Building a better stress test tool

The Minneapolis Fed created the COVID-19 stress test tool to help the public assess the health of the banking sector now, in the midst of the economic uncertainty created by the global pandemic, and going forward. The tool lets users adjust several economic variables, such as unemployment, gross domestic product, and the stock market, and see how they impact capital levels of the nation’s largest banks.

We believe publicly available bank assessment tools like ours can help protect taxpayers. They provide the public with better information about the resilience of big banks which, in turn, leads to more transparency and, we hope, better public policy.

That is why the Minneapolis Fed recently brought together leading experts in stress testing and banking regulatory policy for a virtual conference. With a keynote address from Sir John Vickers, University of Oxford, and a panel of experts moderated by Ann Saphir, Reuters, the conference focused on how to empower the public to assess the health of large banks.

Conference speakers and participants had many recommendations to improve the Minneapolis stress test tool, and others like it, and make it a more valuable resource for the public. We encourage you to watch the conference and, to provide you with a taste of what we heard, a few key recommendations are provided below:

1. In order to make the health of large banks more transparent for the public, we need to give people additional background information and context for bank stress tests. Stress tests are highly technical and complex, so it is important to be very clear about why they are important and to explain what the results mean—or, more importantly, what they don’t mean—to different stakeholders.

2. The public wants more detailed information from stress tests, such as the names of the banks that fail the test. Some will want even more. Tests should aim to provide bank-specific information that emphasizes the key macroeconomic variables or individual loan portfolios and business lines (e.g., auto lending, credit card loans) that drive the results.

3. The results from these tools can become stale quickly, so the underlying data and the scenarios should be updated regularly. The bank data should be refreshed as new information becomes available. The macroeconomic scenario projections should be updated to account for new potential shocks to the banking system.

4. Sometimes the information the public wants is not the variables that go into the stress test but, instead, the rules that govern the stress test. Those rules can significantly impact the stress test’s results. For example, some tests allow banks to continue paying out money to investors, while others do not. Tools should allow the public to vary these rules. Put another way, the tool should allow the public to evaluate the effect of regulatory policy on the results from stress tests.

5. Some users of the test will be very sophisticated. The tool should give these users options to explore more advanced topics that are currently omitted from the official stress tests conducted by the Federal Reserve. These include such things as the impact from “fire sales,” liquidity shocks, feedback loops among banks or within a single institution across time, and more sophisticated and realistic measures of core bank operations.

   These same sophisticated users may want to extend the framework to perform “reverse” stress tests. This form of stress testing identifies and assesses the specific tail risk scenarios that would result in an individual bank failing. Rather than measure how banks perform in a bad recession, a reverse stress test seeks to determine what specific shock will cause a bank to fail.

6. Finally, many participants suggested incorporating bank health assessments that derive from financial markets in the information we make available to the public. There are several ways to do this, with many participants noting that measures from the stock market could help (e.g., price-to-book ratios, option-implied volatilities, abnormal equity returns, or “distance-to-default” measures using Merton-type models).

The Minneapolis Fed will continue working on these and other related topics in the coming months. Importantly, we will continue to make the underlying data, software programs, and models available to the general public wherever possible. We look forward to updating interested parties on our efforts to improve the Minneapolis stress test tool. Please feel free to contact us with additional ideas and feedback.