Data tracker · Sahm rule
Is the Sahm rule triggered right now?
No. The real-time Sahm rule read -0.07 in August 2026, which puts it 0.57 points below the 0.50 point trigger it fires at. The rule compares the three-month average unemployment rate against its lowest three-month average of the previous year. It has fired for every one of the 9 recessions dated since 1960, though for 7 of them only after the downturn had already started.
Monthly readings, no live figure on this page 800 observations since Dec 1959 Data verified as of 2026-09-09
Latest print, August 2026
-0.07Percentage points above the trailing twelve-month low in the three-month average unemployment rate. The rule fires at 0.50.
Real time, not revised
This is the real-time series, which computes the rule from the unemployment figures as they were published in each month rather than as the Bureau of Labor Statistics later revised them. Seasonal factors are revised every January. A revision can move a past month across the 0.50 line after the fact, so a page built on the revised series describes a signal nobody could have acted on at the time.
Where that sits across 67 years
The numbers
The Sahm rule reads -0.07 in August 2026
Every reading here comes from the August 2026 print, which is the last month the real-time series covers. The recession counts run against the 9 business cycle peaks the National Bureau of Economic Research has dated inside that window. The LBMA gold series runs to the same month, so nothing here pairs two different dates.
Real-time Sahm rule reading
-0.070.13 in Aug 2025
The three-month average unemployment rate minus the lowest three-month average of the previous twelve months, in percentage points. A change here is measured in points rather than percent, so the card prints the level a year earlier instead of a return.
Distance from the 0.50 trigger
−0.57below the trigger
Claudia Sahm set the threshold at 0.50 percentage points in her 2019 Hamilton Project proposal, where it was the switch that would start automatic stimulus payments. The rule fires at exactly 0.50 rather than above it.
Months since the rule last fired
23last at 0.50 in Sep 2024
The reading has sat below the line for 23 consecutive months, running back to October 2024. The sparkline marks every month at or above the trigger over the last five years.
Months at or above 0.50
177 of 80022.1% of the record
Since December 1959, across 14 separate firings. The rule spends roughly a fifth of its life above the line, which is most of what a reader needs to know before treating a firing as rare.
Recessions flagged before the peak
2 of 97 flagged only afterwards
The National Bureau of Economic Research has dated 9 recessions inside this record. The rule fired for 9 of them, and for 7 of those the contraction had already been running for 2 to 4 months.
Sources: Claudia Sahm, real-time Sahm rule recession indicator. Federal Reserve Bank of St. Louis, NBER based recession indicator for the United States. LBMA Gold Price (PM London auction, USD/oz). Retrieved 2026-09-09
The long record
The Sahm rule has fired 14 times since 1959
The full real-time series at every month end, with the 0.50 trigger drawn across it and every dated recession shaded underneath. The reading ran from -0.37 in November 1983 to 9.50 in June 2020, having sat at 0.30 as recently as March 2020.
Real-time Sahm rule gap, percentage points
Monthly readings, December 1959 to August 2026
Aug 2026 print
-0.07The dashed rule is the 0.50 threshold, and everything above it is a month the rule was firing. The solid line at zero is where the three-month average unemployment rate matches its own twelve-month low, which is why the series spends long stretches slightly negative while unemployment is still falling.
Firings against recessions, same 67-year axis
Every firing and every recession on one axis
The 9.50 reading of June 2020 is 19 times the trigger, so on the chart above every crossing near the line is pressed into the bottom of the plot. These two rows carry the same months at a readable scale. Each upper bar starts the month the rule fired and each lower bar starts the month after the business cycle peak, so a bar that begins to the right of the one beneath it is a signal that arrived late.
14 firings against 9 recessions. The grey bars are the 2 firings no recession followed, in Nov 1976 and Jul 2024. Recessions: Apr 1960 to Feb 1961. Dec 1969 to Nov 1970. Nov 1973 to Mar 1975. Jan 1980 to Jul 1980. Jul 1981 to Nov 1982. Jul 1990 to Mar 1991. Mar 2001 to Nov 2001. Dec 2007 to Jun 2009. Feb 2020 to Apr 2020. Peak and trough months come from the National Bureau of Economic Research through the FRED indicator, which counts a recession from the month after the peak through the month of the trough
Sources: Claudia Sahm, real-time Sahm rule recession indicator and Federal Reserve Bank of St. Louis, NBER based recession indicator for the United States. Data verified as of 2026-09-09
Download the Sahm rule monthly dataset as CSVReading it
What a Sahm rule reading of -0.07 means
At -0.07 in August 2026 the rule sits 0.57 points below its trigger, where it has stayed for 23 months running back to October 2024.
