93282 Indonesia Urban Poverty Analysis and Program Review NICHOLAS BURGER,PETER GLICK, FRANCISCO PEREZ-ARCE, LILA RABINOVICH, YASHODHARA RANA, SINDUJA SRINIVASAN, AND JOANNE YOONG February 2012 OPEN PUBLICATION Acknowledgements The authors would like to thank the management and field staff of SurveyMeter, notably Bondan Sikoki, Wayan Suriastini and Dani Alfah. From the World Bank we thank Judy Baker, Fatima Shah,Vivi Alatas, Amri Ilmma and Matt Waipoi. John Strauss and Bondan Sikoki facilitated work with the IFLS data, while Arie Kapteyn and Krishna Kumar provided important feedback and input into the study. All remaining errors are our own. 2 Glossary Abbreviation Bahasa English ASKESKIN Asuransi Kesehatan Masyarakat Miskin Health insurance for the poor Bappenas Badan Perencanaan Pembangunan National Development Planning Nasional Body BIA Benefit incidence analysis BLM Bantuan Langsung Masyarakat Community Block Grants BLT Bantuan Langsung Tunai Unconditional cash transfer BOP Biaya Operasional Operational Funds BPS Badan Pusat Statistik Statistics Indonesia BSM Beasiswa untuk Siswa Miskin Scholarships for the poor CCT Conditional Cash Transfer GDP Gross Domestic Product GFC Global Financial Crisis (starting Fall 2008) GOI Government of Indonesia Jamkesda Jaminan Kesehatan Daerah Local level health insurance for the poor Jamkesmas Jaminan Kesehatan Health Insurance Scheme for the Masyarakat Population JPS Jaring Pengaman Sosial Social Safety Net JSLU Jaminan Sosial Lanjut Usia Social cash transfer for the elderly JSPACA Jaminan Social Penyandang Cacat Berat Social cash transfer for the disabled Kemdiknas Kementerian Pendidikan Ministry of National Education, Nasional MONE Kemenag Kementerian Departemen Ministry of Religious Affairs, Agama MORA Kemenkes Kementerian Kesehatan Ministry of Finance, MOF 3 Abbreviation Bahasa English Kemensos Kementerian Sosial Ministry of Social Affairs, MOSA ND Pembangunan Lingkungan Neighborhood Development NTS National Targeting System OPK Operasi Pasar Khusus Program for sale of subsidized rice for the poor Permukiman Kelurahan Kelurahan Settlement P2KP Proyek Penanggulangan Kemiskinan di UPP Perkotaan PKH Program Keluarga Harapan Hopeful Family Program PL Poverty Line PNPM- Program Nasional Pemberdayaan National Community Mandiri Masyarakat Mandiri Empowerment Program PPP Purchasing power parity Raskin Raskin Beras Miskin Program for sale of subsidized rice for the poor Rp Indonesian Rupiah SA Social Assistance SD Sekolah Dasar Elementary School SMP Sekolah Menengah Pertama Junior Secondary School SP Pocial Protection SSN Social Safety Net SUSENAS Survei Sosio-Ekonomi National Socio-Economic Survey Nasional TKPKD Tim Koordinasi Penanggulangan Regional Poverty Reduction Kemiskinan Daerah Coordinating Team TNP2K Tim Nasional Percepatan National team for accelerating Penanggulangan Kemiskinan poverty reduction 4 Abbreviation Bahasa English UCT Unconditional Cash Transfer UPP Urban Poverty Project (P2KP) Urban Poverty Project (P2KP) 5 Table of Contents I. INTRODUCTION ................................................................................................................. 13 II. ANALYSIS OF URBAN POVERTY ................................................................................. 16 1 Economic Trends, Urbanization and Poverty .................................................................... 16 1.1 The New Order Regime .............................................................................................. 16 1.2 The Asian Financial Crisis ......................................................................................... 16 1.3 Post-AFC Recovery .................................................................................................... 17 2 Data Sources for Poverty Analysis and Benefit Incidence ................................................ 18 2.1 SUSENAS .................................................................................................................. 18 2.2 Indonesian Family Life Survey .................................................................................. 20 3 Methodological Issues in Poverty Analysis ....................................................................... 20 3.1 Choice of Welfare and Poverty Measures .................................................................. 20 3.2 Choice of Poverty Line ............................................................................................... 21 4 Trends and Patterns in Urban Poverty 2002-2010 ............................................................. 22 4.1 Evolution of Poverty................................................................................................... 22 4.2 Poverty Severity and Depth ........................................................................................ 23 4.3 Alternative Poverty Lines and International Comparisons......................................... 23 4.4 Urban Poor as a Share of Total Poor .......................................................................... 24 4.5 Regional Location of the Urban Poor ......................................................................... 24 4.6 Patterns and Trends in Non-Monetary Poverty Measures .......................................... 24 5 Correlates of Poverty – A Profile of the Urban Poor ......................................................... 27 5.1 How Do the Urban Poor Compare To Other Groups? ............................................... 27 5.2 How Does Poverty Vary Across Different Types Of Urban Households? ................ 29 5.3 Multivariate Analysis of Household Consumption .................................................... 29 6 Individual Poverty Dynamics 2000-2007 .......................................................................... 31 6.1 Poverty Transitions ..................................................................................................... 32 6.2 Consumption mobility ................................................................................................ 33 6.3 Correlates of Movements in and Out Of Poverty ....................................................... 34 6.4 Poverty Transitions and Migration to Urban Areas.................................................... 37 7 Vulnerable Urban Sub-Groups .......................................................................................... 38 7.1 Rural-Urban Migrants................................................................................................. 38 6 7.2 Slum and informal settlement residents...................................................................... 40 7.3 Informal Workers ....................................................................................................... 42 7.4 Child Labor and Urban Street Children ...................................................................... 44 8 Qualitative Analysis of Urban Poverty .............................................................................. 44 8.1 Site Selection .............................................................................................................. 45 8.2 Methodology............................................................................................................... 47 8.3 Perceptions of Poverty ................................................................................................ 48 8.4 Are There Different Causes and Consequences of Poverty by Gender? .................... 51 8.5 What Strategies Are Used For Coping with Inadequate Resources? ......................... 53 8.6 What Do The Urban Poor Want from The Government? .......................................... 54 8.7 How Do the Urban Poor Perceive Government Assistance Programs ? .................... 55 III. PROGRAM REVIEW AND BENEFIT INCIDENCE ................................................... 57 1 Overview of Programs Serving the Urban Poor in Indonesia............................................ 57 1.1 Social Protection Programs and Basic Needs ............................................................. 57 1.2 Urban Infrastructure Programs ................................................................................... 59 1.3 Microcredit ................................................................................................................. 60 1.4 PNPM-Urban .............................................................................................................. 61 2 Benefit Incidence Analysis for Selected Programs ............................................................ 61 2.1 Methodology for BIA ................................................................................................. 61 2.2 Programs Examined .................................................................................................... 63 3 Program Coverage by Expenditure Quintile ...................................................................... 64 3.1 Social Protection Programs and Basic Needs ............................................................. 64 3.2 Education and Health Services ................................................................................... 65 3.3 Infrastructure and Basic Services ............................................................................... 67 3.4 Credit for businesses ................................................................................................... 68 4 Progressivity of Programs .................................................................................................. 68 5 Geographical Benefit Incidence ......................................................................................... 71 5.1 Patterns of Benefits in Urban vs. Rural Areas and Across Regions ........................... 71 5.2 Patterns in benefits by area income level ................................................................... 72 IV. CONCLUSIONS ................................................................................................................ 74 1 Summary of Findings ......................................................................................................... 74 2 Implications for Policy....................................................................................................... 77 2.1 Targeting of Social Protection .................................................................................... 78 7 2.2 Improving Access to Education and Health Care for the Urban Poor ........................ 78 2.3 Improving Access to Credit ........................................................................................ 79 2.4 Insuring Households Against Income Shortfalls ........................................................ 80 2.5 Urban Infrastructure ................................................................................................... 80 2.6 Specific Implications for PNPM-Urban ..................................................................... 81 TABLES AND FIGURES .......................................................................................................... 85 APPENDIX 1 SUMMARY OF PROGRAMS SERVING THE URBAN POOR IN INDONESIA .............................................................................................................................. 126 REFERENCES.......................................................................................................................... 134 8 9 Tables Table 8.1 Locations of Study Sites (Non-ND locations) .............................................................. 46 Table II.4.1 Trends in Poverty 2002-2010 .................................................................................... 86 Table II.4.2 Poverty Headcount, Poverty Gap, and Poverty Severity 2002-2010 ........................ 87 Table II.4.3 Urban Poverty by Region, 2010 ................................................................................ 88 Table II.4.4 Trends in Enrollment 2002-2010 (percent) ............................................................... 89 Table II.4.5 Trends in Immunizations and Attendance of Medical Professional at Birth, 2002- 2010 (percent) ............................................................................................................................... 90 Table II.4.6 Trends in Household Access to Water, Sanitation, and Electricity, 2002-2010 (percent) ........................................................................................................................................ 91 Table II.4.7 Trends in Housing Characteristics, 2002-2010 (percent) ......................................... 92 Table II.5.1 Characteristics of Poor and Non-Poor Urban Households, 2010 (percentages unless otherwise indicated) ...................................................................................................................... 93 Table II.5.2 Characteristics of Poor and Near-Poor Urban Households, 2010 (percentages unless otherwise indicated) ...................................................................................................................... 94 Table II.5.3 Characteristics of Extreme Poor and Other Poor Urban Households, 2010 (percentages unless otherwise indicated) ...................................................................................... 95 Table II.5.4 Characteristics of Urban Poor, Rural Poor and Rural non-Poor Households, 2010 (percentages unless otherwise indicated) ...................................................................................... 96 Table II.5.5 Rates of Poverty, Extreme Poverty, and Poverty/Near Poverty by Household Head Characteristic (percent) ................................................................................................................. 97 Table II.5.6 Household Log per Capita Expenditure Regressions, 2002 and 2010 ...................... 98 Table II. 5.7 Household Log per Capita Expenditure Regressions with Community Fixed Effects, 2002 and 2010 ............................................................................................................................... 99 Table II.6.1 Transition Matrices 2000-2007 ............................................................................... 100 Table II.6.2 Correlations Between Annual Per Capita Expenditures and Mobility, with Correction for Measurement Error.............................................................................................. 101 Table II.6.3 Poverty Transitions and Individual and Household Characteristics, Urban 2000 sample ......................................................................................................................................... 102 Table II.6.4 Correlates of Poverty Transitions and Change in Consumption of Urban Households ..................................................................................................................................................... 103 Table II.6.5 Migration and Urbanizing Correlates of Poverty Transitions ................................. 105 The omitted education category is "no school level completed" ................................................ 104 Table III.3.1 Participation in Raskin, Jamkesmas, and Credit Programs by Expenditure Quintile, 2010 (percent) ............................................................................................................................. 106 10 Table III.3.2 Participation in Unconditional and Conditional Cash Transfer Programs by Expenditure Quintile, 2007 (percent) ......................................................................................... 107 Table III.3.3 Net and Gross Enrollment by Expenditure Quintile, 2010 (percent) .................... 109 Table III.3.4 Completed Immunizations and Birth Medic Attendance by Expenditure Quintile, 2010 (percent) ............................................................................................................................. 110 Table III.3.5 Urban wage employees: receipt of job related benefits by expenditure quintile, 2007 (percent) ............................................................................................................................. 111 Table III.3.6 Access to Basic Services by Expenditure Quintile, 2010 (percent) ...................... 112 Table III.3.7 Housing Quality by Expenditure Quintile, 2010 (percent) .................................... 113 Table III.3.8 Urban households: Share reporting being victims of crimes by expenditure quintile, 2007 (percent) ............................................................................................................................. 114 Table III.5.1 Shares of Raskin and Jamekemas Benefits received by poor/near-poor and others by region, 2010 ................................................................................................................................ 115 Table III.5.2 Distribution of Provincial Participation Rates of Urban Poor/Near-Poor in Raskin and Jamkesmas, 2010.................................................................................................................. 116 11 Figures Figure II.4.1 Urban consumption distribution and poverty lines ................................................ 117 Figure II.4.2 Urban Share of Population and Share of Total Poor, 2002-2010 .......................... 117 Figure II.4.3 Trends in Urban Poverty by Region, 2002-2010 ................................................... 118 Figure III.4.1 Concentration Curves for Social Assistance and Credit Programs ...................... 119 Figure III.4.2 Concentration Curves for Education .................................................................... 120 Figure III.4.3 Concentration Curves for Maternal and Child Health .......................................... 121 Figure III.5.1.a Share of poor/near poor receiving Raskin by region, 2010 ............................... 122 Figure III.5.1.b Share of poor/near poor receiving Jamkesmas health card by region, 2010 ..... 122 Figure III.5.2a Share of Urban Poor/Near Poor and Others Receiving Raskin, by Region 2010 (percent) ...................................................................................................................................... 123 Figure III.5.2b Share of Rural Poor/Near Poor and Others Receiving Raskin, by Region 2010 (percent) ...................................................................................................................................... 123 Figure III 5.3a Share of Urban Poor/Near Poor and Other with Jamkesmas Health Card, by Region, 2010 (percent)................................................................................................................ 124 Figure III 5.3b Share of Rural Poor/Near Poor and Others with Jamkesnas Health Card, by Region 2010 (percent)................................................................................................................. 124 Figure III.5.4 Share of Urban Poor/Near Poor and Others Receiving Raskin, by Quintile of Province Median per Capita Expenditure 2010 (percent)........................................................... 125 Figure III.5.5 Share of Urban Poor/Near Poor and Others with Jamkesmas Health Card, by Quintile of Province Median per Capita Expenditure 2010 (perce............................................. 125 12 I. INTRODUCTION Since the Asian Financial Crisis, Indonesia has made significant strides in reducing both rural and urban poverty. However, although urban poverty fell to just below 10% in 2010, adding the vulnerable near-poor essentially doubles this figure. Furthermore, non-monetary indicators such as education and access to safe water have traditionally lagged behind those for other countries of the region with similar per capita incomes. These problems are likely to grow, driven in large part by the rapid urbanization of Indonesia itself. Currently about half urban, with urban areas containing slightly more than a third of the country’s poor, the population is expected to be around 70% urban by 2030, at which point the majority of Indonesia’s poor will likely be in cities. As part of a review of the Program National Pemberdayaan Masyarakat-Urban (PNPM-Urban), the Government of Indonesia’s largest anti-poverty program for urban areas, in 2011 the World Bank commissioned a study that combined a background analysis of urban poverty, a review of current programs that serve the urban poor and a process evaluation of the PNPM-Urban itself. The present document is a draft report of the first two of the three components described above, while the process evaluation of PNPM Urban, has been prepared previously (Burger et. al., 2011). Overview of this report The constraints as well as opportunities facing the urban poor are likely to be significantly different from those facing the poor in rural areas. To date, however, there has not been a study of the urban poor in Indonesia that analyzes trends and patterns in urban poverty, the characteristics of the urban poor and dynamics of poverty, and the effectiveness of programs to address their needs. This document attempts to fill that gap, using both quantitative analysis of secondary data sources and analysis of newly collected qualitative (focus group and interview) data to provide a comprehensive picture of urban poverty in Indonesia. Using nationally-representative household data from SUSENAS and the Indonesian Family Life Survey, our quantitative analysis addresses rates and patterns of urban poverty across Indonesia and over time, characteristics of the urban poor, correlates of urban poverty and household-level poverty transitions. We focus on changes and patterns both in monetary measures of well-being (per capita household consumption) and in a range of non-monetary welfare measures such as schooling, housing characteristics and sanitation, and child immunizations. We examine the evolution of these factors since 2002, thus capturing the period of economic growth following the Asian Financial Crisis of 1998-2001. For reference and perspective, we compare findings with those for rural areas. We also provide a broad overview of government programs that cover the urban poor, before turning to a quantitative benefit incidence assessment of several of the most significant programs currently or recently in place: Raskin (rice for the poor), Jamkesmas (health insurance), BLT (unconditional cash transfers), PKH (conditional cash transfers), and credit programs directed toward the poor (PPK and PNPM). For programs such as Raskin that explicitly target the poor 13 (or more precisely, the poor plus near poor as defined above) we investigate the extent of exclusion and inclusion errors of targeting. We also consider programs such as education and maternal and child health services, and basic infrastructure services such as provision of safe water and electricity as well as community level public infrastructure investments. While these do not target the poor specifically, they nonetheless have important impacts on welfare and poverty, including intergenerational poverty. For these services we consider program coverage or participation by per capita expenditure quintile, or the share of the target population (which could be all individuals or a sub group such as children or mothers) receiving the benefit in each quintile, using formal methods of benefit incidence analysis to consider their progressivity Recognizing that quantitative analysis based on large-scale survey data is often incomplete, the results above are complemented by two other analytical exercises. Firstly, to address the fact that some important marginalized subgroups are not easily captured by household surveys, we provide further discussion of vulnerable urban sub-populations, including migrants, slum dwellers, informal workers, and street children, based on a survey of recent studies. Secondly, to provide a current picture of urban poor and their interactions with government programs, including PNPM-Urban, we conducted a qualitative study based on focus groups and interviews with poor residents in 16 urban kelurahan (wards) across Java, Sumatera, and Sulawesi. The focus groups covered respondents’ perceptions of their own poverty and its causes, and barriers to moving out of poverty; strategies for coping with inadequate resources (both permanent and temporary shortfalls); differences in the causes and impacts of poverty for men and women; perceived needs, including forms of government assistance and participation in and perceptions of assistance programs (including efficiency, fairness, value of benefits, convenience, corruption, etc.) These focus groups and interviews were integrated into the larger process evaluation of PNPM-Urban, and should be regarded as important background context for the findings of that companion study. Organization of the report The report is organized as follows. Chapter II covers the first component of this study, the urban poverty analysis. After a brief discussion of recent economic history and poverty in Indonesia, we discuss our data sources and methodological issues for the poverty analysis. We then turn in sections II.4-II.5 to examine trends and patterns in urban monetary and non-monetary measures of well-being, regional patterns in urban poverty, and the profile of the urban poor (correlates of poverty) as well as near-poor and extreme poor, all using SUSENAS data. In II.6 we investigate the individual dynamics of urban poverty using the 2000 and 2007 waves of the IFLS panel survey. Section II.7 reviews the state of knowledge on potentially highly vulnerable populations in Indonesia’s urban areas—migrants, slum residents, informal workers, street children, and child workers. Finally, in II.8 we present the results of our qualitative study of the urban poor. 14 Chapter III presents the program review and benefit incidence part of the study. We begin in Section III.1 with an overview of programs in Indonesia to benefit the urban poor. We then turn to a discussion of the methods used to analyze the distribution and benefit incidence of programs. Section II.3 analyzes coverage of various programs, and II.4 presents a more formal benefit incidence analysis. Finally, II.5 considers targeting from a geographical perspective. Chapter IV provides a summary of the findings of the study and draws out implications for policy to assist the urban poor in Indonesia. The latter bring together findings from the different parts of the study as well as the PNPM-Urban process-evaluation, exploiting the many complementarities between them. Several key implications for policy emerge from this synthesis. 15 II. ANALYSIS OF URBAN POVERTY 1 Economic Trends, Urbanization and Poverty Indonesia’s recent economic history divides into several very distinct periods of growth and crisis: the long Suharto era of 1965 to 1998 (the New Order Regime); the Asian Financial Crisis (1998-2001), and resumed growth since 2002, punctuated by the recent Global Financial Crisis. 1.1 The New Order Regime The three-decade long New Order Regime was one of sustained economic growth, averaging 6.7 percent per year (Thee 2010). This growth was associated with (and generated by) significant structural change, notably a shift from agriculture and commodities to the industrial and services sectors. Manufacturing was the fastest growing sector during 1975-96, and in the last decade or so of the period (1985-1996) it emerged as the largest sectoral source of employment growth. A series of trade and exchange rate reforms in the mid-1980s led to major increases in foreign and domestic investment in labor-intensive export-manufacturing. The manufacturing share of total exports rose from only 5% in 1981 to more than 50% by 1996. In key respects, therefore, Indonesia was following the successful East Asian development model of export-led growth. The rate and pattern of economic growth led to dramatic reductions in the number of poor in Indonesia, from 54.2 million to 34.5 million between 1976 and 1996, and in the rate of poverty, from 40.1 to 17.7 percent (BAPPENAS, 2006). Accompanying the structural changes in the economy was a steady increase in urbanization, with the urban share of the population increasing from 22.3% to 30.9% over 1980-1990, and to 42.0% by 2000 (Firman et al 2007). Much of this was due to reclassification of rural areas to urban, but these changes also corresponded to shifts to urban (non-agricultural) employment activities (World Bank 2006). Non-monetary measures of poverty also improved dramatically over the period. These were brought about not just by rising household income, but by major increases in public spending made possible by the windfall from oil revenues in the decade following 1973. The period from the mid-1970s to early 1980s was marked by massive investments in education, through school construction, increases in the supply of teachers, and increases in public education expenditures. Enrollment in primary school doubled from 13.1 million in 1973 to over 26 million in 1986, leading to net primary enrollment rates above 90% (World Bank 2006). Public investments in primary healthcare also rose sharply. Among other improvements in terms of access and outcomes, infant mortality rates dropped from over 94 deaths per thousand births in the mid- 1970s to 48 in 1995. 1.2 The Asian Financial Crisis The Asian Financial Crisis of 1997-98 had a devastating effect on the Indonesian economy and on the country’s progress in poverty reduction. GDP dropped 13 percent in 1998. The loss of 16 jobs and a major increase in the price of rice (following a massive currency depreciation) plunged millions back into poverty: the poverty rate jumped sharply to 23.4% in 1999. Evidence indicates a rise in malnutrition as well (World Bank 2006). The severe economic contraction, which was the worst among the countries of the region, led to a political crisis that ended President Suharto’s thirty-two years of rule and precipitated rapid democratization as well as decentralization. 1.3 Post-AFC Recovery The post crisis period (since 2000) has seen a resumption of steady growth of income and reduction in poverty, though at about 5% per year the growth in real income has not matched the rate in the years prior to 1997. Poverty reduction has similarly been steady but slower. The national poverty rate only fell back to its pre-crisis level of 17% in 2003, but with continued moderate progress it declined to 14.2% in 2009 and 13.3% in 2010. The structural pattern of growth was also strikingly different from the earlier period, being driven by commodity exports rather than the manufacturing sector and manufacturing exports; indeed, the latter have been stagnant, reflecting significantly reduced investment in this sector. Growth has also been driven by robust internal demand. These structural differences may reflect a number of factors, including growing competition from the manufacturing sectors of lower wage countries such as Vietnam, ‘Dutch disease’ effects of the rise in commodity prices (and consequent real appreciation of the rupiah); and the restrictions of the Labor Law of 2003 and other high costs of doing business in Indonesia (Wie and Negara 2010; World Bank 2010a). However, a slowdown of manufacturing growth has also occurred in other middle income countries of the region so there is a significant regional dimension to this phenomenon (Hill, Ardiyanto, and Aswichayono 2010). The reduction in manufacturing growth has potentially important implications for changes in poverty, and especially urban poverty. The growth of this sector has been the pathway to improving incomes and escaping poverty in other Asian countries and certainly was a factor in Indonesia’s rapid poverty reduction until the AFC. The weak expansion of this sector since the AFC may conversely explain the slower progress toward poverty reduction in this period, though slower economic growth overall also plays a role. On the other hand, given the sensitivity of global demand for manufacturing exports to income, Indonesia’s relatively low reliance on manufacturing exports, as well as its strong reliance on internal demand, has served to shield it from the impacts of the Global Financial Crisis of 2007-8. Indeed, the Indonesian economy has continued to grow fairly robustly since 2007 (averaging 5.6% per year in 2008-2010). Again, this growth has been driven by both internal demand and strong commodity exports, while manufacturing and manufacturing exports have been stagnant (Hill et al. 2010). Urbanization has continued apace in Indonesia in the years since the AFC. Although one reaction to the crisis was said to be a return of individuals or households to rural areas (and to agriculture), SUSENAS data show a steady rise in the urban share of the population, to about 48% in 2009. This is about the level that would be predicted given Indonesia’s per capita income. The share is projected to be about 70 percent by 2030 (Sarosa 2006). 17 Another important development since the crisis with implications for poverty has been the implementation of a slate of social protection programs that have continued to evolve and expand, with new ones introduced over time. These include Health Insurance for the Poor (Askeskin, now Jamkesmas); subsidized rice allocations (Raskin); unconditional cash transfers (BLT or Bantuan Langsung Tunai, provided in 2005 and again in 2008-9, both times to offset impacts of fuel price increases); conditional cash transfers (PKH or Hopeful Families Program, started in 2007) for very poor households, and several other programs directed at the extreme poor of particularly vulnerable populations such as the elderly and disabled.1 Overall expenditures on social assistance have increased significantly since 2005 and now account for about .5 percent of GDP, although this is still well below the average for developing countries (1.5%) and for other countries in East Asia and the Pacific (1%). (World Bank 2011). An important parallel development in the GOI’s poverty reduction strategy has been the increasing focus on community driven development (CDD). The companion paper to this report (Burger et al. 2011) provides a review of the CDD movement and the variety of CDD programs that have long existed in Indonesia, as well as their expansion and consolidation through the PNPM-Mandiri (National Program for Community Empowerment), starting in 2007. Since 2010 a new governmental body, TNP2K, (National Team for Accelerating Poverty Reduction) has been charged with developing and coordinating the country’s poverty reduction strategy. TNP2K uses a three part framework used to categorize programs and policies: Cluster 1 includes social protection programs such as RASKIN and Jamkesmas; Cluster 2 focuses on empowerment and includes primarily PNPM; and Cluster 3, which is much smaller than the previous two clusters in terms of resources, includes programs for increasing incomes in the longer term via credit for micro and small scale enterprises. These programs are discussed in more detail when we examine the coverage of programs for the urban poor in Part II of this report. 2 Data Sources for Poverty Analysis and Benefit Incidence 2.1 SUSENAS The main data source used in this report is the National Socioeconomic Survey (SUSENAS), a nationally representative annual household survey collected by the Indonesian national statistics agency Badan Pusat Statistik (BPS). There are two main SUSENAS surveys, a cross-sectional survey conducted in July and an annual panel survey carried out in March. The July surveys collect samples of about 250,000 households. The size and design of the survey enables analysis at the kabupaten (district) level. A detailed consumption module is collected in 3-year intervals, the last time in 2008. Smaller consumption modules are administered in the intervening years but because of the difference in the questionnaire, consumption aggregates from these years are not comparable to (and less reliable than) those in the years with the full module. The March survey is an annual panel survey on a smaller number of households, first started by following 10,000 1 See Sumarto and Bazzi (2011) for a comprehensive account of the evolution of social protection policy in Indonesia since 1998. 18 households in the 2002 July SUSENAS for three years. Two subsequent three-year panels have been drawn from the July survey samples. In 2007 the sample size for the March surveys was increased to 65,000 households to allow reliable estimates of poverty at the province (both rural and urban) level. The panel collects detailed consumption data on an annual basis, so also can be used to track poverty year to year. While the July surveys with expanded consumption data initially appear most promising for this report, some inconsistencies preclude the use of this data for conducting disaggregated urban- rural analysis over time. As the March and July surveys are representative and just a few months apart, poverty rates in the two surveys for a given year should be generally be similar, particularly for years when the full consumption module was also administered in the July survey. National poverty rates generated from July surveys using BPS consumption aggregates, sampling weights, and poverty lines generally match the national poverty rates for March. However, with the exception of 2008, the urban poverty estimates from July and March are quite different, and rural poverty rates also differ (the July surveys show much lower urban, and somewhat higher rural poverty than the March survey). Discussions with World Bank staff in Jakarta as well as with BPS staff suggest that different weighting schemes and definitions for rural and urban populations were applied during the collection and analysis of July and March surveys until 2008 (when the approaches were harmonized)2, leading to fundamental inconsistencies in the poverty rates prior to that date. Other issues also remain3 These consultations established that comparisons of July data pre- and post-2008 are not valid. In view of this, we rely on March surveys for our analysis of poverty. It should be noted that use of this panel for analyzing patterns and trends in urban poverty has some drawbacks. First, given the smaller samples, analysis can only be conducted at the province (and rural/urban) level. Second, there are some concerns about representativeness. When following the same households over time, changes in incomes or poverty due to economic conditions are difficult to distinguish from changes due to the life cycle (earnings rise or fall with age and experience, household demographics change). Over the short three-year cycle of each panel sample this is probably a minor concern. There is also a problem of attrition. One attempt to link two consecutive years of the panel found an attrition rate of about 15% of households. Even though BPS replaces lost households with other households in the same cluster, this attrition (or conversely, loss to follow- up) may well be non-random with respect to variables of interest, thus reducing the representativeness of the sample over time. However, bearing these caveats firmly in mind, the March surveys are the best available data for the purposes of examining rural-urban poverty over time, and hence are used for our analyses. This is also consistent with BPS and World Bank practice. 2 We are grateful to World Bank Jakarta staff, and especially to Amri Ilmma, for very helpful discussions about SUSENAS as well as for providing us with the data. 3 This still left some discrepancies unexplained, for example, in some years such as 2004 the urban population shares were the same (43.2%) for March and July, but urban poverty rates were still much lower in July (8.6% vs. 12.1%). A further problem identified by World Bank Jakarta staff is the presence of discontinuities in the per capita consumption distribution at or near the poverty line in the July 2008 data, such that for some districts there are few households just below the line while for others there are few just above the line. Our own examination of the data confirmed these discontinuities, which remain unexplained. 19 2.2 Indonesian Family Life Survey We use the Indonesian Family Life Survey (IFLS) to examine the long run dynamics of poverty among urban households. For such analysis, high quality panel data are required. The IFLS is one of the leading panel surveys from developing countries and, being first fielded in 1993 with the most recent wave in 2007-8, is one of the longest running such surveys. RAND has conducted the IFLS since inception, in cooperation with various Indonesian and U.S.