Method unchanged since Nov 17, 2026 · what changed then · full methodology below
Credit Conditions describes where bank, mortgage and consumer credit stand: whether borrowing is unusually slow, and whether late payments are unusually high.
It asks one question of 6 figures on bank business loans, mortgages and consumer credit: Is borrowing unusually slow, and are late payments unusually high?
It describes where credit stands; it does not forecast. Credit prices and bank lending standards are read by Forward Stress, the site's early warning, on State of the Economy.
Not covered: nonbank and private credit, securitized lending, small-business credit and lending standards for households.
Each figure links to its page on this site, or to the Federal Reserve's own release where the site has none.
| reading | series · publisher | FRED id | cadence | first figure | archive of first-published figures from |
|---|---|---|---|---|---|
| Borrowing: bank business loans | Bank commercial and industrial loans, adjusted for the Fed's reclassifications (Federal Reserve H.8) | H8B1023NCBCMG | monthly | Jan 1948 | Aug 2019 |
| Late payments: bank business loans | Business loans 30+ days late or not accruing, all commercial banks (Federal Reserve) | DRBLACBS | quarterly | Q1 1987 | Q2 2011 |
| Borrowing: mortgages | Household mortgage debt, all lenders (Federal Reserve Z.1, Financial Accounts) | HHMSDODNS | quarterly | Q4 1946 | Q2 2010 |
| Late payments: mortgages | Single-family mortgages 30+ days late or not accruing, all commercial banks (Federal Reserve) | DRSFRMACBS | quarterly | Q1 1991 | Q2 2011 |
| Borrowing: consumer credit | Consumer credit outstanding, growth adjusted for breaks in the series (Federal Reserve G.19) | TOTALSLAR | monthly | Jan 1944 | May 2012 |
| Late payments: credit cards | Credit cards 30+ days late or not accruing, all commercial banks (Federal Reserve) | DRCCLACBS | quarterly | Q1 1991 | Q2 2011 |
Each reading's latest figure is ranked within that series' own last 10 years, using only the figures published by that date. The rank is its percentile rank: the share of that window's figures below it, a tie counting half.
Borrowing reads Slow at the 20th percentile or below, the lowest fifth. Late payments read High at the 80th percentile or above, the top fifth.
At the other end, borrowing at the 80th percentile or above reads Fast, and late payments at the 20th or below read Low. Any other reading counts as neither Slow nor High.
With fewer than 5 years of history a reading has no state and reads “too short to rank”.
Each card also says which way its figure moved over 12 months. It reads “steady” when the change is smaller than 0.25 times the standard deviation of 12-month changes over the last 10 years. The direction is shown beside the state; it never changes it.
Borrowing counts as unusually slow when at least 2 of its 3 readings are Slow. Late payments count as unusually high when at least 2 of their 3 readings are High.
Each question needs at least 2 readings with a state; otherwise the page reads “not enough data”.
The two answers together give one reading:
The reading is worked out with the lines at the 20th and 80th percentiles, and again with both lines moved 5 points each way (25th and 75th, then 15th and 85th).
When either moved setting gives a different reading, the month is “on the line”: a small change in one figure would move it. The headline then names both readings, and the page says which figures decide it.
When the other reading differs on both questions, the headline names the month's own reading and the two that are one question away. The far reading is named only as what both changes together would give.
Some readings have related measures, called checks, that only say which way they moved:
When a check moved the other way from its reading over the same 12 months, the card says “Mixed evidence” and names it. A steady check does not count. Mixed evidence never changes a state or a count.
Each figure is ranked as it was first published, from the archive of earlier releases where one exists. The table above gives the month each archive begins.
Before a reading's archive begins, the page uses today's revised figures, and their publication dates are estimates.
The Time Machine shows how the page would have read at the end of any closed month since February 1996, by today's rules. A month whose 6 figures are all as first published is a “Replay”.
A month that uses any of today's revised figures is a “Reconstruction”, says how many, and carries a small “reconstructed” tag beside its headline.
The record holds 370 months, from February 1996 to November 2026, each read by the same rules.
Since 1996: Healthy in 77% of months, Cooling 10%, Strained 5%, Squeezed 8%. 14% of all months were on the line.
The page counts no episode and makes no claim about what followed any reading.
Credit Conditions was introduced in November 2026, ahead of the end of its trial, by the owner's ruling. Its rules are frozen until the review on May 10, 2027.
The rules of the 6 readings have been fixed since Nov 9, 2026, and the rules of the headline since Nov 11, 2026. Any change to a rule would be a redesign, and the trial would start again.
A shadow record of the page has been kept daily since Nov 10, 2026: each reading's state, direction and figure, with the dates it covers and was published.
The review will read that record to see whether every release appears within one business day of publication, the two dates each figure carries are right and no state changes without a release that explains it.
It cannot show whether the page predicts anything, and it is not asked to.
The page confirms where tighter credit has shown up; Forward Stress, which reads credit prices and lending standards, is the site's early warning.
The late-payment figures are quarterly and arrive months after their quarter ends; the line under the two answers names the oldest figures on the page.
Each figure is ranked against its own last 10 years, so a long period of strain moves the lines a reading is judged by.
The New York Fed's figures on loans moving into serious delinquency are not used until permission to publish them is clear.