The Consumer Financial Protection Bureau, under fire from both industry and Congressional constituencies for using unexplained statistical methodologies to reach conclusions that auto lenders have engaged in unlawful credit discrimination, issued a report on September 17 that provided details of its methodologies and argued for their validity.
The CFPB and other federal regulators have generally taken the position that a lender can be shown to have engaged in unlawful discrimination against protected groups through use of a statistical analysis. The legal theory behind this approach is controversial, as the CFPB takes the position that a lender can be shown to have discriminated on the basis of race, ethnicity, or gender based simply on an after-the-fact statistical analysis, without there being any proof whatsoever that the lender intended to discriminate.
Compounding the controversy is that the statistical methods used by the CFPB and other regulators are, at best, imperfect. As noted by the CFPB in its report, most auto lenders are prohibited by law from collecting demographic data on borrowers. Therefore, the CFPB’s statistical analysis depends on various attempts to identify the gender, ethnicity, or race of a consumer based on “proxy” data. According to the report, the “proxy” data used by the CFPB include the consumer’s first name and address. In the report, the CFPB touts its “proxy” approach as providing a high degree of “correlation” or accuracy in identifying the demographic characteristics of a consumer, ranging from a “correlation” of as little as 0.06 for “American Indians/Alaska Natives” to 0.83 for “Asian/Pacific Islanders.” However, none of these “correlations” are “perfect” according to the CFPB, meaning that the CFPB itself admits that its approach does not come close to 100 percent accuracy.
As noted in two previous blog entries in March and July, the CFPB has been under fire for failing to explain its statistical approach, even though it was using its unexplained approach to support accusations of credit discrimination that have extracted, so far, more than $100 million in penalties and consumer “remuneration” from auto lenders. The release of the report may quell accusations that the CFPB has been breaching its own promises of transparency in this area. However, the report is unlikely to settle the more fundamental controversy over the basic legal foundation of the CFPB’s approach – that a company can be proven to have discriminated based on statistics alone. Finally, the many critics of the methodology are unlikely to be satisfied by the report, given the admitted substantial shortcomings of the CFPB’s proxy methodology.