&EPA
Docket ID No. EPA-HQ-OAR-2021-0317
September 2022
Supplementary Material for the Regulatory
Impact Analysis for the Supplemental Proposed
Rulemaking, "Standards of Performance for
New, Reconstructed, and Modified Sources and
Emissions Guidelines for Existing Sources: Oil
and Natural Gas Sector Climate Review"
EPA External Review Draft of Report on the Social
Cost of Greenhouse Gases: Estimates Incorporating
Recent Scientific Advances
-------
EXTERNAL REVIEW DRAFT
Report on the Social Cost of
Greenhouse Gases:
Estimates Incorporating Recent Scientific Advances
September 2022
National Center for Environmental Economics
Office of Policy
Climate Change Division
Office of Air and Radiation
U.S. Environmental Protection Agency
Washington, DC 20460
-------
Table of Contents
List of Figures ii
List of Tables iii
List of Abbreviations iv
Executive Summary 1
1 Background 4
1.1 Overview of SC-GHG Estimates Used in EPA Analyses to Date 4
1.2 Recommendations from the National Academies of Sciences, Engineering, and Medicine 8
1.3 Accounting for Global Damages 10
2 Methodological Updates 16
2.1 Socioeconomic and Emissions Module 18
2.2 Climate Module 26
2.3 Damage Module 37
2.4 Discounting Module 52
2.5 Risk Aversion 62
3 Modeling Results 66
3.1 Social Cost of Carbon (SC-CO2), Methane (SC-CH4), and Nitrous Oxide (SC-N20) Estimates by
Damage Module 66
3.2 Omitted Damages and Other Modeling Limitations 70
3.3 Distribution of Modeled Climate Impacts 77
4 Using SC-GHG Estimates in Policy Analysis 80
4.1 Combining Lines of Evidence on Damages 81
4.2 Application of SC-GHG Estimates in Benefit-Cost Analysis 82
5 Summary 84
References 86
A. Appendix Ill
A.l. Additional Discussion of Scientific Updates in IPCC's Sixth Assessment Report Ill
A.2. Consumption Rate of Interest and Integration into Benefit-Cost Analysis 112
A.3. Derivations of the SC-GHG Values for use in Analyses 115
A.4. Annual Unrounded SC-CO2, SC-CH4, and SC-N20 Values, 2020-2080 120
A.5. Additional Figures, Tables, and Results 122
A.6. Valuation Methodologies to Use in Estimating the Social Cost of GHGs 127
-------
EXTERNAL REVIEW DRAFT
List of Figures
Figure 2.1: The Four Components of SC-GHG Estimation 17
Figure 2.1.1: Global Population under RFF-SPs and SSPs, 1950-2300 22
Figure 2.1.2: Long-run Projections of Growth in Global Income per Capita under RFF-SPs and SSPs, 2020-
2300 23
Figure 2.1.3: Net Annual Global Emissions of Carbon Dioxide (C02) under RFF-SPs and SSPs, 1900-230025
Figure 2.2.1: Global Atmospheric Concentrations of Carbon Dioxide (C02), 1900-2300 31
Figure 2.2.2: Global Mean Surface Temperature Change, 1900-2300 31
Figure 2.2.3: Global Mean Surface Temperature Anomaly from a Pulse of Carbon Dioxide (lGtC) by Model,
2020-2300 33
Figure 2.2.4: Global Sea Level Rise in FACTS and BRICK, 1950-2300 36
Figure 2.3.1: Research on Climate Impacts, 1990-2021 38
Figure 2.3.2: Annual Consumption Loss as a Fraction of Global GDP in 2100 Due to an Increase in Annual
Global Mean Surface Temperature in the three Damage Modules 51
Figure 2.4.1: Distribution of the Dynamic Discount Rates 60
Figure 3.1.1: Distribution of Social Cost of Carbon Dioxide (SC-C02) Estimates for 2030, by Near-term
Ramsey Discount Rate and Damage Module 69
Figure 3.2.1: Population, Temperature, and Sea Level Rise in 2100 74
Figure 3.2.2: Global Ocean pH and Ocean Heat, 2020-2300 75
Figure A.3.1 The Difference Between using a Certainty-Equivalent Rate and Constant Discount Rate to
Discount Climate Benefits from Future Reductions in GHG Emissions Back to the Year of the
Analysis 119
Figure A.5.1: Net Annual Global Emissions of Methane (CH4) under the RFF-SPs and the SSPs, 1900-2300
122
Figure A.5.2: Net Annual Global Emissions of Nitrous Oxide (N20) under the RFF-SPs and the SSPs, 1900-
2300 122
Figure A.5.3: Global Atmospheric Concentrations of Methane (CH4), 1900-2300 123
Figure A.5.4: Global Atmospheric Concentrations of Nitrous Oxide (N20), 2020-2300 123
Figure A.5.5: Global Temperature Anomaly from a Pulse of Methane (lMtCH4) Emissions, 2020-2300 124
Figure A.5.6: Global Temperature Anomaly from a Pulse of Nitrous Oxide (lMtN2) Emissions, 2020-2300
124
Figure A.5.7: Dynamic temperature response of 256 climate science models (the CMIP5 ensemble) and
seven lAMs 125
Figure A.5.8: Distribution of SC-CH4 Estimates for 2030, by Damage Module and Discount Rate 126
Figure A.5.9: Distribution of SC-N20 Estimates for 2030, by Damage Module and Discount Rate 126
-------
EXTERNAL REVIEW DRAFT
List of Tables
Table 2.2.1: Summary Statistics for Equilibrium Climate Sensitivity under Reduced-Complexity Climate
Models and IPCC statements 29
Table 2.2.2: Summary Statistics for Transient Climate Response under Reduced-Complexity Climate
Models and IPCC Statements 30
Table 2.3.1: Current Coverage of Climate Damages in DSCIM 43
Table 2.3.2: Current Coverage of Climate Damages in GIVE 46
Table 2.4.1: Average Real Return on 10-Year Treasury Securities 58
Table 2.4.2: Calibrated Ramsey Formula Parameters 60
Table 3.1.1: Social Cost of Carbon (SC-C02) by Damage Module, 2020-2080 (in 2020 dollars per metric ton
of C02) 67
Table 3.1.2: Social Cost of Methane (SC-CH4) by Damage Module, 2020-2080 (in 2020 dollars per metric
ton of CH4) 68
Table 3.1.3: Social Cost of Nitrous Oxide (SC-N20) by Damage Module, 2020-2080 (in 2020 dollars per
metric ton of N20) 68
Table 3.1.4: Sectoral Disaggregation of Social Cost of Carbon (SC-C02) for 2030 under a 2.0% Near-Term
Ramsey Discount Rate (in 2020 dollars per metric ton of C02) 70
Table 3.2.1: Scope of Climate Science, Impacts, and Damages Included in the Updated SC-GHG Estimates
73
Table 4.1.1: Estimates of the Social Cost of Greenhouse Gases (SC-GHG), 2020-2080 (in 2020 dollars per
metric ton) 81
Table 5.1: Implementation of National Academies Recommendations in this Report 85
Table 4.2.1: Unrounded SC-C02, SC-CH4, and SC-N20 Values, 2020-2080 120
Table 4.2.2: Unrounded SC-C02, SC-CH4, and SC-N20 Values, 2020-2080 (continued...) 121
-------
EXTERNAL REVIEW DRAFT
List of Abbreviations
AR
Assessment Report of the United Nations Intergovernmental Panel on Climate
BRICK
Building Blocks for Relevant Ice and Climate Knowledge
ch4
Methane
CIAM
Coastal Impact and Adaptation Model
CMIP
Coupled Model Intercomparison Project
C02
Carbon Dioxide
DICE
Dynamic Integrated Climate and Economy
DSCIM
Data-driven Spatial Climate Impact Model
ESM
Earth System Models
ECS
Equilibrium Climate Sensitivity
E.O.
Executive Order
FACTS
Framework for Assessing Changes To Sea-level
FaIR
Finite Amplitude Impulse Response
FUND
Climate Framework for Uncertainty, Negotiation, and Distribution
GHG
Greenhouse Gas
GDP
Gross Domestic Product
GIVE
Greenhouse Gas Impact Value Estimator
GMSL
Global Mean Sea Level
GMST
Global Mean Surface Temperature
1AM
Integrated Assessment Model
IWG
Interagency Working Group on the Social Cost of Greenhouse Gases
MAGICC
Model for the Assessment of Greenhouse Gas Induced Climate Change
N20
Nitrous Oxide
PAGE
Policy Analysis of the Greenhouse Gas Effect
PPP
Purchasing Power Parity
RC
Reduced Complexity
RCP
Representative Concentration Pathway
SC
Social Cost
SLR
Sea-level Rise
SP
Socioeconomic Projections
SSP
Shared Socioeconomic Pathways
TCR
Transient Climate Response
-------
Executive Summary
This report presents new estimates of the social cost of carbon (SC-C02), social cost of methane (SC-CH4),
and social cost of nitrous oxide (SC-N20), collectively referred to as the "social cost of greenhouse gases"
(SC-GHG). These estimates reflect recent advances in the scientific literature on climate change and its
economic impacts and incorporate recommendations made by the National Academies of Science,
Engineering, and Medicine (National Academies 2017). The SC-GHG allows analysts to incorporate the net
social benefits of reducing emissions of greenhouse gases (GHG), or the net social costs of increasing such
emissions, in benefit-cost analysis and, when appropriate, in decision-making and other contexts. The SC-
GHG is the monetary value of the net harm to society from emitting a metric ton of that GHG to the
atmosphere in a given year. The SC-GHG, therefore, also reflects the societal net benefit of reducing
emissions of the GHG by a metric ton. The SC-GHG is the theoretically appropriate value to use when
conducting benefit-cost analyses of policies that affect GHG emissions.
Since 2008, the EPA has used estimates of the SC-GHG in analyses of actions that affect GHG emissions.
The values used by the EPA from 2009 to 2016, and since 2021, have been consistent with those
developed and recommended by the Interagency Working Group on the SC-GHG (IWG), and the values
used from 2017-2020 were consistent with those required by Executive Order (E.O.) 13783. During that
time, the National Academies conducted a comprehensive review of the SC-CO2 and issued a final report
in 2017 recommending specific criteria for future updates to the SC-CO2 estimates, a modeling framework
to satisfy the specified criteria, and both near-term updates and longer-term research needs pertaining
to various components of the estimation process. The IWG was reconstituted in 2021 and E.O. 13990
directed it to develop a comprehensive update of its SC-GHG estimates, recommendations regarding
areas of decision-making to which SC-GHG should be applied, and a standardized review and updating
process to ensure that the recommended estimates continue to be based on the best available economics
and science going forward.
The EPA is a member of the IWG and is participating in the IWG's work under E.O. 13990. While that
process continues, this EPA report presents a set of SC-GHG estimates that incorporates numerous
methodological updates addressing the near-term recommendations of the National Academies. The
report takes a modular approach in which the methodology underlying each of the four components, or
modules, of the SC-GHG estimation process - socioeconomics and emissions, climate, damages, and
discounting - is developed by drawing on the latest research and expertise from the scientific disciplines
relevant to that component. The socioeconomic and emissions module relies on a new set of probabilistic
projections for population, income, and GHG emissions developed under the Resources for the Future
Social Cost of Carbon Initiative (Rennert et al. 2022a). The climate module relies on the Finite Amplitude
Impulse Response (FaIR) model (Millar et al. 2017; Smith et al. 2018, IPCC 2021b), a widely used Earth
system model recommended by the National Academies, which captures the relationships between GHG
emissions, atmospheric GHG concentrations, and global mean surface temperature. The socioeconomic
projections and outputs of the climate module are used as inputs to the damage module to estimate
monetized future damages from temperature changes. Based on a review of available studies and
1
-------
EXTERNAL REVIEW DRAFT
approaches to damage function estimation, the report uses three separate damage functions to form the
damage module. They are:
1. a subnational-scale, sectoral damage function (based on the Data-driven Spatial Climate Impact
Model (DSCIM) developed by the Climate Impact Lab (CIL 2022, Carleton et al. 2022, Rode et al.
2021)),
2. a country-scale, sectoral damage function (based on the Greenhouse Gas Impact Value Estimator
(GIVE) model developed under RFF's Social Cost of Carbon Initiative (Rennert et al. 2022b)), and
3. a meta-analysis-based damage function (based on Howard and Sterner (2017)).
The discounting module discounts the stream of future climate damages back to the year of emissions
using a set of dynamic discount rates that have been calibrated following the Newell et al. (2022)
approach, as applied in Rennert et al. (2022a, 2022b). This approach uses the Ramsey (1928) discounting
formula in which the parameters are calibrated such that (1) the decline in the certainty-equivalent
discount rate matches the latest empirical evidence on interest rate uncertainty estimated by Bauer and
Rudebusch (2020, 2021) and (2) the average of the certainty-equivalent discount rate overthe first decade
matches a near-term consumption rate of interest. Uncertainty in the starting rate is addressed by using
three near-term target rates (1.5, 2.0, and 2.5 percent) based on multiple lines of evidence on observed
market interest rates. This approach results in three dynamic discount rate paths and is consistent with
the National Academies (2017) recommendation to use three sets of Ramsey parameters that reflect a
range of near-term certainty-equivalent discount rates and are consistent with theory and empirical
evidence on consumption rate uncertainty. Finally, the value of aversion to risk associated with damages
from GHG emissions is explicitly incorporated into the modeling framework following the economic
literature.
The estimation process generates nine separate distributions of estimates - the product of using three
damage modules and three near-term target discount rates - of the social cost of each gas in each
emissions year. To produce a range of estimates that reflects the uncertainty in the estimation exercise
while providing a manageable number of estimates for policy analysis, in this report the multiple lines of
evidence on damage modules are combined by averaging the results across the three damage module
specifications. Table ES.l summarizes the resulting SC-C02, SC-CH4, and SC-N20 estimates for emissions
years 2020 through 2080.
The modeling implemented in this report reflects conservative methodological choices, and, given both
these choices and the numerous categories of damages that are not currently quantified and other model
limitations, the resulting SC-GHG estimates likely underestimate the marginal damages from GHG
pollution. The EPA will continue to review developments in the literature, including more robust
methodologies for estimating the magnitude of the various direct and indirect damages from GHG
emissions, and look for opportunities to further improve SC-GHG estimation going forward.
2
-------
EXTERNAL REVIEW DRAFT
Table ES.l: Estimates of the Social Cost of Greenhouse Gases (SC-GHG), 2020-2080 (2020 dollars)
SC-GHG and Near-term Ramsey Discount Rate
SC-CO2
SC-CH4
SC-N2O
(2020 dollars per metric ton ofC02)
(2020 dollars per metric ton ofCH4)
(2020 dollars per metric ton of N20)
Emission
Year
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2020
120
190
340
1,300
1,600
2,300
35,000
54,000
87,000
2030
140
230
380
1,900
2,400
3,200
45,000
66,000
100,000
2040
170
270
430
2,700
3,300
4,200
55,000
79,000
120,000
2050
200
310
480
3,500
4,200
5,300
66,000
93,000
140,000
2060
230
350
530
4,300
5,100
6,300
76,000
110,000
150,000
2070
260
380
570
5,000
5,900
7,200
85,000
120,000
170,000
2080
280
410
600
5,800
6,800
8,200
95,000
130,000
180,000
Values of SC-CO2, SC-CH4, and SC-N20 are rounded to two significant figures. The annual unrounded estimates are available in
Appendix A.4 and at: www.epa.gov/environmental-economics/scghg.
-------
EXTERNAL REVIEW DRAFT
1 Background
A robust and scientifically founded assessment of the positive and negative impacts that an action can be
expected to have on society facilitates evidence-based policy making. Estimates of the social cost of
carbon (SC-C02), social cost of methane (SC-CH4), and social cost of nitrous oxide (SC-N20) allow analysts
to incorporate the net social benefits of reducing emissions of each of these greenhouse gases, or the net
social costs of increasing such emissions, in benefit-cost analysis, and when appropriate, in decision
making and other contexts.1 Collectively, these values are referred to as the "social cost of greenhouse
gases" (SC-GHG) in this document. The SC-GHG is the monetary value of the future stream of net damages
associated with adding one ton of that GHG to the atmosphere in a given year. The SC-GHG, therefore,
also reflects the societal net benefit of reducing emissions of the gas by one ton. The social benefits of
abatement are an aggregated measure of the affected individuals' willingness to pay to avoid those
damages. The SC-GHG is the marginal social benefit of GHG abatement and is, therefore, the theoretically
appropriate value to use when conducting benefit-cost analyses of policies that affect GHG emissions.2
Estimates of the marginal social cost will differ by the type of GHG (such as C02, CH4, and N20) and by the
year in which the emissions change occurs.
In principle, the SC-GHG includes the value of all climate change impacts (both negative and positive),
including (but not limited to) changes in net agricultural productivity, human health effects, property
damage from increased flood risk, changes in the frequency and severity natural disasters, disruption of
energy systems, risk of conflict, environmental migration, and the value of ecosystem services. In practice,
because of data and modeling limitations, which prevent full representation of harmful climate impacts,
estimates of the SC-GHG are a partial accounting of climate change impacts and, as such, lead to
underestimates of the marginal benefits of abatement.
1.1 Overview of SC-GHG Estimates Used in EPA Analyses to Date
The academic literature has published estimates of the social cost of carbon and other GHGs since at least
the early 1990s. As early as 2002 researchers began conducting reviews that combined lines of evidence
across early SC-C02 estimates (Clarkson and Deyes 2002). The EPA began regularly incorporating SC-C02
estimates in regulatory impact analyses following a 2008 court ruling in which an agency was ordered to
1 Note, for example, that EPA has recommended use of SC-GHG estimates in environmental impact statements under
NEPA when appropriate. See e.g., Letter from EPA to USPS, on the Final Environmental Impact Statement for Next
Generation Delivery Vehicle Acquisitions, Feb. 2, 2022.
2 These estimates of social damages should not be confused with the estimated costs of attaining a predetermined
emissions or warming limit. Specifically, there is another strand of research that investigates the costs of setting a
specific climate target (e.g., capping emissions or temperature increases to a certain level). The expected marginal
cost of GHG abatement associated with meeting a specific climate target can be useful in evaluating policy cost-
effectiveness but is not an alternative way to value damages from GHG emissions in benefit-cost analysis. For more
on how these concepts (e.g., a predetermined target-based approach and a damage (SC-GHG) based approach) can
be used when designing climate policy and in policy evaluation, see, for example, Hansel et al. (2020); Stern et al.
(2022); Aldy et al. (2021); and Gundlach and Livermore (2022).
4
-------
EXTERNAL REVIEW DRAFT
consider the SC-C02 in the rulemaking process. Specifically, the U.S. Ninth Circuit Court of Appeals
remanded a fuel economy rule to the Department of Transportation for failing to consider the value of
reducing C02 emissions when determining the appropriate level of the fuel economy standard, stating
that "while the record shows that there is a range of values, the value of carbon emissions reduction is
certainly not zero."3 The SC-C02 estimates initially presented in EPA analyses in 2008 and early 2009 were
derived from the academic literature.4
Beginning in September 2009, EPA's regulatory impact analyses applied SC-CO2 estimates that were
developed through a U.S. Government interagency working group (IWG) process. The IWG was launched
in early 2009, under the leadership of the Office of Management and Budget (OMB) and the Council of
Economic Advisers (CEA), to ensure that Federal agencies had access to the best available information
when quantifying the benefits of reducing C02 emissions in benefit-cost analyses. The IWG included
technical experts from the EPA and other federal agencies. The IWG first developed an interim set of SC-
C02 estimates based on an average of estimates published in the peer reviewed academic literature.5 The
EPA chose to use these interim estimates in multiple regulatory impact analyses and sought public
comments to inform the estimates for future use.6 In 2010, the IWG published a Technical Support
Document (TSD) with a set of four updated SC-CO2 estimates recommended for use in regulatory analyses
in addition to guidance on using the estimates (IWG 2010). Three of these values were based on the
average SC-CO2 from three widely cited integrated assessment models (lAMs) in the peer-reviewed
literature - DICE, PAGE, and FUND7 - at constant discount rates of 2.5, 3, and 5 percent. The fourth value
was included to represent the potential for lower-probability, higher-impact outcomes from climate
change, that would be particularly harmful to society and thus relevant to the public and policymakers.
For this purpose, it selected the SC-CO2 value for the 95th percentile at a 3 percent discount rate. Absent
3 Ctr. for Biological Diversity v. Nat'l Highway Traffic Safety Admin., 538 F.3d 1172,1200 (9th Cir. 2008).
4 For more information, see "Technical Support Document on Benefits of Reducing GHG Emissions"
(httpsi//www.regulations.gov/document/EPA-HQ-QAR-2008-0318-0078), prepared for EPA's July 2008 Advanced
Notice of Proposed Rulemaking for Regulating Greenhouse Gas Emissions Under the Clean Air Act, and EPA's May
2009 Regulatory Impact Analysis for the Renewable Fuel Standard Program (RFS2) Proposed Rule.
5 The IWG used a meta-analysis of SC-CO2 estimates (Tol 2008) as the starting point for the development of the
interim estimates recommended in 2009. With that starting point, the IWG filtered the existing SC-CO2 estimates in
the meta-analysis by using those that (1) were derived from peer-reviewed studies; (2) did not weight the monetized
damages to one country more than those in other countries (i.e., no equity weighting); (3) used a "business as usual"
climate scenario; and (4) were based on the most recent published version of each of the three major integrated
assessment models (lAMs): FUND, PAGE, and DICE. See EPA and DOT (2009) for more discussion of how the filtered
estimates were combined to form a set of five recommended interim values.
6 See, for example, EPA and DOT's joint September 2009 Proposed Rulemaking to Establish Light-Duty Vehicle
Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards (EPA and DOT 2009).
7 The DICE (Dynamic Integrated Climate and Economy) model by William Nordhaus evolved from a series of energy
models and was first presented in 1990 (Nordhaus and Boyer 2000, Nordhaus 2008). The PAGE (Policy Analysis of
the Greenhouse Effect) model was developed by Chris Hope in 1991 for use by European decision-makers in
assessing the marginal impact of carbon emissions (Hope 2006, Hope 2008). The FUND (Climate Framework for
Uncertainty, Negotiation, and Distribution) model, developed by Richard Tol in the early 1990s, was originally used
to study international capital transfers in climate policy and was subsequently widely used to study climate impacts
(e.g., Tol 2002a, Tol 2002b, Anthoff et al. 2009, Tol 2009).
5
-------
EXTERNAL REVIEW DRAFT
formal inclusion of risk aversion in the modeling, considering values above the mean in a right skewed
distribution with long tails acknowledges society's preference for avoiding risk.
The EPA chose to update the set of SC-C02 estimates used in regulatory analyses following a May 2013
update of the IWG SC-C02 estimates (IWG 2013). The 2013 IWG SC-C02 update incorporated new versions
of the lAMs used in the peer-reviewed literature but did not revisit other IWG modeling decisions (i.e.,
the discount rates or harmonized inputs for socioeconomic and emission scenarios and equilibrium
climate sensitivity). Improvements in the way damages are modeled were confined to those that had been
incorporated into the latest versions of the models by the developers themselves in the peer-reviewed
literature.8
In June 2015, the EPA began using estimates of SC-CH4 and SC-N20 from Marten et al. (2015), which were
consistent with the methodology underlying the IWG's estimates of the SC-CO2 estimates. The Marten et
al. estimates were first applied in sensitivity analyses in regulatory impact analyses of proposed
rulemakings with CH4 and N20 emission impacts.9 Following the completion of an external peer review of
the application of these estimates to federal regulatory analysis, the estimates were used in the main
analysis of other proposed rulemakings with CH4 emissions impacts (EPA 2015a, 2015b).10 In August 2016,
the Marten et al. SC-CH4 and SC-N20 estimates were adopted by the IWG in an addendum to the IWG's
TSD (IWG 2016a, 2016b).11 The IWG recommended these estimates as a method for improving the
analyses of regulatory actions that are projected to influence CH4 or N20 emissions in a manner consistent
with how C02 emission changes were being valued.
Over the course of developing and updating the SC-GHG estimates that have been used in EPA analyses,
there were extensive opportunities for public input on the estimates and underlying methodologies. There
was a public comment process associated with each proposed EPA rulemaking that used the estimates,
and OMB initiated a separate comment process on the IWG TSD in 2013. Commenters offered a wide
range of perspectives on all aspects of the process, methodology, and final estimates, and submitted
diverse suggestions for improvements. The U.S. Government Accountability Office (GAO) reviewed the
development of the IWG SC-CO2 estimates and concluded that the IWG processes and methods reflected
three principles: consensus-based decision making, reliance on existing academic literature and models,
and disclosure of limitations and incorporation of new information (GAO 2014).
8 The IWG subsequently provided additional minor technical revisions in November of 2013 and July of 2015, as
explained in Appendix B of the 2016 TSD (IWG 2016a).
9 The SC-CH4 and SC-N2O estimates were first used in sensitivity analysis for the Proposed Rulemaking for
Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles-Phase
2 (EPA and DOT 2015).
10 For a discussion of public comments received on the valuation of non-CC>2 GHG impacts in general and the use of
the Marten et al. (2015) estimates, see, e.g., EPA (2012a, 2012b, 2016a, 2016b), EPA and DOT (2016).
11 In 2021, the EPA developed analogous estimates of the social cost of hydrofluorocarbons (SC-HFCs) that are
consistent with the methodology underlying the SC-CO2, SC-CFU, and SC-N2O estimates. See, for example, EPA's final
Regulatory Impact Analysis for Phasing Down Production and Consumption of Hydrofluorocarbons (HFCs) for more
information (EPA 2021a).
6
-------
EXTERNAL REVIEW DRAFT
In 2015, as part of the IWG response to the public comments received in the 2013 solicitation, the IWG
announced a National Academies review of the IWG estimates (IWG 2015). Specifically, the IWG asked
the National Academies to conduct a multi-discipline, two-phase assessment of the IWG estimates and
offer advice on approaching future updates to ensure that the estimates continue to reflect the best
available science and methodologies. The National Academies' interim (Phase 1) report (National
Academies 2016a) recommended against a near-term update of the SC-C02 estimates within the existing
modeling framework. For future revisions, the National Academies recommended a broader update of
the climate system module consistent with the most recent, best available science and offered
recommendations for how to enhance the discussion and presentation of uncertainty in the SC-C02
estimates. In addition to publishing estimates of SC-CH4 and SC-N20, the IWG's 2016 TSD revision
responded to the National Academies' Phase 1 report recommendations regarding the presentation of
uncertainty. The revisions included: an expanded presentation of the SC-GHG estimates that highlights a
symmetric range of uncertainty around estimates for each discount rate; new sections that provide a
unified discussion of the methodology used to incorporate sources of uncertainty; a detailed explanation
of the uncertain parameters in the FUND and PAGE models; and making the full set of SC-C02 estimates
easily accessible to the public on OMB's website.
In January 2017, the National Academies released their final report, Valuing Climate Damages: Updating
Estimation of the Social Cost of Carbon Dioxide and recommended specific criteria for future updates to
the SC-C02 estimates, a modeling framework to satisfy the specified criteria, and both near-term updates
and longer-term research needs pertaining to various components of the estimation process (National
Academies 2017). A description of the National Academies' recommendations for near-term updates is
provided in Section 1.2 below. Shortly thereafter, in March 2017, President Trump issued Executive Order
(E.O.) 13783, which called for the rescission and review of several climate-related Presidential and
regulatory actions as well as for a review of the SC-GHG estimates used for regulatory impact analyses.12
Further, E.O. 13783 disbanded the IWG, withdrew the previous TSDs, and directed agencies to "ensure"
SC-GHG estimates used in regulatory analyses "are consistent with the guidance contained in OMB
Circular A-4", "including with respect to the consideration of domestic versus international impacts and
the consideration of appropriate discount rates" (E.O. 13783, Section 5(c)). The EPA's benefit-cost
analyses following E.O. 13783 used SC-GHG estimates that attempted to focus on the specific share of
physical climate change damages in the U.S. as captured by the models (which do not reflect many
pathways by which climate impacts affect the welfare of U.S. citizens and residents) and were calculated
using two default discount rates recommended by OMB Circular A-4 (2003), 3 percent and 7 percent.13
12https://www.federal register.gov/documents/2017/03/31/2017-06576/promoting-energy-independence-and-
economic-growth
13The EPA's regulatory analyses under E.O. 13783 included sensitivity analyses based on global SC-GHG values and
using a lower discount rate of 2.5%. OMB Circular A-4 (2003) recognizes that special considerations arise when
applying discount rates if intergenerational effects are important. In the IWG's 2015 Response to Comments, OMB—
as a co-chair of the IWG—made clear that "Circular A-4 is a living document," that "the use of 7 percent is not
considered appropriate for intergenerational discounting," and that "[t]here is wide support for this view in the
academic literature, and it is recognized in Circular A-4 itself." OMB, as part of the IWG, similarly repeatedly
confirmed that "a focus on global SCC estimates in [regulatory impact analyses] is appropriate" (IWG 2015). See
Sections 1.3 and 2.3 for further discussion on both issues.
7
-------
EXTERNAL REVIEW DRAFT
All other methodological decisions and model versions used in SC-GHG calculations under E.O. 13783
remained the same as those used by the IWG in 2010 and 2013, respectively.
On January 20, 2021, President Biden issued E.O. 13990 which established an IWG and directed the group
to develop an update of the SC-GHG estimates that reflect the best available science and the
recommendations of the National Academies (2017).14 In February 2021, the IWG recommended the
interim use of the most recent SC-GHG estimates developed by the IWG prior to the group being
disbanded in 2017, adjusted for inflation (IWG 2021). As discussed in the February 2021 TSD, the IWG
concluded that these interim estimates reflected the immediate need to have SC-GHG estimates available
for agencies to use in regulatory benefit-cost analyses and other applications that were developed using
a transparent process, peer reviewed methodologies, and the science available at the time of that process.
The February 2021 update also recognized the limitations of the interim estimates and encouraged
agencies to use their best judgment in, for example, considering sensitivity analyses using lower discount
rates. The IWG published a Federal Register notice on May 7, 2021, soliciting comment on the February
2021 TSD and on how best to incorporate the latest peer-reviewed scientific literature in order to develop
an updated set of SC-GHG estimates. The EPA has applied the IWG's interim SC-GHG estimates in analyses
published since the release of the February 2021 TSD (see, e.g., EPA (2021b, 2021c)) and has reviewed the
comments submitted to the IWG in developing this report.
1.2 Recommendations from the National Academies of Sciences, Engineering, and
Medicine
As previously mentioned, in 2015, the IWG requested that the National Academies review and
recommend potential approaches for improving its SC-C02 estimation methodology. In response, the
National Academies convened a multidisciplinary committee, called the Committee on Assessing
Approaches to Updating the Social Cost of Carbon. In addition to evaluating the IWG's overall approach
to SC-C02 estimation, the committee reviewed its choices of lAMs and damage functions, climate science
assumptions, future baseline socioeconomic and emission projections, presentation of uncertainty, and
discount rates.
