Methods: when the settlement statistic is missing

Supporting material for Matt's article. The source checks and calculations below are version 3 (v3) of this research. Source observation date: October 9, 2026. This is a selected document comparison with one retained public-page illustration. It does not reconstruct actual trades or payments.

Question and evidence design

The question is whether a scheduled economic release supplies the exact statistic a contract requires and what the reviewed terms provide when it does not. The evidence has four distinct parts:

1. Historical agency publication records from October through December 2025.
2. An announced BLS limitation affecting October 2026 year-over-year CPI changes, observed before that release.
3. Four PM-US product templates and the September 30, 2026 rulebook, read as a dated document comparison.
4. Retained Kalshi CPI and unemployment terms, plus an October CPI page notice/outcome observed during the independent review.

The templates were selected from the earlier discovery research because they cover the same four economic families as the publication examples. They illustrate differences within one exchange. They are not a random sample, a complete venue census, or proof of historical applicability. The v3 analysis acquired no venue data. It reuses the independent review's bounded page reads and source files. An event identifier or matching symbol does not bind the exact governing template version. A current public index likewise establishes availability, not applicability to a position.

The article's invented numerical examples are a separate, explicitly hypothetical category. The observed 4.6-to-4.5 unemployment revision is paired with a hypothetical threshold; the Kalshi formula is a reproduction using published index values, not a verified execution record. Inputs_v3.json labels each evidence type and its assumptions. The calculations demonstrate mechanisms without assigning a probability to an exchange decision.

Publication units and coding

One historical row represents a statistic, reference period, and original release slot. Dates use the US Eastern publication calendar. The original scheduled date is compared with the actual date on which the relevant data or replacement report became available.

Row	Original date	Availability	Coding decision and source
September CPI	2025-10-15	2025-10-24	Delayed publication; CPI FAQ question 1
September payrolls	2025-10-03	2025-11-20	Delayed publication; revised BLS calendar and CPS FAQ question 1
October payrolls	2025-11-07	2025-12-16	Initial October establishment estimates in a later report; revised calendar and December 16 release's establishment-survey shutdown box
October unemployment	2025-11-07	Null	Household observations not collected retrospectively; CPS FAQ question 2
October headline CPI	2025-11-13	Null	Headline figure unavailable; not a claim that every component index is absent; CPI FAQ questions 2–3 and 16
November all-items CPI-U, seasonally adjusted one-month change	2025-12-10	Null	Statistic unavailable in December 18 release, Table A; original date from revised calendar
Q3 GDP advance slot	2025-10-30	2025-12-23	Replaced by initial estimate; BEA opening note
Q3 GDP second slot	2025-11-26	2025-12-23	Same replacement report as preceding row; BEA opening note

Primary references: BLS revised calendar, December employment release, CPS shutdown account, CPI shutdown FAQ, BEA replacement release.

Calendar subtraction yields 9, 48, 39, null, null, null, 54, and 27 days. The separate November CPI report event runs from December 10 to December 18, a delay of eight days. It is not assigned to the unavailable one-month statistic. That release reports a 0.2% seasonally adjusted September-to-November change and a 2.7% unadjusted 12-month change. Table A's all-items monthly entry is absent, while some component entries exist. December 18 CPI release. No unavailable value is converted to zero or an open-ended delay. No mean delay or frequency is reported. The two GDP rows preserve the original scheduling information while referring to one replacement event. These intervals contain weekends and holidays and are not trading-day counts.

The original evidence is retained in the discovery research records, which are not published. The v3 input file records the hash of those observations. V3 checks that the original seven rows remain identical and adds the November monthly row separately. Hash-bound review excerpts support the new row. The check does not automate the judgment used to code a publication.

Prospective CPI guidance and arithmetic

Question 16 of the CPI shutdown FAQ says that BLS will omit October 2026 12-month changes for indexes with unavailable October 2025 values. The page shows a September 16, 2026 modification date. The statement remains prospective in this article. The reference period must not be confused with the month in which a release is published.

The CPI release schedule lists October 2026 data for November 10, 2026, at 8:30 a.m. The November calendar confirms the entry and explicitly labels its times Eastern. Both were read October 9. This is the scheduled time, not proof that the release will occur then or contain a particular statistic.

