Last week, Federal Regulators Issued a Joint Statement on the Use of Alternative Data in Credit Underwriting.
Within their statement, they affirmed the consumer benefits of alternative data in its use in lending:
"… the use of alternative data may improve the speed and accuracy of credit decisions…using alternative data may enable consumers to obtain additional products and/or more favorable pricing/terms based on enhanced assessments of repayment capacity. These innovations reflect the continuing evolution of automated underwriting and credit score modeling, offering the potential to lower the cost of credit and increase access to credit…"
However, the marketing buzz surrounding ‘alternative data’ has resulted in multiple, uneven definitions.
For example, the statement defined alternative data by what it is not: “alternative data means information not typically found in a consumer’s credit files” collected by bureaus though used in underwriting.
Further, the statement focused on one specific type of alternative data — the use of cash flow data in underwriting. Research, such as a recent report by FinRegLab, has demonstrated that the addition of cash flow data to underwriting expands access to affordable credit to more applicants.
However, cash flow data (albeit important) isn’t the only type of ‘alternative’ data out there.
So, what are other commonly used ‘alternative’ data types?
And, what does the notable silence on these other types of alternative data potentially mean for lenders?
Enlarging the universe of alternative data for underwriting
Based on what I’ve seen building credit models, here’s a short (non-exhaustive) list of alternative data sources increasingly used beyond the traditional credit report for underwriting:
CRM (customer relationship management) data: lenders sit on troves of data on their own customers that can be used in a credit model. These data points range from the purchase history on credit cards to data on how customers interact with lender, such as their medium, timing, and responsiveness to communications. Discover, for example, uses information on whether a customer has shopped at a discount store as a variable. To date, this is probably the least explored source of alternative data and adding credit signal to models.
Public records data like bankruptcy and liens: the original ‘alternative data,’ companies like LexisNexis collect public records data such as bankruptcy and lien information that is commonly used within underwriting.
GIS, geo-spatial, or mapping data: insurance companies are increasingly evaluating the use of geo-location and mapping data to underwrite property insurance policies like homeowners and rental.
Mobile phone or web data: in the up-and-coming digital lending space, mobile phone or website tracking information is emerging in its use in underwriting. Lenders like Tala solely rely on credit signals derived from mobile phone data. On the e-commerce side, lenders like JD.com and Affirm use cart and order data in their underwriting.
Alternative reporting agencies: Clarity (now owned by Experian) is the most well known of these alternative agencies that collect repayment data on alternative loan products like payday and installment loans. Even in the case of underwriting prime US populations, the inclusion of this data has been found to increase accuracy and its usage is becoming more widespread among mainstream lenders and not just payday providers.
As I said, this is not meant to be an exhaustive list. In auto lending, dealer and car-specific information is increasingly included in underwriting. Banks can now use customer data from one loan product to underwrite the customer for another. And, many more!
What people aren’t using
Social media data: there’s a myth that using ‘alternative data’ means using very personal data like social media data for underwriting. Though some fintech start-ups claim to approach underwriting in this way, they operate mainly internationally and have remained niche… likely for the reason that social media data has yet to be proven to be predictive of credit risk.
What this all means for regulators
Though the joint statement affirmed the use of alternative data overall, especially in tandem with a strong compliance management system, there was a noticeable silence on other types of alternative data sources.
Lenders are already moving ahead with the usage of other types of alternative data where they find credit accuracy and/or the coverage of more people increases. This is sure to continue, especially with this tacit endorsement of regulators of its usage.
However, the lack of specifics on the types of alternative data represents a missed opportunity. Through the power of their megaphone, regulators can encourage the responsible usage of data sources with proven potential while while discouraging those that are still nascent, untested, or risky. In light of this announcement, how should lenders think about fair lending and other risks in this new era of big data in underwriting?
Ultimately, it comes back to lending basics — lending wisely and having a strong risk management system. Experiment but be intentional. Have systems to identify and resolve risks. We’re in a new era with new types of data and more sophisticated algorithms. There’s no alternative to that.
#fintech #underwriting #alternativedata #regulation #machinelearning