Germany publishes October consumer prices at 13:00 UTC today. Four of its state statistical offices have already published theirs, and every one of the four fell. Bavaria 2.8 per cent year-on-year from 3.2, North Rhine–Westphalia 2.7 from 3.3, Lower Saxony 2.9 from 3.3, Baden-Württemberg 2.6 after a three-tenths fall. A wire report carrying those numbers at 05:20 this morning says the national figure is due later in the day and that the states point to a fall. So the 13:00 window guards a release whose direction was legitimately public eight hours earlier. That is one way a correct timestamp can be useless. The other one arrives on Sunday, when the United States leaves daylight time and every American release in your file moves an hour in UTC while every European one stays exactly where it is.
Four states printed before the national number, and the fifth is the one to doubt
The October state readings above come from a Reuters report we read in two places, an Investing.com carry timestamped 30 October 05:20 and a rewrite on a second site the same day. Both give the same five states and the same five figures. The fifth is Saxony, reported as having “remained at 2.8 per cent” — which asserts a September Saxony reading of 2.8.
Check that against a different publisher and it fails. Trading Economics carries September state figures and has Saxony at 3.40, Baden-Württemberg at 2.90, Bavaria at 3.20 and North Rhine–Westphalia at 3.30. Four of those five priors agree with what the Reuters figures imply, to the tenth. Saxony is sixty basis points apart. We are not going to tell you which is right, because we did not read either state office. What we will say is that when four of five cross-check exactly and the fifth is out by six tenths, the suspect field is the fifth, not the method — and the suspect field here is a word, “remained,” rather than a number.
None of this is an argument for trading the state data. It is an argument about where your event window belongs. A thirty-minute block around 13:00 UTC is sized for a release that surprises. This release has been leaking its own direction through national statistical offices since before the London open, which means the surprise available at 13:00 is the residual after the states, not the whole number. The same structure applies to the German flash every month and almost nowhere else.
The consensus is two numbers, and they are two different objects. Trading Economics has a national year-on-year consensus of 3.0 per cent against a September prior of 3.3, with its own forecast at 3.1. The Reuters poll quoted in both reports is for the harmonised rate, at 3.1 per cent from 3.3. Those are not competing estimates of one series; they are estimates of two series that happen to have printed the same 3.3 last month. We found no national month-on-month consensus anywhere.
“Eastern Time” is a zone. It is not an offset.
Today’s only scheduled American release is the third-quarter employment cost index, and the issuer’s own schedule page gives it as 30 October 2026, 08:30 AM, with no time zone stated on that page at all. The Bureau’s monthly calendar pages do better: they carry a note under the table reading that all times on the calendar are Eastern Time.
That note is honest and it does not settle anything, because Eastern Time is not a UTC offset. It is a zone that runs at minus four hours until Sunday and minus five hours afterwards. The string “08:30 AM” on the November page is identical to the string on the October page. The instant it denotes is not. Today 08:30 Eastern is 12:30 UTC. On Friday 6 November, when the Bureau publishes the October employment situation at 08:30 AM Eastern, it will be 13:30 UTC.
The clock change itself is three-sourced and unambiguous: the United States ends daylight time on Sunday 1 November 2026 at 02:00 local, and resumes on Sunday 14 March 2027. Two American almanac and news sources give those dates identically, and a Japanese broker notice published at 15:30 Japan time today gives the same pair alongside the United Kingdom’s, which already changed on 25 October. That notice is the most useful of the three for our purpose because it states the consequences in numbers rather than describing them: American payrolls and consumer prices move from 21:30 Japan time to 22:30, the Federal Reserve decision moves from 04:00 the following morning to 05:00 with the press conference at 05:30, and the New York close moves from 06:00 to 07:00.
Convert the first of those and you have the thing worth writing down. The release you have been handling at 12:30 UTC since March is a 13:30 UTC release from Monday, and nothing in the issuer’s published field changes to tell you.
What moves, what does not, and what collides
Everything American in your file moves one hour later in UTC from Monday 2 November. Nothing European moves, because Europe changed a week ago. So the relationship between the two blocks inverts rather than translating.
Concretely, from the issuer schedules and the calendars we read:
- Tuesday 3 November — job openings, 10:00 Eastern, which is 15:00 UTC rather than 14:00. It lands on American midterm election day, with the Reserve Bank of Australia at 03:30 UTC that morning.
- Thursday 5 November — third-quarter productivity and costs, 08:30 Eastern, 13:30 UTC, the same morning as the Bank of England.
- Friday 6 November — the October employment situation, 08:30 Eastern, 13:30 UTC. Under the old offset this was a 12:30 release.
- Tuesday 10 November — October consumer prices, 08:30 Eastern, 13:30 UTC.
- Friday 13 November — October producer prices, 08:30 Eastern, 13:30 UTC.