The number is a gap rather than a level, which is the part most summaries skip. It compares the last three months of unemployment against the best three-month stretch of the previous year. So a country with 3% unemployment and one with 8% can print exactly the same figure, while one of them is in trouble and the other is fine.
A reading of -0.07 means the labour market is back at its own twelve-month best. It has been below zero since July 2026, which reads better than a bare "not triggered" and is also ordinary, because 265 of the 800 months here sit below zero and 33.1% of the record is spent there.
That construction is also why the rule can fire without a recession following. The three-month average only has to rise 0.50 points off a low. A labour market that was unusually tight has a long way it can loosen before it is weak in any absolute sense. That is roughly what happened in July 2024, when the reading reached 0.57 over 3 months. No business cycle peak has been dated anywhere near it.
The rule fires at exactly 0.50 rather than above it, which matters more than a rounding convention should. One of the 2 firings that led nowhere touched the line and went no further. Nothing about the labour market changes as a reading crosses from 0.49 to 0.50. A firing that clears the threshold by a hundredth of a point carries far less information than one clearing it by a full point, and the record below keeps the two apart by printing the highest reading each firing reached.
The rule gives 2 to 4 months of hindsight rather than notice, so whether any of this belongs in your own account depends on how close you are to withdrawals and what the rest of it already holds. See which companies fit your account
The record
The rule was late to 7 of 9 recessions
The rule fired for every one of the 9 recessions whose peak falls inside this record. Only 2 of them got a signal before the peak, and the other 7 were flagged between 2 and 4 months after the contraction had already begun.
| Recession | Length | First firing | Warning or lag | Highest reading | Gold, next 12 months |
|---|---|---|---|---|---|
| Apr 1960 to Feb 1961 | 10 mo | Dec 1959 | 4 mo early | 1.60 | no price |
| Dec 1969 to Nov 1970 | 11 mo | Oct 1969 | 2 mo early | 2.03 | −6.3% |
| Nov 1973 to Mar 1975 | 16 mo | Mar 1974 | 4 mo late | 3.27 | +2.5% |
| Jan 1980 to Jul 1980 | 6 mo | Apr 1980 | 3 mo late | 2.03 | −6.8% |
| Jul 1981 to Nov 1982 | 16 mo | Nov 1981 | 4 mo late | 2.53 | +5.2% |
| Jul 1990 to Mar 1991 | 8 mo | Nov 1990 | 4 mo late | 1.20 | −4.8% |
| Mar 2001 to Nov 2001 | 8 mo | Jun 2001 | 3 mo late | 1.40 | +17.7% |
| Dec 2007 to Jun 2009 | 18 mo | Apr 2008 | 4 mo late | 3.90 | +1.4% |
| Feb 2020 to Apr 2020 | 2 mo | Apr 2020 | 2 mo late | 4.00 | +3.8% |
Length is peak to trough in months. The highest reading column covers the contraction only, peak month excluded. Gold is the LBMA monthly close twelve months after the first firing, blank where the price series does not reach
Two of those rows carry the whole argument for the rule as an early warning, and they are the two oldest in the table. The peaks at April 1960 and December 1969 were both flagged first, by 4 and 2 months. Every recession since has been dated to a peak that had already passed by the time the rule fired. The lag has been stable at 2 to 4 months, with a median of 4.
The comparison there is against the peak date, which the National Bureau of Economic Research assigns in retrospect rather than at the time. A lag of 4 months against a date nobody held that month is a different thing from a lag against what was knowable then. Measured against the alternative of waiting for the committee to publish a date, the rule is quick. The claim it cannot carry is the one about warning you ahead of a downturn, and the 56 years since December 1969 are the evidence against it.
