Every correction this desk has issued, dated, permanent. Newest first.
Washington's cap-and-invest chapter now carries a section titled Linkage, effective 24 October 2026. WAC 173-446 lists 86 sections against 78 in every capture the desk holds from 11 June to 11 October. All eight added sections were filed under one order, WSR 26-19-139 (Order 24-02), filed 23 September 2026 and effective 24 October 2026. Section 173-446-090 names no linked jurisdiction. It provides that Ecology will begin accepting compliance instruments issued by any linked jurisdiction on an effective date Ecology determines, with public notice at least 90 days in advance, and that the public notice is what lists the names of the linked jurisdictions. The same section allows Ecology to suspend or revoke a linkage if a linked jurisdiction repeals or suspends its programme, and to remove and retire allowances from current and future budgets to hold stringency. The other seven sections cover exchange clearing house and federal power marketing administration registration, allowance distribution to waste-to-energy facilities, a voluntary renewable electricity reserve account, transfer mechanics involving clearing houses, utility-to-FPMA transfers of no-cost allowances, and voluntary retirement transfers. As of 16 October 2026 the desk has found no public notice of a linkage effective date.
the 8 October morning edition stated that the APCR Auction #8 summary report 'published on schedule as Ecology publication 26-14-073, registered by the desk the same day'. The registration claim was false: this desk's primary register held publications 26-14-070 and 26-14-072 only, and did not hold 26-14-073 until 10 October 2026. The RESULT that edition reported was correct - the auction did go entirely unsold - and its SOURCE was correctly and openly stated in the same paragraph as Carbon Pulse citing Ecology, with the edition saying in its own words that 'the desk holds that retrieval and not Ecology's own'. What was wrong was the desk's statement about its own holdings. A reader can check a price against a market; a reader cannot check whether this desk holds a document, so a false claim about provenance damages the one thing they must take on trust. The desk now holds it: 26-14-073, fetched 10 October 2026, three pages, registered with its figures quoted verbatim.
the 9 October evening edition said the Washington APCR Auction #8 summary report 'remains owed to readers here until it prints'. It had printed on 7 October 2026 as Ecology publication 26-14-073. The accompanying statement that the report was 'not on our tape' was TRUE when that edition published and is NOT withdrawn - this desk did not hold the document until 10 October. Only the claim that it had not yet been published is corrected. The desk now holds the report and carries its figures as its own: 172,348 allowances offered at auction, all at the tier 1 price; none sold; total qualified bids divided by total allowances available for sale 0.00. Ecology will release the Auction Public Proceeds Report on 28 October 2026, which this desk's forward calendar did not carry and now will.
The 7 October morning and 7 October evening editions both printed the California LCFS weekly credit transfer price as moving -1.9% against USD 83.77 for the week of 2026-09-14. Both figures were wrong in the same way, and the sign was inverted: the true move was +2.9%, from USD 79.90 for the week of 2026-09-21 to USD 82.18 for the week of 2026-09-28, on volume up 46.3% (1,059,374 MT to 1,549,760 MT).
WHAT CAUSED IT. The desk's store holds the current weekly price and a short history of prior weeks in two separate fields. The current price is refreshed by the weekly collector; the prior-week history was populated by hand on 23 September and has never had a writer, so it froze there. For two weeks that did no harm, because the frozen history's newest entry was still the week immediately before the current one. It stopped being harmless when the current price advanced to the week of 2026-09-28: the prior-week pointer then skipped the week of 2026-09-21 entirely and compared two weeks apart, which inverted the sign.
WHAT HAS BEEN DONE. The history is refreshed from CARB's own Weekly LCFS Credit Transfer Activity workbook, read directly (544 weeks), and the week of 2026-09-21 is confirmed there at USD 79.90, range 56.00 to 86.00, volume 1,059,374 MT. The prior-week pointer now resolves to it. The missing writer is a separate open defect and is recorded as one.
