Every first Friday morning, the Bureau of Labor Statistics publishes the Employment Situation, and every news channel treats it like a verdict on the economy. The reality is less tidy: the report is built from two separate surveys, each a statistical estimate drawn from a sample, with a three-week lag built into the design.
Here is how the jobs report is put together, who produces it, what each number measures, and why the figure you read in March can change in August.
Table of Contents
- How the Jobs Report Is Put Together at a Glance
- Which U.S. Agency Produces the Jobs Report?
- What Does the Jobs Report Measure?
- Total nonfarm payroll employment
- The unemployment rate
- Labor force participation rate
- Average weekly hours and average hourly earnings
- Job openings and quits
- How Does the Establishment Survey Count Jobs?
- Sample selection and coverage
- Why the 12th of the month decides the numbers
- Why job switchers get counted twice
- From sample to national total
- How Does the Household Survey Measure Unemployment?
- The three classifications
- Why the unemployment rate is not the share of adults without a job
- Why the two surveys can disagree
- How Are Seasonal Adjustments Made?
- Which industries get adjusted
- How the factors are estimated and updated
- Why Do the Jobs Report Numbers Change After Release?
- The two monthly revisions
- The annual benchmark revision
- Other revision sources
- How to tell a revision apart from real change
- How Is the Jobs Report Published?
- Why the lag exists
- What to check in the release itself
- What Does the Jobs Report Not Tell You?
- Key Limits and Revisions of the Jobs Report
- Sampling error is built into every figure
- Small monthly changes often mean nothing on their own
- Coverage and model assumptions matter
- Interpretation, not just data
- Frequently Asked Questions
- Why are there two surveys in the jobs report?
- When is the jobs report reference period?
- Why can the jobs report be revised?
- Does a seasonal adjustment mean the economy improved?
- Why does the unemployment rate differ from the share of people without a job?
- How should readers interpret a small change in payroll employment?
- What to Do With the Jobs Report
How the Jobs Report Is Put Together at a Glance
The report is assembled from two surveys that answer different questions about the same labor market, then layered with seasonal adjustment and a revision process that never fully closes.
| Feature | Establishment survey (CES) | Household survey (CPS) |
|---|---|---|
| What it surveys | Businesses and other employers | Households and the people in them |
| Sample size | About 119,000 establishments covering hundreds of thousands of worksites | About 60,000 eligible households |
| Counts | Jobs, not people | People, not jobs |
| Reference period | The pay period that includes the 12th of the month | The reference week that includes the 12th of the month |
| Headline measure | Total nonfarm payroll employment | Unemployment rate |
| Strength | Detailed by industry, state and occupation | Covers people the payroll survey misses, such as gig and self-employed workers |
| Weakness | Misses unpaid and informal work | Smaller sample, so noisier month to month |
Which U.S. Agency Produces the Jobs Report?
The Bureau of Labor Statistics, a statistical agency within the Department of Labor, produces the report. The employer-side data comes from the Current Employment Statistics program, commonly called CES, and the household data comes from the Current Population Survey.
The CES program is a sample survey, not a census of every employer in the country. That single design choice explains most of what follows: the estimates carry sampling error, they are estimated before they are published, and they are revised when better information arrives.
What Does the Jobs Report Measure?
Five numbers get quoted constantly, and they do not all come from the same survey. Knowing which is which prevents most of the confusion in news coverage.
Total nonfarm payroll employment
This is the headline jobs number, and it comes from the establishment survey. It counts payroll jobs at private businesses, government agencies, nonprofits and the military, excluding farm work, private household employment and the self-employed.
The unemployment rate
This one comes from the household survey. It is the share of the civilian labor force without a job who is actively looking for one. People not looking are not counted as unemployed at all.
Labor force participation rate
Also from the household survey, this is the share of the civilian population 16 and older who are either employed or actively seeking work. A falling participation rate can hold the unemployment rate steady even as hiring slows.
Average weekly hours and average hourly earnings
These come from the establishment survey, which captures pay for jobs held by covered workers. Median usual weekly earnings for full-time wage and salary workers come from the household side.
Job openings and quits
These belong to separate BLS programs, the JOLTS survey, released on a different schedule. They are not part of the monthly jobs report even when articles bundle them together.
How Does the Establishment Survey Count Jobs?
The establishment survey counts jobs, not people, using payroll records employers submit for the pay period that includes the 12th of the month. That single fact has consequences people rarely hear explained.
