As stress testing has emerged as a key element of post-crisis financial supervision, it is worth taking a step back from the details of current implementations to consider broader trends and some fundamental topics that might benefit from additional research.

Seeing beyond recent history, that is, not “fighting the last war,” can be challenging. Supervisory guidance was sparse about how to conduct stress tests and use them for managing risk and capital. As a result, there was no consistency in scenarios and metrics across firms, which made horizontal reviews across banks difficult for supervisors.

Still, today’s approach to stress testing remains essentially microprudential; it focuses on the resilience of individual banks to specific shocks, rather than on the broader and more complex macroprudential question of how stress might be transmitted among firms, across financial markets, and into the real economy.

Stress testing for internal risk management is now an essential part of the bank supervision toolkit. Bank supervisors can also use stress tests to probe the financial system for weaknesses, emerging risks, common exposures, and interconnections. Stress tests can help reveal data gaps for further supervisory investigation and assess the risk management capabilities of individual firms. Of course, the task must be undertaken with the understanding that the exercise is based on trying to evaluate future events.

The next crisis will likely differ from the last, and contagion effects in crises are complex and difficult to evaluate before they occur. For these reasons, we argue that the next generation of stress tests should include second-order effects across firms and markets, and should expand to include the impact upon funding and liquidity.

Although much has been learned during the post-crisis process of implementing new bank supervisory stresstesting capabilities, we have identified important questions that would benefit from research and policy analysis. We advocate an exploration of this research agenda, as well as a broader move toward implementation of more dynamic, network-focused modeling embodied in stress testing Version 3.0.

We do not yet fully understand how to design equally challenging scenarios for diverse types of financial institutions.