Contributions and applications in loan loss provisioning, stress testing, and visual analytics

Abstract

This cumulative dissertation summarizes and discusses six research articles that are either published in academic journals and conference proceedings or submitted for review. The topics described are cross-disciplinary and can be allocated to Accounting, Finance, and Information Systems Research. In Accounting, we analyze the methodological differences between ratings and lifetime default risk to develop a proof for the use of rating changes for the determination of significant increases in credit risk in accordance to the impairment requirements of the International Financial Reporting Standards. Our results and findings contribute to more transparency with regard to decision-relevant information for stakeholders of financial statements. In Finance, we combine machine learning techniques with cointegration analysis to produce adequate projections of macroeconomic variables for stress testing exercises. Our results and findings have practical relevance for risk managers in the financial services industry and help to validate the execution of stress tests and to ensure compliance. In Information Systems Research, we develop a general process model and visualization framework to identify and highlight unusual data in subsets for further investigation. Our process model and visualization framework empower domain experts and data analysts to jointly gain and discuss insights from underlying data. Our results and findings show that both our process model and visualization framework contribute to interactive visual analytics, storytelling, and well-founded decision support.