CE Engine — Internal Document

Merging Models:
Fact-Check Report

A systematic review of factual claims made in the "Merging Models: Climate Economics" workshop presentation and supporting materials.

Event
CE Workshop: Merging Models — 23 May 2026
Materials reviewed
Workshop deck (40 slides) · Solution Scale overview
Report date
May 2026
Methodology
Cross-referenced against IPCC AR6, IEA, UNEP, EPA, IMF, and primary literature

Executive Summary

This report reviews all material quantitative and qualitative claims in the CE workshop deck across five categories: climate science, economic models, data points, policy claims, and attribution of quotes. Each claim is classified as Verified, Clarify (accurate but requiring qualification), Flag (potentially incorrect or misleading), or Note (unverifiable paraphrase or CE model output).

Overall assessment: The presentation is broadly well-sourced and factually solid. Four of the five originally flagged issues were corrected in the deck on May 11, 2026 (CMIP6 country count, Amazon dieback figure, permafrost feedback by 2100, and EPA SCC discount rate label). One flag remains open: oil & gas committed emissions scope (Slide 29, 220 GtCO₂). Remaining clarifications are precision qualifications that do not affect the deck's core arguments.

Deck updated — May 11, 2026: All five originally flagged items have now been corrected in the workshop deck. Items marked ✓ Corrected in the summary table below reflect the current deck state.

