Growth and labour
Non farm payrolls is the monthly count of US jobs added or lost, released with the unemployment rate and wage growth in the same report. It is the labour half of the Fed's mandate, and it is the release most likely to reprice the front end of the curve on a day that is not a CPI day.
Also known as US jobs report, payrolls, NFP
US non farm payrolls rose by +162,000 in August 2026, against +21,000 in July 2026. The unemployment rate was 4.1 per cent and the participation rate 61.6 per cent.
Average hourly earnings rose +0.3 per cent on the month and 3.1 per cent on the year. Private payrolls accounted for +127,000 of the total.
Release schedule
The next US Employment Situation report is scheduled for 2 October 2026 at 08:30 New York time, which is 13:30 in London. That date is read from the BLS Employment Situation release schedule each time this page is built, so it follows the publisher rather than being stored here.
Non farm payrolls is one line inside that report. The Bureau of Labor Statistics sets the schedule a year in advance and this page reads that schedule directly, which matters more than it sounds: the release is commonly on the first Friday of the month, but it has not always been, and a page that assumed the rule would have been wrong through the disrupted 2025 calendar. The date above comes from the publisher, not from a rule.
Because the Fed has two jobs, and this is the report that scores the second one.
Inflation tells the Fed whether it is failing on prices. The labour market tells it whether it can afford to do anything about that. A central bank that is worried about inflation but watching unemployment climb behaves very differently from one looking at the same inflation with a labour market that will not crack.
That is why payrolls day moves the front end of the Treasury curve so violently. The market is not repricing the economy, it is repricing how much room the Fed has, and the two year yield is the instrument that carries that repricing first.
It is also the most information dense release of the month. CPI is one number with a core underneath it. The Employment Situation is a jobs count, an unemployment rate, a participation rate and a wage number, published together, drawn from two different surveys, and free to disagree with each other.
The market cares about which of the four numbers changes the Fed's problem.
The headline jobs count is the number that gets the airtime and it is the one most often over read. It is an estimate with a confidence interval wide enough to swallow most of the surprises traders react to, and it gets revised twice.
The unemployment rate comes from a completely different survey. Payrolls come from the establishment survey, which asks employers how many people are on the books. Unemployment comes from the household survey, which asks people whether they have work and whether they are looking. The two can move in opposite directions in the same month without either being wrong.
Wage growth is where the labour report becomes an inflation report. Average hourly earnings rising faster than productivity is the mechanism that turns a tight labour market into persistent services inflation, which is the part of inflation the Fed can least easily ignore.
Participation decides how to read the unemployment rate at all. Unemployment falling because people found work is a strong labour market. Unemployment falling because people stopped looking is a weakening one. The rate is identical in both cases.
And revisions decide whether last month's story was true. A headline that beats while the prior two months are revised down by more than the beat is a weaker report than the number on the screen.
Scenario analysis
Hawkish combination
This is the combination that reprices the Fed hardest, because it moves both halves of the mandate the same way.
Strong hiring says the economy can take tighter policy. Accelerating wages say tighter policy is still needed. The front end sells off, the curve flattens, the dollar firms on the widening differential, and gold struggles as the real return on the alternative rises.
What matters next is whether the long end follows. If the two year sells off and the ten year does not, the market believes the Fed will hold tighter for longer without changing where rates end up. If both sell off, the market has changed its view of inflation itself, which is a much bigger move.
Note that strong payrolls with COOLING wages is a different report entirely. That is the combination the Fed wants: activity without price pressure. It is usually read as risk positive rather than hawkish, and it is the case most often mislabelled as a hawkish beat.
The trap
This is the report most likely to produce a first move that reverses.
A headline that beats expectations while the previous two months are revised down by more than the beat is, in net terms, a weaker report. The initial algorithmic reaction trades the headline. The considered reaction trades the net.
The tell is the front end. If the two year sells off on the headline and then gives it all back within the hour, the market has finished doing the arithmetic and the first move was noise.
The mirror image is worth the same attention: a weak headline alongside upward revisions to the prior months can be a stronger report than it looks, and it produces the same reversal in the opposite direction.
Dovish combination
This is the combination that adds cuts back into the curve, and the front end takes most of the move.
Because the two year spans the window the market is repricing, it falls further than the ten year, the curve steepens, and the dollar usually softens as the differential narrows. Gold typically gets a clean tailwind, since the real return on the alternative is falling.
The participation rate decides how dovish it really is. Unemployment rising while participation also rises is people entering the workforce faster than jobs are being created, which is a loosening labour market rather than a failing one. Unemployment rising while participation falls is genuine deterioration and reprices the curve faster.
The version that catches people out is a weak report the Fed publicly ignores. If officials push back within days, the move reverses, and anyone who traded the first hour is left holding a position the curve no longer supports.
Market impact
Bond yields
The front end moves first and moves most.
The two year Treasury prices the expected path of policy over the next two years, and payrolls day is a direct input into that path. A surprise that changes how many moves the market expects shows up in the two year within seconds, well before the ten year has finished deciding what it thinks.
The ten year moves less because a single labour report rarely changes the market's view of where rates settle over a decade. That difference is visible in the curve: a hot report usually flattens it as the front end sells off harder, and a weak report usually steepens it as the front end rallies.
