Jack and Joel are two identical consumers looking to buy a car. Both Jack and Joel get loans to purchase 2014 Honda Civics. Each for $12,000. Jack chose a Civic that was advertised on the dealer’s website for the last 5 weeks. However, Joel chose a Civic that wasn’t actually advertised on the dealer website. Which one, Jack or Joel, is more likely to default on their car loan in the next 12 months?
It's somewhat remarkable, it turns out Joel is much more likely to default! The chart below demonstrates this. Each bar refers to a loan portfolio. For each of the portfolios analyzed loans were defaulting 33% to 74% more frequently when the car purchased had not been advertised on the dealer’s website prior to sale.
The phenomenon held up in all cases: for portfolios composed predominantly franchise dealer originated loans and also for portfolios mostly consisting of independent originated dealer loans. It held up for small regional lenders and it held up for large national lenders. In all cases there was a significant increase in default rate on loans for cars that were not advertised online!
What’s further fascinating is this additional risk is not captured by existing underwriting and loan pricing controls. The increase in loss severity is approximately equal to the increase in default rate. It's tantamount to saying the risk of non-advertised vehicles is a historically un-captured unique signal.
What is the reason for this phenomenon? We suspect it's measuring a unique characteristic about the consumer that is not reflected in their credit history. Basically, this answers the question “did the consumer do homework on the vehicle they are purchasing”. Or similarly, was the consumer possibly flipped into a vehicle they didn’t know about prior to arriving at the dealership. Is the consumer somewhat impetuous? It turns out "yes" creates more risk.
So now we have this really interesting phenomenon of increased risk for purchasing un-advertised cars. It is agnostic of consumer credit (it’s not picked up by credit risk models) and it's agnostic of the dealer selling the car (it’s not picked up dealer score cards). That means it may even apply to direct lending as well as indirect lending....
Let's answer that. We are looking for more auto lenders that want to retro test this phenomenon with their historical direct and indirect originations. There is no PII involved and it's very easy to do. All we use is a VIN and sale date to run the test. Please contact me to discuss. josh@generalforensics.com
I hope you enjoyed and found this interesting. We have a lot of fun doing work like this with our lender partners. I’d love to hear your feedback and thoughts in the comments section below! And thank you to Joel Kennedy for your hugely valuable insight and the great discussion.
Thank you for reading!
Comments