Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Markets move on surprises, not on levels. Two tools help measure them.
The first is the consensus forecast for each data release, typically an average or median of analyst estimates collected by data providers and shown on economic calendars. Comparing the actual figure with consensus gives the surprise. A reading above consensus is positive for the economy's outlook, and usually for its currency, though the reaction also depends on policy implications.
The second is a surprise index, which aggregates many data surprises into a single measure for an economy. Citigroup's surprise indexes are a well-known example. A positive reading means data have been coming in better than expected on balance, and a negative reading means worse.
These indexes tend to mean-revert, since analysts adjust their forecasts after repeated misses. Persistent positive surprises lead to higher expectations, which makes further positive surprises harder.
For traders, the information is useful for context. A currency whose economy has been surprising to the upside may have momentum, until expectations catch up. A sharp turn in the index can mark a change in the data cycle.
Be careful with interpretation. Consensus figures can be stale, and revisions matter.
Use surprise data alongside rate expectations and technical context, not as a standalone signal.
Practical step: for the next three releases, record the consensus, the actual figure, and the price reaction.
Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
The mean reversion point is subtle and useful. After a long run of positive surprises, forecasters catch up, and the next releases have a harder bar to beat. That's why the surprise index tends to swing above and below zero rather than staying high. For trading, it means a currency whose economy has been beating forecasts for months isn't necessarily set for more upside surprises. The positive news may already be priced in, with forecasts raised to match. Some traders watch for a surprise index turning down from a high level as an early warning that the data run is fading, before the currency itself has reacted.
It’s Fairman 
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Recording consensus, actual and reaction for a few releases is a nice way to learn how much surprise your pair actually needs to move. Take US CPI for example. Note the consensus, the actual, the surprise in tenths of a percent, and how many pips EURUSD moved in the first 15 minutes after 1:30pm Lagos. After six or eight releases you'll see that a 0.1 percent miss on core CPI often moves the pair a lot more than a similar miss on headline. You'll also see that some releases get completely ignored because something bigger is dominating, like a Fed meeting the following week. That kind of note is useful for planning.
It’s Fairman 
-
FXS Prop Desk
- Posts: 1689
- Joined: Wed Oct 07, 2026 9:52 am
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
The Citi Economic Surprise Index is probably the best-known surprise index. One practical use is to compare the indexes of two economies: when the US index is rising and the eurozone index falling, EUR/USD often weakens, since the data is beating expectations more often in the US. The indexes are mean-reverting, which also helps to identify extremes.
Calendar sites differ in how they compute consensus, so using one source consistently matters.
Calendar sites differ in how they compute consensus, so using one source consistently matters.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Putting every surprise into the same units makes the record far more useful across releases. Dividing the miss by the typical size of past misses for that release gives a score that can be compared between payrolls and CPI.
Say payrolls usually miss consensus by about 75 thousand in either direction. A print 40 thousand above forecast is then only about half a typical miss, while a core CPI miss of 0.1 on a release that usually lands within 0.1 is a full one, which fits the point above about core moving the pair more.
Revisions belong in the same row. A headline beat of 40 thousand paired with a downward revision of 50 thousand to the prior two months is a net miss, and the first spike often reverses once that is read.
The surprise index itself is built along these lines, with older surprises fading out over a few weeks. That decay is part of why it swings back toward zero, alongside forecasters catching up.
For the 15 minute reaction column, is the move measured from the release price or from the minute before? Spreads widen in that last minute, so the starting quote can shift the pip count by a meaningful amount.
Say payrolls usually miss consensus by about 75 thousand in either direction. A print 40 thousand above forecast is then only about half a typical miss, while a core CPI miss of 0.1 on a release that usually lands within 0.1 is a full one, which fits the point above about core moving the pair more.
Revisions belong in the same row. A headline beat of 40 thousand paired with a downward revision of 50 thousand to the prior two months is a net miss, and the first spike often reverses once that is read.
The surprise index itself is built along these lines, with older surprises fading out over a few weeks. That decay is part of why it swings back toward zero, alongside forecasters catching up.
For the 15 minute reaction column, is the move measured from the release price or from the minute before? Spreads widen in that last minute, so the starting quote can shift the pip count by a meaningful amount.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
The spread of forecasts behind the consensus changes how big a miss really is. A median hides whether the forecasters were bunched together or scattered across a wide range.
