October's Inflation Rate Is Not a Number. It Is a 26 Basis Point Band, and the Real Yield Did All the Work.
A government inflation statistic looks like a number and is really a ratio. The basket is priced every month and the level is published as an index; the 12-month rate is that index divided by its own value twelve months earlier, minus one. The denominator is not decoration. Without it there is no rate, only a level, and a level tells a bond market almost nothing, because a bond pays in dollars that lose value at a rate, not at a level. October 2025 is missing from the consumer price file. Not revised, not restated: absent, the row does not exist. So the October 2026 print, published and complete at the index level, cannot be expressed as a 12-month rate. That is the data blackout arriving late, and it arrives at the one moment when the market is arguing about whether the last leg up in yields is inflation or the price of duration. Here is what the gap does to the arithmetic, why the band is 26 basis points wide rather than a point, and why the entire 29 basis point move in the 10yr over four sessions landed on the real yield while the breakeven sat still.
The rate needs a denominator, and the denominator is the missing row
Consumer prices rose 0.35% in October from September, on the seasonally adjusted index, and 0.65% the month before. Core rose 0.30%. Those are month-on-month changes and they need only two adjacent levels, which is why they survive the gap intact.
The 12-month rate needs a level from a year ago. The October 2026 index is 337.476. October 2025 is not in the file. September 2025 is 324.245 and November 2025 is 325.063, so the honest statements are that prices are up 4.08% since September last year and 3.82% since November last year, and that the 12-month figure is somewhere between.
The band is 26.0 basis points wide. On core it is 19.0: 2.68% against November 2025 to 2.87% against September 2025. That is not a rounding error and it is not a modelling choice. It is the width of the information that the missing month was carrying.
The bound holds on one assumption, stated plainly: that the October 2025 index sits between its September and November neighbours. For a monthly price index in a low-variance year that is close to certain, and it is still an assumption rather than an observation. If October 2025 had printed below September, the true rate would sit above the top of the band; above November, below the bottom. The band is a fence built around a hole, not a measurement of what fell into it.
The gap is not a revision. Payrolls kept the month; prices and unemployment did not
A revision rewrites a number. This is a different failure: three of the four monthly series used here skip October 2025 entirely, and one does not. Nonfarm payrolls carries an October 2025 observation at 158,408 thousand. The consumer price index, the core index and the unemployment rate all jump from September 2025 to November 2025 with no row between them.
That asymmetry is the useful part, because it rules out the explanations that would otherwise be available. A rebenchmarking would move every series. A change of seasonal factors would restate the levels rather than delete a month. A file that keeps the payroll month and drops the price month is recording a collection gap in the household and price programmes specifically, and the practical consequence is that the hole is permanent in the series as published: November 2026 will get its 12-month rate back on 10 December, because November 2025 exists, and October 2026 will never have one.
This is why the gap belongs in a plumbing piece rather than a statistics piece. The market does not price the index level. It prices a rate, and every instrument that references inflation, from TIPS to the breakeven to the Fed's own preferred core measure, is denominated in that rate. One missing month removes the anchor for one month of pricing, and the anchor is not restored retroactively.
The breakeven is the instrument that needs the print. It did not move
The 10yr breakeven is the difference between the nominal 10yr yield and the real yield on the 10yr TIPS: the inflation compensation a buyer demands for holding government paper instead of an inflation-protected version of it. It is the market's own 10-year inflation forecast, and it is the one instrument on the curve that is explicitly a statement about a rate rather than a level.
It printed 2.20% on 16 November, the 215th lowest of the 218 sessions in this file for 2026, against a year low of 2.18% on 2026-06-24. On 6 November it was 2.23%. Four sessions later it was 2.24%. The whole move was 1 basis points.
The nominal 10yr went from 4.86% to 5.15% over the same four sessions, 29 basis points, and the real yield went from 2.63% to 2.91%, 28 of them. That is 96.6% of the move in the leg that needs no statistical print. The 10yr real yield is a traded price observed every session from TIPS. The breakeven is a subtraction between two traded prices, and its inflation content is only as good as the rate it references.
