Stress tests were instituted at many central banks (such as the Bank of England, the Federal Reserve, and the European Central Bank) as an integral part of their regulatory framework a few years after the financial crisis, which took place in 2007-2008. This financial crisis showed that banks and other financial institutions are able to default, bringing down other financial institutions, and thereby affect the real economy. Stress tests aim to assess the resilience of the financial system and individual institutions therein. The stress test outcome-based regulatory actions are meant to prevent such a crisis happening again.
Although the current regulatory stress tests have brought back confidence in the financial system, have given insight into the previously opaque balance sheets, and have provided a useful mechanism for recapitalizing banks, stress tests have a few significant shortcomings due to which these are not able to accurately assess the actual systemic risk in the financial system. First, stress tests do not take into account interconnections (e.g. interbank exposure, common asset holdings, derivative exposures) between financial institutions. Therefore, contagion (i.e. spread of stress), which was the main driver of defaults in the financial crisis, is not captured. Instead, the stress tests shock balance sheets. Second, the stress tests focus on one type of financial institution (e.g. bank stress tests, pension fund stress tests, insurer stress tests), instead of taking all the tightly interconnected types of financial institutions in the financial system into account.
To address these shortcomings we are collaborating with the Bank of England, and some other central banks, to develop a system-wide stress test that does take into account different types of financial institutions and their interconnections. In particular, based on empirical observations and data sources (e.g. EU, UK) we are developing an agent-based model (ABM) of the financial system. We model:
- Agents (as a balance sheet composed of financial contracts)
- Financial contracts (stipulating the interconnections between financial institutions)
- Binding constraints (such as those coming from regulation, e.g. leverage constraint, these constrain financial institution in their behavior)
- Profit-seeking behavior
- Financial markets (i.e. the venues where two counterparties match and agree upon financial contracts).
Given that the model is initialized with regulatory data on the financial system, the model can be used for stress testing purposes to assess the stability of the financial system. In particular, one can define a systemic risk measure and assess the sensitivity of this measure to various parameters in the financial system. For example, one can assess how the stability is influenced by various regulations (such as the level of the minimum leverage of banks), the structure of the multiplex network (e.g. does crowding into a particular asset class magnify systemic risk?), and the path of the credit intermediation chain (i.e. does a more complex chain of credit creation via various nonbanks increase systemic risk compared to a simpler credit intermediation chain?).