NOT CORRECTED, because it was right: the 6 October evening edition printed -4.6% against USD 83.77 for the week of 2026-09-14. At that date those two weeks were consecutive and the comparison was correct.
neither edition reached readers - the 9 October morning was held at this desk's source-freshness gate, which found the BC LCFS monthly print it carried (CAD 172.00, August 2026) superseded by the September figure of CAD 213.96, a 24.4 percent move the publisher had issued the previous afternoon, and the 8 October evening carried that same superseded print and was retired rather than published a day late
The 2 October morning edition was headlined "Washington's APCR Auction #8 passes without a summary report on our tape, two days after sale", and framed the absent result as a gap.
There is no result yet, and none is due. Ecology's own schedule - 2026 Reserve Auction Potential Dates, publication 2514101 - sets the September 2026 APCR auction (the pre-compliance deadline auction) on Wednesday, 30 September 2026, and its final column, "Auction Certified, Summary Released", gives Wednesday, 7 October 2026. The summary report is five days away, on Ecology's published timetable.
Corrected: the Pre-Compliance APCR Summary Report for Auction #8 is scheduled for 7 October 2026 per Ecology's auction schedule. The desk should have said the report is due 7 October, not that it was missing from our tape.
Not withdrawn. The auction date, the marks and the Washington budget-table discussion were correct and stand.
A result not yet held before its scheduled release date is a result that is DUE, not one that is missing. The desk has added that rule to the note prompt and a lint check behind it, so an absence claim cannot be written about an event whose release date is still in the future.
The 2 October morning note attributed Washington 2027 demand of 56.7 Mt to Ecology's adopted budget table. That figure is a desk estimate; the table sets supply only (47.3 Mt in 2027).
The adopted rule language for Chapter 173-446 WAC (adopted 23 September 2026) states the annual allowance budgets. It states no demand figure. The desk's demand rows are its own projection from covered emissions, and the sentence placed that projection under Ecology's name by citing the rule beside it.
Corrected: Ecology's adopted budget table sets 2027 supply at 47.3 Mt; the desk estimates 2027 demand at 56.7 Mt. The supply figure, the auction dates and the marks were correct and stand.
What changed. The 23 September evening edition is corrected on seven points. Each figure printed was correct for the date it named; it was not the latest release, or a mechanism was stated backwards.
1. EU ETS, the Market Stability Reserve. The edition read: "with the MSR contingent supply that has not released." The Commission's publication of the 2025 total number of allowances in circulation (C(2026) 3496 final, Brussels, 28 May 2026) states 1,023,494,202, inside the 833 million to 1,096 million band, so 190,494,202 allowances go INTO the reserve from 1 September 2026 to 31 August 2027. Release happens only below 400 million. Corrected: "with the MSR taking in 190,494,202 allowances from 1 September 2026 on the 2025 TNAC of 1,023,494,202; release happens only below 400 million."
2. UK ETS. The edition read: "Last print GBP 58.68, 31 July 2026 ... 54 days old." The August 2026 row, GBP 59.34, is on the gov.uk CCM table. Corrected: "Last print GBP 59.34, August 2026 official monthly average UKA settlement (Dec futures), CCM table."
3. Australia. The edition read: "Last print AUD 37.50, 15 May 2026, CER QCMR March quarter 2026, 131 days old." The Clean Energy Regulator's June quarter 2026 report, published 9 September 2026, states: "The ACCU spot price rose after Q2 reaching $39.03 on 14 August 2026." Corrected: "Last print AUD 39.03, 14 August 2026, CER QCMR June quarter 2026 (published 9 September)." The 37.50 was the post-quarter price of 15 May quoted in the March-quarter report.
4. BC LCFS. The edition read: "Last print CAD 135.80, July 2026 monthly average credit transfer price. No fresher mark." The September 2026 edition of BC's credit market data workbook carries August 2026: CAD 172.00 across 10 transfers and 62,428 credits. Corrected: "Last print CAD 172.00, August 2026 monthly average credit transfer price (10 transfers, 62,428 credits)."
5. RINs. The edition read: "RIN prints (D3 2.5820, D4 2.0639, D5 2.3402, D6 2.2615, 20 July 2026, EPA EMTS) are 65 days old ... [confirm: nesting inversion ...]." EPA's RIN Trades and Price Information dashboard, reloaded 17 September, runs to the week of 24 August 2026: D3 2.48, D4 1.9769, D5 1.8376, D6 1.7677, with no nesting inversion; the confirm flag is withdrawn. In EPA's restated series the week of 20 July's D4 is 2.0641. Corrected: "RIN prints, week of 24 August 2026, EPA EMTS via EDAP: D3 2.48, D4 1.9769, D5 1.8376, D6 1.7677; nesting consistent."
