Earlier this week, the Consumer Financial Protection Bureau released a circular that could have widespread ramifications across the consumer credit industry, especially for users of AI, big data and engineered variables.
What does it say?
The circular emphasizes the need for specificity and accuracy when providing reasons for loan denials (also referred to as adverse action codes). On its face, this is nothing new: the circular simply reiterates the legal mandate that loan denial reasons be accurate and specific.
On the other hand, the circular also shines a spotlight on a widespread industry practice: mapping variables, including engineered and non-conventional variables, to the well-understood reason codes in the CFPB’s model adverse action notice templates. For instance, if a lender uses behavioral data, such as a consumer’s visits to online gambling sites, merely citing ‘purchasing history’ or ‘disfavored business patronage’ as a denial reason might now be seen as too ambiguous. This suggests the CFPB’s intent to challenge lenders who might be using certain data types perceived as ‘intrusive’ or ‘creepy’ and then hiding behind the vagueness of Appendix C’s reason codes.
Why Is it a Big Deal?
The practice of mapping variables to Regulation B reason codes is common in the lending industry. This is for several reasons including customer understanding – a belief that consumers might better understand traditional reason codes as opposed to highly specific or technical reasons derived from complex models. But the CFPB contends that this approach can oversimplify and obscure the true reasons for a loan denial.
How will behaviors need to change?
The circular could have a profound effect on the types of data lenders might choose to use and how they use them: While the circular doesn’t ban any specific type of data, it underscores the risks of using unconventional variables, engineered variables and variables that might be predictive of credit risk but lack an intuitive relationship to it. Engineered variables which often combine multiple raw data points or transform them in some way can be particularly challenging to explain in clear terms.
Similarly, variables that don’t have a clear connection to creditworthiness, even if they are predictive of risk, can be perplexing for consumers to grasp. When loan denials are influenced by these kinds of variables, lenders will have to deconstruct them to their fundamental components to offer a clear and specific reason. This could deter the use of certain variables if they prove too complex to explain easily or where they cannot be effectively deconstructed into a “primary” reason for the credit decision.
The Upshot: The CFPB’s circular is a call to action that raises a key question: will we see a retreat from certain data types or will the industry pioneer new ways to communicate complex decisions? Only time will tell, but one thing is clear: the status quo is no longer an option.
