Assistant Professor Florian Gunsilius arrived on the University of Michigan campus during the early stages of the Coronavirus outbreak. Originally hailing from the small village of Denkendorf, Germany, he completed four years studying Management, Philosophy, and Economics at the Frankfurt School of Finance and Management in Frankfurt am Main before completing his education in the United States. Most recently, Dr. Gunsilius was a Visiting Postdoctoral Associate at the Department of Economics at MIT. Just prior to his stint in Massachusetts, he graduated from Brown University with a Ph.D. in economics, receiving the Joukowsky family foundation dissertation award in the social sciences.
Drawn to the University of Michigan for its renowned research and scholarship, Professor Gunsilius has found Ann Arbor to be a gem, even if the coronavirus has put a damper on his explorations. “It is much bigger than the other campuses where I have been, and I can honestly say that it is one of the nicest places I have lived so far. It is quiet enough, but has all the amenities one would expect. And the surrounding nature is gorgeous.”
Professor Gunsilius is currently interested in geometric and nonlinear approaches to causal inference. In his latest paper, Distributional Synthetic Controls, he extends a classical approach for obtaining causal effects in policy evaluation, called ‘synthetic controls,’ to individual effects. “When trying to assess the impact of some policy change on the population in Michigan, say, we can use a combination of other states where the policy has not been put into place, like Wisconsin, Washington, Colorado, etc. in order to obtain a ‘synthetic Michigan’ which is the hypothetical Michigan had it not had the policy change. This is the classical idea of synthetic controls which has been introduced roughly 10 years ago. The classical method can only provide an average of the causal effect over all individuals living in Michigan, however. Oftentimes, one is interested in how a policy affects only a subset of all families, like the poorest families. My new method allows the researcher to look at that.”
This fall, Dr. Gunsilius is teaching Econometric Analysis I (ECON 671/STATS 505). It is designed as the first rigorous course for graduate students to lay the foundation for econometric analysis. Next semester, he will be teaching a graduate topics class, “Advanced Econometrics II” (ECON 679), where the class will dive into nonparametric approaches in Econometrics, Statistics, and Machine Learning. The emphasis is to expose students to the mathematical underpinnings in these areas. When asked about one thing that he would like students to thoroughly understand, he said, “I would like students to understand the underlying intuition of what a linear regression is. Linear regression, for better or worse, is still the main workhorse in applied economics, and I think it is important that students understand its properties, and especially its limitations.”
If you’d like to learn more about Professor Florian Gunsilius, please click here.