Interview with Stefano DellaVigna

Behavioural Science is a rapidly expanding field and everyday new research is being developed in academia, tested and implemented by practitioners in financial organisations, development agencies, government ‘nudge’ units and more. This interview is part of a series interviewing prominent people in the field. And in today's interview the answers are provided by Stefano DellaVigna.

Stefano is the Daniel Koshland, Sr. Distinguished Professor of Economics and Professor of Business Administration at the Haas School of Business, UC Berkeley. He is a Fellow of the American Academy of Arts and Sciences and of the Econometric Society, an Alfred P. Sloan Fellow (2008-10), and a Distinguished Teaching Award winner (2008). He specializes in Behavioural Economics and has published in international journals such as the American Economic Review, the Journal of Political Economy, and the Quarterly Journal of Economics. He has been a co-editor of the American Economic Review, one of the leading journals in economics, from 2017 to 2023.

---

How did you get into behavioral science? My undergrad program in economics taught mostly pretty standard economics. There was lots of math, statistics and macro, but before taking all of these, there was a class in the philosophy of science. We learned about Karl Popper, Thomas Kuhn, how to (dis)prove a theory and the empirical revolution. But then at some point, the teacher said: ‘there's one thing that is important above all, which is to not believe anything they'll tell you in the economics classes. They will tell you people maximize utility. They do all these optimizing choices. And it's all totally wrong.’ So, keep in mind, we haven't even taken those classes. And this professor comes in and says instead what you should believe you should read Kahneman and Tversky and Herbert Simon. And this was in ‘92/’93, so actually pretty early. This professor (Riccardo Viale) was right on a lot of it. I especially fell in love with Herbert Simon. His autobiography, ‘Models of my life’ is one of my favorite readings.

Herbert actually came to give a lecture series in Bocconi. And I remember the first time he gave four lectures. In lecture 1, there were 200 people in the room, but by the 4th lecture, there were only 20 of us. And it was really interesting to see him pioneer the topic.  Then I took the standard economics classes and I was not buying into any of it at the beginning. For my undergraduate thesis, I definitely wanted to do behavioral economics and I tried to do some of it and then by the time I got into my PhD, that's really all I wanted to do. I want to understand human behavior. It was it was a very lucky coincidence that this philosophy of science class turned out to be indoctrination into early behavioral economics, and it left a lasting impression in me.

What is the achievement that you are proudest of in terms of your career?

I would say kind of two things that go together. One thing is that I have this tendency to get ideas in different areas. So I studied crime and charitable giving and voting and asset prices. And that is not an advisable career path – I wouldn’t recommend it to people. It's very difficult because every time you’re restarting - it just really keeps you learning and it keeps you engaged. So I am very grateful to behavioral economics that it gave me an excuse to be able to do that. Because if you're a labor economist, you can't do that.  But behavioral economics kind of tells you so long as it's behavioral in some way, you can do it. So I feel lucky and proud that in a special department like the economics department here in Berkeley and in our special community of behavioral economists we can do this.

And related to that, I feel proud of the work I've done in what you might call empirical behavioral economics. Which is using data to study be behavioral phenomena. And that data can be gained through anything, experimental, observational, field data. Things such as voting data for example. Trying to show that you could add those behavioral insights across fields that didn’t use to have such easy accessible data. That's really kind of the overarching agenda that still motivates me to this day.

---

You said that the journey that you've had into your own career, you wouldn't really recommend if someone were to start now. What would you recommend someone starting out in behavioral science? I try to recommend to people to follow their passion and to really understand where there passion lies. What motivates them? What is the intersection between the ideas and the skills they have and what they feel motivated to do, but also what the discipline will take to.

And so my main advice can sound completely trivial, but it goes back to the Greek that said ‘know yourself’. And so, for example, I've never been as fascinated by supply and demand and that's fine, I learned that about myself. But another thing I learned about myself is that I lean on the obsessive compulsive side. I just, cannot have 10 emails sit in my inbox. That comes with pluses and minuses. When I write things, I tend to be overly organized; boring often. But if I know that, I can work with it. If a co-author tells me ’this kind of goes a little overboard’, I realise that’s just my OCD. It took me years to figure out what it was. Similarly, where was I getting my ideas? Some people say they get their best ideas in the shower, but that doesn't work for me. My best ideas come from when I confront another brain; if I'm talking to another person and I see things from their perspective and vice versa. It's really helpful if you learn who you are, how you work and you can lean into it and be aware of when it becomes a limitation to your growth and be aware of how it can be a weapon.