The formula uses two values of the same series, with consistent geography, adjustment, and reference base:

percent change = (later index / earlier index - 1) × 100

For a 12-month change, the two values must refer to the same month one year apart. A monthly change uses adjacent months. The BLS calculation guide supplies the general method. The previous-year denominator is not the index's normalization base, such as 1982–84 = 100. Changing that normalization is a separate matter from a missing October observation.

The article assumes an invented current value of 309 and previous-month value of 308. Decimal arithmetic gives a one-month change of approximately 0.324675%, displayed as 0.325%. With an independently available previous-year value of 300, the annual change is 3%. With that denominator missing, the annual result remains null. Display rounding in this example is not a proposed contract precision rule.

BLS's carry-forward method used September survey prices for commodities/services and April observations for the rent panel. For seasonal adjustment, BLS approximated missing October values with the geometric mean of September and November. Neither procedure supplies a published all-items October index. Most nonsurvey data were recoverable, allowing some component publication.

The downstream effects have distinct statuses. The November all-items monthly gap is observed in Table A. FAQ question 7 states the omission of December two-month changes for even-month cities. Question 8 explains the unusual April 2026 rent-panel calculation: the sixth root of a 12-month price relative rather than a six-month relative, with the percentage change obtained from the resulting relative. The article attributes this explanation to BLS rather than claiming to reproduce its full housing calculation. Question 16's October 2026 limitation remains prospective as of October 9. These examples are not an exhaustive inventory of downstream effects.

The missing-data factsheet, modified December 17, 2025, describes a geometric-mean approximation subject to the parties' agreement. It labels such values and derived changes unofficial. The article uses this distinction to explain why a private calculation does not by itself satisfy a contract's source requirement. No interpolated official CPI series was created for this study.

Product documents and versions

Page numbers below are one-based PDF pages, counting the cover letter. They are not article page numbers. The rulebook's printed pagination is one lower than its PDF page number.

ID	Document	Version identity	Relevant pages
P01	NFPC	April 3 filename and cover; 183,498 bytes	2: underlying change, parameters, sources, expiration. 3: delay, previous-period/review choice, settlement, clarification.
P02	URC	April 3 filename and cover; 188,085 bytes	2: rate definition and parameters. 3: sources, expiration, delay, and fallback. 4: clarification.
P03	CPIC	April 6 filename; March 31 cover; 206,983 bytes	2: measure and source. 3: initial release, expiration, delay/cancellation. 4: no-data/equivalent-source branch and clarification.
P04	GDPC	April 6 filename; March 31 cover; 205,494 bytes	2: measure, estimate, and source. 3: expiration, specified release, delay/cancellation. 4: boundaries, review, and clarification.
P05	PM-US rulebook	Cover September 30, 2026; downloaded October 9; 827,274 bytes	PDF 14 / printed 13: rule 1.5. PDF 81 / printed 80: rules 10.3–10.4.

The four product PDFs were freshly downloaded during discovery and matched the October 4 retained originals byte for byte. This analysis reuses those bytes. The rulebook is a new narrow verification download. Full hashes are in the claim/source map and the retained packet manifest. Neither cover dates nor intended listing dates establish a particular contract's effective terms.

Actual unemployment vintage example

BLS first reported seasonally adjusted November 2025 U-3 as 4.6% on December 16. Its January 9, 2026 Table A reports 4.6% as first published and 4.5% after updated seasonal factors. The retained FRED excerpt, last updated October 2, 2026 and inspected October 9 Eastern, shows November at 4.5% and October unavailable. Initial report, revision table, FRED series.

The article's at-least-4.6% threshold is invented. The observed first value meets it and the revised value does not. This supports vintage sensitivity without asserting an actual contract outcome. The templates' specified-period delayed-release clause is separate from the unresolved original/revised vintage of an earlier-period fallback. The synthetic 4.2% example remains for the latter mechanism, so the article does not imply that November's observed revision answers the fallback-vintage question.