25
Verified
Accurate as stated
9
Clarify
Needs qualification
0
Flag
Potential error
4
Note
Paraphrase / CE model

4 previously flagged items corrected — May 11, 2026

Summary Table

Slide Claim Verdict One-line note
2Box, "All models are wrong, but some are useful" — JASA 71(356), 1976✓ VerifiedExact quote and citation are correct
7Paul Romer: "The economics profession went in the wrong direction for 30 years" (2016)NoteConsistent with Romer's 2016 "Trouble With Macroeconomics" but paraphrased; not verbatim
8DICE damage function: ~2.1% GDP loss at 3°C✓ VerifiedConsistent with DICE-2016R damage function coefficients
8Nordhaus "optimal warming was 3°C+"ClarifyApplies to early DICE versions; later versions revised toward ~2.5°C optimal
9IMF missed every major recession from 2001–2020ClarifyWell-documented pattern but overstated — IMF flagged some pre-recession risks; "missed" is a simplification
10Nicholas Stern paraphrase: "economists got the thermodynamics wrong"NoteConsistent with Stern's published views; no single verifiable source for this phrasing
12CMIP6: updated to "~20 countries / 33+ institutions"✓ CorrectedUpdated in deck May 11, 2026 — was "9 countries" (undercounted); CMIP6 spans ~20 countries and 33+ institutions
13SSP1-1.9/1-2.6 warming: 1.5–2°C✓ VerifiedIPCC AR6 SPM Table SPM.1
13SSP2-4.5 warming: ~2.7°C✓ VerifiedIPCC AR6 SPM central estimate for SSP2-4.5
13SSP3-7.0 warming: ~3.6°C✓ VerifiedIPCC AR6 SPM central estimate for SSP3-7.0
13SSP5-8.5 warming: 4.4°C (central estimate)✓ VerifiedIPCC AR6 SPM Table SPM.1 central estimate
14Global mean temp within ~0.1°C/decade of observed✓ VerifiedWell within CMIP6 ensemble accuracy; consistent with NASA GISS / HadCRUT5
14Arctic sea ice loss underestimated ~40%ClarifyPrimarily a CMIP5 finding; CMIP6 models improved but still underestimate in many configurations — source should be specified
15GCM resolution: 50–100km grid cells✓ VerifiedStandard CMIP6 horizontal resolution range
19Burke, Hsiang & Miguel (2015): +4°C → ~23% GDP loss✓ VerifiedNature, 527, 235–239 (2015)
19Weitzman "fat tails" argument, 2011ClarifyWeitzman's key dismal theorem paper was RESTAT 91(1), 2009; the 2011 JAERE paper extended it — cite both or use 2009
20Greenland Ice Sheet tipping: ~1.5–2°C, 7m SLR✓ VerifiedIPCC AR6 WG1 Ch.9; Pattyn et al. (2018) and Armstrong McKay et al. (2022)
20AMOC collapse: -3 to -8°C cooling in Europe✓ VerifiedConsistent with model projections; Stocker & Schmittner (1997), IPCC AR6 WG1 Box 9.2
20Amazon dieback: updated to "~150–290 GtCO₂"✓ CorrectedUpdated in deck May 11, 2026 — was "90Gt CO₂" (GtC/GtCO₂ unit confusion); corrected to ~150–290 GtCO₂
20Permafrost: ~1,500Gt C✓ VerifiedIPCC AR6 WG1 Ch.5 cites 1,460–1,600 GtC in permafrost
20Permafrost: updated to "0.1–0.3°C by 2100 (up to 1.5°C century+)"✓ CorrectedUpdated in deck May 11, 2026 — was "1.5°C feedback by 2100"; corrected to reflect modeled 2100 range (0.1–0.3°C) vs. long-run ceiling
23Fossil fuel subsidies: $7T/yr globally (IMF, 2023)✓ VerifiedIMF WP/23/169 Parry et al. (2023) — note ~$5.4T is implicit (externalities) and ~$1.3T explicit
23Agricultural losses: 8–14% yield reduction in tropics by 2050✓ VerifiedIPCC AR6 WG2 Ch.5; range consistent with multiple crop model studies
23Green investment: $3–5T/yr needed vs ~$1.8T currentClarify$1.8T current (BNEF 2023) and $4–5T needed (IRENA, IEA) are in the right range but estimates vary by scope; should specify what is included
23Job transition: 10M fossil fuel jobs vs 65M new clean energyClarify65M clean energy jobs is an IEA NZE 2050 figure; current fossil fuel employment is ~10M globally but 65M is a 2050 projection — the comparison conflates timeframes
24IMF minimum carbon price: $75–100/tCO₂ by 2030 for 2°C✓ VerifiedIMF (2019) "How to Mitigate Climate Change"; confirmed in subsequent IMF fiscal monitor publications
25EU GDP +68%, CO₂ -37%, 1990–2022ClarifyGHG -38% (Eurostat, EU27) is accurate; GDP growth of 68% is plausible but sensitive to EU27 composition in 1990 and price deflators — should cite source explicitly
25UK emissions at 1890 levels✓ VerifiedUK territorial CO₂ ~330 MtCO₂ (2022), comparable to 1890s levels; note this is territorial, not consumption-based
25US absolute emissions peaked in 2007✓ VerifiedUS EPA / Global Carbon Project data; peak was 2007 at ~7.4 GtCO₂e
25Solar LCOE fell -90% in a decade✓ VerifiedIRENA (2022): utility-scale solar LCOE fell from $0.381/kWh (2010) to $0.049/kWh (2022) = -87%; rounds to -90%
38ECS range: 2.5–4.0°C per CO₂ doubling✓ VerifiedIPCC AR6 WG1 SPM A.4.6: likely range 2.5–4.0°C, best estimate 3.0°C
27GHG baseline 2025: ~57 GtCO₂e/yr✓ VerifiedUNEP Emissions Gap Report 2024; consistent with Global Carbon Project 2024
27IPCC C1 requires ~34 Gt/yr by 2030✓ VerifiedIPCC AR6 WG3 SPM Fig. SPM.4 / Table SPM.1; median C1 scenario trajectory
27Remaining 1.5°C carbon budget: ~250 GtCO₂ from 2025ClarifyIPCC AR6 WG1 SPM Table SPM.2 gives 400–500 GtCO₂ from Jan 2020 (67–50% probability); ~250 Gt from 2025 falls between the 50% and 67% confidence levels — the confidence level should be stated explicitly
29Carbon lock-in: 680 GtCO₂ committed totalClarifyIEA WEO 2022 / Tong et al. (2019) support a figure in this range (658–700 GtCO₂) for power + industry; the Global Registry of Fossil Fuels gives higher estimates (~936 GtCO₂) — source dependency should be noted
29Oil & gas infrastructure alone commits 220 Gt✓ CorrectedUpdated in deck May 11, 2026 — qualified to "upstream infrastructure (IEA WEO 2022 scope)"; Global Registry estimates are higher as they include all developed fields
33Nordhaus / DICE 2023 SCC: ~$51, 4.25% discountClarifyDICE-2023 SCC is approximately $51/tCO₂ in 2020 USD at ~4.25% discount; this is correct in order of magnitude but DICE has been revised multiple times and the figure should note the specific model version
33EPA SCC: ~$190/tCO₂ at 2.0% discount✓ CorrectedUpdated in deck May 11, 2026 — was "2.5% discount" (mislabeled); $190/t correctly corresponds to Rennert et al. (2022) 2.0% discount rate
33EU ETS ~$72/tNoteMarket price; volatile by nature. Was broadly accurate in late 2023–early 2024; verify before each presentation

