The Appendix to the current version of the Interagency Fair Lending Examination Procedures (published by the Office of the Comptroller of the Currency (OCC), Federal Deposit Insurance Corporation, Federal Reserve Board, Office of Thrift Supervision, and National Credit Union Administration) provides guidance for augmenting a statistical analysis of an institution’s fair lending practices with comparison of the files of minorities whose applicants are denied with the files of white applicants whose applications are approved. The Appendix even states (at 6) that appraisal of an institution’s self-evaluation procedures should consider whether the self-examination procedures include such comparisons. The recommending of such file comparisons dates at least as far back as April 1993, where it appeared in the Interim Procedures for Examining Racial and Ethnic Discrimination in Residential Lending, issued by the OCC’s Department of Compliance Management.
Such comparisons, while providing no useful information concerning the fairness of a lender’s practices, can be expected to commonly yield misleading information about those practices. Whenever there are large numbers of applicants from different racial groups and there is any flexibility in a process – say, of a nature whereby less than 100% of applicants with certain characteristics are offered the desired outcome – one commonly will find some situations where minority applicants who experience an adverse outcome have the same characteristics as white applicants who experience a favorable outcome. Were one to make the effort, one commonly would also find some situations where white applicants who experience an adverse outcome have the same characteristics as minorities who experience a favorable outcome.
Of course, when there is discrimination against a group there will exist a statistical pattern adverse to that group with regard to comparative proportions of applicants with a particular group of characteristics who experience the favorable outcome. But such pattern can only be identified in a statistical analysis. Indeed, other than in cases where bias is explicitly manifested, discrimination can only be soundly identified statistically. Statistical analyses, however, must be carried out with recognition of the statistical issues addressed on other subpages of the Lending Disparities page and other pages of this site.
The reader may question that agencies of with the presumptive expertise of those identified in the first paragraph would devote resources to a fundamentally unsound exercise. It should be borne in mind, however, that these same agencies have at least since 1994 been encouraging lenders to take actions that will reduce the frequency of adverse lending outcomes; but, unaware that reducing the frequency of adverse lending outcomes will tend to be increase relative differences in those outcome, those agencies have continued to monitor fair lending compliance on the basis of relative differences in adverse lending outcomes (and regarding the size of the difference as an indicator of the likelihood of discrimination). Hence, the agencies have encouraged lenders to engage in conduct that makes them more likely to be sued. See, e.g., ’Disparate Impact’: Regulators Need a Lesson in Statistics” (American Banker, June 5, 2012) and “When Statistics Lie,” (Legal Times, Jan. 1, 1996). See generally the Lending Disparities page and its sub-pages. So the recommendations of these agencies regarding file comparison are among a number of things about the proof of discrimination that these agencies do not understand.
The documents mentioned in the first paragraph are focused on examination of files of rejected and approved loan applicants. That focus reflects the concerns about racial differences in mortgage rejection rates that were the principal fair lending issue in the 1990s. According to an Update on Fair Lending Exams made available on BankersOnline.com, more recent file comparisons by regulators have involved the files of approved applications, with the focus on differences in terms. Such change accords with the recent focus on racial differences in assignment to subprime status and racial differences in loan terms and fees, such as were the subjects of $335 million and $175 million settlements of the Department of Justice’s lending discrimination cases against Bank of American’s Countrywide Financial Corporation and Wells Fargo Bank (which cases are the subjects of the United States v. Countrywide and United States v. Wells Fargo subpages).
These comparisons, however, raise the additional issue involving the ignoring of offers to applicants who declined loans. A sound statistical analysis would have to include these applicants, as discussed on the Partial Picture Issues subpage.