American car loans are not going bad faster. They are going bad for longer
The share of car debt sitting in serious delinquency has never been higher. The rate at which car debt enters serious delinquency is lower than it was in 2009. A stock and a flow are not the same number, and only one of them is in the headlines.
Stock photo
Auto loans are entering serious delinquency more slowly than they did in 2009. The share of American car debt sitting in serious delinquency has never been higher.
Both sentences are true. Both come from the same table, published on the same day, by the same institution.
The Federal Reserve Bank of New York released its Quarterly Report on Household Debt and Credit for the second quarter of 2026 this month. Page 12 of the underlying workbook says 5.49 percent of auto loan balances are 90 or more days delinquent, which is higher than any reading since the series began in 2003. Page 14 says 3.00 percent of balances moved into that condition over the past four quarters, against 3.48 percent in the year to June 2009.
Asked about the auto loan delinquency rate on 26 August, Google's own answer box returned "surged to historic highs", a figure "near 6.9%", and "the pace of acceleration has begun to stabilize", in one paragraph, drawn from four different sources measuring four different things. Every clause was defensible. Together they described nothing.
What is the auto loan delinquency rate right now?
There is no single answer, and the disagreement is not a dispute. Five official measures are current, each counting a different population: balances against loans, all borrowers against subprime, a standing stock against a year of new entries. The lowest reads 1.68 percent and the highest reads about 6 percent.
| Measure | What it counts | Source | Latest |
|---|---|---|---|
| Share of auto balances 90+ days delinquent | Dollars sitting in serious delinquency, including balances already charged off | New York Fed, 2026 Q2 | 5.49% |
| Transition into 90+ days delinquent | Dollars that entered serious delinquency over four quarters | New York Fed, 2026 Q2 | 3.00% |
| Transition into 30+ days delinquent | Dollars that fell at least one month behind over four quarters | New York Fed, 2026 Q2 | 7.87% |
| Share of loans 60+ days past due | Accounts, not dollars, seasonally adjusted | Philadelphia Fed, 2025 Q3 | 1.68% |
| Subprime share of loans 60+ days past due | Accounts of borrowers scoring below 620 at origination | Philadelphia Fed, late 2025 | about 6% |
A sixth measure exists at the Federal Reserve Board, whose staff use the same credit bureau panel as the New York Fed but count balances at least 30 days past due while excluding severely derogatory balances. Their series runs only to the third quarter of 2025.
None of these is the wrong number. Reading them as one series is the error, and it is the same error the credit card delinquency rate invites for the same structural reason.
Why does the record high not mean more people are falling behind?
Because a standing stock and an annual inflow move independently. The 5.49 percent figure counts every dollar currently sitting in serious delinquency, however long it has been there. The 3.00 percent figure counts the dollars that arrived during the past year. A pool can rise while the tap runs slower, if the drain runs slower still.
Set the two side by side at the worst moment of the last credit cycle and at this one.
| Flow into 90+ over four quarters | Standing share at 90+ | Implied years of residence | |
|---|---|---|---|
| 2009 Q2 | 3.48% | 4.47% | 1.29 |
| 2026 Q2 | 3.00% | 5.49% | 1.83 |
In the year to June 2009, more auto debt went seriously bad than in the year to June 2026, and less of it was sitting there. That is the whole puzzle, and it resolves into a single sentence: delinquent car loans are being resolved more slowly than they used to be.
The third column is this publication's own arithmetic, not anybody's published statistic, and it deserves its caveat here rather than in a box at the bottom. Dividing a point-in-time stock by a trailing-year flow assumes a steady state that neither series is in, and both series are shares of the same balance base, which is what makes the division meaningful at all. It is a magnitude check. It establishes an order, not a duration.
It also does not establish a record. The same ratio reads 2.28 at the end of 2010 and 2.09 at the end of 2019, both above today. What is unusual now is the combination: a record standing share arriving while the inflow sits below its own post-pandemic peak.
Is somebody else saying this, or is it just an arithmetic trick?