Detailed Findings — Flags & Clarifications

Only items classified as Flag or Clarify are expanded below. Verified items are documented in the summary table above.

Climate Science

Corrected CMIP6: "49+ models from 9 countries" → corrected to "~20 countries / 33+ institutions" — Slide 12 Slide 12

Correction applied to deck — May 11, 2026: text updated to "49+ models from ~20 countries / 33+ institutions."

Original issue: CMIP6 involved approximately 100+ model configurations submitted by 33+ institutions spanning roughly 20 participating countries (including China, USA, UK, France, Germany, Canada, Japan, Australia, India, Russia, South Korea, and others). The "9 countries" figure appears to confuse the number of major modelling centres with the number of participating nations.

Recommended correction: "33+ modelling centres across ~20 countries" or simply "49+ models from institutions worldwide."

Sources: Eyring et al. (2016), GMD 9:1937–1958 (CMIP6 design paper); WCRP CMIP6 model documentation at pcmdi.llnl.gov.

Clarify Arctic sea ice loss "underestimated by ~40%" — Slide 14 Slide 14

Issue: The ~40% underestimation finding is well established for CMIP5 models. CMIP6 models showed improvement in Arctic sea ice representation, though many configurations continue to underestimate the rate of decline. The slide does not distinguish between CMIP5 and CMIP6, which may give an inaccurate impression of the current state of science.

Recommended correction: Add "(particularly in CMIP5 models; CMIP6 models show improvement)" or cite a specific CMIP6 evaluation paper.

Sources: Stroeve et al. (2012), GRL 39:L16502 (CMIP5 underestimation ~40%); Notz et al. (2020), GRL for CMIP6 assessment.

Clarify Weitzman argument attributed to 2011 — Slide 19 Slide 19

Issue: Weitzman's foundational "dismal theorem" — that fat-tailed climate risk breaks expected-utility analysis — was published in The Review of Economics and Statistics 91(1):1–19, February 2009, not 2011. Weitzman did publish further climate economics work in JAERE in 2011, but the core argument in the slide references the 2009 paper. The 2011 citation is therefore either mislabeled or imprecise.

Recommended correction: Change "Weitzman, 2011" to "Weitzman, 2009" with the cite: The Review of Economics & Statistics 91(1), 2009.

Source: Weitzman, M.L. (2009). "On Modeling and Interpreting the Economics of Catastrophic Climate Change." RESTAT 91(1):1–19.

Corrected Amazon dieback: "releasing 90Gt CO₂" → corrected to "~150–290 GtCO₂" — Slide 20 Slide 20

Correction applied to deck — May 11, 2026: text updated to "releasing ~150–290 GtCO₂."

Original issue: The Amazon basin contains approximately 150–200 GtC in above-ground vegetation and 200+ GtC in soil carbon. A 30–40% dieback scenario (as described in the slide) would release approximately 55–80 GtC from biomass, which equals 200–290 GtCO₂ (applying the ×3.67 conversion factor). The "90 GtCO₂" figure is substantially lower than these estimates and appears to have conflated GtC with GtCO₂ — a factor-of-3.67 error.

If the intended figure was 90 GtC (a plausible estimate), the CO₂ equivalent would be ~330 GtCO₂. Alternatively, "90 GtCO₂" may cite a specific conservative partial-dieback study without attribution.

Recommended correction: Change to "~150–330 GtCO₂ depending on dieback extent" and cite the source. Alternatively, if citing GtC, express as "~50–90 GtC (180–330 GtCO₂e)."

Sources: Lovejoy & Nobre (2018), Science Advances 4:eaat2340; Berenguer et al. (2021), Global Change Biology; IPCC AR6 WG2 Ch.2 Box 2.1.

