A public firm's bankruptcy prediction is an important financial researchproblem because of the security price downside risks. Traditional methods relyon accounting metrics that suffer from shortcomings like window dressing andretrospective focus. While disclosure text-based metrics overcome some of theseissues, current methods excessively focus on disclosure tone and sentiment.There is a requirement to relate meaningful signals in the disclosure text tofinancial outcomes and quantify the disclosure text data. This work proposes anew distress dictionary based on the sentences used by managers in explainingfinancial status. It demonstrates the significant differences in linguisticfeatures between bankrupt and non-bankrupt firms. Further, using a large sampleof 500 bankrupt firms, it builds predictive models and compares the performanceagainst two dictionaries used in financial text analysis. This research showsthat the proposed stress dictionary captures unique information fromdisclosures and the predictive models based on its features have the highestaccuracy.