and community health workers—were gathered during the review process. The program was the subject of different evaluation efforts, which resulted in at least four published and widely disseminated reports. In Korogocho, beneficiary households’ average number of meals per day increased from 1.61 to 2.53 over a period of 11 months. Household food insecurity, measured through the Household Food Insecurity and Access Scale, decreased from 97 percent at the beginning of the program to 73 percent at its end. While calorific intake and dietary diversity improved (for example, household diet diversity increased from 3.4 at baseline to 7.1 10 months later)94, others factors affecting malnutrition (hygiene, sanitation, and care) remained unsatisfactory due to the poor living conditions in the slums. An analysis of coping strategies adopted by the target population showed that, after the cash transfers, beneficiaries had incurred less debt, resorted less to removing children from school, and reduced engaging in transactional sex. There is no evidence of cash transfers acting as disincentives to work. For instance, the transfers had a significant impact on individuals on anti-retrovirals. An effective anti-retroviral treatment requires a regular and adequate diet, which was not affordable to many HIV- positive beneficiaries before the program. The cash transfers allowed them to have more regular and adequate food consumption and to resume their treatment. As a result, many were able to go back to work. The regularity of the transfers was an incentive for half of the beneficiaries to participate in informal savings groups whose members make regular contributions and receive a one-off payout in turn95. In general, community relations did not deteriorate as a result of the transfers. In some cases they actually improved, as recipients could lend to nonrecipients. This situation had a positive impact on informal sharing systems, in which households are supposed to pay back the support received in time of need. The cash transfers restored the payback capacity of previously indebted households, reinserting them into these informal mutual-help systems. The midterm evaluation in Korogocho undertook a limited study of the likely economic multiplier of the cash transfer and, through two different methods, placed this at slightly over 2—meaning that for every $1.00 of cash delivered, there would be an increase of income in Korogocho of over $2.00. Qualitative analysis, however, shows that most of the transfer value (60 percent) was spent on the purchase of food and nonfood items from local traders, which probably captured most of the secondary benefits of the program. The evaluation observed that, for most recipients, gains from safety nets would not be sustained if the cash transfers stopped, since there had been no noticeable easing of the economic situation and the transfers were not sufficient to generate a step-change in livelihoods.                                                              94 Households broadened the number of food items consumed, but did not necessarily diversify the food groups consumed—they tended to vary the food items within the same food group (for example, rice instead of maize). The midterm evaluation in Korogocho found that complementary services such as education, supplementation, messaging through mobile phones, or soft conditionality on clinic attendance would improve the nutritional impact of cash transfers in urban contexts.  95 In these “merry-go-round” informal savings groups, members make regular contributions, receive a one-off payout in turn, and may have access to interest-free loans. This form of saving is not cumulative—members do not make net gains in cash terms. Instead, they gain security for their money and impose self-discipline, since they cannot spend the money until their turn. However, only 24 percent of respondents had been able to save money after the transfers stopped. 128   Lessons learned With a highly heterogeneous and mobile population and loose administrative boundaries for informal settings, data for slums can rapidly become outdated or of limited relevance. For instance, the experience in Nairobi raised an important question as to what level of geographic disaggregation is necessary and appropriate for urban targeting. The use of administrative boundaries presents the risk of aggregating different slums and masking changes at the slum level. One alternative would be to identify, within larger urban areas, noncontiguous zones with similar levels of vulnerability. Some research suggested dividing the city into areas with similar characteristics in terms of population density, service provision (both government and commercial), and infrastructure access (transport, communications, housing, etc.). The government is currently implementing the National Social Protection Policy that was approved by the cabinet in 2012. Design of the national categorical cash transfer programs should consider urban poverty and vulnerabilities more deliberately96. Relatedly, decentralization associated with the reinforcement of local governance and community participation may improve the understanding of urban poverty challenges, and offer an opportunity for national safety net programs to reach and better meet the needs of the poorest urban communities and slums. Strengthening local governments would require the establishment of decentralized coordination platforms and technical capacity building. Similarly, it is critical to coordinate and align targeting across existing safety net programs. While many of these programs tend to use proxy means testing, it would be desirable to use the same targeting approach. Two main challenges exist in this regard. First, the urban poor tend to move in and out of poverty, including during the lifespan of a program. This requires retargeting on a regular basis, which can be costly and time-consuming. Second, proxy means testing requires resources and time, which make it less suitable to rapid-onset crises. Applying community-based targeting in poor urban contexts has proved to be very challenging as well. Given the densely populated and fluid nature of urban areas, communities are hard to define and members may lack power, confidence, and knowledge of their neighbors. The experience with M-Pesa shows that urban contexts offer a highly conducive environment for e- payments, including ample network connectivity, large coverage of mobile phone agents, and familiarity with the technology. Mobile phones can also improve social networks and urban-rural linkages as they allow money to be sent and received. Technology potentialities could be exploited to improve registration and identification processes and information management. The Nairobi case study highlights the challenges that agencies can face in graduating beneficiaries. Linkages and alignment with existing government programs need to be built into the initial design, and graduation of beneficiaries requires clear and measurable indicators and communication strategies. Links to other interventions that fall outside the remit of social protection, such as slum upgrading in urban development programs, would also need to be closely examined. In terms of crisis response, existing frameworks use data collection and analysis tools (such as the Household Economy Approach and the Integrated Food Security Phase Classification) that have not been tested in or adapted to urban contexts. Further work should involve the identification of thresholds and the development of decision-making frameworks to define an urban crisis. The current analytical gap makes it difficult to establish consensus as to when an urban area moves from a chronic                                                              96 These national categorical cash transfer programs are the Older Persons Cash Transfer, Cash Transfers to Orphans and Vulnerable Children, the Hunger Safety Net Programme, and the Persons with Severe Disability Cash Transfer.  129   to an acute crisis, and to respond or scale up/down existing safety nets in those contexts. 4.5 The Programa de Apoyo Temporal al Ingreso (PATI) in urban El Salvador97 Rationale The recent food, fuel, and financial crises had significant impacts on El Salvador’s socioeconomic condition. The economic growth rate declined to about −3 percent in 2009. Some key sources of economic growth such as maquila exports declined by about 23 percent in the same year. Unemployment rose dramatically, while remittances—which represented nearly 18 percent of gross domestic product (GDP) in 2005–08—registered a sharp contraction, as did fiscal revenues, which dropped by around 11 percent in 2009 alone. The crisis particularly affected urban areas, exacerbating an ongoing urbanization of poverty process. By 2008, 58 percent of the poor lived in urban areas; in contrast, in 2002, the number of poor people in rural areas was twice that of the urban poor. In response, the newly elected government unveiled a comprehensive Anti-Crisis Plan in 2009. The plan included investments in a comprehensive social protection system, the Sistema de Protección Social Universal (SPSU), which comprises both contributory and noncontributory components. On the noncontributory side, the government established the Comunidades Solidarias program by executive decree in October 200998. The program had two main pillars aimed respectively at vulnerable rural and urban populations: Comunidades Solidarias Rurales, which absorbed an earlier conditional cash transfer program; and Comunidades Solidarias Urbanas (CSU), a new set of interventions targeted to the urban poor. The government backed up these programs with resource commitments: between 2008 and 2013, spending on social protection increased fivefold, although still representing less than 1 percent of GDP. CSU represented the first social assistance program tailored to urban areas. Within this context, the Temporary Income Support Program (Programa de Apoyo Temporal al Ingreso, PATI) was conceived and introduced as a CSU flagship program, implemented by the Social Investment Fund for Local Development (Fondo de Inversión Social para el Desarrollo Local de El Salvador, FISDL). PATI has two core objectives: (1) to provide short-term income support to poor and vulnerable individuals in urban areas, and (2) to increase beneficiaries’ employability. Although PATI comprises three modalities, this case study focuses on the experience and features of the urban PATI modality only99. Targeting                                                              97 The discussion largely draws from Rodriguez-Alas et al. (2014) produced for this review.  98 Comunidades Solidarias (both rural and urban) has four program components addressing, respectively, human capital, basic social infrastructure, income generation, and institutional management. The latter is a new component that seeks to strengthen local government capacities for program management and oversight, as well as promote community participation. 99 PATI’s three modalities are PATI Ida, PATI Productivo, and PATI Urbano. The target population for PATI Ida included those rural areas most affected by Hurricane Ida (2009). The intervention operated with a projected budget of $23.5 million until the end of 2013, and targeted some 28,000 participants. PATI Productivo was introduced in 2011; it reached 3,640 participants in 2013 with a budget of almost $3 million. Its objective is to strengthen the labor orientation of participants exiting PATI. Its approach includes an expanded training component, technical assistance, and grants for starting up small businesses. 130   PATI Urbano (for simplicity hereafter, PATI) was first pilot tested in November 2009 in two municipalities, San Martín and Ahuachapán, covering 303 and 338 participants, respectively. The pilot phase lasted six months and was financed with $430,000 through central and local government resources100. Following the pilot, the program was rolled out in March 2010, with a total budget of $37.1 million (or 0.26 percent of GDP). By 2014, PATI reached 30,473 beneficiaries in 25 municipalities with a high incidence of poor and marginalized—or precarious—urban settlements (asentamientos urbanos precarios, AUPs). The concept of AUPs is central to the targeting process and is detailed below. But it is important to first understand how poor urban areas are identified by the government—a process that serves as an entry point for PATI and other programs, such as Bonos de Educación (targeted to students living in extremely precarious AUPs) and Pensión Básica Universal (old-age pension). This section reviews, step by step, the government urban poverty diagnostic process and discusses how PATI’s process built on that pre-existing platform. Underlying the process is the Urban Poverty and Social Exclusion Map, a rigorous statistical and geospatial effort led by the Ministry of Economy. Using this map, AUPs are identified through the following five steps: (i) precarious households are identified using the Unsatisfied Basic Needs method based on housing indicators; (ii) precarious blocks are identified on the basis of whether more than half of their households are themselves precarious; (iii) AUPs are identified by grouping neighboring precarious blocks with a minimum of 50 precarious households; (iv) AUPs are divided into four levels of precariousness (low, moderate, high, extreme) based on a cluster analysis conducted on results from ranking an index of residential marginality and an index of social exclusion. As a result of this process, some 2,508 AUPs were identified in urban El Salvador. Approximately 19 percent of AUPs show a level of extreme precariousness, and 32 percent a high level (figure 37); and (v) AUPs are ranked within each cluster using an index of socioeconomic stratification. Figure 37. Clustering of AUPs by level of precariousness Extreme High Moderate IMARES index  Low IEXSC index  The geographic targeting of areas for PATI also involves several steps and is based on information provided by the map. First, the program selects municipalities with (1) a high incidence of AUPs with                                                              100 Of this $430,000, about $30,000 was contributed by the municipalities 131   extreme and high levels of precariousness, and (2) high levels of violence according to the Registry of Violence of the Ministry of Justice and Public Safety. For the latter, a normalized index was created including crime incidence for every 100,000 people (homicides, kidnapping, theft, rape, etc.), weighted by the level of crime (with homicides given the most weight)101. These maps are overlaid to identify possible PATI municipalities. Within the selected municipalities, the program chooses AUPs with extreme and high levels of precariousness as defined by the map. The number of participants is set in proportion to the total population of the municipality and based on the index of socioeconomic stratification estimated by the map. Once the selected AUPs have been identified, PATI officials apply several categorical and administrative criteria to determine individual eligibility. These include (1) living for a minimum of six months in the selected AUP, (2) being at least 16 years of age, (3) lacking employment in the formal sector, and (4) not studying or receiving training during working days and hours. Some self-selection also takes place, in that the transfer level is set lower than the minimum wage for unskilled workers in urban areas. Potential candidates preregister for the program; this is followed by home visits by the FISDL, the municipal government, and community leaders. In these visits, data are collected on demographics, income, education, employment, health, housing, utilities, and asset conditions. A prioritization exercise is conducted, and applicants are ranked by score. Participants are then selected based on their ranking, with the actual number of beneficiaries determined by quotas assigned to the AUPs. Finally, community leaders and municipal committees validate the list of selected persons through on- site visits, and ensure that the participants reflect the ranking and prioritization of the established criteria. In general, no more than 5 percent of the preselected individuals are replaced during this verification. Evidence shows that PATI’s targeting has been effective. About 72 percent of PATI beneficiaries belong to the two poorest income quintiles, while 46 percent lives in extreme poverty (compared to 32 percent among non-PATI households in the same municipalities). The majority (63 percent) of PATI beneficiaries are women; the average formal schooling for beneficiaries was 6.2 years. It proved challenging to reach the youngest cohort (ages 16–24), as shown by their share of total participation (39 percent) and the average age of participants (28 years). Benefit structure The program provides $100 per month (29 percent of beneficiaries’ monthly income), which is, as noted, lower than the minimum wage for unskilled workers in urban areas ($175 per six hours). Individuals can only participate in PATI in a single six-month cycle. The transfer is based on two conditions: (1) participation in light work and activities and (2) attendance of job skills training. Work activities involve between 30 and 50 participants who work six hours a day, five days a week, for six months. These activities generally range from maintenance of light community infrastructure (tending small sidewalks, sign painting, trench digging, slope protection, etc.) to less intensive work such as cleaning public sites, painting school walls, or handicrafts. Estimates indicate that the share of transfers out of total costs ranges from 60 to 76 percent, making PATI a safety net–oriented public works program.                                                              101 This procedure was recently simplified and is now only based on homicide and imprisonment rates. 132   PATI’s employability component includes a total of 64 hours of training provided by El Salvador’s Professional Training Institute (Instituto Salvadoreño de Formación Profesional, INSAFORP). This training is complemented by 16 hours of training in job seeking and self-employment provided by the Ministry of Labor. INSAFORP’s training is informed by three assessments: (1) diagnosis of business and employment opportunities, (2) potential supply of employment from businesses, and (3) training needs of beneficiaries. Based on this information, INSAFORP has elaborated a training plan that attempts to match participants’ demands for activities with the jobs required at the municipal level and the demand for particular occupational profiles by local enterprises. This matching initially represented a challenge for INSAFORP in terms of meeting the surge in training demands created by PATI, and adapting its curriculum to work with people with heterogeneous educational levels— including those without literacy. INSAFORP accordingly increased its internal capacity. It also developed a comprehensive monitoring system to keep track of the process and the number of people trained102. Another challenge stems from a basic mismatch between the labor market and participants’ demand for training. There is weak correlation between the type and profile of opportunities available and participants’ expressed demand: for example, bakery work features at the top of the participants’ list of requested skills training (figure 38), but at the bottom of the market’s needs. Figure 38. Trainings demanded by PATI participants, 2013 Bakery and pastry Cooking Electricity, car maintenance Cosmetology Dressmaking Handicrafts Food processing Auto mechanics Natural cleaning products Jewelry 0 5 10 15 20 25 30 35 Applications (%)   As mentioned, the Ministry of Labor focuses on implementing 16 hours of training in basic life skills, job-seeking techniques, and self-employment orientation. The training is delivered in advance of INSAFORP’s technical courses, and is conducted by counselors and psychologists with a labor orientation. The Ministry of Labor also offers intermediation services through job promoters in its offices located in each department (province) as well as other strategic municipalities. The job promoters use the Employment Intermediation System to match job seekers with employment opportunities; PATI participants are automatically registered in this system.                                                              102 One of the results of the PATI pilot was the revision of the disbursement mechanism from the FISDL to INSAFORP, making it an output-based disbursement, instead of being based on the number of participants trained. 133   Institutional arrangements Institutional coordination for PATI occurs at three levels: (1) national strategic and technical coordination, managed by the SPSU Intersectoral Technical Committee and chaired by the Technical Secretariat of the Presidency; (2) inter-institutional coordination, involving the SPSU Intersectorial Committee, the CSU Intersectoral Committee, and the PATI Technical Roundtable comprised of the FISDL, INSAFORP, the Ministry of Labor, the National Commission for Small and Medium Enterprises, and the Salvadoran Institute for the Advancement of Women; and (3) local municipal coordination, involving CSU municipal coordination committees and PATI municipal commissions. Once an AUP has been confirmed for the program, municipalities convene with local community leaders, and AUP representatives are selected by the PATI municipal commission. Subsequently, communities prepare PATI project proposals for submission to the FISDL. These proposals comprise a description of community and training activities, the expected number of participants, projects’ expected duration, and the co-participation (for example, materials and tools) required from the municipality. At this point, the FISDL, the municipalities, and INSAFORP assemble for the first meeting of the PATI commission to approve projects. The PATI commission, the PATI municipal coordinator, INSAFORP, and the FISDL meet periodically and regularly. The FISDL meets with PATI coordinators every four months to share information and coordinate actions; the FISDL president, mayors, and technical teams meet twice a year; and individual teams at the local level meet periodically. PATI deploys both a national and a local communication strategy. In the latter case, each municipality develops a plan to disseminate program information according to its resources and available media (community radio, newspapers, churches, etc.). The FISDL supports dissemination of information nationally with materials that explain program eligibility, the selected AUPs, and participant benefits and responsibilities. Initiatives such as “PATI-mania” fairs have raised awareness about the program’s purpose and results. Results A rigorous impact evaluation on a range of poverty and employability dimensions was recently finalized. The main results suggest that over the short term (immediately after particpants’ exiting the program), PATI increases labor market participation by an average of 4.6 percent (figure 39). The program’s impact is greater over the medium term (one year after exiting the program), with increased labor market participation of between 5.5 and 8.6 percent depending on the comparison group. Analyzing the data by subgroup shows that increased participation in the labor market occurred only among women, with a 5.9 percent increase in participation in the short term, and between 6.6 and 12.3 percent in the medium term. A similar effect is seen for youth—but not for adults—with a 9.0 percent increase in the short term and between 8.4 and 16.5 percent in the medium term. Finally, the impact is only experienced by individuals with a minimum of nine years of schooling. The program has a positive impact on probability of working, but only over the medium term.   134   Figure 39. PATI’s impact on labor participation a. Economically Active b. Working (salaried or self-employed) Source: Beneke de Sanfeliu and Acosta (2014) With respect to participants’ perception of readiness to seek a job, the evaluation shows that this perception increased over the short term by 6.6 percent, especially among women (8.4 percent), adults (9.2 percent), and those with a lower level of education (7.6 percent). However, there is no impact in the medium term. In contrast, in terms of perception of readiness to start a business, youth (10.3 percent) and those with more schooling (9.6 percent) showed the most gains. Here again, the impact is only in the short term; in the medium term, the impact on young people is negative (−7.7 percent), meaning that the initial gain disappeared for this group. In terms of income, in the medium term, participants earned on average $17.90 more per month, which is a 22.0 percent increase over the end of the program. This effect is concentrated among men (between $56.70 and $96.80). Similarly, one year after leaving the program, extreme poverty among participants was reduced by 9.6 percent. The fact that results on poverty, labor, and other qualitative indicators (such as self-esteem) were not significant in the medium term may indicate that the training was not long enough or not necessarily oriented to the most relevant local job opportunities. It may also indicate that local conditions outside the program’s scope could have played a role, as reflected in participants’ negative perception of the economic situation after they left the program. The fact that the program does not offer a systematic exit to a more permanent job or opportunities to access credit for microbusiness can also have a negative impact on participants’ perception of the possibility for sustained escape from poverty and enhanced quality of life in the long run. The impact on social relations—as measured by participation in community organizations—seems to have continued over the medium term. Although the initial level of participation was very low (8.7 percent), participation in the program increased by 3.7 percentage points. The program, however, had no sizable effect on perceptions of safety. This finding is not unexpected, given that PATI was not directly designed for that purpose, although it operated in AUPs with high levels of violence. Another significant finding was that participants, although in a precarious context, seem to be more satisfied regarding their communities’ public areas, many of which were improved by PATI. This perception was especially prevalent among the people who will probably make most use of these areas—youth and adults. However, this increase in general satisfaction with the neighborhood occurred only in the short term, suggesting the need to strengthen continuation of the work so that the benefit lasts. 