A plain-English guide to economic surprise indices.

Every piece in the analysis chain we have been running, going back to the original data-not-dots call, has used the same building block: the US growth surprise index. We have been writing "US growth surprises have rolled over" and "the surprise gap versus G10 has widened" without ever explaining what the index actually is or how it is built. This is that explainer. The headline number is the Citigroup Economic Surprise Index, or CESI. It is published daily, available on Bloomberg, Reuters, and MacroMicro, and it is the closest thing FX desks have to a one-number summary of the data flow.

What it measures

An economic surprise, in this index, is the gap between an actual data release and the Bloomberg-median consensus economist forecast. If the May ISM Manufacturing PMI consensus was 52.5 and the print was 54.0, the surprise was +1.5 points, scaled by the historical surprise distribution for that series. Each release contributes a single signed number to the index on the day it lands.

The point of normalising against the historical distribution is that a 1-point upside ISM surprise is not the same event as a 1-point upside CPI surprise. The standard deviations of those surprises differ; without normalisation, you would over-weight the noisy series and under-weight the cleaner ones. The CESI converts every release to a z-score on its own historical distribution before aggregating.

Construction in three lines of math

For each release i on day t:

1. Surprise = (actual − consensus) / σ_i, where σ_i is the standard deviation of past surprises for that release.
2. Weighted contribution = w_i · surprise, where w_i is the historical FX impact of a one-standard-deviation surprise in that release.
3. CESI = sum of weighted contributions over the trailing three months, with exponential decay so that recent surprises matter more than older ones.

The three-month rolling window is the key design choice. It is long enough to smooth out single-print noise; short enough to capture a regime change in five to eight releases. The exponential decay means today's ISM has more weight in the index than the ISM from two months ago, which mimics how markets actually forget.

How to read the number

• CESI = 0. Releases are coming in roughly in line with consensus. The data is doing what economists thought it would do. This is the modal regime.
• CESI > 0. Releases are beating on average. The activity side is stronger than analysts expected. Forward growth forecasts will be revised up; real yields will drift higher; the currency will firm.
• CESI < 0. Releases are missing on average. The activity side is softer than analysts expected. Forward growth forecasts will be revised down; real yields will drift lower; the currency will weaken. This is the regime we have been documenting in the US for most of the spring.
• The cross-section. The index is published per country and per region. The number FX traders care about most is the spread between two regional CESIs: US minus euro-area, US minus G10 ex-US. That spread is the relative-growth differential the dollar smile framework uses as its horizontal axis.

What today's print does to the index

Plotted: an illustrative twelve-month path of the US CESI. The spring decline matches the deterioration we have been writing about; the snap-back on the right corresponds to today's upside ISM surprise.

One upside ISM surprise of the magnitude we got today contributes roughly 10 to 15 index points to the trailing three-month window, depending on the exact decay weights. That is meaningful: it moves the index from "comfortably negative" back toward zero, but it does not by itself flip the regime. The regime call takes another two to four upside surprises of similar size to be confirmed.

Why the index leads, not lags

The surprise index is in some sense already an aggregate of the most recent forecast revisions; economists who are missing a series in one direction quickly update their model. The practical consequence is that the CESI tends to lead actual growth-rate revisions and OIS-implied policy moves by two to four weeks. Reading the index level is reading where the consensus is going to revise to, not where it is today.

There is a related, more academic measure published by the Federal Reserve Board, the Macro Surprise Index, which uses a smaller basket but a more careful uncertainty weighting. It correlates above 0.85 with the CESI in the post-2010 sample. For desk work, the CESI's daily publication cadence wins; for academic work the Fed index is cleaner.

What the index does not tell you

1. The level of growth. CESI tells you how the growth is doing versus expectations, not what the growth is. A 2 percent GDP economy with a CESI of +50 means analysts had thought it would be 1 percent and were wrong; a 4 percent GDP economy with a CESI of -50 means analysts had thought it would be 5 percent and were wrong. The index is about the gap, not the absolute.
2. Composition. A positive CESI driven by hot ISMs and soft NFPs is a different regime from one driven by hot NFPs and soft ISMs. The index hides which releases are doing the work; for that, look at the disaggregated contributions, available in the same data feed.
3. Persistence. A single big surprise (today's ISM) lifts the index but does not change the trajectory. Watch the slope, not the level, when judging regime change.
4. Across countries. Cross-country CESI differences are noisy because the country panels include different release calendars. The spread is informative on a three-month average, not on a single day.

Related reading

• Today's analysis applies this framework directly: ISM came in hot. The dollar took it.
• The thesis the surprise index has been backing: The dollar is weaker on the data, not the dots
• The cross-asset regime the surprise index reads through: A plain-English guide to currency correlations
• How the surprise gap connects to dollar regimes: A plain-English guide to the dollar smile