For five consecutive runs this desk has carried a single figure for the French–German ten-year yield spread: 130.3 basis points. It appeared in our files as a level, the way a central-bank policy rate appears as a level, and at each slot we logged the question of whether it was still there and then wrote about something else. Today we went and looked. It is not at 130. It was at 140.6 five days ago, at 127.4 yesterday, and it is somewhere between 135.5 and 138 as London hands over. Our number was not a level. It was the middle of a thirteen-basis-point range.
The series, and what it does to a figure we published five times
Four observations, each with its publisher and its stamp, because the whole point is that they disagree about the state of the world on different days rather than about the same day.
- 2 October, 03:36 UTC: 140.6 basis points, described as the widest since the 2012 euro-area debt crisis, with the French ten-year at 4.935 per cent and the German at 3.529. One publisher. We checked the identity: 4.935 minus 3.529 is 1.406, so the spread and its two legs agree exactly.
- 6 October: 127.4 basis points, French ten-year 4.72 per cent, German 3.45, described as sixteen basis points narrower on the session. A different publisher. Here the identity gives 127.0 against a stated 127.4, a four-tenths gap that is what rounding two yields to two decimals does and not an error.
- 7 October, 07:46 UTC: 135.5 basis points, French ten-year 4.858 per cent, up from 4.703 the previous day. A third publisher, which independently characterises the prior level as about 127 — so two readers agree on yesterday to within half a basis point, which is the cleanest corroboration this file has had on this series.
- 7 October, 09:20 UTC: about 138 basis points, same publisher, same French yield, later filing. So it widened a further two and a half basis points inside the London morning.
The arithmetic on our own figure: 130.3 sat 10.3 basis points below the high of the window and 2.9 above the low. The four-session range is 13.2 basis points — down 13.2 from 2 October to 6 October, then back up 10.6. If you held a view that depended on the spread being “around 130”, you were right on two of those four days by accident and the number never rested anywhere.
We are not going to dress this up as a vendor failure. Every one of those readings is probably correct for its moment. The failure is ours, and it is a specific and common one: we filed a fast-moving price as a reference constant because it arrived in a sentence rather than on a chart.
The German leg barely moved, so this is not a euro-area measure today
The spread is routinely described, including in the reports we read this morning, as a barometer of European credit risk. Over this particular window it is nothing of the kind, and you can show that with subtraction.
Take the 135.5 spread and the 4.858 per cent French yield from the 07:46 UTC filing. That implies a German ten-year of 3.503 per cent. Against the 3.529 reported on 2 October, the German leg has moved 2.6 basis points — in the opposite direction to the spread. Over the same five sessions the French leg moved 7.7 basis points, also lower. The denominator is effectively static and the numerator is doing all of the work.
That implied German yield is ours, derived, not reported: none of the three publishers printed a German ten-year for today, and we are flagging it as a derivation rather than a quote. But the direction of the conclusion does not depend on the last decimal. A measure of two things in which one of them does not move is a measure of one thing. Today the spread is a French political and fiscal story wearing euro-area clothing, and a system that treats a widening as a signal about the currency bloc is reading a single sovereign’s paper through a two-legged instrument for no reason.
The euro paid six pips a basis point today and paid nothing for a data beat yesterday
This is where it becomes a sizing question rather than a bond-market observation.
Yesterday morning this desk published that German August industrial production came in at plus 2.0 per cent month on month against a consensus of plus 0.5, a beat of a point and a half and a 3.1-point swing from July’s minus 1.1 — and that the euro was lower on the day anyway, near 1.1225 and down 0.30 per cent. We filed it as a headline that produced no flow.
Today, with no euro-area release of any consequence, euro–dollar fell to 1.1198 by 09:20 UTC, between fifty-six and sixty-five pips off its Tokyo high depending on which print you take, and euro–yen fell to 177.13, some 117 pips off its Tokyo high. Against a spread widening of 10.6 basis points from yesterday’s level, that is roughly 6.1 pips of euro–dollar per basis point. We offer that ratio as a single day’s measurement and emphatically not as a coefficient — one session is one session, and we will re-measure it rather than reuse it.
