When Justice Department fights bias by the numbers, they're his numbers.

WASHINGTON -- When the Department of Justice attacks a bank's fair lending record, Bemard R. Siskin provides the ammunition.

Mr. Siskin, a statistician who runs the Center for Forensic Economic Studies in Philadelphia, has been the consultant in nearly all of the Justice Department's fair-lending prosecutions.

His studies were the basis for several suits, including a complaint against Decatur Federal Savings and Loan that the institution agreed to settle. He also prepared a similar analysis in a case against Shawmut National Bank.

Most recently, he prepared a study of Chevy Chase Federal Savings Bank, a Maryland thrift that signed a consent decree last month agreeing to open more off'tees in minority areas in Washington. Chevy Chase settled the case, even though his analysis did not show any lending bias.

And his job is not done. The onetime Temple University professor said he's working on several current analyses for future prosecutions. He declined to name the institutions he's reviewing.

But bankers clearly appeared concerned about Mr. Siskin's work, which is genetically known as regression analysis. For example, bankers asked more than a half dozen questions about regression analysis at a recent Consumer Bankers Association conference.

One of Mr. Siskin's parme,s, Leonard A. Cupingood, said Mr. Siskin is considered a pioneer in the regression analysis area.

"He was really one of the first people to apply statistics to the courtroom;' Mr. Cupingood said.

Paul Hancock, who runs the housing section of the department's civil rights division, praises Mr. Siskin, calling him an "expert in his field."

He even gets support from his former employer. Burt Holland, the chairman of Temple's statistics department, said Mr. Siskin's reputation is known throughout the East Coast.

Over the past decade, he has participated in more than 100 cases, most of them involving employment discrimination claims. Bankers probably are wise to avoid confronting Mr. Siskin in court. Stanley A. Twardy, a former U.S. Attomey who now works at Day, Berry & Howard in Condecticut, said jurors are prone to believe testimony from people like Mr. Siskin.

"As with any other person qualified as an expert, jurors tend to accord them extreme deference," Mr. Twardy said. "So, if an expert reaches a conclusion, then the jurors don't look at how he reached that conclusion."

But, litigation can be a dirty business, and Mr. Siskin has picked up his share of battle scars. Several federal judges have questioned the integrity of his work, writing in published opinions that he was not a credible witness.

"The court seriously questions the academic honesty of Dr. Siskin, in part, because he testified one way at trial knowing that his own computer data -- later filed under court order -- showed otherwise," U.S. District Judge Wilbur D. Owens Jr. of Georgia wrote in the 1986 employment discrimination case of Calloway v. Westinghouse Electric Corp.

Others questioned his methodology. Judge Harlington Wood Jr. of the U.S. Court of Appeals for the 7th Circuit wrote in the 1988 case EEOC v. Sears Roebuck & Co. that Mr. Siskin did not include enough variables in his study to prove employment discrimination.

Mr. Siskin ran into similar problems in Rhode Island. U.S. District Judge Bruce M. Selya, since elevated to the federal appellate court in Boston, wrote that Mr. Siskin failed to include variables, such as prior experience, in his analysis. These other factors could explain why certain professors were not promoted, the judge wrote.

"He excluded salient variables even though he knew of their importance, failed to include all individuals in his study by reason of incorrect assumptions, and compounded his sins by overreaching in a vain try to explain away results which did not support the defendants' theory of the case," Judge Selya wrote.

Mr. Siskin takes this criticism in. stride, saying "it's going to happen."

He said judges have mentioned his work in more than 95 opinions. Of those, judges commented favorably in 90 and expressed no opinion in two. That's an acceptable ratio, he said.

Mr. Siskin said his methodology is sound, arguing that it is impossible for him to skew the outcome because the computer assesses all the variables without fear or favor. All he does is input the information, he said.

"Ultimately, in a sense, there is a truth to the data," he said. "Either it is there or it isn't. You can't make it up."

Mr. Siskin said his system uncovers subtle biases that many banks may not even know about. He said he rarely finds cases where banks rejected qualified minority applicants.

Instead, he said, he learned that banks often accept unqualified white applicants and reject unqualified minority applicants. Mr. Siskin said that even though both groups don't deserve loans, the computer determines that banks are discriminating because they employ different standards for both groups.

To uncover this information, Mr. Siskin and his staff of 30 people utilize sophisticated computers and the same software that social scientists use for their studies

In simple terms, he tries to identify the factors a bank uses to make a decision on a loan and then uses the computer to see how much weight the lender places on each. With that information. he can "reprocess" rejected minority applicants to see if the computer model would approve them.

If the bank does discriminate, then the computer will catch the institution because it will approve some of the rejected applicants.

"It's never going to predict perfectly," Mr. Siskin conceded. "But the models tend to predict pretty well."

Mr. Siskin said he often alters the program, at a bank's request, to include additional variables. He said this hasn't yet changed the program's results.

While the programs may give the government the answer to which banks discriminate, they are expensive to run.

Several people familiar with the Justice Department probes said the Decateur study cost $1 million. Mr. Siskin does concede that the Decateur case, as the first fair-lending probe, was expensive. Since then, the costs for these studies has dropped, he said. He declined to provide exact figures.

Mr. Siskin entered the consulting arena on a full-time basis in 1984 when he helped create the Center for Forensic Economic Studies. He sold off his half of the business two years later to National Economic Research Associates Inc., a national research firm.

"They made me an offer I couldn't refuse," he said.

He didn't stay for long, however. He said the tremendous amount of travel finally caught up with him, causing him to rejoin the center in 1991.

Throughout this period, he said, he consulted for the government on employment

discrimination cases. These days, he said, is working for the Federal Bureau of Investigation, evaluating claims that the bureau does not treat minority agents fairly.

He taught at Temple from 1973 to 1984 as a tenured professor of statistics, and co-authored five books, all on statistics. And, he's addressed dozens of groups, varying from Women's Law Caucus to the Institute of Industrial Research.

Mr. Siskin said his fair-lending career should be short-lived.

"I think this area is going to go away soon," he said. "I think the government is going to continue to prosecute, but I think the banks are going to respond and make the problem go away."

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER