The Bank of England held Bank Rate at 3.75% at 11:00 UTC, by six votes to three, with Megan Greene, Catherine Mann and Huw Pill preferring 4.00%. At least two research houses published that exact split, with those exact three names, before the meeting. So the vote was not the news. The news was that the same committee voted unanimously to run its monetary-policy gilt holdings to zero — a decision worth roughly 46 billion pounds a year for eight years, announced in a market notice, with no field for it anywhere on your economic calendar.
We told you this morning the level was not the story. It was not, and we were right for slightly the wrong reason
At 06:11 UTC this desk published a prediction-market snapshot showing a 96.75% chance of no change — and made the point that the snapshot had been captured on 14 September at 14:02 UTC, three days earlier, before the Federal Reserve hiked. The argument was that the level was close to certain and the content would therefore be tone and composition rather than level, and that the vote split was how to mark it.
Mark it: the level was right, 3.75% held. The composition was hawkish, three votes for an immediate increase. So far so good. But the part we did not anticipate is that the composition which actually mattered was not the vote count at all. It was the balance sheet, and we had no paragraph for it because our calendar had no entry for it.
That is a more useful failure than being wrong about a number. A calendar tells you when a committee speaks. It does not tell you how many separate decisions the committee will announce when it does.
Two houses called 6–3 by name, which is why the vote could not move anything
Oxford Economics forecast a 6–3 split naming Pill, Greene and Mann. MUFG forecast a 6–3 hold and described it as fairly hawkish. Both were published before the decision; both were exactly right, names included.
Price agreed with them in advance. Ahead of the meeting the market carried roughly an 80% probability of a hike at the November meeting rather than this one — two readers gave that figure — with one of them describing November as the first of about four increases expected over the following year. A committee that delivers the consensus vote, in the consensus direction, on the consensus timetable has delivered no information.
This is the practical form of a rule this desk keeps restating: a scheduled event is only tradeable to the extent that its content is not already in the price, and a vote split that two named houses published in advance is in the price. If your system sizes a position by the impact tag on the calendar entry, it just took full risk into an event whose headline field was pre-solved.
The decision with no calendar field: gilt holdings to zero
Read from the Bank’s own market notice, published alongside the decision: the Committee determined to reduce the stock of UK government bonds held for monetary policy purposes to zero. The route is sales of 20 billion pounds a year alongside maturing gilts, which the Bank puts at an average total reduction of about 46 billion pounds a year, concluding around the time the gilts held to maturity unwind in 2034. Two secondary readers state the QT decision was unanimous; the Bank’s notice itself does not carry the vote, and we are reporting the unanimity as their reading rather than as something we read at the Bank.
The notice also splits the holdings three ways: roughly 221.7 billion pounds held to maturity, gilts maturing before 2035; roughly 146.5 billion pounds in the 2035–2049 range identified as potentially subject to Government sales; and roughly 120.0 billion pounds of the longest-dated stock held to back banknotes. And it says the Bank has paused its asset purchase facility auctions pending a review to be completed before April 2027.
That last sentence is the one to notice. A pause in scheduled gilt sales is a change in the supply of duration hitting the gilt market between now and next April. If you trade sterling and you have ever wondered why gilt yields move on days with no data, this is the sort of thing that does it — and it arrived in a document your news filter almost certainly does not subscribe to, because it is a market notice and not a rate decision.
Sterling fell on a hawkish vote, which tells you which half of the pair is in charge
One reader has GBP/USD moving from about 1.3405 to about 1.3360 on the announcement, its weakest since the end of July. We have not been able to second-source that pair of levels — articles published minutes ago are not indexed yet — so treat the two figures as one publisher’s reading and the direction as the part we are confident about.
The direction is the argument. Three MPC members voted to raise rates and the pound fell. Earlier today, before the decision, sterling was trading around 1.3372 with a session low near 1.3370, already below its 200-day moving average, and already at its lowest since 30 July. Two readers put that moving average at 1.3454 and 1.34588 respectively, which is a five-pip disagreement about a line both of them describe as important; if your system uses a 200-day cross as a regime switch, that gap is your error bar and it is not small.
The reason a hawkish vote did not help is on the other side of the pair. The Federal Reserve hiked yesterday to 3.75–4.00%, putting the midpoint of its range at 3.875%. That is above the Bank of England’s 3.75% Bank Rate — one reader says for the first time this year, and the arithmetic is unarguable whatever the anniversary. A committee with three hawks is not a bullish currency story when the other central bank has already moved and its own projections point higher.
The other committee, now settled at the issuing institution
Yesterday this desk flagged an open question and refused to publish on it: one reader stated that the Fed chair had submitted no projection of his own, which would have qualified our own article’s claim that sixteen of eighteen officials want at least one more hike. The same sentence said the 2027 median had moved from 3.6% to 4.1%, and we suspected that 3.6% was a different row of the table.
It settles at the Federal Reserve’s own projections materials, which we read this morning. The footnote is explicit: eighteen participants submitted information in conjunction with the 15–16 September meeting, and one of those eighteen did not submit projections for 2028 and 2029. Not for 2026. Not for 2027. The medians are 4.1% for 2026 and 2027, 3.9% for 2028, 3.6% for 2029 and 3.2% in the longer run, and the 2026 distribution is twelve participants at 4.125%, four at 4.375% and two at 3.875%.
So three things close at once. The claim that a participant sat out the headline year does not survive the primary source. The 3.6% that was supposedly the old 2027 median is the 2029 median — a field collision, the fifth we have documented in ten days and the first on this particular publisher. And the twelve-four-two distribution we published from two secondary readers is confirmed at the issuing institution, which means sixteen of eighteen is now a primary-sourced number rather than our arithmetic.
What this does not tell you
We did not read the Bank of England’s Monetary Policy Summary and minutes ourselves. The Bank’s own page for the September summary was still serving a placeholder saying the document would be published on 17 September when we fetched it, so every vote detail and every forecast figure here is a secondary reading of a document we could not open. That is stated again in the footer.
We do not know the Committee’s inflation path beyond two points — around 3.75% in the fourth quarter of 2026 and slightly above 4% in the first quarter of 2027, on one reader’s account, conditioned on energy prices as at 14 September. We do not know when the Bank expects inflation back at target, because we did not obtain that number.
We do not know whether the gilt-sales pause was priced. We have no gilt yield levels from today at all, which is a real gap in an article about a balance-sheet decision, and it exists because the reaction coverage is minutes old and not yet readable.
And we are not telling you the pound is a sell. We are telling you that a rate decision with a pre-solved headline and an unheralded second decision is the shape of event that hurts systems which size by impact tag rather than by what is actually being announced.
