Bankruptcy prediction and the discriminatory power of annual reports: empirical evidence from financially distressed German companies

The structural and linguistic characteristics of companies’ annual reports (e.g., their length, complexity, and linguistic tone) and the qualitative information they contain (e.g., on the risks a company potentially faces) provide useful insights that can help increase the accuracy of predicting bankruptcy. In this study we use a sample of German companies that we compiled through propensity score matching to examine what type of textual information allows us to discriminate accurately between companies that are likely to go bankrupt and companies that, although financially distressed, are likely to remain solvent. Our findings provide empirical evidence that both the structural and linguistic characteristics of annual reports and the qualitative information they contain help discriminate between companies that are effectively bankrupt and companies that are solvent but financially distressed. Furthermore, this study provides empirical evidence that the “management obfuscation hypothesis” is valid because the tone of annual reports produced by bankrupt companies is quantifiably less negative than that of reports produced by companies that, although financially distressed, are likely to remain solvent.