Automated Underwriting and Subprime Lending: Oil and Water?
There’s an old urban legend/joke that goes something like: Back in the 1950s, when the U.S. and Russia were racing to see who could get into space first, the U.S. threw tons of money at developing a pen that could be used in zero gravity. The Russians? They used pencils.
The joke is appropriate in a lot of situations when one is attempting to highlight that the advancements in technology may not always be the most poignant solution.
Take subprime auto loan underwriting, for example.
A recent article in Automotive News highlighted the growth in subprime loan originations in the automotive industry. The article pointed to a number of reasons for the increase, but one in particular struck me as fascinating.
Credit evaluations are becoming less automated … many more credit applications are being evaluated by people rather than by a computer. Chase and Toyota Financial Services say they don’t use automated decision making.
Lenders and technologists have been trumpeting the arrival of automated underwriting for years. The same issue of Automotive News spotlighted a new BMW program where credit applications are being taken via iPads. (Guess they read my blog.)
I remember covering the subprime mortgage space a decade ago and listening to one lender’s discourse about how automated underwriting would never work in the sector because every loan had a story behind it, explaining why each particular borrower had a poor credit score. Those stories could not be analyzed by a computer, the lender said.
Is this true? Why is technology unable to account for the underlying reasons affecting a borrower’s credit score?
Lenders can spend millions of dollars building a pen that works in space, or they can simply have humans underwrite subprime loan applications. But it seems to me that in the long run, a space pen never gets worn down to a nub. The long-term cost savings associated with automated underwriting should create an economic benefit for attempting to solve the problem.
