Nonfarm payrolls

Nonfarm payrolls (NFP) is the US Bureau of Labor Statistics estimate of wage and salary jobs on nonfarm establishment payrolls. The familiar headline is usually the seasonally adjusted change from the previous month, rather than the total level. For quantitative research, the distinction matters twice: the units are jobs, and the number available on announcement day can differ from today's revised history.

NFP comes from the monthly Current Employment Statistics (CES) establishment survey. It covers private nonfarm businesses and government employers. Employment refers to workers who received pay for any part of the pay period containing the 12th of the month, including paid leave. It is not a count of everyone who worked at any time during the entire month. The BLS technical note describes the reference periods and coverage.

The Employment Situation release also contains the Current Population Survey (CPS), a household survey used to estimate employed people, unemployment and labour-force measures. The two surveys answer different questions.

The BLS comparison details these differences. Private-household workers are another group included in CPS employment and excluded from CES. The unemployment rate and labour-force participation rate therefore cannot be calculated from NFP alone. A divergence between the surveys is not automatically a contradiction.

Payroll growth is a net change in the estimated stock of jobs. A positive headline does not count gross hires, unique people entering work, or newly created vacancies. Hiring and separations can both be large while the net change is small.

Suppose a fictional seasonally adjusted payroll series rises from 150,000 to 150,100, with the table reporting thousands of jobs. The first level represents 150 million jobs. The difference is 100 thousand, or 100,000 jobs. It does not mean 100 people found work. No annualisation or price deflation is involved: these are monthly employment levels and their difference.

Seasonal adjustment estimates and removes recurring patterns, such as school-calendar and holiday effects, to make month-to-month comparisons more informative. It does not make an estimate error-free or remove every unusual event. Sampling and nonsampling error remain: a small positive estimate need not statistically establish that employment rose. Use the same seasonal-adjustment basis for both levels; subtracting an unadjusted prior month from an adjusted current month has no useful headline interpretation. The CES calculation methodology explains the adjustment and annual benchmark process.

Initial estimates are revised as additional employer reports arrive. The previous two months are routinely revised, and annual benchmarking aligns payroll estimates with more comprehensive employment counts derived primarily from unemployment-insurance tax records. Seasonal factors can also change. Calling a monthly estimate “final” after its routine revisions does not mean its entire history will never change. See the CES benchmark technical notes.

Now let a later fictional release revise the same two monthly levels to 149,980 and 150,120 thousand. Their difference is 140 thousand, or 140,000 jobs. The estimated monthly change has been revised up by 40,000 jobs. That revision describes old months; it is not the following month's job gain.

For an announcement event study, the headline economic surprise is the announced change minus a clearly specified expectation recorded before the release. Revisions to earlier months are separate information arriving at the same announcement. A larger-than-expected headline is not a mechanical instruction about bond prices, equities or monetary policy: interpretation also depends on the rest of the release and the prior market state.

For a backtest, store the observation month, release timestamp, published vintage, units, seasonal-adjustment status and exact series definition. Join data by when they became available. Substituting the later 140,000 figure into a decision made when only 100,000 was published creates look-ahead bias. The CES FAQ explains that the main database holds the latest published statistics and points to vintage files and archived releases.

The MV real-time macro forecasting capstone uses payroll employment among the predictors of US industrial production, with historical vintages and release-availability checks. Membership and sign-in are required. It is an advanced application of timing discipline, rather than a standalone introduction to labour-market measurement.