The Federal Reserve publishes four credit card delinquency rates. Only one of them is still going up

Two numbers describe American credit card delinquency this year. One is 13.12 percent. The other is 2.92 percent.

Both come from the Federal Reserve. Both are current. Both are right.

The first is in every headline written about household stress since May, and in most of the social posts underneath them, and in Google's own answer at the top of the results page. The second is the one that bank supervisors and bank investors actually watch. Asked on 6 August, Google's AI Overview gave both, placed them in adjacent sentences, and explained the gap as "higher stress among lower-income groups and younger borrowers".

That is not the reason. The two numbers count different things, and the difference between them is written down, in a supervisory policy issued in 1999 and in a data dictionary on page 43 of the New York Fed's own report.

What is the credit card delinquency rate right now?

There are four published answers and they are all official. Balances 90 or more days delinquent are 13.12 percent of card debt. Balances at least 30 days past due, excluding those already charged off, are 4.03 percent. On bank balance sheets the rate is 2.92 percent. Card loans are being written off at an annual 3.84 percent.

The first two are built from the same dataset. That is the part worth sitting with. The New York Fed's Consumer Credit Panel, a 5 percent sample of Equifax credit records, produces both 13.12 percent and 4.03 percent, and the entire distance between them is a definitional choice about what to include.

This shape recurs. The IRS audit rate is a partial count frozen on a fixed date and rises about 45 percent after it is first published. The going-concern warning rate fell for sixteen years mostly because its denominator shrank. On credit cards the moving part is neither the date nor the denominator. It is the definition.

Why do two Federal Reserve measures of the same thing differ by a factor of four?

Because one of them keeps counting a balance after the lender has given up on it. The New York Fed's 90-plus measure includes a category called severely derogatory, which its data dictionary defines as any earlier delinquency state combined with a reported repossession, charge off to bad debt, or foreclosure. The Board's staff measure removes that category deliberately.

The definitions are published and they are short. The New York Fed's report says that "90+ days late" means the share of balance "that is either 90-day late, 120-day late or severely derogatory", and that severely derogatory means "any of the previous states combined with reports of a repossession, charge off to bad debt or foreclosure".

The Board's economists, using the same panel, state their choice in the note under figure 1 of their November 2025 paper: the rate "measures the fraction of balances at least 30 days past due, excluding severely derogatory balances." They exclude the charged-off bucket while using a wider window, 30 days rather than 90, and still land at 4.03 percent against 13.12.

You can watch the effect in the New York Fed's own table. For all household debt at the end of March, 0.31 percent of balances were 90 to 119 days late, 1.30 percent were 120 or more days late, and 1.75 percent were severely derogatory. Those three add to 3.37, which is the published all-loans 90-plus figure of 3.36 to within a rounding step. Nine percent of the total is the fresh part. The rest has been sitting there.

The card-only version of that decomposition is not published, and this piece does not invent one. What is published is the pattern across dates, and it is stable: at the end of 2019, 90.8 percent of the 90-plus balance was 120 days late or worse; at the end of 2023, 88.4 percent. A measure whose contents are nine-tenths old is not measuring this quarter.

How long does a delinquent balance stay in each measure?

On a bank balance sheet it cannot stay more than six months. Federal supervisory policy requires an open-end retail loan to be classified Loss and charged off once it is 180 cumulative days past due, at which point it leaves the numerator and the denominator together. In the credit bureau data, nothing forces it out at all.

The rule is the Uniform Retail Credit Classification and Account Management Policy, issued by the banking agencies in February 1999 and revised at 65 FR 36903 in June 2000. Its operative sentence: "Closed-end retail loans that become past due 120 cumulative days and open-end retail loans that become past due 180 cumulative days from the contractual due date should be classified Loss and charged off."

A credit card is open-end credit. So the bank measure is self-emptying by construction, and the charge-off rate in the table above is the drain.

The arithmetic agrees. The New York Fed reports the flow of balances into serious delinquency at an annualised 7.10 percent. Set that against a standing stock of 13.12 percent and the average balance sits in the 90-plus category for roughly two years. Run the same calculation on the bank series, using the 8.61 percent annualised flow into 30-day delinquency against a 2.92 percent stock, and the implied residence is about four months, comfortably inside the six-month ceiling the policy imposes.

Neither figure is an estimate of anything. Both assume a steady state that neither series is in, and the second borrows a credit bureau flow to explain a bank balance sheet stock, which is a rough thing to do. They are magnitude checks, and what they establish is only this: the two headline numbers are separated by a difference in how long a bad balance is allowed to be counted, and that difference is on the order of years against months.

There is a second, smaller drain on the bank number that nobody mentions. The same policy lets an institution re-age a delinquent open-end account back to current status, once in twelve months and twice in five years, if the account has existed at least nine months and the borrower "has made at least three consecutive minimum monthly payments or the equivalent cumulative amount". The policy adds that funds may not be advanced by the institution for that purpose. Three payments can retire an arrears position from the delinquency statistics without the arrears being paid.

Are credit card delinquencies still rising in 2026?

By the measure in the headlines, yes. By every measure of what card borrowers are doing now, no, and they have not been since 2024. The bank delinquency rate peaked in the second quarter of 2024, bank charge-offs in the third, the Board staff series in the third, and the flow of balances into new delinquency in the second. All four are lower today.

Here is the turn, quarter by quarter.

The first five rows give the 2024 peak. The last gives the same quarter for comparison, because that series has not peaked. Five series down, one up. The New York Fed's own report describes the first quarter in flat terms: transition into early delinquency "ticked down for credit cards, from 8.7% annually to 8.6%", and 4.8 percent of all outstanding household debt was in some stage of delinquency, roughly unchanged from the quarter before. The Board staff series has been falling year over year since the first quarter of 2025.

