The US car maker's finance unit announced on Friday that it would start to implement machine learning to “further enhance its consistent and prudent lending practices across the credit spectrum”. The move follows a study conducted with ZestFinance, a fintech company specialising in credit data and underwriting, that revealed machine learning could help more accurately predict the risks of lending to customers with little or no credit history.
“We worked with ZestFinance to harness the capability of machine learning to analyse more data and to analyse our data differently. The study showed improved predictive power, which holds promise for more approvals, enhanced customer experiences and even stronger business performance, including lower credit losses,” Ford said in an August statement.
The move to incorporate more technology and alternative sources of data comes at a time when credit scores are quickly losing their shine as the sole indicator of a borrower’s ability to repay debt. According to a May report by the Center for the New Middle Class, a research group at subprime lender Elevate, 109m Americans have a credit score below 700, with another 53m falling under the “credit invisible” category. This has created a growing need for credit products that do not heavily rely on credit scores or credit histories as measures of creditworthiness.
Among the broader group of underserved consumers are millennials, which Ford said are the “fastest-growing segment” of new car buyers. According to the company’s estimates, new vehicles bought by millennials represented 29% of all US sales, with the number expected to grow to 40% by 2020. However, they are a demographic that does not always have the necessary credit records needed in traditional methods of underwriting. Alternative data and credit scoring would unlock this consumer demographic, benefiting both lenders and borrowers seeking credit.
Low or no credit borrowers have long been perceived as too risky by many lenders. A report from Kroll last week suggested that an increase in the number of auto loans made to lower quality borrowers had been behind an increase in delinquencies. Incorporating alternative data will not only provide a better insight into the risk of default, but also help lenders offer rates and repayment terms better suited to borrowers’ true ability to repay.
It is hoped that alternative data, with its promise of predicting risk better than traditional credit scores, will improve lending conditions for subprime and underserved borrowers. Given the growing number of Americans falling into the subprime category, more lenders should also follow Ford in seeking alternative data for underwriting or risk locking themselves off from a large and growing segment of consumers.