Table of Contents Foreword iv Abstract v Acknowledgments vi 1. The City Poverty AssessmentYet Another Academic 1 Report of My Overstudied City? 2. Why Care About the Poor? 2 3. City Poverty Indicators 3 Types of Indicators 3 Data Sources 8 City Poverty Information Strategy 12 4. The City Poverty ProfileA Basic Snapshot 13 5. More Than a SnapshotChanges Over Time 20 6. Beyond Indicators and the ProfileUnderstanding the 23 Meaning of Poverty Reduction and Policy Impacts 7. City Finances and the Poor 24 8. City Anti-Poverty Programs 28 9. City Growth and Poverty Reduction 32 10. Concluding Remarks 34 References 35 iii Foreword Cities and towns are increasingly becoming the primary locus of poverty in many countries. Rural-urban migration and low urban mortality rates have contributed to the rapid population growth of cities in many parts of the world. With such rapid growth also comes an increasing concentration of poverty in urban areas. In parallel, more and more countries assign local governments increased responsibility in fighting poverty. With decentralization, the responsibility of local social policy goes beyond the execution of centrally designed and funded education and health programs. In many countries, local policy makers today decide on tax rates, expenditure policies, development of new assistance programs, incentives for local economic development, land and zoning laws, etc. The formulation of pro-poor local policies requires good information and analyses. Local governments and their partners have both an opportunity and a need to understand the determinants of poverty and impediments for its reduction. This paper is an introduction to how such local information on poverty can be gathered and analyzed. The paper is not meant to be a comprehensive guide to information sources, poverty indicators, and methodologies. Rather, it aims to provide local policy makers with a broad overview of the type of questions typically asked and answered in City Poverty Assessments. As the nature of poverty differs widely between cities and countries, so will the content of such poverty assessments as they have to be adapted to local needs. Michael Walton Director, Poverty Reduction Poverty Reduction and Economic Management Network iv Abstract This paper provides an introduction to the concept of and tools used in City Poverty Assessments. There is no standard content to such assessments; rather, they need to be adapted to the specific needs of the city involved. Several aspects of urban poverty touched on in this paper will be irrelevant to certain circumstances, while others not mentioned here will be crucial. The thrust of City Poverty Assessments is to provide city policymakers with good and thorough information about the situation of the city's poor, the key determinants of poverty, the functioning of city anti-poverty programs, the distribution of city finances, and the link between poverty and city growth. Many of the tools used when developing City Poverty Assessments are valuable planning tools in and of themselves, such as poverty maps, institutional maps, tracking of the incidence of taxes and expenditures, and rapid service satisfaction surveys. Further, the very process of preparing a City Poverty Assessmentwhich includes collecting information, analyzing it, and discussing it with all relevant actors, including the poorwill be of major importance in forming new and more effective partnerships for city poverty reduction. v Acknowledgments This paper was written as course material for the Urban and City Management Program of the World Bank Institute. We would like to thank Mila Freire and Alexandra Ortiz, the organizers of the poverty module of the pilot course in Toronto (May 1999), for their feedback on the paper. Discussions with participants of the courses in Toronto and Singapore as well as comments from Tim Campbell, Janice Perlman and Michael Walton are gratefully acknowledged. vi The City Poverty Assessment: A Primer 1. The City Poverty A ssessmentYet A nother A cademic Report of My Overstudied City? Over the last few decades, city managers and activists in many countries have faced an increasingly urgent need to respond to the plight of the urban poor as cities, rather than rural areas, become the loci of poverty. Migration, low urban mortality rates, and high overall fertility rates have all played a part in shifting poverty from rural to urban areas since the beginning of the 1970s. Today, poverty in both Latin America and Eastern Europe is primarily urban; a similar transition will occur in East Asia in the near future. Concurrently, cities around the world have grown tremendously. Rapid urbanization places enormous pressure on cities to use their limited resources to meet or facilitate the increased demand for water, sanitation, electricity, basic education and health, housing, and transport. Urban poverty has placed further demands on cities to address health hazards due to air pollution and contaminated water, crowding, traffic congestion, poverty-induced violence, and inequality. City managersas well as the poor themselvescan pinpoint the most pressing problems in their municipalities and know where help is most neededbe it in the supply of basic services, nutrition, social assistance programs, or employment creation. Moreover, many cities have been subjected to a host of studies, which examine in detail deprivation levels or the deficit of service provision in certain locations. Given this level of knowledge and available information, is another study needed that revolves Box 1. City Poverty Assessments Are around where the poor live and what they need? That may well be, but the City Poverty Assessment ! Resources for up-to-date (CPA) is not simply another study. Rather, it is a information on current poverty and social development in the city; tool for urban planning, providing crucial and up- to-date information on what city managers (as well ! Management tools for city planning; as many actors in the private and nonprofit sectors) and need to know when developing city policies, ! Monitoring and evaluation devices programs, and projects against poverty. to assess the effectiveness of city CPAs allow decisionmakers to look at anti-poverty programs and projects. poverty from a much broader perspective than is possible by simply asking, "Where are the poor, and what do they need?" City Poverty Assessments can provide feedback to city managers on topics as diverse as city finances, city employment and growth, the effectiveness of social programs, and infrastructure priorities, among others. The following questions are integral parts of a comprehensive CPA. ! Who benefits from city social expenditures? What is the share of total expenditures reaching the poor? Do subsidies reach the intended target? To what degree do the poor benefit from public services on a day-to-day basis? 1 ! Who pays local taxes? Is property tax evasion on the part of the rich so extensive that the poor actually pay a large proportion of total property taxes? ! Are anti-poverty projects in the city successful? Are the people that have benefited from such programs better off than others? ! What type of city growth is beneficial for the poor? Which sectors of the city economy do the poor rely on most, and which ones would help them overcome poverty? ! Do city regulationssuch as property titling, demarcation, land regulation, or labor market legislationhelp or hurt the poor? There is no standard format or content of a CPA; rather, CPAs include all the information relevant within a specific context. For example, some cities might find that they have to find ways and means to reduce violence, especially in the marginalized areas. Other cities might need to look at the distributional impact of specific projects, such as an export processing zone that would create jobs for the city. This paper aims to provide city managers, policymakers, and analysts with a general introduction to CPAs. It includes examples and descriptions of the different components of the City Poverty Assessment, such as poverty indicators and data sources, the poverty profile, municipal finance and incidence analysis, evaluation of anti-poverty programs, and city growth. But before describing the details of the CPA, the next section briefly revisits why we need to think about the poor in the city context when planning policies and programs. 2. Why Care A bout the Poor? Obviously, a comprehensive review of the poverty situation in a city has its costs, in terms of time and financial resources. And that's why it is important to question the very necessity of such exercises. The CPA starts from the implicit judgement that poverty is undesirable. Most people will agree with this normative statement as they view `development' closely linked with guaranteeing minimum standards of living for everybodylike freedom from hunger, a healthy and long life, literacy etc. Further, reducing poverty is also important for urban development in other respects. First, reducing poverty through purposefully designed city policies will also tend to decrease city inequality and thereby social tensions within the city. Second, helping the poor reach their own potential, e.g. through education and in gainful employment, will help the city reach its growth and prosperity potential. City policies will impact on different population groups in different waysestablishing these links is the task of the City Poverty Assessment. The `poor', although far from being a homogenous group, will have needs and opportunities distinct from other groups in the city. For example, city investment in education will reach and benefit different groups depending on whether this investment takes place in primary schools or universities. Similarly, increasing local tariff rates for electricity or water can have a very different impact on households depending on whether they have access to such public utilities or not. In certain cities in Latin America, the poor might not feel the impact of such price rises as they have no such access. But if they have access, how hard would it be for the poor to pay such higher prices? Would they potentially forgo other very important expenditures, such as sending their children to school? It is therefore important to analyze 2 the impact of current city policies or Box 2. Common Income/Consumption Poverty proposed changes in city policies on Indicators different population groups separately. ! Poverty rate (headcount rate): percentage of the population not able to finance a basic basket of 3. City Poverty Indicators goods ! Extreme poverty rate: percentage of the population Types of Indicators not able to finance a food basket of goods with their total income Poverty has many faces, and different cities will find that they need to select ! Poverty gap: deficit of all poor persons relative to indicators appropriate to their individual the poverty line circumstances. In most cities, household ! Income distribution, e.g. the Gini-coefficient income or consumption of the most ! Quintile dispersion ratio: average per capita income marginalized will be a crucial determinant of of richest 20 percent of city population (quintile 5) poverty levels. This information can be divided by average income of poorest 20 percent of supplemented with data on access to a basic city population (quintile 1) set of services ranging from water, electricity, and sanitation to children's school attendance. Further, non-income dimensions of poverty include, for example, the incidence of crime and violence, or discrimination against specific population groups in the city. This subsection describes different types of indicators cities might find useful tracking poverty levels and understanding the correlates of poverty. Not all such indicators are strictly linked to poverty, but they all play a role in pro-poor city planning. Also, the combination of many of these indicators gives city policymakers important insight into the nature of deprivation. Income and Consumption Indicators The most commonly used poverty measures are those based on the per capita income or consumption of a household. Such monetary indicators aim to assess whether households can afford to buy a very basic basket of goods at a given point in time. There are many ways to define the value of this basic basket, but these methods all assume that the basket contains a minimum of goods essential for the household: food (often distinguished by their nutritional contributions), housing, water, clothing, transport, etc.1 The value of this basic basket of goods is then called the "poverty line," and a large literature exists on how best to define such a poverty line.2 Simply put, however, the poverty line helps distinguish the segment of the population that has an adequate income or consumption level from the segment that does not. Several standard indicators are derived by applying such poverty lines to data on income and consumption of households. The most common indicator is the poverty rate 1 In most cases, this basic basket of goods mirrors the actual consumption pattern of households. In some cases in Latin Americafor example, Peruit is also instead determined by a group of experts. See Ravallion (1994). 