effects of the integration of the variability of key accounting aggregates in failure prediction models: Application to the Belgian cas.

Abstract

Bankruptcy prediction models have been developed and studied for many years. They are indeed of major interest since bankruptcies are numerous and have negative effects on many people more or less close to the company going bankrupt. The interest of these models is therefore to predict bankruptcy on the basis of different indicators, most often financial, whose evolution could distinguish healthy companies from those that go bankrupt. Moreover, anticipating a bankruptcy could at best make it possible to avoid some of them and at worst to mitigate the negative effects of bankruptcy. The objective of this thesis is to investigate whether adding the variability of the different financial indicators often used within prediction models has an effect on the efficiency of the model. Given the importance that these models can have, it is interesting to investigate whether this can improve the results. In the first part of this work, the theoretical one, we first analysed the key concepts that are used throughout this research. Then, we have detailed what has already been done concerning bankruptcy prediction models and the study of variability by means of a literature review. We explain a major difference in vision between authors who try to predict bankruptcy and those who try to prevent it and finish by explaining the statistical model we use for this research. In the second part of this work, the practical one, we start by detailing all the choices we made to carry out this research within a detailed methodology. We then analyse the various results obtained with the help of our statistical analysis. This allows us to make comparisons and analyse the effect of adding variability within a bankruptcy prediction model.