The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Using Statistics Wisely Remarks by Lawrence B. Lindsey to the Mortgage Bankers Association of America New York City January 14, 1994 Thank you very much. It is a pleasure to be here today to discuss the issue of discrimination in home lending and what we can do about it. Discrimination attacks basic values we hold dear. does it tear at the fabric of our democratic society, Not only it also tears at the fabric of our faith in capitalism and the market. One of the great advantages of the market is that it is theoretically color blind. If that turns out not to be the case, then the foundations of our economic system as well as our political system are at risk. So overcoming discrimination is a fight that we as a society must win. How we fight that battle, however, is as important as its outcome. Actions taken without sufficient planning and thought often produce unintended consequences. Sometimes these consequences are as pernicious as the wrong we initially intended to right. The twentieth century is replete with examples of the havoc that can be wrought by those who believe that the ends always justify the means. This is particularly true when the means themselves are poorly understood. And frankly, statistics are among the least understood tools society has at its disposal. So, the focus of my remarks today will be on appropriate use of statistics in the battle against lending discrimination. Statistics have played a major role in our consideration of the mortgage discrimination problem of late. Their role as an enforcement tool is just now beginning, and is likely to increase dramatically in the years ahead. I approach this issue, frankly, as someone who loves statistics. I make my living using them. But, it is because of my familiarity with statistics that I am well aware of their limitations. For understanding the limitations of statistical analysis may be key to solving the underlying problem of discrimination and establishing truly equal credit opportunities for all Americans. While statistical analysis can highlight inequity, it cannot eliminate it in mortgage lending, or in any other field. That must be done on an individual basis, on the front lines. In lending, discrimination must be battled at the level of the applicant and the loan officer. However, the use of statistics can, and has, provided a baseline from which to start. data. Take for instance, the use of HMDA While community activists, bankers, regulators and legislators are all familiar with the limitations of the HMDA data, the HMDA data still indicate that there is a racially based problem in mortgage lending. Having said that, two important qualifications are in order. First, it is widely acknowledged that the HMDA data exaggerate the extent to which approval rates differ for racial reasons. When economic factors other than income are incorporated into the analysis of HMDA data, the disparity between black and white approval rates is sharply reduced. For example, the study by the Federal Reserve Bank of Boston indicates that even in the absence of discrimination, the rejection rate for minorities would be roughly twice that for white applicants. 2 The reasons for this disparity are due to criteria not included in HMDA data statistics, as well as income differentials. Second, the evidence of race-based differences in loan approvals is overwhelmingly of a statistical nature, based on racial averages. Discrimination is very hard to document by examining specific loan applications, such as during the bank examination process. Accepting this fact is difficult for those who seek simple, straight-forward explanations for the racial disparities. It's always easier when there's a smoking gun and an identifiable culprit. We learned from the Boston study, for example, that what I would call "old style" discrimination was not present. That is, clearly qualified applicants of any race were approved for loans and clearly unqualified applicants of any race were rejected. The days when members of minority groups who meet all of a bank's criteria for lending are rejected anyway seem to be gone. I believe that is why so many bankers believe so strongly that they do not discriminate. However, what the study also found was that a careful statistical comparison of applicants who were less than ideal indicated that imperfect white applicants were more likely to be approved than imperfect black applicants. Disturbingly, this occurred even though the institutions in question all have stated policies against discriminatory practices. In sum, such differential treatment apparently affected about 7 out of every 100 minority applicants. I believe that the magnitude of this discrimination, although distressing, suggests that the problem can be corrected with some carefully focussed adjustments in institutional behavior. Last April, the Federal Reserve Bank of Boston put out a pamphlet on these remedies called Closing the Gap; A Guide to Equal Opportunity Lending which I commend as important reading for all individuals in the financial services industry. Let me also stress that as long as behavior exists which appears outrageous to reasonable individuals, the threat of legislative and/or regulatory action, with all of its attendant burdens remains likely. Banks have a responsibility not only to end the practice of discrimination, but end the appearance that discrimination is occurring as well. As long as large numbers of minority customers remain dissatisfied with the treatment they receive, greater regulation remains a likely prospect. Or, as President Jordan of the Federal Reserve Bank of Cleveland has argued, "This problem is not solved until everyone agrees it is solved." This suggests that greater focus on sensitivity by loan officers is key. A friend of mine from my White House days told me of the experience she and her husband had in applying for a mortgage. Both are black and earning salaries at least commensurate with what Fed governors get. When they showed their tax return to the loan officer, he looked at them, at their tax return, at them again, and said, y'all are doing pretty well aren't you? Such unprofessional behavior has no place in any 4 service business. Regrettably, that is far from the only anecdotal story of discrimination I have heard. The existence of such stories, along with the difficulty in documenting discrimination on a case-by-case basis is leading to an increasing reliance on statistics based enforcement. The potential unintended consequences from this approach are enormous. Left unchecked, a total reliance on statistics in credit enforcement may ultimately lead to a complete replacement of bank judgment and reason regarding loan approval with statistical rules. I fear that in some instances, the use of statistics to establish discrimination may already go too far. At the Federal Reserve we are trying to avoid that problem. Our approach is to use computer based statistical models as a part of our examination process. However, these models are only used to select particular loan applications to examine more closely. The statistical models in and of themselves will not, and should not, be used to determine whether discrimination exists. Instead, the computer will select individually matched pairs of actual applications which will be studied further. We believe that this will improve the efficiency of the examination process by reducing randomness in selecting applications to be examined. The potential overuse and abuse of statistics creates problems in at least two ways. First, the use of statistical models as the sole criteria, especially when the details of such models are unknown to the banks being examined, means that no 5 bank can know what rules it actually has to comply with. It would be like replacing the speed limit on our nation's highways with some computer determined "Conditions Adjusted Velocity" formula in order to enforce traffic laws and not tell motorists what the Conditions Adjusted Velocity formula was. Laws can only work if people know what they have to do to obey them. Second, the likely result of statistics based examination of loan approvals is statistics based approval of loans. This, in turn is likely to work against individuals who do not meet the "normal criterion" of a one-size-fits-all statistical rule. One need only look at the historic performance of the secondary market to see that minorities and other disadvantaged groups find themselves only further disadvantaged by such inflexible practices. So great care must be taken in the years ahead. The unflagging efforts of bankers to eliminate both the practice and the perception of discrimination will be critical to success. Remember that you are a service business and that the customer - all customers -- come first. But, all parties involved in this volatile and emotional issue must practice in their professions what physicians, in taking the Hippocratic oath, practice in theirs -- above all do no harm. Above all, this means that any regulatory or legislative "fix" must be carefully and thoroughly considered. The potential for pernicious, albeit unintended, consequences is too great. 6
@FedFRASER