Some fintechs think including more data and analyzing it with more advanced algorithms could solve the problem. Others say it’s time to build whole new systems.
9/12/2022
Credit scoring doesn’t work well for a lot of people. Those with low incomes, people of color and immigrants have been historically sidelined by the current system. Some of the nation’s most successful struggle with low credit scores despite their wealth. And if your fortune comes from crypto, you might as well be invisible to the current system.
Many startups see a business opportunity here. Finding a way to accurately underwrite those excluded from the existing system could mean unlocking a whole new customer base for lending and investment products. It also entails a lot of risk, as inaccurately assessing borrowers’ ability to repay can result in costly defaults.
Credit scoring is broken, but fixing it isn’t easy. FICO “is the dumbest system except for all the other ones,” Bullpen Capital general partner Eric Wiesen told Protocol. “Someone should clearly do this, but I am yet to find anyone successful at it.”
More data might be the answer. Rent or utility payments are not typically integrated into credit scores, for example. Some fintechs argue that doing so could improve many consumers’ scores.
Some fintechs are collecting that kind of payment data and making it interpretable for the main three credit bureaus, Equifax, Experian and TransUnion — and hopefully thereafter integrated into a credit score. Fannie Mae, meanwhile, began considering rental payment history last year, while California, Colorado and Washington, D.C. have passed bills encouraging landlords to report positive rental payment history, and Delaware recently launched a rental payment reporting pilot program with stimulus funds. “Buy now, pay later” companies are also working to get their on-time payment data integrated into credit reporting.
Zest AI, meanwhile, primarily uses credit bureau and LexisNexis data to create its own score in place of FICO — a score that the company argues has less bias and is more accurate because its algorithm can include many more factors. “FICO, for example, uses credit utilization, and they might use credit utilization at a point in time,” chief legal officer Teddy Flo told Protocol. Zest’s model examines credit utilization over time instead.
But consumer advocate Rachel Gittleman, financial services outreach manager at the Consumer Federation of America, argues that credit scoring is broken not because it doesn’t factor in enough information, but rather because it factors in too much. Much of the information that is already tracked in a credit report is not predictive, she argues, like medical debt, which is a poor predictor of a consumer’s ability to repay because people do not choose to get sick and take on that debt in the first place. Adding more data doesn’t solve the problem of poor data that is already being considered, and it’s crucial that consumers are the ones to decide whether these additional data points are considered in their credit scores.
Further, the credit bureaus themselves have a reputation for misusing consumer data and making difficult-to-correct errors, and are the source of most credit scoring problems for consumers, she says. An alternative to FICO, she told Protocol, “doesn’t take power away from that existing system.”
That’s why some fintechs are instead attempting to rewrite the whole system. A fintech called Trust Science also uses AI to evaluate exponentially more data points than FICO does, but also collects and stores its own data, like a credit bureau. Many others tailor their underwriting process for the specific lending products they administer. Line, a company that issues small lines of credit for emergency expenses without a credit check, also uses its own proprietary process. “We feel people are better than a number,” Line CEO Akshay Krishnaiah told Protocol.
Most of these processes depend on reviewing transaction records from a user's bank account or a business's accounting software. FinRegLab research suggests that such cash flow underwriting can increase accuracy, particularly when paired with a review of traditional credit reports and especially for consumers without a credit history.
The space is loosely regulated, and the quality of underwriting varies. In a joint statement in 2019, the Federal Reserve, CFPB, FDIC, NCUA, and Office of the Comptroller of the Currency said cash flow underwriting is subject to fair lending laws, the Fair Credit Reporting Act and other prohibitions against unfair or deceptive reporting, while the CFPB will likely write a data-sharing rule for financial services within the next year.
RealOpen — a crypto home brokerage founded by Christine Quinn of “Selling Sunset” and her husband, Christian Dumontet — is using a similar concept to prove the stability of the crypto holdings their clients use to buy homes. Their newly debuted scoring product, RealScore, requires clients to link their crypto wallets so the company can review the diversity of tokens in a home offer and assess whether its stability is “excellent,” “fair” or “risky.”
Though it doesn’t help clients procure loans with crypto — the brokerage only presents cash offers — it demonstrates the need for alternative scoring mechanisms. RealScore, Quinn said, started as a manual calculation. Systematizing that process gives it more legitimacy and makes it easier to interpret for real estate agents and brokerages they interface with that aren’t as comfortable with cryptocurrency. RealScore, opens the business to “broader market adoption,” Dumontet said.
If fiat cash flow underwriting is loosely regulated, crypto is the Wild West. Clear regulatory guardrails and industry standards are necessary to affirm any type of credit scoring is fair, transparent and sustainable, consumer advocates warn. “When we’re talking about companies that are not regulated or are not receiving oversight from the government, using new forms of underwriting — well, it has red flags all over the place,” Gittleman told Protocol.
Protocol link: https://www.protocol.com/fintech/credit-scoring-alternatives
Comments