Economic Data Releases and Market Reaction

At 8:30 a.m. New York time on a Friday, the Bureau of Labor Statistics posts a number. Within about a second, ten-year Treasury note futures have traded through several price levels, the two-year yield sits five basis points higher than it did at 8:29:59, and the S&P 500 futures have gapped. Nobody read the report. The number itself was the trade.

The reflex is to think the market moved because payrolls were strong. It did not. It moved because payrolls were stronger than the market had already assumed. Had every economist forecast exactly the figure that printed, the price would not have budged, however spectacular that figure was. The forecast is already in the price; only the error is new.

The calendar, and its conventions

US macro trades off a small, rigidly scheduled set of releases. The timing matters as much as the content, because it tells you which instrument is liquid when the number lands.

Release	Agency	Time (ET)	Cadence
Employment Situation (payrolls, unemployment rate, average hourly earnings)	BLS	8:30 a.m.	Usually first Friday
Consumer Price Index	BLS	8:30 a.m.	Mid-month, for the prior month
PCE price index (the Fed's target measure)	BEA	8:30 a.m.	Late month
Retail sales	Census	8:30 a.m.	Mid-month
GDP (advance, second, third estimates)	BEA	8:30 a.m.	Quarterly, three vintages
ISM Manufacturing PMI	ISM	10:00 a.m.	First business day
FOMC statement / Chair's press conference	Fed	2:00 / 2:30 p.m.	Eight times a year

Everything at 8:30 lands in the pre-open, so price discovery happens in CME futures — ten-year note futures and E-mini S&P futures — not in cash equities. Everything at 10:00 or 2:00 lands mid-session, into a live order book.

Measuring the surprise

The consensus is the median forecast from a survey of economists compiled by Bloomberg or Reuters and published before the release. Raw surprise is actual minus consensus, but a 79,000 miss on payrolls and a 0.1 miss on CPI are not comparable until each is divided by how big misses usually are.

$$\text{surprise} = \frac{\text{actual} - \text{consensus}}{\sigma}$$

In words: take the forecast error, then express it in units of the typical forecast error for that release. A surprise of $+1.5$ means the number came in one and a half typical misses above what people expected. This standardisation is what lets you compare a payrolls beat with a retail sales miss, and it is what makes the reaction roughly linear.

Worked example: a payrolls beat

Consensus is +175,000 nonfarm payrolls. The print is +254,000. Over the past decade the typical miss has been on the order of 55,000 jobs.

1. Raw surprise: $254 - 175 = 79$ thousand jobs.
2. Standardise: $79 / 55 \approx 1.4$ standard deviations.
3. Apply the beta. Historically the two-year yield moves roughly four to six basis points per standard deviation of payrolls surprise. Take five: $1.4 \times 5 \approx 7$ basis points higher.

That is the first-order move, and it is essentially mechanical. The second-order move — whether the curve bear-flattens, whether equities fall on the higher discount rate or rally on the stronger growth — depends on the regime, and that is where the discretionary judgement lives.

Worked example: the in-line print that moves the market

Core CPI is reported to one decimal place, which quietly destroys information. Consensus is +0.3 percent month on month; the print is +0.3 percent. Identical. And yet yields rise.

The index tells the real story. Last month's core index was 320.500 and this month it is 321.590, a rise of $321.590 / 320.500 - 1 = 0.34\%$. Economists' unrounded forecasts had averaged 0.26 percent, which would have printed 321.333 — also "0.3 percent". Same headline, 8 basis points apart on the month.

Annualise both: $1.0026^{12} - 1 \approx 3.2\%$ versus $1.0034^{12} - 1 \approx 4.2\%$. A full percentage point of annualised inflation hid inside a print the calendar recorded as exactly in line. That is why desks track the unrounded index and the whisper number, and why the tape can move hard on an apparent non-event.

Why the same surprise moves markets differently

The map from surprise to price is stable in sign but not in size, because the reaction runs through what the surprise implies for policy.

• Regime. In 2021–2023, with the Fed fighting inflation, strong growth data pushed equities down: good news meant higher rates for longer. In a growth scare the same print rallies equities. The sign on stocks flips; the sign on front-end yields almost never does.
• Position in the calendar. A payrolls report ten days before an FOMC decision carries far more weight than one the day after, and whichever release owns the narrative — CPI in an inflation fight, claims in a slowdown — crowds the others out.

Revisions and the honesty problem

The first print is an estimate built on incomplete survey returns. Payrolls are revised twice over the following two months, then benchmarked annually against near-universal unemployment insurance records. The preliminary benchmark revision published in August 2024 lowered the level of payrolls as of March 2024 by 818,000 jobs — more than several months of reported job growth, arriving long after the market had traded every one of those months. The market reacts to the vintage it sees, not to the truth, which is also why any backtest of macro data against returns must use the series as it stood at the time.

Key terms

• Consensus — the median of surveyed economists' forecasts, published ahead of the release.
• Surprise — actual minus consensus, standardised by the historical size of misses.
• Event window — the fixed interval around a release, commonly 30 minutes, used to measure the reaction cleanly.
• Whisper number — the unrounded expectation that differs from the published consensus.
• Vintage — a data series as it stood on a given date, before later revisions.