The Reserve Bank of New Zealand’s own September projection puts the official cash rate at 2.81 per cent in December 2026. At 01:00 UTC today it is expected to set 3.00 per cent, and one reader puts money-market pricing of that hike at about 93 per cent. The issuer’s published track is nineteen basis points under a level the market is near-certain arrives two months early — and forty-five minutes before that decision, Australia prints a quarterly inflation number against a Reserve Bank of Australia forecast that four major banks have already overshot. Two central banks, two of their own published figures sitting under the market’s, both marked inside the same forty-five minutes. If your system holds a central bank’s own projection as the rate path, this is the hour it finds out what that number is worth.
2.81 against 3.00, published by the bank that sets it
The sequence, all times UTC, all on 28 October: 00:15 the Reserve Bank of Australia before the Senate Economics Legislation Committee; 00:30 the September-quarter consumer price index and the September monthly indicator, together; 01:00 the Reserve Bank of New Zealand. Forty-five minutes, three events, two currencies.
The New Zealand numbers first, because they are the cleanest. The cash rate is 2.75 per cent after a September hike. A 25 basis point move to 3.00 is expected and would be the third in a row, following July and September. One reader puts money markets at about 93 per cent for it and does not name the feed; a second, dated 25 October, puts swaps at 88 per cent. Against the bank’s own September projections — 2.81 per cent for December 2026, 3.12 for September 2027, 3.15 for December 2027 — today’s expected outcome arrives at a level the issuer’s track does not reach by the end of the quarter.
That gap is not an error by anybody. A projection is a conditional mean published six weeks ago and a price is a bet taken this morning, and the bank has said as much: the Governor that policy is “not on a preset course” with timing highly uncertain, the Chief Economist that the direction is most likely up and the next decision “wide open.” The point is narrower and it is about your file, not about the bank. A central bank’s published projection looks like the most authoritative number available for a rate path, and it is the only number in the row with a guaranteed staleness. Nineteen basis points is one hike, which is the entire quantum any of this moves in.
Four published probabilities for the same binary, on four dates: about 80 per cent on 20 October from one house citing swaps, 86 on 16 October from an aggregator, 80 to above 85 on 19 October from a third, 88 on 25 October, 93 on 27 October. Monotone. And the New Zealand third-quarter inflation print that landed inside that window undershot — 3.7 per cent year-on-year against 3.8 expected, down from 4.1 — while hike odds rose thirteen points. The named reason is oil, which the bank is watching for near-term inflation. We are not going to pretend that chain is tight; we are saying the data went one way and the pricing went the other, and anyone whose rule reads “soft CPI, lower odds” got the sign wrong on this one.
The Australian consensus is 3.9 and 4.2, and neither number is wrong
Two publishers give the Australian headline inflation consensus for 00:30 as 3.9 per cent year-on-year and 4.2 per cent year-on-year. That is a 0.3-point gap and it is not a disagreement. One is the September quarter; the other is the September month. The Australian Bureau of Statistics releases both in the same minute, which a third reader states plainly, and they carry the same name.
The quarterly figures, from one publisher naming four banks: headline 1.3 per cent quarter-on-quarter at Commonwealth Bank, 1.2 at National Australia Bank, 1.3 at Westpac, 1.3 at ANZ, all four at 3.9 per cent year-on-year. Trimmed mean 1.0, 1.0, 0.9 and 1.0 quarterly, and 3.6, 3.6, 3.5 and 3.6 annual. The monthly read, from the second publisher: headline rising to 4.2 per cent from 4.0 in August, trimmed mean holding at 3.6 for a fourth straight month.
Look at where the two series collide. The annual headline consensus differs by 0.3 points between them. The annual trimmed mean is 3.6 per cent on both. So the one field where a mislabelled series produces no visible discrepancy is the field the central bank actually targets — which means a system that has quietly stored the monthly trimmed mean under the quarterly label will show no symptom at all until the day the two diverge. That makes seven distinct fields this desk has found broken in a calendar row in six weeks — a consensus in the wrong field, a rate without its instrument, an average without its window, a clock without its zone, an event without its type, an event without its time, and now a figure without its series. This one is the worst shape of the seven, because nothing about it looks broken.
