Randomized Approaches to Development: An Interview with Economist Esther Duflo

From Left to Right: Burkle Center Director Kal Raustiala, UCLA Professor Arnold C. Harberger, Deputy-Director Alexandra Lieben, MIT Professor Esther Duflo, Holt Alden, UCLA Professor Edward Leamer, Arturo Siguenza

Last week, Generation Editor-in-Chief Holt Alden and Editor Arturo Siguenza spoke with world-renowned economist and Massachusetts Institute of Technology (MIT) Professor Esther Duflo during her visit to UCLA. Professor Duflo is the co-Founder and co-Director of the Abdul Latif Jameel Poverty Action Lab (J-PAL), and focuses on development economics and poverty alleviation. The interview covers a broad range of topics, including the use of randomized control trials, effective policies for governmental organizations like USAID, and advice for students looking to work in international development. The following is a transcript of the interview, edited for brevity and clarity.


Arturo Siguenza (AS): In your experience using randomized evaluations, has there been a development puzzle that has been particularly challenging, or that you consider the biggest puzzle you’ve tackled?

Esther Duflo (ED): Not really, in the sense that most things get partitioned away into smaller things that we can approach, so it’s hard to talk about one big problem. But there are things that are… if you ask me whether there are domains that I understand better after fifteen years of conducting randomized trials, than others, then that is true. For example, one key problem in developing countries is the quality of healthcare, that is really terrible, I couldn’t tell you that I have a good idea of what to do about it. I have maybe a better idea of what the nature of the issue is, but I don’t really have any idea, or the beginnings of an idea, of what solutions would be to improve the quality of healthcare. Whereas there are other domains, like for example in education, where I have a better sense of the problem and what could be done, even if I am aware that there are political constraints to implement what would be done. So there are certainly issues that are a bit thornier than others, and healthcare is certainly one of them.  

Holt Alden (HA):  Tied into that, you have a lot of experience with a whole host of different types of [development] variables. Are there any that have changed your perspective on poverty, specifically, or that you really want to focus on going forward?

ED: I tend to jump from question to question, partly because, as a reaction to the fact that a lot of development work, in particular in international institutions and other things, tend to be very siloed. So people are education experts, or health experts, or they are agricultural experts, but in fact when you are looking at, you know, a household as a unit, they are involved in all of these things and all of these decisions tend to be taken together. So the fact that they have health issues has impact on their business, and the fact that their business is very risky has impact on their health.

[blockquote align=”none” author=”Esther Duflo”]So I never really saw myself as becoming a specialist in one topic, as opposed to always trying to keep in mind how all of these things are interconnected. Because each of my projects is reasonably narrow, and one thing at a time — not to become narrow as an economist — I keep circling around, shifting perspectives, shifting projects, as a way of keeping honest.[/blockquote]

HA: Do you feel like that has been effective for you in your studies? Have you been able to relate them together?

ED: Yeah absolutely. That is kind of what we tried to do in Poor Economics, our book, with Abhijit Banerjee. The book is structured by chapter, which are by theme, but at the end there is really an effort to say ‘what have we learned’.

AS: Your work has been focused on utilizing randomized control trials. However, it can be argued that this removes the local context and local demands from these projects by universalising the development issues. What is your take with balancing the local context with your randomized trials?

ED: So I would almost say the opposite, that the randomized control trials are extremely local because they are always conducted in one particular area. I guess you could conduct them in several particular sites together but, even then, each site you would have to know pretty well and understand pretty well. In fact, the most common criticism of randomized control trial is the opposite, that it’s very local, that being so local it loses the ability to generalize from one place to another, so it’s funny that you should ask the question the other way [laughs].

It’s basically impossible to conceive of a good project without having a very good sense of the reality in the field.

[blockquote align=”none” author=”Esther Duflo”]And I think that’s one thing that makes development economists, in particular people who run randomized control trials, different from many other people, in particular bureaucrats in developing countries or people who work in large organizations. We often have a much better sense of the particular realities that people confront in a particular problem.[/blockquote]

AS: Have you ever had challenges from a local bureaucrat that is opposing what you are trying to do?

ED: Not really. The worst that can happen, and that can be really bad, is sluggishness, when you need something and it doesn’t happen. And the best that can happen is actually cooperation. We have great projects that we have done in coalition with local governments where you have people who are actually very committed to what they want to do and very creative. So one big goal of J-PAL as an organization is to nurture champions within bureaucracies, identify them, and then work with them.

[blockquote align=”none” author=”Esther Duflo”]So people are education experts, or health experts, or they are agricultural experts, but in fact when you are looking at, you know, a household as a unit, they are involved in all of these things and all of these decisions tend to be taken together.[/blockquote]

HA: With the randomized control trials, similar to what happens in medicine, have you experienced any trial or variable that works so well that you’ve ended the experiment? Are there any ethical dilemmas that you’ve faced in your randomized control trials?

