Nonfarm payrolls and the jobs report
One report, two surveys, and they routinely tell different stories. Knowing which one produced which headline is most of what it takes to read a jobs day.
At a glance
- Published by
- US Bureau of Labor Statistics, monthly
- Released
- Usually the first Friday, 8:30am ET
- The headline
- Change in nonfarm payroll employment, in thousands
- Revisions
- The prior two months are revised with every release
One report, two different surveys
The Employment Situation report contains two independent measurements taken by different methods, and almost every confusion about jobs day comes from mixing them up.
The establishment survey asks businesses how many people were on their payroll. It produces the headline payrolls number and average hourly earnings. It counts jobs, not people - somebody with two jobs is counted twice - and it excludes farm work, the self-employed and unpaid family workers, which is what the word nonfarm is doing there.
The household survey asks people about their own employment. It produces the unemployment rate, the participation rate and the count of people employed. It counts people, includes the self-employed, and is drawn from a far smaller sample, which makes it noisier month to month.
Because they measure different things with different methods, they can and do diverge for months at a time. Neither is wrong when that happens; they are answering different questions.
Payrolls come from businesses and count jobs. The unemployment rate comes from households and counts people. Same report, different surveys.
Who acts on it
The jobs report is the labour half of the Federal Reserve's mandate arriving in a single number, which is why it is the second most market-moving US release after CPI. A weakening labour market is historically the fastest route to rate cuts; a hot one keeps policy tight.
Within the report, the market's attention has shifted over time. In a growth-anxious cycle the headline payrolls figure dominates; in an inflation-anxious one, average hourly earnings can matter more, because wage growth is what makes services inflation sticky.
- Nonfarm payroll changeEstablishment survey, the headline
- 45%
- Unemployment rateHousehold survey
- 25%
- Average hourly earningsEstablishment survey
- 15%
- Revisions to prior months
- 10%
- Participation and hours
- 5%
Source: US Bureau of Labor Statistics release structure
How the number is built
The establishment survey covers roughly 120,000 businesses and government agencies, a very large sample by survey standards, which is why the payroll number is comparatively stable. The household survey covers about 60,000 households - large for a household survey, small next to the establishment one, and the reason the unemployment rate jumps around.
Two mechanics inside the establishment survey regularly become news in their own right. The birth-death model estimates jobs created and lost at businesses too new or too recently closed to be in the sample; it is a model, not a count, and at turning points it is systematically wrong in a knowable direction. And the sample fills in over time, which is why the first estimate of a month gets revised twice.
Once a year, the whole series is benchmarked against near-universal unemployment insurance tax records. That revision is much larger than the monthly ones and has repeatedly rewritten the story of an entire year after the fact.
- Establishment survey: about 120,000 businesses, counts jobs.
- Household survey: about 60,000 households, counts people.
- Every release revises the prior two months.
- An annual benchmark revision reconciles the series with tax records.
Behind the subscription
The rest of this entry is the part that changes a decision: what moves the price, which contract sets it, who ships it and where that can be cut off.
What moves the print
The seven forces behind a payrolls surprise - from seasonal adjustment and the birth-death model to strikes, weather and government hiring.
Where the jobs view is priced
The instruments that carry a labour-market view, and the private data releases that trade as previews of the official one.
How the report travels
The order in which a jobs number reprices markets, and why the second paragraph of the release sometimes matters more than the first.
How this shows up in prediction markets
Which figure settles a payrolls market, why the revision policy is the fine print that decides it, and the checks that beat the consensus.
Included with a subscription
Create an account to unlock the full entry — price drivers, trading venues, trade flows and the live markets attached to it.
Frequently asked questions
- Why do payrolls and the unemployment rate sometimes disagree?
- Because they come from two different surveys. Payrolls count jobs at businesses; the unemployment rate counts people in households and includes the self-employed. Different samples, different definitions, and they can point opposite ways for months without either being wrong.
- What is the birth-death model?
- An estimate of jobs at businesses too new or too recently closed to appear in the survey sample. It is a model rather than a count, and at economic turning points it is systematically off in a predictable direction - which the annual benchmark revision then corrects.
- Why does the jobs report get revised so much?
- The first estimate is published before all responses arrive, so the following two releases update it as the sample fills in. Once a year the whole series is benchmarked against unemployment insurance tax records, which is a much larger correction.
- Which week does the report actually measure?
- The pay period containing the twelfth of the month. Events outside that window - a strike or a hurricane the following week - do not appear in the print at all, which is a frequent source of confused commentary.
Primary sources
Related entries
Prediction markets carry real risk of loss. Nothing on Market Guy is financial advice — it is research tooling to help you think, not a signal to trade.