Corporate Credit Risk Analysis Utilizing Textual User Generated Content - A Twitter Based Feasibility Study

Irrecoverable receivables resulting from insolvent debtors endanger the own liquidity. Therefore, corporate credit risk analysis should be continuously improved in order to diminish bad debt. We analyse in how far user generated content (UGC) contains evidence concerning the financial stability of companies and hence, can enhance the information base for corporate credit risk analysis. For this purpose, we compare data from the microblogging platform Twitter related to ten insolvent and ten solvent German companies. We utilize techniques from content analysis for the quantification of textual data. Results from independent t-tests indicate, that the amount of UGC is significantly higher and the sentiment is significantly worse for insolvent companies prior to the date of bankruptcy than for solvent companies in the same time span. Furthermore, we apply the k-Nearest-Neighbour algorithm in order to classify companies as prospectively insolvent or solvent based on sentiment scores derived from UGC. Results show, that a classification accuracy above randomly expected values can be achieved. The classification accuracy increases, when UGC published closer to the date of insolvency is used. Future research should focus on how to utilize our findings and improve processes of corporate credit risk analysis while integrating UGC