By: Michael W. Ross
In an article published last month in Law360 (and reprinted in our Consumer Finance Observer periodical), our lawyers highlighted the increasing focus of government enforcement authorities on how companies are using “alternative data” in making consumer credit decisions. For example, the article highlighted that – as stated in a June 2019 fair lending report from the CFPB – “[t]he use of alternative data and modeling techniques may expand access to credit or lower credit cost and, at the same time, present fair lending risks.” Regulators have continued to focus on this area, including on the benefits and risks of using alternative data in lending decisions.
Earlier this month, the CFPB posted a widely reported-on blog entry on the benefits of using alternative data in lending decisions. The CFPB blog post provided an update to the public on the agency’s first and only no-action letter, issued to Upstart Network, Inc. in 2017. In that letter, the CFPB stated it had no intention of taking action against Upstart under the Equal Credit Opportunity Act (ECOA), which prohibits discrimination in lending, for using certain alternative data sources – particularly information about a borrower’s education and employment history – to make credit decisions. To obtain that letter, Upstart committed to implementing a risk management and compliance plan that included a process for analyzing the potential risk that its use of alternative data could lead to impermissible discrimination against protected classes of consumers.
The CFPB’s blog post reported on the results of Upstart analyzing almost two years of data from its risk management process. Its data showed that Upstart’s model approved 27 percent more applicants than would have been approved by a traditional underwriting model (i.e., one that did not use alternative data and machine learning), and led to 16 percent lower average APRs for approved loans. The CFPB also reported that expansion of credit occurred “across all tested race, ethnicity, and sex segments,” and resulted in particular increases in approval among applicants under twenty-five, those with incomes under $50,000, and those with “near prime” credit scores. These results hearken back to a report by the Philadelphia Federal Reserve in 2017 concluding that the use of alternative data in credit decisions (in that case, relying on data from another FinTech lender, Lending Club) expanded access to credit in underserved areas at a lower cost than would otherwise be available.
The news of Upstart’s results was widely reported, as the use of alternative data in consumer lending remains a hot topic that regulators and legislators are continuing to watch closely.
 Government agencies and legislators also continue to focus on the potential risks of alternative data. In June, for example, Senators Warren and Jones wrote a letter to various government regulators highlighting concerns that using algorithms in underwriting decisions could lead to unlawful discrimination.