In its final report (National Academies 2017), the National Academies committee recommended that the
IWG pursue an integrated modular approach to the key components of SC-CO2 estimation to allow for
independent updating and review and to draw more readily on expertise from the wide range of scientific
disciplines relevant to SC-CO2 estimation. Under this approach, each step in SC-CO2 estimation is
developed as a module—socioeconomic projections, climate science, economic damages, and
discounting—that reflects the state of scientific knowledge in the current peer-reviewed literature. In the
longer term, it recommended that the IWG communicate research needs and priorities to its member
agencies to stimulate research on ways to improve accounting of interactions and feedbacks between
these components. In addition, the committee noted that, while the IWG harmonized key inputs across
three lAMs, shifting to the use of a single climate module in the nearer-term (2-3 years) and eventually
transitioning to a single framework for all modules will enhance transparency, improve consistency with
the underlying science, and allow for more explicit representation of uncertainty. It recommended these
14https://www.federal register.gov/documents/2021/01/25/2021-01765/protecting-public-health-and-the-
environment-and-restoring-science-to-tackle-the-climate-crisis
8
-------
EXTERNAL REVIEW DRAFT
three criteria also be used to judge the value of other updates to the methodology. It also recommended
that the IWG update SC-C02 estimates at regular intervals, suggesting a five-year cycle.
Regarding the key components of the SC-CO2, the committee recommended the following improvements:
Socioeconomic and emissions projections: Use accepted statistical methods and elicit expert
judgment to project probability distributions of future annual growth rates of per-capita gross
domestic product (GDP) and population, bearing in mind the potential correlation between
economic and population projections. Use expert elicitation, guided by information on historical
trends and emissions consistent with different climate outcomes, to project emissions for each
forcing agent of interest, conditional on population and income scenarios. Additional
recommendations were offered pertaining to the time horizon, inclusion of future policies,
disaggregation of scenarios, and feedbacks from the damage module to the socioeconomic
module.
Climate science: Adopt or develop a simple Earth system model (such as the Finite Amplitude
Impulse Response (FaIR) model) to capture the relationships between C02 emissions, atmospheric
C02 concentrations, and global mean surface temperature change over time while accounting for
non-C02 forcing and allowing for the evaluation of uncertainty. Adopt or develop a sea level rise
component in the climate module that: (1) accounts for uncertainty in the translation of global
mean temperature to global mean sea level rise and (2) is consistent with sea level rise projections
available in the literature for similar forcing and temperature pathways. The committee also
noted the importance of generating spatially and temporally disaggregated climate information
as inputs into damage estimation. It recommended the use of linear pattern scaling (which
estimates linear relationships between global mean temperature and local climate variables) to
achieve this goal in the near-term.
Economic damages: Improve and update existing formulations of individual sectoral damage
functions when feasible; characterize damage function calibrations quantitatively and
transparently; present spatially disaggregated market and nonmarket damages by region and
sector in both monetary and natural units (incremental and total) and discuss how they scale with
temperature, income, and population; and recognize any correlations between formulations
when multiple damage functions are used.
Discounting: Account for the relationship between economic growth and discounting; explicitly
recognize uncertainty surrounding discount rates over long time horizons using a Ramsey-like
approach; select parameters to implement this approach that are consistent with theory and
evidence to produce certainty-equivalent discount rates consistent with near-term consumption
rates of interest; use three sets of Ramsey parameters to generate a low, central, and high
certainty-equivalent near-term discount rate, and three means and ranges of SC-C02 estimates;
discuss how the SC-C02 estimates should be combined with other cost and benefit estimates that
may use different discount rates in regulatory analysis.
Additional details on the National Academies' near-term recommendations are provided in Section 2
below. The National Academies' final report also provided longer-term recommendations pertaining to
each module and identified research priorities for addressing these recommendations.
9
-------
EXTERNAL REVIEW DRAFT
In focusing on the four categories above, the National Academies left various topics for future research.
For example, the report pointed to future research that might enable more robust methods of capturing
the benefits of reducing climate risks. While the National Academies report did not explicitly address
methods to account for the disproportionate climate damages that may accrue to lower-income
individuals in SC-GHG estimates, it did outline ways to present evidence on the possible distributional
effects of climate change. The National Academies point to the importance of presenting spatially
disaggregated results that could, in turn, enable methods that would better identify vulnerable
populations and those most at risk. Additional discussion of these dimensions can be found in Section 3.3
of this report.
1.3 Accounting for Global Damages
Benefit-cost analyses of U.S. Federal regulations have traditionally focused on the benefits and costs that
accrue to individuals that reside within the country's national boundaries and that accrue to regulated
industries, regardless of the nationality of the owners of affected physical assets.15 This approach reflects
the fact that for most regulations, those are the two groups primarily affected. It does not reflect any
other scientific, legal, or other rationale. The default recommendation in OMB's Circular A-4 (2003) is that,
an "analysis should focus on benefits and costs that accrue to citizens and residents of the United
States."16 However, OMB Circular A-4 states that when a regulation is likely to have international effects,
"these effects should be reported"; and though the guidance recommends this be done separately, the
guidance also explains that "[different regulations may call for different emphases in the analysis,
depending on the nature and complexity of the regulatory issues."17 The National Academies advised that
"[i]t is important to consider what constitutes a domestic impact in the case of a global pollutant that
could have international implications that affect the United States" (National Academies 2017, p. 13).
There are many reasons, as summarized in this section - and as articulated by OMB and in IWG TSDs (IWG
2010, 2013, 2016a, 2016b, 2021) and the 2015 Response to Comments (IWG 2015) - why the EPA uses
the global value of climate change impacts when analyzing policies that affect GHG emissions, which have
global effects. Courts have upheld the use of global estimates of the SC-GHG, partially in recognition of
15 It is customary in the benefit-cost analyses of U.S. Federal regulations to include the full compliance costs that
accrue to entities operating in the U.S„ even if those costs are fully or partially borne by owners, employees, or
consumers that reside outside of the U.S.
16 OMB's Circular A-4 (2003) provides guidance to Federal agencies on the development of regulatory analysis
conducted pursuant to Executive Order (E.O.) 12866.
17 Circular A-4 also explains "You will find that you cannot conduct a good regulatory analysis according to a formula.
Conducting high-quality analysis requires competent professional judgement." For example, as noted above,
benefit-cost analyses have historically often included compliance costs that are ultimately borne by owners,
employees, or customers that reside outside of the U.S. It may therefore also be relevant that Circular A-4 generally
recommends consistency in the analytical treatment of costs and benefits. ("The same standards of information and
analysis quality that apply to direct benefits and costs should be applied to ancillary benefits and countervailing
risks" (OMB 2003).)
10
-------
EXTERNAL REVIEW DRAFT
the diverse ways in which U.S. interests, businesses, and residents are impacted by global climate
change.18
Unlike many environmental problems where the causes and impacts are distributed more locally, GHG
emissions are a global externality making climate change a true global challenge. GHG emissions
contribute to damages around the world regardless of where they are emitted. The global nature of GHG
pollution and its impacts means that U.S. interests are affected by climate change impacts through a
multitude of pathways and these need to be considered when evaluating the benefits of GHG mitigation
to the U.S. population. For example, climate change will directly impact U.S. interests that are located
abroad (such as U.S. citizens, investments, military bases and other assets, and resources in the global
commons (e.g., through changes in fisheries' productivity and location)). An estimated 9 million U.S.
citizens lived abroad as of 2020,19 and the U.S. direct investment abroad position totaled $6.15 trillion at
the end of 2020.20 Nearly 40% of U.S. pension assets' equity holdings are in foreign stocks.21 Climate
impacts occurring outside of U.S. borders have a direct impact on these U.S. citizens and the investment
returns on those assets owned by U.S. citizens and residents. In addition, the U.S. has over 500 military
sites abroad across 45 foreign countries.22 Climate change impacts (such as sea level rise) occurring in
these locations already affect U.S. military infrastructure and will continue to lead to increased
expenditures to maintain bases' viability and readiness (USGCRP 2018a). Failure to do so can lead to
impacts on mission execution and increased security risks. As one example, "...the United States has
important defense assets located in...the Marshall Islands, and Palau, all of which are vulnerable to these
[climate] hazards. Additionally, competitors such as China may try to take advantage of climate change
impacts to gain influence" (DoD 2021). The timing and severity of climate events are already affecting
missions in some cases and these risks are expected to increase. For example, in the Marshall Islands, the
Ronald Reagan Ballistic Missile Defense Test Site, "a pillar of U.S. Strategic Command" used for detecting
foreign missile launches, may be "uninhabitable in mere decades" according to a recent study conducted
by the Center for Climate and Security's Military Expert Panel (CCS 2018).
The U.S. economy is also inextricably linked to the rest of the world. The U.S. exports over $2 trillion worth
of goods and services a year and imports around $3 trillion.23 According to recent data, over 20% of
18 Zero Zone, Inc. v. Dep't of Energy, 832 F.3d 654, 678-79 (7th Cir. 2016) (rejecting a petitioner's challenge to DOE's
use of a global social cost of carbon in setting an efficiency standard under the Energy Policy and Conservation Act,
holding that DOE had reasonably identified carbon pollution as "a global externality" and concluding that, because
"national energy conservation has global effects, . . . those global effects are an appropriate consideration when
looking at a national policy.").
19 U.S. Department of State's Bureau of Consular Affairs, https://travel.state.gov/content/dam/travel/CA-By-the-
Number-2020.pdf.
20 BEA Direct Investment by Country and Industry 2020, https://www.bea.gov/news/2021/direct-investment-
countrv-and-industry-2020.
21 Based on Thinking Ahead Institute's 2022 Global Pension Assets Study , available at:
https://www.thinkingaheadinstitute.org/research-papers/global-pension-assets-study-2022/.
22 Based on data from U.S. DOD BASE STRUCTURE REPORT-FISCAL YEAR 2018 BASELINE: A SUMMARY OF THE REAL
PROPERTY INVENTORY DATA. See Figure 1.
https://www.acq.osd.mil/eie/Downloads/BSI/Base%20Structure%20Report%20FY18.pdf
23 BEA National Income and Product Accounts Table 1.1.5.
11
-------
EXTERNAL REVIEW DRAFT
American firms' profits are earned on activities outside the country.24 Climate impacts that occur outside
U.S. borders will impact the welfare of individuals and the profits of firms that reside in the U.S. because
of their connection to the global economy. This will occur through the effect of climate change on
international markets, trade, tourism, and other activities. Supply chain disruptions are a prominent
pathway through which U.S. business and consumers are, and will continue to be, affected by climate
change impacts abroad. The impact of international supply chain disruptions can be severe. For example,
severe flooding in Thailand in 2011 disrupted production of components for global companies including
computer disk drives and cars (USGCRP 2018a, DoD 2021). As a result, U.S. consumers faced higher prices
for many electronic goods. The U.S.-based firm Western Digital alone posted $199 million in losses and a
51% drop in hard drive shipments, and U.S. vehicle production had to be temporarily halted or reduced
considerably by at least two manufacturers (USGCRP 2018a). As climate change increases the severity and
frequency of extreme weather events, it increases the risk of supply chain disruptions. Recent research
finds the "probability of a hurricane of sufficient intensity to disrupt semiconductor supply chains may
grow two to four times by 2040" and the "probability heavy rare earths production is severely disrupted
from extreme rainfall may increase 2 to 3 times by 2030."2S
Additional climate change-induced international spillovers can occur through pathways such as damages
across transboundary resources, economic and political destabilization, and global migration that can lead
to adverse impacts on U.S. national security, public health, and humanitarian concerns (DoD 2014, CCS
2018). As articulated in a landmark 2007 study by retired three- and four-star Generals and Admirals - and
echoed in the Department of Defense's (DoD) 2014 Quadrennial Defense Review - the projected effects
of climate change act as a "threat multiplier" that will exacerbate many stressors and instabilities that
already exist in some of the most volatile regions of the world (CNA 2007, DoD 2014). A follow-up study
emphasized that beyond being a threat multiplier, climate change impacts will also "serve as catalysts for
instability and conflict" (CNA 2014). For example, in Sub-Saharan Africa regional environmental stressors
exacerbated by climate change can help to transform resource competition into ethnopolitical conflict
and enable the involvement of transnational terrorist groups (such as Al Qaeda in the Islamic Maghreb
(AQIM) in Mali in 2012) (CNA 2014). More recent DoD reports reiterate these concerns, concluding that
the impacts of climate change "could stress economic and social conditions that contribute to mass
migration events or political crises, civil unrest, shifts in the regional balance of power, or even state
failure," with results that affect the national interests of the U.S. (DoD 2021). The key takeaway from the
National Intelligence Council's (NIC) 2021 National Intelligence Estimate is that "climate change will
increasingly exacerbate risks to US national security interests as the physical impacts increase and
geopolitical tensions mount about how to respond to the challenge" (NIC 2021). The NIC finds "the
increasing physical effects of climate change are likely to exacerbate cross-border geopolitical flashpoints
as states take steps to secure their interests", and as intensifying physical effects "out to 2040 and beyond
will be most acutely felt in developing countries, which we assess are also the least able to adapt to such
changes.,.[t]hese physical effects will increase the potential for instability and possibly internal conflict in
24 Bureau of Econ. Analysis, National Income and Product Accounts Table 6.16D,
httpsi//apps,bea,gov/iTable/iTable,cfm?reqid=19&step=2#reqid=19&step=2&isuri=l&1921=survey.
25 https://www.mckinsey.com/business-functions/sustainability/our-insights/could-climate-become-the-weak-link-
in-your-supply-chain.
12
-------
EXTERNAL REVIEW DRAFT
these countries, in some cases creating additional demands on US diplomatic, economic, humanitarian,
and military resources" (NIC 2021).
As described by the National Academies (2017), to correctly assess the total damages to U.S. citizens and
residents, one must account for these spillover effects on the U.S. For more discussion and examples of
international spillover effects, including the ways that climate change spillovers are exacerbating existing
risks and creating new security, health, and humanitarian challenges for U.S. interests, see for example,
NIC (2021), DoD (2021), USGCRP (2018a), Freeman and Guzman (2009), Howard and Livermore (2021),
Schwartz (2021), and IPCC (2022).
A robust estimate of climate damages to U.S. citizens and residents that accounts for the myriad of ways
that global climate change reduces the net welfare of U.S. populations does not currently exist in the
literature. At present, the only quantitative characterizations of U.S. damages from GHG emissions are
based on the share of modeled damages that physically occur within U.S. national borders as represented
in current lAMs. Such estimates provide an underestimate of the climate change damages to the citizens
and residents of the U.S. because these models do not fully capture the range of climate change impacts
and exclude important regional interactions and spillovers discussed above. In addition, a 2020 GAO study
observed that "[according to the National Academies, the integrated assessment models were not
premised or calibrated to provide estimates of the social cost of carbon based on domestic damages, and
more research would be required to update the models to do so" (GAO 2020). Further, the National
Academies observed that existing models "focus primarily on global estimates and do not model all
relevant interactions among regions....More thoroughly estimating a domestic SC-C02 would therefore
need to consider the potential implications of climate impacts on, and actions by, other countries, which
also have impacts on the United States" (National Academies 2017, p. 13).
In addition to accounting for the ways that climate change impacts occurring outside of U.S. borders affect
U.S. populations, it is also important to consider how changes in U.S. emissions affect the GHG emissions
of other countries. This is relevant because the global nature of greenhouse gases means that damages
caused by a ton of emissions in the U.S. are felt globally and that a ton emitted in any other country harms
those in the U.S. This is a classic public goods problem because each country's reductions benefit everyone
else and no country can be excluded from enjoying the benefits of other countries' reductions. As
discussed by EPA and other members of the IWG in the 2015 response to comments (IWG 2015), in this
situation, the only way to achieve an efficient allocation of resources for emissions reduction on a global
basis—and so benefit the U.S. and its citizens and residents —is for all countries to consider estimates of
global marginal damages. Therefore, international GHG mitigation activities taken in response to U.S.
policies that reduce emissions will also provide a benefit to U.S. citizens and residents. A wide range of
scientific and economic experts have emphasized the issue of reciprocity as support for assessing global
damages of GHG emissions in domestic policy analysis (e.g., Kopp and Mignone 2013, Pizer et al. 2014,
Howard and Schwartz 2017, Pindyck 2017, 2021, Revesz et al. 2017, Carleton and Greenstone 2022).
Kotchen (2018) demonstrates how a country's decision to internalize global damages in domestic
policymaking can be individually rational (i.e., in the country's own self-interest) because of the
13
-------
EXTERNAL REVIEW DRAFT
reciprocally induced emissions reductions occurring in other countries.26 Carleton and Greenstone (2022)
discuss examples of how accounting for global damages in past U.S. regulatory analyses may have
contributed to additional international action. Houser and Larson (2021) estimate that under the Paris
Agreement, other countries pledged to reduce 6.1 to 6.8 tons for every ton pledged by the U.S.
Assessing global marginal damages of GHG emissions in U.S. analyses of regulatory and other actions
allows the U.S. to continue to actively encourage other nations, including emerging economies, to also
assess global climate damages of their policies and to take significant steps to reduce emissions. Many
countries and international institutions have either already explicitly adapted the IWG's estimates of
global damages in their domestic analyses (e.g., Canada27, Israel28), developed their own estimates of
global damages (e.g., Germany29), or have taken note of the IWG estimates in their assessments of climate
policies (e.g., India's National Green Tribunal30, the Australian Capital Territory31, New Zealand32, and the
International Monetary Fund33). In 2016, Mexico announced its intention to "align approaches [with the
26 Kotchen (2018) not only details the "efficiency argument in support of all countries internalizing the GSCC [global
social cost of carbon] for domestic policy," but Kotchen (2018) also introduces the concept of countries having a
"preferred" social cost of carbon (PSCC) for setting global climate policy and shows that all countries' PSCC exceeds
the marginal damages to its own populations. The PSCC is shaped by a country's expected benefits from other
countries' emission reductions. Kotchen's study shows that in some countries the PSCC can even exceed the value
of global marginal damages (e.g., in small island nations for whom the benefits of stringent worldwide abatement
based on a high PSCC would exceed the increase in its own abatement costs due to a high PSCC). Kotchen offers
illustrative estimates of the PSCC for several countries and regions based on research using a regionalized version of
the DICE model (Nordhaus 2015). In this analysis Kotchen finds the U.S. PSCC to be nearly 75% of the value of global
marginal damages. And as Kotchen has further clarified, "depending on the U.S. government's diplomatic strategies,
its expectations of international reciprocity, and the international distribution of costs, it can be rational for the
United States to adopt the full global SCC values for use in policy-making." (Kotchen 2021, comment number OMB-
2021-0006-0018, available at: https://www,regulations,gov/comment/QMB~2021-0006-Q018). Such arguments for
accounting for the global value of climate change impacts in analysis of policies affecting U.S. GHG emissions, based
on the U.S. derived benefits from reciprocally induced emission reductions elsewhere, are distinct from and
additional to arguments above based on spillover effects and U.S. interests beyond our geographic borders.
27 Envt. & Climate Change Canada, Technical Update to Environment and Climate Change Canada's Social Cost of
Greenhouse Gas Estimates at 13 (2016), http://publications.gc.ca/collections/collection_2016/eccc/Enl4-202-2016-
eng.pdf.
28 Israel Ministry of Envtl. Protection, Green Book on External Costs of Air Pollutants (2020),
https://www.gov.il/BlobFolder/publicsharing/pc_external_costs_of_air_pollution/he/public_comments_2020_Ext
ernal _air_pollution_costs_pc_accessible.docx
29 See GAO (2020) for a discussion of Germany's SC-GHG values.
30 Additional Report of Committee on Environmental Damage Assessment Due to Air Pollution Caused on Account
of Explosion & FIR...in the Matter of OA No. 22 of 2020, at 12 (2020),
https://cpcb.nic.in/NGT/ADDITIONAL REPORT Air OA 22 of 2020-SEP~2020.pdf.
31 Austl. Cap. Terr. Env't, Plan, and Sustainable Dev. Directorate, ACT Climate Change Strategy 2019-2025 (2019),
https://perma.cc/487H-BHHC: see also Rovingstone Adv. Pty Ltd., A Social Cost of Carbon for the ACT, Report
Prepared for the ACT Government (2021),
https://www.climatechoices.act.gov.au/ data/assets/pdf file/0006/1864896/a-social-cost-of-carbon-in-the-
act.pdf (recommending adopting the U.S. valuations of the social cost of carbon).
32 Ministry of Transp., Preliminary CBAfor Vehicle Fuel Efficiency Standard (2018), https://perma.cc/Y7SS-3AG2.
33 Benedict Clements et al., IMF, Energy Subsidy Reform: Lessons and Implications at 8 (2013),
https://www.imf.org/en/Publications/Books/lssues/2016/12/31/Energy-Subsidy-Reform-Lessons-and-
lmplications-40410 .
14
-------
EXTERNAL REVIEW DRAFT
U.S. and Canada] to account for the social cost of carbon and other greenhouse gas emissions when
assessing the benefits of emissions-reducing policy measures"34, and references to global estimates of
climate damages can be found in Mexican regulatory analyses in 2017.35 However, the bilateral technical
discussions to help implement the announced plan did not occur over 2017-2021 during the time U.S.
federal regulatory analyses stopped focusing on SC-GHG estimates that reflect global damages.
EPA and other members of the IWG found previously and restated in their February 2021TSD that because
of the distinctive global nature of climate change that analysis of Federal regulations and other actions
should center on a global measure of SC-GHG (IWG 2021). This is the same approach that was
recommended by OMB and other members of the IWG and used by EPA and other agencies in regulatory
analyses from 2009 to 2016. It is also consistent with guidance in OMB Circular A-4 that "[different
regulations may call for different emphases in the analysis, depending on the nature and complexity of
the regulatory issues," and National Academies' guidance that "it is important to consider what
constitutes a domestic impact in the case of a global pollutant that could have international implications
that impact the United States." In the case of this global pollutant, for all the reasons articulated in this
section, the assessment of global net damages of GHG emissions allows analysts to fully disclose and
contextualize the net climate benefits of domestic policies that reduce GHG emissions. The extent that
analysis relying on these SC-GHG estimates is considered in setting the stringency of future regulatory
actions and other policy decisions would be guided by the statutes under which those decisions are
promulgated.36,37 The EPA will continue to review developments in the literature, including more robust
methodologies for estimating the magnitude of the various direct and indirect damages to U.S.
populations from climate impacts occurring abroad and reciprocal international mitigation activities.
34 https://obamawhitehouse.archives.gov/the-press-office/2016/06/29/leaders-statement-north-american-
climate-clean-energy-and-environment
35 See, e.g., Secretaria del Medio Ambiete y Recursos Naturals, Que Establece Los Limites Maximos Permisible de
Emision de Monoxido de Carbono....Anexo: Beneficios (2017), https://perma.cc/N6YHZYTM (citing the Working
Group's estimates); Secretaria del Medio Ambiente y Recursos Naturales, Aviso Mediante elCual Se Dan a Conocer
los Parametros para el Calculo de las Emisiones de Bioxido de Carbono (C02) en los Vehiculos Automotores Ligeros
Nuevos con Peso Bruto Vehicluar Que No Exceda Los 3857 Kilogramos, Que Utilizan Gasolina o Diesel como
Combustible Cuyo Ano-Modelo SEA 2017 (June 15, 2016), https://perma.cc/HV8H-62GU (referencing "beneficios
globales para las emisiones evitadas de C02").
36 For example, as the Supreme Court stated in Motor Vehicle Manufacturers Ass'n. v. State Farm Mutual Auto. Ins.
Co., 463 U.S. 29, 41-43 (1983): "Normally, an agency rule would be arbitrary and capricious if the agency has relied
on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the
problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so
implausible that it could not be ascribed to a difference in view of the product of agency expertise." This requires
agencies to "examine the relevant data and articulate ... a rational connection between the facts found and the
choice made."
37 Public comments received on the February 2021 TSD argue that key U.S. statutes explicitly require or allow
consideration of global climate damages in decision making. See, e.g., discussion within comments submitted by the
Institute for Policy Integrity and the attachments and literature cited therein (comment number OMB-2021-0006-
0074, available at: https://www.regulations.gov/comment/OMB-2021-0006-0074) (discussing, for example, how
the National Environmental Policy Act requires that "public laws of the United States shall be interpreted and
administered in accordance with the policies set forth in this chapter, and all agencies of the Federal Government
shall...recognize the worldwide and long-range character of environmental problems").
15
-------
EXTERNAL REVIEW DRAFT
2 Methodological Updates
The SC-GHG is commonly estimated with the use of integrated assessment models (1AM). In the broadest
sense lAMs are "approaches that integrate knowledge from two or more domains into a single
framework" (Nordhaus 2017a). The literature on "lAMs" is vast and spans many sciences, e.g., earth
sciences, biological sciences, environmental engineering, economics, and sociology. lAMs have been used
to study environmental problems and their connection to economic systems for nearly 40 years (e.g.,
Freeman 1979, 1982; Mendelsohn 1980; Nordhaus 1993a, 1993b). The National Academies defined lAMs
used to study climate change as "computational models of global climate change that include
representation of the global economy and greenhouse gas emissions, the response of the climate system
to human intervention, and impacts of climate change on the human system" (National Academies 2017).
These lAMs vary significantly in structure, geographic resolution, the degree to which they capture
feedbacks within and between natural and economic systems and include valuation, and application.
Those that are used to estimate the SC-GHG are reduced-form in nature and combine climate processes,
economic growth, and feedback between the climate and the global economy into a single modeling
framework, providing a holistic view of the system, and include a valuation of climate change damages.
Other climate change lAMs, often called detailed-structure lAMs, include structural representations of
the global economy with a high level of regional and sectoral detail, and were originally developed for
analyzing the impact of policy and technology on greenhouse gas emissions (e.g., Edmonds and Reilly,
1983). These types of lAMs are increasingly being used to examine different climate change impact sectors
and interactions between sectors and regions but do not yet comprehensively link physical impacts to
monetized economic damages as needed for SC-GHG estimation (National Academies 2017).
As illustrated in Figure 2.1, the steps necessary to estimate the SC-GHG with a climate change 1AM can
generally be grouped into four modules: socioeconomics and emissions, climate, damages, and
discounting. The emissions trajectories from the socioeconomic module are used to project future
temperatures in the climate module. The damage module then translates the temperature and other
climate endpoints (along with the projections of socioeconomic variables) into physical impacts and
associated monetized economic damages, where the damages are calculated as the amount of money the
individuals experiencing the climate change impacts would be willing to pay to avoid them. To calculate
the marginal effect of emissions, i.e., the SC-GHG in year t, the entire model is run twice - first as a
baseline and second with an additional pulse of emissions in year t. After recalculating the temperature
effects and damages expected in all years beyond t resulting from the adjusted path of emissions, the
losses are discounted to a present value in the discounting module. Much of the uncertainty in the
estimation process can be incorporated using Monte Carlo techniques by taking draws from probability
distributions that reflect the uncertainty in parameters.
The SC-GHG estimates used by the EPA and many other federal agencies since 2009 have relied on an
ensemble of three widely used lAMs: Dynamic Integrated Climate and Economy (DICE) (Nordhaus 2010);
Climate Framework for Uncertainty, Negotiation, and Distribution (FUND) (Anthoff and Tol 2013a, 2013b);
and Policy Analysis of the Greenhouse Gas Effect (PAGE) (Hope 2013). In 2010, the IWG harmonized key
inputs across the lAMs, but all other model features were left unchanged, relying on the model
developers' best estimates and judgments. That is, the representation of climate dynamics and damage
functions included in the default version of each 1AM as used in the published literature was retained.
16
-------
EXTERNAL REVIEW DRAFT
The SC-GHG estimates in this report no longer reiy on the three lAMs (i.e., DICE, FUND, and PAGE) used
in previous SC-GHG estimates. Instead, this report uses a modular approach to estimating the SC-GHG,
consistent with the National Academies' near-term recommendations. That is, the methodology
underlying each component, or module, of the SC-GHG estimation process draws on expertise from the
scientific disciplines relevant to that component. Under this approach, each step in the SC-GHG estimation
improves consistency with the current state of scientific knowledge, enhances transparency, and allows
for more explicit representation of uncertainty. This section discusses the methodological updates in each
of the four National Academies' recommended modules in addition to other updates in the modeling
framework, such as the explicit incorporation of risk aversion.
Figure 2.1: The Four Components of SC-GHG Estimation38
Source: National Academies of Sciences, Engineering, and Medicine (2017)
38 In Figure 2.1, the different shading for non-monetized impacts signifies that those impacts are outside the scope
of the modeling and is not intended to suggest that non-monetized impacts are less relevant than monetized
impacts,
17
-------
EXTERNAL REVIEW DRAFT
2.1 Socioeconomic and Emissions Module
The first step in the SC-GHG estimation process is the development of projections of socioeconomic
variables and GHG emissions at the spatial and temporal resolution required by the climate and damage
modules. Socioeconomic trajectories are closely tied to climate damages because, holding all else equal,
increases in population and income will increase GHG emissions and lead to a greater willingness to pay
to avoid climate change impacts. Within the SC-GHG estimation process, projections of GHG emissions
serve as inputs to the climate module, and projections of GDP and population serve as inputs to the
damage function and discounting modules. Disaggregation of these inputs is required when greater spatial
and/or temporal resolution is required for the damage module. Finally, because GHG emissions and their
effects are long lived, it is necessary to project these variables far into the future and address the many
complex uncertainties associated with such projections.
SC-GHG estimates used in the EPA's analyses to date have relied on the socioeconomic and emissions
projections selected by the IWG in 2010. The IWG elected to use socioeconomic and emissions projections
based on deterministic scenarios that, at the time, were recently updated, grounded in multiple well-
recognized models, used in climate policy simulations, and spanned a plausible range of outcomes for
these variables. The socioeconomic and emission projections included five deterministic reference
scenarios based on the Stanford Energy Modeling Forum EMF-22 modeling exercise (Clarke, et al. 2009;
Fawcett, et al. 2009). Four of these scenarios represented business-as-usual (BAU) trajectories, while the
fifth scenario assumed that substantive actions would be adopted to reduce future emissions. The SC-
GHG estimates gave equal weight to each scenario. The IWG also elected to use a time horizon extending
to 2300 to try to capture the vast majority of discounted climate damages. Running the lAMs through
2300 required extrapolations of the projections after 2100, the last year available for projections from the
EMF-22 models.39
The National Academies 2017 final report included several recommendations for how to approach
updating the socioeconomic module to reflect newer information. The National Academies (2017)
recommended that socioeconomic scenarios used to estimate the SC-GHG should: "extend far enough in
the future to provide inputs for estimation of the vast majority of discounted climate damages"; "take
account of the likelihood of future emissions mitigation policies and technological developments";
"provide the sectoral and regional detail in population and GDP necessary for damage calculations"; and,
"to the extent possible...incorporate feedbacks from the climate and damages modules that have a
significant impact on population, GDP, or emissions" (National Academies, 2017, p. 15). The National
Academies acknowledged that it would not be possible to meet all these criteria in the near term.
39 These inputs were extrapolated from 2100 to 2300 as follows: (1) population growth rate declines linearly,
reaching zero in the year 2200; (2) GDP/ per capita growth rate declines linearly, reaching zero in the year 2300; (3)
the decline in the fossil and industrial carbon intensity (CO2/GDP) growth rate over 2090-2100 is maintained from
2100 through 2300; (4) net land use CO2 emissions decline linearly, reaching zero in the year 2200; and (5) non-CC>2
radiative forcing remains constant after 2100. See IWG (2010) for more discussion of each of these assumptions. In
2016, the IWG added more specificity to the assumptions regarding post-2100 baseline CFU and N2O emissions in
order to calculate SC-CH4 and SC-N2O. See IWG (2016b) for more details.
