Fig. 3: Aleatory uncertainty risk premia with and without underlying epistemic uncertainty.

This plot shows what the canonical social planner would be willing to pay to avoid aleatory uncertainty when there is also underlying epistemic uncertainty. To compute these risk premia, we first obtain an ensemble of temperature trajectories by solving the deterministic EBM for a distribution of values of the equilibrium climate sensitivity (ECS). We assume that the ECS is log-normally distributed with a most likely value of 3 °C and Pr(2 ≤ ECS ≤ 4.5) = 0.66, in line with the IPCC’s fourth and fifth assessments, and otherwise uses the same physical assumptions as in Fig. 2. Next, we obtain a second ensemble by solving the stochastic energy balance model for the same distribution of ECS values. Both ensembles reflect the same epistemic uncertainty, but only the second incorporates aleatory uncertainty as well. The risk premia shown here are the difference between the expected utility of damages for the two ensembles (Supplementary Note 5). These risk premia can be decomposed into two parts: the darker portion of each bar shows the risk premium when all uncertainty is aleatory (same as in Fig. 2), while the lighter portion shows the additional risk premium arising from an interaction between aleatory and epistemic uncertainty. This risk interaction effect arises because a high draw from the ECS distribution produces both greater mean warming and greater variability, which makes the high draws disproportionately more costly. This results in damage distributions with a fatter right tail.