Making supervisory stress tests more macroprudential: Considering liquidity and solvency interactions and systemic risk

In the run-up to the financial crisis, banking supervisors largely followed a microprudential approach towards assessing banks. As such, many of the "first-generation" stress tests used by bank supervisors after the crisis focused on solvency risks. Some supervisors also considered liquidity risks, but these risks were often viewed as independent of solvency risks. Additionally, authorities' stress tests often did not consider the potential interlinkages in the banking system or ways in which bank behaviour might collectively prove destabilising to the financial system. However, the failure to adequately model interlinkages and the nexus between solvency risk and liquidity risk within and across banks led to a dramatic underestimation of the risks to, and vulnerabilities of, financial systems in many economies. The prior Basel Committee working paper 24 contains a summary of case studies, which discusses some liquidity and solvency interactions at large banks. Building on the experiences of different countries, this paper suggests that authorities should emphasise developing integrated liquidity and solvency stress tests (as opposed to stand-alone liquidity stress test exercises). The paper offers several approaches to incorporating liquidity effects and their interactions with solvency that differ in their level of comprehensiveness and sophistication. In particular, the paper offers contributions to three key areas. First, micro stress tests provide a basis for developing and enriching stress tests by considering channels in addition to the standard credit channel through which shocks can be transmitted. Second, an analysis of estimated interactions between liquidity and solvency risks, using both regulatory and market-based measures, at the micro level will help improve stress testing models for individual banks. Finally, the third layer - network analysis and agent-based models - prove useful for broadening stress tests, as these models consider contagion through common exposure, interbank funding relationships and the endogenous behaviour of banks.