As the debate over minority lending heats up, the devil may prove to be in the details.

It turns out that regulators, legislators, and lenders overlooked one possibility: that the numbers used to judge mortgage-lending performance could be inaccurate.

"All of the emphasis has been on policies," said Lucy Griffin, a compliance consultant for banks. "The caption coming to the bank CEO is |make fair lending a priority.' The message of |get your numbers right' is being lost."

Indeed, as bank examiners have begun to scrutinize the data more closely - comparing the information lenders submit with actual loan files - they have been shocked by the number of errors.

A Federal Reserve official said last month that the Fed has begun warning lenders about the problem and even threatening to levy civil money penalties against violators.

Problem May Be Widespread

Fed officials said they do not know how pervasive the problems could be. But public advocates who have scrutinized data supplied under the Home Mortgage Disclosure Act and the Community Reinvestment Act said the problem could be widespread.

"We've always been very suspicious of the data," said Hubert J. Van Tol, executive director of the Mid-South Peace and Justice Center in Memphis.

His organization has scrutinized Memphis lending data by census tract and found that in at least 50% of neighborhoods, the figures disclosed by banks conflicted sharply with data gathered from other government records.

Study in Pittsburgh

Likewise, in Pennsylvania, the Pittsburgh Community Reinvestment Group spotted widespread errors in data filed by 12 area banks. After comparing 1990 HMDA figures with county records, the group found that error rates ranged from 17% at the most accurate bank to a whopping 87% at the least accurate.

After sharing the findings with the banks, which had hired the organization to do the review, accuracy improved significantly. In 1991, error rates fell to between 5% and 50%.

Community groups say they have found every kind of error imaginable. The most common mistakes: miscoding loans to whites as minority loans, mischaracterizing loans in upperclass neighborhoods as low-income neighborhood loans, and including non-HMDA loans in the HMDA data base. In at least one case, for example, a loan to a car wash was counted as a home loan.

"We do not take at face value the loan application records that financial institutions supply to federal regulators," said Stanley A. Loeb, an official at the Pittsburgh community group.

The possibility that the HMDA data may have significant errors puts regulators in a tough spot. Given the magnitude of the HMDA effort, they are angry that some banks are not making a good-faith effort to submit accurate information. They want to send a message to the lenders that they need to shape up.

Data Base in Question

But in making an issue of the problem, they expose themselves to criticism on the integrity of the data base, which has been used as evidence that there is widespread discrimination against minority borrowers.

"In my mind, the regulatory agencies have a real dilemma," said David C. Fynn, corporate compliance officer of National City Bank. "They need to emphasize the accuracy of the HMDA data, but every time they talk about inaccuracies, they undermine its credibility."

Added Ms. Griffith: "The regulators are taking on two burdens they don't like: First, it takes a lot of time to correct errors, and that's expensive. Second, they risk taking pressure from Congress. And that's not a pleasant experience."

Some banks have taken strong precautions to make sure their data are as accurate as possible. National City Bank, for example, uses software developed by a private firm to check for accuracy. It also pulls a sample of loan files to compare with the computer-generated data. An error rate in excess of 3% generates a more thorough check of the records.

Careful review has enabled the agency to submit to regulators HMDA data that are "significantly accurate," Mr. Fynn said.

But facing a regulation already so costly and burdensome, many banks may be reluctant to implement a thorough quality-control program.

"This could potentially double the cost" of compliance, Mr. Fynn said. And regulators, ceaseless tinkering with HMDA rules makes careful review of the records difficult, he added.

Pitfalls in Handling Data

In making sure reviews are accurate, compliance experts recommend that banks review how their loan application records are developed and emphasize the importance of careful compilation of the data.

"Every time the data is handled, there's an opportunity for error," Ms. Griffin said.

And just as important, the experts say, is developing a system of frequent and thorough self-auditing, comparing data from loan application records with individual loan files. While time-consuming and tedious, this internal review could stave off complaints from bank examiners.

The possibility of prolonged scrutiny from regulators, and possibility even civil money penalties, should be enough to convince banks to look more closely at the integrity of their data, regulators said.

And if banks do not do this themselves, public advocates and community groups may do the job for them. John Taylor, executive director of the National Community Reinvestment Coalition, said the 250 groups in his organization are waiting in the wings to take action if the industry and regulators don't.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.