VIA ELECTRONIC SUBMISSION Monica Jackson Office of the Executive Secretary Consumer Financial Protection Bureau 1700 G Street, NW Washington, DC 20552 Dear Ms. Jackson: May 19, 2017 The undersigned, a group of companies in the financial services technology industry that help consumers and small businesses improve their financial well-being and manage their financial needs, which we call the Consumer Financial Data Rights Group ( CFDR Group ), respectfully submits the following response to the Consumer Financial Protection Bureau s ( CFPB ) Modeling Techniques in the Credit Process ( RFI ). The CFDR Group appreciates the opportunity respond to the CFPB s RFI and supports the Bureau s efforts to explore developments in the credit market and ways to expand access to affordable credit for U.S. consumers. The Use of Data from Digital Financial Records Benefits Lenders and Consumers It is useful to distinguish digital financial records ( DFR ), essentially, digitized versions of consumer bank statements, from other forms of alternative data. Today, DFR can be accessed with a consumer s permission directly from such consumer s bank account. Thanks to advances in data aggregation, lenders can access standardized forms of this consumer-permissioned data directly from the ledger of most financial institutions in the U.S. in real time. This type of data differs from alternative data obtained housed at credit bureaus or reported directly by a consumer. Using DFR in credit decisioning results in improved outcomes for consumers and the lenders that serve them. The use of bank account data to determine a credit applicant s creditworthiness and ability to pay is well established; nearly every mortgage lender in the U.S. reviews the bank statements of applicants as part of the mortgage underwriting process. As consumer-supplied bank statements have not typically followed a standard format, review has historically been a time-consuming manual process that is uneconomical for small dollar loans. For this reason, companies that offer credit cards and other small dollar loan products today tend to leverage fully-automated underwriting processes using credit scores and traditional credit data. With the digitization and standardization of bank statements in the form of DFR, more in-depth data analysis is now possible, which can particularly assist credit invisible and thin-file applicants. Such information, when considered in combination with traditional underwriting variables, provides greater insight into an applicant s credit risk profile than using traditional methods alone. This results in expanded credit access, increased predictiveness of
creditworthiness, lower costs (which can be passed on to the consumer), greater transparency, more timely information and more convenience for the consumer. The use of DFR data in underwriting has the potential to expand access to credit for the 45 million U.S. consumers who lack traditional credit scores. 1 As the CFPB notes in the RFI, this group is disproportionately comprised of Black,Hispanic, low-income and young consumers. Without a traditional credit profile, these underserved demographics face limited access, higher prices, and exclusion from mainstream and high quality credit options. This may constrain their ability to borrow at competitive rates or to leverage credit to build assets, particularly when it comes to purchasing a home. Moreover, instead of relying on consumer s self-stated income at the time of the credit application, which may not be accurate or complete, DFR data allows lenders to verify an applicant s real ability to pay. This can help lenders determine whether an applicant can actually afford a loan, rather than focusing only on his or her propensity to repay it. This level of insight provides a holistic view of an applicant s financial situation, rather than a static snapshot of one aspect of his or her credit lifecycle. Additionally, DFR data lowers risks for lenders, which can decrease the cost of credit for consumers. The use of consumer-permissioned DFR information also increases transparency in the lending process. By providing direct consent to the use of their data, consumers gain more visibility into the information that is being passed to lenders than would be available if such information were collected by third parties. Thus, not only does using alternative data in underwriting increase access to affordable credit, it does so in a transparent and secure way that keeps consumers empowered and informed about their credit application process. Many of the issues with traditional data and other forms of alternative data do not apply to DFR information: 1. Accuracy and Data Quality. Because DFR data comes directly from the ledger of a financial institution, it is subject to regulatory oversight, subject to internal/external audit, and eventually incorporated into public financial report. DFR, therefore, has a high degree of accuracy and integrity. 2. Privacy. DFR data obtained by way of data aggregation is in read-only format, consumer-permissioned, and can be accessed only after a consumer has already 1 CFPB, Data Point: Credit Invisibles (May 2015), available at http://files.consumerfinance.gov/f/201505_cfpb_data-point-credit-invisibles.pdf 2
