Fed's Tarullo Outlines Broad Changes to Stress Test Regime

Register now

WASHINGTON — The Federal Reserve is slated to unveil substantial changes to its supervisory stress testing program, including dropping the qualitative requirements for all but the largest banks and instituting a higher minimum capital requirement.

In a speech at Yale University on Monday, Fed Gov. Daniel Tarullo — who oversees the central bank's supervisory committee — outlined the long-awaited changes to the Comprehensive Capital Analysis and Review program, which he described as "the cornerstone of post-crisis prudential regulation."

"In short, the [Global Systemically Important Banks] will see their capital requirements rise," Tarullo said. "All other CCAR firms will see some reduction in their capital requirements. And firms that have less than $250 billion in assets and do not have extensive international or nontraditional banking activities will also transition to a more tailored set of capital planning expectations outside the CCAR process."

After more than a year of review and extensive conversations with banks, public interest groups and other stakeholders, Tarullo said the agency has prepared concrete proposals to ensure that the program continues to advance the goals of safety and soundness in the global financial system.

"We regard the stress testing and CCAR programs to have made substantial contributions to protecting the safety and soundness of the nation's largest financial firms and to promoting the stability of the financial system as a whole," Tarullo said. "But if a stress testing regime is to continue to play this role, it must be dynamic. After five years of post-crisis stress testing we believed it was important for us to conduct a more thorough assessment of the program."

The Fed has conducted two separate stress tests on all bank holding companies with more than $50 billion in assets every year since 2011 — the Dodd-Frank Act Stress Test and CCAR. Both tests take a bank's balance sheet and other internal data and run them through hypothetical stress scenarios to see how the banks' capital levels perform over nine consecutive quarters.

Under DFAST, the Fed uses a standardized capital management program that does not take a firm's distribution plans into account. No matter how poorly a bank's capital performs, it cannot fail the test. For CCAR, however, the Fed uses the banks' own capital management plans, giving a more accurate depiction of how the firm would perform in a stress event. What is more, the banks must conduct internal stress tests using their own models, and the Fed assesses their performance and the adequacy of those internal tests as well. A bank can fail CCAR both quantitatively — that is, by falling below the minimum capital requirements — and qualitatively, by having poor or underperforming internal assessments.

The changes Tarullo outlined have three broad features: changes to the minimum capital requirement, dropping the qualitative assessment for all but the largest banks, and broader macroprudential considerations for the annual stress scenarios.

The Fed will propose to eliminate the qualitative testing requirement for firms with "less than $250 billion in assets that do not have significant international or nonbank activity" — a substantial concession, given that most firms that have failed CCAR in recent years have failed on qualitative grounds rather than quantitative ones.

"As noted earlier, many of these firms have already met supervisory expectations," Tarullo said. "We do not intend for less complex firms to invest in stress testing capabilities on par with the most complex firms and, given their profile, we feel these firms can maintain the progress they have made through the normal supervisory process, supplemented with targeted horizontal reviews of discrete aspects of capital planning."

Tarullo said the Fed will also raise minimum capital requirements. To date, the CCAR evaluations have required banks to meet certain capital thresholds — 4.5% of common Tier 1 equity, for example. But those minimums have not taken into account various additional capital requirements that the Fed has developed, including the 2.5% capital conservation buffer and any capital surcharge that might be required for the eight U.S.-based global systemically important banks. The GSIB surcharge can vary from 1% to up to 4%, depending on the bank's size, complexity and reliance on short-term wholesale funding.

The new capital requirement, Tarullo said, will be made up of the 4.5% Common Tier 1 capital minimum, the entire GSIB surcharge (where applicable), and an additional measure called the "stress capital buffer." The stress capital buffer would be set based on each individual bank's capital performance in the previous stress test cycle and would consist of the greater of 2.5% or the difference between the bank's pre-stress capital level and their post-stress minimum capital.

For example, if a bank went into the stress test with 11% capital the previous year and showed a post-stress minimum capital level of 6% common Tier 1 capital, the bank would handily pass the quantitative CCAR test under today's rules. Under the modified rules, however, that same bank would have to meet not only the 4.5% common Tier 1 minimum, but also an additional 5% stress capital buffer the following year, making the bank's post-stress minimum capital level 9.5%. That would not include any GSIB capital surcharge that might apply.

However, if that same bank went into the stress test with 11% capital and had a post-stress minimum capital level of 8%, the bank's stress capital buffer would be 3%, meaning its total capital requirement of the next year would be 7.5% (notwithstanding any GSIB surcharge). The stress capital minimum has to be at least 2.5%, Tarullo said.

Finally, Tarullo said the central bank is considering a "relatively modest revision" to the annual stress scenario-design framework in order to make the tests more macroprudential in their scope. Among those would be revising the unemployment rate to be less severe during stress events and replacing the current assessment of home prices with a variable assessing personal disposable income.

"The result would be a more countercyclical path of house prices and increased transparency into how we determine the severity of this very important element of the scenario," Tarullo said.

Most of the changes Tarullo laid out would not take effect until at least the 2018 stress test cycle, with the exception of the exemption of banks with less than $250 billion in assets from the qualitative stress test requirement. The Fed was slated to publish a proposed rule Monday that would enact that exemption in time for the 2017 stress tests.

Tarullo also noted that there are "areas in which transparency considerations need to be applied" — a nod to longstanding concerns raised by some banks that the Fed's process of developing scenarios and development of internal models is too opaque and may potentially violate federal administrative procedure laws.

"Transparency of the scenarios and results gives investors and analysts valuable information about the condition of the tested banks, thereby contributing to market discipline," Tarullo said. "It also allows the public to evaluate the job that the Federal Reserve is doing. For example, analysts can compare our loss estimates for specified portfolios under specified stress conditions with their own evaluations — an exercise that can inform both the analysts and us."

He said the agency already discloses its scenarios and the process by which it develops its scenarios and releases substantial data on the results of the stress tests themselves. The Fed will begin to release "more granular disclosures, such as providing more details on different components of projected net revenues," he said.

But with respect to the models themselves, Tarullo said the agency is "considering further steps" to disclose additional details about the models — such as descriptions of changes made to the models well in advance of the stress testing cycle. But the Fed is not going to publish the models in their entirety, he said.

"We do not intend to publish the full computer code in the supervisory model that is used to project revenues and losses," Tarullo said. "Full disclosure would permit firms to game the system — that is, to optimize portfolio characteristics based on the parameters of the model and take risks in areas not well captured by the stress test just to minimize the estimated stress losses."

For reprint and licensing requests for this article, click here.