4. EXAMPLES OF MAPPING HUMAN WELFARE AND POVERTY

Examples of mapping human welfare and poverty can be found at international and national level. No concerted multi-institutional effort to produce a global poverty map has been initiated yet. Different activities by international organizations such as IFAD, UNDP, and the World Bank, selected CGIAR Centers, and individual countries, however, can become the starting point for international poverty mapping exercises. The following review shows what progress has been made by selected institutions and the differences in approaches, resolution, and country coverage.

4.1 INTERNATIONAL EFFORTS

Three international sources, IFAD, UNDP, and the World Bank, can provide information on poverty and human well-being at country level. In some of their reports, human well-being and poverty indicators have been mapped or are at least listed at coarse subnational level.

IFAD has produced national level poverty indicators and prepared a descriptive table on the location of the rural poor in 56 countries. This information is summarized in Appendix 4, page 84.

UNDP has published country level Human Development Reports and has calculated composite indicators (Human Development Index, Human Poverty Index, and Capability Poverty Measure) at subnational level for selected countries. Appendix 5, page 90, is a bibliography of publications and gray literature used for this effort. It includes country studies describing the conditions and trends in human development and background papers summarizing the methods for disaggregating UNDP's composite indexes.

The World Bank has completed poverty assessments for 35 countries. Although the majority of these assessments do not map poverty indicators at a very detailed level, the listed publications usually provide data disaggregated by urban and rural characteristics and by broad geographic regions. They are a useful entry point to understand poverty and human welfare issues for a country. This bibliography of poverty assessments can be found in Appendix 6, page 95.

4.2 ACTIVITIES BY CGIAR CENTERS

No coordinated poverty mapping effort involving all CGIAR centers is currently underway. All CGIAR Centers were contacted to inquire about their activities to map human welfare indicators or study the relationships between poverty, environment, and land use. Five of the 16 Centers provided information about ongoing or planned projects.

Of the five respondents, CIAT has made the most significant investments in mapping human welfare indicators. The CIAT Hillsides Project started in 1995 with poverty assessments in Colombia and Honduras. In 1996, poverty mapping exercises for Honduras and Nicaragua were initiated and additional GIS staff was hired in 1997 to expand upon these efforts (Leclerc, 1997).

CIAT has produced a working paper reviewing different poverty assessment methods (Oyana, 1997). The paper outlines poverty mapping activities for Honduras and proposes a composite human needs index, which combines an educational attainment index (enrollment ratio and adult literacy rate) with a shelter quality index and a health status index. Research for Honduras is currently or will be conducted at different scales including departments (18 units), municipalities ('municipios' - 291 units), villages (3,792 units), and selected watersheds. CIAT staff have proposed to produce various other working papers that are summarized under the following titles: 'Tools and Methodologies to Handle Poverty Data and Studies,' 'Lessons Learned: Poverty Mapping in Latin America,' and 'Census Data: A Means to Understand Spatial Complexities.'

1. Develop a spatially integrated poverty database.
2. Produce maps at various scales.
3. Examine relationships between poverty, land use, and policy variables.
4. Communicate results and return data to data providers at national, municipal, and village level in appropriate formats.

So far, poverty mapping activities for Honduras have made the most progress. CIAT envisions using multiple measures of well-being based on household expenditures (and a designated poverty line) and composite indexes measuring unsatisfied basic services and human needs with the best available data. The composite indexes will include variables related to shelter (quality of housing and type of dwelling), biophysics (land cover, slope, agro-climate, soil, and protected areas), infrastructure (roads, electricity, and telephones), education (enrollment, number of schools, student-teacher ratio, etc.), health (infant mortality, safe water, sanitation, number of physicians, and primary care facilities), and nutrition (per capita food consumption).

By the end of September 1997, the following data sets were assembled for Honduras: Population census for 1974 and 1988, agricultural census for 1974 and 1993, health status variables for subnational administrative units, and data used for the 'Fondo Hondureno de Inversion Social' (FHIS), the latter data at municipal level of four variables: population density, infant malnutrition, access to water, and access to sanitation.

