|
|
|
|
 |
 |
  |
|||||||||||||||||||||||||||||||||
 |
|
||||||||||||||||||||||||||||||||||
Tool 2: Wellbeing monitoring step-by-step Step 2: Develop local poverty indicators and formulate survey questions This section describes how to develop locally specific indicators for measuring poverty and wellbeing. These indicators are then used to formulate the survey questions. The approach described here uses the Nested Spheres of Poverty (NESP) model that was introduced in Part I. A review of that section might be helpful before continuing. A. Define poverty and wellbeing locally Organise several focus group discussions (see Tool 3) to create local definitions of poverty and wellbeing. Focus groups can include members of a single community, or members of subgroups within a community, like men, women, youth or ethnic groups; the decision on focus group composition depends on the locally relevant criteria related to livelihoods and wellbeing. Focus groups could consist of only local government representatives, or these could be grouped together with development agencies and NGOs active in the region. The understanding of poverty tends to be different between groups, such as women and men (Box 18).
Similarly, government officials often have different views on poverty to community members. During the focus group sessions, instruct the participants to identify the most important aspects of both poor and well-off households. If focus groups are not feasible, interview several key informants who are representative of the community. Once the monitoring team has a full list of aspects of poverty and wellbeing, it should organise them using the following spheres of the NESP model:
If a sphere is not well represented, the facilitators may ask some probing questions to determine whether some aspects were overlooked, or if informants really did not find the sphere very relevant for explaining wellbeing. However, care must be taken to avoid influencing or guiding the informants. B. Prepare a list of possible indicators Using the gathered information, the monitoring team should prepare a long list of possible indicators. As some indicators might be generally applicable, it is also worthwhile checking lists with poverty indicators from other sources, such as:
If some indicators are already measured through existing monitoring programmes, avoid duplication of work and make best use of this data by coordinating with these programmes! Next, check the indicators to verify whether they comply with the SMART criteria (Box 19).
Be aware that poverty indicators only work as long as they can be linked to an underlying poverty cause or condition. When that causal link changes, the indicator may no longer be relevant and may have to be replaced. For instance, if access to schooling ceases to be a constraint for any family, then a different indicator is needed to measure education. As a rule of thumb, poverty indicators should be reviewed at least every 5 years. However, consider changes to indicators with care—changing the indicators can make it difficult to compare over time and monitor change. Note changes to monitoring methods immediately, as people are likely to forget as time passes. The long list of indicators tested in our project can be found online at the project website (http://www.cifor.cgiar.org/povertyindicators). C. Quantify indicators and formulate questions Discuss how the indicators can be weighted or quantified. Quantifying makes it possible to compare poverty data within or across communities. For quantifying an indicator, turn it into a monitoring question and find two or three answers that cover the range from ‘good’ (3 points) through ‘intermediate’ (2 points) to ‘critical’ (1 point). The questions and answers should be simple, clear and unambiguous. Local language should be used where people have problems in understanding the national language. Note that each question should be a closed question (see Box 20).
A closed question has a limited number of answers, and the respondent must pick one. The use of closed questions is best because the comparison of answers is easier. Preparing good questionnaires is an art that requires a lot of experience. Some questions might be sensitive or generate biased answers. For instance, people may avoid answering a question about their annual income, or may be unable to do so accurately, but would have less of a problem with specific questions about recent expenditure figures and living costs. In order to avoid such biases, local governments should seek external assistance from experienced social scientists, the government statistical service, universities or NGOs. D. Shorten the list of indicators Having more indicators allows for a wider range of information. However, too many indicators mean long, exhausting interviews and more complicated analyses. Three indicators per wellbeing sphere is ideal. With nine spheres, the total number of indicators would be 27. This would mean 27 questions, which can reasonably be asked in an interview of approximately 30 minutes. Use a field test for shortening the original long list, prepared under step 2B. Choose 5–10 communities of various sizes, ethnic composition and location. Prepare a questionnaire with all the indicators and test them in the communities, using standard survey techniques. Next, group the results by wellbeing sphere. For instance, combine data from all questions related to health, data related to all questions on knowledge and so on. Add up all the combined scores by sphere (health is one sphere, knowledge another sphere, see Part I) into one figure, until there are nine figures that correspond to the nine spheres of the NESP model. Next, prepare subtotals by adding up data of different combinations of three questions related to a single sphere as you finally need only three indicators per sphere. Test the subtotals of the different combinations of questions in a correlation test (e.g. Spearman’s rank correlation test) against the total value for all households. Choose those subtotals that show the highest correlation with the full set of each sphere (ideally the correlations coefficient, r, should be greater than 0.8). High correlation shows that the subsets represent the full set, because some indicators are correlated with others. For instance, a household that has a satellite antenna almost certainly also has a TV and access to electricity. Thus, ‘having a satellite antenna’ might be an indicator that actually represents all three test indicators.
© 2007 Center for International Forestry Research (CIFOR) |
|
||||||||||||||||||||||||||||||||||
