The Sahm Rule: Real-Time Recession Indicator Explained

The Sahm rule is the cleanest real-time recession indicator in the modern US business-cycle toolkit. It was developed by economist Claudia Sahm in 2019 while she was at the Federal Reserve, and is now published monthly by the Federal Reserve Bank of St Louis as FRED series SAHMCURRENT. The rule triggers a recession signal when the three-month moving average of the US unemployment rate rises 0.50 percentage points or more above the lowest three-month average from the prior twelve months. As of the September 2026 print, the indicator stands at 0.00, far below the 0.50 trigger threshold, having receded from its July 2024 crossing as the unemployment rate came down from 4.5 percent. On the current series it has crossed 0.50 within a few months of the start of every NBER-dated US recession since 1970, and twice (2003 and 2024) without one.

The rule's value comes from its real-time character. NBER dates recessions retrospectively, typically 6 to 18 months after the cyclical peak, once data revisions have settled and the broader six-indicator framework can be applied with confidence. The retrospective character of NBER dating is appropriate for the methodological care it represents, but it is unhelpful for policy responses that need to be triggered when a recession is starting, not confirmed a year later. The Sahm rule fills that gap. Its design intent (articulated in Sahm's 2019 Hamilton Project paper) is to trigger automatic fiscal stabilisers (stimulus payments, extended unemployment insurance, fiscal transfers) as soon as labour-market deterioration begins, rather than after a confirmation lag of many months.

How the Calculation Works

The Sahm rule calculation has three steps, all using publicly available BLS unemployment data. First, take the three-month moving average of the headline US unemployment rate (the U-3 measure published monthly by the Bureau of Labor Statistics in the Current Population Survey). Second, find the lowest three-month average from any point in the prior twelve months, which functions as a rolling reference floor capturing the most recent cyclical low. Third, subtract the second number from the first. If the result is 0.50 percentage points or more, the rule has triggered.

A worked example using the September 2026 reading. The three-month moving average for July to September 2026 unemployment is approximately 4.13 percent (the monthly rates were 4.1, 4.1 and 4.2). The lowest three-month average from the prior twelve months is the June to August average, also approximately 4.13 percent (4.2, 4.1 and 4.1). The difference is 0.00 percentage points, the current reading. A zero or negative value means the three-month average is at or below its recent low rather than rising from it; it is the opposite of a recession signal. For the indicator to approach 0.50, the three-month average would have to rise by more than half a point above that floor, which would require a renewed and sustained increase in unemployment.

The three-month moving average is used rather than a single month's reading because monthly unemployment data is noisy. Seasonal anomalies, weather events, single-firm layoff announcements, and other transient factors can produce significant month-over-month moves that do not represent a real change in labour-market conditions. The three-month smoothing reduces noise while preserving the timing sensitivity that real-time application requires. The rolling-12-month-low reference is used rather than a fixed historical floor because the rule needs to be applicable across different cyclical regimes; the floor adapts to whatever the recent labour-market low has been, so the rule operates symmetrically across business cycles.

The Historical Track Record

On FRED's current-vintage series (SAHMCURRENT), the rule crossed 0.50 shortly after the start of every NBER-dated US recession since 1970, never before it. The first crossing in each case: March 1970 (the 1969-70 recession that NBER dates from December 1969, three months earlier); July 1974 (the 1973-75 recession from November 1973, eight months earlier); February 1980 (the 1980 recession from January 1980); November 1981 (the 1981-82 recession from July 1981); October 1990 (the 1990-91 recession from July 1990); June 2001 (the 2001 recession from March 2001); February 2008 (the Great Recession from December 2007); and April 2020 (the COVID recession from February 2020). That lag of one to eight months is still far shorter than NBER's own announcement lag.

The rule is not free of false positives. The same series crossed 0.50 in July 2003, during the jobless recovery after the 2001 recession, and again in July 2024, when it reached 0.57 on the real-time series; no recession followed either crossing. A lower threshold would fire more often: at 0.30 the current series would also have flashed in December 2023 and November 2025. A higher one would catch recessions later than the one-to-eight-month lag above.