-based collaborators. The 2007 IFLS has a sample size of 13,536 households, of which 7,094 are urban.4 The IFLS takes strenuous measures to re-contact panel households and individuals, including those that may have migrated from their original locations, and as a result has one of the lowest attrition rates of any such panel survey. We are able to match about 90% of individuals across the 2000 and 2007 surveys, which is remarkable given the length of time between rounds. In this respect, it is better suited to understanding individual poverty dynamics than the SUSENAS panels. A disadvantage of the IFLS is that it is not fully nationally representative; only about half of Indonesia's provinces are included in the sampling frame. Still, these provinces account for more than 80% of the population. Those that are not covered in the IFLS are mostly Eastern provinces. These areas tend to be poorer than the covered provinces, which may be considered a drawback for poverty assessment. While there are a few important urban centers (e.g., Ambon, Jayapura) in these provinces, on other hand, they are primarily rural, and thus for assessment of urban poverty, potentially less important. 3 Methodological Issues in Poverty Analysis Analysis of income or monetary poverty – its incidence, patterns, and trends--requires that we (1) have a measure of welfare for the household or individual and (2) define a poverty line such that households below the line are poor and those above are non-poor. 3.1 Choice of Welfare and Poverty Measures The most commonly used welfare measure in the empirical literature is per capita household consumption; the use of consumption expenditures is preferable to income as it is much less variable and is measured with less error. As is well known, simply dividing household total household consumption expenditure by household size ignores economies of scale in household consumption or differences in the age structure of the household (or in adult equivalences) that may affect consumption needs. Still, the simple per capita measure remains overwhelmingly the most common one used in poverty analysis including for Indonesia and was used for example in the World Bank’s 2006 Poverty Assessment (World Bank 2006). Using the given poverty line (discussed below), we then compute the standard poverty headcount measure; the poverty gap index (which measures the mean shortfall of consumption relative to the poverty line); and the poverty severity index (which captures inequality among the poor). A 4 See Frankenberg and Karoly (1995), Frankenberg and Thomas (2000), and Strauss et al. (2004, 2009) for more details about the four waves of the IFLS. 20 well-recognized shortcoming of the simple poverty headcount ratio is that it is insensitive to changes in the depth of poverty. For example, if the consumption of all people below the poverty line were to be cut in half, the headcount ratio would be unchanged. The poverty gap index remedies this by computing the average of the difference between the income of the poor and the poverty line, expressed as a percentage of the poverty line. The absolute value of mean poverty gap times the number of poor also provides an estimate of the amount of resources needed to alleviate poverty via transfers. A limitation of the poverty gap index, in turn, is that it does not distinguish between those who are just poor (just below the poverty line) and those who are very poor (far below the line). The poverty severity index gives more weight to the very poor by taking the square of the poverty gap, i.e. of the distance from poverty line, for each poor person, and averaging this over all poor. When a transfer is made from a poor person to someone who is poorer, this index registers a decrease in aggregate poverty.5 3.2 Choice of Poverty Line The choice of a poverty line may matter substantially for conclusions about the share of Indonesians who are poor, reflecting the level of income or expenditure defined by the line as well as the shape of the distribution of expenditures around the line. BPS defines an absolute poverty line based on the food energy (FEI) intake method. Price per calorie is determined by dividing expenditures on a basket of 52 food items for a reference group (households just above the poverty line) by the number of calories implied by the basket. This is then multiplied by the ‘minimum’ required 2,100 calories to get the cost of meeting this requirement, with a standard adjustment made for non-food items. This is done at the province level every year and this method allows the composition of what is consumed at the poverty line to change each year—and to be different across rural urban areas and across provinces. This method has some disadvantages: it makes comparisons of poverty levels across time (and regions) somewhat problematic, since the change in the poverty line captures both changes in prices and changes in the relative quantities of different items in the basket (see World Bank 2006). An alternative method for constructing an absolute poverty line is the ‘Cost of Basic Needs’ approach. The key difference with FEI is that the same basket is used across regions and over time; the change in the poverty line reflects on change in prices of this fixed bundle, making comparisons across space and time conceptually straightforward.6 Unlike FEI, it requires price information for different regions and periods, but this is available for Indonesia, and several studies by Bank and Indonesian economists (Suryadarma et al. 2005; Pradhan et al. (2000) have created CBN poverty lines with earlier SUSENAS surveys. However, the CBN method requires a disaggregated analysis of SUSENAS consumption data to construct a poverty line, which is beyond the scope of the present analysis. Therefore for our main analyses, we adopt the FEI approach, using poverty lines constructed by BPS with some relatively minor adjustments by 5 The headcount, depth, and severity measures are contained in the well-known Foster, Greer and Thorbecke (FGT) metric (Foster, Greer, and Thorbecke 2004). 6 In fact BPS refers to its method as cost of basic needs rather than FEI. This is strictly true, as both approaches as based on the expenditures required to purchase some minimum levels of food and non-food. However, researchers generally refer to the BPS approach as the FEI method, not CBN. 21 World Bank analysts. This also allows the present analysis to be consistent with prior and ongoing work on poverty conducted by both BPS and the World Bank. For comparison, we also employ alternative poverty lines that are international standards for poverty analysis - a limitation of using the poverty lines calculated by BPS is that they do not provide such internationally comparable measures of poverty. For this purpose the World Bank estimates purchasing power parity (PPP) rates to convert currencies into common dollar terms. Unlike market exchange rates, the PPP exchange rate shows the numbers of units of a country’s currency needed to purchase, in that country, the same amount of goods and services that US $1 would buy in the US. These exchange rates are computed based on prices and quantities for each country collected in benchmark surveys, the last of which were carried out in 2005, including in Indonesia. The standard poverty lines that are generally used for international comparisons are US$1.25 PPP (the mean poverty line for the poorest 15 countries in the 2005 International Comparison Program, hence a measure of extreme poverty), and US $2.00 PPP (See Ravallion and Chen 2008). To update the 2005 $1.25 and $2.00 PPP poverty lines to the 2010 SUSENAS consumption data, we use the CPI data published by BPS. These poverty lines correspond to 224,134 and 358,613 current (2010) Rp per month, respectively; the national BPS urban poverty line is 233,281 Rp per capita per month.7 Finally, we also consider the share of the urban population that is near poor, defined as having per capita expenditures less than 20% above the poverty line. The poor and near poor thus defined constitute the target population for the main social protection programs of GOI. 4 Trends and Patterns in Urban Poverty 2002-2010 4.1 Evolution of Poverty Table II.4.1 shows that poverty incidence has declined in both rural and urban areas of Indonesia since 2002, at the same modest average rate of 0.63 percentage points per year. The decline has been steady with the exception of 2006 (for both rural and urban areas); in that year the rate of poverty actually rose. Two events in 2005-6 may have contributed to the temporary increase in poverty: a reduction in the fuel subsidy in October 2005 that tripled the price of kerosene, and increases in rice price (by 33 percent between February 2005 and March 2006) resulting from a ban on rice imports. Analysis by the World Bank (2006) suggests that the impact of the fuel price increase was largely offset by the introduction of the unconditional cash transfer (UCT) program for poor and near-poor households (which was in fact introduced to reduce the impacts on the poor of fuel price increases). Instead, the main factor behind the 2006 spike in poverty appears to have been the increase in the price of rice, which makes up to about 25% of the consumption basket of the poor. Since a larger share of rural residents than urban would be likely to be net sellers of rice it is somewhat unexpected to see that the percentage increase in poverty was the same in rural and 7 This is the population weighted average of the province urban poverty lines provided by BPS. 22 urban areas (0.018). Since then poverty has resumed its decline, falling an average of 1.3 percentage points per year since 2007 in rural areas and 0.9 percentage points per year in urban areas. These improvements have reduced urban poverty from 15% in 2002 (when poverty still exceeded pre-crisis levels) to just under 10% in 2010. Rural poverty declined from 21.6% to 16.6% in the same period. With similar declines in urban and rural areas, the sizable rural-urban gap in poverty rates - 6.7 percentage points in 2010 - has remained largely unchanged since the start of the period. While the rate of poverty reduction since 2002 and after the AFC has fallen short of that achieved in the several decades prior to the AFC, these sustained reductions represent a substantial achievement that is relatively undiminished by the Global Financial Crises of 2008-2009. 4.2 Poverty Severity and Depth Table II.4.2 presents the evolution of the all three poverty measures for urban and rural areas since 2002. Both poverty depth and severity have declined as the overall poverty rate has fallen. For urban areas, the gap figures suggest that on average that the poor’s consumption levels are currently (in 2010) only 2 percentage points below the poverty line, compared to about 2.5 in 2002, while the decline in the severity index indicates reduced inequality among the poor. These simultaneous developments are consistent with economic growth having affected those at the lower end of the income distribution relatively evenly. 4.3 Alternative Poverty Lines and International Comparisons In urban areas the share of the population living on less than $1.25 per day is 8%, and the share living below $2 per day is 33% (Figure 4.1). The share below the BPS poverty line (the ‘poor’ in BPS’ terms), as already noted, is about 10%. While relatively few urban Indonesians are in extreme poverty by international standards, about a third remain under the $2 standard. Fully 23% of the urban population falls between the national poverty line and this international standard. The near-poor – those with less with consumption than 20% above the national poverty line - form 18.13% of the urban population. Given that the national poverty line is close to the international measure of extreme poverty, it is appropriate to also consider higher thresholds for analysis and policy. International comparisons of poverty are usually made for a country as a whole, and estimates for urban populations alone are not systematically published. However, we may place Indonesia as a whole in perspective by noting that the share below the $2 per day line is 48% for the country overall. From this perspective, Indonesia does not do very well for a middle-income country. This share is about the same as the average for all low and middle income countries combined reported by the World Bank (47%). Regionally, it is high compared to countries with broadly comparable income levels. The Philippines has a GDP per capita somewhat lower than Indonesia’s (3925 vs. 4429 in PPP US$) and a comparable population share is 45%. Vietnam has a per capita GDP of PPP US$3130 but the share is only 38.5%. 23 4.4 Urban Poor as a Share of Total Poor The importance of urban poverty as a policy and social problem depends not just on the urban poverty rate but also on the absolute numbers of urban poor, as well as the share of total poor that is urban. In 2010, there were an estimated 31 million poor in Indonesia of which 11.1 million, or 37%, were urban. The share of urban in total poor has remained at this level— one third or slightly more—since 2002. Figure II.4.2 shows that it has risen slightly since 2004, in keeping with the combination of a constant ratio of rural to urban poverty rates and increasing urbanization (from 43% to 48%) The share of urban poor in total poor in Indonesia, already substantial, will almost certainly rise with the higher levels of urbanization in years to come. 4.5 Regional Location of the Urban Poor The urban poor are concentrated above all in highly-urbanized and densely-populated Java and Bali (Table II.4.3). More than two thirds of the urban poor are found here and these poor also account for about 43% of all poor in this region. Sumatera has about 20% of the urban poor in Indonesia, accounting for about a third of the poor of this region, where only 39% of the population is urban. Other regions are both relatively rural and less populated, so both the share of these regions’ urban poor in total urban poverty, and the share of urban to all poverty with in these regions, are low. Urban poverty varies substantially across regions (as does rural poverty). It is relatively low in Java (under 10%) and higher in Sumatera (12%). The exceptionally high rate of urban poverty for Nusa Tenggara (24%) is noteworthy, though given its low population, this region still accounts for a very small share of all urban poor. This Eastern region overall is very poor, with rural poverty also very high (21.5%) but not higher than urban, in contrast to all other regions. Figure II.4.3 shows a consistent pattern of declines across regions from 2002 to 2010 that are statistically significant in each region other than Papua. Most regions show a temporary increase in 2006 reflecting the impact of the rice price increase. The numbers need to be treated with some caution for years before 2007, since the earlier March surveys had only 10,000 households nationwide; for some regions the urban samples are very small and unlikely to be representative. For the 2007-2010 period (which uses the larger national sample of over 60,000 households), statistically significant reductions in urban poverty rates are seen in every region other than Nusa Tenggara. 4.6 Patterns and Trends in Non-Monetary Poverty Measures This section provides an important complementary analysis of trends in non-income measures of well-being in urban areas. To capture the multiple dimensions of households’ quality of life, indicators of household living standards are examined across critical domains related to human capital (health and education) and basic services (water, sanitation and electrification). The results suggest that urban areas have seen improvements across all these domains over time, although (with the exception of health services) these gains are smaller than those achieved 24 in rural areas, which began the decade with larger deficits. Overall the trends are impressive and important, especially in light of the fact that Indonesia has lagged in non-monetary welfare measures compared with other countries of the region with similar levels of income (Suryahadi et al. 2011). Education and Maternal/Child Health: Table II.4.4 presents the overall trends in current school enrolment for children aged 7-14 from 2002 to 2010. Nationally, enrollment rates have risen steadily and stabilized since 2007, from 92% in 2002 to 96% in 2010. This is primarily as a result of gains in rural areas, as enrollment in urban areas was already close to universal at the start of the period. Current enrolment overall remains slightly higher in urban areas, but the increase in rural enrolment between 2002 (89%) and 2010 (94%) has halved the gap from over 6% in 2002 to less than 3% in 2010. Focusing on urban areas alone, net primary enrollment has remained almost unchanged at about 92% since 2002; there was little room for improvement (this applies equally to rural areas)8. There were marginal improvements in net enrollment at the junior secondary level, to 73% in 2010. However, net upper secondary enrollment increased steadily from 49% to 60% over 2002-2010. Looking across regions in 2010, urban primary enrolments are close to universal (above 95%) in all regions, and urban secondary enrolments are also fairly consistent across the different areas. This is in contrast to rural enrolments, which display considerably more variation and are especially low in Papua and Maluku. Table II.4.5 shows trends in several indicators of health service provision, beginning with rates of full immunization (receipt of complete BCG, Polio, DPT and measles vaccinations) among children aged 12-24 months at the time of the survey (available since 2004). In 2004, only 42% of children in this age group in urban areas were fully immunized (and just 30% in rural areas). These rates increased significantly between 2006-2008, to 65%, and have since remained fairly steady. The change reflects strong overall improvements for DPT and Polio immunizations over the period; however, urban areas lead rural areas by a margin of ten or more percentage points for full immunization rates over the entire sample period. Regional disparities across urban areas in 2010 are fairly marked: the highest coverage of 82% coverage is found in Nusa Tenggara but the lowest is 59%, in Papua. Table II.4.5 also shows the percentage of births attended by a health professional, focusing on births within the last year. Nationally, these statistics have improved from 68% to 84%. While much of this gain is driven by rapid improvements in rural areas, the increases in urban areas are steady and solid: In 2002, 82% of births in urban areas were attended while by 2010, 94% were attended. Looking across urban areas in different regions in 2010, all show a relatively high fraction of attended births - from 86% in Maluku to 100% in Papua. Basic Services: At the household-level, we examine the adequacy of basic services, potable water, sanitation and electricity (Table II.4.6). Nationally, access to safe or improved water 8 Net enrolment is the ratio of the number of children of official school age (as defined by the national education system) who are enrolled in primary school to the total population of children of official school age. For purposes of this study, the definition is based on the official ages of 7-12 for primary, 13-15 for junior secondary and 16-18 for upper secondary. Net enrollment can be less than 100% because of late school entry (so some 7 years olds for example have not started school), and early dropout, so some children 12 or under are no longer attending. 25 (which includes piped water, public tap/standpipe, tube-well/borehole, protected dug well, and protected spring) appears to have increased since 2002 from 78% to 86%. In urban areas the change is from 89% to 95% (and 69% to 78% in rural). We note that the trend may potentially overestimate the increase in access to safe water, particularly for urban areas, as from 2008, the SUSENAS survey introduced a separate response category for ‘branded bottled water’. This presumably safe source is used disproportionately by better-off and urban households but can be included in the safe water group only for the analysis for 2008-2010; in previous years, users would likely have been included under one or another sources in the ‘unsafe’ group. That said, the comparison of urban and rural areas in 2010, which remains valid, shows that the share of households with access to safe water in urban areas is much higher (by 18%) than in rural areas. For sanitation, we focus on access to more advanced forms of waste disposal, toilets and septic tanks. For both these indicators, again, urban growth has been positive although the largest improvements have been posted in rural areas. The vast majority of urban households (91%) in 2002 reported that they had a toilet, and this number has grown to 95% in 2010, with very little regional variation across urban areas. The fraction with a septic tank for waste disposal has also increased, from 64% to 74%. In this case there is regional variation: we find urban areas in Nusa Tenggara are least likely to report having septic tanks (55%), as opposed to 80% or more in Sumatera, Sulawesi and Papua. The table also shows that nationally, rates of electrification have increased from already fairly high initial levels (87% in 2002 to 92% in 2010). In urban areas, there has been virtually no change, as households almost universally reported having electricity in 2002 (98%) and continue to do so over the sample period up to 2010 (99%). Finally, trends in housing quality are shown in Table II.4.7. For most of the period, SUSENAS provides information on the materials used for walls, floors and roof9. We also create an overall index of ‘good quality’ housing equal to 1 if all three characteristics are of good quality. This measure corresponds roughly to the ‘durable’ or ‘permanent’ housing criterion used by UN- Habitat (2010). It is hard from the data to assess the actual quality of materials used. For example, a large share of Jakarta’s housing is thought to be self-built (Mercy Corps 2008), so the use of a given material may correspond to structures that are of good or poor quality overall. Therefore the measure should be interpreted as indicating simply that at best a minimum level of housing quality is met. The table indicates that the quality of housing in urban areas is substantially better than in rural areas: 91% of urban household live in dwellings with at least a minimum level of quality compared with 73% of rural in 2010. The sources of the rural-urban gap are in floor and wall material. For example, in 2010, 17% of rural households have earthen floors compare with about 5% of urban dwellers. Notably, while housing quality remains higher in urban areas, there have been larger improvements since 2002 in rural areas—a pattern seen in many of the other measures of welfare above. 9 We consider ‘good’ and ‘poor’ quality materials to be as follows: for floor, earthen is poor, any permanent covering is good; for walls, brick and wood are good, bamboo poor; for roof, concrete, tile, shingle, iron sheeting, and asbestos are good, palm fiber is poor. 26 5 Correlates of Poverty – A Profile of the Urban Poor 5.1 How Do the Urban Poor Compare To Other Groups? Understanding the characteristics of the urban poor helps policymakers understand the factors that lead to poverty and potentially provides guidance for program targeting. In Table II.5.1 we compare urban poor and non-poor households in 2010 along a range of characteristics. This table and subsequent tables also show tests for differences among groups.10 Occupation: There are clear differences between urban poor and non-poor in terms of the nature of livelihoods. Heads of poor households are more likely to be self-employed (52% vs. 45%) and less likely to be wage employees, indicating a greater importance of informal sector work among the poor.11 Also noteworthy is the breakdown by broad occupation sector. Some 39% of heads of poor households are involved in agriculture and extraction industries – an indication of the semi-rural nature of many areas classified as urban— compared with only 13% for non-poor. Interestingly, the poor are less likely to be in services than the non-poor, though this is not surprising given that the public sector is included under this category. Similar shares of poor and non-poor (22%) are in industry. It is also clear that unemployment (or rather, not currently working) is not a factor that is strongly associated with being poor: about 20% of non-poor and 19% of poor household heads are currently not working. This lack of association is commonly found in developing countries, reflecting the fact that remaining out of work for any sustained period is a ‘luxury good’, something that poor individuals, with less savings or external support, cannot afford. Education: Not unexpectedly, heads of poor households have less schooling than non-poor household heads. 35% of the former have no education certificate, meaning they have less than a completed primary schooling, compared with 14.5% of non-poor heads. 13% of poor heads have a secondary diploma compared with 34% percent of non-poor. Of course, these numbers also indicate that many non-poor household heads also have low education. On average, other adults in the household also have less schooling (6.10 vs. 6.51 years). Demographics: The age of the household head, an indicator of stage in the lifecycle, differs little by poor/non-poor status. On the other hand, poor households are larger, by about 1 person on average (approximately 5 vs. 4), primarily reflecting a larger number of children in poor households. This too is a common pattern and reflects the higher fertility rates among poorer women.12 10 All tests are adjusted for sample clustering, as are regression models later in this report 11 Some analyses treat the self-employment/wage employment distinction as equivalent to the informal/formal sector division. While the large majority of self-employed would indeed be informal workers (using various definitions), many wage workers would also be informal. Unfortunately, the occupational data in the 2010 SUSENAS do not permit a division of wage workers into informal and formal (or for the matter, public and private). 12 Note, however, it can in part reflect the use of per capita consumption as the welfare measure. This tends to overstate the relative well-being of larger households, which tend to have proportionately more children, to the extent that children consume fewer resources than adults. The simple per capita measure treats their consumption as 27 Female Headship: Finally, it is noteworthy that female headship is not higher among the poor than the non-poor and in fact is slightly lower. This goes against a common perception about female headed households. But it is likely that women who lack means will live with others (e.g., extended family) rather than independently, and further, we are not controlling for other characteristics that may be associated with female headship. We examine this factor in a multivariate context below. Table II.5.2 compares characteristics of the urban poor to the urban ‘near-poor’ (non-poor households with per capita consumption less than 20% above the poverty line). As noted, by virtue of their proximity to the poverty line, households in this group are considered vulnerable to becoming poor, and GOI considers this group as well as those below the line as the target population for social protection policies. The near poor make up a non-trivial share of the urban population—8.26% - almost as large as the share of the poor. This makes the group of concern for social protection and policy much larger than considering just those below the poverty line, as we are now considering 18.13% of the urban population. There are some differences between the groups, particularly in terms of occupation. Heads of near poor households are slightly less likely to be self-employed and slightly more likely to be wage employed than heads of poor households. The difference is larger for agriculture/extraction (39 vs. 29%). As we might expect, the differences between near-poor and poor show the same pattern of differences as between poor and all non-poor households, but are less pronounced. We also examine the characteristics of the urban ‘extreme poor’, defined here as those in the bottom 5% of the urban per capita consumption distribution. With an urban poverty rate of about 10%, the extreme poor as just defined will essentially comprise the poorer half of the poor. This group is in the greatest need, and may have different characteristics than somewhat less poor urban households. However, Table II.5.3 shows that the characteristics of the urban extreme poor differ very little from those of the rest of the urban poor – differences in means are generally quite small and not statistically significant. Finally, we briefly compare the urban poor to the rural poor (Table II.5.4). Most of the differences between rural and urban poor are as we might expect. The rural poor are somewhat less educated than their urban counterparts, though mean household size is the same, that is to say, relatively small. Poor rural household heads are somewhat more likely to be working than urban household heads. Unsurprisingly, occupational patterns differ substantially, with the poor in rural areas relying much more heavily on agriculture/extraction than their urban counterparts (78% vs. 39%) and less on industry and services, as well as being more likely to be self- employed (71% vs. 52%) and less likely to be wage-employed. These figures indicate, again not surprisingly, that programs that aim to help the poor in urban and rural areas via their livelihoods need to be designed differently. equal to that of adults, but if it is less, households with more children will be better off than those with fewer children for the same income and household size. 28 It should also be noted that while occupational differences distinguish rural poor and urban poor, they do much less to differentiate rural poor from rural non-poor. For example, 63% of rural non-poor household heads are involved in agriculture and 67% are self-employed (the two categories largely overlap of course), which is only moderately below the figures for rural poor while being far lower than for the urban non-poor. Thus the occupational profiles of rural poor and non-poor are quite similar. Just as in urban areas, in rural areas no single non-income characteristic can be relied on to differentiate the poor and non-poor. 5.2 How Does Poverty Vary Across Different Types Of Urban Households? Next, we reverse the perspective of the preceding analysis and compare poverty incidence for urban households or individuals with different characteristics. This approach is potentially more useful from a targeting perspective as it indicates which characteristics, if any, are particularly strong proxies for income poverty. We do this in Table II.5.4 using the 2010 SUSENAS. It should be recalled that the overall urban poverty rate is just under 10%. Some 20% of households in which the head has less than a completed primary education are poor (and 35% poor or near-poor), compared with 14% for completed primary and less than 4% for secondary. With respect to occupations of the head, poverty rates are very high for agricultural wage and self-employment (23 and 24 percent, more than double the overall rate) and lowest for wage employees in services (4.5%; as noted this includes government employment). The poverty rate for female-headed households, at 9.6%, is essentially the same as for other urban households. Overall, this and the preceding tables point to clear associations of monetary poverty with different characteristics of urban households. While the patterns we observe are generally as expected, taken individually the relations of these factors to poverty are not strong enough to identify the poor and inform targeting. For example, even among households where the head is involved in agriculture or extraction—a characteristic with perhaps the strongest association with poverty among the variables discussed—some 75% percent are not poor. In practice, the GOI has used a combination of such measures for proxy means testing to determine benefit eligibility, which in turn is combined with community-based targeting and geographical targeting; the mix of these approaches varies by program.13 5.3 Multivariate Analysis of Household Consumption Since many of these characteristics discussed above are correlated (e.g., education, occupation) multivariate regression analysis is useful for providing a clearer picture of the relationships between these factors and household well-being. We discuss household log per capita 13 See Poverty Group, World Bank Jakarta Office (2010) and Word Bank 2011 for a description and critique of these approaches. For example, to determine eligibility for the for the 2005 BLT or unconditional cash transfer, BPS carried out a survey gathering information from households on floor type, wall and roof type, toilet facility, electrical source, primary source of income, educational attainment of household head, and other factors. (Sumarto and Bazzi 2011) 29 expenditure functions in urban areas for 2002 and 2010. The regressions estimate the partial correlation coefficient between household consumption and individual household characteristics. While the estimates of these welfare regressions are often assumed to show the ‘determinants’ of consumption or poverty, in fact they can only be said to show associations; causality from explanatory variables to outcomes is difficult to infer as there may be unobserved factors affecting both, or reverse causality (e.g., from income to ill health). Still, when interpreted with appropriate caution, the results provide insight about the factors that lead to poverty and potentially indicate means by which policy can improve the wellbeing of, or opportunities for, the poor. We also note that many location-related determinants of poverty cannot be included in the regressions as they are not measured in the data. These can include local economic conditions and variations in effectiveness of programs to assist poor households. Since these factors may be correlated with included regressors such as education and occupation, inferences about the association of these regressors and consumption may be misleading. Therefore we also estimate community fixed effects models in which both dependent and independent variables are subtracted from their cluster mean values. This differencing eliminates the impacts of any community level factors that enter the undifferenced model linearly and uses within-community variation to estimate impacts (or associations) of education and other characteristics. We focus first on the models using the 2010 SUSENAS. The regression results in Table II.5.6 are largely consistent with expectations as well as the descriptive comparisons of poor and non- poor above. Household expenditure is very strongly associated with education of the head. For occupation of the head, the base category is self-employment in the service sectors. Relative to the base, employment in agricultural or extraction, whether for own account or wage, is associated with lower household consumption. Being self-employed in industry or manufacturing implies higher consumption, while wage employment in industry (much of which is likely informal) is associated with lower consumption. Households with more members, especially children, are poorer. There are also substantial regional differences controlling for household characteristics. Urban households in Sumatera have lower consumption than households with similar characteristics in urban Java, the base category. Households in urban Kalimantan and Sulawesi are better off than similar households in urban Java. A notable difference from the descriptive comparisons of poor and non-poor households is that female headship is negatively associated with consumption after controlling for other factors. Also noteworthy is the quadratic impact of head’s age, reflecting lifecycle effects whereby earnings and consumption rise with age and experience and then fall. The fixed effects estimates (Table II.5.7) do not alter these findings qualitatively though a few parameter estimates differ in magnitude. While returns to primary schooling completion are similar, the coefficients on post primary levels are about 20-25% smaller when controlling for community fixed effects. This suggests that part of the apparent returns to education among urban households reflects economic and labor market conditions where those with post- primary education reside. 30 The comparison of parameter estimates for 2002 and 2010 in both Tables II.5.6 and II.5.7 suggest that the relationship of these factors with urban household consumption has not changed in recent years. Coefficients are quite similar for the two years other than for a few region dummies, with few statistically significant differences. The fixed effects estimates suggest an increase in the returns to secondary and post-secondary education, thought the magnitudes of the changes are modest. In sum, the descriptive analysis and expenditure regressions point to some strong and generally- expected associations of various household characteristics and urban poverty or consumption— or if we are willing to make stronger assumptions about causality, they point to a number of key determinants of urban poverty and consumption. These include education and occupation, as well as (and with less claim to causation) household structure. Hence they shed light on the longer-term characteristics of individuals and households which are associated with being in poverty. At the same time, these standard analytical approaches do not shed much light on the notion of vulnerability, namely, how households become poor or conversely, escape poverty, which we address in the next section. 6 Individual Poverty Dynamics 2000-2007 As described earlier in this study, a large number of individuals in Indonesia are near-poor, a group that is often referred to as at risk of becoming poor because they can easily cross the poverty line when faced with an unfavorable shock. Previous research has found that a large fraction of Indonesia’s population move from non-poverty into poverty, and from poverty to non-poverty (World Bank, 2006; Sumarto and Bazzi 2011). Thus, the fraction that is poor in at least one out of two different points in time is substantially larger than the fraction that is poor when looking at a single point in time. To study transitions in and out of poverty, we require panel data which allow us to observe the same individuals at more than one point in time. In this section, we use the 2000 and 2007/8 rounds of the IFLS to investigate the extent of movement of urban households in and out of poverty over time and to identify the characteristics associated with these movements. As discussed earlier, while the IFLS, unlike SUSENAS, does not cover all areas of Indonesia (though the areas it does cover include more than 80% of the population), the superior performance in terms of reduced attrition make it the better choice for panel analysis: given the very strong efforts at follow-up, attrition is very low in the IFLS and we are able to match close to 90% of individuals across the two surveys despite the seven years between them. We therefore analyze longer-term (over seven years) dynamics rather than more short-term transitions that a year-to-year panel would reveal. Using the IFLS panel data, we (i) construct poverty transition matrices for various types of urban households, (ii) estimate the extent of mobility in consumption, and (iii) estimate multivariate models to understand the association of changes in poverty status with specific household and environmental characteristics. The period under analysis corresponds at least roughly to the period we have been considering for the analysis so far using the SUSENAS March surveys (2002-2010). In a previous study, using the 1993 and 2000 waves of IFLS, McCulloch, Timmer 31 and Weisbrod (2007) find that, among Indonesians living in urban areas who were non-poor in 1993, 17% were living below the poverty line in 2000. This occurred despite an overall reduction in poverty during that time period, during which a far larger share of the sample went from being poor to non-poor. Here, in addition to a more recent period of analysis, we also go further and describe the household and individual characteristics that are associated (possibly causally) with these transitions. 