2 For explanations, see Deaton (1997) and Ravallion (1994). 3 (also termed poverty incidence or the headcount rate) which describes the percentage of the city's population whose per capita incomes (or expenditures) are below the poverty linei.e., the population that cannot afford to buy a basic basket of goods. Another commonly used indicator is the poverty gap. The gap measures the income shortfall of poor people relative to the poverty line. Or, put another way, it characterizes how many resources are needed to bring the poor to the poverty line. The gap is therefore a much more powerful measure than the pure headcount rate because it takes the distribution of the poor below the poverty line into account.3 For example, two cities might seem equally worse off if they have the same poverty rate. However, in one city almost all of the poor have a per capita income very close to the poverty line. In the second city, the poor are extremely poor with barely any income at all. The poverty gap measure will therefore show that poverty in the latter city is higher than in the former. A third frequently used indicator is the extreme poverty rate. This measure compares per capita household income against a basket of goods that generally only includes food items. The percentage of the city population not able to afford this very austere basket of goods is termed extremely poor. In addition to these absolute measures of poverty, city policymakers should also assess the distribution of income in the citya relative measure. The most commonly used measure here is the Gini-coefficient of inequality, a measure that varies between 0 (complete equality of incomes) to 1 (complete inequality, with one person having all the income and all the others Box 3. Health and Education Outcome none).4 To determine the gap between the rich and Indicators the poor, analysts also calculate a quintile dispersion ! Under-five malnutrition rate ratio. This ratio is the average income of the richest ! Infant mortality rate, under-five mortality 20 percent of the city's population divided by the rate average income of the poorest 20 percent.5 ! Maternal mortality rate Health and Education Outcome Indicators ! Life expectancy of the city population Health and education outcome indicators ! Incidence of specific diseases directly measure the degree of well-being the city's ! Literacy rate of the population population has attained that is, the degree to ! Years of schooling of different age groups which people can lead healthy and long lives, and the level of education they have reached. Several such outcome indicators concentrate on children, as the group in society for whom it is of great importance to be well-fed and healthy since their whole lives will depend on their first few years. A widely used indicator in this context is the percentage of all children below the age of five that are malnourished; i.e., children not having grown sufficiently for their age 3 Another poverty measure that is closely related to the poverty gap is the poverty severity index. Like the poverty gap, this is a measure of the depth of poverty; the index gives more weight to those extremely poor people far away from the poverty line. See, e.g., Ravallion (1994). 4 See discussion in Deaton (1997) and Ravallion (1994). 5 This inequality indicator can be calculated using other income brackets as well. For example, the decile dispersion indicator uses the ratio between the average income of the richest 10 percent of the population to that of the poorest 10 percent. 4 (chronic malnutrition), or children that do not weigh enough relative to their height (acute malnutrition).6 Other health outcome measures are the rate at which children die as infants or in their childhood, the maternal mortality rate, and the city's overall life expectancy. Education indicators include the percentage of citizens who can read and write, or also achievement indicators such as standardized exam scores for students finishing a certain school level. Access and Service Satisfaction Indicators Box 4. Access Indicators Proportion of city population without access to: Access indicators measure access to a set of basic infrastructure and social services. They can be ! Water and sanitation considered input indicators (as opposed to the ! Electricity outcome measures discussed above) since they do not necessarily reveal whether a desired impact was ! Garbage collection actually achieved. For example, even though all city ! Schooling residents might have access to a primary health care ! Health centers and hospitals clinic, this will not ensure low infant mortality rates. Similarly, although food distribution programs might ! Public or private transport etc. reach a large proportion of the children in a city, ! Social programs (e.g., nutrition, social malnutrition levels might still be high if, e.g., assistance, childcare) contaminated water causes widespread diarrhea for infants. Nevertheless, such access indicators are very important since they determine the degree to which city programs are available to different population groupswhich may not be a sufficient, but which is often a necessary, condition for improving the lives of the poor. Such access indicators are especially relied upon in Latin America, where they comprise the unsatisfied basic needs indicator (necesidades basicas insatisfechas) used by many national statistical institutes as well as the United Nations Commission for Latin America. With variations, the unsatisfied basic needs indicator includes access of the population to basic services such as water, electricity, and sanitation, and educational attainment. Access information is very important in understanding the actual distribution of program beneficiaries within a city. For CPA purposes, therefore, access to social programs covering nutritional aid and social assistance are also useful indicators. However, access says little about satisfaction and the quality of services. Many urban dwellers might have access to certain services but the quality of the service may be poor. Having a public water connection in the dwelling will not help much if water is not running. Power cuts in electricity, infrequent garbage collection, schoolrooms without teachers, or primary health care centers without medicine all will decrease the benefit the population derives from such services. That's why it can be very insightful to go beyond access and explore the satisfaction with services in the city context. Such satisfaction indicators can explore the costs, quality, quantity, staffing and maintenance of city services. 6 The World Health Organization maintains a Global Database on Child Growth and Malnutrition which contains comparative information, data sources, and analytical information on this topic. 5 Non-income Deprivation Indicators Box 5. Non-income Deprivation Several other indicators shed further light on Indicators poverty; some are closely associated with income ! Unemployment measures. One such indicator is unemployment, which may or may not cast people into poverty, depending ! Violence rate in the city (different on whether good unemployment insurance exists or forms of violence) if the unemployed can count on some other form of ! Child labor support. Even if not directly linked to income- poverty, unemployment is undesirable given its ! Discrimination in workplace or public life (access to city institutions like the impact on household income, self-esteem, or justice system, police treatment); personal health. Other such non-income deprivation exclusion indicators; vulnerability indicators could be crime rates (robbery, homicide, indicators domestic violence, etc.); child labor; and discrimination. If, for example, a certain ethnic or gender group is discriminated against in the workplace (in terms of receiving lower wages) or in public life (in terms of being denied access to city institutions such as the justice system or in terms of differentiated treatment by the police), this is also deprivation. As can be inferred, non-income deprivation indicators generally tend to be more difficult to capture than the income, outcome, or access indicators listed above. Combining the Indicators Graphs 1 and 2 show some of the above-mentioned indicators for a number of cities in Asia. Graph 1 plots household access to water and sewerage services (i.e., input indicators) and child mortality rates. Reading from left to right, infrastructure access tends to decline (at least for water), while child mortality outcomes show an irregular pattern. Although studies have shown that (private) water and sanitation access do improve health outcomessuch as the malnutrition rate and also, probably, the child mortality rate (see Huges and Dunleavy 2000)the simple comparison in graph 1 does not reveal this. Graph 2 plots a measure of income inequality (the quintile dispersion ratio, see box 2 for more information) and a crime indicator (murder rate per 100,000 citizens). Unlike graph 1, a clear pattern emerges, with lower inequality cities showing lower murder rates. Although the pattern is consistent with recent research (Fajnzylber et al. 1998), causality between the two indicators cannot be inferred. Many other factors might be responsible for a city's murder rate, with the possibility of this graph's data depicting a spurious or accidental correlation only. 6 Graph 1. Infrastructure and Health Indicators in Selected Asian Cities 90 Child mortality (below 5) per 1000 80 Household sewerage connection (%) Household water connection (%) 70 60 50 40 30 20 10 - Lahore, Colombo, New Delhi, Medan, Ulaanbaatar, Mysore, Chennai Lucknow, Pakistan Sri Lanka India Indonesia Mongolia India (Madras), India India Source: UNCHS (1999) Graph 2. Income Inequality and Crime Indicators in Selected Asian Cities 16 Murder rate per 100,000 14 Income disparity (Quintile 5/Quintile 1) 12 10 8 6 4 2 0 New Delhi, Surabaya, Lahore, Jakarta, Semarang, Medan, Banjarmasin, Colombo, Sri India Indonesia Pakistan Indonesia Indonesia Indonesia Indonesia Lanka Source: UNCHS (1999) 7 Data Sources The above-mentioned indicators of poverty and social development can be derived from a variety of data sources. Some are quite standardized and available in every country and city; others are less readily available. This subsection briefly describes these data sources; table 1, at the end of this subsection, correlates these sources to the various indicators. Population Census A population census is carried out for all households in a country to obtain basic information on the population, its demographic structure, and location. Since the population census is carried out across millions of households, the information it gathers is of necessity generally limited. Nevertheless, housing and basic service access, education levels, and employment by sector are typically included; this allows policymakers to gather