The Credit Card Delinquency Number Everyone Quotes Is Measuring Old Debt, Not New Distress
There is a statistic circulating this month that sounds like the leading edge of a consumer credit crisis: roughly one in eight dollars of outstanding US credit card balances is at least 90 days past due, a share not seen since the aftermath of the financial crisis. On August 11, the Federal Reserve Bank of New York published two documents at once. One was its Quarterly Report on Household Debt and Credit for the second quarter of 2026. The other was a research post explaining why that headline share has become a poor guide to how many households are actually falling behind.
The post, "How Distressed Are Consumers? Reconciling Diverging Credit Card Delinquency Measures," appeared on the New York Fed's Liberty Street Economics blog and is credited to Donghoon Lee, Daniel Mangrum, Joelle W. Scally, Tejas Sinha and Wilbert van der Klaauw. Its subject is not the level of consumer distress but the construction of the measures used to describe it, and it is worth reading closely before quoting any of them.
Two different things get called the credit card delinquency rate. The first is a stock measure: the share of outstanding balance actively reported on credit reports that is 90 or more days past due at a given moment. The second is a flow measure, which the authors define as the balance on loans that became 90+ days past due in the present quarter divided by the balance of loans that were less than 90 days past due in the previous quarter, annualized using a four-quarter moving sum. The first counts how much bad debt is sitting there. The second counts how much new bad debt arrived. They answer different questions.
The stock measure has climbed steeply. According to the Liberty Street post, it rose from 7.6% of outstanding balance in the third quarter of 2022 to 12.8% in the first quarter of 2026. That 12.8% figure belongs to the first quarter, not the second; the post draws it from the Quarterly Report's own credit card series rather than from the second-quarter figures released the same day. It is the number behind the Great Recession comparisons.
The flow measure tells a different story. In the New York Fed's August 11 press release, the transition rate into 90+ days delinquent for credit cards was 6.97% in the second quarter of 2026, against 6.93% a year earlier. That is essentially unchanged. The release's own headline says as much: household debt balances decreased slightly, and credit card delinquency transition rates remained steady. The rate at which American cardholders are newly falling seriously behind has been broadly flat for close to two years, even as the stock share rose by roughly two-thirds.
The reconciliation the researchers offer is a change in reporting behavior. In their words, between 2004 and 2012 only about 40 percent of borrowers' charged-off debts were still being reported one year later; by 2024, this figure had doubled to 80 percent. A charged-off balance is debt the lender has already written off as uncollectible. It is not new distress. It is old distress that used to disappear from credit reports and now stays visible, accumulating in the denominator and the numerator of the stock measure quarter after quarter.
The authors tested the explanation by removing charged-off balances from the calculation. When they do, they write, the stock delinquency rate falls in line with both their flow delinquency rate and the Call Report delinquency rate, the latter being the Federal Reserve Board's series drawn from lenders' own balance sheets and measured at 30 or more days past due rather than 90. The post is clear that it was the stock series alone that had drifted: the Call Report and flow series were already very similar to each other, both having levelled off, while the stock line had been rising steadily since about 2023. It falls back among them once the accounting artifact is stripped out. Their summary of what the corrected picture shows is that the pace of credit card delinquency is elevated but has been largely stable since 2024.
The distinction has not been universally lost. Marketplace covered the release on August 11 under the headline "Credit card delinquencies approach Great Recession levels," and its story cited the roughly 13% stock figure while also noting that new delinquencies have largely stabilized after a spike a couple of years ago and that many borrowers are still carrying old credit card debt they fell behind on. That story linked to the Liberty Street post directly. The framing and the correction are travelling together, which is more than can usually be said for a statistical caveat.
None of this is an all-clear, and the New York Fed did not present it as one. Joelle Scally, an Economic Policy Advisor at the New York Fed, is quoted in the August 11 release saying: "Delinquency rates across most products have held steady over the past two years. Still, new delinquencies for auto loans and credit cards remain at elevated levels, a trend we'll continue to monitor." Both halves of that sentence are load-bearing. Steady is not the same as low, and the flow rate that the reconciliation vindicates is itself sitting at an elevated level rather than a benign one.
The same artifact runs in the opposite direction elsewhere in the data, and this is where the reporting-versus-reality problem becomes most vivid. The flow into serious delinquency for student loans fell to 7.83% in the second quarter of 2026 from 12.88% a year earlier, per the press release. That looks like a dramatic improvement. But the quarterly report states that the student loan delinquency rate increased to 10.6% of balances 90+ days past due, up from the 10.3% observed in the first quarter of 2026. The stock rose while the flow collapsed. The release itself singles out student loans as the exception to its steady-delinquency story, attributing it to the continued impact of the re-reporting of defaulted student debt causing some distortions. That is a reporting effect, not borrowers catching up on payments.
The quarterly report itself is the quieter document. Total household debt fell by $13 billion, or 0.1%, to $18.8 trillion in the second quarter, the figure given to one decimal place in both the release and the report. Mortgage balances declined $74 billion to $13.117 trillion and student loans fell $7 billion to $1.651 trillion, while home equity lines of credit rose $13 billion to $459 billion, auto loans rose $28 billion to $1.713 trillion and credit cards rose $21 billion to $1.263 trillion. Aggregate delinquency across all stages was 4.7% of outstanding debt at the end of June, down 0.1 percentage point from the prior quarter.
Several outlets described the quarterly drop as the first in six years, including The Epoch Times, whose August 11 headline read "US Household Debt Posts 1st Quarterly Decline in 6 Years: New York Fed." That characterization does not appear in the New York Fed's own report, which says only that aggregate nominal household debt balances declined in the second quarter of 2026, by $13 billion. But the arithmetic behind it is sound rather than merely journalistic: the Bank's August 6, 2020 release, headlined "Total Household Debt Decreased in Q2 2020, Marking First Decline Since 2014," recorded a decline of $34 billion, or 0.2%, to $14.27 trillion in the second quarter of 2020, and that is the last quarter before this one in which balances fell. The six-year framing is a media calculation, but one the Bank's own record supports.
Underneath the totals, the flows are mixed rather than uniformly improving. The overall transition into 90+ days delinquent across all debt categories was 2.57% in the second quarter against 2.91% a year earlier, but that aggregate is flattered by the student loan swing described above. Mortgage transitions moved the other way, to 1.52% from 1.29% a year earlier, and auto transitions ticked up to 3.00% from 2.93%. Lenders kept extending: $505 billion of mortgages and $211 billion of auto loans were originated in the quarter, and aggregate credit card limits rose $85 billion. Roughly 55,000 individuals had new foreclosures added to their credit reports and about 137,000 consumers had a bankruptcy notation added.
One further caution applies to anyone comparing borrower credit quality across this data. A footnote on the first page of the second quarter report states that beginning in 2026Q1, the credit score on pages 6, 7, 8 and 9 of the Quarterly Report on Household Debt and Credit is the VantageScore 4.0, and that previous reports use Equifax Risk Score 3.0. That is a break in the series, not a change in borrower behavior, and origination score comparisons that straddle it are not like-for-like. It is the same lesson as the charge-off finding, arriving from a different direction: before a number is used to describe the economy, it is worth establishing what the number is counting.