6.1 Poverty Transitions We first establish the prevalence of movements into and out of poverty between 2000 and 2007- 8. This first aspect of the analysis shows the share of individuals changing their poverty status (escaping poverty or falling into poverty) between 2000 and 2007/2008. We create the poverty indicator using the IFLS household consumption data aggregates for these years. To deflate 2007-8 expenditures for comparison to 2000, we use price indices from published BPS price data. While the aggregate price indices for the largest cities in Indonesia would allow us to do this easily, these are based on the average bundle of goods for the population overall and not for the poor in particular. For our analysis, it is more appropriate to use price indices based on the bundle of goods consumed by the poor (Gibson 2007). This is particularly important when the price of food is changing differently than the general price index, since a larger share of the poor’s budget goes into food items. Therefore we use expenditure shares on food and non-food items among the poor (based on the BPS poverty line) as weights for the food and non-food price components of the overall price index. The shares were calculated using the SUSENAS data for 2004, the midpoint of the period we are studying. This share was then used to reweight the price indices, and then deflate the 2007 IFLS data to December 2000 prices.14 The analysis focuses on changes in the per capita consumption of an individual’s household (and poverty, based on household consumption) rather than changes in the individual’s own income. There are two reasons for this. First, it allows us to use data on expenditures rather than income; the latter, as is well known, is considerably more prone to measurement error or temporary variation, and measurement error in the welfare indicator is the key concern in panel studies of income and poverty changes. Second, poverty can only be measured at the household level with standard household survey data, which do not indicate the allocation of total expenditure (or income) among household members. Table II.6.1 shows the poverty transition matrices for 2000-2007. The first panel shows the transition matrix for the sample as a whole (urban and rural), while the second panel concentrates on those individuals who were living in urban areas during the 2000 survey. As can be seen from the first matrix, there is substantial overall movement in and out of poverty. Approximately the same number of people moved into poverty (4,943) as moved out of poverty (4,752) between 2000 and 2007. All told, some 26% of the sample changed their poverty status. In percentage terms, 11.5% of those who were not poor in 2000 were catalogued as poor in 2007, while fully 64% of those who were poor escaped poverty. The latter figure is fairly 14 The prevalence of poverty estimated with the IFLS is lower than with the SUSENAS. This is not surprising because the IFLS does not cover all the provinces, and the provinces that were not included tend to be poor. 32 remarkable, though the likelihood of significant measurement error in consumption guarantees that we are overestimating the movement out of poverty as well as into it. The small percentage of non-poor individuals transitioning into poverty is somewhat misleading. The large majority of observations in 2000 are above the poverty line and most are far from the line, unlike the minority that is below the line; hence only a small share of the non-poor would be expected to cross into poverty, even among those whose consumption fell. A more relevant analysis is to consider the share of the near-poor that fall into poverty. If we consider this group (as in our SUSENAS analysis) to be those with per capita consumption less than 20% higher than the poverty line, we see that 821 of these individuals or 22.5% of the overall group of near-poor transitioned into poverty between 2000 and 2007, suggesting a fairly high risk of falling into poverty for those who are near poor. It might seem surprising that most people that are classified as poor in 2000 were not classified that way seven years later. However, this is consistent with prior studies of earlier time periods. For example, McCulloch et al. (2007) also find that most people described as poor in 1993 did not fall under that category in 2000.15 The second panel shows transitions specifically for the urban 2000 sample, that is, those who were living in urban areas at the beginning of the period. 1,041 individuals become poor (6% of the 2000 non-poor) while 632 (73% of the poor) transitioned out of poverty; all told 33% of the urban sample changed their status. If we consider the group of near-poor urban observations in 2000 as defined above, we see that 17.7% of this group was classified as poor in 2007, Thus, the urban sector experienced a higher rate of transitions overall, with a greater share going from poor to non-poor than in the country as a whole. We calculated the same transitions for the period of 1997-2000 in order to have a point of comparison.16 In general there are somewhat fewer observed movements in either direction over the poverty line (which is natural due to the shorter period of comparison). Overall, 57% of the poor left poverty (compared to 64% in 2000-2007) and 10.6% became poor (compared to 11.5% in 2000-2007). The ratio of those becoming poor to those leaving poverty was higher in 1997- 2000 than 2000-2007, which is not surprising as the former period corresponds to the start and aftermath of the Asian Financial Crisis. It is interesting to note that even in that period of increasing overall poverty, there were significant movements out of poverty. In the urban sector, roughly the same proportion became poor over the period of 1997-2000 as in 2000- 2007 (6.78%) but far fewer escaped poverty (59%). 6.2 Consumption mobility We complement the analysis above with an estimate of household consumption mobility (a continuous measure), by estimating a consumption mobility index. The index equals one minus the correlation of per capita expenditures of an individual in two different points in time. A 15 Although their study did not weight by survey-design probabilities and is thus not strictly comparable to this analysis 16 More detailed results are available from the authors 33 higher correlation means that income is less likely to change, and thus the mobility index takes a lower value. Note that this is not intended to measure income or consumption growth (positive or negative), as both high positive growth and large drops in consumption contribute to a higher mobility index. Like the poverty status transition matrix above, this measure is also biased towards finding a high degree of mobility because of the spurious (uncorrelated) variation caused by measurement error. However, in contrast to changes in discrete poverty status, methods have been developed to account for the bias when examining changes in continuous consumption measures. We use the method proposed by Gibson and Glewwe (2005) in which the observed correlation in consumption across years is divided by a “reliability index” to obtain the corrected correlation , and the corrected mobility index equals one minus the corrected correlation. The reliability index has previously been estimated for earlier rounds of the IFLS (Gibson and Glewwe, 2005; Glewwe, 2005)17. Table II.6.2 shows that the correlation between the logarithm of real per capita consumption between 2000 and 2007 equals 0.498, and the corrected correlation estimate is 0.67. Thus the corrected consumption mobility index is estimated at 0.33, or about one third lower than the uncorrected measure. This result suggests that the previous poverty transitions, which are based on the same consumption measures, are also likely to be significantly overestimated. At the same time, even the corrected mobility measure indicates substantial income mobility over time (again, in any direction). As seen in the first four rows of Table II.6.2, the raw correlation for urban per capita consumption is higher than for rural per capita consumption, which could be explained by an actual higher correlation (lower mobility in urban areas) or simply a higher level of measurement error. Correcting for the reliability ratios resolves this problem. The corrected correlation in urban areas between 2000 and 2007 equaled 0.73 versus 0.59 in rural areas, implying less movement of consumption levels in urban areas (which could be explained by a higher consumption level at the starting point since consumption cannot be lower than zero). The corrected mobility indices are respectively 0.27 and 0.41. Similarly, the corrected mobility indices for the 1997-2000 periods were 0.24 and 0.33. Note that the mobility indices for the 1997-2000 are smaller in magnitude, which is not surprising due to the shorter time span. 6.3 Correlates of Movements in and Out Of Poverty Having established the existence of substantial movements across the poverty threshold, we now turn to describing the characteristics that are correlated with those movements to provide a more complete picture of poverty transitions. It bears emphasizing that we are generally not able to interpret these correlations causally. 17 To calculate the corrected mobility index for two points in time requires data from an additional point in time. That is, to estimate the consumption mobility between 2000 and 2007/2008, we also need consumption data for a previous period, which the multi-round IFLS fortunately can provide. The IFLS has two previous rounds (1993 and 1997) which have been used to create the reliability index needed to correct correlations and mobility indices. 34 As motivation, we first present transition matrices for specific groups in the urban sample in Table II.6.3, which shows the proportions of individuals with a given characteristic that moved in and out of poverty. Panel A shows the correlation of education category (primary or no schooling; completed junior secondary schooling, and completed senior secondary schooling and beyond) with poverty transitions. As expected, higher levels of schooling are associated with a lower probability of being poor at one point in time (as can be seen in Column 1), a higher probability of escaping poverty and a lower probability of becoming poor. About seven percent of the non-poor who lacked a schooling degree higher than primary became poor in 2007. Among those who were poor, those who had completed junior secondary schooling were 7 percentage points more likely to escape poverty than those with primary or less (81% versus 75%). Among those with senior secondary schooling, 82.2% were no longer poor in 2007. It is important to note, however, that a lower probability of becoming poor for the better educated does not necessarily imply that those with higher levels of education experience on average greater improvements in consumption. A lower probability of becoming poor can arise simply because having more education is associated with being further from the poverty line in the first place. For example, if individuals with education at the senior secondary level were on average farther from the poverty line, even a large reduction would leave them still classified as non-poor. This can be seen in the differences in transitions between those with senior secondary and beyond, on the one hand, and junior secondary on the other. Only 2.59% of those with senior secondary and beyond became poor, compared to 4.53% of those with junior secondary only. This is despite the fact that the average approximate proportional increase in consumption among those with the highest level of education was lower than for those with junior secondary only (17 percent versus 14 percent). As can be seen in Panel B, individuals who were unemployed in 2007 were no more likely to become poor, or any less likely to escape poverty, than others. This is in line with what was found in the analysis of the SUSENAS data above, where we observed no association of per capita consumption with the household head’s not being employed. On the other hand, urban business owners (defined as those who have enterprises that employ at least one worker besides themselves) were less likely to be poor in 2007, less likely to become poor, and more likely to escape poverty than both the urban self-employed and wage employees. Despite this, the group of business owners was the one for which consumption grew the slowest during this time period (Panel C). Consumption growth was greatest on average for those who were wage employed in 2000. Panel D shows that individuals living in households headed by a woman were more likely to become poor and have smaller expenditure growth than those living in households where the head was male. However, to what extent this is because of the gender of the head or to other variables correlated with it cannot be determined from these descriptive tables. We turn now to the multivariate regression analysis of these transitions on the sample of adults residing in an urban locality in the year 2000. This allows us to analyze the correlation of the variables of interest with the transitions conditional on other observables with which they may be 35 correlated. The first column of Table II.6.4 shows the results for probit models where the dependent variable is being poor in 2007. Columns 2 and 3 show probit models where the dependent variable is an indicator of movement into and out of poverty, respectively. The independent variables include individuals’ characteristics (measured in 2000) including gender, age, education level, and employment status as well as the characteristics of the head of the household in which they resided at the time. Rather than the probit coefficients themselves, marginal effects, or the change in the probability of the outcome for a change in the independent variable, are shown in the table columns 1 to 3. The fourth column presents results for a regression model where the dependent variable is log change in household consumption. All the models include controls for the province of residence at baseline, in addition to the regressors shown. The standard errors are clustered at the community (kelurahan or desa) level. In many respects, the results are consistent with the transition matrices: individuals with higher levels of schooling were less likely to become poor and more likely to escape poverty, but this was not due to any effect of education on expenditure growth, as the last column suggests. The education level of the head of household was as strongly correlated with escaping poverty as the education level of the individual him or herself. This analysis also confirms that unemployment was not an important cause of movements into poverty. However, the employment type of the head of household is important. Individuals who lived in a household whose head was a business owner or was a full time employee were less likely to become poor than those living in households whose head was self-employed. However, in contrast to the case without controls, being in a female headed household is associated with lower poverty in 2007 and a higher likelihood of moving out of poverty—even though it was associated with lower consumption growth.18 Finally, while we have examined the association of poverty with shifts into unemployment, on the one hand, and with occupation at the start of the period, on the other, it is also of interest to see if changes in poverty status or consumption are associated with changes in occupation status during the 7 year interval of the panel. In additional regressions (not shown) we included indicators for the following broad occupational shifts: self-employment to business owner, and to wage employment; wage employment to self-employment, and to business owner; business owner to self-employment and to wage employment (the base category are those who do not change occupational status). These variables were included for both the individual himself or herself and for the head of household of the individual. There were very few significant results for these variables, which for some cases may reflect the small number of individuals making a particular occupational shift. Individuals’ own occupational transitions were not significantly associated with becoming poor or becoming non- poor, though shifting from self-employment to wage employment was associated with a fall in consumption. For household head occupational transitions, shifting from business ownership to wage employment is negatively associated with an individual becoming poor and is positively associated with an increase in consumption. On the other hand, the head shifting from wage 18 This appears contradictory, but it should be kept in mind that the estimate for the impact of female headship on consumption growth pertains to the whole sample. What matters for poverty transitions is what happens around the poverty line. 36 employment to self-employment is associated with a lower probability of becoming poor and a higher probability of becoming non-poor. This transition may capture individuals (heads of households) moving from poorly paid informal wage work to better paid self-employment. It should be kept in mind that causality may run from a change in income to a switch entry into self-employment (since income may be a source of start-up capital for a business). Also, the head of household may be a different person in the two periods (for example, the individual may be changing households), in which case a recorded occupational ‘shift’ does not capture an individual’s actual transition. Ultimately, there is not a clear story in these data with respect to links between occupational change and poverty transitions. 6.4 Poverty Transitions and Migration to Urban Areas Migration to urban areas is a very important factor in transitions out of poverty. Table II.6.5 shows the results of probit models for poverty transitions including migration defined in alternate ways. The first column shows that households (rural or urban) who had undergone a migration of any type between 2007 and 2007 were less likely to be poor at the end of the period. In the first panel, migration is defined generally to include any movement to a different locality (whether within urban areas, within rural areas or between rural and urban). Among individuals living in non-poor households in 2000, those who migrated as just defined were 5.5 percentage points less likely to become poor than those who did not. Further, among poor households in 2000, those who migrated were 18.8 percentage points more likely to escape poverty by 2007. In general, the per capita consumption of migrants increased by 20 percentage points more than that for the average Indonesian in the sample. This very large effect might come as a result of individuals migrating because of a job or other economic opportunity. It is also important to keep in mind that migration may be highly selective on traits that would lead to greater growth in incomes and consumption, which is not controlled for in the models. We further investigate whether these results come about from movements specifically from rural to urban areas. Panel B shows results for poverty transition probits where the independent variable of interest is change in rural to urban status between 2000 and 2007. Individuals who moved from rural to urban areas were less likely to be poor in 2007, less likely to become poor from 2000 to 2007 and more likely to escape poverty, and enjoyed an increase in per-capita consumption of 23 percentage points relative to average. Overall, we observe generally very positive changes in economic status for those who change from the rural to urban sector. It should be understood changes from ‘rural’ to ‘urban’ status as we have been defining this transition can happen in two ways, only of which involves physical migration. A rural inhabitant can move to an urban area, but rural areas can also grow and urbanize, so that long term residents in these areas become urban residents. The above results include both types of changes. We therefore investigate the distinct associations with poverty transitions of these different ways of becoming urban. Columns 1 to 4 of panel C in Table II.6.5 show probit estimates of the effects of being a “rural to urban migrant” and a “rural to urban non-migrant”. Both variables are correlated with a lower probability of being initially poor, a lower probability of becoming poor, and a higher increase in average per capita consumption. However, the magnitude of the effects 37 is stronger for physical migration from rural to urban sectors than for being in a rural area that urbanized. For instance, relative to not experiencing either kind of change to urban status, the probability of becoming poor is 13% lower for migrants compared to 8% for passive urbanization. Similarly, those who migrated from rural to urban areas increased their per capita consumption by 54 percentage points more on average than those who did not change from rural to urban status, while those who lived in areas that become urbanized saw their consumption increase by only 8 percentage points relative to the base group. In sum, this analysis of poverty transitions indicates that there is substantial variation of expenditures over time, which translates into large numbers of individuals moving in and out of poverty between 2000 and 2007. Part of the observed mobility is spurious due to measurement error, but there is still significant real mobility. Expenditure mobility is lower in the urban sector, but it is still substantial. This means that a larger fraction of the urban population will be poor at one point in their lives than are poor at any one time; only the latter is captured in a standard cross section household survey. Most factors that affect poverty also affect transitions into and out of poverty. Put another way, factors that help position individuals far from the poverty line are those that reduce the probability of falling into poverty—in part simply by placing them well above the line. Migration to urban areas seems to have the largest beneficial impact (to the extent it can be interpreted causally) on consumption and poverty of the factors considered 7 Vulnerable Urban Sub-Groups In this section we discuss specific sub-populations in urban Indonesia that may be particularly poor or especially vulnerable to falling into poverty. These groups, such as recent urban migrants, slum dwellers, or child laborers, are either not identified in the household surveys used in this study, or are not covered adequately enough to permit a detailed analysis of their situation. However, other more specialized surveys have been conducted in the recent past for some groups, and there are also a number of ethnographic qualitative studies. We briefly review this literature here, focusing for the most part on studies conducted since 2005. 7.1 Rural-Urban Migrants According to the Intercensal Population Survey (Supas), in 2005, 11 million of the 44 million people living in municipalities, or 24 percent of the total urban population were long-term rural- urban migrants (meaning their current residential location is different from their birth location), and another 5% were recent migrants (whose current location is different from five years before). Hence, nearly one in every four urban residents has migrated from a rural area during his or her lifetime. Java and Sumatra are the two provinces that absorb the largest number of long- term rural-urban migrants: Java alone contains 61% of all long term rural-urban migrants and 60% of all recent migrants. The major migrant enclaves in these two provinces are Jakarta in Java and Medan and Batam in Sumatra. In other regions, Samarinda and Balikpapan in Kalimatan and Makassar in South Sulawesi also draw large numbers of migrants from rural areas. (Resosudarma et al., 2009b). 38 The large numbers of rural migrants settling in Indonesia’s urban centers have made this group important subject for research, and potentially, policy. Most studies find migration to be beneficial to the socioeconomic status or earning of the migrants. Resosudarmo et al. (2009a) use the Rural–Urban Migration in Indonesia (RUMiI) survey conducted in 2008 to compare the socio-economic condition of long-term and short-term migrants to their local counterparts (non- migrants) in the cities of Medan, Tangerang, Samarinda, and Makassar. They find that migrants’ household income is slightly higher than non-migrants, and health status is at par with their non- migrant counterparts. As the authors note, they do not—in common with several other studies discussed here--control for the potential self-selection of migrants, which can bias their results. Deb and Seck (2009) arrive at a similar conclusion about the economic well-being of migrants using panel data from the Indonesia Family Life Survey (1993, 1997 and 2000). When they compare families with at least one migrant to non-migrant families, they find that migration improved either income or consumption for migrant families. However, they also find high prevalence of morbidity among migrants and poor emotional well-being in migrants and/or their families. The study by Lu (2009) also suggests deleterious effects on the psychological health of migrants. Using the IFLS longitudinal data for 1997 and 2000, Lu finds that when compared to people in their originating rural regions, migrants suffer from poor mental health, which may be due to reduced social support. In addition, the author argues that migrants might not accrue health gains from their improved earnings since they send money back home. Research has also examined the well-being of children of migrants to urban areas of Indonesia. In general, migration has been found to have a positive effect on the educational attainment and health outcomes of children of migrants. Resosudarmo and Suryadarma (2010) use the RUMiI survey data to examine the effects of migration on children of migrants who were between 5 and 15 years old when they moved to urban areas with their parents. They use an instrumental variables procedure to control for migration selection, using a measure of district-level propensity to migrate, calculated from the Indonesian intercensal survey, as an instrument to predict parents’ migration decision. In adulthood, education attainment was 4.5 years higher than for children who had not migrated. There was a 15 percentage lower probability of being underweight but no effect on adult height. Deb and Seck (2009), using the IFLS, also find migration to have a positive effect on the well-being of migrant children in Indonesia’s urban centers. Further, when Resosudarmo et al. (2009a) compare children of migrants to their local counterparts in urban centers (without control for selection), they find that the educational attainment of migrant children does not lag behind and that there is only weak but not robust evidence that migrant children are more likely to be underweight. Therefore there is evidence that rural-urban migration in Indonesia leads to improved economic outcomes for migrants relative to staying in their original location and some evidence that adult migrants and their children fare as well as long term urban residents. Evidence for health effects is more mixed, with possibly deleterious effects on adult migrants’ health. Unfortunately, these studies do not shed light on the conditions and potential vulnerabilities facing recent migrants in terms of access to services and other aspects of the areas within cities in which they settle. 39 7.2 Slum and informal settlement residents The definition of ‘slum’ is not straightforward. In some discussions, it is taken to be synonymous with informal settlements, but in others it is not. UN Habitat (2010) defines slum dwellers as ‘a group of individuals living under the same roof in an urban area with at least one of the following four basic shelter deprivations: lack of access to improved water supply; lack of access to improved sanitation; overcrowding (three or more persons per room); and dwellings made of nondurable material’. A home meeting one or all four of these conditions would then be classified as a slum household. This definition is commonly used as it is relatively easy to operationalize, since the necessary information can be obtained from the appropriate surveys. However, as Baker (2008) notes, the definition lacks a spatial dimension: it is not based on the characteristics of the area in which a person lives, but rather on the characteristics of their household. Thus the proportion of the population in slums is measured simply as the share of households lacking improved water and sanitation, adequate living area and other facets of decent housing. Hence some of the literature purportedly addressing slums is discussing conditions of the general urban poor rather than conditions of residents in areas considered to be slums, though obviously these categories strongly overlap. Nonetheless, if we consider this definition, the most recent UN Habitat report (2010) notes that the share of urban residents living in slums in Indonesia declined markedly between 2000 and 2010 (from 34.4 to 23 per cent). Indeed, Indonesia ranks as one of the most successful countries in ‘slum reduction’, although the use of this metric means that welcome improvements may reflect the gains in income and reductions in poverty that have been achieved over the period rather than the impacts of Indonesia’s ambitious programs for slum upgrading. Unlike the definition of slum above, ‘kampung’ which denotes informal settlements, is an area- based definition. Kampung, however, is not interchangeable with slum (Wilhem 2011; McCarthy 2003). While generally poor, these areas encompass a mix of poor and non-poor residents. For example, Wilhelm (2011) suggests that fully 60–70% of the city dwellers in Jakarta live in kampungs; obviously many of these or even a majority would be non-poor households. As Wilhelm points out, for the non-poor as for the poor, kampungs have some advantages including cheap housing as well generally easy access to work in the center city. Other estimates (McCarthy 2003) put the share of Jakarta residents in kampungs far lower. Geographically, kampungs (and slums) in Jakarta and other Indonesian cities are generally very densely populated inner city neighborhoods that are often located along riverbanks, canals, or railways, with attendant flood risk (discussed below) (Wilhelm 2011; Astuti 2009;McCarthy 2003, World Bank 2011a). Many of these settlements are unplanned and unregulated, with a poorly defined legal status that leads to insecurity of tenure for residents. Still, many residents are homeowners. For example, 68% of the residents of Karet Tensin, a downtown Jakarta kampung described by McCarthy 2003) reported that they owned their homes. Most of the rest pay rent. Squatters–households with no formal arrangement for use of land and not paying rent—make up a much smaller share of residents in informal settlement (for example 7% in Karet Tensin). However, many or most homeowning residents, not to mention renters, do not have formal title to the land, especially in newer settlements (Supriatna 2011; Steinberg 2007). A recent World Bank study reports that in much of Jakarta city more than 50% of the 40 land parcels are unregistered with the government and do not have title (World Bank 2011a). This leaves informal communities vulnerable to eviction for urban development. Indeed, by one account there were more than 86 evictions of low-income communities in Jakarta in the period 2001-2005 (Mercy Corps 2008). This also has other important consequences for residents – notably, requirements such as a land title, 30% down payment and a proof of income can prevent residents of slums or kampungs from accessing formal credit lines (ADB, 2010). Slums or kampung are not merely areas of low incomes or insecure land tenure. Due to their unplanned nature (and governmental neglect), residents’ access to public goods and services can be severely limited. Infrastructure to provide basic amenities such as water, drainage and electricity are underdeveloped. Safe water and sanitation are particularly important concerns. With regard to sanitation, Indonesia’s historically low investments relative to the region in public sanitation infrastructure (World Bank 2011b) has left the entire urban population poorly served. However, those in kampungs bear more of the burden of this underinvestment due in part to the sanitation and health effects of flooding in the context of inadequate sanitation infrastructure.19 With limited government provision, slum residents often access services through personal connections or informal intermediary service providers to whom they often pay higher fees. A study of Sidomulyo-Kricak, an urban village in northwest Yogyakarta, found that while less than half of the houses had electricity meters, the remaining households obtained electricity via negotiated informal arrangement with neighbors (Dovey and Raharjo 2007). Investigation of one Jakarta kelurahan found that people who obtain water from neighbors who have access to a water service provider incur higher costs than what they would have paid for individual access from other sources (ADB 2010).20 Corruption and crime are argued to be particularly serious problems in slum areas, in significant part because of the lack of government services also includes public security. As ADB (2010) notes, unregulated service provision in slums is often in the hands of local mafias or other powerful groups, and residents have few mechanisms to seek redress for poor treatment or other grievances. In cases where the government uses intermediates to supply and monitor public goods, residents often have to pay bribes. For example, in North Jakarta, applicants paid US $20- $30 for government-issued identification cards even though the cards were meant to be free. Furthermore, the inadequate supply of public goods and services such as safe water can have serious welfare implications for slum residents. Semba et al. (2009), using data from Indonesia’s nutritional and health surveillance system (NSS) surveys from 1999 through 2003, find that purchase of ‘inexpensive’ drinking water by families in slum areas is associated with higher under 5 mortality and morbidity (diarrhea) controlling for household and location characteristics. Yet, while access to basic services is problematic, for many these inner city urban areas may still be relatively attractive options because they are close to other services such as schools and health clinics as well as to their employment. Furthermore, access may be 19 With respect to water, Chomistriana (2011) asserts that only about 45% of the total demand for water in Indonesia’s urban slums is met, but the source of this figure and the definition of ‘total demand’ are not provided. 20 A recent newspaper account cites many urban residents paying up to $1 daily to buy clean water for drinking and cooking purposes (IRIN, 2010). 41 poor relative to more affluent urban areas, but could be superior to rural areas from which many residents come (Wilhelm 2011). With respect to location, the placement of slums or kampung in low lying areas near water (coastal areas, canals, and rivers) creates environmental risks that are exacerbated by climate change. The insecurity associated with poverty is exacerbated by weather-related vulnerabilities, as has been much noted for Indonesia, and particularly for Jakarta. Much of the city lies below sea level and is vulnerable to tidal flooding, storm surges, and changes in sea level due to climate change. This is the case especially for North Jakarta, which experiences the highest rates of both poverty and flooding (World Bank 2011a). Economic costs to the poor due to the disruption and damage caused by frequent flooding are substantial, in part because many residents of informal settlements also have their livelihoods based there, which are disrupted. Location-related health hazards from contamination of drinking water (also caused by pollution from sewage and garbage dumping) can also spill over to surrounding communities. It is clear from the existing research that, because of locational factors, environmental and health risks are faced disproportionately by those least above to absorb them financially. At present, we are not aware of systematic or representative studies in Indonesia that fully quantify the environment-related costs to the poor vis-a-vis other groups. However, qualitatively, observers have commented on the resiliency of these communities in dealing with persistent environmental hazards (World Bank 2011a; Wilhelm 2011). Well-established social networks and local organizations form the basis of collective action for early warning systems and evacuation procedures. While these systems should not be a substitute for more formal structures and government responses to flooding and other environmental hazards, they may function importantly as complements. Through local leaders and organizations, kampung residents in Jakarta are able to establish networks with formal structures and articulate their needs for aid and assistance (Wilhelm 2011). In theory, these complementarities could be extended constructively to the organization of preventive measures, such as construction of flood control channels, providing one of the key rationales for exploring community-driven strategies in slum area development. A more detailed discussion of slum upgrading programs is provided as part of the program review in Chapter III, Section 1.2. 7.3 Informal Workers Indonesia’s informal economy provides work for more than one-fourth of its employed population, and possibly significantly more. Casual and unpaid workers account for 23% of the urban work force based on analysis of Indonesia’s 2007 National Labor Force Survey (Sakernas) (Cuevas et al. 2009). Alternative definitions of informal employment would include permanent workers in informal enterprises and lead to a larger share for informal employment. Also using Sakernas (from 2009), Klaveren et al.(2010) extend the definition of informal workers to include self-employed workers (own-account workers and temporary/ unpaid workers), contributing family workers and casual wage workers. By the most inclusive definition, up to 72% of employed women and 68% of employed men work in informal jobs. Informal employment dominates as a source of livelihood for residents of low income urban neighborhoods. In a small survey of Karet Tensin, Java (McCarthy 2003) only about one 42 in five residents surveyed worked in the formal sector, defined as being paid a wage on a weekly or monthly basis. Informal work has a number of disadvantages, one of which is its often temporary nature, leaving workers more vulnerable to loss of work and shortfalls in household income. Informal workers also lack access to the job-related benefits, whether from government or private employers, that are provided to formal employees. In an ILO survey of 2,068 informal workers, 80% of the interviewed workers did not have social security, 60% were unaware of government provided social security schemes (such as Jamsostek), and 80% did not have any formal insurance (ILO, 2010). In contrast, about half of formal employees had these benefits. In case of illness, injury or old age, almost all informal workers relied on their families and communities to take care of them. Poverty rates are higher for informal workers than formal sector workers (Cuevas et al., 2009), and since the poor are more likely to be in informal activities, they are also less likely to receive job-related benefits than better off workers, as demonstrated below in Section III.3.2 using IFLS data. Informality may mean not just exclusion from benefits, but also a lack of legal protection or rights, as these workers are not officially recognized. Given their ‘informal’ existence and the government’s attitude towards them, many self-employed are forced to pay illegal taxes to officials as well as payments to criminals (Dimas, 2008). While few studies have considered the situation of different kinds of informal workers, specific categories of workers such as street vendors may be particularly disadvantaged. Based on the 2000 Population Census, roughly 1.7 percent of Jakarta city’s population, 1.5 percent of Bandung’s population, 2.4 percent of Bogor city’s population and 1 percent of Surabaya’s population was working as street vendors (Yatmo, 2008). In 2003, when Suharto (2003) interviewed 178 street vendors21 in Bandung Metropolitan Region, he found that, on average, vendors made Rs. 53, 686 ($6.7) per month. After accounting for unpaid labor (mostly family members), these workers can have a net profit margin as low as 1% (cited in Dimas, 2008). The powerlessness of this group is well illustrated by an episode in 2008, the “Year of Visit Indonesia”, when Surabaya’s mayor decided to make Surabaya a model orderly city. He ordered the local police to remove street vendors and pedicab drivers, thereby pushing out many informal workers, who lacked any formal mechanism to protest (Peters, 2010). More constructively, some local governments have recently tried to integrate street vendors into the formal market by placing them in strategic business locations in cities. Press accounts suggest that this has increased earnings for the vendors while also providing residents with more low cost food sources.22 21 In his study, Suharto defines street vendors as those who meet most of the following characteristics 1. Operate in public spaces that are not designated for business purposes; 2. Sell food, goods and services; 3. Form backward linkages with other industries operating in the formal sector; 4. Unlicensed and do not pay taxes; 5. Involve family members; 6. Are small and mostly have own account workers; 7. Do not offer benefits to workers provided in the formal sector; 8. Characterized by limited capital and infrastructure. 