Clause interpretation

Payroll and unemployment. Unless their payout criteria provide otherwise, the default expiration date is the earlier of the relevant official release and three months after the scheduled release if none occurs. The default time is 11:59:59 p.m. Eastern, subject to the specified contract terms. The delay provision uses the first value for the requested period reported before expiration. If the period's data remain unavailable at expiration, the exchange can select the most recent earlier period with available data or initiate review.

NFPC's underlying is the change in total nonfarm payroll employment. Reading its fallback in that context points to an earlier period's reported change. It does not prescribe an employment level or a new difference across the data gap. This is a contextual reading of two clauses, not an observed exchange ruling. The fallback does not expressly resolve whether an earlier period's original or revised vintage controls. The article preserves that uncertainty.

Sources. NFPC p. 2 and URC pp. 2–3 name designated government agencies as primary sources and include Trading Economics and repositories/aggregators such as FRED as secondary sources. CPIC pp. 2–3 and GDPC pp. 2–3 list official statistical agencies. Their additional/substitute-source authority is distinct from a presently designated source. An agency listed for another country does not establish an equivalent US CPI statistic. Existing contract terms or a prior designation may matter; a new Rule 10.3 action is not asserted as the only possible route. A repository can reproduce a missing official observation as missing and a revised observation as revised.

CPI. The normal input is the first official non-preliminary publication of the selected statistic. The delay branch extends expiration to actual publication unless the contract terms specify otherwise. Cancellation or indefinite delay permits fair-price settlement or review. A separate no-data clause requires both absence of the relevant statistic before expiration, including extensions, and no equivalent statistic from an alternative designated source. The article does not decide when that branch activates for a particular upcoming contract.

GDP. The selected estimate is part of the input. Delayed release extends expiration to the actual release. Cancellation or indefinite delay permits fair-price settlement or review. Unlike CPIC, this template does not contain the same separately labeled no-data-by-expiration clause. The condensed comparison must not manufacture that provision or decide that BEA's replacement automatically qualifies as a named estimate.

Hierarchy and review. Rule 1.5 gives product specifications priority over conflicting general rules and specific contract terms priority over conflicting product specifications. Rule 10.3 allows specified modifications, including expiration changes, with notice. The defined term Modification on PDF p. 9 / printed p. 8 makes a qualifying Rule 10.3 change part of Contract Terms upon issuance. Not every exchange notice is established as such a modification. Rule 10.4 permits designated officers or their delegates to review reliability/transparency circumstances before settlement and ties settlement to completion and determination. It also mentions reversing an outcome for obvious error. The product documents contain finality language, so this analysis does not make an absolute claim that every declared outcome is irreversible. No fixed maximum review duration was established in the provisions examined.

All four templates include clarification or review authority. CPIC and GDPC also use the phrase “Market Outcome Review Process” in additional contingencies, while other clauses and rule 10.4 use “Contract Outcome Review Process.” The article uses the plain term “outcome review” and cites the actual section rather than inventing two separate procedures.

Worked-example assumptions

The unemployment example assumes a single-month binary contract, maximum gross payout of $1, exact reported precision of 4.2%, no eligible requested-period figure by expiration, no conflicting override, and an exchange decision to select the earlier-period branch. The earlier available figure is 4.2%.

Under at least 4.2%, Yes receives $1 and No $0. Under the separate above 4.2% condition, Yes receives $0 and No $1. These are alternative contract conditions, not two simultaneous holdings or a hedge. Review is a permitted alternative with no assumed payout. Neither the missing month's rate nor a trader's profit is estimated.

The separate fair-price example assumes that the applicable contract permits settlement at last fair market prices and that the price determined for Yes is $0.62. This wording does not establish a near-last-trade method. The rulebook's separate trade-review valuation concept is not imported into this settlement example. A complementary $1 payoff gives No $0.38. A holder of 100 Yes contracts receives $62. Purchase costs of $40 and $70 yield respectively +$22 and −$8 before fees. Fees, rebates, financing, partial fills, exits, and fee reversals are excluded. There is no market-price input, calibrated fair-value model, or claim that the exchange will choose this price. The example's purpose is to separate gross settlement from return of each holder's entry cost.

Bounded Kalshi comparison

The independent review retained two undated, two-page public templates: CPI and U3. Their page 2 payout criteria specify, respectively, a CPI ratio formula and the last available month's unemployment data when data are unavailable at expiration. Both page 2 Contingencies clauses preserve discretionary review and payout-determination authority. Their normal settlement-date language also contains a review exception. This study does not treat either venue as wholly mandatory or wholly discretionary.

