🌊 Would a "super El Niño," possibly the strongest on record, cost the world more than $5.6 trillion? A headline posing that question ran on Yahoo! News Japan on September 26. Trace the number back and you end up at a 2023 study estimating the cost of an El Niño from the 1990s. For that same event, the World Bank's figure was less than one-hundredth as large.

The headline number isn't in the article

The piece came from TBS CROSS DIG with Bloomberg, a business news site run by Japanese broadcaster TBS. The loss figure in the headline, about $5.6 trillion at the September 26, 2026 exchange rate, appears only there. The body never mentions it and never explains where it came from. It cites just two dollar amounts, both from a study in the journal Science: $4.1 trillion tied to the 1982–83 El Niño, and about $5.7 trillion tied to the 1997–98 one.

Only the second one lands near the headline figure; $4.1 trillion doesn't match. The most natural reading is that the headline was based on the $5.7 trillion.

If so, the headline number is not a forecast for this year's El Niño. It is a backward-looking estimate of what was lost in the five years after an El Niño nearly three decades ago. The article body describes it in the past tense, but the headline on its own reads like a projection for now.

Not one year of damage, but five years of slower growth

The study was written by Christopher Callahan and Justin Mankin of Dartmouth College and published in May 2023. El Niño is the periodic warming of surface waters in the eastern equatorial Pacific; it recurs every few years and reshuffles weather around the world.

Its cost has usually been tallied as damage in the year it hits: washed-out roads, failed crops. Callahan and Mankin did something different. They compared growth in per capita GDP (a country's output divided by its population) before and after El Niño events across countries from 1960 to 2019. Rebuilding eats money and time, and investment that would have fed future growth shrinks. The authors added up that missing growth year after year and counted it as loss.

According to Dartmouth's announcement, growth in the tropics lags for five years or more, and Callahan said the downturn could last as long as 14. In the US, GDP in 1988 and again in 2003 came in roughly 3% below where it would have been without those events, while Peru and Indonesia had GDP more than 10% lower in 2003. The paper also subtracts the rebound from the La Niña (El Niño's cool counterpart) that followed each event, and runs a version of its model that excludes currency-crisis years.

One event, a 100-fold gap

Counted this way, the 1997–98 El Niño cost $5.7 trillion in cumulative losses through 2003. The paper gives a 95% confidence interval (a range that should contain the true value 95% of the time) of $2.3 trillion to $9.2 trillion.

Earlier estimates for the same El Niño were in a different league.

Estimate Loss from the 1997–98 El Niño What it counted
Earlier study cited by the paper $36 billion Total costs
World Bank (as reported by AP) $45 billion Costs to governments
Callahan and Mankin $5.7 trillion Income lost through 2003

$5.7 trillion is more than 100 times $45 billion. The paper itself says its figure is two orders of magnitude above the earlier one.

The three estimates use the same word, "loss," for different things: the price of what broke, or the growth that never came afterward.

What economists doubt

A 2023 blog post from the Peterson Institute for International Economics (PIIE), a Washington think tank, named two economists who questioned the estimates: Matthew Kahn of the University of Southern California and David Ubilava of the University of Sydney.

Shortly after the paper came out, Kahn posted on X that the authors, who are not economists, had ignored the Lucas critique. That is a classic argument in economics: relationships drawn from past data may not hold in the future, because people and businesses learn and change their behavior. Replying in that thread, Ubilava said he doubted that an El Niño from five or more years back could still be weighing on growth. Ubilava himself co-authored a 2017 study finding that El Niño lowers growth in developing countries. His objection is to how long the effect lasts.

In the Associated Press coverage, University of Cambridge macroeconomist Kamiar Mohaddes said some countries actually grow faster during El Niño, that the damage faded quickly where it did hit, and that the estimates were too big.

Others took the opposite view. In the same AP article, Stanford economist Marshall Burke said the study makes a compelling case that growth slowed in places like Peru. Michael McPhaden, an oceanographer at the US National Oceanic and Atmospheric Administration (NOAA), said he had long thought El Niño damage estimates were far too low.

The estimators got their math wrong, too

On July 20, 2026, PIIE's Cullen Hendrix published an estimate for the current El Niño. If it matches 1997–98 in strength, he wrote, first-year losses in 2027 would be about $986 billion, and six-year losses about $17.6 trillion, with a range of $7.2 trillion to $28.5 trillion.

On August 18, 2026, the post was corrected. The six-year multiplier had been built on how big the hit was in the fifth year, rather than on the losses summed over all six. And the first-year figure had been computed from how many countries sat in each group, not from the combined size of their economies. The corrections pushed the estimates up to the figures above.

Owning the error in public was the right call. Even so, the top of the corrected range is nearly 4 times the bottom. The headline reads "almost $1 trillion or at least $7 trillion." One post contains $986 billion, $7.2 trillion, $17.6 trillion and $28.5 trillion, and which of those a reader picks up changes the story entirely.

The range never makes the headline

Headlines favor numbers that are big and precise. When two estimates differ by 100 times, the bigger one survives, and the confidence interval gets left behind. The paper doesn't hide its uncertainty. It projects $84 trillion in 21st-century losses, and states that the 95% range runs from $453 trillion in losses to $80 trillion in gains. The ranges drop off somewhere between the paper and the headline.

When a past estimate spreads as if it were a forecast, it risks eroding trust in climate reporting more broadly. If the number misses, suspicion tends to fall on the science rather than the headline.

None of this means climate change is being exaggerated. The paper went through peer review (evaluation by other researchers) before Science published it, and its method has a real rationale. Some researchers, McPhaden among them, think the older estimates were the ones that understated the damage, and this paper sits on that side. Hendrix, for his part, stresses that first-year losses are largely baked in but the losses that follow can be reduced.

The same pattern in self-driving math

Japan's government estimates that about $52 billion in public and private investment in self-driving technology will generate about $1.19 trillion in economic ripple effects. We took that number apart in another article. In both cases a few assumptions produce an enormous figure, and by the time it becomes a headline, the assumptions have fallen away.

Think of the last "trillion-dollar loss" or "trillion-dollar boost" you saw in the news where you live. Did the story say how many years it covered, or what exactly was being counted?

References