Two articles combine to highlight the good, the bad, and the ugly associated with utilizing social data for making lending decisions. This article highlights the good that will result from a startup called FriendlyScore that utilizes social data to weigh credit risk. On the other hand, this article describes the ugly horrors that will undoubtedly result from a Facebook patent indicating that Facebook will utilize social data to assist in lending decisions. Neither article identifies the scope of the challenge.
In the FriendlyScore review, the analytics utilizing social data provides a solid underpinning for lending decisions:
“FriendlyScore aims to offer its new approach by using social media data that’s provided by the consumers to serve as a verification system to help those borrowers gain the credibility needed to be seen as a low-risk borrower.
‘Young borrowers want access to loans, and lenders want to make sure they are protected from fraud and high-risk loans. FriendlyScore fills this gap by creating a scorecard based for creditworthiness using over 800 hard data points and social textmining (NLP). An accurate credit score can now be generated simply with a Facebook profile,’ FriendlyScore CEO Maciej Dolinksi said in a company news release.
According to the company, FriendlyScore has a consumer base of 6,000 and ‘works with lending partners by integrating an open API to pull predictive data based on repayment feedback and combines that with real-time, text-extracted social media analysis to generate a credit score.’ ”
Hopefully FriendlyScore has applied for a patent for this, as the next article identifies Facebook has already applied for a patent on similar technology:
“If there was any confusion over why Facebook has so vociferously defended its policy of requiring users to display their real, legal names, the company may have finally laid it to rest with a quiet patent application. Earlier this month, the social giant filed to protect a tool ostensibly designed to track how users are networked together—a tool that could be used by lenders to accept or reject a loan application based on the credit ratings of one’s social network.
You could be denied a loan simply because your friends have defaulted on theirs.
In short: You could be denied a loan simply because your friends have defaulted on theirs. It’s the kind of digital redlining that critics of “big data” collection have been warning of for years. It could make Facebook a lot of money, and it could make the Web even less safe for poor people. And it could be just the beginning.
This article goes on to describe the banks history of discriminatory lending practices and suggests that utilizing social networks will have a similar outcome.
Both articles greatly oversimplify the process by which banks will evaluate any new credit rating approach. For example, lenders will need to prove to regulators that the decisions driven by these new tools don’t have a disparate impact on any protected class (in either design or outcome). Any database used to make a credit decision will likely also need to meet rules regarding the consumers’ right to dispute inaccuracies in their credit report, which might be sufficient to sink the use of social data for credit decisions.
From an analytic perspective, there would be a set of logical, structured data elements that lenders could easily mine and incorporate into their models, such as a long term friendship with an individual that filed for bankruptcy, but even these would need to be justifiable to both regulators and consumer groups.
More difficult to justify to regulators and consumer groups would be the use of unstructured data elements that might prove to be predictive in the future. If, for example, it is discovered that individuals that use all capital letters in their status updates are high risk, it will need to be tested and then regulators satisfied that there are no negative consequences of the approach. This will almost certainly be a harder sell, and will likely prove more difficult to mine and analyze. Perhaps more importantly, it will also be easier for consumers to change if they learn it impacts their credit rating.
Lastly, there is the cost associated with becoming a regulated credit rating agency. Will Facebook investors be comfortable with Facebook investing heavily in a highly regulated business? In all, Mercator doesn’t expect to see social data integrated into a credit offering in the near term and when it is done, it will be introduced by an existing credit rating agency that has the deep pockets and regulatory experience needed to introduce a statistically valid new attribute into the existing rating process
Overview by Tim Sloane, VP, Payments Innovation at Mercator Advisory Group
