查看文章

Understanding how quantitative and qualita-tive data affect corporate bankruptcy prediction is of great significance to all firm's stakeholders. Particularly, annual report narratives are the main official textual releases through which management can signal inside information to their stakeholders. Nevertheless, there is a lack of research integrating different parts of annual report narratives together with financial and market data for bankruptcy prediction. Therefore, this research aims to investigate the effect of integrating information from different sources (ie, Management Discussion and Analysis (MD&A), auditor report, dividend policy section, financial and market variables) on the overall accuracy of the corporate bankruptcy prediction model. To this end, two main models are suggested for corporate bankruptcy prediction. First, hazard models which will be used to evaluate the incremental ability of narrative disclosures to predict bankruptcy beyond quantitative variables. Second, a hybrid machine learning model; where Genetic Algorithm (GA) will be applied to choose features for Support Vector Machines (SVM) classification algorithm. Finally, these integrated models, based on quantitative and qualitative variables, will be compared with a well-established benchmark model; namely, the Shumway (2001) hazard model that combines financial and market variables.