photo: ®Uplift, BRAC program

About the authors: Anne H. Hastings is a Global Advocate for Uplift and was the Director of Fonkoze when she brought the graduation program to that institution. Steven Werlin is the Communications and Learning Officer for Fonkoze’s graduation program, Chemen Lavi Miyò (the Pathway to a Better Life).

Eliminating poverty stands as the first of the 2030 Sustainable Development Goals (SDGs) that 193 countries voluntarily committed to achieving. A growing movement is arguing that to achieve this goal, we must immediately and urgently focus our anti-poverty efforts on the poorest of the poor because they are the most difficult to reach and have the toughest time making their way out of poverty.

They face barriers to changing their lives beyond those faced by most of the poor: they often have many children who are not in school, go days at a time with little to eat, and have few or no productive assets that would help them make a living – like livestock, tools or land. They often have no hope of a job because they have neither job skills nor literacy and there are few, if any, jobs to be had.

Yet evidence suggests that the world is making incredible progress in reducing poverty. Poverty and even extreme poverty have declined markedly in past decades, according to the World Bank. (1) Recent estimates suggest that, in 2013, 10.7 percent of the world’s population lived on less than US$1.90 a day (the prevailing threshold of extreme poverty) down from 35 percent in 1990 and 42 percent in 1981.(2) In all likelihood, it has continued to fall. Homi Kharas of the Brookings Institution calculates that someone emerges from extreme poverty every 1.2 seconds.(3)

The Rationale for “Targeting”

So why do we need to focus on the poorest of the poor? According to The Economist, the struggle to eliminate poverty is about to get much harder, very quickly.(4) The main reason poverty has decreased so rapidly over the past two decades is the tremendous progress made in China, Indonesia, and India. Most of the poorest people now live in Sub-Saharan Africa and South Asia, regions that have endured many more failures than successes. The Economist has argued:

“With more destitute inhabitants than any other region, sub-Saharan Africa now drives the global poverty rate.”(5)

Economic growth is weak, governments are fragile, conflict is rampant, welfare systems are often non-existent and, perhaps most importantly, both the proportion and the intensity of poverty are greater.

It would be a mistake, therefore, to assume that a rising tide will lift all boats: that just because poverty has been decreasing, it will easily be eliminated. We can ill afford such optimism. Overall improvements say nothing necessarily about what is happening at the very end of the spectrum. Reducing poverty generally and helping the poorest and most excluded – those we describe as living in “ultra-poverty”(6) – are distinct undertakings. Many anti-poverty programs do not even reach people living in ultra-poverty because they are invisible.

Their lives are undocumented. They lack birth certificates and do not appear in government records. Their neighbors ignore them, and they themselves withdraw from the communities surrounding them due to shame, fatigue, and fear of rejection. In Haiti, as in many Sub-Saharan African countries, for example, overall levels of extreme poverty have declined while the rate in rural areas has remained unchanged.(7) There is a need, then, for specific interventions that can support those struggling to leave ultra-poverty.

If meeting the first SDG requires that we reach those households in conditions of ultra-poverty, how do we do that? Any remedy must have specific and transparent strategies with which to identify these households and must be held accountable for doing so.

The Problem with “Targeting”

Graduation programs, many of which are based on a model originally developed by BRAC in Bangladesh, identify households in ultra-poverty by first recognizing the poorest regions in a country and then the poorest households within those regions. Some conditional and unconditional cash transfer schemes attempt to accomplish the same thing in different ways.

Important as it is, any such strategy, usually referred to as “targeting”, is controversial, however, drawing criticism that generally falls under one of three themes:

1. Thinking of people as “targets” conceives of them as passive recipients rather than “actors” who, with assistance, can make their way out of poverty. It demeans them as incapable of their own agency.(8)

2. Targeting is part and parcel of a neoliberal concept of social policy that prioritizes low taxation and limited social spending and therefore favors targeting people in ultra-poverty as a means of reducing costs. In other words, it “rations” benefits to the needy when it should be ensuring entitlements to all citizens.(9) This is the rights-based approach now taken by the International Labour Organization and most United Nations agencies.
3. Targeting mechanisms in general are inaccurate, expensive, and engender conflict within communities.(10)

Each of these criticisms deserves careful attention. The first may seem to address specifically the use of the word “targeting,” but we think it is bigger than that. Any process by which an organization or a government takes responsibility for choosing those it will serve subjects those chosen to its selection process. Assisting people living in ultra-poverty means facing, together with them, their isolation, their lack of confidence, skills and opportunities, as well as their lack of hope that tomorrow will be better.

