In a recently proposed regulation, the US Environmental Protection Agency introduces an updated approach to estimating the social cost of carbon that incorporates important scientific and statistical advances.
On November 11, 2022, the US Environmental Protection Agency (EPA) proposed a rule to regulate emissions of methane from the oil and gas sector. While the proposed actions themselves are worthy of discussion, in this blog post, we focus exclusively on a key technical feature of the announcement: EPA has introduced a new approach to estimating the social cost of carbon (SCC) for a sensitivity case in its regulatory analysis.
EPA’s use of this new SCC is noteworthy because it marks the first time a government agency has put forward an updated approach to estimating the SCC that responds to Executive Order 13990, which was announced on January 20, 2021. This executive order initiated an update to the SCC, an important number that the federal government uses in various policy applications.
EPA adheres closely to the order in its updated SCC, and the agency will receive valuable feedback on the proposal, through both the upcoming public comment process and a formal, independent scientific peer review process that EPA also has announced. These signs all suggest that EPA’s approach will provide an important input to the ongoing government process to update the official estimate of the SCC. In short, EPA’s update is comprehensive, quite detailed and technical, and likely to be influential. It’s a worthy subject to unpack.
What the Social Cost of Carbon Means for the New Methane Rule
Before we dive further into the details, let’s first review how EPA’s analysis fits into the new rulemaking that regulates methane emissions. EPA’s proposed rule would reduce the emissions of methane, a powerful greenhouse gas. In the analysis that is required before EPA can implement the rule—an analysis which includes an assessment of the rule’s costs and benefits—EPA has followed standard government practice to quantify the benefits of the greenhouse gas reductions that are projected to result from the rule. To quantify these benefits, EPA uses the SCC, which is an estimate of the societal harms that result from each incremental ton of greenhouse gas emissions released into the atmosphere. “SCC” is a general term that often is used to include greenhouse gases other than carbon dioxide; in this specific case, EPA applies the social cost of methane in its benefit-cost analysis.
For EPA’s central analysis of the proposed methane rule, the agency does two things: It applies the interim estimates of the social cost of methane that were established under Executive Order 13990, which has its foundation in the SCC estimation process that was developed by the original Interagency Working Group on the Social Cost of Greenhouse Gases in 2010. EPA also includes a new sensitivity analysis that incorporates a comprehensive update to the way the agency estimates the SCC, including for carbon dioxide, methane, and nitrous oxide. This updated estimate allows for EPA to receive public comment on the new methodology, inform potential future modifications to the SCC estimation process, and support the potential usage of the SCC in the final rule.
In its sensitivity analysis, EPA updates each of the four major steps of SCC estimation: socioeconomic projections, climate modeling, translation to economic damages, and economic discounting. In doing so, the agency draws heavily on peer-reviewed and published work from the SCC Initiative, a multi-institution collaborative effort led by RFF and the University of California, Berkeley. This work includes the RFF-Berkeley Greenhouse Gas Impact Value Estimator (GIVE) model, which was recently published in the journal Nature.
EPA also relies on updated research related to climate damages that has been conducted by the Climate Impact Lab and researchers Peter Howard and Thomas Sterner. In a major step forward for transparency, the computer code used for the sensitivity has been built using the open-source Mimi software platform (another output of the SCC Initiative), making the code free and easily accessible to download, replicate, and evaluate.
EPA’s Updated Estimation Process for the Social Cost of Carbon
Let’s look at the improvements EPA has included in each of the steps involved in estimating the SCC, in turn.
Estimates of damages from climate change are influenced in no small part by socioeconomic outcomes and by the uncertainty that those outcomes will occur—namely, uncertainty surrounding future population size, economic growth, and emissions levels. To represent uncertainty across future socioeconomic outcomes in the SCC, the government has, since 2010, used a set of five different scenarios, treating each one as equally likely.
This approach has been criticized—for example, by a landmark 2017 panel of the National Academies of Science, Engineering, and Medicine (NASEM)—for not incorporating the uncertainties in the full set of relevant socioeconomic variables and for not reflecting the recent research on the full range of future climate scenarios. The SCC estimates that are based on the original methodology put forth by the Interagency Working Group, therefore, may not reflect damage calculations that are based on the full range of potential future outcomes. The NASEM panel recommended an approach to address these issues that uses a combination of statistical methods and input from experts to generate a set of long-term projections for each socioeconomic variable.
In its sensitivity analysis, EPA moves away from the simpler, scenario-based approach; instead, the agency implements the RFF Socioeconomic Projections, which are peer-reviewed, probabilistic projections that RFF has developed with our research partners and that implement the recommendations from NASEM. The RFF Socioeconomic Projections comprise projections of population; per capita economic growth; and emissions of carbon dioxide, methane, and nitrous oxide that have been generated using a combination of statistical modeling and input from experts.
