UNDERWRITING AND CREDIT SCORING

Underwriting of many online lenders looks quirky at best to some traditionalists in the financial sector. Marketplace lending platforms often start with consumer credit risk scores, credit histories and debt-to-income ratios, and augment this to varying degrees with other data sources and analytics.

The marketplace lending platform Prosper, for example, uses credit scores and other standard consumer risk indicators as its basis for determining eligibility and credit spreads. They supplement these data with results from a proprietary loss-forecasting model called the Prosper Score. Proprietary modeling incorporates factors such as online referral channels as well; a borrower who arrives at Prosper via a debt education site may qualify for a cheaper rate. Prosper maintains that people coming from referral sites repay loans at a lower loss rate. For this reason, the platform uses the referral information in pricing. Prosper also employs data scientists who analyze other factors that inform pricing, among them credit availability and investor demand.

Lending Club’s approach is a little more straightforward and focuses on Fico scores, debt-to-income ratios and credit histories to price loans.³¹ Other marketplace lenders use Big Data explicitly to determine the risk of borrowers instead of relying on consumer credit scores. New York-based OnDeck, which specializes in small business loans, uses reviews on Yelp and Google Places as inputs in its underwriting model. One platform claims its machine-learning algorithm analyses 15,000 pieces of social media data to price loans, according to an investor in marketplace loans.³²

Within the online lending industry, purely computer-driven models based on machine learning and Big Data have raised some eyebrows. Established underwriting methods have worked relatively well for decades, even though they lack algorithms and social media data. The loss of transparency that the proprietary algorithms of each online lending platform introduce is another cause for concern. As a result, some buyers of loans liken online lenders to “playing Wizard of Oz” with their black box underwriting methods.

Certain rules prohibit lending decisions based on information gathered from social media profiles. For example, the Equal Credit Opportunity Act (ECOA) prohibits lenders from making a lending decision on the basis of an applicant’s race, color, sex, age, religion, national origin, marital status, or the fact that all or part of the applicant’s income comes from a public assistance program.³³ As with any other emerging technology used in com­merce, lenders should ensure their use of that technology is consistent with fair lending laws.

2.11