22 In central Jakarta, the municipality’s empowerment project has provided street vendors with space and facilities in the city’s business locations. One of the reasons vendors said their earnings increased was that they no longer had to pay gangsters for protection (The Jakarta Post, 2010) 43 7.4 Child Labor and Urban Street Children Child labor is a common response to household poverty, while also (through reductions in schooling) contributing to the intergenerational transmission of poverty. According to data published by BPS (2010), 2.4 million boys and 1.6 million girls aged 10-17 in Indonesia overall were engaged in work. 28% of the boys and over 34% of the girls who worked did so for more than 40 hours per week (cited in Klaveren et al., 2010). The Indonesia Child Survey, conducted in 2009, shows that among child workers, those in urban areas tended to work more than those in rural areas. 25% of child laborers in rural areas worked at least 40 hours a week while the same percentage of workers in urban areas worked at least 56 hours a week. Median working hours were also slightly higher for girls at about 28 hours per week compared to boys at 26 hours a week (BPS and ILO-IPEC, 2010). Bessell (2009) conducted interviews in 1994, 1995 and 1999 with 121 working children aged 10-16 in Jakarta. Almost half had moved from rural or smaller urban areas in search of economic opportunities (some with and some without their families). The cost of education was a significant factor in these children leaving school, and a significant number of them worked to add to the minimal earnings of their families. While child workers are vulnerable to injury and abuse, the most clearly at-risk groups of children in urban settings are street children and children who work as domestic workers. According to data collected by the Jakarta Social Affairs Agency (cited in Primanita, 2010), there are more than 4,000 children living and working on the streets in Jakarta alone . Beazley (2002) interviewed and examined the lives of 12 street girls, aged between 12 and 20 in Yogyakarta. He found that girl street children are more at risk than boy street children because in addition to other risks, they are also may incur abusive discrimination by boy street children. Child domestic workers are also susceptible to both exploitation and violence, primarily due to the hidden nature of their work in homes, where employers are able to exert control over their lives and movement. Muhammed Ally (2005) conducted forty-five interviews with former and current domestic workers (aged eleven and older) in urban areas of Java and Sumatra. Girl domestic workers generally worked fourteen to eighteen hours a day, seven days a week and ‘several’ (numbers not reported) reported being physically and sexually abused. In the Bessell (2009) study above, children who had left home cited multiple reasons for doing so, whether to escape poverty, to help their families, or to increase their independence. This suggests that general poverty-alleviation and safety-net programs are an important component of a strategy to mitigate children leaving home, but a more specifically targeted policy response is also called for. A more detailed discussion of current policy initiatives to address this group is forthcoming in Chapter III. 8 Qualitative Analysis of Urban Poverty To complement the quantitative poverty profile, we also conducted a qualitative study of the urban poor in a number of locations across Indonesia. The field work for the qualitative poverty analysis was conducted in tandem with the PPNP Urban process evaluation discussed in The work was carried out in collaboration with Survey Meter, a research and survey firm based in Yogyakarta, and the Indonesian partner implementing the IFLS. Burger et al. (2011) describe 44 the logistics of the fieldwork in detail, but we reiterate the main details below, focusing primarily on the aspects of the fieldwork that are relevant to the urban poverty study. 8.1 Site Selection There are many potential criteria for choosing sites but several were key, namely to have broad geographical representation as well as variation in the level of poverty. Further, for the PNPM evaluation, it was desired to include several sites that participated in the pilot for the Neighborhood Development (ND) project communities. 13 ‘non-ND’ kelurahan or urban wards were chosen from the sample of urban communities in the Indonesian Family Life Survey.23 Sampling from the IFLS sites made it possible to stratify on community wealth (reliable kelurahan-level data on poverty was otherwise difficult to obtain.) Given the coverage of the IFLS, the sample of communities chosen is reasonably representative of urban communities across the country.24 Regional stratification was performed based on discussions with the World Bank team in the U.S. and PNPM-Urban specialists at the World Bank in Jakarta, as well as other experts. The division of the 13 non-ND urban kelurahan by region was as follows: 3 from West Java, 3 from East Java, 3 from Central Java, including two field test sites in Yogyakarta and one from the province of Central Java, 2 from North Sumatra, and 2 from South Sulawesi. This distribution was chosen so as to achieve an appropriate geographical representation, based both on urban population and on the allocation of PNPM-Urban funds across regions. To stratify on wealth within these regions, a community wealth index was created from the IFLS community survey25. To choose the West and East Java sites, we categorized kelurahan into three levels of wealth based on the value of this index: less than or equal to the 33rd percentile of the national distribution of the index for all urban kelurahan in the IFLS (i.e., ‘poor’); between the 33rd and 66th percentiles (`moderate’); and above the 66th percentile (‘wealthy’). We then randomly selected one kelurahan from each category in West Java, and one for each category in East Java. For Sulawesi and North Sumatra, from which we were selecting two sites each, we carried out a similar procedure that selected one kelurahan from below and one from above the 50th percentile of the wealth index (`wealthy’). In Central Java, where we had only one site after the two field test sites were (randomly) selected for Yogyakarta, so we simply selected one site randomly. Replacement sites were also chosen randomly within each region/wealth stratum.26 23 We are grateful to John Strauss and Bondan Sikoki, PIs of the IFLS, for allowing us to sample from the IFLS community lists and providing access to the kelurahan IDs for this purpose. 24 We noted earlier that the IFLS does not cover all provinces. However, the small sample size as well as logistical considerations in any case prevented sampling in all provinces for this data collection activity 25 The index is based on the following variables from the IFLS community survey: an indicator of whether the predominant road type in the kelurahan is cement/asphalt, the percentage of homes with electricity, an indicator of whether there is a sewage system in place, the percentage of homes that receive water from a pipe, and the share of households receiving the rice ration. These variables were standardized and summed to create the index. 26 The selection process itself worked in the following way: within each combined region/wealth stratum, we assigned a random number to each kelurahan and then ranked the kelurahan in the stratum by that number. The site with the highest number in each was selected. The site with the second highest was chosen as a “replacement” site. 45 Table 8.1 Locations of Study Sites (Non-ND locations) Kelurahan Kabupaten/Kota Province Wealth index Triharjo Kulon Progo D.I. Yogyakarta High Karo Pematang Siantar North Sumatra High Cengkareng Timur Jakarta Barat Jakarta High Tambakrejo Surabaya East Java High Pancuran Gerobak Sibolga North Sumatra Moderate Astana Cirebon West Java Moderate Kauman Surakarta Central Java Moderate Lirboyo Kediri East Java Moderate Antang Ujung Pandang South Sulawesi Moderate Ngestiharjo Kulon Progo D.I. Yogyakarta Low Wiroborang Probolinggo East Java Low Rantepao Tana Toraja South Sulawesi Low Hulu Banteng Lor Cirebon West Java Low The site selection process ensured variation in wealth and geographic location, but the selected sites also varied along other dimensions. Although all kelurahan that participate in PNPM-Urban are designated officially as urban environments, the degree of urbanicity varies substantially. For example, Cengkareng Timur in Jakarta is a highly urban environment with features typical of a large city: compact multi-story housing, a substantial commercial sector, and high population density. In contrast, other sites were less urban. Pancuran Gerobak is a hillside kelurahan in a relatively small coastal city while Karo, is a low-density urban neighborhood at the edge of a relatively large inland city. Some kelurahan even appear rural: Triharjo, in particular, has low population density, a substantial agriculture sector, many unpaved roads, and few shared basic services (e.g., toilets, springs for washing/bathing). The other kelurahan we selected in Kulon Progo is also relatively ‘rural’, although it is less agrarian and has a somewhat higher density of homes. Two kelurahan within the kabupaten of Cirebon in West Java exemplify the diversity of areas where PNPM-Urban operates. Hulu Bateng Lor is a mainly agricultural area on the edges of the kabupaten while Astana is very close to the center of the city. Despite being on the outskirts of a large city, Hulu Bateng Lor retains a small-village, rural feel. Although some residents travel to work in the city, most of the economic activity revolves around agriculture. Astana, though close in to the city center, does not have the density of other urban areas. Most families live in houses (single-story) rather than in multi-story buildings, and there are small canals next to sidewalks, so cars do not pass next to most houses. However, the edge of the kelurahan yields directly onto one of the busiest avenues of the city of Cirebon. Although some residents work within the kelurahan (primarily attending small shops), most of them work all around the city. In East Java, the three selected sites showed strong variation in degrees of urbanicity and in differences between the poor and non-poor. In Probolinggo, Wiroborang has a moderately dense urban population living in small, single-family homes, many of which were rented, and some 46 proximate heterogeneity among poor and non-poor households. Tambakrejo, Surabaya, showed a strong contrast between the poor and non-poor: the non-poor live in a low density urban neighborhood characterized by multi-story, single-family houses on wide, well-drained streets, many with cars parked inside the gate; many of the poor squat on railroad company land alongside the tracks in high density slums with limited basic services. Lirboyo, Kediri, is a large peri-urban area characterized by low population density, mixture of agriculture and small businesses. Roads are generally paved but there are few shared basic services beyond electricity. In South Sulawesi, the sites showed dramatic differences in levels of urbanicity. Antang, Ujung Pandang, is a highly urban environment with a mix of poor and near-poor living in moderately dense conditions (small, single family homes built up against one another) with limited basic services. In contrast, Rantepao, Tanah Toraja, is a peri-urban/rural area with low population density, a mixture of agriculture and small businesses, and few shared basic services beyond electricity. Finally, in consultation with World Bank Jakarta staff, three more sites were chosen from the Neighborhood Development pilot sites with the objective of having a subsample with a range of experience (relative success) of the program. Two sites were in Central Java and one in West Java. All of the sites were within medium or large-sized cities, including a kelurahan within Bandung, one of the three largest cities in Indonesia. 8.2 Methodology In each community a rapid survey of poor households was conducted in conjunction with a uniform screening procedure designed to recruit the appropriate participants for the focus groups and in depth interviews. The primary criterion for respondent selection for both this study and a second set of PNPM-oriented focus groups conducted during the same site visits was that individuals be from poor households. The selection of households for the screening was based on the communities’ lists of poor households developed as part of the PNPM-Urban implementation. Separate groups of participants for focus groups and interviews were recruited for the PNPM study and the poverty study. In each community there was a poverty focus group for men and one for women, in addition to one in depth interview. The focus group and in depth interview protocols for the poverty analysis were designed to address the following areas:  Respondents’ perceptions of their own poverty and its causes, and barriers to moving out of poverty;  Strategies for coping with inadequate resources (both permanent and temporary shortfalls);  Differences in the causes and impacts of poverty for men and women  Perceived needs, including forms of government assistance.  Participation in and perceptions of assistance programs (including efficiency, fairness, value of benefits, convenience, corruption, etc.) 47 Protocols as well as the screening procedure were extensively revised based on the findings of pilot testing. Note-takers for the focus groups and interviews later expanded their notes using a prepared form with the help of audio recordings. These detailed notes were input directly on laptops and subsequently were translated by SurveyMeter and sent to RAND. The coding team at RAND used the text management software Atlas.ti to code the transcripts along the key themes outlined above. The output of this process was used for the analysis. Further details of the field work procedures and data analysis can be found in Burger et al (2011). 8.3 Perceptions of Poverty Various questions about people’s experiences and perceptions of their own situation were posed during the focus groups and in-depth interviews. Respondents were asked (1) what they thought were the greatest challenges facing them and their communities, (2) the reasons households fall into or escape poverty, and (3) what the main needs of the poor in their communities are. Views on the challenges faced by themselves and their neighbors were varied, but three were most frequently mentioned:  expenses exceed income, which does not keep pace with the price of basic necessities;  difficulty in finding jobs/security of employment;  expenses related to schooling constitute a financial burden to households. Lack of capital for small business and poor infrastructure (including housing, roads, water and sanitation) were also mentioned, but less frequently. Box 1 highlights some typical comments from a sample of locations on the first theme. Th Box 1: The Urban Poor Perceive the Cost of Living to be Increasing e firs  “Here, everything is expensive. For example, if we earn 40,000 per day, it is t of spent to eat, pay rent, pay for water, not to mention the children's school fees. So the if it is only 40,000 [Rp], we get minus earnings in one day” (man, Sibolga) thr  “The daily life is rather hard, for example the [cost of] petroleum. The daily ee needs are expensive. It is not like when Mr. Suharto became the president. It mo used to be cheap, the daily needs--rice [was] only 300 rupiahs. The salary of 700 st rupiahs [could] be used to fulfill the daily needs for one month. But now co everything is difficult.” (male, Probolinggo) m  “In the past, Rp. 10,000 was able to buy anything. Now Rp. 10,000 is only able mo to buy cooking oil. It can’t buy rice”(male, Pemantang Siantar) n co mp laints, that expenses are greater than incomes, is of course essentially a definition of monetary poverty. It was repeated across most kelurahan in our sample. It is noteworthy that most comments were phrased in terms of costs of necessities rising faster than incomes, rather than simply incomes being too low. This suggests that many respondents feel they are worse off than before (real incomes are lower). This question was also explicitly posed to participants. Indeed, participants overwhelmingly felt that the community’s economic situation had 48 worsened in the last 5 years. Of all focus group and in-depth interview participants who answered the question of whether the extent of poverty had changed in their community in the last 5 years or so, 72 said poverty had increased, 20 said it had decreased and 26 said that it had remained the same. Many of those who felt poverty had increased attributed this to a rise in the cost of living, while others felt that limited job opportunities and loans for business were also to blame. The finding that the majority of our focus group participants feel that poverty has increased in the last five years is intriguing in light of the fact that, based on the national surveys, urban poverty has been declining steadily if moderately over the period. The contradiction could reflect that a certain subset of the poor (who have suffered stagnant or declining real incomes in the last several years) tends to predominate in the focus groups, or that people tend to notice rising costs of living more than rising incomes (both are assuming of course that the Susenas data are accurate). The second most frequent complaint relates both to the lack of jobs and the lack of long term employment (Box 2). The latter points to vulnerability – to unemployment, income shortfalls, and poverty—as a major concern. Although somewhat less frequently than men, women also talked about unemployment or job insecurity, especially among youth and heads of household, as key problems in their communities. While some suggested that lack of education hampers opportunities, many also noted that even when young people have managed to complete their studies (high school or beyond) they often cannot find jobs. Box 2: Lack of Fixed Employment is Perceived as Major Challenge  “[Poverty] is increasing now because the job availability is different than before. Now it is difficult finding a livelihood and job” (man, Astana).  “We have many children who have to go to school. But we don’t have [a]fixed job.” (man, Pemantang Siantar)  “It is the economic sector, since there are many jobless people in this RW and they live in a dirty area…In RW 3. there are many jobless people and unskilled laborers.” (man, Jakarta Barat)  “there are a lot of young people here who do not know where to channel their ideas. Hence things like [getting] drunk and fighting which are negative… it is rare here for youth to have high school education. Most of them only get the first elementary school and get unemployed” (woman, Cirebon)  “There are many, many graduate people who don’t have jobs” (woman, Surabaya)  “Even with a bachelors degree, you still have difficulty getting a job (woman, Cengkareng, Timor) Finally, the problem of high (and increasing) schooling expenses was frequently noted, both by women and men. Many respondents singled this expenditure item out from the household’s overall expenses. School fees were the most commonly cited education-related expenditure that was placing financial burdens on families, although books and uniforms were also cited. This was true not only for parents who opted for private schooling, but also for parents whose children attended public school, where fees are nominally free (Box 3). Some participants explained that there are costs and fees involved in certain school exams for their children. 49 These comments indicate that individuals place a high value on education for their children but find schooling expenses to be a significant burden, even though these costs presumably do not constitute a large share of overall household expenditures. A large body of research (see Glick 2008) points to the sensitivity of enrollments to school fees and other education costs. One reason why households may find it difficult to meet these expenses is that outlays for fees and other school expense such as books and uniforms charges typically come all at one point in time. Box 3: Education is A High but Expensive Priority  “It’s about children’s school fee and daily life necessities…Now is not like in the past. They say that school fee is free but we must pay [more] than in the past. We must pay when we re-register in each school year” (women, Kediri)  “In public school its free but in private we still have to pay” (woman, Kulon Progo)  “Less income, a lot of spending. These last few months, my child's school fee rises.” (man, Sibolga  “The children are willing to continue to higher education. But we cannot pay the school fee. School fee for four children is so much. It cannot be compared to our income.” (man, Pemantang Siantar)  “BOS (Bantuan Operasional Sekolah - school operational assistance) aid only covers up to junior high school degree. It hasn’t covered senior high school degree” (man, Surabaya) As noted, two other themes that came up, though with less frequency that the three just discussed, were poor infrastructure and lack of capital. With regard to infrastructure, for example, a man in Cengkareng Timur noted that “there are many jobless people in this RW and they live in a dirty area. People from RT8 until RT1 live by the side of the river and all the houses are improper to live in”. The women of Jakarta Barat, D.K.I. Jakarta provided a litany of concerns: damaged roads, poor housing, no place for littering, and drainage problems. Men in Cirebon, West Java also discussed infrastructure problems in their communities: lack of wells and thus of access to clean water, a dirty and contaminated river, uncovered ditches, no lighting on walkways, and damaged roads. One male participant explained: “The river around here is dirty and contaminated”; another one added that “it gets salty in the dry season”. Concerns about the river and waste disposal were repeated in a few other sites as well. In Astana, a woman complained: “Nearly all of the community throws their waste into the river. In the rainy season, it was flooding all over. If possible, as household mothers, to make the river cleaner, there should be a recycle bin so that people do not throw their waste to the river”. A man in Kebumen said: “In my area, there is a cow slaughterhouse. It is in the north side of my house. And the waste runs through my sewer near the kitchen of my house.” This respondent noted that formal mechanisms seemed not to resolve the problem, which has larger public health consequences “I have informed this to Pak RT, Pak RW and even to the kelurahan. But they keep throwing the waste in the same sewer. The smell is disturbing. It is very strong and the flies are so many. I 50 think they should have their own septic tank so they don’t harm others. I and my family members once suffered from skin disease”. Finally, several, mostly male, participants talked about lack of capital to start or run businesses (including for instance buying and working a field). A man in Surakarta, Central Java explained: “The problem is we want to improve ourselves, but we lack funds. For example, if we want to start a business, we have an idea. But we lack the funds.” For some, loans were a way out of poverty for the unemployed: “The point is that the unemployed want to run a business, but there is no capital” (man, Astana). Another, in Cirebon, said: “The poor people have no funds. If they want to buy land or work on the land, we have no money”. Another man, in Tana Toraja, South Sulawesi noted that even when loans were provided, the levels provided were inadequate: “For example, we [were supposed to] have 10 million [Rp] as capital from the government. But the fund given is only 1 or 2 million. I actually have an idea to run a business but the capital is not enough. It is not even enough to buy the desk, and not enough to buy the other instruments. So, the loan would be in vain. Therefore, we need more information about how to get a loan to run a business.” When asked about temporary rather than permanent economic shocks, most respondents mentioned specific educational expenses and poor health. While one-off or unforeseen educational expenses reflect the general concern with education above, family members needing medicines, having accidents or requiring surgery and other forms of treatment were also mentioned by many. Box 4: Ill-health is a Major Reported Cause of Temporary Shortfalls  “My husband used to have a business. It went off since he was sick and no children could continue running the business. The standard of living was going down at that time. Now it’s kind of going back to normal, since all of my children are working” (Female,Tana Toraja).  “Because my child was sick, I borrowed money from a moneylender, of course with interest. So I became poorer.” (Participant, Sibolga) 8.4 Are There Different Causes and Consequences of Poverty by Gender? In both the male and female focus groups and interviews, participants were asked if and in what ways the experiences, causes and consequences of poverty differed for men and women. One of the key foci of this discussion was whether, and why, women are at a disadvantage relative to men in terms of economic security. Opinion was split on whether there are differences in the extent to which men and women were at risk from poverty and insecurity of employment, income, or access to food. With respect to how these views differed between male and female focus groups, slightly more men than women 51 felt that there were no gender differences in insecurity and the risk of poverty. Women’s views, on the other hand, were somewhat more varied; they were more divided on whether it is men or women who are at greater risk of falling into poverty, or whether the risk is the same. Those who argued there are no differences stressed that families, not individuals, face particular risks and challenges. Of these, a few respondents (men in particular) argued that the family’s situation depends primarily on the man: “If the man is poor, automatically the whole family is poor” (man from Pekalongan); “If the man is not working then he cannot give money to his wife” (man from Kulon Progo). A few participants also said that risk is equal for men and women because any individual may possess the characteristics that would put them at risk, for instance lack of skill or indolence. One female respondent, from Sibolga said: “Every lazy person has the same risk of becoming poor”. Another, from Pemantang Siantar said that the extent to which someone is vulnerable to economic shocks “depends on the person” and not necessarily on the gender. Of those who felt that women are more vulnerable to poverty, the explanations offered tended to differ for the male and female focus groups and there was also variation within male and female groups. Men and women both referred to women’s greater expenses, but characterized these differently. A group of participants, men in particular, felt the reason women were more likely to be poor was their “higher needs”; that is, the fact that unlike men, who have few needs, women require cosmetics, clothes, jewelry and so forth. Women also sometimes said that had more needs than men, but they tended to explain this very differently, saying it was because they have to take care of children, while men “only take care of themselves” (female from Kediri, East Java). Others, both men and women, offered explanations that alluded to women’s lower educational achievement and more limited income-generating options. This, some argued, was especially true for widows and single women who have children in their care. Some respondents argued that that it was easier for women than men to get by. A female respondent in Astana, for example, felt that “Women [have it] easier in getting jobs. Women can sell things in the crowd, for men it is difficult.” Further, there was some feeling that the situation has been changing in favor of women. A small number of respondents, both male and female, said that it was easier now for women to find jobs and engage in income-generating activities (such as selling goods in a market) than a few years ago. Another female respondent, in Kota Bandung, West Java, agreed: “Women can borrow money. Now women [have it] easier to get jobs, while for men it is difficult.” A man from in Kulon Progo said that in his community, “now there is a cigarette and wig manufacturer that can help [women], because the workers are all women.” Many respondents, particularly women, felt that the psychological impacts of poverty were very different for men and women—and harder for women. They commonly remarked that while men were often lazy, apathetic or careless in their spending habits (using their money to buy cigarettes and coffee, or to gamble, for instance), women were much more concerned with meeting the family’s daily needs. Many of the women explained that for them, the risk of poverty weighs more on their minds than on their husbands’. A female participant from Ujung Pandang echoed this sentiment: “We ladies are more concerned [with] things in the house and 52 the family than men. When we lack things. we will get a headache.” A third woman explained: “Men usually don’t understand daily needs. If there is no money, it doesn’t matter [to them] as long as there is food. But for women, if there is no money, they [try to] find it, and worry about not giving pocket money to the children. Men are too relaxed for this kind of thing, unlike women who always think. When a child asks for pocket money, men will just sit quietly but women will try to borrow from others because they feel pain if they do not provide it.” Another woman felt that the difference is also one of patience and adaptability: “Women are usually patient in facing poverty. While men are usually impatient and easily get angry” This view was shared by other women, and a few men. 8.5 What Strategies Are Used For Coping with Inadequate Resources? Borrowing money was the most widely cited way to cope with income shortfalls . Participants were broadly divided between those who said they would borrow from a usurer, and those who said they would borrow from family or neighbors. Many participants noted that because loans from usurers incur very high interest, the loans often end up making them more poor than they were initially, even if the money helps them deal with temporary shocks or crises. Others mentioned that people sometimes borrow money, knowing they were unlikely to be able to repay it, but shouldering the risk out of necessity: “Imagine a person who doesn’t work. He tries to get a job but still he doesn’t get it. So, he borrows money, although he cannot return it. Whether he can return the money or not is not what he thinks about first. The most important thing is that he can stay alive.” Coping strategies that were mentioned somewhat less frequently than borrowing money were selling or pawning belongings, reducing the number or quality of meals, and working odd jobs. A female participant from Probbolingo said “I sold plates, a legacy from my mother, and got Rp 7,000.-. I used it to buy 2 kilos of rice” A respondent from Kediri, East Java said: “To survive, people usually sell their belongings, pawn land, borrow money from the bank. That’s when it’s urgent. If it’s not urgent, it will not be like that.” With respect to the impact of income shortfalls on food intake, one participant from Kebumen, Central Java talked about instances when he had been in such critical situations: “I only eat moderately, just rice and chilies, because borrowing money is risky”. A female FG participant, also from Kebumen, Central Java said that when people find themselves in financial strain, they might “have meals twice a day… or buy cheap packaged rice from the street vendor.” Another participant, from Surabaya, East Java explained: “We usually eat three times a day, [but] because of the lack [of money] it is reduced to two times a day. We don’t have breakfast in the morning, [we] directly do lunch.” Other coping strategies mentioned by participants included asking relatives for assistance other than loans, including taking children in; moving away or abroad to find work; removing children from school and using available government programs/ asking government for assistance such as Jamkesmas (Health Card), or SKTM (village poverty letter). 53 8.6 What Do The Urban Poor Want from The Government? Focus group participants were asked to rank their own or their communities’ needs in a group exercise. In the individual in-depth interviews, respondents were asked to state needs without ranking. Finally, we note there was some ambiguity in the categorizations; for example, support for daily needs and cash assistance in the women’s list undoubtedly overlap. Overall, male and female responses were similar, except that men placed jobs more highly. Based on the rankings, the most strongly perceived needs by gender were: Men Women 1. credit for enterprises 1. credit for enterprises 2. support for daily needs 2. cash assistance 3. assistance for education 3. assistance for education 4. jobs 4. support for daily needs Respondents voiced a strong need for the government to supply credit either to start a business or expand existing businesses. A common feeling among participants was that that even if they were skilled and had the acumen to run a business, they would not be able to use their skills or training because of lack of access to credit. One respondent from Sibolga, gave the example of an acquaintance: “He works in the field of construction and can make a brick building. But he needs capital and therefore he should get assistance for the business in the form of capital loans from a bank through the village board.” It is noteworthy that this apparent unmet need for credit is found in a country well known for its highly developed commercial banking and microfinance sector. However, both anecdotal reports and more formal survey-based evidence has found that the participation of Indonesia’s poor in microfinance is low, suggesting limits to the outreach of both commercial firms and NGO and government programs (Johnston and Morduch 2008). Our focus group results are consistent with these findings. The second commonly-perceived need, support for daily needs (and also cash assistance), is also consistent with the expressed concern regarding the cost of daily life. Typical comments were “the government should think of us, the poor people. For example by giving subsidies of rice.” Or “Basic needs like rice and kerosene: for food we can eat simple food with Tempeh or more luxurious food with meat [for which] Rp 5.000 is enough, but it’s difficult if we don’t have kerosene”. Finally, and also consistent with perceived challenges, education-related assistance from the government ranked highly. Comments included: “Skills training for businesses so people know how to market and develop their business”; “Free education now only reaches until SMP [lower secondary school] while the SMA [senior secondary school] and higher students have to pay.” One respondent in Cirebon specifically requested a Balai Latihan Kerja. Other respondents 54 also mentioned particular education needs such as computers for children and religious education. Although respondents generally found their physical environment to be a problem in other discussions, infrastructure assistance did not rank highly among expressed needs from government. Among other needs, some respondents mentioned assistance with housing: “It is difficult to live in Surabaya, East Java if we do not have our own house. If possible, we want to have houses for free”. Only a few respondents mentioned the need for community infrastructure programs such as roads, drainage and water facilities. 8.7 How Do the Urban Poor Perceive Government Assistance Programs ? Participants were aware of a wide range of anti-poverty programs. In total, at least 40 national and local programs were mentioned in the focus groups and in-depth interviews. Of these programs, the most discussed were, unsurprisingly, the more important national social protection and poverty programs: Raskin (subsidized rice), BLT (unconditional cash transfers), Jamkesmas (health insurance), and PNPM (community empowerment.). When participants were asked about which programs they found most useful, they were divided in their opinions. Roughly one-fourth of those who answered this question said Raskin was the most beneficial to them, by helping them fulfill one of their most important daily needs. About a fifth of respondents stated that they preferred infrastructure programs because the felt that these programs had the largest impact by being useful for the most people. The same proportion of participants also cited all programs for being ‘most useful’ since they each catered to a different need and hence every one of them was useful. Programs that only a few participants mentioned as the most useful were BLT and Jamkesmas. This is perhaps surprising given the widespread coverage of these programs, and in particular, the apparent effectiveness of BLT in transferring significant resources to households during periods of fuel price increases (World Bank 2011c). In the case of BLT, the responses may reflect simply that this program, which was last in operation in 2008-9, was not foremost on participants’ minds. Respondents were asked about both the positive and negative aspects of government programs Most were often grateful for the programs they benefitted from. With regard to negative aspects of programs, commonly-cited grievances were corruption or poor targeting, inadequate assistance, bad service and poor access to services. Across all types of programs, the common themes expressed were that the program helped them meet daily needs and/or helped reduce expenses, and/or provided much needed capital. For instance, respondents felt that programs such as Raskin, BLT, and PHK (conditional cash transfer for the very poor) helped them “to fulfill [their] daily needs,” and “help [with the] education fee, and nutritious food for babies or children under five.” Likewise, respondents considered programs such as Jamkesmas, the social component of PNPM, and education assistance useful since these programs helped them to reduce their overall expenses, e.g., on health care. 55 However, respondents were also clear about the shortcomings of the programs, mentioning that most programs were tainted with corrupt practices and politicization which resulted in poor targeting. For instance, a woman in South Sulawesi elaborated the case of corruption in the distribution of BLT in her community: “I’ve seen many rich people in the queue waiting their turn to receive BLT. They are not government employees as government employees shouldn’t receive BLT, but they are not poor at all. Also the Lurah has cut the donation before it arrives in our hand. We only receive 200 thousand [Rp]. We actually pay the Lurah a certain amount of the donation to get his signature to be able to receive the donation.” Similarly, respondents highlighted the case of corruption in the distribution of Raskin in Cirebon, West Java and Kota Bandung, West Java, Jamkesmas in Tana Toraja, South Sulawesi and PNPM in Cirebon, West Java among others. In general, participants talked about a culture of corruption, collusion and nepotism in which those closer to people in power were more likely to get good assistance . These strong responses are consistent with previous analysis of targeting of social protection in Indonesia, the benefit incidence analysis in Part II on this report, as well as accounts of the political control over distribution of benefits at village or community levels (Sumarto and Bazzi 2011; World Bank 2011). In the views of participants, corruption and the resulting leakages lead to inadequate assistance for intended beneficiaries. With respect to Raskin, it was noted that since well-off people are also given rice assistance even though they are not on the government distribution lists, the overall quantity of rice that is allocated to poor people is reduced. Participants also stated that inadequate assistance has stemmed from policymakers not making a thorough assessment of the needs situation on ground when structuring some of the programs. For instance, one respondent in Kota Bandung, West Java noted that Raskin distribution was based on poverty lists that the government updated infrequently, hence were out of date. Along with issues of fairness and adequacy, respondents complained about bad service and difficulties in accessing some of these programs. ‘Bad service’ primarily meant ill-treatment by program administrators. For instance, respondents across several locations noted that when they went to hospitals with their Jamkesmas card, “staff sometimes provide services with grim faces,” patients were placed in the cheapest rooms, and they often received slow treatment. Similarly, during Raskin distribution, people who were late to pick up their installments often did not receive it and even though people were sick they had to stand in long queues to receive assistance. While most people felt that they could easily access programs such as Raskin and PNPM, they faced several challenges while accessing programs such as KUR (microcredit) which required collateral, and Jamkesmas, which required a fair amount of paperwork. Overall, the above-mentioned four factors were found to be problematic in most programs and according to participants across all locations they significantly reduced the overall benefits of these programs. 