U of T Statistical Sciences: Congratulations on your position at the Harbin Institute of Technology! How are you finding the transition to teaching?
Kevin Haosui: I’m associated with the Institute of Advanced Study in Mathematics, and our institute actually has very little teaching duties. Basically one course per year.
I just got on board this year, so I haven't really gotten a chance to teach any courses yet. I’m probably going to teach a full course next semester or maybe the summer semester in 2023.
So what are you working on right now?
Currently I'm mostly working on my interest and transition to mathematical economics. I've been working on a field called general equilibrium theory, which people think of as the mathematical foundation for microeconomics. [I] study the existence of equilibrium and the social welfare properties of equilibriums in various kinds of models.
I also work on something called economic dynamics: using Markov decision processes to characterize mis-specified models. This part is related to statistics because it's closely related to econometrics. So yeah, those are the main areas in economics I've been working on.
How did you find your way into statistics?
I did my undergraduate degree at the University of Toronto, and in my third year I took several courses in the math department in pure math and the foundation of mathematics. Stuff like set theory and math logic and so on. And I got quite interested in that.
I did a summer research project with professor Frank Tall in 2010, about 12 years ago. Towards the end of my bachelor's, I asked him if there were any applications of logic or set theory to other areas of mathematics. He told me about a tool called non-standard analysis, so I took that word and did a bit of research. I thought that it would be pretty hard to get a job if I were to only concentrate on set theory and logic because it's a very small group, so I figured that it may not be a bad idea to do a combination of logic and some more applied mathematics. And so I thought of statistics and probability theory.
So that’s how I got started. I worked closely with Bill Weiss. He was one of my postdoc advisors. He's also a logician in the math department and was doing non-standard analysis.
When I got into the PhD program at the University of Toronto, I worked with Jeffrey Rosenthal in the stats department because he was working on Markov processes. And then in 2014, I met with Dan Roy and we started doing things in statistical decision theory.
So yeah, I think that's how I got started. And even today, it’s still my major research tool and it’s also my major research focus. I guess I came to statistics from math logic, so it's a bit of a different path compared to most of the other alumni.
What was it about U of T's PhD program and statistics specifically that made you want to continue there?
I did my undergrad here, and U of T is like my comfort zone. I like U of T and it will always have a special place in my heart, so that's obviously one of the reasons.
U of T also has an extremely strong group of set theory and logic researchers. Actually, I think it probably has one of the strongest groups of logicians outside of the US, and even if you wanted to count the US, there are at most two or three comparable schools.
Since I wanted to maintain a research record in this area, U of T seemed to be a natural choice. And also, at the time that I was applying for PhD programs, I was pretty busy with a research paper with Frank Tall, so I really didn't spend any time preparing for the GRE. Actually, I didn't take any of the GRE tests. That basically meant that I couldn't go to any of the American universities. So, for my PhD, I only applied to one school, and when I applied to U of T, I applied to both math and statistics.
They both gave me an offer, but I thought it might be easier to get a job if I have a PhD in statistics because at that time it was obvious that the more applied you are, the more likely you're going to get a job.
Let's say there's a student who is thinking of doing a PhD but doesn't know if it's right for them. What would you advise them?
I think, if someone wants to do a PhD, it should be goal oriented. For example, for me, I always wanted to do research and I always wanted to do research that I liked. And I liked the combination of logic and statistics.
But I also want to point out that nowadays in the job market it is extremely difficult for people to find a job if they’re doing pure theory. Even if your research is related to statistics, if it's very theoretical, it will be much harder to get a job compared to someone doing more applied stuff. So that's why I say if you choose to do a PhD -- because a PhD is four or five years, and it's a big effort -- you have to be goal-oriented.
If your goal is to go into industry or find a job in a reasonably good university, you have to think about your research carefully, right? So for example, you might want to do things like causal inference or biostatistics. Those are more applied research branches, which would bring you more opportunities when you finish your PhD. That would be my advice.
I think U of T stats is a very strong and very good department for this reason -- because it is very diverse. Basically whatever research you want to do, you will find the suitable supervisor here. So yeah, I would definitely encourage people to apply to U of T if they're interested in doing a PhD.
So you chose to do this kind of like, dual route where you did some theory, and also kept one foot in applied stuff. Now that you're on the other end of your PhD, do you ever work in industry projects at all? Or do you stay mostly within academia?
