The monthly jobs report is the Employment Situation news release, published each month by the U.S. Bureau of Labor Statistics (BLS). It reports national employment estimates from two separate surveys, one of employers and one of households. Policymakers, financial markets, businesses and researchers use the figures because they are among the earliest indicators of economic conditions each month, according to BLS.
The two surveys behind the report
The first survey is the Current Employment Statistics program, known as the payroll survey or establishment survey. It collects data from employers and measures employment, hours and earnings in the nonfarm sector, with industry and geographic detail. BLS describes it as best known for providing a highly reliable gauge of monthly change in nonfarm payroll employment.
The second is the Current Population Survey, known as the household survey. It measures the labor force status of the civilian noninstitutional population, with demographic detail. The national unemployment rate is its best-known statistic.
BLS says both surveys are needed for a complete picture of the labor market. The employment numbers differ because the surveys define employment differently and use different survey and estimation methods.
How the payroll survey works
The payroll survey is a monthly survey of individual worksites run by businesses and government agencies. The sample is drawn from a list of unemployment insurance tax accounts across the United States. According to a BLS page last modified Feb. 11, 2026, the sample covers about 119,000 businesses and government agencies, representing about 622,000 worksites.
The survey counts nonfarm wage and salary jobs, not people. A person with two jobs is counted twice. It leaves out agriculture, private households and the self-employed. The reference period is the employer’s pay period that includes the 12th of the month. That period could be weekly, biweekly, monthly or another length.
BLS says the entire sample is redrawn each year, and about one-third of it is rotated out and replaced with newly selected businesses. A supplemental sample of business births is selected midway through the year. Businesses first answer by computer-assisted telephone interview. After the first five months, many switch to cheaper methods they start themselves, such as web reporting or electronic data interchange.
The estimate has two parts. Reports from the sample show month-to-month employment change at continuing businesses. A statistical birth–death model then accounts for firms that open or close but are not yet reflected in the sample, because there is a lag before a new establishment appears on the sampling list.
Before publication, automated checks flag data that may be wrong, and BLS re-contacts respondents when needed. Analysts then look for errors and outliers. The figures are also seasonally adjusted, which removes the effects of normal yearly events such as holidays and weather so that underlying trends are easier to see.
How the household survey works
The household survey covers about 60,000 eligible households, according to BLS. The sample is made up of addresses, not names, drawn from the Census Bureau’s Master Address File. The U.S. Census Bureau conducts the survey for BLS, using telephone calls and in-person visits to ask about work activity during the previous calendar week. The reference week is generally the one that includes the 12th of the month.
The survey covers the civilian noninstitutional population. That leaves out people in the Armed Forces and people living in institutions such as nursing care facilities and correctional institutions. Each person age 16 and over in the sample is classified as employed, unemployed or not in the labor force.
Employed people are those who worked as paid employees, ran their own business, profession or farm, worked at least 15 hours without pay in a family business or farm, or were temporarily absent from a job. Unemployed people had no job during the reference week, were available for work, and had made specific efforts to find work in the prior four weeks. People laid off and expecting to be recalled count as unemployed even if they have not looked for work.
The labor force is the employed plus the unemployed. The unemployment rate is the number of unemployed people as a percentage of the labor force. People who fit neither group are counted as not in the labor force.
BLS weights each person’s answers by the inverse of the probability of being sampled, adjusts for nonresponse, and aligns the results with independent population estimates called population controls. A final step reduces sampling error further. The Census Bureau adjusts the controls each year, and BLS introduces the adjustments with January data. BLS typically does not revise earlier household data for these changes.
Why the two employment totals differ
The payroll survey estimates jobs, while the household survey estimates employed people. A person with two jobs counts twice in the payroll survey and once in the household survey. The household measure also includes groups the payroll survey leaves out, including the self-employed, agricultural workers, private household workers and unpaid family workers. For these reasons, BLS says, the household employment level is higher than the payroll level.
BLS also creates an adjusted household series for research. It removes the groups not covered by the payroll survey and adds multiple jobholders to approximate a jobs count. According to BLS, this adjusted series tracks the payroll measure much more closely, though occasional divergences still occur. The agency says the adjustment is imperfect because some of the needed data are not available.
Other differences also play a role. Employer pay periods can be longer than the household reference week. Payroll employment reflects only people who received pay in the period, while the household survey counts people on unpaid leave as employed. BLS says it cannot adjust for the possible effect of the differing reference periods.
Revisions and margins of error
Because both surveys are samples, their estimates carry sampling error. The household survey has a smaller sample and a larger range of error. On the page last modified Feb. 11, 2026, BLS listed the size of a monthly change in employment needed to be statistically significant at the 90% confidence level as about 650,000 for the household survey and about 122,000 for the payroll survey. BLS says short-term changes, especially monthly ones, should be read with sampling error in mind.
Payroll estimates are revised after first publication. The first preliminary estimates come out about three weeks after the reference period. A second preliminary estimate follows the next month, and the final sample-based estimate comes two months after the first release. After that, the figures stay unchanged until the annual benchmark.
In benchmarking, BLS re-anchors the payroll estimates to a nearly complete count of employment based mainly on unemployment insurance tax records for March. The process helps correct for sampling and modeling error. BLS says that historically, benchmark revisions have been very small for total nonfarm employment. For household data, BLS says seasonal factors are applied to the current month immediately, and the most recent five years of data are revised at year’s end.