6. Energy legs. The edition read: "Energy legs, EIA, 15 September: Brent USD 130.80/bbl, WTI USD 107.02/bbl, ULSD Gulf Coast USD 5.206/gal, Henry Hub USD 2.97/MMBtu." EIA's release of 23 September runs to 22 September: Brent spot 114.89, WTI 96.41, ULSD Gulf Coast 4.992, Henry Hub 2.90 (next release 30 September). These are EIA spot prices, a different instrument from ICE Brent futures. Corrected: "Energy legs, EIA spot, 22 September: Brent USD 114.89/bbl, WTI USD 96.41/bbl, ULSD Gulf Coast USD 4.992/gal, Henry Hub USD 2.90/MMBtu."
7. CA LCFS posting date. The edition read: "report posted today." CARB posted the workbook for the week of 14 September on Tuesday 22 September 2026, on its stated cadence. Corrected: "posted Tuesday 22 September."
Not withdrawn. The auction clearings, the ICE futures marks, the WCA day-over-day line and the RGGI largest-move line were correct and stand.
The desk's data collection missed newer releases for five series; the collectors and the fact check now check for the latest release.
What changed. Four evening editions - 11 September evening, 14 September evening, 15 September evening, 16 September evening - attributed the Washington Auction #15 settlement price of USD 39.50 to Ecology publication 26-14-062. That publication is the Auction #15 Notice: it states the 2026 annual auction floor price (USD 27.92) and no settlement price. The correct source is Ecology publication 26-14-063, the Auction #15 September 2026 Summary Report, issued 9 September 2026, whose Table 1 states a settlement price of USD 39.50 against the auction floor price of USD 27.92.
The price was right in every edition, and the auction date with it. What was wrong was the document named beside them: a scheduling notice where a results report was owed. Nothing derived from the figure changes, and a reader who acted on the level has nothing to re-check. A reader who followed the citation to verify it would have found a document that does not contain the number - which is the reason this correction exists.
The desk has since fetched and registered 26-14-063.
the 2 September evening note flagged that a secondary outlet carried the same figure as our EEX auction print, labelled as an ICE Dec-26 weekly benchmark, and queued the tape's auction tag for reconciliation. Reconciled 13 September against EEX's own 2026 auction report: the 2 September row (Auction 4, Period CAP3 PL, successful) states the clearing at exactly the tape's figure - register EEX-EUA-Auction-Report-2026-row-20260902, which holds the row verbatim. The tag was CORRECT. The identical secondary figure was that outlet's own labelling or a two-decimal coincidence - a collision class the desk has measured at roughly one in twelve in this price band - and coincidence with an exchange value is expected furniture around a public auction result, not evidence of a mislabel. Nothing to retract; the queued question is closed.
What we published. The 28 August morning note said a supplemental proposal to permanently end EPA regulation of power-plant carbon 'entered White House review on Monday', and the 4 September morning note said it 'entered OMB review on 24 August'. Both statements presented the date as desk fact.
What the regulator's own docket shows. OIRA's public dashboard of EO 12866 reviews (read and retained 7 September) lists two pending EPA instruments in this area: the Carbon Pollution Standards Repeal (RIN 2060-AW55, a final rule, under review since 14 May 2026 - that half of our coverage verifies) and a second instrument received 18 August 2026: RIN 2060-AX00, 'Repeal of Greenhouse Gas Standards for Fossil-fired Electric Generating Units'. No pending review carries an endangerment title, and nothing on the docket arrived on 24 August.
Where our date came from. A press account characterised the second instrument as a supplemental endangerment proposal arriving 'on Monday'; the specific date of 24 August was an inference layered on that account, not a fact from any docket. The desk published the inference as the fact.
What we carry now. The final repeal (AW55) pending since 14 May 2026, past the standard 90-day review window - a default, not a deadline; and the second instrument (AX00) received 18 August. Where the endangerment question is discussed, the characterisation is attributed to the press account that made it, never to the docket. The dated claims in this correction are checked against the retained docket text, and the watchlist entry that carried the inferred date now carries its verification date and source.