How they initially engage in a program matters less than where that engagement leads. Graduation programs, as opposed to older forms of aid, aim specifically to help their participants take on long-term responsibility for managing their own livelihoods and lives. Their agency grows with participation and completion of the graduation program.

With respect to the second criticism, we agree that countries should ideally distribute public resources in ways that benefit everyone, at least in countries with a functioning set of social protection systems, such as affordable healthcare, pension systems, and other such supports. But elsewhere it seems appropriate to focus investments, especially where the lack of resources entails severe consequences, like hunger. In addition, households have different needs as they move along the poverty spectrum. Making the same entitlements available for everyone does not mean that all will have the capacity to use those entitlements effectively.

Equal and equitable are not necessarily the same. And where resources are most scarce, equal distribution can leave too little available for the neediest. For instance, among the 14 countries with the highest burden of ultra-poverty, the revenue per capita varies from US$1.85 in Ethiopia to US$781.00 in India. (11)

The third criticism requires a more detailed response, which we provide shortly. But in any case, we set out from the fundamental assumption that eliminating ultra-poverty with whatever resources are likely to be available will require accurate identification of those who need the closest accompaniment and the biggest “push” in order to change their lives. We cannot afford the luxury of the pessimism expressed by Nicholas Freeland, who calls such an assumption “delusional.”(12) So, whether one speaks of “targeting” or “rationing” or, as we tend to say in Fonkoze’s program in Haiti, “selection,” finding a good way to reach the very poorest remains a fundamental part of the struggle to end poverty.

Selection Errors and Proxy Means Testing

We acknowledge that accurate identification of households viewed as being in ultra-poverty is difficult. Both our own experience in Haiti and relevant studies tell us that much. As Dean Karlan and Bram Thuysbaert explain, it is hard both to establish the right criteria for selection (given the multi-dimensionality of ultra-poverty) and to identify the households that meet the chosen criteria. Measuring income is perhaps most difficult given that those in ultra-poverty earn whatever income they have from informal sources and are often paid in-kind. Moreover, potential recipients may not wish to give fully candid answers to questions about their livelihoods.(13) In our experience in Haiti, some people tried to hide their poverty, often out of shame, and others exaggerated their poverty, hoping to qualify for benefits.

Strategies to identify the poorest can run afoul of errors of inclusion and of exclusion. The former occur when households are identified incorrectly as qualifying for services, and the latter when a selection process misses households that do qualify. Including those who do not need the program can substantially increase program costs, and excluding those who do need the program guarantees a program’s failure from the start.

In Haiti, we worry about both sorts of errors. But rather than starting from the assumption that accurate selection is impossible, we look to the more promising methods of finding those families truly mired in ultra-poverty. There are clear distinctions among the various methods that have been or could be employed, and those differences hold varying levels of promise.

Many critics reserve particular ire for proxy means testing (PMT), a method that identifies a relatively small number of markers to calculate the likelihood that a household belongs to a particular wealth category. Kidd and his colleagues describe the approach in some detail:

Conventional means tests assess eligibility for social assistance schemes by verifying whether an individual’s or household’s actual financial resources fall below a predetermined threshold. The PMT methodology, on the other hand, tries to predict a household’s level of welfare using a statistical model. It was developed to address the concern that undertaking a conventional means test based on measuring incomes would be difficult in developing countries, since only a small proportion of the population are in the formal economy, meaning that governments cannot easily obtain information on their incomes. (14)

Characteristics like quality of housing, educational attainment, or location are entered into an algorithm that weights each factor differently. Together they serve as indicators, or proxies, of a household’s total means.

