An empirical evaluation of four financial distress prediction models for Greek firms: is there a 'most appropriate' model?

In this paper, four financial distress prediction models for Greek firms are tested. Relevant analysis is based on a sample of 37 financially distressed (18 listed and 19 non-listed) and 226 non-distressed (48 listed and 178 non-listed) firms. The superiority of a particular model relates to its predictive accuracy and expected loss of misclassification errors in a range of likely values for the prior probability of financial distress and the cost ratio of Types 1 and 2 errors. We find that: a) rates of correct predictions are unstable when models are used to predict financial distress in periods following the one that was considered to estimate them; b) if a model is found to be the most superior, it does so for almost all likely values of cost and prior probabilities ratios; c) no single model can be considered absolutely appropriate to predict the financial distress of Greek firms as superiority of models differs between non-listed and listed firms.