On July 23, Nature reported Dartmouth climate researcher Justin Mankin's estimate that this El Niño could cause "$10 trillion in economic losses by 2032." The figure is real, attributed and traceable to peer-reviewed work, which already puts it ahead of most numbers in circulation this month. It also measures something other than what "losses" implies, and it rests on a forecast peak that has not happened yet.
Most of the numbers attached to this event describe the ocean. This one describes the economy, which makes it harder to check and easier to repeat. It is worth checking anyway, because it is about to be everywhere.
The chain is short and clean. Berkeley Earth hosted a press briefing on July 23. Nature's James Dinneen wrote it up the same day. Gizmodo picked it up on July 27. By this week the figure was circulating without the sentence that qualified it.
The quote in Nature is conditional, and the condition is the whole story: "If the forecasts are anything near correct, the current El Niño could cause '$10 trillion in economic losses by 2032'." Mankin put the load-bearing clause first. Most of what has been written since has kept the number and dropped the clause.
Two things follow from the framing. It is a spoken estimate given to a reporter rather than a published analysis of this specific event, so there is no methods section to inspect for the 2026-27 case. And it is an extrapolation from prior peer-reviewed work, which does exist and can be inspected: Callahan and Mankin, Science 380, 1064-1069 (2023). Nature cites that paper as reference 1 and nothing else for the economics.
This is the substitution that matters most, and it is the one that headline writers make without noticing.
The Callahan and Mankin approach does not tally storm damage. It estimates how much smaller national economies are in the years after an El Niño than they would have been in a counterfactual world without the event, using country-level growth data and each country's exposure to ENSO teleconnections. The finding that made the paper notable was persistence: the drag does not clear when the Pacific cools. Nature's summary of the prior result is that the 1997-98 event "shrank global economic growth by US$5.7 trillion in the years that followed."
Foregone growth and physical damage are different quantities. They are not interchangeable, they cannot be added together, and they are not comparable to the insured-loss totals that reinsurers publish after a disaster season. A reader who encounters "$10 trillion in losses" and pictures flattened infrastructure has been given the wrong mental image, not by Mankin, but by the compression that happens between a research paper and a headline.
There is a second consequence. Because the estimate accumulates over years, the window matters. This one runs to 2032. It is not a 2026 bill, and it is not a 2027 bill.
A separate Peterson Institute analysis, also built from Callahan and Mankin's 2023 framework, estimates approximately $686 billion in contemporaneous losses and $3.1 trillion cumulatively over five years if the event resembles 1997-98. Its cumulative estimate is roughly one-third of Mankin's spoken $10 trillion estimate, although the horizons and event-strength assumptions are not identical.
The gap therefore illustrates how implementation, scaling and time horizon can substantially change the result, even when analysts begin with the same underlying research. That is the useful lesson, and it is invisible when one figure travels alone: two careful groups working from the same 2023 paper land a factor of three apart, before anyone disagrees about the science.
It would be easy to file this under hype. That would be wrong, and worth saying plainly.
Mankin attached a condition, named a horizon, and stands on published work that a reader can go and read. Nature cited the paper. Compare that with the physical claim traveling alongside it: Hausfather's 3.6°C is specifically a detrended Niño 3.4 anomaly, and by the time it reached general coverage it had become "peak temperatures in the tropical Pacific Ocean" rising "about 6.5 degrees Fahrenheit above average," a phrase that describes a much larger region and a different quantity. We have written before about how the same week in the same box yields different numbers in different datasets, and about how fast a CPC bulletin becomes a headline.
The problem with the $10 trillion figure is not that it is wrong. It is that it is being quoted without the method that gives it meaning, and a growth projection stripped of its method reads like a damage estimate.
The economics rides on the physics, so the physics is where the near-term test sits. NOAA's July 9 diagnostic discussion put an 81 percent chance on a very strong El Niño during October to December and a 97 percent chance that El Niño conditions persist through early spring 2027. Michelle L'Heureux, who leads NOAA's ENSO prediction team, was quoted in the same Nature article saying forecasters are "seeing all the markers of very vigorous atmosphere-ocean coupling across the equatorial Pacific Ocean."
The relative weekly Niño 3.4 index, which is the version CPC reports in its ENSO products and the one this site tracks, reached +1.4°C in the week centered July 22. NOAA's conventional weekly table, which does not net out the tropical mean, put the same box at +2.2°C for the same week. Neither weekly value is directly comparable to CPC's seasonal RONI strength probabilities or to Hausfather's detrended forecast peak, but both confirm that the oceanic warming is already substantial.
Four different numbers are in circulation for what is loosely called "how strong this El Niño is," and they are not interchangeable. Keeping them apart is the difference between reading the coverage and being misled by it.
| Value | What it is | Dataset and averaging |
|---|---|---|
| +2.2°C | Conventional weekly Niño 3.4, week centered July 22 | OISSTv2.1, weekly, no tropical mean removed |
| +1.4°C | Relative weekly Niño 3.4, same week | OISSTv2.1, weekly, tropical mean removed |
| +0.5°C | Official seasonal RONI, April to June 2026 | ERSSTv5, three-month running mean |
| 3.6°C | Forecast peak, not an observation | Detrended monthly Niño 3.4, multi-model median |
The first two rows deserve a note, because they are the same box in the same week from the same dataset, and they differ by 0.8°C. That gap is not a revision or an error. It is the tropical-mean anomaly being netted out. Across all four Niño regions for that week the conventional and relative readings differ by a nearly uniform 0.7 to 0.8°C, which is the signature of subtracting a basin-wide warm background rather than of any change to the underlying observations. Since February 2026, CPC's ENSO products report the relative version, while the conventional weekly table it has always published continues to carry the unadjusted numbers. Both are live, both are correct, and quoting one against the other is how a reader ends up thinking the ocean warmed 0.8°C in a week.
The weekly and seasonal figures are measuring the same ocean on different clocks and against different baselines, which is why a weekly reading above +2°C sits alongside a seasonal RONI at the +0.5°C threshold without either being wrong. We have written at length about why the same week in the same box yields different numbers in different datasets and about why NOAA changed the index it monitors with.
The 81 percent is a strong statement from the agency, and it is the foundation the $10 trillion sits on. It is also a probability, not a result.
For the physical side of the same event, see our note on the widening east-west gradient, and on the consumer end, what El Niño does and does not do to a grocery bill. For the 1.7°C figure also discussed at the briefing, see our earlier piece on monthly temperature spikes against the Paris threshold.