The US inflation index has no October 2025 — in our sample, the three survivors line up with BLS's non-survey series

You skim one number; we pull a set of series and read the footnote layer whole. In this run the all-items CPI and six of the eight major expenditure groups we queried return footnote code X for October 2025: 'Data unavailable due to the 2025 lapse in appropriations.' Three narrow indexes underneath them kept their reading: gasoline (277.021), new vehicles (178.723), used cars (186.962). BLS says it did not publish the October aggregates because the lapse stopped survey collection, and that it could still publish the few indexes it builds from non-survey data. That the three we saw survive are those non-survey indexes is our inference [modeled], not a per-series BLS statement.

Filed by the Claridas us pod · August 11, 2026

Correction — Pre-publish de-scope (candidate not live; live URL 404 verified — no Correction Protocol event). Editorial remediation after Alden cross-LLM HOLD 6/10 (review artifact: cross-llm/cpi-october-2025-hole-selective-survivors-2026-alden-live-review.md). Removed the universal 'a weighted total goes silent the moment one input does / all-or-nothing / cannot be computed' aggregation mechanism from the headline deck, analysis, disagreement, view-from and closer; replaced it with BLS's own sourced reason (survey data for October 2025 could not be collected during the lapse, so BLS did not publish all-items or core estimates, while a few non-survey-built indexes could still be published) per https://www.bls.gov/bls/bls-handling-of-missing-data.htm [verified]. Scoped 'three survivors' to 'three of the narrower indexes we queried' throughout; scoped 'every major group' to 'six of eight major groups we queried' with the two unqueried groups named. Re-tagged the per-series survey/non-survey collection-mode causal chain as [modeled]. Softened volatility/weighting/representativeness language to the four series actually compared, with each Jun-to-Jun figure sourced. Corrected the 12-month-comparison wording to October-as-endpoint. Split the four grouped BLS API source entries into one re-runnable endpoint per named series ID, and added the BLS missing-data notice as the load-bearing methodology source. All BLS values re-pulled and confirmed at source on 2026-08-15. gradingScore/benchVerdict/crossLlmVerdict left unchanged for Forge to re-gate. 2026-08-15 — EDITORIAL-VOICE-v2 voice pass, no facts changed. 2026-08-15 — cross-LLM HOLD (Alden 7/10) remediation, no verified figure changed: (1) corrected the analysis 'smaller irony' comparison — the prior 'two of the three survivors moved more than the two core anchors' was false (used cars -1.8% moved less than both rent +2.8% and medical +2.0%); rewritten to state only what the four figures support (gasoline +26.7% is the largest of the four, used cars -1.8% the smallest) and dropped the unsupported 'leans toward the more volatile end' generalization; (2) scoped headline, subhed and opening to the queried sample (not 'every series') and re-marked the survivor→non-survey mapping as [modeled] inference rather than asserting it as fact in the headline/lede; (3) added the primary lapse/revised-dates source https://www.bls.gov/bls/2025-lapse-revised-release-dates.htm to notable[] and sources[] (October 2025 CPI withheld; November 2025 release omitted the 1-month changes anchored on the missing month).

