In September the Bernie Sanders campaign rolled out a plan that would revamp how credit scores, credit bureaus, and, in turn, lending would work in America. This plan would establish a “secure public credit registry” within the Consumer Financial Protection Bureau that would eventually replace the existing system. Currently the three major credit bureaus and many smaller agencies collect data on Americans (with or without their permission). When people apply for loans, apartments, and jobs, the bureaus turn that data and scores computed from it over to banks, landlords, and employers. Unlike the current opaque system, Sanders’ public credit registry would “use a public, transparent algorithm to determine creditworthiness that eliminates racial biases in credit scores.” Sanders’ proposal would also remove medical debt from consumers’ credit reports and would end the use of credit report checks for housing and job applications.
Each of those suggestions, if implemented, would probably result in a fairer, better lending system. But the plan offers a false promise: that we can “eliminate racial biases in credit scores.” It’s a worthy goal, given that across the 60 biggest cities in the U.S., predominantly nonwhite neighborhoods have credit scores that are, on average, 76 points lower than those cities’ predominantly white neighborhoods. But just changing the formulas used to calculate credit scores will not close those gaps much, if at all. As long as our economy is racially biased, our credit scores will be too.
To see why that’s true, we have to understand the many ways credit scores operate. As Josh Lauer, author of Creditworthy: A History of Consumer Surveillance and Financial Identity in America, said when I spoke to him by phone, “We have this idea that credit scores are one single number.” But, Lauer said, “there are hundreds of scores.” Some are specialized scores built by credit bureaus for specific tasks, like predicting whether you’ll default on a particular type of loan. Others are built by the banks themselves, using a mixture of data from credit bureaus and the bank’s own data collection efforts to predict not just whether you’ll default on a loan but whether you’re the type of customer who will frequently use their products, keep your account open for many years, or carry a high balance on a credit card. Some credit scores are relatively well understood by the public—FICO, the familiar credit score that lenders can purchase from Equifax, Transunion, or Experian, is calculated from a generally known list of variables (by way of a confidential formula). By contrast, the credit scores built by the banks themselves for proprietary use are black boxes.
These specialized scores are entirely mysterious to the public and may often be unfair. (It’s not completely clear whether Sanders’ plan would prohibit banks from building their own credit scores.) As Cathy O’Neil reported in Weapons of Math Destruction, American Express has created internal risk scores based on where customers shop, and “it was up to the unhappy Amex customers to guess which establishment had poisoned their credit.” There’s every reason to believe that what Amex is doing is typical of the industry, and it is certainly emblematic of the dangers of this system. While these internal credit scores are monitored for racial bias by the Office of the Comptroller of the Currency and the CFPB, an OCC spokesperson told me by email that the office takes a “risk-based approach to fair lending supervision.” Regulators are forced to prioritize their attention; even with more staff, it’d be impossible to deeply scrutinize every credit score made by every bank for racial bias.
An infinite amount of bank supervision still wouldn’t be enough to close the gap in credit scores between black and white Americans because, frequently, bias isn’t created by the credit score calculation—it’s described there. Imagine, for a second, the simplest, most transparent, arguably fairest credit score possible. It might be a statistical model that predicts your likelihood of defaulting on a loan, based only on (accurate) information about whether you’ve paid your past bills on time. That simple, “fair” approach would still produce scores that were systematically better for white people. Even if the credit score is fair, the life experiences it crunches into a number are not. People of color in the U.S. face rampant employment discrimination and wildly disparate treatment by the criminal justice system, which at times results in unfair prison time (or worse). It’s hard to pay your bills on time if you’re fired because your boss doesn’t like your natural hair, or if you’re thrown in jail for a crime you didn’t commit. And of course, these obstacles to timely bill payment are compounded by the massive wealth gap between black and white Americans. If you “haven’t accumulated assets you can fall back on during hard times,” Lauer pointed out, a job loss or an episode of bad health is more likely to cause a drop in your credit score—and a low credit score can cost more than your ability to borrow money.
The idea that credit scores are enumerations of our moral worth or “level of responsibility” is embedded in our cultural fabric. This is why employers see fit to use credit scores in hiring decisions, a practice that I, like Sanders, believe is unfair and should be outlawed. But even in their narrowest use—forecasting the likelihood someone will default on a loan—credit scores are still a reflection of some things that an individual can control and many other things that they can’t. I challenge you to name a single attribute that would influence or predict whether someone can repay a loan that is not in some way related to the circumstances of that person’s birth. Put simply, if we don’t change the underlying discriminatory forces that make Americans differentially able to weather financial shocks, black Americans will continue to have lower scores than white Americans.
It’s not even clear what it would mean to have credit scores free of racial bias. Would it mean historically advantaged and disadvantaged groups have the same average credit scores? Would it mean the models themselves just don’t contain any information related to a consumer’s race (something already prohibited by the 1974 Equal Credit Opportunity Act)? Or something else altogether? For that matter, should we have credit scores at all, or should every consumer be treated identically by banks? (I would only recommend that final option if we wanted to freeze consumer lending entirely, since it would be hard to offer anyone credit at affordable rates if people could repeatedly default on loans without consequence. While some might advocate for such a freeze, it’s considerably more radical than anything Sanders has proposed). The Sanders team points to research by the Urban Institute that offers suggestions for policies that would help reduce the racial gap in credit scores—for example, changing how local governments levy fines and fees—but that report stops short of implying that the credit scoring algorithms themselves are structurally flawed.
Sanders recognizes that getting rid of private credit reporting agencies isn’t a panacea. In a speech in March, he said, “Do we still have a long way to go to end the institutional racism which permeates almost every aspect of our society? Absolutely.” A member of Sanders’ policy team told me by phone that Sanders’ credit reporting proposal is intended to work in concert with other policies to promote racial equity, like his suggested jobs guarantee and his criminal justice reform plan.
Maybe I’m being a curmudgeon to point out that the proposal, as written, could probably not reach its goal of eliminating racial bias in credit scores. It is good, perhaps even necessary, to aspire toward ambitiously idealistic futures; Sanders often reaches high, and perhaps does not always expect his grasp to exactly match that reach. But it’s important to reckon with what tweaking the algorithms behind credit scores can and cannot accomplish. Algorithms are pretty easy to change. But the society they model still comes up short.
Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.