The useful discipline is to watch the two year rather than the headline. If payrolls beat and the two year does not move, the market did not learn anything about the Fed, and the rest of the reaction is unlikely to hold.
Currencies
The dollar reacts through the rate differential, not through the jobs number.
A strong report lifts the dollar when it widens the gap between expected US rates and expected rates elsewhere. The same report can leave the dollar flat if other central banks are moving in the same direction, or if the front end had already priced the outcome.
That is why payrolls sometimes produces a large dollar move and sometimes almost none. The variable is not the size of the jobs number, it is how much the number changed the differential.
The cleanest dollar moves come from combinations, not headlines. Strong payrolls with accelerating wages widens the differential twice over, because it moves both the growth and the inflation side of the Fed's problem at once.
Gold
Gold trades the real return on the alternative to holding it.
A hot labour report pushes yields up and rate cut expectations out, which raises the real return on holding Treasuries instead of gold, and gold tends to struggle. A weak report does the reverse.
The wage number matters more to gold than the jobs count, because gold responds to real yields rather than nominal ones. Wage growth that lifts inflation expectations alongside nominal yields can leave real yields roughly unchanged, and gold with them.
The reaction that catches people out is a weak report that raises recession risk. Gold can then rally on the safe haven bid and on falling real yields at the same time, which produces a larger move than the jobs number alone would suggest.
Equities
Equities read payrolls through the discount rate and through earnings, and the two can point in opposite directions.
Good news for the economy is good for earnings and bad for the discount rate. Which one dominates depends on where the market thinks the Fed is. When the concern is inflation and tightening, a strong labour report is usually read as hawkish and equities fall. When the concern is a slowdown, the same report is read as resilience and equities rise.
Long duration equities are the most sensitive, because their value sits furthest out in time and is therefore most exposed to the front end repricing.
The combination that hurts equities most is strong wages with soft payrolls: the inflation side worsens while the growth side weakens, and both halves of the valuation argument move the wrong way at once.
The most common error is treating the headline count as the report.
It is one of four numbers, it carries a wide margin of error, and it is revised twice before it settles. Traders who read only the headline are reacting to the noisiest part of the release.
The second error is assuming the unemployment rate confirms the payroll number. It comes from a different survey with a different sample and a different question. When they disagree, that is information about the labour market, not a mistake in the data.
The third is ignoring revisions. A print of plus two hundred thousand alongside minus one hundred and twenty thousand of revisions to the prior two months is a net eighty thousand. The market does that arithmetic within seconds and the screen headline does not.
The fourth is reading a strong labour market as automatically dollar positive. It is dollar positive only while it moves rate expectations. A strong jobs number that changes nothing about what the Fed is expected to do rarely holds its first move.
The fifth is forgetting participation. An unemployment rate can fall for a good reason or a bad one, and the participation rate is what separates them.
The last is trading the first print of the spike. The first sixty seconds are algorithmic and frequently reverse once the composition of the report is read properly.
The view here changes when the relationship between the labour market and Fed pricing changes, not when a single month surprises.
If payroll growth slows for three consecutive months while the unemployment rate stays flat, that is late cycle labour hoarding rather than a downturn, and the front end should price patience rather than cuts.
If the unemployment rate rises while participation is also rising, the labour market is loosening for a benign reason and the Fed can wait. If unemployment rises while participation falls, that is a genuine deterioration and the curve should reprice faster.
If wage growth reaccelerates while payroll growth slows, the Fed loses the easy trade off it has been relying on, and that combination is the one most likely to force a hawkish repricing out of a weak headline.
And if a run of revisions consistently moves in one direction, the level of employment the market thinks it is looking at is wrong, which changes the starting point for everything downstream.
Official reference
The US Bureau of Labor Statistics publishes the Employment Situation at 08:30 New York, usually on the first Friday after the reference month. The figures here are read directly from the BLS public API.
Payrolls, private payrolls, manufacturing payrolls, average hourly earnings and average weekly hours come from the establishment survey of employers. The unemployment rate and the participation rate come from the household survey of people. They are different samples answering different questions, which is why they can disagree in the same month. All series shown here are seasonally adjusted.
The monthly jobs change is calculated as the difference between the current and previous month's employment level, and only a month present in every series is used, so no figure on this page is paired with a number from a different reference month.
Revisions are recorded rather than retrieved. The BLS API serves the current vintage of every series, with no record of what was first printed, and the only public vintage database for these series requires a key this desk does not hold. So the first published value for each month is written to a ledger the first time it is seen and never overwritten. Where no first print was recorded, the revision fields are left empty rather than estimated.
No market consensus or forecast figure is shown on this page. Expectations are not published by the BLS and this desk has no licensed consensus provider, so quoting one would mean sourcing a number we cannot stand behind. Everything above is official data.
One quick step
Drop your email and the desk will send you the details.
KenMacro
Free community
Thousands of traders staying ahead of the macro, together. It's free, and you're one click away.
Instant access. Free, always.
Three thousand traders already get it. The full week ahead, in your inbox before the London open.
Free, weekly, no spam. Unsubscribe in one click.
The next Sunday brief lands in your inbox at 12:00 BST. In the meantime, head to macro insights for the desk reactive coverage.