How Is the Jobs Report Calculated? Surveys, Payrolls, Revisions

The monthly jobs report, formally called the Employment Situation, is calculated from two separate surveys run by the Bureau of Labor Statistics: a household survey of about 60,000 households that produces the unemployment rate, and an establishment survey of roughly 119,000 businesses and government agencies that produces the headline nonfarm payroll number.¹ The raw responses from those surveys then pass through statistical models, seasonal adjustments, and eventual revisions before the numbers you see in the news arrive. Understanding how the jobs report is calculated means following that pipeline from survey response to published figure.

The Two Surveys That Produce the Numbers

The household survey, formally the Current Population Survey, is run jointly with the Census Bureau. Field representatives contact roughly 60,000 households each month by phone and in person, asking about work activity during a specific reference week, usually the calendar week containing the 12th of the month.¹ This is where the unemployment rate, the labor force participation rate, and demographic breakdowns come from. Because it surveys people rather than employers, it picks up the self-employed and gig workers who never appear on a formal payroll.²

The establishment survey, or Current Employment Statistics program, takes the opposite approach. It asks employers about payrolls. The CES collects data from approximately 119,000 businesses and government agencies covering about 622,000 individual worksites.³ Employers report how many workers appeared on their payrolls, hours those workers were paid for, and what they earned. This is the survey behind the nonfarm payroll figure that dominates release-day coverage.

Both surveys reference the same time window. Neither alone tells the full story, and the two can point in different directions in any given month.

How the Unemployment Rate Is Calculated

The household survey sorts every person 16 or older into one of three mutually exclusive groups: employed, unemployed, or not in the labor force. You count as employed if you did any paid work during the reference week, even a single hour. People temporarily away from a job because of vacation, illness, or a family obligation also count as employed.

To count as unemployed, three conditions all have to hold: you had no job during the reference week, you were available to work, and you made at least one active effort to find a job in the prior four weeks.⁴ “Active” matters. Browsing job listings doesn’t count. Sending a resume, filling out an application, or contacting an employer does. People on temporary layoff who expect to be recalled qualify as unemployed even without a fresh search.

Everyone else falls into the third bucket. Retirees, full-time students not looking for work, stay-at-home parents, and people who have given up searching all land there. Whether someone receives unemployment insurance has nothing to do with how they’re classified; the survey never asks about benefits.

The arithmetic from there is simple. The labor force equals employed plus unemployed. The official unemployment rate, known as U-3, is unemployed divided by the labor force, expressed as a percentage.⁴

Two related figures from the same survey fill in what U-3 misses. The labor force participation rate is the share of the civilian population aged 16 and older that is either working or actively looking.⁵ The employment-population ratio divides employed people by the total civilian noninstitutional population.⁴ A falling unemployment rate paired with a falling participation rate can mean people are leaving the job market rather than finding work.

The Six Measures of Labor Underutilization

The BLS publishes six unemployment measures, U-1 through U-6, each broader than the last. U-3 is the middle one and the headline. U-1 counts only people unemployed for 15 weeks or longer. U-2 limits the count to job losers and people who finished temporary positions. U-4 adds discouraged workers who stopped searching because they believe no jobs are available. U-5 adds everyone marginally attached to the labor force. U-6 is the broadest and includes part-time workers who want full-time hours but can’t get them.⁴ U-6 routinely runs several percentage points above U-3, and the gap gives a sense of hidden slack even when the headline rate looks low.

How Nonfarm Payrolls Are Counted

The establishment survey counts jobs, not people. Hold two positions at different companies and you appear twice in the payroll data. That’s intentional. The metric measures labor demand from the employer side rather than individual employment status.

Several categories of workers are excluded from the count. Farm workers are left out because agricultural employment swings wildly with growing seasons and follows different tax reporting rules. Private household employees such as nannies and housekeepers, and unincorporated self-employed workers, also fall outside the count because they don’t appear on establishment payrolls. The survey covers only workers who receive a paycheck from a business or government agency.³

The Birth-Death Model

New businesses open and existing businesses close every month, but the CES sample frame can’t capture brand-new companies in real time. It takes months for a new employer to show up in unemployment insurance tax records, which form the survey’s sampling frame. To account for jobs at businesses too new to be surveyed, the BLS adds a birth-death model adjustment to the raw payroll count.⁶

