iPhotoMarketplace lending platforms such as LendingClub and Prosper have made significant competitive inroads against traditional banks in recent years by bringing together people who want to borrow with investors ready to bankroll them.
But lending platforms, also called peer-to-peer lending, must address a major design problem: Sophisticated investors have been gaming the system by applying specialized screening tools to scoop up the choicest loans with the lowest default rates, leaving less experienced investors with less attractive loans to choose from. After these lower-grade loans perform poorly—that is, the borrowers fall into arrears with payments or default altogether—these less savvy investors may flee the platform.
“If the sophisticated investor picks all the good fruit, the unsophisticated investor will do poorly and [potentially] leave the platform,” says Harvard Business School Assistant Professor Boris Vallée. “Platforms need to be attentive about this potential problem when thinking about their design. They want as many investors as possible, both sophisticated and less so, [and] they want to make sure it’s a sufficiently fair game so everyone can participate.”
Can lending platforms make their systems more equitable for all investors? It’s an important question because if enough disgruntled investors leave a platform, the result can include fewer loans available, and at potentially higher rates, which in turn makes the overall platform less competitive.
In their new working paper Marketplace Lending: A New Banking Paradigm? Vallée and Yao Zeng, an assistant professor of finance at the University of Washington, address these issues from the perspective of what platforms can do to level the investing playing field.
The key variable to control, Vallée and Zeng found, is the amount of information available about loan applicants. When platforms share a lot of information about applicants with potential investors—data such as income, debt level, and credit history, and even whether the loan is financing a wedding, for instance—experienced investors can precisely pin down the safest loans to back. If platforms limit the amount of information available, the playing field is more level and a wider variety of investors have a better shot at choosing cream-of-the-crop loans.
Not your parents’ bank
Peer lending markets have quickly gained a fatter slice of the consumer-lending business over the last decade. In 2016, these loans represented a third of unsecured consumer loan volume in the United States. And their revenues are expected to grow 20 percent each year over the next five years.
Here’s how loans on these platforms work, a process that differs significantly from a traditional lending bank. Once an application is filed, the lending platform collects information on borrowers through online self-reporting and by pulling credit reports. The platform automatically requests supporting documents to verify some of this information for a fraction of the applicants.
Borrowers and platform employees have no personal interactions; no physical meetings, phone interviews, or web chats. The platform pre-screens loans based on the borrower’s information, and either accepts or rejects the application. If accepted, borrowers are assigned a grade depending on their risk, which sets the interest rate for the loans.
After the pre-screening, successful applicants are listed on the platform and some information is provided to investors to help them with their decisions.
Investors then pick the loans they want to fund. This represents the biggest departure from the traditional banking model, where banks do the legwork on behalf of investors, and investors may have little or no information about loans—and in fact may not know which loans their money is supporting.
“The lending model has fundamentally changed,” Vallée says. Investors in the traditional banking system had no information about applicants nor any idea how their capital would be deployed. In addition, decisions were made by bankers using largely subjective criteria, or, as Vallée puts it, “He had a soft shake so I’m not going to lend to him.”
The shift to automation and data science makes this model obsolete. Behavioral scientists tell us that where there is human interaction, people will have biases that influence their decisions, typically in the wrong way.
“Now decisions are increasingly based on hard information,” Vallée says. “As a sophisticated capital provider, you have 10 million data points that help you decide on how to optimally allocate your money, and everything is automated. That’s the future of banking.”
The robot edge
Investors in marketplace lending who do their due diligence use quantitative models like one developed by LendingRobot, with its automated investment and sophisticated screening tool based on historical data provided by the platforms. Investors using this tool can execute a purchase order at high speed online, which is important for quickly grabbing high-demand loans.
These sophisticated investors see a payoff from doing their homework: Their loans perform much better than loans chosen by less experienced investors.
The researchers looked at all transactions executed by LendingRobot users for a three-year period between January 2014 and February 2017, including more than $120 million invested on the two major lending platforms, LendingClub and Prosper. They found that using the LendingRobot screening model paid off by reducing the average loan default rate by more than 20 percent compared to the average level on the platforms.
If peer lending platforms are looking to entice a healthy mix of investors, Vallée says that they may want to provide a lot of information initially to attract investors in the first place. Then as the platforms mature, they can scale back the amount of information they share over time in an effort to level the playing field among investors.
In an apparent effort to prevent sophisticated investors from maintaining a consistent edge, LendingClub in November 2014 removed 50 of the 100 information variables that it had previously shared with investors. The researchers found that this “shock to the information set” had a huge effect: The outperformance of sophisticated investors dropped by more than half.
“As a marketplace, you get sophisticated investors interested in playing the game by giving them some information. They bring their capital, and you can even learn from them,” Vallée says. “But in the long run, giving them all the information you have is too much.”
