WASHINGTON - Banks may have some new ammunition on the fair-lending front.
Two economists from the Office of the Comptroller of the Currency will release a study Monday concluding that statistically based lending- discrimination models do not work.
Their findings suggest that most of the studies that have criticized bank lending practices - including the now-famous 1992 study by the Boston Federal Reserve Bank - are flawed.
The Justice Department relies on similar statistical models when it investigates fair-lending complaints.
"Any statistical model for disparate treatment should be subject to corroboration by file review, whether it is done by our agency, the Justice Department, or a private interest group," OCC economist Mitchell Stengel said in summarizing his study.
The paper is one of a series of reports that economists at the banking agencies have released during the past year questioning earlier findings of widespread lending discrimination. Those reports criticized the methods used in past studies, and argued that minorities default more frequently on home loans than whites.
However, community activists have released a number of studies showing that banks reject proportionally more minorities than whites for home loans. Also, their research concludes that banks provide less credit per person in minority areas than in white ones.
The OCC study casts doubt on many of those community group reports. Mr. Stengel and Dennis Glennon reviewed three banks. They applied to each a uniform statistical model and one tailored to the bank's underwriting criteria. They then compared the results and checked them against the loan files.
They first found that tailored models are better at picking up signs of discrimination then broader tests.
But they said these results evaporated when researchers reviewed the loan files.
They concluded that broad formulas that can apply to more than one institution don't work. They also said that Home Mortgage Disclosure Act data can produce inaccurate samples because they include people who bought foreclosed homes from the bank and people who applied to special CRA programs for credit.
More research must be conducted to find formulas that work, they said.
Stephen M. Cross, deputy comptroller for consumer and fiduciary compliance, said bankers should heed the study's conclusion when reviewing self-testing information.
"Statistical modeling can be a useful adjunct to the supervisory process or the self-assessment process," Mr. Cross said. "But it is unlikely that we will, or that bankers should, rely on it exclusively."
The Comptroller's report was welcome news to banking industry advocates.
"We consider the research by the authors of the Comptroller of the Currency report very important," said Mike ter Maat, an economist at the American Bankers Association. "The research does cast doubt on simplistic modeling techniques which have been used frequently in the past, particularly over the past couple of years."
Lawyers representing banks in fair-lending cases said the study further bolsters the industry's argument that it doesn't have a lending discrimination problem.
"This study, like several others that have come out in recent months, reinforce the need to take with a large grain of salt the statistical analyses that have been conducted by community groups and other interested parties to trumpet the notion that lending discrimination is widespread," said Andrew Sandler, a partner at Skadden, Arps, Slate, Meagher & Flom.
Community groups and regulators have relied on these reports during the past two years to push for stronger enforcement of the Fair Housing Act and the Equal Credit Opportunity Act.
But housing activists questioned the study. John P. Relman, director of the fair-housing project at the Washington Lawyers' Committee for Civil Rights and Urban Affairs, said the OCC economists "fundamentally misunderstand" what studies by his group and others show.
"It is simply wrong to say there is no basis upon which to compare the performance of banks," he said.
The community groups use statistical formulas to control for differing underwriting standards, he said.
"If you find after controlling for those variables that race is still a factor, you can say to a judge or jury that bank A or bank B may be discriminating," Mr. Relman said.
But David Horne, an economist at the Federal Deposit Insurance Corp., said the study does discredit those types of reviews. "This says that ... even when you get a race effect, it does not mean there is lending discrimination," Mr. Horne said. "This absolutely devastates the Boston Fed study."
