From the article:

The authors of a highly publicized study predicting climate change would cost $38 trillion a year by 2049 have retracted their paper following criticism of the data and methodology, including that the estimate is inflated.

“The economic commitment of climate change,” which appeared April 17, 2024, in Nature, looked at how changes in temperature and precipitation could affect economic growth. Forbes, the San Diego Union-Tribune and other outlets covered the paper, which has been accessed over 300,000 times. It has been cited 168 times, according to Clarivate’s Web of Science.

But after two commentaries published this August raised questions about the study’s data and methodology, the researchers revisited their findings. “The authors acknowledge that these changes are too substantial for a correction,” the retraction notice, published today, states.

[...]

Authors of the first “Matters Arising” commentary published August 6 noted the article projected the global gross domestic product would be lowered by 62 percent by 2100, “an impact roughly 3 times larger than similar previous estimates.” The authors of the critique also pointed out the PIK authors had used a dataset for Uzbekistan with “anomalies.” By removing the Uzbek dataset, the estimate in the original paper “aligns closely with previous literature,” the critique reads.

In a second Matters Arising, published a week later, Christof Schötz, a researcher at the Technical University of Munich in Ottobrunn, Germany, and a researcher at PIK, argued the analysis in the paper “underestimates uncertainty … rendering their results statistically insignificant when properly corrected.”

Here is Andrew Gelman's commentary:

It still seems crazy that the data from a single mid-sized country could have such a big effect of a global estimate. That’s something that the original researchers should’ve been aware of, and what it suggests to me is that there is a larger methodological problem that this didn’t get looked at automatically during the research process.