Synthesis of Evidence Yields High Social Cost of Carbon Due to Structural Model Variation and Uncertainties

Academic Background

Climate change is one of the most severe challenges facing the world today, with profound impacts on the economy, society, and the environment. To quantify the economic costs of climate change, the academic community has proposed the concept of the “Social Cost of Carbon” (SCC), which represents the total cost to society of emitting one additional ton of carbon dioxide. The SCC is a key metric for evaluating the benefits of emission reduction policies and is widely used in climate and energy policy analysis. However, estimating the SCC involves complex climate system models and economic impact models, with significant uncertainties and controversies.

Despite the increasing number of studies on the SCC in recent years, these studies are often fragmented and lack systematic integration, making it unclear how different model structures influence SCC estimates. To address this gap, Frances C. Moore and colleagues published a research paper titled Synthesis of Evidence Yields High Social Cost of Carbon Due to Structural Model Variation and Uncertainties in PNAS, aiming to comprehensively assess SCC estimates and their drivers by synthesizing existing literature and expert surveys.

Source of the Paper

The paper was co-authored by Frances C. Moore, Moritz A. Drupp, James Rising, Simon Dietz, Ivan Rudik, and Gernot Wagner, affiliated with the University of California, Davis; the University of Hamburg; ETH Zurich; the London School of Economics and Political Science; and other institutions. The paper was published in PNAS on December 17, 2024.

Research Process

1. Literature Review and Data Collection

The research team first conducted a systematic review of 147 SCC-related papers published between 2000 and 2020, collecting 1,823 SCC estimates. These estimates covered different model structures, parameter settings, and sources of uncertainty. The team also recorded relevant variables for each estimate, such as emission year, discount rate, damage function, and economic scenarios.

2. Expert Survey

To contextualize the SCC estimates from the literature, the research team surveyed 176 authors of SCC papers, gathering their perspectives on SCC estimates. The survey results showed that experts generally believe the SCC estimates in the literature are too low, primarily due to incomplete model structures, insufficient damage characterization, and the use of high discount rates.

3. Random Forest Model

To address the biases in the literature, the research team used a Random Forest Model to reweight the SCC estimates from the literature, aligning them more closely with expert assessments of model structure and discount rates. The Random Forest Model generated a synthetic SCC distribution by training on the SCC estimates and their explanatory variables from the literature.

Key Findings

1. SCC Distribution in the Literature

The SCC estimates in the literature are widely distributed and exhibit a significant right skew, with a mean (truncated) SCC of $132 per ton of CO₂ for 2020 and a median of $39. Analysis of variance (ANOVA) revealed that persistent damages, Earth system representation, and distributional weighting in the model structure significantly influence SCC estimates.

2. Expert Survey Results

The expert survey indicated that experts believe the SCC estimates in the literature are too low, primarily due to incomplete model structures, insufficient damage characterization, and the use of high discount rates. The experts’ average SCC estimate was $142 per ton of CO₂, more than double the literature estimate.

3. Synthetic SCC Distribution

The synthetic SCC distribution generated by the Random Forest Model showed a mean SCC of $283 per ton of CO₂ for 2020 (5% to 95% range: $32 to $874), higher than most official government estimates, including the 2023 update from the U.S. Environmental Protection Agency (EPA).

Conclusions and Implications

By synthesizing the literature and expert surveys, this study reveals the high SCC estimates and their drivers. The results suggest that SCC estimates in the literature are generally too low, primarily due to incomplete model structures, insufficient damage characterization, and the use of high discount rates. The synthetic SCC distribution generated by the Random Forest Model aligns more closely with expert assessments, providing a more accurate reference for climate policy analysis.

Research Highlights

1. Comprehensiveness: The study synthesizes 1,823 SCC estimates from 147 papers, combined with expert surveys, offering the most comprehensive SCC assessment to date.
2. Innovation: The research team used a Random Forest Model to reweight SCC estimates from the literature, generating a synthetic SCC distribution that aligns more closely with expert assessments.
3. Policy Implications: The findings suggest that current SCC estimates may underestimate the true costs of climate change, providing policymakers with a more accurate reference.

Additional Valuable Information

The research team has made detailed data and code available for other researchers to replicate and extend the study. All data and code can be accessed on the Zenodo platform.

Through this study, we have gained a better understanding of SCC estimates and their drivers, while also providing a more robust foundation for future climate policy analysis.