Kidd shares the results of a series of studies of its efficacy. Kidd focuses – to our mind correctly – only on errors of exclusion. While wasting money on services that are not really necessary is undesirable, failing to reach out to families who need those services seems, to us, far worse. If the studies he cites are typical, then exclusion is a regular feature of government programs that depend on PMTs. He mentions error rates as low as 56 percent in Cambodia and as high as 93 percent in Indonesia. If PMTs are missing more than half of the poorest families in the best case, then one has to wonder why they are being used at all.

A Better Selection Process

PMTs are not the only way to target. Fonkoze’s approach in Haiti – like those used by many who have copied or adapted the BRAC graduation program – combines a method of gathering information about community members, via an open public meeting, with a two-step verification that uses program-specific inclusion and exclusion criteria.

Fonkoze does not rely simply on consideration of income, nor an analysis of consumption, to define who is “poorest”. It goes beyond these two measures to look at access to quality health and education, as well as clean water and sanitation. Our approach rests on an understanding that poverty has many dimensions.

This approach is not perfect but it is the best we have seen partly because it allows consideration of a range of the various dimensions of poverty. Our own small study of the process, undertaken by the Institute of Development Studies (IDS), is underway, but preliminary indications are that it selects families who are, on average, significantly poorer than those selected by the PMT most commonly used in Haiti.(15)

Karlan and Thuysbaert’s study of a similar approach – they refer to it as the TUP (as “Targeting the Ultra-Poor” was the name of the original BRAC program) – in Honduras and Peru showed mixed results, but in the end they were able to conclude that:

Overall, the comparison unveils three insights into the TUP selection process. First, when judged using five different poverty metrics, the TUP process typically performs better than random selection. Second, the TUP process, compared to PPI (Progress out of Poverty Index)(16) and the Housing index, leads to selecting households with less land and less valuable livestock. Third, the pattern demonstrates that the TUP process performs best for measures that are easily observable to the community; i.e., the TUP process leads to selection (based) on assets, and less so on consumption or education. (17)

Identifying families in ultra-poverty through a combination of community participation and careful verification against fixed criteria has two very distinct advantages: 1) it ensures that we select those households that are most in need and can most benefit from the program, and 2) it builds a sense of ownership and buy-in within the community. (18)

The process begins with geographical targeting to identify the poorest districts within the country using statistical data from the World Bank, the World Food Program, or other analyses of vulnerability and poverty within the country. This is supplemented by discussions with stakeholders, such as local governments and perhaps microfinance institutions, in those regions.

The next step is taken within the communities themselves, which typically consist of about 50-80 households. If the community is larger, we have to divide it into two given the process depends on participants’ accurate knowledge of their neighbors; this knowledge is less accurate as the number of households increases. We use an exercise called “Participatory Wealth Ranking” (PWR). First, we visit a community informally, looking for local leaders who are willing and able to publicize a meeting and motivate as many of their neighbors as possible to attend. We provide written invitations that he or she can distribute.

On the day of the meeting, participants are asked to draw a map of the community, which they often trace in the dirt with a stick. They identify all the community’s landmarks and place a numbered marker for each household. During this activity, we have a staff member putting each family’s name on an index card. Our staff members also identify participants who seem knowledgeable and respected. We generally try to pick out five or six.

While one of our staff members copies the map, another offers refreshments to the participants. Meanwhile, the third draws the five or six aside for a second activity. In that smaller group, we take the first two index cards, and ask which family is wealthier, putting the cards in separate piles. We then compare the third and then the fourth card, and then the rest, one at a time, organizing the cards into about five different piles. When this is finished, we ask the participants to analyze each pile and to identify the traits that all the families in the pile share. If there are families that seem out of place, they can be moved into the correct pile at this time.

Finally, staff members visit each and every household in the lowest two categories in order to assess their eligibility for the program according to a series of program-specific inclusion and exclusion criteria. We use multiple, sometimes-overlapping criteria as doing so gives us the best chance of finding simple and verifiable ways to evaluate a family. Staff members also use the map produced at the PWR meeting to identify any households that the PWR might have missed. We have found that some of the poorest households go unmentioned at these meetings. These families are invisible, even to their closest neighbors.