EPA’s shift to using the RFF projections for its socioeconomic modeling comes with three important benefits. The first is that the RFF projections provide thousands of future socioeconomic trajectories that span the full range of uncertainty in future socioeconomic outcomes, rather than providing just five scenarios to sample. This breadth allows EPA to account for the uncertainty in those future outcomes and their probabilities.
The second benefit is that the RFF projections natively incorporate uncertainty about future climate policy, among other uncertainties. This forward-looking strategy brings EPA’s approach to socioeconomics into alignment with its standard approach of estimating the incremental costs and benefits of a rule against the future evolution of policy.
The third benefit is that the RFF projections were designed specifically to account for uncertainty across the very long timescale that’s required to calculate the SCC. The longer timescale in the RFF projections removes the need for EPA to arbitrarily extend existing estimates that originally were developed for shorter timescales.
Climate System Modeling
Another important element of SCC calculations is the representation of the climate system that’s used to estimate what the climate will look like in response to changes in emissions levels. The previous process of estimating the SCC relied on averaging the results of three models (DICE, FUND, and PAGE), each of which incorporates its own representation of the climate system, and none of which are up to date with current scientific advancements in the field.
For its new analysis, EPA follows the recommendation from NASEM to use a state-of-the-science representation of the climate system. Though simplified in comparison to more comprehensive models, this update represents the range of results provided by more complex models for variables such as global temperature.
Damage functions translate changes in global temperatures to dollar-value impacts of climate change on society, such as for human health, agriculture, and sea level rise. To update the damage functions, EPA uses three independent lines of evidence: the RFF-Berkeley GIVE model, the Climate Impact Lab’s Data-driven Spatial Climate Impact Model (DSCIM), and a recent meta-analysis by Peter Howard and Thomas Sterner. EPA averages the results from these three models to produce an SCC estimate for a given range of discount rates. (We’ll elaborate on the discount rate in the next section.) Averaging across damage functions is a reasonable approach when alternative damage functions can account for the overlapping impacts of climate change and are based on independent lines of evidence, both of which largely describe the damage functions that EPA uses. For example, four of the five categories of impacts accounted for in DSCIM also are included in GIVE. With the minor exception of sea level rise, the two models draw on completely independent lines of research to assess climate impacts.
The part of EPA’s analysis that use the GIVE damage functions and Howard and Sterner’s meta-analysis match those in our recent paper, which provided a new SCC estimate; those results also are publicly available in RFF’s SCC Explorer data tool. The GIVE model estimates damages for the following four key sectors: temperature-driven mortality, agricultural markets, energy expenditures for heating and cooling buildings, and coastal impacts from sea level rise. The Climate Impact Lab’s DSCIM assesses damages for the same sectors, plus labor productivity. The damage function from Howard and Sterner is an aggregate damage function that is based on a meta-analysis of the climate literature.
EPA’s SCC estimate therefore aligns with recent research and reflects the global nature of greenhouse gases. The estimate includes climate impacts, regardless of which country is affected or in which country those impacts originate. In contrast, the Trump administration sought to narrow the focus of the SCC to climate impacts that occur within US borders, using approximations that previous modeling efforts were not designed to produce. For example, the damage function in the study by Howard and Sterner is global in nature and therefore is unable to project domestic-only impacts. RFF scholars have discussed these issues previously in more detail in formal comments as part of the rulemaking process.
While our recent work allows a projection of damages at the national level, sufficient evidence doesn’t yet exist to comprehensively model the many international spillovers and interactions of climate impacts, such as effects on financial flows and international trade. For these reasons, in addition to others that EPA discusses in its report, a domestic-only SCC estimate that’s produced with existing models would be incomplete at best.
Analysts use discount rates to weigh the dollar value of future impacts against the dollar value of impacts experienced today, which makes the discount rate an important element in estimating the SCC. With its recent proposal, EPA makes two changes in its approach to discounting.
First, the agency moves from a discount rate of 3 percent to instead focus on a rate of 2 percent, with additional sensitivity analyses at 1.5 percent and 2.5 percent. The 2 percent value is in line with the consensus of recent economic and finance research; this value also corresponds with the approach we took in our recent paper. The lower discount rate reflects the decades-long decline in market rates of interest that has occurred in the nearly 20 years since the 3 percent benchmark was devised. While nominal interest rates have risen a fair amount this year, the 2 percent discount rate that the EPA uses represents a real (i.e., inflation-adjusted) discount rate, and real rates remain quite low. For example, despite recent market swings, as of November 2022, real Treasury yields remain below 2 percent.
More important, perhaps, is that EPA has changed the way it applies discount rates. Now, the agency appropriately prices risk with the standard framework that also is used to price risky financial assets. EPA refers to this approach as the “Ramsey” framework, named for economist Frank Ramsey who conducted the relevant research way back in 1928. The Ramsey framework captures the intuitive idea that an investment—whether the investment is a financial asset or an investment in mitigating climate change—is more valuable if it “pays off” particularly well in potential futures that experience bad outcomes (e.g., weak economic growth). Conversely, an investment is less valuable if the reverse is true (i.e., if the investment primarily pays off when the benefit is needed less).
