Two surveys, two units
Nonfarm payroll employment comes from the establishment survey. It counts jobs on participating employers’ payrolls. A person holding two covered payroll jobs can contribute two jobs to the total. The survey excludes categories such as most self-employment that can appear in household employment.
The household survey asks about people and their labor-force status. An employed person is counted once even if that person holds more than one job. Its population, survey design and definitions differ from those used for establishments. It is the source of the U-3 unemployment rate and participation measures.
The unemployment denominator
The U-3 unemployment rate divides unemployed people by the civilian labor force. The labor force includes employed and unemployed people, under the survey’s rules. It is not the entire U.S. population. Children, institutionalized people and other people outside the survey universe are not simply classified as unemployed.
People without a job who are not actively seeking work generally are outside the labor force, with specified treatment for temporary layoffs. The employment-population ratio uses the covered population as its denominator, while participation measures the labor force relative to that population. Each ratio describes a different relationship.
Why changes can diverge
Monthly sampling variation, coverage differences, population controls, seasonal adjustment and later revisions can all affect the published histories. A month when household employment and payroll employment move differently is not resolved by forcing one change to equal the other. Inspect the specific sources and the longer trend.
Job openings, hires and quits introduce another distinction: openings are a stock at the end of the month; hires and quits are flows during the month. Hires alone are not net job growth because separations also occur. Use each measure for the question it was designed to answer and keep its period and unit in the citation.
How the monthly report is assembled
Both measures appear in the BLS Employment Situation release, usually published on the first Friday of the month. The household figures come from the Current Population Survey (CPS), conducted by the Census Bureau for BLS. The payroll figures come from the Current Employment Statistics (CES) program, which surveys businesses and government agencies about the employees on their payrolls.
The two surveys also use different reference periods. The CPS asks about activity during the week that includes the 12th of the month. The CES asks employers about the pay period that includes the 12th. Pay periods can be weekly, biweekly or monthly, so a payroll count and a household count labeled with the same month do not describe exactly the same days. Holiday timing and weather during the reference week can also affect the two counts differently.
How analysts read the two series together
Analysts often favor the payroll survey for month-to-month employment change because its sample is much larger and its sampling error is smaller. They often favor the household survey for demographic detail, unemployment and participation, which payroll records cannot measure. Neither survey is simply the correct one; each answers its own question with its own uncertainty. BLS publishes measures of sampling error for both, and small monthly changes often fall within that range.
Payroll estimates are benchmarked each year to counts derived largely from unemployment insurance tax records, while household estimates are tied to population controls. When the two diverge for several months, analysts look at multiple jobholding, self-employment, population-control adjustments and sampling error before drawing conclusions. A single month of divergence is rarely informative on its own, and a gap that persists for many months deserves a closer look at the source notes.
Reporting errors and correct citations
A common error is writing that “the economy added 200,000 workers” when the figure is a change in payroll jobs. Because one person can hold two jobs, a payroll gain does not equal a gain in employed people. Another error is to subtract payroll job gains from the change in unemployed people as if both came from the same survey. Mixing sources this way produces a number with no consistent definition.
A precise citation names the survey, measure, adjustment and month. For example: “Total nonfarm payroll employment, seasonally adjusted, change from prior month, CES, BLS.” For household data: “Unemployment rate (U-3), seasonally adjusted, CPS, BLS.” Including the release date also helps readers, because payroll estimates for recent months are revised in later releases and the figure a reader checks later may differ from the one first reported.
Worked example: one town, two counts
Suppose a hypothetical town has a civilian noninstitutional population of 1,600 people age 16 and older. Of these, 1,000 are employed, 50 are unemployed and actively seeking work, and the rest are outside the labor force.
- Of the 1,000 employed people, 100 are self-employed and hold no payroll job, so 1,000 − 100 = 900 people hold at least one payroll job.
- If 150 of those 900 hold two payroll jobs, the payroll count is 900 + 150 = 1,050 jobs, while the household count of employed people stays at 1,000.
- The labor force is 1,000 employed + 50 unemployed = 1,050 people. The unemployment rate is 50 ÷ 1,050 ≈ 4.8%.
- The participation rate is 1,050 ÷ 1,600 ≈ 65.6%, and the employment-population ratio is 1,000 ÷ 1,600 = 62.5%.
- If 20 employed people each take a second payroll job, payroll jobs rise by 20 to 1,070 while household employment, unemployment and the unemployment rate do not change.
Takeaway: The payroll and household counts here differ by 50 even though both are accurate for their own units. Real surveys add sampling error, different coverage and different reference periods.
Frequently asked questions
Why can payroll jobs rise while the number of employed people falls?
The payroll survey counts jobs and the household survey counts people. Jobs can rise if existing workers take second jobs, while household employment can fall if fewer people report working. The surveys also differ in coverage, such as self-employment, and both carry sampling error, so their monthly changes can move in opposite directions.
Which survey does the unemployment rate come from?
The unemployment rate comes from the household survey, the Current Population Survey. Employers cannot report on people who do not work for them, so the payroll survey cannot measure unemployment. The headline U-3 rate divides the number of unemployed people by the civilian labor force.
Are people with two jobs counted twice?
In the payroll survey, a person holding two jobs with covered employers can appear on two payrolls and be counted twice as jobs. In the household survey, that person is counted once as employed. BLS publishes a separate household measure of multiple jobholders for readers who want to track the difference.
Does the payroll survey include self-employed workers?
Generally no. Nonfarm payroll employment covers jobs on employer payrolls and excludes most self-employed people, unpaid family workers, private household workers and farm workers. The household survey asks people directly about their work, so it can count self-employed people as employed.