The November 6, 2023 Kalshi filing is historical evidence of a shutdown-related extension, described as the earlier of underlying release or six months after the shutdown ends. It is not proof that every later event used an unchanged version. The October 2025 CPI page independently displays related extension language. V3 makes no exact expiration or settlement-date calculation for that event.

The article uses only the October CPI page case, KXCPI-25OCT. The review observed it on October 9 Eastern (October 10, 01:46 UTC). Its retained notice names September as the fallback period even if November data are available before expiration and specifies percentage conversion and rounding to the nearest tenth. The retained page excerpt displays 0.2. The screenshot independently shows the September/November and rounding language, but is a partial viewport, not a complete page archive.

The reproduced calculation uses published unadjusted all-items CPI-U levels, on the same 1982–84 base:

100 × [(324.800 / 315.301)^(1/12) - 1]
= 0.2476550778... percent
= 0.2 percent to one decimal

September 2025 source, September 2024 source, retained CPI terms URL, event page.

The formula implies a constant monthly rate from an annual ratio. It is not the actual October monthly observation or September's reported 0.3% seasonally adjusted monthly change. The PDF's normal underlying is seasonally adjusted; the reproduced index inputs are not seasonally adjusted. The event notice establishes the period, not the exact series or numerical values used by the exchange. Agreement of the formula with the displayed outcome does not prove those execution inputs.

The comparison lacks the notice's publication timestamp, the exact event-bound terms version, a reliable settlement date, and a complete execution record. The page's payout timeline conflicts with retained member-comment chronology. Neither the timeline nor the comments are accepted as proof of the actual payment date. No claim follows about what a holder knew at entry or which input determined that holder's payment.

Why the other three page reads do not become settlement cases

Retained observation	What remains unidentified	V3 treatment
October U3 page displays 4.4	September and December both reported 4.4; the page does not identify the selected month	No claim that a holder was paid on September data
November CPI page displays 0.2	The reproduced formula yields 0.225115%, rounding to 0.2, while BLS's published two-month change is also 0.2	Diagnostic arithmetic retained here; no route inferred
October payroll threshold results	Several numerical inputs can produce the displayed threshold outcomes	No claim that the page proves use of −105,000 or its settlement date

The November diagnostic uses 324.122 and 315.493 from the November 2025 and November 2024 releases. It illustrates non-identification, not another verified formula settlement. All four source pages were acquired by the independent review. V3 adds no venue reads.

Reproduction and checks

Save inputs_v3.json and recalculate.py from the sources page in one folder, then run:

python3 recalculate.py

The script uses Python's standard library, with date subtraction and 48-digit Decimal arithmetic, including the fractional power in the CPI fallback formula. It reads only the published inputs and writes results in the form of arithmetic_v3.json. With a saved copy of that file in the folder, it reports whether the two match byte for byte. Review the inputs as well as the outputs. A successful run verifies calculations, not the source coding or legal interpretation.

The retained v3 script, calculate_v3.py, runs the same calculation lines. Before it calculates, it checks the input rows and source excerpts against hashes of retained research files that are not published. It also writes TABLES_v3.md.

The v3 source audit also checks links, the author split, exact PDF locators, source hashes, hypothetical labels, null treatment, and agreement of the formula calculations with a separate floating-point calculation. The rulebook clause page was rendered and inspected. Source excerpts from the web tool can be partial and are not represented as full archived HTML. The retained PDFs and BEA HTML are complete downloaded files.

What could change the conclusion

A superseding source notice could change the prospective CPI passage. An applicable contract override, exchange clarification, or ruling could change a scenario's treatment. Those developments would require revising the relevant application, not erasing the historical distinction between a release and its statistics.

A claim that these rules caused a loss, delayed an actual payment, or created a trading opportunity would need separate position and execution evidence. The retained page notice/outcome does not fill that gap. The supported conclusion is that release histories and written fallback mechanisms differ, with one observed notice and displayed outcome consistent with a reproduced formula. The exact execution route remains unverified.