18
-------
EXTERNAL REVIEW DRAFT
However, the report suggested initial steps for how to achieve these goals and overcome several
limitations in the methodology used to date. Specifically, they recommend:
(1) working with demographers to extend existing probabilistic population projections beyond 2100,
validated and adjusted by expert judgment;
(2) generating probabilistic projections of annual growth rates of per-capita GDP with an appropriate
statistical technique, informed by expert judgment;
(3) using a set of emissions projections generated by an expert elicitation, conditioned by the set of
scenarios of future population and income; and
(4) developing projections of sectoral and regional GDP and regional population using scenario
libraries, published projections, detailed-structure economic models, or other sources.
Resources for the Future Socioeconomic and Emissions Projections (RFF-SPs). Based on a review of
available sources of long-run projections for socioeconomic variables and GHG emissions necessary for
damage calculations, the socioeconomic and emissions projections recently developed under the
Resources for the Future Social Cost of Carbon Initiative (Rennert et al. 2022a) stand out as being most
consistent with the National Academies' recommendations. These projections (hereafter collectively
referred to as the RFF-SPs) are an internally consistent set of probabilistic projections of population, GDP,
and GHG emissions (C02, CH4, and N20) to 2300. Consistent with the National Academies'
recommendation, the RFF-SPs were developed using a mix of statistical and expert elicitation techniques
to capture uncertainty in a single probabilistic approach, taking into account the likelihood of future
emissions mitigation policies and technological developments, and provide the level of disaggregation
necessary for damage calculations. Unlike other sources of projections, they provide inputs for estimation
to 2300 without further extrapolation assumptions. Conditional on the modeling conducted for this
report, this time horizon is far enough in the future to capture the majority of discounted climate damages
(see discussion in Section 3). Including damages beyond 2300 would increase the estimates of the SC-
GHG. As discussed in Section 2.5, the use of the RFF-SPs allows for capturing economic growth uncertainty
within a calibrated utility approach to discounting.
The RFF-SPs were developed as follows. The country-level population projections are based on Raftery
and Sevcikova's (2021) extension to the Bayesian methodology that the United Nations has used since
2015 for population forecasting (UN 2015). The extension combines the United Nations statistical
approach with expert review and elicitation to extend the projections to 2300.
The economic growth projections extend research by Muller et al. (2022), who refined a foundational
statistical methodology for generating internally consistent long-term probabilistic growth projections at
the country level. Specifically, Muller et al. were the first to extend the approach provided in Muller and
Watson (2016) for estimating global economic growth. These probabilistic economic growth projections
are combined with the results of a formal expert elicitation of 10 leading growth economists, conducted
individually via videoconference in 2019-2020. As part of the elicitation, the experts first quantified their
uncertainty for a set of calibration questions, the results of which were used to performance-weight the
experts in their final combination. The elicitation focused on quantifying uncertainty for a representative
frontier of economic growth in OECD countries. The combined results from the experts were then used
to inform econometric projections based on the Muller et al. (2022) model of an evolving frontier (also
based on the OECD), in turn providing country-level, long-run probabilistic projections.
19
-------
EXTERNAL REVIEW DRAFT
GHG emissions are projected using expert elicitation techniques.40 A separate panel of 10 experts41 was
asked to provide uncertainty quantiles for four emissions variables in five benchmark years and to indicate
the sensitivity of the C02 emissions responses to five GDP per capita trajectories.42 Responses were
requested under a case incorporating views about changes in technology, fuel use, and other conditions,
including the evolution of future policy.43 The projections from the RFF-SPs represent a state-of-the-art
set of probabilistic socioeconomic and emissions scenarios based on high-quality data, robust statistical
techniques, and expert elicitation. In addition, they cover a sufficient time horizon for estimating the SC-
GHG and incorporate uncertainty over future background policies. As such, the RFF-SPs are consistent
with the National Academies' recommendations on socioeconomic and emissions scenarios.
Other Sources of Socioeconomic and Emissions Projections. The RFF-SPs represent a significant
advancement over the now outdated and deterministic EMF-22 scenarios and offer improvements over
other recently developed socioeconomic and emissions projections. The only other probabilistic
projections identified in this review are a library of scenarios generated using MIT's Emissions Prediction
and Policy Analysis (EPPA) Model, coupled with expert elicitation (Abt Associates 2012, Marten 2014).
These projections have the advantage that they rely on a comprehensive computable general equilibrium
(CGE) model that captures key feedbacks and interdependencies across the sources of uncertainty.
However, they were generated in 2012 and do not incorporate changes in the economy, emissions trends,
and policies adopted over the past decade.
Other socioeconomic and emissions projections developed since the EMF-22 exercise are deterministic
and do not provide global projections over a time horizon sufficient for SC-GHG estimation. The most
prominent deterministic projections come from the database of Shared Socioeconomic Pathways (SSPs)
and Representative Concentration Pathways (RCPs).44 The SSPs and RCPs are the result of a scenario
development effort that started in the late 2000s to replace the Special Report on Emission Scenarios
(SRES) scenarios from the 1990s (used in the IPCC Third Assessment Report). The two components, SSPs
and RCPS, were designed to be complementary. RCPs set pathways for GHG concentrations and,
40 For greenhouse gases other than CO2, CFU, and N2O that are needed as inputs to FaIR (e.g., CF4, C2F6, HFCs, CFCs,
HCFCs), emissions are projected using SSP2-4.5 from AR6. This scenario is also used to calibrate FalRl.6.2 and is
nearest to the RFF-SP median emissions for carbon dioxide and methane.
41 The experts were nominated by their peers and/or by members of the RFF Scientific Advisory Board, and have
expertise in, and have undertaken, long-term projections of the energy-economic system under a substantial range
of climate change mitigation scenarios. More information about the experts is provided in Rennert et al. (2022a).
42 Specifically, the experts were asked to provide quantiles (minimum, 5th, 50th, 95th, maximum, as well as
additional percentiles at the expert's discretion) for (1) fossil fuel and process-related CO2 emissions; (2) changes in
natural CO2 stocks and negative-emissions technologies; (3) CFU; and (4) N2O, for five benchmark years: 2050, 2100,
2150, 2200, and 2300.
43 See Rennert et al. (2022a) for a detailed discussion of the survey methodology and the full elicitation protocol.
44 Some organizations also regularly produce forecasts of key socioeconomic variables and emissions, but these tend
to be only for a few decades or some countries or regions (e.g., IEA, EIA). Some 1AM researchers have constructed
deterministic projections using disparate sources. For example, the inputs used in the latest version of the DICE
model, DICE 2016, include economic growth projections based on a survey by Christensen et al. (2018), population
data from the United Nations, and CO2 emissions projections from Carbon Dioxide Information Analysis Center, with
simple assumptions for extending each series post-2100 (Nordhaus 2017b).
20
-------
EXTERNAL REVIEW DRAFT
effectively, the amount of warming that could occur by the end of the century.45 Many possible socio-
economic futures may lead to the same RCP, so the SSPs are scenarios of projected socioeconomic global
changes through 2100, based on potential future changes in quantitative elements, including population,
education, urbanization, GDP, and technology. There are five SSPs, each consisting of a set of quantified
measures of development and an associated narrative storyline. The storylines provide a qualitative
description of plausible future conditions that drive the quantitative elements. Pairings of these
illustrative SSP scenarios with RCPs have been widely used by the IPCC, the global scientific community,
and researchers spanning a wide range of disciplines. For modeling exercises requiring emissions
projections beyond 2100, such as for SC-GHG estimation, researchers commonly use emissions extensions
provided by the Reduced Complexity Model Intercomparison Project (Nicholls et al. 2020). When
population and economic growth projections beyond 2100 are necessary, researchers have used various
methods to extend the SSPs to 2300, ranging from simple extrapolation assumptions (e.g., CIL 2022,
Benveniste et al. 2020)46 to empirically derived projection methods (e.g., Kikstra et al. 2021).47 Use of
deterministic scenarios, such as the SSP-RCP pairings, would prevent the SC-GHG estimates from
capturing important aspects of climate risk, including its relationship to broader socioeconomic
uncertainty, and from valuing that risk in a way that is consistent with economic theory and observed
human behavior related to risk aversion.
Figure 2.1.1 and Figure 2.1.2 present the RFF-SP projections of population and economic growth through
2300. These figures also include a comparison to the SSPs that have been used in IPCC reports and other
applications.48 The SSP projections beyond 2100 (dashed) are based on the extrapolation method used in
Benveniste et al. (2020) for all SSPs. To illustrate the sensitivity to this assumption, projections based on
the SSP extrapolation method employed by the Climate Impact Lab (CIL 2022) are also displayed for SSP2
and SSP3. The mean (black solid line) and median (black dotted line) of the RFF-SP population projections
follow an increasing trajectory through 2100, consistent but slightly higher than the SSP2 and SSP5
projections, peaking at 11.2 billion people (Figure 2.1.1). This is followed by a slow decline to under 10
billion by 2300. Except for SSP1—which follows an optimistic storyline on sustainability and stabilizing
population—all the SSPs lie within the RFF-SP distribution throughout the modeling horizon—with SSP3
in the upper tails of the distribution.
45 Four RCPs were used in the IPCC Fifth Assessment Report (2014a) that span a range of radiative forcing (watts per
m2) in 2100 and are named for that forcing above the pre-industrial level (RCP2.6, RCP4.5, RCP6.0 and a high-end
no-mitigation RCP8.5). The SSPs took longer to develop. The SSPs were published in 2016 and updated in 2018. The
are available at: https://tntcat.nasa.ac.at/SspDb/dsd?Action=htmlpage&page=10. The SSPs and some additional
RCPs are being used in the IPCC Sixth Assessment Report (2021a). The three additional RCPs include RCP1.9 (which
focuses on limiting warming to below 1.5C), RCP3.4 (an intermediate pathway between RCP2.6 and RCP4.5), and
RCP7.0 which represents medium-to-high end of emissions range and is a baseline outcome rather than a mitigation
target.
46 In the components of their modeling that require extrapolation of GDP and population beyond 2100, when using
SSPs, Climate Impact Lab (CIL 2022) modeling assumed GDP per capita growth and the level of global
population remain constant at 2100 levels through 2300. Benveniste et al. (2020) generates country level extensions
to 3000, based on the assumption that population growth declines linearly to 0 in 2200, and is held constant
thereafter; GDP per capita growth is assumed to decline linearly reaching 0 in 2300.
47 Kikstra et al. (2021) develop regional extensions based on the assumption that regional GDP per capita and
population growth rates (in PAGE model regions) converge toward the global mean.
48 Figures 2.1.1 and 2.1.2 contain all Tier 1 SSPs from IPCC AR6. Tier 2 scenarios, such as SSP4, were not considered.
21
-------
EXTERNAL REVIEW DRAFT
Figure 2.1.1: Global Population under RFF-SPs and SSPs, 1950-2300
35
30
o 25
o 20
¦M
JU
3
Q_
S. 15
"ro
-Q
_o
u 10
5
1950 2020 2100 2200 2300
Year
— Historical (AR6) — RFF-SPs — SSP1 — SSP2 — SSP3 — SSP5
Benveniste (2020) ¦ — ¦ CIL(2022)
RFF-SP projections based on RFF-SPs (Rennert et al. 2022a). Black lines represent mean (solid) and median (dotted) lines along
with 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges. SSP data through 2100 from International Institute for
Applied Systems Analysis (IIASA) SSP Database (Riahi et al. 2017). SSPs beyond 2100 (dashed) are based on two recent
extrapolation methods: Benveniste et al. (2020) and CIL (2022).
Figure 2.1.2 presents the economic growth projections from the RFF-SPs along with comparisons to the
SSPs in AR6.49 The mean (black solid line) and median (black dotted line) economic growth rates are
relatively flat until 2100 at 1.6% and then decline through-out the next century. The mean economic
growth rate levels off again after 2200 at 1.1%. The RFF-SP economic growth projections are lower but
most consistent with SSP2, i.e., the "middle of the road" scenario in which economic trends follow
historical patterns. All the SSP storylines lie within the RFF-SP distribution throughout the modeling
horizon. One notable difference between the RFF-SPs and the SSPs is the high near-term growth rates in
the SSPs. Published in 2017, the SPPs economic growth projections are based on historical data through
2010. Between 2005 and 2010 the historical average annual growth rate was nearly 3%. The SSPs
predicted an average annual growth rate between 2010 and 2019 of 2.89-2.96% (Riahi et al. 2017),
whereas in the past decade average global per capita growth rates have been closer to 2% (World Bank
49 The growth rates (and the uncertainty bounds around the RFF-SPs) shown in Figure 2.1.2 are plotted in a time-
averaged manner to accurately present the underlying year-on-year correlations that exist within each
scenario/storyline.
22
-------
EXTERNAL REVIEW DRAFT
2021). The estimated growth-rates in the RFF-SPs are long-run growth rates, built to eliminate short-run
fluctuations.
Figure 2.1.2: Long-run Projections of Growth in Global Income per Capita under RFF-SPs and SSPs, 2020-
2300
ro
CD
O
(N
O
(N
CD
C*L
o
I—
u
Q_
ro
U
CL
Q-
Q
U
00
CD
i—
-------
EXTERNAL REVIEW DRAFT
the RFF-SPs more closely implement the near-term recommendations from the National Academies on
economic growth and population projections than do the SSPs.
In the SSPs and the mean RFF-SPs, global emissions of C02 peak at some point this century and decline
toward zero emissions (in some cases negative emissions). These emission peaks for the SSPs are based
on simplistic assumptions about net emissions reaching zero in 2250. The RFF-SP projections are based
on expert elicitation, where the experts were asked to incorporate their views on the evolution of future
policy. This is consistent with the National Academies' (2017) recommendations to "take account of the
likelihood of future emissions mitigation policies." Because the RFF-SPs are probabilistic they reflect the
uncertainty in future policy and when this peak would occur. In the mean RFF-SP projection the peak
occurs this decade. In some of the higher emissions scenarios this peak in emissions does not occur until
near the end of the century.
Figure 2.1.3 presents the RFF-SP projections for C02 emissions through 2300 along with a comparison to
a range of SSP-RCPs from AR6 (Figure A.5.1 and Figure A.5.2 in the Appendix present the same information
for CH4 and N20 emissions through 2300). For SSP-RCP pairings presented in the figure, emissions
projections beyond 2100 are based on the commonly used extensions provided by the Reduced
Complexity Model Intercomparison Project (Nicholls et al. 2020). The post-2100 SSP projections are based
on simplistic assumptions about when global emissions reach zero (2055 for SSP1-1.9, 2075 for SSP1-2.6,
2250 for SSP2-4.5, SSP3-7.0, and SSP5-8.5) and how global emissions reach this point after 2100. In the
mean RFF-SPs (black solid line) global C02 emissions continue to rise in the term near but peak at 42 GtC02
in 2026. Both the RFF-SP median and the mean track closely with SSP2-4.5, which is often described as a
"middle of the road" SSP storyline. The SSP5-8.5 projection is the only SSP-RCP pairing with C02 emissions
projections outside the 1st to 99th percentile range of RFF-SPs. The RCP8.5 emissions scenario is a high
emissions scenario in absence of climate change policies (Riahi et al. 2017).50 As mentioned above, the
RFF-SPs explicitly account for the likelihood of future climate policies.51 While the SSP-RCP scenarios offer
plausible storylines that imbed these assumptions within their trajectories, the RFF-SPs have a significant
advantage in that they assign probabilities to these future policies and their outcomes, account for
adoption of cleaner technologies and fuel sources, and explicitly link socioeconomic growth scenarios to
emissions.52
50 While all the RCP emissions scenarios peak and begin to decline by, or shortly after, the end of the century, it is
important to note that CO2 concentrations, and therefore temperatures, will not stabilize until CO2 emissions decline
to zero (Matthews and Caldeira 2008).
51 Specifically, Rennert et al (2022a) states: "...experts viewed low economic growth as likely to reduce emissions
overall but also lead to reduced global ambition in climate policy and slower progress to decarbonization. For median
economic growth conditions, experts generally viewed policy and technology evolution as the primary driver of their
emissions distributions, often offering a median estimate indicating reductions from current levels but with a wide
range of uncertainty. Several experts said high economic growth would increase emissions through at least 2050,
most likely followed by rapid and complete decarbonization, but with a small chance of substantial continued
increases in emissions."
52 Throughout all stages of the SC-GHG modeling process, we compared the intermediate and final outputs across
the SSP-RCP socioeconomic and emissions storylines and the RFF-SP probabilistic scenarios. In all cases (global mean
surface temperature, sea level rise, and even the final SC-GHG estimates) the RFF-SPs lie within the full range of the
SSP-RCP storylines and are most consistent with the SSP2-RCP4.5 pairing.
24
-------
EXTERNAL REVIEW DRAFT
Figure 2.1.3: Net Annual Global Emissions of Carbon Dioxide (C02) under RFF-SPs and SSPs, 1900-2300
u
-t—>
u
in
c
o
\n
U)
100
* 50
-------
EXTERNAL REVIEW DRAFT
2.2 Climate Module
The next step in the SC-GHG estimation process is to estimate the effect of emissions on physical climate
variables, such as temperature, and to ensure that the outputs from the climate model are at the spatial
and temporal resolution required by the damage module. This means the climate module must:
(1) translate GHG and other forcing agent emission projections into atmospheric concentrations,
accounting for the uptake of C02 by the land biosphere and the ocean and the removal of other
greenhouse gases through atmospheric reactions, deposition, and/or other mechanisms;
(2) translate concentrations of greenhouse gases and other forcing agents into radiative forcing;
(3) translate forcing into global mean surface temperature response, accounting for heat uptake by
the ocean, and
(4) generate other climatic variables, such as sea level rise (SLR), that may be needed by the damage
module.53
Together, with the projections of associated socioeconomic variables, the results from the climate module
serve as inputs to the damage module.
As discussed in section 1.1, the methodology underlying SC-GHG estimates used in the EPA's analyses to
date has included a representation of climate and other earth system dynamics as provided in the default
version of the DICE, FUND, and PAGE lAMs. The only climate variable that was harmonized across these
three previous models was equilibrium climate sensitivity (ECS) - a measure of the globally averaged
temperature response to increased radiative forcing (generally, the equilibrium temperature response
resulting from a doubling of atmospheric C02 concentrations). Each 1AM was run using a probability
distribution for the ECS, calibrated to the Intergovernmental Panel on Climate Change's (IPCC) Fourth
Assessment Report (AR4) (IPCC 2007a) findings using the Roe and Baker (2007) distribution.54 All other
aspects of the modeling - such as the representation of the carbon cycle and its parameterization, sea-
level rise, regional downscaling of temperature, and treatment of non-C02 greenhouse gases - varied
across the three lAMs and were used as the model developers had designed them.
To implement a modular approach to updating the representation of climate and other Earth system
dynamics in SC-GHG estimation, it is helpful to review the available climate models capable of meeting
the climate module requirements outlined above, the conclusions of recent scientific assessments
published since the IPCC's AR4 report, the public comments received on individual EPA proposed
rulemakings and the IWG's February 2021 TSD (IWG 2021), and the National Academies (2017)
recommendations related to the climate module.
The conclusions of recent scientific assessments (e.g., IPCC 2014a, 2018, 2019a, 2019b, 2021a; USGCRP
2016, 2018a; and the National Academies 2016b, 2019) bolster the science underlying the modeling of
53 This module could in future iterations also generate estimates of other climatic variables (e.g., precipitation
changes) as well as non-climate mediated impacts of GHG emissions if needed as inputs to future damage functions.
As discussed in Section 3.3, the only non-climate mediated effect included in SC-GHG estimates used by the EPA to
date are plant fertilization effects from elevated CO2 concentrations. Other non-climate mediated effects of GHG
emissions that have not yet been incorporated into SC-GHG estimation are discussed in Section 4.2.
54 The IPCC's Fourth Assessment Report (IPCC 2007b) was the most current IPCC assessment available at the time
when the IWG calibrated the ECS distribution.
26
-------
EXTERNAL REVIEW DRAFT
climate dynamics. Recently, in August 2021, the IPCC released the Working Group (WG) 1 contribution to
the IPCC Sixth Assessment Report (AR6) (IPCC 2021a). The IPCC (2021a) report brings together the most
up-to-date physical understanding of the climate system and climate change. The report includes updated
IPCC AR6 consensus statements on key climate parameters that are relevant for SC-GHG estimation,
including equilibrium climate sensitivity and transient climate response. For equilibrium climate sensitivity
(ECS)55, the AR6 assessment finds, with high confidence, that the best estimate is 3°C with a likely range
of 2.5°C to 4°C.56 AR6 also concludes that "it is virtually certain that ECS is larger than 1.5°C, but currently
it is not possible to rule out ECS values above 5°C" (IPCC 2021a). For the transient climate response (TCR),
AR6 finds that the best estimate of TCR is 1.8°C, and it is very likely between 1.2 and 2.4°C.57 Additional
discussion of scientific updates in AR6 is provided in the Appendix. In particular, Section A.l contains a
summary of the IPCC's understanding of C02, CH4, and N20 greenhouse gas radiative efficiency,
atmospheric lifetimes, and chemistry in AR6 relative to AR4, which was the basis of the simplified lifetime
and forcing equations underlying the IWG estimates used by the EPA and other federal agencies to date.
Reduced-complexity climate models (RC models) offer meaningful improvements over the current
representation of climate dynamics in existing lAMs (Nicholls et al. 2020). RC models are highly
parameterized, computational emulators of the climate system. RC models are different from the highly
complex and computationally demanding Earth system models (ESMs), which are the state-of-the-art
tools for climate projections. However, the use of RC models may be preferred over ESMs for certain
applications for at least three reasons: (1) the computational efficiency of the RC models allows for
hundreds or thousands of simulations in a relatively short timeframe, (2) the adjustability of model
parameters allows for the exploration of uncertainty, and (3) because RC models do not model year-to-
year variability they allow for the estimation of the difference between emission scenarios that would be
smaller than that variability (Sarofim et al. 2021a). RC models have a long history of use in climate science
assessments, 1AM modeling applications, and analyses of climatic processes. They are ubiquitously used
to support model inter-comparisons and diagnostics because of their ability to emulate different ESM
components and variables, explore uncertainties in key climate parameters, analyze scenarios to provide
concentration and temperature inputs to lAMs and other models, and estimate climate sensitivity when
55 ECS is defined as "the equilibrium (steady state) change in the surface temperature following a doubling of the
atmospheric carbon dioxide (CO2) concentration from pre-industrial conditions" (IPCC 2021a).
56 The AR6 assessment finds "[b]ased on multiple lines of evidence, the very likely range of equilibrium climate
sensitivity is between 2°C (high confidence) and 5°C (medium confidence). The AR6 assessed best estimate is 3°C
with a likely range of 2.5°C to 4°C (high confidence), compared to 1.5°C to 4.5°C in AR5, which did not provide a
best estimate" (IPCC 2021a). In IPCC statements, the terms "likely", "very likely" and "virtually certain" are defined
to correspond to probabilities of at least 66% (16.6-83.3 percentile), 90% (5-95 percentile), and 99% (0.5-99.5
percentile), respectively (IPCC 2007c). In IPCC reports a level of confidence is expressed using five qualifiers (very
low, low, medium, high, and very high) based on the type, amount, quality, and consistency of evidence (e.g.,
mechanistic understanding, theory, data, models, expert judgement) and on the degree of agreement across
multiple lines of evidence. Statements in the AR6 WG1 report that include "best estimate" are not specific on its
definition.
57 TCR is defined as "the surface temperature response for the hypothetical scenario in which atmospheric carbon
dioxide (CO2) increases at 1% yr-1 from pre-industrial to the time of a doubling of atmospheric C02 concentration"
(IPCC 2021a), thereby being a measure of the speed as well as the magnitude of the climate response. AR6 states
that "Based on process understanding, warming over the instrumental record and emergent constraints the best
estimate TCR is 1.8°C, it is likely 1.4 to 2.2°C and very likely 1.2 to 2.4°C" (IPCC 2021a).
27
-------
EXTERNAL REVIEW DRAFT
coupled with historical climate observations (Nicholls et al. 2020, Nicholls et al. 2021, Sarofim et al.
2021a).
One of the most widely used RC models is the Finite amplitude Impulse Response (FaIR) climate model
(Millar et al. 2017, Smith et al. 2018) to generate projections of global mean surface temperature (GMST)
change. The FaIR model was originally developed by Richard Millar, Zeb Nicholls, and Myles Allen at Oxford
University, as a modification of the approach used in IPCC AR5 to assess the GWP and GTP (Global
Temperature Potential) of different gases. It is open source, widely used (e.g., IPCC 2018, IPCC 2021b),
and was highlighted by the National Academies (2017) as an RC model that satisfies their
recommendations for a near-term update of the climate module in SC-GHG estimation. Specifically, it
translates GHG emissions into mean surface temperature response following the steps outlined above
and represents the current understanding of the climate and GHG cycle systems and associated
uncertainties within a probabilistic framework. The FaIR model's projections of future warming are
consistent with more complex, state of the art ESMs and can, with high confidence, be used to accurately
characterize current best understanding of uncertainty, is easily implemented, and is transparently
documented.
The updated SC-GHG estimates presented in this report rely on FaIR version 1.6.2 as used by the IPCC
(2021a, 2021b). An alternative version of the model, FaIR 2.0, was recently published (Leach et al. 2021)
that offers some advantages with respect to simplicity and the inclusion of a flexible, state-dependent
methane lifetime, but is less preferable for SC-GHG estimation at this time because it is not yet able to
track ocean heat uptake (which is used as an input to help project future sea level rise in some models
such as BRICK); importantly the calibration of its uncertain parameters is based on historical data but has
not yet been adjusted to be consistent with the AR6 evaluation of climate characteristics such as the IPCC
assessed likely range of 2.5 to 4°Cforthe climate sensitivity. FaIR 1.6.2 also has advantages over the latest
versions of other RC models, including the Model for the Assessment of Greenhouse Gas Induced Climate
Change (MAGICC; Meinshausen et al. 2011) and the Hector model, a U.S. Government-developed model
(Hartin et al. 2015).58 MAGICC is widely used in science research, policy analysis, IPCC reports, and the
latest version, MAGICC 7.5.1, has been calibrated to AR6 findings. However, the model itself is not open
source and, therefore, less preferable to FaIR in terms of transparency and reproducibility. The Hector
model has some additional complexity and features that could be helpful in future SC-GHG updates. For
example, it can emulate ocean acidification, permafrost, and land carbon cycles (Woodard et al. 2021).
However, Hector has not yet been calibrated to the AR6 assessed climate characteristic ranges, and the
current version of Hector has no suggested parameter sets for use in uncertainty analysis. Table 2.2.1
shows summary statistics for the ECS from the FaIR 1.6.2 model used in this report and other RC models
58 FaIR and MAGICC were among the four RC models examined in IPCC (2021a), along with Oscar (Gasser et al. 2020),
and Cicero-SCM (Skeie et al. 2021). Each of these were calibrated based on agreement with observations such as
historical temperatures, ocean heat uptake, CO2 concentrations, and airborne fraction. The WG1 report compares
distributions from the calibrated models to assessed values of metrics such as ECS and TCR. The latter two RC models
are dropped from detailed consideration in this report because Cicero-SCM does not have a carbon cycle
representation, and Oscar did not match projected future temperatures from the Coupled Model Intercomparison
Project (CMIP) and other projections. Thompson (2018) also identified FaIR, MAGICC, and Hector as being good fits
to the National Academies' recommended criteria for the climate module.
28
-------
EXTERNAL REVIEW DRAFT
and compares them to IPCC statements. For reference, Table 2.2.1 also includes the assumed distribution
used in IWG SC-GHG estimates to date. Table 2.2.2 shows similar information for the TCR.
Taken together, FaIR 1.6.2 is a fitting RC model to serve as the basis for an updated climate module in SC-
GHG estimation. It provides, with high confidence, an accurate representation of the latest scientific
consensus on the relationship between global emissions and global mean surface temperature under the
wide range of socioeconomic emissions scenarios discussed in Section 2.1. It also offers a code base that
is fully transparent and available online (unlike MAGICC), and the uncertainty capabilities in FaIR 1.6.2
have been calibrated to the most recent assessment of the IPCC (which importantly narrowed the range
of likely climate sensitivities relative to prior assessments) (unlike FalR2.0 or Hector at the present time).
Table 2.2.1: Summary Statistics for Equilibrium Climate Sensitivity under Reduced-Complexity Climate
Models and IPCC statements
Percentiles and Other Summary Statistics
5%
16.6%
Mode3
Median
(50%)
Mean
83.3%
95%
FaIR 1.6.2 :
2.05
2.37
2.78
2.95
3.18
3.87
5.03
FaIR 2.0.0 (Leach et al. 2020)
1.94
2.36
3.24
4.74
6.59
MAGICC7 (IPCC 2021a)
1.93
2.97
4.83
Hector2.5 (Nicholls et al. 2021)
1.84
2.16
2.85
3.90
5.45
AR6 statement (2022)
2.00
2.50
3.00
4.00
5.00
AR5 statement (2014)
> 1.00
1.50
4.50
<6.00
IWG to date (Roe & Baker (2007),
calibrated to AR4) (2010)
1.72
2.00
2.34
3.00
3.50
4.50
7.14
AR4 statement (2007)
2.00
3.00c
4.50
a Mode calculated after rounding to 2 decimal places.
b AR6 offers a "best estimate" but is not specific on which statistic for central value most closely corresponds to "best".
c AR4 offers a "most likely" value. As noted in IWG (2010), strictly speaking, "most likely" refers to the mode of a distribution
rather than the median, but common usage would allow the mode, median, or mean to serve as candidates for the central or
"most likely" value and the IPCC report is not specific on this point.
d Results from FaIR 1.6.2 were estimated using the 2,237 constrained parameter sets.
29
-------
EXTERNAL REVIEW DRAFT
Table 2.2.2: Summary Statistics for Transient Climate Response under Reduced-Complexity Climate Models
and IPCCStatements
Percentiles and Other Summary Statistics
5%
16.6%
Mode3
Median
(50%)
Mean
83.3%
95%
FaIR 1.6.2
1.36
1.49
1.60
1.81
1.85
2.20
2.46
FaIR 2.0.0 (Leach etal.2020)
1.30
1.48
1.79
2.15
2.44
MAGICC7 (IPCC 2021a)
1.27
1.88
2.61
Hector2.5 (Nicholls et al. 2021)
1.42
1.58
1.82
2.08
2.29
AR6 statement (2022)
1.20
1.40
1.80
2.20
2.40
AR5 statement (2014)
1.00
2.50
3.00
AR4 statement (2007)
1.00
3.00
a Mode calculated after rounding to 2 decimal places.
b Results from FaIR 1.6.2 were estimated using the 2,237 constrained parameter sets.
