Teaching

Resources

Courses I Teach/Taught

The following descriptions refer to the courses’ most recent versions. They evolved (slightly) over time.

This is the first of two compulsory econometrics courses for first-year master students. It starts by discussing the basics of probability theory and (classical) statistical inference (point estimation, uncertainty quantification and hypothesis testing, in finite samples and asymptotically). It then proceeds to the ordinary least squares estimation of the linear regression model, including a treatment of endogeneity and instrumental variables as a possible remedy. Finally, the course discusses maximum likelihood estimation and applies it to the linear regression model and to models of binary and censored outcomes. This includes a discussion of numerical optimization methods and asymptotic properties obtained via extremum estimation theory. Assessment is based on two exams and bi-weekly problem sets. By analyzing foundational models and concepts in-depth and emphasizing the practical (numerical) implementation of inference procedures, the course allows students to easily apply their knowledge to non-standard settings, tailored to their application of interest.

This is the second of two compulsory econometrics courses for first-year master students. It starts with a brief treatment of causal inference, introducing the potential outcomes framework and the concept of an ideal experiment. The focus is on comparing non-parametric causal inference approaches and parametric models like the linear regression. Next, the course discusses inference with time series data. It covers time series regressions, univariate time series models as well as reduced-form and structural vector autoregressions. The final part of the course is devoted to the basics of panel data analysis (incidental parameters problem, pooled OLS, fixed effects, random effects). Assessment is based on two exams and bi-weekly problem sets. By analyzing foundational models and concepts in-depth and emphasizing the practical (numerical) implementation of inference procedures, the course allows students to easily apply their knowledge to non-standard settings, tailored to their application of interest.

This course discusses further topics in econometrics, building on the foundational concepts introduced in two compulsory econometrics courses for master students. First, it introduces Bayesian inference and applies it to the linear regression model (yielding Ridge- and Lasso-estimation and providing the basis for machine learning methods), to panel data models (correlated random effects) and to autoregressions. This includes a discussion of model selection and numerical sampling methods. Second, the course treats multivariate and nonlinear time series models (incl. vector autoregressions, dynamic factor models and models with time-varying parameters like regime-switching-, stochastic volatility- and conditional heteroskedasticity-models), as part of which it discusses cointegration and state space model estimation. Third, the course covers causal inference methods, both in the context of randomized controlled trials and natural experiments. As part of this, it discusses regression discontinuity designs, matching methods and difference-in-differences estimation. Time permitting, the course may briefly cover Machine Learning methods and non-parametric regressions (kernel smoothing methods, regression trees and random forests, neural networks). Assessment is based on problem sets and an individual project, where students apply a method from the course to their application of interest.

Building on the compulsory course “Statistical Literacy”, this course introduces the fundamentals of econometrics. It discusses OLS regressions and the basics of panel data techniques, which are commonly used by social scientists to answer real world questions. The emphasis lies on understanding the assumptions behind these methods, on developing the skills to read and understand quantitative academic papers, and on analyzing interesting datasets using statistical software.