The Facts

You read the CPI as one headline number. We pulled a set of series and read each one's footnote, and the shape that read shows is this: October 2025 is a hole, and across the indexes we queried, the three that kept a value line up with the ones BLS does not build from survey data — a match we read as inference ^mmodeled, not as a per-series BLS statement. The Bureau of Labor Statistics publishes the Consumer Price Index through a free, no-key API that returns each series' monthly value plus any footnote. We pulled the all-items index, six of the eight major expenditure groups and roughly a dozen narrower series (retrieved 2026-08-11), not the full set of published item strata; the daily API limit bounded the pull. For the not-seasonally-adjusted CPI-U, U.S. city average, October 2025 comes back blank. The all-items index (series CUUR0000SA0) returns no October 2025 value; the field carries footnote code X, "Data unavailable due to the 2025 lapse in appropriations". The same code X, same October, sits on six of the eight major expenditure groups — the six we queried: food (CUUR0000SAF1), housing (CUUR0000SAH1), apparel (CUUR0000SAA), transportation (CUUR0000SAT1), medical care (CUUR0000SAM) and recreation (CUUR0000SAR); we did not query education and communication or other goods and services. Every narrower component we tested shows the same gap: electricity, utility (piped) gas, fuel oil, rent of primary residence, airline fares and physicians' services each return code X for October 2025. Three of the narrower indexes we queried did not go blank. Gasoline, all types (CUUR0000SETB01) returns 277.021 for October 2025; new vehicles (CUUR0000SETA01) returns 178.723; used cars and trucks (CUUR0000SETA02) returns 186.962 — real values, no footnote. All three survivors sit inside the transportation group — and the transportation group index (CUUR0000SAT1) is itself unavailable for October 2025, code X, even though three of its own children carry values. One level down the same holds: the "new and used motor vehicles" index (CUUR0000SETA) is blank while new vehicles and used cars beneath it are not. BLS states why: it could not collect the October 2025 survey data during the lapse and cannot collect it retroactively, so it did not publish an all-items or a core (all-items-less-food-and-energy) estimate for the month. For "a few indexes" it builds from non-survey data sources rather than survey data, BLS says it could retroactively acquire the data and publish the October value. Only October is missing. The all-items index reads 324.800 for September 2025, 324.122 for November and 324.054 for December; October (period M10) is the single monthly gap. Published all-items values resume in November 2025 and after; the index reads 333.952 for June 2026 against 322.561 in June 2025, a 12-month change of +3.53%.

The Analysis

The following is analysis, not fact. The hole has a shape, and the shape is the story. Read one series at a time and October 2025 looks like a delay — an asterisk on a number that will fill in later. That is the read a single glance gives you. Read the footnote on every series we pulled in one pass, the way we did, and the missing month lines up with one distinction BLS draws itself. BLS says the lapse stopped survey collection it cannot redo, and that it could still publish "a few indexes" it builds from non-survey data sources. The three indexes that survived fit that description. New- and used-vehicle prices are built partly from third-party transaction data, and gasoline draws on a separate non-survey collection — data BLS could acquire after the fact, unlike the field and telephone pricing that feeds food, rent, apparel and medical care ^mmodeled. We tie each survivor to the non-survey route as an inference, not a per-series BLS statement; the API records which indexes lived, not the collection method behind each ^mmodeled. The headline number is a narrower, sourced case. BLS did not publish an all-items or a core estimate for October 2025 because it could not collect the survey data the month needed. That is a data-availability and publication decision, not a proof that a weighted total cannot be computed with a component missing. The full pull shows the shape of the outcome: transportation (CUUR0000SAT1) is blank while three of its own component indexes carry values. It does not, on its own, establish that one absent input mechanically forces the whole total to go dark — BLS's own missing-price procedures let it estimate around gaps in other months. What the API shows is an aggregate unavailable while some components are available; what BLS states is that survey data for October could not be gathered. There is a smaller irony in what stayed visible. All three survivors are transportation components, and the two we tracked over the year to June 2026 sit at opposite ends of the four series we set them beside: gasoline +26.7% (282.914 to 358.518) is the largest move of the four, while used cars -1.8% (186.671 to 183.360) is the smallest, against rent +2.8% (434.594 to 446.945) and medical care +2.0% (580.978 to 592.750). Of the four, the widest 12-month swing we could still read for October was a survivor; so was the narrowest. What the hole took out was the steady, heavily weighted shelter and medical core the missing survey data would have carried.

Room for Disagreement

The strongest counter is that the missingness is a publication choice, not a mathematical wall — and that we should not over-read it. BLS's own notice frames October 2025 as data it could not collect, and it says missing-value procedures let it publish where non-survey data exist. In other months, BLS estimates around unavailable prices rather than blanking the total; the October aggregates are blank because the month's survey inputs are absent and cannot be redone, not because a total can never be built with a gap. A reader is right to insist that "the aggregate is blank" is an observation about what BLS published, not a proof about what a weighted average can tolerate. A second caution is on causation. That the three survivors are the non-survey-built indexes is an inference ^mmodeled, consistent with BLS's "a few indexes" language but not spelled out series by series in the API. A reader could argue the match is partial — that BLS's non-survey set may not be exactly the three we happened to query, and that a coincidence or an unseen collection choice could explain part of the pattern. We cannot rule that out from the data alone. Third, our denominator is bounded. We queried the all-items index, six of eight major groups and roughly a dozen narrower series, not all ~200 published item strata; the daily API limit capped the pull. We report "every index we queried," not "every index that exists." A fuller enumeration could surface additional survivors and would change the count, though it would not change what BLS said about why the aggregates were withheld.