important information at a very small, disaggregated level within the city, including descriptive statistics of the housing stock; population access to basic services such as water, electricity, and sanitation; and employment patterns in different city subsections. In most countries, the population census is carried out by a national statistics institute, which can then provide municipalities with data tailored to local information needs. Since the census covers the country's entire population, it is very costly; most countries consequently conduct a census only once every decade. A census can thus provide cities with important data for planning in the years directly following its implementation, but its utility diminishes thereafter. Household Surveys Household surveys are a very important resource for CPAs. Unlike the population census, surveys only interview a subsetgenerally quite a small fractionof all city households. This subset (sample) of households is carefully chosen so that the results of the survey accurately describe general living conditions in the city. Sampling should be based on mapping of actual settlements, including informal ones. The sample size (the number of households interviewed) will vary according to a number of factors: ! Indicator to be measured. For example, it is much more difficult to estimate the average income of households than to estimate the percentage of households with water connection; therefore, the sample size required to determine the former must be larger. ! Level at which the information is needed. A citywide average electricity connection rate will require that fewer households be interviewed than to determine averages in each of 20 subdistricts in the city. ! City population. The larger the population of the city (or of a geographical subunit within the city), the larger the sample size has to be. However, this relationship is all but linear with the required sample size increasing only very little with the city population. As a rule of thumb, cities that have carried out household surveys have had a minimum sample size of 1,500 to 2,000 households. Many different types of household surveys exist at the national level. Often, sample design of the national household surveys allows for quite accurate derivation of city statistics. Most countries use employment surveys to gather information on employment and unemployment patterns and fluctuations. These employment surveys also include questions 8 about household income, housing features, and demographic information about the household (size, age of members, etc.); they are therefore useful information sources on income-based poverty indicators and access indicators. National level Demographic and Health Surveys (DHS) are special household surveys geared to explore the incidence of diseases and use of health facilities. They also often collect anthropometric data (height, weight, and age of children which can be used to calculate malnutrition rates) and basic data about housing conditions and educational attainments. A third important national level survey is the Living Standard Measurement Study (LSMS) survey. It is specifically geared toward the measurement and analysis of poverty. This instrument, piloted in Peru and Cote d'Ivoire in 1985, collects information on household expenditures and income, health, education, employment, agriculture, the ownership of assets such as housing or land, access to services and social programs, etc. Besides these national level surveys are city level surveys, notably service satisfaction and needs surveys. These surveys go beyond the access assessments of typical household surveys and ask city residents in-depth questions about the quality of the services they receive and their needs. Such questions can include: whether households receive the service continuously or with interruptions, whether the service is provided on time and in good quality, whether households think that the service improved over the last year, and which type of service the city should expand or reduce. These questions aim at assessing the needs of the population. Often, such needs assessments can be very important planning tools for city policymakers, especially if they demonstrate conflicting priorities for different parts of the population. For example, several small Colombian municipalities conducted such inquiries in 1995 with support from the World Bank.7 Similarly, several cities in India, among them Bangalore (Paul 1998), have instituted a report card on which citizens can assess the quality of services and rate the implementing municipal agencies.8 Household surveys can also capture public perception data. For example, Jacobi (1994) employed a survey in Sao Paulo to understand environmental problems at the household and neighborhood levels along with respondents' perceptions regarding the nature and cause of these problems and the best means for their resolution. Mensah and Whitney (1991) captured perceptions associated with public and domestic refuse and its disposal and the relationships of these perceptions to gender, educational level, and ethnicity in Techiman, Ghana. Similarly, Egunjobi (1989) queried households on perceived environmental problems in Ibadan, Nigeria. Multi-topic city surveys attempt to combine many of the features and advantages of the different types of surveys mentioned above.9 Tailor-made to city needs, they collect both 7 See Fiszbein (1997). Closely linked to such service satisfaction surveys are beneficiary assessments that collect information from the participants and beneficiaries in specific programs. For an introduction see Salmen (1995). 8 See also the World Bank's Core Welfare Indicator Questionnaire (World Bank 1997). Such questionnaires are monitoring surveys normally fielded at the national level, but they are designed to be representative at the city level. 9 Many municipalities and researchers have also used single-topic city surveys to answer specific question of importance. For example, Alam et al. (1998) looked at energy user patterns in Hyderabad, India; Ruan (1993) conducted a social networking survey in Tianjin, China; Gupta and Baghel (1999) quantified the levels and differentials of infant mortality in Calcutta and Raipur City, India; and INEI (1997) reported on a Lima, Peru, crime survey. 9 quantitative and qualitative data. Quantitative data primarily involve household welfare (such as incomes and assets), but can also cover service access, education levels, health service utilization, public transport use, etc. The qualitative data are drawn from citizen assessments of the quality of city services and programs. One such multi-topic survey was recently conducted in Cali, Colombia (see box 6, "A City Survey for Cali"). Box 6. A City Survey for Cali With financial support from the Bank-Netherlands Partnership program, the Encuesta De Acceso y Percepcion de los Servicios Ofrecidos por el Municipio de Cali (EPSOC) (Survey of Access and Perception of Municipal Services) was conducted in Cali in September 1999. The survey built on a pilot survey in Kampala, Uganda, and was adapted to the situation in Cali with the help of the local public university. An experienced Colombian survey firm, the Centro Nacional de Consultoria, carried out data collection and tabulation. The survey covered 1,912 households in Cali, representative at several geographic areas within the city. The "value added" of the survey (as compared to existing data sets) is that it was tailored to the city, and thus included very specific questions about Cali. Also, it combined quantitative information (such as household income) with qualitative information on the population's priorities and satisfaction with existing programs. Specifically, the survey featured modules on housing, access to and satisfaction with basic services, access to and satisfaction with education and health, the labor market, food security, participation in city affairs, and the population's preferences and priorities. Data collection and processing were handled quickly within a period of three months since the results were needed by the municipal government as input to a City Development Strategy. The survey's total costs were less than $US40,000, a relatively modest price, which should allow the municipality to field such surveys regularly in the future. Source: Hentschel (forthcoming). Participatory Assessments Participatory poverty assessments (PPAs) are tools for consulting the poor directly and systematically. PPAs can capture, through qualitative and other flexible research techniques, dimensions of poverty that are not always addressed in household surveys. They also involve the city population to a higher degree than do household surveys through different methods: in town-hall meetings, certain groups or representatives can discuss city poverty problems and policies; communities can rank what they consider to be the causes of poverty; individual interviews can investigate the problems of women and children in households; or citizens can map out new streets or infrastructure in actual planning exercises. Participatory assessments can help policymakers determine the type of indicator important for the poor, be it of the housing, employment, or income dimension. These assessments can obtain a certain type of information other sources normally cannot capture, for example, the incidence and effect of domestic violence. Administrative Data Administrative data can show how a city uses its resources for (or against) the poor, and are therefore essential to a thorough CPA. Such data include information on establishments such as schools or hospitals, costs and expenditures by function, tax income by source, and staffing statistics. 10 Table 1 summarizes the different data sources that can help in assessing and monitoring city poverty indicators. As can be seen from the table, several indicators can be assessed with a variety of sources, while others depend on unique ones. Table 1. Indicators and Data Sources Indicator Data Source Income poverty indicators - Poverty rate (incidence), poverty gap, - National level household surveys (Living poverty severity; extreme poverty rate Standard Measurement Surveys, (incidence); income inequality measures Employment Surveys) if representative at city level; Multi-topic city surveys Health and Education Outcome Indicators - Under-five mortality rate, infant - Specialized national household surveys such mortality rate, maternal mortality rate, as DHS, or LSMS (if representative at city life expectancy level) - Malnutrition rate of children - National levels (DHS, LSMS), Nutrition surveys, Height census - Literacy rate, years of schooling - Most surveys and censuses Access and Service Satisfaction Indicators - Water, electricity, sanitation, garbage - Various household surveys, population collection census, administrative data - School and health facility - Various household surveys, some population census, administrative data - Social programs (nutrition, social - Specialized household surveys (LSMS, Multi- assistance) topic surveys, administrative data), participatory assessments - Service satisfaction - Specialized city surveys (service satisfaction and needs survey, Multi-topic city survey) Non-Income Deprivation Indicators - Unemployment - National employment surveys, LSMS, Multi- Topic City Surveys - Violence - Violence surveys (only certain types of - Child labor violence can be measured), Multi-topic city - discrimination surveys participatory appraisals - National surveys, Living Standard Measurement Surveys, Multi-topic City Survey - Participatory assessments, household surveys (not directly but through application of models) 11 City Poverty Information Strategy Not all cities are going to need or want to monitor all of the indicators described in subsection 3.1, nor are they going to need or have the resources to use the various data- gathering methods described in subsection 3.2. What all cities do need is some kind of information strategy that enables them to assess and monitor changes