None of this means the level is comfortable. The 13.12 percent reading ties the first quarter of 2011 and is the highest since, and in a series that starts in 2003 the only readings above it are the four quarters of 2010. Bank charge-offs at 3.84 percent are well above the 1.63 percent trough of late 2021. What the fresh-flow series say is that the machine feeding that stock slowed two years ago and the stock is still digesting what it was fed.

Small banks are the exception worth naming. At institutions outside the 100 largest, the card delinquency rate is 6.43 percent against 2.80 percent at the top 100, and the charge-off rate is 8.10 percent against 3.67. Those banks have also improved, from a delinquency peak of 7.86 percent at the end of 2023, but they are running at more than twice the rate of the large issuers.

Who is actually falling behind?

Not a cross-section of the country. Using the Board staff measure at the third quarter of 2025, the card delinquency rate was 16.28 percent for subprime borrowers, 5.67 percent for near prime and 1.28 percent for prime. It was 6.55 percent in low-income census tracts against 2.87 percent in high-income ones, and 5.10 percent for borrowers without a mortgage against 2.99 percent for those with one.

Those spreads are the substance behind the phrase Google's summary reaches for. The stress is real and it is concentrated. What the summary gets wrong is the arithmetic: the distance between 13.12 and 2.92 is not the distance between poor borrowers and rich ones, because both numbers already average across everybody. It is the distance between two definitions.

The credit score distribution is drifting with it. Subprime borrowers, scored below 620, were 20.01 percent of the population with a credit report at the third quarter of 2025.

What do Americans actually pay on a credit card now?

The Federal Reserve surveys commercial bank card rates quarterly. In the second quarter of 2026 the average rate across all accounts was 20.94 percent, and across accounts assessed interest it was 22.15 percent. Both are down slightly from their 2024 peaks of 21.76 and 23.37 percent. Both are roughly seven points above where they sat before 2020.

The G.19 terms of credit series averaged 13.32 percent on all accounts across 2015 to 2019. The bank prime rate on 4 August 2026 was 6.75 percent, per the Board's H.15 release, which puts the average card rate 14.19 points above prime.

Revolving consumer credit outstanding stood at $1,296.9bn in May 2026, not seasonally adjusted. That is 45.0 percent above the April 2021 trough of $894.6bn and 18.8 percent above the December 2019 level. The New York Fed's separate count of credit card balances, measured on credit reports rather than on lender books, was $1.25tn at the end of March, after a seasonal fall of $25bn. Two institutions, two universes, two numbers for the same debt, which by this point in the piece should be the expected result rather than a surprise.

Where this reading could be wrong

The stock-versus-flow argument rests on the assumption that the severely derogatory bucket accumulates rather than clears, and the direct evidence for that is published for total household debt, not for cards alone. If card issuers purge derogatory reporting faster than other lenders do, the effect is smaller than the total-debt decomposition suggests.

The New York Fed says as much about its own data, in a caution that belongs in any piece using it: "Not all creditors provide updated information on payment status, especially after accounts have been derogatory for a longer period of time. Thus the payment performance profiles obtained from our data may to some extent reflect reporting practices of creditors."

There is also a live distortion in the total-debt figures. Student loan delinquency reporting resumed in 2025, and the student loan 90-plus share jumps from 0.53 percent to 7.74 percent in a single quarter, which lifts the all-loans series without anything having changed for a card borrower. The card series is unaffected, and the composition figures above are given at 2019 and 2023 as well as 2026 for that reason.

One more caveat, on the source everyone quotes. The chart page ranking fourth on this search publishes the New York Fed series accurately, value for value, under the title "US Credit Card Accounts Delinquent by 90 or More Days". The New York Fed publishes it as a percentage of balance. A percentage of accounts and a percentage of balance are different statistics, and on a product where the delinquent balances are systematically larger than the average balance, they are not close. That single word is why the figure travels through social media as a statement about how many people are three months behind, which is not what it says.

The next reading arrives on 11 August, when the New York Fed publishes the second quarter. The bank figures for the same quarter follow later in the month.

Frequently asked questions

What is the current credit card delinquency rate? It depends on which measure you want. Balances 90 or more days delinquent were 13.12 percent in the first quarter of 2026. The delinquency rate on credit card loans held by commercial banks was 2.92 percent in the same quarter. Balances at least 30 days past due, excluding charged-off balances, were 4.03 percent in the third quarter of 2025.

Why is the credit card delinquency rate 13 percent in the news and 3 percent at the Federal Reserve? The 13 percent figure counts balances that already carry a reported charge off, repossession or foreclosure, and nothing removes them. The bank figure counts loans on bank balance sheets, and supervisory policy forces a card loan off the books once it is 180 days past due, so no balance can sit in it longer than six months.

Are credit card delinquencies at a 15-year high? The 90-plus balance share is. At 13.12 percent it ties the first quarter of 2011, and the only higher readings in a series that begins in 2003 are the four quarters of 2010. Every measure of current payment behaviour peaked in 2024 and has fallen since.

Is 13.12 percent the share of people behind on their cards? No. It is a share of dollars, not of accounts or of borrowers. The New York Fed publishes it as a percentage of balance, and delinquent balances tend to be larger than average balances.

What is the average credit card interest rate now? 20.94 percent across all commercial bank card accounts in the second quarter of 2026, and 22.15 percent across accounts assessed interest, per the Federal Reserve's quarterly survey. The peak was 21.76 percent in the third quarter of 2024.

Where does the data come from? The bank figures come from Call Reports, published in the Board's Charge-Off and Delinquency Rates release. The balance-share figures come from the New York Fed Consumer Credit Panel, a 5 percent sample of Equifax credit records. The two tables compiled for this piece are linked above.