Most of the outputs are still preliminary, use census data, and include selected education variables and disaggregated demographic data by age groups at village level. Based on these data an educational achievement index was calculated for each village and 'municipio.' This index was then compared with the demographic characteristics of villages, population density, and poverty levels at 'municipio' and village level.

Preliminary maps classify villages along an urban-rural continuum and group municipalities according to different education variables (number of schools, enrollment, number of teachers, number of students repeating selected grades, educational achievement, literacy rate) and population density. Other exploratory research has looked at the relationship between a basic service index and a water balance risk index for different seasons and conducted sensitivity analysis of educational variables at multiple scales ('municipio' and 'aldea' [community] level).

Honduras will be one of the case studies for CIAT's work on participatory poverty assessment. Based on methods developed in Africa and Colombia, CIAT expects to develop and test a method to extrapolate results from a participatory assessment (well-being ranking). See 5.3.3, page 62, for a detailed discussion.

Most of the poverty mapping work in Nicaragua is still in the initial phase of assembling core data sets. Various thematic maps have been produced. They include poverty measures from the latest LSMS survey and maps showing population without access to basic needs (education, water, and health services) for 214 'cordobas.'

As in Honduras, CIAT expects to use the same participatory approach in Nicaragua. This has been planned for the Rio Callico watershed in Matagalpa.

In addition to the work in the two Central American countries, CIAT is testing other proxy indicators for poverty. One example, combined farm size, based on digitized cadastral maps, with a digital elevation model for selected watersheds in Colombia and Peru.

CIAT is currently exploring the potential for a poverty map at continental scale. Researchers have started to review existing country efforts to map poverty and human welfare indicators. Staff have obtained digital data sets from poverty assessments in Venezuela, Ecuador, and Peru, and expect to obtain such data for Bolivia. Based on preliminary findings, the continental work will attempt to map poverty at first subnational level. The greatest challenge seems to be integrating different national data sets that are based on different methods and poverty definitions (Hyman, 1997).

The Center has prepared a proposal to analyze the relationships between rural poverty and environmental degradation in the Altiplano (Winters, 1997). It envisions a set of case studies that will be carried out with selected partners in Bolivia and Peru. Methodologically, CIP intends to use maps for presentation and identification of field sites, but not specifically for analysis. The Center and its partners intend to carry out the following workplan:

1. Use remote sensing images to identify environmental degradation for two time periods capturing three major environmental problems: salinization, soil erosion, and removal of woodlands.
2. Examine the depth and severity of poverty in these areas. Data will be collected with household surveys and combined with national poverty data (quality of life index) at the province level for Bolivia and the district level for Peru.
3. Examine the links between the quantity and quality of household assets, environmental degradation, and existing policies (macro-economic, sectoral, institutional, etc.).
4. Try to extrapolate from these case studies to the whole Altiplano watershed and make recommendations on appropriate changes of selected policy variables.

ICARDA has prepared a background paper on rural poverty and natural resources in the "Dry Areas" falling under ICARDA's mandate (Rodríguez, 1997). The paper's objective was to review the severity and distribution of rural poverty, identify natural resource constraints, and relate these indicators to the production value of ICARDA's mandate commodities. The paper used national rural poverty figures from IFAD's global poverty assessment. For countries without poverty data, a linear regression between total poverty and infant mortality was used to estimate missing poverty rates. ICARDA developed a Rural Poverty Indicator (RPI), which combined per capita GDP (as an estimate for income) with a Gini coefficient of income or expenditure inequality. Countries were then ranked by multiplying the RPI with the number of poor, to help set priorities for ICARDA.

