Poverty Measurement Discussion Group, Paper #2

Welcome to the second edition of the Microcredit Summit Campaign's Poverty Measurement Discussion Group. This paper will be made available by E-mail, on our website (www.microcreditsummit.org/papers/papers.htm) and by post to those that request it. As of January 14, 1998, the Poverty Measurement Discussion group had 505 participants.

Background and Purpose

The Microcredit Summit Declaration and Plan of Action is very specific about which clients of microcredit programs "count" toward the goal of reaching 100 million of the world's poorest families. In the Declaration and Plan of Action the "poorest" clients are defined as those living in developing countries who were in the bottom half of the population living below their country's poverty line when they took their first loan. In industrialized countries, the Declaration and Plan of Action defines the Summit's target group as all those living below their country's poverty line when they took their first loan.

The specificity of the Summit's goal provides an important clarity of purpose and a daunting challenge. How are we to know how many clients were very poor when they took their first loan?

The Poverty Measurement Discussion Group arose out of this challenge and the need to monitor progress toward the fulfillment of the Summit's goal. Its purpose is to discuss and disseminate the best practices being used in the field for measuring and monitoring poverty levels of clients in a cost-effective manner. We acknowledge that discussion thus far focuses on finding the poorest families in developing countries. We welcome discussion on methods of targeting poor families in industrialized countries.

 Structure of Discussion Group

To launch this discussion, we circulated two pieces in the Poverty Measurement Discussion Group Paper #1. These pieces were excerpted from the "Targeting the Poorest and Covering Costs" Meet the Challenge Session at the Microcredit Summit. These excerpts, by Professor David Gibbons of CASHPOR (Malaysia) and Mr. John de Wit of the Small Enterprise Foundation (South Africa), discuss the benefits of using the Housing Index and Participatory Rural Assessment Wealth Ranking (PRA Wealth Ranking) and the limitation of using small loan size in targeting the poorest families in a community.

In the pages that follow, you will find:

1. Very brief quotes from the Poverty Measurement Discussion Group Paper #1 "Targeting the Poorest and Covering Costs," describing the three poverty measurement strategies mentioned in the article: the Housing Index, Participatory Rural Assessment Wealth Ranking; and the use of small loan sizes.

    

2. Excerpts from comments on the Poverty Discussion Group Paper #1 that were received from discussion group participants between 14 October 1997 and 1 December 1997. These comments are grouped according to which poverty measurement methodology they address. Following each set of comments you may find responses from Gibbons and de Wit. We'd like to thank Professor Gibbons and Mr. de Wit for the time they spent reviewing a draft version of this paper and then offering their thoughts.

    

3. Suggestions from the moderator for future discussion.

Excerpts from "Targeting the Poorest and Covering Costs," from Poverty Discussion Group Paper #1 (originally excerpted from a Microcredit Summit Meet the Challenge Session, February 1997)

The Housing Index

David Gibbons, CASHPOR, Malaysia
[The Housing Index is] an observational method producing an indicator which we have found is highly related to poverty. So there are three dimensions of the house, and we can look at it from the roadside. We don't have to conduct any interviews. We just go up and down the lanes in the village and map the houses which appear to be qualified. We look at the (1) size, we look at the (2) physical condition or building materials, and we look at the (3) material of the roof.

The material of the roof is very interesting because that turns out to be a simple but powerful way of identifying the very poor, as distinct from the poor, in most countries of Asia. I'm talking about thatched roofs, roofs made out of woven bamboo, roofs made out of twigs, roofs made out of plastic sheeting. These are temporary roofing materials, which always have holes in it, always leak, always create a health problem for the household; nobody wants to live under a temporary roof unless they have to. So the people living under these temporary roofs are nearly always the very poor. Now if you combine that with small size of house and very simple building materials mud, jute sticks, things like that then you are very close to identifying most of the very poor....

Participatory Rural Assessment

David Gibbons, CASHPOR, Malaysia
[W]e bring the whole village together to find out who are the very poor, who are the not so poor, who are the non-poor, through participatory methods.... Interestingly we find when we compare the cost- effectiveness of these two methodologies, the house index versus the PRA Wealth Ranking, they come out almost exactly the same. It takes about five minutes for an experienced field assistant to use the house index properly, and it takes about five minutes per household to use PRA Wealth Ranking properly, if you break it down.