What we published. The 4 September evening note said the linear reduction factor 'steps to 4,4% in 2028 while our rows carry 4,3% through'.
What the rows actually carry. The desk's own supply rows step: the annual decrement is 86.0 Mt flat across 2028-2030 - the 4.4-percent era - and the 2026-to-2027 edge is a larger composite move, not a pure linear-reduction step. The sentence describing our rows as flat 4.3 percent was false about our own arithmetic; the error's direction made our published supply path look looser after 2028 than the rows we compute from.
The formatting half. '4,4%' and '4,3%' reproduce the source Directive's decimal-comma typography, held verbatim in the desk's register; prose renders in the desk's locale (4.4%, 4.3%), and the register keeps the source's exact bytes.
What stays the same. The EU base remains blocked on the missing Union-wide quantity decision, which the rows disclose; this corrects the description of the rows, not the rows.
And the fix that follows from it. Decomposing the edge exposed a scope discontinuity in our own construction: the 2026 model point included the 24.9 Mt aviation component while the 2027-on rows never did. The rows now measure stationary+maritime throughout - the base the linear reduction factor applies to - so the 2026-to-2027 step reads -84.0, the clean 4.3-percent-era decrement, and the aviation component is held beside the rows rather than inside one year of them. Whether the demand side's scope matches is under review and will be stated when settled.
the GBP 58.68 July official monthly average printed since 2 September was reconciled against gov.uk's CCM table on 7 September and CONFIRMED - the table's Aug-26 row lists July at 58.68 (register GOVUK-UKETS-Markets-CCM-2026-09-07). Nothing to retract; the reconciliation had been owed since 2 September and is now on the record.
What we published. The 4 September morning note carried 'No citable REC price marks on our tape' in the same section as 'three of twenty tracked rows carry a sourced traded price.'
What the store says. Three rows carry sourced prices. One - the DC solar carve-out price - is the DC Public Service Commission's own reported average, a regulator print and citable: the desk holds one citable REC price mark, and the first sentence was wrong. The other two carry a commercial vendor's assessment whose licence terms the desk has not yet determined; those prices are withheld, exactly as the desk's other vendor-assessed figures are, until a determination lands - and the second sentence should have carried the word withheld.
What changed at the desk. Each priced row now carries its licence disposition; the per-class figure that derived from the vendor assessment is withheld on the methodology page with its reason; and a class whose only price is withheld does not support a firmness call.
What we correct, and where. Five editions - the 1 September morning and evening notes, the 2 September morning and evening notes, and the 3 September evening note - rested the desk's new RGGI deficit story on a private allowance bank of roughly 67 million short tons, and two of them (2 September evening, 3 September evening) called that bank 'load-bearing'. Every one of the five printed 'July 2025' beside the figure, and four of the five named its source alongside. The date was disclosed each time; what was never made was the judgment. This desk labels a price past its cadence a last print, not a mark - and applied no such standard to a stock, so a fourteen-month-old balance-sheet figure rode as current evidence in five editions with its age in plain sight. Disclosure without judgment was the failure: printing a date beside a number does not make the number current.
Why nothing caught it. The current figure existed before any of the five editions ran. Potomac Economics published its special Report on the Supply and Demand for RGGI CO2 Allowances on 21 August 2026; the desk first read it on 4 September, fourteen days later, because the market monitor's publication page was not watched at all - the report is a special publication, the first of its kind, so no scheduled cadence existed to miss, and the publisher our RGGI evidence leans on was simply not on the watch list. The failure also sat upstream of the staleness: the desk never held the document the 67 million figure came from - Potomac's 2025 annual report was cited, not fetched - where the current figures come from a report the desk has fetched and keeps, with every figure in this correction checked against that document's own text.
The current figures. The 21 August report states that the surplus of allowances in circulation fell from 71 million short tons at the end of the fifth control period to 55 million at the end of the second quarter of 2026, and projects 45 million at the sixth-control-period compliance deadline (1 March 2027). Of that surplus, 18 million - 32 percent - is held for compliance purposes; holdings for investment fell from 60 million to 37 million over the year.
