WASHINGTON -- Buoyed by the success of a pilot program, the Federal Reserve has integrated a computer-based statistical analysis into its fair-lending reviews.
The new program will enable examiners to more easily pinpoint mortgage applicationss where bias may have played a role in lending decisions. It will also help them decide whether race was a significant facto in a bank's overall lending pattens.
For large state-member banks and mortgage subsidiaries of bank holding companies, the new procedure will mean tougher and more detailed scrutiny of minority lending decisions in routine compliance exams.
Referrals Seen Rising
It will also likely mean more referrals from the Fed to the departments of Justice and Housing and Urban Development for further investigation.
"We're getting more and more statistically sophisticated as we go along," Fed Governor Lawrence B. Lindsey recently said about the model.
Fed officials have been developing a computer model to help detect discrimination since 1992, when the Boston Fe released a landmark study on lending bias. That study used regression analysis to study the role of race in leading decisions.
During the last year, examiners have field-tested a model developed by the New York Fed. It was integrated into the agency's exam procedures for larger lenders last month.
Lenders Know the Score
All of the bank regulators have implemented new programs to more carefully scrutinize lenders for illegal discreimination. But the Fedhs new approach is the first to rely so heavily on computerized regression analysis.
Lenders are well aware that regulators are beefing up their procedures to catch discrimination. But they fear that as regulators and law enforcement officials rely more heavily on statistical analyses, all judgement eventually will be squeezed out of lending decisons.
"We've got to strike a balance between ensuring equal treatment and not wiping out the loan officer's ability to use his or her judgment," said Virginia Stafford, a spokeswoman for the American Bankers Association.
Denial Rate Just One Aspect
Until recently, rgulators focused only on minority denials when checking for discrimination in exams. After deciding that this process didn't pick up differences in treatment between minorities and whites that could be discriminatory, regulators adopted a new technique: comparative loan file analysis.
In that procedure, applications by minorities and whites with simillar backgrounds are compared to see if they have been treated differently.
The Fed's new model allows computers, rather than examiners, to go through the bank records to fidn the matched pairs. This wil enable examiners to scrutinize many more - and more closely matched --pairs than they do now. And a more comprehensive look at the files will give them a better chance of detecting a pattern of discrimination.
A Chance to Explain
Under the new procedures, eaminers will review the files of applicants identified by the computer as possible victims of discrimination. Examiners will also give banks a chance to explain the disparate treatment -- a step activists have critized.
"I don't think going over it with the bank and giving them an opportunity to come up woth a reason after the fact is the way to do it," said Deepak Bhargava, legislative director for Acorn.
So far, Fed officials have found no suspected pattern of discrimination using the model. Agency officials are currently reviewing some cases for possible referral to the Justice Department, though.
And while they are confident enough of the model's accuracy to use it in exams at larger banks, Fed officials are still struggling to interpret some of the results. For example, the computer often pinpoints instances where white applicants appear to have treated less favorably then blacks.
Examiners are taking those cases into account when deciding whether bias is widespread, Fed officials said. Instances of disparate treatment among both whites and blacks would suggest that the problem is random and not illegal, they said.