At present, no high resolution poverty maps have been produced. The GIS staff has integrated relevant national data sets related to poverty, human well-being, natural resources, agricultural production, and labor force, and will provide thematic maps based on these data to ICARDA's staff. ICARDA, upon request and in cooperation with national programs, will create or use data at higher resolution (provincial, regional, or watershed). Possible research questions include analyzing the correspondence between natural resource endowment and incidence of poverty and examining the impact of agricultural research on human well-being (Thomas, 1997).

IFPRI has published two studies that estimate poverty and malnutrition, respectively, by agro-ecological zones. Their estimates were generalizations of survey data (Broca and Oram, 1991; Sharma et al., 1996).

The first study by Broca and Oram on the location of the poor was initiated by the TAC in 1990 which was looking for the most up-to-date and consistent poverty estimates by agro-ecological zones. The authors of the study employed two definitions of poverty:

1. A household is absolutely poor if it cannot meet 80 percent of its required daily allowance of calories (data came from nutrition surveys).
2. A household is absolutely poor when its food expenditures is less than 80 percent of the amount required to purchase the required daily allowance of calories (data came from expenditure surveys).

The first definition was the preferred one, but when no other data were available, the second definition was used. Agro-ecological zones are based on FAO definitions.

Since household surveys in developing countries (in this case nutrition and expenditure surveys) have not been consistently stratified by agro-ecological zones, the study had to generalize from one or a few sites to a whole agro-ecological zone, sometimes covering more than one country of a geographical sub-region. For example, poverty estimates for 33 sub-Saharan countries are based on surveys from 22 sites in nine countries (Burkina Faso, Ethiopia, The Gambia, Kenya, Niger, Nigeria, Rwanda, Sudan, and Zambia). Most estimates for Asia used poverty measures at the first subnational level (states, provinces, and regions) which were provided by different national assessments. The poverty estimates for Latin America are based on national figures.

The second study by Sharma et al. used anthropometric surveys that were then generalized to the same FAO-based agro-ecological zones. Anthropometric data of children was interpreted as a proxy for poverty. The relationship between poverty and anthropometric data is especially strong for countries at the lower end of the income range.

While the Sharma et al. study had the same problem of generalizing from a limited number of surveys, it had the advantage of using an internationally accepted and comparable indicator of well-being (the Broca and Oram study is an amalgam of different poverty definitions). Neither study had access to geo-referenced survey data and sampling weights, making it impossible to calculate meaningful averages and standard errors for agro-ecological zones. Section 5.2, page 51, will demonstrate how geo-referenced survey data can be used to calculate such statistics. It will also highlight the great spatial variability within agro-ecological zones, thus making generalizations from a few sample sites questionable.

ILRI and IFPRI are conducting a joint project examining policies for sustainable land management in Mixed Crop Livestock Systems for the Highlands of East Africa (Ehui, 1997). The research will be initially implemented in Ethiopia with possible extensions into Uganda and Kenya. The goal is to understand the extent and main causes of land degradation and suggest policies to improve soil and water management, increase agricultural productivity, and reduce poverty. The project description did not explicitly mention maps and mapping activities related to poverty and human well-being.

Back to CONTENTS

4.3 COUNTRY ASSESSMENTS

Policy makers in different countries are using maps or data disaggregated by subnational administrative units of human well-being and poverty indicators to guide the allocation of public resources, plan infrastructure development, design poverty alleviation programs, etc. Methods and applied scales vary from country to country. This makes it difficult to integrate and compare such national poverty maps to generate an international view.

Four examples, three in Latin America, one in Africa, will be summarized. The El Salvador study uses a simple aggregation of various variables. The Nicaragua and Peru study show how multivariate statistical methods (small area estimation) can be used to produce disaggregated poverty maps. The South Africa study combines income-based poverty measures and a Human Development Index at subnational level.

