US Jobs by Sector: Payrolls, ADP, Claims and Openings
1. Why is there more than one US jobs number?
Because nobody counts every job in the country every month. Each measure takes a different slice of the workforce, counts it a different way, and lands on its own day. So the US jobs tracker draws all of them on one chart, on one axis, at the same instant, and you can see for yourself where they disagree.
The headline comes from a survey of about 119,000 businesses and government offices. That is nonfarm payrolls, and it is what the news means by "the economy added 150,000 jobs". The unemployment rate comes from somewhere else entirely, a survey of about 60,000 households. A private payroll company counts the payslips it processes for its own clients and publishes ADP two days earlier. Filings for unemployment pay give you jobless claims every Thursday, and a fourth survey counts unfilled vacancies, which is JOLTS.
That is five numbers in all, taken from four different groups of respondents and published on four different days. None of them is a version of any other. They were built to answer four different questions, and the interesting months are the ones where they stop agreeing.
2. What does the monthly jobs report actually count?
Filled positions, taken from company payroll records. Not people, and not vacancies, so holding two jobs gets you counted twice.
"Nonfarm" is the only piece of jargon in the name, and it means what it says. Farms are out, and so are private households and the armed forces. Those exclusions are old and practical rather than meaningful, and they are not the reason the number ever surprises anybody.
The margin of error is about plus or minus 130,000 jobs, wider than a typical monthly change. So a print of 40,000 against an expected 150,000 leads the news as a bad miss when statistically it is almost indistinguishable from the forecast. That is why the three month average is the figure most professionals watch, and why the tracker offers it as a button rather than making you work it out.
The count is also net: everyone hired minus everyone who left. A quiet month can mean almost nothing happened, or that a great deal happened in both directions and cancelled out. The headline alone cannot tell you which, and that is the gap JOLTS fills.
3. Why can jobs go up while unemployment goes up too?
Because they come from two different surveys that are not measuring the same thing. Nothing else on this page trips people up as often.
The jobs number counts positions at businesses; the unemployment rate counts people in households. Someone holding two jobs adds two to the first and one to the second, while the self employed and farm workers are missing from the first and present in the second. The two surveys are not even the same size, about 119,000 respondents against about 60,000, so they do not move with the same steadiness from one month to the next.
So a month where jobs rise and unemployment rises with them is not a contradiction, and it is not an error. It is two rulers built for two different jobs, and it happens regularly. The tracker keeps them apart on purpose: counts of people go on the chart, and percentages like the unemployment rate sit in their own row of figures above it, because a percentage and a count cannot honestly share a scale.
4. What is ADP, and should you treat it as a preview?
No, and treating it as one is the most common mistake people make. ADP is a payroll processing company. Each month it counts the jobs at the firms whose wages it processes and scales that up, so these are real payslips rather than survey answers. That sounds like the stronger method until you remember whose payslips they are. The official survey samples a much wider slice of the economy than ADP's own client base.
ADP also covers private jobs only. It never sees a government payroll, so it tells you nothing about the public sector.
The comparison is still worth building, because the two line up sector for sector. The tracker puts the official count and ADP's count side by side across all ten private sectors, this month and over the last year, so you can see whether a disagreement runs across the economy or comes from one industry. In practice the gap between them in a given month is routinely larger than the headline number itself. Treat ADP as an early second opinion, not as a forecast of what lands on Friday.
5. What do jobless claims tell you that payrolls cannot?
Speed, and a much more direct kind of evidence. Claims are not a survey and not an estimate; they are an administrative record of forms people actually filed. They land every Thursday and cover a week that ended five days earlier, which makes them the fastest signal on the page by a distance. Payrolls arrive monthly and describe a week in the middle of the month before.
There are two of them, and the difference matters. New claims count people filing for the first time, which measures people losing work. Continuing claims count people who filed earlier and are still drawing pay, which measures people not finding work. Read side by side they separate two problems that one number blurs together. Steady new claims with rising continuing claims means a jobs market that is not firing people but is not hiring them either, a slow squeeze that a healthy looking headline can hide for a long time.
Claims have a real blind spot. They only count people who qualify for state unemployment pay and go through the process of claiming it, so the self employed and most gig workers never appear at all.
6. What is JOLTS, and why do openings and quits matter?
JOLTS is a monthly survey of about 21,000 workplaces. It asks how many roles were unfilled on the last working day, how many people started, and how many left, and those three answers take the net headline apart. Hires and leavers are both far bigger than the monthly jobs figure, because that figure is only the difference between them. A jobs market where nothing is moving looks nothing like one that is churning hard, and the headline treats the two the same.