Figure 2.2.1 shows the projected future atmospheric concentration59 of C02 through 2300 based on the
RFF-SP emissions projections that are used as inputs into FaIR 1.6.2. Atmospheric concentrations increase
over time due to the accumulation of annual emissions, with excess C02 from the atmosphere moving
into the ocean and ecosystems slowly over time until eventually a new equilibrium is reached.60 Figure
2.2.2 shows the corresponding projection of global mean surface temperature. The ranges in these figures
reflect uncertainty in both emissions and physical climate processes that are consistent with the latest
projections coming out of the Sixth Assessment Report (IPCC 2021a).
59 Atmospheric concentration refers to the amount of a gas in the atmosphere. For CO2, it is measured in parts per
million (ppm). Pre-industrial concentrations of CO2 were 280 ppm, and concentrations this high have not been seen
in at least 2 million years.
60 Figures A.5.3 and A.5.4 in the Appendix show projected atmospheric concentrations of methane (CH4) and nitrous
oxide (N2O). Cm and N2O concentrations are higher than at any time in at least 800,000 years. While CO2, once
emitted into the atmosphere through combustion, is not destroyed but rather moves between the ocean,
ecosystems, and atmosphere, other gases like CH4 and N2O are destroyed through reactions in the atmosphere.
30
-------
EXTERNAL REVIEW DRAFT
Figure 2.2.1: Global Atmospheric Concentrations of Carbon Dioxide (C02), 1900-2300
2000
0
1900 " 2000 " 2100 " 2200 " 2300
Year
Future atmospheric concentrations of carbon dioxide (C02) are based on the range of annual emissions projections from the
sampled RFF-SP scenarios used as inputs into FaIR 1.6.2. FaIR 1.6.2 is run with the full, AR6 calibrated (constrained) uncertainty
distribution. Therefore, the uncertainty ranges in this figure represent both emissions and physical carbon cycle uncertainty. Mean
(solid) and median (dashed) lines are shown along with the 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges.
Figure 2.2.2: Global Mean Surface Temperature Change, 1900-2300
8°
1900 200 0 2100 2200 2300
Year
The range of global mean surface temperature change relative to pre-industrial (1850-1900) as calculated by FaIR 1.6.2
corresponding to the C02 concentrations from Figure 2.2.1 and the accompanying figures for CH4 and N20 in the Appendix.
Uncertainty comes from emissions uncertainty from the RFF-SP projections and physical climate uncertainty from FaIR. Mean
(solid) and median (dashed) lines are shown along with the 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges.
31
-------
EXTERNAL REVIEW DRAFT
Because the SC-GHG is calculated based on the impact of a marginal pulse of emissions, it is particularly
relevant to investigate how the climate model responds to small changes in emissions. The response of
the climate to a pulse of GHG emissions (i.e., C02, CH4, or N20) is calculated by using a reference scenario
(baseline) and subtracting the temperatures of that reference scenario from a second scenario
(perturbed) that is identical in all dimensions except for the marginal increase in emissions for the one
year and one gas being examined (i.e., all characteristics of the model run, emissions levels of other gases,
etc., are held constant for the duration of the perturbed model run). Figure 2.2.3 shows the temperature
response resulting from a pulse of C02 emissions in 2030 under the three RC models considered in this
report.61 The FaIR, MAGICC, and Hector model outputs all exhibit similar dynamics in the timing of peak
warming in response to a pulse of emissions. For most gases, a pulse of emissions leads to a peak in
temperature within a few years following the pulse of emissions. Then, as the radiative forcing declines
and the ocean heat uptake increases, the marginal increase in temperature begins to decline at an
increasing rate. However, as illustrated in Figure 2.2.3, the temperature response to a pulse of C02 is a
little more complicated. When the rate of decrease in radiative forcing slows such that the rate of decline
in ocean heat uptake exceeds it, atmospheric warming resumes leading to a sustained increase in
temperature.62 The temperature dynamics of these models represents a significant scientific
advancement over the temperature responses underlying the climate components of the three lAMs used
in the IWG SC-GHG estimates. Specifically, Dietz et al. (2021) showed that the initial response of DICE,
FUND, and PAGE to a pulse of C02 emissions was slower than the response of FaIR calibrated to 256
models involved in the fifth phase of the Coupled Model Intercomparison Project (CMIP563),
demonstrating that FaIR and related models can better emulate the high-resolution global climate
models. This is an important feature when estimating the SC-GHG as discussed in Section 2.4 (near term
marginal damages are discounted less than damages far in the future). Additionally, Dietz et al. (2021)
found that for the long-term response (200 years after the pulse) FUND and DICE 2016 were higher than
the FaIR emulations and the response of PAGE was consistently lower. (See Figure A.5.7 in the Appendix.)
61 Figures A.5.5 and A.5.6 in the Appendix show the temperature response resulting from a pulse of Cm and N20
emissions.
62 A more detailed explanation of the temporal temperature response resulting from a pulse of greenhouse gas
emissions is as follows. The atmospheric concentration response from an emissions release is the highest at time
zero and declines thereafter as the gas either decomposes in the atmosphere or cycles into other reservoirs. The
radiative forcing is directly related to the increased concentration. However, the temperature response is a function
of the accumulation of energy due to the radiative forcing, minus the heat that the ocean takes up as the atmosphere
warms and the increased heat that is radiated to space due to a warmer planet. For most gases, this balance between
radiative heating from the gas and heat uptake by the ocean leads to a peak in temperature within a few years of
the emission as the radiative forcing declines and the ocean heat uptake increases. The decline in temperature lags
the decline in radiative forcing, as the heat that went into the ocean is eventually released. However, the response
to a pulse of CO2 is a little more complicated: because the elevated concentrations resulting from a pulse of CO2
emissions decreases quickly to start as CO2 cycles into the ecosystems and surface oceans, but then the decrease
slows as the timescale becomes dominated by deep ocean mixing and slows further when it is dominated by
sedimentation. When the rate of decrease in radiative forcing slows such that the rate of decline in ocean heat
uptake exceeds it, atmospheric warming resumes creating a second peak in temperature (Millar et al. 2017).
63 CMIP is the Coupled (sometimes, Climate) Model Intercomparison Project. CMIP creates a framework for
consistent application of climate models to a common set of scenarios, and with a common set of outputs, to
facilitate assessment of these models and provide consistent inputs to impacts assessments. CMIP5 is the fifth phase
of CMIP and was timed to provide important scientific input to the IPCC AR5 assessment.
32
-------
EXTERNAL REVIEW DRAFT
As described in Section 2.3, all three of the approaches to damage function estimation in this report use
only GMST as an input to the damage module. For the two more disaggregated approaches, any needed
regional or more finely spatially disaggregated temperature projections are created internal to the
damage module.
Figure 2.2.3: Global Mean Surface Temperature Anomaly from a Pulse of Carbon Dioxide (lGtC) by Model,
2020-2300
0.0030
O 0.0025
m
O
(N
C
y 0.0020
Year
— FaIR 1.6.2 — HECTOR 2.5 — MAC ICC 7.5.3
The mean global temperature response resulting from a pulse of emissions of C02 in 2030 as projected by FalRl.6.2, Hector 2.5,
and MAGICC 7.5.3. This represents the difference between a reference scenario (using SSP2-RCP4.5for the figure) and the same
scenario including the pulse of emissions. The emission pulse size is 1 GtCfor carbon dioxide. Mean (solid) and median (dashed)
lines are shown along with the 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges.
Sea Level Rise. In addition to temperature change, two of the three damage modules used in this report
require global mean sea level (GMSL) projections as an input to estimate coastal damages. Those two
damage modules use different models for generating estimates of GMSL. Both are based off reduced
complexity models that can use the FaIR temperature outputs as inputs to the model and generate
projections of GMSL accounting for the contributions of thermal expansion and glacial and ice sheet
melting based on recent scientific research. Absent clear evidence on a preferred model, the SC-GHG
estimates presented in this report retain both methods used by the damage module developers.
33
-------
EXTERNAL REVIEW DRAFT
The first damage module used in this report (discussed in Section 2.3.1) projects GMSL using an
implementation of the Framework for Assessing Changes To Sea-level (FACTS). FACTS is a flexible
computational framework, that can mix and match components of different models in order to further
explore uncertainty that is being used for the IPCC AR6 SLR projections (IPCC 2021c, Garner et al. 2021).64
In this damage module, FACTS is used to project sea level rise, relying on the parameterizations based on
the two approaches that the IPCC characterized as "medium confidence", and assuming that those two
approaches were equally likely. This leads to a slightly narrower projected SLR range than the likelihood
bounds from the IPCC medium confidence approach (given two distributions, the IPCC used the outermost
probability for any given likelihood estimate). The choice of using only the medium confidence
parameterizations leads to the lowest future sea level rise projections available from the FACTS model;
the parameterization excludes the possible contributions from marine ice cliff instability (MICI) and from
ocean forcing on basal melt rates that was also assessed to be low confidence by the IPCC.
The additional sea level rise resulting from the emissions pulse is estimated using what is known as a
"semi-empirical"65 sea level model (Kopp et al. 2016), which was cited by the National Academies as a
potential approach for estimating sea level rise from an emissions pulse (National Academies 2017). The
semi-empirical model is driven by the same probabilistic GMST projections from FaIR used in the non-
coastal sectors. It is calibrated based on historical data and has its own probability distribution that is
generally lower than that seen in the FACTS projections. The FACTS projections account for a best
understanding of future contributions to SLR from numerous sources but cannot be applied to an
individual emissions pulse. Thus, to bias-correct the semi-empirical model's projections, each probabilistic
draw is quantile-mapped to an equivalent probabilistic draw of the FACTS projections within each SSP-
RCP. The magnitude of the SLR impact of an emissions pulse is not changed, but the baseline SLR in the
absence of the pulse is adjusted such that it is consistent with the probabilistic distribution from FACTS
for each SSP-RCP. To model SLR in the RFF-SPs, for which no FACTS projections are available for bias
correction, an additional quantile-mapping step is taken. This is detailed in CIL (2022).
The second SLR model used in this report, Building Blocks for Relevant Ice and Climate Knowledge (BRICK),
is a semi-empirical modeling framework that simulates GMSL. Changes in global mean surface
temperature drive changes in GMSL. The model includes contributions to GMSL from the Greenland and
Antarctic ice sheets, thermal expansion, glaciers and ice caps, and land water storage (Wong et al. 2017,
Vega-Westhoff et al. 2019). The parameterizations for the BRICK model include assumptions about
64 Additional information about the IPCC AR6 SLR projection methods can be found at:
https://sealevel.nasa.gov/data_tools/17.
65 Semi-empirical models are a form of reduced complexity process models. These models are known as semi-
empirical because they are based on equations that embody physical understanding and calibrated to historical data.
Semi-empirical models are a commonly used approach in the literature. The Kopp et al. (2016) model is based on a
set of three differential equations: one to relate a change in sea level to a difference between projected atmospheric
temperature and a theoretical equilibrium temperature, one to determine the change in the theoretical equilibrium
temperature over time, and one to address the additional sea level rise from the climate response to long-term
orbital changes. The parameters in these three equations are then calibrated against estimates of historical warming
and sea level over the past millennia. The Kopp et al. model agreed well with process-based model and expert
surveys available at the time. Semi-empirical models calibrated solely on historical data will not include processes
that were not active over the historical calibration period, such as MICI processes (which are often not included in
process-based models either).
34
-------
EXTERNAL REVIEW DRAFT
Greenland and Antarctic melt that are consistent with the IPCC AR6 projections that include MICI.
Inclusion of processes like MICI have the largest effects after 2100, and for the warmest scenarios, such
that inclusion in the RCP8.5 scenario leads to an average increase of 15% in SLR by 2100 and 50% by 2150
(relative to 1850-1900, Table 9.10, IPCC 2021c). By 2300, inclusions of MICI processes for the RCP8.5
scenario results in SLR of 9.5 to 16.2 meters, which is substantially larger than the no ice-sheet
acceleration assumption which yields a rise of 1.7 to 4.0 meters (Table 9.11, IPCC 2021c).
Figure 2.2.4 shows the projected global sea level change resulting from the FACTS- and BRICK-based SLR
models, as implemented in the two damage modules discussed in Section 2.3. FACTS and BRICK have
similar projections of SLR rise through the end of the century. BRICK, as expected, projects greater SLR in
the out years because of its inclusion of accelerated melt processes for the Antarctic and Greenland ice
sheet, consistent with the IPCC forecasts that include MICI processes. By 2300, BRICK estimates an average
of 4 meters, while the implementation of FACTS used in this report generates SLR projections of 2 meters,
on average. This difference in the out years is due to the choices of (a) relying only on IPCC's "medium
confidence" SLR processes, and (b) taking an equal weighting rather than an outer envelope when
combining multiple probability distributions. In the absence of a probabilistic assessment of the likelihood
of these processes, this report retains use of both approaches.
In addition to surface temperatures and atmospheric concentrations, FaIR also calculates C02 uptake in
the world's ocean as part of its carbon cycle calculation and generates projections of measures of ocean
acidification (pH and ocean heat). The impacts of ocean acidification are not captured in the SC-GHG
estimates presented in this report because functions that translate the pH and ocean heat outputs from
FaIR into monetized global damages are not yet available in the damage module. However, given current
understanding of the impacts of C02 emissions on the growth and survival of shellfish and coral reefs,
coupled with the availability of market and nonmarket valuation studies on the ecosystem services they
provide, it is likely feasible to develop damage functions that include ocean acidification impacts in future
SC-GHG updates. See section 3.2 for more discussion of damages associated with ocean acidification and
other impacts of climate change that are not captured in this report.
35
-------
EXTERNAL REVIEW DRAFT
Figure 2.2.4: Global Sea Level Rise in FACTS and BRICK, 1950-2300
FACTS
(U w t
.!2 O
aC o
> A 3
a; O
—1 Lf)
ra co
O *=
1
a:
0
1950
2020
2100
Year
BRICK
2200
2300
10
8
E
"~~
I/) O
IB
-Q >
O +3
U-g 2
OL
1950
2020
2100
Year
2200
2300
The range of global mean sea level rise relative to pre-industrial (1850-1900) as calculated by FACTS (top) and BRICK (bottom).
Uncertainty comes from emissions uncertainty from the RFF-SP projections, physical climate uncertainty from FaIR, and parameter
uncertainty underlying each SLR module. Mean (solid) and median (dashed) lines are shown along with the 5th to 95th (dark shade)
and 1st to 99th (light shade) percentile ranges.
36
-------
EXTERNAL REVIEW DRAFT
2.3 Damage Module
The damage module contains the core "damage functions" in the SC-GHG estimation process. Damage
functions translate changes in temperature and other physical impacts of climate change into monetized
estimates of net economic damages. The damage functions capture multiple net damage pathways that
can be broadly divided into market and non-market pathways. Some net economic damages are
experienced through markets, such as changes in net agricultural productivity, net energy expenditures,
and property damage from increased flood risk. Examples of net damages experienced through the
nonmarket pathways include changes in net mortality rates and changes in ecosystem services, including
those provided by biodiversity.
As discussed above, the SC-GHG estimates used in the EPA's analyses to date have maintained the damage
functions contained in the default version of the DICE, FUND, and PAGE lAMs as used in the peer-reviewed
literature. Specifically, the damages functions underlying the IWG SC-GHG estimates used since 2013 are
taken from DICE 2010 (Nordhaus 2010); FUND 3.8 (Anthoff and Tol 2013a, 2013b); and PAGE 2009 (Hope
2013).66 These models all take stylized, reduced-form approaches to estimating monetized damages as a
function of temperature change and sea level rise. They use a suite of underlying studies to calibrate their
damage functions. FUND 3.8 takes a regional bottom-up approach to specify the damage function by
calibrating to or building up disaggregated pieces consisting of 14 separate damage categories or sectors
using studies and assumptions relating to each sector. Damages in DICE 2010 are an aggregate based on
a calibration of sectoral damages (Nordhaus and Boyer 2000) and scaled using aggregate damages. PAGE
2009 employs a regionalized hybrid approach with an estimate of four categories of damages: economic,
sea-level rise, nonmarket, and discontinuities.
The National Academies' recommendations for the damage module, scientific literature on climate
damages, updates to models that have been developed since 2010, as well as the public comments
received on individual EPA rulemakings and the IWG's February 2021 TSD, have all helped to identify
available sources of improved damage functions. The IWG (e.g., IWG 2010, 2016a, 2021), the National
Academies (2017), comprehensive studies (e.g., Rose et al. 2014), and public comments have all
recognized that DICE 2010, FUND 3.8, and PAGE 2009 do not include all the important physical, ecological,
and economic impacts of climate change. The climate change literature and the science underlying the
economic damage functions have evolved, and DICE 2010, FUND 3.8, and PAGE 2009 now lag behind the
most recent research.
The challenges involved with updating damage functions have been widely recognized. Functional forms
and calibrations are constrained by the available literature and need to extrapolate beyond warming
66 The damages functions underlying the IWG SC-GHG estimates used from 2010 to 2013 came from earlier versions
of each model: DICE 2007 (Nordhaus 2008), FUND 3.5 (Narita et al. 2010), and PAGE 2002 (Hope 2006). The newer
versions of each model that have been used by the IWG since 2013 included a number of updates related to their
damage functions. For example, DICE 2010 included a re-calibrated damage function with an explicit representation
of economic damages from sea level rise. Updates in FUND 3.8 included revised damage functions for space heating,
SLR, and agricultural impacts. PAGE 2009 added an explicit representation of SLR damages, revisions to ensure
damages do not exceed 100% of GDP, a change in regional scaling of damages, revised treatment of potential abrupt
damages, and updated adaptation assumptions. See IWG (2013) for more discussion of each of these changes.
37
-------
EXTERNAL REVIEW DRAFT
levels or locations studied in that literature. Research and public resources focused on understanding how
these physical changes translate into economic impacts have been significantly less than the resources
focused on modeling and improving our understanding of climate system dynamics and the physical
impacts from climate change (Auffhammer 2018). Even so, as illustrated in Figure 2.3.1, there has been a
large increase in research on climate impacts and damages in the time since DICE 2010, FUND 3.8, and
PAGE 2009 were published. Along with this growth, there continues to be wide variation in methodologies
and scope of studies. Comparability issues across both methods and studies create challenges for
synthesizing the current understanding of impacts or damages.
Figure 2.3.1: Research on Climate Impacts, 1990-202167
70
60
¦ ^ 50
~o
-I—J
oo
O 40
CD
_Q
E
30
20
10
0
l Newer research
I Studies underlying DICE 2010
I Studies underlying PAGE 2009
Studies underlying FUND 3.8
1990 1994 1998
Source: Greenstone (2016), updated in 2021.
2002
2006
2010
2014
2018
Approaches to developing a damage module for SC-GHG estimation can be generally grouped into two
broad categories: those that estimate a damage function by calibrating to or building up disaggregated
pieces, and studies that estimate an aggregate global damage function directly. The more disaggregated
approach typically involves spatially explicit and sector-specific modeling of relevant processes and then
67 In many cases, the three lAMs used different studies for calibration. This is particularly true of FUND, which used
studies relating to different subsectors of the model, whereas DICE and PAGE did not have as detailed a sectoral
breakdown. That means that summing across these different models is likely valid in all but a few isolated cases. The
blue bars include studies uncovered from a comprehensive literature review in the economics literature (and a few
others in public health or relevant disciplines) by the Climate Impact Lab through early 2016. Each of the studies
counted in blue was determined by CIL to have employed a research design that allowed for the causal interpretation
of results (Greenstone 2016).
38
-------
EXTERNAL REVIEW DRAFT
aggregates regional or sectoral damages.68 Alternatively, the aggregate global damage function
estimation approach often relies on meta-analysis techniques (e.g., as in recent versions of DICE (DICE
2013R and DICE 2016)) or total-economy empirical studies that econometrically estimate the relationship
between GDP and a climate variable, usually temperature (e.g., used in part in the most recent version of
the PAGE model (PAGE 2020 (Kikstra et al. 2021)). There are also more complex ways to estimate damage
functions directly (e.g., that have been used in extensions of DICE) and through expert elicitation (e.g.,
Pindyck 2019, Howard and Sylvan 2021). Based on a review of available studies using these approaches,
the SC-GHG estimates presented in this report rely on three damage functions. They are:
1. a subnational-scale, sectoral damage function estimation (based on the Data-driven Spatial
Climate Impact Model (DSCIM) developed by the Climate Impact Lab (CIL 2022, Carleton et al.
2022, Rode et al. 2021)),
2. a country-scale, sectoral damage function estimation (based on the Greenhouse Gas Impact Value
Estimator (GIVE) model developed under RFF's Social Cost of Carbon Initiative (Rennert et al.
2022b)), and
3. a meta-analysis-based global damage function estimation (based on Howard and Sterner (2017)).
Each is discussed in turn.
2.3.1 Damage Module based on the Data-driven Spatial Climate Impact Model (DSCIM)
DSCIM was developed by the Climate Impact Lab (CIL). CIL is a multidisciplinary consortium of climate
scientists, economists, computational experts, researchers, and analysts building empirically derived,
local-level estimates of the net damages from climate change and empirically based SC-GHG estimates.69
The DSCIM modeling runs performed for the estimates presented in this report are described in the
September 2022 DSCIM User Manual (CIL 2022). DSCIM monetizes climate damages for nearly 25,000
global impact regions using econometric methods that account for local conditions, including adaptation
investments, when estimating the effect of climate change on sector specific outcomes. These local
damages are aggregated to develop an estimate of global damages as a function of global temperature
changes. The damage functions for DSCIM are constructed through a five-step process. First, researchers
collect and harmonize historic climate and socioeconomic data for each sector. Second, using variation in
short-run weather and cross-sectional variation in the long-run average climate and socioeconomic
conditions, they econometrically estimate the effect of changes in local climatic conditions on sector-
specific outcomes, accounting for the adaptive effects of climate and socioeconomics, which can alter the
sensitivity of outcomes to local climate. Third, they use a revealed preference approach to infer the
adaptation costs incurred by populations as they adapt to warming, drawing on research by Guo and
68 There are also multisectoral, multiregional economic computable general equilibrium (CGE) models. CGE models
calibrate to region-sector impact estimates but account for more interactions among regions, impacts, supply, and
demand factors.
69 The Climate Impact Lab team combines experts from the University of California, Berkeley, the Energy Policy
Institute at the University of Chicago (EPIC), Rhodium Group, Rutgers University, University of California, Santa
Barbara, and University of Delaware. More information on the individual researchers and institutions involved in the
Climate Impact Lab can be found at: http://www.impactlab.org/.
39
-------
EXTERNAL REVIEW DRAFT
Costello (2013) and Deryugina and Hsiang (2017).70 Fourth, they project sector-specific outcomes and
associated monetized damages into the future by combining the econometric results with a probabilistic
ensemble of high-resolution downscaled climate projections from 33 global climate models71 and
aggregate the local damages to global damages. Finally, they use these projections to estimate global
damages as a time-varying reduced-form function of global mean surface temperature. The advantage to
this approach is that global damage estimates reflect the empirically derived local impact relationships,
and account for the uncertainty in economic growth, temperature change, and adaptation. For the DSCIM
model runs in this report, the outputs of the socioeconomic module (Section 2.1) and the GMST output
from the climate module (Section 2.2) are used as inputs in DSCIM.72
At present, DSCIM includes the estimation of climate damages occurring in five sectors or impact
categories: health, energy, labor productivity, agriculture, and coastal regions (CIL 2022).73 Table 2.3.1
summarizes key elements of DSCIM's damage function estimation methods in each of these five sectors.
The health component includes the value of net changes in hot- and cold-related mortality risk (Carleton
et al. 2022). The building block of the global mortality damage function is the estimation of temperature's
impact on mortality rates using historical data. The mortality data is assembled from various sources at
the subnational spatial scale74 for 40 countries covering 38 percent of the global population.75 Temporal
coverage for each country ranges from 13 years (1997-2010) to over 40 years (e.g., 1968-2010 for the
U.S.) across the sample. The age-specific mortality-temperature response is estimated as a linear function
of nonlinear daily grid-level temperature and precipitation data transformations. This specification,
together with the inclusion of fixed effects to account for any time-varying trends or shocks to age-specific
mortality rates unrelated to climate, allows them to isolate the impact of year-to-year, within-location
variation in temperature and rainfall on mortality. Additionally, this model recovers the effect of climate-
driven adaptation (e.g., more cooling systems) and income growth on the shape of the temperature
mortality relationship, as observed in the historical record using cross-sectional variation in long-run
average conditions. These econometric estimates are combined with high-resolution projections of
climate, income, and demographics to compute age-specific projected impacts of climate change under
70 The method for estimating the costs of adaptation reflects that people invest in adaptive behaviors and
technologies until the costs of doing so just equal the protective benefits. The protective benefits are observed
through the changes in the estimated sensitivity of outcomes to temperature (or rainfall or sea level rise) as the
climate gradually warms. The estimated measures of these benefits are used to back out the costs of the adaptation.
See Carleton et al. (2022) for more discussion.
71 See CIL (2022) for a detailed discussion of the ensemble of climate projections.
72 To incorporate the RFF-SPsfor model runs performed for this report, DSCIM uses an emulator approach that allows
for the estimation of probabilistic socioeconomics in DSCIM's highly complex and disaggregated damage system.
The emulator weights the outcome of annual global aggregate damage functions that are estimated using the suite
of SSP-RCP combinations according to how closely the socioeconomics characteristics each year match those
contained in the RFF-SPs. See CIL (2022) for more details.
73 CIL plans to update DSCIM regularly with representation of additional sectors (CIL 2022).
74The mortality data is at the second administrative level (e.g., county), first administrative level (e.g., state), or
somewhere in between.
75 Carleton et al. (2022) also have data from India (which increases coverage to 55% of the global population) but
are unable to include it in the main estimation of the mortality-temperature response function due to the absence
of age-specific mortality statistics. Instead, the authors use the India data to assess external validity of their
extrapolation methods and find the model generates conservative predictions of mortality impacts of climate change
in India, a hot and poor region of the globe.
40
-------
EXTERNAL REVIEW DRAFT
multiple emissions scenarios at the scale of ~25,000 global regions. While the main specification of DSCIM
employs an age-adjusted valuation approach for monetizing net health damages (inclusive of adaptation
costs), in the results presented in this report, the projected changes in premature mortality are monetized
using country-level population-average measures of the willingness-to-pay for mortality risk reductions.76
The energy component includes energy expenditures from temperature-related changes in electricity and
direct fuel consumption across residential, commercial, and industrial end-uses (Rode et al. 2021). Rode
et al. provide the first estimate of the global impact of climate change on total energy consumption using
globally comprehensive data, accounting for economic development and adaptive behavior. Energy
consumption data for electricity and other fuels is compiled from the International Energy Agency and is
available at the country-by-year level for 146 countries from 1971 to 2010. Daily historical climate data
are aggregated to annual, country-level observations following the method in Carleton et al. (2022), which
preserves local-level nonlinearities in the relationship between energy consumption and temperature.
Modeled energy responses to temperature changes reflect income changes and climate adaptation (e.g.,
installation of air conditioning in areas that currently have little penetration and more frequent operation
of existing air conditioning equipment). Similar to Carleton et al., the modeled energy-temperature
relationship for a local impact region is a function of conditions at that location. This allows the authors
to compute the additional impact of climate change on energy consumption, net of local factors (e.g.,
income) that will change in the future. Using the same income and climate projections as in Carleton et
al. (2022), Rode et al. compute projected impacts of climate change on electricity and other fuels
consumption under multiple emissions scenarios at the scale of ~25,000 global regions. To value these
impacts, the results presented in this report use country-level energy prices from the International Energy
Agency's (IEA) World Energy Outlook and Energy Prices and Taxes dataset. Prices are extrapolated into
the future based on the growth rates projected in the U.S. Energy Information Administration's Annual
Energy Outlook 2021. Specifically, based on the AEO projections, prices are assumed to grow at an annual
rate of -0.27% and 0.82% for electricity and other fuels, respectively. See CIL (2022) for more discussion.
The labor productivity component of the model captures the value of labor losses, as measured in labor
disutility, from responses in daily temperature (Rode et al. 2022). Evidence shows that workers in
industries such as agriculture, construction, manufacturing, transport, and utilities reduce their hours
worked when outdoor temperatures deviate from average temperatures.77 Daily variation in weather for
seven countries representing about 30 percent of the global population is used to econometrically
76 Specifically, projected changes in premature mortality in the U.S. are monetized using the same value of mortality
risk reduction as in the EPA's regulatory analyses ($4.8 million in 1990 (1990USD)) and adjusted for income growth
and inflation following current EPA guidelines and practice (EPA 2010) and consistent with EPA Science Advisory
Board (SAB) advice (see e.g., EPA 2011, OMB 2003), resulting in a 2020 value of $10.05 million (2020USD). Valuation
of mortality risk changes outside the U.S. is based on an extrapolation of the EPA value that equalizes willingness-
to-pay as a percentage of per capita income across all countries (i.e., using an assumed income elasticity of 1). The
use of a benefits transfer approach based on a positive income elasticity is consistent with the approach used in the
default version of the models and published studies used in this report (e.g., Rennert et al. 2022b, Carleton et al.
2022, Diaz 2016), and other academic literature. See Appendix A.6 for more discussion.
77 See Rode et al. (2021) for a listing of literature across many disciplines that have studied the effects of temperature
on worker performance and labor, dating back to Huntington (1922).
41
-------
EXTERNAL REVIEW DRAFT
estimate subnational labor supply responses to temperature changes. The labor response is estimated to
be an inverted U-shaped relationship, with lost labor occurring at extreme hot and cold temperatures, for
high-risk, weather-exposed sectors and low-risk sectors. The labor supply temperature response is
projected globally and over time, following Carleton et al. (2022) and Rode et al (2021). It includes
predicted shifts towards less weather-exposed industries as a function of average income per capita and
long-run average temperature, analogous to other forms of adaptation accounted for in Carleton et al.
(2022) and Rode et al (2021). The value of lost productivity is monetized as the compensating wage
increase needed to offset the temperature change's disutility.
DSCIM captures the net production impact of climate change in the agriculture sector by computing
projected impacts for six globally and regionally important staple crops that represent two thirds of global
crop caloric production: maize, wheat, rice, soybean, sorghum, and cassava (Hultgren et al. 2022). The
DSCIM reduced-form econometric approach simultaneously captures the combined impact of biophysical
crop responses and producer decision-making to account for the costs, benefits, and adoption rates of
producer adaptations as they are observed in practice around the world. This contrasts with prior analyses
that rely on agronomic process-based models to explicitly characterize the biophysical processes to
project yields. DSCIM accounts for several types of adaptation. First, the model allows for within-crop
adaptations such as varietal switching and other changes in production methods, such as irrigation,
fertilization, and planting dates. Second, in the monetization step, the results are multiplied by 0.45 to
account for crop switching and trade protective effects, from frictionless trade within continents and
global trade networks, based on an average of the estimates in prior research documenting these
quantities (e.g., Rising and Devineni 2020; Costinot et al. 2016; Gouel and Laborde 2021; Stefanovic et al.
2016). The DSCIM results presented in this report also account for the fertilization benefits of C02
emissions on crop yields based on established estimates in the literature (Moore et al. 2017).
