The Bank of Japan published September producer prices at 23:50 UTC and one line of that release carries the same month of import prices with two different signs. On a contract-currency basis import prices rose 1.0 per cent on the month. On a yen basis they fell 0.2 per cent. The gap is 1.2 percentage points and it is not a price at all — it is the exchange rate, doing enough work in one month to flip the sign of the number Japanese importers actually pay. Exports have the same shape: up 1.4 per cent in contract currency, up 0.5 per cent in yen. And the headline this desk published a month ago as a beat that was really a slowdown is now the prior.
The print, from the issuer, not from a wire
Japan’s domestic producer price index came in at 135.7 on the 2020 base, up 6.8 per cent on the year and down 0.3 per cent on the month. August is shown at 7.6 per cent year-on-year, marked revised. The release went out at 08:50 Tokyo time, which is 23:50 UTC yesterday.
We read this at the Bank of Japan’s own release rather than through a summary, and the reason is visible in the difference. The wire note that reached us first gave 6.8 per cent and said it was slower than the previous month, with no prior figure, no monthly change and no index level. The issuer gives all three, plus the export and import series that make this piece.
On 11 September this desk published that the 7.6 per cent August reading, which everyone had filed as a beat, was in substance a slowdown. That figure is now the prior, and September took another 80 basis points off it. More than that, the monthly change is negative. Japanese producer prices did not merely decelerate in September; they fell.
One month of import prices with two signs
The interesting number is not the headline. It is that the same release reports import prices up 1.0 per cent month-on-month in contract currency and down 0.2 per cent month-on-month in yen. Two signs, one month, one dataset, one page.
Mechanically there is nothing wrong here, and that is the point. The contract-currency index measures what the goods cost in the currency the contract is written in. The yen-basis index measures what they cost after conversion. The difference between them is the currency, and in September the currency moved enough to turn a 1.0 per cent increase into a 0.2 per cent decrease.
Exports show it in the same direction and smaller: up 1.4 per cent in contract currency against up 0.5 per cent in yen, a 0.9-point gap. Both wedges point the same way, which is what you would expect if one variable is responsible for both.
Now the refusal. The obvious reading is that the yen was stronger on September’s relevant average than on August’s, by something on the order of a point. We are not asserting that, because the release does not publish the average rate it used and we have not reconstructed one. If you are building anything on the sign flip, the number you need is a monthly average exchange rate, and it is not in this document.
Twenty-four and a half against six point eight
The year-on-year figures are where the cost-push story actually lives, and they are further apart than the monthly ones by an order of magnitude. Import prices on a yen basis are up 24.5 per cent on the year. Domestic producer prices are up 6.8 per cent. That is a 17.7-point wedge. Exports in yen are up 17.9 per cent, an 11.1-point wedge against the same domestic figure.
Whatever is arriving at the Japanese border is not reaching the factory gate at anything like the same rate, and the monthly domestic figure has now gone negative while the annual import figure is still running near a quarter. For anyone trading the yen on an inflation thesis, those two series have been telling different stories for long enough that the choice of which one you quote is doing most of the work in your argument.
What we are not going to do is convert that into a policy call. A central bank with a 1.25 per cent policy rate and a meeting on 29 and 30 October is not obviously moved by a producer price series decelerating from a very high base, and this desk still carries five incompatible market probabilities for that meeting — roughly 10, 12, 17, 24.8 and 25 — plus a sixth reading of a different kind, a path of three increases by July next year. When your own file holds five numbers for one binary event, you do not have a view. You have a spread.
The session took the print the wrong way round, and Tokyo finally produced something
A softer producer price print is, on the lazy reading, yen-negative. The yen rose. One publisher framed it as strength “despite” the data, which is honest about the direction and silent about the cause.
Dollar–yen opened the Tokyo session near 158.60, ran 158.65 at the top and 158.19 at the bottom, and was in the 158.30s by 05:46 UTC. That is a 46-pip range. Yesterday, at this slot, the reopened Tokyo session produced eleven pips and nothing worth publishing, and this desk wrote down that a reopened session is not automatically a session with a story. Today it is. Two slots, two opposite outcomes from the same premise, which is a reasonable argument for not having the premise.
Euro–yen ran 176.98 to 176.70, with the low 24 pips below the 176.94 close one reader published for Monday. Euro–dollar ran 1.1153 to 1.1177 and was near 1.1175 in the early European session, about 20 pips above Monday’s close. So the yen is firmer against both and the dollar is softer against the euro: the yen leg is real but it is sitting on top of a broader dollar give-back, and sizing a yen trade as if the whole move were Japanese would overstate it.
Intervention is still the live tail. The 159.04 high of 24 September remains the level this desk carries as the point at which official concern sharpens, and at 158.19 the pair is 85 pips below it. One publisher reports traders on alert for action to support the yen. That alert is cheaper to hold this morning than it was on Monday, which is the practical consequence of a 46-pip session going the way it went.
What this does not tell you
The sign flip is arithmetic, not a finding about the economy. Two indices built on the same goods with different conversion rules will diverge whenever the currency moves, and the only new information is how much it moved. We have not quantified that, and we have said why.
The August figure is marked revised at 7.6 per cent, which happens to be the value we published on 11 September. We have not established whether the original print differed, so we are not claiming a revision changed anything — only that the issuer labels it revised.
The 6.8 per cent is a preliminary figure. Producer price series get revised, and the whole of this piece’s monthly argument rests on a 0.3 per cent decline and a 1.2-point gap, both small enough that a revision could move them materially.
The Tokyo range is one publisher’s reading, taken twice about forty minutes apart. A second publisher corroborates the current level and Monday’s close but not the extremes, and that second publisher carries its own spot price at two values 1.4 pips apart on one page. And the stated driver across the session was crude oil, which three sources priced across a band wide enough this morning that we have not used it as a cause.
Finally, no consensus figure for this release was available to us before the print. We are therefore not scoring anyone’s forecast, and we are not pretending this was a surprise or a confirmation. It is a number, read at the issuer, with its own prior attached.
Related
- Economic calendar and event windows — the Japanese releases and the 29–30 October Bank of Japan meeting.
- Signals — yen-cross behaviour through the Tokyo session.
- EA presets — sizing when a cross move is mostly the other leg.
