To build on an earlier article unpacking a new Mercator Advisory Group white paper, Credit Scoring, Fintech, and Consumer Loans: Why AI Scoring Models Do Not Replace the FICO Score, PaymentsJournal sat with Brian Riley, Director of the Credit Services Advisory Practice at Mercator Advisory Group, to hear more about the efficacy of FICO’s scoring system and why lenders should be judicious in their use of alternative lending to open the sales funnel.
The Chicken and Egg Dilemma, but for Credit
The objective of the FICO Score is the same it has always been: to assess the creditworthiness of consumers. Those consumers fall into three broad categories: those who are scorable, those who are unscorable, and those who are “credit invisible.” The vast majority of adult U.S. consumers (more than 80%) are able to be scored using traditional credit bureau data; roughly the other fifth of the population are split between being unscorable — consumers with insufficient or out-of-date credit histories — and credit invisible — consumers with no credit bureau records at all.
It is to the benefit of both individual consumers and lenders for consumers to be considered scorable; consumers because they can access lines of credit and participate more fully in the lending economy, and lenders because they want to bring in ever more consumers. “Lending is a risk-based business,” said Riley. Lenders want to open the sales funnel to as many people as possible while still ensuring that those people will be able to repay their loans.
However, in trying to reach those 20% of people who are unscorable or credit invisible, there is a “chicken and egg” problem: consumers cannot get a credit score if they do not have a credit history, but consumers also have difficulty applying for lines of credit if they do not already have a credit score. The question then becomes, how does one open the sales funnel to consumers who seem credit inaccessible? “Sometimes you need to innovate,” said Riley, “and you have to do it under controlled circumstances.”
Compliant Alternative Data Can be Useful
The controlled experiment Riley referred to is the use of alternative data to bring consumers into the lending sales funnel. Alternative data might include whether a consumer pays their telco bills or their rent on time, or other pieces of data that are not typically credit tradelines and operate outside existing credit reporting, but may correlate with creditworthiness. Riley noted that using alternative data can make sense when evaluating a credit applicant who is on the cusp of being scorable and for whom additional details to round out their profile could prove beneficial.
The problem occurs when fintechs begin to rely heavily on unregulated data — which can vary wildly between lenders — as a foundation upon which to assess creditworthiness. Alternative lenders might also consider a consumer’s social media posts, their SAT/ACT scores, what college they attended and what degree they earned, their online behavior, and their employment history. These data sources can easily be subject to bias, not to mention inaccuracy and privacy breaches. “We’ve seen fintechs that claim they have 1,500 or 2,000 different variables to bring customers in,” Riley added. Integrating that many different data points is not helpful or necessary, and it is simply not possible to ensure the same integrity and rigor that comes with the time-tested FICO Score.
“One of the important things about the FICO Score is that it requires the use of data that are already approved and specified by various federal regulations,” Riley explained. Those data regulations are connected to the five components of a FICO score — payment history, amount owed, length of credit history, credit mix, and recency of new credit applications. These variables produce a common number that is easily understood across the financial spectrum.
“If you are looking at the risk associated with a 720 FICO Score from a consumer who uses credit cards, that 720 equates to what their risk would be if they were doing auto loans, or personal loans, or any other vehicle,” Riley pointed out. “That is one real positive here — being able to have that consistent risk measure throughout.”
FICO Bedrock: Responsible Lending Through Risk Mitigation
As lenders attempt to bring new credit applicants into the fold, they are making choices about how much risk to assume — not just regarding how much risk there is that a consumer will not pay back their loans, but also regarding how much risk to take on in choosing metrics to determine consumer risk profiles. That is to say: using alternative scoring to determine credit risk is itself risky. Some fintechs, such as Upstart, combine alternative scoring with artificial intelligence to try to improve and expand underwriting, but their AI models may not accurately reflect credit risk during poor economic conditions.
“The economy is getting rocky now,” Riley cautioned. “This is not the time to throw out the scoring system that has reliably served the industry for quite a while.”
Indeed, the FICO Score has been in use since 1989, and FICO has been around since 1956. The FICO Score builds an analytic model based on every piece of regulated data furnished from the top three credit bureaus and produces an independent and quality risk score that works no matter how the economy is doing. This is particularly well-suited to asset-backed securitizations (ABS), which are regulated by the Securities and Exchange Commission (SEC) under standards for statistical rating organizations from the Credit Agency Reform Act of 2006.
These guidelines and oversight procedures are critical for ensuring responsible lending. “We’re not saying [alternative scoring] is poorly advised by any stretch of the imagination,” Riley clarified. “But it is not necessarily the foundation upon which you want to build your business.” Instead, alternative scoring is best used as a tool to augment the ever-reliable FICO Score. That is how lenders can both bring previously credit unscored or invisible consumers into the market while still mitigating risk.
Alternative data are just that — alternative. The primary method of credit scoring has been, and will continue to be, the FICO Score.