Squeezed a couple things into one post to avoid clutter.
Newest Leaders in Lending podcast:
Listened to it, good discussion on personalizing marketing for financial institutions but didn’t gain any new information pertaining specifically to UPST (other than Unison credit union is probably a partner. A little discussion on CUSOs in the podcast too. They are small, at $284M in assets and located in Wisconsin).
UPST competitor, Lending Club, reported Q3 results today.
Loan origination volume increased 14% QoQ. Will need to listen to earnings call later today to parse any granularity discussed about expected loan seasonality and current macro environment.
One last thing on Trustpilot for UPST
Organic review number count Jan 2021 23 Feb 2021 16 Mar 2021 22 Apr 2021 28 May 2021 31 Jun 2021 29 Jul 2021 40 Aug 2021 42 Sep 2021 44
On quarterly comparisons, we get 61 organic reviews for Q1, 88 for Q2, 126 for Q3.
Q2 was 44% sequential growth over Q1 (actual Q2 loan transaction number was 68% growth over Q1)
Q3 is 43% sequential growth over Q2. (So, will actual Q3 loan transaction number be at least 60% growth over Q2?)
Take it with a grain of salt. To me it’s another pseudo-indicator that supports my belief that loan transaction number and revenue growth for Q3 will exceed at least 30% QoQ. But… if it’s another 60% QoQ growth…
The auto loan market is incredibly inefficient and ripe for disruption by Upstart.
A consumer report research article was released today. They pulled data from loan securitization transactions and analyzed about 858,000 loans made by 17 major lenders, including banks, financial firms affiliated with automakers, and companies that cater to lower-credit consumers.
The main takeaway:
Auto loan consumers are receiving mispriced loans that are mispriced NOT just because of the use of FICO scores, but far beyond that. They are being mispriced by perverse incentives, where the dealer and lender try to get away with whatever they can - because if the borrower defaults, they can just repossess the car [1.7 million, or 13% of nonprime loaned vehicles, were repossessed in 2019].
To me this indicates that the auto loan market is WAY more inefficient than the personal loan market.
The personal loan market is mostly inefficient because of the over-reliance on FICO scores.
But if the auto market has both FICO inefficiencies AND artificial lender/dealer inefficiencies, then this is truly a greenfield space for UPST’s AI models to disrupt and operate within.
1.) A credit score doesn’t necessarily dictate the terms of the loan offered. Borrowers in EVERY credit score category—ranging from super-prime, with scores of 720 and above, to deep subprime, with scores below 580—were given loans with APRs that ranged from 0 percent to more than 25 percent.
3% of borrowers in the superprime/prime credit score group had APRs of 10 percent or greater!
There is a huge opportunity for auto refinance across all credit score bands, for sure.
2.) Many auto loans are given out to borrowers who shouldn’t have gotten a loan in the first place.
The report found that lenders verified the income reported on credit applications just 4 percent of the time. Employment history was verified at an even lower rate.
I am sure UPST’s AI verification models would be far superior than the current industry rates here - and permit proper pricing of loans.
3.) Delinquencies (overall) are surprisingly common, with over 5 percent of all loans delinquent, but this is likely covered up by the pandemic stimulus. In Q4 2019, 30 day delinquency rates for nonprime loans were at 11.2% (that is crazy!). This is all very indicative of ridiculously mispriced loan APRs.
And they all know it - a June 2021 survey of over 100 auto lenders/servicers showed that 75% of them believe subprime auto loan performance will deteriorate in the coming year.
4.) Many auto loans trigger a feedback loop: if they are so highly mispriced and offer an APR that is way above the actual risk of the borrower, that high APR causes monthly payments to become too high and directly contributes to the borrower defaulting. If the borrower were offered a true APR priced to their actual risk in the first place, they may not have ever defaulted.
An industry publication found that a payment-to-income ratio of more than 14 percent leads to “50 percent higher defaults at every credit score level.”
An example given in the report:
Lamar bought a used Toyota Tacoma truck at a local dealer in early 2019, putting down a $2,500 cash payment and receiving a $3,500 credit for trading in his older vehicle.
Lamar, 55, says he earns about $80,000 annually as an insurance account manager and that before the loan he’d never held debt of any sort. The dealer told him he needed to pay a 17 percent APR to finance the purchase over six years from Santander.
Payments were $900 per month.
Lamar made it work for a while, but by spring 2020, he fell behind. The lender gave him one extension, he says, but the truck was repossessed in June.
I’m pretty sure UPST would not have mispriced his loan to 17% APR and could have kept the monthly payment within reason!
5.) There is a racial disparity: Nonwhite borrowers pay more on average for auto loans than similarly situated white consumers. Certainly something that UPST’s AI can help close the gap, as it has with the personal loan market. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3301009
6.) Consumer report analysis couldn’t figure out why exactly markups are so high for borrowers. (But as I said above, I think it’s multifactorial - dealers can markup loans and charge finance fees themselves, and then if lenders charge too high they can just repossess the car if they default and that car can be exchanged to other subprime borrowers over and over again)
At least 80 percent of car financing is arranged through dealers, [which explains why UPST decided to position themselves at the dealership, with Prodigy] who serve as intermediaries for lenders, according to a 2020 paper published by the National Bureau of Economic Research. In a typical arrangement, the dealer submits a borrower’s information to lenders, receiving loan offers in return. Dealers then can then legally increase, or “mark up,” the interest rate, and they have been shown to typically do so by 1 to 2 percent. The arrangement isn’t good for consumers, regulators and experts say: Dealers aren’t required to show consumers the offers they received, meaning they might not provide customers with the best deal.
But while dealer markup has been cited before as one factor behind APR variation among similar auto loan customers, it doesn’t necessarily explain the disparities CR identified. For one thing, dealer markups are generally capped regardless of APR, says Paul Metrey, senior vice president of regulatory affairs at the National Automobile Dealers Association, and “there is no financial incentive for dealers to present longer-term or more expensive credit options to consumers.”
So what drives the variation in the loan data CR reviewed?
To assess it, CR statisticians built a modeling tool that looked at APR, controlling for the borrower’s payment-to-income ratio, when the loan was issued, whether a co-borrower was present, the length of the loan, the amount of equity in the car, or whether the purchaser received financial incentives on the loan, which might include a 0 percent interest introductory period.
None of these characteristics could fully explain the wide disparity in APRs offered.