On February 21, 2017, the Consumer Financial Protection Bureau (“CFPB”) published a “Request for Information Regarding Use of Alternative Data and Modeling Techniques in the Credit Process” (the “RFI”). In the RFI, the CFPB sought information regarding the use and potential misuse of alternative data and modeling techniques in the credit process. The CFPB received an earful, including multiple responses from various marketplace lending participants, trade groups, consumer advocacy groups, and the major banking trade associations. While the comment period only closed on May 19, 2017, the industry is watching the CFPB closely to determine what, if any, actions the CFPB takes in response to the RFI.
The comments from various market participants highlight the current uncertainty in the industry regarding the regulation of the use of alternative data in lending decisions. If the CFPB publicly supports the use of alternative data, the affirmation could well open the floodgates for the widespread adoption of alternative data-based lending models throughout banking, including not only the marketplace lending participants but also credit models designed to be used in traditional bank lending as well.
In general, the RFI sought information regarding the use of “alternative data” as opposed to “traditional data” in making lending decisions. Under the definitions set forth in the RFI, “traditional data” refers to data assembled and managed in the traditional consumer reporting agency files, including repayment histories, credit inquiries, and public record matters such as judgments and bankruptcies. Alternative data, by contrast, refers to virtually any data that is not traditional data. While a thorough analysis of the types of alternative data available for consideration is beyond the scope of this article, the focus of the RFI was clearly on the use of alternative modeling techniques wherein lenders utilize multiple sources of data to make credit-based decisions through computer generated analysis, including algorithms and other programming techniques.
Through the comments, it appears that the lending industry is primarily concerned about the CFPB’s potential regulatory determinations relating to a few primary aspects of the use of alternative data. The areas of concern include (i) the application of the Equal Credit Opportunity Act and Regulation B, (ii) the CFPB’s subjective determinations under the Unfair, Deceptive, or Abusive Acts and Practices (UDAAP) standard, and (iii) the CFPB’s comments regarding a lender’s duty to verify a particular alternative data-based model.
The Equal Credit Opportunity Act (“ECOA”) prohibits discrimination in a credit transaction on the basis of race, ethnicity, sex, religion, national origin, or other prohibited bases. While lenders certainly acknowledge that they cannot intentionally discriminate on these prohibited bases, lenders have rejected the application of ECOA to claims arising from the use of non-discriminatory factors that result in a “disparate impact” upon protected classes of individual consumers. To date, the courts have held that the disparate impact analysis applies to claims relating to mortgage lending under the Fair Housing Act. However, disparate impact claims have not been widely recognized by courts in connection with ECOA claims, though some uncertainty continues in the area.
If the CFPB publishes guidance indicating that disparate impact claims are not recognized under ECOA, lenders may more readily accept alternative data-based lending models where the application of the model may have a disparate negative impact on a protected class. In any event, even in the absence of a disparate impact standard, lenders would need to be able to demonstrate that the alternative data model itself did not rely upon an assessment of prohibited factors in determining whether or not to extend credit.
In addition, the comments from the lenders demonstrate that further guidance from the CFPB regarding the application of UDAAP to alternative data would be helpful in facilitating alternative data-based lending models. Specifically, lenders cite the vague and subjective standards used by the CFPB in the past to find practices in violation of the UDAAP standard. Lenders, of course, seek more objective concrete standards to follow in order to comply with the UDAAP requirements. This type of standard may, however, be elusive given the nearly limitless sources of information available for consideration in making lending decisions.
Finally, the comments illustrate that lenders are concerned about the potential model validation requirements that may be imposed by the CFPB on alternative data-based lending models. In many cases, the lending models rely on multiple factors programmed into various algorithm and machine learning models generated by computer programmers. Given the complexity of the computer models, they are difficult to analyze and explain to regulators (and difficult for lenders to evaluate without retaining significant data analytic resources). For that reason, lenders would like guidance from the CFPB as to the model validation requirements that will be imposed in the event a bank adopts an alternative database lending model. Given the uncertainty over the current model validation requirements, many lenders are likely standing “on the sidelines” awaiting further guidance in order to ensure their lending programs will not be unduly criticized.
As noted above, the comment period for the CFPB’s RFI expired on May 19, 2017. The CFPB has not published any indication as to when it will provide further guidance on alternative data-based lending models. However, in a speech to the CFPB’s Consumer Advisory Board on June 8, 2017, Director Cordray acknowledged that alternative data-based lending models may help “invisible” consumers (those currently outside of the traditional credit reporting system) establish credit through the consideration of alternative data points such as payments made on rent or cell phone bills. The Director also acknowledged the closure of the RFI’s comment period and advised that the CFPB would have “more to say before long” on the use of alternative data models.
At this point, lenders would be advised to continue to follow these developments closely and to proceed with caution in connection with any alternative data-based lending models. Given the guidance to date, best practices would include closely reviewing the data sources and variables being utilized by an alternative data-based lending model in order to be in a position to explain the model to the lender’s regulator. The lender should also secure appropriate resources to assist the lender in analyzing the data and determining its accuracy, validity, and non-discriminatory bases. Finally, lenders should monitor the results of the alternative data-based lending model in order to detect any discriminatory treatment that cannot be explained utilizing business-based factors.
Written by Davenport Evans lawyer Keith Gauer, as published in the Independent Community Bankers of South Dakota newsletter, June 2017 edition. Sign up for ICBSD e-news here. Keith Gauer is a financial services attorney at Davenport, Evans, Hurwitz & Smith, LLP in Sioux Falls, SD. He can be reached at email@example.com or 605-357-1256.
Davenport, Evans, Hurwitz & Smith, LLP, located in Sioux Falls, South Dakota, is one of the State’s largest law firms. The firm’s attorneys provide business and litigation counsel to individuals and corporate clients in a variety of practice areas. For more information about Davenport Evans, visit www.dehs.com.