ED: Not really, of the first type — that things work so well we’ve had to stop the trial — because the nature of the data collection… I think it might change now with tablets and stuff like that where we get data coming back soon… but until fairly recently the only way we could get data is going out into the field and doing a survey. The data is coming in too slowly to really understand something…

HA: Or after the experiment…

ED: After the fact anyway. So I have not found myself in a situation where I have thought, “Oh geez this is working so well I should stop.” But there are particular issues of other nature that arise. There have been some projects that I haven’t done because I thought that they would be difficult to do ethically. And in these cases it’s never the fact that I randomize is a problem, but it’s the fact that we always randomized when there is some presumption that it might not work — but at least some presumption that it might work — so you have at least some uncertainty to justify the experiment. But there are cases when you feel that in the process of collecting the data you would learn things that could prompt, or should prompt immediate action, but you wouldn’t want to intervene at the time because that would jeopardize the experiment. That’s the type of situation I have avoided, where sometimes I have not done projects.

AS: In the same sphere, with ethical challenges that may arise, a lot of development projects have the tendency of going fad-to-fad. Have you had any experience dealing with this tendency? Or using methods to dissuade this?

ED: It’s about trying to get away from that, trying to generate evidence on things, things that might be fad or might not be fad. So one of the things we run into, we evaluated microcredit programs which certainly have been a fad, or at least were very popular maybe about ten years ago. We collected one evaluation, and found mediocre results, not very bad and not very good. So everybody said “Well, maybe that doesn’t generalize”, we also said “Maybe that doesn’t generalize, maybe it was just that it was not good, who knows.” So we waited and then some people generated other studies like that. When there were seven studies together, we published them all together in one place and then there was no escaping that that thing isn’t working so well. I think it did some something to the fad.

AS: Did you face any problems with people that were very invested in microfinance?

ED: Yeah, microcredit has been an issue because the people who are… it’s not the case in all programs, there are a lot of organizations that just do things because they want to make a difference but if they find out what they do doesn’t work they move to something else. That doesn’t trouble them more than that, they just want to be helpful. They are not married to one program.

[blockquote align=”none” author=”Esther Duflo”]But microcredit is different. A lot of people who are in microcredit are deeply invested in it and there have been two kinds of reactions. There have been reaction of organizations saying well that’s useful, we need to change what we are doing. That doesn’t mean people don’t need credit, it means that we need to restructure the offering to better meet the financial needs of the poor. And then there are people who just were upset and tried to make it go away. But it’s not going away [laughs]. I can’t make it go away.[/blockquote]

HA: Do you normally work more with private organizations or more with governments, especially on the ground?

ED: So early on, when we just started, we worked mostly with NGOs, but now we work a lot with governments. I don’t know what the balance is across projects but certainly if you look at the most recent projects, there is a lot of government work.

HA: Have you felt that [government work] has been more or less effective, or less fast?

ED: It is less fast, it is more work. It is like pulling teeth. But I don’t know if you could call it more or less effective. Working with NGOs has a lot of advantages, in particular the fact that you can, if you have a good partnership, you can control pretty well what’s happening on the ground because they know pretty well what’s happening on the ground. So you know what you’re doing. If you’re interested in evaluating an impact of a particular action — if you manage to do it well — then working with an NGO is much better.

On the other hand, when we move to policy many things happen. The scale changes, the targeting changes, the quality of implementation changes and you’re not going learn that by working with NGOs. If you want to know the effect of policies… if you want to learn about people it might be more effective to work with an NGO and manipulate just one thing and manipulate it very well. That might be a first step in some developing-a-policy type of continuum. But if you’re interested in changing policy, that first step might be necessary but it’s not sufficient. It serves different purposes. But it’s definitely more involved by the nature of the of the exercise.

AS: In J-PAL have you focused on working more with NGOs or with government organizations?

ED: So there has been this shift. Generally we were working more with NGOs and now there is a lot of work with governments. But it is also a question of age. Younger people work more with NGOs because they need results quicker and they have less money. Once you are a little established you can take more risks and you have more time. Even within J-PAL if you looked at the age of the people who run projects, it would be pretty well correlated, I’m guessing.

AS: Have you seen a type of development project that has been receiving more funding recently? That has been more popular?

ED: There is a lot of interest in governance issues. Governance, corruption, et cetera. So that’s an area that gets easy funding. There’s a lot of interest in gender issues. There’s a lot of interest in crime and post-conflict instability, like fragile states. These are the areas that have a lot of funding lately.

