Fig. 2: Estimated damage from a marginal unit of past or future emissions.

a, Estimates of HD-CO₂, calculated as per-tonne cumulative impacts of a 1 Gt pulse of CO₂ emitted in a given year, from 1990 to 2020, under different fixed discount rates. b, Estimates of FD-CO₂, or the cumulative damages after 2020 of each of these pre-2020 emission pulses, under the same discount rates, assuming damages end in 2100. The post 2020 damage estimates for a pulse in 2020 are estimates of the SC-CO₂ in 2020. Numeric values are provided in Extended Data Table 1, and confidence intervals in Extended Data Fig. 5. Estimates account for the lagged effect of temperature on growth (Extended Data Figs. 3 and 4). c,d, Spatial distribution of HD-CO₂ and FD-CO₂ from a 1 tCO₂ emission in 1990, under a 2% discount rate, expressed as a percentage of total country GDP in 2020, showing impacts through 2020 (c) and projected impacts for 2021–2100 (d). Blue (red) shading indicates cumulative benefits (damages), whereas countries in grey have no data. Country borders are outlined if probability of damages or benefits exceeds 90%, accounting for both climate and econometric uncertainty. e, Estimate of the SC-CO₂ (here a 1 tCO₂ pulse in 2020) under three sets of analytic choices: impacts end in 2100, comparable to estimates in b; temperature has no effect on economic growth after 2100 but impacts cumulate through 2300 (see Supplementary Methods and Supplementary Fig. 1 for a schematic); and growth impacts continue through 2300. For each, SC-CO₂ is computed under two discounting schemes: a fixed 2% discount rate (red) and Ramsey discounting calibrated to a near-term rate of 2% (purple). Confidence bands account for both econometric and climate uncertainty (see Extended Data Fig. 5 for more details).