When you think about the greatest invention of the 20th century, a few key technologies might come to mind: the personal computer, internet, antibiotics, or the electric guitar. But, for Douglas Merrill, founder and chief executive of Zest Finance, nothing compares to the Fico score.
“The Fico score is the most important innovation of the 20th century and no one gives them any credit for it,” Merrill told Auto Finance News. “But then, in 1980, credit availability basically just flatlines. The right question is: What happened there?”
Zest tries to solve this issue by using algorithmic techniques, artificial intelligence, and machine learning he studied during his time as the chief information officer at Google before leaving in 2008. The company’s newly launched auto finance software aims to provide credit to consumers traditionally considered too risky by adding more points of information than a traditional Fico score provides.
Fico scores also helped to limit discrimination in auto finance, and the Zest platform aims to bring more data to that process, as well. Zest estimates it can decrease discrimination in the industry by providing tools that will allow lenders to comply with the Consumer Financial Protection Bureau’s rules on disparate impact, using the guidelines of the BISG model.
“There is always some disparate impact, but the tricky part is, the law doesn’t say you can’t have any disparate impact, it says you can’t have any ‘material disparate impact,’” said Merrill, who worked at the company that developed the model — Rand Corp. — for five years.
The platform can take the predicted economic outcomes of a protected class and overlap the actual economic impact the lender is having on that group of people. For example, Chinese borrowers are expected to perform one way in a portfolio, so do the lender’s terms lineup with that prediction? Or are Chinese borrowers being adversely affected by the financing practices — intentionally or unintentionally.
This is something sports analysts have done with statistics for years. For example, you might take NBA star LeBron James’ past history of hitting three-point shots for the Cleveland Cavaliers and form a model that predicts how many he’s expected to make in 2017. You could then see if he’s performing below or above that expectation, and isolate the factors impacting his performance.
The CFPB does not state that all protected classes need to have equal approval rates, it only requires that lenders do not discriminate knowingly, and to the extent that there is unintentional discrimination, Merrill said. Lenders have to mitigate it, he added.
Zest would look at a lender’s predicted and actual outcomes to determine if — for example — Chinese borrowers are experiencing “material disparate impact” under the financial institutions current practices, he said. If those protected classes are being economically affected by the lender’s practices then they have the tools to mitigate it.
“So, you can do this often enough that even if you made a mistake, and you’re building some disparate impact over time, you’ll know in a week instead of a month and fix it right away,” he said.