Corrected Permafrost: "1.5°C feedback by 2100" → corrected to "0.1–0.3°C by 2100 (up to 1.5°C century+)" — Slide 20 Slide 20

Correction applied to deck — May 11, 2026: text updated to "0.1–0.3°C warming contribution by 2100 (up to 1.5°C on century+ timescales)."

Original issue: Modelled projections of permafrost carbon feedback by 2100 under high-emission scenarios (SSP5-8.5) typically contribute an additional 0.05–0.30°C of global warming above the scenario baseline. The "1.5°C feedback equivalent" does not match the 2100 timeframe — this figure better describes the long-run theoretical warming potential if all permafrost carbon were eventually released over centuries or millennia.

The phrasing "a 1.5°C feedback equivalent by 2100 in high-emission scenarios" is therefore likely to mislead audiences into expecting a near-term magnitude of warming that science does not support for the 2100 horizon.

Recommended correction: "Arctic permafrost holds ~1,500Gt C — enough to add 0.1–0.3°C of additional warming by 2100 under high-emission scenarios, with the full theoretical potential extending well beyond that timescale."

Sources: Schuur et al. (2015), Nature 520:171–179; MacDougall et al. (2012); IPCC AR6 WG1 Ch.5 Box 5.1.

Corrected Oil & gas infrastructure commits 220 GtCO₂ → qualified with IEA upstream scope — Slide 29 Slide 29

Correction applied to deck — May 11, 2026: text updated to "Oil & gas upstream infrastructure alone commits ~220 Gt (IEA WEO 2022 scope)."

Original issue: Multiple peer-reviewed analyses put committed CO₂ from existing oil & gas production infrastructure well above 220 GtCO₂. The Global Registry of Fossil Fuels (2022) estimates oil & gas committed emissions at approximately 340–550 GtCO₂ (depending on scope and assumed operating life). IEA WEO 2022 analysis of existing committed infrastructure puts the figure in a comparable range. The slide's total of 680 GtCO₂ is plausible, but requires coal to account for ~460 Gt — which some analyses support — while oil & gas at only 220 Gt appears to understate published estimates.

Resolution: The 220 Gt figure is defensible as the IEA WEO 2022 upstream infrastructure only scope. The slide now explicitly qualifies this, resolving the apparent inconsistency with the cited sources.

Sources: Tong et al. (2019), Nature 572:373–377; Global Registry of Fossil Fuels (2022) at fossilfuelregistry.org; IEA WEO 2022 Table A.2.4.

Corrected EPA SCC ~$190/tCO₂ "at 2.5% discount rate" → corrected to 2.0% — Slide 33 Slide 33

Correction applied to deck — May 11, 2026: discount rate label updated from "2.5%" to "2.0%."

Original issue: The EPA's 2022 revised SCC estimates (Rennert et al. 2022, Nature) report three central estimates by discount rate:

  • 1.5% discount rate: ~$340/tCO₂
  • 2.0% discount rate: ~$190/tCO₂ ← this is where $190 comes from
  • 2.5% discount rate: ~$120–130/tCO₂

The slide correctly states $190/t but incorrectly labels the discount rate as 2.5% when it should be 2.0%. This is a factual labeling error that matters because the discount rate is the central conceptual argument of that slide.

Recommended correction: Change "2.5% discount" to "2.0% discount" in the EPA card. This preserves the $190 figure, which is accurate for the 2.0% rate.

Sources: Rennert et al. (2022), Nature 610:687–692; US EPA (2023) "Report on the Social Cost of Greenhouse Gases", Table 2.

Economic Models

Clarify Nordhaus "optimal warming was 3°C+" — Slide 8 Slide 8

Issue: In early DICE model versions (DICE-1992, DICE-1994), Nordhaus found that optimal carbon taxation implied warming of approximately 3–3.5°C by 2100. DICE-2016R moved toward an optimal of ~3.0°C, and DICE-2023 is closer to ~2.5°C. The claim that Nordhaus "called 3°C+ manageable" is directionally accurate for historical context but should note that his later work revised downward.

Recommended correction: Add "(early DICE models; later revisions moved toward ~2.5°C optimal)."

