Billions of dollars are spent each year on efforts to alleviate poverty in the developing world. Yet there have been remarkably few systematic attempts to examine what actually works. All too often, nongovernmental organizations and foreign aid agencies give away money based on intuition, or on whatever interventions are en vogue at the time. That’s a recipe for aid programs that fail as often as they succeed.

To improve the success rate, three MIT professors founded the Abdul Latif Jameel Poverty Action Lab. Since 2003, its researchers have used randomized trials to figure out which interventions help the poor the most. Their global network of researchers has made some surprising discoveries: Giving mothers in India a bag of dried beans when they bring their children in for immunizations doubles the rate of children who get all their shots. Telling teenage girls in Kenya to abstain from sex doesn’t lower their HIV infection rate, but telling them to avoid sex with older men — who are much more likely to carry the disease — significantly reduces their chances of infection. In the Philippines, a significant number of women are willing to sign up for special savings accounts that don’t allow them to withdraw money for a year. When the year was up, those who opted for the special accounts had 80 percent more money saved than that those who didn’t.