Climate Change May Cost 6x More Than the Nobel-Winning Estimate
Adrien Bilal, France’s best young economist of 2026, swapped one variable in the standard climate damage model. The estimated cost of warming jumped from 1-3% to over 20% of global GDP per degree.
In 2018, William Nordhaus received the Nobel Prize for showing that 1°C of warming would cost the global economy somewhere between 1 and 3% of GDP. His DICE model, the workhorse of climate economics for three decades, put the Social Cost of Carbon at roughly $31 per ton of CO₂. Those numbers shaped every major climate policy framework from the Paris Agreement to the Inflation Reduction Act. They were reassuring. Manageable. Wrong.
A word on DICE, because understanding the model is essential to understanding what Bilal overturns. DICE (Dynamic Integrated model of Climate and the Economy) is built on the Ramsey growth model, the foundational framework of modern macroeconomics (published in 1928, still taught in every graduate programme worldwide) in which society trades off consumption today against investment for tomorrow. The model determines the optimal savings rate across generations: how much should we consume now, and how much should we set aside to build productive capacity for the future? Nordhaus’s innovation, in the 1990s, was to graft a climate layer onto this framework: economic activity generates CO₂ emissions, which accumulate in the atmosphere, which raises temperature through a simplified two-box climate module, which in turn reduces economic output through a “damage function.” That damage function is the crux. In DICE, damages are modelled as a quadratic function of temperature: output falls by a percentage proportional to the square of the temperature increase. The model then optimises over time, finding the carbon tax path that maximises welfare across generations (Barrage and Nordhaus, 2024). Its elegance made it canonical. Its weakness was the calibration of that damage function, which relied on empirical estimates derived from local temperature variation. The method was rigorous, but it was measuring climate change through a keyhole: capturing what happens when a given country is a bit warmer in a given year, while missing what happens when the entire planet warms simultaneously.
Adrien Bilal, a 35-year-old French economist at Stanford (X2010, Princeton PhD, previously at Harvard), has just been awarded the 2026 Prix du meilleur jeune économiste by Le Monde and the Cercle des économistes. His central paper, co-authored with Northwestern’s Diego Känzig and forthcoming at the Quarterly Journal of Economics (Bilal and Känzig, 2024), finds that 1°C of global warming reduces world GDP by over 20% in the long run. The implied Social Cost of Carbon exceeds $1,200 per ton. A business-as-usual trajectory to +3°C by 2100 would destroy roughly half of global output.
Six to twenty times larger than the Nordhaus damage estimate: over 20% of GDP per degree of warming, against the 1-3% that earned a Nobel. (Bilal laid out the reasoning with remarkable clarity in a recent interview on France Culture’s Entendez-vous l’éco?, well worth 40 minutes of your time.)
One Variable, One Blindspot
Behind what turns out to be fairly involved econometrics, the core idea is almost embarrassingly simple. For two decades, the empirical literature on climate damages has relied on country-level temperature variation. A given country is a bit warmer than usual in a given year; researchers measure the effect on that country’s GDP. This panel approach, pioneered by Dell, Jones and Olken (2012) and refined by Burke, Hsiang and Miguel (2015), is econometrically clean. Country fixed effects (statistical controls that absorb all permanent differences between nations: geography, institutions, culture, baseline climate) eliminate confounders.
But it captures the wrong thing.
When France happens to be 0.5°C warmer than average, that says very little about extreme weather. Wind speed does not change. Precipitation patterns hold. The heatwave risk barely moves. Country-level temperature fluctuations are mostly noise from the perspective of the climate system. And adding up all those country-level fluctuations does not reconstitute the effect of global warming, because what matters is not that individual countries are locally warmer but that the planet as a whole is warming, which triggers entirely different physical mechanisms.
Global mean temperature operates through the oceans, which cover two-thirds of the Earth’s surface and which the country-level approach simply cannot see. When the planet warms, oceans absorb heat. Atmospheric humidity rises. Precipitation regimes shift. Cyclone intensity increases. Heatwaves become more frequent, longer, and deadlier. These are systemic reorganisations of the Earth’s energy balance, and they are precisely the mechanisms through which climate change destroys economic value. The previous literature’s reliance on country-level land temperature could not capture what happens when that water warms, which is why the estimates were so small.
