By Samara Trilling
August 10, 2020 at 5:00 am ET
In April, federal mortgage regulators announced relief for borrowers who cannot make their mortgage payments because of economic fallout from the coronavirus.
At a time when millions of Americans have lost their jobs and federal agencies are scrambling to respond quickly, this was an excellent step. One of the most productive things we can do for the economy and for public health is to help keep Americans in the homes they already have. And up until now, it’s been mostly working: Despite confusing forbearance rules that varied from bank to bank, many borrowers have successfully been able to defer mortgage payments and stay in their homes. As COVID-19 cases surge in new places, it’ll be crucial to extend this forbearance support.
But there is another reason why this mortgage relief is so crucial: Banks might use data on who defaults during this pandemic to decide who gets mortgages in the future.
It’s bad enough that anyone who defaults on a loan during this time will take a credit hit that makes it harder for them to get a loan in the future. What’s scarier is that this could also hurt new borrowers who’ve never taken out a loan before. A flood of defaults today could poison lending algorithms against new borrowers in the future — especially gig economy workers, restaurant employees and anyone whose credit looks like someone who defaulted on a loan during this pandemic.
The way our lending system works, people who default on their mortgages today become data points in the algorithms that decide the mortgages of tomorrow. According to a 2019 University of California Berkeley study, 45 percent of the nation’s largest mortgage lenders now allow you to apply for a mortgage online, with an algorithm making all or part of the decision about whether you get a loan and at what interest rate — and the number of online lenders is growing.
These lending algorithms work by comparing each new borrower to other borrowers with similar characteristics. They’re the mortgage equivalent of the ads you may have seen online: “People like you also bought this.” Here, it’s “people like you repaid their loan” — or didn’t. It’s guilt by association: If you look like someone who defaulted in the past, you could get denied a loan, even if you have personally never missed a payment.
In order to prevent virus-related defaults from impacting this equation, state and federal mortgage regulators should ban lending algorithms from using default data from the coronavirus epidemic in future mortgage decisions. This is the most direct way to prevent borrowers’ temporary financial problems stemming from this crisis from affecting future loan decisions.
All of the gig economy workers, independent contractors, restaurant workers, educators and others who lose their jobs due to coronavirus and then default on a loan will otherwise become data points to feed these algorithms. And this means that in the future, anyone who has a similar credit profile to these people may find it harder to get a mortgage.
Ordinarily, a person’s credit history can help banks predict whether they will default on a future loan. But at a time when many people with good credit histories are losing their jobs and finding it hard to make mortgage payments, that calculation is no longer useful. If we use coronavirus-related default data to teach algorithms that people with good credit histories are also prone to default, we’ll end up with paranoid algorithms that will predict doom for future homebuyers who really could repay a mortgage.
Existing relief measures — and passing legislation like the HEROES Act, which suspends negative credit reporting during the pandemic — will help current borrowers weather this storm. But they won’t protect future borrowers who might be unfairly denied a loan or charged a higher interest rate because of their similarity to people who defaulted on loans during this crisis.
Federal regulators can act quickly to instruct banks not to use current default data in future loan decisions. In doing so, they will be protecting current borrowers from losing their homes — and protecting the future mortgage borrowers we’ll need to rebuild our economy.
Samara Trilling is a software engineer researching machine learning fairness policy and builds next-generation city master planning software at Sidewalk Labs.
Morning Consult welcomes op-ed submissions on policy, politics and business strategy in our coverage areas. Updated submission guidelines can be found here.