For me, almost never. But I would like to point out that the industry market is huge because after I finished my PhD, when I was doing my postdoc at UC Berkeley, I received quite a few emails from hedge funds and all these kinds of companies offering me interviews, and they tell you that the financial composition is huge and so on. So I think, if you do really well as a stats PhD and if you want to go to industry, there would be plenty of opportunities.
What was your postdoc like at Berkeley?
Oh, it was really good. Most of my PhD is based on this work from a math logic analysis, and there are maybe like, 10 to 20 people in the whole world who are well versed with this technique. And my postdoc advisor, Robert Anderson, is certainly one of them. He has a joint appointment in the Berkeley math department and the Berkeley economics department.
He was also my PhD dissertation external examiner. I think we really hit it off because we talked for maybe 15 minutes and he decided to offer me a postdoc at UC Berkeley. So I went to Berkeley and I worked with him. The first two years we were mostly working on Markov processes and so on.
So it was really working on stuff in my expertise. But towards my last year, I thought I should learn a bit. I wanted to expand my research expertise. I told Bob that I wanted to do some work in economics, so then he gave me some problems to work on. He also sent me to Johns Hopkins University to talk with Ali Khan, who is also a well-known economist. It was overall a very good experience at Berkeley.
It sounds like it.
Yeah, but I have to say UC Berkeley campus is no comparison to the U of T campus. The UC Berkeley campus is not very good looking. And if you're at UC Berkeley, you’ll once in a while you receive emails like “you guys have to be careful because there are violent crimes near the campus” and so on.
So the city is not like Toronto. I think, in terms of the city, people say San Francisco is great, but I like Toronto better. But in terms of the post experience, I thought it was a very good experience.
Do you have any long-term research goals?
Yeah, I think maybe for the next three or four years, my focus will still be in economics. For example, we have a big project using a general equilibrium theory to model climate change.
Classical general equilibrium theory was mostly developed before the seventies, and before that period, climate change and environmental problems were not really considered an issue. People were basically only concerned about the productivity of the economy and so on. But now it’s become a popular issue. We are thinking about coming up with a model that will cope with environmental issues and climate change and so on.
And other than that, I’m interested in economics. In economics people always say that people's behavior is rational and that we're influenced by price and so on. But really, a lot of people's behavior is influenced by social norms. We would like to conduct extensive studies in this, which would bring us to the field of behavioral economics, where you conduct social experiments to find out what people's preferences actually are and how people act under certain circumstances.
Yeah, so this is gonna be quite different from the research I have been doing because the research I've been doing has been all very theoretical. We hope to combine studies in behavior and experimental findings to come up with a solid theoretical framework that’s consistent with our findings to model a social norm. So that’s one of my goals in the next couple of years.
And so how did you get interested in behavioral economics?
Oh, I got interested in this because I read a paper by a famous game theorist, Ariel Rubenstein.
He argued that a lot of economists have been developing and working on models where people's behavior is characterized by the price or by their budget and so on, and these are of course correct, but he argues that in reality people's preference and people's behavior are also heavily influenced by the social norms.
He gives a very good example, which is that maybe all of us go to dinner and we order a cake. All of us want to have as much cake as possible. But the social norm says that if there's N people, then each of us shouldn't take more than N slices of the cake. So that's a social norm, and human behavior is bound by these social norms. So he developed like a novel but incomplete model on social norms.
The math is not difficult, and I think the model is pretty intriguing. So I spent some time on it and extended the model, and I think it's quite interesting. I’d like to use the model to think about the formation of social norms. I've also talked to several behavior economists in China and in the US and yeah, it seems to be pretty interesting to me.
Do you have any hobbies?
I like playing soccer. I'm a big soccer fan. I’ve been staying up quite late recently because of the World Cup.
Who's your World Cup team?
I don't really have a team, I just enjoy watching it. It could be Morocco, because my favorite player’s [Adele Taarabt] in Morocco, but he wasn’t on the national team this year. I also play Texas hold ‘em.
Oh nice! Are you any good at poker?
I think I'm fine, but gambling is technically illegal in China so you can't play, you know, that much. But I used to drive all the way to Niagara Falls or Casino Rama when I was in Toronto.
No way, that’s great.
I think a lot of professors in statistics or in math generally play Texas Hold ‘em.
Right, chance and probability and so on.
Yeah, it's not pure probability, but I think it's a very fun game because the rules are simple, but to actually be good at it, it’s quite sophisticated. And there's also this matter of luck. So in that way it’s more exciting than games like chess, which is purely deterministic.