Non-Farm Payrolls: What the NFP Report Actually Measures

One US data release is described more often than any other, and usually in the same three sentences: it lands on the first Friday, it measures jobs, and it moves the dollar. Payroll employment is also one of the monthly measures used to date the economic cycle, which is why the series carries weight beyond the day of the release.

The first is frequently untrue and the second is less precise than it sounds. What follows is the structure of the release as its publisher describes it, including the margin of error printed alongside it and the revisions that change the number after the market has traded it.

What the Non-Farm Payrolls Report Actually Measures

Non-farm payrolls is not a standalone report. It is one series inside the Employment Situation news release, published monthly by the US Bureau of Labor Statistics.

That release presents statistics from two separate surveys. Payrolls comes from the establishment survey, which collects data each month from the payroll records of a sample of non-agricultural businesses.

BLS describes the scale of that sample precisely: about 119,000 businesses and government agencies, representing approximately 622,000 individual worksites, an active sample covering roughly 26 per cent of all non-farm payroll jobs.

Employees on non-farm payrolls are those who worked or received pay for any part of the reference pay period, including people on paid leave.

The reference period is the detail that changes how the number should be read. For the establishment survey it is the pay period including the 12th of the month, which may or may not line up with a calendar week.

A release published in the first week of a month is therefore describing employer payrolls around the middle of the previous month, not conditions on the day it appears.

One further mechanical point explains a great deal about the report later on. People are counted in each job they hold, so the same person on two payrolls contributes two jobs to the total.

Why the First Friday Rule Is Not Reliable

Almost every trading guide states that the report is released on the first Friday of the month. The publisher’s own calendar contradicts that often enough to make the habit expensive.

The BLS schedule of releases for the Employment Situation lists twelve releases falling in calendar 2026. Four of them are not on the first Friday.

The remaining eight do fall on a first Friday, so the rule of thumb is right about two thirds of the time. That is the hit rate that makes it dangerous, because it works often enough to feel dependable.

Every release on the schedule is timed at 08:30, and the BLS calendar states that all times shown are Eastern Time. The time is far more stable than the date.

The remedy is small: check the published schedule for the month rather than the day of the week. Our guide to the economic calendar covers how these entries are structured.

The Confidence Interval That Changes How You Read the Number

The establishment survey is a sample, not a census, and BLS publishes the resulting margin of error in the same technical note. The figure is larger than most traders expect: the confidence interval for the monthly change in total non-farm employment is on the order of plus or minus 122,000.

BLS then works the example itself. If the estimate increases by 50,000 from one month to the next, the 90 per cent confidence interval on that change runs from minus 72,000 to plus 172,000, and because that range includes values below zero the agency states it could not be said with confidence that employment had increased at all.

The contrast it draws matters just as much. If the reported rise were 250,000, every value inside the interval would be greater than zero, and it becomes likely that employment genuinely rose.

None of this makes the release irrelevant. Prices move because participants act on the headline, and that reaction is real whether or not the underlying difference is statistically meaningful.

What it changes is the interpretation. A small beat is evidence about positioning and expectations, not, on its own, evidence about the labour market.

The unemployment rate carries its own band. BLS states that at an unemployment rate of around 6.0 per cent, the 90 per cent confidence interval for the monthly change in the rate is about 0.3 percentage point, which places a one-tenth move inside the noise.

Payrolls and the Unemployment Rate Come From Different Surveys

The two headline numbers in the release are routinely discussed as though they measure the same thing from two angles. They do not, and the differences are structural.

The household survey is a sample of about 60,000 eligible households conducted by the US Census Bureau for BLS. It classifies each person aged 16 and over as employed, unemployed, or not in the labour force, and it is the source of the unemployment rate. The establishment survey asks employers about payrolls instead.

Those differences are enough to produce a report in which payrolls rise and the household measure of employment falls, without either survey being wrong. A month in which second jobs increase raises payrolls without adding a single employed person to the household count, and a month in which self-employment grows does the reverse.

Commentary that treats a divergence between the two as a contradiction is usually describing a definitional gap that BLS documents openly.

The Figure Is Revised After You Have Already Traded It

The number that moves the market is a first estimate, and BLS is explicit that it is not final.

Estimates for the two most recent months are based on incomplete returns, and for that reason they are labelled preliminary in the tables. BLS states that it is only after two successive revisions, when nearly all sample reports have been received, that an estimate is considered final.

Seasonal adjustment adds a second source of movement. The establishment survey uses concurrent seasonal adjustment, calculating new factors each month and using them to adjust the three most recent monthly estimates.

So a single release can change the current month, the previous month and the month before that, for two reasons at once: more sample reports arriving, and recalculated seasonal factors.

A third layer operates on a longer cycle. Sample-based estimates are adjusted once a year, on a lagged basis, to universe counts of payroll employment from the administrative records of the unemployment insurance programme.

BLS calls the difference between the March sample-based estimates and the March universe counts a benchmark revision, and describes it as a rough proxy for total survey error. The absolute average over the prior ten years is 0.2 per cent, ranging from minus 0.4 to plus 0.3 per cent.

One more component sits inside the monthly estimate. To account for jobs at new firms the sample cannot capture in time, BLS uses a two-part net birth-death procedure, part imputation and part an ARIMA model built from unemployment insurance records over the past five years.

What Changes in the Market Around the Release

The mechanical effects of a scheduled release are more predictable than its direction, and they are the part a trader can prepare for.

Liquidity thins in the moments before a fixed-time release and the quoted spread widens, because the firms providing prices face the same uncertainty as everyone else. The gap between the price requested and the price filled is covered in our guide to slippage.

Order behaviour changes too. A stop order becomes a market order once triggered, so it is filled at whatever is available rather than at the level written on it, which is why the choice between order types matters more around a release than at any other time.

Position size is the variable that remains under your control when spread, execution and direction are not. The method for setting it is covered in our guide to position sizing.

Why the report is watched at all comes down to what it feeds into. Employment data is one input to the central bank policy stance, and the reaction usually reflects a shift in rate expectations rather than a view about employment itself.

Who This Page Is Not For

This page does not offer a way to trade the release. It names no expected outcome for any month, gives no entry or exit level, and describes no strategy for positioning before or after publication.

Anyone looking for a rule that converts a beat or a miss into a direction will not find one here, because the relationship between the surprise and the price reaction is not stable enough to state as one.

The material is aimed at a reader who wants to know what the number is, how precise it is, and how it changes afterwards.

Frequently Asked Questions