Event Debrief: Social Cost of Carbon Could be Significantly Underestimated, Says Environmental Economist Frances Moore

Key Takeaways

• The social cost of carbon (SCC) is an important accounting exercise to estimate of the future costs to society caused by anthropogenic climate change driven by carbon emissions.
• There is significant variation in published studies on the SCC and experts expect that the true SCC is likely a lot higher than what is currently reported in the literature. Accounting for this variation and for recent modeling and parametric improvements yields a higher expected SCC.
• A recently updated EPA study raised previous estimates from $50/ton to closer to $210/ton, but may still be underestimating the true SCC.

On November 13, Harvard Kennedy School hosted the environmental economist Frances Moore, whose recent work has focused on forecasting future damages and costs associated with anthropogenic climate change, something known as the “social cost of carbon” (SCC). While it is difficult to put a price on the disruption to lives, homes, and livelihoods by climate change, incorporating SCC into policymaking and corporate decision-making can be crucial to inform the cost-benefit analyses used to evaluate policies and investments that would reduce society’s carbon emissions. And current studies in the literature may be significantly underestimating the true cost, according to Moore.

As damages from climate disasters mount around the world, the cost of carbon emissions is becoming more and more evident. Last year, the 10 largest extreme weather events (including record floods in Pakistan, Hurricane Ian in the United States, and more) resulted in an estimated $169 billion in damages. While the idea that there is a social cost associated with carbon emissions is broadly catching on, it is difficult to establish how much. In 2007, U.S. federal agencies were ordered by the courts to consider the SCC in rulemaking processes, with a U.S. Ninth Circuit Court of Appeals remanding a DOT fuel economy rule for failing to assign a dollar value to CO2 emissions, stating that “while the record shows that there is a range of values, the value of carbon emissions reduction is certainly not zero.” 

“It’s essentially an accounting problem,” according to Moore, Associate Professor and the Hurlstone Presidential Chair in the Department of Environmental Science and Policy at the University of California Davis. She pointed to three primary models used today: the DICE model, FUND, and PAGE. Each attempts to convert the socio-economic impact of a changing climate into common units (discounted dollars) and add them up. These models are currently used within the United States, Canada, Germany, and several other countries to inform policymaking and carbon pricing. The large majority of SCC estimates, including current U.S. federal values (~$50/ton on average), come from these three models.

However, growing critiques in the literature on SCC focus on these models’ dated representation of the climate system and expected damages as well as other limitations arising from significant uncertainties and model structure. Moore walked through several recent innovations in SCC model assumptions related to: the earth system and carbon cycles, the inclusion of climactic tipping points, the limited substitutability of certain goods, allowing transient changes in temperature to have persistent effects, accounting for the greater welfare loss of damages in poor regions, distinguishing risk and time preferences, allowing for learning over time, and accounting for model uncertainty. Numerous adjustments to these factors within the baseline modeling has resulted in a proliferation of different estimates.

To make sense of the variation, Moore and her team compiled 147 studies from the published literature into one dataset to form a single SCC distribution. Their goal was to identify the structural model assumptions and parametric variation driving the variation in results, so that they could account for these differences and craft a “comprehensive” estimate closer to the “true social cost of carbon.” They then paired this literature survey with a qualitative survey sent to the authors of the papers they compiled, asking each one what they thought was the central value reported in the literature and what they think is the “true” central value. The consensus response was that authors expected a central value of $50 per ton of CO₂ in the literature, versus a “true” value of $120/ton in reality, indicating a perceived discrepancy between published papers (many of them outdated) and what the experts would expect based on new methodologies or advancements. This suggests a downward publication bias, leading studies to report lower SCCs than might otherwise be expected.

Although the distribution of 2020 SCC values the team constructed “does not have a clear statistical interpretation,” according to Moore, they identified several main sources of variation, including different damage function parameters, tipping points, limited substitutability, pure time preference, growth damage, distributional weights, endogenous adaptation, etc. As a result, they were able to train a predictive “Random Forest Model”—a machine learning algorithm that can flexibly handle nonlinear variation—to produce “synthetic” SCC distributions that match the structural and normative assessments from the surveyed experts. Far from the current consensus in the literature of a SCC of $50/ton (as reported in the expert surveys), Moore’s team’s model produced distributions with a much higher mean of $330/ton and a median of $298/ton.

While this may seem like a shocking difference, the U.S. Environmental Protection Agency recently proposed an updated SCC of their own in 2022, with a mean of around $210/ton and median of around $160/ton, which is also significantly higher than the central tendency in the extant literature. The EPA notes this is due to “incorporating recent scientific advances” and updated methodology. Overall, Moore noted that while significant variation remains due to parametric and climate uncertainties, combining multiple elements of recent SCC improvements yields a much higher SCC than previously supposed. This is an important finding not only for updating the existing academic literature, but also with important implications for policymaking around the world as we seek to properly understand the future cost to society associated with continued carbon emissions, informing cost-benefit analyses of reducing emissions in the near-term or even the formation of a price on carbon.