Joint Comments re: Validation and Approval of Credit Score Models

Dear Mr. Pollard:

The undersigned Bankers Associations are pleased to submit the following commentspursuant to the Federal Register notice of proposed rulemaking dated December 21, 2018,implementing Section 310 of the Economic Growth, Regulatory Relief, and Consumer ProtectionAct of 2018 (Pub. L. 115–174).

The Proposed Rule would establish a framework for the housing Government Sponsored Enterprises (GSEs) to consider consumer credit scoring models in additionto the FICO score now serving as the current GSE and industry-wide standard.The Proposed Rule is mandated by Section 310 of the Economic Growth, Regulatory Reliefand Consumer Protection Act, one of the major accomplishments of the 115th Congress. In 2017,when the FHFA issued a Request for Information (RFI) on possible changes to the GSE’s creditscoring practices, the response of the banking industry was all but unanimous in urging eitherextreme caution or expressing outright opposition. As the current Proposed Rule appropriatelynotes: a “central theme from RFI respondents was that the operational challenges of implementinga multi-credit score approach would outweigh any benefits.”

Chief among these operational challenges are the costs that would result from having tointegrate a new, more-complex, and ongoing government-directed credit scoring system, and theresulting impact of those increased costs on the borrower’s ability to afford a mortgage. For thisreason, it is essential that a comprehensive cost-benefit analysis be done as a part of this process.

The American Bankers Association raised such concerns in response to the RFI and we areunaware of any market developments altering these previously indicated industry views.

As discussed in further detail below, our general view is that the major provisions of theProposed Rule are well-balanced and crafted as necessary to comport with the underlying statute.The current system relying upon FICO credit scoring has been in place for over 30 years and hasperformed very well in terms of its most important aspect, namely serving as a reliable indicatorof loan performance. Over the course of the recent mortgage crisis, FICO scores remained adependable source of objective data while other underwriting tools such as loan-to-value and debtto-income ratios misrepresented credit quality, especially in the non-bank lending sector.

Claims are now being advanced that “alternative credit scoring models” could qualifymillions of additional mortgage borrowers as contrasted with current industry standards. The basis for this appears to be the potential inclusion of individuals with sparse, inactive or even nocredit history at all. Qualifying eligible borrowers is an important public policy goal, and onewhich the banking industry strongly supports, but it is not accomplished best by lowering standardsto promote additional scoring outcomes. That actual result is much more likely to simply increasethe number of consumers with credit scores that are effectively lower, and who might then bedeemed subprime borrowers subject to the higher rates and fees, which were widely viewed ascontributory to the mortgage crisis. We support the development of predictable, reliablealternative models that can score borrowers even those who have spare or limited credit history. Ifsuch models can be developed they must be tested and implemented in an orderly process that doesnot add unnecessary costs for borrowers or pose undue risks for lenders or borrowers.

A more positive approach and a market development already underway is the utilization ofdata sources traditionally outside the scope of payment histories monitored by the credit reportingagencies. These include telephone and cable bills, rent payments and other indicators of anindividual’s reliable financial performance.

Another more recently announced initiative toward the goal of expanding credit availabilitythrough measured means is the incorporation of bank data into existing credit scoring algorithmsto create broader credit scoring input factors. While not intended as a substitute for the traditionalcredit score, under this approach, consumers would have an option to contribute records of positivechecking account practices and other asset information to supplement their current scores. Forthese reasons, we support §1254.11 of the Proposed Rule allowing the adoption of pilot programsthat can promote continued innovation in the credit decision-making marketplace.

Experience from the last mortgage crisis convincingly demonstrated that stretchingborrower qualifications has very real downside ramifications to borrowers, lenders and the publicat large. A very apt analogy is the credit rating “shopping,” which has been widely recognized asa leading contributor to the collapse of the private-label MBS market in 2007-08.

For these reasons it is especially important to cautiously proceed, both as to the FHFAdesignation of a credit scoring assessment process and any set implementation period for industryto effectively comply. The Proposed Rule properly noted industry concern in the feedback to the RFI, which recommended a phase-in period of approximately 18-24 months as necessary forindustry to adapt to a new credit scoring system.

Further analysis, in light of the many additional factors discussed at length in the ProposedRule, demonstrates that the adoption of multiple alternative credit scoring models is even morecomplicated than earlier examination. The first step will have to be the complex business judgmentas to whether a given bank would want to assimilate alternative scoring models just for itsmortgage originations intended for sale to the GSEs or across its full lending program. That issue,of course, turns on what options the third-party computer program vendors, which will benecessary in almost all situations, plan to offer and what the pricing parameters will be.

Upon completion of the initial rulemaking, there are indeed two distinct execution periods:(1) Outside program vendors will have to develop and market their compliant product range and(2) user banks will need time to select and assimilate any new program into their existing systems.In this context, the previously suggested 18-24-month preparation period strikes us as being on theshort side. In addition, this will, of course, be an ongoing process as new credit scoring companiesmay seek designation. For this reason, we also support §1254.5 of the Proposed Rule and wouldsuggest that it be a minimum of seven years between renewal periods

In general, our principal concern regarding this entire process is that the additional costsfactored into the mortgage origination process (i.e. new systems, employee training, recordkeeping, etc.) will become just one more factor driving community banks out of the mortgageorigination business. We have already seen how increased liability, narrow margins, and a vastlyexpanded regulatory environment have driven many community banks from mortgage lending.

There is also the inevitable confusion from the use of a different system which incorporates,as is now the case, the same numeric parameters as FICO. Section 1254.7(b)(4) would allow theGSEs to establish additional criteria for purposes of assessing credit score models, but a betterapproach would be further specification that new models be distinct so as not to be confusing toeither consumers or industry users.

The Proposed Rule also leaves open the question of potential variances or other lack ofalignment between the two GSEs in terms of their validation and approval outcomes.This would, in and of itself, be an immensely complicating factor for nearly all mortgage market participantsfrom primary market providers across the spectrum to secondary market investors.

Section 1254.6 sets forth a fair lending certification that any credit scoring model and thecredit scores it produces comply with “federal fair lending requirements.” This requirement isappropriate, but must recognize that federal fair lending requirements are in a considerable stateof flux. Since the Department of Housing and Urban Development initially finalized its 2013 rule implementing the Fair Housing Act's disparate impact standard, several Supreme Court caseshave been decided by way of delimiting that rule. These include, in 2015, Texas Department ofHousing and Community Affairs v. The Inclusive Communities Project, Inc., and, in 2017, Bankof America Corp. et al. v. City of Miami.

As a consequence of these more recent cases, HUD has already initiated an Advance Noticeof Proposed Rulemaking, seeking public comments on necessary revisions to its Disparate EffectsStandard Impact to bring them abreast of current law under the Federal Housing Act. Thisagency action is also being mandated by specific litigation on this subject. Our recommendationwould be that §1254.6 be modified to include the word “current” before the phrase “federal fairlending requirements” to ensure the fair lending certification requirement addresses the significantdifferences between the 2013 HUD Rule and the subsequent and controlling Supreme Courtprecedents.

We appreciate the opportunity to submit these comments. If you have questions or wouldlike to discuss any of these comments in more detail, please contact American BankersAssociation’s Senior Vice President and Senior Counsel, Joseph Pigg, at [email protected] or on202-663-5480. Thank you.

Download Comment Letter to view full list of signatories