Recent reports to regulators illustrate the difficulties banks have had predicting the riskiness of certain business activities during volatile periods.
J.P. Morgan & Co. and Bankers Trust Corp. both said this week that their own statistical models, used to measure the degree of risk they take in activities such as trading, essentially broke down during the third quarter as those businesses proved far more unstable than had been forecast.
"Unprecedented market volatility during the August and September period caused actual results to be more volatile than estimates predicted," Morgan said Monday in its quarterly filing with the Securities and Exchange Commission.
Bankers Trust was even more specific in its filing on Tuesday. "On five days during the quarter ended September 30, 1998," it said, the corporation experienced losses exceeding the maximum its model predicted it could suffer in its trading position in a single day.
Both banks reported weaker results for the third quarter because of market-related losses and a slowdown in investment banking and underwriting. At Morgan earnings fell 60%, to $156 million. Bankers Trust posted a $488 million loss for the quarter.
"The models are not set up to deal with extremes," said Stephen Biggar, an analyst at S&P Equity Research. "The models didn't take into account that the scenarios would be worse and would last for a sustained period."
Analysts said Morgan and Bankers Trust were not alone. Other companies, such as BankAmerica Corp., experienced similar problems.
Banks are under intense scrutiny by regulators and investors over their risk management procedures. A report released Wednesday by the Federal Reserve Board raised concerns about banks' credit-risk models. Analysts said they expect the Fed to take a similar look at market risk soon.
In market-risk modeling, banks have developed their own statistical models that measure the potential for losses in certain risky business activities. Banks use these risk models to calculate the capital they need to allocate to those businesses.
Risk management experts, however, said banks are relying too heavily on forecasting. "The senior bankers got their noses up against the computer screens and didn't step back to look at what was going on in the real world," said George Davis, president of Scarborough Partners, a consulting firm.
"They began to believe in their models, but the assumptions were wrong," Mr. Davis said.
Analysts said there may be a lesson to be learned.
"Banks have argued that they can effectively manage risks because their models are so good," said Lawrence Cohn, an analyst at Ryan, Beck & Co. "This is especially embarrassing for Bankers and Morgan, because they have always been considered to be on the cutting edge in this area."
Morgan and Bankers Trust each said that over the long run, their models have been statistically accurate. But during the months of July, August, and September their calculations were not on target.
Bankers Trust said it had been aggressively reducing its exposure to market risk-particularly in emerging markets-as a result of third-quarter volatility. It reduced its trading-related market risk by 55% during the first nine months of this year, from $112 million in trading assets in a 10-day period at Dec. 31, 1997, to $49.9 million at the end of September.
"The effective management of market risk continues to be a key focus for the firm," Bankers Trust said in its filing.
Mr. Cohn said banks have so far been able to police themselves in this area, having convinced regulators that they have the sophistication to accurately assess risk.
But, Mr. Cohn said, evidence of failure may spur the Federal Reserve Board and other regulators to curtail banks' market-related activities.
"It's clear that the Fed wants more capital allocated to these businesses, but it's not clear how much they want," Mr. Cohn said. "It'll be a balancing act as the Fed tries to foster liquidity in the credit markets and simultaneously tries to restrain banks' market risks."