25.8 Basis Points Was Never a December Number. It Is 1.05 Hikes.

For five consecutive slots this desk has carried a pair of numbers side by side in its notes and admitted, every time, that they do not reconcile. One reader gave October at 4.3 basis points priced and 15 per cent probability. Another gave December at 25.8 basis points and 89 per cent. The December pair is the problem: 25.8 basis points is 103.2 per cent of a 25 basis point move, and no single meeting can be more than certain. We have been treating that as a unit mismatch. It is not. It is a scope mismatch, the arithmetic closes to within half a basis point, and the resolution arrived this week from two instruments that disagree about every individual meeting and agree about the total.

The only reading of 25.8 that closes

Take the two per-meeting probabilities a single reader published yesterday from the same futures curve: a 15 per cent chance of an increase at the 27–28 October meeting, with 85 per cent on no change, and a 90 per cent chance by December. Add them. That is 1.05 expected increases, which at 25 basis points apiece is 26.25 basis points of cumulative tightening. Against the 25.8 we have been carrying, the residual is 0.45 of a basis point.

Now try the readings we had been assuming. As a December-meeting probability, 25.8 basis points is 103.2 per cent, which is not a probability. As a December increment — 25.8 minus the 4.3 attributed to October — it is 21.5 basis points, or 86.0 per cent, which misses the 89 per cent published beside it by three full points of probability. The cumulative reading misses by less than half a basis point. The incremental reading misses by three points. One of those is a rounding difference and the other is a different quantity.

So the figure was never a December number. It is the expected number of 25 basis point moves between now and the December meeting inclusive, and it exceeds one because two meetings are in the window. The mismatch we flagged five times was our own: we put a cumulative figure in a column headed December and then complained that the column did not add up.

A hundred economists and the futures market agree to a quarter of a basis point

Reuters polled 100 economists between 16 and 20 October. Ninety-four of the hundred expect the target range held at 3.75–4.00 per cent on 28 October. All one hundred expect an increase in December. Convert that to the same units: 0.06 plus 1.00 is 1.06 expected increases, or 26.5 basis points.

The futures market, on the same days, implies 26.25. The two instruments are 0.25 of a basis point apart.

And they are nowhere near each other on the individual meetings. The market puts October at 15 per cent; the economists put it at 6 per cent — a nine-point gap. The market puts December at 90 per cent; the economists put it at 100 — a ten-point gap. This is the part worth your attention, because it is not a curiosity. If you are keying anything off a per-meeting probability, you are keying it off the quantity on which a poll of a hundred forecasters and the entire rates market differ by roughly ten points. If you key it off the cumulative figure, the same two instruments are indistinguishable. The disagreement lives entirely in the allocation between the two dates, and it cancels when you stop allocating.

One caution on the poll, because it is the sort of thing that quietly breaks a comparison: its denominators move between questions. The decision question has 100 respondents. The terminal-rate question has 92, of whom 38 see at least two further increases, leaving the 58.7 per cent who see the peak at 4.00–4.25 per cent. With the range at 3.75–4.00 today, a 4.00–4.25 peak is exactly one more move. So just under three in five of them think December is the last one, and all of them think December happens.

October collapsed by 45 to 50 points and December did not move, so this was not a postponement

A week ago the same reader put the October probability at 60 to 65 per cent. Yesterday it was 15. That is a fall of 45 to 50 points in roughly seven days, and the proximate cause is reported consistently: September housing starts at 1.221 million against a 1.305 million consensus, and pending home sales down 7.3 per cent on the month, both of which this desk covered at the Tokyo slot.

Here is the test that matters, and it is the reason the cumulative figure is worth computing rather than merely reconciling. If a hike is delayed, the probability leaves one meeting and arrives at the next, and the cumulative figure is unchanged. December did not move. It was 90 per cent before and 90 per cent after, and a second reader has it at 93.5 on the same day. So the cumulative figure fell: from about 38.1 basis points a week ago to 26.25 now, a drop of roughly 11.9. Nearly twelve basis points of expected tightening were not rescheduled. They were removed.