opted-in to share the information with the lender and has provided his or her bank account credentials to do so. 3. Lost transparency, control and ability to correct. Consumers have full visibility into their own DFR, and, to the extent inaccuracies exist, regulated financial institutions, with controls and oversight, are incentivized to correct errors. In fact, it is easier in most cases for a consumer to access her or his DFR than to access a traditional credit report. 4. Ability to change credit standing through behavior. DFR data is specific to an individual consumer, unlike education or occupation data. Models that use DFR data also apply proper weighting to recent information and allow consumers to recover from a negative financial event much faster than traditional models. 5. Unintended or undesirable side effects. The fundamental economic information found in a consumer s DFR is inexorably tied to that consumer s ability to pay. 6. Discrimination. Underwriting models that incorporate DFR data should not include as inputs any prohibited basis as defined in Reg B and the Equal Credit Opportunity Act ( ECOA ) and may be empirically derived, demonstrably and statistically sounds (EDDSS). Industry practices that are currently employed with respect to traditional data models (e.g., validation, internal controls, monitoring, testing and audit) are equally applicable to underwriting models that incorporate DFR data. Existing industry practice is sufficient to mitigate discriminatory outcomes or fair lending risk. Existing Market Practice Will Prove Equally Effective when Used to Test Alternative Data and/or Modeling Techniques for Fair Lending Compliance The CFPB alludes in its RFI to the concern that the use of alternative data and modeling techniques may lead to violations of ECOA. ECOA and its implementing Regulation B, prohibit a creditor from discriminating against an applicant for credit on the basis of race, religion, sex, age, color, national origin, marital status, receipt of public assistance and exercise of certain legal rights. 2 Even if a creditor has no intent to discriminate, violations of ECOA and Regulation B could result if a lending practice has a disproportionately negative impact on a protected class, unless that practice meets a legitimate business need that cannot reasonably be achieved by means that are less disparate in their impact. 3 While ECOA and Regulation B apply to the credit decision regardless of whether traditional or alternative data and/or modeling techniques are used, we believe that current 2 12 C.F.R. Part 1002 Supp. I, 1002.4 4(a)-1. 3 12 C.F.R. Part 1002 Supp. I, 1002.6 6(a)-2. 3
market practice would be equally effective in testing for fair lending compliance regardless of the underlying data or modeling techniques used. Most non-mortgage lenders do not collect prohibited basis information in order to test for fair lending compliance, and instead rely on proxies to test for fair lending compliance. This same methodology could be used to test for fair lending compliance when alternative data and/or modeling techniques are used. In fact, the addition of alternative data to traditional data might result in a more robust data set that would make testing by proxies more effective. 4 Conclusion The emerging technology that allows for individual cash flow analysis of DFRs sets the stage for a radical change in how consumers creditworthiness could be assessed for small dollar lending. The availability and access to a consumer-permissioned DFR in credit decisioning results in improved outcomes for consumers and lenders alike, with potentially far-reaching implications for the financial resilience and stability for American families, particularly families that have been traditionally underserved by mainstream financial services. The results of leveraging consumer-permissioned DFRs in the credit lending process can include greatly expanded credit access for consumers, increased predictiveness of creditworthiness, lower costs (which can be passed on to the consumer), greater transparency, more timely information and more convenience for the consumer. Without continued access to consumer-permissioned DFRs, underserved consumers will continue to remain blocked from accessing affordable credit that might otherwise have been available if had lenders had the ability to access and to assess a consumer s more complete financial picture. The CFDR Group would welcome the opportunity to further discuss this important issue with the CFPB at your convenience. Please do not hesitate to contact Steven Boms (sboms@yodlee.com) at (202) 997-0850 if the CFDR Group can be of any further value in this discussion. 4 Devin G. Pope and Justin R. Syndor, Implementing Anti-Discrimination Policies in Statistical Profiling Models, 3 A.E.J. Econ. Pol. 206 (2011). 4
Sincerely, Affirm Betterment Envestnet Yodlee Kabbage Oportun Petal SnapCheck SoFi Wealth 5