This presentation is a short survey dedicated to the 50th anniversary of Edward Altman's publication of the Z-Score (Journal of Finance, 1968). It aims at indicating what we have learned from Altman’s novel idea of quantitatively describing the financial condition of a company and how this idea developed methodologically during the following years.
Edward Altman, Professor Emeritus, New York University, Stern School of Business, celebrates this 50th anniversary worldwide, together with many of his followers, in the form of seminars, publications, conference appearances etc. Our presentation lies along these lines. Professor Altman is a 2015 honorary doctor at my university – SGH Warsaw School of Economics. SGH holds an annual series of lectures dedicated to Altman.
The objective here is to present selected questions on bankruptcy prediction modelling with the Altman Z-score as the starting point. In the presentation, we emphasize not only his amazing career and the acceptable accuracy of the Z-score but also a stream of competing approaches from other areas of data analysis.
Data/ Methods
The literature on Altman approach to provide early warning signals of company financial distress is so vast that any new survey of those topics is destined to be novel.
Bankruptcy prediction models might be classified as part of financial microeconometrics (Gruszczynski 2008, 2012). Here, an attempt is made to categorize various approaches to prediction. Accent is put on novel methodologies established within the expert systems area. We begin by commenting on a paper by Altman, et al. (2017) with international empirical analysis of the Z-Score, mostly for European countries.
New approaches to bankruptcy modelling have emerged in all areas of data analysis. This has been documented e.g. in 2017 by Barboza, et al. Another novel approach is the use of the Z-Score in the methodology of counterfactuals, (Subrahmanyan 2017). Attempts of Polish authors in this field are also presented, (e.g. Zieba et al. 2015) about the use of ensemble-boosted trees in bankruptcy prediction.
Results/ Expected results
The aim of the presentation is to indicate major new approaches in bankruptcy modelling as well as its relevance for traditional models like the Z-score. Multivariate discriminant analysis and logit are competing with machine learning models, (Barboza, et al. 2017). The new results obtained for Polish data on bankruptcy are also presented.