This climate study made a big error. One piece of data was to blame.

A year ago, a paper published in the journal Nature made a sweeping claim: The world economy was already on track to lose 19% of global gross domestic product by 2050, compared with what it would have been without climate change. By 2100, under a high emissions scenario, it predicted that global GDP would be roughly 62% lower than without climate change.

Those numbers were a whopping three times higher than previous estimates — sparking concern that climate change would hinder the global economy much more than expected. The paper made waves across social media; according to one analysis by the U.K.-based outlet CarbonBrief, it was the second-most cited paper in the media in 2024. The paper’s analysis and dataset, meanwhile, have been used for financial planning by the U.S. government, World Bank and other institutions.

There was just one problem, according to a new analysis: The paper’s findings were flawed.

A new commentary published Wednesday in Nature found that the massive damages predicted by the paper were predicated on data errors stemming from one country — Uzbekistan.

With Uzbekistan removed from the dataset, the predictions dropped substantially — from 62% GDP loss in 2100 to 23% and from 19% by 2050 to 6%, said Solomon Hsiang, director of the global policy laboratory at Stanford University and one of the authors of the new commentary.

“Everybody who works with data has some responsibility to look at the data and make sure it’s fit for purpose,” Hsiang said.

The authors of the original paper, however, argue that their analysis still holds.

Karl Ziemelis, chief applied and physical sciences editor at Nature, wrote in an email that the journal was reviewing the study, and “appropriate editorial action would be taken once the matter was resolved.”

“Science has worked, and always will work, through a process of constant interrogation and review, whether that be during the course of research, in peer review or in post publication assessment,” Ziemelis added.

Hsiang and his co-authors, graduate students Tom Bearpark and Dylan Hogan, discovered the error by removing one country at a time from the dataset. Every other country they removed only slightly changed the GDP predictions. But when Uzbekistan was taken out, the results changed dramatically.

They then looked closer at the Uzbekistan data. The paper’s dataset showed the country’s GDP plummeting dramatically in the year 2000, losing almost 90%. Then in 2010, it showed the GDP climbing in some regions by over 90%. Other years also showed wild oscillations.

According to the World Bank, the country’s growth over the past 40 years has actually been quite modest — ranging from a 0.2% loss to a 7.7% growth.

Those swings were so dramatic in the initial paper’s data that they dominated the underlying model, which connected temperature and precipitation changes with economic growth. That resulted in a model that showed GDP would take sharp hits from climate change.

Hsiang found the results surprising. “When you have a lot of data points, the idea that a small country could be so influential is not intuitive,” Hsiang said. That’s why, he said, it’s essential to rigorously test the results and the data.

The authors of the original paper, scientists at the Potsdam Institute for Climate Impact Research in Germany, have a different take. The error in the Uzbekistan data, they said, was due to problems with how the original data was processed. In an additional analysis, they corrected the Uzbekistan data, and also changed how their model controlled for underlying economic trends.

“We find actually that by doing that, our estimates in general get more robust,” said Maximilian Kotz, a postdoctoral researcher at the Potsdam Institute.

Using this altered method, they found results that agreed closely with their original findings. Instead of 19% damages by 2050, for example, the new analysis shows 17%.

“We are grateful, and I think it’s a good part of the scientific process that they’ve pointed out these issues,” said Leonie Wenz, professor of environmental economics at the Technical University of Berlin and another author of the initial study. “But importantly, the main conclusions of the paper hold, and there are only slight changes to the estimates.”

That change, however, required modifying the methodology of the paper, and critics are still skeptical.

“Science doesn’t work by changing the setup of an experiment to get the answer you want,” Hsiang said. “This approach is antithetical to the scientific method.”

Concerns about the large magnitude of the GDP changes were already clear in the peer review process. In a peer review document, posted publicly by Nature, one anonymous reviewer wrote, “I find all of this well explained and fairly convincing, yet, purely subjectively, I have a hard time in believing the results, which seem unintuitively large given damages aren’t perfectly persistent.”

After some back-and-forth with the authors, the reviewer later approved of the paper going to publication.

Hsiang says the study is an important example of how science corrects itself, and that “The fact that there’s one car accident due to driver error doesn’t mean that we think cars are fundamentally dysfunctional.”

“If people are at all skeptical about how science functions — the answer is well, our team discovered this issue, and we believe transparency is super important,” he said. “That’s the ethos of science.”