The staffing recession ended in January. It lasted 45 months
Temp help employment has risen seven months running. Across the 45 months from the March 2022 peak to the December 2025 trough, it never rose twice in a row. Not once. Here's what turned, what hasn't, and what it means for your desk.
The short version
Temp help employment is the staffing industry's own economic indicator: temps are the first workers hired in a recovery and the first released in a contraction.
It moves before almost anything else in the labour market does, which is why it is worth reading closely even in a quiet quarter. Here's where it stands.
- 7 consecutive months of growth in temp help employment. In the 45 months before January, the series never rose twice in a row.
- 45 months from peak to trough. The longest staffing downturn in the 36 years BLS has published this series.
- -22.5% peak to trough. Shallower than the financial crisis or covid, and more than twice as long as either.
- 1.577% of American workers are temps. At the 2022 peak it was 2.089%. Headcount is recovering. Share isn't.
Seven straight months of growth, after 45 months of none
The trough was December 2025, at 2,451.4k, and every single month since then has printed higher than the one before it, without one interruption.
The industry has added 53,600 jobs off the bottom of that trough, which works out at a gain of 2.2% over seven months.
Bars run from a 2,440k floor rather than from zero, so the shape of the recovery stays legible.
Seven months is a modest run, so why does it matter?
Because of what came before it: between March 2022 and December 2025, this series did not post two consecutive monthly gains. Not one pair, in 45 months.
In other words, every uptick in the downturn got handed straight back the following month, which is why nobody could call the bottom while standing in it. There was never a shape that looked like one.
Seven consecutive gains is the first thing this cycle has produced that a previous recovery would recognise.
BLS revises the two most recent months with every release and re-benchmarks the whole series once a year. June and July 2026 are still preliminary, so a seven-month streak could yet be revised to six.
45 months is the longest downturn on record, and not the deepest
Three previous contractions show up in this series, and all three of them bottomed faster than this one did. Two of them fell further.
This is the shape that operators spent four years trying to describe to their boards and could never quite put a name to.
There was no crash to point at, and revenue didn't fall off a cliff in one quarter that everybody could agree on.
It ground down for very nearly four years, which is longer than the dot-com bust and the financial crisis put together.
A shallow, long decline is harder to run a business through than a deep, short one.
You can cut for two quarters. Cutting for fifteen is a different exercise entirely.
And recovery in this industry can be slower still: after the April 2000 peak, temp help took 164 months to get back to where it started. Thirteen and a half years.
Temp is recovering. Perm placement isn't.
Employment services splits into three at the six-digit NAICS level, and the three have not been moving together.
Temp help has turned, while placement agencies and executive search, which is the line most contingency perm desks actually bill against, is still falling.
In other words, the recovery you’ve read about may not be the one that reaches your desk.
If you bill perm fees, your line is 56131, and it’s down 4.0% with no turn in it yet.
So why would temp turn first?
The usual explanation is that employers hedge: when demand returns but confidence hasn’t, you take a contractor rather than commit to a headcount.
That mechanism is plausible and it matches the sequencing in previous recoveries, but this is a two-series correlation and it should be read as one.
That said, almost nobody publishes this split, which is a shame, because the two halves of the industry have not shared a direction since 2022. It's the difference between a headline that applies to you and one that doesn't.
Temp penetration is 1.577%, barely above the covid floor
Headcount is one story, and the industry's share of the workforce is a second one that runs alongside it. The second one is less kind.
The temp penetration rate is temp help employment as a percentage of total nonfarm payrolls, and it currently sits at 1.577%. At the March 2022 peak that figure was 2.089%, so roughly a quarter of the industry's relative weight has gone.
Here's the comparison that lands it: in April 2020, with the entire economy shut, penetration was 1.495%.
Six years later, in an expansion, with temp employment rising seven months running, the industry's share of American employment is eight hundredths of a point above its pandemic floor.
Headcount is recovering.
Relevance, so far, is not.
Something absorbed that missing half-point of penetration: in-house talent teams, contractor platforms, or work that simply stopped being done.
This data can’t tell you which, and anyone who says otherwise is guessing.
How to read the next release for your own desk
A national number is a weather report for the whole country, not a forecast for the street your desk sits on. Here's how to make it useful.
Find your line first. If you bill temp or contract, watch 56132, and if you bill perm fees, watch 56131.
They’ve been telling different stories for two and a half years, so averaging them tells you nothing.
Check the streak, not the month. One monthly gain means nothing in this series, which was the whole lesson of 2022 to 2025.
Two consecutive gains is a signal. Seven is a trend.
Ignore the first print. The two most recent months get revised every release, so a number that is three months old is a number that is true.
Every figure on this page can be pulled from primary sources in about ten minutes, for free, without a single paid subscription.
The temp series is TEMPHELPS on FRED, which exports to CSV without an account, and the six-digit detail comes from the BLS public API, which needs no key for recent history.
- What agency recruiters actually talk about, where sentiment turned about a year before this data did.
- Recruiting complaints 2026, the same method applied to the candidate side.
How this was built
Sources. Temp help employment and total nonfarm payrolls come from FRED series TEMPHELPS and PAYEMS, which mirror the BLS Current Employment Statistics survey.
The six-digit NAICS detail comes from the BLS public API: CES6056131001, CES6056132001 and CES6056133001. Everything is seasonally adjusted, in thousands of jobs.
Definitions. A downturn episode runs from a local peak in the series to the lowest month recorded before it recovers. Penetration is temp help divided by total nonfarm payrolls, and a gain streak counts consecutive months in which the seasonally adjusted level rose.
This isn’t a perfect test. Three limits are worth stating plainly here, before anyone else gets the chance to state them for us.
The NAICS-6 table covers 30 months rather than 30 years, because the BLS public API returns roughly three years of history without a registered key.
Seasonal adjustment does real work in a seasonal industry, and the adjustment factors themselves get revised.
And employment is a headcount measure rather than a revenue one: bill rates, margins and hours per assignment don't appear here at all. An agency can be busier and poorer at the same time, and this series would show only the busier half.
What’s missing on purpose. The ASA Staffing Index is the other well-known weekly measure of temp and contract employment, and it’s not here because the underlying series isn't machine readable without an ASA participant login. We only publish numbers that we can re-verify from the original source every time this page updates.
Final thoughts
Forty-five months down, seven months up.
The staffing industry has its recovery, and it’s a smaller one than the headline suggests.
- Temp turned in January. Perm hasn’t turned at all.
- Penetration is still within a rounding error of the covid floor, so the industry got its jobs back without getting its share back.
- Two consecutive gains is the bar in this series. Anything less is noise.
One thing to do this month: work out whether your revenue sits in 56131 or 56132, then track only that one. We'll update this page after each jobs report.