A critical examination of three methodological issues – accounting information (ratios) used, sampling procedures, and expected cost of prediction errors – related to corporate bankruptcy prediction models and their implications to the model's usefulness

In this paper we examine three methodological issues associated with corporate bankruptcy prediction models. Literature review reveals that there are six groups of accounting ratios used to predict bankruptcy. Despite their shortcomings, ratios and, more specifically, traditional ratios are the most popular input of predictive models while the use of other forms of information, such as qualitative variables, faces major computational and interpretational limitations. In the vast majority of cases, models are based on non-random samples and produce biased predictions. Despite the solutions proposed by the relevant literature, this problem still remains unresolved. Another major issue related to corporate bankruptcy prediction is the expected cost of prediction errors. Although this cost is often difficult to compute, users with different attributes may find different models more useful. Our analysis of the abovementioned issues provides with clear guidelines for further research which will lead to the improvement of corporate bankruptcy prediction models.