Kelsey Jack is an associate professor in the Business and Public Policy group at the Haas School of Business, UC Berkeley. Her research is in environmental and development economics, with work on different forms of adaptation to climate change, programs to conserve natural resources and protect public goods, behavioral factors in environmental decision-making, and much more (including for our measurement method fans, a paper on using remote sensing in RCTs).
1. We usually like to start by asking the interviewee to tell us a bit about how they decided to become a development economist, and what drew them to the area of work that they focus most on. Can you tell us a bit about your pathway to becoming a development economist, and what particularly draws you to work on environmental issues?
The apparent tension between long term environmental sustainability and immediate economic needs – at the household level – was my first moral conundrum. I was 12. My parents and I spent a month in Madagascar. It was my first time in Africa and the juxtaposition of incredibly unique ecosystems and dire economic conditions for rural households was striking. In particular, many these households depend on slash and burn agriculture (tavy, in Madagascar) so environmental destruction is central to their livelihoods. It blew my 12 year old mind. Up to that point, things in the world had seemed pretty black and white, good or bad. Madagascar presented me with a new problem where both goals seemed equally important but somehow at odds. It’s the problem I’m still working on.
After college, where I focused more on international solutions, like UN conventions, I decided I needed to spend more time in the places where these issues were playing out, so I worked for an environmental organization in Lao PDR for two years. That is where I decided I wanted to learn to do research, but also wanted to stay very applied. I applied to graduate school to work on environment and development, and had the good fortune to be in Cambridge, Massachusetts in the early 2000s when RCTs were really catching on in development. For me, this felt like a tool that I could use to better understand the environment-development challenge and test potential solutions, while keeping a close connection to policy and practice.
2. You have several papers where you are working with utilities and using their administrative data, such as with water and electricity utilities in South Africa. Such data can be particularly attractive for graduate students, since they are less time-consuming and costly than running fieldwork, and can offer new windows into adaption behavior. What is your advice in approaching a new utility and persuading them to share data? Is there a key example you use in these discussions, or lessons you have learned over time?
Yes! Administrative data can be really exciting, particularly when combined with ancillary data collection. More on that below. A first condition is access. I’ve seen this work in a few different ways, but it almost always requires some in-person interaction and a period of building trust. I have found it helpful to consider what is in it for the utility partner – why should they take their time to help a researcher? And then consider ways to help make it in their interest. The more collaborative and less extractive the relationship, the more likely it is to work.
I have found that many of the things that we do as part of the research process, before getting to the final paper, can be of great value – and that a little bit of additional effort can go a long way. For example, offering to write up a codebook or guide to the variables in the dataset can both help the researcher by ensuring that the interpretation of the data is correct and can help the utility since they may lack such a guide for their own internal use. Other examples include producing summary statistics, sharing code that they can easily adapt, and identifying outliers or other suspicious patterns in the data – these are all things that typically won’t add much time to the work the researcher already plans to do, but may be of real value to the partner. As a more extreme example, I have worked closely with the municipal government in Cape Town to help make their data more accessible to researchers. A few years ago, I wrote up the experience with my long time collaborator there, Hugh Cole, and a former post doc, Derek Strong. It’s part of a handbook that is a more general resource for researchers interested in working with administrative data.
Figuring out where to focus effort requires taking the time to understand the utility’s problems, the ways in which they are constrained, and what outputs might be of value. In my experience, this usually pays off.
3. However, administrative data often can be a lot harder to work with than it seems from the outside, and from the nicely cleaned figures and replication files one sees in finished papers. What have been some of the biggest surprises you have found in attempting to use administrative data in developing countries?
I completely agree! It can be a steep learning curve. One common theme in my experience is that the administrative data alone (or one administrative dataset) may be sufficient to test hypotheses, etc. but that often adding a little bit of field work can be really helpful. This can look like surveys around key sources of variation (e.g., a discontinuity in an eligibility rule) or introducing a random shock and measuring its impacts in the admin data. Access to data that allows for these kinds of add ons can be a slightly harder sell to the partner, but is really valuable for research.
That doesn’t quite answer your question, though – one thing that I have often been surprised by is the lack of common identifiers across related datasets from the same partner. This is often an artefact of different datasets emerging for different purposes, through different data generating processes, with different software systems used to manage different parts of operations. It’s easy to be lulled into thinking that data coming from the same source would be able to communicate but that’s not the case! Fortunately, with some blood, sweat and tears, this challenge can often be overcome.