The Cost Containment Reserve, corrected with it - same market, same window. The 2 September morning note described Virginia's Auction 73 tranche as '1.148 million Cost Containment Reserve allowances, which are released only if the CCR trigger price is reached'. The mechanism is stated correctly; the framing presented as an open question a trigger that had already fired. At Auction 71 on 11 March 2026, bids above the 2026 CCR trigger price of $18.22 exceeded the initial offering, over 7.8 million CCR allowances were sold, the 2026 reserve was fully depleted, and the auction cleared at $24.99 (Potomac's Auction 71 market monitor report). No edition of this note had said so, and the 21 August report treats the remaining CCR allowances as sold at Auction 73 given recent price levels. The reason nothing contradicted the conditional framing is structural, and it is the same failure as the bank: the desk held the CCR's trigger condition but no record of whether it had fired - trigger state existed for no mechanism the desk tracks - so a sentence implying the trigger had not yet been reached had nothing to collide with.
What it changes. The cushion under the deficit rows is smaller than the figure we printed and it is falling: 55 million at mid-2026, projected 45 million by the 1 March 2027 compliance deadline - and the reserve that would have added supply above the trigger this year is already spent. That strengthens, not weakens, the tightening read the rows carry; this is a correction of the evidence we showed, not a reversal of the view.
What changed at the desk. The market monitor's publication page is now watched for any new report. Stock figures carry their dates and cadences as data, and a stock past its cadence is labelled exactly as a stale mark is. Trigger state is now held per mechanism where a primary states it, with its source and as-of date, and recorded as absent where none does - so whether a mechanism has fired is a fact the desk stores rather than a tone a sentence implies.
What changed. Washington auction and reserve-sale dates were cited to Ecology publication 25-14-100 - the 2026 Cap-and-Invest Summary of Expected Dates, an annual schedule published in 2025, which is why its number begins 25. From the 4 September edition the citations are the governing auction notices themselves: Ecology publication 26-14-062 (Auction #15), 26-14-068 (ECR Auction #1) and 26-14-072 (APCR Auction #8). A schedule states a plan; the auction notice governs the event - the desk cites the governing document and keeps the summary as corroboration.
Why it is logged. The citation moved between editions without a log entry, and the same publication was described two different ways in three days. One description is now settled - 25-14-100 is the 2026 Summary of Expected Dates (published 2025) - and citation changes are logged like any other change a reader might notice.
The desk's mark for Washington Cap-and-Invest (WCA) changes basis with this note.
What changed. Until this note the WCA mark was the most recent Ecology auction settlement - Auction #14, held 3 June 2026 (Ecology publication 26-14-039, the Auction #14 Summary Report: "Ecology held an auction of allowances on June 3, 2026", which also states the USD 64.56 current-vintage settlement the desk carries). From this note it is the most liquid futures settlement - the contract carrying the most open interest, which on the 2026-09-01 activity file is December 2026 at 85% of the WCA curve - stated as such on every appearance with the contract month and its open-interest rank named, and the auction settlement stays on every surface as the primary print beside it.
Why. Washington auctions clear quarterly, so an auction print describes the market in the week it prints and drifts from it for the three months afterwards - the mark this note replaces was a quarter old. This is the same shelf-life argument that moved RGGI off its auction print on 27 August, applied to the same auction cadence. The most liquid contract reprices daily and is where the market actually trades - the front month is not, because open interest rolls out of a contract as it nears expiry, which is why the desk stopped marking off it; the auction remains the primary print, and when an auction is the news the auction leads.
What it means for reading the note. The mark will move daily and by small amounts; a change in the mark is no longer evidence that an auction has happened. Each mark states what it is and which contract month it is.
What we did, in order, because a figure that disappears and returns without an account is worse than one that never moved.
1. The suspicion. Our RIN prints for the week of 20 July showed D4 at 2.0639 against D6 at 2.2615 - the biomass-based diesel RIN trading two tenths of a dollar BELOW the conventional one. A D4 RIN can be retired against a D6 obligation, so a lasting D4 discount is an arbitrage. Set against the only other RIN observation on our tape - a different publisher, six weeks earlier, showing D4 four cents ABOVE D6 - it looked like the collector had swapped the labels.
