• A village in rural Indonesia. A new study by MIT researchers suggests local residents make accurate judgments of their neighbors’ poverty levels.

    Image: flickr/p2-r2

    Full Screen

Wisdom of crowds

In countries that lack financial records, how can we tell who is truly poor? An innovative study suggests: Ask the neighbors.


Press Contact

Marta Buczek
Email: mbuczek@mit.edu
Phone: 617-253-2702
MIT News Office

In developing nations, anti-poverty programs face a basic hurdle: Who, exactly, is poor enough to qualify for the aid being given out? Emerging states often lack the official records, such as income and tax documents, that are used to make those judgments in wealthier countries.

Now, a novel study co-authored by two MIT economists has identified a surprisingly effective way of deciding who, in the developing world, is especially poor: Let the citizens sit down and decide among themselves.

The study, done by Abhijit Banerjee and Benjamin Olken of MIT’s Department of Economics along with three colleagues at other institutions, is based on fieldwork conducted in 640 Indonesian villages. Indonesia, like many other developing countries, lacks the comprehensive data needed for means testing; that is, it cannot create objective measures of personal wealth to indicate which citizens need aid most badly.

There are two striking results in the new paper, to be published in the American Economic Review later this year. First, when citizens are asked to make collective judgments about the relative wealth of their neighbors, the outcomes are very close to those produced by objective measures. Second, citizens are far more satisfied by the results when they are consulted than when they are left out of the process.

In theory, that means governments such as Indonesia’s could use the knowledge of local residents to help them distribute aid, while enhancing the state’s legitimacy in the eyes of its citizens.

“It might be easier than we thought to hand over this whole process to the communities,” says Banerjee, the Ford International Professor of Economics and a co-founder of MIT’s Abdul Latif Jameel Poverty Action Lab (J-PAL).

“I think there’s a much bigger role for community-based targeting of these programs than people would have expected,” adds Olken, an associate professor of economics.

Power to the people

The field experiment, conducted in collaboration with the Indonesian government and the World Bank, offered a small amount of cash to households deemed below the poverty line. In one-third of the villages, the government used a proxy means test involving household assets to create an objective measure of wealth. Another third of the villages used the “community method,” as the researchers call it: With the assistance of facilitators from the research project, citizens from certain neighborhoods would gather at meetings and subjectively rank the wealth of the local inhabitants. The final third of the villages employed a hybrid of both methods.

The economists also used a research firm to survey all the citizens about their spending habits. Using this spending data as a baseline for wealth, the results show that the proxy means test is the best method for estimating wealth — but only by a slim margin.

On the other hand, the community method led to 60 percent fewer complaints among villagers about the results of the programs, compared to the proxy means test.  

“People are much happier with the list generated by the community method,” Olken says. “The proxy means test does a slightly better job of assessing the objective things, but the community method does much better job of matching people’s own individual self-assessment of who is poorer or wealthier, and thus generates higher satisfaction.”

Explaining how the villagers’ assessment of poverty differs from the means test is hard, but the survey data contain some clues. Widows, for instance, tend to be considered poorer than the means tests indicate; so are households with a relatively larger portion of children. People with lower education levels also tend to be considered poorer.

Policy upshot: Is it worth the tradeoff?

The policy implications of the study seem clear: In exchange for a slightly less effective method of determining who needs the most help, governments can gain significant amounts of credibility by empowering local citizens to help distribute aid. In a country such as Indonesia, which has made a transition toward democracy since the ouster of its former president, Suharto, in 1998, matters of public credibility are highly significant.

“The question of legitimacy always comes up whenever you have somebody from outside deciding who’s poor,” Banerjee says. “Here, even the attempt to produce modest checks and balances made people much less happy.” Moreover, Olken says, village leaders often “don’t want to be in the business of picking winners and losers.”

On the other hand, as Olken notes, Indonesia’s president, Susilo Bambang Yudhoyono, promised to reduce poverty levels during his 2009 re-election campaign, and the government may feel that the extra bit of accuracy deriving from means testing is worth pursuing.

Olken says the researchers found no evidence of “elite capture,” the attempt by local leaders to manipulate the rankings for the benefit of their own relatives or political allies. However, the research team is studying the results of a follow-up experiment to see if the problem occurs when larger aid grants are at stake.

The absence of elite capture caught the eye of other development economists, including Tim Besley, Kuwait Professor of Economics and Political Science at the London School of Economics, who calls the work “a very nice and persuasive study” in the field.

“When you decentralize anti-poverty programs on the ground, the idea that elites will use that for capturing a program is a big policymaker’s concern,” Besley says. “Therefore, evidence that tells us this is not happening in particular contexts is enormously useful.”

The study, Besley thinks, could encourage similar research and policy analysis in other countries, although any future results might well differ from those found in Indonesia. “It’s very difficult to take a country-specific study and then say it applies somewhere else,” he notes. “But … having a well-done and interesting study like this will, without question, get policymakers discussing the issues … and I think the knowledge base evolves in that way.”

Besides Banerjee and Olken, the authors of the paper, “Targeting the Poor: Evidence from a Field Experiment in Indonesia,” are Vivi Alatas of the World Bank; Rema Hanna of Harvard University’s Kennedy School of Government; and Julia Tobias of the National Team for the Acceleration of Poverty Reduction in Indonesia.

Financial support for the project came from a World Bank – Royal Netherlands Embassy trust fund.


Topics: Developing countries, Economics, Government, International development, Policy, Social sciences

Comments

The article seems to assume that the survey of spending habits gives the "truth" about relative affluence, that the proxy measure is less accurate in that it disagrees with the "truth" defined by that survey, and that the community opinion is still less accurate because it differs more from the survey's "truth". People lie to surveys about their spending and indeed they can provide erroneous answers even when trying to be honest. (I recall an epidemiologist friend telling me that he had once told a surveyor that none of his children had died, when indeed one had, because he forgot. There are errors noted in doctors records as to whether their male patients are circumcised.) It may well be that neighbors have a better idea of the relative affluence of people in a village than any outside survey would reveal. Perhaps the best approach would be to use some kind of Baysian procedure utilizing several sources of information to get a best approximation of "truth" against which to evaluate each indicator.
Back to the top