On this page
- Start with jobs, not people or new hires
- Check the series and the month before the headline
- CES payroll jobs and CPS employment are different measures
- Separate a forecast surprise from a prior-month revision
- An initial estimate is not a permanently final number
- Read the wider report without turning it into a market forecast
- A statistical release does not define a contract's result
- Common questions about nonfarm payrolls
- Keep a short release-reading record
- Sources and scope
Start with jobs, not people or new hires
US nonfarm payrolls measure payroll jobs, not the number of different people working or the number newly hired. The familiar monthly headline usually reports the estimated change in those jobs. The unemployment rate comes from a different survey and answers a different question. [2][3]
To read a release, identify the series, reference month, units, seasonal adjustment and estimate vintage (the release it comes from) before comparing it with a forecast. A positive jobs change can miss a forecast, while a revision can lower an earlier month's estimate without describing a new loss this month.
Check the series and the month before the headline
The Bureau of Labor Statistics (BLS) publishes the Employment Situation with separate establishment and household survey tables. For the payroll headline, look for the Current Employment Statistics (CES) measure of total nonfarm payroll employment. A table may show the employment level or its change from the previous month; those are different quantities. [1][2]
Total nonfarm includes government as well as private payroll jobs. Federal, state and local government are not excluded simply because the measure is called nonfarm. The CES measure excludes agricultural employees, private-household employees, unincorporated self-employed workers and unpaid family workers. A person with jobs on two nonfarm payrolls can be counted twice. [2][4]
- Series
- US CES total nonfarm payroll employment, monthly change; private and government jobs.
- Reference month
- Month B, a fictional month. The CES reference pay period includes the 12th; it is not the later publication day.
- Estimate vintage
- Release R1, the fictional first estimate for Month B.
- Unit and adjustment
- Thousands of jobs; seasonally adjusted. A displayed +150 means +150,000 jobs.
- Comparison
- A fictional pre-release forecast of +180 thousand for that same Month B series and adjustment.
Month B and Release R1 are teaching labels, not an actual BLS release or schedule. The reference-period definition is from BLS. [2]
Seasonal adjustment accounts for recurring patterns such as holidays and seasonal business activity. It does not remove all uncertainty or make a sample estimate an exact count. Keep the adjustment basis consistent when comparing months or matching a forecast. [4]
A headline labelled ADP private payrolls is not the BLS total-nonfarm series used here. Keep its publisher and coverage attached to it; do not substitute it for the BLS result merely because both headlines concern jobs.
CES payroll jobs and CPS employment are different measures
The payroll survey asks establishments about jobs. The Current Population Survey (CPS), or household survey, measures people's labor-force status. Their coverage and reference periods differ, so their monthly employment changes need not agree. [2][3]
| Reading question | CES establishment survey | CPS household survey |
|---|---|---|
| What is counted? | Nonfarm payroll jobs; a worker on two payrolls can count twice. | Employed people; a person with more than one job counts once in employment. |
| What is the coverage? | Nonfarm business and government payrolls, with the exclusions described above. | The civilian noninstitutional population aged 16 and over; its employment coverage also includes categories outside CES, such as self-employment and agricultural work. |
| Which outputs matter here? | Payroll employment levels and changes; also hours and earnings measures. | Employment, unemployment, labor force and participation measures. |
| Which period should you check? | The reference pay period, including the 12th of the month. | The survey's reference week, rather than the publication day. |
| Can it supply the unemployment rate? | No. Neither its jobs level nor its monthly change is the CPS unemployment-rate denominator. | Yes. The unemployment rate is unemployed people as a percentage of the labor force, under CPS definitions. |
The comparison follows the BLS technical note, FAQ and methods guide. The labor force consists of employed and unemployed people under the household survey's definitions; it is not the entire population. Do not divide a payroll-jobs change by population to calculate unemployment. [2][3][4]
Separate a forecast surprise from a prior-month revision
Suppose a fictional Release R1 reports a seasonally adjusted increase of 150,000 total nonfarm payroll jobs for Month B. A fictional forecast recorded before the release was an increase of 180,000 for the same series and month. The difference is +150,000 − (+180,000) = −30,000 jobs. Employment still increased in this example; the result was 30,000 jobs below that forecast.
Now suppose the same release changes the estimate for the earlier Month A from +100,000 to +80,000. The revision is +80,000 − (+100,000) = −20,000 jobs for Month A. That is a change to an earlier estimate, not a statement that 20,000 people were newly fired when Release R1 appeared.
- Before Release R1: Month A's first estimate, published in Release R0, is +100. The separately recorded forecast for Month B is +180.
- At Release R1 — new Month B estimate: +150 actual estimate versus +180 forecast gives −30, or −30,000 jobs relative to the forecast. Month B's estimated change is still positive.
- At Release R1 — revised Month A estimate: +80 replaces the earlier +100, giving a −20 revision, or −20,000 jobs relative to Month A's first estimate.