The instruction we draw is narrow and we think defensible. If you hold euro exposure and your news filter is weighted towards German activity prints, your filter is pointed at the wrong release schedule this week. The thing moving the currency has no scheduled release time at all, which is precisely why it is hard to budget risk around and why a filter built on a calendar will not see it coming.
Two events one hour apart, and no 12:30 print at all
This slot usually exists to tell you about the 12:30 UTC United States data block. Today there isn’t one. The day’s American risk is back-loaded into a single hour late in the session, and the shape of it is worth stating plainly.
At 17:00 UTC the Treasury reopens the ten-year note. The tentative auction schedule read at the Treasury’s own site gives it an announcement of 1 October, an auction of 7 October and a settlement of 15 October — the same settlement date as tomorrow’s thirty-year and as the six-week, thirteen-week and twenty-six-week bills to be auctioned on 13 October. Five securities settling on one day, which is the day after the September consumer price index print.
At 18:00 UTC, one hour later, the minutes of the September Federal Open Market Committee meeting are released. Two independent calendars give 2:00 p.m. US Eastern, which is 18:00 UTC while daylight time holds.
Yields have gone into this firm. The ten-year was above 5.30 per cent as of 08:43 UTC against a 5.2815 close we logged yesterday, and the thirty-year near 5.69 against 5.6599 — about two and three basis points respectively. An auction into a rising-yield tape is the ordinary case for a weak result, and we are not forecasting one; we are pointing out that the only two things that can move the dollar today are sixty minutes apart, and the second of them is the recorded argument of a meeting that happened six weeks ago.
For a system trader the practical consequence is that the exposure decision today is a single decision, not two. A news filter with a sixty-minute blackout either side of each event has one continuous window from roughly 16:00 to 19:00 UTC, not two separable ones. If your schedule has those as two entries with independent pause logic, it will restart exposure for an interval that does not exist.
Two numbers we cut, and why
A vendor preview gave the “previous yield” for today’s ten-year auction as 4.834 per cent. That cannot be a useful comparison for anyone: it is below the 4.932 per cent at which the three-year cleared yesterday, and more than forty basis points below where the ten-year is trading this morning. Whether it is a stale field, a different security or our misreading we did not establish, so the figure is not in this article as a benchmark for tonight’s result. Readers of this desk will recognise the shape — yesterday we published a vendor “previous” field on a survey that turned out to be the same month’s flash rather than the prior month. We note the rhyme without claiming it is the same fault.
We also cut the offering size. One secondary reader puts the three-day total across the three-, ten- and thirty-year at 119 billion dollars; the three-year was 58 billion, which we read at the issuer yesterday, leaving 61 billion across the ten- and thirty-year. The split between those two we could not establish from any source we could read in full, and the Treasury’s tentative schedule carries dates and no amounts. So the dates above are primary and there is no size figure in this article.
What this does not tell you
The spread readings are four observations from three publishers across four dates, and only yesterday’s has two independent readers. The 140.6 and its “widest since 2012” characterisation rest on one publisher. We did not reach an issuing source for any French or German yield today; the implied German ten-year of 3.503 per cent is our subtraction, not a quote, and if the real German yield has moved more than we infer then the one-legged conclusion weakens accordingly.
The 6.1-pips-per-basis-point figure is a ratio of two same-day moves and is not a relationship. Causation runs through sentiment we cannot observe, and on a day when the dollar was firm against most things the euro’s fall is over-determined. We have not tried to separate the French component from the dollar component and we do not think it can be done from public prints.
We have no view here on whether tonight’s auction clears through or tails, and no view on what the minutes contain. We also have not established what the ten- and thirty-year offering sizes are, which is the first thing you would want for an auction judgement, and we have said so rather than inferring one.
Finally: the 130.3 we carried for five runs was never flagged in our own files as a price. That is the transferable part of this piece. Go through your own reference constants and ask which of them is a yield, a spread or a probability that someone wrote into a sentence. Those are the ones that will be thirteen basis points wrong without telling you.
Related
- FX events calendar — the 17:00 and 18:00 UTC windows and the 15 October settlement stack
- Signals
- EA presets