Use of Small Loan Sizes to Target Very Poor Clients

John de Wit, SEF, South Africa
So when asked to answer the question, "If you want to reach the poor, can you simply offer a small loan size?" what we find is that it didn't work in our case... Well, the people who are better off joined the program because there's no other access to credit there except from loan sharks. They're also desperate for credit, and they have very legitimate needs. But they're coming and taking small, small loans, inappropriate for their own needs, in the hopes that one day you will give them a bigger loan... Now the clients that are very much poorer, why don't they come to you in large numbers? And the reason is, they're intimidated by the wealthier clients....

Commentary on Poverty Discussion Group Paper #1

Thank you to those who submitted the 56 comments that we received. The comments below are excerpted from those received between 14 October 1997 and 1 December 1997. Comments received after 1 December are being reviewed and will be incorporated into Poverty Discussion Group Paper #3, to be distributed in May 1998.

Comments on the Housing Index, received 14 October 1997 to 1 December 1997

The following two comments point out the limitations of the housing index in urban slums.

• Dr. Jacob Kurien, Dept. of Economics, Loyola College, India
  The housing index could be deceptive, more so in the slums of metropolitan cities and towns. High costs of living coupled with prohibitive rents compel people to live in dilapidated homes. In Indian cities of Bombay, Calcutta, Delhi and Madras it is quite common to find television sets and other electronic gadgets in the homes of slum dwellers.

   

• Patricia Fuertes, DESCO, Peru
  ... I'm afraid a house index alone would be misleading in different contexts. For instance, in Peru, there are places like Lima, in which the weather conditions allow the poor and the poorest people to live in quite similar precarious building conditions. Since it never rains, and there is no important incidence of winds, one can find that their houses show very similar deficient roof building materials, but within some of those houses, televisions, radios and the like can also be found and these cases are not the exception to the rule but are quite common. So we cannot tell by the housing condition alone that we are dealing with the poorest....

   

• Ramesh Bellamkonda, India
  In the area of the first field office of BSS, around Hebbur, Tumkur District, about 100 km from Banglaore City in Southern India, our first walk-through survey of a nearby village revealed the drawback with housing index. An employee of BSS program who was with me thought that the good houses were due to a government housing program and therefore not indicative of their resident's wealth.

   

• Ed Tongson, Haribon Foundation, Philippines
  The housing index is indeed a simple cost-effective indicator targeting the poorest of the poor. Other measures we use in coastal communities is the presence of a latrine. Some houses do not have a latrine which indicates the inability of households to provide for sanitation. Children in these areas are affected by gastro-intestinal diseases because of the lack of sanitary facilities, including potable drinking water. If they have a well, it may be polluted from coliform bacteria coming from household wastes. These houses sometimes have a latrine as a seperate cubicle from their household, and is easy to detect without asking the house occupant. Those who don't have a latrine are very poor, and rely on community latrines provided by government.

Comments on the Housing Index from Gibbons and de Wit, received January 1998

In the following section, Professor Gibbons shares the results of a CASHPOR Workshop in China testing the effectiveness of a Housing Index that had been adapted to local conditions.

• A CASHPOR Workshop in Yixean County, Hebei Province, in China tested whether a House Index adapted to local conditions could predict the poverty status of households, as measured by per capita income. All but one of the households [that were] means-tested, confirmed the poverty rating based on the House Index.

  This validation of the House Index in this experiment although on a small sample in only one area is important. Identifying the target poor through the House Index is from six to ten times cheaper than trying to establish the income and assets of a family through interviews. It also requires far less training for field staff.

  This is how the workshop participants in Yixean did it: In field work, they adapted the House Index to an area where severe winters make solid walls and cement and thatch roofs essential, even for the poorest. They deleted "condition of the house" [on House Index check-list] and replaced it with "material of the compound wall." It is customary in this part of China to surround the house compound with a high wall of the strongest materials one can afford. The non-poor had high, brick walls and the poor, lower walls of mud. "Material of the roof" was replaced by "material of the walls." The rich had brick walls and the poor, only mud.

  This adapted House Index was then applied, with 20 households being indexed by each participant. [The participants] had been divided in three groups. In all, 60 houses were indexed. Although the indexers had little training, there was only very small variation in the scores within each group. This indicated that the Index was a reliable tool for targeting, since when used by different fieldstaff it tends to produce the same overall results.

  The Index scores of the 60 houses were used to divide them into "poor" (six points or less), "border-line" (seven points), and "non-poor" (eight points and more). From these, 18 houses were selected randomly; 12 from the "poor" category and 6 from the "non-poor" category.

  Income data was collected from the 18 houses by interviewers who did not know the scores of these houses on the House Index. The results, as defined by the official per capita income poverty line, were matched with the results obtained through the House Index.