HA: If you were in charge of something like USAID, or some other government agency here, is there some step that the U.S. could take in order to better support developing countries?

ED: I was a member of the President’s Global Development Council, so I can tell you what we said. There were three pillars. One was focusing on environment and climate [change], because it’s going to become a poverty issue and a lot of the problem is generated in the US. It needs international leadership, it’s not something that countries can really deal with on an individual basis. The second was public-private partnership. I have less to say about that, that’s less my area of interest. And the third one was innovation.

So it was, basically, saying look, the aid budgets are small. They are becoming smaller and smaller, both in absolute value and as a fraction of what developing countries have to spend because of how the world is going. Therefore the share of the US in the aid budget is falling, because there are countries like Brazil and China who are now spending a lot of money on other countries.

[blockquote align=”none” author=”Esther Duflo”]So if we don’t do anything, if we continue building a hundred schools here, and paying some beltway bandit to do some half-hearted agricultural project there… we are doomed to irrelevance. The only way to become relevant and really helpful is to take the lead on generating innovation.[/blockquote]

Not just innovation as in widgets, but innovation as in policy innovation. How do you do that? By devoting a considerable amount of your own budget –instead of just doing things– doing things to evaluate them. Or by scaling things up that have been scientifically proven to be effective in helping countries to scale them up. Or by supporting countries to create such innovation units within their own government. So that’s sort of the things we are pushing. I think if I were in charge of USAID I would try to push on the third, on this innovation lever. There is something at USAID called the Development Lab, I think, and they are very proud of it and rightly but it’s about two people [laughs].

AS: Commenting on something you said about not just policy innovation but also tech innovation, has there been a particular technological innovation that has really come up in development?

ED: Cell phones. Generally I’m reasonably skeptical of technology but the cell phones has just… in the space of very few years, it is everywhere and it’s really changed the way that everything works. So cell phones, coming next is the smart phone. I really think it’s going to make a difference. Then, already there but for now just in Kenya, is mobile money. It’s not just in Kenya, its Kenya, Bangladesh, Pakistan, but it’s really huge in Kenya. In Kenya, however you look at it, it seems to be changing lives.

AS: Has there been a particular method of disbursing, or trying to get cell phones out there to people in developing countries? Or has it just grown on it’s own?

ED: No, you know Nokia is a very cheap phone — has very cheap phones. They just spread, people find it useful. You think of innovation that has really commercially succeeded, you think of the cell phone and mobile money and microcredit, to some extent. But microcredit, I don’t know if it has… it’s just a commercial success.

HA: Regarding the cell phones, I think that has definitely been pushed and helped by processes of globalization and has been a really positive benefit. Do you think that globalization overall has helped or hurt impoverished nations? Or is it sort of a balance?

ED: I have no idea. I’m not very good at this type of question, I don’t think anybody knows. What we do know is that there are some distinct positive benefits, like the cell phone, but we also know there are distinct negative impacts on distribution. When you’re looking at India and the districts that were most affected by the rapid liberalization in the ‘90s, these places became comparatively poorer compared to other places less affected by the removal of the trade barriers. So we do know it affects distribution in the poor countries and the rich countries. I think these are the two things we know.

I don’t know the extent to which.. people are suggesting that as countries become closer together and trade more, they grow more. This worked on the Panama Canal and the advent of plane transport, which made some countries which were previously effectively very far away now closer by, and seem to be trading more and becoming richer. So there is some indication in that sense. I think the question of trying to put this together and asking whether it’s a positive or a negative doesn’t make sense. Because whatever is negative has to be addressed anyway and in principle could be. It shows something of our inability to redistribute gains. But there is no theorem that trade produces no losers unless you are able to redistribute. That seems to have been a big lesson.

AS: Are there particular organizations that you’ve seen on the rise? Startups or organizations like USAID who have had major projects that have been making a difference lately?

ED: You want me to name specific organizations doing specific things?

AS: If you can.

ED: An organization that has gotten a lot of attention, and I think rightly, is GiveWell. It is organizing to provide cash transfers to people in Kenya using mobile money. So they are really exploiting the existing mobile money to provide cash transfer with minimal screening and no monitoring whatsoever. They have done an evaluation of that work and they show that people seem to be spending the money perfectly reasonably. So I think it was a useful result and they got a lot of attention for it, rightly.

Not up and coming, but I am actually a big fan of the World Bank. I think that they have a lot of very good economists, they are trying hard. They are a bit disparate, they are doing this and they are doing that, but at the same time they have a lot of people that know what they are doing. By and large, in particular lately, they have supported innovation so that is nice.