| Firing | Months above | Highest | What followed | Gold, next 12 months |
|---|---|---|---|---|
| Dec 1959 to Jan 1960 | 2 | 0.77 | 4 months before the Apr 1960 peak | no price |
| Aug 1960 to Nov 1961 | 16 | 1.70 | Recession already running, dated from Apr 1960 | no price |
| Oct 1969 | 1 | 0.50 | 2 months before the Dec 1969 peak | −6.3% |
| Feb 1970 to Oct 1971 | 21 | 2.37 | Recession already running, dated from Dec 1969 | +10.8% |
| Mar 1974 | 1 | 0.50 | Recession already running, dated from Nov 1973 | +2.5% |
| Jul 1974 to Jan 1976 | 19 | 3.83 | Recession already running, dated from Nov 1973 | +6.9% |
| Nov 1976 | 1 | 0.50 | No recession followed | +22.9% |
| Apr 1980 to Apr 1981 | 13 | 2.03 | Recession already running, dated from Jan 1980 | −6.8% |
| Nov 1981 to Jun 1983 | 20 | 2.53 | Recession already running, dated from Jul 1981 | +5.2% |
| Nov 1990 to Nov 1992 | 25 | 1.53 | Recession already running, dated from Jul 1990 | −4.8% |
| Jun 2001 to Oct 2002 | 17 | 1.63 | Recession already running, dated from Mar 2001 | +17.7% |
| Apr 2008 to May 2010 | 26 | 3.90 | Recession already running, dated from Dec 2007 | +1.4% |
| Apr 2020 to Mar 2021 | 12 | 9.50 | Recession already running, dated from Feb 2020 | +3.8% |
| Jul 2024 to Sep 2024 | 3 | 0.57 | No recession followed | +36.0% |
Runs separated by fewer than 3 months below the trigger count as one firing. A firing is a miss once the NBER record runs 24 months past it with no peak dated
Gold against a control
Gold returned +4.5% in the year after a trigger
Gold rose over the following 12 months after 9 of the 12 firings the price series is long enough to measure, at a median of +4.5%. A month drawn at random from the same years returned +6.2% over the same horizon.
After a firing
+4.5%Median gold return over the 12 months after the first month of a firing, across 12 firings running from October 1969 to July 2024. Gold finished higher in 9 of them.
After any month at all
+6.2%The same measure run from every one of the 689 months both series share, which is the comparison that turns a number into evidence. Gold finished higher in 62.1% of those windows against 75.0% after a firing.
The difference
−1.7%Percentage points of median return separating the two, on a sample of 12 firings. A gap that small on a sample that small is not a finding, and anyone selling gold on the strength of a labour market signal is working from a shorter record than this one.
The spread inside those 12 firings is far wider than the median suggests, running from −6.8% to +36.0%. That range covers most of what gold has ever done over a year. Both ends of it came from firings that look much the same on the chart. So the signal is not separating the good twelve months from the bad ones in any way a buyer could act on, and a median drawn from 12 observations would not settle the question either way.
2 of the 14 firings carry no gold figure at all. They fall before the LBMA series begins in April 1968, so there is no published London price to measure them against. That removes the two oldest firings from the gold column, which leaves it describing a shorter record than the rest of the page.
Method
How is the Sahm rule calculated?
Every figure is computed at build time from three committed monthly series, so the numbers are in the HTML before any script runs and the tables work with JavaScript turned off.
The reading is FRED series SAHMREALTIME, the real-time construction of the rule. It uses the unemployment rate as it stood in each month's published release, rather than the revised series recomputed after the Bureau of Labor Statistics updates its seasonal factors each January. That is the version anyone acting on the signal would have held at the time. FRED publishes the revised construction separately as SAHMCURRENT, and the two disagree about individual months, so a figure quoted from one will not always match a figure quoted from the other.
Recession dating comes from the NBER indicator published as FRED series USREC. Its own definition counts a recession from the month after the peak through the month of the trough, so a business cycle peak is the month before the indicator first reads 1. Reading it the other way would move every lag on this page a month earlier, in the direction that flatters the rule. All 9 peak and trough pairs derived that way match the reference table the National Bureau of Economic Research publishes, which was last updated on 14 March 2023.
Firings are grouped by month key rather than by position in an array. Runs separated by fewer than 3 months below the trigger count as one firing, because a reading that dips to 0.47 for a month and comes back has not un-fired the rule. A firing counts as a miss only once the recession record runs 24 months past it with no peak dated. That waiting period is a convention this page chose and states here, rather than one tuned until the record looked better than it is.
Gold is the LBMA monthly close, taken at the month a firing began and again 12 months later, matched on the calendar month rather than by counting rows forward. The control beside it runs that same measure from every month the two series share. Both samples stop 12 months before the end of the price series, because a window that has not closed yet cannot be scored.
Limits
What the Sahm rule cannot tell you
The sample is 9 recessions in total, which is small enough that the entire record fits into a single table. The rule identifies a downturn that has usually already started, and most of what follows comes back to one of those two facts.