2. The withholding. We took all four prints off the tape rather than publish figures we could not check, and rather than publish them re-sorted into the order we expected, which would have been a market invented from a hypothesis.
3. What we could not do, and why that was the real problem. Our collector read the newest week out of the export and discarded the rest. There was no retained copy, so there was nothing to adjudicate against. The desk could not tell whether its own figure was wrong.
4. The adjudication. We changed the collector to keep the export, pulled a fresh one, and read the whole series: 839 weeks back to June 2010, under the same source. D4 sits ABOVE D6 in 92.2% of those weeks, at a median premium of seventeen cents. A swapped label is a property of the source and would have inverted the entire history; it did not. A swap that began recently is refuted the same way - the last 26 weeks carry ten normal weeks interspersed among the inverted ones, and a swapped label cannot produce those. The disputed week is genuine: the third most inverted week in 838, inside a historic range that has been wider.
5. The restoration. The four prints are back, with a flag on the D4/D6 pair marking the week as rarer than one in a hundred and inviting a cross-check against a commercial assessment. They render as last prints with their age, as they did before.
What did not change, and should not be read as recovered. These are prints for the week of 20 July. They were never a current mark and are not one now. The renewable-diesel margin channel composed on them before the withholding and does so again; nothing was regained that the desk did not already have.
What we learned, which is the part worth publishing. The desk was one retention step away from being able to answer this in minutes, and instead spent a day unable to answer it at all. A collector that parses and discards leaves its own output unauditable. Ours keeps the file now.
The desk's mark for EU ETS (EUA) changes basis with this note.
What changed. Until today the EUA mark was the latest EEX auction clearing price. From today it is the front-month futures settlement, stated as such on every appearance with the contract month named, and the EEX auction clearing stays on every surface as the primary print beside it - the same pairing RGGI, California-Quebec and UK ETS carry.
Why. An auction clearing price and a futures settlement are different instruments by design: the auction is a primary issuance event and the futures settle is the traded market's daily close. Both are real, citable exchange prices, and nothing published on the auction basis was wrong - EEX auctions clear most weekdays, so the EUA print never had the shelf-life problem the quarterly RGGI print had. What the change buys is one basis across the complex: every compliance market on this tape now marks off the instrument a compliance buyer transacts against, and the auction remains the primary print it always was. It also retires the last scraped proxy from the desk's internal cross-checks - a mirror series was standing in for a settlement the desk could not previously reach, and it can reach the settlement itself now.
What it means for reading the note. The mark and the primary print will usually sit close together and will sometimes disagree by more than a percent - that is basis between an auction and the secondary market, not an error, and the desk no longer treats it as one. Each mark states what it is and which contract month it is; when an auction is the news, the auction leads.
The desk published a bearish reading of softer crude for California LCFS credits this morning. The chain runs the other way, and the correction is worth stating in full because the error was in the reasoning rather than in any number.
What the note said. That crude softening into late August "widens the diesel-minus-feedstock spread, supporting renewable diesel volumes and credit generation", and that with the 2026Q1 bank drawing down, "deficit quarters are bullish, softer crude is the offset". Two forces, pointing opposite ways, netting to neutral.
Why that is backwards. Diesel reprices with crude. Feedstock - used cooking oil, tallow, soybean oil - does not follow it down on the same timescale: it is priced by its own supply and by competing demand from food and oleochemicals. So when crude falls, the renewable-diesel margin over feedstock COMPRESSES rather than widens. A producer facing a thinner margin needs MORE value from the environmental credits - LCFS and the D4 RIN - to clear the same volumes, which is demand for credits, not supply of them. Softer crude is therefore supportive for LCFS credits. It points the same way as a drawing bank, and there is no offset.
This desk's own model already said so. The renewable-diesel channel in the desk's flow-channel layer composes its legs by sign, and it carries ULSD with a POSITIVE sign to the margin: diesel down is adverse for the margin. Its stated conclusion is that a widening margin is price-bearish for credits, so a compressing one is price-bullish. The note contradicted the model it sits beside. That is the part that matters: this was not an unknowable call, it was an internal inconsistency that nothing in the pipeline was set up to catch, because every gate the desk runs tests whether a claim is TRUE and none of them tested whether two of the desk's own claims agree.
