Machine-learning models for bankruptcy prediction: do industrial variables matter?

Daniela Bragoli,Camilla Ferretti,Piero Ganugi,Giovanni Marseguerra,Davide Mezzogori,Francesco Zammori
DOI: https://doi.org/10.1080/17421772.2021.1977377
2021-10-07
Spatial Economic Analysis
Abstract:We provide a predictive model specifically designed for the Italian economy that classifies solvent and insolvent firms one year in advance using the AIDA Bureau van Dijk data set for the period 2007–15. We apply a full battery of bankruptcy forecasting models, including both traditional and more sophisticated machine-learning techniques, and add to the financial ratios used in the literature a set of industrial/regional variables. We find that XGBoost is the best performer, and that industrial/regional variables are important. Moreover, belonging to a district, having a high mark-up and a greater market share diminish bankruptcy probability.
economics
What problem does this paper attempt to address?