Reading America's first mandatory corporate emissions filings from the other side
On November 10, 2026, roughly 2,600 companies with more than $1 billion in annual revenue that do business in California will file the first economy-wide mandatory corporate greenhouse-gas emissions reports in United States history, under California SB 253.
A large industry serves the companies doing the filing. Almost nobody serves the other side: the investors, counsel and risk teams who will want to know who failed to file, and whose filing does not match what they have been saying elsewhere.
The corpus itself does not exist yet. The inconsistencies do. What follows was assembled from public data available today — company sustainability reports, SEC filings, CDP responses, and EPA's facility-level Greenhouse Gas Reporting Program — to show the kind of work I will be applying to the real filings in November.
The claim. PBF Energy's 2024 Sustainability Report states: "Since PBF Energy's operations in 2013, we have achieved a 30% reduction in absolute Scope 1 and Scope 2 GHG emissions across our operations" (p. 9). The CEO letter repeats it on p. 4, alongside a note that the company "approximately doubled our refining capacity through strategic acquisitions." The same claim appears in PBF's 2026 proxy statement filed with the SEC: "Measured against our operations in 2013, as of 2024, we have achieved a 30% reduction in absolute Scope 1 and Scope 2 GHG emissions across our operations."
The record. From EPA's Greenhouse Gas Reporting Program, using emissions PBF itself reported:
| Basis | 2013 | 2024 (company-reported) |
|---|---|---|
| The three refineries PBF operated in 2013 — Toledo, Delaware City, Paulsboro | 7.42 Mt CO2e | — |
| PBF's actual operations (six refineries) | — | 10.46 Mt CO2e (Scope 1) |
Measured against the operations PBF actually had in 2013, its emissions rose by roughly 40 percent.
The arithmetic that lands near −30%. Take PBF's six current refineries and roll them back to 2013 under whoever owned them then: Torrance (ExxonMobil, 3.04 Mt), Martinez (Shell, 3.47 Mt), Chalmette (ExxonMobil/PDVSA, 1.47 Mt), plus PBF's own three (7.42 Mt). That gives 15.4 Mt, and against 2024's 10.46 Mt it produces −32%. Components are rounded; the unrounded total is 15,405,850 tonnes.
I could find no disclosure defining the baseline that way. Not in the 2024 Sustainability Report, the 2025 TCFD Report, the data supplements, the 10-K, or the proxy. Three further details:
The strongest defence. The CEO letter mentions doubled capacity two sentences earlier, and report boilerplate notes that measurement standards are still developing. A Scope 2 explanation does not survive contact with the numbers: a genuine same-company −30% would require roughly 8.75 Mt of Scope 2 across three refineries in 2013, about 20 TWh of purchased power. PBF's Scope 2 today is about 0.86 Mt across six refineries.
Why it matters in November. SB 253 will require PBF to publish an entity-level Scope 1 and 2 figure on a fixed boundary, eventually assured. That creates a public benchmark against which this framing can be tested by anyone.
Sources. PBF Energy 2024 Sustainability Report, pp. 4 and 9. PBF Energy DEF 14A (2026), SEC accession 0001140361-26-009843. EPA GHGRP facility data, reporting years 2013 and 2023. PBF Energy 2025 TCFD Report. PBF Energy 2022 Sustainability Report footnotes.
NRG set a goal of cutting US emissions 50% by 2025 against a 2014 base, and announced early success twice. Its 2024 SASB report says the company was "thrilled about achieving our goal two years in a row," adding that results "could be impacted by volatility within the power markets." In the goal year itself, the FY2025 10-K reports 47 percent, using that same phrase, and never uses the word "miss."
The reason two baselines are in play at once. NRG's sustainability-linked notes (3.875% due 2032, with $480 million outstanding of a $1.1 billion issue) test the same goal as a fixed absolute cap tied to the original 2014 base: 2025 emissions at or below 31.7 Mt CO2e. NRG reported roughly 30 Mt, so the covenant was satisfied, as its Q2 2026 10-Q states. The 47 percent figure measures that same tonnage against a 2014 baseline the company has since recalculated downward to 57 Mt.
Same year, same tonnage, two answers. Base-year recalculation is permitted under the GHG Protocol; the point is how much the answer depends on which vintage of the baseline is being used. A fixed public series is what ends that ambiguity.
