In 2003, French-American economist Esther Duflo co-founded the Abdul Latif Jameel Poverty Action Lab (JPAL) at the Massachusetts Institute of Technology (MIT) with Abhijit Banerjee and Sendhil Mullainathan. In just over ten years, JPAL has carried out 568 field experiments – or Randomised Control Trials (RCTs) – in 56 countries, nearly 100 of them in India alone. RCTs test whether a particular development intervention – smartcards for benefit transfers, incentives for immunisation – actually works. Ms. Duflo, recognised as one of the world’s top economists, has won numerous awards including the John Bates Clark Medal and a MacArthur Fellowship.
In an interview with Rukmini S , she spoke about what it’s like to bring evidence into Indian policy, and which policies are doomed to fail.
I don’t think many people in India know the scale of JPAL’s work here – both the number and scale of the evaluations going on. Are Indian state governments getting more open to having such RCTs conducted in their states?
Yes, >Karthik Muralidharan’s RCT covers 20 million people in Andhra Pradesh. There is another evaluation I was involved with >which is in Bihar and covers 20 million people. Earlier, >we did one with Rajasthan police which covered about eight million people.
I don’t think they were ever against it but it was just not in the type of things they used to consider. From governments in India, I haven’t seen, to be honest, objection to the idea of experimenting. Some times, they have objections to the programme itself, but not to the idea of trying something. It’s only in the last few years that we have been trying with different government departments and we have now several experiments - >one on pollution auditing in Gujarat , the Rajasthan police one, >two MGNREGA experiments , one >on schools in Haryana and now a series of partnerships with the government of Tamil Nadu. So across states, across sectors, we find a remarkable level of openness to the idea that it’s worth trying things out.
With Tamil Nadu, is the partnership only for evaluations, or will the government also incorporate the findings into policy?
There is an MoU saying broadly speaking that the government and JPAL agree to work together on evaluations. The proposal comes together between a government department and a team of researchers brought by JPAL from all over the world. Then, the proposals are submitted to a steering committee which meets twice a year, which may approve the proposals or not. And then the government will ask the departments to fund whatever is approved, and a significant amount of money is approved both for the evaluation and the programme. In principle, these evaluations can be in any area which interest both the government and the team. What has been explored are proposals in education, in the management of non-communicable diseases like diabetes and livelihoods.
Presumably the intent is for this then to feed into policy. I don’t think the government would invest so much money and effort into this if they were not going to use it for policy. The first steering committee met a few days ago. There were a few proposals before it and they were well-received, so we are on track.
Much of your work has found that the actual job that a government employee has to do is so enormous that there is a tacit understanding that she will only do a part of it. If the new government in India wants to improve public service delivery and reduce corruption, do you think they first need to demythologise a public servant’s actual job?
If they could pronounce that word, it would be a good place to start! I think that’s absolutely right but I don’t know whether anybody would have the courage to do that, because it requires to say in black and white – even though it’s already the case in practice – that we are not able to deliver what we promise to deliver. So right now if you take the example of >nurses , on paper the government is delivering a fantastic system of primary healthcare to the poor. In practice, that system is different, but nobody says it.
A government would have to say: Forget it, we are not going to have this wonderful three-tier system, with the nurse knowing the first name of every person in the village. You’re going to have your sub-centre that’s 10 kilometres from your place, you’re going to have to figure out your way to get there, but at least it’s going to be open, and it’s not going to be open at 8 am because you need time for the nurse to get there, and it’s going to be open from 10 to 4, and this is what she can do.
I think it’s difficult to admit that it will not be someone who knows everyone and who convinces people to was their hands and get sterilised and teach the women about weaning foods. These are things that it would be lovely to have but we are not equipped to. But to admit that would require so much political courage that I doubt it will happen. But I think if it could then yes, at least then we could hold people accountable to a reasonable standard of expectation and that’s the first step.
…The combination of improving the logistics, say with UID, and also simplify and clarify the work of government employees, will help.
Are there any major Indian government programmes that you can see in this way are doomed to fail?
The Right To Education. Not in its entirety – I think it has good pieces to it – but largely. The idea that private schools and community schools need to have such-and-such buildings and infrastructure and pay their teachers that much – it’s not just failing, it’s hurting. Usually things fail and it’s just a waste of money. Here it’s like, the consequences, if it’s adopted, will be hurtful. So that’s one.
The Public Distribution System in its current form is not working. The objectives are the wrong ones, the way it physically works is terrible, the level of corruption would need to improve a lot to be outstanding. The idea that you should push grain on to people – to start with, it seems to me on the basis of all that we know of people’s nutritional problems, that is the least of their concerns. So that’s the big one.
Then the diesel subsidy - although we haven’t done any randomised research, just looking at the numbers, it is a redistribution from the poor to the non-poor, and it’s terrible for the environment. There is no case for it, except that it’s there.
Of the policy lessons that JPAL has come up with in the last ten years, I know deworming programmes in schools have been taken up by state governments. But other work, like on improving teacher attendance, which is over ten years old, has not been taken up. Does it get frustrating that what your evaluations are finding is not making it to policy?
Yes, with >deworming now the national government has also announced that it is taking it up. In the case of the >absent teachers , I’m not sure what could be scaled up….In the two experiments that we did, it didn’t really work at a level where I would be comfortable recommending scale-up.
When you run a very large experiment, you can recommend the scale-up more easily. But in others cases when you do the experiment, and with NGO partners etc, it takes some work to translate it into policy. So take for example education. We did plenty of experiments with Pratham and Pratham knows how to do it as well if they do it themselves. But it took a while, it took about three or four replications with different states to arrive at a model that works within the government. Now we have one, one that we tested in Haryana and it works. So now we are confident to say you can replicate it elsewhere, and so far at least there are glimmers of interest from various state governments, so there’s no reason to be frustrated.