  On the income test, seventeen of the eighteen households fell into the "poor" or "non-poor" catagories as predicted by the House Index. Only one household living in a "poor" house as categorized by the Index, turned out to be "not poor" by the income measure. The workshop concluded that the House Index, adapted to the local environment, is a valid and cost-effective indicator of household poverty status, except for a minority of borderline cases.

  What is becoming clear about the House Index is that it can be a valid and reliable tool for targeting a substantial proportion of poor households, but that in order for this to happen in Asian contexts the Index must be adapted to the locality. This is well worth doing as the saving on the cost of targeting can be as much as six to ten times.

   

• Initially SEF used a housing index to identify the very poor, but we have seen some problems with this methodology in our context and so have begun to replace the housing index method with PRA Wealth Ranking ... In the experience of SEF working in rural areas of South Africa, where housing is one of the key indicators of poverty and wealth, a system based only on external housing conditions has proved to be unacceptably inaccurate [in targeting very poor households]....

Comments on Participatory Rural Assessment, received 14 October 1997 to 1 December 1997

The following two comments point out some potential weaknesses in PRA Wealth Ranking as an accurate poverty measurement tool.

• SAFWCO, Pakistan [Regarding] PRA ... listing of major assets by representatives of households and reading it out before villagers and asking them to classify: There are chances of wrong listings [of household assets] and lobbying by villagers which may cause dispute among the villagers.

   

• Dr. Jacob Kurien, Dept. of Economics, Loyola College, India
  ... The greatest hurdle [to PRA Wealth Ranking] is in collating reliable information on the size and market value of physical assets and livestock. Besides no proper legal document or records are maintained with respect to ownership rights. One has therefore to rely on oral evidences, which is inevitably biased and subjective.

   

• Richard Smiley, USA
  In some of the African villages I've seen, comparing houses via housing index would certainly be effective in targeting the poorest. But in other villages, this would not be effective because most of the families have houses with similar shapes, sizes and construction.... With this in mind, it seems that it would be preferable to promote the PRA method as a cost-effective strategy for targeting the poor. It certainly has the built-in benefit of having the input of the community members in determining what qualifies as being poor within their community. One danger with this method (as described in the excerpt) would seem to be that it would be possible for households to get classified as being poor which are actually not in such a bad situation.... Whether these families would be willing to reveal all their assets is open to debate...

  It should be mentioned that both these indicators are relative poverty measures. A household the PRA Wealth Ranking method identifies as being poor is poor in comparison to other households in that village. The household may actually be doing well in comparison to households in the next village but this method would not account for that....

  This suggests that there really is a two-stage approach to poverty identification taking place: 1) Identify the poorest areas on a macro basis (say by using health indicators such as the under-five mortality rate or economic indicators such as income); 2) Go into the areas identified and use the relative poverty measures described above to identify the poorest within these areas. It is important we focus on developing sound policy for step 1 as well as step 2.

Comments on Participatory Rural Assessment from de Wit and Gibbons, received January 1998

Mr. de Wit outlines the Small Enterprise Foundation's PRA Wealth Ranking methodology

1. No attempt is made to gain information about the actual wealth of people. What is measured is relative wealth. Thus there is no need to ask households to list their assets and values. We have found that people are often sensitive about giving actual details of their wealth particularly in public fora. We work from people's own perceptions of poverty and wealth. An initial discussion talks about concepts and the characteristics of poor people, and what differentiates people for example the very poor from the poor. We also list households onto individual cards and then ask people to place them into piles according to which wealth category they fall, poorest, slightly better off, etc.

    

2. We do not "bring the whole village together". Most of the villages in which we work have over three thousand people; some have more than double that. It would be logistically impossible to involve all of the people. We therefore use a small number of small reference groups who inform us about the village. Initially, we contact village representative structures to discuss our involvement in the village and to set up logistics. Next we hold a participatory mapping exercise with 50 to 100 people where the whole village is mapped on the ground and every household is listed; a range of people from the village will be involved in this. Then, we set up several reference groups of 3-6 people from each section of the village. It is these reference groups who do the ranking.

    

3. Cross-checking is vital to the process. The opinion of any one reference group can be biased. We use a minimum of three reference groups per village section. Consistency between the results is usually very high and this ensures that the results give an accurate perception of the wealth status in the village based on people's own perception. It is highly unlikely that three consistent results can be subject to bias. Poor correlation between results reflects poor ranking procedure or manipulation of the results by participants, and these are discarded and repeated. The small number of inconsistent results are checked by visiting those households individually if they apply for a loan.