in the poverty indicators they have selected. This strategy will allow them to collect and analyze the data they require in a systematic and comprehensive manner; from this, they can then develop tailored and appropriate poverty-reduction policies and programs. Box 7 lists the steps involved in preparation of a city poverty information strategy. Box 7. Preparation of City Poverty Information Strategy Involves " Selection of most important poverty indicators and poverty information for the city; " Planning of intervals in which indicators need to be monitored; " Review of all available data sources and their data collection agencies; " Assessment of city's capacity to monitor and gather information on its own (or degree to which it should be developed); and " Selection of partnerships (e.g., national statistical institute, NGOs) that may conduct data collection and/or analysis for city. Many options exist for city planners and policy makers to gather important data on poverty and social development. Some of these are: ! Use existing household surveys and other data sources. Many countries field large employment or living standard measurement study surveys that are frequently representative at the city level. Many other data sources may exist thatif pieced togethercould give a good account of the city's poverty situation. The population census is one such data source; there are also numerous in-depth case studies and beneficiary information that non-governmental organizations (NGOs), public programs, and other organizations might possess. ! Link a city survey to a national household survey. Even if data from a national survey can be used to derive city level indicators, more detailed information (such as by geographical breakdown within the city) is often desirable. Rather than conduct a completely new survey, the city could negotiate with the national statistical institute to apply a city-specific module when it implements the national household survey. This will save considerable cost and ensure that the collected data can be linked to some of the other variables routinely collected by the statistical institute (such as income or expenditure data). ! Conduct multi-topic city surveys, service satisfaction surveys, and participatory assessments. The costs of customized multi-topic city surveys, service satisfaction surveys, and participatory assessments can be moderate if these are designed with a clear purpose and implementation time frame Multi-topic city surveys are probably the richest analysis tools informing the City Poverty Assessment. Service satisfaction and needs surveys generally do not contain income or expenditure information, which reduces costs of fielding the survey significantly, but limits the analysis of welfare and its correlates. 12 Participatory assessments collect opinions quickly and at low cost through an effective use of group methodology. ! Partner with other organizations to gather information. A large number of organizations are active at the city level in poverty reduction programs. Several of these may be able to collect data for policymakers through ongoing operations. For example, organizations working on housing issues may have detailed information on the quality of the housing stock in specific areas of the city. 4. The City Poverty ProfileA Basic Snapshot The most important first step in the analysis of poverty is to construct a poverty profile. Such a profile has several aspects, including who is poor, where the poor are in the city, how they earn their living, their access to and use of government services, and their living standards with regard to health, education, nutrition, and so on. Dimension and Geographic Location in the City Poverty dimensions in the city can be assessed using some of the indicators discussed in the previous section. And, since poverty has many faces, it is crucial to get to these different dimensions by describing welfare in the city with, for example, income poverty indicators but also the incidence of diseases, crime, or malnutrition. The poverty profile can help in comparing one city to another within a country. The CPA for Cali, Colombia, for example, started by looking at extreme income poverty, income poverty, inequality, and service provision levels in Cali as compared to other Colombian towns. Table 2 shows that, from this perspective, Cali does not stand out from other large cities in Colombia. Table 2. Income Poverty and Inequality of Colombian Cities, 1998 Poverty Rate Extreme Rate Inequality (Gini) Access to Sewerage (percent) Bogota 35.7 5.3 0.537 98.6 Medellin 36.5 5.9 0.514 99.6 Cali 36.6 6.4 0.542 99.5 Barranquilla 49.5 17.2 0.567 79.9 Bucaramanga 30.4 4.7 0.487 99.7 TOTAL 32.2 6.6 0.553 96.3 Source: Santamaria (1999) Like countries, cities are comprised of very different areassome affluent and some poor. The CPA can help identify areas in which a high number of poor and extremely poor people are concentrated. As an example, table 3 shows the distribution of income poverty in Karachi, Pakistan. As can be seen, the poverty rate varies greatly between the various parts of the city, and is highest in the Rural Fringe and lowest in the geographical areas characterized as affluent by the authors. Note that this table also shows another important dimension of poverty, geographical concentration: how many of the total poor live in a certain location. (This is shown in column 2 which could also be labeled "contribution to the total poor.") As observed, the Rural Fringe, the area with the highest incidence of poverty, is not the area with the highest share of the total poor, since relatively few people live there. 13 Table 3. Poverty Clusters in Karachi, Pakistan, 1990 Clusters Poverty Rate Poor Households in Cluster / Population Share Total Poor Households in City Old City 61 18 12 Old Settlements 56 10 7 Korangi 49 12 10 Site 59 27 18 Service Areas 54 11 8 Rural Fringe 67 7 4 Total Affluent 15 16 41 Source: Altaf et al (1993) Information on the spatial distribution of poverty within the city can be used to construct city poverty maps. Such maps can be of considerable value to governments, NGOs, and multilateral institutions interested in strengthening the poverty alleviation impact of their spending. For example, they can be used to guide the division of resources among local administrative units within the city as a first step in reaching the poor. Many countries, especially in Latin America, have constructed such poverty maps, most often using an unsatisfied basic needs indicator as the underlying welfare measure. Recently, some countries have started to construct spatially disaggregated income maps as well; these can then be combined with service deficiency information.10 Multi-topic city surveys can be used to produce such maps if they are designed to achieve sufficient precision at a local level.11 For example, the EPSOC survey in Cali (see "A City Survey for Cali" in box 6) lent itself to the construction of such a map (see graph 3). As can be seen, poverty rates differ significantly across Cali being highest in the western part of the city (Aguablanca) and at the eastern side, a steeply sloped populated area (called Ladera). 10 For information on poverty maps and how they are constructed, see the Web sites for the United Nations Environment Program at Arendal , the World Resources Institute , and the World Bank on poverty . See Hentschel et al. (2000) for a description of a method to link survey and census data to obtain spatially disaggregated estimates of income poverty. 11 Sample surveys are designed so that they achieve a specific precision of a target variable, such as income, at a certain geographical level or by socioeconomic group (see, e.g., Grosh and Munoz 1996). Since samples do not cover the whole population, all statistics derived from them should be accompanied by computations of the standard errors associated with them (see, e.g., Ravallion 1994). 14 Graph 3. Poverty in Cali, 1999 1 Km Poverty Headcount Rates Low: <20% Middle: 21% - 40% High: 41% - 50% Very High: 51% - 70% Extremely High: > 70% Source: Hentschel (Forthcoming) Characteristics of Poor Households The poverty profile should also describe the living circumstances of the poor in the city. This description can show how characteristics of poor households can vary within the cityfor example, the income poor in one area might not have access to basic services, while those in other parts of the city might. Table 4 presents an example from Karachi. Although household size, number of children per household, housing structure, and employment characteristics do not vary much between the extremely poor groups in the Old City and the Old Settlements, there is variation with regard to education. This information could help city planners by indicating an educational deficiency in the Old City. Table 4. Characteristics of Very Poor Households across Poverty Clusters in Karachi, Pakistan Old City Old Settlement Household size 10.2 9.7 No. of children 4.5 4.9 Housing structure - permanent 33% 23% - semi-permanent 62% 74% - impermanent 5% 3% Access to services - piped water 45% 63% 15 Old City Old Settlement - electricity 91% 88% - gas 52% 49% Employment sector - industry 16% 25% - services 73% 69% Education - adult males literate 47% 56% - adult females literate 22% 40% - adult males high school 14% 27% - adult females high school 6% 16% Source: Altaf et al (1993) Characteristics of Poor Households Compared to Non-Poor Households Characteristics of living circumstances can also be used to compare poor and non- poor groups. Such a comparison will show where the living characteristics of the poor are similar and where they differ from better-off groupsfor example, the fact that a large proportion of the poor do not have access to water might not be a distinguishing factor if the whole city population lacks such access as well. Table 5, derived from the EPSOC survey in Cali, distinguishes five different income quintiles in the city's population: 1 represents the poorest 20 percent of the population and 5 the richest 20 percent. Such a distinction goes beyond the "poor/non-poor" divide imposed by a poverty line and is able to provide a much more complete picture of living standards. As shown in table 5, several household characteristics do not vary at all by income class, while others do. The incidence of unemployment, incidence of hunger, or access to social programs (such as health insurance) are strongly correlated with household per capita income. Similarly, secondary school attendance increases with income quintile. Primary school attendance, however, and access to services such as electricity, sewerage, and water vary little by income quintile. Note, however, that even if such access to services is universal in Cali, differences remain. These pertain to where hygiene facilities are located (in the house or outside) and whether public water is used by a household alone or shared with neighbors. Often, the population in Cali's slums shares a single public standpipe across dozens of households. The poor in these areas are thus very vulnerable to water supply stoppages of these few standpipes. Table 5. Characteristics of Income Poverty, Cali, 1999 Income Quintile Total 1 2 3 4 5 Labor market Unemployment rate 35.9 22.4 18.4 11.8 5.8 17.1 Education Years of household head1 6.4 6.6 7.3 8.4 10.3 8.0 Food security Family member with hunger2 34.2 22.8 16.9 11.7 5.1 18.1 Access to nutrition programs 4.4 3.7 3.1 1.9 0.5 2.7 Housing 16 Income Quintile Total 1 2 3 4 5 Rented 37.9 41.0 38.2 36.0 35.5 37.7 Titled 77.5 79.2 89.2 92.8 93.0 86.9 Access to basic services Electricity connection 99.5 100 100 100 100 99.9 Hygiene facility 93.9 98.8 98.6 99.6 100 98.2 in house 73.3 81.1 84.3 89.6 95.1 84.7 Public water 99.7 99.2 99.8 99.8 100 99.7 single use 69.7 77.9 81.7 86.0 94.9 82.1 shared use 30.3 22.0 18.3 13.8 5.1 17.9 School attendance 6-11 years 91.1 93.8 95.3 97.9 100 94.8 private 32.6 31.9 39.7 62.3 72.2 100.0 public 67.7 69.0 58.5 36.3 25.0 55.7 male 86.9 95.1 92.8 100.0 100.0 93.9 female 95.4 93.6 97.0 96.7 100.0 96.1 12-18 years 58.4 68.9 69.9 73.1 85.0 70.1 private 43.1 50.1 54.8 56.6 74.9 100.0 public 56.4 50.9 44.1 42.0 22.3 43.2 Male 60.5 72.8 67.1 69.1 87.9 71.0 Female 57.0 65.6 72.7 76.0 82.1 69.3 Reason for not attending Costs 59.5 57.1 43.7 36.7 13.4 48.8 work 2.1 7.2 9.9 18.8 17.2 8.6 Health Sick in last 4 months 28.9 24.3 26.6 21.3 19.3 24.1 When used medical facility Public health post 37.6 29.1 27.7 13.6 7.5 22.7 Public hospital 21.1 16.1 16.9 8.8 8.9 14.2 Health insurance Affiliated 54.0 58.8 60.8 73.1 82.4 65.8 Violence Family member been victim of 20.2 23.8 19.7 20.6 25.4 21.9 assault, robbery, or violent acts 1 Mean education years for the whole population 18 years and older. 