Troubling Timeline of NGFS’s Most Recent Climate Estimate

In early November 2024, the Network for the Greening of the Financial System (NGFS), a consortium of central banks and other regulatory supervisors across the globe, announced that a new climate damage function had been incorporated into its climate scenario toolkit. Based on an academic article in Nature, the new climate damage function implies dramatically higher economic losses from worsening climate conditions: a 19% loss of global real income by 2050 and a 60% loss by 2100. These revised estimates of economic collapse are unprecedented in size. The very large economic damage opens the door to more binding bank transition targets, penalties on banks for inadequate compliance with climate change risk expectations, such as fines on banks from the ECB, and potentially additional bank capital requirements for climate risks.

The following chronology raises troubling questions about the integrity of the process at NGFS and Nature that produced that estimate.

Chronology

Sometime in 2023 or 2024: “The economic commitment of climate change” by Kotz et al. enters peer review at Nature. The paper encounters skepticism that the results will hold up to scrutiny, as one of the three referees (Referee #2) states: “Further the robustness of the results to alternative specifications in the main text seems to be – ‘other literature suggests the work is robust.’ But we also know from previous work that robustness in one specification will not carry to another. So, the reluctance to demonstrate that the results are robust to alternative specifications could be a concern.”

Another referee is brought in, who opines that “purely subjectively, I have a hard time in believing the results, which seem unintuitively large given damages aren’t perfectly persistent. In many ways this is an unfair critique because it is just an intuition. Yet, it is worth probing further the possible sources of bias, because we know from the experience with Burke, Hsiang and Miguel (2015 in Nature) that publishing numbers in high-impact journals, which subsequently essentially get discredited, can create a lot of confusion. This is what R2 is getting at, and yet in my view R2 hasn’t found a ‘smoking gun’.” Thus, the new referee recommends publication.

April 17, 2024: “The economic commitment of climate change” by Kotz et al. is published in Nature. The article receives comprehensive media coverage.

May 8, 2024: “Spatial correlation in economic analysis of climate change” by Christof Schotz is submitted to Nature. The paper raises serious methodological problems with the key statistical claims in the Kotz paper. Correcting for those errors, the Schotz paper demonstrates that economic damage is effectively zero rather than the large values reported in the original paper.

The existence of this paper is not reported to the public.

Sept. 16, 2024: “Data anomalies and the economic commitment of climate change” by Bearpark et al is submitted to Nature. The paper shows that a serious data error in the Kotz paper had inflated the economic damage estimates by 200%. The paper also demonstrates that the original paper had underestimated the uncertainty of its results.

The existence of this paper is not reported to the public.

Oct. 22, 2024: Three officials at the ECB who work with the NGFS publish “Economic losses from climate change are probably larger than you think: New NGFS scenarios.” The article states that the model from the Kotz paper has been incorporated into NGFS scenarios.^[1] The article includes no reference to the two papers raising concerns about the paper.

Nov. 5, 2024: The NGFS formally announces incorporation of the new damage function.^[2]

The NGFS publishes a 50-page technical document on the new damage function, explaining and defending it: “Damage functions, NGFS scenarios, and the economic commitment of climate change: an explanatory note.”^[3]

On or about Nov. 8, 2024: The NGFS hosts a launch event for the new NGFS scenarios and damage function that includes one of the authors of the Nature paper and another member of the Potsdam Institute explaining their contributions to the scenario toolkit.^[4]

Dec. 9, 2024: BPI publishes the first publicly available study on the Nature paper.^[5] Like the Schotz paper, which was not public at that point, the BPI study finds serious statistical issues with the lag length selection procedure as well as the statistical significance of the coefficients, for different but similar reasons than those advanced by Schotz. The BPI study similarly concludes that the economic damage should have been very small or zero. The BPI study calls for the NGFS to disavow the model.

Sometime in late December 2024: Nature posts a warning that the methodology and data of the Nature paper are in question (https://www.nature.com/articles/s41586-024-07219-0). However, the warning is backdated to Nov. 6, 2024, making it appear as if the warning was posted prior to the BPI study, when in fact it came after. (See appendix for how the backdating was discovered.)

Feb. 27, 2025: The author of BPI study emails Leonie Wenz, the corresponding author of the Nature study, highlighting the BPI analysis of the paper’s statistical problems. There is no reply.

March 3, 2025: The author of the BPI study follows Nature’s published paper critique procedure and emails Nature editors, formally informing them of the criticisms in the BPI study, and that the lead author had been informed and failed to reply. The BPI author (still unaware of the existence of two independent analyses raising concerns) asks Nature to confirm that the issues raised by the BPI study will be addressed and also offers to submit a formal academic report if needed. The Nature editors never reply.

Aug. 6, 2025: The Bearpark et al. paper is published by Nature. The authors of the Kotz et al. Nature paper simultaneously post a revised version in which, after fixing the data problem and changing some features of the model, they adjust their model to reach essentially the same large economic damage results.

Aug. 13, 2025: The Schotz paper is published by Nature. Simultaneously, the authors of the Kotz et al. Nature paper release another update, again maintaining essentially the same large economic damage results as the original paper.

Aug. 14, 2025: A second BPI study is published. The BPI study points out the serious failures in process and procedure at the NGFS: In a well-functioning process, the NGFS should have known of the data and methodological problems in the Nature paper that had been reported well before they finalized the damage function in their climate toolkit. The new BPI study also does a preliminary model review of the revised Nature paper, finding that the statistical problems had not been corrected.^[6] The NGFS continues to be silent on the issue.

On or around Aug. 28, 2025: About eight months after Nature posted a public warning that the methodology and data of the paper were in question, and nine months after the BPI study explained the statistical problems with the paper, the NGFS finally puts up a warning on the climate damage function for its users on its scenario website. However, it informs banks and other users that neither the NGFS, its member institutions, nor anyone acting on its behalf is responsible for any reliance on its scenarios or documentation, according to the NGFS licensing agreement. The NGFS takes no responsibility for the quality of its scenario toolkit. The NGFS states that it will wait for Nature’s decision before it decides what it will do next, effectively denying any independent responsibility for ensuring that the climate models that it uses to regulate banks are sound.^[7] Meanwhile, the original documentation of the NGFS damage function remains on the NGFS site, unchanged.

Sept. 4, 2025: The Potsdam Institute puts up an FAQ explaining the process for the paper going forward, suggesting that the revised paper will undergo a new peer review. But the paper has not been withdrawn in the meantime.^[8]

Appendix:

Here is how the warning appears on Sept. 15, 2025.

The wayback machine, available at https://web.archive.org/, takes periodic snapshots of webpages so that users can see how they have been changed or revised. If we put the internet address of the website for the Nature article into the wayback machine, we get the following output for the final months of 2024.

The Nature warning is dated as having been posted on Nov. 6, 2024. However, if we click the Dec. 26 date, circled in red, we can get a screenshot of what the paper website looked like on that date, shown below. As is apparent, the warning was not posted on Dec. 26, 2024, but was rather posted sometime after that, but backdated to Nov. 6, 2024.