Poverty maps are not only influenced by the selection of a conceptual approach to define poverty and by the choice of a specific poverty indicator. The data collection method itself can determine the resolution of the poverty map and the type of analysis to conduct.
A brief review of different data collection methods will highlight the pros and cons of various subjective and objective methods and the trade-off between survey and census data. A short section summarizing major sources for international poverty maps will show that the pool of existing data for a global poverty map is limited. Additional investments in data collection and modeling need to be made to produce maps with higher resolution, more comprehensive poverty measures, and a wider international country coverage.
Methods to collect data for poverty assessments can be grouped into two approaches, bottom-up and top-down approaches. Bottom-up approaches solicit active participation of the poor, incorporate their perspectives into the assessment, and generally are more qualitative in nature. Top-down approaches rely more on questionnaires, collect information via a survey or census, and tend to be more quantitative in nature. The following summary is based on reviews by Lok Dessallien and Oyana, respectively. For more detail refer to Lok Dessallien (1996), Oyana, (1997), Kingsbury et al. (1995), United Nations (1984), and United Nations (1991).
Figure 2, page 22, summarizes some of the major data collection methods and presents them within a two-dimensional space. The horizontal axis depicts a continuum of methods ranging from case studies to census. A movement along this axis toward the right reflects an increase in sample size to complete coverage of the population, lower frequency of data collection, and an increase in cost and effort to collect and process data. The vertical axis, a continuum from subjective assessment to direct measurement, represents a movement from more subjective to more objective methods. Most bottom-up approaches can be found in the two left quadrants. The top-down methods are in the upper right quadrant.
The upper right quadrant illustrates some of the methodological challenges for poverty mapping. While census data permit easy aggregation to appropriate subnational units and a fine resolution map, they provide only selected and sometimes outdated indicators of well-being. Surveys yield more up-to-date and relevant indicators, but require additional processing and modeling efforts to overcome the limitations of small sample sizes and produce poverty maps of adequate resolution.
Major bottom-up approaches in Figure 2 include intensive anthropological and sociological methods (ethnographic and participant observation), participatory and rapid appraisal methods, and beneficiary assessments (systematic consultations with project beneficiaries and stakeholders). Knowledge, attitudes, and practices (KAP) studies have been used primarily in health and family planning to identify decision making patterns, perceptions, and awareness. They combine formal questionnaires with sampling methods and use qualitative approaches relying on key informants and focus groups. Oyana mentions the 'E Delbecq-Delphi' method which has been applied for vulnerability mapping in Bangladesh. It relied on experts who were familiar with the study area and included technical and development specialists, government officials, and village elders. KAP and 'E Delbecq-Delphi' are hybrid methods, employing bottom-up and top-down components (Oyana, 1997).
Bottom-up approaches bring with them the advantage of allowing participants to apply their own criteria to define poverty, thus making them the main stakeholder of poverty assessments. This in turn provides a better foundation for identifying solutions and implementing interventions. A participatory survey is usually less costly than a household survey and produces outputs faster. It provides micro-level information and identifies nuances of poverty that become very important when analyzing causes of poverty.
The disadvantage of bottom-up approaches is that they use relatively small samples that make it difficult to extrapolate results and compare different surveys. A second major limitation is that the quality of participatory approaches varies greatly with the skills of the facilitators and the established level of trust between facilitators and participants.
Two studies, one in Honduras and the other in Tanzania, provide instructive examples of participatory methods. The Honduras study proposes a method to quantify and extrapolate local perceptions on poverty (see 5.3.3, page 62). The Tanzania study, carried out by the World Bank, was able to bridge some of the methodological gaps usually found between these two approaches (Narayan, 1997).
The participatory poverty assessment for Tanzania sampled 100 villages which are part of Tanzania's National Master Sample framework. National-level studies are conducted in these framework villages to permit generalizations to the nation's rural area as a whole and to make findings from different surveys comparable. The study employed three methods to collect data: participatory tools (community mapping, group discussions involving wealth ranking, trend and price analysis, gender analysis, and Venn diagrams of village-level groups and institutions), key informant interviews (tried to answer similar questions as elicited at group sessions), and household surveys. The household surveys consisted of two questionnaires: The first asked questions related to social capital and the second tried to capture household consumption and expenditures.
The study's work on social capital is a good example how to measure structural conditions that determine poverty, discussed in 2.2 under the enabling environment dimension of well-being. Based on the household survey, the study developed a Social Capital Index, which represented the average of both the number and characteristics of groups to which a household belongs.
Both the conventional household survey and the participatory approach yielded similar aggregated results measuring poverty. The participatory assessment provided more subtle and detailed observations related to social capital, gender, seasonality, and access to water that were not picked up by the household consumption survey. The study's conclusion sees participatory poverty assessments as a useful tool for interim poverty monitoring between major surveys.
