Lesson 4 of 7 · 8 min
Surprise versus forecast
The headline is only part of a release. Learn to size a surprise, count revisions to earlier months and read a report whose parts disagree.
By the end you'll be able to
- Calculate a surprise and judge whether it is large next to a normal forecast miss
- Count revisions to earlier months when you read a release
- Recognise a mixed report and why its first move often reverses
Free course
Enrol in “Trading the News” to start this lesson
Enrolling is free. Your progress, quiz scores and certificate are kept in your account, on every device.
- 7 short lessons with quizzes and interactive tools
- Lesson videos for members
- A certificate when you finish
Payrolls come in at 230,000 against a forecast of 170,000. The dollar rises for a minute, then falls for twenty. The headline was a beat; the rest of the report was not.
How big is big?
| Time | Cur. | Impact | Event | Previous | Forecast | Actual |
|---|---|---|---|---|---|---|
| 13:30 | USD | ●●● | Non-farm payrolls (jobs added) | 185k | 170k | — |
Move the actual jobs figure 10k from the forecast, then 50k, then 80k, and compare the verdicts. Reactions are illustrative.
The surprise is actual minus forecast. Its size only means something next to how wrong forecasts usually are. US non-farm payrolls forecasts are often off by tens of thousands, so a 10,000 miss is noise. CPI forecasts are usually within a tenth of a point, so a 0.1 miss already moves price and 0.3 is large.
Some calendars also show the range of economists’ forecasts, from the lowest to the highest. An actual inside that range surprises fewer people than one outside it, even if the gap to the consensus is the same.
Revisions change the past
Many releases also revise earlier figures. The US jobs report revises the previous two months every time; GDP is published as a first estimate and revised later. Your calendar shows this as a change to the “previous” column, for example from 185k to 150k.
A revision counts as part of the surprise. If this month beats by 60,000 but the two previous months are revised down by a combined 70,000, the economy created 10,000 fewer jobs than the market believed an hour earlier. Read it as:
Jobs surprise = (actual − forecast) + (revised previous months − first-reported previous months)
When the parts disagree
A report has several numbers, and they can point different ways. The jobs report has the headline, the unemployment rate and average hourly earnings (wage growth). CPI has headline and core (excluding food and energy). Markets often give most weight to the part the central bank watches: core inflation, or wages when inflation is the worry.
When the parts disagree, the report is mixed. Fast trading systems read the headline first, so the first move follows it; then people read the detail and the move can reverse. The jobs data lesson showed a strong headline with weak wages doing exactly this.
Try it
Illustrative: forecasts 170k jobs and 0.3% wage growth; one "typical miss" is taken as 60k jobs or 0.1 point on wages. Real reactions depend on what the market was already positioned for.
Set the revisions to zero and the 230k headline counts in full, so the report turns mixed against the soft wages. Then raise wages to 0.4% as well and it becomes a clean strong report.
Quick check
Quick check
Question 1 of 4
Payrolls beat the forecast by 40k, but the two previous months are revised down by a combined 50k. What is the jobs surprise including revisions?
Question 2 of 4
Which surprise is more likely to move price sharply?
Question 3 of 4
The jobs headline beats strongly, but wage growth misses. What should you expect in the first half-hour?
Question 4 of 4
Where does a calendar usually show a revision?
Go deeper: Non-farm payrolls: what it is and how it moves forex · What is CPI? How inflation data moves forex · How to read an economic calendar