Very self-servingly — but I should still cite them — I’m a big fan of IPA, that’s Innovations for Poverty Action. I’m saying that self-servingly because that’s of course our very close cousin. I think they are great, they are an NGO — which we are not — helping people conduct evaluations of their program. They are working in a lot of countries in the world setting up ambitious projects. That would be my three.

[blockquote align=”none” author=”Esther Duflo”]I’d also cite Pratham, which is a massive education NGO in India. When I said in education we try to understand what should be done, even if there are barriers to actually do it, I think Pratham is the one who figured out how to do it. If someone can do it, they can do it.[/blockquote]

HA: Do you have any new ideas or experiments that keep you up at night? That you think will really help innovate, like you’ve been saying, in developing countries? What’s the next step you want to dive into?

ED: I have a two year old and a four year old, and they are the ones who keep me up at night, nothing else. [Laughs all around] I try to not be bothered by other things. There are two things that interest me: one thing is very abstract, one thing is very concrete.

One thing is so concrete it’s not economics anymore, it’s really more cognitive psychology. In fact I’m working with cognitive psychologists; I’m very interested in small kids — that might be related to my having a two year old and a four year old — I’m very interested in how small kids learn mathematics. And the reason why I’m very interested in this question is because… a central problem of education in developing countries is that poor kids arrive in school with a small delay, just because they have not gone to preschool, or have not gone to a good preschool.

[blockquote align=”none” author=”Esther Duflo”]But then the system in developing countries, partly because of colonial history, partly because of insistence in elite bias… it is the total opposite of “no child left behind”. It is many, many children left behind.[/blockquote]

The teacher focuses on the top of the class and doesn’t pay attention to the children who are lagging behind. So a small lagging behind at the beginning turns into not being able to read at the end of primary school and exiting the school system.

My hope is to really harness the best of science –that’s why I went to see the psychologists and said you’ve got to work on some real problems, not just labs– to harness the best of what they know to construct the best curriculum possible for preschool. Then haul it out everywhere such that poor kids arrive with a slight leg-up compared to the richer kids, instead of a slight delay. Therefore this advantage will accumulate over time, that’s kind of my hope. So that’s one thing that I’m very, very interested in. And again, that doesn’t keep me up at night — nothing does — but it would.

On the other end of the spectrum, I’m very interested in the plumbing of public policy. That’s the one thing that might seem too abstract, literally how the government machine works. So, when I talk about plumbing, it could really be the plumbing of how money travels from point A to point B without leaking everywhere in the system. Suppose I come up with some great ideas, either they can be profitable, so market entity is going to take care of it and that’s fine, or they are not but we still want to do it because they have some social value. Then government is going to have to run it. Then the issue is, how is government going to be able to run it without destroying it in the process or losing all the money in the process, et cetera. The issue is, “Can I help government implement what they want to implement anyway?” As opposed to giving them ideas — you should do this, you should do that — how do you help government implement what they want to do anyways. Starting with fund flows, so I have a project on fund flows. Literally, how does the money go from the center to a local place and how does that affect corruption and stuff like that? And also, organizations of monitoring systems and the like. That might seem terribly unsexy but in the end the big ticket items will be there I think.

AS: One final question to wrap up, here at UCLA we also have an International Development Studies program. What advice would you give to young IDS students in pursuing a career in development?

ED: Once you finish your studies, you should find yourself a job in a developing country, if possible in the field somewhere and not in the capital. Spend a few years there and understand how things really work until you contemplate your next step. I think that would be the most important thing.

[blockquote align=”none” author=”Esther Duflo”]I think there are too many people who work in development that — I think there are bureaucrats in developing countries who just have absolutely no idea what it is. Get yourself out there.[/blockquote]

AS: That’s the biggest problem that we’ve learned. There are all these organizations telling developing countries what to do when they’ve never set foot in the country.

ED: It’s not just that, it’s also bureaucrats sitting in Delhi, dreaming up some systems. Going back to the plumbing issue, I was talking about healthcare, saying the quality of healthcare is terrible. So one of the issues is the nurses in the public centers never show up to work. So you write the nurses are really bad people, but then you look at them and their job description, and its some kind of … so she’s supposed to arrive at the center at 8 o’clock and be there from 8 to 8 and do field visits in the meantime, she doesn’t have any vehicle, she has ten villages that she has to visit. It’s 50 degrees centigrade, she has to persuade people to get sterilized, which, of course, they don’t want to get sterilized. So someone has been sitting down saying what is the ideal job description for one of these nurses without thinking one thing about how hard it is. So the way we describe it in the book with Banerjee [Poor Economics] is the three I’s: policies for developing countries are plagued by ideology, ignorance, inertia. You have no idea of the realities in the field, and once that sort of thing is in place it never goes away.

Please Post Your Comments & Reviews

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.