Photo: Fonkoze Program

The criteria we use include the following:

• Food insecurity with hunger, such that a household regularly goes days at a time without even a single cooked meal
• A lack of productive assets, like livestock, land under cultivation, or business capital
• The presence in the household of a woman who has dependents
• A dependence on income from begging or day labor
• School-age children who are not in school
• A lack of external support from another family member or another organization

Staff present to management a list of the households they believe are eligible for the program and a supervisor again visits those households to verify the information they have been given. This verification visit is as important as the first visit, and often new information is discovered.

We make the criteria as clear and as verifiable as possible, both to facilitate eventual community buy-in and to minimize errors. The IDS study of our selection process initially showed apparent inclusion errors regarding 8 percent of the families we selected, but further investigation of all these cases showed that selection was probably appropriate. We found, rather, that the PMT tool used for comparison had weighted factors that distorted the households’ circumstances.(19) We have less evidence concerning possible exclusion errors but given we are able, in principle, to work in an area more than once, we can use the period of our initial intervention to find any additional families that we might have left behind.

In summary, the process does a reasonably good job of enabling us to exclude both those who do not qualify based on their relative wealth and those whom our program cannot help due to their age or infirmities. It creates what we believe to be the ideal situation, where a program for qualified families in ultra-poverty is one piece of a comprehensive social protection system that offers appropriate support to all those who need it.

For example, those whom we cannot help because of their age or disability could be helped by a government pension system. Those, however, who do not qualify (perhaps because they are supported by a family member living in the U.S. and so have sufficient food and their children are in school) might be better assisted by a microfinance institution that could teach them how they can best invest the money they receive.

We in Haiti are frustrated at our inability to find alternative services for those who do not qualify for our work. On the one hand, the government does not currently have a safety net for the aged or the severely disabled. On the other hand, even pro-poor microfinance institutions are often unwilling to accept families who are too well off to need graduation; the small loan sizes or remoteness involved can make them too expensive to serve.

Of course, there can be barriers to any selection process. For instance, the households can be so far apart that it is virtually impossible to identify anything that could be defined as a community. In Haiti, the rural population often is not really organized into distinct villages, so the divisions we draw in our selection units can be arbitrary. Sometimes the politics of the community might make it impossible to assemble a group that represents all the differing factions. Only a very few of the communities we have tried to work in have refused to cooperate.

What matters most is adherence to three core principles of the process for identifying families that need services designed for people living in ultra-poverty:

1. Consider local dynamics. Walk through the entire community to identify the most marginalized, and get a physical ‘lay of the land’. This identifies community assets, enterprises, and families on the physical and social periphery.
2. Engage the community. Involve community members in the process of identifying various levels of community well-being in ways that go beyond involving one or two key informants who might select only their friends and families.
3. Verify your information. Cross-reference insights from the community through simple surveys or verification processes.

It is possible to apply these principles even in countries with established national registries and databases that gauge levels of poverty as a basis for anti-poverty programming. For instance, one can leverage these national databases and conduct activities that build community engagement and support, while ensuring that needy households are not omitted. A simple household verification survey can ensure that database information is accurate before including households in these comprehensive graduation approaches.

Is this approach sufficiently cost-effective, scalable, and viable for governments?

The most salient questions about the method we have described are its expense, its scalability, and whether governments can implement it with their own personnel. We do not yet have an example of this approach covering an entire country, or of a government trying it on its own.