Quitting is the signal most worth learning. People resign when they believe they can do better somewhere else, so the share of people quitting rises when workers feel secure and falls when they do not. It turns well before layoffs do, and it is one of the clearest reads on confidence you can get.
The order rarely changes. Openings fall first, quits fall next, then hiring slows, and layoffs and claims only climb at the end. A market can be cooling for a year while the headline count stays positive, and if the headline is all you watch, you find out last.
What holds JOLTS back is timing. It arrives about a month behind everything else on the page, which makes it both the deepest view and the stalest. The economic calendar shows when each one lands.
7. Which sectors are actually adding the jobs?
The headline number is worst at exactly this question, which is why the tracker exists. Every private job in the country sits in one of ten sectors, and those ten add up to total private employment exactly. Private plus government adds up to the headline exactly. Government itself splits into federal, state and local, and those three add up too.
That is not a rounding claim. These are counts of people, so the parts simply sum to the whole: nothing is weighted, nothing is estimated, and there is no leftover to explain. The contribution panel groups the eleven pieces into six bands and shows how many jobs each band added or lost, and the six always come to precisely the headline figure.
What the breakdown usually shows is how much work "the economy added jobs" is doing. Health and education has added more than any other sector since 2022, by a wide margin. Leisure and hospitality drove much of the recovery while it clawed back what it lost in 2020. Manufacturing and information can be shedding jobs in the same month a headline nets to a positive number, and that headline tells you none of it.
There is a growth view too, where every line starts at 100. Health and education employs tens of millions and mining a few hundred thousand, so raw job counts will never show you which of the two grew faster. Start both from the same point and it is obvious.
8. Why are government jobs counted separately?
Because they answer to something completely different. Company hiring follows demand and borrowing costs. Government hiring follows budgets, elections and policy, and averaging the two together hides both.
Government here means civilian federal, state and local employees, including public school and public hospital staff. The armed forces are not in it. The split between the three surprises most people: local government is by far the largest, comfortably bigger than federal and state combined, and most of it is school staff. So a swing in the government line is usually about local education budgets or a seasonal adjustment around the school year rather than anything happening in Washington. The tracker separates all three, so you can see which one moved before you decide why.
9. Why do the jobs numbers get revised so much?
Because the first figure is published before all the responses are in, and it gets corrected as they arrive. That happens in two rounds. Each monthly figure is revised twice over the following two months as more of the sample reports in, and then once a year the whole series is rebenchmarked against near complete tax records covering almost every job in the country. Those annual corrections have run to several hundred thousand jobs.
So a first print is a good estimate, not a final answer. The tracker follows the corrections rather than freezing the first version, which means a figure on the page can move after you have read it. That is the data being honest rather than the page being unstable. It is also why a three month average deserves more weight than any single print, and why anyone who builds a strong story on one month should make you suspicious.
10. How do you use the tracker, and is it free?
It is free, with no login, no key and no paywall. There are three controls, and it is easiest to take them in order.
The presets
There are seven of them, each phrased as a question you might actually have. Start here splits the headline into company and government jobs. Two counts of the same thing puts the official private number against ADP's. The rest cover which sectors are growing, public jobs, whether the market is cooling, whether layoffs are starting, and how much people move.
Drawing your own lines
You can draw up to 5 at once from 50 series, and each one carries a plain one line description next to its own switch, so you never need to know the jargon before you click. Dash patterns separate the lines as well as colour does, which keeps the chart readable if you are colour blind, and the labels sit at the end of each line and push each other apart when two land on the same value.
The window, and how far back it runs
Five buttons set the window: this month, the average month over three, the average month over six, the whole year, and growth since the start. A separate control sets how far back the axis runs.
It opens on five years rather than ten on purpose. April 2020 is minus 20.5 million jobs in a single month, roughly a hundred times a normal print, and on a ten year default it draws one spike and flattens everything else into the zero line. The ten year button is there when you want it, and the spike is worth seeing once.
Every line begins where its own survey begins, not where the chart begins. ADP starts in 2010 and the JOLTS series in December 2000, so those lines start part way along, and the page says which ones and when rather than leaving you to wonder whether something broke. Where a figure genuinely was never published, the solid line stops and a faint dashed bridge carries the shape across, with the value left empty in the chart, the table and the feed. We do not write invented numbers into data.
The whole thing works with JavaScript switched off, including a 24 month table. The full history back to 1990, every sector and every window come through the feed, written up for developers, and there is an endpoint your own AI assistant can read directly. It sits alongside the rest of the market data pages. What the jobs numbers mean for prices is on the inflation tracker, and what they mean for borrowing costs is on the Treasury curve and the FOMC page. If the jargon is the barrier, the guides assume nothing, and each release lands in the news feed as it happens. Live market data next to it is what the paid tiers add. The jobs page stays free either way.