Comment: Statistical Anomaly Penalizes Fair-Lending Effort

Beginning with a 1993 Washington Post series on racial disparities in mortgage rejection rates at banks in the Washington area, there has been increasing recognition that the disparities may result from the even-handed application of traditional lending criteria, which minorities have a harder time satisfying.

In March 1994, 10 federal agencies involved in monitoring fair-lending laws issued a policy statement placing greater emphasis on the disparate impact lenders' policies may have on minorities. The statement noted, for example, that minimum loan amounts may have greater adverse effects on minorities and indicated that banks would have to justify their use of such criteria.

Banks now face a perplexing dilemma. There is reason to expect that the more banks relax their lending criteria, the greater the disparities will be.

The statistical reason is rudimentary. When two groups differ relative to some desired outcome, the disparity in failure rates will rise as the overall failure rate shrinks.

The tendency can be easily illustrated with data on test scores. If a cutoff is lowered, for all who take the test, the disparity in pass rates declines, but the disparity in failure rates increases. This is why the lowering of cutoffs is universally regarded as a way of reducing the discriminatory effect of tests, though it increases the disparity in failure rates.

Among higher-income mortgage applicants (where overall failure rates are relatively low), the disparities in approval rates tend to be smaller, and the disparities in rejection rates tend to be larger, than among lower- income applicants (where overall failure rates are relatively high).

The tendency is also observed in data on particular banks. Banks with high rejection-rate disparities tend to have low disparities in approval rates. They also tend to be the banks where blacks, like whites, are most likely to have their applications approved.

Regrettably, however, it seems that practically no one monitoring bank lending practices understands this tendency. Consider the recent experience at NationsBank, one of the country's largest and fastest-growing financial institutions. An outspoken supporter of the Community Reinvestment Act, NationsBank has aggressively marketed in low-income and minority neighborhoods and has adopted programs to improve the likelihood that minority applicants will qualify for loans.

But it has not fared well in a number of recent studies and may not fare so well in court. In August 1995, the Black Caucus of the Teamsters union released a study on Washington and three other cities where NationsBank has major market share. Finding that NationsBank had one of the worst records for racial disparities in rejection rates in each city in each of three income categories, the study called on the Justice Department to investigate the bank's lending practices.

Yet the study overlooked entirely the fact that most of the time NationsBank tended to have lower acceptance-rate disparities and lower overall black rejection rates than the average for all banks in each of the cities.

The study also found that racial disparities in rejection rates were greater in the higher-income categories, a fact the study interpreted as refuting NationsBank's claim that its large rejection disparities resulted from aggressive marketing in low-income communities.

In reaching that conclusion, however, the study overlooked the fact that rejection-rate disparities are almost always greater in higher-income categories. Shortly after the Teamsters study was released, 11 plaintiffs represented by the Washington Lawyers Committee for Civil Rights and Urban Affairs filed suit, alleging that NationsBank and its subsidiaries had discriminated against black mortgage applicants throughout the Washington area.

Though the complaint cited the findings in the Teamsters study, it was principally based on a study of large lenders in the District of Columbia area that the Washington Lawyers' Committee had itself released at the end of 1994.

That study found that, at NationsBank, the black rejection rate was 5.4 times the white rate, the largest disparity among 13 banks reviewed. Focusing on rejections for poor credit, the area where the Department of Justice had found discrimination most likely to occur, the study also found that NationsBank had by far the worst record of the four banks with large numbers of minority rejections that provided information on reasons for the denials.

NationsBank denied black applicants mortgages because of poor credit ratings 15 times as often as whites; at none of the other banks were blacks denied loans for poor credit more than four times as often as whites. Efforts to adjust for applicant income did not materially change these figures. Yet though unmentioned in the text of the study, a table showed that NationsBank had the third-lowest black rejection rate of the 13 banks.

A table also showed that the black rejection rate for poor credit at NationsBank was only 10.3%, while at none of the other three banks on which similar data were presented was that rate below 24%.

The court refused to certify the case as a class action for technical reasons. Nothing, however, would preclude other members of the putative class from filing new suits and making a timely motion for class certification. If that happened, the bank would seem to have much to fear in court. The case will be tried by a jury. The statistical tendency described above has eluded experts. It is hard to expect much greater understanding from a jury.

Since the case probably presages a great deal of litigation based on similarly appealing - but seriously misleading - statistical disparities, how the courts and jury deal with such statistics warrants close attention.

Also warranting attention is how the court and jury deal with the types of evidence detailed in the complaint of the 11 plaintiffs. All had problems in their credit histories. But all claimed they had reasonable explanations for the credit problems that led to the rejections but were given no opportunity to explain.

The complaint maintains that whites would have gotten loans in similar circumstances. Assuming that persons of all races with like credit histories had, say, even a 10% or 20% chance of receiving credit, one ought to be able to find among the almost 14,000 whites in the area to whom NationsBank did grant mortgages a fair number whose credit histories appear no better than those of the rejected black applicants.

The evenhanded exercise of flexibility by loan officers would be expected to increase further the racial disparities in rejection rates. So until there is more widespread understanding of the statistical issue, banks willing to do the most to promote lending to minorities may find themselves increasingly the objects of misguided enforcement efforts.

Mr. Scanlan is of counsel with the Washington law firm Akin, Gump, Strauss, Bauer & Feld.

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