Extracting Key Phrases as Predictors of Corporate Bankruptcy: Empirical Analysis of Annual Reports by Text Mining

Bankruptcy predictions have been one of the most interesting topics for accounting researchers. Most bankruptcy prediction models are developed by using financial ratios. However, signs of the changing financial position of a company may appear in the nonfinancial information earlier than we can identify the changes in the financial numbers. In recent years, analysis of qualitative information has become remarkably important, because frequent changes in accounting standards have made it difficult to compare financial numbers between years. In this study, we analyzed the sentences in financial reports in Japan and extracted key phrases/descriptions to predict bankruptcy. Our research revealed that if some particular expressions appear together with the word “dividend” or “retained earnings” in the same section of an annual report, they were effective in distinguishing between bankrupt companies and non-bankrupt companies.