Evaluating Aid Programs
MIT economists judge which methods combat poverty the best
By Davin Wilfrid
In 2003, governments spent $68 billion on programs designed to help the worlds poorest people, according to the United Nations. But how can we tell whether these programs are actually making a difference? The answer, according to economists at MITs Poverty Action Lab, is to compare communities that participate in a particular aid program with similar communities that do not.
Economics professor Abhijit Banerjee, one of the labs founders, uses an example to explain his groups thinking. To determine whether new roads lead to greater economic well-being, he says, its not good enough to measure the well-being of towns with roads against the well-being of towns without roads. Instead, one should take a set of towns where roads are being planned and randomly decide the order in which the roads will be built. Then the comparison of the sites where the roads are built early and the sites where they are built later gives a valid measure of their impact.
Often, people just ask, How do you feel now that the road is built? Our view is that that often gives you the wrong answer, Banerjee says. Now we have a reasonable way of running a horse race between these different things, where for every ten dollars spent, this one really improved peoples lives and that one didnt.
When Banerjee first started studying aid programs, he was disturbed by the lack of solid grounds for evaluating them. You would look at even a simple question, like, Is it the case that having more textbooks or teachers improves student achievement?, and without hard evidence, we really didnt have any answers, he says.
Banerjee founded the Poverty Action Lab in June 2003, along with economics professor Esther Duflo, PhD 99, and Sendhil Mullainathan of Harvard University. The lab now supports the two MIT professors and 13 researchers who work globally to implement randomized trials, and it collaborates with six faculty affiliates at other universities.
Sometimes the researchers findings square nicely with the aims of a particular program. For example, one trial conducted in rural Kenya showed significant increases in achievement among girls whose schools were selected to participate in scholarship programs. But evaluations can also indicate that programs arent working. A second Kenyan trial showed that children who were taught worm-prevention measures fared no better than those who werent.
Once the Poverty Action Lab researchers have collected enough information on different programs, they plan to divide their findings into themes, such as how to get children to go to school. The lab has 15 completed or ongoing studies, whose subjects range from affirmative action programs in India to racial discrimination in the job application process in Chicago and Boston.
Other news from around campus