What changes. The California LCFS desk view moves from neutral to neutral-to-constructive into the 8 September weekly print. The deficit quarters and the compressing margin now point the same way. The leg the desk still cannot price is the RIN side, and that is stated rather than assumed away.
What does not change. No mark, no print and no forward row moves. The CA LCFS credit print, its date and its range are as published. This is a correction to a read, and the read was ours.
The rule, stated first because it is the change. The reference contract for every mark on this tape is the one carrying the most OPEN INTEREST. It is named on the mark line with its rank and the date the ranking was read, and it moves when liquidity moves. Not the nearest expiry, not the newest vintage, not the contract a compliance buyer is assumed to use.
What was wrong. Between 27 August and today this desk rebased its compliance markets onto the FRONT MONTH, and justified it as the contract a compliance buyer transacts against. That was an argument, not a measurement, and the measurement disagrees on every market we checked. Ranked by open interest on the 1 September activity file, the December 2026 contract leads all five: it holds 54% of the EU curve's open interest, 82% of the UK curve, 78% of California-Quebec, 57% of RGGI and 85% of Washington. The front month we had been quoting ranked third or fourth on four of them, and SIXTH on Washington, where it held two tenths of one percent of the curve. On UKA and RGGI the December contract carries roughly twenty times the open interest of the strip we were marking.
Why this happens, since it is not obvious. A contract weeks from expiry has had its open interest rolled OUT of it. Compliance markets concentrate in the December contract, which is why December is the benchmark everyone quotes. The front month is the nearest contract; it is rarely the traded one. A mark on an illiquid strip is a price nobody can trade, and that is the objection to it - not that it was inaccurate, but that it was not a price you could act on.
Why now rather than at the roll, which is the fair question. The September contract expires within weeks. Had the desk said nothing, these marks would have moved to December on their own at the roll, and the change would have arrived with no announcement and no reason given - a basis change by attrition. We would rather make the change deliberately, say why, and state the rule that governs the next one. That also means the next roll is an expected event rather than another correction.
This is the second correction on the EU mark in two days, and that deserves a plain word. This morning's correction moved EU ETS from an auction clearing to a futures settlement - a change of INSTRUMENT TYPE. This one keeps the futures settlement and changes WHICH CONTRACT is read. They are different questions, and the first correction did not address the second. It is also not an EU correction: one rule is being adopted across the whole tape and five marks move on the same morning.
Alberta is held out, and the reason is the rule working. The Alberta EPC rebase queued for 4 September does not happen. The feed that carries Alberta prices has no open-interest column at all - not missing for this contract, absent from the data. Under a rule that selects on liquidity, a market with no liquidity data gets no mark, so Alberta keeps the administered TIER headline price as its reference. That is the settled position rather than a pause: the data the desk receives for these contracts does not contain open interest, so there is no liquidity for us to read and no reference we could justify. We would rather say that plainly than pick a contract and imply we had a reason.
What does not change. No forward row, no supply or demand figure, no auction print. Every primary print keeps its own date and basis. The mark lines now carry their contract, their open-interest rank and the date that ranking was read, so a reader can see the reference is the traded one rather than take our word for it.
from today this desk marks UK ETS off the front-month UKA futures settlement, the same basis RGGI and California-Quebec moved to on 27 August. The official gov.uk monthly average - our UKA mark until now - stays on every surface as the primary print beside the futures mark, the same pairing those markets carry.
Nothing published before today was wrong: the monthly average is an official settlement statistic and remains one. But a monthly average describes a month, and a month-on-month change quoted against it could not be checked by a reader against anything published daily. The front-month settlement gives the UKA line a daily, same-basis series - the delta becomes a verifiable number rather than an assertion.
The basis line names the instrument: front-month futures settlement with its contract month, priced in sterling. Where the two bases sit side by side they are labelled - a settlement mark and an official monthly average are different facts about the same market, and the desk carries both.
The desk's RGGI demand rows change base with this note, and the change is large enough - about 35% - that it deserves a plain account.
What the old base was. From 19 August the demand rows ran off ~60 million short tons, described in our own records from the day it was written as an assumption pending verification: a level read from a dashboard, flagged as an estimate, with the exact registry pull listed as owed. It was a placeholder that carried a placeholder's label. What it was not was a measured figure.
