The nightmare of last summer’s Equifax data breach is resurfacing. This week the Wall Street Journal reported that the data thieves stole was incredibly detailed and far more extensive than the company reported publicly. As an already chaotic tax season approaches, IRS experts fear that because thieves have almost 150 million records containing names, Social Security numbers, birth dates, and addresses, they will be able to redirect hundreds of billions of dollars in tax refunds to their own accounts.
Traumatizing as such breaches are, most of us have come to accept the risk of identity theft as an awful but worthwhile price to pay for the conveniences of modern life. We share our personal information with online retailers, credit card issuers, and email providers in exchange for indispensable services. Sometimes that data leaks out. But how does the calculus change when our most sensitive information is harvested, then lost, by companies, like Equifax that we never chose to do business with in the first place? Especially if those companies aren’t providing value for us, or are in fact causing trouble behind the scenes?
“Credit bureaus don’t exist for the benefit of the consumer,” says Zaydoon Munir, a former data executive at Experian from 2004-2010 and founder of RevolutionCredit, a behavioral credit data and scoring company. “It’s a black box. You don’t have control of your file, you can’t contribute data to it.”
Munir says credit files are built to manage risk for lenders, but, as backward-looking instruments, they can impose enormous penalties for slip ups that are extremely difficult to challenge and lock millions of deserving consumers out of basic economic opportunities. For example, “thin file” borrowers, or borrowers who haven’t established a lengthy credit history with one of the big for-profit bureaus, face either being shut out from credit or paying much higher rates.
“If you’ve missed a few payments, you could be the Warren Buffett of the future, the system doesn’t care, you’re being penalized,” Munir says. Munir fled Iraq in the 1980s and he only half-jokingly asserts that this guilty-till-proven-innocent approach is something Saddam would have devised.
How big is the problem? Thanks to the thin file classification, about a quarter of American adults have credit scores down in the 600s, or even no score at all. Unsurprisingly, people most in need of economic opportunity are more likely to fall into this group–a group that pays more for credit if they can get it at all. People of color are especially disadvantaged by backward-looking systems as they have been shown to be more likely to have access only to predatory first-credit experiences. High rates lead to higher likelihoods of default and then to a lifetime of bad credit, making it a vicious cycle that can be devastating for consumers.
But if backward-looking credit scoring is imperfect, does that really mean we can run a credit-based economy without it? Increasingly, entrepreneurs are combining big data and machine learning with insights from behavioral economics to show it can be done. A new generation of credit applications has launched, in the past five years, that are based on forward-looking data and predictive models rather than strictly on past performance. They allow consumers to proactively demonstrate credit potential. Perhaps most promisingly for those trying to stop data theft, they don’t require massive aggregation of personal information.
Highly predictive data models aren’t just good for consumers, of course. By better predicting behavior, they expand the customer base for lenders by identifying millions of underserved consumers who can be reasonably trusted.
“We can predict creditworthiness in about 20 seconds based on data that’s already sitting on a customer’s device,” says Shivani Siroya, CEO of Tala, which, thus far, has delivered more than 2 million loans in Kenya. “We can look at where they generally travel through geolocation, analyze their social networks, even ask questions we know the answer to, so we can see how honest they are.”
Siroya began Tala in 2012 by lending out about $80,000 of her own money. She says she did it pretty much blindly just to collect data so she could begin to establish patterns that might predict creditworthiness. Her repayment rate was a fairly dismal 70%. But as her algorithms churned through loan after loan, that rate has increased to 92%, which is significantly better than a traditional bank using a traditional credit bureau like Equifax can achieve. Because her company makes credit available to people who could never hope to walk into a bank and get a loan, Siroya says that Tala is now the third most popular mobile app in Kenya behind Facebook and WhatsApp. Over the last year, the startup has deployed more than $100 million in loans.
“We look at people by what they’re doing in their daily lives, not some payment they might have missed three years ago,” says Siroya.
Munir, the former Experian executive, agrees with this approach. He says a blemish on a credit file doesn’t always mean what it seems to. “Imagine someone goes through a divorce or has a crisis. They’re likely to miss a payment here or there. But we now know from data that their future behavior is likely to get more conservative, not less. The credit file isn’t smart enough to see that. It will assume your future will look like your past.”
Munir’s company, RevolutionCredit, which he founded after leaving Experian, is designed to solve the thin file problem and lower credit costs by raising the scores of the 30% of consumers with credit between 550-720.
“After I left Experian, I fell in love with the work of Nobel Prize-winning psychologist Daniel Kahneman,” Munir tells me. Kahneman, along with his collaborator Amos Tvsersky, revealed that human beings regularly behave in ways that defy normal conceptions of rationality, but their behaviors are still predictable if you know what to look for.
Munir says that there are 120 million U.S. consumers with scores between 550 and 750. Looking at macro behaviors in credit markets, we know that about 30% of them, or 36 million people, deserve better rates. The trick is to figure out which ones.
RevolutionCredit, like Tala, is using machine learning to identify the non-obvious signals of creditworthiness. So far, 500,000 users have played RevolutionCredit’s financial literacy quizzes and games. Just making it through the games, Munir says, provides a valuable sorting function, identifying those with real intent to improve their financial situations. Deeper analysis on how diligently users optimize mock budgets and navigate in-game scenarios creates an even more fine-tuned picture.
“In Uruguay we had a lender that used to decline 80% of applicants. Using RevolutionCredit, they now only deny 40% and their repayment rates have remained stable,” Munir says. RevolutionCredit now claims 4 of the top 10 American credit issuers as customers.
For now, tools like Tala and RevolutionCredit aren’t ready to overthrow Experian, Equifax, and TransUnion. In their early phases, they’re oriented toward allowing consumers to qualify for better terms and bringing more people into formal credit markets, which are still dependent on traditional backward-looking models. But the more users enter these predictive markets, the more credit risk can be fairly assigned through systems in which consumers control their own data.
Since the Equifax breach, lawmakers, several of whom lost their own data in the hack, have been trying to put restrictions on credit bureau behavior, forcing them to be more transparent and secure. The new rules would require bureaus to offer free services like credit file locking, a cumbersome and feeble tool that, at best, will make our data marginally more secure. The question I think they should be asking is not how do we plug the holes in a fundamentally flawed industry, but how do we look to entirely new models that keep top-level financial data in the hands of those it actually belongs to. Data thieves are embracing bold new technologies. Our financial system must too.
“Technology can create a view into a consumer without the bureaus,” Munir says. He envisions a day when a credit file looks a lot more like a LinkedIn profile, something verifiable and reality-based but controlled by the individual, not the platform. This would force fundamental change in the big credit bureaus–changes that might go beyond just better protecting our data to making markets more equitable, transparent, and affordable.
Jonah Sachs is a digital marketing pioneer who created some of the internet’s first viral social change campaigns. He writes about storytelling, creativity and the work of groundbreaking innovators. His latest book Unsafe Thinking: How to be Nimble and Bold When You Need it Most will be published in April 2018.