The Jobs Report Explained: Payrolls, Unemployment and Wages

The Bureau of Labor Statistics writes, most months, that the unemployment rate “was little changed” at whatever it printed. That phrase is a statistical statement. The household survey is drawn from a sample, the sample carries error, and the technical note published with every release names the confidence interval around each figure, which is wide enough that a tenth of a percentage point frequently cannot be distinguished from no movement at all. The agency is telling readers to ignore it. Readers make something of it anyway.

That tension runs through the whole release, which is published at 8:30 am Eastern, usually on the first Friday of the month, and which is the most closely watched monthly data in the United States. Most of the confusion around it dissolves once you know that the headline jobs number and the headline unemployment rate come from two separate surveys that are under no obligation to agree.

Two surveys, one release

	Establishment survey	Household survey
Who is asked	Businesses and government agencies	Households
Headline output	Nonfarm payroll change	Unemployment rate
Counts a person with two jobs	Twice	Once
Counts the self-employed	No	Yes
Also produces	Hours worked, average hourly earnings, industry detail	Participation rate, U-6, reasons for unemployment, demographics
Sample size	Much larger	Much smaller, so noisier month to month

The establishment survey asks employers how many people were on the payroll for the pay period including the twelfth of the month, and it is the source of the figure reported as the economy having added so many jobs. It works from payroll records. A person with two jobs counts twice. The self-employed do not appear at all.

The household survey asks people about their own situation, and it exists because for most of American history nobody could say how many people were out of work. Estimates of unemployment during the Depression varied by millions depending on who was doing the counting, which is an uncomfortable position for a government making policy, and a regular national survey of households was established around 1940 to settle the question. It remains the only source of the unemployment rate, the participation rate and the broader measures of underemployment, and its sample is far smaller than the payroll survey, so its monthly movements are noisier. Never read one month of it alone.

The rate, worked from the numbers

The calculation is simple. The simplicity is exactly where it misleads.

unemployment rate = unemployed / labor force

labor force = employed + unemployed

Unemployed means without a job, available for work, and having actively looked in the past four weeks, so someone who has stopped looking sits outside the labor force altogether, counted in neither the numerator nor the denominator.

Take an illustration. Suppose 161.0 million people are employed and 7.0 million are unemployed. The labor force is 168.0 million, and the rate is 7.0 / 168.0 = 0.0417, or 4.2 percent.

Now suppose 400,000 of those unemployed people give up the search. Nobody was hired. The unemployed count falls to 6.6 million and the labor force falls to 167.6 million, so 6.6 / 167.6 = 0.0394, or 3.9 percent.

The rate improved because people left. That is why participation is published beside it. Participation is the labor force divided by the civilian population aged 16 and over. A falling unemployment rate accompanied by falling participation is a weaker report than the headline suggests, and a rising unemployment rate accompanied by rising participation often means people are returning to look for work because they believe jobs exist.

U-3 is the headline rate above. U-6 adds people working part time who want full-time hours, along with those marginally attached to the labor force, and runs several points higher by construction, so the spread against U-3 tends to widen while the labor market is softening, before the headline rate has moved at all.

Wages, and why they belong to the inflation story

Average hourly earnings is reported as a monthly change and against the same month a year earlier, and the annual figure is the one worth watching, since the monthly series is noisy enough to reverse itself regularly.

Wage growth is where this release joins the inflation release. Services are labor-heavy, so when wage growth runs persistently above productivity growth, firms tend to pass the difference into prices, and that is the component of core inflation a central bank finds hardest to influence without damaging employment. A hot wage print can move Treasury yields as much as a hot payroll print, which surprises people the first time they see it.

One trap deserves naming. Average hourly earnings is an average. The mix of jobs moves it. A wave of lower-paid hiring pulls the average down while nobody’s pay was cut, and a wave of low-wage job losses pushes it up while nobody got a raise, an effect conspicuous in 2020, when the average wage jumped simply because the jobs lost first were disproportionately at the bottom of the scale.

Revisions, and the benchmark nobody reads

The first payroll estimate counts the businesses that had responded by the cutoff, more arrive late, so each release revises the two prior months, and those revisions are printed near the top of the document for anyone who reads that far.

Behind them sits a larger correction. Once a year the BLS benchmarks the whole payroll series against state unemployment insurance tax records, which cover nearly every employer, and the level of employment is restated accordingly. The gap being corrected comes largely from the model that estimates jobs created and destroyed by businesses too new or too recently closed to be surveyed, since a firm cannot be asked about its payroll before anyone knows it exists. That model works well in steady periods and poorly at turning points, which is precisely when people most want the number to be right.

Reading the release in order

Start with payrolls and the revisions together, as a single number, because separating them is how people talk themselves into a view the data does not support. Then the unemployment rate and participation, together. Then average hourly earnings year over year. Then the industry table, since gains concentrated in government and health care carry different information from gains in construction and manufacturing, the first being funded by budgets and the second by demand.

Average weekly hours is the quiet line. Employers cut hours before they cut people, because rehiring is expensive and reputationally awkward, so a workweek that has been shortening for several months is an early softening signal that reaches the headline count much later.

Good news as bad news, and when the sign flips

The market reaction depends entirely on what the market is worried about that month.

When the dominant worry is inflation and policy, a strong report is taken badly. Strong hiring means the economy can absorb tighter policy, expected rates rise, and higher discount rates lower every equity valuation through the arithmetic in how interest rates affect stocks, which is the familiar inversion in which good news is bad news, a fixture of the late 1990s as well as of the post-2021 tightening cycle.

When the dominant worry is a slowdown, the sign flips. A weak report raises the odds of cuts, and that support can outweigh the damage to expected earnings, though only up to a point, because once the market believes a downturn has actually begun, weak data stops being read as good news of any kind, and the two interpretations can swap places within a single quarter.

The front of the Treasury curve is the cleanest place to watch the reaction, since it prices the expected policy path directly and reprices within seconds, and the yield curve tool shows how the shape shifts from one week to the next.

Where it misleads

Seasonal adjustment does enormous work here. January and September carry huge seasonal factors because of post-holiday layoffs and the school year, and the adjusted figure can differ from the raw count by more than a million jobs, which means the number being discussed is substantially a modelling output. Strikes, hurricanes and government shutdowns distort individual months, and the BLS names them in the text of the release for anyone reading past the tables.

The unemployment rate lags by construction. Firms cut hours, then stop hiring, then lay off, so watching the rate for the beginning of a downturn amounts to watching for confirmation. That lag is the entire reasoning behind the Sahm rule and the other measures collected in recession indicators, and it is why a committee following the structure described in how the Federal Reserve works cannot simply wait for the employment data to tell it what to do.

Put this release in its monthly context with the economic calendar guide, or follow employment into the broader growth picture in GDP explained. The Fed and the economy quiz covers the two surveys and the rate arithmetic.