135   Lessons learned PATI was introduced in El Salvador as a crisis response initiative. It has deliberately targeted the poorest people living in the riskiest and most violent urban area environments. There is a general consensus that PATI is an effective safety net given its ability to provide time-bound support to participants. Its targeting performance and results in terms of poverty mitigation and employment have been significant in the short run, but less so in the medium term. Over the years, however, the entry point provided by an acute economic crisis and natural disaster has been narrowing, exposing the structural challenges that define the context within which PATI operates. These challenges include chronic levels of unemployment, basic or absent levels of education, and widespread crime and illegal activities. While PATI was meant to be a crisis response mechanism, the program has evolved into a suite of interventions in disadvantaged urban contexts (with an income incentive attached) beyond the provision of a temporary income support program. PATI indeed appears to have carried out its objective of providing temporary protection to the income of the poor and vulnerable populations in urban areas by increasing work income—especially for youth, men, and those with more education. Long-lasting results require complementary interventions. As an intermediate option, policy makers might consider unbundling programs like PATI and streamlining support according to different profiles and objectives. In particular, the definition of PATI’s future policy direction could be particularly powerful if framed in conjunction with, or in the context of, an urban strategy combining elements of urban development and social protection—making PATI a more dynamic safety net in complex and rapidly changing urban environments. In this way, PATI and similar programs could evolve from deploying a standard approach to presenting a more diversified, multipronged operational strategy. Similarly, there needs to be more forceful (perhaps fiscal) incentives to ensure that the private sector engages more in high-poverty and violence-stricken areas. Promoting concepts of social responsibility and sensitization are important advocacy initiatives in a number of municipalities, but these alone may not be sufficiently powerful forces of change in highly resource-scarce environments. The agenda for countries like El Salvador, therefore, may include revisiting some of the institutional bottlenecks that perpetuate the vicious cycle of low demand–low supply of higher-skill labor. The level of municipal capacities, readiness, and entrepreneurship varies significantly. Municipalities with stronger leadership and management capacity take advantage of the program to strengthen local capacities, establish links to other initiatives, and generate synergies. The establishment of closer linkages between municipalities and communities tends to enhance the former’s understanding of relevant issues at the grassroots level, as well as to empower and give voice to communities to better connect them to bureaucracies. Several of these aspects were not only recognized by the executing agencies during PATI’s implementation, but were also put into practice. For instance, the program increasingly provided more employment counseling, workplace linkages, seed capital, and mentoring, especially for youth. Program officials established connections with municipalities and other sectors (education, finance, and agriculture) to offer job alternatives for former PATI participants in supplementary programs. This circumstance has highlighted the need for specific attention to be paid to that population at risk of violence; it is expected that the program could be adapted in the future to respond more comprehensively to that population. 136   4.6 Family Rewards: A U.S. Experimental Conditional Cash Transfer Program103 Rationale In 2007, inspired by Mexico’s Oportunidades (now Prospera) program, New York City launched an experiment to test a conditional cash transfer (CCT) program, called Opportunity NYC–Family Rewards, in six high-poverty urban areas. Family Rewards was the first comprehensive CCT program in a higher-income country. It tied cash rewards to the completion of particular activities and outcomes in three domains: children’s education, family preventive health care, and parent employment. The program was available to about 2,400 families and has been the focus of a careful ongoing random assignment evaluation. Eligible families that volunteered for the program were assigned at random to either a group that received the Family Rewards intervention or to a control group that did not. In 2010, the operational phase of this pilot program concluded—as scheduled— after a planned three-year run, although the evaluation is continuing. It is important to note at the outset that poverty in New York City is vastly different from poverty in Mexico and other countries in which CCT programs operate. Moreover, unlike those other locations, New York City has a well-developed social safety net made up of a variety of programs and policies that offer cash income as well as subsidy payments for food, housing, and medical care. Consequently, Family Rewards is best viewed as supplementing a substantial system of social protection. It offered extra resources to families that continued to have very low income despite the other assistance they received. The early evaluation results showed that the pilot program produced some initial promising effects on poverty reduction and on a number of human capital outcomes, offering a reason to continue experimenting with this approach. At the same time, features of the model that did not work as well pointed to a number of ways in which the Family Rewards approach could be strengthened. Building on the early evidence from the first trial, the model was revised considerably. In 2011, a new version of Family Rewards was launched in two locations: the Bronx, New York, and Memphis, Tennessee. That model is being tested through a new randomized trial. The original and revised versions of the model are referred to here, respectively, as Family Rewards 1.0 and Family Rewards 2.0. The origins of Family Rewards 1.0 date back to 2006, when New York City officials began to explore bold new ways of using financial incentives to address some of the root causes of poverty, particularly poverty that continued from one generation of a family to the next. The city’s Center for Economic Opportunity (CEO), a unit within the Office of the Mayor, initiated the program after learning about successful efforts with CCT programs in Latin America and a growing number of lower- and middle- income countries throughout the world. The CEO was inspired by the basic principle of CCTs: structuring cash transfers so as to promote human capital development for all family members, while simultaneously alleviating immediate poverty and hardship. The two-generation focus was particularly appealing, and the initiators wondered whether some form of a CCT program could work in their own city. They began to explore the idea of a trial project. The CEO engaged MDRC, a not- for-profit, nonpartisan social policy and education research organization headquartered in New York City, to head up a design process that ultimately led to the launch of the Family Rewards                                                              103  This case study was produced by James Riccio and draws on information presented in earlier published research reports by MDRC, with contributions from Cynthia Miller, Nadine Dechaussay, David Greenberg, Stephen Nunez, Zawadi Rucks, Nandita Verma, and Edith Yang.  137   demonstration project104. Targeting The Family Rewards 1.0 model was tested in six community districts—two each from the Bronx, Brooklyn, and Manhattan. These areas were chosen because they are among New York’s most persistently disadvantaged communities, suffering high rates of poverty and unemployment even when economic conditions in the city as a whole were good. For example, in 2006, the official poverty rate in the city was 18 percent. In the six Family Rewards community districts, the official poverty rate averaged 35 percent; it approached or exceeded 40 percent (a level many experts define as extreme poverty) in three of those communities. The unemployment rate across the districts was also disproportionately high, averaging 19 percent, compared with 5 percent citywide. The proportion of residents 25 years of age or older without a high school diploma averaged 43 percent across the six districts, compared with 28 percent for the city as a whole. Family Rewards was targeted toward families in these districts with incomes at or below 130 percent of the federal poverty level. This is the same eligibility standard used for food stamps, free lunches under the National School Lunch Program, and a number of other benefit programs that serve very low-income families, making it a widely accepted benchmark for identifying families in need of government cash transfer programs. Rather than create an entirely new income determination process, the designers of Family Rewards decided to use receipt of free school lunches (according to school records) as the income eligibility indicator for the program. Eligible families had to have at least one child in the fourth, seventh, or ninth grade. These grades were selected because they are at or near the start of critical transition points in education. Once a family volunteered for the study, all children in the family who were school age or younger were eligible for the program. All parents and children had to be legal residents of the United States. A majority of the families (81 percent) that enrolled in Family Rewards were one-parent families at the time of random assignment. About 47 percent of the families were Hispanic/Latino; most of the remainder (51 percent) were black non-Hispanic/Latino. Just over half of the parents (53 percent) were employed, with about 37 percent working full time, but at low-wage jobs. About a third (32 percent) had only a high school diploma or General Educational Development (GED) certificate, and about 18 percent had an associate’s or bachelor’s degree; 50 percent had not completed high school and did not have a GED certificate. In the first year of the program, the procedures for engaging families, educating them about rewards, verifying claims, and making payments were being refined, concurrent with some sample members being enrolled into the study. Not surprisingly, some aspects of program delivery suffered as a result, especially in initial efforts to orient and explain the complex set of rewards to families. However, as the program matured and staff gained more experience in operating it, many of these early challenges                                                              104 To support and assist that exploration, the CEO entered into a partnership with the Rockefeller Foundation, which envisioned that the project could provide a new opportunity to help low-income New Yorkers while building evidence for a poverty reduction strategy that would have national and international importance. With a special grant from the Rockefeller Foundation, the CEO asked MDRC to help design the project, including a rigorous evaluation. Subsequently, MDRC engaged Seedco in the planning process, and the three organizations worked closely together to come up with a plan for a CCT demonstration project. During this planning process, they conferred with officials and researchers involved with the Prospera program, meeting with them in New York City and visiting their program in Mexico. The design team also sought guidance and feedback from the World Bank; the Inter-American Development Bank; and experts in universities, foundations, other social policy organizations, and various New York City government agencies. 138   were overcome. By the second year of implementation, program operations were much improved; and the model was being operated in a way that was generally consistent with its designers’ vision. Benefit structure New York City’s original program included an extensive set of rewards with the following conditions: (1) education-focused conditions, which included meeting goals for children’s attendance in school, achievement levels on standardized tests, and other school progress markers, as well as parents’ engagement with their children’s education; (2) health-focused conditions, which included maintaining health insurance coverage for parents and children, as well as obtaining age-appropriate preventive medical and dental checkups for each family member; and (3) workforce-focused conditions aimed at parents, which included sustaining full-time work and participation in approved education or job training activities. The program offered a set of 22 different incentives during its first two years, ranging in value from $20 to $600. By rewarding a wide range of activities, the program gave families many different ways in which to earn money, and it was able to avoid attaching overly large amounts of money to any one activity or outcome. Based on assessments of the program’s early operational experiences—including the complexity of administering so many different rewards—along with preliminary impact evidence, a number of rewards were discontinued for the third year. These changes were made to simplify the program, lower its costs, and make it easier to replicate should it prove to be successful105.2 The program allowed families to receive cash rewards totaling several thousand dollars per year over a three-year period. The actual amount families received depended on the particular rewards they earned (some carried higher payments than others) and the number of rewards they earned. In addition, larger families could earn higher payments because each child’s actions could earn education and health rewards. In general, payments were made directly to the parents. However, some education-related payments for high school students were paid directly to the students; depending on the reward, either the entire payment was made to the student (for example, for passing a Regents exam) or split with the parents (for example, for meeting the attendance standard). As noted above, the Family Rewards model differs in important ways from CCT approaches in other countries. In many countries, CCT programs function as the main government-sponsored safety net, or as an important component of it, and they most commonly tie payments only to children’s school enrollment and attendance and to routine health checkups. Family Rewards includes many more conditions and rewards. In the education domain, it is unusual in rewarding children’s school achievement, including test score results, not just school enrollment and attendance. Its work-related component for parents is also distinctive. As a short-term intervention layered on top of an already well-developed social safety net, Family Rewards serves as a supplemental program rather than a core welfare system, as in Mexico and elsewhere. Family Rewards is also unusual in being operated by private, nonprofit agencies rather than by the government. Seedco, a nonprofit community and economic development agency, served as the main implementing agency. It assembled a network of local organizations in the designated community districts to assist in implementing Family Rewards. Called “neighborhood partner organizations”, these nonprofit                                                              105 The discontinued incentives included the attendance reward for elementary and middle school students, rewards to parents for discussing their children’s annual English language arts and math test results with teachers, rewards for obtaining library cards, all health insurance rewards, and rewards for making doctor-recommended follow-up visits. 139   community agencies recruited and enrolled eligible families into the research sample and served as the face of the program in the communities106.3 They provided ongoing customer service to participants who requested assistance, such as in making claims for rewards or for information about other services in the community. Neighborhood partner organizations also conducted informational workshops on how to earn and claim rewards in each of the domains in which incentives were offered. Seedco maintained a telephone helpline and website to provide additional information to families. Once Seedco verified that families earned rewards (which it did using a combination of automated data from city agencies and special coupon book forms submitted directly by participants), it initiated a process of transferring payments electronically into participants’ newly opened or existing bank accounts—or, if they preferred, onto stored value cards (prepaid cards, like gift cards or prepaid phone cards, that are not connected to any individual account holder). To provide families with a safe banking option, New York City officials worked with several banks and credit unions to develop special Opportunity NYC accounts that carried no fees and came with debit cards that were impossible to overdraw. The reward payments were made every two months, and families could access the money at any time through any automated teller machine (ATM). Envisioned as an “incentives-only” intervention, the program model did not provide social services or case management. It made no provision for staff to work with families to develop personalized action plans for pursuing education, health care, or employment goals; and staff members did not provide ongoing counseling to families to address personal problems that made it difficult for them to take full advantage of the program. The program also did not provide any direct services such as tutoring, test preparation, job search classes, or skills training. However, it did include an information and referral component, whereby the implementing agencies (Seedco and the neighborhood partner organizations) referred families (upon request) to other agencies in the community that provided relevant services. Results Overall, Family Rewards made payments to virtually all participating families. It transferred substantial amounts of cash—over $8,700 per family, on average, over the three-year period, with many families receiving considerably more. Reward amounts averaged over $3,100 during each of the first two years and $2,700 in the third year (when several rewards were discontinued). A majority of families—approximately 57 percent—earned at least $7,000 over the life of the program. The top 20 percent earned more than $13,000 in reward money107.4 The families among the top 20 percent of earners tended to be larger (giving them more opportunities to earn rewards) and less disadvantaged. For example, the parents were more educated, more likely to be employed, and more likely to be married; and the families were less likely to be receiving government transfer benefits. In-depth interviews suggest that parents who were top earners may                                                              106 These organizations were Urban Health Plan and BronxWorks (formerly Citizens Advice Bureau) in the Bronx; Brownsville Multi-Service Center and Groundwork, Inc., in Brooklyn; and Catholic Charities and Union Settlement Association in Manhattan.  