Markets trade surprises, not levels

The most common beginner mistake in macro is to read a number, decide whether it is "good" or "bad," and expect the market to agree. Unemployment at 4.2% is neither good nor bad for a trade. What matters is that the consensus expected 4.0%, so 4.2% is a miss, and the price had already been set on the assumption of 4.0%.

Prices are forward-looking. By the time a release prints, every forecaster, every model and every trader with a view has already pushed the price to where they think the number will land. The release does not tell the market what the economy is doing. It tells the market how wrong it was.

Discounting

"Priced in" is the shorthand. If the Fed is expected to cut by 25 basis points and it cuts by 25, the two-year yield barely moves, because the cut was already in the yield. If the Fed holds when a cut was expected, the two-year jumps, even though "holding" sounds like the less dramatic action.

The same logic applies to earnings, to elections, to OPEC meetings, and to the weather. The question is never "what happened" but "what happened relative to what was already in the price."

Measuring the surprise

Analysts who publish forecasts get averaged into a consensus number. The surprise is actual minus consensus. To compare surprises across different releases, divide by the typical forecast error for that series (its standard deviation), which gives a standardised surprise in "sigmas."

The ISM surprise in that table is the big one, even though 3.5 index points sounds like nothing. A one-tenth miss on core CPI is a full sigma because core CPI is forecast to a tenth and the errors are small. Some data vendors publish the standard deviation of forecasts alongside the consensus; where you cannot get it, the range of forecasts (high to low) is a rough substitute.

Key idea: The market reaction is to actual minus expected, scaled by how predictable the series usually is. A small miss on a tightly forecast number can move more than a big miss on a noisy one.

Levels still matter, slowly

Surprises drive the first hour. Levels drive the next quarter. If unemployment has risen from 3.5% to 4.2% over a year, that trend matters for the central bank's reaction function regardless of whether any single print beat or missed. The trap is trading the level on release day, when the surprise dominates, or trading the surprise on a swing horizon, when the level dominates.

A practical split:

• Intraday and event trades: care about surprise, sigma, and what was positioned.
• Swing and position trades: care about the level, the trend of the level, and whether the central bank has noticed.

When the consensus is not the real expectation

Sometimes the published consensus is stale and the market has moved on. If three regional Fed surveys all collapsed the week before ISM, traders expect a weak ISM even if the consensus (collected days earlier) still says 50. A print of 49 might then be a relief rally rather than a sell-off. The "whisper number" is the informal expectation the desk is actually trading against; Module 3 covers how to estimate it from price action and from the more recent inputs.

You can also infer expectations from the price. If the two-year yield fell 15 basis points in the week before a CPI print, the market has partly priced a soft number, and a soft number will move less than the sigma table implies.

A worked example

Consensus NFP is 150k. Print is 150k, exactly on the number. Unemployment ticks up from 4.0% to 4.1% (consensus 4.0%). Average hourly earnings come in at 0.2% m/m against 0.3% expected.

The headline is "in line." The internals are soft: higher unemployment and lower wage growth. The two-year yield falls, ES bounces, the dollar dips. A trader who only read the headline would be confused. A trader who lined up each component against its consensus would have the reaction before the first candle closed.

Try it: Take the next three releases on the calendar you care about. Before each, write down consensus, the forecast range, and your guess at which direction the market is leaning. After the print, note actual, the standardised surprise, and the first five-minute move in the instrument you trade. Three data points will not prove anything, but the habit of comparing to expectations rather than to "good or bad" is the whole skill.

Recap

• Prices are set on expectations; releases reveal the error in those expectations.
• Surprise = actual minus consensus; standardise by the typical forecast error to compare releases.
• Surprises dominate the first hour, levels and trends dominate the following weeks.
• The consensus can be stale; recent related data and pre-release price moves tell you what is really expected.
• Read every component of a report against its own consensus, not just the headline.