The surprise compares an estimate with a forecast for Month B. The revision compares two estimate vintages for Month A. Do not add −30 and −20 and label the sum one month's payroll change or surprise.
A forecast or consensus is not measured by BLS. For a real comparison, record its publisher, collection or aggregation method when available, observation time, target month, series and adjustment basis. This example supplies no real consensus or latest payroll value. “Actual” here means the reported estimate used in the comparison, not an error-free count.
An initial estimate is not a permanently final number
In ordinary monthly CES processing, an initial estimate is revised twice as additional reports and updated seasonal factors become available. BLS also benchmarks payroll estimates against administrative employment records later. Completion of the two monthly revisions does not prevent a later benchmark revision. [2][4]
A revision changes the statistical picture of its reference period. Keep the old and new publication vintages beside the same month, rather than moving the revision into the current month's change. When comparing historical trends, check whether all observations come from the same updated series instead of mixing first estimates with later revised values.
Both sampling and nonsampling error matter. Responses, coverage, modelling and seasonal adjustment can affect the estimate; one rounded headline is not a perfectly precise census. This guide does not supply a universal error band or a confidence threshold for deciding whether a monthly change is meaningful. Those require the relevant BLS series and reliability information. [4]
Read the wider report without turning it into a market forecast
Payroll jobs, unemployment, participation, hours and earnings describe different aspects of the labor market. Read their units and populations before treating them as one story. A change in a wage measure is not another count of jobs, and a household labor-force measure is not interchangeable with an establishment payroll estimate. [1][4]
The Federal Reserve considers broad labor-market information in pursuing its employment and price-stability goals. One payroll surprise does not mechanically determine the next policy decision or the direction of the dollar, stocks or gold. A forecast comparison is also different from a probability estimate. Our guide to reading CME FedWatch probabilities explains the separate model and its assumptions. [5]
Use the BLS release gateway to check the publication you are reading. Do not assume that every release will arrive on the first Friday or at a fixed clock time where you live. This guide gives no live release timetable.
A statistical release does not define a contract's result
If you are reading an event contract, check its exact source, reference period, threshold, deadline and treatment of initial or revised data. A question about a first estimate is different from one tied to a later revised figure. The wording of the specific contract controls that distinction; a familiar “NFP” label is not enough.
The prediction-market guide covers prices, questions and payouts. Nothing in this statistical explanation establishes that a particular payroll market exists, is available to you or has any specific resolution procedure. Understanding a release does not establish individual legal, age or account eligibility.
Trading real-money event contracts can cause losses and incur fees. A correct reading of a release does not make a position safe, provide reliable income or offer a way to recover earlier losses.
Common questions about nonfarm payrolls
Why can payrolls rise while unemployment also rises?
They come from different surveys with different coverage and units. Payrolls count jobs, while the unemployment rate relates unemployed people to the household survey's labor force. Changes in employment and the labor force can produce different movements. Neither headline can be calculated from the other alone. [2][3]
Does below forecast mean jobs fell?
No. In the fictional example, +150,000 jobs is below a +180,000 forecast by 30,000, but it is still a positive estimated monthly change. Below forecast and below the previous month's employment level are different comparisons.
Are downward revisions new job losses?
Not necessarily. A change from an estimated +100,000 to +80,000 revises the earlier month's increase downward by 20,000. It does not describe 20,000 new dismissals in the current month. Identify the month, whether the figure is a level or change, and both estimate vintages.
Is ADP the same as the BLS payroll report?
No. Do not replace a BLS total-nonfarm figure with a headline labelled ADP private payrolls. Their source and scope are different, and BLS total nonfarm includes government jobs. A comparison needs the actual series labels, not just the word payrolls.
Can one release tell me the Fed's next decision?
No. A payroll report is one source of labor-market information, not a policy commitment or a fixed rule for a rate decision. It also cannot guarantee a profitable trade or a reliable income. [5]
Keep a short release-reading record
- Copy the exact US series and distinguish total nonfarm from private payrolls.
- Record the reference month, publication vintage, jobs or thousands of jobs, and adjustment basis.
- Label a level, monthly change, forecast difference and revision separately.
- Keep any forecast's publisher and observation time; match its series and month.
- Read the wider report and its uncertainty without inferring a guaranteed policy or market reaction.
Explore HypaForecast markets
Browse the market list, or inspect the linked 5-minute and 15-minute BTC market pages. These discovery links do not establish that a contract is currently offered, tradable or available to you, and they are not NFP examples.
Sources and scope
Source evidence checked on 18 October 2026 is reused here. BLS release notes, FAQs and a bounded methods-guide extract support the statistical definitions and revision limits. The Federal Reserve source supports broad policy context only. All Month A/Month B, R0/R1, forecast and jobs-change numbers in the worked example are fictional. No live release, real consensus, market response or individual eligibility was verified. The source-check date is not this article's publication date.
HypaForecast publishes educational guides and operates a prediction market. This article explains US employment statistics; its discovery links do not establish product availability or recommend a transaction.
Share this guide