Corporate governance and financial distress: lessons learned from an unconventional approach

Using a and a unique set of Italian non-listed Unlikely to Pay (UTP) positions, that consist in the phase that precedes the insolvency but where it is still possible for the company to succeed in restructuring, this paper aims to analyze the relationships between corporate governance characteristics and fnancial distress status. We compare the performance of corporate governance variables in predicting corporate defaults, using both the Logit and Random Forest models, which previous researchers have deemed to be the most efcient machine learning techniques. Our results show that the use of corporate governance variables – especially with regards to CEO renewal and stability in the composition of the board of directors – increases the accuracy of the Random Forest technique and infuences the success of the turnaround process. This paper also confrms the Random Forest technique’s ability to signifcantly outperform the Logit model in terms of accuracy.

Using a and a unique set of Italian non-listed Unlikely to Pay (UTP) positions, that consist in the phase that precedes the insolvency but where it is still possible for the company to succeed in restructuring, this paper aims to analyze the relationships between corporate governance characteristics and financial distress status. We compare the performance of corporate governance variables in predicting corporate defaults, using both the Logit and Random Forest models, which previous researchers have deemed to be the most efficient machine learning techniques. Our results show that the use of corporate governance variables – especially with regards to CEO renewal and stability in the composition of the board of directors – increases the accuracy of the Random Forest technique and influences the success of the turnaround process. This paper also confirms the Random Forest technique’s ability to significantly outperform the Logit model in terms of accuracy.