There is a mechanical reading here and a structural one, and the difference matters for what to do about it. Mechanically, when the spot inflation anchor is an interval instead of a point, a repricing that wants to be about inflation has nothing firm to grip, so it expresses itself through the leg that is continuously observed. Structurally, the real yield is where duration and fiscal risk have been pricing all year, so a move that lands entirely there is consistent with a market that never thought this was an inflation story in the first place.
The real leg is carrying more of the nominal yield than at any point this year
The real yield was 56.5% of the nominal 10yr on 13 November, the highest share of the 218 sessions in the file, from 54.1% on 6 November and a low of 43.2% on 2026-03-09. At the start of the year the share was 46.3%.
On the day's levels that is a 10yr nominal of 5.12% against a real yield of 2.91% and a breakeven of 2.20%, each dated separately: the nominal and the breakeven are 16 November closes, the real yield is 2026-11-13, the last session in the file. Two thirds of the coupon the government pays is now compensation for the time value of money rather than for expected inflation, and that was not true in March.
The real yield at 2.91% is the 5th highest of 218 sessions in 2026, 4 basis points below the year's high of 2.95% on 2026-10-05 and 119 above the low of 1.72% on 2026-02-27. This is not an outlier print at the top of the year's range. It is a market sitting near the top of the range with the inflation leg switched off.
What the band does to the next three prints
The calendar is what makes this urgent rather than academic. PCE for October lands on 25 November, nonfarm payrolls for November on 4 December, and CPI for November on 10 December. The CPI print on 10 December will have a real base, because November 2025 is in the file, so the 12-month rate comes back that morning at a point rather than an interval.
That creates a clean test. If the November rate resolves inside 3.82% to 4.08%, the band was the right way to read the gap and the market's decision to price around it through the real yield was correct. If it resolves outside, then the true October rate was outside the band too, the fence was in the wrong place, and the whole decomposition in this piece needs redoing rather than defending.
The instrument to watch is not the index. It is the gap between the two legs. A breakeven that starts rising while the real yield holds is a market re-anchoring on a rate it can finally see. A real yield that keeps rising while the breakeven stays at the year's low is a market that has stopped pricing inflation altogether and is pricing the balance sheet, and that is a different trade with a different hedge.
What would change the read
A breakeven above 2.36% before 10 December, with the real yield flat, would kill the central claim. That level is the 27 October breakeven, from before the last leg of the selloff, and reclaiming it would say the inflation leg is doing the work and the real-yield decomposition is measuring the wrong thing.
Or a real yield back below 2.63% - the 6 November level, before the four-session move - with the breakeven still under 2.24%. That would make the 28 basis point rise a duration event that reversed, not a repricing that stuck, and the 56.5% share would be a spike rather than a level.
Or a November CPI on 10 December that resolves above 4.08%. That would put the true October rate above the top of the band, which would mean the missing month was carrying more inflation than its neighbours imply, and the 26 basis point fence was too tight rather than merely unobserved.
Notes
1. Index levels: CPIAUCSL and CPILFESL, seasonally adjusted, US Bureau of Labor Statistics via FRED, observations to 2026-10-01. October 2025 is absent from both series and from UNRATE; PAYEMS carries it at 158,408 thousand.
2. The 12-month band: 337.476 divided by 324.245 gives 4.08%, and by 325.063 gives 3.82%. The geometric midpoint of the two base levels gives 3.95%. The band is a bound under the assumption stated in the text, not a confidence interval.
3. Yields: par curve from data/treasury_par.csv, US Treasury daily par yield curve, close of 2026-11-16. Real yield from DFII10 and breakeven from T10YIE, constant maturity, FRED; the real series last observed 2026-11-13 and the breakeven 2026-11-16. A single reported close, not an index print.
4. Real share: 10yr TIPS real yield divided by the 10yr par yield in the same session, 218 sessions from 2026-01-02 to 2026-11-13. Where the two series do not share a date the session is dropped rather than carried forward.
5. Calendar dates from tools/econ_calendar.py: PCE 25 November, Employment Situation 4 December, CPI 10 December, FOMC 9 December.
Investment Grade Now Costs More Than the Government's Own Duration Risk. It Took 3 Weeks. — the earlier test of whether the compensation is for default or for duration
Reserves Built $30.5bn on a $24.1bn Cheque — the same day's funding accounting, and the snapshot where this series went null