Say twenty payrolls forecasts run from 90 to 210 thousand with a median of 150. A print of 190 then lands inside the range and was expected by several forecasters, while the same 40 thousand beat against forecasts all between 140 and 160 sits outside every one of them.
The second case tends to produce the cleaner move, because positioning was built around a narrow view. Some calendars show the high and low forecast next to the median, which is enough to note the width each time.
Stale forecasts are the other issue. If eight of the twenty were filed before a strong jobless claims week, the live expectation is probably above the printed median, and a print that matches the median can trade as a small miss.
Does anyone record the forecast range alongside the surprise? Pairing it with the standardised score would show whether wide ranges really do soften the first move on the pair.
Say twenty payrolls forecasts run from 90 to 210 thousand with a median of 150. A print of 190 then lands inside the range and was expected by several forecasters, while the same 40 thousand beat against forecasts all between 140 and 160 sits outside every one of them.
The second case tends to produce the cleaner move, because positioning was built around a narrow view. Some calendars show the high and low forecast next to the median, which is enough to note the width each time.
Stale forecasts are the other issue. If eight of the twenty were filed before a strong jobless claims week, the live expectation is probably above the printed median, and a print that matches the median can trade as a small miss.
Does anyone record the forecast range alongside the surprise? Pairing it with the standardised score would show whether wide ranges really do soften the first move on the pair.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Markets do not react to a number, but to the gap between that number and what was already priced in. The published consensus is a median of economist forecasts, usually collected days earlier, so it can be stale by release time.
A whisper number is the informal expectation that circulates among desks and positioning surveys in the final hours. It has no official source and no fixed definition, which makes it unreliable as a precise input. Still, when price has drifted sharply in one direction before a release, the implied expectation has probably moved away from consensus.
That explains a familiar oddity: a result that beats consensus can still produce a drop in the currency. If the market had already positioned for a bigger beat, the print is a disappointment relative to the real expectation. The forecast table alone cannot reveal this.
Scalpers sometimes compare the pre-release move with the consensus deviation to judge how much is priced in. The dispersion of forecasts also matters: a wide spread of estimates means any single outcome is less likely to shock. Conclusions here remain probabilistic, never certain.
A whisper number is the informal expectation that circulates among desks and positioning surveys in the final hours. It has no official source and no fixed definition, which makes it unreliable as a precise input. Still, when price has drifted sharply in one direction before a release, the implied expectation has probably moved away from consensus.
That explains a familiar oddity: a result that beats consensus can still produce a drop in the currency. If the market had already positioned for a bigger beat, the print is a disappointment relative to the real expectation. The forecast table alone cannot reveal this.
Scalpers sometimes compare the pre-release move with the consensus deviation to judge how much is priced in. The dispersion of forecasts also matters: a wide spread of estimates means any single outcome is less likely to shock. Conclusions here remain probabilistic, never certain.
-
GrahamW1972
- Posts: 46
- Joined: Thu Oct 22, 2026 4:54 am
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
shadow trader, mine went a bit differently when i tried it. i logged about eight releases last month, just consensus, actual and what the pair did, and the stale consensus thing barely showed up... the beats mostly moved it the way you'd expect, apart from one payrolls where it dropped on a beat and i just put that down to bad timing on my part. didnt bother with whisper numbers, cant see how you'd record them without making them up. fair play though, the pre-release drift is definately something ill look at next, cos i reckon i missed it by only looking at the print. bit like Leeds, you can have the numbers right and still get stung by what everyone else already knewShadow Trader wrote:Markets do not react to a number, but to the gap between that number and what was already priced in.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Consensus Forecasts and Surprise Indexes: Measuring What Was Not Expected
Ha, the Leeds line is about right. Eight releases is a small log, but useful. If the stale-consensus effect barely showed up and most beats moved the pair the way you'd expect, then for those prints the published number was still informative enough. The one payrolls that dropped on a beat is the classic trap: the print looks good, but the market had already climbed into it, so the actual surprise was smaller than the headline gap. Whisper numbers are hard to record without inventing them, so I wouldn't force that either. Pre-release drift is the bit worth adding next, because it tells you what was already priced before the number landed. If you log the move from, say, two hours before until the print, alongside consensus and actual, you get a cleaner read on whether you were trading the surprise or just late to a move everyone else already knew. Worth a month of that and see what the journal says.