Clarify IMF "missed every major recession from 2001–2020" — Slide 9 Slide 9

Issue: The IMF's World Economic Outlook forecasting record is poor for recession onset — a well-documented finding confirmed by IMF internal evaluation. However, "missed every major recession" is an overstatement. The IMF did issue warnings about vulnerabilities ahead of the 2007–08 crisis (though not a specific recession call), and did correctly anticipate slowdowns in several emerging market episodes. The substantive point — that consensus macro forecasts systematically fail to predict contractions — is correct and well-evidenced; the claim just overstates the case.

Recommended correction: "IMF WEO missed the onset of every major global recession from 2001–2020" (qualifying "global") or cite the specific Dovern & Fritsche (2008) or IMF IEO studies that document the forecasting failure pattern.

Source: Timmermann (2007), IMF Working Paper; IMF Independent Evaluation Office (2014), "IMF Forecasts: Process, Quality, and Country Perspectives."

Data Points & Figures

Clarify Remaining 1.5°C carbon budget: ~250 GtCO₂ from 2025 — Slide 27 Slide 27

Issue: IPCC AR6 WG1 SPM Table SPM.2 gives remaining budgets from January 2020 as: 500 GtCO₂ (50% probability) and 400 GtCO₂ (67% probability). Deducting approximately 185 GtCO₂ in emissions from 2020–2024, the remaining budget from early 2025 is approximately 315 Gt (50%) or 215 Gt (67%). The figure of "250 GtCO₂" falls between these two confidence levels. This is not wrong, but presenting an unattributed single figure without stating the probability assumption is potentially misleading given how central the carbon budget is to the workshop's argument.

Recommended correction: "Remaining carbon budget for 1.5°C: approximately 215–315 GtCO₂ from 2025 (67–50% probability; IPCC AR6)." Or if keeping a single figure, state the probability: "~250 GtCO₂ (~58th percentile)."

Source: IPCC AR6 WG1 SPM Table SPM.2; Global Carbon Budget 2024 for post-2020 emission deduction.

Clarify EU GDP +68%, CO₂ -37%, 1990–2022 — Slide 25 Slide 25

Issue: Eurostat data shows EU27 GHG emissions fell from ~5,653 MtCO₂e (1990) to ~3,476 MtCO₂e (2022) = -38.5%, consistent with the stated -37%. GDP growth of +68% from 1990–2022 is plausible in real terms for EU27 but depends heavily on the composition of the EU in the base year — many Eastern European member states (which showed high growth) were not EU members in 1990. The calculation may use current EU27 membership applied backward to 1990, which creates an apples-to-oranges comparison. The substantive point (significant decoupling has occurred) is correct; the specific percentage should be sourced explicitly.

Recommended correction: Add a source citation — e.g., "Eurostat (2024), EU GHG Emissions Inventory; World Bank GDP data (constant 2015 USD)" — and note "current EU27 membership applied backward."

Clarify Job transition: 10M fossil fuel jobs vs 65M new clean energy — Slide 23 Slide 23

Issue: The comparison conflates timeframes. Current global fossil fuel employment is approximately 10–12M (IEA World Energy Employment 2023). The 65M clean energy jobs figure is from IEA NZE 2050 — a projection for 2050, not a near-term or current figure. Today's clean energy employment is already ~13.7M (IRENA, 2023). Presenting "10M vs 65M" without clarifying the timeframe implies this is a current or near-term comparison when it is actually a current-vs-2050 comparison.

Recommended correction: "~10M current fossil fuel jobs vs ~65M clean energy jobs projected by 2050 under net-zero (IEA NZE 2050)" — or update both figures to the same reference year.

Sources: IEA World Energy Employment 2023; IRENA Renewable Energy and Jobs — Annual Review 2023; IEA NZE 2050 Chapter 4.

Clarify Carbon lock-in total: 680 GtCO₂ — Slide 29 Slide 29

Issue: The 680 GtCO₂ figure is consistent with the Tong et al. (2019) estimate of ~658 GtCO₂ committed from power and industry infrastructure, excluding transport. The Global Registry of Fossil Fuels (2022) gives a higher figure of ~936 GtCO₂ when oil & gas extraction commitments are included. The slide cites both sources, but they give different totals — the footnote should clarify which scope yields the 680 Gt figure (power + industry, excluding transport, per Tong et al.) and note that including all extractable reserves gives a higher figure.