2 In the past year. Source: Hentschel (forthcoming). Graphs can show the interconnections among different city poverty indicators. Graph 4, for example, shows how access to basic services varies with income in Rio de Janeiro. Each chart in the graph distinguishes three different areasthe municipality of Rio, the metropolitan area, and all urban areas in Brazilthus allowing for comparison of Rio's basic service access to that in larger areas of the country. On the horizontal axis, 10 income deciles are portrayed with 1 being the poorest and 10 the richest. In general, poorer residents have much less access to any of these services than the richa stark contrast to the Cali example above. Graph 4 also shows that the poor in metropolitan Rio (deciles 1 and 2) are 17 comparatively better off than the poor in other urban areas because their access to services is higher. Graph 4. Access to Services, Rio de Janeiro, 1997 No Access to Bathroom No Solid Waste Collection 10 30 (%) 25 (%) 8 20 6 Share Unserved 15 4 10 2 Share 5 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Decile Decile No Water Supply Inadequate Sanitation 14 30 (%) (%) 12 25 10 20 8 Unserved 15 6 Unserved 4 10 Share 2 Share 5 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Decile Decile RIO Municipal RIO Met. Brasil Met. Source: World Bank (1998) Satisfaction Levels with City Programs If relevant data are available, part of the poverty profile can address the satisfaction levels of the city population (and the various subgroups) with programs and services offered by the city. Such information can be collected through separate inquiries (e.g., Paul 1998) or be part of a multi-topic city survey. The EPSOC survey in Cali included a module in which citizens rated municipal services. Results are presented in table 6. Table 6. Dissatisfaction with Basic Social and Infrastructure Services, Cali, 1999 Income Quintile Average 1 2 3 4 5 Dissatisfaction with Education (students) 9.1 9.2 7.0 9.5 8.2 8.6 Electricity service 8.2 11.6 9.3 5.4 6.2 8.1 Water service 8.7 8.6 7.2 12.3 7.2 8.8 Garbage collection 9.2 7.7 10.1 12.8 11.2 10.2 Health (those using 24.7 16.2 17.9 16.0 17.5 18.4 them) Sewerage 33.8 23.1 21.9 26.0 20.1 25.0 Environmental 61.2 66.8 60.5 64.3 60.9 62.7 cleanliness Source: Hentschel (2000) It is interesting to observe that satisfaction levels as depicted in table 6 vary significantly across services but very little across income groups. Education, electricity, water, and garbage collection services are viewed as satisfactory by a large majority of the 18 population. Dissatisfaction with health servicesespecially of public health servicesis considerably higher, with about one-quarter of the population discontented with the service offered. The city's sewerage service obtained similarly low ratings, and two-thirds of the population view the environmental situation as especially negative. Differences between income quintiles are low. This is important information for the city, since it means that policy changese.g., giving more emphasis to environmental cleanlinesswould be welcomed by a large majority of citizens, regardless of their economic status. Important Dimensions of City Life The two angles portrayed abovecharacteristics of the life of the poor in different parts of the city and characteristics across different income classescan be used to examine a broad variety of subjects that can be customized to individual city circumstances. Some of these could include: ! Environmental and health conditions of different population groups. Many assessments of inner-city health conditions suggest that death and disease rates for infants and children are between 2 and 10 times higher in deprived areas of cities as compared to nondeprived. This finding is closely connected to the fact that the city's poor are much more often affected by the results of environmental pollution than are the better off (see box 8, "Urban Living Conditions"). Qualifying and understanding these often hazardous living conditions is key to understanding the problems of the poor. Box 8. Urban Living Conditions: Selected Cases Bombay, India: People living in slums and the homeless are often the worst victims of industrial pollution in the cities. They are the worst affected by the insufficiency and poor quality of water, by inadequacies of drainage, sanitation, and household waste removal facilities and, in general, by unhealthy living and working environments. Slums are located typically in areas that are not meant for human habitation, for instance, in low-lying areas, on hillsides, on marshy land, near garbage dumps and under high-tension wires. The area is flooded in in the high tide and, when the tide recedes, it leaves behind all kinds of toxic waste, including carcassses of cattle and pigs in the swamp that surrounds the new tenements. Use of slum shelters as workplaces adds to health risks. In the Dharavi slum of central Bombay, there are about 400 leather-processing units which are a major source of air and water pollution. Water is a primary medium for the transmission of diseases, the most imporant of which are typhoid, cholera, hepatitis, polio-myelitis, dysentery, amoebiasis, and infection by intestinal protozoa. Slums lack systems for disposing of excreta, sewage, sullage (water from washing and bathing) and solid wastes. Source: Swaminathan (1995) Rio de Janeiro: By choice of location and by political process directing efforts to clean the environment toward richer areas, the poor are more affected by adverse environmental conditions. More of the poor live in the northern part of the municipality of Rio, which is affected by serious, and health threatening, air pollution. They live closer to heavily polluted water bodies, such as Guanabara Bay, which leads to health risks, for example, for bathing children. Many poor neighborhoods, especially favelas, are located on lands exposed to natural hazards (landslides, flooding, etc.). The health costs of pollution particularly affect low-income households which typically live in more polluted areas and lack the resources for protective expenditures and investments. 19 Environmental improvements will, therefore, often more than proportionately benefit the low- income population. Measures that improve environmental conditions and generate benefits primarily for the poor, such as the extension of basic sanitation, are an obvious top priority. On the other hand, it would be misguided to try to address income inequalities through environmental improvements that would not otherwise be a priority. For example, investments in secondary and tertiary sewage treatment would in many locations not likely be a priority for the poor who might instead prefer faster expansion of sewage collection or better health care and education. Source: World Bank (1998) ! Income and expenditure patterns of different population groups. Various population groups derive their income from various sources; they also spend it on different items. Exploring these differences is vital in assessing how changes in the price of goods will affect the well-being of the most marginalized groups. For example, if it is found that the city's poor spend almost 10 percent of their income on public transport while the richer groups do not depend on public transport at all, it can be seen that a strong price increase in public transport will cause substantial problems for the poor. The reverse, however, might be true as well: inhabitants of the richer parts of the city might be the exclusive users of public transport services, while the poor use informal transportation or simply walk. In this circumstance, any price increase will be borne by the well-off in the city and will not much affect the poor. Similar analyses can be carried out for other public services as well as for food and clothing price increases. ! Relative poverty risks of different groups within the city. Different population groups within a city might be at different levels of risk for being poor. A poverty profile should analyze the degree to which poverty and deprivation are linked to certain personal characteristics of citizens (e.g., gender, age, or ethnicity). To make comparisons easier, analysts sometimes use a relative poverty risk which sets the likelihood of someone being poor in relation to all other groups in society that do not have this characteristic (alternatively, the average poverty rate can be used; see box 9, "Poverty Risk in Rio de Janeiro, 1998"). In Peru, a recent poverty analysis showed that a person with an indigenous background was 50 percent more likely to be poor than a person with a non- indigenous background (World Bank 1999). Box 9. Poverty Risk in Rio de Janeiro, 1998 A breakdown of the poor in Rio shows that certain characteristics of the household head are associated with a higher probability of being poor. In particular, the percentage of poor (compared to the overall poverty rate) is higher for particularly vulnerable groups, including female-headed households (29% higher poverty rate); young households, headed by under 25 year olds (105% higher); uneducated households, heads without formal schooling (85% higher); blacks (55% higher) and unemployed (230% higher) and informal sector workers (26% higher). Source: World Bank (1998) 5. More Than a SnapshotChanges Over Time For most city policymakers, changes in city living conditions over time are even more important than the snapshot of poverty described in the preceding section. This emphasis 20 has two dimensions. First, changes over time give feedback as to whether a city is moving in the right direction. While poverty might be extremely low, an increase in a city's poverty has to concern policymakers. Conversely, even if the city's poverty levels are high, a reduction in poverty at least shows that improvements are possible and are in fact being made. Second, changes over time provide insight on the factors that help people grow out of poverty or fall into it. Such factors can be good "hooks" for the development of city anti-poverty programs. Poverty Changes over Time Measuring the quantitative changes of poverty over time requires, first of all, a rock- solid definition of the poverty indicators employed. This might seem easier than it actually is. While the measurement of access variables is generally quite straightforward, the most commonly used poverty indicatoran income-based poverty rate or poverty gapis much more difficult to define in precisely the same way in different years. Household surveys that measure the monetary welfare measure (income or consumption) must follow the same sampling procedures in different years; they must ask exactly the same questions and record the same prices which are then used for adjusting nominal incomes. Such comparability is not always easily achieved. Table 7. Changes in Income Poverty and Inequality of Colombian Cities, 1994_98 Change in Change in Inequality Ext. Poverty Rate (Gini coefficient) Bogota -2.5 -0.02 Medellin +3.8 -0.06 Cali +0.8 +0.04 Barranquilla -7.8 -0.19 Bucaramanga +2.2 +0.01 Source: Santamaria (1999) Given standardized and consistent indicators, comparisons over time can offer many additional insights. Take the example used earlier for Cali (table 2). That table showed that poverty, extreme poverty, and inequality levels in the city were clearly not the worst