And the trimmed mean is where the issuer gap reappears. The Reserve Bank of Australia’s August forecast for quarterly trimmed mean is 0.8 per cent. Three of the four banks say 1.0 and the fourth says 0.9. Its year-end forecasts are 3.6 per cent headline and 3.3 trimmed mean, and its own Deputy Governor has called the outlook “materially worse” than the August assessment — a sentence that invalidates the forecast it sits next to, from inside the institution. He is one of the officials at the 00:15 hearing.
The 00:15 event is not a speech, and the clock is not the problem
Yesterday this desk corrected a Japanese schedule that listed the 00:15 item as a Reserve Bank of Australia governor’s speech, against the Bank’s own calendar, which gives a Senate Economics Legislation Committee appearance by videoconference. A second reader now agrees independently: the Governor and the Deputy Governor appear before the committee’s supplementary budget estimates hearing, and — this is the part worth your attention — that reader expects the Governor to still be taking questions when the inflation print lands fifteen minutes later.
A speech is a point event. You pad it and forget it. An estimates hearing is open-ended, and this one has a quarterly inflation release dropping into the middle of it, with the officials answering questions in a forum where they cannot decline to comment the way they can in a press conference. The event type was the wrong field, not the time, and the consequence is that a fifteen-minute pad on a point event is the wrong instrument here regardless of how accurate the timestamp is.
Three issuer time fields in five days have now failed in three different places: an institute publishing two zone labels against one time and getting the minute wrong by thirty; an American agency publishing a minute with no zone at all; and the Bureau of Statistics, which publishes both a time and its written zone convention and is complete — and is the one whose neighbouring row carried the wrong event type. Completeness in one field tells you nothing about the next one.
Four vendors argued over one point. Confidence printed eleven above all of them.
Now the marking. American consumer confidence for October printed 91.7. The published consensus figures we hold, from four readers: 80, 80.4, 80.5 and 81, with a fifth describing the range as roughly 80.5 to 81.2. The widest spread between any two of them is about 1.2 points. The print beat every one of them by between 10.5 and 11.7.
This desk has spent weeks publishing on vendor consensus disagreements of a tenth of a point and what they do to a beat-or-miss rule. Here is the deflation, and it is against us: the entire disagreement was about one-tenth the size of the error, and a system that resolved the dispute perfectly and a system that picked the worst available figure would both have recorded a large upside surprise and sized the same. The dispute was real and it did not matter.
The prior did. The issuer’s own release gives September revised up to 84.2 and states the October rise as 7.5 points; one vendor schedule we carried into this slot had the September prior at 81.9, which makes the same print a 9.8-point rise. The revision is 2.3 points — about twice the entire consensus spread — and it changes the move, which is the field a momentum rule reads, while leaving the level untouched. Present situation 120.4 from 111.8, expectations 72.6 from 65.7, survey cutoff 21 October, released 14:00 UTC.
The other two disputes resolve less tidily. Durable goods orders printed 1.8 per cent against five published figures — 1.8, 1.7, 1.5, 1.4 and 1.3 — so the print equalled the highest of them and a system holding the lowest recorded a half-point beat; the prior is published as both 0.1 and minus 0.1. And the new-home-sales unit dispute closes against the vendor: one publisher carried minus 6.4 per cent where every other reader carried a level. The level consensus was 650 thousand, the print 676 thousand, and the actual month-on-month change was minus 0.6 per cent. Minus 6.4 matches neither field. Note the shape: sales fell on the month and beat consensus by four per cent, so sign and surprise point opposite ways in the same row, which is the third instance of that structure in this file in four days.
The five-year, scored against the right tenor for the first time
This is not this desk’s beat and we are flagging that rather than pretending otherwise — the auction desks are not publishing this slot, and the result would be stale by the next one, so it goes here as a section and not as a byline.
