Predictors of financial distress: Differences between financial and non-financial small and medium-sized enterprises

Banks use financial distress models to assess the creditworthiness of firms. This is particularly crucial for small and medium-sized enterprises (SMEs), as they often face strong financial constraints. While most existing models are developed for non-financial firms, understanding distress in financial firms is equally important. This study compares predictors of financial distress in financial and non-financial SMEs using a unique dataset of Norwegian firms from 2000 to 2020. We apply logistic regression and a machine learning method known as Light Gradient Boosting Machine (LightGBM) to identify key financial ratios and evaluate predictive performance. The results show that revenue generation and operational efficiency are the most influential for non-financial SMEs, whereas profitability and asset structure are more important for financial SMEs. Out-of-time validation demonstrates that models perform best when trained and tested within the same firm group. Applying a model estimated on one group to the other results in substantially weaker performance, especially when transferring from non-financial to financial firms. These findings underscore the importance of sector-specific models in enhancing risk assessment and early warning systems.