An Evaluation of Selection Techniques for Bankruptcy Prediction Models: Does the Position in the Global Value Chain Matter?

This study introduces a comparison analysis in the application of statistical and artificial intelligent techniques to solve the bankruptcy forecasting problem faced by small and medium size firms. Relying on a comprehensive bankruptcy database of 7,104 Belgian SMEs, we show that intelligent techniques yield superior prediction performance when predicting bankruptcy. Interestingly, our empirical results are sensible to the position of the firm in the global value chain (GVC), with prediction performance of key variables being lowered for firms behaving upstream in the production process, providing support to upward contagion effects of bankruptcy along the GVC.