Separately, NRG's own materials carry two different FY2024 Scope 1 figures: 25,035,662 tonnes in a KPMG-reviewed statement, and 27,457,767 in its SASB index. Both are equity-share-based generation inventories, 9.7 percent apart, unreconciled. A report footnote states that its numbers "do not match what is listed in the 10-K."
Sources. NRG Energy FY2025 10-K, Item 1 greenhouse-gas section and Note 12. NRG Energy Q2 2026 10-Q, SEC accession 0001013871-26-000020. Second Supplemental Indenture, 8-K accession 0001104659-21-108598. NRG 2024 SASB Index. NRG FY2024 Statement of GHG Emissions (KPMG-reviewed).
Republic Services, Waste Management and Waste Connections are among the largest methane reporters in the coming corpus, and all three derive landfill emissions from models rather than measurement, under 40 CFR 98 Subpart HH. Republic's own FY2025 SASB filing states it plainly: "In the absence of viable measurement methods, landfill emissions are based on projection models."
The largest aerial survey of US landfills yet published — Cusworth et al., Science, 2024, covering more than 200 sites — found emission rates averaging 1.4 times the GHGRP figures, reaching 1.4 to 2.7 times at measured sites, with 52 percent of surveyed landfills showing observable point-source plumes. EPA-coauthored follow-up research finds operators tending to elect the method that yields the lower number: one assumes 72 percent average gas-collection efficiency, the other 47 percent.
CARB accepts GHG Protocol-conformant figures, so the same modelled numbers will flow into the SB 253 filings, initially without assurance.
Public context on the three companies: Republic appeared in NASA/JPL California super-emitter flight data, and its Sunshine Canyon landfill was placed under a South Coast AQMD abatement order in March 2025 following more than 150 notices of violation. EPA found 51 exceedances at Waste Management's Prairie Hill landfill where operator monitoring had found few to none. Waste Connections' Chiquita Canyon elevated-temperature event has produced an EPA Finding of Violation in June 2024, a California DTSC endangerment order in April 2025, a $480.8 million closure charge, and roughly 11,700 civil plaintiffs.
Sources. Republic Services FY2025 SASB Report, p. 3. Cusworth et al., Science 383:1499 (2024). Stark, Tian and Krause, EPA Office of Research and Development, landfill model investigation (2024). South Coast AQMD abatement order, March 2025. Waste Management and Waste Connections FY2025 10-K filings. CARB SB 253 workshop materials, August 2025.
Eastman Chemical weakened both halves of its climate commitment inside consecutive SEC filings. Item 1 of the FY2024 10-K reads: "Eastman has committed to reduce its absolute scope 1 … and scope 2 … emissions by approximately one-third by 2030 … in order to achieve carbon neutrality by 2050." The same sentence in the FY2025 10-K reads: "… by 30 percent by 2035 … in order to achieve net-zero operations by 2050." Deadline moved five years, ambition softened, in an annual report.
The published progress figure moved with it. Eastman's 2024 Sustainability Report claimed a 22.3% reduction against its 2017 base; the 2025 report claims 17.4%. The cause is a 452,787-tonne upward restatement of 2023 emissions, disclosed in a single footnote in a separate document — "For comparability, historical values for 2023 were updated to reflect an identified Scope 1 immaterial data error correction" — with no magnitude, cause or affected sites given. That restatement erased 5.4 points of previously claimed progress, and the figure it revised had already carried PwC review-level assurance.
In fairness, and this matters: on a restated like-for-like basis Eastman improved slightly year over year, from 16.9% to 17.4%. Its physical progress did not reverse. What reversed was the published history. Note also that the working 2017 baseline of roughly 8,415,000 tonnes does not match the 8,086,164 figure Eastman gave CDP, and the current baseline has never been published.
Underneath the flat headline the components moved in opposite directions. Scope 1 rose 600,388 tonnes, up 10.0%, with no explanation offered anywhere in the report. Market-based Scope 2 fell 639,577 tonnes, down 63.3%, as renewable energy purchases went from 162,710 to 1,361,655 MWh — the assurance statement shows location-based Scope 2 at 871,833 against market-based at 371,443, a 500,390-tonne gap closed by certificates. In the same report, the renewable purchased-electricity goal was retired, alongside the SO2, NOx, energy-productivity, gender-parity and US racial-equity goals. Eastman holds no SBTi validation and states it is awaiting the initiative's chemical-sector guidance.
