At 14:30 UTC yesterday the Federal Reserve Bank of Dallas released its Texas Manufacturing Outlook Survey for September. The general business activity index printed 9.8, down from 11.6. Six of the seven sub-indices underneath it went up, two of them by double digits. Forex Factory and Minkabu do not carry the release at all — Minkabu’s New York wrap says in terms that no major US indicator was released — and FXStreet, which this desk praised eleven hours ago for having the time right to the minute, still shows the actual value as “n/a”. Nearly nine hours after the print, the only place the number exists is the issuing bank’s own web page.
The headline is the one number pointing the wrong way
Here is the report, as the Dallas Fed published it. General business activity 9.8, from 11.6 in August. Production 29.5, from 16.1 — up 13.4 points. New orders 30.7, from 22.0. Employment 15.1, from 8.0. Prices paid for raw materials 52.2, from 44.1. Prices received 27.6, from 22.7. Wages and benefits 27.4, from 21.1. And company outlook 8.7, from 19.2 — down 10.5 points.
Six of seven sub-indices rose. The two that fell are the headline and the outlook. The bank’s own summary line is that output growth “accelerated sharply” while the outlook moderated, and the numbers support both halves of that sentence without strain.
Now consider what a calendar row would have shown you. One field, labelled Dallas Fed Manufacturing Index: 9.8, previous 11.6. A reader who saw only that row would book September as a deterioration in Texas manufacturing. The report says production accelerated by more than thirteen index points and orders piled up, and that the thing which deteriorated was how factory owners feel about the next six months while paying eight points more for raw materials. Those are not the same story and they do not trade the same way. If you are running a filter that reacts to a headline index against its prior, this release is a live example of a row whose sign is opposite to its content.
The aggregator that had the time right still has no print
Yesterday morning this desk published an article headlined “Nothing Prints Today,” sourced to two calendars that agreed. Yesterday at 11:52 UTC we corrected ourselves: the release existed, the Dallas Fed’s own survey page gave 9:30 a.m. Central, and FXStreet’s event page gave 14:30 UTC. We called FXStreet the independent second reader that agreed with the primary to the minute, and we were right about that.
It is now nearly nine hours past the release and that same FXStreet event page carries previous 11.6, consensus n/a, actual n/a. The countdown has run out and the field is empty. Forex Factory and Minkabu never had the row. Minkabu’s summary of the New York session lists Cook, Hassett and Lagarde and states that there were no major US economic indicators — which is, on its own terms, wrong, and wrong in a way no reader of that page could detect.
So the rule we amended yesterday holds, and it holds harder than we wrote it. We said that a negative claim about a scheduled series is settled at the issuing institution, not at a second aggregator, because two aggregators agreeing may only mean they share a gap. The extension is that carrying the correct release time is no evidence that a vendor will carry the release value. Those are two separate pieces of coverage and this event splits them cleanly. If your process reads the print off a calendar page, the print for this release does not arrive.
The day’s biggest number never traded
Yesterday morning this desk published a 4.5-billion-euro option expiry at EUR/USD 1.1400 for the 14:00 UTC New York cut, sourced to a mirror of an option-expiry page which called it “the standout” and said it was “both exceptionally large and almost sitting on top of current price.” We said in print that where spot sat at 13:55 and at 14:30 was the whole test. Marking it now: EUR/USD’s entire range for 28 September was 1.1353 to 1.1395. The high was five pips below the strike. The low was forty-seven pips below it. Spot was at 1.1371 at 02:00 JST, twenty-nine pips below.
Price never reached the strike. Not at the cut, not at any point in a forty-two-pip day. The largest single notional on the board was never in play, and the test we set has a null answer: you cannot observe a pin at a level price does not visit. That is worth saying plainly rather than quietly dropping, because the expiry was published here as a reason to expect something around 14:00 and nothing around 14:00 was attributable to it. A strike five pips away is close enough to look decisive on a page and far enough to be irrelevant on a tape.
Tomorrow at 04:30 the decision is made and the probability is three weeks old
The Reserve Bank of Australia announces at 04:30 UTC — 14:30 Sydney — with Governor Bullock’s press conference an hour later at 05:30 UTC. One preview gives the timing as 05:30 BST and 00:30 EDT, which is the same instant, and the three renderings agreeing is a small mercy on a channel that has spent a week on timezone faults.
The cash rate is 4.35% and the expected outcome is 4.60%. All four major Australian banks forecast the hike. Money markets are described as having it fully priced. Headline CPI printed 3.5% year on year against 3.3% expected and the trimmed mean 3.6% against 3.5%, both above the 2–3% target band, and the Governor has said some of the upside risks to inflation appear to be materialising.
This desk will not publish a current probability for this decision, because it does not have one. What it has is 80% from interbank cash-rate futures dated 7 September, and about 90% attributed to one of the four banks in an article dated 21 September. Those figures are three weeks and eight days old respectively and they are not in conflict — they are two points on a rising series, and the containment check that would turn them into a disagreement fails on the dates alone. Reporting them as a spread would be manufacturing one. So: no probability.
What is worth your attention is the one thing the four banks disagree about. Two of them expect a unanimous vote; a third expects a split. Two of them flag the risk of a follow-up to 4.85%. The level is decided and the vote count and the terminal are not, which means the tradeable content of this decision is in the statement wording and the presser, not in the number that hits the wire at 04:30. If you are sizing this event, size the 05:30 window, not the 04:30 one.
What this does not tell you
The Dallas Fed figures are a single primary source read once. They are the issuing bank’s own page, which is as good as sourcing gets, but no second reader has confirmed the sub-indices and the aggregators that would normally provide one do not carry the series. We published no consensus for this release yesterday and we still have none, so we cannot tell you whether 9.8 was a beat or a miss — only what it was and what moved underneath it.
The EUR/USD extremes come from one publisher’s session summary. A forty-two-pip range is narrow enough that a five-pip gap to the strike is within the distance two vendors routinely disagree by on a session high, so the claim that price never touched 1.1400 is a claim about one publisher’s tape rather than about every tape. It would take a second reader to make it a fact, and we have not got one.
The RBA figures are a preview and a bank-forecast summary, not the decision. Consensus being unanimous among four houses is not the same as being right; this channel has spent three weeks documenting weeks where every point estimate in the file missed. And the FXStreet field may well be populated by the time you read this — the claim above is about its state nearly nine hours after the release, which is when a system that trades the print would have needed it.
Related
- FX events calendar — release times in UTC, including the 04:30 RBA and the Wednesday month-end block.
- Signals — how we treat a headline index whose sign disagrees with its internals.
- EA presets — news-window configuration for a release your calendar feed may not carry.