56 III. PROGRAM REVIEW AND BENEFIT INCIDENCE 1 Overview of Programs Serving the Urban Poor in Indonesia In this section we describe the key programs that serve the urban (and generally also, rural) poor. As indicated, a three part framework is used by the GOI to categorize programs and policies: Cluster 1 consists of social protection programs; Cluster 2 focuses on community empowerment and includes, primarily, PNPM; and Cluster 3, which is much smaller than the previous two clusters in terms of resources, includes programs for increasing incomes in the longer term via credit for micro and small scale enterprises. The set of programs that serve the urban (and usually also, rural) poor is described systematically in Appendix Table 1, which shows the main needs addressed, the target population, the main instrument (e.g., subsidized rice, cash) and describes key aspects of the program. The Table as well as the summary figure below show that the number and range of major programs is remarkable, and indeed this has been a source of problems due to difficulties in coordination as well as inconsistencies in targeting success. 1.1 Social Protection Programs and Basic Needs Much of Indonesia’s current social safety net is rooted in the massive new system of social protection programs, Jaring Pengaman Social (JPS), that was created in 1998 to alleviate the 57 negative impacts of the Asian Financial Crisis and ensure social stability in the midst of both economic crisis and major political change. JPS programs included the sale of rice at subsidized prices; nutritional supplements for infants and children; scholarships for elementary and junior secondary students from poor families, and block grants to health centers and to schools. Beginning in 2005, these programs have evolved from relatively ad-hoc temporary measures towards a more permanent system of social assistance (Donaldson 2011), while shifting from donor financing to being part of the GOI budget. Today, Raskin, which provides highly subsidized rice to poor households, is the largest assistance program in Indonesia. Initially called the OPK program, it was revamped as Raskin, under which the Bureau of Logistics (Bulog) distributes low-quality rice to distribution points throughout the country. Local governments determine eligibility based—in principle--on needs, and eligible individuals can purchase limited amounts (currently 14 kilograms) at a below- market price (Rp1,600 versus Rp 5,060). The target for the Raskin program in 2010 was to reach 17.5 million households. The target population is defined as poor households as well as vulnerable households, i.e., those living on less than 20% above the poverty line. Raskin is thus viewed as a core component of the basic social protection programs. A second pillar of social protection for the poor is social health insurance. GOI began providing health insurance for the poor in 2003 with a program that eventually became Askeskin (Health Insurance for the Poor) and then in 2008, Jamkesmas (Health Insurance Scheme for the Population). Jamkesmas currently provides health service fee waivers for 18.2 million households, making it the largest permanent program in terms of coverage in the country. Like Raskin it is targeted to poor and vulnerable households. A health card is distributed to eligible households (based on need, determined at the district level). The card entitles holders to free services in health centers and hospitals. After Raskin, Jamkesmas is the largest assistance program in Indonesia in terms of expenditures. Together the two programs account for 73% of total expenditures on social protection. Other important public-health programs include commitments to universal vaccination and delivery assistance. The third major program, BSM (scholarships for the poor) was begun in 1998 as part of the JPS and later renamed. Currently managed by the Ministry of Education (Kemdiknas), BSM provides cash transfers to low income students in public secular schools at each of the different levels. The target for 2010 was 2.77 million students in elementary schools; almost one million in junior secondary; 349,000 in senior secondary and 306,000 in vocational schools. At the same time, the BOS (school grants) program was initiated in order to provide schools with direct assistance. The BOS scheme currently also includes efforts to improve the quality of education via improvements in school-based management. The BLT (Bantuan Langsung Tunai) was a temporary unconditional cash transfer program that twice was used to offset the impacts of rising fuel costs on the poor (arising from the reduction of the subsidy for fuel), first in 2005 and again 2008-09. The 2005 program provided support for over 19 million households, making it the largest cash transfer program in a developing country. Recipient households received approximately Rp 1.2 million each and Rp 900,000 in 2008/9. The BLT, while considered successful, has been discontinued, and a conditional rather than unconditional approach to cash transfers has been adopted. Program Keluarga Harapan (PKH) is 58 a conditional cash transfer program for very poor households that was initiated in 2007. To determine eligibility, BPS, in conjunction with local governments, determines the poverty status of families with children. Those deemed “very poor” are eligible for the program. They receive cash if the mother attends pre and post-natal health checkups, has childbirths attended professionally, brings young children (newborns and toddlers) to professional health check-ups, and enrolls older children in school (verified through school records). The amount of the transfer varies depending on the number of dependents in the household, and ranges from Rp 600 thousand to Rp 2.2 million disbursed four times a year. The target for 2010 was to reach 816 thousand households, and in 2011 the program achieved nationwide scale with a goal of 1.17 million households by 2014. We noted in Section III.7 that members of particularly vulnerable groups will likely be helped by the overall expansion of these flagship programs that target the very poor generally. At the same time, some of the smaller programs listed above have been recently adopted to serve special needs of specific groups. For instance, the Indonesian government has launched the National Action Committee against Child Labor with the announced aim of eradicating child labor by 2022. Under this initiative, the government has recently introduced innovative programs such as mobile classes for street children and student drop-outs. In 2009, the Ministry of Social Affairs introduced the PKSA (Program Kesejahteraan Sosial Anak or Child Social Welfare Service Program) to help children at greatest risk, including street children. PKSA combines conditional cash transfers with service provision. Other currently-small but potentially important components of the overall safety net are JSLU, which provides cash and services to the vulnerable elderly, and JSPACA, which assists the disabled. 1.2 Urban Infrastructure Programs Indonesia has a long history of programs of municipal and slum improvement efforts, dating back to the 1960s or in some sense to the colonial period. The Kampung Improvement Program (KIP), started in 1969 in Jakarta and heavily supported by the World Bank and other donor agencies, is considered one of the most important and successful slum upgrading projects in the world. Over its 25-year duration, some 15 million kampung residents in urban areas across the country benefited from improved footpaths, roads and drainage, garbage bins and collection vehicles, safe drinking water through public taps, public washing and toilet facilities, neighborhood health clinics, and primary school buildings. The project is widely viewed as having had significant overall impacts in terms of quality of life as well as poverty reduction and improved health (Buckley and Kalarickal 2006, World Bank 2005). The success of KIP has been attributed to strong political commitment as well as community support, though the actual degree of local participation has been questioned (Setiawan 1998). Despite its successes, sustainability and maintenance were problems with KIP projects, which generally had a short term focus. KIP was also not concerned with broader, larger scale infrastructure needs of kampungs, and in some cases led to displacement of some poor residents due to increases in land values (Supriatna 2011; Agrawal 1999). In the 1980s the focus of the Government shifted toward a broader, city-wide approach to upgrading, an approach that coalesced into the Integrated Urban Infrastructure Development 59 Program (IUIDP). While IUIDP absorbed the KIP, the focus of IUIDP, unlike KIP, was not exclusively on slum upgrading. Further, IUIDP projects typically had components covering many sectors such as water supply, sanitation, roads, and housing, rather than standalone projects along one of these dimensions. The program also emphasized improved city level management and coordination across sectors and agencies (ADB 2000). After 1999, programmatic approaches shifted fundamentally once more, this time reflecting the broader transition toward decentralization as well as community driven development (CDD). Designed in many respects along similar lines as PNPM, the Neighborhood Upgrading and Shelter Sector Project (NUSSP), financed by the Asian Development Bank, included both slum upgrading projects and new housing development components. NUSSP, which ran from 2003 through 2010, was implemented in 32 cities and districts within 17 provinces. The upgrading component was designed to be highly participatory. Indeed, it was essentially modeled on the UPP, predecessor to PNPM Urban, and included local project identification, a community-based organization and/or community self-help group (BKM) as well as facilitators to assist the communities and the BKM. Access to housing improvement was facilitated through a housing micro finance program. The ADB’s review of its spending indicates that the NUSPP improved more than 6,800 hectares of urban slum areas in 803 neighborhoods and benefited about 800,000 poor urban households (ADB 2011). Several other major programs have been directed at slum improvement. A noteworthy current example is Healthy Places, Prosperous People (HP3), a program supported by Mercy Corps Indonesia, which is directed at improving the neighborhood quality of slums in North Jakarta. Currently, government policies toward upgrading are organized under the Indonesia Slum Alleviation Policy and Action Plan (SAPOLA) funded by World Bank, UN Habitat, and Aus- AID forming Cities Alliance (Cities Without Slums). This project supports the development of a National Slum Upgrading Policy and a National Slum Upgrading Action Plan, which have as a focus the enabling of local governments to improve living conditions in urban slums. 1.3 Microcredit Among developing countries, Indonesia has a uniquely long, rich and largely successful history of microfinance that goes back over more than a century (the country’s first microfinance program, the Badan Kredit Desa, was established in 1898). The range of programs extends from Bank Rakyat Indonesia’s (BRI’s) large Unit Desa network to very small, localized initiatives. With respect to government programs with an explicit Cluster III mandate, prior to 2004, there existed a number of government microcredit programs under different ministries, including the KUBE (Kelompok Usaha Bersama) under the Ministry for Social Affairs; the Revolving Fund for Smallholders with Profit Sharing or Syariah as well as conventional mechanisms under the Ministry of Cooperatives and SME; and various other rural and agricultural schemes coordinated by the Ministry of Marine Affairs and Fisheries and the Ministry of Agriculture (Suryahadi et al, 2010). In 2004, via a Presidential Decree, the government launched the scheme KUM-LTA or Kredit Usaha Micro Layak Tanpa Agunan (Collateral-free Microenterprise credit), which aimed to increase the availability of credit to microenterprises by having state-owned enterprises provide credit-guarantees to private agencies. Under KUM-LTA therefore the government takes 60 more of a market-supporting approach to microfinance. In addition, the Revolving Loan Fund under the PNPM-Urban program of community block grants, noted in the next section and discussed in more detail in IV.2.6 below, constitutes a major microcredit scheme in its own right. 1.4 PNPM-Urban Finally, as also previously described, the creation of PNPM in 2007, consolidated earlier community driven development (CDD) programs (including the UPP or P2KP program in urban areas). The Government of Indonesia’s nationwide massive PNPM program (National Program for Community Empowerment) encompasses all three clusters of the current TN2PK framework within a community driven development approach: social protection; infrastructure investment to promote pro-poor growth; and credit for small and micro enterprises. Currently the first two components dwarf the third, and the infrastructure component is larger than the social component. PNPM Urban, and in particular the infrastructure component, is discussed in detail in a separate study for this project (Burger et. al. 2011). The current thrust in PNPM is to increase the role of infrastructure and more generally community economic development, including through expansion in the size and scope investments as in the Neighborhood Development Pilot program (Ochoa 2011). For a more detailed description, the reader is referred to Burger et. al. (2011)’s process assessment of PNPM in urban areas. We note that the assessment generally shows that the CDD- based infrastructure activities under the PNPM-Urban program have successfully provided relatively high-quality small-scale infrastructure nationwide. However, in addition to the suite of programs described above, residual demand for social assistance provided via PNPM-Urban is substantial, as is the demand for more enterprise-oriented microcredit. 2 Benefit Incidence Analysis for Selected Programs In this section of the report, our objective is to quantitatively analyze how well a selected range of programs serve Indonesia’s urban poor using the tools of benefit incidence analysis (BIA). Benefit incidence analysis strictly defined seeks to assess whether public spending is progressive, that is, whether it improves the distribution of welfare, proxied by household income or expenditures. More pointedly, it examines whether this spending serves to redistribute income to the poor. A related but slightly different focus, also to be considered in the analysis, is to examine whether programs are well targeted to the poor. 2.1 Methodology for BIA The first step in BIA is to define the measure of the benefit of the program in question. For our analysis we use a simple but frequently applied method of representing the benefit with a binary (0, 1) indicator for whether a particular public service is used or the household participates in a program (e.g., the rice ration). The binary approach has some disadvantages: in particular, it imposes the assumption that the benefit is the same for all recipients, thereby ignoring, among other things, quality differences that may be associated with income level. On the other hand, it 61 is very simple to implement since we only require household survey data on program participation. Further, the binary method has been shown to yield results very similar to more complex and often equally questionable approaches, such as using the unit cost of provision derived from budgetary data or valuation based on compensating variations estimation (Younger, 1999; Sahn and Younger, 2000). The next step is to rank individuals in the population from poorest to richest, using (in our case as in most analyses) household expenditures per capita as a measure of individual welfare. The final step is to compare public services with regard to their progressivity, statistically using dominance tests and graphically using benefit concentration curves. Two measures of progressivity can be defined. The more standard definition, which can be called “expenditure progressivity” or simply “progressivity”, involves comparing the distribution of the benefit to the distribution of welfare (expenditures). If the benefit concentration curve is at all points above the curve for household expenditures — that is, if it ‘dominates’ the expenditure curve — the benefit is said to be progressive. In such a case, the distribution of the service or benefit is more equal than the distribution of income. Hence if the benefit is viewed as a form of income transfer, one can say that it serves to make the distribution of welfare more equal than in the absence of the transfer. The second measure we will use can be called (following Sahn and Younger 2000 and Glick and Razakamanantsoa 2006) “per capita progressivity”. This is a stricter definition of progressivity that is based on a comparison of the distribution of benefits with the distribution of the population rather than of expenditures. For any poverty line (any point along the expenditure distribution), the share going to the poor so defined must be greater than their share of the population, that is, a disproportionate share of the benefit must go to the poor. This focus allows us to assess whether a program is well targeted to the poor. Related but different ways of assessing targeting are to see what share of the poor are not getting the benefit (undercoverage or exclusion error) and the share of non-poor who do get it (leakage or inclusion error). We consider this as well for several key programs below. We also consider benefit incidence and targeting from a geographical perspective, by comparing the share of poor in different regions and provinces to their share of total benefits of specific programs (or equivalently but more intuitively, the share of poor in different regions or provinces getting the benefit). We present tables showing participation rates in different programs for the target population for each program, by quintiles of household per capita expenditure: enrollment rates and vaccination rates of children, the share of recent mothers who had a medical professional attend at the birth, and the like. This is of course an obvious way to look at the distribution of benefits and is sometimes referred to as benefit incidence among target beneficiaries. The difference between this and standard benefit incidence analysis should be noted. Standard BIA is concerned with how government spending policies affect the distribution of income or welfare in the population. It treats benefits, valued in monetary terms, essentially like an income transfer. From this perspective, it is appropriate to consider the entire population. Public expenditures on services received by persons in the poorest quintile (a health care consultation, a secondary school 62 enrollment) in essence transfers income to that quintile, improving the overall distribution of income in the society. Of course, public expenditures are not conceived purely or even primarily as income transfers. They are also designed to address the needs of particular target populations—education for children, maternal care for pregnant women, and so on. Participation or coverage rates like these implicitly compare benefits to ‘needs’. Distributional issues can obviously be considered through this lens too, by comparing participation rates in different income or expenditure quantiles.27 The two approaches often lead to different conclusions because the target population for many benefits is not evenly distributed across income groups. In Indonesia as in most developing countries, poorer expenditure quantiles tend to have a disproportionate share of target populations for many benefits (children, pregnant women, infants), largely reflecting differences in fertility and family size by income level. Therefore while the poorest 20 percent of the overall population might receive a disproportionate share of immunizations, the rate of immunizations among children might be lower for this group than for wealthier quintiles. 2.2 Programs Examined The selection of programs for the BIA is driven primarily by data gathered in the most recent surveys available. The 2010 SUSENAS and 2007/8 IFLS both contain information on participation in a number of programs to assist the poor and on the use of other services such as education that while not necessarily targeted to the poor nonetheless have important impacts on poverty and welfare. The SUSENAS core module gathers information on participation in several programs targeting the poor or near poor: (1) Raskin (2) Jamkesmas and (3) business credit through government programs.28 Further, as noted in section II.4.5, SUSENAS also collects information on maternal and child health services such as child vaccinations and birth attendance by medical professionals, as well as school enrollment at different levels. These are not anti-poverty programs per se but are standard outcomes for BIA and as noted have strong welfare implication. We will also examine the benefit incidence of these services. We will do this in the standard way, examining how these benefits are distributed across different income groups in the population overall, but also with a coverage perspective that considers the share of the target population (such as young children in the case of vaccinations) within each expenditure quintile that is covered by the program or service (Glick and Razakamanantsoa 2006). It should be noted that these services, and in particular education and birth attendance, need not be provided 27 Participation rates can also be expressed in terms of how the benefits received by a quintile compares with its share of total ‘needs’—i.e., of the target population for the benefit. For example, for immunizations, a quintile’s share of total immunizations over its share of the target population (all young children) equals the ratio of the quintile’s participation rate (share of children immunized) to the average participation rate (the share for all quintiles). Hence if the participation rate for the quintile equals the average participation rate (a ratio of 1.0), the quintile’s share of the benefit is the same as its share of the population in need of it. If the participation rate of the quintile is lower (higher) than the average, it receives a share less than (greater than) its share of the target population. 28 Including (separately) credit received in past year from government programs, from Kecamatan Development program Urban Development Project (PKK), and from PNPM. 63 exclusively by the public sector. As BIA is concerned with the distribution of public expenditure it should count only government services. Unfortunately, SUSENAS does not distinguish public and private providers for these services, so results will need to be interpreted with care. Since the wealthy are usually more likely than the poor to use private providers, public spending may be more pro-poor than our calculations, which include both, suggest. The 2007/8 IFLS household survey gathers information on participation in several of the same programs as SUSENAS (Raskin, Jamkesmas) but also several others, notably the BLT (Unconditional Cash Transfer). Further, the IFLS administers a comprehensive community survey with information on a number of community level public goods, notably, infrastructure. These are potentially important determinants of welfare (indeed, this assumption underlies the PNPM focus on infrastructure). We will also examine the incidence of these goods or services. The main variables of interest include the following: public infrastructure for sewage and sanitation, public lighting, public library, and asphalt or paved roads. In the case of a community public good, we assume that each person in the community benefits equally from its provision. This is a possibly incorrect assumption for some variables; for example, within a kelurahan, street lights may be installed only in wealthier neighborhoods. There is no way to go beyond this assumption without an actual mapping of public goods within communities. Still, the results should be informative. For example, if poor people are more likely to live in communities that are overall poorer (as we would expect) and such communities lack the resources to obtain or provide their own public goods, we would see a substantially non- progressive distribution of these public goods. 3 Program Coverage by Expenditure Quintile We first look at participation rates or coverage for anti-poverty programs by household per capita expenditure quintile. We present results for urban areas as well as for rural areas and for the total population. The quintiles are constructed based on the distribution of expenditures in the area considered, rather than just the national distribution, so that each quintile represents 20% of the population of interest (urban, rural, or national). Note that the first quintile of the urban population more or less corresponds to the share of individuals below the poverty line plus those less than 20% above it, which as seen earlier is just over 18%. Therefore we can take this group to represent the poor plus near poor, or the target population for most social protection programs 3.1 Social Protection Programs and Basic Needs As shown in Table III.3.1, 70% of individuals in the bottom 20% of the urban expenditure distribution received subsidized rice under Raskin in the three months prior to the survey. This share is larger than for all other quintiles, suggesting the benefit is somewhat targeted to the poor, but the benefits clearly extend well beyond this targeted group. More than half of those in the 2nd quintile received subsidized rice, and even in the 4th urban quintile close to 20% of individuals benefit from this program, while some 30% of individuals in the lowest quintile did not receive the benefit. These large exclusion and inclusion errors are in accord with other quantitative analysis of targeting of Raskin (World Bank 2010b) as well as the complaints raised in focus groups and interviews in the qualitative analysis discussed earlier. As 64 Sumarto and Bell (2011) note and as the focus groups and interviews suggest, local officials and organizations are able to use their power to distribute rice to non-beneficiary households. However, it is important to note that these leakages, large as they are, remain lower in urban areas than rural areas. In rural areas, fully 38% of those in the highest quintile report purchases under Raskin compared with 5% of in urban areas. Since rural areas are poorer (e.g., the poorest 40% in rural areas are poorer than the poorest 40% in urban) we would expect to see this pattern to some extent, but still the data suggest better, if highly imperfect, targeting of Raskin in urban areas. A similar pattern is seen for the health insurance program: disproportionate benefits for the poor but significant leakages to better off groups. Some 38 percent of the poorest urban quintile benefit from Jamkesmas, compared with 27% of the next poorest. Participation is low in the highest two quintiles (11% and 4%). With more than 60% of the target population (poor and near poor households) not getting the benefit, the program is not succeeding well in the objective of providing this group with access to health care. Also noteworthy is the fact that while the official target population for these two programs is the same—poor and near poor, or slightly less than the poorest 20% of the urban population—the shares of this group that enjoy these benefits are very different (much higher for Raskin—70% vs. 38%). However, this is not surprising given that each program is implemented by a different agency and uses different targeting approaches and databases of the poor. They also have different possibilities at the local level for leakages, with the distribution for Raskin rice likely to be especially prone to these outcomes. The 2007/8 IFLS provides information on receipt of both conditional and unconditional transfer programs. The questions about BLT, the unconditional cash transfer program, refer to receipt both in the past year and any time previously. Although the survey extended into 2008, it was too early to capture the more recent implementation of the BLT program in 2008-9. The results in table III.3.2 instead refer to participation in the 2005 implementation, which as noted earlier was implemented to offset the impact of increases in kerosene prices arising from the first reduction of the fuel subsidy. In urban areas, targeting of BLT was roughly similar, if slightly better, than for Jamkesmas, with some 39% in the first quintile receiving the transfer and about 21% in the 2nd, and relatively few in the upper two quintiles. Exclusion of poor and near-poor households appears to have been very high, as almost two thirds of urban households in the lowest urban quintile reported not getting the transfer. PKH, the conditional cash transfer, is more limited in scope as it is designed to target very poor households with children and at the time of the last IFLS round, the program was new. Only about half a percent of households overall, both in urban areas and nationally, reported participation. That said, the program, though limited, does seem to be targeted toward poorer households, as Table III.3.2 shows. 3.2 Education and Health Services As emphasized above, due to data limitations we cannot distinguish use of public and private education and health services. We expect the large majority of such services to be publicly provided, but also for wealthier households to make greater use of private providers (e.g., private 65 schools). Therefore the distributions we show, which consider all providers, will likely underestimate the relative benefits accruing to the poor from specifically public services Table III.3.3 shows net and gross enrollments29 for different school levels. For primary school, enrollment is close to universal across quintiles in both urban and rural areas. However, gaps emerge at post primary levels. 66% of junior secondary age children in the poorest urban quintile are enrolled compared with 77% in the richest; gross enrollments follow a similar pattern. The gap is much larger for senior secondary school, from 40% net enrollment in the first urban quintile to 69% in the highest quintile. Enrollment at post-secondary levels is very low overall and is especially heavily skewed toward the well off. Patterns by expenditure quintile are similar for rural areas, but with lower enrollment levels across the distribution. In sum, while Indonesia has made impressive strides in achieving close to universal primary enrollment, significant gaps persist for higher school levels. Table III.3.4 shows results from the 2010 SUSENAS for the share of children age 12-23 months who have received complete courses of vaccinations in Polio, measles, BCG and DPT. There are notable differences across quintiles, with 60% complete vaccinations for children in the poorest urban quintile compared with 74% in the wealthiest. At the same time, perhaps most striking is that even among children in the highest group, fully one quarter have did not get all four vaccinations. In contrast, virtually all recent births among the highest two urban quintiles were attended by a medical professional. Even for the poorest quintile the share is high (85%) while well short of universal. In rural areas there are more serious deficits for poorer households, for both indicators. Individuals also may receive health care and other benefits through their employment. While these are not public services (though for public employees one may think of them this way), for a fuller perspective on the distribution of services it is important to understand how these benefits are distributed across income groups. The SUSENAS surveys do not collect detailed information on job characteristics, but such questions were asked in IFLS. Table III.3.5 shows the receipt of a range of job benefits by urban per capita consumption quintile. It should be noted that the sample here is all wage-employed persons, not all individuals. With the exception of meals provided on the job, the incidence of employment related benefits clearly rises with expenditure quintile. Most strikingly, only 14% of urban wage earners in the poorest quintile get any kind of health related benefits, compared with over half in the highest quintile. Almost no employees in the bottom quintile enjoy pension benefits compared with 27% in the top quintile. These patterns reflect that poorer wage workers are more likely to be informally employed (indeed, receipt of these benefits is sometimes used as an indicator of formality of employment), as observed in the literature discussed in Section II.7. It is also important to note that the share of employed individuals in wage employment itself rises slightly with quintile. As the self-employed do not get job benefits, the allocation of benefits across the overall urban workforce thus even more disproportionately favors the better off employed than the table 29 As mentioned earlier, gross enrolment equals total enrolments in the level divided by number of children at official age for the level, multiplied by 100. Net enrolment is the total enrolments of children of official age for the level only, divided by the number of children of official age and multiplied by 100. 66 suggests. The distribution of job benefits favoring those in higher quintiles exacerbates the inequalities dues to simple income differences, and points to the need for public provision of health and other services to be especially targeted to those who are not well off. 3.3 Infrastructure and Basic Services We also examine the access by quintile to basic services such as water and electricity (from the 2101 SUSENAS) as well as community infrastructure (from the 2007 IFLS). It should be kept in mind that the former are usually, but not always, publicly provided. Private provision is certainly the case for some kinds of ‘safe water’, for example, high quality (branded) bottled water. In Table III.3.6 in can be seen that access to safe water is quite high for urban households across the expenditure distribution but there are still differences: some 87% of individuals in the poorest quintile have safe water compared with 95% or above in all other quintiles. Note that the differences across expenditure groups are larger in rural areas, with lower access overall to safe water as noted in section 4. Piped and good quality bottled water is much more highly concentrated among the better off. Having a toilet is distributed across urban quintiles in very similarly to the safe water indicator. In contrast, access to electricity is almost universal in urban areas. Again, variation between poor and well-off show up more strongly in rural areas. The relatively high access by urban residents in poorer quintiles to safe water, not to mention electricity, suggests that even in poorer urban neighborhoods residents are able to secure basic services. This may seem at odds with the evidence discussed in Section II.7 of highly inadequate public infrastructure in slums or informal settlements. However, that literature also suggests that slum residents do access utilities informally, though potentially at significantly higher cost. It is also possible that the SUSENAS sampling fame does not well capture the newest or poorest unplanned settlements. Table III.3.7 shows the variation by quintile in the ‘good quality’ housing indicator, as well as specifically the presence of permanent (non-earthen) floor. Obviously, the caveat above about inclusion of private services applies especially strongly here: while government housing programs exist, housing generally remains a private good. Nonetheless, the patterns for housing quality are of interest and conform to those for the basic services just discussed, in particular, the lower level overall in rural areas as well as greater variation across income levels compared with urban areas. Even among the poorest urban quintile, 79 percent of individuals live in houses of at least minimum quality (and 89% have houses with permanent flooring), and for the richest quintile these characteristics are close to universal. In contrast, only 57% of the poorest 20% in rural areas have housing of at least minimum standards (72% have permanent flooring), with 89% in the richest rural quintile having at least minimum quality housing. As noted earlier, the IFLS community survey provides information on a number of community- level infrastructure public goods. We found relatively little variation across urban expenditure quintiles in receipt of benefits from these investments, which for this purpose is assumed to accrue to an individual if he or she lives in a community where they are present. Table III.3.8 illustrates with two examples: the community having a sewage system, and having the predominant road type be of high quality construction, that is, asphalt or cement. There is a small level of variation in the sewer system indicator: 69% of individuals in the poorest urban quintile live in communities with this infrastructure compared with 77% of the highest quintile. 