When Duflo won the 2019 Nobel Prize in economics (along with Michael Kremer and her research partner and husband, Abhijit Banerjee), she became the second woman—and at 46 the youngest person ever—to receive the honor. The trio popularized the idea of testing poverty-reducing policy interventions through randomized controlled trials. A professor at MIT, Duflo is a cofounder and codirector of its Abdul Latif Jameel Poverty Action Lab (J-PAL) and a coauthor of the newly updated Poor Economics. HBR: Why did you become an economist? Duflo: For the longest time, I didn’t want to. When I was little—eight or nine—I was very aware of poverty because my mother was a pediatrician volunteering for a nongovernmental organization helping child victims of war. I was focused on finding a way to help people who were poor, but I had no concrete plan. I followed the pattern of a good student in France, studying history at university and adding economics as a double major because a charismatic professor told me it would be useful. Still, at the time, I did not think it was very relevant or practical. In my fourth undergraduate year, I went back to thinking, What about this plan of making the world a better place? I thought maybe I should go into politics or public administration, but I wisely decided to mull it over. I took advantage of an opportunity to be a teaching assistant in Russia for one year, at the height of the transition from communism to capitalism, and economists were everywhere—advising the central bank, the ministry of finance, the ministry handling the privatizations. Because I spoke Russian, I was useful to that work as an interpreter and researcher, and though I was low on the chain, it gave me a chance to see how powerful and influential these people could be. I thought, Oh, this is what economists do. I can be an academic, which suits my personality better than being a politician, and at the same time, actually affect the world. Economists can take more time to think through ideas and get to answers that they think are correct and scientifically backed and then share them. When you’re in the day-to-day of running a government, or an NGO, or a company, you don’t often have that luxury. That’s when I decided to do my master’s in economics and applied to study in the United States. I’ve devoted my career to developing the tools and the infrastructure to answer the question of which policies are effective, which are not, which should be scaled up, and which should be scaled down. And that work—not just my own but the policies that we have spawned through the network of J-PAL—will have touched 650 million people by the end of this year. Why did you push economics research toward randomized controlled trials? If you ask what can be done to improve the lives of people in poverty, too often the answer coming from people in the field, policymakers, or NGOs is one thing: access to credit or open trade or aid or education or fill in the blank. But poverty is multidimensional. For example, with education, are you too poor to go to school or is school not available? If you’re in school, are you learning anything, and if you’re not, why not? It can be divided into better- and better-defined problems. So maybe you are wondering whether having an AI tutor in schools would make a difference for kids. We know how to answer these kinds of questions from medicine: You take a sample of people, choose half at random to receive vaccines and then see whether those who got them are less likely to get sick or be hospitalized. If so, your vaccine was effective. You can do the same thing for policy. You take a random sample of classrooms and introduce an AI tutor in half for six months and then compare student test scores with those from classrooms that didn’t have one. That’s better than comparing schools that have independently introduced AI tutors and those that haven’t because in real life, absent experiments, the schools that innovate are very different from those that don’t. Maybe they have an enterprising head teacher or more government support. Sometimes you’re lucky and get something almost as good as a randomized control trial. But if you’re ambitious and want answers, you usually have to create your own. Why did it take so long to bring that kind of scientific experimentation to economics? The methodology was well known—from medicine and also agriculture. There were even early policy experiments in the United States in the 1970s and 1980s on health insurance and negative income tax. But they stopped, and I think the idea was that it was too difficult, too expensive, especially in the developing world. But in fact, it’s not that hard, and it’s less costly than having the wrong policy for a long period of time, so it’s definitely worth doing. When we started, we did everything by hand. But then we created J-PAL to make it easier. Now if a young person wants to do an experiment, she has support. Speaking of support, have you found economics to be a hospitable field for you as a woman? There were not too many women when I started, probably because it wasn’t very hospitable. I can’t say I’ve personally suffered because I’ve always been in supportive environments. But it is an issue. Something in the culture contributes to the fact that women are less likely to get their PhDs. If they do, they are less likely to continue in academia. If they become an assistant professor, they are less likely to get tenure. At every step, we lose more women. Economists have a tendency to be laissez-faire, not wanting to intervene too much. But about 10 years ago we realized there are things we can do to make the field better for women. For example, most women don’t like and don’t do well with the very aggressive seminar, and so now at MIT, for the first 10 minutes of a seminar, the speaker can speak uninterrupted, after which people can start asking questions. A group of researchers also found that women are interrupted more often in seminar and more aggressively. Of course this is harder to legislate against, but if you know that’s the case, everyone can at least pay attention. These small things can help even the playing field. It’s also a matter of personality, though. There is no gender determinism. Most women tend to not like being as confrontational or as competitive as most men. But it turns out that I don’t mind. How did you and Abhijit form the kind of collaborative partnership that yields a Nobel? You’re right to emphasize collaboration, and it goes well beyond the two of us. This work requires a lot of different skills: You need to be creative for the experimental design, organized for keeping all the trains running on time, analytical, able to write a nice model and interpret it, and so on. That