Each topic in this seminar consists of a recent paper at the frontier of current research. All papers are applied papers that focus on exciting applications over a broad range of topics. Each paper also takes you to the methodolgical frontier by using data analysis methods that slightly extend the basic methods you should have encountered in a previous lecture course (see requirements below). The idea is to learn about the topics, but also understand a few new methods. The new methods are close enough to what you know already that everybody should be able to understand them after the seminar. 

The Seminar requirements consist of a paper, a presentation, and active participation in class discussion. The presentations will be blocked in June/July. I am flexible regarding timing and will decide together with the participants when the presentations will take place. 

Applications for seminars is organized via web-based central seminar matching. Please sign in with your university mail account on following website:

http://econ.mathematik.uni-ulm.de:3838/semapps/stud_de/

Relevant deadlines for application are displayed within the tool. If you urgently require a place in a seminar it is recommended to sign in in several lists. According to the timeline you will receive feedback which seminar you are registered for.

We have seminar places for up to 15 master students.

The application concerns the seminar in general, there is no application for specific seminar topics possible! After you are registered for the seminar, topics will be allocated taking into account student’s preference. You will be informed about this and all further steps in more detail via e-mail after you are registered for the seminar.

 Ths seminar requires a basic understanding of prediction and machine learning methods (cross-validation, LASSO, random forests) and/or methods of causal inference (treatment effects, diff-in-diff sn instrumetal variable estimation) on the level that is acquired e.g. in the classes  Data Analysis (Gebhardt) or Market Analysis with Econometrics and Machine Learning (Kranz).

In this semester's seminar we focus on methods at the intersection of causal inference and machine learning. In particular, approaches that use machine learning methods to otpimaly choose (complex functions of) covariates to estimate treatment effects. These approaches are especially useful for a later career outside academia. In such a career, you will often be asked to estimate a causal effect even though you have neither proper randomization, nor instrumental variables, nor regression discontinuities and maybe not even panel data. In these situations, you must rely on covariates to get close to a causal effect. These methods can help you to make the best of such a situation.

The focus of this semester's seminar will not be the theory and proofs behind these methods, but each participant will get (possibly different, depending on the number of participants) data sets and we will all apply the same methods to each of them and discuss the outcome. The data sets may be drawn from different fields (labor, finance...) and you are welcome to make suggestions.

For writing the seminar paper please follow the guidlines  Richtlinien (Prof. Gebhardt). If you are interested in writing your seminar paper with LaTex, you may find a short introduction here.

The Seminar requirements consist of a paper, a presentation, and active participation in class discussion. The presentations will be blocked in June/July. I am flexible regarding timing and will decide together with the participants when the presentations will take place.