Finally, the coastal component of DSCIM estimates damages resulting from sea level rise inundation in
coastal regions. As described in Section 2.2, the GMSL projections are based on the probabilistic FACTS
model that is being used in IPCC's AR6 report (Kopp et al. 2016, Garner et al. 2021). To generate a damage
function relating GMSL to welfare loss, probabilistic local mean sea level (LMSL) projections are used as
inputs to an updated version (Depsky et al. 2022) of the Coastal Impact and Adaptation Model (CIAM)
(Diaz 2016). These projections come from LocalizeSL (Kopp et al. 2017), using AR5 emissions trajectories.
The updated CIAM model (pyCIAM) estimates highly localized SLR related damages (Diaz 2016). CIAM is a
deterministic optimization model that chooses the least-cost adaptation strategy for each of the 9,000
coastal segments defined in the Sea Level Impacts Input Dataset by Elevation, Region, and Scenario
(SLIIDERS, Depsky et al. 2022)78 after accounting for local physical and socioeconomic characteristics.79
78 The SLIIDERS dataset provides details on local physical and socioeconomic characteristics. The original CIAM uses
12,148 coastal segments in the Dynamic Interactive Vulnerability Assessment (DIVA) database. The use of 9,000
segments in DSCIM is just the result of Depsky et al. (2022)'s re-optimization of the coastal segment choices (e.g., in
the original CIAM inputs, 10% of the 12,000 global segments were in French Polynesia).
79 In CIAM the adaptation choice set includes: (1) retreating inland from the coastline, (2) protecting coastal
communities and infrastructure, or (3) taking no adaptive measures. The decision maker first selects the lowest-cost
combination of these and then chooses the degree of investment in coastal defense against several different return
periods, under the assumption of perfect foresight about SLR conditions. Ongoing research is being developed by
Diaz and collaborators to refine the foresight assumptions and the resulting coastal damages from SLR.
42
-------
EXTERNAL REVIEW DRAFT
Damages are then estimated as the costs associated with the selected adaptation strategy plus the
residual damages due to inundation, wetland loss, and flooding.
Table 2.3.1: Current Coverage of Climate Damages in DSCIM
Sector
Health
Damage
Categories
Represented
Heat- and cold-
related mortality
Energy
Labor
Productivity
Expenditures for
electricity and
other direct fuel
consumption
Labor disutility
costs from labor
supply responses
to increased
temperature
Agriculture
Coastal
regions
Production
impacts for six
crops: maize, rice,
wheat, soybeans,
sorghum, and
cassava
Impacts of SLR as
realized through
inundation,
migration,
protection, dry
and wetland loss,
and mortality and
physical capital
loss from SLR
Empirical Basis for
Damage Function
Estimation
Subnational annual
mortality statistics for 40
countries covering 38%
of global population;
1990-2010 or longer for
most countries
Annual country-level
energy consumption
data (residential,
commercial, and
industrial) by energy
source for 146 countries,
1971-2010
Daily worker-level labor
supply data (minutes
worked) from 7
countries representing
nearly 30% of global
population
Subnational crop
production data for over
12,658 sub-national
administrative units
from 55 countries
Numerous empirical
findings are used to
parameterize the CIAM
process model for 9,000
coastal segments. (Low
levels of SLR in the
historical record prohibit
the use of a fully
empirical model)
Accounting for
Adaptation
Accounts for adaptative
effects of income
growth and estimates
the costs of adaptive
investments using a
revealed preference
approach
Accounts for both
climate- and
socioeconomics-driven
adaptive responses
Accounts for shifts in
workforce composition
to less weather-exposed
industries
Documentation
Carleton et al.
(2022)
Rode et al. (2021)
Rode et al. (2022)
Accounts for CO2
fertilization effects,
varietal switching,
changes in production
methods (e.g., irrigation,
fertilization, planting
dates), crop switching,
and trade effects
Reflects retreat or
protective infrastructure
and costs under an
optimal adaptation
scenario with perfect
foresight of SLR
Hultgren et al.
(2022)
Kopp et al. (2016)
and Garner et al.
(2021) for SLR; Diaz
(2016) and Depsky
et al. (2022) for
damages
43
-------
EXTERNAL REVIEW DRAFT
2.3.2 Damage Module Based on the Greenhouse Gas Impact Value Estimator (GIVE)
The second damage module used in this report istaken from the GIVE integrated assessment model (1AM).
GIVE is an open-source 1AM developed under the Resources for the Future Social Cost of Carbon Initiative
in collaboration with dozens of researchers from private and public institutions across the globe, spanning
a wide range of disciplines (Rennert et al. 2022b). The model was developed in direct response to the
National Academies (2017) recommendations surrounding needed improvements in the estimation of the
SC-GHG. The damage function component of the model is structured in such a way that it can
accommodate additional damage sectors underlying the estimation of the SC-GHG, making it particularly
attractive for incorporating future research and findings.80 Moreover, the model can accommodate
components with differing temporal and spatial resolutions. The model can be estimated deterministically
(fixed parameter) or in a Monte Carlo (random parameter) setting, sampling from socioeconomic, climate,
and damage function distributions to allow for uncertainty within and across each of its components. In
the model runs performed for this report, the outputs of the socioeconomic module and the GMST
projections from the climate module described above serve as inputs to the damage function components
of GIVE.
At present, GIVE includes estimation of climate damages occurring in four sectors or impact categories:
health, energy, agriculture, and coastal regions.81 The damage functions reflect recent scientific
advancements in the peer-reviewed literature. Table 2.3.2 summarizes key elements of GIVE's damage
function estimation methods in in each of these four sectors. The health damage function is based on a
recent study authored by a collaboration of public health, epidemiology, climatology, and economics
experts in response to the 2017 National Academies' recommendations (Cromar et al. 2022). The authors,
along with an additional panel of convening experts, conducted a systematic review and meta-analysis of
health impacts related to climate change. Then, regionally resolved all-cause mortality estimates from
increases in temperature were generated through a random-effects pooling of studies that were
identified in the systematic review.82 Net changes in mortality risk associated with increased average
annual temperatures were estimated for all global regions varying in their effect size and uncertainty
across each of the 9 regions. The resulting changes in premature mortality are mapped to country-specific
baseline mortality projections and rates such that premature mortality from global climate change is
unique to all 184 countries. Uncertainty in the mortality damage function is parametric and sampled from
the region-specific coefficient that relates GMST to changes in premature mortality. The GIVE model
monetizes the projected changes in premature mortality using country-level population-average
measures of the willingness-to-pay for mortality risk reduction (Rennert et al. 2022b), consistent with
methodology used in the DSCIM model runs presented in this report and described above.
80 The GIVE model is built on the Mimi.jl platform, an open-source package for constructing modular integrated
assessment models, www.mimiframework.org. GIVE is written using the Julia programming language which allows
for extremely fast estimation times.
81 The modular nature of GIVE offers a straightforward way to add other damage functions and sectors. For example,
nonuse biodiversity losses are currently under development based on an approximation of Brooks and Newbold
(2014).
82 A total of 33 unique health studies, most of which were extensive multi-locational studies, were included in Cromar
et al. (2022). Studies were predominately from North America, Europe, and East Asia and thus some of the more
populous parts of the world were underrepresented (Cromar et al. 2022).
44
-------
EXTERNAL REVIEW DRAFT
The energy damage function component of GIVE is based on a recent multidisciplinary study that
estimates the relationship between changes in building energy expenditure (net heating and cooling
expenses) and changes in local temperature and climate (Clarke et al. 2018). That study used the Global
Change Analysis Model (GCAM) that models regional changes in heating and cooling expenditures as a
proportion of regional gross domestic product resulting from changes in regional temperatures. That is,
for each of the 12 GCAM regions, Clarke et al. (2018) find an approximately linear relationship between
degrees of temperature change and net change in energy expenditures. Reflecting this, the climate-
expenditure relationship from Clarke et al. (2018) is estimated within GIVE by a regional linear regression
that yields region-specific damage functions to estimate changes in net energy expenditures within each
of the 184 countries in the model.
The agriculture damage function component of GIVE follows Moore et al. (2017). It is derived using (1) a
meta-analysis of over 1,000 published temperature-yield response estimates from 55 unique studies, and
(2) an open-source computable general equilibrium (CGE) model that estimates the welfare consequences
(as equivalent variation) of climate-induced productivity changes, accounting for adjustments in
agricultural markets including trade patterns, consumption, and production. The productivity changes (for
maize, rice, wheat, and soybeans) are based on biophysical crop impacts documented in the literature.
Productivity impacts include both within-crop adaptations (e.g., varietal and planting date changes) as
well as C02 fertilization using estimates of the size of these effects from the meta-analysis. Welfare
changes at 1, 2 and 3 degrees of warming calculated from the CGE model give damage functions for 140
regions. GIVE maps the regions to all 184 countries for country-level effects on crop production. Within
GIVE, the non-parametric uncertainty provided in Moore et al. (2017) is converted to parametric
uncertainty and used in the Monte Carlo estimation.
The fourth damage sector in GIVE connects the BRICK sea level rise (SLR) model (Wong et al. 2017) and
the CIAM model (Diaz 2016) to estimate SLR induced coastal damages from temperature change. As
described in Section 2.2, GMST and ocean heat content from FaIR 1.6.2 are used as inputs to BRICK to
generate projections of GMSL. As in the damage module described above based on DSCIM, the GMSL
projections are downscaled to a 1-degree grid (Slangen et al. 2014) and used as inputs to CIAM to estimate
local adaptation decisions and their associated costs.83 Since CIAM is a deterministic model, uncertainty
in coastal damages is the result of uncertainty in BRICK that arises due to the RFF-SP probabilistic emission
scenarios and sampled climate and sea-level parametric uncertainty.
83 As noted in Section 2.3.1, CIAM includes 12,148 unique coastal segments. Of these 11,835 correspond to countries
included in the GIVE model. See Rennert et al. (2022b) for a full description.
45
-------
EXTERNAL REVIEW DRAFT
Table 2.3.2: Current Coverage of Climate Damages in GIVE
Sector
Health
Energy
Agriculture
Damage Categories
Represented
Heat- and cold-
related mortality risk
Expenditures for
space heating and
cooling in buildings
Welfare changes
from temperature
driven changes in
production of four
crops: maize, rice
wheat, and soybeans
Coastal
regions
Empirical
Basis/Methodology
Pooled effect estimates
(36 studies across 9
regions) for changes in
temperature on
mortality risk, by region
Regional costs of energy
consumption,
temperature, and
climate
Meta-analysis of 1010
yield effect estimates
from 55 studies and
computable general
equilibrium (CGE) model
of trade
Impacts of SLR as
realized through
inundation,
migration, protection,
dry and wetland loss,
and mortality and
physical capital loss
from SLR
Accounting for Adaptation Documentation
Numerous empirical
findings are used to
parameterize the CIAM
process model for
11,835 coastal segments
Observed responses to
changes in temperature
are assumed to persist into
the future
Implicit in the regional
relationship between
increases in energy
expenditures and
temperature
Explicit in the estimation of
the damage function
through assumed changes
in on-farm, within-crop,
management practices.
Adaptive adjustments in
agricultural markets
through changes in crops,
trade, consumption, and
production patterns.
Reflects retreat or
protective infrastructure
and costs under an optimal
adaptation scenario with
perfect foresight of SLR
Cromar et al.
(2022)
Clarke et al.
(2018)
Moore et al.
(2017)
Wong et al.
(2017) for SLR;
Diaz (2016) for
damages
The damage functions in DSCIM and GIVE represent substantial improvements relative to the damage
functions underlying the SC-GHG estimates used by the EPA to date in reflecting the forefront of scientific
understanding about how temperature change and SLR lead to monetized net (market and nonmarket)
damages for several categories of climate impacts. The models' spatially explicit and sector-specific
modeling of relevant processes allows for improved understanding and transparency about mechanisms
through which climate impacts are occurring and how each sector contributes to the overall results,
consistent with the National Academies' recommendations. DSCIM addresses common criticisms related
to the damage functions underlying current SC-GHG estimates (e.g., Pindyck 2017) by developing multi-
sector, empirically grounded damage functions.84 The damage functions in the GIVE model offer a direct
implementation of the National Academies' near-term recommendation to develop updated sectoral
damage functions that are based on recently published work and reflective of the current state of
knowledge about damages in each sector. Specifically, the National Academies noted that "[t]he literature
on agriculture, mortality, coastal damages, and energy demand provide immediate opportunities to
84 Note that Pindyck has consistently noted that modeling and damage category considerations are not a reason to
abandon the social cost of greenhouse gases; Pindyck has consistently supported updating the IWG's past estimates
(Pindyck 2013, 2017, 2019, 2021).
46
-------
EXTERNAL REVIEW DRAFT
update the [models]" (National Academies 2017, p. 199), which are the four damage categories currently
in GIVE. A limitation of both models is that the sectoral coverage is still limited. Neither DSCIM nor GIVE
yet accommodate estimation of other categories of temperature driven climate impacts (e.g., storm
damage, morbidity, conflict, migration, biodiversity loss); damages that result from physical impacts other
than temperature and SLR (e.g., changes in precipitation, ocean acidification, non-temperature-related
mortality such as diarrheal disease and malaria); or many feedbacks and interactions across sectors and
regions that can lead to additional damages.85,86 DSIM and GIVE do account for the most commonly cited
benefits associated with C02 emissions and climate change - C02 crop fertilization and declines in cold
related mortality. As such, the GIVE- and DSCIM-based results presented in this report provide a partial
estimate of future climate damages resulting from incremental changes in C02, CH4, and N20. DSCIM and
GIVE developers have work underway on other sectors that may be ready for consideration in future
updates (e.g., morbidity and biodiversity). DSCIM and GIVE are structured so that future research can be
reasonably incorporated into their damage modules.
2.3.3 Damage Module Based on a Meta-Analysis Approach
Given the still relatively narrow sectoral scope of the recently developed DSCIM and GIVE models, this
report includes a third damage function that reflects a synthesis of the state of knowledge in other
published climate damages literature. Studies that have employed meta-analytic techniques87 offer a
tractable and straightforward way to combine the results of multiple studies into a single damage function
that represents the body of evidence on climate damages that pre-date CIL and RFF's research initiatives.
Meta-analysis is a common tool in empirical research. Within the climate change literature, meta-analyses
have been used to analyze physical and sector impacts (e.g., Moore et al. 2017, Hoffmann et al. 2020,
Cromar et al. 2022) and to directly estimate aggregate global damage functions. The first use of meta-
analysis to combine multiple climate damage studies was done by Tol (2009) and included 14 studies. The
studies in Tol (2009) served as the basis for the global damage function in DICE starting in version 2013R
(Nordhaus 2014). The damage function in the most recent version of DICE, DICE 2016, is from an updated
meta-analysis based on a rereview of existing damage studies and included 26 studies published over
1994-2013 (Nordhaus and Moffat 2017). Howard and Sterner (2017) provide a more recent peer-reviewed
meta-analysis of existing damage studies (published through 2016) and account for additional features of
the underlying studies. They address differences in measurement across studies by adjusting estimates
such that the data are relative to the same base period. They also address issues related to double
counting by removing duplicative estimates. Dependence across climate-damage estimates can arise over
time due to the common practice of calibrating climate-model damage functions based on previous
estimates in the climate damage literature. Howard and Sterner's review identified 35 studies that meet
85 The one exception is that the agricultural damage function in DSCIM and GIVE reflects the ways that trade can
help mitigate damages arising from crop yield impacts.
86 See Section 4.2 for more discussion of omitted categories of climate impacts and associated damages.
87 Meta-analysis is a statistical method of pooling data and/or results from a set of comparable studies of a problem.
Pooling in this way provides a larger sample size for evaluation and allows for a stronger conclusion than can be
provided by any single study. Meta-analysis yields a quantitative summary of the combined results and current state
of the literature.
47
-------
EXTERNAL REVIEW DRAFT
their a priori selection criteria, of which 15 were dropped because they duplicated studies already in the
sample. Their final sample is drawn from 20 studies that were published through 2015.
Howard and Sterner (2017) present results under several specifications, and their analysis shows that their
estimates are somewhat sensitive to defensible alternative modeling choices. Howard and Sterner's main
specifications vary across two dimensions: (1) whether the sample includes estimates from studies that
consider large temperature changes (i.e., above 4°C), and (2) whether the econometric specification
explicitly accounts for different damage channels underlying the studies, such as studies that attempt to
account for the effect of climate impacts on economic productivity, and whether or not the estimates of
those damage channels should be additive to the primary damage estimate in the model.
Regarding the first dimension, this report focuses on a specification that includes estimates across the full
range of temperature changes considered in the underlying studies. Howard and Sterner's reasoning for
considering only estimates for temperature changes below 4°C is that, in their modeling, most present
value damages occur before 2100 and at or below 4°C.88 Applying the same logic would lead to the
opposite conclusion in the current modeling framework. After incorporating major advancements in the
socioeconomics, climate and discounting modules, as discussed in sections 2.1, 2.2, and 2.4, a significant
share of the temperature anomaly distribution exceeds 4°C based on RFF-SPs and FAIR1.6.2 over the
modeling horizon (2020 to 2300) and a significant amount of estimated discounted damages occur after
2100 (see Section 3). The coefficient estimate on the temperature variable in the specification in Howard
and Sterner (2017) used in this report (i.e., the specification that includes estimates of damages at all
temperatures, including those above 4°C) is smaller in magnitude than in the specification which limits
the analysis to studies that estimate damages at temperatures less than 4°C. Thus, the specification used
in this report reflects a more conservative estimate of the relationship between temperature and climate
damages, and thereby leads to a lower estimate of the SC-GHG, all else equal.
Regarding the second dimension, this report focuses on Howard and Sterner's estimation of combined
damage channels—the primary damage coefficient in their model. This choice, to exclude the coefficients
on catastrophic and productivity effects, is consistent with the authors' recommendations in the
published paper and follows the method Nordhaus (2019) uses to adjust the default damage function in
DICE 2016 to reflect the findings of Howard and Sterner's meta-analysis. The authors' rationale for
excluding the estimated coefficients on the control variables89 for catastrophic damages and productivity
impacts in the primary specification of the damage function was "because of their mixed [statistical]
significance and volatility across the various specifications." The catastrophic damages coefficient is
identified by five older studies which, while illustrative about the potential importance of such effects, are
not grounded in empirical evidence or explicit modeling of tipping elements and other effects
contemplated by the authors to lead to catastrophes.90 There is a need for improved methods for
88 As noted in the published paper, "...the majority (approximately two-thirds) of the 2015 SCC estimate for DICE-
2013R correspond to impacts occurring this century....for which estimates for approximately 4°C or less are more
germane" (Howard and Sterner 2017, p. 220).
89 These control variables indicate Howard and Sterner's categorization of whether the underlying damage estimates
account for potential for "catastrophic" impacts or account for the effects of climate change on economic growth.
90 The 5 studies from which Howard and Sterner (2017) take damage estimates that were considered to include
catastrophic damages were: Nordhaus (2014), Nordhaus (2008), Weitzman (2012) via Ackerman et al. (2012),
Ackerman at al. (2012) adjusting Hanemann (2008), Meyer and Cooper (1995).
48
-------
EXTERNAL REVIEW DRAFT
quantifying and incorporating these types of important elements of damages in future updates (e.g.,
through modeling specific tipping points and earth system feedback effects). See section 3.2 for further
discussion of these considerations.
Productivity damages in Howard and Sterner (2017) are identified by four studies (2 statistical and 2 CGE)
and the coefficient on the productivity indicator is estimated to be positive but not statistically different
from zero in any of the specifications.91 There is an ongoing investigation in the literature of whether
temperature effects on the economy are only temporary or persistent—with empirical findings sensitive
to model specification. Over the past decade, a host of empirical studies have found evidence of
temperature changes having persistent effects on the economy (e.g., Dell et al. 2012; Burke et al. 2015;
Deryugina and Hsiang 2017; Burke and Tanutama 2019; Colacito et al. 2019; Henseler and Schumacher
2019; Kahn et al. 2021; Kumar and Khanna 2019; Bastien-Olvera et al. 2022); this is an important finding
because even small changes in economic growth rates accumulate into large economic impacts overtime.
However, other recent studies have failed to identify conclusive evidence of persistent effects of
temperature changes (Newell et al. 2021, Kalkuhl and Wenz 2020). Given that the question of impact
persistence remains largely unresolved in the empirical literature to date, and given the statistical
insignificance of the estimated coefficient on the productivity indicator in the published Howard and
Sterner meta-analysis, the SC-GHG estimates presented in this report do not rely on Howard and Sterner's
specifications that include productivity effects. This is consistent with the authors' recommendations in
the published paper, to only consider the inclusion of the productivity impact in sensitivity analysis.92
However, this potentially important effect is worthy of additional study and the EPA will continue to follow
advances in the literature on methodologies for identifying productivity effects of climate change. Finally,
unlike Howard and Sterner (2017), the model runs performed for this report do not adopt a 25% adder
(as used in the DICE model (e.g., Nordhaus 2017b)) to account for unknown or missing damages for the
meta-analysis based damage module. Taken together, this report uses the most conservative damage
function specification (that excludes duplicate studies) from Howard and Sterner (2017).93
2.3.4 Comparing the Three Damage Modules
Each of the three damage modules - based on DSCIM, GIVE, and the Howard and Sterner (2017) meta-
analysis - is separately estimated in combination with the socioeconomics, climate, and discounting
modules described elsewhere in this section. The sectoral damage modules in GIVE and DSCIM are based
91 The term "productivity" used in the Howard and Sterner (2017) damage function is distinct from the empirically
grounded micro-economic labor productivity described in the DSCIM damages model. Instead, productivity in
Howard and Sterner (2017) relates to the ongoing debate about persistence in damages as measured by changes in
economic growth over time.
92 Howard and Sterner (2017) conclude that "...given the debate over the impact of climate change on productivity
and economic growth (Dell et al. 2012; Burke et al. 2015; Howard [and Sylvan] 2015), we recommend conducting an
analysis of sensitivity to the inclusion of the productivity impact."
93 This specification of Howard and Sterner's results (i.e., using the estimated temperature coefficient in specification
7 presented in Table 2 of their paper) is also provided as an alternative damage function option in the GIVE model
(Rennert et al. 2022b). That is, when the Howard and Sterner (2017) damage function is used within the GIVE model,
the other damage sectors (agriculture, mortality, energy, and coastal) are turned off and the Monte Carlo simulation
samples from all relevant model parameter distributions including those underlying the Howard and Sterner (2017)
meta-analysis damage parameters.
49
-------
EXTERNAL REVIEW DRAFT
on different underlying information, data sources, and estimation methods.94 GIVE and DSCIM are both
independent lines of evidence from the meta-analysis-based damage module since the studies underlying
each sectoral damage modules in GIVE and DSCIM are not included in Howard and Sterner's (2017) final
sample of studies. Figure 2.3.2 illustrates the shape of the damage function across the three models.
Specifically, the figure presents projections of total damages from climate change in 2100 as a function of
GMST change. The points represent each trial of the Monte Carlo simulation where the socioeconomic
and climate module parameters are consistent across damage modules (i.e., the first trial of DSCIM takes
the same socioeconomic pathways and climate parameters as the first trial of GIVE and the meta-analysis-
based damage function). The global damage functions shown here are generated using estimated
damages in 2100 (the points) and regressing on temperature and temperature squared in 2100 at the
mean (solid line), and quantile regressions at the median (dashed lines), 5th to 95th (dark shade) and 1st to
99th (light shade) percentiles.
As seen in Figure 2.3.2, there are notable differences between the damage functions. On average, DSCIM
estimates lower damages but predicts a more rapidly increasing damage function beyond 4 degrees
Celsius, compared to GIVE that has increasing but consistent damages throughout the temperature range.
The meta-analysis-based damage function reflects the explicit quadratic nature of the published Howard
and Sterner (2017) damage function. Section 3 presents the resulting SC-GHG estimates based on each
damage module combined with the socioeconomic and climate modules and discusses the importance of
omitted climate impacts and associated damages.
94 Only one component of the methodology for calculating coastal damages is common across the two models. Both
DSCIM and GIVE rely on the CIAM model developed by Diaz (2016) to estimate the economic damages resulting from
projections of SLR.
50
-------
EXTERNAL REVIEW DRAFT
Figure 2.3.2: Annual Consumption Loss as a Fraction of Global GDP in 2100 Due to an Increase in Annual
Global Mean Surface Temperature in the three Damage Modules
DSCIM
10%
6%
4%
2%
0%
GIVE
o
u
M—
o
NP
0s
20%
15%
10%
"5 5%
-O
o
2 4
M eta-Ana lysis
15%
10%
5%
0%
Global Surface Temperature Change
Relative to 1850-1900 (°C)
GDP loss functions are generated using estimated damages in 2100 (points) and regressing on temperature and temperature
squared at the mean (solid line), and quantile regressions at the median (dashed lines), 5th to 95th (dark shade) and 1st to 99th
(light shade). 5,000 of the 10,000 points for each module are randomly selected to simplify the presentation of damages. DSCIM
estimates damages relative to global mean surface temparatures between 2000-2010 and was normalized here to 1850-1900 to
be consistent with GIVE and the Meta-Analysis. GIVE and the Meta-Analysis presented here include the full uncertainty underlying
each module in the Monte Carlo analysis, DSCIM observations present climate and socioeconomic uncertainty (no statistical
uncertainty from the underlying damage functions). The IPCC (2021a) notes that present day global mean surface temperatures
in the year 2020 are around 1.1 °C above preindustrial (1850-1900) levels.
51
-------
EXTERNAL REVIEW DRAFT
2.4 Discounting Module
GHG emissions are stock pollutants, where damages result from the accumulation of the pollutants in the
atmosphere overtime. Because GHGs are long-lived, subsequent damages resulting from emissions today
occur over many decades or centuries, depending on the specific GHG under consideration.95 In
calculating the SC-GHG, the stream of future marginal damages, as estimated by the damage modules
discussed in Section 2.3, is calculated in terms of reduced consumption (or monetary consumption
equivalents). Then that stream of future damages is discounted to its present value in the year when the
additional unit of emissions was released. Given the long time horizon over which the damages are
expected to occur, the approach to discounting greatly influences the present value of future damages.
Arrow et al. (1995) outlined two main approaches to determine the discount rate for climate change
analysis, which they labeled "descriptive" and "prescriptive." The descriptive approach reflects a positive
(non-normative) perspective based on observations of people's actual choices - e.g., savings versus
consumption decisions over time, and allocations of savings among more and less risky investments.
Advocates of this approach generally call for inferring the discount rate from market rates of return
because "no justification exists for choosing [a social welfare function] different from what
decisionmakers actually use" (Arrow et al. 1995).
In addition, the Kaldor-Hicks potential compensation test - one theoretical foundation for the benefit-
cost analyses in which the SC-GHG will be used - suggests that market rates should be used to discount
future benefits and costs. This is because the market interest rate would govern the returns potentially
set aside today to compensate future individuals for the climate damages that they bear (e.g., Just et al.
2004). The word "potentially" indicates that there is no assurance that returns will be set aside to provide
compensation, and the very idea of compensation is difficult to define in the intergenerational context.
On the other hand, societies provide compensation to future generations through investments in human
capital and the resulting increase in knowledge, infrastructure and other physical capital, and the
maintenance and preservation of natural capital.
In contrast, the prescriptive (normative) approach specifies a social discount rate that formalizes the
normative judgments that the decision-maker wants to incorporate into the policy evaluation. That is, it
defines from the decision-maker's perspective how interpersonal comparisons of utility should be made
and how the welfare of future generations should be weighed against that of the present generation.
Ramsey (1928), for example, argued that it is "ethically indefensible" to apply a positive pure rate of time
preference to discount values across generations.
Additional concerns motivate adjusting descriptive discount rates. Future generations' preferences
regarding consumption versus environmental amenities may not be the same as those today, raising
concerns about using the current market rate on consumption to discount future climate-related
damages. Furthermore, markets for relatively riskless assets with a maturity similar to an
intergenerational horizon, akin to the horizon over which climate change impacts are realized, do not exist
(Gollier and Hammit 2014). Others argue that the discount rate should be below market rates to correct
95 "GHGs, for example, CO2, methane, and nitrous oxide, are chemically stable and persist in the atmosphere over
time scales of a decade to centuries or longer, so that their emission has a long-term influence on climate. Because
these gases are long lived, they become well mixed throughout the atmosphere" (IPCC 2007b).
52
-------
EXTERNAL REVIEW DRAFT
for market distortions and uncertainties or inefficiencies in intergenerational transfers of wealth
(Schwartz and Howard 2022).
Further, a concern about discount rates developed using both the descriptive and prescriptive approaches
is that they tend to obscure important heterogeneity in the population. For instance, many individuals
smooth consumption by borrowing with credit cards that have relatively high rates. Some are unable to
access traditional credit markets and rely on payday lending operations or other high-cost forms of
smoothing consumption. This behavior may reflect rational intertemporal preferences, or it may reflect
other factors such as present bias, lack of financial literacy, and other distortionary effects of poverty
(Haushofer and Fehr 2014; Lusardi and Mitchell 2014). Nevertheless, whether one puts greater weight on
the prescriptive or descriptive approach, the high interest rates that credit-constrained individuals accept
suggest that some account should be given to the discount rates revealed by their behavior.
The EPA's analyses rely primarily on the descriptive approach to inform the choice of a discount rate for
SC-GHG estimation, consistent with the rationale outlined in IWG TSDs (e.g., IWG 2010, 2021) and EPA's
economic analysis guidelines (EPA 2010). With a recognition of its limitations, the IWG found this
approach to be the most defensible and transparent given its consistency with both the standard
contemporary theoretical foundations of benefit-cost analysis and the approach recommended by OMB's
existing guidance.
In 2010, the IWG specifically elected to use three constant discount rates: 2.5, 3, and 5 percent per year.
The 3 percent rate was included as consistent with the default recommendation provided in OMB's Circular
A-4 (OMB 2003) guidance for the consumption rate of interest. The IWG found that the consumption rate
of interest is the correct discounting concept to use when the future damages from climate change are
estimated in consumption-equivalent units, as is done in the lAMs used to estimate the SC-GHG.96 The 3
percent rate was roughly consistent with the average rate of return for long-term Treasury notes
calculated at the time the OMB guidance was published. The upper rate of 5 percent was included to
represent the possibility that climate-related damages are positively correlated with market returns,
which would imply a certainty-equivalent97 risk-adjusted rate higher than the consumption rate of
interest. The low rate, 2.5 percent, was included to incorporate the concern that interest rates are highly
uncertain over time, which would imply a risk-free certainty equivalent rate lower than the consumption
rate of interest. Additionally, a rate below the consumption rate of interest would also be justified if the
return to investments in climate mitigation is negatively (or weakly) correlated with the overall market
rate of return. The use of this lower rate was also deemed responsive to certain judgments based on the
prescriptive or normative approach for selecting a discount rate and related ethical objections about rates
of 3 percent or higher. Further details about selecting these rates are presented in the 2010 TSD (IWG
2010).
Based on a review of the literature and data on consumption discount rates, the public comments received
on individual EPA rulemakings, and the February 2021 TSD (IWG 2021), and the National Academies (2017)
96 Appendix A.2 provides additional detail on why the consumption discount rate is the appropriate rate to be used
in estimating the SC-GHG.
97 The certainty-equivalent discount rate is the certain discount rate that is equivalent to an uncertain discount rate
in terms of the discount factor over a particular horizon. See National Academies (2017) for more explanation of this
and other discounting terminology.
