06 October 2022 |

Credit Scores Perpetuate Inequities. Can Fintech Do Better?


Credit scores are supposed to be gender and race-neutral. 

What a cute idea. Now, let’s come back to reality and remember that’s impossible.

President Gerald Ford signed the Equal Credit Opportunity Act (ECOA) in 1974 to ensure that banks use credit history, not any part of a person’s identity, in deciding whether to approve a loan. 

It also granted women the right to obtain credit cards separate from their husbands for the first time. 

So credit reporting agencies began calculating number scores in 1989, using a formula created in 1959 by two white guys, also known as a FICO score. 

The idea was that reducing a human’s life experiences with money to a single number is the fairest way to establish if they’re credit-worthy. 

Sounds fair and unbias!

Not quite. 

People and communities of color have been disproportionately targeted for high-cost, predatory loans and other risky products that lead to higher delinquency and default rates than non-predatory loans. 

Consequently, POC are more likely than their white counterparts to have damaged credit

Its systemic inequity is so firmly entrenched into the system that FICO even tried defending itself by writing a blog saying the credit-scoring model is color blind. ðŸš©ðŸš©ðŸš©ðŸš©ðŸš©

Wells Fargo rejected half of its Black applicants in mortgage applications. According to Zest AI, half of all white people have a FICO score over 700, but only 1 in 5 Black people do

In 2019, African Americans were denied mortgages at a rate of 16%, and Hispanics were denied at 11.6%, compared with just 7% for white Americans, according to data from the Consumer Finance Protection Bureau. 

An Iowa State University study published the same year found that LGBTQ couples were 73% more likely to be denied a mortgage than heterosexual couples with similar financial profiles.

Using data gathered by a FICO score is just a mirror of inequities from the past. By using this data, we’re amplifying those inequities today. 

Examples of exclusion: 

  • Redlining the practice of keeping applicants of color, particularly Black people, from getting mortgages and building wealth through home equity (Officially outlawed by the 1968 Fair Housing Act, it continued informally, more recently in subprime lending that ultimately led to the meltdown of the global economy). 
  • Employment discrimination has, even today, kept about half of Black Americans living paycheck-to-paycheck.
  • Auto loans impact BIPOC most because many dealers charge more and place higher interest rates. 

This is the moment when fintech companies can swoop in and save the day like a true hero. 

That’s almost true.

Fintech algorithms have not entirely removed discrimination, but two silver linings have emerged.

  1. Fintech companies are helping to decrease discrimination thanks to the rise in competition with regular lenders. 
  2. Fintech algorithms discriminated 40% less on average than face-to-face lenders in loan pricing and did not discriminate at all in accepting and rejecting loans, according to a study by Berkeley’s Haas School of Business.

But credit underwriting software is far from perfect. For example, the study found that fintech lenders still charged Black and Hispanic borrowers higher interest rates than whites.

So yes, algorithms are less biased than loan officers

But there should be zero biases

Another issue is that fintech users are still affluent customers

This is a significant area of improvement. 

Better.com’s average client earns over $160,000 a year and has a FICO score of 773. However, as of 2017, the median household income among Black Americans was just over $38,000, and only 20.6% of Black households had a credit score above 700, according to the Urban Institute.

Unless we change these figures, we can’t boast about improving access to credit for more people. 

Fintechs Stepping In

What if credit scoring agencies factored in thousands of other data points like payments for phone bills, utilities, subscriptions, or rent?

Companies like SaverLife and Padsplit have entered the game to help on the housing front: 

  • SaverLife: Offers new ways to promote saving and help Americans save money. Specifically, it supports low and moderate-income families to build a habit of saving money and ultimately building wealth. Serve 600,000 people currently. 
  • Padsplit: Helps people find homes, rooms, and apartments to rent. Offered co-living homes in its platform marketplace in partnership with hosts, primarily serving members with a $25k income. 

Zest AI partners with the world’s largest financial institutions (like Freddie Mac and Discover Bank) to tiny credit unions in rural America to help use machine learning to underwrite consumers for credit. 

Think of it like this: You would get a fuzzy image if someone could only use 25 data points to describe you. But if someone could tell you in 1,000 data points, you would get a much better picture of credit risk

AI in digital lending has caught the attention of regulators. But, of course, just because something is automated doesn’t mean it cannot have biases

But with the proper safeguards and being super intentional with the tech, underwriting via AI is an improvement from traditional ways

For example, Zest AI participated in a study with Financial Health Network where they examined one of its customers’ use of the technology, Suncoast Credit Union.

As a result, they saw a 28% increase in approvals for women and people of color with no added risk, which kept those folks from turning to payday lenders and other much more high-cost sources of credit.

The big hurdle to overcome is the power of alternative data sources because all of the data points from the alternative data sources are not available to every consumer. They’re still emerging

For Zest AI, complete documentation makes this less daunting. So unlike other scores where you have no idea what features are used to underwrite your loans, every assessed model comes with a 300-page document that lists every single model and exactly how it works. 

Sure, fintech has its work cut out. But some wins show we’re heading in the right direction

For example, a student out of NYU did a recent study on businesses that got PPP loans through fintechs and went into brick-and-mortar stores of traditional financial institutions like Bank of America and Wells Fargo. 

She found that people with PPP loans granted via fintechs were distributed much more evenly according to gender and race than the access to loans through in-person meetings and financial institutions. 

Fintechs played an essential role in extending PPP loans to Black- and Hispanic-owned businesses, she found. 

The digitization of finances leveling the playing field, and expanding access by reducing the impact of racial biases on credit decisions is such a significant development that often goes overlooked. 

So what are you doing in your fintech company to ensure it’s never overlooked?