HEALTH/NUTRITION/POPULATION (HNP) AND POVERTY SEMINAR REPORT
HEALTH/NUTRITION/POPULATION (HNP) AND POVERTY
Topic: Determining Whether HNP Projects Reach the Poor: What Can
We Learn from the Microcredit Movement?
Presenter: Manohar Sharma
Commentators: Adam Wagstaff and Abdo Yazbeck
Date and Time: February 25, 2002, 12:30 – 2:00 PM
The session was the thirty-second in the HNP/Poverty Thematic Group’s seminar
program. Participants included approximately 30 staff members of the World Bank and
other Washington area development agencies. Thematic Group Co-Coordinator Dave
Gwatkin served as moderator.
The purpose of the seminar was to introduce the hnp community to a technique
used by microcredit institutions to determine whether their loans are reaching
disadvantaged population groups, and to assess the applicability of the approach to
assessing how effectively hnp activities are reaching the poor. The presenter, Manohar
Sharma, is a Research Fellow at the International Food Policy Research Institute (IFRPI)
in Washington. He was coordinator of the IFRPI team that developed the technique for
Consultative Group to Assist the Poorest (CGAP), a support body for microcredit
institutions that is located at the World Bank. After Manohar’s presentation of the
technique itself, Adam Wagstaff and Abdo Yazbeck of the World Bank staff opened a
discussion of the technique’s relevance for the hnp sector.
Manohar started by explaining the reason for developing the technique, known as
the “CGAP Poverty Assessment Tool.” He explained that microcredit institutions are
committed to reaching two, sometimes conflicting objectives: attaining financial
sustainability, and reaching the poor. Progress toward the first of these objectives can be
and is routinely monitored through the information readily available from the institutions’
financial records. But these records do not contain the information needed to determine
the socioeconomic status of loan recipients relative to that of the population at large. As
a result, the records are inadequate to assess whether the loans are going to poor people,
and a special effort is required to measure progress toward the second objective.
The CGAP poverty assessment tool is a plan for such an effort, in the form of
detailed guidelines for the design, implementation, and analysis of data from a simple
household survey in order to compare the socio-economic status of households that
receive and do not receive loans from microcredit institutions. It is intended to be easy
enough to for non-specialists to use, applicable to a wide range of cultures, inexpensive
(under $10,000 per survey), and rapid (such that results can be available within three
months of a survey’s initiation).
The tool features a questionnaire designed to determine household socioeconomic
status. After considerable thought and experimentation, the tool developers settled upon
a set of 30-35 questions covering five dimensions of socio-economic well-being: human
resources (e.g. education level and type of employment); dwelling characteristics (e.g.
type of roof and/or walls, presence of electricity), household assets or possessions (e.g.
amount of irrigated land owned, presence of possessions like radios and/or fans),
household food security situation (e.g. number of meals per day, consumption of luxury
or inferior foods), and additional dimensions specific to individual countries (e.g.
remittances received from overseas household members). Each country questionnaire
typically employs some 15-20 of these questions, selected on the basis of information
gathered through local focus group surveys and similar methods.
The questionnaire is used in to gather information from 500 randomly-selected
households: 200 households that receive loans from the microcredit institution being
studied, and 300 households in the institution’s service area that do not receive loans.
Once gathered, responses to the different questions are weighted, using weights produced
by applying the statistical technique of principal components analysis to the data set, and
added together in order to produce a wealth score for each household.
The non-client households are then ranked from lowest to highest in terms of the
wealth scores, and are divided into three groups of 100 households each: low, middle,
and high. The wealth scores of the highest and lowest household in each of the three
groups are taken to establish the range for that group; and the proportion of client
households whose wealth scores fall within the range of each non-client group is
calculated. The result is a socioeconomic profile of client households relative to nonclient
households that can be used to assess how well the microcredit institution being
studied is reaching the poor. (For example, a microcredit institution with 50% of its
clients in the lowest population group might be said to reach the poor more effectively
than one with only 25% of its clients in that group.)
After thus describing the poverty assessment tool, Manohar presented some
illustrative results drawn from the several instances where the tool has been applied.
These showed very wide differences in performance. For instance:
— 58% of loans provided through a microcredit program Andhra Pradesh,
India named SHARE went to households in the poorest group, much higher that the
comparable figure for loans made available through ACODEP, one of Nicaragua’s largest
— In South Africa, households in the poorest group received 52% of loans
from two microcredit programs that focused especially on the poor. Only 15% of loans
issued by a third program that did not focus on the poor went to this group.