53
-------
EXTERNAL REVIEW DRAFT
recommendations for updating the discounting module, this report uses a new set of discount rates that
reflect more recent data on the consumption interest rate. The approach presented in this report
continues to rely on a descriptive approach to discounting but more fully captures the role of uncertainty
in the discount rate in a manner consistent with the other modules. Specifically, rather than using a
constant discount rate, the evolution of the discount rate over time is defined following the latest
empirical evidence on interest rate uncertainty and using a framework originally developed by Ramsey
(1928) that connects economic growth and interest rates. The Ramsey approach explicitly reflects (1)
preferences for utility in one period relative to utility in a later period and (2) the value of additional
consumption as income changes. The resulting dynamic discount rate provide a notable improvement over the
constant discount rate framework for SC-GHG estimation. Specifically, it provides internal consistency within the
modeling and a more complete accounting of uncertainty98, consistent with economic theory (Arrow et al. 2013,
Cropper et al. 2014) and the National Academies (2017) recommendation to employ a more structural, Ramsey-
like approach to discounting that explicitly recognizes the relationship between economic growth and discounting
uncertainty. The following sections provide an overview of the Ramsey discounting formula and then describe the
calibration of the new set of dynamic discount rates.
2.4.1 The Ramsey Formula
The Ramsey formula for discounting is derived from work by Frank Ramsey (1928) and others (Cass 1965,
Koopmans 1963) on the optimal level of consumption and saving. The formula describes the optimal
consumption discount rate as a function that explicitly reflects: (1) preferences for utility in one period
relative to utility in a later period (called the "pure rate of time preference"); and (2) the value of additional
consumption as income changes. These factors are combined in the equation
rt = P + r\gt, (.2.4.1)
where rt is the consumption discount rate in year t, p is the pure rate of time preference, r| is the elasticity
of marginal utility with respect to consumption, and gt is the representative agent's consumption growth
rate in year t."
The pure rate of time preference, p, is the rate at which the representative agent discounts utility in future
periods due to a preference for utility sooner rather than later. The elasticity of marginal utility with
respect to consumption, r|, defines the rate at which the well-being from an additional dollar of
consumption declines as the level of consumption increases. In this context, it is common to assume that
98 As noted in Circular A-4, "the longer the horizon for the analysis," the higher the "uncertainty about the
appropriate value of the discount rate" (OMB 2003).
99 The economic framework in this report implicitly assumes an exogenous fixed savings rate. With this assumption
consumption growth and income (GDP) growth are equivalent. A more restrictive assumption that leads to the same
result would be to assume that the savings rate is zero and consumption is equivalent to income. Relaxing the fixed
savings rate assumption would require adding further complexity to calculate the optimal savings rate in each year.
54
-------
EXTERNAL REVIEW DRAFT
well-being can be described by an isoelastic utility function, where utility, u, is a power function with
respect to consumption, ct, such that
7-77
u(ct) = y-— ¦ (2.4.2)
I — rj
This function implies that the elasticity of marginal utility with respect to consumption is a constant value
(i.e., for a given percent increase in baseline consumption the benefit of an additional unit of consumption
decreases proportionally). The per capita consumption growth rate, gt, defines the projected change in
consumption per capita over time. Under the common assumption of a constant savings rate, gt would
be expected to change with income over time.100 When using the Ramsey formula to estimate the SC-
GHG, the per capita consumption growth rate, gt is calculated net of baseline climate change damages as
estimated by the damage modules described in Section 2.3.
The use of the Ramsey formula provides internal consistency within the modeling between the socio-
economic scenarios and the discount rate. With uncertainty in the per capita consumption growth rate,
the Ramsey discount rate becomes a dynamic parameter within the modeling framework that reflects
how uncertainty about future conditions has implications for how future impacts are valued. Gollier
(2014) showed that when there is uncertainty in future consumption growth, the distribution of discount
rates defined by the Ramsey formula will have a certainty-equivalent risk-free discount rate path that
declines overtime, under standard assumptions about individual preferences. This is particularly true
when shocks to consumption growth are positively correlated over time, as they are in the probabilistic
scenarios described in Section 2.1. The declining certainty-equivalent risk-free discount rate implied by
the Ramsey formula reflects that additional climate change damages are a greater burden to society in
future states of the world with relatively lower economic growth. Damages in low economic growth states
of the world are given greater weight than if those same damages were realized in a future state of the
world with relatively higher economic growth, all else equal (Gollier and Weitzman 2010). The declining
certainty-equivalent discount rate implied by the Ramsey formula is also consistent with the empirical
literature on discount rates under uncertainty (e.g., Newell and Pizer 2003, Bauer and Rudebusch 2021).101
The use of the Ramsey formula also provides internal consistency when accounting for the effect of
correlations between climate change damages and economic growth. The correlation between climate
change damages and future economic uncertainty is important in determining the appropriate discount
rate. If climate change damages are positively correlated with economic growth (e.g., if the willingness to
pay to avoid climate impacts increases with income or emissions), then the risk of climate change impacts
being worse than expected is greater when the world is relatively wealthier than anticipated. In this case,
less weight should be placed on those future impacts. Conversely, if climate change damages are
negatively correlated with economic growth (e.g., if less adaptation is available at lower incomes or if
climate damages slow economic growth), then the risk of climate change impacts being worse than
100 More information on the derivation of the Ramsey formula can be found in Dasgupta (2020).
101 The approach employed in this report should not be confused with applying an exogenously specified declining
discount rate. There are similarities, in that incorporating economic uncertainty in the Ramsey equation yields a
declining certainty-equivalent discount rate. However, the application of an exogenously specified declining discount
rate would fail to capture the way in which correlations between uncertain climate damages and uncertain economic
growth affect estimates of the SC-GHG.
55
-------
EXTERNAL REVIEW DRAFT
expected is greater when the world is relatively less wealthy than expected. In this converse case, more
weight should be placed on those future impacts. Using the Ramsey formula for discounting in conjunction
with probabilistic scenarios and modeling climate change damages under uncertainty ensures that the
correlation between climate change damages and economic growth within the model is appropriately
captured in the SC-GHG estimates. It allows for an internally consistent approach to capturing these
effects, and exogenous adjustments to the discount rate are not required.
Incorporating dynamic discount rates through the application of the Ramsey formula remains widely used
in the peer reviewed literature and is consistent with the National Academies' (2017) recommendations
on discounting. It provides important improvements over the use of a static discount rate and
incorporates connections between important components of the modeling. While offering an important
improvement, the Ramsey formula is an approximation of complex economic processes and future
research may provide methodological advancements that further improve the representation of those
processes within dynamic discount rates.
2.4.2 Calibration of Discount Rate Distributions
The National Academies (2017) recommended that the IWG "choose parameters for the Ramsey formula
that are consistent with theory and evidence and that produce certainty-equivalent discount rates
consistent, over the next several decades, with consumption rates of interest." The SC-GHG estimates
presented in this report adopt a descriptive approach to calibrating the Ramsey parameters, meaning that
the parameters are calibrated based on observed interest rate data, consistent with the National
Academies' recommendation. Specifically, the parameters are calibrated following the Newell et al. (2022)
calibration approach, as applied in Rennert et al. (2022a, 2022b). Under this approach, the parameters
are calibrated such that the decline in the certainty-equivalent discount rate path matches the latest
empirical evidence on interest rate uncertainty estimated by Bauer and Rudebusch (2020, 2021). The
parameters are also calibrated such that the average of the certainty-equivalent discount rate over the
first decade matches a specified near-term consumption rate of interest. As described below, given the
uncertainty about the appropriate starting rate, three near-term target rates (1.5, 2.0, and 2.5 percent)
are used based on multiple lines of evidence on observed interest rate data. The calibration of the
parameters is carried out using the same probabilistic socioeconomic scenarios presented in Section 2.1
to ensure internal consistency. This approach results in three discount rate paths and is consistent with
the National Academies (2017) recommendation to use three sets of Ramsey parameters that reflect a
range of near-term certainty-equivalent discount rates consistent with theory and empirical evidence on
consumption rate uncertainty, and uncertainty surrounding long-run socioeconomic and emissions
projections.
Specifying the near-term target rates. The near-term certainty-equivalent discount rate is calibrated
based on observed interest rate data. Estimates of the risk-free consumption interest rate - used to
represent temporal preferences in benefit-cost analysis - have generally focused on historical returns to
long-term Treasury securities backed by the faith and credit of the U.S. Government. In particular, the
estimates of the consumption interest rate published in OMB's Circular A-4 in 2003 are based on the real
56
-------
EXTERNAL REVIEW DRAFT
rate of return on 10-year Treasury Securities102 from the prior 30 years (1973 through 2002). However,
there has been a substantial and persistent decline in real interest rates over the past four decades. Recent
research has found that the decline in real interest rates reflects a reduction in the equilibrium real
interest rate, suggesting that lower real interest rates are expected to persist (Bauer and Rudebusch
2020). These changes indicate the need for new estimates of the near-term consumption rate of interest
that incorporate recent data.
From 2003 onwards, it is possible to use the 10-Year Treasury Inflation-Protected Securities (TIPS)103 as a
measure of the real rate of return on 10-Year Treasury Securities. Prior to the TIPS introduction, nominal
returns on Treasury securities needed to be adjusted for inflation. To use the consumption interest rate
as an estimate of social preferences for trading off consumption over time, the inflation adjustment
should reflect investor expectations about inflation over the maturity period to produce an estimate of
the tradeoff investors believe they are making. There are multiple approaches to adjusting the nominal
rate for inflation expectations over the maturity of the security at the time of purchase. Three measures
of inflation expectations are considered. The first is a ten-year moving average of the consumer price
index (CPI)104 prior to the year of the security issuance. This measure assumes that recent trends in
inflation inform expectations over future inflation. The second is a ten-year moving average of inflation
expectations as measured by the Livingston Survey, which is a survey of forecasters about key economic
variables.105 This approach has been used in the economics literature to measure inflation expectations
when examining real rates of return (e.g., Newell and Pizer, 2003). The third is the perceived inflation
target rate (PTR) from the Federal Reserve's FRB/US model. The PTR is an expectation of long-run inflation
estimated from the Survey of Professional Forecasters (SPF). For years before the inception of the SPF,
the PTR is estimated econometrically.106 The PTR has also been used in the economics literature as a
measure of inflation expectations when examining real rates of return (e.g., Fuhrer et al. 2012, Bauer and
Rudebusch 2017, Bauer and Rudebusch 2020).
102 Board of Governors of the Federal Reserve System (US), Market Yield on U.S. Treasury Securities at 10-Year
Constant Maturity, Quoted on an Investment Basis, Series name: DGS10, retrieved from FRED, Federal Reserve Bank
of St. Louis; httpsi//fred.stlouisfed.org/series/DGS 10
103 Board of Governors of the Federal Reserve System (US), Market Yield on U.S. Treasury Securities at 10-Year
Constant Maturity, Quoted on an Investment Basis, Inflation-Indexed, Series name: DFII10, retrieved from FRED,
Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/serjes/DFM10
104 U.S. Bureau of Labor Statistics, Consumer Price Index for All Urban Consumers: All Items in U.S. City Average,
Series name: CPIAUCSL, retrieved from FRED, Federal Reserve Bank of St. Louis;
https://fred.stlouisfed.org/series/CPIAUCSL
105 Federal Reserve Bank Philadelphia, Consumer Price Index seasonally adjusted, rate of growth over the period
from the last monthly or quarterly historical value to the month that is 12 months beyond the survey date or four
quarters beyond the survey date, Series name: G_BP_To_12M; https://www.philadelphiafed.org/-
/media/frbp/assets/surveys-and-data/livingston-survev/historical-data/meangrowthrate.xlsx. Additional
information available at https://www.philadelphiafed.org/surveys-and-data/real-time-data-research/livingston-
survey
106 Board of Governors of the Federal Reserve System (US), Trend price inflation measured using survey data on
ten-year inflation expectations, Series name: PTR;
https://www.federalreserve.gov/econres/files/data only package.zip. Additional information on the FRB/US
model and the PTR are available from the U.S. Federal Reserve at https://www.federalreserve.gov/econres/us-
models-about.htm
57
-------
EXTERNAL REVIEW DRAFT
Table 2.4.1 presents the average real return on 10-Year Treasury securities for two time periods. The first
is a 30-year period (1991-2020) following the approach taken by OMB (2003) in developing Circular A-4.
The second is 48-years long (1973-2020) and includes all the years originally used by OMB (2003) in
developing Circular A-4 as well as more recent data (2003-2020). The average real returns are lower under
the shorter time period, reflecting the decline in real interest rates over recent decades.107
Table 2.4.1: Average Real Return on 10-Year Treasury Securities
Time Period
Inflation Measure
1991-2020
1973-2020
Consumer Price Index (CPI)
1.55%
2.12%
Livingston Survey
1.62%
2.48%
Perceived Inflation Target Rate (PTR)
1.98%
2.80%
The consideration of more recent versus older data depends on whether the downward trend in real
interest rates is due to structural changes in the economy that are expected to persist. Bauer and
Rudebusch (2021) estimate the current equilibrium real interest using three empirical models for the
interest rate process that allows for an evolution in the equilibrium real interest rate over time. Using a
time series of 10-Year Treasury securities they estimate current equilibrium real interest rates of 1.3, 1.9,
and 2.4 percent.108 When using a longer time series of long-term government securities, Bauer and
Rudebusch (2021) estimate current equilibrium real interest rates of 1.5, 2.3, and 3.0 percent.109
Other government assessments of consumption interest rates suggest a focus on a similar range. The U.S.
Congressional Budget Office's Long-Term Economic Projections forecast real rates on 10-Year Treasury
securities returning to levels of 2.0% and higher over the next couple of decades (CBO 2021a, 2021b). The
most recent Social Security Administrations Trustees report (SSA 2021) uses three estimates of the long-
run real interest rate of 1.8%, 2.3%, and 2.8% based on their assessment of interest rates over the next
couple of decades.
The empirical evidence on central tendencies for the consumption interest rate is also consistent with
recent surveys of economists and technical experts on the appropriate discount rate. Drupp et al. (2018)
surveyed economists who have published at least one paper on discounting in a leading economics journal
107 The average real return on 10-Year Treasury securities has, in general, trended downwards since the 1990s. The
average real return on 10-Year Treasury securities in the period 2001-2020 was 1.1 percent and in the period 2011-
2020 it was 0.2 percent. Based on empirical evidence, Bauer and Rudebusch (2021) utilize the year 1991 as a
breakpoint when considering potential shifts in long-run mean of the interest rate process, which coincide with the
start of the 30-year period considered in Table 2.4.1. The focus on a 30-year period is also consistent with the
approach used by OMB (2003) used in developing guidance on consumption discount rates in Circular A-4. In
addition, under the Ramsey approach used in this report, the certainty-equivalent discount rate for the first 30 years
remains close to the near-term target, suggesting shorter time periods may not be adequately capturing the interest
rate characteristics over the relevant time period.
108 Time series of 10-Year Treasury securities from 1968-2019 with a PTR based inflation adjustment. When using 1-
Year Treasury securities Bauer and Rudebusch (2021) find lower estimates of the equilibrium real interest rate
ranging from 0.7 to 1.3 percent.
109 Bauer and Rudebusch (2021) use a time series of long-term government securities from Newell and Pizer (2003),
updated to include more recent data, that spans 1798-2019 and uses a ten-year moving average of the Livingston
Survey CPI expectation as inflation adjustment after 1954.
58
-------
EXTERNAL REVIEW DRAFT
aboutthe appropriate social discount rate, finding a mean of 2.3% and a median of 2%. Howard and Sylvan
(2020) surveyed experts who have published at least one article related to climate change in a leading
economics or environmental economics journal about the appropriate discount rate for calculating the
SC-GHG, also finding a mean of 2.3% and a median of 2%. Pindyck (2019) also surveyed economists on
discounting and other topics related to the SC-GHG and found a mean discount rate of 2.7% and a median
of 2.0%.
The National Academies (2017) recommended the use of "three sets of Ramsey parameters, generating a
low, central, and high certainty-equivalent near-term discount rate, and three means and ranges of SC-
C02 estimates." Recent studies have found empirical evidence suggestive of a structural break in the
interest rate process sometime during the 1990s that has been associated with declining equilibrium
interest rates in recent decades (e.g., Del Negro et al. 2017, Christensen and Rudebusch 2019, and Bauer
and Rudebusch 2020). Based on empirical evidence, Bauer and Rudebusch (2021) utilize the year 1991 as
a breakpoint when considering potential shifts in long-run mean of the interest rate process. Given the
evidence of structural shifts in the interest process beginning in the 1990s, and the precedent for using
1991 as a reasonable and empirically formed breakpoint, this report places greater focus on the range of
mean interest rate estimates from 1991-2020 presented in Table 2.4.1. To cover that range, this report
includes a half a point spread in certainty-equivalent near-term target rates of 1.5 to 2.0 percent. Given
the potential value in considering a longer time series, this report also considers a third near-term target
rate of 2.5 percent reflective of the average of the Table 2.4.1 estimates using the longer time series110,
which is also consistent with the lines of evidence above suggesting a consumption interest rate of slightly
above 2 percent. Therefore, considering the multiple lines of evidence on the appropriate certainty-
equivalent near-term rate, the modeling results presented in this report consider a range of near-term
target rates of 1.5, 2.0, and 2.5 percent. This range of rates allows for a symmetric one point spread around
2.0 percent.
Calibration of Ramsey parameters. Calibration of the Ramsey parameters follows Rennert et al. (2022a,
2022b) using the specified set of near-term discount rates to generate a certainty-equivalent discount
rate path. Rennert et al. (2022a, 2022b) apply the Newell et al. (2022) calibration approach to the same
set of probabilistic socioeconomic scenarios presented in Section 2.1 and adopted in this report. The
Ramsey parameters, p and n, were calibrated to meet two conditions. First, the average certainty-
equivalent rate over the first 10 years is equal to the near-term target rate. Second, the shape of the
certainty-equivalent discount rate path over the time horizon fits the empirical estimates of Bauer and
Rudebusch (2021).111 The resulting calibrated values of the Ramsey formula parameters are presented in
Table 2.4.2.
110 The average across the estimate in Table 2.4.1 form the window 1973-2020 using different approaches to adjust
for inflation is 2.47 percent, which rounded to one significant digit is 2.5 percent.
111 Additional details of the calibration methodology are available in Newell et al. (2022).
59
-------
EXTERNAL REVIEW DRAFT
Table 2.4.2: Calibrated Ramsey Formula Parameters
Near-Term Target
Certainty-Equivalent Rate
P
n
1.5%
0.01%
1.02
2.0%
0.20%
1.24
2.5%
0.46%
1.42
Source: Rennert et al. (2022b)
Figure 2.4.1 presents the resulting distribution of time-averaged discount rates using the calibrated p and
r| associated with each of the three near-term target rates. The mean and 95th percentile range of the
discount rate used to discount climate damages back to 2020 for the RFF-SPs probabilistic growth
scenarios are presented using dashed and dotted lines. The solid lines illustrate the certainty-equivalent
risk-free discount rate that would lead to the same average discount factor over a specific time horizon
as using the full distribution of dynamic discount rates to calculate a distribution of discount factors. This
path is the same as the calibrated certainty-equivalent risk-free term structures presented in Rennert et
al. (2021a).
Figure 2.4.1: Distribution of the Dynamic Discount Rates
Certainty-Equivalent Path
— — Mean
¦ ¦ ¦ ¦ 95th Percentile Range
o
(N
O
f\l
O
(J
CD
CO
c
3
o
u
in
Q
O
"D
in
=5
s.
6.0%
4.0%
2.5%
2.0% 2.0%
1.5% —
0.0%
-2.0%
e«
2020
2100
2200
2300
Year
The range of the dynamic discount rates used to discount climate damages back to 2020 in any one year for the three near-term
target rates is summarized by the mean (dashed lines) and 5th to 95th percentiles (dotted lines). Also shown here is an illustration
of the corresponding certainty-equivalent risk-free path (solid lines) implied by the calibration procedure described in Section 2.4.2.
During the calibration, Newell et al. (2022) place additional constraints on the rates in each trial such that rates are allowed to go
negative but cannot remain negative for the duration of the time period (2020-2300).
60
-------
EXTERNAL REVIEW DRAFT
While the certainty-equivalent path illustrates the declining certainty-equivalent risk-free discount rate
implied by the Ramsey formula, it is important to emphasize that this does not illustrate the discount rate
used to estimate the SC-GHG values. First, an exogenous, certainty-equivalent declining discount rate is
not used to discount climate damages; each scenario is discounted using the calibrated p and r| values
presented here and the specific consumption growth rate for that scenario. Second, the consumption
growth rate used for discounting is net of baseline climate damages for each model (Kelleher and Wagner
2019).
The calibration approach and resulting Ramsey parameters presented above are consistent with the
National Academies' (2017) recommendation to use a descriptive calibration based on empirical interest
rate data. The resulting parameters presented in Table 2.4.2 are also within the ranges of values of p and
r| used in the peer-reviewed literature, including many studies that state their parameter choices are
based on prescriptive reasoning. For example, the IWG (2010) noted that most papers in the climate
change literature adopt values for r| in the range of 0.5 to 3, although not all authors articulate whether
their choice is based on prescriptive or descriptive reasoning (IWG 2010). The IPCC AR5 report found
values of r| in the literature in the range of 1 to 4 (IPCC 2014b). Values between 1 and 1.45, consistent
with the calibrated range in Table 2.4.2, have been commonly used in recent peer-reviewed studies
(Lemoine 2021, Hansel et al. 2020, Glanemann et al. 2020, Tol 2019, Dietz and Venmans 2019, Nordhaus
2018, Burke et al. 2018, Adler et al. 2017). The Drupp et al. (2018) survey asked economists about the
most appropriate values for r|, and found a median (mean) value of 1 (1.35), and a mode value (i.e., the
most frequently provided response) of 1.
With respect to the pure rate of time preference, the calibrated values presented in Table 2.4.2 are also
within the ranges of p used in the peer-reviewed literature. The vast majority of papers in the climate
change literature adopt values for p in the range of 0 to 2 percent per year, with most studies in the lower
end of the range (IPCC 2014a). The selection of rates on the lower end of that range tend to emerge from
ethical concerns. Some have argued that to use any value other than p = 0 would unjustly discriminate
against future generations (e.g., Arrow et al. 1995, Stern 2006). When Drupp et al. (2018) surveyed
economists about the most appropriate values for p, the experts' responses had a median (mean) value
of 0.5 (1.1) percent, and a mode value of 0. However, even under the case of intergenerational neutrality,
a small positive pure rate of time preference may be appropriate to account for the probability of
unforeseen cataclysmic events (Stern 2006).112 Furthermore, it has been argued that very small values of
p can lead to an unreasonable rate of optimal savings (Arrow et al. 1995), particularly with r| around 1
(Dasgupta 2008, Weitzman 2007).
Regardless of the theoretical approach used to derive the discount rate(s), there remain inherent
conceptual and practical difficulties of adequately capturing consumption trade-offs over many decades
or even centuries. While this report relies on the descriptive approach for selecting specific discount rates
based on observed preferences for temporal tradeoffs of consumption, the EPA is aware of the normative
dimensions of both the debate over discounting in the intergenerational context and the consequences
of selecting one discount rate over another.
112 Stern (2006) assumes a pure rate of time preference of 0.1%. This reflects a 91% probability of the human race
surviving 100 years.
61
-------
EXTERNAL REVIEW DRAFT
2.5 Risk Aversion
The impacts associated with GHG emissions present substantial new risks and exacerbate existing risks to
human health and welfare (USGCRP 2018b, NIC 2021). This raises the question of how to account for
individuals' preferences over these risks in the valuation of climate damages. Individuals are typically not
indifferent between a situation with a certain outcome and a situation with a risky outcome whose
expected value is the same as the certain outcome. That is, in most decision-making processes individuals
tend to be risk averse. This is evident by the existence of voluntary insurance markets where individuals
demonstrate a positive willingness to pay to reduce risk exposure.
U.S. regulatory benefit-cost analyses to date commonly assume risk neutrality (i.e., zero risk aversion).
This assumption is justified in cases where idiosyncratic risks can be pooled across regulations, are
uncorrelated with baseline economic uncertainty, or are shared across large populations (OECD 2018).
However, the largest climate change risks are collective in nature, affecting large shares of the population,
and, therefore, may not be diversifiable (Heal & Kristrom 2002). The marginal damages are also expected
to be correlated with baseline consumption (inclusive of baseline climate change damages) and may add
to society's overall risk (National Academies 2017, Dietz et al. 2018). Therefore, in the case of climate
change risk reductions, individuals are expected to have a positive willingness to pay for that reduced risk
exposure beyond the value of the mean damages. The peer reviewed climate economics literature has
demonstrated the importance of accounting for risk aversion in estimates of the SC-GHG (e.g., Anthoff et
al. 2009, Cai et al. 2016, Lemoine 2021, van den Bremar and van der Ploeg 2021).
In the EPA's analyses relying on the IWG SC-GHG estimates to date risk aversion was incorporated through
adjustments to the discount rate and through consideration of the fourth estimate reflecting the 95th
percentile for a 3% discount rate. However, in the IWG's 2010 TSD, the IWG acknowledged the limitations
of these approaches to provide a unified framework for valuing risk changes. For the SC-GHG estimates
presented in this report, the value of risk associated with marginal GHG emissions is explicitly
incorporated into the modeling following the economic literature and consistent with the National
Academies' (2017) recommendations.
Assuming a time separable welfare function for a population of size Lt with representative agent utility
u(-) and per capita consumption ct, the SC-GHG is defined as
where At are the marginal damages associated with emissions in a given year. That is, the SC-GHG is the
expected marginal changes in utility normalized by the marginal utility of consumption to convert to a
SC-GHG =
(2.5.1)
u(co)
62
-------
EXTERNAL REVIEW DRAFT
willingness to pay in monetary units. Setting aside uncertainty in future populations for ease of exposition,
a second order Taylor expansion of u around E[ct\ allows the SC-GHG to be decomposed as
SC-GHG « Jq -^^{uXE[ct])E[At\ + 1-u "(E[ct])E[At]Var(ct) + Cov(u\ct),At)}dt. (2.5.2)
Expected Precautionary Insurance
Damages
The first term in the braces on the right-hand side of equation (2) is the change in utility from the expected
marginal damages, which drives the willingness to pay for the expected marginal damages. The second
two terms incorporate the way in which risk impacts the SC-GHG estimates and have been referred to as
the precautionary and insurance channels, respectively (Kimball 1990).113 The precautionary term
captures the result that climate damages are more impactful when consumption is lower, all else equal,
leading the returns to mitigation to increase with uncertainty in future consumption. The insurance term,
also referred to as the risk premium, captures the covariance between marginal utility along the baseline
and marginal damages. This term incorporates the degree to which mitigation provides a hedge against
future economic uncertainty, sometimes referred to as the "climate beta" (e.g., Dietz et al. 2021). In other
words, the precautionary channel represents the willingness to pay to avoid the additional climate change
risk itself and the insurance channel represents the willingness to pay to avoid the broader change in
society's risk based on how climate change damages intersect with economic growth.
The IWG SC-GHG estimates used by EPA to date have focused on explicitly quantifying the first component
in equation (2). Incorporating the precautionary and insurance channels into the estimation requires
probabilistic socioeconomic scenarios, which were not available at the time those estimates were
developed. Instead, the IWG partially incorporated the impact of risk into the estimates through
adjustments to the discount rates. The motivation for using a lower 2.5 percent discount rate to capture
risk in future economic conditions was premised on the precautionary channel. The motivation for using
a higher 5.0 percent discount rate was premised on the insurance channel if there is a positive covariance
between economic conditions and climate change damages.114 The fourth value (the 95th percentile at a
3 percent discount rate) was included to represent the extensive evidence in the scientific and economic
literature of the potential for lower-probability, higher-impact outcomes from climate change, which
would be particularly harmful to society. Absent formal inclusion of risk aversion in the modeling,
considering values above the mean in a right skewed distribution with long tails acknowledges society's
preference for avoiding risk.
Accounting for risk aversion more explicitly in the analysis allows valuation of the precautionary and
insurance channels based on the specific evidence of future economic uncertainty and the correlation
with marginal climate change damages presented in Sections 2.1 and 2.3. That is, the value of risk aversion
is incorporated into the SC-GHG estimates based on the marginal climate change risk reductions identified
by the modeling as opposed to through exogenous adjustments. Explicitly incorporating risk aversion into
113 The second and third components on the right-hand side of equation (2) are sometimes also referred to as the
diversifiable and non-diversifiable components of risk valuation (OECD 2018).
114 If there is a negative covariance between economic growth and climate change damages a downward adjustment
in the discount rate would be warranted.
63
-------
EXTERNAL REVIEW DRAFT
the analysis requires a functional form for the representative agent's utility function. The most commonly
used utility function in the climate economics literature and one consistent with the approach to
discounting identified in Section 2.4, is the isoelastic utility function, u(ct) = cj 11 /(1 — rf), where utility
is a power function with respect to consumption. If the utility function is assumed to follow an isoelastic
function, the definition of the SC-GHG in equation (2.5.1) reduces to the expected value of the marginal
damages discounted using the Ramsey formula,
SC - GHG = E [/0T £?"(P+?Wt)tdtdt], (2.5.3)
where gt is the time averaged per capita consumption continuous growth rate through time t. Therefore,
by discounting via the Ramsey formula as detailed in Section 2.4 and incorporating uncertainty throughout
the modeling process as detailed in Sections 2.1-2.3, the SC-GHG estimates incorporate the climate risk
through the precautionary and insurance channels.
Within the isoelastic utility function, the single parameter, 77, has a role in reflecting both intertemporal
and risk preferences which can present challenges in calibrating the utility function. As noted in Section
2.4, the calibrated values for 77 presented in Table 2.4.2 are consistent with the calibrated range (1 to 1.45)
that has been commonly used in recent peer reviewed studies employing an isoelastic utility function
(Lemoine 2021, Hansel et al. 2020, Glanemann et al. 2020, Tol 2019, Dietz and Venmans 2019, Nordhaus
2018, Burke et al. 2018, Adler et al. 2017). However, while that range of values may be appropriate for 77
in its role representing intertemporal preferences, they may be too conservative for 77 in its role
representing risk preferences. Some have suggested that values of 77 between 2 and 10 would be required
to match empirical and experimental evidence on rates of risk aversion (Crost and Traeger 2014, Jensen
and Traeger 2014, Cai et al. 2016, Cai and Lontzek 2019, Daniel et al. 2019, Okullo 2020, Lemoine 2021,
Jensen and Traeger 2021, Van den Bremer and Van der Ploeg 2021). To address this calibration challenge,
some recent SC-GHG studies have used alternative utility function specifications (e.g., Epstein-Zin
specifications) that allow for the separation of intertemporal and risk preferences (Cai et al. 2016, Daniel
et al. 2019, Cai and Lontzek 2019, Okullo 2020, Lemoine 2021, Van den Bremer and Van der Ploeg 2021).