107  To put these amounts in perspective, the federal poverty level for a family of three (for example, a single parent with two children) in 2009 (roughly midway through the program period) was $18,310. Thus, families of that size and income level that received $3,000 in reward payments in a year would increase their annual income by about 16 percent. Similarly sized families with income below half of the poverty level (or below $9,155 for the example cited above), which some experts would define as living in extreme poverty, would boost their income by 33 percent. Put another way, a reward amount of $3,000 would add about 21 percent to the total wages ($14,560) of a single parent who was paid $8 an hour for working 35 hours per week for an entire year.  140   have been better organized, better able to handle the verification procedures associated with the program, and more likely to track their families’ performance against the conditions they needed to meet in order to earn rewards. Family Rewards reduced current poverty and material hardship (its main short-term goal), but those effects weakened after the cash transfers ended. The proportion of families in the Family Rewards group whose incomes were below the official poverty line during Year 3 of the program was lower by 12 percentage points than for control group members. This effect was due mostly to the reward payments they had earned in the program, and the reduction dissipated after those transfers ended. The extra income they received during the program period helped families reduce a variety of material hardships, and those effects persisted into the early postprogram period, although they were attenuated. For example, the proportion of families that experienced food insufficiency (as indicated by parents responding on the study’s 42-month survey that their families sometimes or oftentimes did not have enough to eat) dropped from over 20 percent in the control group to about 15 percent in the program group. Program group families were less likely than control group counterparts to report not having enough money to pay their rent sometime in the past year. They were more likely to report having enough money to make ends meet and that their financial situation had improved over the prior year. The reductions in hardships were largely concentrated among families that were living in severe poverty at the time they entered the program. Among that group, the program caused a 9 percentage point reduction in the likelihood of reporting food insufficiency after the program ended, and about an 11 percentage point reduction in the likelihood of not paying their full rent in the past year. Family Rewards did not improve school outcomes for elementary or middle school students. For these students, the analysis found few positive effects on attendance rates, scores on standardized tests, or other school outcomes during the program period. Perhaps the model’s limited approach for these children—of rewarding only attendance (which was already high, leaving little room for improvement) and standardized test scores (rather than more immediate performance indicators, such as good report card grades)—might partially explain the lack of improved outcomes. The program did not have an educational payoff for this group. Although Family Rewards had few effects on school outcomes for high school students overall, it substantially increased graduation rates and other outcomes for students who were already stronger readers. Students who were behind educationally when they entered Family Rewards did not experience educational gains from the program. In contrast, those who entered better prepared for high school—and who may have been in a better position to take advantage of the incentives offered—do appear to have benefited. Although subgroup findings tend to carry less statistical certainty than full-sample results, a number of other studies of education-focused incentive programs have similarly found more positive effects for more capable students. Family Rewards had particularly strong effects on students in the ninth-grade cohort who had scored at or above the basic proficiency level on their eighth-grade standardized English language arts test (which primarily tests reading skills) before random assignment. For this subgroup, which made up almost one-third of the overall sample of ninth graders, Family Rewards appears to have improved a range of school outcomes. These include an 8 percentage point increase in the likelihood of graduating from high school within four years (a gain of 12 percent above the 67 percent graduation rate among control group students who were proficient in English language arts at the beginning of the study). The program did not help students who were less prepared for high school. Specifically, the analysis 141   found no pattern of statistically significant impacts on educational outcomes for students in the ninth- grade cohort who had scored below the proficiency threshold on either the eighth-grade English language arts or math exam before random assignment. Family Rewards did not increase families’ use of preventive medical care, which was already high; and it had few effects on health outcomes. The health-related incentives of the program were designed to encourage low-income families to adopt better preventive health care practices. It turned out that a higher proportion of families than the program’s designers had expected were already receiving health insurance coverage and practicing preventive health care. This finding may reflect the success of efforts by New York State and New York City to expand access to health coverage in the years leading up to and during the study period. Perhaps for that reason, Family Rewards had few noteworthy health-related impacts. However, it produced large increases in families’ use of dental care services, leading to increased dental care for parents and children alike. For example, parents in the program group were 10 percentage points more likely than control group parents to report having seen a dentist for any reason in the prior year, and about 12 percentage points more likely to have had two or more dental checkups in the past year. Strong positive effects were also observed among both high school students and younger children. Family Rewards did not substantially improve parents’ employment and earnings. The program did increase the likelihood of self-reported full-time employment. However, it did not increase employment in or earnings from (according to administrative data) jobs covered by the unemployment insurance system, which captures wages reported by employers to the agency that administers the state’s unemployment insurance system. The lack of positive effects in this domain stands in contrast to previous work incentive programs. It may be that the added income families received from the education and health rewards offset the program’s work incentives for some participants, especially those who would have the most difficult time finding jobs in a tough economy. Indeed, subgroup analyses found that the program had a small but statistically significant negative effect on labor market outcomes for parents who entered the program with lower education levels and other disadvantages. In other words, they worked and earned less than they would have in the absence of the program, according to unemployment insurance records. For example, those without a high school diploma or GED certificate had an average quarterly employment rate in Year 3 that was 3 percentage points lower than that of their counterparts in the control group, and they earned an average of $1,790 less (a reduction of almost 8 percent). Lessons learned The evidence available so far on Family Rewards shows that a CCT approach in urban areas in one large American city can reduce immediate poverty and material hardship and promote at least some improvements in some forms of human capital investment, especially for certain subgroups. At the same time, the specific model tested in New York City left many important outcomes unchanged. The evaluation of Family Rewards is continuing, and the final story remains to be written. Further evidence will be available in the next evaluation report to be completed in 2015, which will present findings on the program’s effects over five to six years after random assignment. In the meantime, it seems reasonable to draw at least three general conclusions: (i) it feasible to operate a comprehensive CCT program and target it well to low-income families in high-poverty urban areas; and (ii) the Family Rewards model has not demonstrated its value enough to scale it up as a broader antipoverty policy in its original form. Because of its success in reducing short-term poverty and material hardship while achieving at least some improvements in human capital development, continuing to experiment 142   with a CCT approach in the United States has merit. With these conclusions in mind, New York City’s CEO and MDRC joined forces again to design and test a next-generation version of Family Rewards. Family Rewards 2.0 built on the lessons of the original New York City demonstration and incorporated several important modifications. It was launched in the Bronx, New York, and Memphis, Tennessee, in the summer of 2011 and concluded its operational phase in December 2014. The new model included a streamlined set of financial rewards (8 rewards rather than 22), more frequent payments, and a new family guidance component to try to help more parents and students meet the conditions that enable them to earn rewards. It is hoped that these refinements to the model will make it a more effective intervention. The new program targets families with children who were set to enter ninth or tenth grade in the upcoming school year, since effects on education were found in this study only for the older students. However, once enrolled in Family Rewards 2.0, all of the family’s school-age children are eligible for the health-related rewards. The program also targeted Temporary Assistance for Needy Families (TANF) and food stamp (SNAP) recipients, in order to target resources to the neediest families and to consider how a CCT program might supplement or interact with these safety net programs. Like the original model, Family Rewards 2.0 is being carefully tested using a randomized control trial. Information on early impact findings is planned for release in late 2015. 4.7 Urban Voucher Program in Gaza108 Rationale The Palestinian Territories in general, and the Gaza Strip in particular, embodies a context of complex, protracted crisis with deeply intertwined political, economic, and humanitarian dimensions. With only 3 percent of the population living in rural settings, Gaza is almost entirely urban. Not only are 81 percent of the households in urban settings (and 16 percent in urban refugee camps), but these areas are some of the most densely populated in the world, with an average of 4,350 people per square kilometer. From 2004 to 2011, poverty prevalence in the strip soared from 30.2 percent to 38.8 percent, with a spike up to 49.5 percent in 2007. Such levels are higher than the Palestinian Territories’ average (25.8 percent) and more than twice the average in the West Bank (17.8 percent). In early 2014, the unemployment rate reached 45.1 percent, including 60.8 percent among women. Youth unemployment in the Middle East is less than half that in Gaza, i.e., 27.9 percent versus 45.6 percent, respectively. As a result, social assistance, including externally financed social assistance, represents a key source of household income. According to estimates, the share of Palestinians relying on safety nets increased from 10 percent in 2000 to 80 percent in 2014. The Ministry of Social Affairs spearheaded the formulation of the Social Protection Sector Strategy for 2014–16. The strategy encompasses a range of interventions, including cash, food assistance, and social service programs. Among these, the Palestinian National Cash Transfer Programme (CTP) was established in 2010 by merging two large safety net programs, the Social Safety Net Reform Project and the Special Hardship Case program. Accounting for about 1 percent of gross domestic product, the CTP provides quarterly cash payments ranging from $146 to $468 according to household size and vulnerability.                                                              108 The discussion largely draws from Creti (2014b) produced for this review. 143   The Social Safety Net Programme coordinated by the United Nations Relief and Works Agency for Palestine Refugees in the Near East (UNRWA) provides basic food and cash transfers to 21,000 refugee households (see box 31 for a discussion of Gaza’s refugee population). Also, UNRWA’s cash- for-work program was established in 2001 and provides short-term employment positions ($420 per month) for food-insecure refugee families. Large-scale food assistance is provided to nonrefugee populations in the form of unconditional food transfers, school feeding, and electronic voucher transfers. These support about 24 percent of the Gaza population and are mainly managed by the World Food Programme (WFP) in close coordination with the social protection and food security sectors. Box 31. Gaza’s refugee population Most Palestinians in Gaza are UNRWA-registered refugees. There are around 1.2 million registered refugees in Gaza, accounting for almost 70 percent of the population. Besides their high number, refugees in Gaza have another characteristic that distinguishes them from other urban displaced populations. In other urban contexts, refugees usually have to deal with vulnerabilities related to the fact that local authorities do not accept their presence and do not provide for their needs. In Gaza, refugees have actually led to the expansion of built-up areas with the acceptance and support of national and international authorities. Furthermore, refugee camps have been gradually upgraded by the UNRWA and by refugees themselves, who have built more permanent structures and service infrastructure. Eventually, refugees residing outside the camps have come to outnumber those living inside them. This case study reviews the experience of the voucher program in Gaza, which was introduced in October 2009 as a pilot crisis response intervention within a larger WFP emergency operation. The program, implemented in close partnership with international and national nongovernmental organizations (Oxfam and the MA’AN Development Centre), envisioned a value-based voucher targeting 15,000 people (2,335 households). By January 2012, the program doubled its coverage to 30,000 people, driven by rising needs as well as an expanding resource base dedicated to vouchers. In 2011 and 2012, a number of programmatic innovations were tested and introduced, such as the transition from paper-based to electronic vouchers, the inclusion of new commodities redeemable through the voucher109, and a program variant combining the provision of vouchers and in-kind food. By January 2013, another 30,000 beneficiaries were covered, with 20,000 participating in the pure voucher program and 10,000 in the combined version. In 2014, program officials were positioning the voucher system as a platform for a range of humanitarian and development projects. In the West Bank, the voucher system was leveraged to provide food vouchers to refugees through the UNRWA. This allowed the UNRWA to manage the program through its network of field staff, while utilizing the voucher system and network of stores. The program was expected to reach 32,000 refugees by the end of 2014. During the 51-day conflict in Gaza in the summer of 2014, the Palestinian Authority partnered with the WFP and the United Nations Children Fund (UNICEF) to provide a voucher to conflict-affected households that could be redeemed for food as well as water and sanitation items in local stores. As schools reopened following the conflict, school uniforms were added to the voucher. Through this initiative, 300,000 people were provided with food, 84,000 people received water and sanitation items, and 14,000 children received school uniforms. The WFP also allowed two international nongovernmental organizations to use the voucher system for their projects.                                                              109  In 2012, the food basket consisted of nine commodities: bread, cereals, eggs, labaneh, milk, pulses, rice, vegetable oil, and yogurt. Others were subsequently added—such as olive oil, canned fish, tahini, and condiments (duqqa and zatar)—based on programmatic review of beneficiary preferences.  144   In the fall of 2014, a new voucher system was launched that will further enhance the ability to use the program as a platform for a range of humanitarian and development work. The system uses electronic “wallets” for different sectors, such as food security; water, sanitation, and hygiene; education; and health. Approximately $9 million was injected in the Gaza economy during the conflict through the vouchers for food assistance. Since 2011, the voucher program has injected $84 million in the Palestinian economy. Targeting The voucher program’s core objective is to enhance the food consumption and dietary diversity of food-insecure populations. The decision to use a value-based voucher was driven by three factors: (1) to provide beneficiaries with access to a balanced food diet through existing market mechanisms, including staples (such as bread and wheat flour rice) and protein-rich foods (such as dairy products and eggs); (2) to give beneficiaries choice and flexibility when accessing their entitlement; and (3) to ensure that cash is directly injected into the local economy and benefits local production, employment, and small businesses. The initial eligibility determination was based on having incomes below a certain threshold as measured by a proxy means test (PMT) formula110. The first targeting exercise was conducted in 2010, with a recertification process in 2013–14. The PMT method is commonly used—with some slight variations—by a range of actors involved in social assistance, including the Ministry of Social Affairs, the World Bank, the WFP, and the UNRWA. Since 2013, the PMT-based analysis has been complemented by the Food Consumption Score index, a measure of dietary diversity. Interventions are devised based on predetermined combined cutoff points, including relative and deep poverty lines (figure 40).                                                                110  Households with a PMT score below 6.39 are classified as below the deep poverty line, those scoring between 6.39 and 6.57 are classified as below the relative poverty line, and those scoring above 6.57 are classified as above the relative poverty line. The deep poverty line was defined by the Palestinian Authority as about $495 a month for a family of two adults and three children; the relative poverty line was about $594 a month for the same type of family.  145   Figure 40. Matching targeting measures, poverty lines, and interventions From this standpoint, the eligibility of voucher participants is based on them being below the deep poverty line as measured by the PMT and having a borderline or poor Food Consumption Score. Households that fall below the deep poverty line and have high consumption and dietary gaps are enrolled in the combined in-kind and voucher arm of the program. In 2014, the voucher program reached 50,000 nonrefugee beneficiaries living below the deep poverty line (table 23). These beneficiaries represent 3.0 percent of the total Gaza population, 10.0 percent of the nonrefugee population, and 14.5 percent of the population living below the deep poverty line. Recertification of beneficiaries is conducted every two years in alignment with other national safety nets. Reviews suggest that the PMT would be an appropriate indicator for exiting or graduating from the voucher program. For instance, the Ministry of Social Affairs recently graduated some households from the CTP to microcredit and loan programs. 146   Table 23. Benefit structure in Gaza’s voucher program Household Selection Criteria Food Consumption PMT Score Beneficiaries Modality Transfer Below deep poverty Good 170,000 Food basket Food ration (2,100 line calories) Below deep poverty Borderline and poor 50,000 Value-based $12.50/person/mont line voucher h Below extreme Poor with high 10,000 Combined in-kind Ration of fortified poverty consumption and dietary + voucher wheat flour and gap $9/person/month Benefit structure During the pilot, the value of the voucher was calculated based on the local market value of the in- kind food ration. Food prices were regularly monitored so as to check the devaluation of the voucher over time. Program reviews conducted in 2011 and 2012 recommended that the value of the voucher be adjusted to price inflation. The average transfer currently is $15 per capita per month, or about 12 percent of the average monthly income of the poor (about $120). The transfer is adjusted by household size, ranging from $15.20 per month for households of one or two members to $97.20 per month for households of nine or more members. Currently, the voucher value is expressed in U.S. dollars but implemented in the local currency (NIS). As a result, in the absence of adjustment, the risk of inflationary costs is borne by the beneficiaries, while the risk of exchange rate fluctuation is absorbed by the implementing agency. Yet the voucher program includes a 10 percent allocation for contingency against price inflation and fluctuation of the exchange rate. Initially, the benefit took the form of paper vouchers distributed monthly in booklets of four. In July 2011, these were replaced with a debit card (e-voucher) called a sahtein. The system allows users to purchase groceries from the assigned proximity shop. E-vouchers can be used for bread, canned fish, cereals, eggs, labaneh, milk, olive oil, pulses, rice, tahini, vegetable oil, wheat flour, white cheese, yogurt, duqqa, and zatar. This list comprises Palestinian and imported items. Given the perishable nature of some of the items, the e-voucher is topped up four times a month, but the planning and monitoring cycles are monthly, and the shop payment cycle is bimonthly. At the beginning of each month, market-based ceiling prices are agreed upon by the WFP, cooperating partners, and the shops. This measure is aimed at protecting beneficiaries from price volatility and reducing the risk of speculation. Voucher redemption is controlled by an electronic terminal. After having their identification verified111, beneficiaries swipe their card and the shopkeeper enters the products purchased (value and quantity) using a menu on the terminal. The terminal produces a receipt for the beneficiary and another for the shopkeeper. The voucher program has been complemented by a nutrition-awareness campaign and gender                                                              111  The program encountered virtually no barriers related to identification. Almost everyone in Palestine has an identification number; those rare cases of households without a Palestinian identification could use Jordanian or Egyptian documents.  147   empowerment activities (including strengthening informal networks and promoting the development of neighborhood groups through community events); these are implemented in partnership with Oxfam and a local nongovernmental organization, Ard El Insan. These complementary activities were not conditional and were intended to serve as a pilot for learning, sharing, and later potential mainstreaming through government programs. Participation is monitored through an attendance register. These complementary measures have, to some extent, translated into improved diets through different approaches to ingredients and cooking. For example, several beneficiaries reported using substantially less cooking oil because of the nutrition training. Qualitative evidence also shows that, while the voucher program has a positive social impact on men by restoring their perceived role as household breadwinner, the empowerment activities give women an opportunity to increase their mobility and social engagement outside the family. Institutional arrangements One of the main concerns regarding the design and implementation of the voucher program was to avoid a disproportionate accrual of benefits to large retailers. For this reason, the program aimed to involve small proximity shops, preferably those beneficiaries already used to purchasing their food items on a daily basis. However, beneficiaries were only allowed to redeem their vouchers at selected preassigned shops, limiting their ability to choose among retailers. This condition was lifted during the 2014 Gaza emergency. The rationale for restricting the portability of the voucher to a predefined shop was mainly due to potential monitoring and reporting difficulties. It proved difficult to monitor the voucher exchanged in different shops located in different governorates, especially with regard to double counting. Proximity shops were selected according to a ranking score based on a number of criteria, including sufficient stock capacity, availability of adequate cold storage facilities, good hygienic conditions, and possession of a bank account and trading license. Some of these criteria, particularly those related to storage, ruled out small shops. Medium-size shops tended to meet most of the criteria. A total of 23 such shops were selected over 2009–11; this pool was increased to 60 by 2013, and to 90 in 2014. The program introduced a feedback mechanism by putting a box in each shop where beneficiaries could post complaints, requests, or suggestions. This mechanism was successful, and beneficiaries used it—for instance—to request to be moved to a closer shop. Feedback was addressed either case by case or through workshops with participants organized by the implementing partners. Monitoring and evaluation The voucher program has been regularly monitored through various tools, such as postdistribution household visits to beneficiaries based on a random sample; shop visits to check on the physical space, quality of products, packaging and expiration dates, storage conditions, the electronic terminal, display of lists of commodities and ceiling prices, and the suggestions/complaints box; and electronic monitoring of sales values and quantities, backed up with paper records and a biweekly reconciliation. During the early phases of the program, the postdistribution monitoring covered 5 percent of the beneficiary households every month; this was subsequently reduced to 1 percent. The decreased monitoring workload was made possible due to the introduction of the e-voucher, which allowed for more secure management and closer monitoring of transactions. A baseline survey (October 2009), two postdistribution surveys (August 2010, February 2011), and a midterm review (March 2011) were conducted to monitor the process and impact of the pilot phase as well as provide operational 148   recommendations. These were followed by a programmatic review in 2012 and a 2013 study of the multiplier effects of the voucher program. The efficiency of targeting based on income eligibility criteria (PMT threshold) was very high, with 93 percent of the targeted group falling within the poverty line threshold. In terms of food consumption, 88 percent of the beneficiaries moved to a good consumption level, with less than 1 percent remaining in the poor consumption group. Vouchers were twice as cost-effective as in-kind food assistance in moving beneficiaries from a poor food consumption score. Those improvements in food security are correlated with