Sample selection and coverage
Employers are selected through probability sampling and grouped into industries, with nearly every employer in a sample industry eventually asked to report. Small businesses, government agencies and nonprofits are all in scope. Participation is mandatory for sampled employers under federal law, and BLS follows up with non-respondents rather than simply recording zeros for them.
Why the 12th of the month decides the numbers
Fixing the reference period to the pay period containing the 12th gives every establishment a comparable counting date, whether that employer pays weekly, biweekly or twice a month. Without a fixed date, a payroll cut that landed on a different day each month would swing the totals for reasons that have nothing to do with hiring.
Why job switchers get counted twice
Because the unit is a job, a payroll record, someone who changes employers during the reference period can show up as one job at the old employer and one at the new one. Both payrolls report it, so total payroll employment rises even though the number of people with jobs did not. The effect is small in any given month and self-corrects over time, but it is real.
From sample to national total
Responses are summed and weighted up to national totals using the sampling design, with non-responding establishments imputed. The result is an estimate, published first as a preliminary figure for the reference month.
How Does the Household Survey Measure Unemployment?
The Current Population Survey interviews about 60,000 households about what the people in them did during the reference week, the week that includes the 12th of the month. Everyone in the household 16 and older is classified.
The three classifications
A person is employed if they did any work for pay or profit during the reference week, including as an employee or in their own business, or if they had a job but were temporarily absent. A person is unemployed if they had no job, were available to work, and had actively looked for work in the past four weeks. Everyone else is outside the labor force, which includes students, retirees, homemakers and anyone who stopped looking for work.
Why the unemployment rate is not the share of adults without a job
This is the most common misreading of the report. The unemployment rate uses the labor force as its denominator, not the adult population. Someone who stopped searching months ago and never resumed is counted as out of the labor force, so a weak job market can push the unemployment rate down as people exit the count.
Why the two surveys can disagree
The payroll survey counts jobs from employer records. The household survey counts people and includes workers the payroll survey never sees, such as self-employed contractors, unpaid family workers and people working without a formal arrangement. When employment growth is concentrated among those workers, payrolls and household employment can move in different directions, and the two unemployment rates will not match.
How Are Seasonal Adjustments Made?
Seasonal adjustment removes patterns that repeat every year so the adjusted change is closer to what is actually happening. Without it, a December-to-January comparison would look like a collapse when it is just holiday staffing ending.
Which industries get adjusted
Adjustment is applied by industry, and the sectors that move most are the ones that matter most: retail, education and health, leisure and hospitality, construction and tourism. Any recurring pattern in the unadjusted data, such as summer hires at resorts or the back-to-school cycle, gets its own factor.
How the factors are estimated and updated
Factors are estimated using a statistical model over long historical series and then recalculated every year as new data arrives. A revised factor changes how past months are presented, which is why seasonal adjustment is itself a source of revision rather than a fixed correction.
An adjusted number is an estimate of underlying trend, not a more accurate observation. That distinction matters, and the rest of this guide builds on it.
Why Do the Jobs Report Numbers Change After Release?
Revisions are a designed part of the system, not a sign that something was fabricated. The first print is made with incomplete data; better data replaces it. In one well-documented case, the preliminary estimate proved to be off by 911,000 jobs for the year before March 2025, announced on September 9, 2025.
The two monthly revisions
The second release revises the most recent month once more, and the third release revises the two prior months. These usually come from late payroll reports and corrected employer records.
The annual benchmark revision
Once a year, CES estimates are re-based to the Quarterly Census of Employment and Wages, a near-complete count compiled from state unemployment insurance tax records. The gap between the two becomes a correction spread across recent months. The 911,000-job figure cited above was the product of such a benchmark. An earlier round produced an 818,000-job adjustment covering March 2023 through March 2024.
Other revision sources
Businesses open, close, grow and get reclassified into different samples. Modeled estimates for business births and deaths, often called the birth-death model, get replaced as real employer data arrives. And a smaller monthly change, 258,000 jobs across May and June 2025 in one widely reported revision, came largely from updated seasonal factors rather than from employers being miscounted.
How to tell a revision apart from real change
A revision rewrites history for a month that already happened. A change in the current labor market shows up as a new direction in the newest numbers. When both appear in the same week, headlines usually blur them together, which is exactly when the distinction is worth keeping straight.
How Is the Jobs Report Published?
The Employment Situation is released on the first Friday of every month at 8:30 a.m. Eastern, covering data from the pay period that included the 12th of the previous month. That is roughly a three-week window between counting and publication.