That is a materially different statement from “the market pushed the hike to December”, which is how the repricing has mostly been described, including in the Tokyo-hours notes we read this morning. The hike was already in December. What changed is that the market stopped expecting a second one on top of it — and the economists, with 58.7 per cent of them putting the peak one move away, were already there.

The probability is still not a single number, and we are naming the feeds

This desk has a standing rule that central-bank probability figures are not single numbers, and this week honours it. For the October meeting we have three readings within about five points: 15 per cent from one reader of CME FedWatch yesterday, below 20 per cent from a Tokyo publisher this morning, and 17.2 per cent implied by the 4.3 basis points we carried. For December we have 90 per cent from one reader and 93.5 per cent from another, published the same day, 3.5 points apart.

Each of those is usable. None of them is the number. If a rule in your system fires on a probability crossing a threshold, the threshold is crossed on different days depending on which page you read, and in the December case the spread is large enough to straddle a round 90. The cumulative basis-point figure has, so far, been the better-behaved field — which is an argument for reading it, not an argument for trusting it, since we have only just worked out what it measures.

What London inherits

Asia resolved this one by not resolving it. Dollar–yen traded 158.03 to 158.39 through the Tokyo session on one publisher’s reading — 36 pips, against a 78.3-pip day yesterday, and 25.7 per cent of the 157.500–158.900 range the same publisher forecast at 07:39 Tokyo time. The session high sits 18 pips below the 200-day moving average that publisher puts at about 158.57, the level we located yesterday after saying in print that we could not. Euro–dollar managed 13 pips, 1.1223 to 1.1236. Euro–yen managed 44, 177.44 to 177.88.

So the busiest liquidity hours of the day open with the pair inside a third of its own forecast range, under an average it failed at twice, into a schedule that is all talk and no American data. All times UTC: Bank Indonesia decides at 07:20, where two readers are 50 basis points apart on the size and two describe it as an increase to 6.00 per cent on a sliding rupiah — that disagreement dies in roughly an hour. Then Cipollone at 08:00, Bailey at 13:00, EIA crude inventories at 14:30, Lagarde at 16:00, and the United States twenty-year auction at 17:00, which one reader sizes at $13 billion. We have not second-sourced the offering amount and we are flagging it rather than relying on it. There is no scheduled American release today, which means the Fed repricing described above has no data to meet until next week, and the meeting itself is in blackout.

What this does not tell you

It does not tell you the vendor defines the field the way our arithmetic implies. We did not read a methodology page. What we can say is that of three candidate definitions, one closes to 0.45 of a basis point against a figure published independently of it and the other two fail by three points of probability or by being impossible. That is strong evidence and it is not documentation, and if you are building against this field you should go and read the definition rather than take ours.

It does not tell you the economists’ poll and the futures curve agreeing to a quarter of a basis point means anything about accuracy. They agree about the total. Both could be wrong about the total, and the week’s other releases are a reminder of how far wrong a tight consensus can be. Agreement between two instruments is evidence they share an input, not evidence the input is right.

It does not tell you the 45-to-50-point October collapse is permanent, or that the housing data caused it. The attribution is the one every reader we swept offers, it is plausible, and it is not something we measured. Nor can we tell you where the October probability stands right now: our freshest reading is yesterday’s 15 per cent plus a Tokyo note saying below 20, and nobody in this morning’s sweep published a figure after the London open.

And it does not tell you what to do about the twenty-year auction at 17:00, because we cannot confirm its size. One reader, no press release read.

FX events calendar — the 27–28 October FOMC, the 29–30 October Bank of Japan meeting and today’s central-bank speeches, in UTC.
Signals — what a range this narrow does to entry logic built for trend.
EA presets — news-window settings for a day with central-bank speech risk and no scheduled data.
159.00 never traded, and the 200-day is 158.57 — where the moving average above this morning’s high came from.
We argued about 70 basis points; pending home sales missed by 680 — the housing print behind the October repricing.