increased consumption of milk and other dairy products, and eggs. At the same time, vouchers do not seem to compromise access to and frequency of consumption of staple foods such as cereals and oil, at least for non–extremely poor households. A reliable and sufficient source of food or income is a major factor in reducing stress in the home. Such decline in stress reduced the risk of domestic and gender-based violence. The 2012 review reported an increased sense of dignity felt by program participants as a result of using vouchers and being able to shop as customers as opposed to collecting rations. A study of the voucher program carried out in 2014 via survey confirmed previous findings of the program’s strong direct and secondary economic impact on beneficiary households, participating retailers, and local dairy producers whose commodities are redeemed through the e-vouchers. Key findings of this impact study include the following: (i) household dietary diversity was improved 9 percent compared to that of in-kind recipients; (ii) every voucher dollar generates $0.40 of additional sales at participating shops; (iii) participating shops have stimulated the local economy through $772,000 in investments; (iv) a total of 485 new jobs in participating shops and affiliated producers have been created since the start of the program; and (v) over $64,000 in value-added tax revenue for the government has been generated every month. About 65 percent of participating processors attribute increased sales to greater distribution through the program. Lessons learned The Gaza Strip presents a range of challenging issues, many unique. In a context of long-standing protracted crisis, it is often difficult to separate program-specific issues from those of the broader economic and political realms—especially involving issues of sustainability, labor markets, and the role of social assistance. In other words, the constrained urban environment in Gaza makes safety nets play a fundamental social, economic, and even humanitarian role for the strip as a whole. Yet even under such difficult circumstances, the social protection agenda has advanced remarkably, including in terms of the interventions provided (for example, by the CTP), the technology supporting them (for example, the voucher program’s e-system), and the level of policy and operational coherence across programs. For instance, a review found that the CTP’s targeting accuracy was the top-performing one of 29 countries studied. This level of accuracy provided a solid basis for an overall targeting harmonization process across interventions, including the voucher program. A challenge of the application of the PMT in Gaza is the potential exclusion of the so-called “new poor”—those households that have recently fallen into poverty due to the blockade. These households have limited income, but are unlikely to be targeted through the PMT because they maintain a good asset base. It is important to find mechanisms to identify and protect this vulnerable group before it further slips into poverty. Also, while there has been considerable progress in establishing a common 149   base for targeting different interventions, more work needs to be done to harmonize monitoring and evaluation systems. An issue observed in the Gaza Strip is the difficulty in distinguishing between households and families. High population density and limited housing mean that several households from the same family commonly live in the same property. These families may maintain their individual behaviors, such as eating separately, but they often share resources. The household/family distinction is key at the poverty assessment stage, as well as in determining the transfer value. The urban context offers the opportunity to engage with the community-level support networks crucial for the population in dealing with Gaza’s challenges. Distribution locations (like banks or shops) and complementary activities can be used to raise awareness and channel feedback. Also, awareness campaigns represent an opportunity to reinforce women’s empowerment and protection. In the long run, it might be worth investing in an e-voucher system that allows beneficiaries the freedom to redeem their vouchers from a wide range of shops. This would require investments in monitoring and reporting associated with program portability. 4.8 Social Pensions in Urban Delhi112 Rationale The National Capital Territory of Delhi has witnessed a massive transformation over the past decade, characterized by higher urbanization, income growth, decline in poverty levels, and higher employment. For instance, in 2010/11 Delhi’s gross domestic product rose by about 19 percent over the previous year. At the same time, the prevalence of poverty declined from about 13 percent in 2004/05 to about 10 percent in 2011/12. A prevalent view in the Indian policy sphere, as reflected in the recent Task Force on Urban Poverty report, is that exclusion from social assistance programs is a function of geographic isolation as well as a lack of sociopolitical capital. The recognition that vulnerable populations at risk of program exclusion reside in slums has driven administrative reforms in the targeting and eligibility criteria used to determine social assistance. The urban poor face distinct problems not faced by the rural poor. Broadly speaking, urban poverty is characterized by the lack of access to basic facilities and services. Low-income residents live in small, overcrowded settlements with poor, unhygienic conditions like slums, or live as squatters. In addition, those in informal settlements—hovering at the margins of legality—tend to be concentrated in informal occupations, making access to work- based benefit packages and formal social insurance more difficult. A study by the Institute of Rural Management estimates that nearly 57 percent of Delhi’s workforce is informal, and most of this informal workforce resides in slums. There is no comprehensive policy framework for urban social protection, especially for rights- centric entitlements. The major obstacle in this regard is the lack of legitimacy of the urban poor, a lack that relates to various issues arising from the characteristics of those settled in slums. Most slum inhabitants are migrant workers and face challenges such as social discrimination (based on ethnicity), uncertainty of employment, lack of legal representation, illegality of homes, and lack of identification—all of which means that the urban poor face a distinct lack of visibility, both in                                                              112 The discussion largely draws from Bhattacharya et al. (2014) produced for this review. 150   general and in the political sphere. Like other Indian states, Delhi has over 40 social protection programs administered by nine different departments that target the poor and vulnerable113. Each of these programs has its own target population, implementation, and reporting arrangements. From the point of view of the poor citizen seeking public assistance, this multiplicity entails understanding various eligibility rules, making several applications, and engaging with multiple authorities. More systematic documentation of state- sponsored initiatives can provide vital learning opportunities for states wishing to reform. The Delhi government’s Mission Convergence initiative, influenced by Kerala’s Kudumshree model, aims to ensure such a systematic approach. This case study reviews one particular category of safety nets: social pensions or unconditional cash transfers for widows, the elderly, and the disabled poor. Targeting Social pensions are provided to poor rural and urban residents by the government of India within the National Social Assistance Programme and under three schemes: the Indira Gandhi National Old Age Pension Scheme, the Indira Gandhi National Widow Pension Scheme, and the Indira Gandhi National Disability Pension Scheme. As of December 2012, these three schemes covered 26 million beneficiaries belonging to families living below India’s official poverty line with total budgetary support of around $1.2 billion. In addition, several state governments in India spend substantial resources from their own budgets in supplementing the coverage and transfers of the central government. The government of Delhi, which has been providing pensions since 1975, uses different criteria from the national targeting guidelines to determine eligibility. In particular, the state does not make strict use of the below-the-poverty-line criterion for eligibility determination. To be eligible for the old age, widow, or disability pension, an individual’s household cash income must be below a specified threshold. The applicant also must be a resident of Delhi for a minimum of five years. The three programs offer $16 per month to each pension beneficiary (increased in 2008 from $9.60). These payments are made through direct transfers to bank accounts held by beneficiaries. Table 24 sets out the eligibility criteria and benefits under the three pension schemes.                                                              113 In addition to social pension programs, school scholarships cover 0.5 percent of Delhi households. A school meal program offers wider coverage, reaching 10.7 percent of households. The Public Food Distribution System, which supports 3.3 million people, is particularly noteworthy; its reforms in targeting from 1997 onwards have reversed its pro- urban bias. Swarna Jayanti Shahari Rozgar Yojana is the only program that offers wage/self-employment and skills training opportunities for the urban unemployed, but a negligible proportion of urban households have received such support. 151   Table 24. Features of social pension programs in Delhi Eligibility Criteria Annual Household Cash Program Applicant Age Income Other Characteristics Monthly Transfer Coverage Old age ≥ 60 < $94 — $16; $24 for 385,000 pension those above 70 and minority communities Disability 0–60 < $1,172 Disability $16 48,000 pension (physical or mental) Widow 18–60 < $750 Widow, separated, $16 130,000 pension woman in distress Data from the 2011 Abdul Latif Jameel Poverty Action Lab (JPAL) survey shows that, on average, only 36 percent of those eligible in the poorest quintile participated in the program. World Bank surveys conducted in 2013, which are representative of the elderly and widowed populations in Delhi’s slums, corroborate the low levels of coverage indicated by the JPAL survey. Results show that only a quarter of the elderly and one-third of the widows were covered, which is less than half the coverage rates the state intends based on its eligibility criteria. Exclusion errors are dominant, and results show that ineligible people covered by the old age and widow pensions included about 28 percent and 22 percent of beneficiaries, respectively; about 46 percent of the old age pensioners and about 30 percent of widow beneficiaries have no income source other than those programs. Applicants for old age pensions need to provide documentary proof of residency within the state of Delhi, as well as evidence of disability/age/marital status. Perspective beneficiaries can follow two routes in applying for the scheme: self-reporting with documentary evidence, or with support from public officials and without documentary evidence. The most common problem pensioners applying on their own faced was obtaining supporting documents for the application. Nearly 27 percent of widows and the elderly reported this as their major problem. Filling out forms and understanding the application process are the next major sources of reported difficulty (figure 41).   152   Figure 41. Main difficulties in application process for pension schemes (percent) Others 7.8 8 Elderly Submitting Forms 3.9 3.1 Widows 4.9 Travel to Offices 6.7 Obtaining Signatures 4.2 2.7 14.5 Understanding Application Process 15.7 27.2 Getting Supporting Documents 27.1 Filling Form 25.3 22.6 12.2 Obtaining Form 14.2 The alternative route for those without requisite proof of residence, income, or age is to file an application with no documentary evidence. To do this, an applicant must obtain recorded statements from two officials or witnesses114. About 70 percent of applications are made in this way. On average, pensioners apply for pensions 1.5 times, and have to wait an average of 5.24 months from application to approval. Pensioners also report it taking 6.19 months on average from application to first payment. About 5 percent of pensioners reported paying a fee for documents, signatures, or submitting applications. In general, awareness levels among the pensioners on issues involving payment mode (bank), frequency (quarterly), and transfer size were high, ranging from about 79 to 98 percent of surveyed beneficiaries. About 78 percent of all pensioners withdraw the entire pension amount, and 45 percent share the amount partly with their family. The majority of pensioners (nearly 73 percent) spent the transfer on food. Some 98 percent of pensioners had a bank account. Around 94 percent of pensioners had to open bank accounts to receive pensions. Only about 13 percent of pensioners were asked to show identification during withdrawal. Institutional arrangements The availability and quality of program performance data is a key challenge. Reviews of administrative and secondary sources highlight limited systematic data on coverage and targeting outcomes. Different data sources suggest a mixed picture of coverage. While official data show that schemes covered a majority of the elderly poor, more in-depth data collected through household surveys in slums suggest that only a small fraction of those eligible for the pensions manage to enter the applicant pool and receive benefits. The cumbersome and costly application process emerges as a core bottleneck. Transaction costs of                                                              114  Officials and witnesses must be a member of parliament or a member of the local legislative assembly, a local resident welfare association president or general secretary, a neighbor, a registered self-help group president, general secretary of integrated child development services, or an accredited social health activist worker or officer of the Delhi or central government. Each witness needs to provide his or her own length of stay in Delhi.  153   applying for programs are high due to information asymmetry, and ambiguity regarding application and documentation requirements and their implementation. The lack of service networks and platforms to facilitate the application process imposes greater costs on the poor and discourages them from applying for scheme benefits. Despite variations in geographic size, density of population, and total number of households, each district other than the northwest of Delhi has only one Department of Social Welfare office. Hidden costs such as purchasing forms, photocopying documents, and making multiple trips to the appropriate office (for documents and applications) make the enrollment process transaction intensive and costly for applicants. Given the difficulties faced by the elderly and the disabled in traveling, and the high opportunity cost of time for the urban poor, the application process may discourage citizens from enrolling. Many potential applicants find acquiring the necessary signatures and corroboration from politicians and public officials to be fairly cumbersome. Results from World Bank studies show that 30 percent of slum households reported interacting with a politician, while only 10 percent had ever interacted with a member of the legislative assembly (MLA). Data also suggest that the asset poor are also sociopolitically poor. In other words, those who lack material wealth in slums report a lower likelihood of interacting with elected officials. An applicant’s ability to access bureaucrats is a function of geographic location and type of residence. Thus, those living in informal slums and squatter settlements may have resided in the city for more than five years but cannot gain MLA corroboration as the MLA lacks adequate knowledge of these applicants and their areas of residence. Similarly, the poorest find it costly to access politicians; at the same time, party officials may have weak incentives and skills to travel to all households in slums to solicit applications, scrutinize documentation, and check claims. In practice, applicants negotiate and follow up repeatedly with local political workers to apply through the local MLA and check their application status. This situation is exacerbated by a lack of support staff at the district or subdistrict/ward level who can provide assistance in filling out forms, verify documentary proof, and provide photocopying services. As per state guidelines, priority is to be given to applicants residing in slums or resettlement colonies. Yet there seems to be no clear prioritization protocols based on geographic criteria. Sociodemographic characteristics such as sex, size of household, and number of dependents are important for prioritization. In practice, analysis and discussions suggest that the current nature of prioritization is ad hoc, and that administrators do not seem to rank applicants in terms of their severity of need or poverty, as the paperwork submitted does not allow them to make such assessments. Pensions are sanctioned on a first-come, first-served basis. There is no standardized set of reasons for rejecting applications, and verification protocols are limited115. Interviews and reviews of institutional roles and responsibilities revealed that there is no dedicated team managing and implementing tasks related to social pensions. Staff members at headquarters and district offices are not exclusively involved in scheme-related functions, as the Department of Social Welfare manages other financial assistance schemes and welfare activities. Currently, there is no clear allocation of roles and responsibilities within the department. Ward councilors are elected to manage                                                              115 In theory, a sanction letter/order must be sent from the Department of Social Welfare to the beneficiary indicating the beneficiary’s identification number and details, date of sanction, and an address to contact in case of queries. But beneficiaries do not seem to receive this letter. The common method adopted by beneficiaries is to instead check with each other or with the MLA party office as to whether their pension has been sanctioned. The MLA party offices maintain registers to inform the beneficiaries based on a consolidated acceptance/rejection list (including reasons for rejection) from the Department of Social Welfare. Opening a bank account is fraught with difficulty as different banks have differing interpretations of guidelines for verification and proof of residence. Such difficulties are part of the larger problem of helping migrants access those schemes. 154   municipal mandates related to water, sanitation, and urban infrastructure. Municipalities are also responsible for another set of welfare schemes that are not funded by central assistance. For example, the Municipal Corporation of Delhi manages its own pension schemes for widows, the destitute elderly, orphans, and the disabled. Beneficiaries for these schemes are identified on the basis of referrals and recommendations from municipal ward councilors. Thus, urban municipal bodies play no role in the implementation of social pensions managed by the Department of Social Welfare. Survey data for urban Haryana and Delhi show that, although similar in city structure, those urban contexts differ significantly with regard to social pension performance. Specifically, Haryana covers a large share of its target populations in urban areas as compared to Delhi’s coverage. There are various reasons for this. First, eligibility rules in Haryana seek proof of current domicile for the past one year, while Delhi seeks proof of five-year continuous residence. The latter is much more cumbersome to collate, compile, and submit—thus, coverage in Delhi could be lower due to the additional paperwork burden. Furthermore, in Haryana, three documents suffice for an application (ration card, voter card, and a certificate proving age/death of husband). In Delhi, four documents are needed for old age pensions and five for widow pensions. In Haryana, domicile certification is automated and decentralized through urban citizen service centers. In Delhi, an applicant must approach the relevant district magistrate office, which provides the certificates through the revenue department. Only 22 percent of Haryana pension applicants perceive political connections as being a major help in receiving pension benefits; in contrast, the majority (56 percent) of applicants in Delhi feel political connections matter in this regard. Despite both states requiring corroboration from urban elected representatives on application forms, access to these representatives in urban Haryana is more formalized and clearly codified due to the involvement of urban local bodies and cadres of municipal administrative staffers. The Delhi system is more ad hoc with no formal outreach body or mechanism for the local MLA. Consequently, slum dwellers rely on local party officials to forward applications to an MLA on their behalf. Lessons learned Despite a near-universal targeting regime and relaxed eligibility rules, entry into pension programs in Delhi remains challenging. Clearly, scaling up eligibility on paper does not yield improved coverage in practice. The government should aim to reduce transaction costs involved in accessing documents and relevant functionaries for applicants living in low-income housing and settlements. This improvement can be achieved through systematic identification and scheduled/announced enrollment camps116. Another important lesson is about the need for centralized support at the state level in opening or linking bank accounts so as to avoid inconveniencing beneficiaries. The government may need to consider specifying the type of financial products and services that should be made available to beneficiaries. Many beneficiaries with low-balance accounts have stated that they were not provided with an automated teller machine (ATM) card or checkbook; this made accessing payments difficult,                                                              116 Camps should include all functionaries who are to provide supporting documents and signatures as well as bank representatives. This strategy can be particularly proactive in notified slums where a large share of the population has some form of documentation. Camp performance should be evaluated and tracked by recording the number of citizens issued documents, the number of applications made, and the number of camp-based applications by the Department of Social Welfare. Such a process can create feedback loops within the pension application process that can be leveraged by other safety net programs. These data need to be improved, because the informality of the urban poor is a major hindrance to their social protection. 