Helping people climb the credit ladder: Looking into financial inclusion with FICO, Credit Karma and TomoCredit
- Not all Americans have equal access to credit. But financial institutions and fintechs alike are trying to build a fairer system that is more inclusive of all groups.
- Their approaches differ -- from reworking their current credit decisioning models, using alternative data, and offering access to credit without a credit score.
Once upon a time, credit decisions were made not by intangible algorithms but humans using subjective methods which relied on social hierarchies, judgment of moral character, and hearsay. But by the end of the American Civil War, the foundational elements of modern credit reporting were in place. Namely, mass surveillance by the private sector, information sharing, and a quantitative rating-based system.
Fast forward to today, these tenets have become part and parcel of credit decisioning. However now they are implemented at radically different speeds with improvements in technology.
As swift and non-subjective as the system has become, it still has some ground to cover. Credit does not equally include everybody (yet) nor is it a perfect representation of their ability to pay back. AI used in credit decisioning can be less accurate for lower income families and may even discriminate on the basis of gender and race. For example, in 2021 research found that 54% of Black consumers reported having no credit or credit scores below 640.
But financial institutions and fintechs alike are trying to build a fairer system that is inclusive of all groups. Their approaches however differ, from reworking their current credit decisioning models, using alternative data, or offering access to credit without a credit score. Here’s how a veteran like FICO, seasoned player like Credit Karma, and upstart TomoCredit are making inclusion a priority in the credit system.
FICO : Improving credit scoring
FICO has scored about 90% of the credit eligible US population -- in many ways it's the cornerstone of the credit decisioning industry. The FICO score is a 3 digit number that summarizes the information on one’s credit file. Vice president and general manager, Scores at FICO, Sally Taylor adds that the score doesn't include factors like one’s ZIP code, gender, race, or marital status, but does include:
- a person’s payment history
- amounts owed
- length of credit history
- new credit accounts
- and types of credit used
But for those who go unscored, the firm has started using alternative data to build products, like the UltraFICO Score. This score uses information such as telecom and utilities payments, public records, and checking account data to paint a picture of a consumer’s finances.
Some credit scoring decisions are made by using models that only depend upon Machine Learning (an application of AI), which is an effective computational beast but not infallible. This means that every now and then, depending on how an analytics model is built, subjective biases can trickle into the system. In 2022, a law firm that monitored Upstart’s fair lending practices found that the company had lower approval rates for Black applicants. These biases can keep people out of the system, essentially they are software-based gatekeepers which make credit decisions less inclusive.
This is specifically a bigger problem for those credit scoring models that use Machine Learning only. Due to the nature of AI it is generally hard to discern why a ML model came to a conclusion it did: this is called the ‘black box’ issue. For example, why did the credit scoring models assign the scores that it did? What factors impacted the decision?
This makes reliance on Machine Learning-only models for credit decisioning a question of compliance, transparency, and inclusion. FICO’s answer to this issue is building models that use Machine Learning as well as their own scorecard technology. This varies from other approaches where a ML-only model is deployed and then checks are run on its results to see why a model arrived at a particular conclusion.
Being able to understand why ML-only models arrive at their conclusions, is new and perhaps the most important issue to tackle when black box AI is concerned. The decisions and machinations of these models can be opaque to even those who built it. This opaqueness can prevent firms from understanding why there is bias in the system and in turn, hinder inclusion.
Credit Karma : Inclusion through products
There are many paths to financial inclusion, and building better credit decisioning models is just one of them. Credit Karma uses another approach: it likes playing “matchmaker,” said Rich Franks, head of Lightbox at Credit Karma. The firm’s Lightbox technology allows the firm’s lending partners to use anonymized data on its members to design marketing strategies that make offers to the consumers likely to be approved for a product. On the other hand, consumers can get information about their likelihood for approval. The firm uses badges to communicate their chances of success.
“A three digit credit score can only tell a lender so much,” Franks added. Indicators like consumers’ savings, income, and timely rental payments can be strong measures of a consumer’s ability to pay back.
Credit Karma also offers a slew of parallel products that help consumers make informed decisions, as well as plan for scenarios that may impact their credit score. For example, the firm’s Credit Monitoring service helps consumers spot any errors in their TransUnion or Equifax credit reports. The Credit Score Simulator lets people explore how actions like taking out a loan or letting accounts go past due may impact their credit report.
TomoCredit : Easing access to credit
TomoCredit offers consumers with no or limited credit history access to a credit card. The firm calculates its own score, called the “Tomo Score” and based on banking and other financial metrics, to assign a credit limit. According to the firm’s cofounder and CEO Kristy Kim, the company looked into 30,000 data attributes gathered by user consented data to develop Tomo Score’s recipe.
Those without a Social Security Number can also apply for the credit card, opening up lines of credit to those who have just stepped onto American soil. Their applications are reviewed by the firm on a case by case basis. If an application is rejected, the consumer’s credit score is not negatively impacted but they cannot apply again. According to the company’s website, the firm hopes to build this feature in the future.
However, consumers do have to link a bank account for their Tomo Credit Card. This account has a minimum limit of $800 and Kim says that the company hopes to lower this limit in the future. “As a small startup with limited resources, we had no choice but to come up with a threshold so we can grow sustainably for a long term,” she added. Soon the firm expects to launch a “credit dashboard” that allows consumers to monitor their credit score and get tips to build better credit. Unlike traditional credit cards, customers cannot carry balances on the card from one month onto the next, instead the payment is automatically deducted from your card by the end of the week.
In 2021, the FDIC found that underbanked households were less likely (62.4%) to have credit cards but were more likely to have bank and nonbank personal loans than fully banked households (76.6%). Being underbanked or unbanked is costly, and it also opens consumers up to risky credit products like payday loans or auto title loans.
The implications of being underbanked or unbanked can be amplified by factors of race, geographic location, class, and gender. Labyrinthine as it may be, if firms in the credit space keep building towards inclusion, everyone might finally be able to climb up the credit ladder.
The three approaches above differ radically in improving inclusivity in the credit system in America. From improving foundational problems like how people are scored to building products that make access to credit easy for all, different approaches are needed to ensure that consumers at different stages of their credit journey are included. While It is unlikely that inclusion and equality in the financial system of America is only a product or AI model update away, it is important that organizations make it a priority.