Despite anti-discrimination and fair credit laws, a sizable swath of the U.S. population remains financially underserved, with limited access, if any, to mortgages and other credit products that are associated with full participation in the economic mainstream.
Economists, consumer advocates, policymakers and the banking industry are largely in agreement about the goals and opportunities of greater financial inclusion. Traditional credit scoring and risk management served as an impediment – highly efficient only for applicants falling within standard credit-history parameters and models.
Technologies such as artificial intelligence and innovations in data analytics are enabling lenders to broaden the pool of creditworthy consumers.
Out of the Shadows
Doug Minor, founder and president of Easy Credit Relief and co-author of Anatomy of Credit Scores, gives the example of “a woman who came to me and wanted to buy a house. She was 53 years old but had absolutely no credit,” owing to her upbringing in a family that paid cash for everything and did not go into debt.
“We had to get her to generate credit before she could actually get a loan to buy the house,” Minor says.” She got the loan, but only after the considerable effort that went into lifting her from the segment of the population that Minor and others term “credit invisibles.”
The goal of new alternative credit scoring tools and methodologies is to capture a broader range of data, which could spot trends that suggest good risk but might otherwise go unnoticed.
“The clearer the data is and the more meaningful it is, the more it can help reduce risk,” Minor says. “Obviously, reducing risks is going to help you improve the profitability of your company.”
System Is ‘Gamified’
Co-founder and CEO of OneBlinc, a fintech focusing on the underserved, Fabio Torelli explains via email: “Our primary product for this context is personal loans, and we also offer fee-free checking accounts, and a growing portfolio of credit-building tools.
“In order to fairly serve these individuals, we use alternative scoring mechanisms to determine true creditworthiness and ignore credit scores (which, for this population would normally disqualify them). We also employ risk reduction mechanisms in payment methods, such as payroll deduction, in order to be able to offer credit to a wider audience.”
Torelli believes traditional credit scores have become “gamified,” and increasingly detached from an individual’s actual ability to repay a loan. That created an opening for companies that use alternative scoring methods incorporating AI and advanced analytics.
“Tapping into thousands of alternative data sources creates a wealth of data that we are only starting to truly leverage,” says Torelli, whose background includes stints with Citi, Mastercard and Santander. “Statistical and credit models are being redone from the ground up. Credit players from emerging markets – much like myself, who have cut their teeth lending in economies where there were no credit bureaus and thus had to use other data points to dole out consumer loans – are pouring into the U.S. now.”
Broader Transaction Data
“The traditional models, which are primarily based on data in consumers’ credit files at the credit bureaus, have limitations, given that certain consumer segments – such as minorities and lower-income individuals – tend to have a lighter footprint at the credit bureaus,” says Emre Sahingur, who until recently led modeling and analytics services at VantageScore Solutions. (He is now chief credit officer of J.G. Wentworth but is not speaking in that capacity.)
As a result, “millions of consumers are unscorable by models like FICO,” he adds.
New data elements are key to improved insights, Sahingur explains. An example is consumer-permissioned data on bank-account transactions and cash flows. How often a depositor runs negative balances, as well as the tenure of the banking relationship, can be important risk signals.
Other alternative data inputs related to payday loans, rent and utility payments also improve scoring capabilities, Sahingur says.
Better Credit Bureau Data
A measure of the potential market that advanced technologies can open up is that VantageScore models can score nearly 37 million consumers beyond what conventional models cover, Sahingur says. VantageScore has been using rent and utility data when available, and taking advantage of credit bureau data enhancements.
A more recent focus has been on bringing new, alternative data sets into algorithms to not only improve models’ predictive accuracy, but also to evaluate consumers with limited information in their credit files.
“Minorities are over-represented among these newly scorable consumers,” Sahingur points out. “Other developers have been working on approaches that generate scores in absence of any credit bureau information, primarily based on bank transactions.” Examples include Nova Credit and Petal offshoot Prism Data.
Sahingur also cites credit bureau products like Experian Lift and TransUnion Credit Link that are “increasingly providing access to alternative data assets that can be used by lenders in developing models and decisioning engines.”
Payrolls and Employment
Open banking and payroll data, “when properly analyzed, can provide data points that, until now, were incredibly hard to infer, and confirm – not only in terms of manual operations, but most importantly from a data reliability perspective,” says OneBlinc’s Torelli. “Reliable employment information can be the Holy Grail for many alternative credit models.”
Torelli agrees that the established players have come to realize that there is a growing market for additional data beyond credit scores to supply a full picture of the consumer.
“They have responded accordingly, with supplemental data products for startups and fintechs, such as fraud information, rental payment records, etc. However, their cash cow is still (and will continue to be in the near future) their traditional credit score products, which they will defend and protect aggressively.”