    

4. Related to cross-checking is the level of detail required. The use of three piles [very poor, poor, non-poor] does not give results which are sufficiently detailed for accurate ranking of the village.... SEF allows participants to sort the cards into as many groups as they want to. Good facilitation asking often whether two cards belong in the same group, or if there is a difference usually leads to many piles forming (usually 5 or 6, but sometimes up to 10). A system is used to score each pile and average scores are calculated. Values are assigned to differences between the scores for a single household to determine whether the results for different reference groups are considered consistent or not.

    

• In my experience PRA [Wealth Ranking] is a reliable procedure for listing and valuing household assets. Everyone knows the major assets of everyone else in a typical Asian village. When someone does not declare an asset, someone else always says "What about your cow, etc.?" Everyone knows what was paid also.

  The major problem with PRA [Wealth Ranking] is that it tends to be inclusive rather than exclusive as a targeting tool. As everyone at the meeting knows that there may be some benefit from being included among the poor, participants will tend to include not-so-poor relatives and friends and the latter will press for inclusion. To some extent this can be controlled for by establishing asset cut-offs or creating more categories.

  For example, instead of the common three categories very poor, poor and non-poor which usually results in very few non-poor, five can be used: poorest, very poor, poor, not-so-poor and not poor. If done carefully this should result in a better spread.

  Yes, PRA [Wealth Ranking] tends to measure relative poverty within the village; but it can be adapted to identify the absolute poor, e.g., those usually food deficient, etc. In the Housing Index a leaky, temporary roof is getting pretty absolute.

Comments on using location as a proxy for poverty-level, received 14 October 1997 to 1 December 1997

The following comment describes the use of regional socio-economic indicators to identify the poorest population in a country.

• Ana Luisa Estrada Galarza, FIRA Banco de Mexico, Mexico
  In my institution, the poorest groups of rural producers have been identified as those who come from areas characterized by a high level of marginality, a scarcity of public services, a lack of sources of work and paid employment, low incomes, isolation and geographic dispersion and whose productive activities are for subsistence. In addition to this, they are characterized by other social poverty indicators such as high rates of illiteracy, infant mortality, malnutrition, and short life expectancies....

  As a first step, the Institutio Nacional de Estad¡stica, Geograf¡a e Inform tica [The National Institute of Statistics, Geography and Computer/Informatic Science], has created a document entitled "Niveles de Bienestar en Mexico" [Levels of Well Being in Mexico] where they catalogue different variables that touch upon this subject. Using this classification as a base, on the municipio level (the states of Mexico are divided into administrative units called "municipios"), it is feasible to determine which communities are the most underdeveloped. Perhaps after this point it is necessary to deal with more specific characteristics, but since the communities are so marginalized, you can assume that practically the whole population [in a certain municipio] belongs under the definition "the poorest" of Mexico.

   

Professor Gibbons offers a caution about relying exclusively on location as a proxy for targeting the poorest

• Identification of the poor by area alone nearly always results in heavy leakage to the non- poor, as every area tends to include non-poor households often the leaders. It can be a cost-effective first step, however. Followed by use of a Housing Index and/or PRA [Wealth Ranking], where necessary, it should result in a cost-effective package.

Mr. de Wit makes the distinction between the operational need for the area a program serves to be geographically compact and the strategic need to serve very poor clients

• In our context, rural South Africa, even though a village may lie in a magisterial area or region which, from a macro perspective is very poor, it would still be the norm to find a significant percentage of the population of that village who may be classified as being relatively well off. To simply offer services to any person in such an area would almost certainly result in those services being taken up by the "non-poor."

  People within the potential target population [i.e., women or very poor families] of a poverty-focussed microfinance institution may well live in disbursed geographical areas. However, the viability and impact of the organisation depend on achieving high numbers of clients in a small geographical area. It therefore makes good sense from these two perspectives to use available data (where available and reliable) to focus work in the areas with highest concentration of the target population, the very poor. SEF has adopted a similar strategy to FIRA [see above]. It is important to remember, however, that this method does not select the poor. It is an operational issue, to achieve operational efficiency, rather than a poverty-targeting issue.

   

The following comments discuss the use of client or household interviews and surveys to determine the poverty level of the family.

• SAFWCO, Pakistan
  ...Mostly the method used by our organization's social organizers to identify the poor is that they go in to the streets of the villages and meet the people at their homes, observe the house conditions and ask some related questions about their family members, sources of income, expenses, food consumption and watch their physical appearance and health. During this visit, members of the village organization who personally know the condition of the rural peoples are also accompanied with our social organizers. After the brief discussion with the village organizers, needy people are classified. This method is 90% accurate but it does take more time than the method described by Gibbons and de Wit. In fact, the housing index method suits the condition of our region. I think if we use this method ... it will be more successful not only in the case of identification of the poor but also in cost-effectiveness....

