Actual vs Forecast vs Previous: How Big a Surprise Matters

A 40,000-job miss on payrolls sounds like a lot. A 0.2-point miss on monthly inflation sounds like almost nothing. Set each one against the range of estimates and the release’s own history, though, and the order flips: the jobs miss turns out to be the kind of number that turns up most months, while the inflation miss lands further from the estimates than any miss in the six months before it, which is why the smaller-looking number can do more to rate expectations and to stock prices than the bigger one.

Where does the forecast come from?

The Forecast column is a consensus. Before each release, economists send in their estimates, and a data service reduces them to a median or an average. One number comes out. The spread behind it disappears.

Some calendars and data services publish that spread as well, as the highest and lowest estimate. Look for it. A wide range means economists had little idea what was coming, and a tight one means they broadly agreed.

The consensus can also drift in the days before a release as late estimates come in, so the figure on your screen the night before may differ a little from the one in place at the release minute, and a miss measured against the older number can look slightly bigger or smaller than the one prices reacted to.

The surprise is actual minus forecast. Its raw size tells you little on its own. Judge it against the range of estimates for this release, and against how big the misses on this row usually are.

Did the actual land inside the range?

Start with the payrolls row from the walkthrough on how to read an economic calendar, where 150,000 jobs came in against a forecast of 190,000. Now add the estimate range the consensus was built from.

Plenty of economists expected a number like this. Some expected worse. Anyone who had looked at the range before the release knew a print of 150,000 was well within what the forecasters thought possible, so the miss tells the market something it had already half considered.

Now a hypothetical monthly CPI row. The forecast is 0.3%, with estimates running from 0.2% to 0.4%. The actual is 0.5%.

That sits above every estimate. Not one economist in the survey called it. A reading outside the whole range is a different kind of event from a reading inside it, whatever the raw gap looks like.

How big is a normal miss for this release?

The range describes this month. The release’s history describes how hard it is to forecast at all. Some series jump around from month to month, and economists miss them by wide margins as a matter of course; others move in small steps and get forecast to within a tick.

So take the last six misses on each row, drop the plus or minus sign, and average them.

Dividing a surprise by the usual size of surprises gives a standardized surprise. The payrolls figure comes out at 0.86. Ordinary. It’s a smaller miss than an average month produces. The inflation figure comes out at 3.0, three times the normal miss, on a release that the forecasters had been getting right or nearly right for half a year, and a miss on a row that has been quiet for months is the kind that catches positions leaning the wrong way.

That is the number that moves rate expectations. A trader who sized both misses by their raw units would have braced for the jobs report and shrugged at the inflation report, which gets the two backward and leaves the part of a portfolio most exposed to rates sitting unguarded on the wrong day.

Six months is a short history, and one wild month can drag the average up. A longer run of past misses steadies it.

Can rounding make a miss look bigger than it is?

Yes, most of all on releases measured in tenths of a percent. Headline inflation figures are rounded to one decimal place, so a monthly change printed as 0.3% covers everything from 0.25% up to just under 0.35%. A one-tick miss on the row can shrink a good deal once the decimals come back.

The release itself prints index levels, and those give the unrounded change. Suppose a hypothetical price index moves from 300.00 to 301.08.

The 0.5% print from earlier comes through the rounding intact. Printed as 0.5%, the actual sits somewhere from 0.45% to just under 0.55%, so the unrounded miss runs from 0.15 to about 0.25 point. Even at the low end, 0.15 / 0.067 is 2.25 typical misses.

The same rounding sits inside every past miss in the history too, so a standardized figure is a guide to scale, no more precise than the one-decimal numbers behind it, and a 3.0 and a 2.25 both say large, which is all the reading has to tell you.

Where does Previous come in?

A revision adds to the surprise. When last month’s figure is revised in the same release, the market reprices both months at once, and the headline miss understates the change. The economic calendar walkthrough works a 40,000 miss plus a downward revision into a 70,000 shortfall, and the same logic applies here: add the revision to the miss before you judge its size.

A standardized surprise built on the headline alone misses that part. So check Previous first.

What if one release says two different things?

Many releases carry several numbers. The monthly employment report brings payrolls, the unemployment rate and average hourly earnings in the same minute. They can disagree. Payrolls might miss low while hourly earnings come in above forecast.

The market weights whichever figure speaks to its worry at the time. When inflation is the worry, wages carry the most weight, because rising pay feeds prices and keeps rates high; when a slowdown is the worry, payrolls and the unemployment rate carry it, and weekly initial jobless claims get watched closely in the weeks between reports. Whether a higher reading helps or hurts depends on the row, which is the subject of actual higher than forecast.

Stocks do not all take the same hit when rate expectations jump. Rate-sensitive groups react hardest, and the lesson on stock market sectors sets out which ones those are and why. The lesson on the economic releases that move markets covers which rows deserve this treatment in the first place.

Put together, the routine for any row is short. Subtract. Check the range. Divide by the typical miss. Unround if the number is small. Add the revision. Then ask which figure in the release the market is watching this month, since a 3.0 surprise on a row nobody is worried about can still pass quietly.