Bilal and Känzig’s move is to replace local temperature with global temperature as the explanatory variable. They exploit natural variability in global mean temperature (driven by phenomena like El Niño and solar cycles) and use a local projection approach, a flexible econometric method developed by Òscar Jordà (2005) that traces the cumulative response of GDP at different time horizons (1 year, 5 years, 10 years out) to a single temperature shock, without imposing the rigid dynamic structure of a traditional vector autoregression. This allows them to map the full trajectory of damage across 170+ countries from 1960 to 2019.
The finding: global temperature shocks predict extreme heat events, extreme wind speeds, and extreme precipitation far more strongly than local temperature shocks do. And through those channels, they produce GDP loss estimates an order of magnitude larger than anything the existing panel literature had found.
Two Channels of Destruction
The paper decomposes the damage into two structural channels within a neoclassical growth model (the standard framework in which output is produced from capital and labour, and the economy grows by accumulating capital and improving technology over time).
The first is productivity. When global temperature rises, total factor productivity (TFP, the residual efficiency with which an economy converts inputs into output) falls. Workers produce less. Outdoor labour becomes dangerous. Supply chains slow. Power grids strain. Cognitive performance degrades. This channel is familiar from the existing literature, but Bilal and Känzig find it operates at a much larger scale when driven by global rather than local temperature variation.
The second channel is capital destruction. Floods erode infrastructure. Storms wreck power lines, roads, and ports. Wildfires consume physical assets. This depreciation channel was largely absent from the Nordhaus framework, which focused almost exclusively on productivity losses. Bilal and Känzig show that capital destruction accounts for a substantial share of the total damage, and that it compounds: destroyed capital is not instantly rebuilt, and the rebuilding itself diverts resources from productive investment.
The compounding matters. Because warming is persistent and damages accumulate through both channels, the long-run effects dwarf the short-run impacts. A 1°C shock reduces GDP by roughly 12% at peak in the medium run, but the cumulative long-run effect exceeds 20%. The damage function is not just steeper than Nordhaus assumed. It is a different shape.
The Policy Reversal
Every cost-benefit calculation that climate policy rests on is affected.
Start with the Social Cost of Carbon (SCC), the dollar value of all present and future economic damages caused by emitting one additional ton of CO₂, and in theory the number that should determine the optimal carbon tax. Nordhaus’s DICE-2023 model, co-authored with Lint Barrage, already represented a significant upward revision from earlier vintages, landing in the range of $60-80 per ton of CO₂. Bilal and Känzig’s global-temperature-based estimate exceeds $1,200 per ton. At that level, virtually every decarbonisation intervention currently on the table is a bargain. For context, the EU Emissions Trading System prices carbon at roughly €65-75 per ton today, and the most ambitious IRA decarbonisation interventions cost approximately $80 per ton of CO₂ abated. Those figures sit at roughly 5-6% of what Bilal and Känzig estimate the true global damage to be.
But the more consequential result concerns unilateral action. In a follow-up paper published in the AEA Papers and Proceedings (Bilal and Känzig, 2025), Bilal and Känzig construct the “Domestic Cost of Carbon”: the economic losses within a single country from emitting one ton of CO₂. For the United States: $226 per ton. For the European Union: $216 per ton. Both exceed the marginal cost of decarbonising over 80% of economic activity using current technologies.
This reverses the game-theoretic logic that has paralysed climate policy for three decades. The Nordhaus framework implied that decarbonisation was a global public good: each country bore the full cost but captured only a fraction of the benefit. The rational strategy was to free-ride. Bilal and Känzig show that for large economies, the domestic benefits of decarbonisation alone exceed the domestic costs. The US and the EU do not need a global agreement to justify aggressive climate action. The purely selfish economics already work.
As Bilal put it on LinkedIn after the prize: large economies have an interest in decarbonising, even if they are the only ones doing it. One sentence, and climate policy shifts from a coordination failure to a straightforward capital allocation question.
What This Means for Capital Allocation
If the true cost of carbon is 6x higher than the models underpinning current regulation, then every asset exposed to climate risk is mispriced, and every asset positioned for decarbonisation is undervalued.
