Andrew Foster is the George and Nancy Parker Professor of Economics at Brown University and the Editor in Chief of the Journal of Development Economics. He has worked on a wide range of empirical microdevelopment topics including health, population, agricultural productivity and technology adoption, household structure, and the environment, with much of this work focused on South Asia.
1. We were interested to see you recently note on twitter that your interest in development came about because, as a 13-year old, you were living in Bangladesh with your family during the time of the terrible 1974-75 famine, and while your family was doing fine, you wondered about the impact on others and decided to collect data. We would love to hear more about this experience. How long did you live in Bangladesh for? How did you go about collecting data as a 13-year old, and what did you find?
I lived in Bangladesh from early 1973 until late in 1975 with my family. My father, an epidemiologist, was the Director of the Smallpox Eradication Program there. In Bangladesh, in contrast to Nigeria where we had lived earlier, eradication involved “surveillance and containment” rather than mass vaccination. As I result, I heard a lot at the dinner table about the value of surveillance data. So, when I was assigned to write a class paper on food, it seemed natural to go out and get some numbers. The rise in rice prices was well over by April 1975 and health conditions were improving. But I figured the circumstances would have been worse for those who were consuming most of their income as food. As a result, my “sampling frame” was two easily found groups near my neighborhood in Dhaka: shopkeepers and bicycle rickshaw pullers. I collected monthly income, household size, and food expenditures and asked about their experience in the famine. I was reminded of this work, by a former teacher, when I returned as a graduate student and started to explain my “new” interest in empirical development economics.
2. You have continued working on Bangladesh on and off throughout your career, and also done a lot of your work on India. How has the development of these two countries shaped what you find most interesting to research about them?
One of the things that I most appreciate about development is the ability to work on a wide variety of topics and to incorporate new data and methods. Not only is there tremendous variety in how things function across countries but even within one country over time development creates new things to think about. Early on I was captured by the improbable nature of what I saw as a kid in South Asia: so many people with such variegated life conditions living in close quarters but maintaining a degree of apparent peace. The region still fascinates me, but the questions have changed. Agricultural mechanization, for example, which Rosenzweig and I study in our most recent paper was negligible in the early rounds of ARIS-REDS. Most plowing then was done by bullocks. But now the bullocks are largely “out of business” and there is, in some places, an active rental market for tractors. New issues like groundwater depletion or air quality have become more salient over time and this has required that I find new ways to think about data collection and data integration.
3. Part of this work has involved really long-term panels, including 30-year plus panels in India and a 40-year panel data set in Bangladesh. How did you get involved in such long-term panels, and what are your thoughts of the feasibility of following people versus locations for such long periods of time?
Long term panels were not something I set out to focus on, but the accidents of academic connections and my own evolving interests made me realize that there was a lot of value to be had in working with these long -term multi-purpose panels. My first real introduction to the ARIS-REDS data from India was when Rosenzweig came to the U of Penn, where I was an Assistant Professor, in 1990. Mark had an old envelope of codebooks from the 1967-71 and 1982 rounds but 1982 had never been released. We worked with various people to gain access to the data, but were also told that the sampling frame had been destroyed so no further follow-up would be possible. Later we discovered that the frames were in a dusty closet somewhere. The Matlab data resulted in part from my connection with Jane Menken, a demographer, who helped advise my undergraduate thesis and involved an interdisciplinary team.
On the second part of your question, I don’t think about place-based versus person-based as a binary choice. It is a matter of how you balance different interests. Obviously, there is a messiness to following people over time. They are born, marry, die, move to a different part of the village, and migrate to another region. Places (mostly) stay in one place. So if one is interested in measuring development or the impact of some (place-based) policy there are advantages to thinking about things only in terms of place. But if you want to think about how people’s responses frame the process of development, I think you really have to concern yourself with how people map into groups such as households or firms over time, as these groups have so much impact on the types of choices that individuals make.
4. As far as we are aware, you haven’t worked in Africa at all. Have you ever been tempted to work on an African country?
The scarcity of work on Africa by me is not for lack of interest. But a lot of what motivates me, as you touch on below, is the study of mechanisms associated with larger economic and demographic change. This sort of work is best suited to household data collected over longer periods of time and perhaps space. Data systems of this sort—whether administrative or survey-based—were simply not available in Africa when I was starting my career. Of course, data availability in Africa is changing and I do hope to find the right question, data and partnership in Africa before too long.
I should also say I have worked on Africa in various ways. Most importantly, I am fortunate to have had talented graduate students working on thesis projects from all over the world, including Milusheva in Senegal and Hardy in Ghana. I also, for example, did some work with my current President Chris Paxson, among others, on economic fluctuations and demographic change in Africa as part of an NAS panel in the 1990s. And, of course, as an editor of the JDE, I get to a play a role in the production process of papers from all over.
5. A lot of your work might be characterized as attempting to use a mix of theory and empirics to describe why a certain phenomenon in development is occurring. For example, in recent work with Esther Gehrke, you look at how schooling investments in rural India respond to risk given dynamic complementarities in producing human capital; while in some of your best-known work, you highlight the importance of learning from others and learning spillovers in the adoption of high-yield seeds during the Green Revolution. Such work has yielded rich insights into the process of development, and suggests directions for policy. But we know that many policies that sound good in theory struggle when it comes to implementation, or may have unanticipated impacts. How do you think about the balance between diagnostic research and impact evaluation/designing and testing policies in practice?
I think of development research as a kind of spinning orbit, a quasar if you will, where each element feeds into each other but also emits bits of information relevant to its own particular context. I love getting into the meat of a problem, using data and appropriate theory to unravel a puzzle or think clearly about a difficult issue. Some of what I do is I hope of broader interest outside of the development community. And I surely don’t think my expertise can substitute for the insights of the civil servant or NGO worker who is on the frontlines of development practice. But if what I do helps, however indirectly, to determine whether and how a program gets designed and implemented then I can take some satisfaction in that. Similarly, the process of program design, testing, and implementation can both be of direct relevant to those with primarily humanitarian aims and set the stage for new ways of thinking about a problem. It seems to me that the Development Impact Blog is predicated on the idea that keeping this circle alive and well is worthwhile, and I would endorse that idea.
6. Putting on your editor’s hat for a minute, we imagine you are seeing/will see a huge number of papers that are about COVID-19 policy efforts or that are attempting to understand other development policies and patterns but with data that covers the pandemic. What criteria do you have in mind when you are considering whether such work is likely to be of lasting and more general interest?
We are seeing some papers that are COVID -related but, to date, the numbers to date have been modest—maybe 1 in 20. My primary view is that the JDE is not a good outlet for work with a primary purpose of having an immediate impact on COVID policy. But I do think there will be plenty of room for COVID work that will be of interest to readers with a longer horizon. I can imagine papers that can address questions like: What does the disruption in trade or labor tell us about the ability of local markets to adapt in the sort run? Why might short-run economic disruptions have longer-run effects for some people than others?
I do suspect that any empirical study that uses data collected over the last year will have to confront the issue of COVID. I do not see this as a problem—every paper is contextual in one way or another. But one of the places this issue seems most significant is in our Registered Reports workflow. We anticipate that a number of Phase 2 papers will report on how they altered plans due to COVID. We had always envisioned such exogenous shocks as a natural and necessary feature of open science in the context of development (though not at this scale!) and therefore have room for such revisions. We only require that the paper conform to the extent possible with the original design and that changes are justified and transparently presented.