Monday, August 29, 2011

Poor Economics: Designing robust interventions to fight poverty through randomized experiments

When I hear a professor from MIT saying she knows how to fight poverty, my first reaction is that of skepticism. “Does she even know what poverty is?” That’s how my mind would respond. And yet when I heard the TED talk by Prof. Esther Duflo of MIT titled “Social experiments to fight poverty”, I was thoroughly impressed. It took me some time to understand what exactly in the talk that impressed me so much. Subsequently I bought the book “Poor Economics” which she co-wrote with Abhijit Banerjee. Now, I am slowly beginning to understand why Amartya Sen has said, “A marvelously insightful book by two outstanding researchers on the real nature of poverty”. In this article I want to highlight three things that I find interesting in the book.

Intractable problem, manageable sub-problems: Poverty eradication looks like an unsolvable problem. People from Karl Marx to Mahatma Gandhi have taken a shot at it. However, it is not clear whether we have a handle on it. Silver bullet approach like “give aids” is not helping. Banerjee-Duflo take a different view. They feel that instead of trying to answer the top question, why not look at some sub-problems – each of which might have a definite and practical solution. For example, in the talk, Duflo presents three such sub-problems: How can we have more kids immunized? How can we get more people to use bednets that can reduce malaria affliction? How can we get students to attend school more number of days for a given dollar spent on the cause? The book, like the talk, shows how we can go about systematically addressing these sub-problems.

Randomized experimentation: Let’s take the question of immunization. In Udaipur district of Rajasthan, it was found out that only 1% of the children are fully immunized. The vaccines are there and are available for free. It is not that the parents don’t care about their kids. When their kids get measles parents end up spending thousands of rupees in treatment. So you have empty village sub-centres on one hand, and crowded hospitals on the other hand. So looks like the intention is not translating into action. What do you do? Dulfo & team decided to try random trials in 134 villages in Udaipur district. For one third villages there was no change, another one third villages had immunization monthly camps conducted and the last one third had camp plus a kilo of lentil free for camp participant. As it turned out the immunization percentage jumped by a whopping 37%. Note that this approach is analogous to randomized control trials used in medicine to discover drugs.

Design as if implementation matters: The beauty of Banerjee-Duflo approach is not that it does not make any assumption about the culture, anxiety, aspirations of the poor. The experiments reveal their biases anyway. I call this approach of designing an intervention – design as if implementation matters. Note that the approach does not advocate laboratory experiments – the experiments are performed in-field in actual conditions. I feel that experimentation and immersive research are the heart and the soul of systematic innovation. Banerjee-Duflo approach epitomizes both. Now, I know why the TED talk struck such a chord with me.

I strongly recommend the book for every student of social innovation.

No comments:

Post a Comment