Sources. Eastman FY2024 10-K (filed 14 Feb 2025) and FY2025 10-K (filed 13 Feb 2026), Item 1. Eastman 2024 Sustainability Report; 2025 Sustainability Report, p. 57 and goals tables; 2025 Sustainability Data Sheet, p. 2 note d; 2024 ESG Data Sheet, p. 9 and note c; independent accountants' review report, 2025 report pp. 61–64; Eastman CDP response.
Air Products moved its emissions baseline twice across three consecutive reports, from 2015 to 2023 to 2025. The FY2025 boundary overhaul lifted reported Scope 1 by 47 percent, from 16.9 to 24.9 Mt, largely from facilities previously left outside the inventory. Prior years were withdrawn as "not directly comparable" with no bridge provided, a full year of goal-performance reporting was skipped, and the named "Third by 2030" target was retired without explanation. The replacement target's denominator is non-GAAP adjusted operating income, carrying the company's own note that it "may be influenced by factors unrelated to emissions performance."
Sources. Air Products Sustainability Reports 2024, 2025 and 2026, notably the FY2025 boundary note and FY2026 goals table.
Vistra holds an SBTi-validated target of a 58 percent absolute Scope 1 and 2 reduction by 2028 against a 2018 base. At the end of 2025 it stood 22 percent below that base, with Scope 1 up 7.4 percent year on year in a Deloitte-reviewed statement. Coal retirements have slipped, Baldwin by two years; 4,165 MW of coal carries no retirement date; about 2,600 MW of gas has been added and roughly 5,500 MW more agreed. Reaching the target now requires a 46 percent reduction in three years.
Sources. Vistra 2025 Sustainability Report targets table; Vistra FY2024 and FY2025 Statements of GHG Emissions (Deloitte-reviewed); Vistra FY2025 10-K retirement schedule.
AES reports two different FY2023 Scope 1 figures in its own documents, 499,894 tonnes apart: 33,901,488 in its LRQA-assured report and 33,401,594 in its CDP response, same year and same equity-share boundary. The CDP response claims 100 percent third-party verification and attaches the assurance statement covering the other number, while marking past-year recalculation as "No."
Sources. AES 2023 Improving Lives Report, LRQA assurance statement; AES CDP 2024 response, questions 7.6.1, 7.9.1 and 7.1.3.
California Resources stated in September 2025 that it had "reduced Scope 1 and 2 greenhouse gas emissions by 27% compared to the company's 2020 baseline." Roughly three-quarters of that reduction sits on the Aera Energy side of the combined baseline — assets CRC did not own until July 2024, and the 2020 baseline is 53 percent Aera. CRC's legacy operations fell about 10 percent and were flat into 2023. The scope of its "Net Zero 2045" goal narrowed from Scopes 1, 2 and 3 to Scopes 1 and 2. The release carries neither caveat.
Sources. California Resources 2023 and 2024 Sustainability Reports; CRC Q1 2025 10-Q, "Responsible Net Zero Goal"; CRC 2025 CDP response; CRC press release, 18 September 2025.
CARB's preliminary covered-entity list omits several large California-operating emitters. Checked at parent level and under known California subsidiary names: Vistra, Marathon Petroleum, Valero, Waste Management, AES and PBF Energy all operate significant California facilities and none appears. Absence from the list does not remove a filing obligation, which attaches by statute — but it does mean these companies are not on the regulator's tracking sheet, which makes them the first names worth checking on November 10.
The list also sweeps in companies that appear not to be covered at all. One row, flagged as owing both SB 253 and SB 261, is a Pomona manufacturer of barber-and-beauty-shop equipment with roughly six employees. It is not CF Industries Holdings, the S&P 500 nitrogen producer whose name it resembles; that company does not appear on the list at all. CARB's own header note explains the mechanism: the list was built from "a partial match of company name fields" against Secretary of State filings current only through March 2022.
Where the list does name a company, it frequently contradicts itself. Its 4,160 rows cover 3,127 unique entities, so 1,033 rows are duplicates — about a quarter of the file. More importantly, 262 entities (8.4 percent) carry contradictory coverage determinations, appearing in one row as owing both laws and in another as owing only one. Accenture, Abercrombie & Fitch Management, ADT Services, AECOM C&E, Akzo Nobel Services and Advance Auto Parts are among them. A company cannot reliably learn from the regulator's own list whether it owes a November 10 emissions filing.
