Miércoles 5 de Julio de 2006, Ip nº 160

Measures of success
Por Drake Bennett

LAST WEEK, when it was announced that Warren Buffett had pledged roughly $31 billion-the majority of his fortune-to the Bill & Melinda Gates Foundation, the fact that the world's second-richest man had entrusted his philanthropic legacy to an existing foundation instead of starting one under his own name only added to the public acclaim. But Buffett said he was simply following what has always been one of his tenets: investing in companies whose managers he trusted, and letting them run the business. ``I've got some people who I say can give it away better than I can," he said of the Gateses on Monday.

Buffett, who considers Bill Gates a close friend, is not alone in being impressed by the Gates Foundation's way of doing things. Dedicated to the twin goals of combating disease in the developing world and improving education, the foundation stands out not only for its extraordinary resources-it was already the world's largest charitable organization before Buffett's gift essentially doubled its size-but for a dogged insistence on making sure the vast sums at its disposal are well spent. Gates is seen as the leader of a new generation of philanthropists who believe they can do the most good by bringing to their charitable work the relentless bottom-line thinking that made them their fortunes in the first place. ``The reason that the Gates Foundation has been very successful is that, from the beginning, they have said, `We are going to be sure we get results,"' says Nancy Birdsall, president of the Center for Global Development, a think tank based in Washington, D.C.

And yet measuring the efficacy of philanthropic aid-where a host of social and economic factors come into play-can be a far more complicated calculation than determining a business's bottom line. In recent years, a few economists and social scientists have begun to take up the challenge, funded in large part by institutions like the Gates Foundation and older aid organizations like the World Bank and USAid. The hope is that, in illuminating what works and what doesn't, these researchers can make sure the billions of Buffett and Gates go as far as possible.

To give a sense of the difficulties involved in quantifying aid's impact, Nancy Birdsall uses a hypothetical example of a program to provide meals for poor children at school. Even if the children's health and school performance improves after the program is introduced, she says, ``you can't assume that it's only because of the aid intervention that you got those results." If, for example, the area where the program has been instituted is going through rapid economic growth or families have more money because it's a good year for the local cash crop, the researcher won't be able to pick out the effect of the program from that of the other factors.

Or take the example of microfinance, the provision of very small loans to people whose poverty would otherwise disqualify them from financial assistance-$50, for example, to buy a sewing machine and set up shop as a seamstress. Such programs are seen as a promising avenue for spurring growth in developing countries. But according to Delia Welsh of the Millennium Challenge Corporation, an aid organization set up by the Bush administration, the actual effect of such programs can be hard to measure. ``The sort of people who apply for microfinance programs are very different from the people who don't apply," she points out. For one thing, they're more enterprising. And when they do better financially than their neighbors who didn't apply for microloans, how much is that a result of the program and how much a result of an industriousness that would have found other solutions had this one not presented itself?

Two years ago, hoping to find answers to these kinds of questions, three economics professors-MIT'S Abhijit Banerjee and Esther Duflo and Harvard's Sendhil Mullainathan-got together to start the Poverty Action Lab at MIT. Their goal, says Banerjee, was ``to bring a simple idea from the natural sciences into the social sciences." That simple idea is the randomized trial. Banerjee's work, and that of other like-minded economists, aims to create-or at least to simulate-the sort of controlled experiment that a pharmaceutical company uses to determine the efficacy of a drug.

One way to create that setup is to intervene selectively-for example, to target aid so that it goes to one community but not to its otherwise identical neighbor. Any subsequent difference between them, then, can be attributed to the aid. At that point, its effectiveness shown, the program can be scaled up. Of course, if no difference is found in the testing, then that result is significant, too.

In one of the earliest such studies, in the mid 1990s, Banerjee looked at an Indian antipoverty organization called Seva Mandir, which ran a large-scale tutoring program for poor children. The program was having trouble getting its teachers to show up for class. To ensure some degree of continuity, the organization, with help from the Indian government, set out to hire an additional teacher for each classroom, in the process doubling the number of salaries it had to pay. At Banerjee's insistence, however, the organization decided to only add a teacher to half of the classrooms and then compare the results. Comparing the two groups, Banerjee found no improvement in students' test scores in classrooms with two teachers.

Randomized studies are at their most helpful when they yield counterintuitive results like these. Sometimes that can mean showing how seemingly common-sense solutions fail, while sometimes it can mean showing how less obvious solutions succeed.

Returning to Seva Mandir's truant teacher problem a few years ago, Esther Duflo suggested a cheaper solution and tested it. She gave teachers in 60 of Seva Mandir's schools a camera that marked its photos with time and date stamps and told them to take a picture with their classes every day. The film was collected each month, and bonuses and pay cuts calculated on how many days teachers could document that they were at work. Another 60 schools continued with the old policy, which was sporadic supervision and the threat of firing if a teacher missed too many days. The results were striking: In the camera schools, teacher absenteeism fell by half, and test scores rose dramatically, as did the rates at which the students gained admittance to primary schools. All for a little more than a third of the cost of hiring additional teachers.

Using similar methods, the Harvard economics professor Michael Kremer has found that a more cost-effective way to increase school attendance in Kenya than subsidizing school expenses like tuition, uniforms, or textbooks-strategies that have traditionally been favored-is giving children medication for parasitic worms, thereby keeping them healthier and less likely to have to stay home sick. Other studies have looked at ways of combating corruption or encouraging childhood vaccination or instituting women's education programs.

There are ethical questions involved in withholding, in the name of social science, aid from people who might benefit from it-just as there are ethical questions about using an ill person in the control group of a drug trial and giving him a placebo when the real drug, though yet untested, might cure him.

Evaluation need not necessarily require deprivation. One of the great success stories of development aid is Mexico's Oportunidades program (originally called Progresa), an initiative that pays millions of poor families to send their children to school and take them to the doctor. Since it had to phase in the program over time, the Mexican government decided to use that process to its advantage, in effect creating a controlled experiment by randomly assigning some villages to start receiving benefits 20 months before others. When the two groups were compared, striking improvements showed up in the health and educational achievement of the children in the program.

Banerjee concedes that there are situations in which the withholding of aid is unethical, but he emphasizes the value of knowing what works and what doesn't. One could easily spend even the wealth of the world's two richest men on untested solutions for the world's problems.

``We spend lots of money on all kinds of aid programs that don't seem to do anybody any good," he says. ``That's why the critics of international aid are up in arms all the time. If you're like me and think that a world without aid is too depressing to contemplate, one of the things you need to do is make aid work."


  02/07/2006. Boston.com.