A headline says payrolls rose by 100,000. The next month another article says that same month added 80,000. A third release says 90,000. These fictional estimates refer to one period. Their differences show why the release date matters as much as the month on the chart.
A monthly estimate arrives in stages
The Bureau of Labor Statistics surveys employers for its Current Employment Statistics program. The payroll series estimates jobs on nonfarm employer payrolls. The first monthly estimate uses the information available when the release is prepared. The agency revises that estimate in the next two releases as more reports arrive and estimates are updated. The third figure is the final sample-based estimate in the ordinary monthly cycle.
In our example, the first revision moves the estimate down by 20,000, from 100,000 to 80,000. The next moves it up by 10,000 to 90,000. By then, the estimate is 10,000 below the initial release. An article comparing the latest release with the immediately previous one might correctly describe an upward revision. An article comparing it with the first release might correctly describe a lower estimate.
An annual benchmark asks a different question
BLS also benchmarks payroll estimates using more comprehensive employment counts, largely drawn from unemployment insurance tax records. That process can revise historical figures beyond the two ordinary monthly updates. Seasonal adjustment also aims to separate recurring patterns from the movement readers want to compare, such as the regular rise and fall of holiday employment. BLS explains that some seasonally adjusted figures remain subject to later revision.
A headline about a benchmark can concern payroll levels at a particular date or a revision across a span of months. Check whether the number is a level adjustment or a change from one month to the next. The language can sound similar even when the measures differ.
Record both dates when you save a figure
If you are checking a statement made on release day, use the estimate available on that day. If you are describing what the most recent official data show about the past, use the newer figures and identify their release. When adding consecutive changes, take them from a consistent version of the same series.
The employment report also includes a household survey. It supports separate measures, including the unemployment rate. Keep the series name beside your number. A payroll change and a change in how many people are employed use different surveys and definitions, so their results can move differently.
Sources & further reading
- BLS questions about payroll estimates
Explains the two ordinary revisions, annual benchmarks and the distinction between payroll jobs and household measures.
- BLS guide to the employment report
Sets out what the establishment and household surveys measure and why their totals differ.