Yesterday morning this desk published a finding and flagged it as its own largest exposure: the German two-year yield had risen five to six basis points to 3.28% and the ten-year five basis points to 3.52% on the Mecklenburg-Vorpommern result, while EUR/USD did nothing at all. One house’s morning note, one reader, and we said in print that it needed checking before anyone acted on it. It has now been checked. A second reader has the German ten-year closing Monday at 3.46%, six basis points lower on the day — and attributing that easing to the same election we said had pushed it up.
The second reader says down, not up
Trading Economics carries the German ten-year benchmark at 3.46% for 21 September, a daily change of −0.06, up 0.21 percentage points on the month and 0.71 on the year. Its accompanying text puts the Monday easing down to “a disastrous state election result for Germany’s governing CDU” together with softer oil.
Read that against what we published at 06:22 UTC and the problem is not a rounding difference. Two publishers have the same instrument, on the same day, moving in opposite directions, and both name the same cause. One of them says the election pushed Bund yields up five basis points. The other says the election pushed them down six. That is not a data dispute you can split the difference on.
One thing both agree on, and it is worth holding: this is a high-yield regime for German paper by any recent standard. The same source records 3.57% on Tuesday 15 September as the highest level since June 2009.
The uncomfortable coincidence, which we cannot resolve
The third reader is where this gets awkward for us. Investing.com’s German ten-year page, opened at 23:10 UTC on Monday night, is dated to the Friday 18 September close: 3.5217%, a gain of 4.34 basis points on that day, with a Friday range of 3.4782 to 3.5343. Its two-year page is dated the same way: 3.265%, a gain of 5.2 basis points on Friday, a Friday range of 3.218 to 3.273 and a fifty-two-week high of 3.312%.
Put those next to the numbers we published as Monday’s election reaction. We said the ten-year rose about five basis points to 3.52%. Friday’s close was 3.5217% on a move of about four and a half. We said the two-year rose five to six basis points to 3.28%. Friday’s close was 3.265% on a move of 5.2.
There are two readings and we cannot tell you which is right:
- The figures we published were genuine Monday-morning intraday levels, quoted a few hours into the European session, and the whole move reversed into the close. The Trading Economics number is then correct and our article was true at 06:22 and stale by 16:00.
- The figures we published were Friday’s close and Friday’s move, carried into a Monday note about a Sunday election. In that case we attributed a pre-election move to the election.
The honest position is that a morning research note quoting a level is not a daily change, and we treated one as the other. Whichever reading is right, the claim as we published it — that the news was in the Bund — is not supported by where the Bund closed.
Two bond pages frozen at Friday is the reason you read three
Both of the Investing.com pages were still showing Friday’s close hours after the Monday cash session ended in Europe. That is not an accusation; it is an observation with a direct operational consequence. If your only reader for a yield is a page that has not rolled, you will mark Monday with Friday’s number and never know it. The tell is cheap: the page prints the date it is closed to, and it said 18/09.
One further oddity, recorded once and explicitly not generalised from: that two-year page carried instrument metadata giving a 2.50% coupon and a maturity of 12 December 2025 — a bond that has already matured. The ten-year page’s metadata, 15 August 2033 at a price of 95.69, is unremarkable. We looked at two pages and one had stale reference data. That is an anecdote, not a pattern, and we are not going to pretend otherwise.
The American leg, from the issuer, where there is no ambiguity
The United States Treasury publishes its own daily par yield curve, and it settles this kind of question in one fetch. The closes it carries for the week just gone:
- 14 September — 2-year 4.65%, 10-year 4.97%, 30-year 5.34%
- 15 September — 4.67%, 5.00%, 5.36%
- 16 September — 4.74%, 5.01%, 5.35%
- 17 September — 4.67%, 4.94%, 5.29%
- 18 September — 4.76%, 5.01%, 5.34%
Two things follow. First, the ten-year dipped below 5% on 17 September and was back above it by Friday’s close — which closes, in the affirmative, a line this desk published on 18 September about the ten-year coming back under five and the yen not following it. It did not stay there.
Second, and this is the number worth carrying: on the same 18 September close, the US ten-year was 5.01% and the German ten-year was 3.5217%. That is a spread of about 149 basis points, and both sides of it are dated to the same session rather than assembled from two different days, which is the mistake this article is about. If Monday’s German close at 3.46% stands, the gap widened to roughly 155 basis points before a single American number printed.
The rate-path side is not quiet either. Fed official Collins, speaking at 20:44 UTC on Monday, is penciling in another hike this year and no change through next year, citing energy pressure from renewed Middle East hostilities as the reason she supported last week’s increase. That is one reader on one speaker and nothing here is priced off it.
What reprices today, and what it means for sizing
European rates have a heavy speaking calendar on Tuesday 22 September and almost no data. All times UTC:
- 08:30 — Bundesbank President Nagel
- 11:00 — ECB President Lagarde
- 14:00 — euro area consumer confidence, consensus −16 against a prior of −16
- 19:30 — Nagel again
Two Nagel appearances eleven hours apart, a Lagarde appearance between them, and a consumer-confidence print whose consensus is identical to its prior. The sizing question this poses is not directional. On a day when the only euro-area number on the board is forecast to be unchanged, the entire distribution of outcomes for Bund yields sits inside three unscripted speaking slots, and the one instrument that did not react to Sunday’s election was EUR/USD. A euro position sized off spot volatility on a day like this is sized off the wrong variable.
And Tokyo is shut for a second consecutive day, so anything the yen does against the euro today happens in offshore books with no Japanese cash market underneath it.
What this does not tell you
It does not tell you which of the two readings in section two is correct. We have a Monday close from one publisher, a Friday close from another, and a morning research note whose figures are consistent with both. Settling it needs an intraday German yield series with timestamps, which we do not have.
It does not tell you that Trading Economics is right and the note we quoted was wrong. It tells you they disagree on direction while agreeing on cause, and that we published one of them alone.
The Monday German close at 3.46% is itself a single reader, and we are flagging it exactly as we flagged the figure it contradicts. Do not replace one unconfirmed number with another and call the matter closed.
The 149-basis-point spread is arithmetic on two readers dated to the same close, not a quoted spread product. Nobody trades it at that number.
Nothing here says which way Bund yields go next, and the speaking calendar above is a list of times, not a forecast.
