How it is calculated
Precedent answers one concrete question: when a US economic release comes out, how much and how does the market usually move in the minutes that follow? Everything here is computed from public data, with the same rules for every release.
Dates and times
- Official data (CPI, payrolls, PPI, retail sales, GDP, PCE, jobless claims, JOLTS, durable goods, housing starts): the St. Louis Fed keeps every version of every series exactly as published in ALFRED. The day a new version appears is the release day.
- Fed (rate decision and minutes): the Federal Reserve's official historical calendar.
- ISM and ADP: their dates follow a fixed rule (ISM Manufacturing on the first business day of the month, ISM Services on the third, ADP on the Wednesday of jobs-report week).
Every time is checked against the market: for each release we look for the minute with the most unusual Nasdaq move in a ±20-minute window. If the time is right, the spike lands on the announced minute. The table shows how far that minute exceeds normal (median) and the share of releases with a clear reaction (at least twice normal):
| Release | Time | n | Typical spike | Clear reaction |
|---|---|---|---|---|
| Consumer Price Index (CPI) | 08:30 | 197 | 5.0× | 95% |
| Nonfarm Payrolls (NFP) | 08:30 | 186 | 7.9× | 99% |
| Fed Interest Rate Decision | 14:00 | 128 | 7.1× | 95% |
| FOMC Minutes | 14:00 | 124 | 4.7× | 91% |
| Producer Price Index (PPI) | 08:30 | 196 | 3.9× | 92% |
| Retail Sales | 08:30 | 195 | 3.8× | 95% |
| GDP | 08:30 | 183 | 3.9× | 95% |
| Core PCE Price Index | 08:30 | 185 | 3.2× | 88% |
| Initial Jobless Claims | 08:30 | 809 | 3.5× | 92% |
| JOLTS Job Openings | 10:00 | 188 | 2.5× | 72% |
| Durable Goods Orders | 08:30 | 186 | 3.1× | 88% |
| ISM Manufacturing PMI | 10:00 | 189 | 2.8× | 79% |
| ISM Services PMI | 10:00 | 188 | 2.6× | 77% |
| ADP Employment Change | 08:15 | 190 | 3.7× | 92% |
| Housing Starts | 08:30 | 182 | 2.8× | 80% |
University of Michigan consumer sentiment was tested and removed: there is no reliable free source for its past dates and the spike did not appear at the expected time.
The published figure
Each release uses the figure as it came out that day, not the one revised months later: that is what the market saw. It is compared with the previous figure as stated in that same release. There is no free history of analyst forecasts (the "consensus") yet, so the filters compare the figure with the previous one, not with expectations. Each week's forecasts are shown in the calendar.
Prices
1-minute prices since 2010, in New York time, from the Nasdaq 100, S&P 500 and spot gold CFDs (HistData). They track the NQ, ES and GC futures closely but are not identical: they can differ by a point or so in the most stressed moments. The most recent releases, after HistData's latest month, are measured on the NQ, ES and GC futures (Yahoo Finance 1-minute bars), and the site updates itself every day. Since everything is measured in % from the price just before the release, both sources are comparable.
The figures
- Starting price: the close of the minute before the release.
- Move at 1, 5, 15, 30 and 60 minutes: % difference between the price at that moment and the starting price.
- Typical move: the median absolute move. Half of the releases moved less and half moved more. The median, not the mean, so that a single extreme day does not distort it.
- At today's price: the typical % move applied to the latest available price, to get a feel in points.
- How much more than a normal day: the 5-minute move divided by the median move over those same 5 minutes on the previous 20 business days.
- The first candle sets the direction: share of releases where, at the horizon, price is on the same side the first 1-minute candle after the release set.
- Max excursion: the furthest price travelled, up or down, from the release.
- Path aligned to the first candle: every path whose first 1-minute candle falls is flipped, to see whether that direction tends to continue or fade.
Limits
- Several releases come out at the same time (for example GDP and jobless claims on a Thursday at 8:30). They share the reaction; the "only when alone" filter separates them.
- These are statistics of the past. They do not predict the next release and are not investment advice.