The Federal Reserve Board has published the text of Governor Christopher Waller’s 10 November speech in London, and this desk has now read it in full at the issuer rather than through a wire. Three things come out of it. The first is a retraction: a line we have carried in our own working notes for four consecutive slots, and which rotation twice kept off the page, is not in the speech. The second is a field problem of the kind we keep finding in other people’s data and have now found in a central banker’s own argument — the condition he set for a 50 basis point December move is written in one inflation index, while every inflation figure he uses to justify it comes from a different one. The third is a single sentence about the sequencing of hikes that should change how anything you have keyed to December treats January.
The quote we carried for four slots is not in the text
Since 9 November our internal carry list has held an item described as Waller’s “October hold was a choice” lines, flagged as issuer-sourced and deferred three times for want of a byline. It was due to be printed or formally dropped at this slot. It is being dropped, because the speech does not say it.
What the speech says is narrower and differently shaped. He states that he supported the Committee majority in holding policy at its current setting, and that he was “willing to wait to tighten monetary policy” at the October meeting. His stated reason is that the Committee’s own communication had already led financial markets to price in some tightening — so waiting cost nothing, because the tightening was already in the price.
That is a claim about the sufficiency of communication. It is not a claim that the hold was finely balanced, and it is not the characterisation we were carrying. The two readings point different ways for a trader: a close call implies the next one could go either way, while “the market has already done the work” implies the decision was comfortable and the December move is the one that matters. We had the second mechanism filed under the first description, and we would have published it that way.
The lesson is not that the wire was wrong. We never read the wire’s sentence against the speech. An item sitting on a carry list for four slots marked “issuer-sourced” was in fact sourced to a summary of the issuer, and nothing about its position on the list made that visible. The label was doing the work the reading should have done.
The trigger is written in CPI. The entire inflation case is in PCE.
Here is the whole of the speech’s inflation evidence, as stated for September data: headline personal consumption expenditure prices rose 0.5 per cent on the month and 3.7 per cent over twelve months; core prices excluding food and energy rose 0.3 per cent on the month and 3.1 per cent over the year; and more than 70 per cent of the components of core services prices showed twelve-month increases above 3 per cent. That last figure is the argument. It is a breadth statistic, and breadth is what distinguishes a price level pushed around by a few volatile lines from one that has spread.
Now the condition. He said that if the October consumer price index data due later that day came in similar to the past two monthly readings, with core CPI inflation around 0.3 per cent or even higher, a 50 basis point hike should be considered in December. Core CPI printed 0.3 per cent. The condition fired.
But notice what fired it. The threshold is set on the consumer price index. The diagnosis — the level, the twelve-month rate, the breadth count that carries the whole case — is entirely in the personal consumption expenditure deflator. Those are different baskets, different weights, different compilers and different revision practices. The headline gap on the day was 0.6 points between his own two PCE readings; the gap between the two indices is a separate thing again, and he did not state it.
This is the defective-field class we have been logging against vendors all month, and it is the sixteenth instance in our file: a number that is right, attached to a field that is not the one the argument runs on. The operational consequence is specific. If you hold a rule that says “Waller’s inflation threshold” and it reads a PCE release, it will never fire, because the threshold is not a PCE threshold. If it reads a CPI release, it fires on an index he did not use to argue his case. Both versions are defensible and they are not the same rule, and the row in your calendar that says “Fed inflation threshold” does not have a field for which index it means.
The sentence that should rewire your December-to-January assumption
The most useful line in the speech has nothing to do with the level of rates. On the sequencing of hikes he says that when communication is clear and credible, what really matters is that the hikes are eventually carried out, and that it is not necessary, for example, to raise rates at consecutive meetings. He adds that while the precise timing does not matter, implementing the promised hikes in a reasonable period does.
Read that against how December is actually priced. One reading of the CME FedWatch tool, carried by a wire on Wednesday, put the probability of a 50 basis point increase at 10.4 per cent. We are publishing that as one reading with the feed named, because that is our rule and because the figure reached us through a page summary rather than verbatim text, and it carries no stated measurement time. Our own pre-London piece on Wednesday put the same branch at about 11 per cent after the print. The two are consistent and neither is “the” probability.
The point is what any of those numbers does and does not tell you. A December probability is a statement about one meeting. Waller has just said in terms that a December move carries no implication for January, because the Committee’s obligation in his framing is to the eventual total, not to the spacing. If you run anything that treats a high December probability as raising the probability of a January move — a carry model, a correlated-exposure cap, a news blackout that widens after a hike — the issuer has told you the spacing is free. And his own base case remains 25 basis points: the half point is to be “considered”, which is a word that does a lot of work and should not be read as a commitment.
A number of ours is confirmed, and the issuer flags its own weak source
One carried item closes cleanly. On 10 November we derived, from the payrolls release, that the two prior months had been revised up by about 84,000 jobs in total, and we noted we had not confirmed it against a Board document. The speech states 85,000 jobs added to estimates of the previous two months. Our derivation was 1,000 light on a figure of that size. The item was on its third asking and it is now closed against the institution.
The rest of the labour picture, for the record and all from the same text: real gross domestic product rose 3.4 per cent in the third quarter on the initial estimate, with growth expected to back off somewhat in the current quarter. Unemployment ticked down to 4.1 per cent in October, the same rate as over the summer. October payrolls gained 165,000 against six- and twelve-month averages closer to 80,000 a month, and September is cited at 60,000. More than five million people left jobs in September and more than five million were hired, which is the figure that makes a net 60,000 look like what it is. And the three-month average share of the unemployed out of work for 27 weeks or more has risen from around 18 per cent at the end of 2022 to more than 27 per cent through September 2026.
One figure in the speech deserves naming for how it is presented rather than for its value. On energy, the text says insurance and ship-to-ship transfer costs are “reportedly” boosting oil prices by $20 to $30 a barrel. That is a Federal Reserve governor, in prepared remarks, flagging his own sourcing as secondhand on a $20 to $30 range. We have spent this month writing that an issuer is authoritative about its process and merely assertive about its clock; this is a third category — an issuer being explicitly non-authoritative about a number inside its own argument, and saying so. Treat the figure the way its author did.
What this does not tell you
- We cannot tell you the gap between CPI and PCE inflation on comparable terms, because the speech states PCE figures and a CPI threshold and never puts them side by side. We did not construct the comparison ourselves, because the two indices are not reconcilable from the figures given.
- We cannot tell you the policy rate level or the target range. The speech does not state either, and we are not supplying it from elsewhere in a piece sourced to one document.
- The speech does not mention the neutral rate at all, so the long-running question of where the staff estimate sits is untouched by it.
- The 10.4 per cent figure for the half-point branch reached us through a page summary rather than through verbatim text, and carries no measurement time. We publish it as a single reading with the feed named and would not build a sizing decision on it.
- The October CPI print that fired the condition was 0.3 per cent on the month. We are not restating the year-on-year field, because the release carried none, as we published on Monday and again on Tuesday.
- We cannot tell you whether the dollar index closed at a new high, because three publishers said on Wednesday that it rose and not one of them printed a number. This is the second category of absence we have had to state in print this week, and it is cheaper to say so than to source a level from a chart.
- We formally drop, with no further discussion, the claim of a coordinated Japan–United States intervention that has sat on our carry list untouched for seven slots. It was never second-sourced, it has not improved with age, and leaving it on a list is not the same as having it.
Related
- FX events calendar — release times in UTC, including the December meeting dates.
- Signals — how we frame a reaction function as a sizing question.
- Presets — filter and exposure settings for scheduled policy events.
