US labour market · Historical study
When did the Sahm Rule signal a US recession?
Last updated
The rule measures how far unemployment has risen from its recent low. We compare each new signal with US business-cycle dates to distinguish early warnings, readings during recessions and false alarms.
Most crossings occurred after the recession had begun
Of 13 new signals, 10 fell in months later classified as recessions. One, in October 1969, came three months before the next contraction. Two, in November 1976 and July 2024, occurred during expansions and had no recession onset in the following six months.
All 9 recessions had at least one crossing during the contraction. The median first such crossing came 3 observation months after the first contraction month. A crossing in a given month could only be observed once that month’s unemployment figures were released.
Only two new signals in the sample followed the rule’s publication in 2019. One coincided with the 2020 contraction. The other was the July 2024 false alarm under our six-month definition. That is a small record on which to judge future performance.
Every new threshold crossing
| First month | Entry value | Last consecutive month at or above 0.50 | Timing against NBER dates |
|---|---|---|---|
| 0.53 | Nov 1961 | During contraction that began May 1960 | |
| 0.50 | Oct 1969 | Three months before the January 1970 onset | |
| 0.53 | Oct 1971 | During contraction that began Jan 1970 | |
| 0.50 | Mar 1974 | During contraction that began Dec 1973 | |
| 0.57 | Jan 1976 | During contraction that began Dec 1973 | |
| 0.50 | Nov 1976 | Expansion, no onset within six months | |
| 0.67 | Apr 1981 | During contraction that began Feb 1980 | |
| 0.80 | Jun 1983 | During contraction that began Aug 1981 | |
| 0.50 | Nov 1992 | During contraction that began Aug 1990 | |
| 0.53 | Oct 2002 | During contraction that began Apr 2001 | |
| 0.50 | May 2010 | During contraction that began Jan 2008 | |
| 4.00 | Mar 2021 | During contraction that began Mar 2020 | |
| 0.53 | Sept 2024 | Expansion, no onset within six months |
Two crossings can refer to the same recession. The October 1969 and February 1970 episodes bracketed one onset. In 1974 the series dropped below the threshold between March and July, producing two entries within a single contraction.
How the episodes were counted
The published rule triggers when the three-month moving average of US unemployment rises at least 0.50 percentage points above its lowest three-month average over the preceding 12 months. We use Claudia Sahm’s real-time indicator series, SAHMREALTIME, which uses unemployment estimates available at the time. The separately published current-data version incorporates later revisions. FRED series definition.
Each episode begins when the published value moves from below 0.50 to at least 0.50. Further months above the threshold belong to that episode until the value falls below it. We test 780 monthly observations through the last full calendar year before a missing October 2025 reading breaks the series. The first available reading, December 1959, was already above the threshold. Its crossing date is unknown, so that run is excluded.
We use FRED’s USREC monthly labels derived from the National Bureau of Economic Research chronology. A contraction runs from the month after a business-cycle peak through the trough month. A new signal counts as “during” if its observation month is inside this interval. Otherwise it counts as “before” if the next contraction starts in the following six calendar months. All other crossings count as unmatched, our definition of a false alarm. December 2024 has a full six-month outcome window through June 2025. Monthly labels · NBER peak and trough dates.
The six-month horizon is an editorial convention for this comparison. Shortening it can change which warnings qualify. NBER dates are assigned retrospectively and are used only to evaluate outcomes. Neither the final chronology nor future readings enter the crossing decision. NBER dating procedure.
What a reader could have known at the time
Sahm published the rule in May 2019. The first FRED release archived in ALFRED dates to 6 September 2019. The earlier crossings are historical reconstructions. An economist in 1960 could observe rising unemployment, but could not yet consult this published rule. Sahm’s original proposal · ALFRED release history.
We checked the archived readings for both crossings after publication. April 2020’s reading of 4.00 was available on 8 May, after what NBER later dated as the April trough. July 2024’s 0.53 was available on 2 August, up from 0.43 in June. The 2020 example shows why aligning observation months can make a signal look more timely than it was for a decision-maker.
| Archive date | Latest observation | Prior month | Latest value |
|---|---|---|---|
| 6 September 2019 | Aug 2019 | 0.03 | 0.07 |
| 8 May 2020 | Apr 2020 | 0.30 | 4.00 |
| 2 August 2024 | Jul 2024 | 0.43 | 0.53 |
This audit describes one unemployment indicator in one country. Repeat crossings and a rule developed from historical experience limit what can be inferred from the counts. They do not estimate the probability of a recession over the next six or twelve months. For the Observatory’s separate multi-indicator assessment, see the US economic evidence.
Data, replication and citation
The calculation and interpretation are automated research by Recession Indicator. Corrections can be submitted through our corrections process. Inputs are Claudia Sahm’s Real-time Sahm Rule Recession Indicator, the NBER-based USREC series distributed by the Federal Reserve Bank of St. Louis through FRED, and three dated ALFRED archives. The results file contains derived episodes and source checks.
Download the calculation (.mjs) · Download edition 1.0 results (.json)
Reproduce this edition
With Node.js 22 or newer and curl installed, run node sahm-audit.mjs --download > results.json. Compare the result with edition 1.0. The SHA-256 digest covers the normalised monthly inputs in the study window, including its boundary and follow-up months. A changed digest warrants checking the source history before comparing results.
05ab3ea665f54a47f04d388e58edf801441791610f1738156a8650365850c37a
Recession Indicator (2026). “When did the Sahm Rule signal a US recession?” Edition 1.0, 6 November. Permanent study page.