The 3-month Treasury yield against the 10-year — tested against every recession in its record, and the one cross-asset relationship on this site whose record held.
3-month to 10-year: +0.94 · not inverted
The 3-month to 10-year curve last inverted in June 2025 – July 2025. No recession declared yet — scored in December 2027.
Percentage points, the 10-year yield minus the 3-month, at the close on Sep 24, 2026. The record reads each month's last close.
The 2-year to 10-year curve, the one most often quoted, reads +0.31 (not inverted) — see its page →
Since 1982, 7 of the 9 inversions whose outcome is known were followed by a recession within 18 months. From an ordinary month, a recession followed within 18 months only 17% of the time.
The two that were not: a single inverted month in February 2006 and the 2022–24 inversion, which lasted 26 months. The leads ran from 1 to 17 months: 14, 9, 8, 17, 5, 9 and 1.
It says a recession has usually followed. It has never said when.
Orange: inversions. Grey: NBER recessions. The dashed line is zero. The full record, from 1982, is in the expanded view.
| began | lasted | what followed, and when |
|---|---|---|
| Jun 2025 | 2 months | No recession declared yet — scored in Dec 2027 |
| Feb 2025 | 3 months | No recession declared yet — scored in Aug 2027 |
| Oct 2022 | 26 months | No recession within 18 months |
| Jan 2020 | 2 months | A recession: the NBER peak came one month later (Feb 2020) |
| May 2019 | 5 months | A recession: the NBER peak came 9 months later (Feb 2020) |
| Jul 2007 | one month | A recession: the NBER peak came 5 months later (Dec 2007) |
| Jul 2006 | 10 months | A recession: the NBER peak came 17 months later (Dec 2007) |
| Feb 2006 | one month | No recession within 18 months |
| Jul 2000 | 6 months | A recession: the NBER peak came 8 months later (Mar 2001) |
| Oct 1989 | one month | A recession: the NBER peak came 9 months later (Jul 1990) |
| May 1989 | 3 months | A recession: the NBER peak came 14 months later (Jul 1990) |
The 2-year to 10-year curve, tested the same way, also held: 8 of its 11 inversions since 1976 were followed by a recession within 18 months, against 21% of all months, and it did not invert before the recession that began in Feb 2020 (its series).
The 3-month to 10-year curve came before this site's own stress signal on 2 of its 7 hits, and before the Slowdown stage on 3.
Seven other relationships were tested and did not hold — the quality spread, gold with the dollar, copper/gold, gold/oil, gold/silver, stocks/gold and lumber/gold — and their records are in the methodology.
The curve is the 10-year Treasury yield minus the 3-month, read at each month's last close. A month below zero is an inverted month; consecutive inverted months are one inversion. Each inversion is scored on one question: did an NBER business-cycle peak follow within 18 months? Months already inside a recession are not scored.
The test was written before it was run. An inversion record passes if it has at least 5 scored inversions from 4 separate episodes, a hit rate at least 1.5 times the rate for all months, no more false alarms than hits, the same with any one episode removed, and more hits than 500 random placements of the same number of inversions. The 3-month to 10-year curve passes all five.
Treasury yields are market prices, recorded as they trade and never revised, so this record is what a reader could have seen at each month's end. The NBER's dates are set after the fact; they are used only to score what followed.
The full method, the 2-year curve's record and the tested relationships →
A historical record, not a forecast. Methodology →