How an Economy Is Measured · Lesson 4 of 6
The labour market: two surveys, three numbers
One concept: the employment release's two surveys
Why it matters
The newest employment release on Kitalpha’s calendar was published by U.S. Bureau of Labor Statistics on 2026-09-04 for the reference period August 2026. Behind that one row sit two series: the unemployment rate, whose latest print is 4.1%, and nonfarm payrolls, a count of jobs in thousands. Headlines quote one or the other, and sometimes the two point in opposite directions in the same month. How can employment rise while the unemployment rate also rises? The answer is that the release is not one survey but two.
The concept
The monthly employment release from the Bureau of Labor Statistics is built from two surveys that ask different people different questions. The household survey samples households and asks each adult whether they worked in the reference week, and if not, whether they looked for work. From those answers come the labour force, which is everyone working or looking; the unemployment rate, which is the share of the labour force that is looking but not working; and the participation rate, which is the labour force as a share of the adult population. The establishment survey samples employers and asks how many jobs were on their payrolls in the same week. From it comes the payrolls figure, usually quoted as the change from the month before: jobs added or lost, in thousands.
The two surveys count different things. The household survey counts people; a person with two jobs is one employed person. The establishment survey counts jobs; that same person is two jobs. The household survey includes the self-employed and farm workers; the payrolls count leaves them out, which is why it is called nonfarm. So the figures are not two views of one number but two numbers, and they can diverge for ordinary reasons.
The unemployment rate is a ratio, and a ratio moves when either part moves. It falls when people find work, and it also falls when people stop looking, because someone who is neither working nor looking is not in the labour force at all. It rises when people lose jobs, and it also rises when more people start looking than find work, which is what happens when a growing population or a rising participation rate brings new entrants into the count. A month in which payrolls add jobs and the rate still rises is therefore not a contradiction; it says the labour force grew faster than employment did. Reading the rate alongside the participation rate resolves most such puzzles.
Two further habits make the release readable. First, each figure describes a reference week inside the reference month, not the day of the release; the calendar row carries the period and the date separately so the two are not confused. Second, both surveys are samples, so both figures carry sampling error and both are revised as later survey responses arrive; a payrolls change of a few tens of thousands is within the range of ordinary revision, and a later lesson takes up revisions. Weekly initial jobless claims, which count first-time filings for unemployment insurance, are a third source from administrative records rather than a survey, and they arrive every week rather than once a month.
The chart below plots the unemployment rate over the last five years. Look at how it moves: mostly in steps of a tenth of a point, occasionally in larger turns, and always as a line that a new release extends by one point rather than as a number that stands alone.
The table lists about every 3th observation; the chart plots all 60.
| Date | Unemployment Rate |
|---|---|
| 2021-08-01 | 5.10% |
| 2021-11-01 | 4.10% |
| 2022-02-01 | 3.90% |
| 2022-05-01 | 3.60% |
| 2022-08-01 | 3.60% |
| 2022-11-01 | 3.60% |
| 2023-02-01 | 3.60% |
| 2023-05-01 | 3.60% |
| 2023-08-01 | 3.70% |
| 2023-11-01 | 3.70% |
| 2024-02-01 | 3.90% |
| 2024-05-01 | 3.90% |
| 2024-08-01 | 4.20% |
| 2024-11-01 | 4.20% |
| 2025-02-01 | 4.20% |
| 2025-05-01 | 4.30% |
| 2025-08-01 | 4.30% |
| 2025-12-01 | 4.40% |
| 2026-03-01 | 4.30% |
| 2026-06-01 | 4.20% |
| 2026-08-01 | 4.10% |
Worked example
Take the two series behind the release row: the unemployment rate as of 2026-08-01 and nonfarm payrolls as of 2026-08-01. The steps below read each series’ latest print and the print before it, then take the difference: for the rate, a change in percentage points; for payrolls, a change in thousands of jobs. Every figure comes from the named record on the date shown.
Record: Employment Situation, August 2026 · as of · U.S. Bureau of Labor Statistics
- Unemployment rate in the latest print, from the rate record 4.1%
- Unemployment rate in the print before it 4.1%
- Change in the unemployment rate, in percentage points 0.00 pp A ratio from the household survey: people looking for work divided by everyone working or looking.
- Nonfarm payrolls in the latest print, thousands of jobs, from the payrolls record 159,075
- Nonfarm payrolls in the print before it, thousands 158,913
- Jobs added or lost in the month, thousands 162 A count from the establishment survey of employers: jobs, not people.
Read the two differences against each other. The third line is the household survey’s answer: the share of the labour force looking for work moved by that many percentage points, which can happen through hiring, through job losses, or through people entering or leaving the count of those looking. The sixth line is the establishment survey’s answer: employers reported that many thousand more or fewer jobs than the month before. If the two signs agree, the two surveys tell one story; if they disagree, the participation rate is the first place to look. Neither line says why, and neither line is the whole release.
Faded example
Now a weekly record from a third source: initial jobless claims, the number of people who filed a first claim for unemployment insurance in the week, as of 2026-09-05. The latest week and the week before are given. Complete the last step: the change in claims from one week to the next.
Second record: Initial Jobless Claims (seasonally adjusted) · as of
- Initial jobless claims in the latest week, from the claims record206,000
- Initial jobless claims in the week before207,000
- Tolerance ±1
Reveal the answer and the explanation
-1,000 — Subtract last week's figure from this week's, keeping the sign. Initial claims count people filing for unemployment insurance for the first time in the week, a third source that is neither survey; the sign of the change says whether more or fewer people filed than the week before, nothing more.
Stored on this device only; not graded.
Retrieval check
Mark your confidence before each answer. Every option carries an explanation; read the ones you rejected too.
-
1. A release shows that payroll employment rose and the unemployment rate also rose. The most accurate factual reading of the two figures together is:
Choose your confidence first. -
2. Using the unemployment-rate record, what is the change from the previous print to the latest print, in percentage points? Enter one decimal place, negative if the rate fell.
Source record: Unemployment Rate (as of 2026-08-01)
Tolerance ±0.05 ppChoose your confidence first. -
3. Using the nonfarm-payrolls record, how many thousand jobs were added or lost between the previous print and the latest print? Enter a whole number of thousands, negative for a loss.
Source record: All Employees, Total Nonfarm (as of 2026-08-01)
Tolerance ±1 thousandChoose your confidence first. -
4. The newest release row on the calendar names the Bureau of Labor Statistics as its agency. Within that release, the unemployment rate and the payrolls count come from:
Choose your confidence first.