Fed Study: Bank Risk Models Too Weak for Setting Capital

Credit-risk models at most large banks are unsophisticated and cannot be used to determine capital needs, according to a Federal Reserve Board study due out today.

Shortcomings include too few loan ratings, insufficient data on the accuracy of past grades, and vague definitions of credit-quality ratings, according to the paper.

"The great majority of the 50 largest holding company rating systems are not really well-prepared to support the most advanced credit-risk modeling techniques," Mark S. Carey, a Fed economist who co-wrote the study, said in an interview.

"People find these results very surprising," said William F. Treacy, a supervising financial analyst at the Fed and the study's other author. "They expected a more advanced state for the industry."

Many industry officials and regulators-including Fed Chairman Alan Greenspan-have said that credit-risk models could be used to set capital requirements, much as market-risk models are used to determine how much capital is needed to cover trading risk.

Federal Reserve Bank of New York President William J. McDonough, chairman of the international Basel Committee for Banking Supervision, is expected to include a models-based approach in an overhaul of the global capital rules he is drafting.

Though the study's authors conclude that such a use is "probably feasible," they argue it would require a significant overhaul of credit- risk rating systems. For instance, many banks would be forced to add more grades to their rating systems. Of the 45 banks that provided data, 16 slotted more than half their loans into the same rating category. Five awarded the same grade to more than 80% of their commercial credits.

Regulators would have to validate ratings by spot checking loans to ensure grades are applied consistently across the company, they wrote. The Fed staff members also warned that basing capital on credit-risk ratings could lead banks to underestimate a loan's risk.

Much of the 25-page study is devoted to the current state of credit-risk management at the 50 largest banks. Though all use credit-risk modeling, techniques vary widely. Most apply ratings only to commercial credits and include four categories for troubled loans: special mention, substandard, doubtful, and loss. The typical bank also has five pass grades, ranging from "exceptional" to "watch," which means the loan is at risk of becoming a problem credit.

Banks typically relate their grades to the scales used by Standard & Poor's and Moody's Investor Services. For instance, an average loan would be a BBB credit. For the first time, the researchers linked the bank credit ratings to the S&P ratings. They found that 7% of bank loans qualified for AAA or AA ratings, a surprisingly high number given the common perception that investment grade companies do not rely on banks for financing. Roughly half of the loans were rated below-investment grade.

The study found that banks relied on the judgment of their employees- rather than on statistical formulas-when awarding grades. Though guidelines exist, the bank's credit culture was more important at ensuring grades were consistent.

The study, "Credit-Risk Rating at Large U.S. Banks," is available at www.federalreserve.gov. An early draft was the basis for a supervisory letter on credit-risk modeling issued by the Fed in September.

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER