Fitch Ratings put out a number in January that should have gotten more attention than it did. Subprime auto loan delinquencies, borrowers 60 or more days behind on payments, hit 6.90 percent, the highest reading since Fitch's records began in 1994. Prime borrowers, meanwhile, sat at 0.39 percent over the same stretch. The gap between the two tiers is now more than tenfold, and it's been widening since 2021.
I read that number against my own research rather than against the headline, though I want to be careful about exactly what kind of evidence this is.
What my paper actually tested, and what today's data is consistent with
My paper on securitized auto loans tested how much predictive power a borrower's credit score contributes during calm periods versus stressed ones, using the COVID disruption in 2020 and the rate tightening stretch in 2022 and 2023 as the stress windows. During those periods, the marginal explanatory power that credit score added to the model fell from 0.0416 to 0.0236, close to a halving. Loan to value ratio and payment burden became more predictive over the same stretch, not less.
I have not re-run that hazard model on 2026 vintage data, and I want to be clear that what follows is a reasoned interpretation, not a confirmed replication. What's happening in the subprime delinquency spike right now is consistent with the mechanism my research identified, a genuine affordability squeeze in which score-based risk tiers become less reliable and structural inputs like leverage and payment burden carry more of the real signal. It would take running the current data through the same model to call it confirmed.
Here's what's driving the current spike. Average new car payments have climbed past 750 dollars a month, with close to one in five loans or leases now above 1,000 dollars monthly. Wage growth has slowed. J.D. Power reported that nearly 14 percent of new car buyers had credit scores below 650, the highest share since 2016, meaning lenders extended credit into the subprime tier at a rate not seen in almost a decade, right as household budgets tightened from the other direction.
More than 6% of subprime auto loans are now at least 60 days past due, the highest rate ever recorded.Fitch Ratings, via CBT News
If the mechanism from my research does extend to this cycle, and testing that directly is the natural next step rather than something I've already done, the practical implication for anyone underwriting or investing in this market is straightforward. Pool composition by credit tier is the number every headline leads with, and it may be the least useful number for assessing which 2025 and 2026 vintages are actually at risk. Loan to value at origination and payment to income ratio are the inputs my research suggests carry more signal in exactly these conditions. A servicer or an ABS investor who wants to know which pools are genuinely exposed right now would be better served checking those two numbers before checking the subprime share of the portfolio, though I'd want to validate that against current vintage data before treating it as settled rather than a well-grounded hypothesis.