Tomorrow at 13:30 UTC the US Bureau of Labor Statistics publishes the October consumer price index. It will not contain a year-over-year inflation rate for the headline index. That is not a forecast, a rumour or an inference from a vendor page — it is on the Bureau’s own site, has been since 16 September, and reads: “BLS will not publish 12-month percent changes for October 2026 for indexes where October 2025 data are not available.” For most consumer price series, October 2025 was never collected. One economic calendar we opened this morning is nevertheless carrying a 3.9 per cent headline annual rate as the market consensus for tomorrow, with no caveat attached. If your handling computes a surprise as actual minus consensus, tomorrow hands it a blank where the number should be and a 3.9 where the number should not.
The issuer says it plainly, and said it seven weeks ago
The Bureau keeps a question-and-answer page on the effect of the 2025 lapse in appropriations on the consumer price index. Its last-modified date is 16 September 2026. The relevant statements, verbatim and short:
- There was a “lapse in appropriations from October 1, 2025, through November 12, 2025”.
- “Most CPI operations, including data collection, were suspended during that time period.”
- “BLS did not issue an October 2025 CPI news release.”
- “For most CPI series, BLS could not publish October 2025 indexes.”
- “BLS will not publish 12-month percent changes for October 2026 for indexes where October 2025 data are not available.”
It goes further than a missing month. “Survey data for commodities and services were carried forward to October 2025 from September 2025”, and rents for October 2025 were carried forward from April 2025 — six months back — leaving rent and owners’ equivalent rent unchanged for that month. For seasonal adjustment the Bureau says it approximated the missing October by taking the geometric mean of the September and November index values. The reason given for not simply collecting twice is the plainest sentence on the page: the Bureau “does not have data collection resources to collect 2 months of data in 2 weeks”.
Two of this desk’s standing rules point the same way here and both paid. The first is to settle anything on a fixed cadence at the issuing agency rather than an aggregator. The second is that when the primary source is reachable you quote it and when it is not you say so. This time it was reachable, and the issuer settles the question that four secondary pages disagree about. Worth noting what the same release day also carries, from the same page: the Bureau plans to publish final chained-index estimates for October through December 2025 on 10 November 2026. The missing quarter gets filled in on the morning that its absence breaks the headline.
The time, for completeness, is issuer-confirmed twice. The September release states that the October 2026 news release “is scheduled to be published on Tuesday, November 10, 2026, at 8:30 a.m. (ET)”, and the Bureau’s schedule table carries the same date and 08:30. So the calendar entry is right about when and wrong about what.
Four annual rates for a month that has no annual rate
The field is absent at the issuer. It is not absent from the internet, because it is computable — and four pages have computed it four ways:
- 3.9 per cent, on a calendar page verified this morning, labelled as the market consensus and explicitly not that publisher’s own view. Whose consensus it is, the page does not say. There is no caveat about the missing month anywhere on it.
- 3.74 per cent, a Cleveland Fed nowcast dated 6 November as reported by one reader, which reaches an annual rate by using its own stand-in for October 2025.
- 3.8 per cent headline and 2.6 per cent core, which is that same reader’s own arithmetic rather than a forecast: what you get if October is flat, measured against an average of the September and November 2025 index levels.
- 3.9 per cent headline and 2.8 per cent core from a research house, which states openly that these are comparisons against September 2025 — a thirteen-month change wearing a twelve-month label.
Those four sit between 3.74 and 3.9, which is close enough to look like a consensus and is nothing of the kind. Each one is a different synthetic October 2025: a nowcast’s interpolation, a geometric-style average of the two neighbours, and a base shifted by a month. The spread between them, about 16 basis points, is purely an artefact of which fake month you chose. None of them is what the Bureau will print, because the Bureau will print nothing there.
The point is not that these publishers are being dishonest. Three of the four say what they did. The point is that a field being blank at the issuer does not keep it blank downstream, and a number that is computable will be computed, circulated, and eventually quoted by somebody who has dropped the methodology sentence. By tomorrow afternoon there will be a headline annual inflation rate for October 2026 in general circulation. It will not have come from the agency that measures inflation.
What your filter does with a null
Here is the mechanism, which is the only part of this that is actionable. A news filter keyed to the consumer price index has three ways to handle tomorrow and two of them lose money.
It reads the missing annual rate as zero and computes a 3.9 point miss, which is the largest downside inflation surprise in the history of the series and did not happen. It skips the event because a required field is absent, and you are flat through the release that one page this morning calls the one that decides December. Or it falls through to whichever synthetic annual rate its data vendor supplies, and you are trading against a stand-in month you did not choose and cannot see. The third is the least bad and it is still not a decision you made.
The field that will actually print is the monthly change, and that is where the consensus work should have gone. One reader gives 0.3 per cent headline and 0.2 per cent core as the consensus and names the houses behind the range: Morgan Stanley at 0.19 per cent on both fields, ING at 0.3 on both, Bloomberg Economics at 0.25 and 0.24. Add the Cleveland nowcast at 0.24 and 0.21, a research house at 0.24 and 0.22, and a bank at 0.3 on both, and the spread across every named forecast is 0.19 to 0.3 per cent. Every one of those rounds to 0.2 or 0.3 on a field reported to one decimal place. A print of 0.2 is simultaneously in line with the consensus and a tenth below it depending on whose number you were holding, and nothing in the release will tell you which.