67 In contrast, there is very little difference across expenditure groups in the share living in communities with predominantly asphalt or cement roads, which is generally about 90%. This is typical of the other community infrastructure variables we examined with the IFLS. Therefore, access to basic services and infrastructure variables appear to be both generally high in urban areas and relatively though not completely evenly distributed across expenditure groups. With respect to community level infrastructure, we may not be capturing important differences in quality either across rich and poor communities or across rich and poor households. Further, the IFLS as well as SUSENAS may not be sampling in highly marginal urban areas where services may be particularly lacking. Public safety is an essential public good. As discussed in Section II.7, studies of slums or informal settlements point to a lack of police presence as a problem for residents in these areas. We can examine this issue using the ILFS, which asks respondents if anyone in the household was ever a victim of theft or mugging or robbery. As Table III.3.9 shows, the percentages overall are not high, generally not much above 5% for theft excluding crops and well below that for muggings or robbery. Perhaps unexpectedly, the incidence of crime, while low overall, rises with per capita consumption. This may be because crimes of this sort tend to occur in more affluent neighborhoods, where the houses and individuals are more attractive targets. Further, residents of poor neighborhoods may use informal approaches (neighbors looking out for each other) to substitute for inadequate police services. 3.4 Credit for businesses The 2010 SUSENAS indicates whether the household received credit for businesses they own. While these indicators are not precisely designed to capture microcredit for microenterprises, three sources are specified: credit received in the past year from ‘government’ programs (not further specified); as part of the Kecamatan (subdistrict) Urban Development Project (PPK); and from P2KP (now PNPM Urban). The latter two programs are intended to be directed to the poor. Credit from banks and other private sources is also recorded. As seen in Table III.3.1, receipt of credit in the last year from either PPK and P2KP is overall quite low, going to less than 1 percent of urban households in each case. Of course, we would not expect the shares to be as high as for rice purchases and health insurance, but the limited overall coverage suggest that these programs have a low impact overall . That said, the PPK/PNPM program credit accrues to poorer urban households more than to the better off, while the smaller shares getting credit through PNPM are fairly evenly distributed across quintiles. Credit from other government programs plays a larger role, accounting for 56% of credit from all public sources (including PPK and PNPM). All government sources in turn account for about half of enterprise loans received in the last year. Interestingly, use of credit overall and from government sources specifically is not higher among the better off in urban areas, though this likely reflects that many better off urban residents are wage employees. 4 Progressivity of Programs 68 The tables of participation by quintile address whether spending is more equally distributed among different expenditure groups than the population itself (per capita progressivity, meaning essentially that poorer people benefit disproportionately). Here we present a graphical analysis using benefit concentration curves, focusing on urban areas, that addresses whether program spending is more equally distributed than household per capita expenditures (expenditure progressivity). Figure III.4.1 shows the results for participation in Raskin. The x-axis in the graph plots the cumulative shares of individuals in the urban population, ranked by per capita household expenditures, while the y-axis shows the cumulative shares of the benefits, or the benefit concentration curve. Also depicted, by the dashed green line, is the Lorenz curve for expenditures, which shows the cumulative share of expenditures (or welfare) in the population. The degree of convexity of the Lorenz curve indicates the extent of inequality in consumption in urban Indonesia in 2010—if consumption were equally distributed across the population, the Lorenz curve would merge with the 45 degree line. The benefit curve for Raskin is very concave, indicating that a highly disproportionate share of the benefits accrue to the less well off. As it lies everywhere above (‘dominates’) the 45 degree line, the benefit is per capita progressive. This was already quite clear from the examination of receipt of Raskin by quintile in table III.3.1. Expenditure progressivity is a less strict condition than per capita progressivity, as it requires only that the concentration curve dominates the expenditure curve, i.e., the benefit is more equally distributed than expenditures. This is less difficult to achieve because expenditures themselves are quite unequally distributed. Raskin is clearly progressive in this standard fiscal incidence sense. However, as we saw in the last section, while Raskin is progressive (in both senses) and thus clearly ‘pro-poor’, it is not very well targeted to the poor, since significant shares of the non-poor also receive it. The graphs shows, for example, that the poorest urban quintile, roughly equivalent to target population of poor and near poor, receives about 40% of the benefit30—so that fully 60% accrues to those in the 2nd through 5th quintiles. As we would expect from the earlier discussion in Section 3.1, similar results are seen for Jamkesmas in Figure III.4.1b and unconditional cash transfer in 1c. Given the small numbers involved, we do not plot the concentration curve for the conditional cash transfer program but simple calculations indicate that 70% of the (still very limited overall) PKH benefits in urban areas accrued to households in the poorest quintile. In Figure II.4.2 we examine education. In these graphs we also show the cumulative share of children of school age for the relevant level, that is, the target population for each level.31 Since poorer households have a higher share of children relative to adults, the share of children at the lower end of the consumption distribution exceeds its population share. As seen in figure 2.a, for example, the poorest 20% of the urban population contains slightly more than 25% of all urban children age 7-14. Primary education benefits are clearly expenditure progressive and also appear to be per capita progressive as the benefit curve dominates (visually) the 45 degree 30 The share of the benefit going to the poorest 20% of the population is the height of the concentration curve at 0.20 on the x-axis, or about .40. 31 The age ranges are slightly broader than the official ages for each level. Given late starting and repetition, it is appropriate to consider children several years older than the official range as part of the target population. 69 line. Note, however, that enrollments are distributed very similarly to children of primary school age. This means that primary enrollments rates among primary age children are similar across expenditure levels. While normally this would indicate that the benefit is less progressive than would appear using the standard criteria for progressivity (expenditure and per capita progressivity), primary enrollment is essentially universal so there is no actual deficit among the poor that would require more targeted benefits.32 It should be pointed out, however, that there may be important quality differences in the education provided to poorer vs. better off children. We are unable to address this issue with our data. For junior secondary school a similar pattern is seen—enrollments are proportional to the share of children in different expenditure groups. However, this is not the case for senior secondary and (especially) post-secondary education. For the former, enrollments are clearly expenditure progressive but they are, for most of the distribution, apparently per capita regressive: the poor (and poor children) receive a disproportionately low share of total enrollments. Post-secondary schooling is even, by and large, expenditure regressive, that is, distributed less equally than welfare, as well as per capita regressive. Expenditure regressivity implies that public subsidies for higher education worsen the overall distribution of welfare. However, it should be kept in mind that for education as well as other services we are unable to distinguish public and private providers. As noted, public spending may thus be somewhat more pro-poor than our results show, since they include private providers that the well-off may use proportionately more than the poor. In Figure III.4.3 we examine immunizations and births attended by a medical professional. Again we show the cumulative shares of the target populations, in this case children age 12-24 months (for vaccinations) and under 2 years (attended births). In both cases these services are distributed more equally than expenditures, and also appear to be per capita progressive as the benefit curves lie above the 45 degree line. However, given the concentration of young children in poorer percentiles of the consumption distribution, rates of immunization and medical birth attendance are actually somewhat lower for young children, a pattern already observed in Table III.3.4. Hence these benefits related to maternal and child health are progressive in the standard senses while still somewhat underserving poor children relative to better off children . Finally, Figure III.4.1.d shows the curves for public enterprise credit programs. All the public enterprise credit programs that we study are clearly progressive relative to expenditures. However, they do not target poor households particularly well. The P2PK and other government credit generally lie modestly above the 45% line, while the PKK program is essentially distributed evenly across the consumption distribution. 32 Though it is often argued that public benefits should be more highly focused on the poor while the better off use the private sector, to allow both fiscal savings and better quality of services the government provides to the poor. As noted, we are not able to distinguish public and private providers in our data, and it is likely a disproportionate share of children from better off households attend private schools. 70 5 Geographical Benefit Incidence This section expands the analysis in Section III.3.1 to consider in more depth the geographical distribution of benefits, focusing on the two key programs included in the SUSENAS data, Raskin and Jamkesmas. 5.1 Patterns of Benefits in Urban vs. Rural Areas and Across Regions Figure III.5.1a presents the shares of the poor and near poor receiving Raskin rice by regions in urban and rural areas. As above, the near-poor are defined as having per capita expenditures within 20% of the poverty line. The figure indicates that within a region, the target population is fairly equally likely to benefit whether they are rural or urban. That is, for Raskin, exclusion errors are generally similar in rural and urban areas. The main exception to this is Sumatera where a poor/near poor rural resident is some 15% more likely to participate in Raskin than a similar urban resident. Figure III.5.1b indicates that for Jamkesmas, at least in the major regions of the country, we similarly see that the poor/near poor in rural and urban areas are about equally well targeted—or equally poorly targeted, given the generally low overall participation of poor and near poor in this program. For Raskin especially, far more notable than rural-urban differences within regions is the variation across regions in the share of poor and near poor (urban or rural) getting the benefit. The shares are high in heavily populated Java and much less populated Nusa Tengarra, and substantially lower in Kalimantan and Sulawesi. Thus the probability that a poor or near poor resident of urban Java receives Raskin is close to 80%; for his or her counterpart in urban Sulawesi it is less than 50%, and a similar pattern holds for the rural poor/nonpoor in these regions. For Jamkesmas the regional variation in exclusion errors is less pronounced. Next, we consider how inclusion errors – program leakages to the non-poor--vary between rural and urban areas and across regions. This is examined in two ways. First, Figures III.5.2 and III.5.3 show participation rates for poor/near poor and for others by program and location. For Raskin, there is a sharp difference between rural and urban areas. In urban areas, participation rates of those who are not in the target population (i.e., who are not non-poor or near poor) are generally well below those for the poor and non-poor, usually less than half. (Figure III.5.2a). Thus in urban Java, 79% of poor/near poor benefit compared with 32% of the rest of the urban population. In rural areas, in contrast, participation rates among those who are neither poor nor near-poor are very high, considering that this is not the target population for the program (Figure III.5.2b). Among rural Javanese, 83% of the poor/non- poor receive Raskin rice, similar to urban areas, but almost 70% of those who are not in this group also get the benefit. Jamkesmas overall suffers less from inclusion errors so defined (Figure III.5.3a,b). In urban areas, participation rates among those who are neither poor nor near-poor are less than half those of the target population in every region. For rural areas, the same holds true, or nearly holds, for the larger regions (Sumatera, Java, Kalimantan, and Sulawesi). As noted above, divergence in patterns for Raskin and Jamskemas is not surprising given their separate control and administration structures. In particular, local (village) authorities in rural settings apparently 71 have significant discretion to divert the rice ration to reward those who are not poor enough to be officially eligible. It is also striking for both Raskin and Jamkesmas that the participation rates of the non-targeted population tends to be high in regions where the participation among the poor/near poor is itself high, and low when the poor/non-poor participate less. This pattern tends to hold for both for rural and urban areas. This suggests that a key driver of regional differences in how well the target population is covered are differences in the overall level of benefits available to the population of the region. Although participation rates for these benefits are always lower among the non-target population than among the poor and near poor, the share of the non-targeted group in the overall population is much larger than the latter. What does this imply for the relative levels of program resources going to each group? This question leads to the second and more common definition of leakages, which is the share of the total benefit going to non-targeted individuals, or equivalently in our case, the proportion of beneficiaries who are non-target individuals. Table III.5.1 considers the share of the total benefit accruing to the target population and accruing to others for Raskin and Jamkesmas by region rural/urban division (as throughout this discussion the total benefit is measured simply as the number of individuals participating in the program). In most cases the share of the benefit accruing to the poor and near poor is under 50%, meaning that less than half of a benefit intended for this group actually goes to them. Leakages defined in this way do not appear to be notably worse in rural areas compared with urban . However, it should be kept in mind that the share of the poor/near poor in the population is higher in rural areas, so for similar patterns in participation rates, the target population would get a larger aggregate share of the benefits.33 In comparing inclusion errors of rural and urban areas, it should also be pointed out that the non- poor in urban areas tend on balance to be better off than the non-poor in rural areas. This derives from the fact that relative to poorer rural areas, in urban areas the poverty line is far to the left side of the expenditure distribution, with a larger share of the population above the line (or the poverty line plus 20% for poor and near poor). This larger group of non-poor are also on average better off than the smaller group of non-poor in the rural areas.34 It is plausible that a smaller share of these better-off non-poor would demand and receive the benefit. 5.2 Patterns in benefits by area income level Another way location may matter for one’s chances of participating in programs is through differences by area in average level of income, especially in view of the role played by local authorities in determining eligibility and distributing benefits. For example, more prosperous areas may have more resources to efficiently target the appropriate beneficiaries or simply to cover more people. Mechanisms of transparency may also vary by income level. Or, the poor 33 More refined measures used in World Bank (2010) adjust for this and also weight inclusion errors more heavily for the better off non-poor, but these measures are somewhat less intuitive than the simple ones used here. 34 This assumes the shapes of the consumption distribution are similar. Than one is essentially shifting the distribution to the right, leaving the share to the right of the fixed poverty line larger as well as richer on average. 72 may tend to be more powerful in asserting their rights when they a larger minority, as they would be in poorer provinces (or more simply, poverty might be a policy issue that is higher on the agenda in such areas). In Table III.5.2 we consider the overall variation in the province-level means of the share of urban poor/near-poor participating in these two programs. For Raskin, there is a great deal of variation. In 25% of provinces, less than 44% of poor/near-poor receive the benefit while in an equal number of provinces the share exceeds 67%. For Jamkesmas, there is less variation, but nonetheless in 25% of provinces less than 32% of the urban poor/near poor have a health card while 25% have at least 46% with the card. In Figures III.5.4 and III.5.5 we show again the urban shares of poor/non-poor and others receiving Raskin and Jamkesmas, this time by quintiles of median province level per capita expenditures. That is, we sort Indonesia’s 33 provinces into quintiles by their median per capita expenditures, and then assign each individual in the sample to a quintile based on the province in which they reside. For Raskin, there is an apparent tendency for the participation rates of the poor to be higher in low income relative to high income provinces, consistent with the last hypothesis noted above. However, this is not seen for health insurance coverage. A more consistent pattern, though stronger for Raskin, is that the participation rates of those who are not poor/near poor fall with median province consumption level. A possible explanation of this pattern is the same as given above when interpreting the larger inclusion errors in rural areas, which are poorer. Relative to a very poor province, in a rich province the non-poor are on average better off than the smaller group of non- poor in the poor province, hence a smaller a smaller share of these better-off non-poor might be expected to receive the benefit. One caution should be noted here, namely that given the relatively small sample sizes in some provinces, the measured variation is potentially exaggerated by sampling error. Nevertheless, these findings strongly suggest, as already previously implied that if one is poor or near poor, where one lives is an important determinant of whether one in fact benefits from these forms of assistance. Again the evidence suggests that the distribution mechanisms for the rice program permit a good deal more discretion on the part of local authorities in terms of targeting the poor, or conversely, rewarding the non-poor. 73 IV. CONCLUSIONS In this section we summarize the main findings of the study and following that, draw out the implications for policy, highlighting in particular implications related to PNPM-Urban. 1 Summary of Findings Indonesia has experienced steady, moderate declines in poverty in the last eight years in both urban and rural areas, at a rate of about .63 percent per year in both areas. These improvements have brought the rate of urban poverty from 15% in 2002 (when poverty still exceeded pre-Asian financial crisis levels) to just under 10% in 2010, based on the national poverty line. Although the urban poverty rate thus appears quite low, and is well below that in rural areas, a significantly larger share of the population (18.3% urban and 28% rural) are poor or ‘near-poor’, meaning having per capita consumption less than 20% above the poverty line. This group is considered the target population for the government’s main social protection programs. Further, the national poverty line is only slightly higher than the international $1.25 PPP per day measure of extreme poverty, and 33% of the urban population (and 48% of the total population) remains below the alternative $2 PPP per day standard. Currently about 37% of the country’s poor are in urban areas, a share that is certain to rise as the country goes from its current level of 45% urbanization to a projected 70% by 2030. While the urban poor overall are highly concentrated in Java, urban poverty rates vary substantially across regions: poverty is under 10% in Java, higher in Sumatera (12%), and generally high in smaller and less urbanized Eastern regions. However, declines in urban poverty since 2002 have occurred in almost all regions. Non- monetary indicators of well-being, such as school enrollments, vaccinations, and access to improved water and sanitation, have also generally registered improvements, sometimes substantial, in the last decade, in both urban and rural areas. For example, the share of young children getting all four immunizations (BCG, Polio, DPT and measles) in urban areas rose from 42% to 67% between 2004-2009, with even larger gains from a lower base in rural areas. Since human development indicators in Indonesia have tended to lag behind income growth, these developments are encouraging. Poor and non-poor households differ in important ways. Heads of poor urban households are more likely to rely on self-employment and less likely to be wage or salaried employees than are heads of non-poor households. Not unexpectedly, heads of poor heads of households have less schooling than the non-poor. Poor households are larger, by about 1 person on average, primarily reflecting the negative association of fertility and income. Multivariate models, which estimate the partial correlations of these factors and poverty or expenditures, confirm these descriptive relationships. A notable difference from the descriptive comparisons of poor and non-poor urban households is that female headship is negatively associated with per capita household consumption after controlling for other factors. There are also substantial regional differences controlling for household characteristics, so that otherwise similar urban households are better off in Java than Sumatera but less well off than in Kalimantan and Sulawesi. 74 The characteristics of the urban and rural poor differ markedly but in expected ways, especially in terms of occupation. The poor in rural areas rely much more heavily on agriculture/extraction than their urban counterparts (78% vs. 39% of household heads) and less on industry and services, as well as being more likely to be self-employed and less likely to wage employed. Hence, and not surprisingly, programs that aim to help the poor in urban and rural areas via their livelihoods need to be structured differently. Examination of the 2000-2007 panel from the Indonesian Family Life Survey reveals substantial movements in and out of poverty: some 26% of the 2000 national (rural and urban) sample experienced a change in poverty status from 2000 to 2007, with an approximately equal number becoming poor as became non-poor. Among urban residents, a third of the sample changed status, with 60% more people becoming poor than moving out of poverty. Part of the observed mobility is spurious, reflecting measurement error, but analysis of consumption mobility which can correct for measurement error suggests that there is still significant real mobility. The analysis confirms the need for policy to focus on those who are vulnerable to becoming poor, not just those who are poor. At the same time, it suggests that a large share of people and households who are poor will move out of poverty over time. From a national perspective, rural-urban migration has the largest association with movement from poverty to non-poverty. In focus groups and interviews in 16 urban communities, poor residents indicated that the greatest challenges they face were inadequate incomes, lack of jobs (and especially, secure employment), and high education expenses. The last two factors point to significant vulnerability to economic or unemployment shocks, and the inadequacy of the level and perhaps timing of current assistance to schooling for the poor. To deal with income shortfalls, borrowing from moneylenders or family/neighbors is the most common strategy used. Although views were mixed as to whether men or women are more vulnerable to poverty, a number of respondents (generally women) noted that the impact of household poverty is more significant for women, who have the most responsibility for meeting the daily needs of the family. With respect to what they or their communities needed most in terms of assistance from government, both men and women ranked credit for business the highest. Also highly ranked were support for daily needs, assistance with education expenses, and (for men) jobs. Relatively few participants cited infrastructure or housing as key needs for themselves or their communities. Overall, respondents generally strongly praised social assistance programs such as Raskin, Jamkesmas, and BLT as being helpful to them, but voiced several serious criticisms about how the programs were operated: poor or unfair allocation of benefits (poor targeting), inadequate levels of assistance, bad service, and poor access to services. Dissatisfaction with targeting procedures and outcomes was especially apparent, with complaints of corruption, nepotism, and politicization of the distribution process for different programs. These complaints are consistent with other, mostly rural-based, studies of distribution mechanisms and perceptions of the poor and consistent with the results of the program review. Turning to the review of programs, an unusually large number of programs exists for the poor, with multiple approaches, each oriented toward one of the three clusters of the government’s anti-poverty strategy (social assistance, empowerment, and credit for enterprises). PNPM-Urban 75 uniquely positioned in having a CDD approach that cuts across clusters. To complement the assessment of PNPM-Urban found in Burger et al (2011), the analysis in this report conducted a benefit incidence analysis of the largest of the other poverty programs. In urban areas, Raskin (subsidized rice) and Jamkesmas (health insurance for the poor) are disproportionately – but far from exclusively–allocated to the target population of poor and near poor (approximately equivalent to the poorest fifth of the urban household per capita expenditure distribution). Leakages to better off households are very significant, while at the same time, many of the poor do not enjoy these benefits: 30% of the poorest quintile did not purchase subsidized rice, and 62% do not have a health card. Leakages to those who are not poor or near- poor are more prevalent for Raskin than Jamkesmas. Overall, targeting performance for these two programs, particularly Raskin, is better in urban areas than rural areas. Unconditional cash transfers (in 2005) have a targeting performance similar to Jemkesmas while the still very small conditional cash transfer program PKH appeared in 2007 to be more precisely targeted to the poor. Several credit programs are also intended to be directed at lower income households. Overall, very few households (less than 1%) reported getting credit through either the PPK or P2KP (PNPM-Urban), limiting their overall effectiveness. Credit from other government programs plays a larger role than these sources, and all government sources in turn account for about half of enterprise loans received in the last year by urban households. That said, the PPK/PNPM program credit accrues to poorer urban households more than to the better off. A number of other services provided by government are not explicitly targeted to the poor but have important impacts on poverty and welfare, and are standard foci of benefit incidence analysis. In education, primary school enrollment is close to universal across quintiles in both urban and rural areas. However, gaps emerge at post primary levels. In urban areas, 66% of junior secondary age children in the poorest urban quintile are enrolled compared with 77% in the richest. For upper secondary the difference is very large: 40% vs. 69%. Enrollment at post- secondary levels is very low overall and is especially heavily skewed toward well off households. These gaps between the poor and well-off are even larger in rural areas. Therefore, while Indonesia has made impressive strides in achieving close to universal primary enrollment, significant deficits persist for higher school levels. There are also notable differences across urban quintiles in the shares of children receiving vaccinations for Polio, measles, BCG and DPT, with 60% complete vaccinations for children in the poorest urban quintile compared with 74% in the wealthiest. At the same time, perhaps most striking is that even among children in the highest group, fully one quarter have did not get all four vaccinations. In contrast, even in the poorest urban quintile, most (85%) recent births were attended by a medical professional, though fewer than in the highest two quintiles (close to 100%). Benefit incidence analysis indicates that social protection programs targeted to the poor and near-poor in urban areas are easily considered progressive by standard criteria, but still fall well short of being well targeted to this population. Primary and junior secondary education benefits 76 in urban areas are expenditure progressive (more equally distributed than per capita expenditures) and generally are allocated in proportion to the share of school-age children in each expenditure group. Progressivity starts to disappear with senior secondary: enrollments at this level are more equally allocated than expenditures but they are generally per capita regressive in that the poor (and poor children specifically) account for a disproportionately low share of total enrollments. Post-secondary enrollment is regressive; it is for the most part even less equally distributed than expenditures. Immunizations and births attended by a medical professional are progressive in the standard senses while still somewhat underserving poor children relative to better off children. Targeting performance varies significantly across location, i.e., where a poor (or a near-poor) person lives affects the likelihood that he or she receives benefits. There is also very significant variation at the province level in the share of poor/near poor getting these benefits, though again the variation is stronger for Raskin. For a given region, the target population for Raskin and Jamkesmas is fairly equally likely to benefit whether they are rural or urban. That is, exclusion errors are similar in rural and urban areas for most regions. However, especially for Raskin, there is considerable variation across regions in the share of poor and near poor (urban or rural) getting the benefit. The probability that a poor or near poor resident of urban Java receives Raskin is close to 80% but for his or her counterpart in urban Sulawesi it is less than 50%, and a similar pattern holds for the rural poor/nonpoor in these regions. With regard to inclusion errors or leakages to the non-targeted group, in urban areas, participation rates of those who are neither poor nor near poor are generally well below those for the target population--usually less than half. Thus in urban Java, 79% of poor/near poor benefit compared with 32% of the rest of the urban population. Still, in most regions—for both programs and for rural as well as urban areas--the share of the total benefit accruing to the poor and near poor is below 50% of the total, meaning that less than half of a benefit intended for this group actually goes to them. 2 Implications for Policy Sustained urban (and national) poverty reduction encompasses the development of human capital as well as macroeconomic and sectoral policies that provide employment and skill upgrading. In Indonesia prior to 1997 and in the other successful developers in the region, this was brought about by rapid expansion of the manufacturing sector, and especially, export manufacturing. Economic growth post-AFC in Indonesia in contrast has been driven by commodity demand as well as internal demand, while the manufacturing sector and manufacturing exports have been stagnant. Both economic growth and poverty reduction, while positive, have not matched levels prior to the crisis. Since the growth of labor-intensive manufacturing has been the pathway to improving incomes and escaping poverty (with clear implications for urban poverty especially), recapturing that growth still stands as the clearest path to more rapid poverty reduction for Indonesia. A wide range of policies, beyond the scope of this report to discuss in detail, are relevant to this outcome and to increasing growth and employment generally, including labor regulations, measures affecting overall business investment climate, and trade and exchange rate policy. 77 However, within the scope of our analysis, the results of this report give rise to several implications relevant for general policymaking as well as for PNPM-Urban specifically. 2.1 Targeting of Social Protection As this report (and a number of other analyses) has shown, while social protection programs are pro-poor in that the poor as well as near-poor benefit disproportionately from these programs, there is a fairly extraordinary level of leakage of benefits to better off households (inclusion errors). This is especially the case for Raskin, or subsidized rice purchases. Further, while many non-poor receive benefits, a large share of the targeted group of poor and near poor do not benefit—in the case of Jamkesmas, 60% of individuals in the lowest urban expenditure quintile are in households that do not have a health card. These large exclusion errors are particularly striking to observe in urban areas, where health facilities should be reasonably accessible to the large majority of residents so there should be no reason not to have insurance that covers the cost of care. Further, participation among the urban poor and near poor varies substantially across region and province. BLT or unconditional cash transfers had (in 2005) a targeting performance similar to Jamkesmas while the nascent conditional cash transfer program (PHK) appeared in 2007 to be fairly strongly targeted to the poor. Differences in targeting performance across programs reflects the fact that social protection programs, which currently are managed by different ministries, use different targeting approaches (combinations of means testing, community-based targeting and geographical targeting), rely on separate recipient databases, and sometimes use different definitions of poverty (Alatas 2010). Leakages to the non-poor arise not just from inaccuracies in poverty assessments but from local allocation decisions that reflect pressure from non-poor groups, social norms of sharing, or corruption (World Bank 2010b, Sumarto and Bazzi 2011.) While previous findings on these practices come largely from rural areas, this report finds that these are also true for urban areas. These shortcomings are well-recognized. The GOI has plans to rationalize the system by establishing a National Targeting System (NTS) for all social protection programs directed at households. The NTS will create of a unified database of poor households from which implementing agencies can draw lists of beneficiaries eligible to receive different social programs. Under this system, a poor household should be automatically enrolled in all programs for which it is eligible. Other changes to the structure of social protection policy, such as improved socialization, improved oversight, and better coordination among managing agencies, are recommended in World Bank (2010b, 2011c). 2.2 Improving Access to Education and Health Care for the Urban Poor Programs to benefit the poor go beyond protection, to encompass investments that raise income generating capacity and human capital. Our benefit incidence analysis found while primary education is now essentially universal in Indonesia, gaps still remain between poorer and wealthier households in urban areas (and more so in rural) with respect to enrollment at higher 78 levels. Net enrollment in upper secondary is 20% higher for the richest urban quintile than for the poorest. It is abundantly clear from our focus groups and interviews that poor people both highly value education for their children and find the costs of schooling highly burdensome, possibly related to the timing and discrete nature of school fees and related costs. Greater assistance to cover education expenses was a highly ranked need among the urban poor for both men and women. These perceptions suggest that current approaches including BSM (scholarships for the poor) and BOS (school operational assistance program) are not adequate. BSM in particular suffers from timing problems (disbursements coming months after enrollment) that limit its usefulness to households that are strapped for cash at the time of school enrollment. Redesign of BSM to address these shortcomings would have significant benefits for children (and struggling households) and would likely gain a great deal of popular support. Our analysis of the distribution of health care benefits focused on child vaccinations and attendance of medical professionals at births. While the findings may not be representative of all publicly provided health services, the low rate of complete vaccinations among children in the poorest urban quintile (60%) is disturbing and points to the need for better efforts to reach the poor. While the literature suggests that supply and demand side constraints can prevent the poorest from accessing vaccinations in remote rural areas, this should not come into play significantly in urban or peri-urban contexts. It also should be pointed out that while wealthier urban children do better, they are also far from enjoying complete vaccination coverage. GOI has now implemented universal vaccination and delivery care programs. These findings suggest that program structures that go beyond access to directly address parents’ motivation and provide incentives (such as PKH) may be important to explore for poor urban communities. 2.3 Improving Access to Credit Very few poor or near poor urban households, a significant portion of which are engaged in own account activities, receive business credit from programs that are supposed to be targeted to them. Credit for creating or expanding enterprises was the most highly ranked program need in focus groups of urban poor—for both men and women. This theme emerged both in the focus groups for this study and those for the companion study by Burger et al.’s 2011 on PNPM Urban. This apparently significant unmet demand for credit exists despite Indonesia’s well-developed microfinance infrastructure. Our findings confirm earlier work indicating that these services do not effectively serve the poor. The analysis of Johnston and Morduch (2008) suggests that while a large share of poor households would be creditworthy by the standards of commercial lenders, the loans they would request would be too small to be profitable to these lenders given administrative costs. This points to the need for a continuing and expanded role of government and NGOs in providing credit to the poor. While this report does not extend to a cost-effectiveness study of various channels of microfinance provision, a reconsideration of the phase-out of the Revolving Loan fund of PNPM-Urban is timely. 