Better than Trial and Error

A recent article in the Forbes magazine on randomized trials of development projects (via PSD blog);

“The Indian antipoverty group Seva Mandir runs an educational program that teaches 4,000 kids ages 7 to 10 math and reading and writing in Hindi. Seva Mandir had a problem: The teachers, recruited from the villages, often with only a tenth-grade education, would show up at the school only 60% of the time. While a teacher could be fired for excessive absenteeism, the remote locations throughout the impoverished state of Rajasthan ruled out regular monitoring.

A classic way of dealing with this problem is to throw money at it--which is what officials at Seva Mandir proposed to do. They would hire an additional teacher for each classroom, doubling the cost. But in stepped a group of development economists from the Massachusetts Institute of Technology. For a few years now the group has been testing antipoverty programs using the same scientific technique pharmaceutical researchers use to evaluate new drugs: the comparison of a randomized test group with a control group.

The idea is to divide a targeted population into two groups, then give the aid--microcredit, computers, textbooks, teacher incentives, health care programs--to one group but not the other and compare the results. "We aren't really interested in the more-aid-less-aid debate. We're interested in seeing what works, and what doesn't," says Abhijit Banerjee, a development economist at MIT who (with Esther Duflo of MIT and Sendhil Mullainathan of Harvard) helped found MIT's Poverty Action Lab.”

Abhijit himself was an ‘activist’as a student in India. There course on Evaluating Social Programs is also conducted at Centre for Micro-Finance in Chennai as well.


Powered by Movable Type 5.02

About this Entry

This page contains a single entry by Paul published on June 7, 2006 1:22 AM.

A Reality Check on Iraq was the previous entry in this blog.

Google Spreadsheets Beta Not Ready for PrimeTime is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.