Reading Forecast, Previous, and Actual Data

So the net new information is not +210,000. It is +210,000 minus a 55,000 downgrade to what everyone believed a month ago, which leaves roughly +155,000 of genuinely fresh employment against a 250,000 expectation. The miss is nearly two and a half times the size it appears at first glance, and that is usually why a reaction that looks disproportionate to the headline is not disproportionate at all. Anyone reading only the big bold number is looking at half the release.

Most calendars show the revised prior in small text next to the previous column, or strike through the old figure. Find where your tool puts it before you need it. On employment and GDP series in particular, a two-month revision of 50,000 to 100,000 jobs is routine rather than exceptional.

Units decide whether 0.1 is trivial or enormous

The same series is published in several forms, and mixing them up produces confident nonsense.

• Month on month against year on year. A 0.1 percentage point difference on a monthly inflation print is large, because twelve of those compound into a materially different annual path. The same 0.1 on the year on year figure is a rounding difference in the twelfth month of an already known series.
• Headline against core. Core strips out food and energy because those are volatile and largely outside policy control. When headline and core diverge, the market generally follows core, because that is the number the central bank is targeting.
• Seasonally adjusted against raw. Retail sales in December are enormous in raw terms every single year. The adjusted series is the one with information in it.
• Annualised against quarterly. A US GDP figure of 2.4% is a quarterly change multiplied up to an annual rate. The equivalent European figure of 0.6% is the quarterly change itself. They are almost the same growth rate.

Write your own threshold before the print

Ten minutes before a release you care about, write two lines in your notes: the forecast, and the number at which you would consider the result a genuine surprise rather than noise. For inflation that might be 0.2 percentage points on the monthly core measure. For employment it might be 60,000 either side of consensus.

Doing this beforehand is worth more than it looks, because it converts a snap judgement made during the fastest thirty seconds of the day into a comparison you already made calmly. It also stops you from retrofitting a story onto whatever price did, which is the default behaviour when a chart moves before you have decided what the number meant. If you are holding through the release, the sizing decision belongs in the same set of notes, and the position size calculator converts a reduced-risk decision into lots without arithmetic under pressure. For the broader picture of what these series are measuring, how forex prices move covers why inflation and employment data reach the currency through rate expectations rather than directly.