155   as pensioners had to make multiple trips to their local bank branch to withdraw payments in person. In Delhi, the state government can also consider mobile forms of payment and disbursal to ease constraints posed by opening a bank account. Many migrants in the city find the current safety net regime difficult to navigate and access due to residency requirements. The current eligibility criteria concerning term of residence discriminates against migrant groups. This is a major design gap in the social pension program and its implementation in urban areas. If portable pensions are not fiscally and administratively feasible, the government of India could consider a contributory migrant pension fund for informal workers, in partnership with various pension boards set up under the Social Security for Unorganized Workers Act of 2009. At present, Rashtriya Swasthya Bima Yojna (RSBY), a health insurance program for the poor, is the only central program that allows families that fall below the poverty line to avail themselves of program benefits anywhere in the country. The scheme’s design aims to make insurance available to migrant workers and their families. In Delhi, RSBY partnered with Mission Convergence for enrollment. To enable portability, each household was given a unique national identification number under the RSBY program, allowing the implementing agency—the Ministry of Labour and Employment—to track enrollment and hospitalization data geographically. These data can be used to identify migrant households in the state with elderly and disabled persons in need of further assistance. Governance and political economy pose a serious challenge in the design of a safety net regime. In rural India, safety nets rely on local information and cheaper verification done by village-level authorities or political representatives. However, population density is much greater in urban areas, and there is a limited role for the municipality in Delhi—that is, the state of Delhi cannot use workers from municipal bodies due to their differing roles and mandates. Another important challenge to safety net reform in Delhi has been the unclear nature of the authorizing environment due to the lack of urban local bodies involved in implementation. For instance, in the recent past, the mandate of citizen service centers and nongovernmental organization partners has been modified, with the focus having shifted from facilitating targeting and delivery to training, information, and outreach activities in slums. This shift was made in response to nongovernmental actors assuming a perceived threat that local political forces posed in assuming positions of authority in the program delivery cycle. Based on qualitative data analysis and a seminal survey on slum governance and social networks, this case study has looked at how access to political representatives is elite driven and more likely to occur in notified slums and resettlement colonies. The survey findings, based on responses from a representative sample of slum households in Delhi, highlight the high-powered incentives of politicians vis-à-vis street-level administrators in interacting with the poor. The findings show that only one-tenth of those sampled had ever interacted with an administrative officer, which suggests that bureaucrats are more likely to interact with elite and asset-wealthy slum dwellers. They also indicate that the size and quality of social networks matter for political access: those with broad networks that extend beyond close relatives or villagers gain better access to government officials. Wealthier individuals and those with close relatives in the city prior to migrating had a much higher chance of knowing and interacting with an elected politician. Other nonwealth barriers, such as ethnicity, are also relevant in this regard.   156   4.9 Philippines: Modifying a Conditional Cash Transfer for Urban Areas117 Rationale The Pantawid Pamilyang Pilipino Program (Pantawid Pamilya) is the bedrock of the government’s social protection programs. It was introduced when Filipino poverty reduction was stagnant despite historically high economic growth. The program aims to reduce immediate and long-term poverty by promoting investments in education and health through conditional cash transfers (CCTs). It targets poor households with children aged 0–18. Households are identified through Listahanan, the national household targeting system. Since its inception in 2007, Pantawid Pamilya has rapidly scaled up to become the fourth largest CCT in the world, reaching about 4.4 million households, or approximately 22 percent of the country’s total population118. Implementation is supported by over 11,000 staff members, a centralized management information system, and a budget of about $1.4 billion (0.5 percent of gross domestic product). The program has been subject to robust impact evaluations which have shown it is on track to achieving its objective. Pantawid Pamilya requires that beneficiary households have an address119. It has several additional eligibility requirements such as the presence of children or pregnancy; these are verified during the enrollment process. To address those poor households excluded by the CCT, a modified CCT (MCCT) was piloted; this pilot also served as a bridge to mainstream beneficiaries to the regular program after one year. Table 25 contrasts the main design features and implementation arrangements of the two programs120. Clearly, the pilot builds on the earlier program’s strategy, but has important modifications in terms of targeting, conditions, benefits package, interventions, and implementation modalities. This case study summarizes the key features of the CCT and the MCCT programs, and lessons learned from their implementation in urban areas. Because the MCCT program is a pilot, its design and implementation have been modified since its launch; the information presented here reflects conditions as of early 2014.                                                              117 Case study prepared by Yuko Okamura, Shanna Rogan, Jorge Miranda, Sheryll Naminigit, and Ugo Gentilini. 118 In terms of absolute number of beneficiaries, Pantawid Pamilya follows flagship CCT programs in India, Brazil, and Mexico. 119 This requirement means that the CCT program does not cover or reach a key segment of the most vulnerable and disadvantaged group because of its design: the homeless and others without any housing structure. Also, because beneficiaries are selected from Listahanan, the national household targeting system that contains information on 11 million households (60 percent of all households in the Philippines), some households might not be included because they were missed in the major 2009/10 data collection effort due to their geographic isolation, or location in an area affected by conflict where security and safety concerns did not allow for data collection. 120 Interviews indicated that, in some cases, the term MCCT was deemed misleading, given the considerable differences between the two programs  157   Table 25. Key features of the Pantawid Pamilya CCT and MCCT Feature CCT Modified CCT Started in 2007 (pilot) 2012 (pilot) Pilot areas 2 urban cities and 4 rural municipalities Urban areas (National Capital Region and highly urbanized cities and municipalities in other regions)a Objective To respond to food and fuel price shocks and the To reach out to those who are poor but were not global financial crisis in the context of a social included in the Listahanan enumeration sector reform agenda Coverage 4.2 million households 218,000 households Delivery Implemented by the Department of Social Welfare Two schemes, one run by civil society mechanism and Development organizations (CSOs)b and one managed by the Department of Social Welfare and Development Target Poor households with children 0–18 years old Poor households with children 0–18 years old and beneficiaries and/or pregnant women meeting one of following criteria: Homeless street family Indigenous people in geographically isolated and disadvantaged areas Evacuees or displaced people due to disasters or armed conflictsd Targeting Poor households identified by Listahanan based on Poor households not included in Listahanan; CSOs methods proxy means test use pre-enrollment forms (intake sheets) developed by MCCT program office Conditions Health: (1) age-relevant preventive care in Health and education conditions are modified as accordance with Department of Health protocol, follows: (2) attend a family development session once a Health: more frequent attendance of family month development sessions Education: regular attendance (85% of school Education: gradual increase of attendance rates, days) alternative education models for the homeless Benefits Health: $11.00 Besides health and education grants as per the Education: $6.60 for elementary school child, CCT program, benefits and interventions based on $11.00 for high school child, for a maximum of 3 case management assessment; support includes children per household housing grant, alternative family home, cash for work, training, psychological counseling, livelihood assistance, and referral services Payment Every two months Designed as monthly, but implementers try to frequency synchronize with regular CCT schedule Duration Until child turns 19 or graduates from high school Six months to a maximum of one year, until household is assessed under Listahanan, included in the database, and identified as poor (or not poor) a. The MCCT Homeless Street Families initiative was pilot tested in the National Capital Region; the Families in Need of Special Protection initiative was pilot tested in other regions. b. With the technical support and supervision of the regional project management office, CSOs undertake overall implementation and operation, and are funded to provide 158   the necessary intervention support services and office space for the period of implementation. c. In addition to meeting the criteria, beneficiaries must demonstrate a willingness and commitment to comply with program conditions. d. These “families in need of special protection” are temporarily housed in evacuation centers or transitional shelters. Targeting In general, the Pantawid Pamilya targeting process relies on a census-type data collection strategy (in areas where poverty is concentrated) capturing key variables that would inform a PMT121.4 In areas with a lower concentration of poverty, households can ask to be surveyed. Survey information on potentially eligible households is included in Listahanan. Table 26 shows the results of the first nationwide assessment from 2008 to 2011. The number and distribution of the poor based on Listahanan closely align to official statistics (based on surveys)122. Table 26. Number of households assessed and identified as poor Households Urban areas Rural areas Total Assessed 3.4 million 7.5 million 11 million Identified as 1.4 million 3.9 million 5.3 million poor Listahanan estimates household income using proxy variables highly correlated with household income, including household demographics, education and occupation of household members, housing conditions, access to basic services, and asset ownership. To factor in different characteristics of the poor in urban and rural areas, separate PMT models were employed. Differences include education levels, access to basic services, and occupation. For example, the urban and rural poor use different types of toilet facilities and housing materials (figures 42 and 43). Estimates at the design stage show that urban and rural models have similar levels of predicted exclusion and inclusion errors: around 31–35 percent and 22–25 percent, respectively123. Figure 42. Distribution of poor households by type of toilet facility, urban and rural areas Urban Area Rural Area                                                              121 The PMT is a regression-based tool that determines a household’s welfare status based on observable and verifiable variables such as household composition, socioeconomic characteristics, assets, education, access to basic services, and other household variables as proxies of income. The welfare level is then compared to official poverty thresholds to classify if a household is poor or not. 122 According to the Family Income and Expenditure Survey of 2006, 71 of 100 poor Filipinos live in rural areas. 123 Calculated using 2003 Family Income and Expenditure Survey data for urban and rural areas and compared to 2006 official poverty thresholds. 159   Figure 43. Distribution of poor households by type of materials of wall (%), urban and rural áreas   Urban Area Rural Area Source: GoP (2011) Pantawid Pamilya is considered one of the best targeted cash transfer programs in the world. Benefit incidence analysis shows that 72 percent of CCT households belong to the poorest quintile of the national income distribution, which is over twice the global average for lower-middle-income countries (34 percent). However, spatial analysis based on reporting data points to a challenge in correctly identifying the poor in urban areas—inclusion errors were 15 percentage points higher in urban areas (37 percent) than in rural (22 percent). The Listahanan data collection strategy differs for urban and rural areas. Rural areas with high poverty incidence were completely assessed (full saturation). In urban areas, a more circumscribed approach was adopted, which focused on pockets of poverty alongside on-demand-application (ODA) for areas outside those pockets124. The term ‘pockets of poverty’ indicates geographic areas in the village, city, or municipality where clusters of poor households reside. The Department of Social Welfare and Development assesses urban areas to identify such pockets by examining access to potable water, housing materials, access to electricity, informal settlers, attendance rate among school-age children, malnutrition, inaccessibility by regular means of transportation, vulnerability to environmental hazards, availability of waste collection facilities, and access to schools and health centers. ODA usage in regions with large urban areas was more widespread than elsewhere, suggesting a greater role for ODA in capturing the urban poor125. Listahanan household interviews were conducted door to door, thereby excluding families without housing. As noted, the MCCT program was designed to cover potential beneficiaries such as homeless street families; these were identified by civil society organizations or nongovernmental organizations based on their existing records and grassroots presence. Potential beneficiary households were assessed using a prescreening form and enrolled in the program using a family intake sheet; this sheet then becomes the family’s baseline profile and serves as a basis for case management. Unlike the core CCT program, the MCCT program focuses on poverty hotspots. The methods and protocols used by the two programs to identify poverty pockets and hotspots are described in box 32.                                                              124 ODA provides an opportunity for households that were not assessed during the regular enumeration, but whose residence was identified as within the target areas, to apply for assessment and be included in the database of poor households  125 Out of the total number of households enumerated (11 million), around 8 percent (0.9 million) appealed for and were approved through ODA. Urban regions had a higher ODA share: for example, 16 percent for the National Capital Region. 160   Box 32. Assessing urban poverty: identifying pockets and hotspots To identify poverty pockets for the regular CCT program, several steps were undertaken. The Department of Social Welfare and Development (DSWD) NCR regional director notified city mayors that assessments would be conducted in their local government units. The mayors notified the DSWD social workers in their local government units to alert nearly 1,700 barangay captains about the assessment in orientation meetings. (Barangays are the smallest administrative division in the Philippines.) DSWD NCR conducted presentations to barangay captains in each of the 16 cities and 1 municipality in Metro Manila. The captains followed up with a list of 60,000–65,000 poor households disaggregated by purok, sitio, and street (barangays consist of puroks and sitios; puroks are generally more urbanized than sitios). Household validation was conducted by street, purok, or sitio. About 80 percent of cases were assessed; the remainder were deemed to risky to visit and were not assessed in Listahanan. Validation was guided by 10 criteria provided by the National Housing Targeting Office (NHTO), some of which were of limited applicability in urban areas or were difficult to collect at the barangay level. DSWD NCR teams were thus given flexibility to interpret the criteria broadly as a reference. Administrative maps were complemented with maps drawn by the DSWD NCR team. Enumerators refined the maps by sketching and labeling substreets, indicating the houses in the substreets and the number of households per housing unit. Selected areas were surveyed using a 34-question household assessment form. In instances where no members of the household were available for the interview (30 percent of all cases), the teams would revisit the households at set days and times (8 a.m.–5 p.m. on weekends; 5–8 p.m. on weekdays; 4–7 a.m. for the homeless, as described below); the security risks of the evening sessions required enumerators familiar with the area. Household information was encoded through the online data entry application. July 2011 data show that for the NCR, of 697,443 households assessed, 316,823 were found poor based on the 2009 regional poverty line of $423. The eligibility check routines (ECR-1 and ECR-2) were then conducted to check household eligibility. From those deemed eligible, 217,000 households are registered with Pantawid Pamilya. In identifying poverty hotspots in the MCCT program, each local government unit was asked to list areas of major concentration of homeless street families or hotspots. There are 16 such areas in Metro Manila, including several major streets and the shoreline of the Manila Bay. In 2010, a rapid assessment was conducted by the DSWD NCR to validate hotspots. By 2011, however, those families could not be found due to the Metro Manila Development Authority’s policy of cleaning major streets. Consequently, a number of homeless street families tended to hide in secondary and darker streets, parks, and even cemeteries. To reach and enumerate these people, special teams were formed to work between 4 and 7 a.m., as the homeless scatter and wander in the cities during the day. About 3,000 homeless street families were interviewed. The information provided by the NHTO/NCR was complemented with information provided by civil society organizations. Validation was conducted, including focus group discussions and individual interviews. Eligibility verification (about half of those assessed do not meet the criterion of having children aged 0–18) and possible registration in the MCCT program was conducted on the spot in the case of isolated homeless street families. Source: authors’ compilation based on interviews with DSWD-NCR staff Benefit structure Pantawid Pamilya provides cash in exchange for compliance with age-specific health and education conditions. The former are based on a protocol set out by the Department of Health and includes activities such as the following: pregnant women must receive pre- and postnatal care, and birth delivery must be assisted by skilled health personnel; children aged 0–5 must receive immunization; and children aged 6–14 must take deworming pills. Beneficiary households are also required to participate in monthly family development sessions. In addition to program information, these sessions cover a wide range of topics such as responsible parenthood, preparation for disasters, and community participation. Education conditions require that school-aged children (3–18 years old) enroll in day care or primary or secondary school, and attend at least 85 percent of school days per month (the minimum requirement set by the Department of Education). The MCCT program has modified these conditions. Registered households must attend family development sessions more frequently during the initial phases: weekly for the first two months, bimonthly for the next two months, and monthly thereafter. Education conditions are initially relaxed: 161   on the second to fourth month after registration, children must attend at least 60 percent of classes. The target is increased to 70 percent three months thereafter; for the succeeding months, children should maintain at least an 85 percent school attendance rate. Household composition determines benefit size. Transfers are paid to the beneficiary household upon compliance with program conditions. Each household receives a lump sum health grant of $11.00 per month, plus a monthly education grant of $6.60 for each day care and primary school-age child, and $11.00 for each child in high school. A maximum of three children per household are eligible for education grants for a maximum of 10 months per year. In total, CCT program households receive annual transfers ranging from $132 to $462. On average, this level represents roughly 21 percent of the poor’s income or consumption, which is lower than the global average for adequacy (30.6 percent). Part of the relatively low transfer size is explained by the fact that the grant amount is uniform across the country, with the same amount having less purchasing power in urban areas where the cost of living is higher. While its health and education grants are similar to those provided by the CCT program, the MCCT program envisions an intensive process of case management conducted by civil society organizations. This process determines a customized package of benefits and interventions for each household, including matching available resources to family needs. Civil society organizations assess all MCCT household members in order to design options for interventions and facilitate services that meet the needs of each member. The case management process consists of the following steps: intake or psychosocial assessment; goal setting and treatment planning; implementation/accessing; progress monitoring and evaluation; and transition, follow-through, or mainstreaming. The MCCT program offers additional support for housing in the form of balik probinsya ($1,540) for families that want to return to their province of origin to establish a permanent abode, and rental subsidies of $88 per month for a maximum of six months. MCCT beneficiaries also have access to skills training and cash-for-work assistance ($6.60 per day or 75 percent of the prevailing daily wage rate), or micro capital assistance for sustainable livelihoods ($220 per family). The program offers assistance such as psychosocial counseling and referrals to alternative learning systems or acceleration programs and family camps. Grant amounts for these are based on assessments by Department of Social Welfare and Development field implementers. Institutional arrangements Over the last seven years, the Philippines CCT program has matured significantly, establishing a rigorous business process and implementation structure. While key performance indicators show that implementation of Pantawid Pamilya is on track nationally, regional variation exists. Analysis of CCT process monitoring shows that there is no single region that stands out as having the worst performance throughout implementation. Rather, regional performance and ranking vary depending on indicators of implementation processes. Nonetheless, implementation appears to be consistently challenging in two regions: the urban National Capital Region (NCR, or Metropolitan Manila) and the conflict-affected Autonomous Region of Muslim Mindanao (ARMM). CCT program implementation appears to face specific challenges in Metro Manila. For example, overidentification of the poor is a potential problem here. At the national level, the number of poor identified by Listahanan was very close to official estimates. In the NCR, the number of poor households identified by Listahanan was almost twice the official estimate as based on survey data 162   (figure 44). Enrolling in the CCT program is also challenging in the NCR. At the national level, most eligible households are registered with the program (87 percent); in the NCR, the take-up rate was the lowest nationwide—15 percentage points below average, or 72 percent (figure 45). These low rates may be due to a high incidence of eligible households not registering, no longer residing in the area as of the registration period, or duplication in enumeration. High mobility among poor households in the NCR could be a significant reason for difficulty in keeping track of households. Figure 44. Share of poor households as Figure 45. Take up rate for the CCT program identified by the Listahanan (%) (%) 200 140 Number of Poor Identified by NHTS‐PR / 150 Number of Poor (NSCB estimates) 120 Take up rate (registered HHs over 100 eligible HHs) 100 80 60 50 40 20 0 0 4B 11 12 10 4A CARAGA ARMM 7 6 CAR 2 1 3 8 5 9 NCR AR… CA… NCR CAR XI V XII IV‐A III IX IV‐B VI X VIII I II VII Region Region Source: DSWD (2013a), NSCB (2009) Source: DSWD (2013b) Monitoring compliance in urban areas is also a challenge. Based on monitoring results at the national level, from November–December 2013, about 89 percent of children aged 0–5 complied with health conditionalities. The health compliance rate in the NCR for children under five was as low as 73 percent—the second worst in the country—even though the NCR’s supply-side indicators show that 91 percent of its health facilities meet standards, compared to 84 percent at the national level. Several factors can explain this performance notably that the NCR has the nation’s highest share of eligible children (6 percent) who are not registered with any health facility for proper monitoring. Also, residents in urban areas often have access to multiple health centers and use