Preparation of this review to map rural poverty in developing countries was motivated by a study, carried out by Nelson et al. (1997) on behalf of the Technical Advisory Committee (TAC) of the Consultative Group on International Agricultural Research (CGIAR), on CGIAR research priorities for marginal lands. The study differentiated between marginal agricultural land (MAL) and favored agricultural land (FAL) and estimated their extent and the number of people living in these areas. According to this analysis, about 634 million rural poor are living in marginal lands, of which 375 million (59%) are in Asia. The report also shows that only 25% of the 374 CGIAR projects endorsed in 1997 are fully targeted at poverty alleviation in MAL, suggesting a need "to sharpen its [CGIAR's] strategic focus on poverty alleviation particularly in setting priorities for research related to marginal rural areas."
The relevance of such research targeting the poor was highlighted by a study by the International Food Policy Research Institute (IFPRI) which found that public investments in low-potential rainfed areas, for example with high-yielding varieties, irrigation, and education, would increase agricultural productivity and reduce rural poverty in India, providing a greater gain per unit of additional investment than similar investments in irrigated or high-potential rainfed areas (Fan and Hazell, 1997). Similarly, a study based on the 1992-93 Living Standard Measurement (LSMS) Study survey of Vietnam found that the highest increase in net crop income would occur in Vietnam's two poorest regions: the Northern Uplands and the North Coast (van de Walle, 1996).
These two studies took advantage of disaggregated data on population, incidence of poverty, land use, and infrastructure. But in many developing countries the empirical basis for characterizing and mapping marginal lands is so weak and at times unavailable as to make policy recommendations meaningful. Similar limitations are apparent in the Nelson et al. study: the soil and length of growing period maps used to define MAL and FAL included no information on land cover or use, population data were only available at the first subnational level, and a constant poverty rate was applied for all areas within a country.
In its review of Nelson et al., the TAC (1996) identified the following key limitation in our understanding of the nature and distribution of marginal lands: "the lack of readily available data in a geo-referenced framework, in particular with respect to the incidence and nature of poverty and probability of land degradation by land type." The TAC recommended a "review of available data on poverty and land degradation in relation to these marginal lands."
In response to this observation, United Nations Environment Programme (UNEP)/GRID-Arendal contracted with World Resources Institute (WRI) to carry out a review of poverty mapping (see Appendix 1 Terms of Reference - Poverty Mapping Assessment, page 75). This review is part of an ongoing collaborative project between UNEP and CGIAR to strengthen the use of Geographic Information Systems (GIS) in agricultural research, assist in the production of reliable statistical and cartographic products, for example on poverty and land use quality, and contribute to further development of global databases relevant to agricultural research. WRI had previously mapped human development indicators in West Africa to support regional priority setting by the Abidjan-based Regional Office of the U.S. Agency for International Development (USAID).
1.2 POVERTY MAPS - APPLICATIONS AND USERS
Most national poverty assessments using household and community surveys have compiled data that allow disaggregation by broad categories such as urban and rural areas, socio-economic characteristics such as household types and educational backgrounds, and major geographical regions such as a coastal, forest, and savanna zone. These poverty assessments have helped in| Figure 1 Poverty Map of Uganda | ||
![]() |
||
1.3 SPATIALLY DISAGGREGATED DATA - AT WHAT SCALE OR RESOLUTION?
The final applications of these poverty maps ultimately determine the appropriate scale or resolution. For example, data can be analyzed at the individual, household, village, community, administrative, national, or regional level. Whether comparing countries, setting research priorities, studying causes and effects, developing a baseline for monitoring, or targeting specific project interventions, each application requires a specific resolution for reference units. Data can generally be aggregated from the individual to the macro level, and analysts need to balance detail and coverage required for analysis with the cost of data collection. A coarse resolution or a scale too small neglects the heterogeneity within each unit and provides insufficient detail for decision making, a fine resolution or a scale too large increases the cost of compiling, managing, and analyzing the data. In addition, data at coarse resolution, for example national poverty indicators, usually are more readily available and cover a wider geographic area. It is difficult to pre-determine an ideal resolution or scale that would be a perfect framework to guide all research priorities for marginal lands. The TAC and CGIAR Centers need to define the purpose and specific applications of their poverty mapping more precisely and determine how accurately they want to reflect the spatial distribution of poverty. Ultimately, multiple assessments and scales will be necessary, and the optimal scale will be determined by the loss attached to errors of identification in locating the poor.