Targeting Poor Communities: An Example from Côte d’Ivoire
Implementation teams of rural development projects have to make choices about village selection. Even projects with clear poverty alleviation objectives often lack clear poverty-based criteria for screening villages. In the absence of explicit poverty-related criteria, there may be a tendency to favor richer and less remote villages because they are better organized, easier to work with and more accessible. The poverty alleviation objective can become no more than empty rhetoric in practice - something that happens all too often.
While participatory approaches are useful for poverty ranking within villages, they may be less useful for making wealth comparisons across a large number of villages. For making large-scale comparisons, judicious use of quantitative approaches can also complement participatory approaches to enable development projects to more effectively reach poor people.
The Project Zone
The IFAD-supported Rural Development Project in the Zanzan Region of Cote d'lvoire was designed in 1998. With a population of about 600,000 and over 1,000 villages spread over a large area, the region is comprised of three administrative departments - Bondoukou, Bouna, and Tanda - which are very heterogeneous in terms of population density, economic activity and potential, and income levels. The northernmost department, Bouna, is in the Savannah zone, while Bondoukou and Tanda are transition zones between savannah and forest. Zanzan is among the poorest regions in the country, although agricultural potential does exist and much of the region has strong, but informal, commercial agricultural links with urban areas.
Rural social and physical infrastructure investment in Zanzan has been minimal relative to other regions, seriously hampering agricultural development.
Study Objectives and Methodology
In early 1999, with IFAD support, two economists from the International Food Policy Research Institute (IFPRI) trained a nationally-recruited team to launch a survey in order to more effectively target interventions
to the rural poor in the project zone through development of an initial screening mechanism for village choice.
An additional objective was to test the specific method for reliability, practicality, cost-effectiveness, and clarity for non-economists.
Steps in the Approach
1. Calculate a community-level index of poverty indicators using a statistical model
2. Design a community questionnaire to survey the indicators
3. Design a household expenditure questionnaire to validate the poverty indicators
4. Carry out the community survey in all potential project villages and the household survey in a small number of villages and households
5. Compute and map village scores
Step 1 is to select proxy indicators for poverty using pre-existing survey data. One identifies a limited number of easily observable community-level variables that strongly correlate with income poverty by estimating a regression equation to weight the respective coefficients to arrive at a village-level score.
Cote d'lvoire has a particularly rich set of data on poverty, having been one of the first countries to participate in the World Bank's Living Standards Measurements Surveys (LSMS). Nation-wide surveys were conducted in 1986, 1987, and 1988. In rural areas, data were collected both at household and community levels.
Per capita annual household expenditure was used as the basic measure of welfare. All variables in the LSMS community questionnaire were examined to determine whether or not they were associated with household expenditure levels. The variables that resulted in the strongest statistical model included presence of nuclear and satellite settlements, length of time the village was cut off during the rainy season, distance to a post office, the portion of village girls attending school, and portion of births in clinics.
Proxy variables are easily observable substitutes for variables that are more time-consuming and costly to directly identify. The most direct measure of income poverty is expenditure level, but this is prohibitively costly to measure on a large-scale. Carefully selected community-level variables can serve as indirect measures of poverty if one is reasonably confident that they approximate' the true situation.
Step 2 is to design a community-level survey questionnaire with questions related to the proxy indicators as well as to other community-level information of potential practical value for project implementation. Care was taken to keep the questionnaire short, but to also gather additional information of practical value for project implementation. There were a total of 18 questions on the following topics: geographical background of the village and access problems; presence of community health and education infrastructure and service providers; main types of housing; sources of potable water supply; and presence of development projects and existence of village and sub-village associations.
Step 3 involves also designing a detailed household-level expenditure survey to be carried out in a limited number of villages for purposes of validating the relevance of the proxy indicators to poverty in the project zone. Without this step, it is dangerous to assume that the proxies are valid indirect measures of income poverty in the project zone.
Step 4 involves carrying out the community-level survey in all villages in the project zone, or in all villages with population greater than a pre-determined cut-off point. At the same time, the household expenditure survey is also implemented in a limited number of villages to double-check that the variables derived from the national survey are valid in the project zone.
The IFPRI experts were in-country for 10 days, during which they trained a local team composed of an economist, a statistician (who also served as field supervisor) and 8 enumerators from the region in the technique. The team field tested and finalized the questionnaire, developed data entry and synthesis procedures and carried out the household-level expenditure survey.
The actual time required to conduct village interviews was 10-20 minutes. However, village protocol required a longer stay of as much as two hours to be properly introduced to the village chief and dignitaries, accept hospitality (at a minimum, a drink of water, soda, or palm wine, but sometimes reception of chickens or yams) and answer questions from the villagers about the new project. The most time-consuming part of the exercise was reaching the villages (including fair amounts of time getting lost) rather than completing the questionnaire. In retrospect, the opportunity cost of including a richer set of community-level questions would not have been very high (in terms of data collection and entry, not necessarily in terms of analysis later).
The household-level expenditure survey (for purposes of validation) was carried out in 2-6 villages per department, with 1-2 villages each considered rich, median or poor (as determined by the community survey). In each village, 30-50 households were randomly interviewed. Results of the household survey confirmed the validity of the community survey as there was a good correlation of income poverty as measured at household-level and community ranking.
