Decision Support for Credit Risk Management Using Alternative Data
von Jan Roeder
Datum der mündl. Prüfung:2022-11-17
Erschienen:2023-02-16
Betreuer:Prof. Dr. Jan Muntermann
Gutachter:Prof. Dr. Matthias Schumann
Gutachter:Prof. Dr. Lutz M. Kolbe
Dateien
Name:Dissertation Roeder 2022.pdf
Size:6.48Mb
Format:PDF
Description:Dissertation
Diese Datei ist bis 01.06.2024 gesperrt.
Zusammenfassung
Englisch
The digitization of economic activity of individuals and organizations has been a significant trend in recent decades. This shift creates incentives to collect and process data relevant to the information needs of diverse market participants. In many cases, the captured data no longer correspond to a strict structure as they are semi- or even unstructured. Such data streams may contain signals not captured in established data sources, creating a wide range of opportunities to support the decision making process in various industries. Credit risk management is positioned to be particularly impacted by this development. The information asymmetry between lender and borrower drives the demand for additional signals that can enhance risk assessment. It is also a field of far-reaching significance since the existence of credit risk is deeply embedded into our economic system. The challenge of using alternative data sources lies in extracting relevant signals and linking those to credit risk. Because the goal is to assist decision making, the chosen approach should be interpretable. Additionally, established procedures and findings from the field of credit risk management must be considered. Although existing research offers suggestions for addressing the emerging challenges, a unified and integrated approach has not yet been determined and established. This cumulative dissertation investigates the use of alternative data sources for decision support in credit risk management. The thesis consists of five research studies, each of which belongs to one of the following two research areas. Research Area I is dedicated to systematizing and structuring the subject area. For this purpose, the first study conducts a literature review. It serves as a basis for identifying research gaps and deriving a research agenda. The second study develops a taxonomy to highlight aspects of data heterogeneity. Research Area II consists of empirical analyses that examine how alternative data can be utilized to support decisions in the context of credit risk management. Data sets, such as analyst reports and financial news, represent the foundation for the three studies. Text mining techniques are applied to extract signals that are linked to credit risk measures. An overarching consideration throughout the studies is the interpretability of the applied statistical and machine learning models. The findings indicate a relationship between credit risk and the textual signals that originate from analyst reports and financial news.
Keywords: Credit Risk; Data-Driven Decision-Making; Alternative Data; Decision Support System; Credit Risk Management; Unstructured Data; Text Mining; Information Value Chain; Analyst Reports; Financial News; Sentiment Analysis; Topic Mining; Sentiment Dictionary; Credit Default Swap Spread