Does the headline number move gold?
CPI, the jobs report, jobless claims and retail sales, each split into higher or lower than the previous official figure, with gold's median move, hit rate and sample size. Real data, no forecasts.
US CPI
Next release . The latest figure was 3.7% (September 2026); the next one goes in the “higher” rows if it comes out above that, the “lower” rows if below.
| Releases where… | n | Median move, release hour | Median move, 4 hours later | Release hour closed up | Middle half of release-hour moves |
|---|---|---|---|---|---|
| Headline CPI (y/y) higher than the previous month | 12 | +$0.7 | −$3.3 | 50% | −$2.9 to +$5.5 |
| Headline CPI (y/y) lower than the previous month | 7 | fewer than 8 releases: hidden | |||
| Core CPI (y/y) higher than the previous month | 10 | +$0.7 | −$3.4 | 50% | −$1.5 to +$3.5 |
| Core CPI (y/y) lower than the previous month | 9 | +$19.6 | +$28.8 | 67% | −$2.4 to +$25.9 |
| Headline CPI at or above 3.0% | 9 | +$3.5 | −$3.9 | 56% | −$0.8 to +$25.9 |
| Headline CPI below 3.0% | 10 | +$2.2 | +$3.6 | 60% | −$5.3 to +$9.6 |
19 US CPI releases with hourly gold data (Feb 2025 to Oct 2026). Figures from the BLS as now published. Gold: COMEX gold futures (GC=F) 1-hour bars via Yahoo Finance; release hour = from the close of the hour before to the close of the hour that contains the release. Cells with fewer than 8 releases are hidden. History, not a forecast.
Every release, its date and gold’s move: US CPI event page.
US jobs report (NFP)
Next release . The latest figure was +29k (September 2026); the next one goes in the “higher” rows if it comes out above that, the “lower” rows if below.
| Releases where… | n | Median move, release hour | Median move, 4 hours later | Release hour closed up | Middle half of release-hour moves |
|---|---|---|---|---|---|
| Payrolls higher than the previous month | 8 | −$8.9 | −$13.0 | 25% | −$23.7 to +$3.7 |
| Payrolls lower than the previous month | 10 | +$12.1 | +$3.3 | 70% | −$0.3 to +$34.9 |
| Payrolls +100k or more | 4 | fewer than 8 releases: hidden | |||
| Payrolls below +100k | 14 | +$9.7 | +$3.3 | 57% | −$4.4 to +$34.9 |
| Unemployment rate up on the month | 6 | fewer than 8 releases: hidden | |||
| Unemployment rate down on the month | 5 | fewer than 8 releases: hidden | |||
18 US jobs report (NFP) releases with hourly gold data (Feb 2025 to Oct 2026). Figures from the BLS as now published. Gold: COMEX gold futures (GC=F) 1-hour bars via Yahoo Finance; release hour = from the close of the hour before to the close of the hour that contains the release. Cells with fewer than 8 releases are hidden. History, not a forecast.
Every release, its date and gold’s move: US jobs report (NFP) event page.
Weekly jobless claims
Next release . The latest figure was 194,000 (week ending 24 Oct 2026); the next one goes in the “higher” rows if it comes out above that, the “lower” rows if below.
| Releases where… | n | Median move, release hour | Median move, 4 hours later | Release hour closed up | Middle half of release-hour moves |
|---|---|---|---|---|---|
| Claims higher than the week before | 6 | fewer than 8 releases: hidden | |||
| Claims lower than the week before | 6 | fewer than 8 releases: hidden | |||
| Claims 210,000 or more | 1 | fewer than 8 releases: hidden | |||
| Claims below 210,000 | 13 | +$1.2 | +$2.5 | 54% | −$8.6 to +$15.6 |
14 Weekly jobless claims releases with hourly gold data (Jul 2026 to Oct 2026). Figures from the Labor Department as now published. Gold: COMEX gold futures (GC=F) 1-hour bars via Yahoo Finance; release hour = from the close of the hour before to the close of the hour that contains the release. Cells with fewer than 8 releases are hidden. History, not a forecast.