Knowing your advice, and given your experience, what do you think makes the real difference between a good and a great behavioral scientist in terms of skills?

There are some skills that I think are really good to have. I think every behavioral scientist needs to have a good understanding of statistics. Understand when something is (under)powered, be able to spot things that are errors. Really early on in my career I had ideas I was so excited about, but to be honest, they were going to be statistically underpowered on the particular data sets I was working with. There was no chance. So I shouldn't have done them. And that's the lesson I learned later. So I think just having an idea of what it means to have confounding factors, how the standard errors are correlated or what the variation in your data is. I think that is a great thing that we owe to statisticians.

The second is, through understanding statistics, having realistic expectations of what can be ascertained from the data. I think we learn a lot from having models in the sense that often they give us a sense of magnitude. For example, if I say I'm gonna cut prices by 10 percent and the quantity sold goes up by 4%, I have a sense of an elasticity of like minus 0.4. That's actually in line with what you'd expect. And I wouldn't say 4% is small or not statistically significant, but it's in the range of what you might expect. The work I've done on nudge units, with the idea that you would do a fairly simple, non-invasive intervention and get an effect size that could be one half of a standard deviation, that's just really implausible. Most interventions, unless they're big default changes, will probably not even be able to move the scales much. And so I think having a sense of what are plausible magnitudes is something where models help. Because if you expect a huge effect size, you underpower your study. So you run it with the population too small and then things are very noisy and there is publication bias, and then you end up publishing results that may not replicate. So I think that it's just this is this kind of logic comes out of, you know, my work with the wonderful Elizabeth Linos and the equally wonderful Woojin Kim.

---

Given where we're at now, what realistically do you see as a great challenge for behavioral science to continue growing as a field? 

I think behavioral science in general has done a lot of things right. It embraced a broader view of human behavior, compared to where most economists stood a decade ago. It engages with a variety of important policy relevant outcomes, like household financial choices and has luckily triggered a lot of enthusiasm.

At the moment that the recognition came, mostly you can think about this period as after the Kahneman and Tversky work, and of course the 2017 Nobel Prize for Richard Thaler. But now you run the risk that people think that anything behavioral economists do will like change the world. Now that's not gonna happen. No behavioral economist will tell you that if you read ‘Nudge’ it’ll fix the world overnight. Addressing global warming, for example, is one of the biggest challenges for the world. Nudges will not fix that by themselves. We're gonna need carbon taxes and non-behavioral measures. In addition to trying all of that, you also are gonna get some help from social comparison. So, we should do those too. But nobody has said that we're gonna be able to drastically reduce CO2 emissions, by just having a couple nudges. So I think having a realistic view of what can be achieved is really important. Which is in the spirit of the work that Elizabeth and I have done is to say ‘look, nudges are effective; the ones we looked at increased outcomes by one and a half percentage point, which is 8 percent off of a baseline of 20 percentage point for whether it's paying taxes in time, et cetera.’ But it's not what you might have expected reading the literature.

So now what? You can probably get bigger effect sizes with bigger interventions. So here is an example that I think Richard Thaler calls an ‘ultra’ nudge, which is the work that Raj Chetty and co-authors have done with the ‘Moving to Opportunities’ (MTO) program, which is one of the greatest programs in the world. In the U.S. instead of doing public housing, there’s vouchers for low income people to move to better neighborhoods and set up for a better life. The thing is, a lot of people didn't take them up. They’d get the vouchers and then they don't really find the house. Now that can be  because maybe some landlords are suspicious. So doing a nudge and explaining things more clearly does help. But it has an effect of one or two percentage points. On the other hand, having the ‘ultra’ nudge, which is basically having a staff person. who basically helps you to make the phone call to find the landlord, to contact the lender, to figure out what could be areas for you to go to, that is a tremendous thing. That doubled the rate of placement: a huge effect size. It's not cheap like a nudge. But think of the value in terms of happiness; just think of the earnings opportunities.

So if we can do more, well, let's talk about what ‘more’ would be. It could be more out of the box. It could cost more. So I think that's the future, trying to be honest about the magnitude that one can get. That means finding out that some things are moderately effective, at a very low cost. And then let's keep going and find out what the things that are very effective could be.

