Many fintechs use the terms "financial inclusion" and "financial health" to promote their products or to explain why they beat traditional credit, lending and banking options.
But to what extent are they living up to this promise?
"There is still a long way to go," Adrienne Harris, superintendent of the New York State Department of Financial Services, said during a fireside chat kicking off the
Still, "the potential is there," she said. "It has to be done safely and in a cost-effective way."
Many BNPL lenders can credit their meteoric rise to the unique circumstances of the pandemic. As their market shifts, these companies are forced to find new avenues of growth under the increased scrutiny of the Consumer Financial Protection Bureau.
At the summit, which took place in Washington, D.C., fintech leaders and researchers made their case that fintech has made a difference in enabling more equality in lending and helping community institutions, especially those that reach the underserved, compete with larger players. They pointed to machine learning and artificial intelligence as key tools that made this possible. They also highlighted open questions, such as how consumers feel about the role of machine learning in credit underwriting. The FTA is a group composed of fintechs like the small-business challenger bank Bluevine, the payments company PayPal and the international money transfer service provider Wise. The event was produced by Protocol, an online publication that covers technology (it announced Tuesday that it would shut down).
One topic of conversation was the importance of process automation in financial inclusion.
New York University associate professor finance Sabrina Howell co-wrote a paper about automation and racial disparities in credit access that was finalized in November and will be published in a future issue of the Journal of Finance.
"I was motivated by the observation that fintechs have brought two important innovations to lending" she said, namely algorithmic underwriting that uses machine learning and new data to gauge credit risk, and process automation, or efforts to remove humans from loan originations by automating application intake, income verification, anti-fraud efforts and more.
"These technologies raise important questions about their implications for racial disparities in lending," Howell said on a panel about shaping the future of small business.
Her research showed that more than half of Black-owned businesses that received a Paycheck Protection Program loan got it from a fintech lender, even though fintechs made a small portion of PPP loans overall. Furthermore, the rate of lending to these businesses among banks increased with bank size and with branch-level software spending. When examining a set of small banks that outsourced back-end PPP loan origination to fintechs, Howell and her coauthors found that lending to Black-owned businesses roughly doubled compared to small banks that continued to use manual processes.
There are three possible reasons why automation boosts lending to Black-owned firms, said Howell. One is that by lowering costs, it lets lenders make smaller loans, which are more likely to go to minority-owned firms. Another is that financial institutions with automated processes typically allow for online origination as well, which means areas underserved by bank branches, and with higher shares of minority borrowers, get access. A third is that removing humans from the decisions could reduce racial bias.
Connie Evans, CEO of the Association for Enterprise Opportunity, a membership organization for firms that provide capital and services for underserved entrepreneurs, found these results lined up with her own observations.
"Many of the CDFIs were able to do so much lending during the Paycheck Protection Program because they partnered with fintechs," she said.
Leaders from fintech companies spoke up about the difference they feel technology can make in financial inclusion.
Alex Marsh, global head of public policy at buy now/pay later firm Klarna, argued that Klarna's technology allows for more real-time decisions than what credit card issuers make at the outset when people apply for a credit card.
"We're assessing eligibility every time a customer uses Klarna," he said during a panel about fair and inclusive credit.
Marsh says that Klarna uses a combination of internal data — for example, if someone is overdue for a payment — and credit bureau data to approve customers for each purchase. The company is testing an open banking approach in Europe, where customers can consent to sharing additional information, like transaction data, to help Klarna get a clearer picture of their income and expenditures in marginal cases.
He argued that credit bureau data is often out of date, meaning lenders may make a decision based on circumstances that have changed. "We know how quickly, particularly with the pandemic and with cost of living challenges, someone's circumstances can change in eight weeks," he said. "Open banking can be a disruptor to make better decisions for consumers."
Teddy Flo, chief legal officer at
"A machine learning model is capable of using 300 data points," he said. The company counts Citibank, First National Bank of Omaha and Truist among its clients, as well as community banks and
"Small financial institutions don't have the staff to build a machine learning model," said Flo. "People want instant decisions." For example,
In a February 2022
Kelly Cochran, deputy director of FinRegLab, a nonprofit research organization focused on new technologies and data, said she was encouraged by FinRegLab's research into using cash flow data, largely from bank accounts, to underwrite consumers and small businesses.
"The data was independently predictive and could also be used in conjunction with traditional sources," she said. "It's potentially really useful not just to consumers with no traditional credit scores, but a broader range of consumers who are accessing credit at different prices."
But there are still open questions that fintech has yet to solve.
One is ensuring machine learning models are as unbiased as possible and respectful of fair-lending laws, said Cochran. Another is for lenders to figure out how to make decisions with a wide diversification of data and many options in the models they can deploy. A third is consumers' feelings about using machine learning in credit underwriting models.
"There are so many lenders in this space, many of which specialize in working with underserved communities, that may not have the same technological infrastructure that some other players do," she said. "Providing certainty so some of these lenders can find their place and decide what techniques are useful is an important question."