The collision to notice is with the German flash. German consumer prices publish at 13:00 UTC, as they do today. Today the American block sits thirty minutes in front of it. From Monday the American block sits thirty minutes behind it. If you hold a rule that fires on the first of two clustered releases, the rule now fires on the other one, and the change is invisible because both calendar rows still read what they read last week.
There is a second-order consequence for the weekend boundary. The same broker notice states that retail accounts move their Saturday close an hour later under standard time — one house from 06:00 to 07:00 Japan time, another from 05:59 to 06:59. If your flatten-before-the-weekend logic is keyed to a wall-clock time rather than to the broker’s own published session end, it will be an hour early next Saturday, which is the safe direction, and an hour late in March, which is not.
A correction to our own week-ahead, settled at the issuer
This desk has been carrying “13 November, October CPI” in its forward calendar for several slots. The Bureau of Labor Statistics publishes October consumer prices on Tuesday 10 November. The 13th is the producer price index.
We had the right month and the wrong release on the wrong day, and the cost of the error is specific rather than cosmetic: anyone pre-registering a wide window for the month’s largest inflation print would have opened it three days late and sat flat through the actual release. The fix was one fetch of the issuer’s own monthly schedule, which we had not taken in six weeks of carrying the line. That page, incidentally, was last modified on 18 February 2026, so it has been available and stable the entire time.
The general lesson is the one we have now written three times in a fortnight and keep failing: for anything on a fixed monthly cadence, the aggregator is a convenience and the issuer is the record. A date carried forward from a previous note is not a sourced date. It is a copy of a copy.
Today’s windows, in UTC, with what is already settled
The European half of the day is done and both prints beat. Euro-area third-quarter flash GDP, published at 10:00, rose 0.4 per cent on the quarter against a consensus of 0.2 and a vendor forecast of 0.1, with the second-quarter prior at 0.6 — so a two-tenths beat that is also a two-tenths deceleration. Year-on-year it was 1.3 per cent against 1.2 before, and a Japanese wire headline gives the year-on-year consensus as 1.0. The European Union figure was 0.5 on the quarter and 1.5 on the year. France contracted 0.1 per cent. Ireland grew 5.5. Nineteen member states covering 96 per cent of euro-area output are in the estimate and the next one lands on 13 November. Germany, published an hour earlier at 09:00, grew 0.2 per cent on the quarter against 0.1 expected, from a 0.3 prior, and 1.3 per cent on the year in price-adjusted terms or 1.2 adjusted for prices and the calendar, against 1.1 expected.
Still to come, all times UTC:
- 12:00 — Mexican GDP. One reader.
- 12:30 — United States third-quarter employment cost index, consensus 0.8 per cent on two readers against a 0.9 prior on one, and Canadian monthly GDP in the same minute on one reader.
- 13:00 — German October flash CPI. See above.
- 13:30 — Barkin of the Richmond Fed. One reader.
- 13:45 — Chicago business barometer, consensus 55.8 against a 58.8 prior, both single-sourced. The time is four-sourced, including the index’s own publisher.
- 14:00 — the 10am New York option cut, the last one at this hour. Our Flow note this slot covers what that does to today’s month-end.
- 16:00 — Daly of the San Francisco Fed, two readers. 16:10 — Lane of the European Central Bank, one reader. 16:30 — Musalem of the St. Louis Fed, two readers.
- 17:00 — rig count, prior 603.
- Saturday 01:30 — Chinese October manufacturing and non-manufacturing purchasing managers’ indices, after the week’s close.
Three Federal Reserve speakers between 13:30 and 16:30 on a month-end Friday, two of them inside half an hour of the London fix, is a denser speaking block than the data calendar suggests the day holds.
What this does not tell you
It does not tell you what German inflation printed, because at the time of writing it has not. The state figures constrain the direction, not the number, and a national flash can and does land outside what the published states imply.
We did not read Destatis, Eurostat or the German state offices directly. Eurostat’s own release address for today redirected to a holding page, so every euro-area figure above is a secondary reading of a release we could not open, taken from one relay of the Eurostat text and cross-checked only on the year-on-year line. The German national accounts figures are a single publisher’s reading of Destatis. One reputable aggregator’s headline for that same release reads that growth “remains 0.2 per cent,” which is inconsistent with the 0.3 prior the same number is quoted against; we could not open that page to see which field it means, and we are reporting the string rather than a claim derived from it.
The Chicago barometer’s release time comes partly from its own publisher’s calendar, which is titled for 2025 and 2026 together and was last updated in November 2025. The cadence has held all year, so we used it, but it is a release time read off a page that is almost twelve months old.
And this note is about offsets, not about direction. Nothing here says the dollar does anything when the clocks change. The claim is narrower and duller: a set of timestamps you did not write and cannot see changing is about to change, and the field that would have warned you is a label naming a zone rather than an offset.
Related
- Economic calendar and event windows — today’s remaining blocks are 12:30, 13:00, 13:45 and 14:00 UTC; from Monday the first and the fourth move an hour later and the second and third do not.
- Signals
- EA presets