- 9 recessions is the entire sample. Every lead time, lag and hit rate on this page rests on 9 observations spread across 67 years. A median lag of 4 months describes what happened rather than a distribution anyone can put a confidence interval around.
- The rule confirms rather than forecasts. In 7 of the 9 cases the business cycle peak had already passed by the month the rule fired. Both exceptions are more than 56 years old now. For whether a downturn is coming, our yield curve inversion tracker covers the indicator that leads, and our recession probability tracker covers the model that estimates where the economy stands today.
- Real-time and revised disagree. This page uses the real-time series throughout. The revised construction recomputes past months after each January seasonal-factor update, so a month can cross the 0.50 line retrospectively. A figure here that fails to match one quoted elsewhere will usually differ for that reason rather than because either of them is wrong.
- 1 month is missing from the record. The series carries no observation for October 2025, because no unemployment rate was ever published for that month. FRED shows the same hole in both the rule and the underlying rate. Every calculation here joins on the calendar month, so the gap cannot quietly turn a twelve-month lookback into a thirteen-month one.
- The record starts mid-firing. The series opens at 0.77 in December 1959, already above the trigger. So the first firing may have begun before anything this page can see, and its 4 months of warning ahead of the April 1960 peak should be read with that in mind.
- Gold here is a spot close and nothing more. Every gold figure is the LBMA monthly fix with no dealer premium over spot, no setup fee, no annual custodian fee, no storage fee and no tax. A real outcome inside a gold IRA is worse than the column shown by the size of all five. Which form you hold moves it further: a gold ETF and a mining stock both sit in an ordinary brokerage IRA at brokerage cost, while bullion needs a self-directed IRA, a custodian and a depository, and the metal has to clear the 99.5% purity rule before a custodian will accept it. The price series runs to August 2026, the same month the Sahm series ends, so no figure here pairs two different dates.
- NBER dating can still change. Peaks are assigned in retrospect, so the record this page scores firings against is not final. The July 2024 firing could yet gain a peak near it and move out of the miss column. That is the reason a firing is only called a miss once 24 months have gone by without one.
Sources
Where these Sahm rule numbers come from
- FRED series SAHMREALTIME. Published as Claudia Sahm, real-time Sahm rule recession indicator. 800 monthly observations in percentage points, December 1959 to August 2026. Its series notes carry both the 0.50 point definition and the statement that the Bureau of Labor Statistics revises unemployment annually in January, which is what separates this series from the revised one. Retrieved 2026-09-09
- FRED series USREC, NBER based recession indicators for the United States. Federal Reserve Bank of St. Louis, NBER based recession indicator for the United States. Monthly, 1 from the month after an NBER peak through the trough month, 0 otherwise, through August 2026. Retrieved 2026-09-09
- NBER, US business cycle expansions and contractions. The published reference table, last updated 14 March 2023, used to check the 9 peak and trough pairs derived from the indicator above
- Claudia Sahm, Direct stimulus payments to individuals. The Hamilton Project, 16 May 2019. The proposal the rule comes from, where the trigger starts automatic payments rather than forecasting a downturn
- Bureau of Labor Statistics, Labor Force Statistics from the Current Population Survey. The survey the national unemployment rate comes from, which is the single input the rule is computed on
- LBMA precious metal prices, monthly gold fixes. LBMA Gold Price (PM London auction, USD/oz). 701 monthly closes in USD/oz, April 1968 to August 2026. Retrieved 2026-09-09
Full dataset available as CSV, with the reading, the trigger flag, the recession indicator and the gold close in separate columns, so anyone who prefers a different threshold can rebuild the record themselves. Corrections go to the editorial team and are logged publicly.
The Digest Perspective
9 September 2026
The Sahm rule went below zero in July 2026 and reads -0.07 in August 2026, putting the three-month unemployment average back at its own twelve-month low. That reads better than a bare "not triggered", and it is ordinary, since 265 of 800 months sit below zero.
What the tables cannot settle is what a quiet reading is worth. The rule fired for all 9 recessions dated since 1960, and in 7 of them the contraction was already running by the month it fired. So -0.07 says the labour market has not broken yet, and nothing about the notice you would get.
The Federal Reserve
It wants the labour market loose enough to finish the inflation job, no looser. Because the rule measures a rise off a low rather than a level, a deliberate cooling and the first months of a recession print the same number.
