Value of Social Media Data Disputed In Assessing Consumers’ Creditworthiness

Professionals in the arenas of consumer finance and privacy expressed doubts about a recent wave of news stories and blog posts asserting that lenders will soon use data culled from social media channels to assess the creditworthiness of consumers applying for car loans, credit cards and home mortgages.

Attorneys consulting major lenders acknowledged that social media data could be used as a tool for assessing creditworthiness, but stressed that this strategy generally has not been injected into the scoring models or the internal underwriting practices used by the vast majority of financial institutions and credit companies. They also questioned the usefulness of information pulled from consumers’ Facebook posts and Twitter tweets in the context of assessing credit risk.

“This is not prevalent, and the reason it is not prevalent is so much of the traditional consumer credit underwriting is effectively done through automated scoring models,’’ said Jeffrey Taft, a partner in the Washington office of Mayer Brown LLP, whose practice focuses on banking regulation and consumer finance.

A spokesman for the Consumer Data Industry Association (CDIA), the industry organization representing the big three credit reporting agencies (CRAs) TransUnion LLC, Equifax Inc. and Experian plc, echoed such views.

CRAs Focusing on FICO, VantageScore.

“To my knowledge none of our members are using this,’’ said Norm Magnuson, CDIA’s vice president for public affairs. “They are more into the traditional scoring systems, VantageScore and FICO. Those are the products they use, buy and sell.’’

Attorneys interviewed by Bloomberg BNA pointed to potential legal landmines confronting lenders that might rely on scoring models featuring social media information. Such models could trigger scrutiny under the Equal Credit Opportunity Act (ECOA), which specifies that credit scoring systems be scientifically derived and have no disparate impact on protected classes.

As a practical matter, the attorneys and credit professionals also disputed the wisdom of a social media credit model when current scoring systems, focusing heavily on a consumer’s payment history and previous utilization of credit, appear to be working well.

“No lender is going to disassemble their traditional credit report evaluation protocol to plug in a Facebook evaluation protocol. To me, that just doesn’t make any sense,’’ said John Ulzheimer, president of the credit industry consulting firm The Ulzheimer Group LLC and a former executive of both Equifax and Fair Isaac Corp., developer of the FICO credit scoring system.

“The reason these tools are so commonly used is because the people in operations research and risk management in banks and insurance companies recognize that they actually work,’’ he said.

Nontraditional Lenders.

Ulzheimer said the only niche in the consumer finance industry where the social media scoring model seems to hold some currency is with a very small group of nontraditional lenders. Operating from microfinance and peer-to-peer lending models, these small-scale credit providers are looking at social connections and community affiliations as their basis for assessing risk.

One example is the two-year-old microfinance company Lenddo. Based in New York, but operating primarily in emerging markets, Lenddo uses community-based microfinance techniques with social media data as a basis for extending credit to the unbanked.

“Lenddo’s community members can use their reputation on social networks such as Facebook, LinkedIn, Twitter and Yahoo! to obtain life-improving loans, to use for education, healthcare, home improvement or a small business,’’ the company states on its website.

Credit industry professionals interviewed by Bloomberg BNA conceded that Lenddo’s innovative business model could cause larger lenders to rethink some of their practices, particularly for consumers who do not fit comfortably within the walls of current credit models. In this regard, immigrants, students and unbanked consumers might benefit from scoring systems that consider other factors predictive of their ability to repay loans.

FICO Considering Social Media Strategy.

Anthony Sprauve, a spokesman for San Jose, Calif.-based Fair Isaac, said his company is engaged in a process to assess whether different types of data might enhance its risk analysis products. While Fair Isaac’s FICO remains the most widely used credit scoring system in the country, Sprauve said alternative data pools could prove useful in building a broader picture of a consumer’s creditworthiness.

Under this process, Sprauve said Fair Isaac is considering strategies that fold social media data into its scoring systems, along with payment histories to utility companies and landlords. He said such information could prove particularly beneficial to consumers operating without a traditional credit report.

At the same time, Sprauve said there are fundamental flaws with social media data for the purposes of assessing creditworthiness. Sprauve said the ECOA and the Fair Credit Reporting Act require credit scoring models to rely on sources of data that are pervasive, verifiable and correctable—requirements that raise doubts about the viability of social media data.

“Not everyone is on Facebook, and not everyone is on Twitter,’’ Sprauve said. “And you can basically make up whatever you want on those sites, and LinkedIn as well. And there is no real mechanism to verify or correct information that is wrong or misleading. So those are the things that lead to legal and regulatory questions.’’

Consumers Pushing for Alternative Scoring Schemes.

Joel Winston, a partner with Hudson Cook LLP in Washington, said he sympathizes with Fair Isaac and the CRAs as they examine alternative scoring schemes. Winston said consumer groups have challenged the industry to develop less-restrictive protocols, permitting wider access to credit for consumers falling outside of the traditional scoring models.

But Winston, who previously served as associate director of the Federal Trade Commission’s Division of Financial Practices, said some of the data featured on social media sites raise potential disparate treatment questions under the ECOA. While traditional scoring models avoid certain information about credit applicants, Winston said, Facebook could provide information about race, gender, religion, sexual orientation, national origin, marital status and age.

“I’m not surprised that FICO and the big three credit bureaus are exploring this. It is a way to make money and provide greater access at the same time,’’ Winston said. “But once you get past rent payments and utilities, and you start getting into certain characteristics that people might reveal on their Facebook page, that’s when things get a little risky.’’

CFPB Monitoring Credit Scoring Innovations.

While the Consumer Financial Protection Bureau has not issued any policy statements on social media in the context of access to credit, the agency said it will monitor innovations in this arena.

“We would expect creditors to ensure that their scoring models do not have an unjustified disparate impact on a prohibited basis, meaning that they meet a legitimate business need that cannot reasonably be achieved as well by alternative means that are less disparate in their impact,’’ said CFPB spokeswoman Michelle Muth Person.

Several professionals interviewed by Bloomberg BNA continued to express skepticism about social media, even if Fair Isaac and the CRAs are able to bridge the challenges around disparate treatment. They said any new credit scoring platforms, with or without social media data, would have to be demonstrably superior to current systems to justify the expense associated with a wholesale conversion.

Big Banks Unlikely to Be Pioneers.

The data industry association’s Magnuson said large financial institutions are not inclined to be pioneers in any potential conversion.

“Banks tend to make traditional lending decisions based on data they are familiar with. They are not the first to adopt a new solution,’’ he said. “From their perspective, it is costly to integrate something into their underwriting systems. There is cost associated with purchasing the hardware and software, and then you have to train your employees and make sure you are in compliance with all state and federal laws.’’

Ulzheimer agreed and added that Fair Isaac and the CRAs would have to be able to prove that their new systems can operate as demonstrably sound predictive tools before lenders take on the risk and expense associated with conversion.

“I don’t know that this information is going to be any more valuable to a lender than how a consumer is paying his bills,’’ Ulzheimer said. “That is by far—and this is not a secret—the most predictive attribute with respect to risk assessment. If my Tweets turn out to be a better predictive tool, well then kudos to whoever figures that out.’’

To contact the reporter on this story: Michael Bologna in Chicago at mbologna@bna.com

To contact the editor responsible for this story: Richard Cowden at rcowden@bna.com