services from different providers. The program is not currently designed to track services accessed outside the facilities where beneficiaries are registered. In terms of MCCT program performance, it is still in its pilot phase and subject to continuous design and implementation reviews. Thus, no evaluation or assessment has been conducted. The MCCT Homeless Street Families pilot is relatively small scale, reaching around 600 homeless families in the NCR. These families mostly originate from poor regions and have low levels of education (often less than elementary); most have no source of livelihood. About four out of five are male-headed households, and the rate of long-term homelessness—defined as having been street dwellers for more than 10 years—is about 36 percent. Initial implementation assessments highlighted a number of challenges. Fewer than half of the beneficiaries availed themselves of the six-month housing rental provision. Of those who did, only 30 percent were able to continue renting a house through their own efforts after the program ended. Some 20 percent of households returned to the streets due to their inability to sustain themselves, conflicts with landlords, or delayed provision of cash-for-work activities. A limited fraction of beneficiaries accessed other services: for example, about 32 percent 163   accessed cash-for-work activities; and 16 percent received skills training126, 9 facilitation of employment, and provision of capital assistance. Lessons learned Poverty characteristics differ between urban and rural areas; thus, different data collection strategies to identify the poor through the national household targeting system are in order. In conducting household data collection in urban areas, it is critical to correctly identify pockets of poverty, cover more of them, and allow more time for ODA. Based on lessons from the first round, the second round of Listahanan incorporates more spatially differentiated characteristics in the PMT models. Tracking a mobile population remains a challenge, particularly in urban areas. Currently, Listahanan does not fully accommodate real-time updates, as its mandate is to conduct a large-scale recertification every four years. Improvements have been made, and updates are currently being processed through the grievance redress system. The question remains as to how to handle continuous updates, as households are mobile and their welfare situation changes continuously. As large national programs use Listahanan—including the CCT program and the national indigent health insurance program—it has become increasingly relevant to facilitate updates and changes to household information, including residency. Implementation of the CCT program can be effectively adjusted to urban settings while retaining its integrity as a national program. CCT program design is uniform across the country, but local contexts are accommodated in program implementation. Various examples in the NCR demonstrated how implementation can be adjusted to urban settings, including scheduling shorter family development sessions. To fine-tune the MCCT program, the monitoring and review process should be strengthened. The process of monitoring households that complete the program is not fully established, and the results need to be reviewed thoroughly. Given the chronic nature of urban poverty that characterizes MCCT beneficiaries, exposure to longer-term interventions may be important in helping address those structural challenges. Although the Philippines has an established poverty targeting system and the CCT program as a core social protection program, no one system or program can capture the entire poor and vulnerable population. To deal with these limitations, different approaches beyond the CCT or MCCT must be explored to craft complementary programs and interventions. Finally, clear communication is necessary. The public tends to confuse the MCCT and the CCT initiatives, even though these instruments are designed to serve different groups and have diverse target beneficiaries and implementation procedures. The advantages and limitations of the CCT program need to be clearly communicated to the public to manage expectations, including in terms of building long-lasting pathways out of poverty.                                                                126 Skills training covered soap and fabric conditioner making, candle making, meat processing, and peanut butter making; beauty and nail care; cosmetology; food preparation; electrical installation and maintenance; and housekeeping. Also, beneficiaries were referred to construction companies as masons and welders, to the Manila Manpower Development Center (which offers skills training in hotel and restaurant services, among others), and to Manila’s public employment assistance services. 164   4.10 Program Keluarga Harapan in urban Indonesia127 Rationale In 2007, the government of Indonesia launched Program Keluarga Harapan (PKH), a conditional cash transfer program initially covering about 388,000 poor households and with a budget of $79 million. By 2014, PKH covered 3 million poor households with a budget of $441 million (about 0.05 percent of gross domestic product). PKH was introduced along with other social assistance reforms after the partial removal of general fuel subsidies in 2005, when a portion of the savings from subsidy cuts was reallocated to a set of targeted safety net programs. A temporary unconditional cash transfer program covering poor and vulnerable households was introduced to compensate for increased fuel prices. PKH was introduced as a more permanent solution, aimed at halting the intergenerational transmission of poverty over the long term. Compared to other safety net programs operating in urban and rural areas—notably, Bantuan Siswa Miskin (BSM), Jamkesmas, and Raskin—PKH’s coverage is modest128. Of the national programs considered, Raskin, which provides rice at subsidized prices, has the highest coverage, reaching about 80 percent of the poorest decile (figure 46). The Jamkesmas program, a health care fee waiver for the poor, reaches about half of the poorest 10 percent of the population; while BSM, a scholarship program for poor students, benefits about 15 percent of the lowest decile. Coverage of PKH includes slightly over 6 percent of the bottom 10 percent; coverage is somewhat higher in urban areas. PKH targets the poorest 10 percent. Poverty is assessed through a proxy means test method, the results of which are included and ranked in a national unified database129. In terms of outreach and communication, dissemination of information on PKH has been rather limited. The program has recently improved this situation, producing and disseminating advocacy videos and materials more widely130.                                                              127 The discussion largely draws from Fernandez (2014) produced for this review. 128 Indonesia has had several other large-scale programs in recent years, including unconditional cash transfer programs. Bantuan Langsung Tunai, operating in 2005–08, was Indonesia’s largest program, covering nearly one-third of all households (19 million) and transferring $10 each month per household. In 2013, a new unconditional cash transfer program was implemented—Bantuan Langsung Sementara Masyarakat—which reached 15.5 million households with a monthly transfer of $12 per household over five months. 129 The unified database is the result of a massive data collection effort in 2011, during which 25 million households (around 42 percent of the population) were visited. The surveyed households were selected through a mix of geographic targeting at the district level, census-based poverty mapping, and community-based methods such as peer referrals. Information on household demographics, housing characteristics, and asset ownership was collected. These were then scored with proxy means test methods. Indonesia uses distinct proxy means test models for each of its 500 districts to account for local variations. This approach has been found to increase accuracy significantly relative to provincial urban- rural models. 130 The Ministry of Communication is responsible for overseeing and managing PKH dissemination activities and has been tasked with diversifying and intensifying the spread of information materials through various actors at multiple levels. The PKH central office also has a team in charge of the dissemination strategy. 165   Figure 46. Coverage of KPH and other national safety nets Targeting Eligible PKH participants must possess at least one of the following features: be a pregnant or lactating woman, have children younger than five years old, have children aged 6–15 who are attending primary or middle school, or be a youth aged 16–18 who has not completed basic education. Disbursement of PKH cash transfers is made after verification that the household has fulfilled specific conditions131.4 Program conditions do not vary for urban-rural contexts; also, targeting outcomes do not show particular spatial variations (figure 2). With 51 percent of its urban beneficiaries and 46 percent of its rural in the poorest 20 percent, PKH shows the best targeting outcomes (beneficiary incidence, or share of actual out of total beneficiaries) of Indonesia’s safety net programs.                                                              131 To receive PKH benefits, pregnant or lactating women must complete pre- and postnatal care visits and take iron tablets during their pregnancy, and be assisted by a trained professional during the birth. Children aged 0–5 years must have complete immunization according to health protocols and take vitamin A capsules twice a year, and be taken for monthly or quarterly growth monitoring checkups depending on their age. Children aged 6–15 must be enrolled in primary/secondary school and attend at least 85 percent of school days. Youth aged 16–18 must be enrolled in an education program to complete nine years of schooling.  166   Figure 47. Beneficiary incidence of KPH and other national safety nets Benefit structure From 2007 to 2013, the benefit size defined at the beginning of the program remained unchanged. The yearly transfer per household ranges from $68 to $237, depending on household conditions and composition, with an average of $152. PKH payments represented about 15 percent of the average expenditure of beneficiaries in the bottom decile in 2014. Households also receive a fixed amount ($25) regardless of their compliance with conditions. Cash transfers are disbursed quarterly through direct distribution at post offices132. The value of transfers does not differ between rural and urban areas, even though the cost of living in urban areas is higher, as reflected in a higher urban poverty line and higher average expenditures. Qualitative research conducted in some program areas showed that the urban poor need more cash to meet their daily needs, in addition to covering the costs of meeting program conditions. As program implementation has progressed, payment frequency has been improved, including synchronizing the fixed amount of the benefit ($25) with the beginning of the school year. Alternative payment mechanisms are being considered, including banking facilities for PKH beneficiaries; pilots are being conducted featuring different electronic payment mechanisms.                                                              132 The vast majority of PKH payments (and most other social assistance payments) are distributed through the network of post offices, and most are handled manually. In 2011, a pilot was conducted using savings accounts through BRI Bank. Due to high transaction costs (time, activation requirements, distance, etc.), it has been difficult to scale up. In October 2014, a pilot using electronic funds was launched in four urban and rural districts. 167   The initial cohort of 2007 beneficiaries was expected to graduate out of the program after six years, with those households that had moved out of the bottom decile being incorporated into other safety net programs. However, a recertification survey conducted in 2013 revealed that 71 percent of rural and 72 percent of urban beneficiary households remained poor. A transition phase was designed to extend the participation of poor households for an additional three years, while providing them with a range of integrated interventions. These interventions notably included a series of training modules to be delivered at monthly family development sessions providing guidance to encourage early childhood education, good parenting practices, management of household income, enhanced business planning, and better health and nutrition practices. Implementation of the training program began in 2014133. Institutional arrangements PKH has spent between 14 and 18 percent of its total program costs on administrative expenses134. The unit managing the program within the Ministry of Social Affairs is relatively small in terms of staff and infrastructure as compared to the program’s total number of beneficiaries135. Division of functions is not specialized, and staff manage the entire range of operational functions. At the local level, limited operational capacities and weak coordination between the central Ministry of Social Affairs and the local offices have prevented systematic integration of PKH with other safety net programs serving the same beneficiaries. The program’s aim of improving access to basic health care and primary education has only been partially met. Spot checks showed that PKH beneficiaries did not optimally utilize health services for their under-five children. Immunization rates were generally low, although higher in urban areas due in part to better stocks in health centers. The lack of coordination between the education and health ministries with regard to PKH has contributed to this challenge (for an example of combining safety nets and education, see box 1). Box 33. The poor student allowance program: a case study from Jakarta In 2013, the city of Jakarta launched Biaya Personal Siswa Miskin through the use of a card called Kartu Jakarta Pintar. The objective was to increase enrollment rates of poor students by providing a personal allowance to help cover out-of- pocket expenses. For 2014, the program’s expenditure totaled $93 million and covered 332,465 primary through high school students. The implementing agencies are Bank DKI and the Ministry of Education, which oversee delivery of the cards to schools in Jakarta. The program was developed to address a major drawback in the BSM program: that the transfer represented only a very small amount of the actual costs of attending school. Poor students often remained unable to purchase uniforms and textbooks. The Kartu Jakarta Pintar essentially acts as an automated teller machine (ATM) card for Bank DKI. The government tops up the card with an annual amount of $177 for primary, $211 for middle, and $244 for high school students, disbursed every three months. The initiative is unique to Jakarta; it does not currently exist in other urban areas.                                                              133 A pilot of individual family development sessions will be undertaken to assess both family capacity and productivity in order to identify the most appropriate assistance programs for transitioning PKH families. The pilot looks to contract with civil society organizations to implement support programs for PKH families in functional literacy and numeracy, livelihood promotion, and legal identity. Additional tailored assistance will be delivered via mentors and coaches to a small percentage of households. 134 Administrative costs vary by year depending on coverage. In 2010, these costs represented about 14 percent of total program costs. 135 About 60 staff members manage the program at the central level. In comparison, about 200 staff implement the Philippines’s conditional cash transfer program at the central level—a program similar in size to PKH. 168   Dissemination activities have been rather weak with regard to the various program stakeholders. The lack of information about PKH and how it relates to Indonesia’s social protection system has led to low levels of access to the services for which PKH households are eligible. In 2010, only 69 percent of PKH recipients were found to have access to the Jamkesmas program; 18 percent reported they were receiving BSM benefits. Many officials in relevant sectors at the district and city levels considered dissemination and socialization duties an additional burden because they had not been involved with PKH from the start. In some cases, PKH participants were rejected from receiving program services because they were not enrolled in Jamkesmas. The establishment of the National Team for Acceleration of Poverty Reduction in late 2011 was an important step, as it served to oversee all safety nets and to produce evidence to improve them136. And, with the introduction of the social protection card system for poor households, the fragmentation of social programs was poised for improvement137. The unified database and the social protection card have consolidated existing programs and unified data management systems to improve targeting and beneficiary identification. Over time, there have also been improvements to the PKH management information system; also, some program processes have been improved, such as verification forms and delivery and retrieval mechanisms. Institutional strengthening is still needed on three fronts: (i) strengthening institutional capacity and creating specific units with teams dedicated to handling PKH’s main operational processes; (ii) better involving local governments in accordance with an organized road map and with clear responsibilities with regard to the program; and (iii) increasing the supervision of processes performed at the local level such as verification of conditionalities, payments, and facilitation. Monitoring and evaluation PKH has incorporated monitoring and evaluation since its inception. When the program was designed, a combination of operational spot checks, impact evaluation, and operational monitoring was put in place. A series of spot checks to monitor operations and the capacity of facilitators and related service providers was undertaken by several Indonesian universities in 2009 and 2010. These operational assessments were very useful in detecting bottlenecks in implementation, and the findings were the basis for some of the improvements the program has since undertaken. A randomized controlled trial design was made to assess program impact. Specifically, a baseline survey was executed in early 2007 with participating and nonparticipating households. These households were surveyed again in late 2009, producing panel data to evaluate program impact138.                                                              136 The National Team for Acceleration of Poverty Reduction (Tim Nasional Percepataan Penanggulangan Kemiskinan, TNP2K) was created under the Vice President’s Office to promote coordination, monitoring, and support of Indonesia’s social assistance programs, and livelihood and community programs; as well as to promote financial inclusion initiatives. 137 The social protection card was issued in July 2013 to identify the 15.5 million poor and vulnerable households that comprise the lowest 25 percent socioeconomic status in Indonesia. The card—which must be shown every time the household requests social protection benefits or services—has a bar code, a unique identifying number, and information (names, address) about the household and its members. It can be used to receive benefits under both Raskin and Bantuan Langsung Sementara Masyarakat.  138  The findings show a 10 percent increase in average monthly expenditures over preprogram levels. This increase was used mainly to buy high-protein foods and cover health costs. Clinic visits increased by 3 percentage points, and child growth monitoring by 5 percentage points. Women’s pre- and postnatal visits were 7–9 percentage points higher than in 169   Operational monitoring is managed by the PKH office at the central and district/city levels. Monitoring and evaluation are based on data entry from PKH facilitators and surveys conducted on relevant aspects of the program cycle. Among those indicators continuously monitored are compliance rate, grievances received by type and content, disbursement, and timing of disbursement. Lessons learned While PKH’s design and implementation are not substantially different in urban settings as opposed to rural, there are important features that can shape the opportunities and challenges in these different contexts. For example, anecdotal evidence from the unified database suggests that about 20 percent of households initially surveyed in urban areas cannot be found in the same residence within a period of six months. Also, a study conducted on PKH payments found that urban beneficiaries spend about 22 minutes traveling to the nearest post office or pay point (with a cost equal to 4.0 percent of the average PKH transfer), while those who live in rural areas spend 42 minutes in travel time (5.5 percent of the transfer). Qualitative evidence shows that the overall opportunity cost of attending monthly meetings in urban areas is higher, since most of the mothers have to work to help their husbands meet basic household needs. Field observations suggest that these monthly meetings need to be scheduled at different times, sometimes at night or during weekends, in urban areas. Such adjustments rely on facilitators proactively accommodating participant schedules. While there are more schools in urban areas, it is more difficult to get a place in a school near areas where the poor live. As a result, beneficiary children attend schools outside their neighborhood, increasing transportation costs. Because schools and beneficiaries are more spread out in urban areas, it takes more time for facilitators to cover schools to verify attendance. Conversely, in rural areas, children of a given village are more likely to attend the same school, making the density of beneficiaries per school higher and easier for facilitators to verify. These differences affect program performance. A 2011 qualitative report found that beneficiary compliance depended significantly on the role of the local PKH facilitator. In the urban areas of central Nusa Tenggara, for example, many facilitators were found to be responsible for very large areas, leading to diverging outcomes. In rural areas, facilitators often oversaw just one village and were reportedly more actively engaged in motivating beneficiaries to comply with program conditions. Although PKH benefits were recently increased, their adequacy remains low compared to the average consumption of the target group. Indeed, PKH transfers do not take into account the significantly higher costs of living in urban areas versus rural ones. However, in urban areas the program’s impact has been greater in terms of promoting positive health behaviors and the use of health services, since the availability and quality of health centers is better than in rural settings. Officially, PKH households may access other social protection programs; in practice, this access is                                                              control areas, and deliveries attended by midwives or physicians increased by 5–6 percentage points in PKH areas. Program impacts were more pronounced in urban areas in Java due to the higher availability, better quality, and proximity of health facilities. Breastfeeding, rates of immunization, and diarrhea treatments increased more in urban areas. The likelihood of mothers from beneficiary households completing four prenatal checkups increased by more than 13 percent over preprogram levels; completing the recommended two postnatal visits increased by nearly 21 percent. The likelihood of children aged 0–5 being taken to local health facilities to be weighed increased by 30 percent, and beneficiary households increased the likelihood of completing their children’s vaccinations by approximately 11 percent. There was no significant impact on school attendance for middle and high school students, and no effect on reducing dropout rates. This result is likely because the transfers did not coincide with the academic school year, so parents did not have the funds when needed. Also, the amount received was not adequate to cover additional required school fees.  170   challenging due to limited coordination, capacity, and information. If PKH cash transfers are to make long-term improvements in health and education, recipients must be able to access other relevant programs during and after their participation in PKH139.12 Daily needs such as transport, housing, and food of PKH participants in urban areas are generally more expensive than for those in rural areas. For instance, the cost of midwife-assisted delivery is two to three times more expensive in urban areas than rural, especially when there is no access to Jamkesmas. Only 14 percent of urban wage earners in the poorest quintile receive health-related benefits. Given the overall differences in cost of living as well as opportunity costs, there could be merits in having differentiated features for urban and rural areas. Practical challenges should not be ignored, including possible implications for administrative capacity and migration patterns. PKH facilitators are often tasked with very large coverage areas, and sometimes the ratio of beneficiaries per facilitator is quite high. In 2013, estimates suggested that to serve 3 million households, PKH would need a minimum of 12,500 facilitators; in 2012, there were 6,700. Better linking facilitators with local staff in relevant education and health offices could help in coordinating actions for beneficiaries. Since this program is centrally financed, such coordination will need to come directly from central ministries to local offices and implementers140. Spot checks show that the poorest households value education but cannot finance the costs of educating several children. This situation is particularly acute when PKH transfers are delivered after the start of the school year. More efforts need to be made to integrate PKH with other programs such as BSM; this is especially relevant at the secondary school level. Another option to promote transition to higher levels could be to revise the structure of benefits in favor of higher education. To maximize the impact of individual programs, a convergence strategy could be designed at the central level around the geographic areas where PKH and other safety net programs are linked to livelihood programs, such as the urban community development programs and the KUBE-PKH initiative141. Local governments would implement the integrated strategy at the local level. Having an integrated strategy in front of local governments and communities, plus a clear way of disseminating it, is an important step in creating awareness of the need for such integration. The involvement of local governments is critical when it comes to dealing with the urban poor. Given that Indonesia has decentralized service delivery to the district level, implementation of safety nets is—to some degree—up to the district leaders142.                                                              