Improving bankruptcy prediction in micro-entities by using nonlinear effects and non-financial variables

The use of non-parametric methodologies, the introduction of non-financial variables, and the development of models geared towards the homogeneous characteristics of corporate sub-populations have recently experienced a surge of interest in the bankruptcy literature. However, no research on default prediction has yet focused on micro-entities (MEs), despite such firms’ importance in the global economy. This paper builds the first bankruptcy model especially designed for MEs by using a wide set of accounts from 1999 to 2008 and applying artificial neural networks (ANNs). Our findings show that ANNs outperform the traditional logistic regression (LR) models. In addition, we also report that, thanks to the introduction of non-financial predictors related to age, the delay in filing accounts, legal action by creditors to recover unpaid debts, and the ownership features of the company, the improvement with respect to the use of solely financial information is 3.6%, which is even higher than the improvement that involves the use of the best ANN (2.6%).