The survey covered 17 districts and 1,073 villages. Initially, the team intended to survey only villages with more than 200 inhabitants. However, this idea was discarded because of the small average size of villages in Bouna (only about 130). The decision was therefore taken to visit all villages in the project zone.
Step 5 is to compute and map village scores. Results for individual indicators are also useful to analyze and map. Using the statistical index, an example of how the scoring was calculated for an individual village is shown in the following table.
Example of Village Score Calculation
Number of points example
|Regression constant||Same for all villages||è 164|
|Are these satellite settlements (campements) attached to the main village?||If yes, -10.24 points; if no, 0 points||Yes è -10.24|
|In general, how many months per year is the road cut?||Response multiplied by -7.8 points||1 month è 1 x-7.8|
|How far is the village from the post office or telephone?||If located in the village, 37.45 points; if not, 0 points||10 km è 0|
|About what percent of school-age girls
1= nearly all
2= more than half, but not all
4= less than half
5= Just a few
|Response multiplied by -10.15 points||2, more than ½ è 2 x -10.15|
|Where do the majority of women give birth?
1= at home
2= at a clinic (maternité)
3= in a hospital
|If response is 2 or 3, 30.32 points; if 1 or 4, 0 points||1, at home è0|
Result è164 + -10.24 + (1x-7.8) + 0 + (2x10.15) + 0 = 125.66 points
Results of the Village Scoring Exercise
For ease of presentation, villages were divided into five categories with scores of 50 point intervals. A solid majority of villages (about 60%) were in either category one or two (the poorest categories), confirming the general impression of the Zanzan region as having very poor access to infrastructure and services. Yet results are highly variable between departments. By far, Bouna is the least blessed with nearly 80% of villages in the two lowest categories. In contrast, Tanda has only about one-fifth of villages in category 2 and none in the lowest category. Almost half of the Bondoukou villages are in the second category.
While Tanda is clearly better off, the number of less well-off villages is not negligible: about one-fifth of its villages are in the second category. Almost half of Bondoukou's villages are in this category. This points to the potential usefulness of the approach for identifying pockets of poverty in otherwise better-off zones.
Generally "poor zones" are often assumed to be uniformly poor, thus discounting the need for targeting within those zones. However, the survey team found that variability of village scores (as measured by the coefficients of variation) was significantly greater in Bouna than in the other departments. Twenty percent of villages were in the bottom category and could be classified as "poorest of the poor" while 58% of villages were in the second category. If the goal of a development project is truly to reach the poorest of the poor, this approach can also be of assistance in not only poor villages but also the poorest villages.
Beyond village rankings, it is also possible to provide a rich level of reporting on individual variables for each zone such as access to health, education and communication facilities, transport and water problems, extent of village organization, and involvement with on-going development projects.
An earlier IFAD project design team tried to use minimum distance from a paved road as a major decision rule for village selection, specifying that at least 75% of the villages selected for interventions should be situated more than 5 km from a paved road. This was partly due to the tendency of projects to concentrate activities in villages with easy access, and partly due to analysis from other countries demonstrating links between infrastructure access and rural poverty. Yet these nuances were lost in the debate that ensued. Officials considered it arbitrary and not reflecting local reality. The idea was dropped, and subsequently, the IFAD country portfolio manager was often jokingly referred to as "Mister Five Kilometers."
Potential practical uses include identifying poverty "pockets" and the poorest communities. It can be a powerful supplement (or pre-cursor) to participatory diagnostic and planning approaches. For investments at levels higher than individual villages (district and sub-district) like roads, a mapping of villages by their scores and populations can enable decision-makers to prioritize roads for rehabilitation that reach the maximum number of poor people. The approach can also be used for monitoring and evaluating the equity impact of project interventions.
In the specific context of Cote d'lvoire, the approach appeared to be politically acceptable. An array of indicators was seen as consistent with common-sense notions of poverty. In addition, while variables were aggregated to derive a village score, the individual variables were generally consistent with common-sense notions of poverty. It also mattered very much to ministry technicians that practical uses were obvious and that results were generated quickly.
Strengths and Limitations
Only for initial screening across villages; not a substitute for village-level participatory diagnosis and planning
Income-based, but poverty has many dimensions
Quantitative approaches may be sensitive to
choice of variables and their weighting
Reliable household expenditure and community survey data must already exist
May miss significant numbers of the poor if wealth disparities are greatest within village
Cost-effective in time and money for large-scale exercises (4-5 months duration and 0.5% of project costs)
Appears valid for making poverty comparisons
Can supplement or precede participatory
diagnostic and planning
By making poverty criteria explicit in village selection, helps avoid natural tendency of implementers to work in "easier" villages
Can be use to identify "pockets" of poor villages and "poorest of the poor" villages
As this is a new approach, it is worth considering different options for improving upon it. Could indicators be derived in more participatory ways? Using participatory approaches, villagers in the project area could be surveyed about what they considered to be easily observable characteristics of poverty at community level. If their perceptions were fairly uniform, or varied in ways that could be easily stratified and adapted by zone or ethnic group, questionnaires and indices could be designed using locally-derived variables. This could potentially be more locally reliable, save time and be less demanding in technical expertise. The survey data could also be entered into a geographic information system (GIS). Additional data (including results of participatory exercises) could also be incorporated to enhance project planning. Whatever the technique chosen, one thing is clear: there is a need to introduce more rigor into village selection in self-proclaimed rural poverty alleviation projects.
Africa I Division