A Multi-Group Hierarchical Discrimination Approach for the Identification of Financial Distressed Firms : Evidence from Greece

In the present study we investigate the development of a classification model for discriminating between bankrupt, financially distressed and non-failed firms. The sample consists of 44 Greek firms that filed for bankruptcy over the period 2000-2003, 32 distressed firms that had serious financial problems over the same period but did not filed for bankruptcy (as of end 2003) and 44 non-failed and non-distressed firms. The model is developed through the MHDIS classification technique that distinguishes the groups progressively, starting by discriminating firms from the first group from all the others, and then proceeds to the discrimination between the firms belonging into the other groups. To ensure the proper development and validation of the model we follow a ten fold cross validation approach. Six financial variables covering all dimensions of firm financial performance are used. The results indicate that the characteristics of firms that are important in discriminating between non-failed firms and distressed/bankrupt firms are not the same as the ones that are important in discriminating between distressed and bankrupt firms. The MHDIS model achieves quite satisfactory overall classification accuracy in the validation sample that is also higher than the one achieved by a model developed through ordered logistic regression, which was used for comparison purposes