The study for El Salvador classified those households as poor that cannot satisfy their basic needs such as housing and education and created a 'unsatisfied basic needs index' (Ministerio de Coordinación del Desarrollo Económico y Social, 1995). Fourteen indicators from the 1992 Population and Housing Census were combined to construct the index:

• Infant mortality
• Illiteracy rate of population older than 10 years
• Overcrowding (percentages of dwellings with more than 3 people per room)
• Percentage of dwellings with earth flooring
• Percentage of substandard dwellings/improvised housing
• Percentage of dwellings without drinking water service
• Percentage of dwellings without sanitation services
• Gross school enrollment rate (grade 1 - 6)
• Gross school enrollment rate (grade 7 - 9)
• Proportion of students enrolled in grade 1 - 6 who are older than the required age for this level
• Proportion of students enrolled in grade 7 - 9 who are older than the required age for this level
• Population in 'municipio' as a percentage of country population
• Rural population as a percentage of 'municipio' population
• Percentage of population under 15 of 'municipio'

• Identify the maximum and minimum values of each indicator.
• Define ten equal intervals between the maximum and minimum value.
• Assign a value of 10 to the highest interval and a value of 1 to the lowest interval.
• For each 'municipio' and for each indicator assign a value between 1 and 10 depending into which interval category the respective indicator falls. For example, the maximum value for 14 indicators would be 14 x 10 = 140, representing areas with the highest priorities. The lowest value would be 14 x 1 = 14, representing 'municipio' with the lowest priorities.

Poverty assessments in Nicaragua have used an 'unsatisfied basic needs index' and household consumption expenditure in relation to a poverty line. Examples for the 'unsatisfied basic needs index' include composites of variables on basic education, health service, and access to safe water and an index used for public sector spending ('Fondo de Inversión Social de Emergencia') that combines three principal variables: child malnutrition, access to safe drinking water, and proportion of displaced people in the community. FGT-type poverty measures based on household consumption expenditures have been produced for seven large geographic sub-regions from the most recent LSMS (McKinnon, 1994).

The objective of a more recent study was to produce income-based poverty indicators at the 'municipio' (143 units) level and improve on the methods used for the targeting of public social funds, the 'Fondo de Inversión Social de Emergencia' (Arcia et al., 1996). The study used multivariate regression (small area estimation) to calculate monthly household expenditures for municipalities that did not have a sufficient sample size to calculate robust statistics from the LSMS data. Socio-economic variables on education, infrastructure, health, and nutrition available for all municipios were the independent variables in the regression model.

In Peru, various poverty assessments based on an 'unsatisfied basic need index' and on an income-based poverty line have been produced over the past years. A recent example integrated socio-economic information from the 1993 national population and housing census (Censo de Poblacion y Vivienda - CPV) with 1995 national household survey data (Encuesta Nacional de Hogares - ENAHO) to overcome some of the shortcomings of the previous approaches (INEI, 1996).

Basic needs indexes have the shortcoming of not clearly differentiating between correlation and causality of poverty and implying for example that low access to services means a household is poor. They do not differentiate between the quality of services and do not measure the degree or severity of poverty. Survey data have the problem of a small sample size that make it difficult to extrapolate the data to subnational units.

• Find the statistical relationship between the survey and the census data. This requires a common set of variables and compatible definitions between the two datasets.
• Based on these common variables, develop a prediction model that expresses household income as a function of household characteristics, education, locality, etc.
• Apply this model to the census data and use the census data to predict the proportion of households below the poverty line at departmental and provincial level.

The South Africa example used data from the 1993 Living Standards and Development Survey (LSDS), conducted by the South African Labor and Development Research Unit and funded by the World Bank, and the 1991 Population census (Whiteford et al., 1995). Poverty indicators include income-based measures (headcount index and poverty gap based on household income, and Gini coefficient of income inequality) and a Human Development Index (HDI). The HDI was a composite of life expectancy, literacy, and income. Since life expectancy data were not available by magistral districts, province level data were used. Literacy was defined as the percentage of adults who had completed Standard 5.

The maps and tables show poverty measures for 9 provinces and 371 magistrates. Data were also disaggregated by race, gender, and educational level.