in Colombia and were in line with those of Bogota and Medellin. Table 7 reports changes in these same indicators from 1994 to 1998. As can be seen, Colombian cities had very different experiences in these four years: only Bogota and Baranquilla reduced both extreme poverty and inequality, while both extreme poverty and inequality increased in Cali and Bucaramanga. This ability to track these developments in Colombian cities was made possible by DANE's (the Colombian statistical institute) periodic nationwide application of its Encuesta de Hogares, using the same income definition and ensuring representativeness of survey results at the city level. Qualitative assessments or subjective evaluations of changes in welfare do not require such a strict comparability of statistics and can be illuminating as well. A city survey in Haiphong, Vietnam, captured how the city population assessed its living standards to have changed over the past two years (Luan et al. 1999). The results, reproduced in table 8, show a diverse picture emerging, with over 30 percent of the households surveyed feeling that they were worse off; a slightly smaller proportion felt they were better off than two years ago. The impression that living standards had declined was particularly prominent among the poorest quintile. This could imply that mobility in the city in periods of rapid economic 21 change had been highboth upward as well as downward. A CPA could further analyze the specific characteristics of different groups of households, especially within given income quintiles. Such information can be very important in understanding the dynamics of welfare change. Table 8. Subjective Assessment of Changes in Living Standards in Haiphong, Vietnam Income Quintile Total 1 2 3 4 5 Much better 0.0 0.0 0.5 0.5 2.5 0.7 Better 11.4 20.2 25.5 31.3 44.1 26.5 Same as before 39.3 43.4 41.5 47.0 34.7 41.1 Worse 46.3 35.9 32.0 20.7 18.8 30.7 Much worse 3.0 0.5 0.5 0.5 0.0 0.9 Source: Luan et al (1999) Factors Associated with Poverty Comparisons over time can help policymakers better understand the dynamics of poverty. If data on factors related to poverty are available over time, analysts can determine whether these factors remain strongly associated with poverty or if they lose or increase their importance over time. For instance, the World Bank (1999) reports that in Peru an indigenous person was 40 percent more likely to be poor than a non-indigenous person in 1994 but almost 50 percent more likely to be poor in 1997. Thus, the indigenous population fell further behind the non-indigenous population over time. Landless rural households were, as one could expect, more likely to be poor in 1994 than rural households with land. However, in 1997, such rural landless households were about 5 percent less likely to be poorthus indicating that this factor does not appear to be systematically linked to welfare change. A special case is given if consecutive household surveys interview identical households time and a panel is included in the household survey. Then, more sophisticated methods can be used to link what factors help households grow out of poverty.12 Other Indicators If cities do not have representative household surveys that contain income information at different points in time, welfare developments can be tracked using other indicators. As one such indicator, Wong (1997) used the share of households that possess certain durable consumer goods (refrigerator, washer, color TV) to show welfare developments in Shanghai from 1985 to 1995 (table 9). 12 See Deaton (1997) for an explanation of household panels. Examples of panel analysis can be found in Glewwe and Hall (1995) and World Bank (1999). 22 Table 9. Possession of Durable Consumer Goods in Shanghai, 1985-95 (percentage of households in possession of good) 1985 1988 1990 1993 1995 Refrigerator 20 73 88 92 98 Washer 26 62 72 76 78 Color TV 22 54 77 94 100 Source: Wong (1997) 6. Beyond Indicators and the ProfileUnderstanding the Meaning of Poverty Reduction and Policy Impacts While the relatively objective measurement and description of poverty conditions are important components of a CPA, a variety of other tools can help (and are in fact often necessary) to understand the actual meaning and dynamics of poverty. For example, areas in many cities are known as poor, and the stigma of being poor accompanies their residents wherever they goif they use city facilities or public transport, if they try to send their children to school, or when they search for work. Residents of favelas in Rio experience such geographic discrimination: "the prevalent view among non-favela residents is that they represent a `break' in the cityscape and that this fact reflects the characteristics of the social groups living in them. When one refers to a favela, one implies that such a place is `irregular,' `poor,' `disorganized,' `dangerous,' i.e., full of problems" (Pamuk and Cavalieri 1998). Participatory and qualitative tools can be used to ! assess how poor communities understand and experience poverty, ! explore whether different groups in the community (or individuals in households) face a different set of problems in overcoming poverty than others. ! identify vulnerable groups in times of hardship, ! prioritize poverty indicators among the many that could be employed, ! learn what poor communities view as the main bottleneck in reducing poverty, and ! understand the living circumstances of the poor and the dynamics and causes of poverty. The CPA would ideally also include a detailed assessment of the current (and future) regulatory framework of the city, including an institutional assessment of the access of different population groups to the legal system, the transparency of budget and expenditure decisions, and the accountability of decisionmakers to the public. Policy assessment would include, for example, the impact of zoning regulations in terms of their impact on housing prices on different population groups in the city. Similarly, existing registration procedures for squatter settlements would be analyzed and linkages made to the ability of the illegal squatters to obtain important service provisions. In some cases, these policy reviews might lead to the adoption of different indicators for poverty monitoring than usually proposed, such as the percentage of households without a plot of land or proper title to their house. In other cases, such "contextualization" of poverty can call for a combination of different indicators (see box 10, "Income, Poverty, and Living Conditions in Bombay"). 23 Box 10. Income-Poverty and Living Conditions in Bombay A longitudinal case study in Bombay (i.e. a study that records and observes a community over a period of time), shows that income changes and changes in living conditions need not go hand in hand. Observing the same families in 1987 and 1992, found income variations to be extremely large in both directions, upward and downward. Such high income mobility went hand in hand with very little change in overall living conditions of households, like the health risks they faced from environmental pollution or access to basic services. Hence, the study concludes that poverty can be neither understood nor tackled through a simple focus on income. A more comprehensive approach, which includes housing and living conditions in addition to the income measure, is necessary. Source: Swaminathan (1995). 7. City Finances and the Poor A central part of the CPA is the analysis of city finances. Unlike municipal finance studies, which look at the appropriateness and level of taxes and expenditures, the CPA's emphasis is on fiscal resource distribution and equity. Responsibility of the City and Central Governments The first step in analyzing municipal finance from a poverty perspective is to distinguish which functions are performed by which levels of governmentthe city, the provincial government, and the central government. Programs might work very differently from each other in terms of the types of services they fund and how they try to reach beneficiaries. Similarly, taxes might be shared between different levels of government, or cities might have the freedom to levy certain taxes without much central control. All of this is important background information for understanding how the city can obtain or use its resources in a more pro-poor way. Incidence Analysis Incidence analysis is the main tool for assessing the distributional impact of city expenditures and taxes. It aims to quantify the share of total revenues and expenditures a certain population segment (e.g., the poorest decile or the population in district 1) pays or receives. There are two dimensions to incidence analysis in the city context: type of household and geographical location. Incidence analysis by type of household requires that household survey data be representative to the city level. Using the total expenditure or tax per activity, specific questions of the household survey can then be used to distribute total funds (e.g., to the poorest decile of the population). Incidence by geographical location requires good data from the city planning or budget office as to where actual expenditures in the city went and where taxes were raised. The geographic assessment is generally carried out at the level of administrative subunits. 24 Incidence by Household Type: City Expenditures Household incidence analysis generally begins by establishing user patterns of public services and programs, employing the household survey as the main data source.13 These user patterns provide information on who in the city obtains what share of services. The example in graph 5 shows total use of public health services (left-hand chart) by population deciles in Rio de Janeiro. The most frequent users of health services are deciles 3 and 4 whoalthough their share in the population is each only 10 percentuse 15 or 16 percent of the total health services provided in the city. Rio's more wealthy groups tend to use private rather than public health services, which explains their lower share in total service use. Graph 5. Public Health Facility Use in Rio de Janeiro, 1997 PublicHealth FacilityUse(%) Health FacilityUse(%) Rio de JaneiroDeciles Rio de JaneiroDeciles 1 8 Pr ivate Facili t y Public Clinic Public Hospital 15.2 1 6.2 100% 1 6 90% 13.3 1 4 12.4 80% 1 2 10.5 70% 9.5 9.5 1 0 60% 8 50% 40% 6 4.8 4.8 3.8 30% 4 20% 2 10% 0 0% 1 2 3 4 5 6 7 8 9 1 0 1 2 3 4 5 6 7 8 9 1 0 Deci l e Deci l e Source: World Bank (1998) The second step in calculating incidence is to distribute actual expenditures by user profile. Most often, household surveys will not record the actual benefit received from social programs; thus, assumptions about such benefits must be made. The most common assumption is that benefits for all users are, on average, the samethat is, the actual benefit derived from a health visit, for example, is independent of the user's income status. Applying such a rationale to the above picture of public health service use in Rio results in the incidence of expenditures shown in table 10 (for ease of presentation, incidence is reported by population quintiles). Given the user pattern established above, overall public health expenditures in Rio can be said to be progressively distributed: i.e., they benefit the poor more than the well-off households. If total expenditures in the health sector are known, such shares translate into actual monetary figures which can be aggregated across different social programs as long as the household survey can be used to establish user or benefit patterns. 13 See World Bank (1992) for a discussion of incidence analysis. 