4. I recently listened to a nice interview you did with the UCSD backstory podcast, where you discussed the importance of field discussions even for projects in which no fieldwork is required – talking with people from the context and subjecting your ideas and findings to a “laugh test” of whether they seem plausible. I see the importance of this, but there is also the risk of over-updating based on small samples (I’ve definitely had several occasions where something comes up in these discussions that seems important, but then when we ask in a survey turns out to be idiosyncratic). How do you decide who to talk to for these laugh tests, and whether to focus on the average or unusual experiences?
Yes, this is always a tension – learning from more open ended interactions without pretending like they are a substitute for the larger N work we do. I find value in qualitative work at both the beginning and end of a project. At the beginning, when trying to develop a new idea, I try to spend time talking to people to hear how they articulate the challenge or situation. If I spend enough days doing this, similar stories often emerge. This was how Günther Fink, Felix Masiye and I came up with the idea behind our paper on seasonal poverty and agricultural labor supply – we realized that we had all been hearing the same story from farmers in Zambia, and decided to pursue it. Anecdotes can be really valuable for uncovering a new perspective or idea – but then they need to be validated, which I often try to do at a pilot scale. I’m a big believer in the value of piloting, especially for large-scale surveys or RCTs.
At the end of a project, I try to do what you referred to – the “laugh test.” At that point, the data are in and I’ve analyzed and interpreted the results. Then, do the people underneath the numbers think the story is plausible? Even if they might say it doesn’t apply to them, do they think it explains things more generally? At this stage, small-N anecdotes can still be misleading, but often less so than at the start, because it’s now about interpretation of patterns that have been validated in the larger data. Asking people both about themselves and about others (i.e., does this interpretation make sense for you or your household? what about for your neighbors?) can also help if people are more cleareyed about others than about themselves.
5. A lot of the work I see in environmental development studies the behaviors of households and farmers and how individuals make changes. Chater and Loewenstein argue that a lot of behavioral science and nudge theory has led public policy to focus on individual actions rather than systematic changes. Do you see the same happening in development environment work – where there seems to be more work on how poor people are affected by or adapt to climate change, and relatively less work on government actions, focusing on the largest polluters, or changing energy systems?
In my view, we need all kinds of research. Certainly there will be a tendency for researchers to study the questions where it’s relatively easy to access data and to observe some variation. Some areas of work – such as behavioral science, as discussed in the paper you link – will be better placed to talk about individual decisions, while others – such as macro, trade, political economy, etc. – may be better suited to questions of government actions, corporate decisions or international negotiation.
Climate economics has helped push the subfield (environmental/development) toward bigger questions, and toward methods that can handle large scale datasets, limited variation and out of sample projections. Some great examples are Nick Ryan’s work on renewable energy investments in India or Clare Balboni’s work on sea level rise. More generally, I see a nice diversity of methods in the subfield of environmental-development, ranging from experiments and other applied micro reduced form methods to spatial GE models and other more structural approaches. This is important if we want to advance understanding of big, complex problems and analyze potential solutions.
6. You have worked in quite a few countries (South Africa, Ghana, Malawi, Niger, India, Zambia, Bolivia, and Indonesia among others). What is your process for starting work in a new country and getting up to speed quickly on institutions, data, key development problems, etc?
It can be really challenging to start work in a new and unfamiliar setting. On the other hand, I find it exciting, fascinating and inspiring. I wouldn’t say I have a single approach. It depends a lot on the project and my collaborators. In a number of the cases you listed, I have been lucky to have coauthors with a lot of experience and understanding of the place. For example, in South Africa, Ghana, Niger, Zambia and Indonesia, I’ve worked with coauthors from the country, who inevitably have a far deeper understanding of their home than I will ever have. In other cases, I have worked with international coauthors who have deep experience in a place, or with a locally based implementing partner. I guess the summary point is that I have rarely tried to be the local expert on a project, and have instead benefitted from great collaborators and partners!
Here are our previous Six Questions with interview series:
· Six questions with Chris Udry
· Six questions with Rohini Pande
· Six questions with Mark Rosenzweig
· Six questions with Martin Ravallion
· Six questions with Andrew Foster
· Six questions with Tavneet Suri
· Six questions with Morgan Hardy
· Six questions with Oriana Bandiera
Join the Conversation