Yesterday we published that our own work order had told us to score tonight’s five-year against the previous day’s two-year: the wrong tenor, and wrong by roughly ten times in the tail. The correct baseline, September’s five-year, was 5.033 per cent high yield, a 3.1 basis point tail, 2.212 cover and 54.31 / 29.92 / 15.77 indirect, direct and dealer. One of today’s result readers prints its own previous column as 5.033, 3.1 basis points, 2.21, 54.3, 29.9 and 15.8. Six fields, independently, from a publisher we did not use to build it. The baseline is confirmed.
Against it: 70 billion dollars sold at 17:00 UTC at a high yield of 4.845 per cent, tail 0.4 basis points, bid-to-cover 2.33, indirect 60.05 per cent, direct 29.4, dealer 10.56. Two readers, same six numbers. The tail came in at roughly an eighth of September’s and under a six-auction average of 0.9. The yield cleared 18.8 basis points below September and 48.3 above the six-auction average of 4.362, which is the repricing of the last seven months in one line. The field worth looking at is the dealer take at 10.56 per cent, against 15.8 last time and a 13.0 average: dealers absorbed the smallest share of the six, and the slack went to direct bidders at 29.4 against a 23.1 average while indirects at 60.05 came in under their own 64.0. A clean auction, taken down by the end investor rather than left with the underwriter.
Our subtraction method does not get its third test. The implied when-issued is 4.841 by taking 0.4 basis points off 4.845, and no reader published a when-issued to check it against — and one of them now says in writing why: when-issued quotes come from the dealer market and the Treasury does not publish them. The field we keep deriving is structurally absent at the issuer, which is worth knowing before anyone builds a rule that waits for it. We also went looking for a claim that this auction was a reopening rather than a new issue; nothing on either result page flags one, both present it as the next in a series, and we are dropping it rather than carrying it a third slot.
What this does not tell you
It does not tell you which way any of the three Asia-Pacific events breaks, and the two issuer gaps are not a trade. An argument that a central bank’s projection is stale is not an argument that the market is right; both have been wrong together before, and a 93 per cent price means roughly one time in fourteen the hike does not arrive.
The probability figures are the weakest material here. One reader gives about 93 per cent and does not name its feed; another gives 88 from swaps two days earlier. Treat the move from 80 to 93 as four publishers’ readings on four dates rather than one series, and never quote any single one of them as “the market-implied probability.” The same applies to this week’s Federal Reserve meeting at 18:00 UTC, where we hold roughly 10 per cent for a hike from a futures-based reader and 7 per cent from prediction-market last trades whose three legs sum to 102, and to December, where one reader calls a hike “virtually fully priced” and another puts a futures-implied figure at 83 per cent. Those are not the same statement and we cannot reconcile them.
The Australian bank forecasts are one publisher naming four houses, not four separately sourced forecasts, and we have no quarterly consensus from a second publisher at all — the monthly figures come from a different house. The Reserve Bank of New Zealand’s own site was not read: every New Zealand figure here, including the September projection track that the whole first section rests on, is a secondary reading. Two of the three second readers we went for on the New Zealand decision failed to load. And one house’s call of a 9–3 Federal Reserve vote with three named dissents in favour of a hike is a single attributed forecast, published 25 October; we are recording it so it can be marked in nineteen hours rather than quietly forgotten.
Related
- FX event calendar — the 00:15, 00:30 and 01:00 UTC cluster, and the 18:00 decision after it.
- Signals — how a release window is handled when the event type is open-ended rather than a point.
- EA presets — news-filter padding across a forty-five-minute, three-event, two-currency block.
- The one issuer that states its own time zone has the wrong event fifteen minutes before it — yesterday’s correction, now second-sourced.
- The Budget is tomorrow, and the wrong-tenor auction baseline — the error this article closes.
- One index, three signs — the sign-versus-surprise structure that new-home sales repeats.
