WASHINGTON - To all the statistical studies suggesting racial bias in lending, Anthony Yezer has a simple reply: Phooey.
Mr. Yezer, an economist at George Washington University, maintains that the analytical techniques used for these studies are fatally flawed.
To make his point, Mr. Yezer created an "ideal" bank in his computer. The bank was fundamentally incapable of discrimination. Yet when he put the bank through the standard statistical test for lending bias, it flunked.
That, Mr. Yezer says, is "a false positive" - and one that should blow the credibility of the standard test. That test isolates every factor in the underwriting decision, except race. Experts then look for disparities based on race.
"There is just no excuse for anyone who is informed about the literature to use this technique," Mr. Yezer said. "But, since they are using the technique, even though we told them not to, we have demonstrated that it produces these false positives. It is a double-whammy argument."
The issues are hardly academic. Statistical studies have come to play a major role in fair-lending investigations by the Justice Department and civil rights groups. If Mr. Yezer is right, banking advocates say, the industry stands a much better chance of defending itself in court.
Mr. Yezer, of course, is not the first to question the barrage of fair- lending studies that have come out over the past few years. Many bankers, for example, argue that the studies focus too heavily on rejection rates, thereby punishing banks that reach out to minority communities.
Another common complaint: the data used for the studies, gathered under the Home Mortgage Disclosure Act, are unreliable.
Mr. Yezer, who outlined his views in the November issue of the Journal of Real Estate Finance and Economics, brings a new issue to the table.
Mr. Yezer, along with fellow George Washington economists Robert F. Phillips and Robert P. Trost, says that conventional analysis of lending data has ignored the ability of borrowers to actively improve their application's chances of success. In other words, the analyses have falsely assumed that banks have total control over whether or not they grant a loan.
In fact, banks don't have that control because borrowers can secure gift letters, find co-signers, or reduce their credit card debt. Census data, Mr. Yezer said, show that minorities have a harder time boosting their applications in this manner.
Also, Mr. Yezer said, the standard formula ignores the role that real estate agents and other professionals play in the mortgage equation.
Mr. Yezer maintains his computer work with the ideal bank shows that the current test fails to account for such factors, and thus is flawed.
But the authors of the previous fair-lending studies are unconvinced. They argue that Mr. Yezer himself is working with a faulty assumption: that borrowers know their chances for rejection. Such knowledge is needed for a borrower to know whether he had to enhance his application.
The assumption doesn't make sense because no borrower would apply for a loan if he knew he would be rejected, said Bernard R. Siskin, who runs the Center for Forensic Economic Studies in Philadelphia and serves as a consultant to the Justice Department.
"It is a good piece of work," Mr. Siskin said of the study. "It is an interesting mathematical model. I just don't believe in most cases that the assumptions of the model are appropriate."
William Milczarski, a professor of Urban Affairs at New York's Hunter College who organized the National Community Reinvestment Coalition's recent report on the nation's "Worst Lenders," questioned whether Mr. Yezer was going too far with his argument. He likened it to people who still claim that studies showing a link between smoking and cancer are flawed.
"Whether any statistical evidence is ever satisfactory enough will always be debatable," he said.
Mr. Yezer, however, is not completely alone in his approach. Economists at the Federal Reserve Bank of Cleveland published a similar study earlier last year. Those researchers found that the technique regulators use when they examine banks tends to find bias when none exists.
In Mr. Yezer's view, the shortcomings of the techniques are perhaps most evident in the issue of loan co-signers. He says the more complex analyses, including studies commissioned by the Justice Department, have failed to give any weight to co-signers.
"Presumably any sensible banker will say getting a co-signer reduces the chance of rejection," Mr. Yezer said. "What this shows is that the co- signer has no effect on rejection. That by itself is an absurd result."
Mr. Yezer does not limit his criticism to just the sophisticated studies. He also blasted the numerous reports that community groups have issued during the past several months.
"They have every right to publish any study they want," Mr. Yezer said. "But, I don't understand why people pay attention to it."
The main flaw with these studies, he says, is heavy reliance on Home Mortgage Disclosure Act data - statistics that researchers have been questioning for years. Critics say the information is at best imprecise on such matters as credit problems and relative wealth.
So what technique does Mr. Yezer suggest for probing for discrimination?
For starters, he says, focus on loan applications made through banks that compete in the same neighborhood. If more minorities apply to one bank over the other, than there is a chance that one institution has a lending problem, he said.
Or, he said, groups could look at the default rate on loans to minorities. The rates should mirror the default rate for the entire mortgage portfolio. If the rate is abnormally low, then the bank is picking only the most creditworthy minority borrowers - and thus possibly discriminating, he said.
At best, however, such excercises will yield only indications of potential discrimination, he says.
The only way to truly detect bias, he says, is through bank examinations or programs that send decoy loan applicants into lending offices.
"That's the way to do it," he said.
