Foundations4 min read

The Jobs Report

Monthly employment data from two separate surveys that regularly disagree. Along with CPI it is the highest-impact scheduled release of the month.

MadStockAlerts Research · Updated August 28, 2026

What to take away

  • Released on the first Friday of the month at 8:30 a.m. ET.
  • Nonfarm payrolls come from a survey of employers; the unemployment rate from households.
  • The two surveys measure different things and frequently tell different stories.
  • Revisions to prior months are routine and sometimes larger than the headline.
  • How equities react depends on which risk the market is currently pricing.
  • The two surveys in the release measure different populations and can point opposite ways without either being wrong.

MAD Academy Training Video · 0:46

Two Surveys That Disagree

The payrolls number and the unemployment rate come from two different surveys, which is why they sometimes tell opposite stories.

This lesson is part of a Stock Alerts + Tools plan.

See the library

Two surveys, one release

Establishment surveyHousehold survey
AsksEmployersHouseholds
ProducesPayroll jobs addedUnemployment rate, participation rate
CountsJobs, so one person with two jobs counts twicePeople
IncludesPayroll employees onlyThe self-employed and agricultural workers
Sample sizeLarge; less volatileSmaller; noisier month to month

The two disagreeing is not an error. Strong payroll growth alongside a rising unemployment rate is entirely coherent if the labour force grew faster than hiring, and that combination appears regularly.

What is read beyond the headline

  • Revisions to the previous two months, which can exceed the current month's figure.
  • Average hourly earnings, read as a wage-inflation input.
  • The participation rate, which determines whether a falling unemployment rate reflects hiring or people leaving the workforce.
  • The composition of gains by industry, and how much came from a single sector.
  • The underemployment measure, which counts part-time workers who want full-time work.

Average hourly earnings is the line that most often determines the market reaction when the headline is close to expectations, because it is the one that speaks to inflation rather than to growth.

Why revisions matter

Initial estimates are based on incomplete survey responses and are revised twice as more arrive, then again in an annual benchmark. Revisions of a hundred thousand jobs are common.

That means the number that moved markets was frequently not the number that turned out to be true. It also means a sequence of downward revisions is itself a signal, and one that is visible only by tracking the revisions rather than the headlines.

The number that moved the market, and the number that was true
The number that moved the market, and the number that was true0100200300Revised down by nearly a third, twomonths laterMarAprMayJunJulPayrolls, thousands

Scroll the chart sideways to see all of it.

  • First print
  • After two revisions
The headline is a first estimate from an incomplete survey. It is revised twice, and by then nobody is trading it. Illustrative figures.

The good-news-is-bad-news regime

How equities respond to strong employment data depends on the regime. When the concern is recession, strong data is welcomed. When the concern is inflation and tight policy, the same data implies rates staying higher for longer and equities can fall on it.

The number is identical; the interpretation is set by what the market is currently worried about. Establishing which regime is in force before the release is the only way to have any view about the direction of the reaction.

Why the two surveys disagree

The headline payroll figure and the unemployment rate come from different surveys of different populations, conducted differently, and they can point in opposite directions in the same release without either being wrong.

Establishment surveyHousehold survey
Who is askedAround 120,000 businesses and agenciesAround 60,000 households
What it producesPayroll jobs added, hours, average earningsThe unemployment rate, participation, self-employment
Counts a person twice ifThey hold two payroll jobsNever; it counts people, not jobs
MissesThe self-employed and unincorporated businessesNothing structurally, but the sample is far smaller
RevisedTwice, then annually against tax recordsNot routinely

A period in which payroll jobs rise while the household survey shows employment falling is usually explained by multiple job-holding, by growth in self-employment that payrolls do not capture, or by sampling noise in the smaller survey. The disagreement is informative in itself, and it recurs at turning points.

The birth-death model and the annual benchmark

New businesses do not appear in the establishment survey sample until they are found, and closing businesses stop responding without saying why. The Bureau of Labor Statistics fills that gap with a statistical model estimating jobs created and destroyed by business formation and closure.

The model works reasonably in stable conditions and poorly at turning points, which is exactly where the number matters most. At the start of a downturn it tends to add jobs from assumed business formation that is not occurring, and the error is corrected later.

The correction arrives in the annual benchmark revision, when the survey is reconciled against state unemployment insurance tax records covering nearly all employment. Those revisions have been large enough in some years to change the description of what the labour market was doing, long after every market reaction to the original prints.

None of this makes the monthly figure useless. It makes it an early estimate with a known error structure, and treating a single month's print as a measurement rather than as an estimate is the error the revision history warns against.

Educational content only. MadStockAlerts provides market commentary, research, and educational content. It is not personalized investment advice, and nothing here is a recommendation to buy or sell any security. Trading and investing involve substantial risk, including loss of capital. See the Risk Disclosure and Customer Agreement.