Predicting Bankruptcy: An Investigation of the Applicability of Altman's Z''-Score Model

Søren Skovbjerg Steen & Florian Gaudenz Degen

Student thesis: Master thesis

Abstract

The scientific literature sees a wide range of bankruptcy prediction models, with new and reportedly improved models perpetually coming forth. At the same time, prior research found that models tend to be sample-specific, suggesting that they perform less accurately when applied to firms from other time periods, countries, industries and size classes. Yet, empirical tests of the general usefulness of existing models are receiving not much attention in academia. Taking up these points, this study investigated the applicability of the wellknown Altman Z-score model when applied to a sample of Italian firms in the period from 2011 to 2015, an environment that is markedly di↵erent from the one underlying the model estimation sample. In addition, the models sensitivity to industry aliation and firm size class was analysed by testing it on di↵erent subsamples of firms. Lastly, the accuracy of the model after re-estimating the variable coecients of the original model was assessed. The empirical results revealed that the Altman Z-score model, despite being more than 30 years old, still performed reasonably well in discriminating between bankrupt and non-bankrupt firms. Yet, the study provided evidence that the models accuracy has dropped over the last decades. Moreover, di↵erences were found in model accuracy when applied on di↵erent subsamples, indicating that the model is sensitive to industry and firm size. Re-estimating the variable coecients resulted in higher classification accuracy as compared to the original version of the model. These findings constitute useful information to model users, who are dependent on knowing the properties and boundaries of bankruptcy prediction models.

EducationsMSc in Accounting, Strategy and Control, (Graduate Programme) Final Thesis
LanguageEnglish
Publication date2018
Number of pages129