Statistics Canada publishes September CPI at 12:30 UTC, about an hour after this note goes out. The consensus is 3.2 per cent year on year against a 3.0 per cent prior, and plus 0.2 per cent on the month against a prior of minus 0.1. That second pair is the one worth your attention, because the issuing agency published two August monthly figures and the vendors all print the other one. If September comes in at plus 0.2 unadjusted, that is a 30 basis point swing against the prior in every preview you will read this morning — and dead flat against the number Statistics Canada itself calls the monthly change.
The prior is minus 0.1 and plus 0.2 at the same time, and only one of them is in the preview
Statistics Canada’s August release, read at the agency this morning rather than at a reader, says it plainly: the CPI “fell 0.1% month over month in August” and, on a seasonally adjusted monthly basis, “rose 0.2%”. Both sentences are in the same document. Every September preview in this morning’s sweep carries minus 0.1 as the prior and plus 0.2 as the consensus, which reads as a 30 basis point acceleration and is presented as one.
It may be exactly that. But nobody publishing the consensus says which series the forecast is for, and if the forecasters are quoting the adjusted series — which is the series the agency headlines as the monthly change — then plus 0.2 against plus 0.2 is no acceleration at all. The two readings are 30 basis points apart on the same month of the same index. This desk has offered the unadjusted-versus-adjusted gap as a reconciliation hypothesis before and had it killed by a print. The difference today is that the agency’s own page is reachable, so this is not a hypothesis: both numbers exist, and we can show you where.
The operational point is not which is right. It is that a trader sizing off “a 30 basis point swing” and a trader sizing off “unchanged” are looking at the same two published figures.
Four numbers are called core for August, they span 70 basis points, and the one with a consensus is the one the Bank does not target
For the same August month, the figures on the table are CPI-common at 2.6 per cent, CPI-trim at 1.9, CPI-median at 2.0, and ex food and energy at 2.4. That is a 70 basis point spread across four baskets, and the agency separately gives ex-gasoline at 2.4 per cent after 2.2 in July — numerically identical to the ex-food-and-energy figure while being a different exclusion.
The only one of the four with a published September consensus is ex food and energy, at 2.4 per cent year on year against a 2.4 prior. The three the Bank of Canada actually watches have no September consensus anywhere in this sweep. So the beat-or-miss that moves the Canadian dollar at 12:30 will be scored against a basket the central bank does not target, and the measures that feed the October 28 decision will arrive without a number to beat.
The three core readings are one reader here and are labelled as such. We did not get them out of the agency’s own table; its core table did not come through on this fetch, and we are not going to pretend otherwise.
Our September argument is now everybody’s prior column, which is how that kind of argument actually ends
Five weeks ago this desk spent three slots on two vendors who were 40 basis points apart on Canadian CPI, and published the split verdict: one had the year right and the month wrong, the other had the month right and the year wrong. The resolution was that the question was malformed, because the answer is per field.
Those two numbers — 3.0 per cent on the year, minus 0.1 on the month — are now the uncontested prior in every preview published today, from four different houses, with no disagreement at all. Nobody re-fought it. The dispute did not get settled by anyone winning; it got settled by becoming the baseline. That is worth saying out loud, because it is the normal fate of a vendor disagreement and it is a reason to spend less energy adjudicating one: wait a month and the field itself will tell you which figure the market adopted.
What survived is the technique, not the verdict. The reason we could take the split apart was that we went to the issuing agency for the prior instead of arguing about the forecast. We did the same thing this morning and it produced the two-series finding above.
One publisher priced the Bank of Canada’s year end twice in three hours, eleven basis points apart
At 07:30 UTC, a preview from one house said markets price “nearly 23 basis points of tightening by year-end”. At 10:29 UTC, the same house’s Canadian dollar note cited ING saying markets are fully pricing a 25 basis point December hike plus 9 basis points for October. That is 34 basis points. Two pages, one publisher, two hours and fifty-nine minutes apart, eleven basis points of difference on the same quantity.
Both can be literally true if the market moved, and the Canadian dollar did move — USD/CAD is down about 0.3 per cent and has traded below 1.4100 for the first time in about a month. Eleven basis points of repricing in three hours on a day with no Canadian data yet released would itself be the story, which is precisely why the pair of figures needs naming rather than averaging. If you key a position size off “how much is priced”, you need to know you are reading a number with a timestamp on it, and this morning that number has two.
The policy rate is 2.25 per cent and a hold on October 28 is the broad expectation. Neither reading changes that. Both change what a surprise is worth.
What this does not tell you
It does not tell you what September printed, because at the time of writing it has not. It does not tell you which monthly series the consensus refers to — that is the open question, not an answer, and if the vendors are quoting the unadjusted series then the 30 basis point framing is correct and this note is merely pedantic. It does not give you the three Bank of Canada core measures from the agency’s own table, only from one reader. It does not tell you whether the eleven basis point gap in year-end pricing is real repricing or one of the two pages being stale; we cannot distinguish those from outside. And it contains no view on the Canadian dollar. A release whose prior exists in two versions and whose core consensus covers the wrong basket is an argument for a smaller position through 12:30, not a direction.
Related
- FX events calendar — the 12:30 UTC window and what else clusters into the New York open.
- Signals — how a release window is handled when the consensus field is ambiguous.
- EA presets — news-window settings for scheduled releases.
- Only one 12:30 print, and the Canadian vendors are 40bp apart — the dispute whose two numbers are today’s prior.
- Canada printed 3.0 and minus 0.1: each vendor was half right — the split verdict this note marks to market.