Every release, its date and gold’s move: Weekly jobless claims event page.
US retail sales
Next release . The latest figure was +0.8% (September 2026); the next one goes in the “higher” rows if it comes out above that, the “lower” rows if below.
| Releases where… | n | Median move, release hour | Median move, 4 hours later | Release hour closed up | Middle half of release-hour moves |
|---|---|---|---|---|---|
| Monthly change stronger than the month before | 6 | fewer than 8 releases: hidden | |||
| Monthly change weaker than the month before | 10 | +$6.1 | −$0.5 | 80% | +$1.2 to +$17.3 |
| Sales up on the month | 13 | +$1.0 | −$3.5 | 54% | −$11.3 to +$10.7 |
| Sales down on the month | 3 | fewer than 8 releases: hidden | |||
16 US retail sales releases with hourly gold data (Jul 2025 to Oct 2026). Figures from the Census Bureau as now published. Gold: COMEX gold futures (GC=F) 1-hour bars via Yahoo Finance; release hour = from the close of the hour before to the close of the hour that contains the release. Cells with fewer than 8 releases are hidden. History, not a forecast.
Every release, its date and gold’s move: US retail sales event page.
The short answer
Sometimes there is a lean, but it is not reliable. Across 67 releases, the largest median release-hour move in any bucket was +$19.6 (US CPI: core CPI (y/y) lower than the previous month, n = 9); the middle half of its moves ran from −$2.4 to +$25.9. In 10 of the 11 rows with enough releases, the middle half of release-hour moves runs from a fall to a rise, so the direction was far from dependable even where the median leans one way. What matters to the market is the surprise against expectations, which the “higher or lower than last time” split only roughly captures.
Why compare with the previous number?
Traders normally judge a release against the consensus forecast. Those forecasts come from commercial surveys whose terms do not allow republishing, so we use the comparison anyone can check from official data: the new figure against the previous one, plus a simple threshold (CPI 3%, payrolls +100k, claims 210,000, retail sales up or down). It is a cruder test than a surprise index, and we say so.
How to read the tables
- n: how many releases fell in that row. Rows with fewer than 8 are hidden (9 rows right now) because a handful of releases says almost nothing.
- Median move: the middle release-hour move in COMEX gold, in dollars per ounce, and the move four hours later from the same starting close.
- Release hour closed up: the share of those releases where gold was higher at the end of the release hour. Around 50% means no lean at all.
- Middle half: the 25th to 75th percentile of release-hour moves. If it runs from minus to plus, the bucket did not have a dependable direction.
Method and limitations
- Release times come from the site's checked event list (the BLS, Labor Department and Census Bureau schedules and archives), the same dates as the gold event pages.
- Figures: CPI year-on-year (not seasonally adjusted, never revised), payrolls and the unemployment rate from the BLS public API; weekly initial claims and advance retail sales via FRED from the Labor Department and the Census Bureau. Payrolls, claims and retail sales are revised later, so the figures here can differ from the ones first published on the day.
- Gold: COMEX gold futures (GC=F) 1-hour bars via Yahoo Finance, from mid-2024. Futures trade a few dollars above spot XAUUSD; moves are almost identical.
- Delayed 2025 releases: after the 2025 lapse in appropriations the September 2025 jobs report came out on 20 November 2025 and is matched to September.
Frequently asked questions
Does a hot CPI number make gold fall?
Not reliably. Splitting past CPI releases into higher and lower than the previous month shows small median moves and a wide spread in both groups. The surprise against expectations and the move in yields and the dollar matter more.
Why not use the consensus forecast?
Consensus figures come from commercial surveys whose terms do not allow republishing. The previous official figure is public, so it is the comparison we can show honestly.
Why are some cells hidden?
Any row with fewer than 8 releases is hidden, because a median of a handful of moves is mostly noise.
How often is the study updated?
Every time the site rebuilds. New releases are added automatically once the official figure and the gold hour are in the data.
Educational study only. Not financial advice and not a signal. Past price reactions do not predict future prices. See our risk disclaimer.