The anatomy of credit risk: Traditional models vs. intelligent systems

Credit risk management is a field that continues to require thorough research, owing to the fact that the inability of borrowers to service their obligations has led to the failure of numerous financial institutions. Traditional models, some of which were developed over half a century ago, have been used widely to evaluate counterparty defaults. However, as demonstrated in the latest financial crisis, these models have not been successful in capturing the main risk drivers that have been responsible for the adverse developments in the credit quality of borrowers. The purpose of this paper is to present the evolution of credit risk modeling from traditional models to intelligent systems. We aim to draw useful conclusions, which could add to the credit risk research and its application in contemporary risk management systems. In Section 2, we provide an overview of the most significant developments in the area of credit risk, which date from the 1930s, when the first bankruptcy models were developed and have contributed to the establishment of traditional statistical models. Next, we present the risk measurement framework required in contemporary risk management systems together with several methodologies of traditional statistics. In Section 4, we provide a thorough description of intelligent systems, including traditional methods of artificial intelligence such as expert systems, rule-based systems, and genetic algorithms, as well as methods belonging to the realms of data mining and machine learning, such as neural networks and support vector machines. In Section 5, we contrast traditional models with modern methods that are used for the purpose of forecasting company default events. We conclude with a few final thoughts and focus on areas that require further research aiming to assist scholars and practitioners in improving their risk management systems.