   

• H.S. Shylendra, Institute of Rural Management, India
  For credit-based programmes like Integrated Rural Development Programme (IRDP) implemented by government, the main criteria used is the official monetary poverty line (currently Rs. 11000 per annum) which is supplemented to some extent by caste as a factor. Households have to be surveyed and a list based on income to be prepared by local authorities. The households below poverty line are further classified on the basis of their income, like poor, very poor, very very poor and destitutes. Till recently the emphasis under IRDP has been to reach the poorest among the poor, i.e., those who are in the lower rungs. The households belonging to socially disadvantaged class like the scheduled castes (officially identified) are to be given special preference....

  IRDP is a major national level poverty alleviation programme. There were a lot of hopes on IRDP in the official circle of its role in poverty alleviation. IRDP failed to deliver the expected results. One of the major failures of the IRDP has been the large scale leakage of assistance to non-target groups.... There were many reasons as to why foolproof selection of poor could not be ensured. The surveys conducted to identify the poor for such programmes invariably reveal suppressed income levels of the households. Households, in expectation of the benefits like subsidy and cheap credit, would like to be enlisted as poor. Further, the surveys were conducted by untrained people leading to unscientific income estimation.

  However, two specifications had to a great extent prevented large scale leakage of IRDP assistance to non-target groups. They are (1) the specification to cover compulsorily a certain percentage of households belonging to scheduled castes and (2) [that] the list of below poverty line households [was] to be vetted by the body of village elders. While the former is a specification to ensure self-targetting as most of the schedule caste/tribe households invariably belong to the category of poor; the latter is to utilise the collective wisdom of the village in identifying the poor in a participatory way. The first specification has been applied all over the country and has ensured at least a part enlisting of the poorest; the second specification also has helped in selection of the really needy wherever it has been tried out.

Comments on using a House-to-House Interview of Potential Clients by Field Officer, from Gibbons and de Wit, received January 1998

Professor Gibbons argues the limitations of Household Interviews

• Household interviews are notoriously unreliable and expensive. They should be used only in border-line cases and then by well-trained staff. Increasingly there are socio-economic differences within the Scheduled Castes and Tribes in India. Going to such villages first is still cost-effective; but a Housing Index and/or PRA [Wealth Ranking] must be used within the village. Otherwise the upper half will get most of the benefit.

Mr. de Wit shares SEF's experience with the limitations of assessing the poverty level of each household

• SEF used to use a "poverty-test" method. This was based on a detailed list of proxy indicators of poverty, and involved assessing both the condition of the outside of a potential member's house as well as looking at the condition of the inside (ie., their assets). This was a very time-consuming and expensive process and was replaced by a housing-index test and more recently by the PRA Wealth Ranking method. There is a concern when trying to gather actual income and expenditure data rather than proxies, that there may be great distortion of the data collected. Income and expenditure surveys are notoriously inaccurate and susceptible to under or over-reporting depending on the perceived benefits. A great deal of time is needed to develop relationships and trust before this type of survey is likely to yield accurate results.

Comments on using vocation as a proxy for poverty level, received 14 October 1997 to 1 December 1997

The following two comments outline two ways of using the vocation of the client or the location of her enterprise, as a way of targeting the poorest clients.

• H.S. Shylendra, Institute of Rural Management, India
  SEWA tries to target its members mainly by working with women who are engaged in activities which are pursued mostly by low-income category households. The households (activities) covered are street vendors, house-based workers, rag pickers, labourers, etc. The nature of activity here serves as a proxy for the income levels.

   

• Association des Artisans Senegalais, Senegal
  Many [artisans] are squatters on the property they use for work precarious places and this fact can be used as a reliable indicator of their financial situation. First, they can't build workshops; they use scavenged materials (zinc sheet metal) to demark their work places. For certain professions, these workshops are set up in isolated locations with no access to water, much less electricity. They are often driven off (evicted) and some find it hard to start over. Using this indicator is very efficient since these artisans always cluster in the same area (a piece of unused land). Identifying them thus takes only a few minutes.

Comment on using vocation as a proxy for poverty level, from de Wit, received January 1998

Mr. de Wit points out that the true test of any proxy is how often it gives a "wrong" answer (i.e., identifying a household as very poor when it actually is not).

• Proxies are location specific, and in some situations they may be found to give very accurate information about poverty level. In other situations they may be less accurate. When using a proxy the key is to determine how often cases arise which are "exceptional" (ie., where someone will be wrongly classified), and to agree on a level of inaccuracy which is acceptable to the organisation.