79 2.4 Insuring Households Against Income Shortfalls The GOI has done fairly well in cushioning the impacts of systemic or macroeconomic risk, in the form of the unconditional cash transfer programs of 2005 and 2008-9, both of which were designed to compensate poor households for large increases in kerosene prices (though this program has been instituted only on an ad hoc basis). However, our analysis suggests that vulnerability to idiosyncratic changes or shocks is also important and should be addressed by policy. Analysis of the last two waves of the IFLS suggest substantial movement of households into and out of poverty, hence vulnerability to being poor at least temporarily. Other analysis using three successive years of the SUSENAS panel (World Bank 2006, Sumarto and Bazzi 2011) also points to significant movements between poor and non-poor states. While all such analyses suffer from measurement error problems that exaggerate changes in consumption and hence in poverty, the finding remains that economic insecurity is a very real problem for the poor. This is underscored by responses in focus groups and interviews, even though these tended to mix the problems of losing an income source or having unsteady work with a general lack of jobs. More generally, the qualitative data make clear that temporary shortfalls of income to meet basic needs are a fact of life for many urban poor. Informal workers as noted are particularly vulnerable to periods of reduced income. As noted by World Bank (2011c) and discussed in this study, despite a plethora of assistance programs, very few programs currently exist to protect households against idiosyncratic risks from loss of employment or from illness or other shocks. The initial JPS package of measures to offset the impacts of the Asian Financial Crisis included a range of temporary employment programs. These appear to have been successful at raising household incomes as well as being well targeted (Sumarto and Bazzi, 2011), in part because those suffering income shocks self- selected into jobs programs where the wage level was lower than the market wage. Our analysis suggests that this is a gap in policy that should be addressed, either by renewing forms of assistance successfully used earlier, or increasing the scope and coverage of current programs such as Askesos, a currently undersubscribed government insurance program that provides cash benefits for informal workers in case of sickness, work-related injury or death. 2.5 Urban Infrastructure Programs that address urban infrastructure needs or are directed at slum upgrading have varied along several dimensions, including whether the projects are comprehensive and large scale (e.g., city-wide) or local (neighborhood level); the degree to which issues of tenure are directly addressed; the extent of local government involvement; and the role of community participation. The GOI and donors have funded a number of ambitious infrastructure programs that have improved the lives, and potentially raised the incomes of, millions of poor urban residents in Indonesia. Hence these programs can play a complementary role to the other antipoverty programs discussed in this report. At the same time, the programs have suffered from some common limitations. Observers have noted that land values rise in areas where the programs operate—a reflection of their success— 80 and this leads to higher rents or dislocation of poorer residents (Mulyana 1990; Supriatna 2011). The increasing scarcity of land and concomitantly rising land values in Jakarta and other cities will add to this pressure, and make new housing developments such as those attempted under the NUSSP (which require the acquisition of land) harder to implement. Sustainability and maintenance is another critical and recurring issue. A recent review by ADB (2010) of its multi-subsector urban projects, which span several decades and multiple phases of the government’s approach to upgrading, noted the importance for sustainability of having strong local participation in design, implementation and maintenance. As the report notes, participation is easiest to ensure for smaller neighborhood infrastructure projects where the benefits will be directly realized by residents. For larger municipality-wide infrastructure projects such as urban roads or district-level water supply and sanitation, participation is harder to sustain without substantial prior planning. The findings of our qualitative study also suggest that in spite of the fact that the urban poor benefit from infrastructure investments, government programs that address infrastructure needs are not particularly highly valued by poor residents. In part this may reflect the classic public goods problem, in that individuals tend to undervalue projects that yield collective benefits. For PNPM-Urban Burger et. al. (2011), noted a similar tension between participation and scale even for still relatively modestly sized projects. Community involvement was harder to achieve in Neighborhood Development pilot sites, where the scale of resources was greater and projects were designed to benefit the kelurahan rather than focused on a specific neighborhood. 2.6 Specific Implications for PNPM-Urban PNPM-Urban differs in design from most of the programs considered above in several important ways: the multisectoral (and multi-cluster) focus; the approach to decision-making and allocation, which relies on community decision making and has empowerment as a key goal; the focus on community well-being and development and the fact that it does not target individual households but rather communities. The findings of the present report have some implications for understanding how PNPM-Urban could be made more effective and how it fits into the overall scheme of Indonesia’s anti-poverty strategy for urban areas. It is difficult to fully assess the PNPM urban program in the broader context of Indonesia’s urban poverty strategy without specifying the counterfactual—that is, what would have happened in the absence of PNPM. Doing so is a challenge to rigorous evaluation because of the non-randomized rollout of the program. The one rigorous albeit non-experimental impact evaluation of the UPP program (predecessor program of PNPM) from 2004-2007 by Pradhan et al. (2010) and did not find gains in household consumption levels in program communities measured after a relatively short duration. This finding is in contrast to two rural-focused studies. Voss (2008) carried out an impact evaluation of the Kecamatan Development Program (KDP), the predecessor to PNPM Rural, combining panel household data from 2002 (from SUSENAS) with a 2007 specialized follow-up survey. Using a combination of propensity score matching and difference in difference techniques, Voss reports that households in the lowest income quintile experienced an 11% increase in per capita consumption relative to control areas (consumption effects were not seen 81 robustly across the income distribution). A more recent study (World Bank 2011d) evaluated the PNPM Rural program, also using difference and difference (with late recipient villages forming the control group) combined with matching.35 The estimates suggest that the program raised consumption per capita substantially overall, not just for the poor though more strongly for them. These results may not translate to the urban setting where the overall marginal economic return to small-scale infrastructure improvements may be low. However, larger scale improvements may have more economic effect. The Neighborhood Development project represents an ambitious attempt to recast PNPM Urban as a more comprehensive community development program that potentially could have significant economic impacts (See Burger at. al 2011; Ochoa 2011). It is currently being implemented in an extended pilot study in about 250 kelurahan. It will be important to carefully evaluate the impacts on economic development and incomes or consumption in these communities. Economic impacts could also come, in principle, though employment of residents in project construction. However, while many respondents reported actively working on PNPM infrastructure projects, such labor participation is generally voluntary (Burger et. al. 2011). Participation in paid employment under PNPM was not found to be extensive. In its present form, therefore, PNPM Urban does not play a significant role as a public works employment scheme but this could be reconsidered, particularly in the light of earlier recommendations regarding provision of temporary employment. Another way PNPM-Urban could potentially contribute to economic opportunity hence higher incomes is through the Revolving Loan Fund component of PNPM (urban and rural), which was created to increase access of the poor to loans via the community grant funds. Although much smaller than the social and infrastructure components of PNPM, the RLF nonetheless constitutes an important program in its own right, having led to the creation of thousands of community- based small-scale revolving loan funds over a 10 year period. This report found that general demand for credit for businesses is high among the urban poor, and respondents in Burger et al (2011) specifically call for it to be provided via PNPM-Urban. In view of this need, it is appropriate for PNPM-Urban to reconsider the planned phase-out of the RLF. However, such a response should be cautious given the mixed success of the RLF.36 On the one hand, it clearly brought loans to many poor households. At the same time, a number of important shortcomings have been noted (World Bank 2010c), including mixed or poor loan repayment performance, undercutting long term sustainability; lack of training and inadequate adherence to recognized standards of microfinance implementation and management; and some evidence of crowding out of commercial sources since RLF finance is cheaper and easier to access. At the World Bank notes, the nesting of a microfinance initiative in a larger grant based program (PNPM), with both managed by the same organization, raises potentially inherent difficulties: loan recipients may tend to view the loans as grants, reducing motivation to repay. A cross-country review of microfinance programs funded by community-managed external grants suggests that for these reasons, such programs are prone to failure by the normal standards of microfinance performance (Murray and Rosenberg 2006). The proper role for the RLF in 35 The 2007 baseline survey for this evaluation was the same survey used for the follow-up of the previous KDP evaluation. 36 This discussion is not specific to the urban part of the RLF. 82 serving the credit needs of the urban poor and in reducing poverty should be carefully assessed (and compared to alternative means of providing credit) in order to balance the demands and needs of the poor against the specific program constraints and structure of PNPM-Urban. With respect to the infrastructure component, Burger et. al. (2011) found that PNPM-Urban successfully delivers infrastructure projects of good quality with relatively few complaints about issues such as governance and control, and general satisfaction among residents with project quality and selection, although program familiarity and community participation in project decision-making could be improved. At the same time, findings from interviews and focus groups for the PNPM evaluation as well as from those conducted for this study indicate that improved infrastructure is not high among self-perceived needs of the urban poor. This is an important finding, but it is not as straightforward to interpret as it might seem at first glance. Respondents are likely to undervalue public goods such as infrastructure in favor of direct assistance (training, credit, cash transfers) that yield clear or immediate benefits to them. Turning to the social component of PNPM, in view of the focus of the present study it is especially pertinent to understand the ability of the CDD approach to deliver social protection as well as education and health services to the urban poor. Is it possible, for example, for PNPM to compensate for shortfalls in government provision of schooling, health services, or protection for the poorest (or perhaps to complement government provision)? This question is important in view of the fact that education and health service coverage of the urban poor remains far from complete (as does the coverage of the key national social protection programs). However, most evaluations of PNPM, including the process evaluation conducted as part of the present study, have focused on objectives such as participation, capacity, local governance, and citizen satisfaction, rather than measuring impacts or access to services. An exception is the rural KDP evaluation by Voss (2008) noted above, which examined changes in school enrollment and health service utilization in addition to consumption. Use of outpatient care among household heads rose 11.5%, but no impacts on secondary enrollment rates (which unlike primary had significant room for improvement) were seen, though this may reflect limitations of the panel. Further, the more recent PNPM-rural evaluation (World Bank 2011d) found that outpatient health care increased about 5% relative to controls communities, and these gains appeared to be pro-poor. However, as in the previous evaluation, no gains were seen in school enrollments or transitions to secondary school. While it should be kept in mind these results may not be applicable to CDD programs in urban areas, the findings for health services utilization suggests that PNPM can improve access to services, though the relative costs and benefits of this compared with an expansion of national or sub-national government programs are not known. However, it is also important to note that, at least for the more recent PNPM Rural evaluation, the measured improvements in health care utilization reflected gains in income rather than better service provision, since only a small percentage of funds were actually allocated to health projects. In contrast, PNPM Generasi, a pilot program that is part of PNPM Rural, is directly focused on a core set of health and education results. Villages participating in PNPM Generasi commit to improving twelve basic health and education service delivery or outcome indicators through the 83 block grants. The program combines CDD with a conditional cash transfer approach in that villages are incentivized (in terms of larger block grants in the next cycle) to perform well on these indicators. In a village-randomized evaluation carried out by Olken, Onishi,and Wong (2011), it was found that the program improved a range of health outcomes and program coverage, while having no effects on education. Results were stronger when incentives were in effect. The applicability of these findings to PNPM urban may be limited by the rural focus of the evaluation, but still they point to a number of important questions. One, already noted, is how the cost effectiveness of CDD approaches compares to standard, more top-down approaches to delivering health or other services; this is not known. Another question goes to the essence of the CDD approach. The core PNPM program (PNPM Mandiri, whether rural or urban) is not a program to provide services of a given type. It is a program to provide resources to and empower communities, which decide within broad parameters what to spend the resources on. There is an inherent tension between the achievement of specific aims the government values (such as higher immunization rates or secondary enrollments) and the objective of giving decision making power over resources to communities, whose interests may not align with those of the government (they may, for example, value livelihood programs such as computer training over basic education or health programs). One approach is to continue to give communities wide latitude and accept that they may not choose to invest in health or schooling (or to hope that there are gains in income which translate into improved outcomes in these dimensions, which seems to have been the case for health care in the PNPM rural impact evaluation.). PNPM Generasi instead attempts to deal with this tension by limiting the menu of choices for block grants to education and health services, and further, by incentivizing efforts to ensure good outcomes. For PNPM in urban areas, this may also be a useful approach to attain specific key objectives and to ensure that community efforts integrate well with broader government anti-poverty strategies. At the same time, it is important to weigh the tradeoff between these objectives and true community control over resources which is central to the CDD approach. 84 TABLES AND FIGURES 85 Table II.4.1 Trends in Poverty 2002-2010 National Urban Rural Poverty Change Poverty Change Poverty Change Year Rate (%) Rate (%) Rate (%) 2002 18.14 14.45 21.09 2003 17.44 -0.70 13.58 -0.86 20.23 -0.86 2004 16.66 -0.78 12.13 -1.45 20.11 -0.12 2005 15.97 -0.69 11.68 -0.45 19.98 -0.13 2006 17.76 1.79 13.47 1.79 21.81 1.83 2007 16.58 -1.18 12.52 -0.95 20.37 -1.44 2008 15.42 -1.16 11.65 -0.86 18.93 -1.43 2009 14.15 -1.27 10.72 -0.93 17.35 -1.59 2010 13.33 -0.82 9.87 -0.85 16.56 -0.79 SOURCE: Susenas March surveys, 2002-2010 86 T Table II.4.2 P Poverty Headcount, Poverty Gap, and Poverty Severity 2002-2010 National Urban Rural Poverty Poverty Poverty Poverty Poverty Poverty Poverty Poverty Year Rate Gap Severity Rate Gap Severity Rate Poverty Gap Severity 2002 18.14 3.19 0.85 14.45 2.52 0.69 21.09 3.73 0.98 2003 17.44 3.68 1.01 13.58 2.87 0.84 20.23 4.29 1.14 2004 16.66 3.02 0.83 12.13 2.29 0.65 20.11 3.57 0.96 2005 15.97 2.98 0.87 11.68 2.13 0.63 19.98 3.71 1.08 2006 17.76 3.72 1.22 13.47 2.76 0.90 21.81 4.63 1.52 2007 16.58 2.89 0.80 12.52 2.05 0.58 20.37 3.66 1.01 2008 15.42 2.77 0.76 11.65 2.07 0.56 18.93 3.42 0.95 2009 14.15 2.50 0.67 10.72 1.91 0.52 17.35 3.05 0.82 2010 13.33 2.21 0.58 9.87 1.58 0.40 16.56 2.79 0.75 SOURCE: Susenas surveys 87 Table II.4.3 Urban Poverty by Region, 2010 Jawa Sumatera NT Sulawesi Kalimantan Maluku Papua National Urban poverty rate (%) 9.62 11.59 23.97 6.15 4.70 6.84 5.59 9.87 Region share of all urban poor (%) 67.64 20.43 5.95 2.90 2.36 0.40 0.32 100.00 Urbanization rate (%) 56.69 39.12 30.30 30.73 40.23 27.60 22.80 48.30 Urban share of regional poor (%) 42.90 34.06 32.65 13.71 25.70 9.36 3.51 35.76 Mean daily per capita expenditures 16,982 14,895 13,851 17,981 18,363 13,769 12,544 16,481 SOURCE: Susenas surveys 88 Table II.4.4 Trends in Enrollment 2002-2010 (percent) 2002 vs 2010 2007 vs 2010 p- p- Education 2002 2003 2004 2005 2006 2007 2008 2009 2010 difference value difference value Current enrollment (7-14) national 91.8 92.7 92.5 93.6 94.1 95.9 95.1 95.2 95.6 -3.76 0.00 0.29 0.00 rural 89.2 90.0 90.2 92.3 92.6 94.6 94.1 94.0 94.4 -5.24 0.00 0.16 0.08 urban 95.5 96.8 96.0 95.4 95.9 97.4 96.3 96.7 97.0 -1.44 0.00 0.49 0.00 Net primary enrollment (7-12) national 91.5 92.2 91.0 92.3 92.3 93.6 94.2 91.8 91.6 -0.04 0.65 2.03 0.00 rural 91.4 92.6 90.6 93.2 92.7 94.2 94.4 92.3 91.8 -0.41 0.00 2.39 0.00 urban 91.7 91.5 91.6 91.1 91.9 92.8 93.8 91.0 91.3 0.43 0.01 1.58 0.00 Net junior secondary enrollment (13-15) national 59.6 60.0 61.6 63.4 64.7 66.5 67.3 68.0 69.7 -10.03 0.00 -3.19 0.00 rural 51.5 52.1 54.4 59.5 58.9 61.3 63.6 64.1 66.7 -15.26 0.00 -5.43 0.00 urban 70.3 71.8 71.0 68.8 71.3 72.7 71.5 72.5 73.1 -2.81 0.00 -0.48 0.20 Net senior secondary enrollment (16-18) national 36.0 39.1 40.5 42.0 41.5 45.3 44.8 48.7 52.0 -16.06 0.00 -6.71 0.00 rural 23.4 25.7 27.5 31.7 30.9 33.8 35.4 39.8 43.7 -20.25 0.00 -9.88 0.00 urban 49.0 53.7 55.9 53.6 52.2 56.7 54.0 57.7 60.2 -11.19 0.00 -3.54 0.00 SOURCE: Susenas surveys Note: ages for each school level given in parenthesis 89 Table II.4.5 Trends in Immunizations and Attendance of Medical Professional at Birth, 2002-2010 (percent) 2002/2004 vs 2010 2007 vs 2010 2002 2003 2004 2005 2006 2007 2008 2009 2010 difference p-value difference p-value Immunizations: completed all 4 courses (BCG, Polio, DPT and measles) * national -- -- 35.3 35.1 46.1 53.4 61.3 61.1 60.0 -24.70 0.00 -6.62 0.00 rural -- -- 30.2 29.2 37.2 48.7 57.7 55.7 55.6 -25.34 0.00 -6.86 0.00 urban -- -- 42.1 41.5 55.6 58.9 65.4 67.2 65.1 -23.03 0.00 -6.27 0.00 Medical professional attended birth (last 12 months) national 67.6 66.0 72.8 76.8 73.8 76.1 77.9 80.2 83.6 -16.02 0.00 -7.56 0.00 rural 57.8 52.4 61.1 64.2 61.6 64.7 66.3 70.1 75.6 -17.78 0.00 -10.91 0.00 urban 82.1 87.2 87.7 89.8 87.9 88.2 91.1 91.9 93.6 -11.42 0.00 -5.35 0.00 SOURCE: Susenas surveys * children age 12-23 months 90 Table II.4.6 Trends in Household Access to Water, Sanitation, and Electricity, 2002-2010 (percent) 2002 vs. 2010 2007 vs. 2010* Household 2002 2003 2004 2005 2006 2007 2008 2009 2010 difference p-value difference p-value Access to safe water national 78.1 77.4 78.0 76.2 77.5 75.5 83.4 83.9 86.2 -8.03 0.00 -10.64 0.00 rural 69.3 69.0 70.9 68.9 70.3 71.8 73.3 75.7 77.5 -8.21 0.00 -5.70 0.00 urban 89.1 88.5 87.0 85.1 85.3 79.9 94.3 94.7 95.4 -6.27 0.00 -15.54 0.00 HH has toilet national 72.3 73.1 76.1 80.5 81.5 83.5 83.3 85.2 -12.84 0.00 -3.66 0.00 rural 51.5 56.9 59.0 64.4 67.2 70.6 72.3 73.8 -22.23 0.00 -6.52 0.00 urban 90.7 90.6 92.5 94.1 93.0 94.0 94.3 94.7 -3.97 0.00 -1.70 0.00 Waste disposal - septic tank national 40.1 41.6 44.5 47.6 50.9 54.1 53.8 56.4 -16.28 0.00 -5.52 0.00 rural 20.5 23.5 25.5 27.6 33.1 36.6 38.4 39.5 -19.03 0.00 -6.45 0.00 urban 64.4 66.7 69.7 71.8 70.2 73.0 74.2 74.2 -9.82 0.00 -4.10 0.00 Electricity national 87.7 87.4 87.1 88.6 91.9 92.9 93.4 91.8 -4.15 0.00 0.03 0.66 rural 79.1 79.3 79.4 81.4 85.9 87.4 89.0 85.4 -6.33 0.00 0.48 0.00 urban 98.4 98.3 98.0 97.5 98.6 98.9 99.1 98.8 -0.46 0.00 -0.27 0.00 SOURCE: Susenas surveys Notes: *for variables missing in 2006, difference is across 2002 and 2005 91 Table II.4.7 Trends in Housing Characteristics, 2002-2010 (percent) 2002 vs 2010 2007 vs 2010* p- p- 2002 2003 2004 2005 2006 2007 2008 2009 2010 difference value difference value Roof good quality (concrete, tile, shingle, asbestos, iron sheeting) national 95.0 94.9 94.8 95.3 95.8 96.0 96.1 96.7 -1.69 0.00 -0.91 0.00 rural 92.4 92.0 92.3 92.5 93.0 93.3 93.8 94.5 -2.04 0.00 -1.51 0.00 urban 98.3 98.4 98.1 98.7 99.0 99.0 99.0 99.2 -0.92 0.00 -0.17 0.00 Walls good quality (wood, brick) national 83.6 85.2 85.6 88.0 87.8 88.5 88.4 89.8 -6.14 0.00 -1.95 0.00 rural 76.1 80.6 79.2 82.9 82.3 83.2 83.9 85.0 -8.88 0.00 -2.67 0.00 urban 92.8 94.1 94.4 94.2 93.8 94.3 94.6 94.7 -1.89 0.00 -0.97 0.00 Floor good quality (permanent covering, i.e., non-earthen) national 82.9 83.3 84.2 86.8 87.9 87.9 88.8 -5.88 0.00 -2.07 0.00 rural 75.5 78.2 77.1 80.6 82.2 82.8 83.3 -7.80 0.00 -2.69 0.00 urban 92.3 93.4 94.0 93.2 94.1 94.5 94.5 -2.29 0.00 -1.35 0.00 a Good housing quality national 73.0 73.6 75.0 78.1 79.7 79.7 81.8 -8.84 0.00 -3.74 0.00 rural 61.0 65.1 63.8 67.8 70.0 71.5 72.9 -11.92 0.00 -5.11 0.00 urban 87.8 89.5 90.1 89.0 90.1 90.8 91.1 -3.3 0.0 -2.0 0.0 a Good quality =1 if roof, walls, and floor are all good quality, 0 otherwise. 92 Table II.5.1 Characteristics of Poor and Non-Poor Urban Households, 2010 (percentages unless otherwise ndicated) Characteristics Poor Non-Poor Difference p-value HH head female 14.38 16.16 -1.78 0.05 HH head has no edu certificate 34.84 14.47 20.37 0.00 HH head has primary certificate 37.24 23.20 14.04 0.00 HH head has junior secondary certificate 14.76 16.07 -1.31 0.20 HH head has senior secondary certificate 12.80 33.69 -20.89 0.00 HH head has post-secondary certificate 0.36 12.58 -12.22 0.00 HH head employed 81.38 80.20 1.18 0.25 Hhhead ill (disrupted work > 5 days past month) 8.39 6.06 2.33 0.00 HH head age (years) 49.70 48.26 1.44 0.00 HH size (number) 4.83 3.87 0.96 0.00 Number of children < 15 1.71 1.01 0.7 0.00 Number of men 15-65 1.38 1.31 0.07 0.01 Number of women 15-65 1.45 1.36 0.09 0.00 Number of adults > 65 0.29 0.18 0.11 0.00 Head self-employed 52.34 45.06 7.28 0.00 Head wage employed 38.01 50.33 -12.32 0.00 Head unpaid worker 1.49 1.69 -0.20 0.52 Head sector agriculture/extraction 39.44 13.69 25.75 0.00 Head sector industry 21.49 22.45 -0.96 0.49 Head sector services 39.07 63.87 -24.80 0.00 Sumatera 18.22 15.96 2.26 0.14 Jawa 69.74 71.52 -1.78 0.35 Kalimantan 2.16 4.96 -2.80 0.00 Sulawesi 2.81 4.48 -1.67 0.00 NT 6.50 2.02 4.48 0.00 Maluku 0.31 0.50 -0.19 0.42 Papua 0.26 0.56 -0.30 0.01 # households 2060 24395 SOURCE: Susenas surveys 93 Table II.5.2 Characteristics of Poor and Near-Poor Urban Households, 2010 (percentages unless otherwise indicated) Near Characteristics Poor Poor Difference p-value HH head female 14.38 14.71 -0.33 0.79 HH head has no edu certificate 34.84 29.88 4.96 0.01 HH head has primary certificate 37.24 34.29 2.95 0.12 HH head has junior secondary certificate 14.76 16.45 -1.69 0.22 HH head has senior secondary certificate 12.80 18.16 -5.36 0.00 HH head has post-secondary certificate 0.36 1.23 -0.87 0.01 HH head employed 81.38 81.81 -0.43 0.74 Hhhead ill (disrupted work > 5 days past month) 8.39 7.53 0.86 0.33 HH head age (years) 49.70 49.20 0.50 0.31 HH size (number) 4.83 4.52 0.31 0.00 Number of children < 15 1.71 1.43 0.28 0.00 Number of men 15-65 1.38 1.41 -0.03 0.33 Number of women 15-65 1.45 1.43 0.02 0.39 Number of adults > 65 0.29 0.25 0.04 0.04 Head self-employed 52.34 49.95 2.39 0.22 Head wage employed 38.01 41.46 -3.45 0.08 Head unpaid worker 1.49 2.21 -0.72 0.17 Head sector agriculture/extraction 39.44 28.84 10.60 0.00 Head sector industry 21.49 25.60 -4.11 0.01 Head sector services 39.07 45.56 -6.49 0.00 Sumatera 18.22 19.08 -0.86 0.54 Jawa 69.74 71.36 -1.62 0.35 Kalimantan 2.16 2.40 -0.24 0.58 Sulawesi 2.81 2.53 0.28 0.52 NT 6.50 3.75 2.75 0.00 Maluku 0.31 0.60 -0.29 0.23 Papua 0.26 0.28 -0.02 0.72 # households 1843 20600 SOURCE: Susenas surveys 94 Table II.5.3 Characteristics of Extreme Poor and Other Poor Urban Households, 2010 (percentages unless otherwise indicated) Extreme Characteristics Poor Other Poor Difference p-value HH head female 14.83 13.92 0.91 0.58 HH head has no edu certificate 36.31 33.31 3.00 0.23 HH head has primary certificate 37.58 36.88 0.70 0.78 HH head has junior secondary certificate 14.05 15.50 -1.45 0.39 HH head has senior secondary certificate 11.47 14.17 -2.70 0.10 HH head has post secondary certificate 0.58 0.14 0.44 0.09 HH head employed 80.64 82.15 -1.51 0.38 Hhhead ill (disrupted work > 5 days past month) 7.99 8.81 -0.82 0.55 HH head age (years) 49.62 49.78 -0.16 0.81 HH size (number) 4.94 4.72 0.22 0.01 Number of children < 15 1.85 1.57 0.28 0.00 Number of men 15-65 1.35 1.41 -0.06 0.15 Number of women 15-65 1.45 1.45 0.00 0.98 Number of adults > 65 0.30 0.28 0.02 0.44 Head self-employed 50.28 54.43 -4.15 0.11 Head wage employed 37.27 38.74 -1.47 0.57 Head unpaid worker 1.12 1.86 -0.74 0.22 Head sector agriculture/extraction 38.34 40.52 -2.18 0.45 Head sector industry 21.25 21.73 -0.48 0.83 Head sector services 40.41 37.75 2.66 0.33 Sumatera 16.82 19.56 -2.74 0.16 Jawa 72.36 67.24 5.12 0.03 Kalimantan 2.32 2.00 0.32 0.54 Sulawesi 3.18 2.45 0.73 0.25 NT 4.74 8.18 -3.44 0.00 Maluku 0.37 0.26 0.11 0.39 Papua 0.22 0.30 -0.08 0.64 # households 1063 997 SOURCE: Susenas surveys 95 Table II.5.4 Characteristics of Urban Poor, Rural Poor and Rural non-Poor Households, 2010 (percentages unless otherwise indicated) Urban vs. Rural Poor vs. Rural Poor Non-poor Characteristics Urban Poor Rural Poor Rural Difference p-value Difference p-value non-poor HH head female 14.4 12.8 14.7 1.58 0.12 -1.92 0.00 HH head has no edu certificate 34.8 42.3 31.1 -7.46 0.00 11.17 0.00 HH head has primary certificate 37.2 41.4 38.4 -4.17 0.02 2.99 0.00 HH head has junior secondary certificate 14.8 10.2 13.6 4.52 0.00 -3.35 0.00 HH head has senior secondary certificate 12.8 5.7 13.2 7.06 0.00 -7.45 0.00 HH head has post secondary certificate 0.4 0.3 3.7 0.05 0.71 -3.35 0.00 HH head employed 81.4 87.0 87.5 -5.65 0.00 -0.48 0.41 Hhhead ill (disrupted work > 5 days past month) 8.4 8.9 9.2 -0.52 0.52 -0.33 0.49 HH head age (years) 49.7 49.1 49.3 0.60 0.18 -0.16 0.54 HH size (number) 4.8 4.8 3.7 0.01 0.84 1.08 0.00 Number of children < 15 1.7 1.8 1.1 -0.08 0.07 0.72 0.00 Number of men 15-65 1.4 1.3 1.2 0.06 0.06 0.10 0.00 Number of women 15-65 1.5 1.4 1.2 0.05 0.03 0.17 0.00 Number of adults > 65 0.3 0.3 0.2 -0.01 0.35 0.08 0.00 Head self-employed 52.3 71.0 67.6 -18.68 0.00 3.40 0.00 Head worker/employed 38.0 19.8 25.4 18.22 0.00 -5.63 0.00 Head unpaid worker 1.5 2.5 2.3 -1.01 0.02 0.23 0.48 Head sector agriculture/extraction 39.4 78.3 63.3 -38.83 0.00 14.95 0.00 Head sector industry 21.5 9.2 11.2 12.32 0.00 -2.05 0.00 Head sector services 39.1 12.6 25.5 26.51 0.00 -12.89 0.00 Sumatera 18.2 20.7 24.5 -2.43 0.27 -3.82 0.00 Jawa 69.7 52.8 52.3 16.96 0.00 0.52 0.72 Kalimantan 2.2 3.3 7.1 -1.12 0.05 -3.86 0.00 Sulawesi 2.8 9.6 9.2 -6.78 0.00 0.42 0.51 NT 6.5 6.5 4.7 -0.04 0.97 1.87 0.00 Maluku 0.3 1.8 1.1 -1.51 0.00 0.71 0.02 Papua 0.3 5.3 1.2 -5.08 0.00 4.16 0.00 # households 2,060 5,527 34,534 SOURCE: Susenas survey 96 Table II.5.5 Rates of Poverty, Extreme Poverty, and Poverty/Near Poverty by Household Head Characteristics (percent) Poverty or Extreme Near Characteristic Poverty Poverty Poverty No education certificate 19.85 10.70 34.39 Primary certificate 14.34 7.58 25.52 Junior secondary certificate 8.76 4.34 17.06 Senior secondary certificate 3.78 1.76 8.71 Post-secondary certificate 0.30 0.23 1.27 Not employed 9.50 5.09 17.54 Self-employed, agriculture 22.57 12.19 36.64 Self-employed, industry 11.00 4.68 18.02 Self-employed, services 7.97 3.90 15.71 Wage worker, agriculture 23.86 13.16 38.06 Wage worker, industry 9.27 4.89 19.29 Wage worker, services 4.46 2.15 9.08 Female 9.64 5.03 18.14 Ill 13.52 6.64 23.61 SOURCE: Susenas surveys 97 Table II.5.6 Household Log per Capita Expenditure Regressions, 2002 and 2010 2010. Difference 2010-2002 Variable coefficient SE t-stat coefficient SE t-stat Difference p-value HH head female 0.007 0.049 0.146 -0.059 0.016 3.620 -0.066 0.198 HH head age 0.000 0.007 0.021 0.009 0.003 3.230 0.009 0.221 HH head age squared 0.000 0.000 0.592 0.000 0.000 1.220 0.000 0.314 HH head ill (disrupted work > 5 days) -0.011 0.047 0.236 -0.026 0.019 1.392 -0.015 0.767 Head has primary certificate 0.143 0.026 5.453 0.119 0.013 8.870 -0.024 0.424 Head has junior secondary certificate 0.338 0.035 9.552 0.261 0.015 16.964 -0.077 0.046 Head has senior secondary certificate 0.507 0.037 13.803 0.518 0.017 31.391 0.011 0.782 Head has post secondary certificate 0.938 0.081 11.632 1.009 0.026 38.517 0.071 0.403 Avg years ed of HH members 0.026 0.023 1.145 0.011 0.010 1.031 -0.016 0.535 Avg years ed of HH members squared -0.001 0.002 0.448 0.000 0.001 0.212 0.001 0.746 Head not employed -0.058 0.054 1.064 -0.027 0.025 1.097 0.031 0.605 Head self-employed, agriculture -0.185 0.048 3.872 -0.198 0.018 11.294 -0.013 0.796 Head wage worker, agriculture -0.250 0.075 3.354 -0.174 0.026 6.737 0.076 0.329 Head self-employed, industry 0.084 0.039 2.137 0.063 0.022 2.815 -0.022 0.631 Head wage worker, industry -0.075 0.029 2.589 -0.060 0.014 4.414 0.015 0.639 Head wage worker, services -0.027 0.028 0.943 -0.012 0.011 1.111 0.014 0.638 Number of kids in HH, 0-5 -0.168 0.016 10.287 -0.201 0.006 31.506 -0.032 0.060 Number of kids in HH, 6-14 -0.135 0.010 13.613 -0.166 0.005 34.053 -0.031 0.004 Number of male adults in HH, 15-65 -0.067 0.013 5.270 -0.050 0.005 9.441 0.017 0.213 Number of female adults in HH, 15-65 0.000 0.017 0.028 -0.026 0.007 3.797 -0.026 0.139 Number of seniors in HH, > 65 -0.167 0.032 5.233 -0.136 0.014 9.794 0.031 0.367 Sumatera region -0.040 0.050 0.793 -0.064 0.018 3.554 -0.025 0.642 Kalimantan region 0.174 0.066 2.643 0.167 0.028 5.868 -0.006 0.928 Sulawesi region -0.071 0.070 1.021 0.218 0.029 7.465 0.289 0.000 NT region -0.126 0.063 1.999 -0.056 0.039 1.424 0.070 0.344 Maluku region - - 0.094 0.084 1.119 - - Papua region - - 0.115 0.056 2.067 - - Observations 3212 19588 Adjusted R squared 0.353 0.373 SOURCE: Susenas surveys. Note: Base category for head education is less than completed primary; for head occupation, self-employed in services; for region, Java/Bali. 98 Table II. 5.7 Household Log per Capita Expenditure Regressions with Community Fixed Effects, 2002 and 2010 2002 2010 Difference 2010-2002 Variable coefficient SE t-stat coefficient SE t-stat Difference p-value HH head female -0.043 0.039 1.123 -0.065 0.014 4.629 -0.022 0.600 HH head age 0.012 0.005 2.392 0.018 0.002 7.375 0.006 0.290 HH head age squared 0.000 0.000 1.580 0.000 0.000 5.184 0.000 0.486 HH head ill (disrupted work > 5 days) -0.011 0.034 0.306 -0.008 0.017 0.435 0.003 0.938 Head has primary certificate 0.106 0.018 6.001 0.118 0.012 9.913 0.011 0.603 Head has junior secondary certificate 0.219 0.026 8.333 0.211 0.013 15.855 -0.008 0.796 Head has senior secondary certificate 0.331 0.026 12.690 0.391 0.014 28.307 0.061 0.040 Head has post secondary certificate 0.618 0.039 15.759 0.776 0.018 42.603 0.158 0.000 Avg years ed of HH members 0.044 0.015 2.986 0.026 0.009 2.974 -0.017 0.309 Avg years ed of HH members squared -0.003 0.001 2.080 -0.002 0.001 2.208 0.001 0.486 Head not employed -0.064 0.048 1.337 -0.041 0.023 1.755 0.023 0.665 Head self-employed, agriculture -0.037 0.037 1.009 -0.063 0.015 4.290 -0.025 0.524 Head wage worker, agriculture -0.080 0.052 1.533 -0.130 0.021 6.225 -0.050 0.373 Head self-employed, industry 0.071 0.029 2.431 0.082 0.020 4.192 0.012 0.741 Head wage worker, industry -0.063 0.021 2.972 -0.076 0.011 6.950 -0.013 0.602 Head wage worker, services -0.019 0.021 0.924 -0.011 0.010 1.173 0.008 0.729 Number of kids in HH, 0-5 -0.165 0.012 13.674 -0.189 0.006 32.399 -0.024 0.073 Number of kids in HH, 6-14 -0.128 0.008 16.351 -0.162 0.004 37.036 -0.034 0.000 Number of male adults in HH, 15-65 -0.072 0.010 7.565 -0.067 0.005 14.074 0.006 0.595 Number of female adults in HH, 15-65 -0.065 0.010 6.724 -0.057 0.005 10.627 0.007 0.512 Number of seniors in HH, > 65 -0.117 0.024 4.900 -0.122 0.012 9.995 -0.004 0.869 Obs 22800 Adjust R squared 0.288 Source: Susenas surveys Note: Base category for head education is less than completed primary; for head occupation, self-employed in services; for region, Java/Bali. 99 Table II.6.1 Transition Matrices 2000-2007 Non-Poor in 2007 Poor in 2007 Panel A. All Indonesia Non-Poor in 2000 38,082 4,943 Percentages 89% 11% 85% Poor in 2000 4,752 2,681 Percentages 64% 36% 15% Total 42,834 7,625 85% 15% Panel B. Urban residents in 2000 Non-Poor 2000 15,213 1,041 Percentages 94% 6% 87% Poor 2000 1,736 632 Percentages 73% 27% 13% Total 16,949 1,673 91% 9% Source: IFLS 2000,2007 100 Table II.6.2 Correlations Between Annual Per Capita Expenditures and Mobility, with Correction for Measurement Error Source Indonesia Urban Rural Correlation 1993 and 2000 * 0.4288 0.4362 0.3322 1993 and 1997 * 0.4684 0.4656 0.3785 1997 and 2000 + 0.5280 0.5485 0.4555 2000 and2007 + 0.4888 0.5224 0.4018 Reliability Reliability ratio * 0.73 0.72 0.68 Reliability index * 0.86 0.85 0.83 Mobility Mobility index (1997-2000) + 0.4720 0.4515 0.5445 Mobility index (2000-2007) + 0.5112 0.4776 0.5982 Corrected correlation and mobility Corrected correlation (1997-2000) + 0.72 0.76 0.67 Corrected correlation (2000-2007) + 0.67 0.73 0.59 Corrected mobility index (1997-2000) + 0.28 0.24 0.33 Corrected mobility index (2000-2007) + 0.33 0.27 0.41 * Gibson & Glewwe (2005) + Authors' calculations using IFLS 2000,2007 101 Table II.6.3 Poverty Transitions and Individual and Household Characteristics, Urban 2000 sample Became Became Average log Poor in Poor in Non-Poor change in 2007 2007 in 2007 consumption (1) (2) (3) (4) Panel A. Highest Schooling Level Attained Completed primary or less 10.29% 7.22% 75.10% 16.02% completed junior secondary 5.96% 4.53% 80.77% 17.13% Completed senior secondary or higher 3.57% 2.59% 82.24% 14.44% Panel B. Unemployment Unemployed in 2007 7.47% 5.41% 76.43% 10.04% Not unemployed 10.44% 5.70% 66.04% 12.05% Panel C. Employment Categories Self employed 7.61% 5.89% 77.84% 8.93% Full time wage employee 7.42% 5.23% 75.93% 12.84% Business owner 2.52% 1.95% 80.00% 7.75% Panel D. Gender of Household Head Female 8.01% 6.95% 84.41% 2.71% Male 7.41% 5.12% 74.54% 11.47% All urban All urban All urban Non-Poor in Poor in All Urban Sample residents 2000 2000 residents Based on the sample of matched respondents in the 2000 and 2007-8 rounds of the IFLS. 102 Table II.6.4 Correlates of Poverty Transitions and Change in Consumption of Urban Households Poor in Became Poor in Became Non-poor Change in Logarithm of 2007 2007 in 2007 Expenditure (2000- 2007) (1) (2) (3) (4) Highest degree obtained, primary -0.006* -0.005** 0.015 0.029 [0.003] [0.003] [0.022] [0.021] Highest degree obtained, junior secondary -0.018*** -0.011*** 0.063** 0.032 [0.003] [0.003] [0.027] [0.021] Highest degree obtained, senior secondary -0.027*** -0.020*** 0.075*** 0.033* [0.003] [0.002] [0.027] [0.020] Highest degree obtained, college or above -0.041*** -0.028*** 0.013 [0.003] [0.002] [0.026] Female Dummy -0.001 -0.000 0.013 -0.006 [0.001] [0.001] [0.009] [0.005] Age 0.000 0.000 0.000 -0.001*** [0.000] [0.000] [0.001] [0.000] Age squared -0.000 -0.000** 0.000 0.000*** [0.000] [0.000] [0.000] [0.000] Became Unemployed (2000 to 2007) 0.025* 0.007 -0.123* -0.069* [0.013] [0.010] [0.074] [0.041] Business owner 2000 0.006 0.005 -0.058 -0.061 [0.025] [0.020] [0.235] [0.056] Employed Full-time 2000 0.033*** 0.032*** 0.114** -0.096*** [0.009] [0.009] [0.045] [0.027] Head of Household is female -0.009*** -0.001 0.078*** -0.094*** [0.003] [0.003] [0.022] [0.036] Age of the Head of Household -0.001* -0.001*** -0.003 0.012* [0.000] [0.000] [0.005] [0.006] 103 able II.6.4 – continued orrelates of Poverty Transitions and Change in Consumption of Urban Households Became Poor in 2007 Became Poor in 2007 Non-poor in 2007 Change in Logarithm of Expenditure (200 (1) (2) (3) (4) [0.000] [0.000] [0.000] [0.000] Household's highest degree obtained, primary -0.019*** -0.015*** 0.011 0.022 [0.004] [0.003] [0.027] [0.065] Household's highest degree obtained, junior secondary -0.029*** -0.024*** -0.023 0.007 [0.003] [0.003] [0.036] [0.068] Household's highest degree obtained, senior secondary -0.046*** -0.032*** 0.081** -0.052 [0.004] [0.004] [0.033] [0.070] Household's highest degree obtained, higher education -0.051*** -0.034*** -0.147** [0.004] [0.003] [0.074] Household, business owner in 2000 -0.036*** -0.023*** 0.016 [0.003] [0.003] [0.088] Household, employed full-time in 2000 -0.015*** -0.015*** -0.163*** 0.079** [0.005] [0.004] [0.061] [0.031] ons 23,458 20,744 2,538 19,308 0.032 -squared 0.133 0.120 0.0581 IFLS 2000,2007 mns 1, 2 and 3 are probit models (Marginal effects shown). Column 4: Linear regression coefficients shown nclude province dummies andard errors in brackets *** p<0.01, ** p<0.05, * p<0.1 Probit marginal effects shown controls for province. employment category is "self-employed" The omitted education category is "no school level completed" 104 Table II.6.5 Migration and Urbanizing Correlates of Poverty Transitions Became Poor in Became Non- Log change in Poor in 2007 2007 Poor in 2007 consumption (1) (2) (3) (4) Panel A: Migrant Household Migration (in any direction) -0.086*** -0.055*** 0.188*** 0.204*** (s.e) [0.004] [0.003] [0.014] [0.010] Other Controls YES YES YES YES Obs 42,389 34,924 7,465 41,146 R2 0.0120 0.00941 0.0145 0.011 Non-Poor in Sample All 2000 Poor in 2000 All Panel B: Rural to Urban Movements Change in Status: Rural to Urban -0.097*** -0.091*** 0.089* 0.231*** (s.e) [0.018] [0.012] [0.049] [0.013] Other Controls YES YES YES YES Obs 24,397 19,150 5,247 22,387 2 R 0.0194 0.0173 0.0241 0.0198 Rural Non-Poor Rural Poor in Sample All rural in 2000 2000 All Rural Panel C: Rural to Urban Migration and Non-Migration Movements Rural to Urban Migrant -0.180*** -0.130*** 0.342*** 0.543*** (s.e) [0.012] [0.011] [0.032] [0.020] Rural to Urban Non-Migrant -0.063*** -0.079*** -0.006 0.077*** (s.e) [0.024] [0.017] [0.056] [0.015] Other Controls YES YES YES YES Obs 22,477 17,523 4,954 22,079 2 R 0.024 0.0193 0.042 0.0360 Rural Non-Poor Rural Poor in Sample All rural in 2000 2000 All Rural SOURCE: IFLS 2000,2007 Note: (1) All observations in 2007 (2) All matched respondents who were not poor in 2000 (3) All matched respondents who were poor in 2000 and (4) All matched observations in IFLS 3 and 4 Probit models (Marginal Effects Shown). Other controls include age, gender, indicator for locality of residence in 2000, clustered standard error in brackets (clustered at the community level) 105 Table III.3.1 Participation in Raskin, Jamkesmas, and Credit Programs by Expenditure Quintile, 2010 (percent) 1st 5th Program Quintile 2nd Quintile 3rd Quintile 4th Quintile Quintile All Purchased rice under Raskin in last 3 months national 72.07 61.26 49.33 38.00 21.81 48.50 rural 73.99 70.20 63.43 56.61 37.40 60.33 urban 70.01 51.69 34.24 18.09 5.11 35.83 Has health card (Jamkesmas) national 40.86 32.02 23.93 18.37 10.28 25.09 rural 43.76 36.93 31.29 25.92 16.56 30.89 urban 37.75 26.76 16.06 10.28 3.56 18.88 Used health card in last 6 months (Jamkesmas) national 16.12 12.14 9.23 6.68 3.20 9.47 rural 15.81 13.21 11.67 9.39 5.33 11.08 urban 16.44 10.99 6.63 3.79 0.92 7.76 Obtained business credit through PPK program in last year national 