What the pull found. The awaited data has now landed twice over, by two independent routes. The market monitor's annual report - the compliance regime's own accounting - puts 2024 covered CO2 emissions across the ten RGGI states at 82.0 million short tons. EPA's unit-level emissions data for the same ten states, read under this desk's membership allowlist with every one of its 645 units accounted for, gives 82.9 million short tons - the two routes agree to 1.1%. The rows now anchor on the monitor's 82.0, because it counts covered sources: the same population the cap itself is written against. No population was found for which the old ~60M was the right 2024 ten-state figure.
What changes in the read. Demand near 80 million short tons against a 2027 cap of 69.8 million inverts the RGGI balance: instead of a surplus narrowing toward a 2029 flip, the rows now show a deficit from 2027 - roughly 10 million short tons in the first year, deepening as the cap declines toward 44.3 million by 2030. That is a change in the structural story, not a squeeze call: the market holds a private allowance bank last observed near 67 million short tons (July 2025, market monitor), which covers deficits of this size for years. The repricing question RGGI now poses is the pace of bank drawdown, not the arithmetic of any single compliance year.
What stays the same. The supply ladder is untouched - it was anchored to the Model Rule's own figures throughout. The mark and its basis are untouched. And the discipline that produced this correction is the one that will produce the next: every assumption in the forward rows carries its flag, and when a measurement lands, the desk moves to it and says so.
this morning's note said the amended Table 6-2 removes 117.7 Mt across 2027-2030, and gave the 2027 pre-amendment budget as 240. The removal is 118.3 Mt and the 2027 base is 240.6. The direction, the mechanism and the desk's view are unchanged - the cut is very slightly larger than we stated. CARB's own adopted summary describes it as the 'removal of 118 million allowances'.
How it happened is worth stating, because reading the primary did not prevent it. CARB's Final Regulation Order renders the 2027 cell as '240' in both its PDF and its Word version: the struck old figure and the inserted new one sit adjacent and the decimal is lost, the same way 2026 renders '254' where the codified regulation reads 254.0. The codified table reads 240.6, the pre-amendment schedule steps down by a uniform ~13.35 Mt a year - which 240.0 breaks and 240.6 fits - and our own joint California-Quebec supply row reconciles against Quebec's Decret 1126-2017 to the cent for 2028, 2029 and 2030 and requires 240.6 for 2027.
One consequence for readers following the offset debate: the removal is now exactly the same size as the 118.3 million current-vintage allowances created for the Manufacturing Decarbonization Incentive. They remain two different pools, and the thing that distinguishes them is vintage - the removal comes out of future-vintage budgets, the incentive allowances are created with a current vintage - not their size, which is now identical and settles nothing either way.
The desk's marks for RGGI and California-Quebec (CCA) change basis with this note.
What changed. Until today both instruments were marked off the most recent auction clearing price - the RGGI quarterly auction and the WCI joint auction. From today they are marked off the front-month futures settlement, stated as such on every appearance, with the contract month named.
Why. An auction print is a real, citable transaction price, and nothing carrying one was wrong. The problem is not accuracy, it is shelf life: auctions clear quarterly, so an auction print is an excellent mark in the week it prints and a progressively worse one for the three months until the next. Both ends of that are on this page today. The CCA mark is the 19 August joint auction, results published 26 August, and it sits close to where the front month is trading - an auction print at its freshest. The RGGI mark is Auction 72 of 3 June, a full quarter back, and the traded market has moved well above it. Same basis, same desk, three months apart. Marking off the front month is how the desk stops arguing views off whichever point in that cycle it happens to be standing on.
What it means for reading the note. Marks will move more often and by smaller amounts, because a front-month future reprices daily where an auction reprices quarterly. A change in the mark is no longer evidence that an auction has happened. Each mark states what it is and which contract month it is, and the auction prints remain in the note - they are still the primary print, and when an auction is the news it is the auction that leads.
A correction is held to a stricter standard than the note it corrects: every figure in it is re-derived from a pinned date, and every claim it makes about this desk's own prior labelling is diffed against the notes that carry it. Where a correction has itself been wrong, that is stated in the same block rather than edited away.