The past estimates of the SCC from the Interagency Working Group incommensurately value climate risk, because those estimates used a constant 3 percent rate that did not consider the relationship of the discount rate with payoffs from investments, and therefore did not account for a risk premium that is familiar to students of financial analysis and asset pricing.
A fundamental way to value this risk is to use a “state-contingent” discount rate—that is, a higher discount rate in future states of the world when economic growth is high, and a lower discount rate when economic growth is weak. EPA’s new approach matches the risk-valuation methods we used in our recent paper, which linked the discount rate to the uncertain future pathway of economic growth (technically, “consumption growth,” which is the share of GDP that is actually consumed instead of saved).
We detailed a method of establishing the precise form of that link in a recent journal article (which we summarize in turn in this Common Resources blog post). These articles show how to link the discount rate to uncertain economic growth while also centering around a given target discount rate in the near term, such as EPA’s central 2 percent rate. Accounting for risk is crucial to a proper calculation of the value of investments. This incorporation of risk includes (but is not limited to) the value of mitigating climate change. Indeed, RFF scholarship on discounting has demonstrated that failing to account for risk could generate SCC estimates that are wrong by a factor of two or more.
Comparing the Environmental Protection Agency’s New Estimates
Given the comprehensive nature of EPA’s update and the substantial difference between the value of the interim SCC estimate and the updated SCC estimate, a natural question arises: What are the primary drivers of the change in results between the new estimates and the interim estimates?
Answering this question directly is challenging, because the change in the estimates is a result of scientific improvements in each of the steps of estimation, which can interact with each other in nonlinear ways. Such interactions complicate any attribution of the change to individual scientific improvements.
In our recent work, however, we conducted an analysis using only the GIVE model to assess what primarily drives the increased estimate for the SCC . We found that, while the greatest driver of the change was the shift to a 2 percent discount rate, both the updated damage functions and the updated climate model, socioeconomics, and pricing of risk each also increased the value of the SCC by about one-third.
Given the focus on reductions of methane emissions in the recent EPA proposal, the updated social cost of methane notably changes much less from the interim estimates than the social cost of carbon dioxide.
As shown in our SCC Explorer data tool, the GIVE damage functions and those from Howard and Sterner yield higher SCC estimates than the interim number: at the central 2 percent discount rate, the GIVE damage functions yield an SCC of $185 per ton of carbon dioxide, while Howard and Sterner’s damage functions yield $205 per ton. In comparison, the interim SCC estimate, which corresponds to a 3 percent discount rate, is $51. The social cost of methane rises from $1,500 to $1,900 per ton of methane using GIVE damage functions and to $2,200 per ton using the damage functions from Howard and Sterner.
EPA averages the estimates from these two damage functions with the results of the Climate Impact Lab’s DSCIM to yield the central estimates in the agency’s report. EPA’s estimates, which reflect the SCC for a ton of emissions released in 2020, are shown below in Table 1, alongside the results from the GIVE model for comparison.
Table 1. Estimates of the Social Costs of Carbon and Methane Derived Using Multiple Methods
EPA = US Environmental Protection Agency; GIVE = RFF-Berkeley Greenhouse Gas Impact Value Estimator model; DSCIM = the Climate Impact Lab’s Data-driven Spatial Climate Impact Model; Howard and Sterner = 2017 meta-analysis by Peter H. Howard and Thomas Sterner; Rennert et al. (2022) = publication detailing the GIVE model.
The GIVE model produces SCC estimates that are the same as the three-model average after rounding the numbers: $190 per ton of carbon dioxide (our unrounded value from GIVE is $185), with the estimates corresponding to the other two discount rates (2.5 percent and 1.5 percent) straddling this value. For methane, the three-model average of $1,600 per ton of methane is somewhat lower than the $1,900 value produced by the GIVE model.
Given the focus on reductions of methane emissions in the recent EPA proposal, the updated social cost of methane notably changes much less from the interim estimates than the social cost of carbon dioxide (in terms of percentage). Two primary reasons explain this result.
First, while methane has a much more potent warming effect than carbon dioxide on a per-molecule basis, methane also breaks down more quickly in the atmosphere than carbon dioxide. This faster breakdown makes methane’s warming effects less sensitive overall to the shift to a lower discount rate.
Second, the updated climate model that EPA uses represents the amount of time that methane remains in the atmosphere more accurately than previous models, an update that produces lower long-term damages compared to previous models.
The Significance of the Environmental Protection Agency’s Update to the Social Cost of Carbon
In our view, EPA’s updated approach to estimating the SCC is fully responsive to the near-term recommendations in the NASEM report on the interim SCC estimate, including the quantification of uncertainty throughout the major components of SCC estimation.
EPA’s move to make its analysis fully open source and freely available represents an important step forward for transparency and open science. The proposal of this approach represents an important milestone in the process of moving toward an updated SCC estimation methodology, and we look forward to the next step: engagement with the new estimates by the public, the scientific community, and an independent panel of peer reviewers.