Examples of the top-down approaches that can provide information for poverty assessments are presented in the upper right quadrant of Figure 2. They include population and housing censuses and different types of household surveys based on a probability sample.
Because population censuses are, by definition, comprehensive in their coverage, they are usually only designed to provide information on the structure and distribution of the population, not on poverty. However, they may collect information on educational attainment and sometimes under-five mortality statistics can be calculated from the demographic parameters. Only a few censuses in developing countries have included questions on income. If a country conducts a regular housing census, poverty can be inferred from questions related to type and size of dwelling, water supply, sanitation, and cooking facilities.
Many household surveys, on the other hand, which are based on a sample of the population (typically less than 1%), contain detailed questions on economic indicators of well-being (e.g., consumption, income) or non-economic measures related to health, education, and services. A number of surveys provide internationally comparable data relevant to characterizing marginal populations. They include the Demographic and Health Survey (DHS), Living Standards Measurement Study (LSMS) Survey, and other surveys under the World Bank's Social Dimension of Adjustment Program.
The DHS was established by USAID to provide information on fertility, health, and morbidity. As of 1997, surveys in 59 developing countries have been carried out, often repeatedly, by Macro International, a Maryland-based consultancy. The DHS is a specialized survey. Other specialized (single topic) surveys that provide information relevant for poverty assessments include household budget surveys, labor force surveys, agricultural surveys, and other social surveys such as food consumption and nutrition surveys.
The LSMS was established by the World Bank in 1980 to improve the type and quality of household data collected by statistical offices in developing countries. The first surveys were conducted in 1985. As of 1997, LSMS surveys have been carried out in 31 developing countries. The LSMS is a multi-topic household and community survey. It consists usually of three major modules, a household survey, a community level survey, and a price survey that tries to measure purchasing power.
Other survey instruments used by the World Bank to assess and monitor poverty include Integrated Surveys (IS) and Priority Surveys (PS) under the Social Dimension of Adjustment Program. The IS is an in depth survey, similar to the LSMS, and provides information to assess impacts of structural adjustments on households. A PS is conducted more frequently (ideally annually) and uses a large sample to insure that all population groups are represented. Another, rapid survey, the Core Welfare Indicators Questionnaire, is being developed and field tested by the World Bank (Lok Dessallien, 1996).
The Sentinel Site Surveillance in Figure 1 combines bottom-up and top-down elements and tries to monitor policy impacts. It is based on small samples, uses a minimum of numerical data and interviews, focuses on key informants, and communities participate in questionnaire design, data analysis, and communication of results.
Back to CONTENTS
To produce internationally comparable maps that show poverty measures at subnational level requires consistency in the definition of the indicators used and a wide geographic coverage. National population and housing censuses and international surveys such as the LSMS and DHS can provide the principal variables for such mapping or become the input for subsequent modeling efforts. A potential list of variables from these three sources (Table 7, page 25) makes clear that most of these indicators are captured by the social dimension of human well-being. This restricts international poverty mapping to these available variables or will require additional investments in new data collection or modeling to estimate missing indicators.
The geographic coverage and the timeliness of these data further restricts the universe of countries. Of 141 developing countries, only 59 have conducted a DHS, 31 a LSMS, and 72 have a population and housing census with data collected after 1991 (Table 8, page 25). See Appendix 2 Availability of Survey and Census Data, page 77, for a detailed country list).
There are other international sources providing for example economic measures such as the Integrated Surveys, Community Surveys, and surveys conducted under the Social Dimension of Adjustment Program of the World Bank. Appendix 3, page 80, summarizes all housing surveys from these sources completed in Africa since 1985. Over the past 12 years, most countries in Africa conducted at least one household survey that provides variables of relevance to poverty assessments. However, only two countries, Tanzania and South Africa, provide access to their survey data. All other countries require special Government permission or have not established a data access policy yet.
Table 7 Variables Related to Poverty and Human Welfare - Census, LSMS, and DHS
|
|
|
|
| Health | |||
| Anthropometric measurements | |||
| Child Mortality | |||
| Disability | (selected countries) | ||
| Education | |||
| Literacy | |||
| Educational attainment | |||
| School attendance | |||
| Economics | |||
| Economic characteristics of households | |||
| Occupation | |||
| Status in employment | |||
| Total consumption | |||
| Household income | (selected countries) | ||
| Total household expenses | |||
| Total food expenses | |||
| Access to services | (selected countries) | ||
| Housing | |||
| Type of building | |||
| Number of rooms, floor space | |||
| Water supply | |||
| Sanitation | |||
| Cooking facilities | |||
| Number of occupants (crowding) | |||
|
|
| |||||
|
|
|
|
|
|
| |
| Africa |
|
|
|
|
|
|
| Asia + Oceania |
|
|
|
|
|
|
| Central America + Caribbean |
|
|
|
|
|
|
| South America |
|
|
|
|
|
|
| TOTAL |
|
|
|
|
|
|