Abstract
The assessment of credit default risk is a central task in credit risk management. Especially in corporate banking, financial statement data plays an important role, as it provides information on a company’s asset, financial and earnings position. At the same time, such data is historical and balance-sheet-date-related, which may limit its informative value for future defaults. Against this background, this master’s thesis examines the extent to which selected financial ratios and an AI-supported risk classification can contribute to the assessment of credit default risk.The empirical analysis is based on financial statement data of Austrian construction companies provided by KSV. The sample consists of 50 companies, including 25 insolvent and 25 solvent companies. For each company, three available annual financial statements were considered, resulting in a total of 150 company-year observations. Based on the available data, selected ratios relating to capital structure, liquidity, asset structure, earnings position and debt repayment capacity were calculated and compared between insolvent and solvent companies.
In addition, an AI-supported risk classification was carried out. Each annual financial statement was assessed separately using a standardized prompt. The AI assigned each observation to one of three risk categories: low, medium or high risk. It also provided a binary assessment of whether an increased credit default or insolvency risk existed for the following year. These assessments were then compared with the actual insolvency classification. The results show that the equity ratio, liquidity ratios, working capital and earnings position in particular differ noticeably between insolvent and solvent companies. The AI-supported risk classification also showed recognizable explanatory value, as the actual insolvency rate increased with higher risk categories. The binary AI forecast achieved high precision for positive risk warnings, but did not reliably identify all actual insolvency cases.
Overall, the thesis shows that traditional financial ratios remain an important basis for assessing credit default risk. An AI-supported risk classification can complement this analysis by structuring financial statement data and highlighting potential risk cases. For practical application in credit risk management, however, data quality, transparency, professional review and the combination with further credit-relevant information remain essential.
| Date of Award | 2026 |
|---|---|
| Original language | German (Austria) |
| Awarding Institution |
|
| Supervisor | Stefan Fink (Supervisor) |
Studyprogram
- Controlling, Accounting and Financial Management
Cite this
- Standard