Federal Reserve Board Governor Daniel Tarullo proposed ditching Basel III regulatory capital requirements in favor of the Fed’s stress test capital assessment process in a May speech delivered at the Chicago Fed. Tarullo argued that banks have “gamed” the Basel capital regulations by using aggressive modeling assumptions to reduce their asset risk weights.

While there is reason to be concerned that Basel III gives banks too much latitude, allowing capital regulation to be guided solely by stress tests is a mistake for a number of reasons. Stress tests have a spotty record as a regulatory tool for identifying weaknesses at institutions. And the intrusive nature of stress test regulation reinforces investors' perceptions that some banks are too big to fail.

A stress test-based approach to setting bank capital has two gigantic measurement problems.  First, the macroeconomic scenario must actually anticipate the next financial crisis. Next, it must estimate an institution’s loss.  Both are easier said than done.

Both pre-and post-crisis history offers no evidence that macroeconomic stress tests are effective in preventing future financial disasters.  Prior to the financial crisis, many countries issued financial stability reports with bank stress tests—yet none anticipated the crisis.  The U.S. housing agencies Fannie Mae and Freddie Mac both passed elaborate regulatory stress tests before failing in September 2008. Pre-crisis pan-European stress tests administered by the European Banking Authority gave a green light to a number of banks that subsequently imploded. 

Regulators' record for catching problems with stress tests has failed to improve since the crisis. The Fed’s stress tests have missed the London Whale at JPMorgan Chase and, this past spring, a multibillion-dollar hole in Bank of America’s balance sheet.

But let’s be optimistic and assume that the Fed's stress scenarios manage to accurately anticipate the coming crisis.  Next, the scenario must accurately predict the crisis's effects on individual bank profits and losses. This is difficult because individual bank profits and losses are not highly correlated with changes in macroeconomic indicators.  Most banks have regional concentrations, and some regions may be spared the impact of a national recession while others are hit harder.  Moreover, bank credit losses react to economic conditions with long and variable lags while capital market reactions are famously unpredictable, anticipating two recessions for every one that actually occurs.

For all these reasons, quarter-to-quarter bank profits do not move in lock-step with changes in gross domestic product, inflation, unemployment and other measures of economic health.  The best econometric models, in my extensive experience with stress tests, typically explain only about 25% of the quarterly variation in individual bank profits and losses—meaning that 75% of the variation cannot be predicted using GDP, unemployment, or other business cycle indicators.  Compounding the measurement problem is the fact that these models perform worse in crisis periods and when they are used to forecast beyond the sample period in which they were estimated—precisely the conditions in a stress test.

Because of these measurement issues, it is unavoidable that bank loss predictions from macroeconomic stress tests will have large forecast errors.  Even using the best models, there remains a great deal of uncertainty about how each bank may actually perform in the next crisis. As a consequence, crisis loss estimates are heavily dependent on modeler judgment. The inherent model uncertainty and differences in judgments applied by the Federal Reserve and a bank’s modelers make stress test-based capital regulation especially problematic for bank management and planning.

The Fed's intimate involvement in banks’ stress test capital planning process also reinforces the too-big-to-fail problem. Since the Fed models the operations and exposures of each large banking institution using its own judgment, investors have difficulty believing that the government would impose losses on shareholders and bondholders should a given institution later become distressed.

The Fed’s move to abandon Basel III's capital rules would be especially short-sighted if other nations follow their lead. True, the stress test process gives the Fed enhanced powers to set domestic capital requirements that are higher than those set by Basel regulations. But as an international standard, stress tests fall short.  Without Basel regulations, nothing stops foreign regulators from using their discretion to set stress test capital requirements below Basel capital levels for their banks.  The potential for “gaming” capital adequacy under a stress test regime far exceeds the Basel III gaming that worries Governor Tarullo.

Replacing Basel III capital rules with macroeconomic stress tests is not the answer to concerns that Basel III gives banks too much latitude. Instead, Basel III capital rules should be simplified and improved to prevent the banks from gaming their capital calculations. 

Paul H. Kupiec is a resident scholar at the American Enterprise Institute. He has also been a director of the Center for Financial Research at the Federal Deposit Insurance Corp. and chairman of the Research Task Force of the Basel Committee on Banking Supervision.