The headline number that has been feeding the "consumers are cracking" narrative is a measurement artifact, not a distress signal. Credit card balances 90-plus days delinquent climbed from 7.6 percent in the third quarter of 2022 to 12.8 percent in the first quarter of 2026, a level not seen since the Great Recession. Read literally, that is the kind of figure that reprices consumer-credit risk. Read carefully, it is telling you almost nothing new about how many households are falling behind right now.
The tell is that the flow rate, the share of previously current balances newly transitioning into serious delinquency, has been flat for nearly two years. When the stock of bad debt rises while the flow into bad debt holds steady, the stock is not being fed by fresh distress. It is being inflated by debts that never leave.
Two Delinquency Measures, One Divergence
Three series describe the same borrower and disagree about her health. The New York Fed's stock measure counts the share of all outstanding reported balance that is 90-plus days past due. The flow measure counts new entrants into that bucket each quarter, annualized. And the Board of Governors' Call Report series counts the share of balances on lenders' books that are 30-plus days past due.
Two of those three, the Call Report line and the flow line, have leveled off since roughly 2024. The stock line has kept climbing. That is the entire puzzle, and the resolution is mechanical rather than behavioral.
The mechanism sits in what happens after a loan goes very bad. A lender charges off a delinquent card balance, typically somewhere between 120 and 180 days past due. On the lender's balance sheet, the loan then vanishes from both the numerator and the denominator of any delinquency ratio. It registers as a charge-off once, in that quarter, and disappears. This is why the Call Report measure can look tame: it mechanically expels its worst loans.
The borrower, however, still owes the money. Lenders and their collection agents keep pursuing these debts and keep reporting them to the credit bureaus. The New York Fed's panel, built from those bureau records, keeps counting them. So a charged-off card balance from 2023 that a household has not resolved is still sitting in the stock numerator in 2026, quarter after quarter, stale but never expelled.
Why the Scary Line Is the Slow Line
Run the arithmetic forward and the divergence explains itself. If new delinquencies are entering at a stable rate but old delinquencies are accumulating faster than they are being cured, charged off the bureau files, or written down, the pool of reported 90-plus balances swells even with no deterioration in fresh borrower behavior. The stock rate is a reservoir with a steady inflow and a slow drain. The flow rate is the inflow alone.
This also resolves the counterintuitive fact that the stock measure sits above the Call Report measure despite the Call Report capturing earlier-stage 30-to-89-day delinquencies the stock measure misses. It should not, if both were measuring current distress the same way. It does, because the two are measuring different things: one tracks live loans on a lender's book, the other tracks the durable shadow of debts that have already left the book but not the borrower.
The New York Fed's own reconciliation lands there. The rising stock rate reflects a pool of stale, charged-off debts being reported for longer durations, not a worsening in the incidence of delinquency.
The Broader Ledger Is Quiet
The rest of the second-quarter 2026 household-debt picture argues the same way. Total balances actually fell by $13 billion. Mortgage and student-loan balances declined slightly; other products rose modestly; delinquency rates across most categories stayed roughly stable. This is not the balance sheet of a consumer sector rolling over. It is a consumer sector treading water with a legacy pile of unresolved card debt that keeps showing up in one particular statistic.
The macro backdrop is not screaming stress either. The 10-year Treasury sits at 4.72 percent and the 2-year at 4.25 percent, leaving the 2s10s spread positive at roughly 0.47 percent as of early August 2026. A dis-inverted, upward-sloping curve with the front end still above 4 percent is a restrictive-but-orderly funding environment, not one pricing an imminent consumer credit event. If the market believed the 12.8 percent figure was a real-time distress reading, the front end and consumer-lender spreads would be doing more work than they are.
What Would Break This Reading
The benign interpretation rests on one load-bearing assumption: that the flow rate stays flat. That is exactly where the argument is vulnerable, and it deserves to be stated plainly rather than assumed away.
If new transitions into 90-plus delinquency start rising, the stock-versus-flow reconciliation stops being reassuring, because then both the reservoir and the inflow are growing. A stale-debt overhang is a one-time repricing of a measurement; a rising flow rate is the leading edge of genuine deterioration. The two look similar in the stock line for a quarter or two, then diverge sharply.
The second vulnerability is duration. The stale-debt explanation depends on lenders reporting charged-off balances for longer than they used to. That behavior can reverse. If reporting durations normalize or a wave of these old debts is finally written off the bureau files, the stock rate could fall for reasons that have nothing to do with improving household finances. The number would look better while nothing real changed, the mirror image of today's problem.
The Line That Actually Matters
For anyone pricing consumer-credit risk, the instruction from this data is to stop watching the 12.8 percent headline and start watching the flow rate and the Call Report line, which have leveled off. Those are the near-real-time reads on whether households are falling behind. The stock rate is a lagging accumulation that will keep drifting higher on its own inertia regardless of what borrowers do next quarter.
The thesis breaks the moment new transitions into serious delinquency turn up in the flow series. Until that happens, the widely quoted "worst since the Great Recession" framing is describing the depth of an old reservoir, not the speed of a new leak, and the credit signal worth trading is the quiet one, not the loud one.
