Fed opens black box on stress tests (but not too far)

Complimentary Access Pill
Enjoy complimentary access to top ideas and insights — selected by our editors.

WASHINGTON — The Federal Reserve Board on Thursday proposed to give banks more insight into portfolio performance under its stress testing models in an effort to help banks better understand the Fed’s methodology without giving away too much.

“This enhanced transparency will bolster the credibility of our stress tests and help the public better evaluate the results,” Randal Quarles, the Fed's vice chairman for supervision, said in a statement. “The proposed changes will also generate valuable insight from stakeholders and we look forward to it.”

The central bank issued a handful of proposals outlining several policy changes, some of which Fed Governor (and now chairman-designate) Jerome Powell said would be coming months ago.

Randal Quarles, governor of the U.S. Federal Reserve.

The proposal would give banks “enhanced descriptions of supervisory models, including key variables,” as well as models on loan loss rates on loans and statistics on the loss characteristics for certain loan groups. The proposal would also provide for banks “portfolios of hypothetical loans and the estimated loss rates associated with the loans in each portfolio,” the Fed said.

Under the existing stress testing regime, the Fed makes certain aspects of its modeling process known in an appendix to the released results, but those disclosures are limited to the basic structure of the model — indicating that losses are arrived at based on other models that, for example, assess the probability of default, the expected losses from a default or the bank’s exposure upon default.

The proposed rule would go one step further and disclose some actual equations detailing the probability of default, or losses from default.

“The proposed enhanced model description would include certain important equations that characterize aspects of the model,” the proposal said. “The proposed enhanced descriptions would [also] include a table that contains a list of the key loan characteristics and macroeconomic variables that influence the results of a given model.”

The other proposals released Thursday include amendments to the Fed’s stress testing policy statement related to the development of the hypothetical stress scenarios at the heart of the stress testing regime.

The proposed changes to the policy statement would give the Fed more flexibility in allowing it to increase the hypothetical unemployment rate in its severely adverse stress scenario by a smaller amount — for example, by raising unemployment 3% rather than 5% — if the baseline unemployment rate is already relatively high. They would also provide an explicit path for the nominal housing price index — giving more detail on how the Fed assumes housing prices and values will be affected in a severely adverse scenario.

The Fed also proposed that its policy statement be changed to allow for the incorporation of the costs of short-term wholesale funding in adverse and severely adverse scenarios — a consideration that had not been made in scenarios to date. The proposal said it does not expect such considerations to be included in scenarios “until 2020 at the earliest” but wanted to lay the groundwork for their inclusion.

The Fed’s stress testing regime is widely understood to be among the most consequential post-crisis innovations. It requires all banks with more than $50 billion in assets to submit their balance sheets to the Fed, which examines how those assets would perform under hypothetical future scenarios of varying degrees of economic stress. If banks fail to maintain a minimum capital level throughout each of the scenarios, they can be penalized by blocking their payments of dividends.

Banks have mostly accepted stress testing as part of the regulatory landscape, but have complained about specific aspects of the testing program — namely the opaqueness with which the scenarios are decided and the opaqueness of the models that assess a bank’s performance through those scenarios.

Until now, the Fed has been wary of disclosing too much about its models and its scenario design, arguing that if the industry had a more comprehensive knowledge about the models then it could manage its portfolios to achieve the desired stress test results, potentially making the bank appear better capitalized than it really is.

Quarles echoed this sentiment earlier this month, when he said that banks clearly have a vested interest in figuring out what is in the models and are working to reverse-engineer the models all the time.

“The banks are figuring it out,” he said. “They are very smart people who have devoted lots of resources to figuring out what’s in the model.”

For reprint and licensing requests for this article, click here.
Stress tests Loan-loss provisions Data modeling Randal Quarles Federal Reserve
MORE FROM AMERICAN BANKER