The Federal Reserve Bank of Philadelphia published the same finding in April 2026, from different data, at the level of individual accounts rather than balances. Its Consumer Finance Institute decomposed the 60-plus delinquency rate into newly delinquent loans, loans delinquent across multiple quarters, and redefaulters, and found the growth concentrated in the middle group.
Their report states it without hedging:
the headline auto loan delinquency rate is primarily driven by a decline in the rate of exit from delinquency, rather than a growing proportion of newly distressed borrowers
The authors, Julia Cheney, Bob Hunt, Lauren Lambie-Hanson, Larry Santucci and Justin Zhou, go further in the conclusion: the headline rate "taken at face value, likely overstates the degree to which auto borrowers' financial health is currently deteriorating."
They offer a mechanism. Lenders have expanded loss mitigation, and extensions on subprime loans reached about 3.5 percent, roughly 100 basis points above the level three years earlier, on data the authors take from Intex Solutions. An extension returns a borrower to current status by moving missed payments to the end of the loan. Where the underlying problem persists, the borrower falls behind again. The loan cycles rather than resolving, and it stays in the delinquent population the entire time.
Their conclusion and this one are arrived at independently, from a stock-versus-flow comparison of balances on one side and an account-level decomposition on the other. That is worth more than either alone.
Who is actually behind on their car payments?
Not a cross-section of America. Borrowers who scored below 620 when their loan was written hold 17 percent of active auto accounts and account for nearly two-thirds of all delinquent loans, per the Philadelphia Fed. Their delinquency rate has run near 6 percent since mid-2024, the highest in over 20 years of collection.
Concentration matters for how the aggregate should be read. A national rate at a record can coexist with a prime borrower base whose behaviour has barely moved, and it does.
The vintage effect is the second concentration. Over 62 percent of auto loans in default at the time of the Philadelphia Fed's report were originated between 2021 and 2023, the years when vehicle prices ran furthest ahead of incomes. Federal Reserve Board staff put the mechanism in dollars: monthly payments rose nearly 30 percent between 2020 and 2023, driven more by what cars cost than by what loans cost, and earlier Board work attributed roughly 40 percent of the rise in the two-year delinquency rate between the end of 2019 and the end of 2022 to those payments.
Negative equity keeps those borrowers in place. In the fourth quarter of 2025, 29.3 percent of new vehicle trade-ins involved a loan larger than the vehicle was worth, the highest share since early 2021, on Edmunds figures cited by the Philadelphia Fed. A borrower who cannot sell the car for what is owed on it cannot leave the loan, and a borrower who cannot leave the loan cannot leave the statistic.
Meanwhile the credit being written now is not obviously worse. Loans to borrowers below 620 were 16.1 percent of auto originations in the second quarter of 2026, against 19.5 percent at the end of 2019 and 29.5 percent at the end of 2006. Originations themselves jumped to 211 billion dollars in the quarter, the strongest in the series.
One number does cut the other way. The New York Fed reports that the median credit score on new auto loans fell seven points in the second quarter. That reading is only comparable with the previous quarter, because the report switched from the Equifax Risk Score 3.0 to VantageScore 4.0 at the start of 2026, and the report says so in a footnote. Anyone comparing a 2026 score with a 2024 score is comparing two different rulers.
What is the "32-year record" everybody is quoting?
It is a ratings agency's index of securitised subprime auto loan pools, not a measure of American auto credit. It covers loans bundled into asset-backed securities from subprime issuers, which is a slice of a 1.71 trillion dollar market, and its composition has been revised as issuers were added.
The figure comes from Fitch Ratings and is carried by the outlets that currently rank on this query, including the Yahoo Finance story sitting at the top of the results page. Fitch's own research sits behind a subscription. Requested on 26 August, the page returned a navigation shell and no reading.
So the number is not printed here. This publication does not reprint a figure it could not open, which is a rule that exists because a plausible summary of a document is not the document, and the difference has bitten this desk before.
