Scott Cunningham is a professor of economics at Baylor University in Waco Texas. He has published in economics outlets such as The Review of Economic Studies, Journal of Urban Economics, Journal of Human Resources, Journal of Public Economics, Journal of Development Economics and more. He is the author of Causal Inference: the Mixtape published by Yale University Press in 2021 and co-editor of The Handbook on the Economics of Prostitution (with Manisha Shah) published by Oxford University Press in 2016.
His research focus covers a range of applied topics in health and labor, including sex work, abortion, drug policy, corrections, and mental healthcare. He has taught dozens of in-person and online workshops on causal inference and difference-in-differences to universities and firms across the world including eBay, Twitch, Facebook, HP, University of Oxford, London School of Economics, University of Pennsylvania and countless others.
He will teach the course "Recent Developments in Difference-in-Differences Estimation" at the Spring Seminar in Cologne in March 2024.
How did you become interested in your subject?
Scott: I got interested in causal inference because in graduate school, I wanted to study labor with David Mustard who had been a Gary Becker student. And I loved Becker a lot. I wanted to be just like Becker. A lot of us did and still do. But Becker mainly wrote applied economic theory papers, and I couldn’t do that. Still, we read his theory papers and then read the empirical papers, and very often the best ones were what we’d call causal. They’d use natural experiments in some ingenious way and I got hooked. My dissertation was very Beckerian — the effect of sex ratio imbalances in Black marriage markets due to high rates of Black male imprisonment and the impact that had on risky sex. But it was closely focused on trying to find some variation in the sex ratio that I could use to study the question.
And that really just got me going.
I kept being interested on the one hand in these Beckerian type questions related to risky sex, drug use, abortion — all these things people typically didn’t associate with economics — and on the other hand causal inference. I just was never satisfied with superficial knowledge of the material, but I also was not and never would be an econometrician. And so I had to try and somehow gain deeper understanding while at the same time not being very good at math. Not bad but not great. I came from a literature major background in college and so I was just always looking for some narrative or some metaphor that could help me. And it was just slow groping for a long time.
What lessons can participants draw from your GESIS course?
Scott: I hope we can together learn the importance of the crucial issues that drive all of causal inference which is the credibility of the assumptions underlying a research design like diff in diff or synthetic control. I also hope that you’ll understand how simple diff in diff is and I don’t mean it in a bad way. Instrumental variables is also simple. Many of these methods are simple and my hope is I can help you see that so you can enjoy it. And then I hope you feel empowered.
I want you to feel confident about your comprehension, and capable to do your own work.
What do you enjoy most about being a social scientist?
Scott: I enjoy that my intellectual curiosity is rewarded with the freedom and mandate to answer my own questions. I enjoy that if I see a problem I can go work on it. I love the 250 year history of the economics field too. I love the story of it — our tribe. I love that I belong to it.
What do you think is the most exciting recent development in your field?
Scott: I think the availability of datasets in really large volumes, already digitized, in contexts that you could use to study topics that no one else has before is very cool. I got to spend a decade studying internet sex work because the data existed to do it. That was and is amazing — to get a chance to study things that hadn’t been there before, offer up what you found, let someone else read it and see if it helps them. I do that now with suicide in prisons and jails. My state, Texas, has 120,000 inmates and 100 prisons. And we are using their data to predict using machine learning who’s likely to hurt themselves and then examine whether certain programs can stop it. I find that whole thing amazing. In causal inference I think really just the connecting of Angrist and Imbens at Harvard was great. That seemed to be a special encounter that left us all better off.
We thank Scott for his exciting insights and look forward to his course.