The voluntary run-through drew almost nobody. The complete SB 261 climate-risk docket, all 185 approved submissions, amounts to 5.9 percent of the 3,127 unique entities named on CARB's list. Almost no large emitters took part; the notable exceptions were PG&E, International Paper and PBF Energy.
Disclosure posture is already measurable. Across the 21 large emitters examined here, exactly one 10-K names SB 253 or SB 261 by bill number: International Paper, whose filing also quotes a deadline that has since moved. ExxonMobil, Waste Connections, Linde, Dow, ADM and every power company in the sample are silent on California climate disclosure. ExxonMobil, Chevron, Valero and Waste Connections publish no current CDP responses.
The pre-2026 baseline is unstable. Eight of those 21 companies restated Scope 1 within about fourteen months, ranging from a 50 percent increase to a 26 percent decrease, several while also resetting target baselines. That instability is what a day-one reader will be comparing against.
Sources. CARB Preliminary List of Reporting/Covered Entities, posted 24 September 2025 (4,160 rows; counts computed from the published file). California Secretary of State business registry. CARB SB 261 public docket, approved comments listing. EPA GHGRP facility and parent-company files, reporting years 2021–2023. Company 10-K filings for the 21 emitters, retrieved via SEC EDGAR full-text search.
Company claims were gathered separately from the EPA attributions and without sight of them, so the two sides were assembled independently. Every candidate finding then went through a structured red-team pass built to kill it, run against the source documents rather than my notes. Candidates that failed are retained in full and available on request: a Sempra boundary restatement that looked aggressive turned out to be orthodox GHG Protocol operational-control adoption, transparently dual-reported, and so it is not here.
Where boundaries align, the EPA-based attribution reproduces companies' own reported totals closely. For reporting year 2023, PBF reported 10.78 Mt against 10.76 Mt computed; Phillips 66 reported 31.6 Mt against 31.3 Mt computed, including the step change caused by its DCP Midstream consolidation.
Two traps in the underlying data, both of which will bite anyone building this series.
EPA's parent-company field is the only link between facility emissions and a corporate parent, and it is uncurated. It lags divestitures by years and splits corporate families across name variants. BP is the documented case: EPA labeled the Cherry Point refinery "BP Products North America" in 2021 and 2022, then "BP America" in 2023, while the refinery's actual emissions barely moved, at 2.20, 2.28 and 2.18 Mt. Any series keyed to a single name string therefore misstates BP's own baseline by 2.2 Mt. Prudhoe Bay likewise remained attributed to BP for two and a half years after its sale to Hilcorp. Across the 1,283 US parents emitting more than 100,000 tonnes, 155 of them, or 12.1 percent, show trend differences greater than 10 percentage points depending on whether name variants are grouped. Westlake Chemical reads −36.8 percent one way and +7.4 percent the other.
Second, EPA's published workbooks carry malformed sheet dimensions in several reporting years, which causes standard spreadsheet libraries to silently truncate whole emissions categories. In the 2021 and 2022 files, the oil-and-gas production, gathering, pipeline, distribution and SF6 sheets each read as four rows instead of hundreds, producing spurious upward trends for oil-and-gas-weighted companies. I hit that one in my own pipeline and corrected it; every figure here is post-correction. Every trajectory in this document was verified facility by facility, in each year cited.
Limits worth stating. The 21 companies examined here were selected by US emissions scale and are not a random sample, so the disclosure-posture counts describe that group rather than the whole corpus. EPA data covers Scope 1 at facilities above the 25,000-tonne reporting threshold, so it cannot see Scope 2, Scope 3, or smaller sites. Nothing here alleges unlawful conduct; these are documented inconsistencies in public disclosures, presented for professional evaluation.
About me, and conflicts. I am an independent researcher working on this full time through the November filing deadline. I hold no position, long or short, in any company named in this document, and I accept no compensation from any company named here or from its adversaries. If I find an error in my own work I will correct it in writing and tell anyone who received the original.
On November 10 the filings land. Within 72 hours I will publish a census of who filed and who did not, measured against a cleaned covered-entity universe rather than CARB's raw list, with assurance status, boundary choices and restatement flags across the corpus. Company-specific dossiers follow, applying the Finding 1 treatment to individual filings.
Day-one products and pricing are available on request. If you would like the census, a dossier on a particular company, or simply the underlying workings for anything above, email firstcorpusresearch@gmail.com.