The model compares historical Quarterly Census of Employment and Wages data with and without business openings and closings. The difference between those two data sets represents the net impact of births and deaths. The BLS feeds those historical patterns into an ARIMA time series model to forecast the current month’s adjustment. Starting with preliminary January 2026 estimates, the BLS modified the model to incorporate current sample data into the forecasts, adding a regression component that uses the relationship between the survey’s own measure of employment change and the historical birth-death patterns.⁶

This adjustment is one of the most scrutinized parts of the report. It can add or subtract tens of thousands of jobs from the headline figure each month based on a statistical forecast rather than actual survey responses. It performs reasonably well in stable periods and can miss turning points badly. When the economy shifts suddenly, the historical patterns the model relies on may no longer apply.

Seasonal Adjustment

Employment follows predictable annual rhythms. Retailers staff up for the holidays, schools hire in the fall and release temporary workers in the summer, and construction slows in cold-weather months. Reporting raw numbers would turn a normal January retail drop into an apparent crisis. Seasonal adjustment strips out these expected patterns so month-to-month changes reflect actual shifts in economic conditions.⁷

The CES program uses X-13ARIMA-SEATS software developed by the Census Bureau to perform the adjustment.⁸ The software builds a statistical model of seasonal patterns from years of historical data, estimates how much of the current month’s change is seasonal, and removes that portion. Both surveys undergo seasonal adjustment, each using models tailored to its own data series.

The adjustment is never perfect. Unusual weather, shifted holiday calendars, or one-time events such as a government shutdown can distort the seasonal factors. The BLS updates its seasonal models annually and publishes both adjusted and unadjusted figures.

Why the Numbers Have a Margin of Error

Every number in the Employment Situation report is an estimate based on a sample, not a census of every worker and employer in the country. The establishment survey’s monthly change in total nonfarm employment carries a 90% confidence interval of roughly plus or minus 122,000 jobs.⁹ A reported gain of 130,000 jobs could plausibly represent an actual change anywhere from about 8,000 to 252,000.

That uncertainty is worth keeping in mind when markets react sharply to a single month’s data. A report that misses expectations by 40,000 jobs is well within normal statistical noise. The household survey has its own sampling error, and the unemployment rate can move a tenth or two of a percentage point from sampling variability alone. Trends over several months are always more informative than any single release.

How Revisions Change the Initial Numbers

The initial release contains preliminary estimates because not all employers submit their payroll data before the early deadline. Each month’s payroll figure gets revised in the following two releases as late responses arrive. These monthly revisions are normal, but they can be substantial enough to shift the narrative around a given month.

Once a year, the BLS recalibrates the entire payroll series against a near-complete count of employment drawn from state unemployment insurance tax records, which nearly all employers are required to file.¹⁰ This Quarterly Census of Employment and Wages data covers about 96% of all nonfarm employment, making it the closest thing to a true headcount available.

The benchmark revision adjusts the previous year’s monthly estimates to align with the administrative data, recalculates seasonal factors, and revises the seasonally adjusted series going forward. These corrections are typically published alongside the January data release. The most recent benchmark, released in February 2026, revised total nonfarm employment for March 2025 downward by 898,000 jobs.¹¹ A revision that size is unusually large, and it illustrates why the preliminary monthly numbers, useful as they are for tracking trends in real time, are never the final word.

1
U.S. Bureau of Labor Statistics. Monthly Employment Situation Report – Quick Guide to Methods and Measurement Issues
2
United States Census Bureau. Sampling
3
U.S. Bureau of Labor Statistics. Current Employment Statistics Overview – CES (National)
4
U.S. Bureau of Labor Statistics. Concepts and Definitions (CPS)
5
U.S. Bureau of Labor Statistics. Employment Projections – Concepts
6
U.S. Bureau of Labor Statistics. CES Birth-Death Model Frequently Asked Questions
7
U.S. Bureau of Labor Statistics. Seasonal Adjustment Methodology for National Labor Force Statistics From the CPS
8
U.S. Bureau of Labor Statistics. CES Seasonal Adjustment Technical Notes
9
U.S. Bureau of Labor Statistics. Employment by Industry, Monthly Changes, With Confidence Intervals
10
U.S. Bureau of Labor Statistics. Revisions in State Establishment-Based Employment Estimates
11
U.S. Bureau of Labor Statistics. CES National Benchmark Article