Under the old Nordhaus SCC ($31/ton), IRA-style decarbonisation investments ($80/ton abated) were barely justifiable, requiring governments to internalise benefits to the entire world. Under Bilal and Känzig’s domestic cost ($226/ton for the US), the same investments return nearly 3x their cost to the country that makes them. Under the full global SCC ($1,200+/ton), the return is closer to 15x.
Grid-scale storage, industrial heat pumps, direct air capture, advanced nuclear, electrified transport: if these numbers hold, these are not wagers on future regulation tightening but responses to the economic damage we can now measure, backed by estimates that are, if anything, conservative (the paper does not yet incorporate tipping points, non-linear feedback loops, or biodiversity collapse). The numbers are large enough to reshape discount rates, hurdle rates, and portfolio construction for anyone allocating to energy transition, climate adaptation, or resilience infrastructure.
Open Questions
Three limitations deserve attention, two of which point in the same uncomfortable direction.
First, the time-series approach exploits natural temperature variability over 60 years. Bilal and Känzig’s estimates come from observing what happens to GDP when global temperature naturally swings up by a fraction of a degree in a given year (an El Niño event, a volcanic lull) and extrapolating that response to permanent warming. The open question is whether that extrapolation holds. Maybe economies gradually adjust to slow, predictable warming in ways they cannot during sudden shocks: new building codes, crop switching, air conditioning adoption. That would make the 20% figure an overestimate. But the reverse is equally plausible, and arguably more likely: permanent warming may break systems that recover perfectly well from temporary shocks. Ecosystems do not bounce back from a warm century the way they bounce back from a warm year. Infrastructure designed for a climate that no longer exists degrades continuously. Agricultural zones shift faster than farming communities can follow. The relationship between short-term shocks and long-term trends could cut both ways, but the asymmetry is worth dwelling on: if persistent warming triggers compound effects that temporary fluctuations cannot capture (infrastructure decay, agricultural system collapse, mass migration), then 20% is too optimistic. The climate science literature leans towards the latter.
Second, reinforcing that concern, the estimates do not incorporate tipping points: the collapse of the Atlantic Meridional Overturning Circulation (the deep-ocean conveyor belt that keeps Western Europe temperate), permafrost methane release, Amazon dieback. These are not marginal additions to a smooth warming curve. They are discontinuities: self-reinforcing processes that, once triggered, accelerate warming and its consequences independently of human emissions. Permafrost alone holds an estimated 1,500 gigatons of organic carbon; if even a fraction enters the atmosphere as methane, it dwarfs anything in the emissions scenarios Bilal and Känzig model. If these thresholds are crossed, the 20%+ GDP loss per degree may itself prove conservative. Put differently: this paper, which already multiplies the consensus estimate by six, may still represent a lower bound. That should give pause to anyone who has internalised the Nordhaus numbers as a worst case.
Third, the unilateral decarbonisation result depends on the size of the economy relative to global emissions. It holds for the US and the EU, whose emissions are large enough that their own decarbonisation measurably reduces the global temperature trajectory, and whose economies are large enough that even a fractional reduction in warming translates into substantial domestic savings. The EU is a useful test case here: carbon pricing and emissions targets are set at the Union level (the EU ETS, the Fit for 55 package), but much of the actual decarbonisation effort, from building renovation to transport electrification to grid planning, remains a member-state competence. Bilal and Känzig’s result makes the economic case for the EU-wide framework, but the gap between what Brussels legislates and what member states actually deliver on the ground is where much of that value leaks away. Unlike water pollution or soil contamination, where the damage is local and the incentive to act is immediate, carbon emissions disperse globally: a ton of CO₂ emitted in Warsaw warms Lisbon just as much as it warms Warsaw. That global dispersion is precisely why coordination matters, and why the result does not hold for smaller economies outside such blocs: a country responsible for 0.1% of global emissions cannot move the global temperature needle by cleaning up its own act, however virtuous. For those countries, the free-rider problem holds, and there is no substitute for coordinated international action.
Känzig told the NZZ that when they first saw the results, they were shocked. They checked the model repeatedly. But the numbers held. And the climate scientists they consulted were not surprised at all. The economists had simply been reading the wrong thermometer.
