Data Note
Payroll employment counts jobs, while household employment counts people across a broader range of work. Worked examples show how second jobs and shifts into self-employment can produce different movements without contradictory observations.
An additional job need not represent an additional employed person. This distinction is central to interpreting the two employment measures in the U.S. monthly labor report. A business reports positions on its payroll; a household reports the employment status of its members. Their changes can differ even when respondents describe their circumstances accurately.
The Bureau of Labor Statistics comparison of the surveys identifies the relevant boundaries. The establishment survey counts nonfarm payroll jobs and counts a person separately at each covered employer. The household survey counts employed people once and includes categories, such as unincorporated self-employment, outside the payroll measure. It covers the civilian noninstitutional population aged 16 and older. These distinctions make a direct subtraction of the published totals difficult to interpret.
Consider a hypothetical group of 100 employed people. Each holds one covered payroll job, and ten also hold a second covered job. The group supplies 110 payroll jobs but contains 100 employed people. If five other members take second jobs, the payroll count rises to 115 while the person count remains 100. Payroll growth is approximately 4.55 percent; growth in employed people is zero.
Neither calculation identifies whether total hours increased by the same proportion. A second job could add ten hours a week, replace overtime at the first employer, or substitute for hours that the first employer reduced. The job count records an employment relationship, not a fixed package of working time.
Now suppose three of the original single-job employees leave payroll employment and start unincorporated businesses, remaining employed throughout. The covered payroll count falls from 115 to 112; the number of employed people stays at 100. The movement concerns employment composition. It does not establish a corresponding fall in the number of people working.
These examples contain no survey error, population adjustment or economic contraction. They are accounting illustrations, not estimates of the contribution of multiple jobholding or self-employment to any reported U.S. divergence.
A household observation generally concerns the week containing the twelfth of the month. The payroll survey uses the employer's pay period containing that date. BLS also distinguishes unpaid absence and temporary layoff across the survey concepts. A disruption around the reference date therefore needs examination of its timing and employment classification before the two totals are compared.
There is an additional difference between a level and its monthly change. If a series has a broader coverage, its level may exceed another series persistently. That fact does not explain why their changes diverge in a particular month. The explanation must identify a category or measurement adjustment that actually changed, rather than invoke a standing conceptual difference as sufficient evidence.
A useful comparison consequently starts with a common population and employment concept. BLS publishes an adjusted household research series for this purpose. Adjustment improves comparability; it does not turn two independent measurement systems into identical observations or remove uncertainty from their estimates.
A rise in payroll employment cannot by itself determine the employment-to-population ratio. That ratio requires a count of people and a population denominator. Nor does unchanged household employment imply an unchanged unemployment rate: the labor force may have changed as people entered or left active job search.
For an industry-demand question, covered jobs, hours and payroll earnings can be informative. For labor-force participation or the distribution of employment across people, the household framework addresses the relevant population. A briefing should name the measure that answers its question and examine corroborating evidence before assigning an economic narrative to a one-month discrepancy. Treating every difference as a contest over which headline is correct discards information about the margins on which employment is changing.