Looking forward, what is it that you would still like to achieve?

First, I'm really excited about continuing on the work with Elizabeth Linos and Woojin Kim. Expanding our understanding of evidence based policy. There are a lot of dimensions to this, in terms of bottlenecks. What haven’t we been able to run? Why not? And then if we were able to run something, what are we doing with the evidence? I think it's gonna take a few years to further develop this. We're excited to look into this.

Second, the other thing is something that is happening in economics more broadly, especially in laboratory experiments: we're figuring out that a lot of behavioral biases or even what look like behavioral preferences, even reference dependence to some extent, is just is the noisy perception of the world. For example, I read a contract and I kind of get, I understand only some parts of it. Or as per economic experiment, I see a choice between a lottery and a sure thing. And I have this vague understanding of what a lottery means. And so when you model that, you actually do a lot of progress on understanding behavioral biases. Some people call it noisy encoding, and I'm doing some work that is somewhat related to that, with my colleagues, David Card and Dmitry Taubinsky. There’s a strong focus on decision time here. For example, when you sent me the invitation for doing this interview, I looked at some previous interviews and said ‘yes’, right away. I didn't wait too long, but for some other things, I’d feel the need to ponder it; I might be on the fence. And so basically the decision time conveys some information on strength of preferences. There's a lot more to that, but we're trying to take that into the decision making of experts.

Those are, I think, the two things that I'm excited for the future, as well as continuing the work I've done on forecasts of research results. Which is part of a kind of a different agenda that I have that is methodological. As one of the most frustrating things that can happen to us scientists is that we work for years, and then we present our results and somebody says ‘we kind of knew that already’. It’s dismissive. Thing is, most of the time we didn't know it already! And so how do you counteract that is actually by surveying people when the research results are not yet known, but saying this is what we're going to do. This is our design. What do you think we'll find? Collect the predictions of experts. And then present the actual findings.

So I think kind of all of these three things are constellations of things that I'm working on.

---

The experience that you've just outlined where results are being dismissed as common sense, is that one of your biggest frustrations with the overall field?

I think I'm an optimist, so I see a lot of good things, but this one I feel good about because in some sense we're making progress on it. I feel like we've made a lot of progress on publication bias as well. So those frustrations have gone. The one that remains is regarding diversity. Is the group of behavioral economics representative enough? I don't think that we necessarily have as good a representation of minorities or even women as we definitely want to have. And so you ask yourself questions like ‘are we deterring people?’ At Berkeley, we've always tried to create a positive environment. For example, in our seminars it's not okay to show off or be tough, just try to be constructive. Be less deterring.

Would you say you apply behavioral science to your own life?

I have implicit commitment devices. I set deadlines but know that I’m probably going to be naive about it and set too many deadlines. And then regret it.

But in general I try to know myself, as mentioned earlier. Understanding what kind of type you are, certainly as an advisor, as a mentor. Try to have empathy and also try to identify what kind of help students need because different students need very different help. Some people don't need deadlines. Some people need deadlines the most and some people need to be encouraged and some people get tend to be discouraged because they get into too many projects. It’s important to be a good psychologist to help support people best.Looking back, what would have happened if you hadn't actually been in this philosophy of science class? If you hadn't come into touch with behavioral science?It's a good question. I think in light of what I know now, I also would have loved to be a statistician. I love engaging with data. I think there is just something magical about looking at a data set and trying to figure out the patterns. Basically reading the world.

Last question:  who was or were the most inspiring people in your journey into behavioral sciences? 

A lot of the people that you interviewed would fit in that category. It was so great to see people like Colin Camerer, Katy Milkman, Elisabeth Linos. I don’t think you've gotten to talk to David Laibson, who has always done amazing work. And then Ryan Oprea at Santa Barbara and does some of his work on this cognitive based understanding of the world, like noisy encoding I mentioned before. And then somebody else that I would flag among the young researchers is Kirby Nielsen at Caltech, who has a way of asking some tough questions and using the laboratory to make progress on them.

---

Thank you so much for taking the time to answer my questions Stefano!

As I said before, this interview is part of a larger series which can also be found here on the blog. Make sure you don't miss any of those, nor any of the upcoming interviews!

Keep your eye on Money on the Mind!