If cutting emissions pays for itself, why is it so hard to do?

Why the question matters

Climate-economy models have long found that restricting emissions immediately would be in humanity's economic interest quite apart from any environmental case. That such policies remain difficult to enact suggests the obstacle is not a failure to recognise net benefit.

Once the timing is made explicit the puzzle largely dissolves, and the problem is revealed to be one of distribution between generations rather than of efficiency. That changes which instruments are likely to work — measures paying back within a political lifetime are not merely nice to have, they are what makes action self-sustaining.

The distributional question also runs between countries, not only between generations, and the arithmetic of who actually emits turns out to be more forgiving than the framing of climate and development as rival goals suggests.

What our research finds

• Adaptation returns on investment far sooner than abatement, with abatement's timescale set by discount rates and adaptation's by capital depreciation, so investing in both yields greater near-term and long-term net benefit (Duan et al., 2025).
• Lower-income countries face the largest increases in high-temperature days endured without air conditioning, and would need GDP growth above their historical rates simply to prevent that number from rising (Wongel et al., 2025).
• A technology's past learning rate barely predicts its future one — a correlation of 0.12 across 87 technologies — so cost projections built by extrapolating an observed learning rate carry far more uncertainty than they report (Carlino et al., 2025).
• Reductions in the Green Premium can be translated into equivalent reductions in carbon emissions, making cost-saving innovation and direct abatement comparable in a single unit (Caldeira et al., 2023).
• After Pakistan's 2022 Indus floods, elevated rebuilding would have protected 13 per cent of those affected and relocation 16 per cent, while 70 per cent faced water shallow enough for low-cost measures — at 26 to 63 per cent added reconstruction cost (Schmitt et al., 2023).
• An economically optimal emissions policy does not begin producing net global economic benefits until its break-even year, which arrives long after the generation enacting it has paid the costs — on these results, mostly more than half a century away (Brown et al., 2020).
• Delaying decarbonization in poorer countries until per-capita GDP exceeds $10,000 adds less than 0.3 °C of warming, even though over half the world's population lives below that threshold (Duan et al., 2020).
• Holding all other factors constant, today's relationship between climate and population density implies that climate change may within decades give hundreds of millions of people additional incentive to migrate, largely from warm tropical and subtropical countries towards cooler temperate ones, with India the country with the greatest number of people affected (Chen and Caldeira, 2020).
• Anthropogenic aerosol emissions were cooling the Earth by 0.72 degrees Celsius in 2010 relative to a world without them, and because that cooling benefits hot low-income economies while harming cold high-income ones, it reduced the ratio of per capita GDP between the richest and poorest population-weighted deciles by about 1.0% (Zheng et al., 2020).
• A 1 per cent rise in energy's share of costs induces about a 1.2 per cent efficiency gain over twenty years, making carbon prices save up to 30 per cent more energy by 2120 than models omitting the mechanism predict (Wang et al., 2019).
• The global social cost of carbon has a median estimate of 417 US dollars per tonne of carbon dioxide, of which India alone bears about 21%, at 86 dollars per tonne — the largest share of any country, ahead of the United States at 48 dollars and Saudi Arabia at 47 dollars (Ricke et al., 2018).
• A carbon tax that succeeds collects nothing, so a revenue-dependent government acquires a reason to keep emissions going — an incentive that turns perverse around 2085 in the DICE model (Wang et al., 2017).
• For equal spending and equal immediate cost reduction, research that shifts a technology's learning curve beats research that merely moves along it, most of all for expensive technologies that learn slowly (Shayegh et al., 2017).
• Optimal coastal investment depends on the rate of sea-level rise rather than any particular amount, so adaptation designed for a climate state rather than an ongoing trend misestimates both damages and adaptive capacity (Shayegh et al., 2016).
• The value of temporary carbon storage can be derived from a carbon price path and a discount rate rather than stipulated by a ton-year horizon, and the same reservoir is then worth 97 per cent of permanent storage or nothing at all depending on that path (Herzog et al., 2003).