David Leonhardt has this column in the Times, where he describes the work of the Abdul Latif Jameel Poverty Action Lab at MIT as the most influential work going around on the critical questions facing modern society. Using the methodology of randomized trials, Estehr Duflo, Abhijit Banerjee and Co seeks answers to why so many people in the world are so crushingly poor and what can be done to combat this poverty.
In the randomized trial methodology, a particular anti-poverty strategy would be tested for its efficacy in much the same way as in medical research, by studying its influence on a community (or group) where the particular strategy has been implemented and on another where it has not been. Randomized trials can be very useful in shaping policy by understanding what causes the problems we are trying to cure and which cures work. The results can be analysed to draw meaningful conclusions about the effectiveness or otherwise of the particular policy choice. This in turn can be the basis for advising developing country governments on anti-poverty programs and also in making the most optimal use of aid money.
But as Dani Rodrik points out, there are dangers associated with the generalization possibilities implicit in randomized trials. This arises because there is nothing standard (or uniform) about the settings or context under which the specific society or economy exposed to the trials responds to the anti-poverty particular strategy. The contexts exhibit a mind-numbing diversity between and within developing societies, and their possible responses to the specific strategy or approach could (and will) itself exhibit similar diversity.