Philip Rademacher

My research interests are very diverse and range from educational economics to economic policy. However, all of them have a common denominator: they are tackled empirically and data-based. I like to apply standard econometric techniques, but also focus on the implementation of new methods (like machine learning) and the utilization of other datatypes (e.g. unstructured text data).

The generation of scientific knowledge in the humanities is predominantly based on written sources. Quantitative research approaches are strongly underrepresented, although data on various potential research topics is available (in particular for economic questions). Such research gaps are exciting because quantitative approaches may yield different results than their qualitative counterparts.

Forecasting is actually not far away from economic history. The idea is to use data from the past to make a reliable prediction for the future. In recent years machine learning (ML) models were able to improve predictions in many areas. I am not only interested in the hunt for the best forecasts, but I want to take a closer look: Why are some ML-models better than econometric models? Can economists improve their theoretical understanding with the help of those models?