0.76 0.70 0.74 1.03 0.84 0.82 rural 0.84 0.84 0.92 1.39 1.10 1.02 urban 0.68 0.55 0.56 0.64 0.57 0.60 Obtained business credit through P2KP program in last year national 0.96 0.95 0.86 0.94 0.52 0.85 rural 0.78 0.72 0.83 0.95 0.63 0.78 urban 1.16 1.18 0.89 0.94 0.40 0.92 Obtained business credit through other gov't program in last year national 2.32 2.47 2.33 2.77 2.07 2.39 rural 2.35 2.71 2.59 3.53 2.83 2.80 urban 2.28 2.21 2.04 1.95 1.25 1.95 Source: Susenas survey 106 Table III.3.2 Participation in Unconditional and Conditional Cash Transfer Programs by Expenditure Quintile, 2007 (percent) 2nd Program 1st Quintile Quintile 3rd Quintile 4th Quintile 5th Quintile All Unconditional Cash Transfers (BLT) national 0.504 0.352 0.233 0.152 0.047 0.261 rural 0.572 0.463 0.377 0.257 0.156 0.367 urban 0.390 0.205 0.161 0.074 0.018 0.175 Conditional Cash Transfers (PKH) national 0.009 0.004 0.000 0.001 0.000 0.005 rural 0.014 0.004 0.000 0.000 0.000 0.005 urban 0.008 0.001 0.000 0.001 0.000 0.005 Source: IFLS, 2007/8 107 108 Table III.3.3 Net and Gross Enrollment by Expenditure Quintile, 2010 (percent) Calculated for ages 1st 2nd 3rd 4th 5th Program (years): Quintile Quintile Quintile Quintile Quintile All Primary net enrollment national 7-12 90.8 92.5 92.1 91.8 90.5 91.5 rural 90.2 93.0 92.5 92.2 91.4 91.9 urban 91.5 92.0 91.6 91.2 89.5 91.2 Primary gross enrollment national 7-12 103.5 104.3 103.0 103.0 100.3 102.8 rural 104.1 106.6 105.0 103.9 102.5 104.4 urban 102.8 101.4 100.7 101.7 97.6 100.9 Junior secondary net enrollment national 13-15 60.0 69.0 72.6 74.1 74.8 70.1 rural 54.7 64.9 69.6 73.4 73.1 67.1 urban 65.7 74.0 76.0 75.1 76.7 73.5 Junior secondary gross enrollment national 13-15 76.4 86.1 91.8 94.9 95.4 88.9 rural 69.9 80.6 88.6 93.0 93.4 85.1 urban 83.5 92.8 95.5 97.4 97.7 93.4 Senior secondary net enrollment national 16-18 31.7 46.7 53.7 61.1 66.2 51.8 rural 23.9 33.9 42.5 52.5 63.7 43.3 urban 39.4 58.5 64.0 70.0 68.7 60.1 Senior secondary gross enrollment national 16-18 41.4 60.7 69.2 77.9 84.9 66.8 rural 31.2 43.6 53.9 65.2 82.7 55.3 urban 51.7 76.7 83.5 90.9 87.2 78.0 Post secondary net enrollment national 19-24 0.3 0.8 1.4 2.4 6.0 2.2 rural 0.2 0.5 0.6 1.1 2.8 1.0 urban 0.4 1.0 2.3 3.8 9.3 3.4 Post secondary gross enrollment national 19-24 0.3 0.8 1.5 2.7 6.4 2.3 rural 0.2 0.5 0.6 1.1 3.0 1.1 urban 0.4 1.1 2.4 4.3 9.9 3.6 Source: Susenas survey 109 Table III.3.4 Completed Immunizations and Birth Medic Attendance by Expenditure Quintile, 2010 (percent) Calculated 1st 2nd 3rd 4th 5th Program for ages: Quintile Quintile Quintile Quintile Quintile All Medical professional attended birth national 0-24 months 70.3 82.0 85.7 88.9 93.1 82.5 rural 57.4 72.5 78.5 80.8 88.6 73.7 urban 85.3 93.7 93.6 98.2 99.0 93.0 Completion of immunizationsa 12-24 national months 54.8 58.0 57.9 65.1 69.3 60.0 rural 50.7 52.3 56.7 57.0 66.0 55.6 urban 59.3 65.3 59.1 74.5 73.8 65.1 Source: Susenas survey a Received complete course of vaccinations in Polio, measles, BCG and DPT 110 Table III.3.5 Urban wage employees: receipt of job related benefits by expenditure quintile, 2007 (percent) 1st 2nd 3rd 4th 5th Benefit Quintile Quintile Quintile Quintile Quintile Meals 34.8 28.6 25.5 27.1 27.1 Food (provisions) 4.0 6.0 6.0 5.6 6.9 Housing Benefits 1.6 2.4 3.4 6.7 11.5 Car 0.6 1.2 1.8 2.6 3.9 Transport Support 4.3 8.0 13.3 16.9 22.1 Covers Health Expenses 8.4 12.2 15.6 20.0 26.3 Health Insurance 7.9 15.6 22.2 28.5 37.8 Health Clinic 5.3 7.8 12.5 14.7 16.5 Any health benefits 14.1 23.9 34.0 42.0 51.8 Credit 17.3 19.5 23.5 29.9 35.2 Pension 3.5 6.7 11.6 18.7 27.4 Severance Pay 8.1 12.5 19.4 22.5 24.6 Source: IFLS 111 Table III.3.6 Access to Basic Services by Expenditure Quintile, 2010 (percent) 1st 2nd 3rd 4th 5th Household has: Quintile Quintile Quintile Quintile Quintile All Access to safe water national 78.4 83.9 86.6 89.1 92.0 86.0 rural 68.8 73.7 77.7 80.8 85.4 77.3 urban 88.6 94.9 96.0 98.0 99.1 95.3 Piped water/branded bottled water national 23.5 31.2 36.7 41.6 48.6 36.3 rural 12.2 17.0 21.4 25.1 31.9 21.5 urban 35.7 46.5 53.2 59.2 66.4 52.2 Toilet national 72.5 81.2 85.9 89.5 93.9 85.3 rural 56.5 65.9 73.0 79.5 88.5 74.1 urban 85.0 93.2 96.3 98.0 99.0 94.7 Electricity national 86.4 91.3 92.4 93.4 94.5 91.6 rural 76.1 83.9 86.2 87.7 90.0 84.8 urban 97.4 99.1 99.1 99.4 99.4 98.9 Source: Susenas 112 Table III.3.7 Housing Quality by Expenditure Quintile, 2010 (percent) 1st 2nd 3rd 4th 5th Quintile Quintile Quintile Quintile Quintile All Roof good quality (concrete, tile, shingle, iron sheeting, asbestos) national 93.0 96.1 97.2 98.0 98.7 96.6 rural 88.5 93.3 95.1 96.3 97.6 94.2 urban 97.7 99.0 99.4 99.8 99.8 99.1 Walls good quality (wood, brick) national 82.0 87.8 91.6 94.0 96.6 90.4 rural 77.0 82.4 86.5 90.0 94.2 86.0 urban 87.4 93.6 97.0 98.3 99.1 95.1 Floor good quality (permanent, non-earthen) national 80.4 87.7 90.8 93.0 96.0 89.6 rural 72.8 81.5 85.7 89.0 93.7 84.5 urban 88.5 94.4 96.3 97.4 98.5 95.0 Good housing quality national 68.0 78.9 84.7 88.6 93.3 82.7 rural 57.5 68.5 75.7 81.3 89.0 74.4 urban 79.3 90.0 94.2 96.4 98.0 91.6 Source: Susenas 113 Table III.3.8 Urban households: Share reporting being victims of crimes by expenditure quintile, 2007 (percent) 1st 2nd 3rd 4th 5th Quintile Quintile Quintile Quintile Quintile Theft (except crop theft) 3.1 4.1 4.6 5.2 6.1 Any Theft 3.9 5.3 5.4 6.2 6.9 Mugging or Robbery 0.1 0.3 0.1 0.9 1.5 Source: IFLS Note: Shows percent of households reporting that a household member was ever a victim of the indicated crime 114 Table III.5.1 Shares of Raskin and Jamekesmas Benefits received by poor/near-poor and others by region, 2010 Program Sumatera Jawa Kalimantan Sulawesi NT Maluku Papua Raskin urban poor/near poor 0.44 0.35 0.27 0.29 0.52 0.38 0.19 others 0.56 0.65 0.73 0.71 0.48 0.62 0.81 rural poor/near poor 0.37 0.35 0.28 0.44 0.42 0.45 0.59 others 0.63 0.65 0.72 0.56 0.58 0.55 0.41 Jamkesmas urban poor/near poor 0.39 0.37 0.20 0.25 0.58 0.39 0.20 others 0.61 0.63 0.80 0.75 0.42 0.61 0.80 rural poor/near poor 0.38 0.43 0.31 0.43 0.46 0.43 0.46 others 0.62 0.57 0.69 0.57 0.54 0.57 0.54 Source: Susenas 115 Table III.5.2 Distribution of Provincial Participation Rates of Urban Poor/Near-Poor in Raskin and Jamkesmas, 2010 Percentile 0.10 0.25 0.50 0.75 0.90 Raskin 38.5 44.0 53.7 66.8 78.1 Jamkesmas 24.3 31.7 38.0 46.0 60.2 Source: Susenas surveys 116 Figure II.4.1 Urban consumption distribution and poverty lines Population density 9.9% below national poverty line log pe expenditures Figure II.4.2 Urban Share of Population and Share of Total Poor, 2002-2010 117 SOURCE: Susenas surveys Figure II.4.3 Trends in Urban Poverty by Region, 2002-2010 SOURCE: Susenas surveys 118 Figure III.4.1 Concentration Curves for Social Assistance and Credit Programs (Urban) a. Concentration Curves for Participation b. Concentration Curves for Health Card in Raskin and Use of Health Card 1 1 Cumulative share of benefit .8 .8 .6 .6 .4 .4 .2 .2 0 0 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 Cumulative share of sample, poorest to richest Cumulative share of sample, poorest to richest Per capita Per capita 45% line expenditures 45% line expenditures Raskin Healthcard Healthcard use d. Concentration Curves for C. Concentration Curves for Participation Business Credit in UCT Program 1 1 .8 Cumulative share of benefit .8 .6 .6 .4 .4 .2 .2 0 0 .2 .4 .6 .8 1 Cumulative share of sample, poorest to richest 0 45% line Per capita 0 .2 .4 .6 .8 1 expenditures Cumulative share of sample, poorest to richest PKK-Kecamatan P2KP/PNPM- Development Urban 45% line Per capita expenditures Business Credit Government Program UCT indicator 119 Figure III.4.2 Concentration Curves for Education (Urban) a. Concentration Curves for b. Concentration Curves for School Enrollment-Primary School Enrollment-Junior 1 1 Cumulative share of benefit .8 .8 .6 .6 .4 .4 .2 .2 0 0 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 Cumulative share of sample, poorest to richest Cumulative share of sample, poorest to richest Per capita Per capita 45% line expenditures 45% line expenditures Primary Education Children 7-14 Junior Secondary Children 13-17 c. Concentration Curves for d. Concentration Curves for School Enrollment-Senior School Enrollment- PostSecondary 1 1 Cumulative share of benefit .8 .8 .6 .6 .4 .4 .2 .2 0 0 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 Cumulative share of sample, poorest to richest Cumulative share of sample, poorest to richest Per capita Per capita 45% line expenditures 45% line expenditures Senior Secondary Children 16-20 Post Secondary 120 Figure III.4.3 Concentration Curves for Maternal and Child Health (Urban) a. Concentration Curves for b. Concentration Curves for Child Vaccinations Birth Attendance by Medical Professional 1 1 Cumulative share of benefit .8 .8 .6 .6 .4 .4 .2 .2 0 0 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 Cumulative share of sample, poorest to richest Cumulative share of sample, poorest to richest Per capita Per capita 45% line expenditures 45% line expenditures Complete immun. Children Medics assist with Child 24 months courses children 1-2 ages 12-24 months birth of child ages 1 & 2 or younger 121 Figure III.5.1.a Share of poor/near poor receiving Raskin by region, 2010 Source: Susenas survey Figure III.5.1.b Share of poor/near poor receiving Jamkesmas health card by region, 2010 Source: Susenas survey 122 Figure III.5.2a Share of Urban Poor/Near Poor and Others Receiving Raskin, by Region 2010 (percent) Source: Susenas survey Figure III.5.2b Share of Rural Poor/Near Poor and Others Receiving Raskin, by Region 2010 (percent) Source: Susenas survey 123 Figure III 5.3a Share of Urban Poor/Near Poor and Other with Jamkesmas Health Card, by Region, 2010 (percent) Source: Susenas survey Figure III 5.3b Share of Rural Poor/Near Poor and Others with Jamkesnas Health Card, by Region 2010 (percent) Source: Susenas survey 124 Figure III.5.4 Share of Urban Poor/Near Poor and Others Receiving Raskin, by Quintile of Province Median per Capita Expenditure 2010 (percent) Source: Susenas survey Figure III.5.5 Share of Urban Poor/Near Poor and Others with Jamkesmas Health Card, by Quintile of Province Median per Capita Expenditure 2010 (percent) Source: Susenas survey 125 APPENDIX 1 SUMMARY OF PROGRAMS SERVING THE URBAN POOR IN INDONESIA Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage The rice subsidy program was initially introduced in 1998 as Operasi Pasar Khusus (OPK), renamed Raskin (Beras untuk Orang Consumption and Basic Needs Rice subsidy National High Raskin in 2002. In 2008, Raskin distributed 3 Miskin) / Rice for the Poor million tons of rice to a target of 18.5 million households. The 2010 estimated budget was Rp 8.9 billion. The Indonesian government's overall fuel subsidy program benefits the poor but is highly regressive, Scaling the controversial process of phasing them out Consumption and Basic Needs Fuel Subsidies Fuel Subsidy National Down continues. Reallocating the savings in the budget towards pro-poor policies has been a cornerstone of the Indonesia reform agenda. JSLU and JSPACA form part of the social security system for two important but relatively specialized disadvantaged groups. In 2011 this JSPACA(Jaminan Sosial program has a total of 19,500 beneficiaries. Penyandang Cacat Berat) / Consumption and Basic Needs Cash transfers National Very Low Indonesia has yet to develop a comprehensive Social Security for the Heavily database of disability hence assessment of this Disabled program's coverage against the overall need (population of severely disabled people) remains difficult JSLU and JSPACA form part of the social JSLU (Jaminan Sosial Lanjut security system for two important but relatively Consumption and Basic Needs Usia) / Social Security for the Cash transfers National Very Low specialized disadvantaged groups. In 2011 the Aged program has 13,250 beneficiaries and a budget allocation of Rp 44.7 billion As a compensation for fuel subsidies, a monthly BLT (Bantuan Langsung subsidy of RP 100,000 per month was paid to Consumption and Basic Needs Tunai) / Fuel Price Unconditional cash transfer National Ended approximately 19 million poor households from Compensation Program 2005-2009, but amid concern with targeting and unconditionality, has been discontinued 126 Appendix 1 (Continued) Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Consumption and Basic Needs The government provides IDR 30 million to Co-contributory insurance that financial organizations that manage the funds for Askesos (Asuransi pays cash benefit for informal 3 years and workers contribute IDR 5 ,000 per Kesejahteraan Sosial) / Social National Low workers in case of sickness, month of membership.Total government budget Welfare Insurance work-related injury or death. allocation is IDR 40 billion per year. 2010 membership is approximately 300,000 PKSA is a special conditional cash transfer for vulnerable children first piloted in 2009, targeting . abandoned infants/infants with special needs (5 years or younger), abandoned children (6-18 years PKSA(Program Kesejahteraan Conditional cash transfer with old), street children (6-18 years old), children with Consumption and Basic Needs Sosial Anak) / Child Social savings account targeted at National Low criminal/legal issues (6-16 years old) and disabled Welfare Program children children (0-18 years old) Conditionalities differ for different groups (staying at school, stop working on the street, not getting into criminal behavior etc). The total budget for 2011 is Rp 287.1 billion. 127 Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Consumption/Education/Heath PKH is a conditional cash transfer program primarily designed to improve maternal and neonatal health as well as children education among poor households. First introduced in 2007, the program was piloted in 7 provinces. In 2010 PKH is implemented in 20 provinces, targeting 816,000 very poor households, with budget allocation of IDR 1.3 trillion (0.02 percent of GDP). The realization of that target is IDR 929.4 billion for 774,293 households. In 2011, PKH targets 1,116,000 households in 25 provinces with budget allocation of IDR 1.6 trillion. The program is expected to cover all provinces by 2012 and cover all districts within the provinces by 2014. PKH (Program Keluarga Household conditional cash Currently priorities are given to areas with the Harapan) (urban areas) / transfer based on education / Regional High most need i.e. high number of very poor Hopeful Family Program health outcomes households, but where access to health care and education is available. Beneficiaries consist of households with children younger than 15 years old (or 15-18 years but having not completed 9th grade) and/or pregnant or lactating women. Depending on family structure and their obedience in fulfilling educational and health requirements, households receive RP 600,000 to 2,200,000 per year. Conditionality of the cash transfer includes: (1) children are enrolled in school and attend at least 85 percent of school days. (2) Pregnant and lactating mothers as well as infants of 0-6 years of age regularly visit health facilities for health checks. 128 Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Consumption and Basic The ministry of education, in coordination with Needs/Education six other ministries, launched the School Feeding Program (Program Makanan Tambahan Anak Sekolah) in 2010. The program provides additional food for kindergarten and elementary school students in 27 less developed districts in Indonesia. In 2010 the program targets around 1.4 million kindergarten and elementary students in PMT-AS ( Program Makanan public schools (managed by the Ministry of Tambahan Anak Sekolah) / Additional food supplements Regional High Education) as well as Islamic schools (Managed School Supplementary given to schoolchildren by the Ministry of Religious Affairs). Students Feeding Program receive three meals every week. A budget of IDR 218 billion is allocated, with each meal in Eastern areas of Indonesia estimated to cost IDR 2,600 while that in western parts of Indonesia costs IDR 2,250. The program prescribes that the food provided to students must be obtained locally. A similar school feeding program had also taken place in the early 1990s. Pro-poor scholarships originated in 1998 under BSM (Biasiswa untuk Siswa the Jaring Pengaman Social (JPS)-Scholarship Education Miskin) / Scholarships for the Scholarships National High program with the objective of reducing dropouts, poor especially among girls in rural areas. The Ministry of Manpower and Transmigration (MoMT) oversees the Technical and Vocational Education and Training (TVET) Centers or known as Balai Latihan Kerja (BLK). The BLK provides vocational training and job placement BLK(Balai Latihan Kerja) / services, to formal and informal workers. Courses Technical and Vocational Education/Employment Job-training programs National Low are provided free of charge, though a few BLK Education and Training also provide non-subsidized courses on the side. (TVET) Centers BLK exists in all provinces and at some district level. BLK expansion as a platform for increasing employability is a priority program for 2012. Based on 2011 data there are currently 237 BLKs, to be expanded to 313. Balai Latihan. 129 Appendix 1 (Continued) Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Education School support was initiated 1998 as a companion program to individual scholarships, as the Jaring Bantuan Operasi Sekolah Pengaman Social (JPS)-DBO. The BOS scheme (BOS) / School Operational School support fund National High currently also includes efforts to improve the Assistance quality of education via improvements in school- based management. This program was begun in 2006 in 14 provinces, Expansion of existing SATAP (Sekolah Satu Atap) / and has resulted in the construction of over 1500 Education elementary schools to include Regional High Schools Under One Roof schools (both secular and religious). The focus junior secondary schools has been on the most underserved districts. This CDD program began in 2006, targeting predominantly rural but also urban areas in Organization of community districts and provinces with high maternal and into "alert networks" for data, child mortality (primarily in Nusa Tenggara). notification and referrals; Program Desa Siaga (urban This program is run by the Ministry of Health Health health literacy training by National Scaling Up areas) / Alert Village Program but supported by donors and central/local health workers to promote governments. Initial evaluation suggests that maternal and child health e.g. while demand for health services has increased, safe delivery, family planning effectiveness is limited by a lack of supply response. Jampersal is a new program (started in early 2011) that guarantees Consultation and delivery Free delivery care, including care are provided in health centers or 3rd class Jampersal (Jaminan pre-natal and post-natal wards in hospitals. In its first year of Health Persalinan) / Universal National Scaling Up consultations, universally to implementation, 2011, this program allocated IDR Delivery Care all women. 1.2 trillion with the target of covering 2.6 million deliveries or 60% of the total estimated deliveries (4.8 million). Basic Vaccinations are provided for free for all Vaksinasi Dasar untuk Balita / children of 0-5 years old. These vaccinations Health Free vaccination for Under-5s National High Universal Vaccination include BCG, DPT1-3, HepB3, Polio and Measles. 130 Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Health National social health programs were started in 1994 via a health card program (Kartu Sehat) entitling holders to free health care at subdistrict health centers and public hospitals. In 1998, this program evolved into the health component of the larger social safety net program (Jaring Pengaman Social (JPS) - Kesehatan). Between 2001-2005 Jamkesmas (Jaminan the scheme was paid for by fuel subsidy cuts and Kesehatan bagi Masyarakat Provision of individual health named the PKPS-BBM. In 2005, under the new National High Miskin) / Health Guarantees insurance cards Yudhoyodo administration the program was re- for the Poor housed under PT Askes, government employee insurance agency and renamed Askeskin (Asuransi Kesehatan bagi Keluarga Miskin) in 2005. In 2008, Askeskin was reformed into Jamkesmas. The practice of funding health schemes as an effective and efficient offset to cuts in the fuel subsidy and the sustainability of this practice is a source of concern. Family planning with a community emphasis has Family planning and been a significant movement in Indonesia since Program Kependudukan dan reproductive health-related the 1970s, and is implemented by the National Health Keluarga Berencana National / activities and education National High family Planning Coordinating Board (BKBBN). Program for Family Planning (currently focusing on The family planning program in Indonesia limiting births to 2) graduated from USAID support in 2009. PNPM consolidates a number of pre-existing PNPM-Urban / National Community-directed block community empowerment programs, including Infrastructure Program for Community National High grants grants for public works the PDM-DKE, the PPK, UPP/P2KP and others. Empowerment See Burger et al (2011) Program Pembinaan dan Improvement to target areas Pangembangan Infrastruktur This program covers various central government for infrastructure development Infrastructure Permukiman / Program for National High activities including regulation, guidance and and waste water facilities with Building and Developing financing on-site systems (district / city) Local Infrastructure 131 Appendix 1 (Continued) Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Infrastructure Slum upgrading and new housing development, shelter This ADB-supported project was implemented NUSSP (Neighborhood financing, improvement of Completed with 32 local city governments in 12 provinces. Upgrading and Shelter Sector Regional urban planning and in 2010 ADB estimates shelter improvements for 3.1 Program) / management systems, million poor people instritutional strengthening REKOMPAK (Rehabilitasi Small grants for rebuilding dan Rekonstruksi Masyarakat housing and reconstruction of dan Permukiman Berbasis This CDD program was a response program small scale infrastructure, as Infrastructure Komunitas) (urban areas) / Regional Moderate taking place after the Merapi disaster; some well as other priority Community-Based Settlement activities were brought under P2KP investments; community Rehabilitation and education Reconstruction Project Training and other employment support activities PNPM-Urban (Program Provision of training represent one of initially three lines of activites - Nasional Pemberdayaan programs and sponsored infrastructrure, economic and social activities Employment Masyarakat) / National economic activities, National High sponsored under PNPM-Urban. In some Program for Community development of work groups communities labor is also generated via work on Empowerment and market areas etc the funded infrastructure projects. In 2004, the government launched a scheme of microcredit (KUM-LTA), consolidating previous schemes under various ministries including Provision of collateral-free KUM (Kredit Usaha Mikro) / KUBE (Kelompok Usaha Bersama) under the Entrepreneurship loans up to Rs 5 million to National High Credit for Micro Enterprise Ministry of Social Affairs, the Ministry of microbusiness owners Cooperatives and SME's Revolving Fund for Smallholders and a number of other schemes targeted more at rural areas/agriculture. KUR was started in 2007 and is a collaborative arrangement between the government, assurance Subsidized guarantees for institutions (PT Askrindo and Jamkrindo) and a KUR (Kredit Usaha Rakyat) / Entrepreneurship banks that make loans to National High number of banks (including BRI, Bank Mandiri, Credit for The People microenterprises Bank Syariah Mandiri etc). Up to 2009, the KUR has supported over 2 million individuals with a nominal amount of RP 15 trillion 132 Appendix 1 (Continued) Need Addressed by the Current Program Name Main Instrument Scope Comment Program Coverage Entrepreneurship Loans are one component of the economic support PNPM-Urban / National Microcredit loans via Scaling activities that represent one of initially three lines Program for Community National revolving loan fund Down of activites - infrastructrure, economic and social Empowerment activities sponsored under PNPM-Urban Sources: Suryahadi et al (2011); www.ilo.org. Kemensos - Kementerian Sosial (Ministry of Social Affairs) Kemdiknas – Kementerian Pendidikan Nasional (Ministry of Education) Kemenko Kesra-Kementerian Koordinator Bidang Kesejahteraan Rakyat- “Coordinating Ministry for People's Welfare” Kemen. PU Kementerian Pekeerjaan Umum –Ministry of Public Works Bulog- Bureau of Logistics Kemenkeu – Ministry of Finance Kemenkes – Ministry of Health 133 REFERENCES African Development Bank (2011) Indonesia: Neighborhood Upgrading and Shelter Sector Project ADB Completion Report. Retrieved 24 January from: http://www.adb.org/documents/PCRs/INO/35143- 013-ino-pcr.pdf African Development Bank. (2010). “Indonesia: Has the Multi-subsector Approach been Effective for Urban Services Assistance? Retrieved 24 January from: http://www.adb.org/Documents/Evaluation/Learning-Curves/SES/LC-multisubsector-approach- ino.pdf Agrawal, P. (1999). Urban land consolidation: a review of policy and procedures in Indonesisa and other Asian countries. [10.1023/A:1007150828772]. GeoJournal, 49(3), 311-322 Asian Development Bank (ADB), International Labour Organization (ILO), Islamic Development Bank (IDB). Indonesia: Critical Development Constraints. Mandaluyong City, Philippines: Asian Development Bank, International Labour Organization and Islamic Development Bank, 2010. Asian Development Bank (ADB). Access to Justice for the Urban Poor. Mandaluyong City, Philippines: Asian Development Bank, 2010. Asian Development Bank. Neighborhood Upgrading and Shelter in Indonesia: Asian Development Bank, October 2008. Baker, Judy L. (2008). “Urban Poverty: A Global Overview.” World Bank, Washington D.C. January, 2008. Beazley, H. "'Vagrants Wearing Make-Up': Negotiating Spaces on the Streets of Yogyakarta, Indonesia." Urban Studies 39.9 (2002). Bessell, S. "Indonesian Children’s Views and Experiences of Work and Poverty’." Social Policy and Society 8.4 (2009): 527-40. Buckley, R. M., & Kalarickal, J. (2006). Thiry Years of World Bank Shelter Lender: What Have We Learned. Washington, DC: The World Bank. Burger, Nicholas, Peter Glick, Francisco Perez-Arce, Lila Rabinovich, and Joanne Yoong.A Qualitative Study of the Program Nasional Pemberdayaan Masyarakat (PNPM)- Urban in Indonesia. Draft study for World Bank. RAND, 2010. Central Board of Statistics (BPS), and ILO-IPEC Jakarta. Working Children in Indonesia 2009. Jakarta: BPS-Statistics Indonesia and International Labor Organization, 2010. Chen, Shaohua, and Martin Ravallion, The World Bank, The Developing World Is Poorer Than We Thought, But No Less Successful in the Fight against Poverty, Development Research Group August 2008. Available at: http://wwwwds.worldbank.org/external/default/WDSContentServer/IW3P/IB/2010/01/21/000158349 _20100121133109/Rendered/PDF/WPS4703.pdf. Chomistriana, Dewi. Inclusive Cities. Urban Development Series. Eds. Lindfield, Michael and Florian Steinberg. Manila: Asian Development Bank, April 2011. 134 Cuevas, Sining, et al. Informal Employment in Indonesia. Manila, Philippines: Asian Development Bank, April 2009. Deb, Partha, and Papa Seck. Internal Migration, Selection Bias and Human Development: Evidence from Indonesia and Mexico: United Nations Development Programme, 2009. Dimas, Harlan. Street Vendors: Urban Problem and Economic Potential. Bandung: Center for Economics and Development Studies, Department of Economics, Padjadjaran University, 2008. Donaldson, Dayl S. Social Assistance Policy and Programs Evolution in a Historical Context. Background Paper prepared for the Indonesia Social Assistance Public Expenditure Review, 2011. Indonesia Dursin, Kanis. "Mobile Classes a Lifeline to Dropouts". Jakarta, February 2010. Inter Press Service News Agency. 21 September 2011. Available at: http://ipsnews.net/news.asp?idnews=50272. Firman, T., B. Kombaitan, and P. Pradono. The Dynamics of Indonesia's Urbanisation, 1980–2006', Urban Policy and Research 25 3 (2007): 433–454. Firman, Tommy. The Patterns of Indonesia’s Urbanization, 1980-2007. School of Architecture, Planning and Policy Development, Institute of Technology, Bandung, Indonesia, Urban Policy and Research 25(4): 433-454. Foster, James, J. Greer, and Eric Thorbecke. “A Class of Decomposable Poverty Measures,” Econometrica 52(3) (1984): 761-765. Frankenberg, E., and D. Thomas. (2000). The Indonesia Family Life Survey (IFLS): Study Design and Results from Waves 1 and 2. Report No. DRU-2238/1. Santa Monica, CA: RAND Corporation. Frankenberg, E., and L.A. Karoly. (1995). The 1993 Indonesian Family Life Survey: Overview and Field Report. Report No. DRU-1195/1. Santa Monica, CA: RAND Corporation. Gibson, John, and Paul Glewwe (2005). “Analysis of Poverty Dynamics” in Handbook on Poverty Statistics: Concepts, Methods and Policy Use. United Nations Statistics Division, December 2005. Glewwe, Paul (2005). .How Much of Observed Economic Mobility is Measurement Error? A Method to Reduce Measurement Error, with an Application to Vietnam, Department of Applied Economics, University of Minnesota. Glewwe, Paul, and Hai-Anh Dang (2005). .Has Vietnam’s Economic Growth in the 1990’s Been Pro- Poor? An Analysis of Data from the Vietnam Living Standards Surveys, Department of Applied Economics, University of Minnesota. Glick, P., and M. Razakamanantsoa (2006). “The Distribution of Public Expenditures in Madagascar over the 1990s: Increasing Progressivity Without Economic Growth?” Journal of African Economies 15(3): 399-433. Hartono, Djoni, and Tony Irawan. Decentralization Policy and Equality: A Theil Analysis of Indonesian Income Inequality. Bandung: Department of Economics, Padjadjaran University, October 2008. 135 Hill, Hal (ANU), Dionisius Ardiyanto (ERIA, Jakarta) and Haryo Aswichayono (CSIS) Indonesian Industrialization: Jobless Growth?, Presentation at Indonesia Update Conference 2010, The Australian National University. Available at: http://www.crawford.anu.edu.au/acde/ip/pdf/update/2010/Session_1_-_Hill_&_Aswicahyono.pdf. Integrated Regional Information Networks (IRIN). "Indonesia: Jakarta's Slums Struggle with Sanitation". Jakarta, 16 April 2010. Integrated Regional Information Networks. 21 September 2011. Available at: http://www.irinnews.org/report.aspx?reportid=88833. International Labour Organization. Social Security for Informal Economy Workers in Indonesia: Looking for Flexible and Highly Targeted Programs. Jakarta: International Labour Organization, 2010. Jakarta Post, The. "Affordable Food Courts Benefit Vendors, Office Workers". Jakarta, July 2010. The Jakarta Post. 07 October 2011. Available at: http://www.thejakartapost.com/news/2010/07/24/affordable-food-courts-benefit-vendors-office- workers.html. Klaveren, Maarten van, et al. An Overview of Women’s Work and Employment in Indonesia. Amsterdam: Amsterdam Institute for Advanced Labour Studies, University of Amsterdam, May 2010. Lu, Yao. "Rural-Urban Migration and Health: Evidence from Longitudinal Data in Indonesia." Social Science & Medicine 70 (2010): 412-19. McCarthy, P. (2003). “Understanding Slums: Case Studies for the Global Report on Human Settlements,” The case of Jakarta, Indonesia. University College of London. 2003 McCarthy, Paul. "The Case of Jakarta, Indonesia". 2003. UN-Habitat. 05 October 2011. Available at: http://www.ucl.ac.uk/dpu-projects/Global_Report/cities/jakarta.htm. McCulloch, Neil; Peter Timmer, and Julian Weisbrod (2007). “Pathways Out of Poverty During an Economic Crisis: An Empirical Assessment of Rural Indonesia”. Center for Global Development Number 115, March 2007. Mercy Corps Indonesia. (2008). "Urban Poverty Reduction Strategy." Urban Bulletin No. 1 Muhammed Ally, Sahr. Always on Call: Abuse and Exploitation of Child Domestic Workers in Indonesia. Vol. 17: Human Rights Watch, 2005. Mulyana, A. (1990). “An evaluation of selected social impacts of an urban development program in Bandung, West Java Province, Indonesia”. Belconnen ACT: University of Canberra. Murray, J., Rosenberg, R. (2006) “Rosenberg Community-Managed Loan Funds: Which Ones Work?” CGAP (Consultative Group to Assist the Poor), FocusNote_36 Ochoa, Catalina (2011). The Community Based Neighborhood Development Approach: Evolution, Definition, and Key Aspects Moving Forward. Washington DC: World Bank. Olken, B., Onishi, J., Wong, S. (2011). “Evidence from a field experiment on health and education in Indonesia,”Should Aid Reward Performance? Working Paper, MIT. August 2011 136 Peters, R. "The Wheels of Misfortune: The Street and Cycles of Displacement in Surabaya, Indonesia." Journal of Contemporary Asia 40.4 (2010): 568-88. Pirmana, Viktor. Earnings Differential between Male-Female in Indonesia: Evidence from Sakernas Data. Bandung, Indonesia: Center for Economics and Development Studies, Padjadjaran University, August 2006. Pitriyan, Pipit, and Ahmad Komarulzaman. The Effect of Income Shocks and Credit Constraint on Child Labor Participation and Poverty: The Case of Indonesia: Department of Economics, Padjadjaran University, June 2010. Pradhan, M., Suryahadi, A., Sumarto, S., Pritchett, L. (2000). “Measurements of Poverty in Indonesia: 1996, 1999 and Beyond,” SMERU. Pradhan, Menno, Vijayendra Rao and Cristina Rosenberg (2010). ”The Impact of the Community Level Activities of the Second Urban Poverty Project (UPP).” Draft available at: http://www.aiid.org/uploads/File/projects/12_UPP_impact.pdf?PHPSESSID=843e721de6b3e088044 4023a29242f1b. Primanita, Arientha. "Jakarta Enlists Help to Get Kids Off the Streets". 2010. Jakarta Globe. 06 October 2011. Available at: http://www.thejakartaglobe.com/home/city-enlists-help-to-get-kids-off-the-streets/353492. Resosudarmo, Budy P., and Daniel Suryadarma. The Effect of Childhood Migration on Human Capital Accumulation: Evidence from Rural-Urban Migrants in Indonesia: Australian National University, 2010. Resosudarmo, Budy P., et al. The Socioeconomic and Health Status of Rural–Urban Migrants in Indonesia. Jakarta: The SMERU Research Institute, October 2009. Resosudarmoa, Budy P., Chikako Yamauchia, and Tadjuddin Effendib. "Rural–Urban Migration in Indonesia: Survey Design and Implementation." Research School of Economics, Australian National University, December 2009. Sahn, D., & S. Younger (2000). “Expenditure Incidence in Africa: Microeconomic Evidence,”.Fiscal Studies 21 (3): 329–347. Sarosa, W. (2006) “Indonesia”, pp. 155 – 187; in: Roberts, B. and Kanaley, T., eds., Urbanization and Sustainability in Asia, Case Studies of Good Practice, Manila: ADB, 486 pp. Semba, R.D Saskia dePeeb, Klaus Kraemer, and others. (2009). Purchase of drinking water is associated with increased child morbidity and mortality among urban slum-dwelling families in Indonesia. International journal of hygiene and environmental health ; vol. 212, no. 4 : p. 387-97. Setiawan, B. (1998). Local dynamics in informal settlement development : a case study of Yogyakarta, Indonesia. School of Community and Regional Planning. Vancouver, B.C., University of British Columbia. Doctor of Philosophy. Social Alert International. “Organising Informal Economy Workers in Indonesia” (2005). 19 September 2011. Available at: http://www.socialalert.org/pdf/IndonesieBROCH%5B1%5D.pdf. 137 State of the World's Cities 2010/2011 - Cities for All: Bridging the Urban Divide. (2010). London: UN- HABITAT. Steinberg, F. (2007). "Jakarta: Environmental problems and sustainability." Habitat International 31(3-4): 354-365. Strauss, J., F. Witoelar, B. Sikoki, and A. M. Wattie. (2009). The Fourth Wave of the Indonesian Family Life Survey (IFLS4): Overview and Field Report. Working Paper No. WR-675/1. Santa Monica, CA: RAND Corporation Strauss, J., K. Beegle, B. Sikoki, A. Dwiyanto, Y. Herawati, and F. Witoelar. (2004). The Third Wave of the Indonesia Family Life Survey (IFLS): Overview and Field Report. Working Paper No. WR-144/1. Santa Monica, CA: RAND Corporation. Suharto, Edi. "Accomodating the Urban Informal Sector in The Public Policy Process: A Case Study of Street Enterprises in Bandung Metropolitan Region (Bmr), Indonesia". 2003. 07 October 2011. Available at: http://www.policy.hu/suharto/Index.htm. Sumarto, S., A. Suryahadi, and W. Widyanti, "Assessing the Impact of Indonesian Social Safety Net Programmes on Household Welfare and Poverty Dynamics", European Journal of Development Research, 17 (2005) 155-177. Sumarto, Sudarno, and Samuel Bazzi. Social Protection in Indonesia: Past Experiences and Lessons for the Future, Paper presented at the 2011 Annual Bank Conference on Development Opportunities (ABCDE) jointly organized by the World Bank and OECD, 30 May-1 June 2011, Paris. Supriatna, A. (2011). The Feasibility study of Land Readjustment for Kampung Upgrading In Jakarta. Faculty of Geo-Information Science and Earth Observation. Netherlands, University of Twente. Suryadarma, D., Artha, R.P., Suryahadi, A., Sumarto, S. (2005). “A Reassessment of Inequality and Its Role in Poverty Reduction in Indonesia,” SMERU.. Suryahadi, A., A. Yumna, A. U. Reku Raya and D Marbun. (2011) ‘Accelerating poverty and vulnerability reduction: trends, opportunities and constraints’, in Employment, Living Standards and Poverty in Contemporary Indonesia, eds C. Manning and S. Sumarto, Institute of Southeast Asian Studies, Singapore: 68–89. Suryahadi, A., Reku Raya, U., Marbun, D. and Yumna, A. (2010) ‘Review of Government’s Poverty Reduction Strategies, Policies and Programs”, SMERU Research Report Wie, Thee Kian, and Siwage Dharma Negara. Indonesia Economics Update presentation at Indonesia Update Conference 2010, The Australian National University. Available at: http://www.crawford.anu.edu.au/acde/ip/pdf/update/2010/3_Economic_Update_- _Thee_Kian_Wie_&_Dharma_Negara.pdf. Wie, Thee Kian. Indonesia’s Economic Crises and the Prospects for Rapid and Sustained Growth, RIEBS Vol. 1 No. 1, November 2010, p. 43. Wilhelm, M. (2011). The Role of Community Resilience in Adaptation to Climate Change: The Urban Poor in Jakarta, Indonesia. In Resilient Cities (pp. 45-53). Springer Netherlands. 138 Yatmo, Yandi Andri. "Street Vendors as ‘out of Place’ Urban Elements." Journal of Urban Design 13.3 (October 2008): 387-402. Younger, S. (1999). “The Relative Progressivity of Social Services in Ecuador,” Public Finance Review 27 (3): 310–352. World Bank. (1995). Indonesia - Enhancing The Quality of Life in Urban Indonesia: The Legacy of Kampung Improvement Program. Washington, DC. Report No. 14747-IND World Bank, (2006). Making the New Indonesia Work for the Poor. Washington, D.C. Available at: http://web.worldbank.org/WBSITE/EXTERNAL/COUNTRIES/EASTASIAPACIFICEXT/INDONE SIAEXTN/0,,contentMDK:21152026~menuPK:224605~pagePK:2865066~piPK:2865079~theSitePK :226309,00.html. World Bank (2008). Global Purchasing Power Parities and Real Expenditures, 2005 International Comparison Program. Available at: http://siteresources.worldbank.org/ICPINT/Resources/icp- final.pdf. World Bank (2010a). Fostering Competitiveness to Increase Investment, Trade and Good Jobs. Available at: http://www- wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2010/03/30/000333038_2010033 0020915/Rendered/PDF/534780BRI0101i10Box345611B01PUBLIC1.pdf. World Bank (2010b). Targeting Effectiveness of Current Social Assistance Programs in Indonesia, Targeting in Indonesia Policy Paper Series. Poverty Group, World Bank Jakarta Office, Jakarta. World Bank (2010c). “PNPM Mandiri Revolving Loan Fund Capacity-Building and Sustainability Project Concept Note,” World Bank Jakarta Office, Jakarta. World Bank (2011a). "Jakarta: Urban Challenges in a Changing Climate." Retrieved 24 January, 2011, from http://www- wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2011/10/19/000356161_2011101 9004946/Rendered/PDF/650180WP0Box360ange0Jakarta0English.pdf. World Bank. (2011b). “Lessons in Urban Sanitation Development Indonesia Sanitation Sector Development Program 2006-2010”, Water and Sanitation Program. World Bank World Bank (2011c). “Protecting Poor and Vulnerable Households in Indonesia: Household Social Assistance” Public Expenditure and Program Poverty Group,” World Bank Jakarta Office, Jakarta. World Bank (2011d). “PNPM-Rural Impact Evaluation” Draft. World Bank Jakarta Office, Jakarta. 139 140