The performance of credit scoring models in contexts of strong macroeconomic instability: The role of Artificial Neural Networks
Abstract
The study addresses the problem related to the performance of the tools for forecasting corporate crises in periods characterized by strong macroeconomic instability (financial crises, pandemics, wars, etc.). The results obtained show how the performances of the models decrease over time and how, in a period characterized by strong macroeconomic instability, more evident drops in performance are observed. Particularly, with reference to the hotel sector in Italy, in correspondence with and immediately after the financial crisis of 2008, it emerges that artificial neural networks produce more precise and less volatile predictions than the classical models used in the literature (linear discriminant analysis and logistic regression).
Keywords: Insolvency forecasting, Credit risk, Early warning system, Credit scoring, Artificial Neural Networks