work lends itself very naturally to teams. I also think the field of development tends to attract people who are pretty nice, which helps. But the Nobel in economics recognizes less a particular discovery than it does the broader impact you’ve had on the field, and I think ours comes from the fact that our methodology gathered steam and we spent a lot of energy to make the work easier for others. When we got some grant money, instead of using it on our own research, we invested in creating a network through J-PAL. Originally it was just eight people but from all over—not just MIT, not just the United States. And then we grew. As an economist, your influence doesn’t come from writing a cleverer paper. It comes from facilitating a movement. That’s a different kind of collaboration than the one between two people that you were asking about, but it’s been critical to our success—being willing to take two steps back to focus on the greater good and then, when we were big enough, to cede control and allow each sector of research to be directed by specialists in that field. They do whatever they want now, and we have no idea about it. You are often out in the field, though. Why is that so important? One, it’s the only way to understand what’s going on. Unless you go to Togo, it’s going to be difficult to get a sense of Togo. Two, all projects involve implementation, and so you need to work directly with partners, see what they do, agree on designs, and so on. Three, I just like it. It’s so interesting to talk to people and try to understand how they see things, what their issues are. Being in the field is both necessary and extremely pleasant. Among all the studies you’ve run, does one favorite stand out? It’s hard to choose between your children, but I think my favorite is the one on immunization in Rajasthan, an area of India where only about 5% of kids were getting the five standard vaccines. We tested improving the supply by having regular immunization camps, and that improved the rate to 17%. But in half the villages, we added a small incentive: a kilo of lentils for each vaccine and a set of plates at the end of the course. And that increased the immunization rate to about 37%. It was the most cost-effective intervention to nudge people in the right direction, and some version could easily be put in place everywhere. You’ve been praised for publishing your negative findings. Which of those was your best? One time I was pretty convinced that we had a great intervention: a stove with a chimney to address the problem that a lot of people in India and Africa have: They cook with a clay oven in their house, and the smoke goes everywhere, and people inhale it. We launched a project with Michael Greenstone and Rema Hanna, and we were so sure the effect would be big that we were already starting to anticipate down-the-line impacts on health expenditures, for example. But within a few weeks we realized that people were not using the stoves. The new one was different from what they were used to, and they didn’t like it. I came away thinking, You can not change the way people cook. But there is a coda to the story: A year ago, there was a study introducing electric stoves. I thought, It’s never going to work. But acceptance was very good. So I went from being optimistic to pessimistic, and now I’m learning again. That gets at the question of generalizability. How do you know that an intervention that works in one context will work in another? You can’t answer this question with a single study—but you can with many. On the smokeless stove, there were several studies subsequent to ours, and none found any effect. No one liked it. The willingness to pay was zero. Once you have several findings in several countries, in several settings, you can say, “That seems to be a fact.” Another negative result we came upon early was finding that microcredit had little impact—positive or negative—on the average borrower. At first, we kind of kept it quiet because maybe it was true only where we’d studied it, in Hyderabad, India, where other finance is available. But then we saw the same finding in place after place after place. When you identify a positive intervention, how do you make sure it becomes policy? Some things catch on by themselves, but in a lot of cases you need to work with policymakers, make them aware, and do some amount of coaxing. A lot of J-PAL’s work is in what we call the evidence-to-policy pipeline, and it’s all about how to mainstream an intervention once you know it works. For example, for a few years I’ve been working on a program of mathematics education for very early grades. It started with a pure science experiment that was published in the journal Science. Then you need to create a program that’s easily replicable and try it in actual school environments with regular teachers. Then you look for a good policy window, which we found in India, to introduce it in schools and to hold teacher training. And then you use that to highlight the role that state administrations can play. It’s very patient work, and there is not much academic credit to be gained, but something gets into the system. Especially in a large country like India, suddenly you can get it to millions and millions of children. You have several big jobs. You teach, research, publish, travel, and advise on policy, and you’re raising two kids. How do you balance it all? Not very well. I’m always running from one thing to another. At work the students come first. If they want to meet, I’m available. Everything else competes for attention. At home it helps that Abhijit and I both think we have the best jobs in the world, so we can continue to be passionate about them, and our kids are game. They make fun of us a bit, but they’re also interested and like to ask questions. I don’t feel I need to keep my work from my life. My work is my life. When you win an honor like the Nobel, how do you think about next steps? Going back to my original plan of trying to make a difference, the Nobel presents a huge opportunity. It opens doors. It’s easier to meet policymakers. It helps us be magnets and make things happen. How would you like to see the business world step up to fight poverty? The most immediate and urgent need is to fill the gap in humanitarian aid created by cuts to USAID [U.S. Agency for International Development]. Sixty billion dollars is a lot. But if you take 1% of the wealth of the 3,000 richest people in the world, which is currently estimated at $1.6 trillion, you’d have $160 billion. That’s individuals. If you take corporations, even just the “Magnificent Seven,” [seven tech companies recognized for their market dominance] finding $60 billion to make sure that no kids die of starvation is so possible.
