A Deep Learning Approach for Dynamic Balance Sheet Stress Testing

In the aftermath of the financial crisis, supervisory authorities have considerably improved their approaches in performing financial stress testing. However, they have received significant criticism by the market participants due to the methodological assumptions and simplifications employed, which are considered as not accurately reflecting real conditions. First and foremost, current stress testing methodologies attempt to simulate the risks underlying a financial institution's balance sheet by using several satellite models, making their integration a really challenging task with significant estimation errors. Secondly, they still suffer from not employing advanced statistical techniques, like machine learning, which capture better the nonlinear nature of adverse shocks. Finally, the static balance sheet assumption, that is often employed, implies that the management of a bank passively monitors the realization of the adverse scenario, but does nothing to mitigate its impact. To address the above mentioned criticism, we introduce in this study a novel approach utilizing deep learning approach for dynamic balance sheet stress testing. Experimental results give strong evidence that deep learning applied in big financial/supervisory datasets create a state of the art paradigm, which is capable of simulating real world scenarios in a more efficient way.