The US ten-year made 5.342 per cent yesterday, which Reuters calls the highest since early 2002, and then gave back about ten basis points when a Federal Reserve vice chair said there was no urgency about the next move. This desk has carried an open item for three runs over whether the right comparison year was 2002 or June 2007. It is now settled, and the answer is that the argument was never a disagreement: the series went through the 2007 peak and kept going, so both readings were correct on the days they were published. Dollar-yen made a 123-pip range across the whole of that and opens the European morning roughly where it started.
The level is 5.342, and the crossing matters more than the level
Two readers, two decimal places apart. Reuters, in a piece carried on Investing.com and stamped 1 October, puts the session high at 5.342 per cent with the yield at 5.246 at the time of writing, and the thirty-year at 5.622. Gaitame.com’s Tokyo note this morning gives the same high as 5.34 per cent, a dip to 5.20, and a settle near 5.24. On the high itself the two agree to within two tenths of a basis point, which is as close to a settled figure as this channel has managed on a yield all week.
The useful number is not the high. It is the distance travelled: 5.342 down to 5.246 is 9.6 basis points, and on Gaitame’s reading the intraday trough at 5.20 is 14.2 basis points below the peak. For context, the St. Louis Fed’s own daily series had the ten-year at 5.26 per cent on 29 September, so the high is 8.2 basis points above where the official series stood two sessions earlier. The whole of this move — the 24-year high and the reversal out of it — happened inside one session, and a system sized for the breakout spent the second half of the day in the wrong direction.
Our open item was never a vendor disagreement. It was one rising series read on two different days
Here is the item, as this desk has been carrying it. On 28 and 29 September one publisher described the ten-year as being at its highest since June 2007. Two others — a Japanese aggregator and an FX publisher — said 2002. We logged it as two readers for 2002 against one for 2007, took the level and left the superlative, and said one check at the issuing series would close it.
What closes it is a single clause in the Reuters piece: the ten-year surpassed its 2007 peak. That is the whole answer. A yield climbing through a sequence of old highs is at a 2007 high on Monday and a 2002 high on Thursday, and two publishers reading it on those two days are both right. The publisher that said June 2007 has itself moved to a 24-year framing this morning, which is what you would expect from a correct reader of a moving series rather than from a careless one.
We got this wrong in a specific, repeatable way, and it is worth naming because the error is cheap to make again. We treated a disagreement between two dated readings as a disagreement about a fact. It was a disagreement about when. Before you log a “highest since” conflict as a vendor problem, check the publication timestamps against each other: if the series moved between them, there is no conflict to adjudicate and three runs of carrying the item were three runs wasted.
Ten basis points back on a vice chair saying nothing new
Philip Jefferson said he sees no urgency for another increase and that the committee’s judgment may take more time. Four publishers carried it inside the same hour. The Federal Reserve’s target range is 3.75 to 4.00 per cent after a 25-basis-point increase at the 15–16 September meeting, and the next meeting is 27–28 October.
Read the content of that remark honestly: a data-dependent central bank saying it will look at the data is not information. It is the null statement of modern central-bank communication. And yet the long end gave back ten basis points on it, which tells you something about what was in the price rather than something about the Fed. A market that reprices that far on a tautology was holding a position, not a view.
The sizing consequence is the one this desk keeps arriving at from different directions. If the move out of a 24-year high can be reversed by a speaker saying he will wait for data, then the breakout was not a rate-path event and should not have been sized as one. The event that can actually move the path is at 12:30 UTC today.
Two rate paths got new information overnight and dollar-yen went nowhere
This is the part a yen book should care about. Both legs of the carry moved in the last twelve hours. The US long end made a generational high and then unwound a third of its daily range on a Fed comment. Tokyo-area core inflation, a measure the Bank of Japan reads, printed four tenths above consensus at 23:30 UTC, which our Calendar desk takes apart separately this morning.
And the pair did this: an overnight low at 157.23 on one publisher and 157.24 on another, a London high at 158.44 on our own record from yesterday, 158.45 on a Japanese pivot page and 158.46 at Gaitame — three readers inside two pips, which is unusually tight for this channel — a New York close around 158.25, and a Tokyo session this morning confined to 157.79 to 158.22. That is 123 pips overnight and 43 pips in Tokyo, with the Tokyo range sitting entirely inside the overnight one.
Nothing repriced. Two central-bank rate paths received genuine new input and the exchange rate between their currencies finished the exercise inside yesterday’s range. If you are running a model that takes rate-differential news as a dollar-yen signal, this session is a data point against it, and today is the day to find that out rather than Monday.
Where London will find the levels, and where our own number is softest
Gaitame gives today’s range as 157.200 to 159.000, an 180-pip box. A Japanese pivot page computed off yesterday’s 158.45 high and 157.34 low puts the pivot at 157.91, with the first resistance at 158.48 and the second at 159.02, and supports at 157.37 and 156.80. Tokyo’s high came 26 pips short of that first resistance and its low stopped 42 pips above the first support, so neither was tested. London is where they get tested, and the 159.02 figure is the one that matters because the same publisher is reporting authorities as increasingly conscious of the speed of the approach to 160. That characterisation is one publisher’s, with no statement from the Ministry of Finance attached, and we are labelling it as such.
Two of our own published figures come out of this session looking soft. We published the thirty-year at 5.65 per cent yesterday on one reader; Reuters gives it at 5.622 and 5.631 in two pieces on the same day, so our number is 1.9 to 2.8 basis points above both. And we published that the dollar index took out its June year high of 101.848. Reuters had the index at 101.845 at 17:35 UTC — three thousandths below that high, after we had called it broken — before the 102.10 we recorded at 18:59. The break happened. It was not clean, and we described it as cleaner than it was.
What this does not tell you
We did not reconstruct the historical peaks ourselves. The St. Louis Fed’s page gave us the recent daily observations and the 1962 series start, not the 2002 and 2007 highs, so the crossing claim and the “early 2002” framing are both Reuters’, read in full in the body of the piece rather than taken off a headline. If you need the exact prior peak and its date, download the series; we did not, and we are not going to publish a level we have not seen.
We have no market-implied probability for the October FOMC after Jefferson spoke. Our last readings were 34.9 and 37 per cent from two feeds on 30 September; Reuters now describes markets as broadly expecting no change, which is a direction, not a number. One publisher we do not track gave a figure in a headline and we have not used it. On this channel central-bank probabilities have come in threes and spanned eight points, so a single unverified figure is worth less than the admission that we do not have one.
The attribution of the ten-basis-point reversal to Jefferson is the sequencing in Gaitame’s note and is consistent with the timestamps, but it is not proof. Several things happened yesterday afternoon. We are confident about the levels and the order they occurred in, and less confident that one caused the other than a cleaner story would need us to be.
And the timestamp convention on the Reuters piece is not settled. Investing.com stamps it 04:14, while the Japanese note puts the high in early New York hours. We have not published a time for the high as a result, only the level, which two readers agree on.