And you cannot pre-register which fields will be populated, because the issuer does not publish the list. The Bureau’s page confirms that “a few indexes” are built entirely from non-survey sources and were published for October 2025, but it names none of them. One secondary reader says gasoline, new vehicles and used cars therefore keep their annual changes; another says only monthly changes will be published at all. Those cannot both be right, and the issuer’s own page does not settle it — it also does not mention core inflation anywhere, so whether the core annual rate prints tomorrow is genuinely unknown rather than merely disputed. A filter that needs to know the shape of a release in advance cannot be given it this time.
We were wrong about 15:00 today, and the correction is seven hours old
This desk published at 23:14 UTC last night that 15:00 today held US final September wholesale inventories, forecast 0.6 per cent against a 0.6 per cent prior, and called it “a release whose forecast equals its own prior, which is a news window you can safely not have”. In the same paragraph we wrote that Goolsbee “speaks with no published time, which is its own kind of event”.
Both halves are wrong, and one calendar fixed both. At 15:00 UTC there are four items, not one: wholesale inventories with its tautological forecast, wholesale trade sales with a 1.8 per cent prior and no forecast at all, the Conference Board Employment Trends Index with a 107.56 prior, and Austan Goolsbee, at 10:00 a.m. Eastern. A second calendar independently puts Goolsbee at 23:00 Beijing time, which is the same instant. So the published negative is closed and the answer is yes, there is a time, and it lands in the window we told you to ignore. Thirty minutes later the Atlanta Fed’s GDPNow estimate arrives at 15:30, forecast 3.4 per cent against a 3.4 per cent prior, and at 16:00 the New York Fed publishes one-year consumer inflation expectations against a 3.9 per cent prior.
Note that last number and then disconnect it from the other one. The New York Fed’s prior is 3.9 per cent; the annual consensus one calendar carries for a field that will not print tomorrow is also 3.9 per cent. They have nothing to do with each other. We are naming the coincidence because two identical figures a day apart, both describing inflation, both quotable, is exactly how a wrong number gets into a trading log and stays there.
The rest of the day, in UTC: Bowman at 13:40, the three-month and six-month bill auctions at 16:30 with 4.040 and 4.150 per cent priors, Barkin at 17:15, Bailey at 17:25, and the three-year note auction at 18:00 against a 4.932 per cent previous — the first leg of the week’s coupon supply. One of the two calendars says no high-importance events are scheduled today, which is true of the data and false of the speaker list.
The lesson we are taking is narrow and unflattering: we judged a window from one calendar. A second calendar turned a dead 15:00 into a four-item cluster with a Federal Reserve president in it. One of those pages also states that it was generated with the support of artificial intelligence and reviewed by an editor, which is why we only report what a second source confirms — Goolsbee’s 15:00 is on both, so it stands.
The time zone nobody wrote down
The second calendar states every single time on the page without a time zone. Not an ambiguous zone — none at all. Its China consumer price entry reads 09:30, the German trade balance 15:00, euro-zone Sentix 17:30, Bailey 01:25 on 10 November.
Those are Beijing times, and we know that by reconciliation rather than by reading: 09:30 Beijing is the 01:30 UTC at which China actually published this morning, and 01:25 on the 10th in Beijing is 17:25 UTC, which is where the Eastern-time calendar puts the same speaker. Every entry lines up once you assume the zone. That is the dangerous case. A calendar with wrong times announces itself the first time an event fails to happen; a calendar with correct unlabelled times reconciles perfectly against a second source and then silently shifts everything by eight hours the one time you read it alone.
Which leaves what London actually gets this morning, and it is almost nothing: German September trade data at 07:00 UTC, with a single-sourced 18.0 billion euro forecast against a 19.4 billion revised prior, and euro-zone Sentix investor confidence for November at 09:30. Nothing else is scheduled in European hours. The busiest liquidity window of the day has no release in it worth a stop, the first real event is five and a half hours after the London open, and the event that decides the week is tomorrow and is missing its headline field. We would spend today reading our own calendar metadata rather than the market, and tomorrow checking which fields came back populated before reacting to any of them.
What this does not tell you
The issuer’s page settles that annual changes will be withheld where October 2025 data are missing, and settles that most series are in that position. It does not name a single index either way, and it does not mention core inflation anywhere. So we cannot tell you which fields will be populated tomorrow, and the two secondary readers who claim to know contradict each other. Anyone giving you a definitive field list for tomorrow’s release has more than the Bureau has published.
We did not read the October release, because it does not exist yet. Everything above about its contents is the issuer’s advance statement plus two readers’ interpretation of it, and the interpretations are the weaker half.
The 3.9 per cent calendar does not name whose consensus it is, so we cannot tell you whether a survey behind it exists and has been misapplied, or whether the figure is a carried-forward field in a template. We did not reach the one page most likely to carry a full forecast table for tomorrow — a wire service’s week-ahead piece on US inflation — so our monthly consensus set is assembled from secondary readers rather than from one survey, and the 0.19 to 0.3 per cent range may be narrower or wider than we have it. One week-ahead page we did open turned out to be two weeks stale and covering a different week entirely, which is the reminder to read the dateline before the table.
The market levels on one of these pages come from an undated sidebar snapshot with no timestamp, so we have not used them for anything. We have no option expiry board and no resting-order levels for today, which is the third session running, and no dollar–yen or euro level of our own in this article at all. This is a piece about release mechanics, not about price.
Finally, the claim that the 2025 shutdown is the cause is the issuer’s own, but no page we read describes any 2026 interruption. If you have heard that this is about a current shutdown, that is not what the Bureau says.
