The most violent scheduled release in the currency market, and how to read it properly.
intermediate · 5 min read · 20 XP
Non-farm payrolls produces more volatility per minute than any other scheduled non-policy release. Most traders read one number out of it. The release contains at least five that matter, and the ones that decide the direction of the move are rarely the headline.
Non-farm payrolls is the headline line of the Employment Situation report, published by the US Bureau of Labor Statistics on the first Friday of most months at 8:30am New York time, covering the month just ended.
It counts the net change in the number of paid jobs in the US economy, excluding farm work, private household employees, the self-employed and the military — hence "non-farm".
The report is actually two independent surveys stitched into one release, and confusing them is the most common error in reading it:
| Survey | Who is asked | What it produces |
|---|---|---|
| Establishment survey | Businesses, about their payrolls | The payroll count — the headline NFP number, plus average hourly earnings and hours worked |
| Household survey | Households, about their members | The unemployment rate and the participation rate |
They are different samples with different methodologies, and they routinely disagree — sometimes for months. That disagreement is information, not an error.
The transmission runs through the employment half of the central bank's mandate.
Strong labour market → wage pressure → services inflation → policy stays tight → higher expected path → dollar strength.
Weakening labour market → the case for cuts strengthens → lower expected path → dollar weakness.
Which side of that has more force depends on where the cycle is. When inflation is the binding constraint, a hot payroll number is unambiguously hawkish and dollar-positive. When the bank has already declared victory on inflation and is watching for damage to employment, a weak number becomes the market-moving one and a strong number is shrugged off. Knowing which regime you are in is more than half of trading NFP well.
The regime question, asked before every release
Before payrolls, answer one question: is this central bank currently more afraid of inflation, or of recession? The answer tells you which side of the distribution the market is sensitive to, and therefore whether a beat or a miss produces the bigger move. The two are almost never symmetric.
Read in this order. It takes about ninety seconds and it is what separates a considered reaction from a coin flip.
1. The headline payroll change — versus consensus. This is what the algorithms trade in the first second.
2. The revisions to the previous two months. Every release revises the two prior months, and the revision is frequently larger than the surprise in the current month. A headline that beats by 30k alongside downward revisions of 60k is a net negative report, and the market usually works this out within minutes. This is the single most reliable source of first-move reversals on NFP day.
3. Average hourly earnings, month-on-month and year-on-year. The wage number is the inflation content of the report and often outranks the job count in importance.
4. The unemployment rate, and crucially the participation rate alongside it. A falling unemployment rate is only strong if participation held up. Falling unemployment plus falling participation means people gave up looking, which is a weak report wearing a strong number.
5. The composition. Where were the jobs? Government hiring, leisure and hospitality, and healthcare have very different implications for wage pressure than manufacturing or professional services.
The execution reality
NFP is the worst execution environment on the calendar. Spreads can widen by a large multiple for tens of seconds, depth evaporates, and slippage on both entries and stops is normal rather than exceptional. Some brokers widen stop-distance requirements around the release. A position held through payrolls is not the position you sized — it is that position with a much wider effective risk.
The birth-death model. The BLS cannot survey businesses that opened or closed since the sample was drawn, so it models them. In stable periods this is a reasonable adjustment; at economic turning points it systematically overstates or understates jobs, and it is a recurring source of large later revisions.
Annual benchmark revisions. Once a year the whole series is re-benchmarked against actual tax records. These revisions have at times been very large, retroactively rewriting a year of employment history — and the market does react when the preliminary benchmark estimate lands.
Weather, strikes and holidays. A hurricane, a large strike or an unusual reference-week calendar can move the headline by a large amount for reasons with no economic content. Read the BLS commentary, which flags these explicitly.
Work down the list. The most likely candidates: prior-month revisions large enough to make the net two-month picture negative; average hourly earnings missing badly, removing the inflation implication; the unemployment rate rising despite the job gain, because the household survey disagreed; or job gains concentrated in government and part-time work rather than private full-time. Any of these turns a beat into a soft report, and all of them are visible within two minutes of the release.
What to remember
Non-farm payrolls is two surveys in one release, and the headline job count is only the first of five numbers that matter. The two-month revisions, average hourly earnings, the unemployment and participation rates and the sector composition together decide the direction of the sustained move — which is why the first minute of price action so often reverses.