Photo: Bloomberg News
The Fed’s Stress Tests Need to Be Transparent
Banks are being judged by secret models, in violation of the law that governs agency rule-making.
By Hal Scott and John Gulliver
The stress tests that big American banks face each year are about to get more stressful. The Fed is planning to substantially increase—by an average of 57%, we calculate—the regulatory capital that the eight largest banks in the U.S. need to pass the annual tests.
Had these expected higher capital levels been in effect this year, it is likely that the country’s four largest banks ( J.P. Morgan Chase, JPM -1.23 % Bank of America, BAC -1.15 % Wells Fargo WFC -1.56 % and Citigroup C -1.42 % ) all would have failed the test. As a consequence, they would have been barred from remitting more profits to their shareholders.
The higher capital requirements will diminish these banks’ ability to lend, potentially affecting economic growth. That isn’t all: The Fed’s secretive process for designing stress tests might well be illegal. It likely violates the Administrative Procedure Act of 1946, requiring government agencies to be transparent and publicly accountable.
This newspaper reported Sept. 1 that some of the banks covered by the stress tests are considering suing the Fed on this basis. An analysis released this week by the Committee on Capital Markets Regulation—a nonprofit with members belonging to virtually all segments of the financial industry, as well as leading academics—supports the conclusion that the Fed has acted illegally.
The stress tests are intended to measure whether banks would maintain sufficient capital in a future crisis. In principle, they should make the banking system safer by increasing the minimum amount of capital banks must hold and by enhancing market confidence in their resiliency.
Before conducting each year’s stress tests, the Fed develops a new economic hypothetical. In 2016 it involved a 6.25% drop in U.S. GDP and a crash that erased half the value of the stock market. Although the Fed discloses its assumptions, it does not provide the public with an opportunity to comment on them.
The Fed also develops new models each year for how the hypothetical crisis would affect banks’ capital levels. In 2016 the models predicted losses of $526 billion. But the Fed keeps these models permanently secret, which has a troubling result: Banks must guess at how millions of different assets would respond to the Fed’s assumptions and what this means for the minimum capital they need to pass the test.
Designing a stress-test program can cost a bank between $150 million and $250 million, as this newspaper has reported. Since the secretive process is clearly unpredictable, banks can unexpectedly fail. If they do they are severely punished by the market. After failing the 2014 test, Citigroup’s value promptly fell about 5% (or $8 billion). Citigroup alone spent $180 million on stress tests over the next two quarters.
Such costs are particularly burdensome for regional banks. There are approximately 15 of them with assets over $50 billion, the level that makes them subject to stress tests. They include Fifth Third Bank of Ohio and Regions Financial RF -0.73 % of Alabama, neither of whose possible failure would pose systemic risk.
The legality of the Fed’s process is dubious at best, as the analysis by the Committee on Capital Markets Regulation shows. The models and assumptions constitute agency “rules” under the Administrative Procedure Act, because they are designed in advance of the annual test, are applied to banks in a consistent manner, and have a binding effect.
But the Fed has failed to follow the rule-making process required by the law. There is no legal basis for denying public comment on the economic assumptions that underlie the tests. The law would also require that the Fed disclose its models. Agencies can claim an exemption for “good cause”—if following the prescribed process would be “impracticable, unnecessary, or contrary to the public interest.” However, there is no judicial precedent for applying the exemption here.
Daniel Tarullo, a member of the Fed’s board of governors, argued two years ago that publishing the models would allow banks to “game” the stress tests. In theory, a bank could accumulate assets that would perform well in the test and then quickly liquidate them after passing.
But this argument would have to break new legal ground on the Administrative Procedure Act, which has governed government agency conduct for 70 years. Not only that, but “gaming” concerns could be resolved without the secrecy. Most apparent, the Fed has broad authority to implement and enforce a good-faith or anti-circumvention rule. Since it collects data throughout the year from the banks, it is well positioned to determine if one is attempting to undermine the test.
The Fed should allow public comment on the economic assumptions for the next round of tests. It should also immediately justify its need to keep the models secret. A more transparent process would not only avert a lawsuit from the banks, it would better protect the financial system by increasing the validity and robustness of the stress tests.
Mr. Scott is a professor at Harvard Law School and director of the Committee on Capital Markets Regulation, where Mr. Gulliver is the research director.