These studies can incorporate a higher rate of relative risk aversion without affecting the calibration of
the intertemporal preferences. While these approaches have promise for improving the calibration of risk
preferences, they are relatively new in the climate economics literature, computationally complex, and
require additional assumptions (e.g., timing of uncertainty resolution) for which there is no consensus in
the literature. For these reasons, these alternative utility functions are not used in this report, but they
are worthy of additional investigation, consistent with recommendations of the National Academies
(2017). Furthermore, the use of an isoelastic utility function via equation (2.5.3) remains widely used in
the peer reviewed literature and is consistent with the National Academies' (2017) recommendations on
robustly capturing the value uncertainty through probabilistic scenario, climate, and damage function
models in conjunction with a Ramsey-like approach to discounting. However, because the calibrated
values of 77 using the isoelastic utility function may be low from a risk aversion perspective, the value of
reducing climate change risk included in the SC-GHG estimates will likely be an underestimate, holding all
else equal.
64
-------
EXTERNAL REVIEW DRAFT
When using the damage module based on GIVE and Howard and Sterner (2017), the SC-GHG is calculated
using equation (2.5.3) for a global representative agent. Implicit in the use of a global representative agent
is that all risks can be pooled at the global level. This is the model developers' default approach in GIVE,
and the global nature of the Howard and Sterner (2017) damage module precludes other assumptions.
However, when using the DSCIM damage module, a conceptually similar approach is applied but,
following the model developers' default approach, a different assumption on risk pooling is applied.
Specifically, when the DSCIM damage module is used, it is assumed that risks associated with uncertainty
in the climate response and future socioeconomic conditions can be pooled globally, but damage function
risks (conditional on a given level of climate change and RFF-SP socioeconomic realization) are pooled at
the damage function's impact region level. All else equal, assuming that risk can be pooled across broader
geographic areas reduces the value of risk reductions within the SC-GHG estimates.
65
-------
EXTERNAL REVIEW DRAFT
3 Modeling Results
3.1 Social Cost of Carbon (SC-CO2), Methane (SC-CH4), and Nitrous Oxide (SC-N2O)
Estimates by Damage Module
This section presents the SC-GHG values estimated using the methodological updates described in Section
2. The combination of using three specifications of the damage module over the modeling time horizon115
and three near-term target discount rates produces nine separate distributions of SC-GHG estimates for
each emissions year and GHG. Each distribution consists of 10,000 estimates based on draws from the
distributions of uncertain parameters in each module.116 Given the consideration of multiple lines of
evidence in the damage module and multiple near-term discount rates, the results are first presented
separately for each of the three damage modules by discount rate.117 Table 3.1.1, Table 3.1.2, and Table
3.1.3 show the certainty-equivalent SC-C02, SC-CH4, and SC-N20 estimates, respectively, in ten-year
increments for emissions years 2020-2080 by damage module and near-term discount rate.118 As
expected, estimates based on a higher near-term discount rate are consistently lower, while lower near-
term discount rates result in higher SC-GHG estimates independent of the damage module. There is some
variation in the SC-GHG estimates across the three damage modules. This is expected given that the
damage modules are, at least to some extent, measuring different categories of damages and with
different approaches. The SC-GHG estimates based on the meta-analysis damage module tend to be
higher than those based on damage modules from the DSCIM or GIVE models for C02 and N20. For CH4,
which has a notably shorter atmospheric lifetime than the other two gases, the SC-GHG estimates based
on the GIVE damage module tends to have higher estimates. This suggests differences across the models
as to the damage from climate change in the near-term.
The near-term SC-CO2 estimates reported in Table 3.1.1 are comparable in magnitude to recent published
SC-CO2 estimates that were developed using non-IAM based approaches. For example, Pindyck's (2019)
recent survey of several hundred experts in climate science and climate economics yielded mean SC-CO2
estimates around or above $200 per metric ton C02 for various subsets of his sample of respondents.119
115 As mentioned in Section 1.2, the National Academies recommended that the modeling time horizon "extend far
enough in the future to provide inputs for estimation of the vast majority of discounted climate damages." In the
case of models presented here, the discounted streams of marginal damages in all models and discount rates peak
by the end of the century (2100) and begin to steadily decline through the end of the modeling time horizon (2300)—
capturing the majority of the quantified discounted damages associated with the emissions of a metric ton of CO2,
CH4, and l\l20.
116 Monte Carlo methods are used to run the combined suite of modules 10,000 times. In each simulation the
uncertain parameters are represented by random draws from their defined probability distributions.
117 Estimates in this report are discounted back to the year of emissions and presented as certainty-equivalent values
that account for uncertainty in the socioeconomic scenarios. See Appendix A.3 for more information on how those
transformations were made and Section 4 for how they can be used in analyses.
118 Values in Table 3.1.1, Table 3.1.2, and Table 3.1.3 are rounded to two significant figures.
119 Pindyck's (2019) full sample of respondents yielded mean SC-CO2 estimates above $200/mtCC>2, after dropping
responses where values fell outside the 5th or 95th percentiles. Responses from economists were lower (on average
$174) while the mean SC-CChfor other groups was close to $300. To further illustrate the heterogeneity in responses,
Pindyck (2019) also reported results based on further trimming of responses, e.g., to 10th through 90th percentile
66
-------
EXTERNAL REVIEW DRAFT
Studies using other types of survey techniques have found similar ranges of SC-C02 estimates. For
example, based on the results of a vehicle choice experiment, Hulshof and Mulder (2020) derived a mean
willingness-to-pay estimate for C02 emission reduction of $236 per metric ton C02.120 An earlier vehicle
choice survey by Achtnicht (2012), using a different population and a somewhat different method to
translate the WTP for clean cars into WTP for emission reductions, found car buyers to be willing to pay
between $130 and $372 per metric ton of C02 reduced.121
For all damage modules, the SC-GHG estimates increase over time - i.e., the societal harm in 2030 from
one metric ton emitted in 2030 is greater than the harm in 2020 caused by one metric ton emitted in
2020. Emissions further in the future produce larger incremental damages as physical and economic
systems become more stressed in response to greater climatic change and because income is growing
over time. As income grows so does the willingness to pay to avoid economic damages. The growth rate
of the SC-GHG is generally larger in the case of the DSCIM climate module than the other damage modules.
In the case of longer-lived C02 and N20 emissions, this can lead the SC-GHG estimates based on the DSCIM
damage module to eventually exceed those based on one or both of the other damage modules. This is
reflective of the marginal damages in the DSCIM damage module being more sensitive to baseline climate
change than in the other damage modules (see Figure 2.3.2).
Table 3.1.1: Social Cost of Carbon (SC-C02) by Damage Module, 2020-2080 (in 2020 dollars per metric ton
of CO2)
Near-Term Ramsey Discount Rate and Damage Module
2.5%
2.0%
1.5%
Emission
Year
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
2020
110
120
120
190
190
200
330
310
370
2030
140
150
150
230
220
240
390
350
420
2040
170
170
170
280
250
270
440
390
460
2050
210
200
200
330
290
310
500
430
520
2060
250
220
230
370
310
350
550
470
570
2070
280
240
250
410
340
380
600
490
610
2080
320
260
280
450
360
410
640
510
650
values (which reduces mean SC-CO2 estimates to $147-243/mtCC>2), or to the experts who reported high confidence
in their impact probabilities (which reduced mean SC-CO2 estimates to $108-138/mtCC>2).
120 We convert the results reported in Hulshof and Mulder (2020) to U.S. dollars using December 2017 exchange
rates (1.1836 USD/Euro (httpsi//www,federalreserve,gov/datadownload/Choose,aspx?rel=H10)), the month the
survey was administered.
121 We convert the results reported in Achtnicht (2012) to U.S. dollars using the average exchange rate during the
time period when the survey was administered, August 2007 through March 2008 (1.4502 USD/Euro
(httpsi//www,federal reserve,gov/datadownload/Choose,aspx?rel=H 10)).
67
-------
EXTERNAL REVIEW DRAFT
Table 3.1.2: Social Cost of Methane (SC-CH4) by Damage Module, 2020-2080 (in 2020 dollars per metric
ton of CH4)
Near-Term Ramsey Discount Rate and Damage Module
2.5%
2.0%
1.5%
Emission
Year
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
2020
470
1,600
1,700
850
1,900
2,200
1,500
2,500
2,900
2030
1,100
2,300
2,300
1,600
2,800
2,800
2,400
3,500
3,700
2040
1,900
3,300
2,900
2,500
3,800
3,500
3,300
4,700
4,500
2050
2,700
4,200
3,700
3,400
4,900
4,400
4,300
5,900
5,600
2060
3,500
5,000
4,400
4,200
5,800
5,300
5,200
7,000
6,700
2070
4,200
5,700
5,100
5,100
6,600
6,200
6,100
7,900
7,800
2080
5,100
6,300
5,900
6,000
7,300
7,100
7,100
8,800
8,900
Table 3.1.3: Social Cost of Nitrous Oxide (SC-N20) by Damage Module, 2020-2080 (in 2020 dollars per
metric ton ofN20)
Near-Term Ramsey Discount Rate and Damage Module
2.5%
2.0%
1.5%
Emission
Year
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
DSCIM
GIVE
Meta-
Analysis
2020
30,000
38,000
38,000
49,000
55,000
58,000
81,000
85,000
96,000
2030
40,000
47,000
46,000
63,000
67,000
69,000
98,000
100,000
110,000
2040
52,000
57,000
55,000
77,000
78,000
81,000
120,000
110,000
130,000
2050
64,000
67,000
66,000
93,000
91,000
95,000
140,000
130,000
150,000
2060
77,000
75,000
76,000
110,000
100,000
110,000
150,000
140,000
160,000
2070
89,000
82,000
84,000
120,000
110,000
120,000
170,000
150,000
180,000
2080
100,000
89,000
94,000
140,000
120,000
130,000
190,000
160,000
200,000
For a given near-term target discount rate, the certainty-equivalent SC-GHG estimate is the value applied
to GHG emission changes in benefit-cost analysis (see Section 2.5 for a definition of the SC-GHG). These
certainty-equivalents are calculated over a distribution of SC-GHG estimates reflecting the full range of
quantified uncertainties incorporated into the modeling (see Section 2 for a description of the quantified
uncertainty in each module). Figure 3.1.1 shows the full distribution of SC-GHG estimates for emissions in
2030, where the boxes span the inner quartile range (25th to 75th quantile), whiskers extend to the 5th
(left) and the 95th (right) quantiles. The vertical lines inside of the boxes mark the median of each
distribution, and the points inside of the boxes and dollar estimates on top of the boxes mark the simple
mean (average). In these distributions, the uncertainty that is explicitly characterized includes the
socioeconomics and emissions projections from the RFF-SPs and the GHG concentrations and
temperature changes generated from the FaIR model. Explicit characterization in these distributions of
uncertain parameters in the modeling of SLR and the parametric uncertainty captured in the estimation
of each damage function varies across the three damage modules.
68
-------
EXTERNAL REVIEW DRAFT
It is important to note that the distributions presented here do not fully characterize uncertainty about
the SC-GHG due to impact categories omitted from the models and sources of uncertainty that have not
been fully characterized due to data limitations. These limitations are discussed in Section 3.2 below.
Uncertainty grows over the modeled time horizon. Therefore, under cases with a lower near-term target
discount rate - that give relatively more weight to impacts in the future - the distribution of the SC-GHG
is wider (see Figure 3.1.1). Across damage modules, the DSCIM based runs generate the widest
distribution of results. The DSCIM damage module has a greater degree of curvature in the damage
function mapping temperature to economic damages than the GIVE and H&S specifications (see Figure
2.3.2). The interquartile ranges overlap across the three damage modules.
Figure 3.1.1: Distribution of Social Cost of Carbon Dioxide (SC-C02) Estimates for 2030, by Near-term
Ramsey Discount Rate and Damage Module
$140
$230
DSCIM
$220
2.0%
~
$240
GIVE
Meta-Analysis
$390
$0
$500 $1,000
SC-C02 for 2030 emissions (2020$ per metric ton of C02)
$1,500
Boxes span the inner quartile range (25th to 75th percentiles), whiskers extend to the 5th (left) and the 95th (right) percentiles. The
vertical lines inside of the boxes mark the median of each distribution, and the points inside of the boxes and dollar estimates on
top of the boxes mark the simple mean (average).
Table 3.1.4 provides a disaggregation of the SC-C02 results by sector or impact category for emissions in
2030 under the GIVE and DSCIM based damage modules - alongside the meta-analysis-based damage
module that does not permit a sectoral disaggregation. The GIVE and DSCIM damage modules are
consistent in that net mortality risk increases are the largest share of marginal damages across the sectors
considered in each damage module. However, the share of marginal damages due to net mortality risk
69
-------
EXTERNAL REVIEW DRAFT
increases is larger for the DSCIM damage module compared to the GIVE damage module. Variation across
the two damage modules for the other sectors reflects uncertainty in the underlying scientific literature
and differences in the sectors included in the models (e.g., labor productivity). See Section 2 for detailed
descriptions of the methodological differences across models. The differences in results are the aggregate
effect of these different methodologies.
Table 3.1.4: Sectoral Disaggregation of Social Cost of Carbon (SC-C02) for 2030 under a 2.0% Near-Term
Ramsey Discount Rate (in 2020 dollars per metric ton of C02)
Damage Module
Damage sector or category
DSCIM
GIVE
Meta-Analysis
Health
$179
$104
-
Energy
-$4
$10
-
Labor productivity
$47
-
-
Agriculture
$4
$103
-
Coastal
$3
$2
-
Total
$233
$219
$238
3.2 Omitted Damages and Other Modeling Limitations
The research community's considerable progress in developing new data and methods have helped to
bring the SC-GHG estimates presented in Section 3.1 closer to the frontier of climate science and
economics and address many of the National Academies' (2017) near-term recommendations. However,
the SC-GHG estimates presented in this report still have several limitations, as would be expected for any
modeling exercise that covers such a broad scope of scientific and economic issues across a complex global
landscape. There are still many important categories of climate impacts and associated damages that are
not yet reflected in these estimates due to data and modeling limitations. There is also incomplete
coverage of some categories that are represented, including important sectoral and regional interactions.
Table 3.2.1 below highlights some of these limitations. For important categories within climate science,
impacts and associated damages, and methodology, the table denotes those that the SC-GHG estimates
in this report have been able to incorporate, those only partially incorporated, and those that are not yet
included. For example, the damage modules currently focus on climate change damages driven by
changes in annual average temperatures or sea level rise. The damage modules have not yet explicitly
incorporated damages associated with other changes in the temperature distribution such as variability
and changes in the probability of extreme temperatures throughout the year. Nor have the damage
modules explicitly considered damages associated with changes in precipitation or humidity due to
climate change.
The climate module considered in this report omits some potentially large-scale Earth system changes
(e.g., from tipping elements) or non-climate mediated effects of GHG emissions (e.g., ocean acidification,
tropospheric ozone formation due to CH4 emissions). Climate change impacts described as resulting from
tipping elements are often associated with crossing a threshold in an Earth system, or 'tipping point', after
which a relatively small perturbation in radiative forcing results in a large, often irreversible change in the
climate or other Earth systems (see, e.g., Kopits et al. (2014) for a review of this literature). A few of these
70
-------
EXTERNAL REVIEW DRAFT
processes (e.g., Arctic Sea ice loss and surface albedo feedback, slowdown of the Atlantic Meridional
Overturning Circulation (AMOC)) are captured in the underlying CMIP6 models in which FaIR vl.6.2 was
calibrated to and are thus implicitly reflected in the climate module used in this report (Weijer et al. 2020).
For other processes - such as Amazon Forest dieback, melting of permafrost, changes in the Indian
summer monsoon (ISM) - it is less certain how well their behavior is captured in CMIP6 models or whether
they are implicitly included in FalRl.6.2 (see, e.g., Arora et al. 2020, IPCC 2021d). Lastly, methane hydrates,
Greenland (GIS) and Antarctic icesheet (AIS) collapse are not included at all within FaIR. However, GIS and
AIS are simulated within the sea-level models used in this report.
Recent studies have started to make progress on incorporating more of the tipping elements discussed
above in the estimation of SC-GHG. In particular, Dietz et al. (2021) developed a response function that
maps increases in global mean surface temperature (GMST) to additional warming that is realized through
feedbacks in the underlying biophysical systems such as permafrost thaw, ocean methane hydrates,
Amazon rainforest dieback, GIS and AIS collapse, the AMOC slowdown, and ISM variability. This allows for
an improved, more explicit accounting of the temperature-driven damages resulting from these types of
large-scale feedback effects within SC-GHG estimation. The EPA will continue to follow progress in this
line of research and look for opportunities to better reflect tipping elements and other Earth system
changes and to account for non-climate mediated GHG effects in future updates of the SC-GHG estimates.
Additional discussion of these is provided in Section 3.2.1 below.
The bottom-up damage modules from the DSCIM and GIVE models provide a transparent accounting of
which climate change damages are incorporated into the modules, as discussed in Section 2.3.122 While
the advancements in these newer damage modules is laudable, it is clear that many categories of climate
change damages are not yet represented. Examples include changes in the demand for water resources,
the costs and feasibility of providing safe drinking water, changes in ecosystem services, and the
productivity of the livestock, aquaculture, and forestry industries just to name few.
For those damage categories that are represented, they may only be a partial accounting. For example,
the estimated health damages in GIVE and DSCIM only include temperature- and SLR-related mortality,
and exclude other sources of mortality impacts (e.g., climate mediated changes in storms, wildfire,
flooding, air pollution), and morbidity impacts (e.g., infectious diseases, malnutrition, allergies). Studies
are available on how climate-relate changes impact infectious diseases (Levy et al. 2016, Trinanes et al.
2021, Colon-Gonzalez et al. 2021, Ryan et al. 2019, Ryan et al. 2015, Mordecai et al. 2020) but additional
work in needed to both model metrological conditions (e.g., humidity, precipitation patterns, length of
transmission seasons, and daily temperature ranges) under climate change and link these to infectious
disease damage functions (Cromar et al. 2022). Importantly, none of the damage modules incorporate
cross-country or regional spillovers that occur through migration, national security concerns, tourism, or
supply chain disruptions. The physical and economic pathways that drive many of these omitted or
partially included categories are well documented in key scientific assessments, such as those developed
122 For the GIVE model, Rennert et al. (2022b) illustrate the impact that the updated damage functions have on the
SC-CO2 estimates relative to damages functions used in earlier studies. The authors find the SC-CO2 estimate is
notably larger when using GIVE's updated four-sector damage function ($185/mtC02 in 2020 under 2% Ramsey
discounting compared to using the aggregate top-down damage function approach used in the latest version of the
DICE model (DICE 2016) ($152/mtC02 in 2020), which was stated to be more comprehensive in scope and included
a 25% adder for omitted impacts (holding all else equal in the modeling).
71
-------
EXTERNAL REVIEW DRAFT
by the IPCC (e.g., IPCC 2008, 2014a, 2018, 2019a, 2019b, 2021a) and the U.S. Global Change Research
Program (e.g., USGCRP 2016, 2018a). However, key data and research gaps currently prevent
incorporating these damage categories into global damage modules for the purpose of estimating the SC-
GHG.
While the SC-GHG estimates presented in this report provide numerous methodological improvements
over the previous estimates, as detailed in Section 2, there are opportunities for future improvements.
For example, none of the damage modules explicitly consider potential interactions among damage
categories. For example, the modules do not account for how climate change-mediated impacts to water
supply will interact with climate-mediated changes in the demand for water resources by the agricultural
and electric power sectors that may be in competition in similar water markets.
Equally important to note among the methodological limitations is the valuation of risk aversion in the
updated SC-GHG estimates. As noted in Section 2.5, the SC-GHG estimates provide an improved
accounting of risk aversion over the estimates used in the EPA's analyses to date. However, the approach
relies on an isoelastic utility function in which a single parameter has a role in reflecting both
intertemporal and risk preferences. In this report, the utility function parameter is calibrated based on its
role representing intertemporal preferences leading to lower values than would be expected if it was
calibrated based on its role representing risk preferences. As a result, the SC-GHG estimates likely
underestimate the damages associated with increased climate risk resulting from a marginal ton of
emissions, all else equal. As noted in Section 2.5, to address this calibration challenge, some recent SC-
GHG studies have used alternative utility function specifications (e.g., Epstein-Zin specifications) that
allow for the separation of intertemporal and risk preferences (Cai et al. 2016, Daniel et al. 2019, Cai and
Lontzek 2019, Okullo 2020, Lemoine 2021, Van den Bremer and Van der Ploeg 2021).
Although not all omitted climate change impacts work in the same direction in terms of their influence on
the SC-GHG estimates, taken together, the numerous omitted damage categories, modeling assumptions
that go in the direction of being conservative, and other limitations discussed above and throughout
Section 2, make it likely that the SC-GHG estimates presented in this report underestimate the damages
from GHG emissions. For example, first, as discussed above, many categories of damages are only partially
modeled or omitted altogether in the DSCIM- and GIVE-based damage modules. Second, many
interactions and feedback effects are not yet represented, both in modeling physical earth system changes
(e.g., feedback effects of tipping elements) and economic damages. For the GIVE model-based results,
Rennert et al. (2022b) "expect that, in total, the future inclusion of additional damage sectors and tipping
elements is likely to raise the estimates of the SC-C02, and that therefore the estimates from the present
study are likely best viewed as conservative." Third, as noted in Section 2.3, data limitations have been
pointed out as a likely cause of the estimated response function in DSCIM to be generating conservative
predictions of mortality risk increases in some low income regions. Fourth, under the meta-analysis-based
damage module, the results are based on a Howard and Sterner (2017) specification to which those
authors and other researchers (e.g., Nordhaus and Sztorc 2013, Nordhaus 2017b) have routinely added a
generic 25% increase in recognition of omitted damages that are likely significant. Fifth, coastal damages
in both GIVE and DSCIM are estimated based on an optimistic assumption that optimal, lowest cost
adaptation opportunities will be realized globally under perfect foresight about SLR. Finally, the method
employed to account for risk aversion likely underestimates the damages associated with increased
climate risk resulting from a marginal ton of emissions.
72
-------
EXTERNAL REVIEW DRAFT
Table 3.2.1: Scope of Climate Science, Impacts, and Damages Included in the Updated SC-GHG Estimates123
Climate Science
Impacts and Associated Damages
Temperature change
(
Human Health and Well-being
1
Averages
•
Heat and cold related mortality
•
Extremes
o
Mortality and morbidity from extreme weather events (e.g.,
storms, wildfire, flooding), and sea level rise
4
Variability
o
Mortality and morbidity from climate mediated changes in the
formation of criteria air pollutants (e.g., ozone, PM2.5)
O
Sea level rise
1
Infectious diseases
o
From average temperature change
•
Other morbidity (e.g., malnutrition, allergies)
o
Non-linear effects (e.g., ice-sheet collapse)
1
Displacement and migration
o
Precipitation
o
Labor
1
Averages
o
Labor supply (i.e., hours worked)
•
Extremes
o
Labor productivity (i.e., output per hour worked)
o
Variability
o
Energy
1
Humidity-wet-bulb temperature
o
Energy consumption (e.g., heating, cooling)
•
Large scale Earth system changes (tipping elements,
etc.)
1
Energy production and provision (e.g., hydroelectric, thermal
power generation)
1
Additional changes in temperature
1
Water
o
Sea level rise
1
Water consumption (residential, industrial, commercial)
o
Precipitation
o
Provision of safe drinking water
o
Extreme weather events
o
Water storage and distribution
o
Ecosystems
o
Land
1
Other impacts
o
Coastal land loss from sea level rise
1
Non-climate mediated effects (e.g.)
1
Buildings, transportation, and infrastructure
1
Carbon fertilization (C02)
•
Sea level rise
4
Ocean acidification (C02)
o
Intensity or frequency of coastal storms
o
Tropospheric ozone formation (CH4)
o
Extreme weather inland (e.g., storms, wildfire, flooding)
o
Stratospheric ozone destruction (N20)
o
Environmental conditions (e.g., melting permafrost, air
temperature and moisture)
o
Food production
1
Methodology
1
Agriculture/Crop production
1
Explicit treatment of uncertainty
•
Animal and livestock health and productivity
o
Accounting for adaptation and costs of adaptation
1
Fisheries and aquaculture production
o
Interactions/feedbacks across sectors
o
Forestry
o
Feedbacks from damages to socioeconomics and
emissions
o
Timber, pulp, and paper production
o
Valuation of risk
1
Tourism, recreation, aesthetics
o
Visitation, locations, and opportunities (e.g., recreational fishing,
skiing, scuba diving, scenic views)
o
Ecosystem services
o
Availability and quality of natural capital used in the production of
marketable goods
o
Biodiversity and wildlife habitat (e.g., aquatic environments,
breeding grounds)
o
Other provisioning and regulating services (e.g., water filtration,
wildfire and flood mitigation, medicinal resources, pest control,
pollination)
o
Cultural services
o
Legend
Crime (property, violent)
o
9 Incorporated
National Security
o
4 Partially Incorporated
Military base impacts
o
O Not Yet Incorporated
Military mission impacts from international civil conflict
o
International development, humanitarian assistance
o
Trade and logistics
o
123 Table 3.2.1 presents a general indication of the climate science, impacts, and damages included across the three
damage modules used in this analysis and may not be reflective of any one specific damage module.
73
-------
EXTERNAL REVIEW DRAFT
Supply chain disruption (e.g., from extreme weather) Q
Supply chain transitions (e.g., altering trade routes) O
One way to illuminate the potential magnitude of some omitted damage categories is to consider the
current spatial distribution of global population and climate indicators. Figure 3.2.1 shows that a
substantial portion of the world's population lives in latitudes that are projected to experience some of
the highest temperatures. And although not explicitly captured in the figure, within each country most
Figure 3.2.1: Population, Temperature, and Sea Level Rise in 2100
Population
Expected Population in 2100 (Millions of People)
400 600 800 1000 1200
Average Temperatures (°C)
1400
these populations are located near the
coasts in areas expected to experience
significant sea level rise. The spatial
correlations that exist between
population centers and known damage
pathways highlight how temperature-
and SLR-related damages will impact a
significant share of the world's
population. This further underlines the
significance of impacts not currently
reflected in the estimates, such as
geopolitical and regional tensions,
conflict, scarcity, displacement, and
migration, all of which are issues that
affect an interconnected global
economy.
3.2.1 Further Discussion of Ocean
Acidification and Other Non-
Climate Mediated Impacts of GHG
Emissions
SC-GHG estimation to date has
primarily focused on the climate-
mediated effects: e.g., the pathway
from emissions, to concentration, to
radiative forcing, to temperature, to
climate change, and to economic
damages. However, there are other
impacts of GHG emissions. The only
non-climate-mediated effect included
in SC-GHG estimates to date and those
in this report is the crop fertilization
effects resulting from elevated C02
concentrations.
Average Annual Temperature in 2100 (°C)
10 15 20
Sea Level Rise
Average Change in Sea Level in 2100 (Meters, Relative to 2000)
74
-------
EXTERNAL REVIEW DRAFT
However, there are several other potentially important non-climate mediated GHG effects. These include,
for example, the ecosystem effects of ocean acidification and aragonite undersaturation resulting from
elevated concentrations of C02, the health and agricultural impacts of tropospheric ozone generated
through chemical conversion of methane in the atmosphere, and the health effects of stratospheric ozone
destruction resulting from elevated concentrations of N20. Several studies have investigated these effects
and are discussed here.
Ocean acidification from carbon dioxide (CO2) concentrations. In addition to its effects on temperature
and other climate endpoints, C02 emissions contribute to ocean acidification, which will likely result in
substantial changes to marine ecosystems. The ocean absorbs about 30 percent of the C02 released into
the atmosphere. Higher atmospheric levels of C02 cause the ocean to absorb more, which affects the
carbonate chemistry of seawater. Water and carbon dioxide combine to form carbonic acid, contributing
to ocean acidification (i.e., the pH decreases and the ocean becomes more acidic). As noted in Section 2.2,
the FaIR reduced complexity climate model calculates carbon dioxide uptake in the world's ocean as part
of its carbon cycle calculation and provides projections of pH and ocean heat uptake. Specifically, the
model estimates the changes in pH with a simple function to approximate globally averaged surface ocean
pH from atmospheric C02 concentrations (National Academies 2017) and accounts for uncertainty in the
atmospheric C02 concentrations. Figure 3.2.2 depicts the range of ocean pH and ocean heat that is
predicted by the coupling of the RFF-SPs with FalRl.6.2. Under these projections, mean ocean pH is
expected to decrease by 0.11 pH units by 2100 relative to 2020.
Figure 3.2.2: Global Ocean pH and Ocean Heat, 2020-2300
A B 100
0
2020 " 2100 " 2200 " 2300 2020 " 2100 " 2200 " 2300
Year Year
Uncertainty is represented by the emissions uncertainty from the RFF-SP projections and physical climate uncertainty from
FalRl.6.2. Mean (solid) and median (dashed) lines along with 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges.
One of the impacts of ocean acidification is a reduction in the concentration of carbonate ions available
to calcifying marine organisms to build and maintain skeletons, shells, and other carbonate structures.
Among the affected organisms are mollusks, bivalves, reef building corals, and microorganisms at the base
75
-------
EXTERNAL REVIEW DRAFT
of the marine food web. Commercially valuable shellfish including oysters, clams, and abalone exhibit
reduced growth and survival rates under conditions expected by mid-century (Ries et al. 2009). The
synergistic effects of marine heatwaves and acidification on coral reefs will inhibit corals' ability to recover
from increasingly frequent bleaching events (Klein et al. 2022). The scale of follow-on effects of ocean
acidification on marine ecosystems (including fisheries) resulting from a reduced availability of habitat
and prey is much more uncertain and difficult to quantify.
Studies estimating the economic impacts of ocean acidification necessarily focus on those for which the
biophysical outcomes are best understood. Several studies forecast producer and consumer welfare
losses in commercial shellfish markets in the US (Cooley and Doney 2009, Cooley et al. 2015, Moore 2015),
in Europe (Fernandes et al. 2017, Narita and Rehdanz 2017), and globally (Narita et al. 2012). Some of the
largest forecasted welfare impacts of ocean acidification arise from the recreational and existence value
of coral reefs (Brander et al. 2012, Lane et al. 2013) while other studies include the impacts of lost coral
reef habitat on finfish (Colt and Knapp 2016, Kite-Powell 2009, Speers et al. 2016).The impacts of ocean
acidification are not included in the damage modules used in this report because work remains to upscale
existing regional studies to capture global economic impacts. Among the challenges is accounting for
synergistic effects between temperature and seawater chemistry and how the ecological impacts differ
across economically important species. With the current understanding of pH and temperature effects on
growth and survival of shellfish and corals, and existing market and nonmarket valuation data for the
ecosystem services they provide, we expect that it will be feasible to develop damage functions for some
ocean acidification impacts in future SC-GHG updates.
Tropospheric ozone formation from methane (CH4) emissions. In addition to its climate effects, methane
oxidation in the atmosphere leads to the production of tropospheric ozone, which has harmful effects for
human health and plant growth (USGCRP 2018c). Due to methane's atmospheric perturbation lifetime of
about 12 years (IPCC 2021e), methane is well-mixed globally and therefore the effects on ozone are also
global (in contrast to regional ozone effects from NOx and VOC emissions). Studies have estimated that
half of the increase in global annual mean ozone concentrations since preindustrial times is due to
anthropogenic methane emissions (IPCC 2013).