25 Table 10. Incidence of Public Health Expenditure in Rio, by Population Quintile Population quintile Share of expenditure received 1 22.9 2 31.4 3 19.0 4 18.1 5 8.6 Source: World Bank (1998) Incidence by Household Type: City Taxes The other side of expenditure incidence is an assessment of how such expenditures are paid for. While all citizens can be asked to contribute to local revenues, it is not desirable to have the less fortunate pay the brunt of total revenues. Information on tax payments is rather scarce in most cities, which complicates tax incidence analysis. In some cases, household surveys do contain important informationfor example, if a share of value-added tax (VAT) goes directly into a municipality's account, household surveys can provide good estimates of VAT distribution through information on consumption patterns. But other local taxes, especially the property tax, will not be captured accurately in household surveys (partly because respondents are skeptical as to what the information they provide is going to be used for). Hence, tax registries at the local level will be the major source of information in this regard and should provide an accurate geographical tax incidence by type of tax. Many inferences can be made from such a geographical distribution of taxes. For example, if income from property taxes is highest in the poorest areas of town, it is likely that the more wealthy evade such taxes; consequently, property tax incidence is likely to be regressive. Geographic Incidence Data requirements to assess the geographic incidence of program expenditures are more moderate. Detailed city expenditures accounts, which allow for an identification of recurrent and capital expenditures by different subunits in the city, are generally sufficient, and many cities have good information systems in this regard. The geographical pattern of expenditures can then be compared to the geographical distribution of poverty to establish whether funds flow into the most deprived areas. Graph 6 gives an example from Cali. Here communities are ranked by the share of total poor from left (communities with largest share of total poor) to right (lowest share of total poor). Similarly, the share of social expenditures going to each community can be calculated from administrative data. Graph 6 shows the difference between the share of the total poor per community and the share of social expenditures coming from the municipal treasury. If these expenditures were distributed according to the amount of poor per community, all bars in the graph would be at the zero line. Because the bars toward the left tend to be negative (i.e., the communities obtained less expenditures than they had poor people) and tend to be positive on the right, the 1997 distribution of social expenditures in Cali can be seen as anti-poor. 26 Graph 6. Social Expenditure Distribution against Poverty Distribution, Cali, 1997 20.00 10.00 0.00 -10.00 -20.00 Commun. commun. w/ most w/ least poor poor Source: Hentschel (2000) Budget Priorities The CPA might consider capturing the priorities of the city population for its municipal budget. If appropriately designed, multi-topic city surveys can collect such data. The EPSOC survey in Cali, for example, asked each respondent to answer two questions: ! If the city were to be able to increase funding for one city program, which one should be financed? ! If the city had to cut expenditures on a program, which program should it be? Table 11. Municipal Programs: Priorities for Expansion of Program, Cali, September 1999 Program Income Quintile Average 1 2 3 4 5 Education 31.3 30.9 29.2 32.3 34.8 31.7 Health 19.5 19.9 30.2 23.6 23.9 23.4 Employment and Income 18.9 22.2 18.6 18.7 19.8 19.7 Programs Nutrition Programs 8.8 4.4 5.6 5.6 1.2 5.1 Social Housing 10.4 11.7 8.4 5.9 5.5 4.8 Police 3.2 3.1 2.6 7.2 7.8 4.5 Water 2.1 1.5 1.5 1.9 1.4 1.7 Electric Lighting 1.8 1.8 0.8 0.5 1.5 1.3 Communal Households 1.0 2.6 0.6 1.5 0.6 1.2 (ICBF) Public Transport and Roads 1.4 0.8 0.9 0.8 2.0 1.2 Sports Arenas 1.3 0.9 1.3 0.7 0.8 1.0 Sewerage 0.5 0.1 0.2 0.8 0.2 0.4 Garbage Collection 0.0 0.0 0.1 0.7 0.5 0.3 Source: Hentschel (2000) 27 These questions checked for respondent consistency. Tables 11 and 12 show the results obtained. A clear pattern as to the priorities of Cali's population emerges. If resources were available, education, health, employment generation, and nutrition programs should be the beneficiary programs, according to the respondents. In the reverse case (table 12), these are exactly the programs that should be protected from cuts. Instead, the population suggests cutting expenditures for sports arenas, the police, public transport, and lightingareas relegated to lowest priority when an expansion of programs was probed. Table 12. Municipal Programs: Priorities for Cutback of Programs, Cali, September 1999 Program Income Quintile Average 1 2 3 4 5 Health 0.9 0.2 0.4 0.8 0.5 0.6 Education 0.8 0.6 1.1 0.7 1.1 0.9 Water 2.8 1.0 1.1 0.7 0.3 1.2 Nutrition Programs 0.9 2.5 0.8 3.3 1.8 1.9 Employment and Income 0.5 1.7 3.1 1.9 3.6 2.2 Programs Garbage Collection 0.7 3.8 1.9 1.6 2.6 2.2 Sewerage 1.1 3.3 4.2 3.2 2.1 2.8 Social Housing 3.3 1.1 5.5 5.5 3.7 3.8 Communal Households (ICBF) 10.5 9.5 5.5 5.9 8.0 7.9 Electric Lighting 7.6 14.1 12.5 10.1 9.4 10.8 Public Transport 18.2 12.9 12.9 17.5 18.0 15.9 Police 18.6 17.2 16.6 15.1 12.9 16.1 Sports Arenas 33.9 32.1 34.3 33.6 35.9 33.9 Source: Hentschel (2000) 8. City A nti-Poverty Programs City Poverty Assessments can be used to take an in-depth look at the functioning and effectiveness of existing anti-poverty programs. Institutional Map A good starting point for such an assessment is the preparation of an institutional map (see box 11). Such a map records detailed information about the total supply of social programs and services in the city (see table 13 for an example from Rio). Often, the maps begin with a sectoral assessment of the service or infrastructure supply (table 14 provides an example of the education sector in Haiphong, Vietnam). The maps look beyond the public sector (city and central) to include private, community-based, and nonprofit organizations. Their purpose is to draw a picture of the city's total supply of social and investment programs: who does what where and with how many resources. The maps can be physical, computerized, or both and are important information tools for all actors in the city. Preparation of an institutional map can go hand in hand with an in-depth review of the functions of the local government and its potential role in poverty reduction (see box 12, "Poverty Reduction and Municipal Functions in Rio"). 28 Box 11. Institutional Maps " Who are the main actors (public, private, voluntary) in the provision of social and productive services in the city? " What functions do these actors perform? " Where do they operate? " How much do they spend? " How many people do they reach? " How do they identify their beneficiaries? Table 13. Inputs for Institutional Maps: Information of Public Sector Programs in Rio Program Potential Target Population Households Approximate Reached Cost (1997) Kindergarten program for Households with young 26,055 R$9 m 0-6 year old children children and under R$360 (Creches) monthly household income (about 100,000 households) School Maintenance About 100,000 poor households 2,648 R$0.8 m Program for 7-14 year old with children 7-14 years Youth Training Program About 65,000 poor households 1,702 R$0.3 m for 15-18 year old with children 15-19 years Street Children Program 1,422 R$1.5 m Disabled Support Program Up to 10% of the population 1,702 N/A (530,000) are likely to be affected by some disability Elderly Support Program About 35,000 poor elderly that 2,494 N/A do not receive pensions (women above 60 and men above 65 years) Food Basket Distribution 20,000 R$3.5 m Source: Municipal Secretariat of Social Action. Target numbers from special tabulations of the 1996 PNAD by Sônia Rocha. Box 12. Poverty Reduction and Municipal Functions in Rio Municipal policies toward poverty are complicated, on the one hand, by the dependency on policies of other Government levels (for example, public security and water and sanitation under the responsibility of the State, and growth policy, minimum wage and unemployment policy under the Federal Government), and on the other hand, leakage of municipal services to residents of other, typically poorer, municipalities of the Rio metro area. Traditionally, the role of the municipalities, including its policy toward the poor, has been focused on the provision of urban services, and more recently the provision of basic health and education services. However, there is an increasing recognition of the potential role of a major municipality, such as the municipality of Rio, in fostering local growth and employment, and in establishing an effective system of social protection to complement the basic functions of a municipality. Source: World Bank (1998) 29 Table 14. School and Student Inventory in Haiphong, Vietnam, 1998 NGO QUYEN LE CHAN HONG BANG PRECINCT PRECINCT PRECINCT MAY CAU CAT BI DU HANG NIEM HA LY THUO TRAI CHAI TRE HANG KENH NGHIA NG LY CHUOI Number of schools in the ward Primary 1 2 1 1 1 2 1 1 1 Lower secondary 1 0 1 0 2 1 1 1 1 Upper secondary 0 1 1 0 0 0 1 0 0 Kindergarten 3 2 2 1 1 1 2 2 1 Other schools 0 3 0 0 0 0 0 0 Number of students Primary 1343 2411 1638 572 933 2990 1024 1415 749 Lower secondary 1721 0 1348 0 2267 1909 932 1823 558 Upper secondary 0 2438 1948 0 0 0 1282 0 0 Kindergarten 588 394 357 359 230 320 126 297 335 Number of classrooms and type of building Primary 21 21 24 12 13 37 13 16 21 Permanent 17 8 0 0 4 33 13 0 0 Semi-permanent 4 13 24 12 9 4 0 12 21 Temporary 0 0 0 0 0 0 0 4 0 Lower secondary 22 0 21 0 26 18 11 16 13 Permanent 14 12 12 14 11 16 0 Semi-permanent 8 9 14 4 0 0 13 Upper secondary 0 28 35 0 0 0 14 0 0 Permanent 0 25 10 14 Semi-permanent 0 3 25 0 Kindergarten 18 11 12 10 7 8 6 7 15 Permanent 6 11 7 4 0 0 6 0 0 Semi-permanent 12 0 5 6 7 8 0 7 15 Source: Luan et al. (1999) Institutional maps are key inputs into planning and reorienting city anti-poverty programs. They identify gaps and overlaps in program provisions between different actors and by geographical areas. Combining such maps with detailed information on the location of poverty (see discussion in section 4 on poverty maps) and the output of Service Satisfaction and Needs Assessments (see section 3) allows supply and demand information to be combinedand is thus a pivotal planning tool. Targeting, Coverage, and Benefit Transfer of Social Programs Using information from the institutional mapping exercise together with results from the incidence analysis discussed in section 7, an analyst can determine the targeting, coverage, and benefit transfer of various social programs. Targeting refers to the percentage of a program's total expenditures that goes to the poor. Coverage describes how many of the 30 poor are reached by the program. These are two distinct dimensions of program performance. For example, a program might be well targeted in that almost all of its expenditures may go to the very poor. Its coverage rate might nonetheless be low if only a very few of the poor are reached. Graph 7 brings these two dimensions together to analyze the effectiveness of various of Rio's social programs: the horizontal axis maps program coverage, while the vertical axis maps targeting efficiency. Programs with large coverage and good targeting are therefore in the upper right corner of the graph. Most programs in Rio have either good targeting and low coverage (lower right corner) or good coverage but weak targeting (upper left corner). A third dimension of program effectiveness is actual benefit transfer. Both household surveys and administrative information can provide an estimate of the actual benefit received by program beneficiaries. The size of the bubbles in graph 7 represent such benefit transfers per recipient. In Rio, the program with the largest benefit transfer is basic education. Graph 7. Targeting, Coverage, and Benefit Level of Social Programs in Rio de Janeiro, 1997 Program Benefits to Poor 120 Solid Waste 100 Basic Health Water 80 Sewage Basic Education 60 40 Kindergarten Favela Bairro 20 Bolsa Alimentar 0 0 20 40 60 80 100 120 Targeting (Size of bubbles reprsents per-family cost or benefit) Source: World Bank (1998) Efficiency of Programs Although difficult to measure, an assessment of the economic efficiency of public expenditures is essential.14 Questions to be answered here include: " Are the expenditures that appear to be directed to the poor directed toward high-return activities? " Are the programs operated efficientlyi.e., are a program's administrative and targeting costs justifiable when compared with the benefit derived from the program? 