Abstract: This study evaluates whether feature selection improves machine learning forecasts of German business cycles. Using a high-dimensional dataset with 73 indicators, primarily from the OECD Main Economic Indicator Database, covering a period from 1973 to 2023, Sequential Floating Forward Selection (SFFS) is applied to build compact, explainable, and performant models. The focus is on regularized regression models (LASSO, Ridge, Elastic Net) and tree-based classification models (Random Forest, Gradient Boosting and AdaBoost). SFFS yields models with up to eleven indicators that outperform a standard term-spread probit model—especially during Quantitative Easing. Regularized regressions provide the most accurate recession signals. Feature selection increased the forecasting power of tree-based models, while marginally reducing the performance of regression models. The findings contribute to the ongoing discussion on the use of machine learning in economic forecasting, especially in the context of limited and imbalanced data.

Conferences: I presented this paper during the Workshop “Forecasting in Times of Structural Change and Uncertainty” at the Halle Institute for Economic Research (2024) and during the Doctoral Research Workshop at the University of Düsseldorf (2024).

Discussion Paper: An earlier version of this paper can be found in the DICE Discussion Paper Series.

Abstract: The Coin Hoards of the Roman Empire database currently contains more than 18,000 coin hoards from all areas of the Roman Empire and beyond. This large data collection encourages data-driven approaches to research the ancient world. However, there are several sources of bias in the data that make data-driven research difficult. One problem is that the hoards are recorded differently. For some hoards, archaeological information is available; for others, only numismatic; and for others, not even the exact location of the find is known. This paper uses logistic regression to analyse whether the data quality correlates with some observable hoard characteristics. It can be shown that numismatic data tends to be superior for hoards from the first century AD. In addition, there are significant differences in data quality between hoards from ancient cities and forts versus those in rural areas or outside the Roman provinces.

Conferences: I presented this paper during the FORGE conference at the University of Tübingen (2023). The conference abstract (written in German) can be found here. In addition, I gave a lecture with the title “Make Bias visible!” at the Berlin-Brandenburg Academy of Sciences and Humanities (2025), that discussed the research findings in detail.

This policy report is written in German.

Summary: This policy report analyzes Germany’s decarbonisation at the NUTS‑3 (district) level by constructing regional CO₂ emission data through a shift‑share method based on sectoral emissions and local employment shares. It examines how regional economic growth relates to emission reductions from 2000–2019 and finds a clear negative correlation: districts that reduced emissions more strongly tended to experience weaker growth in employment and GDP. This pattern reflects cases like the Ruhr region, where decarbonisation resulted mainly from the decline of highly emission‑intensive industries rather than from sustainable structural transformation—an outcome that conflicts with broader political and societal goals.

Media Coverage: Many national newspapers in Germany have written about the report, e.g., SPIEGEL, FOCUS, Mitteldeutsche Zeitung, Süddeutsche Zeitung.