Content analysis of XBRL filings as an efficient supplement of bankruptcy prediction? Empirical evidence based on US GAAP annual reports

Most of the bankruptcy prediction models developed so far have in common that they are based on quantitative data or more precisely financial ratios. However, useful information can be lost when disregarding soft information. In this work, we develop an automated content analysis technique to assess the bankruptcy risk of companies using XBRL tags. We develop a list of potential red flags based on the U.S. GAAP taxonomy and assign the elements to 2 categories and 7 subcategories. Then we test our red flag item list based on U.S. GAAP annual reports of 26 companies with Chapter 11 bankruptcy filings and a control group. The empirical results show that in total, the red flag item list has predictive power of bankruptcy risk. Logistic regression results also show that the predictive power increases the nearer the bankruptcy filing date approaches. We furthermore observe that the category 2 red flags (bankruptcy characteristics and influencing factors) have higher discriminatory power than category 1 red flags (earnings management indicators) for one year before the bankruptcy filing date. This difference narrows for two years before the bankruptcy filing date and may turn in favor of category 1 red flags for three years before the bankruptcy filing date.