139 Evidence suggests that the program fragmentation resulting from not using the same targeting mechanism leads to only a very minimal overlap of the main programs to the poorest 10 percent. Only 1.5 percent of the poorest households are beneficiaries of all three main safety net programs targeted to the poor (PKH, BSM, and Raskin). And, although individual transfers represent a small portion of the average expenditure of the poorest 10 percent (for Raskin, 4 percent; for BSM, 8 percent), simulations show that an integrated PKH, BSM, and Raskin transfer would represent around 24 percent of the poor’s average expenditure. This combined transfer would have a better impact on poverty reduction, as well as increase efficiency and reduce administrative costs. 140 The example, in the Philippines the ministries of education, health, and social welfare meet regularly to coordinate actions involving CCT beneficiaries; and guidelines are transmitted directly to the respective local offices. 141 Kelompok Usaha Bersama (KUBE) is an unconditional cash grant of $847 for 10 PKH recipients for the purpose of setting up a microbusiness. KUBE-PKH is being piloted as a way to promote activation and graduation of PKH beneficiaries during the three additional years of transition. 142 Qualitative research suggests that local leaders usually observe the widespread bagi-rata custom of sharing benefits, thereby altering the targeting of social safety net programs, when they are given the discretion to do so. This can take two forms. Some programs experience bagi-roto—a distribution of benefits among the community, such as sharing Raskin 171   Rates of graduation (28 percent on average) are comparable to international standards, although still limited compared to initial national targets. A comprehensive investigation into those factors that contribute to a household’s graduating or transitioning in urban versus rural areas would provide insights into the effects of PKH. In particular, the urban experience of recertification could provide the basis for redesigning PKH according to the area in which is implemented143. There is significant undercoverage of Indonesia’s safety net programs in informal urban settlements. These host some of the poorest households and require specific approaches beyond the unified database, which contains only people with registered addresses. Attention needs to be given to formal identification, including bottlenecks in registration for large inflows of poor urban migrants144. References ADB (2010) “Social Assistance and Conditional Cash Transfers: Proceedings from a Regional Workshop”. Manila. Ajwad, M. I. (2014) “Making Public Works Inclusive for Urban Residents: the Case of Latvia’s Crisis-Response Social Safety Net Program: A Background Note”. Washington DC. Alderman, H. (2002) “Do Local officials Know Something We Don’t? Decentralization of Targeted Transfers in Albania.” Journal of Public Economics 83(3): 375-404. Alderman, H., Gentilini, U., Gilligan, D., Hoddinott, J. and Karachiwalla, N. (2014) “Designing and Implementing Urban Safety Nets: A Review of Selected Issues and Practices”. World Bank and IFPRI. Washington DC. Mimeo. Almansi, F. (2009) “Rosario’s Development: Interview with Miguel Lofschitz, Mayor of Rosari, Argentina.” Environment and Urbanization 25(2): 339-360 Alsayyad, N. (2012) “The Virtual Square, Urban Space, Media, and the Egyptian Uprising.” Harvard International Review 34(1): 58-63. Alzua, M.L., Cruces, G. and Ripani, L. (2013) "Welfare Programs and Labor Supply in Developing Countries: Experimental Evidence from Latin America." Journal of Population Economics 26(4): 1255-1284. Angelucci, M. (2004). "Aid and Migration: An Analysis of the Impact of Progresa on the Timing and Size of Labour Migration", IZA Discussion Paper 1187, IZA, Bonn. Angelucci, M. (2008) “Love on the Rocks: Domestic Violence and Alcohol Abuse in Rural Mexico” B.E. Journal of Economic Analysis and Policy 8(1).                                                              rice equally. Bagi-giliran is also common, whereby a household benefiting from one program is denied participation in another so other households can take the slot. For example, it is common for PKH households not to receive BSM benefits—even though they are eligible—because local leaders feel they are already receiving social assistance and someone else should get a turn. 143 The results of the recertification survey were just released as this case study was being finalized. Their analysis will help address the point raised on urban graduation. 144 It is estimated that approximately 60 percent of Indonesian children under the age of five do not have birth certificates and that around half of them are not registered as citizens. 172   Ardington, C., Case, A., Hosegood, V. (2009) "Labor Supply Responses to Large social Transfers: Longitudinal Evidence from South Africa." American Economic Journal: Applied Economics 1(1): 22-48. Arimah, B. (2010) “The Face of Urban Poverty: Explaining the Prevalence of Slums in Developing Countries.” UNU-Wider, Working Paper No.2010/30. Helsinki. Azevedo, V., Robles, M. (2010) "Simulating the Impact of Policy Changes in Mexico's Progresa/Oportunidades." Journal of Development Effectiveness 2 (2): 263-286. Azuara, O. (2009) "Does Poverty Alleviation Increase Migration? Evidence from Mexico." Universtity of Chicago, MPRA Paper 17599. Chicago. Baker, J. (2008) “Urban Poverty: a Global View.” World Bank, Urban Papers 5. Washington DC. Baker, J. (2012) “Climate Change, Disaster Risk, and the Urban Poor: Cities Building Resilience for a Changing World.” World Bank. Washington DC. Baker, J. (2013) “Indonesia: Urban Poverty and Program Review.” World Bank. Washington DC. Baker, J. and Schuler, N. (2004) “Analyzing Urban Poverty.” World Bank, Policy Research Working Paper 3399. Washington DC. Barrett, C. (2013) “Food Security and Sociopolitical Stability”. (Ed.) Oxford University Press. Oxford. Barrientos, A., and Nino-Zarazua, M. (2010) "Social Assistance in Developing Countries Database Version 5.0." MPRA Paper No. 26403. Beall, J., Goodfellow, T. and Rodgers, D. (2011) “Cities, Conflict and State Fragility.” LSE, Crisis States Research Centre Working Paper 85. London. Behrens, K. and Nicoud, F. (2013) “Survival of the Fittest in Cities: Urbanisation and Inequality.” The Economic Journal 124(581): 1371-1400. Behrman J., Gallardo, J. Parker, S., Todd P. and Velez, V. (2012) “Are Conditional Cash Transfers Effective in Urban Areas? Evidence from Mexico.” Education Economics 20(3): 233-259. Beneke de Sanfeliú, M., and Acosta, P. (2014) "Evaluation of the Temporary Income Support Program in El Salvador." FUSADES and World bank. San Salvador. Bernal, R. and Fernánde, C. (2013) "Subsidized Childcare and Child Development in Colombia: Effects of Hogares Comunitarios de Bienestar as a Function of Timing and Length of Exposure." Social Science & Medicine 97: 241-249. Bhan, G. (2014) “Urban Social Protection in India?” Indian Institute for Human Settlements. Bangalore. Bhan, G., Goswami, A. and Revi A. (2014) “The Intent to Reside: Spatial Illegality, Inclusive Planning, and Urban Social Security.” In Mathur, O. (ed) State of the Urban Poor Report 2014. Oxford University Press. New Delhi. Bhattacharya, S., Prabhakar, A. and Sun, C. (2014) “Social Pensions Program in Delhi: A Background Note.” World Bank, Washington DC. Blomquist, J. (2014) “Urbanizing India: Preliminary Social Protection Trends and Challenges”. World Bank, powerpoint presentation. Washington DC. 173   Bobonis, G. and Castro, R. (2010) “The Role of Conditional Cash Transfers in Reducing Spousal Abuse in Mexico: Short-Term vs. Long-Term Effects”. Toronto and Mexico City. Bobonis, G., Castro, R. and González-Brenes, M. (2006) “Public Transfers and Domestic Violence: The Roles of Private Information and Spousal Control.” University of Toronto. Toronto. Mimeo. Bobonis, G., González-Brenes, M. and Castro, R. (2013) “Public Transfers and Domestic Violence: The Roles of Private Information and Spousal Control”. American Economic Journal: Economic Policy 5(1): 179-205. Burger, N., Glick, P., Perez-Arce, F., Rabinovich, L., Rana, Y., Srinivasan, S., and Yoong, J. (2012) “Indonesia: Urban Poverty Analysis and Program Review”. World Bank, Washington DC. Camacho, A., Cunningham, W., Rigolini, J. and Silva, V. (2014) “Addressing Access and Behavioral Constraints through Social Intermediation Services: A Review of Chile Solidario and Red Unidos.” Policy Research Working Paper No.7136. Washington, DC: World Bank. Centro Nacional de Consultoria (CNC). (2011) “Evaluacion del Programa Familias en Accion en Grandes Centros Urbanos.” Bogota. Chaudhury, N., Friedman, J., Onishi, J. (2013) “Philippines Conditional Cash Transfer Program: Impact Evaluation 2012.” World Bank, Report No. 75533-PH. Manila. Chetty, R., Hendren, N. and Katz, L. (2015) “The Effects of Exposure to Better Neighborhoods on Children: New Evidence from the Moving to Opportunity Experiment”. NBER, Working Paper 21156. Cambdige. Chicago Council. (2013) “Feeding an Urban World: A Call to Action”. Chicago. Chioda, L., de Mello, J., and Soares, R. (2012) "Spillovers from Conditional Cash Transfer Programs: Bolsa Familia and Crime in Urban Brazil." IZA, Discussion Paper series No. 6371. Bonn. Cho, Y. (2015) “Self-Employment and Small-Scale Entrepreneurship Programs in Urban Areas: A Background Note.” World Bank, Washington DC. Cho, Y., Robalino, D, and Watson, S. (2014) “Supporting Self-Employment and Small-scale Entrepreneurship.” World Bank, Washington DC. Christiaensen, L. and Todo, Y. (2013).“Poverty Reduction During the Rural-Urban Transformation: The Role of the Missing Middle”. World Bank, Policy Research Working Paper No.6445. Washington DC Christiaensen, L., De Weerdt, J. and Todo, Y. (2013) “Urbanization and Poverty Reduction: the Role of Rural Diversification and Secondary Towns.” Agricultural Economics 44(4-5): 435-447. Consultative Group to Assist the Poor. (2010) “Technology Program Country Note: Brazil.” Washington, DC. Corral, L., Gomez, A., Ortiz, A. (2006) "Evaluacion del Impacto del Programa Familias en Accion- Subsidies Condicionados de la Red de Apoyo Social.” Bogota, Colombia. Courtin, A. (2015) “U.S and U.K Cities Join Ranks of Fastest Growing Economies.” Fourth Global MetroMonitor, Brookings. Washington DC. Creti, P. (2014a) “Safety Net Programs in Nairobi’s Slums: A Background Note.” World Bank, Washington DC. 174   Creti, P. (2014b) “The Urban Voucher Program in Gaza: A Background Note.” World Bank, Washington DC. Creti, P. (2014c) “The Voucher Program in Urban Burkina Faso: A Background Note”. World Bank, Washington DC. Dabalen, Andrew, Talip Kilic, and Waly Wane. (2008) "Social Transfers, Labor Supply and Poverty Reduction: The Case of Albania." World Bank Policy Research Working Paper No. 4783. Washington, DC: World Bank. Davies, A., Ternström, B. (2013) “Evaluation of five Humanitarian Programs of the Norwegian Refugee Council (NRC) and of the Standby Roster NORCAP.” Ternstrom Consulting. Oslo. Davila, J. (2009) “Being a Mayor: The View from Four Colombian Cities”. Environment and Urbanization 1(21): 37-57. Davila, J. (2013) “Urban Poverty and Mobility: Lessons from Medellín and Soacha, Colombia.” (Ed.) University College London. London. Davila, L. (2014) “The Oportunidades Program in Urban Mexico: A Background Note.” World Bank. Washington DC. de Brauw, A., O Gilligan, D., Hoddinott, J., and Roy, S. (2015) "Bolsa Família and Household Labor Supply. Economic Development and Cultural Change 63(3): 423-457. Dechausay, N., Miller, C., Quiroz-Becerra, V. (2014) “Implementing a Conditional Cash Transfer Program in Two American Cities: Early Lessons from Family Rewards 2.0.” MDRC. New York. Del Ninno, C. (2003) “Reducing Risk and Vulnerability in Urban Areas: What is the Role of Safety Nets?” World Bank. Washington DC. Mimeo. Demombynes, G. (2014) “Study of Vietnam Ho Khau System.” World Bank. Hanoi. Mimeo. Desai, R. (2010) “The Political Economy of Urban Poverty in Developing Countries. Theories, Issues, and an agenda for Research.” Brookings, Working Paper 20. Washington DC. Deshingar, P., Akter, S., Sharma, P., Farrington J. et al. (2010) "The Impacts of Social Protection on Labour Markets and Migration with Particular Reference to the NREGA.” Overseas Development Institute, London Dobbs, R., Manyika, J. and Woetzel, J. (2015) “No Ordinary Disruption: The Four Global Forces Breaking All the Trends”. McKinsey & Company. London. Dobbs, R., Remes, J., Smit, S., Manyika, J., Woetzel, J., Agyenim-Boateng, Y. (2013) “Urban World: The Shifting Global Business Landscape.” McKinsey Global Institute. Seoul, San Francisco and Shangai. Dudwick, N, Hull, K. Katayama, R., Shilpi, F., and Simler, K. (2011) “From Farm to Firm. Rural- Urban Transition in Developing Countries.” World Bank. Washington DC. Dustmann, C., Glitz, A. and Vogel, T. (2010) “Employment, Wages, and the Economic Cycle: Differences between Immigrants and Natives.” European Economic Review 54(1): 1-17. Eisinger, P. (1997) “Cities in the New Federal Order: Effects of Devolution.” The La Follette Policy Report 8(1): 1-24. Evans, B. (2015) “A Review of Urban-Rural PMT Methods: A Background Note”. World Bank, Washington DC. 175   Fay, M., Cohan, L., McEvoy, K. (2005) “Public Social Safety Nets and the Urban Poor”. In Fay, M. (ed) The Urban Poor in Latin America. World Bank. Washington DC. Fernandez, E. (2007) “Implementing the Urban Reform Agenda in Brazil.” International Institute for Environment and Development 19 (1): 177-189 Fernandez, L. (2012) “Design and Implementation Features of the National Household Targeting System in the Philippines.” Social Protection Policy Note 5, World Bank, Washington, DC. Fernandez, L. (2014) “The Program Keluarga Harapan in Urban Indonesia: A Background Note”. World Bank, Washington DC. Fernandez, L. and Velarde, R. (2012) “Who Benefits from Social Assistance in the Philippines?” World Bank. Social Protection Policy Note 4. Manila. Ferre, C., Ferreira, F. and Lanjouw, P. (2012) “Is There a Metropolitan Bias? The Relationship between Poverty and City Size in a Selection of Developing Countries.” World Bank Economic Review 26(3): 351-382. Fitzgerald, E., Corsellis, T., Vivo, G., and Debomy, S. (2014) “Haiti: Rental Support Program Operational Manual.” World Bank. Washington DC. Foster T., Hope R., Thomas M., Cohen I., Krolikowski A., and Nyaga C. (2012) “Impacts and Implications of Mobile Water Payments in East Africa.” Water International 37(7): 788-804. Foster T., Hope R., Thomas M., Cohen I., Krolikowski A., and Nyaga C. (2011) “Mobile Water Payments in Urban Africa.” Oxford University, Said School of Business. Oxford. Galasso, E. and Ravallion, M. (2004) "Social Protection in a Crisis: Argentina’s Plan Jefes y Jefas." World Bank Economic Review 18(3): 367-399. Galasso, E., and Carneiro, P. (2008) "Lessons from the Evaluation of Chile Solidario." World Bank. Washington DC. Galasso, E., Ravallion, M., and Salvia, A. (2001) "Assisting the Transition from Workfare to Work: A Randomized Experiment." World Bank, Policy Research Working Paper 2738. Washington, DC. Gallaso, E., and Ravallion, M. (2004) “Social Protection in a Crisis: Argentina’s Jefes y Jefas.” World Bank, Policy Research Working Paper 3165. Washington DC. Garganta, S., and Gasparini, L. (2012) “El Impacto de un Programa Social Sobre la Informalidad Laboral: El Caso de la AUH en Argentina.” Universidad Nacional de la Plata, Working Paper 133. Buonos Aires. Garrett, J. (2002) “Livelihoods in the City: Challenges and Options for the Urban Poor." USAID. Washington, DC. Garrett, J. and Ersado, L. (2003) “A Rural-Urban Comparison of Cash and Consumption Expenditure.” IFPRI. Washington, D.C. Garrett, J. and Ruel,M. (2000) "Achieving Urban Food and Nutrition Security in the Developing World. ” IFPRI, 2020 Vision Focus 3. Washington, D.C. Gentilini, U. (2013) “Banking on Food: the State of Food Banks in High-Income Countries.” IDS, Working Paper 415. Brighton. Ghani, S. and Kanbur, R. (2013). “Urbanization and (In)Formalization.” World Bank, Policy Research Working Paper 6374. Washington DC. 176   Gharad, B, Chowdhury, S, and A., Mobarak, M. (2014) "Underinvestment in Profitable Technologies: The Case of Seasonal Migration in Bangladesh." Econometrica 82(5): 1671-1748. Glaeser, E. (2010) “Agglomeration Economics”. (Ed.) National Bureau of Economic Research. Cambridge. González-König, G. and Wodon, Q. (2005) "Do Cash Transfers to Farmers Reduce Migration? Procampo in Mexico”. Universidad de Guanajuato, Working Paper 200501. Guanajuato. Government of Colombia (2007a) "Evaluacion del Programa Familias en Accion: Pilotos Centros Urbanos". Bogota. Government of Colombia (2007b) "Manual Operativo Programa Familias en Accion, 2007-2010". Bogota. Government of India (2011) “Report of the Working Group on Urban Poverty, Slums and Service Delivery System”. New Delhi. Government of Philippines (2009) “Poverty Statistics”. National Statistics Coordination Board. Manila. Government of Philippines (2011) “Operations Manual of National Household Targeting System for Poverty Reduction.” Department of Social Welfare and Development. Manila. Government of Philippines (2012a) "Adoption of the Modified Conditional Cash Transfer Under the Pantawid Pamilyang Pilipino Program." NAC Resolution No.06. Manila. Government of Philippines (2012b) "Annual Per Capita Poverty Threshold and Poverty Incidence among Families by Region". National Statistics Coordination Board, available at: http://www.nscb.gov.ph/secstat/d_income.asp. Manila. Government of Philippines (2013) “NHTS-PR Database of Poor Households: Status Report.” Department of Social Welfare and Development. Manila. Government of Philippines (2014a) "Revised Guidelines on the Implementation of the Modified Conditional Cash Transfer for Homeless Street Families." Memorandum Circular 09. Quezon City, Philippines. Government of Philippines (2014b) “Modified Conditional Cash Transfer for Homeless and Street Families: Status report.” Department of Social Welfare and Development. Manila. Government of Philippines (2015) “Briefer on the DSWD Sustainable Livelihood Program (SLP). Overview of the Program: Self-Employment Assistance – Kaunlaran (SEA-K)”. Manila. Government of Romania (2014a) “Elaboration of Integration Strategies for Urban Marginalized Communities in Romania”. Bucharest. Government of Romania (2014b) “Integrated Interventions Tool: Integration Strategies for Urban Poor Areas and Disadvantaged Communities”. Bucharest. Government of Romania (2014c) “The Atlas of Urban Marginalized Areas in Romania.” Bucharest. Government of the U.S. (2000) “Electronic Benefit Transfer Interoperability and Portability Act of 2000”. Public Law 106-171. 2000. 106th Congress. Washington DC. Government of the U.S. (2003) “Food Stamp Program: Electronic Benefit Transfer (EBT) Systems Interoperability and Portability.” Washington DC. 177   Graham, D. (2007) “Variable Returns to Agglomeration and the Effect of Road Traffic Congestion”. Journal of Urban Economics 62: 103-120. Grosh, M., Del Ninno, C., Tesliuc, E. and Ouerghi, A. (2008) "For Protection and Promotion: The Design and Implementation of Effective Safety Nets" World Bank. Washington DC. Gulyani, S. and Talukdar, D. (2008) “Slum Real Estate: The Low-Quality High-Price Puzzle in Nairobi’s Slum Rental Market and its Implications for Theory and Practice”. World Development 36(10): 1916-1937. Gulyani, S., Talukdar, D. and Jack, D. (2010) “Poverty, Living Conditions, and Infrastructure Access: A Comparison of Slums in Dakar, Johannesburg, and Nairobi”. World Bank, Policy Research Working Paper 5388. Washington DC. Gupta, H. (2015) “Home Sweet Home. Housing Practices and Tools that Support Durable Solutions for Urban IDPs.” Norwegian Refugee Council. Oslo. Haddad, L. (2012) “Poverty is Urbanising and Needs Different Thinking on Development”. The Guardian, Development Matters Blog. Post available at: http://www.theguardian.com/global- development/poverty-matters/2012/oct/05/poverty-urbanising-different-thinking-development. Hagen-Zanker, J. and Himmelstine, C. (2012) “How Does Access to Social Protection Programmes Affect the Decision to Migrate?”. ODI, Background Note. London. Hall, S. (2013) “Cash Program Review for IDPs in the Kabul Informal Settlements.” Danish Refugee Council (DRC). Copenhagen. Harari, M. (2015) “Cities in Bad Shape: Urban Geometry in India.” MIT. Cambridge. Harvey, C. (2012) “Cash Transfers in Nairobi’s Slums, Improving Food Security and Gender Dynamics”. Oxfam. Oxford. Haushofer, J. and Shapiro, J. (2013) “Household Response to Income Changes: Evidence from an Unconditional Cash Transfer Program in Kenya”. Princeton University. Princeton. Henderson, V. (2002) “Urbanization in Developing Countries.” World Bank Research Observer 17(1): 89-112. Henderson, V. (2003). “The Urbanization Process and Economic Growth: The So-What Question.” Journal of Economic Growth 8(1): 47-71. Heracleous, M., Gozalez-Flores, M., Winters, P. (2010) “Conditional Cash Transfer and Schooling Decisions: Evidence from Urban Mexico.” Paper presented at 10th Conference on Research on Economic Theory and Econometrics. Milos. Hope, R., Foster, T., Krolokowski, A. and Cohen, I. (2011) “Mobile Water Payment Innovations in Urban Africa”. University of Oxford. Oxford. IDB (2010) “Conditional Cash Transfers in Latina America: Problems and Opportunities.” Washington DC. IDB (2014) “Diagnostico de la Situacion Actual de la Poblacion Urbana en Situacion de Porbreza de Honduras y Recomendaciones de Politicia Publica”. Washington DC. IFRC (2010) “World Disasters Report 2010: Urban Risk” Geneva. Indian Institute for Human Settlements (IIHS) (2013) “Urban India 2011: Evidence”. Bangalore. 