At 06:16 UTC this morning this desk published that the September Treasury selloff is a belly story and not a long-end one: fifty-one basis points at the three-year since 1 September against thirteen at the thirty-year, a ratio of nearly four to one, and the conclusion that a split that lopsided has to be a repricing of the Fed path rather than of term premium. Two and a half hours later the thirty-year traded at 5.44 percent, which is the highest it has been since 2004. Both of those sentences are true. The second one does not falsify the first — it falsifies the inference we hung on it, and the difference between those two failures is the whole of this article.
The arithmetic is unchanged. The reading of it is not.
Nothing in this morning’s table has moved. The US Treasury’s own par yield curve still shows the three-year at 4.46 on 1 September and 4.97 on 23 September, and the thirty-year at 5.27 and 5.40 over the same span. Fifty-one basis points against thirteen. That is the issuer’s published series and it is not in dispute.
What arrived this morning is a level, not a change. Two readers put the thirty-year at 5.44 percent intraday — one wire desk at 08:37 UTC and a Japanese publisher at 09:32 UTC quoting 5.4439 — and both attach the same characterisation, highest since 2004. A third reader, writing earlier in the session, has 5.39 and calls it a twenty-two-year high, which is the same claim to the year.
A market yield and a par curve point are not the same object and we are not going to subtract them to the basis point. What we will say is the direction: the thirty-year is now trading above where the issuer closed it yesterday, and the level it is trading at has no precedent in the working lives of most people reading this.
Moving least and being quiet are not the same property
This is the part we got wrong, and it is worth naming precisely because it is a mistake a system makes as easily as a desk does. We observed that the thirty-year had contributed the smallest move on the curve and treated small as inactive. But a tenor that begins the month at a twenty-two-year high and rises thirteen basis points from there is not inactive. It is pinned against a ceiling that has not been tested in two decades, and the thirteen basis points are what it took to push through.
The correct statement of this morning’s finding is narrower than the one we published. A 51-to-13 split tells you where the month’s marginal repricing happened. It tells you nothing about which tenor is closest to a level that changes behaviour. Those are different questions and we answered the second one with the first one’s data.
Two channels are open and they do not point at the same trade
The explanations on offer for today’s long-end move are not the explanations for the belly’s. One wire attributes the thirty-year to growth, elevated energy prices, inflation and heavier government borrowing, and notes that Brent jumped on the same morning. Another cites yesterday’s US flash surveys — manufacturing 57.0 against a 53.6 forecast, services 58.7 against 55.8 — with one bank’s economists reading them as consistent with a five percent annual GDP run rate, and adds that the prices firms reported paying were rising too.
That second set is a Fed story and it belongs in the belly, where it already is. The first set — energy and issuance — is a term-premium story and it belongs at the long end, where it has now shown up. Both channels are open at once. A trader who reads this month’s curve as one move with one cause is going to size the next print as though the whole curve responds to it, and roughly half of the curve will not.
One reader also reports that every point on the Treasury curve except the two-year now trades above five percent, and that markets are putting better than even odds on quarter-point increases at both the October and December meetings. We are publishing that probability as one publisher’s characterisation with no feed named, because this desk has been caught twice by probability figures that turned out to be one vendor’s number wearing a definite article.
The inversion we closed yesterday may have closed itself today, and we cannot tell you
Yesterday the twenty-year closed at 5.45 on the issuer’s curve and the thirty-year at 5.40 — five basis points inverted, the fourth consecutive session, and this desk published that as a settled observation at 06:16. If the thirty-year is at 5.44 this afternoon, it sits one basis point below where the twenty-year finished yesterday.
We do not know what the twenty-year did today. The par curve for 24 September is not published until after the close, no reader we checked quotes a live twenty-year, and we are not going to interpolate one. So the honest statement is this: the inversion this desk called a pattern yesterday is, on the only figures available, within a basis point of not existing, and the resolution arrives on one fetch of the issuer’s table tomorrow morning. That is a falsifiable claim and it is the right shape for one — it can only be settled by a source that already exists and will publish on a schedule.
What this does to the next two hours
At 12:30 UTC the United States prints jobless claims at a 201,000 consensus against 196,000 prior, and the second-quarter current account at minus 258 billion against minus 227 billion. Neither is a curve event on an ordinary day. On a day when the long end is at a twenty-two-year high on an issuance-and-energy argument, the current account is the more interesting of the two and it is the one nobody watches.
Four Federal Reserve speakers are on the board — one this morning, one in the same minute as the claims print, one twenty minutes later and one at 14:10 UTC. If you run a news filter that blocks around Fed speech, today it will block most of the New York session. That is not an argument for switching it off. It is an argument for knowing, before 12:30, whether your filter treats a speaker as an event window or as a reason to flatten, because those two settings produce very different afternoons on a day like this one.
What this does not tell you
It does not tell you the thirty-year is going higher. A twenty-two-year high is a statement about the past, and the most crowded position in any market is usually the one with the best backstory.
We have no twenty-year quote for today and therefore no live 20s/30s spread; we said so above rather than deriving one. We have no thirty-year auction between now and the close to read demand from, and no when-issued levels for anything. The 5.44 figure is an intraday level from two readers and an intraday level is not a close; it may not survive the session, in which case tomorrow’s par curve will say so and we will publish that.
And the correction in this article is a correction of reasoning, not of arithmetic. Every number we published at 06:16 is still the number. What we withdraw is the sentence that followed them — that a lopsided split settles the question of what kind of repricing this is. It does not. It settles where the month’s marginal basis points went, which is a smaller and duller fact than the one we sold you this morning.