The 2020 COVID experience was a useful real-time test of the rule. The rule had not yet been formally adopted in policy circles when COVID hit, but applied to the COVID data, it crossed 0.50 in April 2020, two months after the February 2020 recession peak NBER later dated. The COVID experience demonstrated that the rule could function during an unprecedented public-health shock as well as during conventional business-cycle downturns.

Why 0.50 Specifically

The 0.50 percentage-point threshold is empirical rather than theoretical. Sahm set it in her 2019 Hamilton Project proposal, where the rule would switch on stimulus payments automatically, so the aim was to fire early in a recession while rarely firing outside one. Lower thresholds catch recessions earlier but fire on labour-market noise: at 0.30 the current FRED series would also have flashed in December 2023 and November 2025. Higher thresholds cut false signals but trigger later in real recessions.

The intuition behind it is that unemployment tends to drift down slowly and rise quickly, so a half-point rise in the smoothed rate has usually meant the rise was gathering pace. 2024 is the reminder that this is a regularity, not a law: the rate rose enough to cross 0.50 and then stopped.

The Live 2026 Reading

The September 2026 Sahm rule reading of 0.00 sits far clear of the trigger threshold. The indicator crossed 0.50 in July 2024 as unemployment climbed off its April 2023 low of 3.4 percent, but it has since receded as the headline unemployment rate peaked at 4.5 percent and came down to 4.1 percent in July and August 2026, before ticking up to 4.2 percent in September. Because the rule measures the rise in the three-month average above its rolling 12-month low, a falling unemployment rate pulls the reading down as the reference floor catches up; it dipped to -0.07 in August and is now exactly at zero.

Whether the indicator climbs back toward 0.50 depends on whether the unemployment rate resumes rising. With unemployment having levelled off and then eased, the rolling 12-month-low reference floor has caught up to current levels, which is precisely why the reading is at 0.00. A renewed and sustained increase in unemployment of more than half a percentage point above the recent floor would be required to approach the trigger; a stable-to-easing jobless rate keeps the reading low. The reading is worth watching in context: hiring is slow, with payrolls down 10,000 in July, up 133,000 in August and up only 29,000 in September on the latest revisions (FRED PAYEMS), so the low Sahm reading reflects few layoffs rather than strong job creation.

Claudia Sahm herself has expressed caution about applying her rule mechanically to the current period. In commentary through 2024-25, she has noted that post-COVID labour-market dynamics differ from earlier business cycles: the labour-force participation rate has been recovering rather than declining, immigration-driven labour-force expansion is contributing to measured unemployment in ways that do not reflect underlying labour-market weakness, and the relationship between unemployment and broader macroeconomic conditions may be different than in earlier cycles. Her caveat is that the rule's historical 0.50 crossing identified recessions reliably, but the post-COVID 0.50 crossing may not have the same meaning.

The Policy Significance

The Sahm rule's design intent was to trigger automatic fiscal stabilisers as soon as labour-market deterioration begins. The rule has been proposed (by Sahm and others, including in the Federal Reserve's policy-tools research and in academic discussions of fiscal-policy reform) as a trigger for stimulus payments, extended unemployment insurance, or fiscal-transfer programmes. The argument is that the political process of legislating recession-fighting fiscal stimulus is slow (months to years) and tends to operate too late in the business cycle to be maximally effective. A rule-triggered automatic response could deliver fiscal support during the recession itself, not after it has ended.

The 2020 CARES Act stimulus payments were not formally triggered by the Sahm rule (Congress passed them in response to the COVID shock directly), but the speed of the policy response (the first round of payments arrived within weeks of the unemployment spike) functionally resembled what a Sahm-rule-triggered automatic stabiliser would have produced. Subsequent academic fiscal-policy discussion has referenced Sahm rule mechanics as a model for recession-trigger design.

Whether the rule is ever formally adopted as an automatic fiscal-stabiliser trigger remains a political question. The technical case for it is strong; the political case requires accepting an automatic spending mechanism that bypasses congressional appropriations, which is constitutionally and politically challenging. As of 2026, the rule remains an informal real-time recession indicator rather than a formal policy trigger.

For the broader indicator framework, see the indicator dashboard. For other forward-looking indicators, see the yield curve indicator and the Conference Board LEI. For the 2020 recession application, see the COVID recession 2020.