Why the lag exists
Payroll records have to be collected, edits applied, non-responses chased, weights applied and seasonal factors run. Agencies choose the schedule knowing that a faster report would mean a smaller or less timely sample. The trade-off is deliberate, which is the honest answer to the common question of why the agency does not simply wait for complete data: waiting months longer would make the report describe an economy that has already moved on.
What to check in the release itself
Beyond the headline number, the technical notes list the survey response rate, the sampling error intervals, any unusual industry movements, and the current revision status of recent months. Those notes are where the report explains itself, and they are where anyone seriously interested should start.
What Does the Jobs Report Not Tell You?
The report describes employment counts and earnings. It does not describe job quality, and it should not be read as if it did.
It cannot tell you whether a job is full-time or a 20-hour arrangement, whether it comes with benefits or requires unpredictable shifts, or whether a worker wants the hours they have. It does not report vacancies, which come from a separate survey, or remote work arrangements. It does not explain why an industry added or lost positions, and it cannot identify labor shortages or the causes of a business cycle.
It also does not cover the self-employed or undocumented workers the payroll survey cannot see, which is one reason the household survey exists alongside it.
Key Limits and Revisions of the Jobs Report
Read the report as a signal inside a range, not a scoreboard. A few limits are worth keeping in mind every month.
Sampling error is built into every figure
Both surveys sample a fraction of the population. For the monthly change in total nonfarm payrolls, the 90 percent confidence interval runs on the order of 130,000 jobs. A headline gain of 100,000 is inside that interval, which means the direction is genuinely uncertain. A change of 300,000 is not.
Small monthly changes often mean nothing on their own
Single prints are noisy. Reading one month’s number and reacting to it is how both readers and markets talk themselves into mistakes. The trend across three months or more carries the signal.
Coverage and model assumptions matter
Estimate quality depends on who answers. Employer participation has weakened over time, and missing establishments are filled in using models rather than observation. A model is a reasonable approximation, not a fact, and the gap widens when the economy changes faster than the model was built for.
Interpretation, not just data
Payroll growth near zero with a falling unemployment rate usually means more people exited the labor force rather than that hiring improved. Reading the two surveys together, in the technical notes, is the minimum diligence the report deserves.
Frequently Asked Questions
Why are there two surveys in the jobs report?
They measure different things. The establishment survey counts payroll jobs from employer records and produces the headline nonfarm payrolls number. The household survey counts people, classifies each adult as employed, unemployed or outside the labor force, and produces the unemployment and participation rates. Using both gives a fuller picture than either could provide alone.
When is the jobs report reference period?
The establishment survey reference period is the employer pay period that includes the 12th of the month. The household survey reference week is the week that includes the 12th. Fixing the date to the 12th makes every employer comparable regardless of pay schedule, and it is why the report describes a period that is already three weeks old by publication.
Why can the jobs report be revised?
Because the first print is an estimate built from incomplete data. Later payroll reports arrive, employers correct their records, new businesses enter the sample, seasonal factors get updated, and the annual benchmark re-bases the estimates on unemployment insurance tax records. Revisions replace early guesses with better information about periods that have already happened. They are not a statement about the current labor market.
Does a seasonal adjustment mean the economy improved?
No. Seasonal adjustment removes patterns that repeat every year, such as holiday hiring or summer tourism work, so the reported change is closer to underlying movement. The adjusted number is still an estimate produced by a model, and because factors are recalculated annually, adjusting an old month can change its value without anything in the real economy having changed.
Why does the unemployment rate differ from the share of people without a job?
The unemployment rate counts only people without a job who are available and actively looking for work. Adults who have stopped searching are classified as outside the labor force and are excluded from the denominator. So the rate can fall in a weak labor market simply because people stopped looking. It is not the share of adults without a job, and the two should never be used interchangeably.
How should readers interpret a small change in payroll employment?
Treat it as inconclusive. The 90 percent confidence interval on the monthly change in total nonfarm payrolls is roughly 130,000 jobs, so a gain or loss smaller than that is within normal sampling variation. Look for a consistent direction across several months, check whether participation and unemployment moved the same way, and read the technical notes for anything unusual before drawing a conclusion from one print.
What to Do With the Jobs Report
Start with the reference period, not the headline. Knowing you are reading about the pay period that included the 12th tells you how old the information is before you interpret any of it.
Next, read the two surveys together. Payroll growth of zero alongside a falling unemployment rate means something different from payroll growth of zero alongside a rising participation rate. Then look at three months rather than one, check the technical notes for anything flagged, and treat any large revision as a better estimate of the past rather than news about today.
Understanding how the jobs report is put together does not make the numbers predictable. It does make them legible, which is the part most coverage leaves out. Updated for 2026.
