Financial Stress Testing

Abstract

Years of turmoil in the banking sector have revealed the need to assess the performance of banks and explore the resilience of the banking system under adverse futures. Banks are not the safe houses everyone believed them to be, as the recent crisis of 2007 showed. Today, banks are highly uncertain dynamically complex systems that are permanently at risk due to internal and external stresses and uncertainties. Although external uncertainties and stresses cannot be controlled and accurately forecasted, system’s exploration under plausible futures can lead to the identification of its weak points. Monitoring a complex system like a bank and identifying vulnerabilities to different types of risks requires advanced tools. Monitoring and analytical tools, like financial stress tests, are developed by regulatory authorities and financial institutions to identify causes and vulnerabilities of the system under adverse future scenarios. Financial stress testing allows assessing the financial system stability or even individual bank’s performance. During the last decade, the evolution in this field is significant, in an effort to enhance the resilience of financial institutions, especially after the severe financial crisis. New methodologies and more elaborated tools are implemented in risk management practices of banks. Common techniques can be summarised in simple sensitivity tests and historical or hypothetical scenario analyses. To this direction, focusing on an approach that could simulate multiple future hypothetical scenarios, exploratory System Dynamics modelling could offer a new tool for a model based exploration in order to support monitoring of bank’s financial state. This research illustrates a pilot System Dynamics approach towards financial stress testing in view of making banks more robust by identifying possible weaknesses. A System Dynamics model was developed to represent the core endogenous operations of a bank. The bank model represents a medium sized commercial bank and not a systemic bank of a country. The level of aggregation provides an example of its balance sheet but it is not an accurate and detailed representation of a bank. Although, further research is needed for a more detailed model, an analysis of plausible scenarios that could have effects on bank’s balance sheet, regarding its net worth and liquidity, is performed. The model is constantly or periodically attacked by unforeseen risks and shocks in order to generate insights into all sorts of plausible bank system behaviours under stress. The use of EMA workbench can assist in the exploration of those behaviours by plotting all the plausible future scenarios. Based on the observed behaviours we identify the causes of undesirable dynamics, vulnerabilities and levers. The combination of uncertainties that lead to those undesirable outcomes is revealed with the use of machine learning algorithms. Using these insights, basic policies are designed and applied in an effort to improve bank performance under particular undesirable scenarios.