But we do have some indications about cost. Some argue that the process is too complicated and staff-intensive to be cost efficient. Karlan and Thuysbaert suggest that in the countries they studied, it costs about US$7.00 per household, just slightly more than using the Progress Out of Poverty Index, a form of PMT.(20)

However, this targeting method allows for many other benefits, such as community acceptance of the decisions and a much deeper knowledge of the community. Ultimately, of course, the cost of the selection methodology depends on the cost-benefit of the intervention for which the selection is being made. While not the topic of this paper, most randomized controlled trials of graduation programs have typically demonstrated their positive cost-benefit.(21)

We also have some positive indications about its scalability. BRAC is assisting the Government of Lesotho in implementing a graduation program, and the Ministry of Social Development has ambitions to cover the entire country using this selection method. Similarly, BRAC is advising the governments of Kenya and the Philippines in applying these programs. According to Aude de Montesquiou and Syed Hashemi, there are some 57 graduation programs underway in nearly 40 countries, of which one-third are led by national governments. (22)

We will not really know whether effective targeting processes can be scaled across whole countries until we try. But we can already draw four important conclusions:

1. We must immediately and urgently focus on people living in ultra-poverty if we expect to meet the first 2030 SDG.

2. To do so, we must have a selection process that can transparently and accurately target those households most in need of support. We are in favor of any process that transparently and effectively identifies the poorest of the poor and the graduation selection process developed by BRAC, and used by many others, is the best we know of.

3. As a community, we must face and overcome the challenge of developing clear, comparable selection criteria, despite what might be dramatic differences in cultural and economic contexts.

4. Finally, we must hold implementers – whether civil society organizations or governments – accountable. They must be able to show that they are, in fact, reaching families living in ultra-poverty and accompanying them as they strive to make their way out of this inhumane form of poverty.

Notes

1. World Bank Poverty Overview.
2. Ibid.
3. “Fewer, but still with us,” The Economist, March 30, 2017.
4. Ibid.
5. Ibid.
6. We define a household in ultra-poverty as one that has a weighted rate of deprivation of 60 percent or more on the Multidimensional Poverty Index developed by the Oxford Poverty and Human Development Institute. Please see more about our methodology [here.]((https://www.ultra-poverty.org/our-methodology/)
7. “Living Conditions in Haiti’s Capital Improve, but Rural Communities Remain Very Poor,” World Bank, July 11, 2014, (Link)
8. Amartya Sen, “The Political Economy of Targeting,” in Public spending and the poor: Theory and evidence, edited by Dominique Van De Walle and Kimberly Nead (Washington, D.C.: World Bank, 1995), 11-24.
9. Stephen Kidd, Bjorn Gelders, and Diloá Bailey-Athias, “Exclusion by design: An assessment of the effectiveness of the proxy means test poverty targeting mechanism.” International Labour Office, (Switzerland: International Labour Organization, 2017). (Link)
10. The Development Pathways blog. “Rationing, not targeting,” blog entry by Nicholas Freeland, April 11, 2017. (Link)
11. World Bank.
12. The Development Pathways blog. “Rationing, not targeting,” blog entry by Nicholas Freeland, April 11, 2017. (Link)
13. Dean Karlan and Bram Thuysbaert, “Targeting Ultra-Poor Households in Honduras and Peru,” December 2015. (Link)
14. Stephen Kidd, Bjorn Gelders, and Diloa Bailey-Athias, “Exclusion by design: An assessment of the effectiveness of the proxy means test poverty targeting mechanism,” ESS Working Paper #56, (Switzerland: International Labour Office, 2017, 1) (Link)
15. Martin Greeley, Liam Kennedy, and Alexandra Stanciu, “Who are the ultra-poor: A Haitian study,” Institute for Development Studies working paper, 2017.
16. The Progress Out of Poverty Index, one example of a PMT. The name was recently changed to Poverty Index. For more information, see povertyindex.org.
17. Karlan and Thuysbaert, p. 18.
18. Vivi Alatas, Abhijit Banerjee, Rema Hanna, Benjamin A. Olken, and Julia Tobias, “Targeting the Poor: Evidence from a Field Experiment in Indonesia,” American Economic Review, 102, no. 4 (2012): 1225, Link
19. Greeley, Kennedy, and Stanciu, op. cit.
20. Karlan and Thuysbaert, p. 33.
21. See, for example, Nathanael Goldberg, “What We Know About Graduation Impacts and What We Need to Find Out”, Policy in Focus, 14, no. 2 (July 2017), 36-39. Link
22. See the Insight on this website by Aude de Montesquiou and Syed M. Hashemi, entitled “The Graduation Approach Within Social Protection: Opportunities for Going to Scale.”