British monthly GDP for August printed at plus 0.1 per cent. One vendor’s calendar carried a consensus of minus 0.1; Deutsche Bank, writing on 8 October over two named economists, had it flat at zero. The print beat the lower figure by two tenths and the higher by one, which is to say it cleared the entire published range of expectations we could read. It also landed at 06:00 UTC, which is 07:00 in London, which is the minute the London session opens — so whatever sterling did in the hour that followed, the release and the open are not separable in it. And the issuer announced in advance that this is the release that reopens the whole time series to revision, which means the 0.4 per cent prior sitting in your calendar may not have survived the print you are scoring against it.
Two published expectations, one tenth apart, and the print cleared both
The two numbers we could actually read behind, rather than off a headline:
- A vendor economic-calendar line for Thursday 15 October at 06:00, consensus minus 0.1 per cent, previous 0.4.
- Deutsche Bank Research, published 8 October through a vendor’s insights desk, led by its chief UK economist with a named colleague: flat on the month, zero per cent, framed as a catch-down after a strong July.
Print: plus 0.1. Against the calendar consensus that is a two-tenth beat; against the bank it is one tenth. The two expectations are themselves only a tenth apart, which is narrow, and both of them were on the wrong side of zero relative to the outcome.
Note what these two objects are, because this desk has been caught by exactly this before. One of them is a consensus — a vendor’s aggregation of forecasts, with no contributor list published on the page we read. The other is one house’s view with its authors named. They are not two samples of the same thing and we are not averaging them. What they jointly establish is narrower and sufficient: no reading we could obtain had Britain growing in August, and Britain grew.
We could not find a third readable expectation. On this desk’s own rule, a two-source range is a range and not a distribution, and the honest summary is that the published expectation set we can defend is two numbers wide.
The prior is not safe, and the issuer said so a month ago
The July bulletin states, in terms, that in that release “there are no periods open for revision” and that the full time series would be open for revision in the 15 October release. That is today.
So the previous value your calendar is showing — 0.4 per cent for July — is a figure published under a closed-revision rule, being used as the reference point for a release published under an open-revision rule. Nothing we read this morning says whether July survived at 0.4. The vendor report of the print describes it as following “July growth of 0.4 per cent” and does not say whether that is today’s July or last month’s July.
This matters more than it sounds. A plus 0.1 after a plus 0.4 is a sharp deceleration. A plus 0.1 after a revised plus 0.2 is something else entirely, and the three-month path — which the issuer also publishes and which no vendor page we read quoted this morning — is where the Bank of England actually looks. If you are running an event filter that scores a print against the prior in the same calendar row, today is a day when that row is internally inconsistent by construction and the issuer told you a month in advance that it would be.
The release has no time at the issuer
We read two of the issuer’s own GDP bulletins. Both give the next release as a date. Neither gives a time. Neither gives a timezone. The 06:00 in your calendar came from a vendor, and the only corroboration we have for it is circumstantial: a vendor article reporting the figure carries a 06:02 stamp, two minutes after.
And there is a second reading of the same issuer that disagrees with the first. At roughly a quarter past six this morning, the issuer’s own standing link for the newest monthly GDP bulletin was still serving July — fifteen minutes after a vendor had published August. We are not going to tell you which of those two is the issuer’s real state. We will tell you that we read both of them within a quarter of an hour and they describe different months, and that the one carrying the number everyone is trading is the vendor’s.
The operational consequence is a timing one. If your news window for this release opens at 05:58 and closes at 06:05 because the calendar said 06:00, that window rests on a vendor’s editorial field and not on anything the Office for National Statistics published. On a release that collides with the London open, a window that is a couple of minutes wrong is a window pointed at the wrong event.
Industrial production beat nothing, because nothing was published
The same release carried industrial production at plus 0.8 per cent on the month and manufacturing at plus 1.2. The page we read gives no consensus for either and no prior for either.
Against the issuer’s own July figures — production plus 0.2, manufacturing plus 0.9 — August is running six tenths and three tenths faster respectively. That is a comparison we constructed from the issuer, not one the vendor offered, and it is the only frame available this morning for two numbers that will nonetheless show up in every post-release commentary as beats.
They are not beats. A figure with no published expectation cannot surprise anybody in the sense that a model means by surprise, whatever it does to price in the following ten minutes. If your system sizes off a surprise index, these two releases contribute nothing to it today and should not be made to.
For completeness on the price: a vendor article timestamped 06:02 UTC put sterling–dollar at 1.3242, up 0.37 per cent on the day. The same publisher, an hour and eleven minutes earlier, described the pair in the mid-1.3200s after bouncing from a year-to-date low around 1.3135, and reported the Bank of England’s chief economist saying that a hike now should not give the impression it would be the start of a series. We are quoting those as two readings with times on them, not as a reaction function.
What this does not tell you
It does not tell you that minus 0.1 was the consensus. It tells you one vendor’s calendar carried minus 0.1. We did not reach a second aggregator, and this desk has been wrong before by treating one vendor’s consensus field as the market’s expectation.
It does not tell you whether July was revised. We flagged the open revision window because the issuer published it; we did not read today’s bulletin, because the issuer’s standing link was still serving last month’s when we looked. That is a limitation in our reading, and the fix is to check the dated bulletin later today rather than to assume either way.
It does not tell you the three-month growth rate, the services contribution or the construction figure. None of those appeared on any page we could read this morning, and they are the components that will decide how this print is actually interpreted.
It does not tell you what sterling will do. The release and the London open happened in the same minute, so the thirty-seven basis points on the day that one vendor quotes at 06:02 contain the print, the open, and an overnight dollar move that was already running before either. We have no way to decompose that and neither does anybody quoting it. There is no directional view in this article.
Related
- Economic calendar and event windows
- Signals
- EA presets
- Eight Readers Agreed on One Field. It Is the One They All Got Wrong. — the same consensus problem, on US CPI.