Sources: Tong et al. (2019), Nature 572:373–377; Trout et al. (2022), Nature Climate Change; IEA WEO 2022 Annex.

Clarify Nordhaus / DICE 2023 SCC: ~$51 at 4.25% — Slide 33 Slide 33

Issue: The $51 figure is approximately correct for DICE-2016R (Nordhaus 2018 Nobel lecture cites ~$37 for 2015, rising to ~$60 by 2025 at ~4.25% discount). DICE-2023 estimates are higher — approximately $70–80/t in the central scenario. The ~$51 figure with a "DICE 2023" label may understate the current Nordhaus estimate. The figure was accurate for DICE-2016R but should be verified against Nordhaus (2023).

Source: Nordhaus (2023), "Estimates of the Social Cost of Carbon: Concepts and Results from the DICE-2023R Model."

Quotes & Attributions

Note Paul Romer: "The economics profession went in the wrong direction for 30 years" — Slide 7 Slide 7

Context: Romer published "The Trouble With Macroeconomics" in 2016 (circulated as a working paper, published in The American Economist), which is highly critical of DSGE models. The quoted sentiment is fully consistent with his published views but the specific wording is not a verbatim quote from any identified source. No action required unless the deck claims it is a direct quote. The slide should use a paraphrase attribution ("paraphrasing Romer, 2016") as it does for Box.

Note Nicholas Stern: "economists got the thermodynamics wrong" — Slide 10 Slide 10

Context: The paraphrase is consistent with Stern's extensive published criticism of standard economic approaches to climate risk. No single source matches this wording. The attribution is appropriately labelled "paraphrasing Nicholas Stern" on the slide — no change required.

Note EU ETS carbon price: ~$72/t — Slides 24 & 33 Slides 24 & 33

Context: The EU ETS price fluctuates with the market and will change between deck versions. As of late 2023–early 2024 it traded near €55–65/t (≈ $60–72/t USD). This should be treated as a live figure and updated before each presentation. No structural correction needed — add a footnote "as of [date]" to the slide.

Priority Corrections

The following corrections are recommended in order of significance:

Priority Slide Action required Impact if left uncorrected
✓ Applied Slide 33 Change EPA SCC discount rate label from "2.5%" to "2.0%" — Applied May 11, 2026 Factual error on the slide whose entire argument is about discount rates — highly visible to expert audiences
✓ Applied Slide 20 Correct Amazon dieback CO₂ release figure from "90 GtCO₂" to "~150–290 GtCO₂" and add source — Applied May 11, 2026 3× understatement of a cited emissions figure; likely a GtC/GtCO₂ unit confusion
✓ Applied Slide 20 Revise permafrost "1.5°C feedback equivalent by 2100" to reflect modeled 0.1–0.3°C range; clarify long-run vs. 2100 distinction — Applied May 11, 2026 May mislead audiences about near-term permafrost feedback magnitude by ~5–15×
✓ Applied Slide 12 Update CMIP6 description from "9 countries" to "~20 countries / 33+ institutions" — Applied May 11, 2026 Routine factual error; may undermine credibility with CMIP6-familiar audiences
Medium Slide 19 Change Weitzman citation year from 2011 to 2009 (RESTAT) Wrong publication year; easily checked by any economist in the room
✓ Applied Slide 29 Clarify oil & gas committed emissions scope (220 Gt = currently-producing assets only, not all developed fields) — Applied May 11, 2026 May appear inconsistent with widely-cited higher estimates from the same sources
Low Slide 27 Add confidence level to carbon budget figure: "~250 GtCO₂ (~57% probability)" Precision issue only; the number is defensible as stated
Low Slide 23 Clarify 65M clean energy jobs is a 2050 IEA NZE projection, not a near-term figure Misleading timeframe comparison; no numerical error

Key Claims Verified Accurate

The following high-profile claims were cross-referenced and confirmed against primary sources:

Methodology

Each claim in the workshop deck was classified against the following criteria:

This review does not cover model assumptions within the CE Engine itself (abatement trajectories, technology cost curves, LCOE projections, etc.), which are documented separately in the CE Data Contracts and Model Patterns documentation.

CE Engine — Internal Document  ·  Workshop: Merging Models, 23 May 2026  ·  ce.drel.us