178   Ishikawa, A. (2014) “The Effect of Mexico’s Conditional Cash Transfer Program on Migration Decisions”. Duke University. Durham. Jha, S., Rao, V. and Woolcock, M. (2007) “Governance in the Gullies: Democratic Responsiveness and Leadership in Delhi’s Slums.” World Development 35(2): 230–246. Johannsen, J., Tejerina, L. and Glassman, A. (2010) “Conditional Cash Transfers in Latin America: Problems and Opportunities.” IDB. Washington DC. Jolmes, R. and Jones, N. (2010) “Rethinking Social Protection using a Gender Lens”. ODI, Working Paper 320. London. Kariuki, M. and Schwartz, J. (2005) “Small-Scale Private Service Providers of Water Supply and Electricity: a Review of Incidence, Structure, Pricing, and Operating Characteristics”. World Bank, Policy Research Working Paper 3727. Washington DC. Kauff, J, Dragoset, L., Clkary, E., Laird, E., Makowsky, L., Sama-Miller, E. (2014) “Reaching the Underserved Elderly and Working Poor in SNAP: Evaluation Findings from the Fiscal Year 2009 Pilots.” Mathematica Policy Research, report for USDA. Washington DC. Keleher, N. (2008) “Identifying the Urban Poor: A review of Household Poverty in Lagos, Nigeria.” World Bank. Washington DC. Kenny, C. (2012). “In Praise of Slums, Why Millions of People Choose to Live in Urban Squalor.” Foreign Policy online. Available at http://foreignpolicy.com/2012/08/13/in-praise-of-slums/. Khera, R. (2011) “The Revival of the Public Food Distribution System and Possible Explanations.” Economic and Political Weekly 46(44):36-50. Kneebone, E., Nadeau, C., and Berube, A. (2011) “The Re-Emergence of Concentrated Poverty: Metropolitan Trends in the 2000s”. Brookings Institute. Washington DC. Kuztets, S. (1966) "Modern Economic Growth: Rate, Structure and Spread". Yale University Press. New Haven. Lall, S., and Deichmann, U. (2011) “Density and Disasters: Economics of Urban Hazard Risk.” World Bank Research Observer 27(1):74-105. Lall. S., and Timins C., and Yu, S. (2008) “Moving to Opportunity: Successful Integration or Bright Lights?” Background paper for the WDR 2009. World Bank. Washington DC. Latapí, A. and de la Rocha, M. (2004) "Evaluación Cualitative del pPrograme Oportunidades. Centro de Investigación y Estudios Superiores en Antropología Social." Mexico City. Le Bach, D, Tran, L., and Nguen, T. (2011) “Social Protection for Rural-Urban Migrants in Vietnam: Current Situation, Challenges and Opportunities.” IDS, Research Report 08. Brighton. Leitmann, J and Baharoglu, D. (1999) "Reaching Turkey’s Spontaneous Settlements: the Institutional Dimension of Infrastructure Provision". International Planning Studies 4(2): 195-212 Leroy, J., García-Guerra, A., García, R., Dominguez, C., Rivera J. and Neufeld, L. (2008) "The Oportunidades Program Increases the Linear Growth of Children Enrolled at Young Ages in Urban Mexico". Journal of Nutrition 1389(4): 793-798. Levy, S. (2008) "Good Intentions, Bad Outcomes: Social Policy, Informality, and Economic Growth in Mexico". Brookings Institute. Washington DC. 179   Lewis, A. (1954) “Economic Development with Unlimited Supplies of Labor.” Manchester School of Economic and Social Studies 22: 139-91. Lindert, K., Linder, A., Hobbs, J., and de la Brière, B. (2007) "The Nuts and Bolts of Brazil’s Bolsa Família Program: Implementing Conditional Cash Transfers in a Decentralized Context." World Bank, Social Protection Discussion Paper 0709. Washington DC. Lucchetti, L., Rigolini, J. and Vakis, R. (2014) “Left Behind: Chronic Poverty in Latin America and the Caribbean”. World Bank. Washington DC. Lucchi, E. (2013) “Humanitarian Interventions in Situations of Urban Violence” ALNAP-ODI, Lessons Paper. London. Lwasa, S. (2004) “Urban Expansion Process of Kampala in Uganda: Perspectives on Contrasts with Cities of Developed Countries”. University of Makerere. Makerere. MacMaster, B., Escobar, M., Angula R., Monroy M., Tamayo, A. (2013a) “Mas Familias en Accion: Manuel Operativo”. Government of Colombia. Bogota. MacMaster, B., Escobar, M., Angula R., Monroy M., Tamayo, A. (2013b) “Rediseno del Programa Familias en accion.” Centro Nacional de Consultoria (CNC). Bogota. Mani, A., Mullainathan, S., Shafir, E, and Zaho, J. (2013) “Poverty Impedes Cognitive Function”. Science 341(6149): 976-980. Marx, B., Stoker T., and Suri, T. (2013) “The Economics of Slums in the Developing World.” Journal of Economic Perspectives 27(4): 187-210 Masters, W.A, Djurfeldt, A, De Haan, C., Hazell, P., Jayne, T., Jirström, M., Reardon, T. (2013) “Urbanization and Farm Size in Asia and Africa: Implications for Food Security and Agricultural Research”. Global Food Security 2(3): 156-165. Mendoza, R., Mari Chrys P., and Barua Yap, D. (2013) "Including Homeless Families and Children in the Social Protection System: A Brief Review of International Experience and Data on a Philippine Pilot Program". Asia Institute of Management, Working Paper 13-18. Manila. Mohiddin L., Phelps, L., and Walters, T. (2012) “Urban Malnutrition: a Review of Food Security and Nutrition Among the Urban Poor.” Nutrition Works. London. Montalvo, J. and Ravallion, M. (2010) “The Pattern of Growth and Poverty Reduction in China.” Journal of Comparative Economics 38(1): 2-16. Montgomery, M. (2009) “Book Review of World Development Report 2009: Reshaping Economic Geography.” Population and Development Review 35(1): 197-208. Moser, C., Winton, A. and Moser, A. (2005) “Violence, Fear, and Insecurity Among the Urban Poor in Latin America”. In Fay, M. (ed) The Urban Poor in Latin America. World Bank. Washington DC. Mumford, L. (1961) “The City in History: Its Origins, It Transformations, and Its Prospects”. Hardcourt. New York. Murthi, M. (2014) “What Can the EU Learn from Poverty Maps?” Future Development: Economics to End Poverty. Blogpost available at: http://blogs.worldbank.org/futuredevelopment/what-can-eu- learn-poverty-maps Nithin, U., Wang, D., and O’Keefe, P. (2013) "Eligibility Thresholds for Minimum Living Guarantee Programs. International Practices and Implications for China." World Bank. Washington DC. 180   O’Keefe, P. (2004) “Social Assistance in China: An Evolving System”. Paper presented at the ANU conference (August 27). Beijing. Olinto, P. and Sherpa, M. (2014) "Targeting Assessment and Ex-Ante Impact Simulations of Addis Ababa Safety Net”. World Bank. Washington DC. Oliveira, V. (2014) "Food Assistance Landscape FY 2013 Annual Report." USDA, Economic Information Bulletin 120. Washington DC. Packard, T, Koettl, J. and Montenegro, C. (2012) "In from the Shadow: Integrating Europe’s Informal Labor". World Bank. Washington DC. Paes-Sousa, R. (2013) “Plano Brasil Sem Miseria: Incremento e Mudanca na Politica De Protecao e Promocao Social no Brasil”. IPC, Working Paper 113. Brasilia. Pantuliano, S., Metcalfe, V., Haysom S., Davey, E. (2012) “Urban Vulnerability and Displacement: a Review of Current Issues”. Disasters 36(1): S1–S22. Parilla, J., Trujillo, J., Berube, A. and Ran, T. (2015) “Global Metromonitor 2014: An Uncertain Recovery”. Brookings Institute. Washington DC. Perlman, J. (2010) “Favela: Four Decades of Living on the Edge in Rio de Janeiro.” Oxford University Press. Oxford. Perova, E., Reynolds, S. and Muller, M. (2013) “Towards a More Comprehensive Domestic Violence Policy in Brazil”. World Bank. Posel, D., Fairburn, J., Lund, F. (2006) "Labour Migration and Households: A Reconsideration of the Effects of the Social Pension on Labour Supply in South Africa." Economic Modelling 23(5): 836- 853. Qin, G., Zhai, Z., and Garfinkel, I. (2010) "How Does Public Assistance Affect Family Expenditures? The case of urban China." World Development 38(7): 989-1000. Rascon, E. and Rubalcava, L. (2008) “Diagnóstico General de las Barreras a la Acumulación del Capital Humano en Zonas Urbanas en México (Perfil de la Pobreza Urbana)". IDB. Washington DC. Rasella, D., R. Aquino, C. Santos, R. Paes-Sousa, and M. Barreto. (2013) "Effect of a Conditional Cash Transfer Program on Childhood Mortality: A Nationwide Analysis of Brazilian Municipalities". The Lancet 382(9886): 57–64. Ravallion, M. and Chen, S. (2013) “Protection versus Promotion: Incentives and Targeting in a Cash Transfer Program.” Paper presented at the Annual World Bank Conference on Development Economics. Washington DC. Ravallion, M., Chen, S. & Sangraula, P. (2007) “New Evidence on the Urbanization of Global Poverty”. World Bank, Policy Research Working Paper 4199. Washington DC. Ravi, K., Keen, M. and Tuomala, M. (1995) “Labor Supply and Targeting in Poverty-Alleviation Programs”. In Dominique van de Walle and Kimberly Nead (Eds), Public Spending and the Poor. Johns Hopkins University Press. Baltimore. Reardon, T, Tchirley, D., Dolisager, M., Snyder, J., Hu, S., and White, S. (2014) “Urbanization, Diet Change, and Transformation of Food Supply Chains in Asia.” Michigan State University. Michigan. Reardon, T., and Timmer, P. (2014) “Five Inter-linked Transformations in the Asian Agrifood Economy: Food Security Implications.” Global Food Security 3(2):108-117. 181   Renkow, M. (2007) “Cities, Towns and the Rural Nonfarm Economy”. In Haggblade, S. et al. (eds) Transforming the Rural Nonfarm Economy. Opportunities and Threats in the Developing World. IFPRI, Washington DC. Ribe, H., Robalino, D., Walker, I. (2012) “From Right to Reality: Incentives, Labor Markets, and the Challenge of Universal Social Protection in Latin America and the Caribbean.” World Bank. Washington DC. Rivera, L., Hernández, B. and Castro, R. (2006) “Asociación entre la violencia de pareja contra las mujeres de las zonas urbanas en pobreza extrema y la incorporación al Programa Oportunidades.” In de la Paz López M. and Salles V. (eds) El Programa Oportunidades examinado desde el género. UNIFEM. Mexico City. Roberts, B. (2014) “Managing Systems of Secondary Cities: Policy Responses in International Development”. Cities Alliance. Brussels Rodriquez-Alas, C., Acosta, P. and Gentilini, U. (2014) “The Urban PATI Program in El Salvador: A Background Note”. World Bank. Washington DC. Ruel, M. and Alderman, H. (2013) "Nutrition-Sensitive Interventions and Programs: How Can They Help Accelerate Progress in Improving Maternal and Child Nutrition?" The Lancet 382(9891): 536- 551. Sabates-Wheeler, R. and Feldman, R. (2012) “Migration and Social Protection: Claiming Social Rights Beyond Borders”. (Eds.) Palgrave Macmillan. Samad, T., Lozano-Gracia, N. and Panman, A. (2013) "Colombia Urbanization Review: Amplifying the Gains from the Urban Transition”. World Bank, Washington DC. Sambodho, P., Fillaili, R. and Sodo, R. (2013) "Integrating Spatial Aspects of Poverty Into Urban Spatial Planning: Solutions for Tackling Urban Poverty”. MSERU. Jakarta. Sanderson, D. and Knox Clarke, P. with Campbell, L. (2012) “Responding to Urban Disasters: Learning from Previous Relief and Recovery Operations”. ALNAP/ODI, Lessons Paper. London. Sanogo, I. (2014) “The Urban Voucher Program for Syrian Refugees in Lebanon: A Background Note”. WFP. Beirut. Satterthwaite, D. (2007) “The Ten and a Half Myths that May Distort the Urban Policies of Governments and International Agencies”. University College London. London. Satterthwaite, D. and Mitlin, D. (2013) “Reducing Urban Poverty in the Global South”. Routledge. New York. Save the Children (2015) “The State of the World’s Mothers 2015: The Urban Disadvantage.” London. SEASDH (Secreteria de Estado de Assistencia Social e Direitos Humanos) (2014) “Guia Para Gestao Municipal do Programa Renda Melhor”. Rio de Janeiro. Skoufias, E. and Di Maro, V. (2008) "Conditional Cash Transfers, Adult Work Incentives and Poverty." Journal of Development Studies 44(7): 935-960. Soares, V., F., Hirata, G. and Perez Ribas, R. (2010) "The Programa Subsidio de Alimentos in Mozambique: Baseline Evaluation." IPC, Research Brief 14. Brasilia. 182   Spence, M., Clarke Annez, P. and Buckley R. (2009) "Urbanisation and Growth". (Eds.) World Bank. Washington DC. Stecklov, G., Winters, G., Stampini, M., and Davis, B. (2005) "Do Conditional Cash Transfers Influence Migration? A Study Using Experimental Data from the Mexican Progresa Program". Demography 42(4): 769-790. Stegman, M., Lobenhofer, J., and Quinterno, J. (2003) “The State of Electronic Benefit Transfer (EBT).” University of North Carolina. Chapel Hill. Steta de Gandara, M.C. (2013) “Conditional Cash Transfers in Urban Aras: The Case of Mexico’s Oportunidades”. World Bank, powerpoint presentation. Washington DC. Subbarao, K., del Ninno, C., Andrews, C., Rodríguez-Alas, C. (2013) "Public Works as a Safety Net: Design, Evidence, and Implementation" World Bank, Washington DC. Swinkels, R. and Turk, C. (2007) “Poverty Mapping in Vietnam”. In Bedi, T. et al. (eds) More Than a Pretty Picture: Using Poverty Maps to Design Better Policies and I.terventions. World Bank, Washington DC. Tesliuc, E., Pop, L., Grosh, M., and Yemstov, R. (2014) "Income Support for the Poorest. A Review of Experience in Eastern Europe and Central Asia.” World Bank, Washington, DC. Timmer, P. (2007) “A World Without Agriculture: The Structural Transformation in Historical Perspective”. American Enterprise Institute. Washington DC. UNDESA (2014) "World Urbanization Prospects." The 2014 Revision. New York. UNHABITAT (2008) “State of World Cities 2008/09: Harmonious Cities”. Nairobi. UNHABITAT (2013) “State of the World’s Cities 2012/2013”. Nairobi. UNICEF (2012) “State of the World’s Children 2012: Children in an Urban World”. New York. USDA (2013) “A Short History of SNAP”. Washington DC. Vasquez, C. (2014) “The Familias en Accion Program in Bogota: A Background Note”. World Bank. Washington DC. Villarosa, F. (2013) “Productive Inclusion in the Favelas of Rio de Janeiro”. Report prepared for the World Bank. Washington DC. Villegas, W. and Moreno, O. (2009) “Analise de la base de datos del piloto del programa familias en accion Medellin.” Bogota. Wang, D., Glinskaya, E. and Song, J. (2014) “The Dibao Program in Urban China: A Background Note”. World Bank. Washington DC. Wang, D., Song, J., and O’Keefe, P. (2012) “Understanding the Hukou System through Quantifying Hukou Thresholds: Methodology and Empirical Findings.” World Bank. Washington DC. Weir, M. (2012) “Building the Local Social Safety Net in an Era of Fiscal Constraint.” University of California. Berkeley. Weir, M., Wolman, H., Swanstrom, T. (2005) “The Calculus of Coalitions, Cities, Suburbs, and the Metropolitan Agenda.” Urban Affairs Review 40(6):730-760. Willman, A. (2014) "Economic Empowerment and Livelihood Interventions, and SGBV." World Bank. Washington DC. Mimeo. 183   Willman, A. and Corman, C. (2013) “Sexual and Gender-based Violence: What is the World Bank Doing, and What Have We Learnt? A Strategic Review” World Bank. Washington DC. World Bank (1979) “World Development Report 1979”. Washington DC. World Bank (2008) “World Development Report 2008: Agriculture for Development”. Washington DC. World Bank (2009a) “World Development Report 2009: Reshaping Economic Geography”. Washington DC. World Bank (2009b) “Global Experiences on Expanding Services to the Urban Poor.” Washington DC. World Bank (2009c) “Guidance notes on Services for the Urban Poor. A practical guide for improving water supply and sanitation Services.” Washington DC. World Bank (2010) “Egypt’s Food Subsidies: Benefit Incidence and Leakages”. Washington DC. World Bank (2011a) “Protecting Poor and Vulnerable Households in Indonesia: Household Social Assistance” Public Expenditure and Program Poverty Group,” World Bank, Jakarta. World Bank (2011b) “Social Protection in Changing India.” New Delhi. World Bank (2012) “Raskin Subsidized Rice delivery. Social Assistance Program and Public Expenditure Review 3.” Jakarta. World Bank (2013a) “China 2030: Building a Modern, Harmonious and Creative Society”. Washington DC. World Bank (2013b). “Planning, Connecting, and Financing Cities – Now: Priorities for City Leaders”. Washington DC. World Bank (2013c) “Global Monitoring Report 2013: Rural-urban Dynamics and the Millennium Development Goals”. Washington, DC. World Bank (2013d) “Mapping the risk of poverty for small areas in the New EU member states”. Powerpoint presentation. Washington DC. World Bank (2013d) “Urbanization beyond Municipal Boundaries. Nurturing Metropolitan economies and connecting Peri-Urban Areas in India”. Washington DC. World Bank (2013e) “Violence in the City: Understanding and Supporting Community Responses to Urban Violence”. Washington DC World Bank (2014a) “Urban China: Toward Efficient, Inclusive, and Sustainable Urbanization”. Washington, DC: World Bank. World Bank (2014b) “Inclusive Cities: Approach Paper”. Washington DC. World Bank (2014c) “Competitive Cities: Reshaping the Economic Geography of Romania.” Washington DC. World Bank (2014d) “Social Inclusion: The Key for Prosperity for All”. Washington, DC. World Bank (2015a) “Ethiopia Poverty Assessment, 2014.” World Bank, Washington DC. World Bank (2015b) “The Growth Challenge: Can Ugandan Cities Get to Work?” Uganda Economic Update, Fifth Edition. Washington DC. 184   World Bank (2015c) “The State of Social Safety Nets 2015”. Washington, DC. Young, A. (2013) “Inequality, the Urban-Rural Gap and Migration”. Quarterly Journal of Economics 128(4): 1727-1785. Zapatero, E. and Villanueva, R. (2014) “The Policy Process of Expanding Safety Nets in Urban Mozambique: A Background Note”. World Bank. Washington DC. 185   Social Protection & Labor Discussion Paper Series Titles 2013-2015 No. Title 1504 Entering the City: Emerging Evidence and Practices with Safety Nets in Urban Areas by Ugo Gentilini, July 2015 1503 Pension Patterns in Sub-Saharan Africa by Mark Dorfman, July 2015 1502 Social Protection in Fragile and Conflict-Affected Countries: Trends and Challenges by Mirey Ovadiya, Adea Kryeziu, Syeda Masood and Eric Zapatero, April 2015 1501 Defining, Measuring, and Benchmarking Administrative Expenditures of Mandatory Social Security Programs by Oleksiy Sluchynsky, February 2015 1425 Old-Age Financial Protection in Malaysia: Challenges and Options by Robert Holzmann, November 2014 1424 Profiling the Unemployed: A Review of OECD Experiences and Implications for Emerging Economies by Artan Loxha and Matteo Morgandi, August 2014 1423 Any Guarantees? An Analysis of China’s Rural Minimum Living Standard Guarantee Program by Jennifer Golan, Terry Sicular and Nithin Umapathi, August 2014 1422 World Bank Support for Social Safety Nets 2007-2013: A Review of Financing, Knowledge Services and Results by Colin Andrews, Adea Kryeziu and Dahye Seo, June 2014 1421 STEP Skills Measurement Surveys: Innovative Tools for Assessing Skills by Gaëlle Pierre, Maria Laura Sanchez Puerta, Alexandria Valerio and Tania Rajadel, July 2014 1420 Our Daily Bread: What is the Evidence on Comparing Cash versus Food Transfers? by Ugo Gentilini, July 2014 1419 Rwanda: Social Safety Net Assessment by Alex Kamurase, Emily Wylde, Stephen Hitimana and Anka Kitunzi, July 2012 1418 Niger: Food Security and Safety Nets by Jenny C. Aker, Carlo del Ninno, Paul A. Dorosh, Menno Mulder-Sibanda and Setareh Razmara, February 2009 1417 Benin: Les Filets Sociaux au Bénin Outil de Réduction de la Pauvreté par Andrea Borgarello et Damien Mededji, Mai 2011 1416 Madagascar Three Years into the Crisis: An Assessment of Vulnerability and Social Policies and Prospects for the Future by Philippe Auffret, May 2012 1415 Sudan Social Safety Net Assessment by Annika Kjellgren, Christina Jones-Pauly, Hadyiat El-Tayeb Alyn, Endashaw Tadesse and Andrea Vermehren, May 2014 1414 Tanzania Poverty, Growth, and Public Transfers: Options for a National Productive Safety Net Program by W. James Smith, September 2011 1413 Zambia: Using Social Safety Nets to Accelerate Poverty Reduction and Share Prosperity by Cornelia Tesliuc, W. James Smith and Musonda Rosemary Sunkutu, March 2013 1412 Mali Social Safety Nets by Cécile Cherrier, Carlo del Ninno and Setareh Razmara, January 2011 1411 Swaziland: Using Public Transfers to Reduce Extreme Poverty by Lorraine Blank, Emma Mistiaen and Jeanine Braithwaite, November 2012 1410 Togo: Towards a National Social Protection Policy and Strategy by Julie van Domelen, June 2012 1409 Lesotho: A Safety Net to End Extreme Poverty by W. James Smith, Emma Mistiaen, Melis Guven and Morabo Morojele, June 2013 1408 Mozambique Social Protection Assessment: Review of Social Assistance Programs and Social Protection Expenditures by Jose Silveiro Marques, October 2012 1407 Liberia: A Diagnostic of Social Protection by Andrea Borgarello, Laura Figazzolo and Emily Weedon, December 2011 1406 Sierra Leone Social Protection Assessment by José Silvério Marques, John Van Dyck, Suleiman Namara, Rita Costa and Sybil Bailor, June 2013 1405 Botswana Social Protection by Cornelia Tesliuc, José Silvério Marques, Lillian Mookodi, Jeanine Braithwaite, Siddarth Sharma and Dolly Ntseane, December 2013 1404 Cameroon Social Safety Nets by Carlo del Ninno and Kaleb Tamiru, June 2012 1403 Burkina Faso Social Safety Nets by Cécile Cherrier, Carlo del Ninno and Setareh Razmara, January 2011 1402 Social Insurance Reform in Jordan: Awareness and Perceptions of Employment Opportunities for Women by Stefanie Brodmann, Irene Jillson and Nahla Hassan, June 2014 1401 Social Assistance and Labor Market Programs in Latin America: Methodology and Key Findings from the Social Protection Database by Paula Cerutti, Anna Fruttero, Margaret Grosh, Silvana Kostenbaum, Maria Laura Oliveri, Claudia Rodriguez-Alas, Victoria Strokova, June 2014 1308 Youth Employment: A Human Development Agenda for the Next Decade by David Robalino, David Margolis, Friederike Rother, David Newhouse and Mattias Lundberg, June 2013 1307 Eligibility Thresholds for Minimum Living Guarantee Programs: International Practices and Implications for China by Nithin Umapathi, Dewen Wang and Philip O’Keefe, November 2013 1306 Tailoring Social Protection to Small Island Developing States: Lessons Learned from the Caribbean by Asha Williams, Timothy Cheston, Aline Coudouela and Ludovic Subran, August 2013 1305 Improving Payment Mechanisms in Cash-Based Safety Net Programs by Carlo del Ninno, Kalanidhi Subbarao, Annika Kjellgren and Rodrigo Quintana, August 2013 1304 The Nuts and Bolts of Designing and Implementing Training Programs in Developing Countries by Maddalena Honorati and Thomas P. McArdle, June 2013 1303 Designing and Implementing Unemployment Benefit Systems in Middle and Low Income Countries: Key Choices between Insurance and Savings Accounts by David A. Robalino and Michael Weber, May 2013 1302 Entrepreneurship Programs in Developing Countries: A Meta Regression Analysis by Yoonyoung Cho and Maddalena Honorati, April 2013 1301 Skilled Labor Flows: Lessons from the European Union by Martin Kahanec, February 2013 To view Social Protection & Labor Discussion papers published prior to 2013, please visit www.worldbank.org/spl. Abstract Most safety net programs in low and middle-income countries are conceived for rural areas. Yet as the global urban population rapidly increases and poverty urbanizes, it becomes of utmost importance to understand how to make safety nets work in urban settings. This paper discusses the process of urbanization, the peculiar features of urban poverty, and emerging experiences with urban safety net programs in dozens of countries. It does so by reviewing multidisciplinary literature, examining household survey data, and presenting a compilation of case studies from a ‘first generation’ of programs. It finds that urban areas pose fundamentally different sets of opportunities and challenges for social protection, and that safety net programs are at the very beginning of a process of urban adaptation. The mixed-performance and preliminary nature of the experiences suggest to put a premium on experimentation, learning and evidence- generation, particularly in key design choices as well as in better connecting safety nets to spatial, economic and social services agendas in urban areas. About this series... Social Protection & Labor Discussion Papers are published to communicate the results of The World Bank’s work to the development community with the least possible delay. This paper therefore has not been prepared in accordance with the procedures appropriate for formally edited texts. The findings, interpretations, and conclusions expressed herein are those of the author(s), and do not necessarily reflect the views of the International Bank for Reconstruction and Development/The World Bank and its affiliated organizations, or those of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgement on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. For more information, please contact the Social Protection Advisory Service, The World Bank, 1818 H Street, N.W., Room G7-803, Washington, DC 20433 USA. Telephone: (202) 458-5267, Fax: (202) 614-0471, E-mail: socialprotection@worldbank.org or visit us on-line at www.worldbank.org/spl. © 2015 International Bank for Reconstruction and Development / The World Bank