US banks are researching use cases of alternative data in the lending space, according to software vendors.
“The risk of using alternative data without some sort of guidance framework is that data can be used intentionally or unintentionally in a discriminatory fashion. So that’s why we are very concerned as an industry about the use of alternative data,” says Brian Costello, VP of data strategy, Yodlee. “And we have to be careful with this because typically we use alternative data in a machine learning, AI scenarios and we’re still figuring out all of the bugs in the AI model.”
“The label 'alternative data' is really tied to the context of using bank statement data, using your transaction data in a lending scenario. I am aware of just a few that are starting to really look at that in their underwriting process. The vast majority of banks I speak with are using it for the affordability check side,” he says.
Naeem Siddiqi, senior advisor, risk research and quantitative solutions, SAS says he is also aware of four financial institutions that are currently at proof of concept stage.
In August, the Consumer Financial Protection Bureau estimated that 19 million Americans' credit history is either insufficient or too old to produce a credit score under current credit scoring models. 26 million have no credit history at all.
To address this, on December 3 five US regulators published a joint statement on financial institutions' use of alternative data in “credit underwriting, as well as in fraud detection, marketing, pricing, servicing, and account management”.
The regulators defined alternative data as “information not typically found in consumers’ credit reports of customarily provided by consumers when applying for credit. Alternative data include cash flow data derived from consumers’ bank account records.”
The statement outlined the need for compliance management programs to be implemented before the use of alternative data in credit underwriting. This would include testing, monitoring, and controls to assess the risks of using such data.
On November 25, New York state governor Andrew Cuomo signed a law to prohibit lenders from using social media to calculate an individual’s credit score. According to the governor’s statement the US General Accounting Office has estimated that 75 percent of credit scores are incorrect as they are based on erroneous information.
The move by the New York regulator was prompted by the Fair Isaac Corporation (FICO) hoping to include social media data as part of credit scoring calculations, according to the statement.
For Costello, the joint statement by the US regulators is a signal to the industry to begin considering alternative data for credit underwriting.
But financial institutions will wait for regulatory clarity before going beyond research stage, according to Siddiqi.
“It depends on regulatory guidance. The bigger financial institutions are more cautious because of reputational risk. And so if people become generally more comfortable sharing their social media data, they may start using it,” he says.