What can be said is what the index is. A record in securitised subprime pools is a real fact about securitised subprime pools. It is not the American auto loan delinquency rate, and every restatement that drops the qualifier makes it sound like one.
Are auto loan delinquencies rising or falling in 2026?
Both, depending on the horizon, and the two are not in conflict. The flow into early delinquency fell across the year to March and rose across the quarter to June. The standing 90-plus share set a record in March and fell in June. The honest summary is that the series has flattened at an elevated level.
| Series | A year earlier | 2026 Q1 | 2026 Q2 |
|---|---|---|---|
| Share of balances 90+ days delinquent | 4.99% | 5.60% | 5.49% |
| Flow into 30+ days delinquent | 7.96% | 7.72% | 7.87% |
| Flow into 90+ days delinquent | 2.93% | 2.97% | 3.00% |
The New York Fed's own summary for the quarter is one clause long: transition into early delinquency "upticked slightly for auto loans and mortgages". Aggregate delinquency across all household debt fell, to 4.7 percent of balances from 4.8 percent.
WardsAuto reported in May that the flow into early delinquency was 7.72 percent in the first quarter, down from 7.99 percent a year earlier. That checks out against the workbook exactly. It is also, three months later, no longer the current reading, which is how a correct number becomes a stale one.
Where this reading could be wrong
The New York Fed's own data dictionary contains the strongest objection to it. Not every creditor keeps updating payment status on an account that has been derogatory for a long time, and the report warns that the resulting profile "may to some extent reflect reporting practices of creditors". A stock measure that fills up partly because lenders stop refreshing stale records would look exactly like a stock measure filling up because loans are resolving slowly.
That is a reporting artefact and a real economic slowdown in resolution producing the same signature, and the balance data alone cannot separate them. The Philadelphia Fed's account-level decomposition is what makes the second reading more likely, because redefaulters and multi-quarter delinquencies are identified as distinct populations rather than inferred from a total.
The other limit is timing. The Board's staff series stops at the third quarter of 2025 and the Philadelphia Fed's charts stop at the fourth. Only the New York Fed series reaches June 2026. Where those sources are quoted above, they describe the run-up to this year rather than this quarter.
What would actually change the picture
Watch the exits, not the level. If charge-offs and repossessions accelerate, the standing share will fall without a single borrower being better off, because the pool drains rather than shrinking at the source. If the flow into 30-plus delinquency keeps climbing off its March low, that is the measure to take seriously, since it counts people falling behind now.
The Philadelphia Fed leaves the open question where it belongs. Loss mitigation practices that hold loans in delinquency rather than resolving them have not been tested against a weakening labour market. Whether lenders extend further or stop extending is the variable that decides which of these numbers moves next.
Frequently asked questions
What is the current auto loan delinquency rate? The most quoted figure is 5.49 percent, the share of auto loan balances 90 or more days delinquent in the second quarter of 2026, from the New York Fed. Measured as the share of loans 60 or more days past due, the Philadelphia Fed puts it at 1.68 percent.
Is the auto loan delinquency rate at an all-time high? The share of balances 90 or more days delinquent reached 5.60 percent in the first quarter of 2026, the highest in a series that begins in 2003, and eased to 5.49 percent in the second. The measures of new delinquency remain well below their 2009 levels.
Why is the auto delinquency rate higher than it was in 2009? Because delinquent loans are resolving more slowly. Less auto debt entered serious delinquency in the year to June 2026 than in the year to June 2009, but more of it is still sitting there, having not yet been cured, charged off or repossessed.
Who holds most of the delinquent auto loans? Borrowers who scored below 620 at origination. They hold 17 percent of open auto accounts and close to two-thirds of delinquent loans.
How much do Americans owe on their cars? Auto loan balances stood at 1.713 trillion dollars at the end of June 2026, up 28 billion on the quarter, across roughly 109 million open accounts.
The quarterly series behind every figure above is published as a CSV, including the residence column, so the arithmetic in the second section can be checked or rejected.
