CFPB Highlights Innovative Uses of Data and Machine Learning in Credit

On Tuesday, the Consumer Financial Protection Bureau (CFPB) released a noteworthy blog post providing an update on key findings about the use of alternative data for credit access in connection with the Bureau’s first no-action letter (NAL).  The post, written by top agency officials, discusses a NAL that the CFPB issued in September 2017 to Upstart Network, Inc., a company with an online lending platform that utilizes “alternative data” and machine learning to make credit decisions.  The NAL required Upstart to submit data to the CFPB about its lending decisions and outcomes, and in its post, the CFPB noted that the data furnished by Upstart is promising:  Upstart’s testing of its own data indicated that its models resulted in approvals for 27% more applicants than traditional models and yielded 16% lower average APRs for approved loans.  And, in a particularly important finding, this result held true “across all tested race, ethnicity and sex segments.”  For this reason, the CFPB, in its post, “encourages lenders to develop innovative means” of providing access to credit – which may include leveraging alternative data sources for consumers’ benefit.      

As background, a NAL is a tool that the CFPB uses to express that “the [CFPB] staff has no present intention to recommend initiation of an enforcement or supervisory action against [a party] with respect to a specified matter.”  NALs are meant to be used with regards to novel product or services “that promise substantial consumer benefit.”  The CFPB may issue a NAL to provide some level of certainty.[1]

The CFPB entered into the Upstart NAL based on a desire to “explor[e] ways that alternative data may be used to improve how companies make lending decisions.”  Generally, lenders rely upon “traditional factors” to determine whether they will give someone a loan and at what rate they will do so.  These factors include information contained in a traditional credit report, such as credit score, income, and payment history.  The “alternative data” utilized by Upstart refers to additional variables generally not reflected in a traditional credit report, such as electronic deposits, withdrawals, and transfers.  The NAL essentially gave Upstart the (nonbinding) green light to use both traditional and alternative data in its lending decisions.  In exchange, Upstart agreed to provide the CFPB with information about its lending decisions and outcomes. 

The results are encouraging for companies looking to use alternative data sources in financial services.  By broadening the set of variables used to assess creditworthiness, the pool of consumers eligible for credit will hopefully correspondingly broaden.  This is important, as the CFPB estimates that 26 million Americans are “credit invisible, meaning they have no credit history with a nationwide consumer reporting agency,” and another 19 million do not have enough credit history to produce a credit score, making it very difficult to borrow money.  And the NAL results show that this can be done in a way consistent with fair lending laws. 

The CFPB has proposed a number of steps moving forward to encourage this kind of consumer-friendly innovation.  For example, the agency has proposed revisions to its NAL policy as a way “to increase participation by companies seeking to advance new products and services . . .”   Navigating the regulatory landscape—particularly around innovative uses of data—remains challenging, but the CFPB’s latest post reinforces how companies can move forward with innovative products while maintaining regulatory compliance. 

[1] Under the CFPB’s current NAL policy, NALs are subject to many limitations, and they are not binding on other agencies, enforcement officials, or courts.