At some point today the Ministry of Finance will publish how much it spent defending the yen between 27 August and 28 September. This desk has told you four times that the release lands at 10:00 UTC. We said at the pre-US slot that we would go and read that at the Ministry before it printed, and we have. The Ministry does not publish a time. Not on the page that describes the disclosure schedule, not on the data index that carries every previous release. The 10:00 UTC we have been quoting is market convention, corroborated this morning at exactly one calendar, and it has never been a fact from the source. The number it attaches to, on the other hand, matters more than anything else on today’s schedule: the previous window cost 15 trillion 399.3 billion yen, the largest monthly intervention Japan has ever recorded.
What the Ministry Actually Publishes, and What It Does Not
The Ministry’s own explanation of the disclosure regime says two things and only two things: the total amount of intervention is disclosed monthly, and the detail — execution dates, amounts, currency pairs — quarterly. There is no day of the month on that page and no hour of the day. The monthly data index is equally silent: it lists every release back to 2023 with its coverage period, and it states no publication time anywhere.
What the index does give you is the convention, and the convention is readable. The most recent release covers 30 July to 26 August 2026 and is dated 28 August 2026. The one before it covers 29 June to 28 July and is dated 31 July. So the pattern is: published on the last business day of the month, covering a window that closes two or three days earlier. Apply that to today and the window is 27 August to 28 September, which is precisely what one Japanese calendar we read this morning says — it also puts the release at 19:00 JST, which is the 10:00 UTC we have been quoting.
That is worth being exact about, because the distinction is the whole point. The period we published is derived from the primary source and is right. The time we published is not in the primary source at all. It is a convention that one aggregator agrees with. If you have a news filter keyed to 10:00 UTC today, it is keyed to a habit rather than to a commitment, and the Ministry has never promised to keep the habit.
Fifteen Point Four Trillion Yen Is the Number You Are Marking Against
The release dated 28 August put the total for 30 July to 26 August at 15 trillion 399.3 billion yen. That is the Ministry’s own figure, on the Ministry’s own page, and it is a record for a monthly window — a fact carried on the same day by the Japan Times, the Nikkei, Jiji and Xinhua, so it is not a reading anyone needs to take on trust.
Set that against what this desk has actually observed in the window reporting today. On 28 September a Ministry official told the market to take the government’s concern about yen undervaluation seriously and dollar-yen fell 135 pips to 156.51. We published that move. We published, the following morning, that 89 of those 135 pips came back inside the Tokyo session. What we have never seen in this window is evidence of money being spent.
So today’s figure is a clean, binary test of something a system trader should care about a great deal: whether the authorities who spent a record amount in August defended the same handle in September with words alone. A small number, or zero, tells you the 135-pip move was free and repeatable. A large number tells you the official who spoke on 28 September was describing an operation rather than threatening one, and that every verbal warning above 157 has an execution desk behind it.
We are not going to guess which. We will say that the market reaction function is not the same in the two cases, and that a strategy sized for a talking ministry is mis-sized for a buying one.
The Long End Is What the Intervention Is Fighting
Whatever the Ministry spent, it spent it against a bond market that moved decisively the other way yesterday. The US 10-year touched 5.2911 per cent in the New York session, which one Japanese publisher describes as the highest since June 2007, before retreating roughly seven basis points to the 5.22 area. The 30-year did something starker: readers put yesterday’s high at 5.58, 5.59 and 5.61 per cent, and all of them call it the highest since 2002.
We are not going to quote you a single 30-year yield, because our sources give three and they are three basis points apart. Take the band. Any argument that depends on whether the long bond printed 5.58 or 5.61 is weaker than it reads, and the useful fact survives the ambiguity intact: the US long end made a twenty-four-year high on the same day the American consumer survey made a twelve-year low.
That is the shape of the problem for anyone selling dollars against yen on official warnings. Intervention can move spot. It does not move a rate differential, and the differential widened yesterday at the end of the curve where Japanese institutional money actually decides whether to hedge.
Williams Said Both Things, and Only One of Them Was New
The New York Fed president spoke during the session and the two lines that came back to Tokyo were that the year-end may warrant one more rate increase, and that there is no need to act hastily after the September decision. Read those as a pair rather than as a headline. The first is a repetition — one more move by year-end has been the working assumption for weeks. The second is the sentence that did work, and one Japanese desk reports October FOMC expectations receding on it.
That is the distinction this desk keeps insisting on: what changed versus what was merely said again. Williams did not withdraw the hike. He detached it from October. If you are running a rate-decision filter with October in it, that is a probability change and not a direction change, and it is the kind of adjustment that shortens the list of windows you need to be flat for without shortening the list of windows that matter.
What This Does Not Tell You
It does not tell you the intervention figure. We know the window, we know the previous window’s record total, and we know roughly when the Ministry has historically published. We do not know today’s number and zero is a live outcome.
It does not tell you the time with certainty either, and that is the honest result of the check we promised. One aggregator says 19:00 JST. The Ministry says nothing. We have now published 10:00 UTC four times on weaker footing than our own hard rules allow, and this paragraph is the correction.
The yield readings are thin. The 10-year’s 5.2911 print and its June 2007 comparison come from a single Japanese publisher at one remove. The 30-year’s high is multi-sourced on the claim and unresolved on the level. And we have still not read a rate-probability feed directly — the October figures in this file remain somebody else’s screenshot, which is why there is no percentage in this article.
Related
- FX events calendar — the 19:00 JST entry, and everything else on a quarter-end Wednesday.
- Signals
- EA presets