One study estimated the monetized increase in human mortality risk from the ozone produced due to
methane emissions to be $800 to $1800 per ton of methane emissions (Sarofim et al. 2017), using a
methodology similar to that of the IWG SC-GHG estimates at the time the paper was written. A more
recent study estimated that sustained reductions of a million tons of methane emissions per year could
prevent about 1,430 premature deaths annually, along with preventing the loss of 145,000 tons of wheat,
soybeans, maize and rice (UNEP 2021). The UNEP results are larger than the Sarofim et al. (2017) estimate
of 239 to 591 premature deaths avoided due to the mitigation of a million tons of methane. UNEP used
an improved methodology to estimate the ozone changes resulting from methane mitigation, but also
used an estimate of the cardiovascular mortality risk due to elevated ozone concentrations that may be
larger than estimates used by the EPA (EPA 2020).
Stratospheric ozone destruction from nitrous oxide (N20) emissions. In addition to its climate effects, N20
has impacts on stratospheric ozone. When N20 is in the stratosphere, high-energy photons break it apart
resulting in the production of nitric oxide (NO). Like the chlorine atoms from CFCs, NO can catalytically
destroy ozone. Because of this reaction, it has become clear that as CFC emissions are eliminated, N20
76
-------
EXTERNAL REVIEW DRAFT
emissions have become the largest anthropogenic contributor to the destruction of stratospheric ozone
(Ravishankara et al. 2009, Portmann et al. 2012, WMO 2018). A recent article (Kanter et al. 2021)
estimated the monetized impacts of the stratospheric ozone loss due to N20 emissions on human health
and crop damages as $2,000 perton N20 (2020 dollars)124, or over 11% of the value of the SC-N20 estimate
for 2020 emissions in the IWG February 2021TSD.
Other effects. As discussed in Section 2, the SC-GHG estimates presented in this report include the
monetized value of carbon dioxide fertilization effects on agriculture. There may be additional benefits of
carbon dioxide fertilization for ecosystems. However, elevated C02 concentrations can also lead to
reductions in the nutrient content (such as protein, iron, and zinc) of some crops, with potential negative
effects on diets (Beach et al. 2019). Elevated C02 concentrations can also change the production and
allergenicity of aeroallergens (Ziska, 2020). These additional impacts have not been monetized.
One approach for accounting for non-climate mediated GHG effects in SC-GHG estimates would be to use
the estimates of the dollar impacts of a ton of emissions of a given gas from existing studies and add those
impacts to the appropriate social cost. Another approach would be to estimate the monetized damages
within the existing SC-GHG modeling framework. For example, as recommended in Kanter et al. (2021),
this might involve estimating the change in stratospheric ozone concentrations over time resulting from
an additional ton of N20 emissions, and then calculating the increase in the risk of health effects resulting
from the increased ozone concentration (e.g., skin cancer morbidity and mortality). The health effects can
then be valued within the framework in the same way that mortality resulting from extreme heat events
or other climate effects is valued.
3.3 Distribution of Modeled Climate Impacts
As discussed in detail in Section 1, benefit-cost analysis of Federal regulations and other actions include
the global net damages from expected changes in GHG emissions. The distinctive global nature of GHG
emissions combined with an increasingly interconnected world means that climate change impacts
occurring on one side of the world can directly and indirectly affect the welfare of citizens and residents
of a country located on the other side of the world through a multitude of pathways. As the prominent
2014 CNA study concluded, the increasing political complexity and economic integration across the world
makes it "no longer adequate to think of the projected climate impacts to any one region of the world in
isolation. Climate change impacts transcend international borders and geographic areas of responsibility"
(CNA 2014).
However, there is heterogeneity in the distribution of climate change damages across the globe and within
the U.S. The SC-GHG by design, and consistent with the economic theory and methods for benefit-cost
analysis, is an aggregation across individuals of their willingness to pay to avoid the marginal damages of
climate change. As such the SC-GHG is not designed to assess the important distributional considerations
124 Kanter et al. (2021) estimate a median value of US$2.66 per kg N2O-N (in 2008 dollars) for the ozone impacts of
N2O emissions. We convert this estimate to $/ton N2O using the N2O-N to N2O factor of 1.57 and adjust for inflation
to 2020 dollars using the annual GDP Implicit Price Deflator values in the U.S. Bureau of Economic Analysis (BEA)
NIPA Table 1.1.9 (specifically, using 2020USD = 2008USD x (113.648 / 94.419, accessed February 7th, 2022). See
https://apps.bea,gov/iTable/iTable,cfm?reqid=19&step=3&isuri=l&select all years=0&nipa table list=13&series
=a&first year=2005&last year=2020&scale=-99&categories=survey&thetable=.
77
-------
EXTERNAL REVIEW DRAFT
of climate change damages.125 Therefore, it is important for the results of analyses using the SC-GHG to
be placed in context with respect to how the impacts of climate change are expected to be distributed
across populations. This section presents the available evidence on the distribution of climate change
impacts based on the results from the SC-GHG modeling above.
The spatial distribution of climate impacts is the result of complex physical and economic dynamics
interacting with the existing heterogeneity in physical and socioeconomic conditions. As discussed at
length in Section 2.3 and emphasized in Section 3.2, the damage modules used in this report do not
capture all of the pathways through which climate change impacts public health and welfare and hence
only cover a subset of potential climate change impacts. Furthermore, the damage modules do not
capture spillover or indirect effects whereby climate impacts in one country or region can impact the
welfare of residents in other countries or regions, as detailed in Section 1.3. Only two modules, the DSCIM
and GIVE damage modules, have spatial resolution that allows for any geographic disaggregation of future
climate impacts across the world. Hence, the results from the SC-GHG modeling in this report are only
able to provide partial evidence of the global distribution of climate change impacts. Conditional on these
critical caveats, the spatial resolution in both models does allow for the calculation of a partial SC-GHG
measure of damages resulting from climate impacts physically occurring within a particular country. For
example, the DSCIM damage module, which includes net impacts on temperature-related mortality,
agriculture, energy expenditures, labor productivity, and sea level rise, estimates damages from climate
change impacts physically occurring within the U.S. of $ll/mtC02 for a 2020 emissions year, rising to
$27/mtC02 for a 2080 emissions year (under a near-term target discount rate of 2%).126 The GIVE damage
module, which includes net impacts on temperature related mortality, agriculture, energy expenditures,
and sea level rise, estimates damages from climate change impacts physically occurring within the U.S. of
$14/mtC02 for 2020 C02 emissions, rising to $24/mtC02 for 2080 C02 emissions (under a near-term target
discount rate of 2%).127 These estimates are not equivalent to an estimate of the benefits of GHG
mitigation accruing to U.S. citizens and residents even for the 4-5 damage categories included in GIVE and
DSCIM. First, due to technical modeling limitations these estimates do not include damages from physical
impacts occurring in all U.S. territories. For example, damages occurring in Guam, a U.S. territory which is
already being affected by climate change, are not captured in these estimates. As highlighted in a recent
DoD report, "[a]t Naval Base Guam, recurrent flooding limits capacity for a number of operations and
activities including Navy Expeditionary Forces Command Pacific, submarine squadrons,
telecommunications, and a number of other specific tasks supporting mission execution" (DoD 2019).
Second, for the reasons discussed in Section 1, these estimates exclude the myriad of pathways through
which global climate impacts directly and indirectly impact the interests of U.S. citizens and residents. For
example, climate change is likely to worsen public health, change migration patterns, and disrupt aspects
of the global supply chain. Changing economic and health conditions across countries will impact U.S.
125 Some analysts (e.g., Azar and Sterner 1996, Anthoff et al. 2009, Anthoff and Emmerling 2019) employ "equity
weighting" to incorporate distributional equity objectives into estimates of the SC-GHG. As noted by Anthoff and
Emmerling (2019), "[e]xisting equity weighting studies assume a social welfare function (SWF) that exhibits
inequality aversion over per capita consumption levels."
126 The analogous DSCIM results for 2020 emissions of CFU and N20 (under a near-term Ramsey discount rate of 2%)
are $22/mtCH4 and $2,900/mtN20, rising to $382/mtCH4 and $8,500/mtN20 by 2080.
127 The analogous GIVE results for 2020 emissions of CFU and N2O (under a near-term Ramsey discount rate of 2%)
are $223/mtCH4 and $4,400/mtN20, rising to $534/mtCH4 and $7,900/mtN20 by 2080.
78
-------
EXTERNAL REVIEW DRAFT
business, investments, and travel abroad. In addition to the economic consequences, unrest and political
instability in foreign countries are expected to have national security ramifications for the U.S. (DoD 2021).
Empirical estimates of some international spillover impacts have started to appear in the academic
literature. For example, as noted in IPCC (2022), "Schenker (2013) estimated that the climate impacts on
trade from developing to developed countries could be responsible for 16.4% of the total expected cost
of climate change in the US in 2100."For these reasons, and those discussed in Section 1, such estimates
of damages from climate change impacts physically occurring within the U.S. do not provide a robust
estimate of damages to U.S. populations.
These GIVE and DSCIM estimates of damages physically occurring within the U.S. are subject to the
broader set of limitations discussed in Section 3.2, includingthe omission of important damage categories.
Additional modeling efforts can shed further light on some of these categories. For example, the
Framework for Evaluating Damages and Impacts (FrEDI) is a modeling framework developed by the EPA
to facilitate the characterization of net climate change impacts in numerous sectors within the contiguous
U.S. and monetize the associated net damages (EPA 2021d, Sarofim et al. 2021b). FrEDI includes 20
sectoral impact categories, many with multiple adaptation scenarios and sub-impacts, across seven U.S.
regions.128 FrEDI was originally developed to calculate impacts through the end of the 21st Century.
Developments are underway to extend the estimates from within FrEDI out to 2300. Results from the
most recent version of FrEDI that show that damages resulting from climate change impacts within U.S.
borders and in sectors not represented in GIVE and DSCIM are expected to be substantial. For example,
under the RFF-SPs and FaIR model outputs used within this report, FrEDI estimates total net damages
(undiscounted) across 20 sectors in 2060 to be over $300 billion annually, growing to over $600 billion per
year by 2090 (2020$).129 Some of the sectors not appearing in DSCIM or GIVE but having large economic
damages estimated in FrEDI for 2090 include: transportation related damages from hightide flooding
128 The FrEDI model uses estimates of physical and economic impacts of climate change by degree of warming
developed using existing sectoral impacts models to project impacts and damages resulting from any emission
scenario. It is designed to synthesize the results of a broad range of peer-reviewed climate change impact and
damage projections, including those derived from econometric approaches and detailed, processed-based
simulation models. These include various impacts to human health, coastal and inland property (e.g., from SLR,
flooding and storms), transportation and other infrastructure, energy demand and supply, water resources, labor,
and winter recreation. Currently, all impacts in FrEDI are based on changes in temperature or SLR, although the
relationship between climate and impacts in the underlying models often includes other factors, such as
precipitation; the framework employ a variety of assumptions regarding adaptive responses to climate impacts. EPA
(2021d) provides a complete list of endpoints and details regarding the scope and assumptions for each sector. For
additional description of FrEDI please see www.epa.gov/cira/fredi and www.githyb.com/USEPA/FrEDl.
129 Inputs to FrEDI include a time series of global mean temperature from the baseline scenario calculated the mean
over an ensemble of 10,000 FaIR vl.6.2, U.S. population in each of the 7 National Climate Assessment regions (i.e.,
Northeast, Southeast, Midwest, Northern Great Plains, Southern Great Plains, Southwest, Northwest) and U.S. GDP
in 2015$ from the RFF-SPs. Regional population was calculated as a percentage of total national population from
FrEDI. FrEDI provides damage estimates in 2015USD. These were brought to 2020USD for this report using U.S.
Bureau of Economic Analysis (BEA) Table 1.1.9 (specifically, using 2020USD = 2015USD x (113.648 / 104.691),
accessed February 7th, 2022). See
https://apps.bea.gov/iTable/iTable.cfm?reqid=19&step=3&isuri=l&select all vears=0&nipa table list=
13&series=a&first vear=2005&last vear=2020&scale=~99&categories=survev&thetable= .
79
-------
EXTERNAL REVIEW DRAFT
($142 billion annually in 2090), premature mortality from climate-driven changes in ozone and PM2.5 ($90
billion annually in 2090), and property damage from hurricane winds ($28 billion annually in 2090).
Due to the limitations associated with the DSCIM and GIVE damage modules these models significantly
underestimate the benefits of GHG mitigation to U.S. citizens and residents. The EPA will continue to
review developments in the literature, including robust methodologies for estimating the magnitude of
the various direct and indirect damages to U.S. populations from climate impacts occurring abroad and
reciprocal international mitigation activities.
Just as there is heterogeneity in the distribution of climate change damages across the globe, the scope
and magnitude of climate change impacts is not uniform across the U.S. Although subnational detail on
the distribution of impacts and associated monetized damages is not available from the SC-GHG modeling
presented in Section 3.1,130 scientific assessment reports and additional modeling efforts can shed further
light on the distribution of damages expected to occur within the U.S. For example, scientific assessment
reports on climate change produced over the past decade by the U.S. Global Change Research Program
provide detailed findings as to the distribution of climate changes impacts across the U.S. (e.g., USGCRP
2016, 2018a). Modeling efforts using a predecessor of DSCIM (e.g., Hsiang et al. 2017) and using the FrEDI
model provide additional information about how damages are expected to be substantial and distributed
unevenly across U.S. regions. For example, of the sectors examined in FrEDI in 2021, the largest source of
modeled damages differed from region to region, with wildfire impacts the largest for the Northwest, air
quality impacts on the East Coast and the Southwest, temperature-related mortality in the Midwest, wind
damage in the Southern Plains, and damages to rail infrastructure in the Northern Plains. In addition, a
growing body of literature is focusing on the disproportionate and unequal risks that climate change is
projected to have on communities that are least able to anticipate, cope with, and recover from adverse
impacts. National Academies of Science, Engineering, and Medicine reports provide evidence of how the
impacts of climate change create potential environmental justice concerns (NRC 2011, National
Academies 2017). For a recent detailed discussion of climate change impacts in the U.S. and their
intersection with environmental justice concerns, see the 2021 Climate Change and Social Vulnerability
report (EPA 2021e).
4 Using SC-GHG Estimates in Policy Analysis
This section discusses how the SC-GHG results presented in Section 3.1 can be used in the EPA analysis of
policies that affect GHG emissions. Section 4.1 presents a combination of the multiple lines of evidence
on damages into a manageable number of values for policy analysis. Section 4.2 describes how the SC-
GHG values are applied to a stream of estimated emissions changes in an analysis.
130 The GIVE damage module is only resolved at the country level, such that subnational detail on the distribution of
impacts is not available. The DSCIM damage module is resolved at a spatial resolution resembling counties, though
that level of detail is unavailable for the model results based on the probabilistic socioeconomic scenarios used in
this report.
80
-------
EXTERNAL REVIEW DRAFT
4.1 Combining Lines of Evidence on Damages
The SC-GHG estimation process in this report produces nine separate estimates of the SC-C02, SC-CH4,
and SC-N2O for a given year, the product of three damage modules and three discount rates. To produce
a range of estimates that reflects the uncertainty in the estimation exercise while providing a manageable
number of estimates to incorporate into policy analysis, the multiple lines of evidence on damage modules
can be combined by averaging the results presented in Table 3.1.1, Table 3.1.2, and Table 3.1.3 across the
three damage module specifications. In assigning equal weight to each damage module specification no
underlying line of evidence is given greater weight than another. As discussed in Section 2.3, the sectoral
damage modules in GIVE and DSCIM are based on different underlying information, data sources, and
estimation methods.131 GIVE and DSCIM are both independent lines of evidence from the meta-analysis-
based damage module since the studies underlying each sectoral damage modules in GIVE and DSCIM are
not included in Howard and Sterner's (2017) final sample of studies.
Table 4.1.1 presents the resulting SC-GHG estimates for each emissions year, gas, and near-term target
discount rate after averaging across three damage module specifications. This table displays the rounded
values; the annual unrounded values for use in calculations are available for all emissions years over 2020-
2080 in Table A.4.1 in the Appendix.
Table 4.1.1: Estimates of the Social Cost of Greenhouse Gases (SC-GHG), 2020-2080 (in 2020 dollars per
metric ton)
SC-GHG and Near-term Ramsey Discount Rate
SC-CO2 SC-CH4 SC-N2O
(2020 dollars per metric ton of C02) (2020 dollars per metric ton of CH4) (2020 dollars per metric ton of N20)
Emission
Year
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2020
120
190
340
1,300
1,600
2,300
35,000
54,000
87,000
2030
140
230
380
1,900
2,400
3,200
45,000
66,000
100,000
2040
170
270
430
2,700
3,300
4,200
55,000
79,000
120,000
2050
200
310
480
3,500
4,200
5,300
66,000
93,000
140,000
2060
230
350
530
4,300
5,100
6,300
76,000
110,000
150,000
2070
260
380
570
5,000
5,900
7,200
85,000
120,000
170,000
2080
280
410
600
5,800
6,800
8,200
95,000
130,000
180,000
Note, given the relatively modest variation in the SC-GHG estimates across the three damage modules in
Tables 3.1.1-3.1.3, the values presented in Table 4.1.1 are similar to what would be obtained under
alternative approaches for drawing on the multiple lines of evidence represented by the three damage
modules. For example, if the estimates for each model were weighted in such in way that the weighted
131 Only one component of the methodology for calculating coastal damages is common across the two models. Both
DSCIM and GIVE rely on the CIAM model developed by Diaz (2016) to estimate the economic damages resulting from
projections of SLR. This small degree of overlap across the two modules is unlikely to affect the representation of
structural uncertainty when pooling estimates across the two damage modules.
81
-------
EXTERNAL REVIEW DRAFT
average is the certainty-equivalent across the models,132 the average (unrounded) SC-C02 in emissions
year 2020 would change by less than 1% for all three near-term discount rates. The SC-GHG estimates
resulting from averaging across the models (as presented in Table 4.1.1) are also similar to the central
estimates presented in Tables 3.1.1-3.1.3. That is, the unrounded estimates based on the DSCIM damage
module for the 2.5% discount rate, and the GIVE damage module for the 2.0% and 1.5% discount rates, in
emissions year 2020 differ from the three-model average estimates by only 2% (2.5% discount rate), -1%
(2.0% discount rate), and -1% (1.5% discount rate).
4.2 Application of SC-GHG Estimates in Benefit-Cost Analysis
The SC-GHG reflects the future stream of damages associated with an additional ton of emissions
discounted back to the year of the emissions. Several steps are necessary when using the SC-GHG
estimates in an analysis that includes GHG emissions changes in multiple future years in addition to other
benefits and costs. First, the gas-specific SC-GHG estimates corresponding to the year of estimated
emissions change need to be applied and discounted to the year of analysis to monetize the emissions.
Second, the monetized GHG emissions impacts need to be incorporated with other costs and benefits
considered in the analysis.
The SC-GHG estimates presented in Table 4.1.1 represents the damages associated with each additional
ton of emissions released discounted back to the year of emissions. To calculate the monetized value of
damages from emissions in year r discounted back to the year of analysis, denoted as year 0, two steps
are required. First, the emissions changes in the future year, xT, are multiplied times the SC-GHG in that
future year, scghgT, to the obtain the future monetized net damages associated with those emissions.
Second, that value needs to be discounted back to the year of analysis to obtain the present value of the
damages, pv0, using the discount factor ST. Mathematically, these two steps can be written as
pv0 = xT ¦ scghgT ¦ ST . (4.2.7)
The correct discount factor to use when discounting the SC-GHG estimates presented in this report is the
certainty-equivalent discount factor, ST. This is because the SC-GHG estimates are certainty-equivalent
values that account for the uncertainty in future consumption per capita. As described more fully in
Appendix A.3, the certainty-equivalent discount factor incorporates the uncertainty in future
consumption using the RFF-SP probabilistic growth scenarios. Discounting the SC-GHG estimates using a
constant discount rate equal to the near-term target rate would not capture the uncertainty in
consumption per capita for that year. This means that precise discounting of a stream of future emissions
132 Specifically, the weight is estimated for each module, near-term discount rate and emission year using: wr m r] =
E\(cTmyv]
M —-3—, where c is consumption net climate change, r is emission year, m is damage module, and 77 is the
elasticity of marginal utility with respect to consumption. The resulting weights given to the damage module based
on DSCIM, GIVE, and Howard and Sterner (2017) are: 0.331, 0.334, 0.334, respectively, under Ramsey discounting
with a 2.0% near-term target rate. These weights are close to an equal weight (0.333) on modules. These three
modules share the same distributions of GDP and have estimates of damage under climate change that are
comparable. Therefore, the distributions of net consumption across the three modules are similar, leading to similar
weights.
82
-------
EXTERNAL REVIEW DRAFT
requires the SC-GHG for each year (provided in Table A.4.1) together with the certainty-equivalent
discount factor for that year.
While applying the certainty-equivalent discount factor would ensure a full accounting of scenario
uncertainty, this process introduces substantial complexity in the calculations, which may not be
warranted in all situations. If the stream of future emissions being evaluated is moderate (e.g., 30 years
or less), the difference between discounting from the year of emissions to the year of analysis using a
constant discount rate equal to the near-term target rate, and discounting using the certainty-equivalent
discount factor, ST will be small. For example, if the year of analysis is 2022 using the near-term target
rate to discount back from the year of emissions instead of the certainty-equivalent discount factor will
underestimate the present value emission reductions by less than 1% for the first ten years of future
emissions. The present value of emission reductions 30 years in the future will be underestimated by
slightly over 2% yielding a conservative approximation to the more complete calculation.133 (The
differences from using a constant discount rate rather than the certainty-equivalent discount factor for
each year in the future are provided in Figure A.3.1.) Therefore, discounting the monetized value of
emission reductions over the first 30 years of the analysis using the near-term target rate provides a close
approximation.
133 This example is based on the SC-GHG estimates using a 2 percent near-term Ramsey discount rate. The
quantitative results will vary slightly across the near-term target rates considered in this report, but the difference
between the two approaches remains relatively small over the first 30 years.
83
-------
EXTERNAL REVIEW DRAFT
5 Summary
This report presents new estimates of the SC-GHG that reflect recent advances in the scientific literature
on climate change and its economic impacts and recommendations made by the National Academies of
Science, Engineering, and Medicine in 2017.
Since 2008, the EPA has used estimates of the SC-GHG in analyses of actions that affect GHG emissions.
The values used by the EPA from 2009 to 2016, and since 2021, have been consistent with those
developed and recommended by the Interagency Working Group on the SC-GHG (IWG), and the values
used from 2017-2020 were consistent with those required by E.O. 13783. During that time, the National
Academies conducted a comprehensive review of the social cost of carbon and issued a final report in
2017 that recommended specific criteria for future updates to the SC-C02 estimates, a modeling
framework to satisfy the specified criteria, and both near-term updates and longer-term research needs
pertaining to various components of the estimation process. The IWG was reconstituted in 2021 and E.O.
13990 directed it to develop a comprehensive update of its SC-GHG estimates, recommendations
regarding areas of decision-making to which SC-GHG should be applied, and a standardized review and
updating process to ensure that the recommended estimates continue to be based on the best available
economics and science going forward.
The EPA is a member of the IWG and is participating in the IWG's work under E.O. 13990. While that
process continues, this report presents a set of SC-GHG estimates that incorporates recent research
addressing the near-term recommendations of the National Academies. The report takes a modular
approach in which the methodology underlying each of the four components, or modules, of the SC-GHG
estimation process - socioeconomics and emissions, climate, damages, and discounting - is developed by
drawing on the latest research and expertise from the scientific disciplines relevant to that component.
Table 5.1 summarizes the key elements of the National Academies' near-term recommendations for each
module and how the methodological updates employed in this report addressed those recommendations.
The modeling implemented in this report reflects conservative methodological choices, and, given both
those choices and the numerous categories of damages that are not currently quantified and other model
limitations, the resulting SC-GHG estimates likely underestimate the marginal damages from greenhouse
gas pollution. The EPA will continue to review developments in the literature, including more robust
methodologies for estimating the magnitude of the various direct and indirect damages from GHG
emissions, and look for opportunities to further improve SC-GHG estimation going forward.
84
-------
EXTERNAL REVIEW DRAFT
Table 5.1: Implementation of National Academies Recommendations in this Report
Near-term National Academies' recommendations Methodological updates employed in this report
Overarching
~ Framework: Adopt a modular approach to allow
relevant disciplinary expertise to shape each
part of the analysis.
~ Scientific basis: Modules should be consistent
with scientific knowledge in the current, peer-
reviewed literature.
~ Adopted a modular modeling framework that unbundled the
socioeconomic-emissions scenarios, climate modeling, damage
function modeling, and discounting to allow each component to
be informed by high-quality science from the relevant disciplines.
~ Selected modeling frameworks and parameters for each module
based on recent peer-reviewed scientific literature and scientific
consensus reports.
~ Uncertainty characterization: Key uncertainties,
including functional forms, parameter
assumptions, and data inputs, should be
adequately represented and uncertainties not
quantified should be identified.
~ Transparency: Documentation should allow
readers to understand and assess the modules,
including which features are evidence-based or
judgment-based. Model code should be
available to researchers.
Socioeconomic module
~ Use statistical methods and expert elicitation for [] Adopted the probabilistic RFF-SPs, which provide multi-century
~ Expanded upon past estimates used by the EPA by incorporating a
quantitative consideration of uncertainty into all modules and
using a Monte Carlo approach to develop SC-GHG distributions
that captures interactions across modules' uncertainties.
~ Documented modeling features in detail, including within
replication instructions and computer code that has been made
publicly available.
projecting probability distributions of GDP,
population growth and emissions into the
future.
Climate module
~ Employ a reduced complexity Earth system
model that satisfies well-defined diagnostic
tests, such as the FaIR model, to represent
temperature change over time, and include sea-
level rise and ocean pH components.
Damages module
~ Improve and update existing damage functions
to reflect recent scientific literature.
Discounting module
~ Incorporate the relationship between discount
rates and economic growth using a Ramsey-like
framework and parameters chosen consistent
with theory and empirical evidence on
consumption interest rates.
projections of population, GDP per capita, and GHG emissions
based on statistical and structured expert judgment methods that
account for future policies and connections between variables.
~ Adopted FaIR 1.6.2 to serve as the basis for an updated climate
module, which provides an accurate representation of the latest
scientific consensus on the relationship between global emissions
and global mean surface temperature under a wide range of
socioeconomic emissions scenarios, complemented by the BRICK
and FACTS models of sea-level rise.
~ Adopted a suite of three updated damage functions (GIVE,
DSCIM, and the meta-analysis), which together represent the
major scientific lines of evidence on the economic impacts of
climate change that are available, capture uncertainty, and, in the
cases of GIVE and DSCIM, provide transparent bottom-up
modeling that map Earth system changes to damages.
~ Adopted a Ramsey discounting approach that endogenously
connects the discount rate and the socioeconomic scenarios and
where the parameters are empirically calibrated based on
observed behavior of interest rates and economic growth.
85
-------
EXTERNAL REVIEW DRAFT
References
Abt Associates, 2012. Development of probabilistic socio-economic emissions scenarios. Report prepared
under EPA Contract No. EP-W-11-003. https://www.epa.gov/sites/default/files/2Q18-
Q2/documents/ee-0574-01.pdf
Achtnicht, M., 2012. German car buyers' willingness to pay to reduce C02 emissions. Climatic Change,
113(3), pp.679-697.
Ackerman, F., Stanton, E.A. and Bueno, R., 2012 (August). CRED v. 1.4 Technical Report. Stockholm
Environment Institute.
http://frankackerman.com/publications/climatechange/CREDl.4TechnicalReport.pdf
Adler, M., Anthoff, D., Bosetti, V., Garner, G., Keller, K. and Treich, N., 2017. Priority for the worse-off
and the social cost of carbon. Nature Climate Change, 7(6), pp.443-449.
Aldy, J.E., Kotchen, M.J., Stavins, R.N. and Stock, J.H., 2021. Keep climate policy focused on the social
cost of carbon. Science, 373(6557), pp.850-852.
Anthoff, D. and Emmerling, J., 2019. Inequality and the social cost of carbon. Journal of the Association
of Environmental and Resource Economists, 6(2), pp.243-273.
Anthoff, D. and Tol, R.S., 2010. On international equity weights and national decision making on climate
change. Journal of Environmental Economics and Management, 60(1), pp.14-20.
Anthoff, D. and Tol, R.S.J., 2013a. The uncertainty about the social cost of carbon: a decomposition
analysis using FUND. Climatic Change, 117(3), pp.515-530.
Anthoff, D. and Tol, R.S.J., 2013b. Erratum to: The uncertainty about the social cost of carbon: A
decomposition analysis using FUND. Climatic Change, 121(2), pp.413.
Anthoff, D., Tol, R.S. and Yohe, G.W., 2009. Risk aversion, time preference, and the social cost of
carbon. Environmental Research Letters, 4(2) 024002, pp. 1-7.
Arora, V.K., Katavouta, A., Williams, R.G., Jones, C.D., Brovkin, V., Friedlingstein, P., Schwinger, J., Bopp,
L., Boucher, O., Cadule, P. and Chamberlain, M.A., 2020. Carbon-concentration and carbon-climate
feedbacks in CMIP6 models and their comparison to CMIP5 models. Biogeosciences, 17(16),
pp.4173-4222.
Arrow, K., Cline, W.R., Maler, K.G., Munasinghe, M., Squitieri, R., Stiglitz, J.E., 1995. Intertemporal
equity, discounting, and economic efficiency. Climate Change 1995 - Economic and Social
Dimensions of Climate Change. Contribution of Working Group III to the Second Assessment Report
of the IPCC, pp. 125-144.
86
-------
EXTERNAL REVIEW DRAFT
Arrow, K., Cropper, M., Gollier, C., Groom, B., Heal, G., Newell, R., Nordhaus, W., Pindyck, R., Pizer, W.,
and Portney, P., 2013. Determining Benefits and Costs for Future Generations. Science, 341(6144),
pp.349-350.
Auffhammer, M., 2018. Quantifying economic damages from climate change. Journal of Economic
Perspectives, 32(4), pp.33-52.
Azar, C., 1999. Weight factors in cost-benefit analysis of climate change. Environmental and Resource
Economics, 13(3), pp.249-268.
Azar, C. and Sterner, T., 1996. Discounting and distributional considerations in the context of global
warming. Ecological Economics, 19(2), pp.169-184.
Bastien-Olvera, B.A., Granella, F., and Moore, F.C., 2022. Persistent effect of temperature on GDP
identified from lower frequency temperature variability. Environmental Research Letters, 17(8)
084038, pp.1-10.
Bauer, M.D. and Rudebusch, G.D., 2017. Resolving the spanning puzzle in macro-finance term structure
models. Review of Finance, 21(2), pp.511-553.
Bauer, M.D. and Rudebusch, G.D., 2020. Interest rates under falling stars. American Economic
Review, 110(5), pp.1316-54.
Bauer, M.D. and Rudebusch, G.D., 2021. The rising cost of climate change: evidence from the bond
market. The Review of Economics and Statistics, https://doi.org/10.1162/rest_a_01109.
Beach, R.H., Sulser, T.B., