Non-farm payrolls

The headline figure of the Employment Situation report, published by the Bureau of Labor Statistics, usually on the first Friday of the month and covering the month before. It comes from the establishment survey, a large monthly survey of business and government payrolls, and it counts jobs on payrolls rather than people in work. Farm work is excluded, along with private household employees, the unincorporated self employed, unpaid family workers and the armed forces, which is what the name records.

The same release carries a second survey and a second set of numbers. The household survey asks people about their own status and produces the unemployment rate, the participation rate and the employment to population ratio. The two surveys count different things on different bases, so they can move in opposite directions in the same month without either being wrong. Average hourly earnings and average weekly hours arrive with the payrolls figure and are read as a measure of wage pressure.

Everything published is seasonally adjusted, and everything is provisional. Each release revises the two months before it as late responses arrive, and once a year the whole series is benchmarked to a near complete count of employment drawn from unemployment insurance records. The revisions are the part most often missed: they regularly exceed the gap between the headline figure and what forecasters expected, so a release can read as a beat on the headline and as a downgrade once the revisions to earlier months are added in.

Because employment is one half of the mandate most central banks are judged against, the release moves the dollar, government bond yields and index futures within seconds of publication, and quoted spreads across affected instruments commonly widen around it. Which part of the release matters most is genuinely contested and shifts with the cycle: when the labour market is the constraint the headline count leads, and when prices are the constraint the earnings component often moves markets more than the count does. Forecasters disagree about the count itself as well, since the models used to estimate jobs at firms too new or too small to survey are an assumption rather than an observation.