Claudia Sahm
She proposed this in May 2019 through the Hamilton Project, as the switch starting automatic stimulus payments once a downturn was already underway. Against that job, the 7 late calls are the design working rather than failing.
The gold sales floor
A firing is a reason to call somebody, so the incentive runs toward treating a threshold crossing as a countdown. The July 2024 firing is the cleanest test of that, because it reached 0.57 over 3 months and no peak has been dated near it.
Where we are probably wrong
The strongest case against all of that is that we are holding the rule to a job it never applied for. A record of 2 misses in 67 years, with a firing for all 9, beats almost anything else in macroeconomics, and confirming a downturn inside 4 months helps anyone who would otherwise wait years for a published date.
We think a quiet Sahm rule tells you the labour market has not broken yet and nothing about whether it will. We would not move a retirement account on this number in either direction, and we would not ignore it the month it fires, because 7 times out of 9 that was already the news.
Which leaves the question that is actually yours. Whether the account survives a downturn arriving on a date nobody can name depends on how close you are to drawing on it and on what it already holds, and neither of those is on this page.
A few questions about your account and your timeline, and no phone number is needed to see matches. We are not a licensed adviser and none of this is a recommendation to buy.
If you would rather not answer anything, the fee page carries what a setup fee, an annual custodian fee and a depository storage fee actually run, and the rollover eligibility checker works out which transfer route your plan allows.
Written by the Gold IRA Digest Editorial Team on 9 September 2026. The reporting above this block is recomputed from the publishers' own files on every build. This block is opinion, it is dated so it can age in public, and earlier versions stay in the changelog rather than being quietly rewritten.
Common questions
Sahm rule questions
- What is the Sahm rule?
- The Sahm rule fires when the three-month moving average of the U3 unemployment rate rises 0.50 percentage points or more above the lowest three-month average of the previous twelve months. Claudia Sahm published it in May 2019 through the Hamilton Project. There it was the switch that would start automatic stimulus payments once a downturn was underway, rather than a forecast of one. The real-time version on this page covers 800 months since December 1959. It has sat at or above the threshold in 177 of them.
- Is the Sahm rule triggered right now?
- No. The reading was -0.07 in August 2026, which puts it 0.57 points below the 0.50 threshold. It last fired in September 2024 at 0.50. It has stayed under the line for the 23 months since October 2024. The unemployment rate this is computed from is published monthly by the Bureau of Labor Statistics, so the figure moves once a month rather than daily.
- Has the Sahm rule ever been wrong?
- Yes. It has fired 2 times in 67 years without a recession following, in November 1976, when it reached 0.50 and no recession was dated in the 24 months that followed, and again in July 2024, when it reached 0.57 and no recession was dated in the 24 months that followed. That is a low miss rate by any standard. It also rests on 9 recessions, which is a much smaller sample than the precision of a 0.50 threshold suggests. Counting the other direction, the rule did fire for every one of those 9 recessions.
- Does the Sahm rule predict a recession before it starts?
- No. Of the 9 recessions the National Bureau of Economic Research has dated inside this record, 2 were flagged before the business cycle peak. The other 7 were flagged only after the contraction had already begun, by 2 to 4 months. The rule was designed to confirm a downturn fast enough to release money into it, and the record is consistent with that job rather than with forecasting. For whether a recession is coming, our yield curve inversion tracker covers the indicator that leads rather than confirms.
- What happened to gold after the Sahm rule triggered?
- Gold rose over the following 12 months after 9 of the 12 firings the price series is long enough to measure, at a median of +4.5%. The control matters more than that figure does. A month drawn at random from the same years returned a median of +6.2% across 689 windows, so the firings did slightly worse than an ordinary month rather than markedly better. The remaining 2 firings fall before the LBMA series starts in April 1968 and carry no figure at all.
How this page is built
Every figure is derived at build time from three committed data files, so the numbers are in the HTML before any script executes and every table works with JavaScript turned off. The three series are joined on a year-and-month key rather than on a position in an array. That matters here because the Sahm file is missing October 2025, and an index-based twelve-month lookback would quietly measure thirteen months for everything after the gap. Business cycle dating comes from one shared helper used by every recession page on this site, so the four of them cannot disagree about whether a peak month counts as part of the contraction.
Gold IRA Digest is an independent publisher. This page is information, not investment advice, and nothing on it is a recommendation to buy or sell gold or any other asset. Past performance does not predict future results. Some links on this site earn us a commission, which never changes what we publish about a company.