Modern empirical research in economics and finance uses computer intensive methods for the analysis of complex data sets. The estimates are based on algorithmic methods to cope with a whopping number of covariates and possible model specifications. The problem is prevalent e.g. in macroeconomic forecasting, in financial econometrics (e.g. portfolio and risk management), and in the evaluation of causal treatment effects of public policies. Consequently, empirical studies in economics and finance rely more and more on machine learning techniques (e.g. lassoing, support vector machines, artificial neuronal nets). However, the algorithmic nature of the estimation results is non-standard and differs from conventional statistical tests.
The goal of this seminar is to acquaint students with the necessary toolbox of machine learning techniques. Students must write an empirical research paper in which they apply a novel and/or advanced method to shed more light on a real-world problem.
The seminar will take place on 1-2 August 2022 (Monday and Tuesday).
Submission of Seminar Papers
- Send your paper as pdf to Verena.Kretz@uni-konstanz.de no later than Friday September 30th, 2022, 08:00 pm
- Zip your source code and the data and upload the zipped file to the online storage service of the university: Cloud Uni Konstanz . Please send the link for the upload to Verena Kretz no later than Friday September 30th, 2022, 08:00 pm
Time and Date
The seminar will take place as a block seminar on the following two days:
- August,1. & 2. 2022, 9:00 - 18:00 in room F425
Econometrics I and Advanced Econometrics (or equivalent). Knowledge of Financial Econometrics; Microeconometrics and/or Time series Analysis is desirable. We expect that students have a decent programming background in either MatLab, R or Python or at least willing to invest sufficient effort to learn one of these languages.
- Kick off Meeting: Thursday, 13 April 2022, 17:00 - 18:30, F 423
- ECTS: 6 credits
- Seminar participants must decide on the topics until May 1, 2022
- There will be topics for which joint work of two participants may be possible.
- For more information contact Winfried Pohlmeier, F319, Tel. 2660, Winfried.Pohlmeier@uni-konstanz.de
Causal Returns to Schooling: Evidence from Machine Learning
Undetected Heterogeneity in Schooling: Evidence from Machine Learning
Hypothesis Testing in the case of many and many Weak Instruments: The Angrist-Krueger study revisited.
How strong is the “Bla Bla?” in Economics and Finance: An analysis to study the language quality of reports in economics and finance.
Which animals are in the factor zoo? Evidence from Index prediction by machine learning methods
Learning from Machines about Portfolio Allocation
Learning from Machines about Financial Risks: Bagging the VaR and ES
Tracking Portfolios by the Lasso: Which Stopping Rule Should be Applied?
Tracking Portfolios by the Lasso: Accounting for Transaction Costs
It’s all in the timing: A generalized ridging strategy to outperform popular portfolio strategies
Team Captain or Cheer Leader: Who is central in School Classes?
Are women the better peers? Accounting for Gender Heterogeneity in Networks
The topics above should be considered only as first suggestions and are supposed to give interested students a flavor of the seminar’s content. We will assign the final topics according to the student’s background and interests.