Adam Wagstaff pointed to the basic similarity between the methods used to assess
household socioeconomic status in the poverty assessment tool in the World Bank
hnp/poverty country reports. However, he noted that there were also significant
differences in the particular questions selected for inclusion in the questionnaires and in
the resulting wealth score, and proceeded from this observation to reflect on the criteria
for selecting questions. Although unresolved conceptual questions about what constitutes
wealth prevents any definitive response to what and should not be included, it is still
possible to suggest at least some guidelines. For instance, one would want to add only
those questions that provide additional information – that is, that have been demonstrated
through pilot inquiries to increase the power of the wealth scores to discriminate between
the poor and the better off. Another consideration would be way in which the wealth
scores would be used: for instance, one would wish to avoid introducing circular
reasoning by including educational status in a wealth score used in a study of how
educational status rises with the level of wealth, whereas in studies designed to measure
other things (say, poor-rich differences in access to water or sanitation), inclusion of
education in a wealth score might be appropriate or at least less inappropriate.
Adam concluded by advocating the regular use of some variation of the poverty
assessment tool in monitoring who benefits from Bank-supported hnp projects. This
could be accomplished by earmarking a modest amount of resources for this purpose in
the budget of each Bank loan. The resources could be used to train and empower people
in the countries concerned to carry out such assessments and apply the results to develop
Abdo Yazbeck praised the poverty assessment tool developers for producing an
approach of clear practical value, which he commended as model for other researchers.
However, he foresaw significant challenges in applying the tool directly to hnp projects
in its present form, because hnp projects lack a readily identifiable client population
comparable to the group of borrowers with whom microcredit institutions retain regular
contact. So at least some modification is likely to be necessary. As an illustration a
modified version of the CGAP approach that might work for hnp projects, Abdo
suggested using a wealth questionnaire for a survey of patients at facilities in a country
where a nationwide household survey including the same wealth questions had recently
been undertaken. This would permit comparison of the socioeconomic status among
facility patients with that of the population at large as measured by the nationwide
survey, enabling one to tell whether the facility had succeeded in reaching the poor. This
approach has been successfully used in India.
Abdo argued against any mandatory earmarking of resources in Bank loans for
this kind of assessment. Instead, he advocated that the Bank actively lead in developing
and disseminating information about equity monitoring methods, but leave it to countries
to take the initiative in deciding to apply them.
The ensuring floor discussion covered several issues. Among them were:
— Choice of Questions. A participant asked about the basis for selecting
the particular questions included in the questionnaire over the many alternatives that
might also have been chosen. For example, why were there no questions about health
status in the section on human resources? Manohar replied that the principal reasons for
not including particular questions included political sensitivity, difficulty of
understand ing on the part of respondents, and the production of inconsistent or
implausible results. The participant who had originally raised the question did not find
these sufficient reasons for omitting health questions, despite the well-known doubts
about the accuracy of self-reported health status. She referred to a five-question
instrument developed by the World Health Organization, which made it possible to
surmount this difficulty.
— Reasons for Reaching or not Reaching the Poor. Another participant
noted that the CGAP tool appeared to provide solid data about whether microcredit
institutions were reaching the poor, but produced no information about why the
institutions were or not doing so. This “why” information, he argued, is even more
important than the “whether” data for developing pro-poor approaches. In response, a
representative from CGAP noted that, despite the lack of formal inclusion of “why”
considerations in the tool’s design, it was nonetheless possible to get a pretty clear idea of
why some of the institutions have reached the poor than others. In general, those
institutions shown to have done best in this regard were those known for their
commitment to poverty alleviation and for the use of targeting mechanisms to this end.
The institutions that had done poorly were those which had given higher priority to
— Use of Wealth Questions for Targeting Purposes. Two or three
participants wondered whether the same questions appearing in the CGAP tool for
assessing who had received loans after the loans had been made could also be used in
advance for determining who should receive loans. That is, could one administer the
same questionnaire to prospective borrowers, and lend only to those whose wealth score
fell below some predetermined point? Although time limitations prevented a full
discussion of this question, a couple of comments suggested a preference for using some
simpler system for identifying poor individuals; or perhaps better yet, an indirect
approach that would focus on serving everyone in selected poor geographic areas rather
than seeking to find and serve only disadvantaged individuals.
* * * * * * * * * * * * * *
Further information about the CGAP Poverty Assessment Tool is available at: