Dr. Guido Imbens, a Dutch American econometrics professor at Stanford University is at the forefront of econometrics research, with his theories being widely used in social and medical sciences. As some of you may already know, he started his econometrics studies here, at the Erasmus University Rotterdam. He is one of the winners of the Nobel memorial prize in economic sciences. Following the award, earlier this fall, at Estimator, we had an online interview with Dr. Imbens. During which he shared some memories of his years spent in Rotterdam, talked through the important aspects of being a researcher, and gave some priceless advice for econometrics students.
To begin with, tell us why you decided to study econometrics?
When I was in high school you had to apply to do a particular university degree, so I had to choose while in high school. I wanted to do something like mathematics, but I didn’t want to just study mathematics. My brother had been already doing it, my sister ended up studying mathematics as well, so I wanted to do something different, but certainly with a lot of mathematics. I took an economics class in high school, where we did just a bit of econometrics. My teacher there gave me this book by J. Tinbergen ‘Econometrics’, and said “Guido, this might be something interesting for you”. Now, thinking back to it, I’m not sure what I made of it, but it looked like it had a fair amount of mathematics, it seemed more applied and interesting than just pure mathematics.
So, I went to a student's open day in Rotterdam, and it sort of played out that it was a very small program, very high quality, and so that appealed to me. This is how I ended up in Rotterdam.
I feel like looking back it was somewhat unusual because I think we had only 60 students coming in the first year and for the economics program there were 1200 students every year. So, from the beginning, we were in a really small class, and therefore it was a very hands-on program, where we could actually talk to the faculty, which was very impressive in that sense.
Even today, the proportion of econometrics students is rather small compared to economics students, with a pretty significant drop-out rate, especially in the first year.
It was true in those days as well. They told us: “Look at the person on your left and the person on your right. Only one of the three of you is going to stay.” I think by the end of the year out of 60 it was about 20 or 30 students left. But you know, dropping out just meant that they would switch to other areas, like economics.
Looking back to your years at Erasmus University, what impact did it have on your career?
There are a couple of things. I still know some people. As a result of being in a small course, a lot of people stayed in contact. I wrote some papers with a few of those who stayed in academics at some point. So that’s one thing. Second thing is that it was an incredibly good foundation for going to a graduate school in economics. The US programs tend to be very mathematical, but having a background already makes it much easier to do well in those programs. But, while in the US, I realized that the economics part of the degree was not quite as current, but the technical part was very good. So, it was very good preparation for going to graduate school.
In general, the vast majority of econometrics students after their studies are seeking to work in the private sector. However, you’ve chosen to become a researcher. What had led you to this decision?
While at the university, I didn't think about what I’m going to do after university. Maybe I should have thought about that more. But you see, neither of my parents had a university degree, so for me, it wasn’t very clear what it would lead to, so it was figuring things out as I went along. At the end of my 3rd year, a friend of mine saw this exchange program in England, with Hull University and he and I both went there to do our masters. While there we accepted the offer to go to the US, to Brown University, and continue our masters there. We were planning to go back, but then we thought why would we go back? We already have our masters, better stay in the US and do the PhD. It just seemed like a big adventure, since there was nobody’s intention of staying here in the US and making a career out of this. But I enjoyed being around the universities and I enjoyed the work, so moving to the US, where I had never been before, was more like an exciting possibility rather than a thought-out plan. There was a lot of luck and kinda just fortuitous events involved.
One thing that did make me consider a research career was that in my 3rd year in Rotterdam there was a visitor from I think from Rochester University at the time, and so he, as part of the visit, taught a class on the general equilibrium theory with infinitely many goods. This was a fairly esoteric topic because we haven’t done anything with general equilibrium, to begin with, and even nowadays, general equilibrium is not a particularly hot topic. But at the time, general equilibrium was an important area, and he was working on extensions with infinitely many goods, and so he taught a whole class on this. And in the first class like four or five students showed up and then after the second class, I was the only one left. Everyone else thought it was just completely crazy. I thought it was kind of interesting, so I stayed with it. And so, once that professor introduced me to the idea of doing a PhD in the U.S., and that it might be feasible, I was very open to doing that.
What are the main challenges that you encounter in your daily life as a researcher?
Well, it's a very fortunate career that can give you a huge amount of freedom to do what you want, but at the same time, things don’t always go well. It is incredibly frustrating when you spend a lot of time working on papers, and then you send them to a journal, and people say: “Nah, we don’t think it is very interesting”. But I always really enjoyed looking for solutions, solving puzzles, things that I don’t understand. Then I feel sort of that there are some insights to be gained. Often that has come from working with people who are doing more applied work. So the work I do in the end is a pure econometric theory, but I always try to interact a lot with people doing empirical work. When I started at Harvard, it was Josh Angrist and Gary Chamberlain, who told me that I need to talk to empirical people, listen to them, go to the seminars, but since then I kind of realized that it is a very effective way of finding good problems that need simpler solutions.
How important do you consider collaboration is in the academic world, and do you think that it should be more emphasized in universities?
Yes, that is absolutely right, these things are hugely important. I talked before about how with some of the things you need to get lucky at times, as well as kind of putting yourself out there in the position to take advantage of things. When I finished at Brown University, my first job was as a junior faculty at Harvard, the way that the job market works there is, typically, that you give seminars and do interviews at annual meetings, then some people get invited for the seminars and then they decide whether to give you a job, so I end up giving a job talk to Harvard, and Josh Angrist had already been there for a year. He actually didn’t like the seminar I gave, and he was opposed to hiring me, but luckily the more senior people there decided that I was good enough, so I ended up there. Josh was only there for one more year because he had to go back to Israel for a couple of years but in that one year, we spent a huge amount of time talking together about the econometrics, about what his views were, and what good questions were. And even though he was only one year ahead of me in terms of his job there, he sort of was way more mature in this research field, and he had these very strong views on what was useful, and what was not useful in econometrics. That helped me to develop my own views, and essentially in that one year, we laid the basis for the 1994 paper, which was kind of the main part of the Nobel Prize citation. So, these collaborations were just hugely important and kind of opened new opportunities. Subsequently, I moved to a lot of new places, but at all these places, that was again very helpful for my career, for my understanding, cause spending a lot of time talking to people about what they think is good research, why do they think it is good to research, what are some good questions, has been very helpful for me to get to my current understanding. I also went to Berkeley for a while, and David Card was my colleague there, and so we had these long conversations about these things. I learned a huge amount from there. That’s sort of where the top US department can make a big difference, they kind of just have a huge concentration of people doing very interesting work, and there is a huge number of things to learn from them.
Were you expecting that your research would be not only relevant in economic fields, but also many others?
The narrow motivation was the Angrist draft lottery paper that he had done for his thesis, looking at the effect of military service on earnings, so he used that Vietnam-era draft lottery as an instrument. And so there it seemed to us very convincing, very treatable, and at that time the context for thinking about it was this paper by E. Leamer ‘Let's take the con out of economics‘.
He writes that nobody really believes in empirical work in economics, or at least nobody believes in nobody else’s empirical work in economics. So at the time, there was just a lot of empirical work that people used for complicated models, but it was just very hard to take it seriously. Then, Angrist’s papers, natural experiment papers seemed very different, it seemed like they were doable. But at the same time there was, is this 1999 paper that said - that if there even if there was an effect on the military service, that you couldn’t get the average effect. So we tried to reconcile these two things, we tried to reconcile the fact that in econometric theories you can’t get the average effect, and that it is still intuitive for us that the draft of the lottery paper was a very credible way of getting estimates out of military services. The local average treatment effect kind of reconciles the problem of not getting an overall average by getting the average for these sub-population compilers. This complication made people hesitate that well, this is not a group you can identify, you can’t tell whether someone is a compiler or not. That created a lot of confusion. People were debating whether that was useful or not useful. Certainly, at the time we did not anticipate that the paper would have such a big impact. I did feel at the time that it was a really interesting result and some people agreed with that, but there certainly was a lot of disagreement on that at the time.
But It did turn out well in the end.
Yes. It is quite funny the way it sorts of very slowly became more and more accepted. It wasn't really that immediately people realized that it took a long time. But now it's gotten there. That is sort of how you see the publication process. And it goes back to the earlier question, whether you want to do risky things, and what kind of work do you want to do. Papers like that often have a hard time getting accepted into journals, because they do something unusual. And that's at some level how you should be doing research, you should be taking risks, and accept that not all of them are going to work out. But the things that do work out are so much bigger and their chances are much more substantial that way.
What is your research focused on now and what do you want to focus on in the future?
Well, in some sense most of my work is still one way or the other about causal inferences, but that has become a much bigger part of econometrics. Although it was always a part of econometrics, not as explicitly as it is now. If you look back at Tinbergen’s work on instrumental variables, that was also about causality, but it wasn't made quite as explicit, and certainly, back in the 70s, 80s, or 60s people were less focused explicitly on that policy effects. I'm still working on various aspects of that, some of it has to do with actually doing experiments, doing experimental designs, there is a huge amount of experimentation going on these days in the tech companies, where they are much more complex than just the A/B tests, because that are typically interactions between the individuals, between agents in some setting. So, they are kind of experimenting in more complex settings. So, there are a lot of interesting questions there. With the big data, variability in the large data sets there is a lot more that we can do. Especially if you have data over time, trying to find ways to combine experimental and observational data can be very insightful. I think all of these areas are very exciting at the moment.
As at times the study can be hard, we wanted to know whether you had any advice for young econometric students?
Yeah, getting through is tough. When I was in a graduate program, in the Ph.D. program, at Brown University, at the end of my first year I was thinking about dropping out. I wasn’t sure, I wasn’t motivated enough to keep doing that. I started applying for some other jobs, but I didn't really get anywhere, so I ended up staying in the program. And that worked out well. So, you need to find some balance with the work and try to find a group of people to go through that whole process together rather than be isolated and try to do it on your own. Try to find some support group, especially coming out last almost two years now with the pandemic it’s been very hard to do that. However, there are always going to be times when being a student, whether it is a bachelor's degree or during a graduate program, it is just very hard, so you are going to need other people to get through that. Especially with an econometrics degree, there are lots of possibilities, there are very interesting private sector jobs these days, with a lot of tech companies. Also going to graduate programs provides you with a lot of opportunities. Econometrics programs, like the one in Rotterdam, prepare you very well for that.
We sincerely thank Dr. Guido for providing us with his time to share his experiences and crucial insights.