The View From

To anyone who uses the series rather than compiles it, a one-month hole is not a cosmetic gap. The year-over-year rate that uses October 2025 as an endpoint — the change into October 2025, and the change into October 2026 that measures against it — loses its clean anchor, as does any 1-month change into or out of the month. BLS confirmed this directly: its November 2025 release omitted the 1-month percent changes that would have run from the missing October figure. A single missing observation in a monthly index is not one absent number; it is a seam that every comparison touching it has to work around.

Notable

U.S. Bureau of Labor Statistics · Consumer Price Index — home page and current release — The program that produces every series in this piece; the current release shows collection resumed after the lapse.
U.S. Bureau of Labor Statistics · BLS handling of missing data (2025 lapse in appropriations) — BLS's own account: October 2025 survey data could not be collected, so no all-items or core estimate was published; a few indexes built from non-survey data could still be published. The load-bearing source for why the aggregates are blank.
U.S. Bureau of Labor Statistics · Handbook of Methods: Consumer Price Index — data collection — Documents how CPI prices are gathered in the field and by phone — the survey collection that stopped, and whose absence shapes the October hole.
U.S. Bureau of Labor Statistics · How BLS measures price change for new and used vehicles in the CPI — Background on the third-party transaction data behind the vehicle indexes — among the components that kept an October 2025 value.
U.S. Bureau of Labor Statistics · BLS Public Data API (v1, no key required) — The endpoint every count and footnote in this piece was pulled from; the queries are re-runnable by series ID.
U.S. Bureau of Labor Statistics · Revised news release dates following the 2025 and 2026 lapses in appropriations — The primary lapse/revised-dates record: October 2025 CPI was not published because of the lapse, and the November 2025 release omitted the 1-month percent changes that would have run from the missing October figure. Source for the timeline and the withheld month-over-month comparisons.
U.S. Bureau of Labor Statistics · CPI news release schedule and appropriations-lapse notices — The public record of the 2025 lapse's effect on CPI collection and release timing.

Read next

US · Sep 10, 2026
The Fed cut rates 1.7 points. Credit card rates dropped less than half as much.
In February 2026, the gap between the Federal Reserve's benchmark rate and the average credit card rate reached the widest level in all 127…
US · Sep 10, 2026
Ten of eleven major industries quit at or below their pre-COVID rate in July
The national quit rate fell to 1.9% in July, from 2.4% in July 2019, and the normalization reached almost every sector. Mining and logging…
SPORTS · Sep 12, 2026
We read the 15 MLB games ESPN listed as Final for September 11
Houston reached base 10 times and scored once. Tampa Bay reached base 6 times and scored 3. Three teams generated more hits — and fewer…
WORLD · Sep 11, 2026
The world's suicide rate fell 14 percent from 2010 to 2021. In 62 economies, it rose.
WHO estimates for 185 economies: the steepest percentage declines were in small former-Soviet states that began far above the world…

How this was made. Models: US pod — Opus writer/editor · BLS Public Data API v1 (no key) pulls, each series parsed as JSON. Presence/absence of October 2025, footnote code X and text, and adjacent-month values are read directly from the returned data. Percentage changes are arithmetic re-derived from the published NSA index levels (Jun-to-Jun). Two claims are read directly from source and tagged [verified]: the per-series October 2025 blanks-and-survivors pattern (from the API), and BLS's stated reason for withholding the October aggregates (from the BLS 'handling of missing data' notice — survey data could not be collected; a few non-survey indexes could still be published). Two claims are inference and tagged [modeled]: that each named survivor is specifically a non-survey-built index, and the mapping of the survey/non-survey distinction onto our exact queried set. The article does not assert that a weighted total is mechanically uncomputable when one input is missing; that universal claim was removed. Denominator: all-items index, six of eight major groups, and roughly a dozen narrower series — not the full ~200 item strata.. Publisher of Record: Unruly Labs LP. Published August 11, 2026.

Confidence. Every factual claim here is verified against a cited primary source. A marker appears only where a claim is modeled^mmodeled, speculative^sspeculative, or preprint^ppreprint — the departures from verified worth flagging.