DRIVERS OF THE COUNTRY-LEVEL SOCIAL COST OF CARBON

Abstract

Climate change is expected to impose heterogeneous damages throughout the world due to varying levels of hazards, exposure, and vulnerability. The social cost of carbon (SCCO 2 ), the net present value of welfare losses from an additional ton of CO 2 , provides a useful synthesis of these damages globally, and marginal damages at the national level highlight the spatial pattern of these damages. This paper presents the PAGE-2025 integrated assessment model with 183 individual countries and five small country groups. We incorporate several critical improvements: a novel mapping of climate uncertainty to country-level warming patterns, an empirical approach to dynamic vulnerability using risk indices, updated abatement cost estimates using NGFS energy model scenarios, and an approximation of the effects of subnational heterogeneity in impacts. These changes result in a global SCCO 2 distribution with a median of $359 per ton of CO 2 . Damages are dominated by market damages, which are high in tropical regions and negative at high latitudes, and nonmarket damages, which are high in temperate regions that show the greatest rates of warming. We then decompose the spatial pattern of country-level SCCO 2 into the role of hazard, exposure, and vulnerability, by calibrating a simple emulator with temperature, population, and income, which captures the country-level SCCO 2 pattern with high precision. The spatial pattern roughly follows GDP, but with tropical countries at twice their GDP-only SCCO 2 level and cooler countries below this level. We use this decomposition to evaluate the role of different modeling decisions, including different scenarios, downscaling, damage assumptions, abatement costs, and assumptions about sub-national vulnerability. Across modeling decisions, the greatest changes in spatial pattern are driven by the use of SSP scenarios (rather than the default RFFSPs) and the updating of nonmarket damages damages to reflect Howard and Sterner ( 2017 ).