Credit card balances at least 90 days delinquent came in at 12.9 percent in the second quarter, about two tenths lower than the quarter before. Serious auto delinquency eased to 5.5 percent from 5.6. Aggregate household delinquency sits near 4.7 percent of outstanding debt.
Those are improvements, and they are real.
In the third quarter of 2022 the credit card figure was 7.6 percent.
Both sentences describe the same series. One is a quarterly direction and the other is a level, and a reader handed either one on its own has been told something true and been left with the wrong impression.
Why the level matters more than the delta here
A two-tenths move in a quarterly series is close to noise. It can come from seasonal patterns in collections, from a change in charge-off timing, or from balances growing faster than the delinquent portion — which improves a ratio without a single household catching up.
That last mechanism deserves saying plainly, because it is the most common way a distress measure improves without distress falling. The figure is a share of balances. If the denominator grows and the numerator holds, the percentage falls. Nobody has paid anything off.
Whether that is what happened this quarter is not knowable from the headline number, which is exactly the problem.
The central bank is arguing with its own indicators
The strongest evidence that the headline is unreliable is that the New York Fed published work in August specifically to reconcile credit card delinquency measures that disagree.
Different series count different things. Some measure the share of balances delinquent, some the share of accounts, some the rate at which loans transition into delinquency. Balance-weighted measures are dominated by large balances, so a small number of heavily indebted households can move them a long way. Account-weighted measures treat a $600 balance and a $26,000 balance identically. Transition rates catch the flow of people newly in trouble and miss the stock of people who have been in trouble for a year.
When a central bank has to publish an explanation of why its own indicators point different directions, the sensible reading is not that one of them is wrong. It is that no single number answers the question people are asking, which is whether households are all right.
What is not ambiguous
Two things, and they run in opposite directions.
Mortgage balances fell by $74bn in the quarter, to $13.1 trillion. That is genuine deleveraging in the largest household liability, and it is not a denominator effect.
And 12.9 percent of card balances remain more than three months late, against 7.6 percent four years ago. Nothing in this quarter's improvement touches the size of that gap.
The picture consistent with both is a divided household sector rather than an improving or deteriorating one: owners with fixed low-rate mortgages paying down principal, alongside a cohort carrying revolving debt at rates that make the balance close to unrepayable. This desk has found the same split from several directions — in auto loans stretched to absorb payments that outran incomes and in the households paying down revolving debt at rates not seen in a decade. Both stories were true at once because they were about different people.
What to watch
Not next quarter's headline rate.
Watch the transition rate into 90-day delinquency, which measures the flow of newly distressed borrowers rather than the accumulated stock, and is the only one of these series that turns before the others. If transitions fall while the level stays high, the cohort in trouble is fixed and slowly clearing. If transitions rise while the level falls, the improvement is arithmetic and the next two quarters will take it back.
The decline of about 0.2 percentage points to 12.9 percent in credit card balances at least 90 days delinquent in the second quarter of 2026; the rise in the same measure from 7.6 percent in 2022 Q3 to 12.8 percent in 2026 Q1; aggregate household delinquency of about 4.7 percent of outstanding debt; the easing of serious auto loan delinquency to 5.5 percent from 5.6 percent; the decline of mortgage balances by $74bn in the second quarter to $13.1 trillion; and the existence of New York Fed research published in August 2026 reconciling diverging credit card delinquency measures are as published by the Federal Reserve Bank of New York in its Household Debt and Credit report and Liberty Street Economics during 2026. The interpretation of what a share-of-balances measure does and does not capture is our own.
Topics moneyconsumer credithousehold financedelinquencybanking
