When even a millennial with a six-figure income and a 750 FICO score can't snag a loan with a decent interest rate, there's something wrong with the system.
That's the attitude held by the founders of peer-to-peer lender Upstart, including Paul Gu, who was one of those borrowers.
Gu a 22-year-old math whiz who created the company's statistical model attempted to get a loan from a bank as well as p-to-p rivals Lending Club and Prosper. Gu had a string of enviable numbers attached to his name, including that 750 credit rating, but less than three years of credit history. Prosper and the bank turned him down flat, while Lending Club offered him a loan with an interest rate above 20%, he says.
That test confirmed to Gu and his fellow co-founders, former Google executives Dave Girouard and Anna Mongayt, that responsible borrowers are being discounted by underwriting models that fail to look beyond credit files. Upstart is attempting to carve out a niche in the increasingly crowded alternative lending market by incorporating educational data and work experience into its underwriting decisions.
"That's when the lightbulb went off," Girouard, Upstart's chief executive and a former president of Google's enterprise business unit, says of Gu's borrowing experience. "If the lender could better understand the borrower, they could much better understand the borrower's creditworthiness."
Other alternative lenders have made significant inroads in harnessing big data to root out high-quality borrowers.
Online small-business lender Kabbage decides whether to lend to merchants based on their social media activity, shipping records and sales data; its fellow small business lender OnDeck also crunches social media data. And LexisNexis Risk Solutions offers companies an alternative assessment of credit risk by analyzing information about consumers' voter registration, education level and residence history.
Meanwhile, Lending Club, which opened in 2007, announced Tuesday that it facilitated $1 billion in consumer and business loans in the second quarter. The company is reportedly planning an initial public offering later this year.
"Anything that explores new ways to underwrite and develop a track record is moving in the right direction," says Mark Schwanhausser, a director at Javelin Strategy & Research. "There's got to be a way to help people get loans who deserve loans."
To get a more complete picture of borrowers, Upstart's website asks people applying for its three-year unsecured loans for information about their job histories, grades, alma mater, area of study and standardized test scores.
"Your GPA is an indicator of how focused and responsible you are, so it's a good indicator of your propensity to pay a loan back on time," Girouard says. Upstart's model favors academic achievers and applicants who major in areas like nursing and computer science, which tend to correspond with low rates of long-term unemployment.
"We also look at all the traditional variables, from income to debt to credit scores," Girouard says, noting that minimum FICO score for Upstart applicants is 640 and that the average score is 710. "It's really kind of a superset of what other lenders look at."
Upstart's approach to lending includes a willingness to shift course in order to meet the demands of the market. Up until two months ago, the company specialized in making ten-year loans to young entrepreneurs in the form of income-sharing agreements.
But when Upstart realized that far more applicants wanted to use the loans to pay for computer-coding boot camps or refinance credit card debt than to start their own business, the company decided to offer more traditional consumer and student loans in amounts ranging from $5,000 to $25,000.
Its loans carry annual percentage rates ranging from roughly 6.7% to 22%; borrowers can defer their first payments for a period of three or six months.
"When we launched the loan product at the end of April, the response was so strong that within a week we decided to sunset the other product," Girouard says. Upstart's income-sharing contracts were growing at a rate of about 10% month-to-month, he says; the new product has doubled month-to-month thus far. The company says it has originated more than $3 million in loans over the last two months and has recorded zero late payments to date on either its new or old loan products.
While Upstart is currently operating at a fairly small scale, analysts say that its wide-reaching, data-driven approach to evaluating borrowers could prove effective. But banks may have trouble translating the model to their own underwriting practices.
"Banks are still stuck with their traditional underwriting models," says G. Michael Flores, the chief executive of consulting film Bretton Woods and a specialist in researching alternative financial services. "Some of that is inertia but the other part is regulatory. To make an unsecured loan to someone with a thin credit file is just asking for regulators' criticism."
Fear of bias charges might also dissuade banks from using educational data, according to Flores. "You can build models based on historical data about how [graduates from] certain schools have performed and their earnings potential," Flores says. "But for a bank to do that, they could unintentionally exclude various segments of the marketplace."
A similar criticism could be levied at Upstart, whose marketing may come across as elitist. The borrowers featured on the company's website have "an Ivy League quality," Javelin's Schwanhausser says.
But banks can find other ways to include a broader range of borrowers in their loan portfolios, according to Schwanhausser.
"In some ways this is a partnership opportunity for banks to take advantage of new platforms as investors," he says, pointing out that banks that buy peer-to-peer loans can balance their risk portfolios by acquiring loans outside of their specific geographical areas. "That's a way banks can learn how to underwrite more creatively and take a proper amount of risk."