Although the headline figure is truly concerning, it is not the most peculiar aspect of the auto loan story of the last two years. It’s the quiet that surrounds it. The financial media has largely ignored the fact that subprime auto delinquencies in January reached a level not seen since Bill Clinton’s first year in office. Bank stocks did not plummet. There was no emergency call from the Fed. Every used car lot in Phoenix, Atlanta, or Chicago’s outer ring will have the same signs as it did two years ago: financing banners, balloon-string pennants, and a salesperson waving you in while wearing a quarter-zip. The pain is real, but it’s localized, and that’s an accomplishment in its own small way.
2008 was the comparison that everyone used for a long time. It made sense intuitively. There was a rhyme between loose credit, an increase in defaults, and Wall Street’s desire for riskier paper-filled asset-backed securities. However, there are significant differences between the behavior of auto loans and mortgages. During the bubble years, a house was seen as a lifesaver. No car ever claimed to be that. As soon as you drive it home, it loses value. The entire ecosystem is predicated on the idea that the collateral is decreasing; lenders are aware of this, and borrowers eventually discover it. That has a peculiar honesty to it.
The borrower is not what has changed. If anything, the borrower’s situation has gotten worse due to rent, groceries, and the average new car’s current $50,000 price tag. The underwriting is what has changed. Ten years ago, no one took fintech lenders seriously. Now, they score people based on their utility history, rent payments, and cash flow. Although the FICO score remains, it is no longer the only voice in the room. Speaking with industry insiders gives me the impression that the outdated credit-score-as-oracle model finally gave way. Forged paystubs are detected by real-time fraud screening services like Point Predictive before the loan is even funded. Platforms for refinancing catch borrowers before they get out of control. None of it gets rid of the default. It simply has it in it.
You can find out who took the damage by looking at the casualty list. In September of last year, deep-subprime expert Tricolor Holdings filed for Chapter 7. That’s the model; instead of showing up on JPMorgan’s balance sheet, the suffering remains inside the lenders designed to handle it. For those who lost their cars, it’s a small act of kindness, but it has a significant impact on the economy as a whole.
It’s difficult to ignore how unromantic the rescue has been as you watch this play out. There was no White House summit, no televised testimony, and no rescue legislation. Over 90% of loans are now auto-decided, review times have decreased by almost 99%, and thin-file borrower approval rates have increased by about a third. This is just a slow, unsexy shift toward better data. Compared to the humans they replaced, the machines are not any more compassionate. Simply put, they are slightly better at saying yes to those who might actually pay and quicker at saying no.
Whether this is true is still up for debate. The model is put to the test in ways it hasn’t been before if unemployment continues to rise and used car values continue to decline. However, the doomsday theory, which the chief economist at Cox Automotive referred to as “comical” in 2018, has now withstood a real, record-breaking delinquency wave. That is not insignificant. The crisis materialized. It didn’t spread.