14 See World Bank (1992) for a review of program efficiency. 31 " In the broader city expenditure program, is there scope for efficiency-related cuts that would free resources for poverty reduction? " Are public expenditures directed at public goods and services that promote broad-based, efficient growth, or are they captured by special interest groups? Evaluation and Monitoring of Impact Social programsregardless of who provides themneed efficient monitoring and evaluation systems so city managers and other policymakers can determine whether the financed programs had the intended impact (e.g., improving health or reducing income poverty). The difficulty here is to develop systems that will distinguish the impact of the specific project from that of other developments. For example, to assess the effect of a nutrition project on child malnutrition, the program's direct nutritional impact must be clearly distinguished from the effect of rising incomes, for instance, on nutrition. The design of appropriate monitoring and evaluation systems is thus a precondition for designing effective city poverty reduction programs.15 9. City Growth and Poverty Reduction CPAs can also assess how a city's general economic performance is linked to poverty reduction. Obviously, a city that stagnates economically and has high unemployment rates will find it difficult to reduce poverty significantly. The poor's connection with the city's economic development works mainly through the labor market. Hence, it needs to be established what the main activities of the poor are, in which sectors they work, and what the most likely additional employment sources would be. In many countries, the poor's main income stems from informal sector activities in commerce and construction. Using household surveys, it is possible to calculate the impact of growth in these and other sectors on employment creation and to infer the resulting impact on poverty reduction. This information will give city managers a basic idea of howand what type ofcity growth will affect poverty reduction. Box 13, "Growth, Employment, and Poverty Reduction in Peru, 1994-97", provides some insightalbeit at the country levelas to the simple analysis that can be performed if consecutive data on poverty and employment are available. 15 For literature on monitoring and evaluation, a basic discussion of terms, and case examples, see the World Bank's impact evaluation Web site . See also Baker (2000). 32 Box 13. Growth, Employment, and Poverty Reduction in Peru, 1994-97 Employment growth in Peru has been closely linked to poverty reduction. Table 15 provides severe poverty rates and employment growth for the various sectors of the Peruvian economy. As the table shows, the three sectors with the highest employment growth ratesconstruction, trade and commerce, and servicesare also the three sectors that achieved the highest percentage decreases in poverty. Much of this employment growth provided families with more hours of work or a second source of income. On the other hand, agriculture/forestry and mining/petroleum/manufacturing had the lowest employment growth rates among the sectors and also showed the lowest percentage reductions in the severe poverty rate. Sectoral growth rates and employment creation are also connected. The table reports that the sectors expanding in Peru over the last few years, agriculture, construction, and trade/commerce, were associated with growing employment rates. While the real growth rates noted here capture the output of formal enterprises only, it can be assumed that if formal sector growth is high, supporting or parallel informal enterprises within the sector should also realize an upswing. On face value, Peru's growth pattern was pro-poor over the time period considered, in that it was driven by the sectors in which the severe poverty rates were the highest. In two of these sectorsconstruction and trade/commercereal growth translated into employment growth and poverty reduction. However, the country's impressive agricultural growth rates did not translate fully into employment creation. The sector's real growth rate is estimated at 23 percent over the 1994-97 period, making it the best performing sector after construction. However, agricultural productivity was seriously depressed at the beginning of the 1990s; consequently, any growth generated in this sector would have been achieved without excessive cost outlays for labor or productione.g., by having the existing workforce work longer hours. This could explain why employment grew slowly during agricultural sector expansion and why poverty reduction in this sector was less than could have been hoped for. Note: All households have been assigned a primary sectori.e., the sector in which the household's main income earner is employed. Source: World Bank (1998). Table 15: Sectoral Poverty Reduction and Growth Rates in Peru, 1997-97 Severe poverty rate Growth % change Sector 1994 1997 1994_97 Employment Real Agriculture and 31.8 26.4 -17.0 10.3 23.4 forestry Construction 25.2 17.4 -31.0 63.9 33.8 Transport and 11.8 10.2 -13.0 18.0 NA communications Trade and commerce 13.8 8.6 -37.5 43.9 22.8 Mining, petroleum, 9.2 8.4 -8.5 7.9 13.7 and manufacturing Services 11.9 8.8 -26.0 21.6 8.4 Total 18.8 14.8 -21.0 19.0 100 Sources: World Bank (1999). Real growth rates from Central Bank of Peru (1998). But employment creation for the poor can also be a function of factors other than city economic performance. Regulations, for instance, can be impediments for the poor to establish their own enterprises. Taxes might deter establishment of enterprises or investment 33 by outsiders. And inadequate transportation might be a major problem, if it doesn't allow the poor to get to places where jobs are. The CPA would analyze all such factors and enable the development of appropriate policy recommendations. 10. Concluding Remarks This paper provides an introduction to the concept of and tools used in City Poverty Assessments. There is no standard content to such assessments; rather, they need to be adapted to the specific needs of the city involved. Several aspects of urban poverty touched on in this paper will be irrelevant to certain circumstances, while others not mentioned here will be crucial. The thrust of CPAs is to provide city policymakers with good and thorough information about the situation of the city's poor, the functioning of city anti-poverty programs, and the link between poverty and growth. Many of the tools developed in the course of a CPA are valuable planning tools in and of themselves, such as poverty maps, institutional maps, tracking of the incidence of taxes and expenditures, and rapid service satisfaction surveys. In the final analysis, the very process of preparing a City Poverty Assessmentwhich includes collecting information, analyzing it, and discussing it with all relevant actors, including the poorwill be of major importance in forming new and more effective partnerships and in understanding city poverty reduction. 34 References16 Alam, M., J. Sathaye, and D. Barnes. 1998. "Urban Household Energy Use in India: Efficiency and Policy Implications." Energy Policy 26:885-91. Altaf, M.A., et al. 1993. "Poverty in Karachi: Incidence, Location, Characteristics, and Upward Mobility." Pakistan Development Review 32:159-70. Baker, Judy. 2000. Evaluating the Impact of Development Projects on Poverty, Directions in Development. Washington D.C.: World Bank. Central Bank of Peru. 1998. Boletin Mensual. Lima. Deaton, A. 1997. The Analysis of Household Surveys. Baltimore: Johns Hopkins University Press. Egunjobi, L. 1989. "Perception of Urban Environmental Problems: A Pilot Study of the City of Ibadan, Nigeria." African Urban Quarterly 4:59-67. Fajnzylber, P., D. Lederman, and N. Loayza. 1998. What Causes Violent Crime? Washington, DC: World Bank. Fiszbein, A. 1997. "The Emergence of Local Capacity: Lessons from Colombia." World Development 25:1029-43. Glewwe, P., and G. Hall. 1995. Who Is Most Vulnerable to Macroeconomic Shocks? Hypotheses Tests Using Panel Data from Peru. Living Standards Measurement Study 117. Washington, DC: World Bank. Grosh, M., and J. Muņoz. 1996. A Manual for Planning and Implementing the Living Standards Measurement Study Survey. Living Standards Measurement Study Working Paper 126. Washington, DC: World Bank. Gupta, H.S., and A. Baghel. 1999. "Infant Mortality in the Indian Slums: Case Studies of Calcutta Metropolis and Raipur City." International Journal of Population Geography 5:353-66. Hentschel, J., J. Lanjouw, P. Lanjouw, and J. Poggi. 2000."Combining Census and Survey Data to Study the Spatial Dimensions of Poverty.", World Bank Economic Review 14: 147- 165. Hentschel, J. Forthcoming. Rapid City Surveys as Tools for Municipal Social Policy Making: An Application in Cali, Colombia. Policy Research Working Paper. Washington, DC: World Bank. 16 The word processed describes informally reproduced works that may not be commonly available through libraries. 35 Huges, G., and M. Dunleavy. 2000. "Why Do Babies and Young Children Die in India? The Role of the Household Environment." Processed. Washington, DC: World Bank, Environment Department. INEI. (Peruvian National Statistical Institute). 1997. Encuesta Nacional sobre Violencia. Lima. Jacobi, P.R. 1994. "Households and Environment in the City of Sao Paulo: Problems, Perceptions and Solutions." Environment and Urbanization 6:87-110. Luan, T.D., N.X. Mai, and V.T. Anh. 1999. "Poverty and Social Issues in Haiphong City." Processed. Hanoi: Institute of Sociology. Mensah, J., and H.A. Whitney. 1991. "Some Third World Environmental Perceptions and Behaviours Concerning Urban Waste: A Survey of Techiman, Ghana." Canadian Geographer 35:156-65. Pamuk, A., and P.F. Cavalieri. 1998. "Alleviating Urban Poverty in a Global City: New Trends in Upgrading Rio de Janeiro's Favelas." Habitat International 22:449-62. Paul, S. 1998. Making Voice Work: The Report Card in Bangalore's Public Service. Policy Research Working Paper 1921. Washington, DC: World Bank. Ravallion, M. 1994. Poverty Comparisons. Chur, Switzerland: Harvood Academic Publishers. Ruan, D. 1993. "Interpersonal Networks and Workplace Controls in Urban China." Australian Journal of Chinese Affairs 29:89-105. Salmen, L. 1995. Beneficiary Assessments: An Approach Described. Environment Department, Social Assessment Series No. 23. Washington, DC: World Bank. Santamaria, M. 1999. "Poverty in CaliBasic Comparisons and Developments." Processed. Washington, DC: World Bank, Poverty Group Swaminathan, M. 1995. "Aspects of Urban Poverty in Bombay." Environment and Urbanization 7:133-43. United Nations Centre for Human Settlements (UNCHS), Global Urban Observatory. 1999. Urban Indicators Wong, C. 1997. "How Many Poor People in Shanghai Today? The Question of Poverty and Poverty Measure." Issues and Studies: A Journal of Chinese Studies and International Affairs (Taiwan) 33:32-49. World Bank. 1992. Poverty Reduction Handbook. Washington, DC. . 1997. "Core Welfare Indicator Questionnaire." Background documentation.. Washington, DC. . 1998. "Poverty in Rio de Janeiro." Processed. Washington, DC. 36 . 1999. Poverty and Social Developments in Peru, 1994 to 1997. Country Report. Washington, DC. Internet Resources Center on Urban Poverty and Social Change, Case Western Reserve University, . Inter-American Development Bank, . International Forum on Urban Poverty, UNCHS Habitat, . National League of Cities, . United Nations Centre for Human Settlements (UNCHS), Habitat, Urban Indicators Programme, and . The United Nations Environment Program at Arendal, . The Urban Institute, . The World Bank poverty page, . The World Health Organization, . World Resources Institute, . 37