Algorithmic fintech lending is less discriminatory against minorities than traditional loan officers, according to a recent study of US mortgages. The findings signal hope that technology could provide financing that’s more fair, but the research also underscores how widespread discrimination remains.
The US housing market has long been prejudiced against minorities. When Latino and Africa-American borrowers are looking to buy a home, they usually end up paying 7.9 basis points (0.079 percentage points) more than whites to take out the mortgage, and 3.6 basis points more when they refinance the debt, according to a National Bureau of Economic Research working paper published this month.
That comes to $765 million in additional interest costs each year. The researchers also estimated that discrimination may have resulted in as many as 1.3 million mortgage applications being rejected between 2009 and 2015.
Algorithms tend to have a better record. Online financial technology companies discriminate, too, but 40% less than loan officers who make decisions face-to-face, the NBER researchers found. They also found no discrimination from the robots when it comes to loan approvals.
To reach their findings, the researchers studied decisions made by more than 2,000 of the largest mortgage lenders in the US (45% of which have complete online or app-based mortgage contracting) from 2012 to 2018. They examined mortgages that were securitized by Government Sponsored Enterprises (GSEs) Fannie Mae and Freddie Mac. Because the GSEs guarantee the loans against credit risk, differing interest rates for mortgages with the similar credit scores and loan-to-value ratios may indicate decisions by lenders that result from discrimination.
The researchers didn’t disclose which companies were included in the study, but pointed out that Quicken Loans, which started fully online lender Rocket Mortgage in 2015, has become the biggest mortgage lender in the US.
Should we turn all the lending over to the algos? Doing so would have challenges. Many people still want to have in-person discussions when it comes to major financial decisions. It seems likely that some consumers would be reluctant to take out financing without a human connection to provide reassurance. Even if technology is making the credit decisions, that face-to-face guidance can re-introduce bias, such as by encouraging or discouraging a potential applicant based on discriminatory factors.
Algorithmic lending, meanwhile, is proliferating online, transforming everything from personal loans to small-business borrowing. Algos can also be embedded with biases, and it may be difficult to unravel which variables are causing statistical discrimination. Jack Dorsey, co-founder of CEO of fintech lender Square, recently said that the “black box” nature of digitized credit decisions is a vulnerability.
The algorithms aren’t “being written in such a way that they can explain the criteria being used, or how they actually made the decision,” Dorsey, who also co-founded Twitter, said in an interview with Quartz. “If we fix those problems then I imagine that we will be able to increase the access, because it will increase the velocity and therefore the scale” of financing available, Dorsey believes.
Online financing isn’t always better. There are signs that some predatory, payday lending strategies have simply shifted to the internet.
Overall, there are distinct indications that the internet is making lending more fair. NBER researchers found that mortgage discrimination has declined in recent years, which could be because comparison websites have made it easier to shop around. “In short, algorithmic lending may reduce discrimination relative to face-to-face lenders,” the researchers wrote. “But algorithmic lending is not alone sufficient to eliminate discrimination in loan pricing.”