The Sahm rule: a recession signal from the unemployment rate

The Sahm rule says a recession has probably begun when the three-month average of the national unemployment rate rises by 0.50 percentage point or more above its low of the previous 12 months. It is named after the economist Claudia Sahm, who proposed it in a May 2019 paper for The Hamilton Project, written while she was a section chief at the Federal Reserve Board.

The rule became famous for a simple reason: in the US data since 1960 it caught every recession within a few months of its start, using nothing but the The unemployment rate: U-3, U-6 and labor force participation. It became more famous in August 2024, when it was triggered and no recession followed, at least as of the data through September 2026.

This guide gives the exact definition, a step-by-step calculation with real numbers, the rule's record on the revised and real-time data, what happened in 2024, and why the rule is a quick test of recession risk rather than a forecast.

Where the rule comes from

Sahm's 2019 paper, Direct Stimulus Payments to Individuals, was about fiscal policy, not forecasting. She proposed that the federal government send lump-sum payments to households automatically once a recession began, instead of waiting for Congress to act. That needed a trigger that was timely, simple and hard to argue with. The one she chose was a rise of at least 0.50 point in the three-month average national unemployment rate relative to its low in the previous 12 months.

The trigger caught on as a recession indicator in its own right. The Federal Reserve Bank of St. Louis publishes it on FRED as the Real-time Sahm Rule Recession Indicator, built from the unemployment rates that were actually available each month.

How it is calculated

Averaging three months smooths the household survey's noise, where single-month moves of 0.1 to 0.2 point are common. Comparing with the lowest average of the previous year, rather than with a fixed level, lets the rule work whether unemployment starts from 3.5% or 6%.

The record since 1960

The chart is our own calculation from the unemployment rate as it stands today, after all revisions. On that data, the indicator first reached 0.50 in these months:

Three patterns stand out.

• It is a coincident signal, not a forecast. Every hit came one to eight months after the peak NBER later identified. Its value lies in speed: NBER announces peaks much later, so the rule says "a recession has probably started" well before the official verdict. The What is a recession? How the NBER dates US recessions guide explains NBER's process.
• Once triggered in a recession, it keeps rising. The indicator reached 2.3 in early 1971, 3.7 in 1975, 2.5 in 1982, 4.0 in mid-2009 and 9.4 in June 2020, when the three-month average jumped to 13.0%.
• False signals happen near the line. The 2003 reading touched exactly 0.50 in July and August, during the slow recovery after the 2001 recession, then fell back. Just before the chart begins, the revised data also show a crossing in November 1959, five months before the April 1960 peak, which faded by February 1960. The 2024 episode peaked at 0.57. Both were barely over the threshold, while true recession signals kept climbing well past 1.0.

Real-time versus revised data

The unemployment rate is revised once a year, when BLS updates seasonal factors, so the indicator you compute today for a past month can differ from what forecasters saw then. FRED's real-time series uses the data as published each month.

The real-time and revised records disagree in other places too. In real time the rule crossed 0.50 in November 1976, with no recession, while on revised data it did not; in 2003 the reverse is true. Real-time readings tend to cross later in some episodes, for instance April 2008 instead of February 2008. The lesson: a rule built on noisy, revisable data has fuzzy edges, and readings within a few hundredths of 0.50 deserve little confidence either way.

What happened after 2024

From its August 2024 peak, the indicator fell back as unemployment stabilized. In the revised data it was 0.40 in October 2024, 0.30 in early 2025 and 0.13 in August 2025, rose to 0.35 in November and December 2025 as unemployment reached 4.4%–4.5%, and then declined to 0.00 by September 2026, with unemployment at 4.2%. FRED's real-time series shows the same shape: 0.50 in September 2024, 0.43 in November 2025 and 0.00 in September 2026.

As of the data used here, NBER has not identified a business-cycle peak after February 2020. That makes 2024 a false positive on the record so far. The data cannot say on their own why the rule fired; the indicator measures a rise in unemployment, and unemployment can rise either because jobs are being lost or because the labor force is growing faster than hiring. That second possibility is the rule's main blind spot.

How to use the Sahm rule

• As a recession-risk check, not a timing tool. A reading near zero says unemployment is not rising meaningfully. A reading climbing through 0.3 to 0.5 says it is, and that history has usually ended in recession, but not always.
• Look at the slope and the size. Past recessions pushed the indicator well above 1. A brief touch of 0.50 that reverses, as in 2003 and 2024, is a weaker signal than a steady climb.
• Check the cause of rising unemployment. Rising Jobless claims: reading the weekly unemployment insurance report and falling payrolls point to job losses; a rising participation rate with steady claims points to labor-force growth.
• Combine with other indicators. The Leading, lagging and coincident indicators: what turns first guide shows how the Sahm rule fits alongside yield-curve, claims and output measures.

Pitfalls and limitations

• It lags the start of recessions. By design it detects, not predicts.
• It ignores why unemployment rose. Immigration, re-entrants or demographic shifts that expand the labor force can push it up without layoffs.
• It is sensitive to the threshold. Readings of 0.50 and 0.57 sit within the error of the household survey, whose monthly change in the rate needs to be about 0.3 point to be statistically significant at a 6% rate.
• Revisions change history. The same month can trigger in real time and not after revision, or the reverse.
• It is one series. It uses U-3 only, not payrolls, hours, claims, output or income.
• The past is a small sample. Nine recessions since 1960 is not a large test set, and rules fitted to history often work less well afterward.

The Sahm rule next to other recession signals

See Yield curve inversion: what it signals and its recession record for the leading signal on the bond side.

How Finabase does it

Finabase does not compute the Sahm rule, publish unemployment data or run an economic calendar. The indicator on this page is our own calculation from the public BLS unemployment rate on FRED, downloaded on October 30, 2026.

In the app you can watch how markets respond to labor news that moves the rule: the Treasury page shows yields from 1 month to 30 years, and the economy and Fed topics in Discover gather web coverage of jobs reports and policy.

Frequently asked questions

What is the Sahm rule in simple terms?

It says a US recession has probably started when the three-month average unemployment rate is at least 0.50 percentage point above its lowest three-month average of the previous year. Claudia Sahm proposed it in 2019 as a trigger for automatic stimulus payments.

Did the Sahm rule trigger in 2024?

Yes. In real time the July 2024 jobs report took the indicator to 0.53, and on today's revised data it reached 0.50 in July and 0.57 in August 2024. No recession followed as of the data through September 2026, when the indicator was 0.00.

Does the Sahm rule predict recessions?

No. It identifies recessions early, usually one to eight months after they begin in our calculation, rather than forecasting them. Its value is that it fires well before NBER's official announcement.

Why use a three-month average?

Because the monthly unemployment rate is noisy. A three-month average filters out single-month blips of 0.1 to 0.2 point that come from sampling error in the household survey.

Summary

The Sahm rule compares the three-month average unemployment rate with its low of the previous 12 months and signals a likely recession at a rise of 0.50 point or more. Proposed by Claudia Sahm in 2019 as a stimulus trigger, it flagged every recession from 1960 to 2020 within one to eight months of its start on revised data. It also misfired: in 2003 on revised data, in 1976 in real time, and in 2024, when the July report took it to 0.53 in real time and it peaked at 0.57 before falling to 0.00 by September 2026 with no recession identified. Treat it as a fast, rules-based reading of rising unemployment, and check why unemployment is rising before drawing conclusions.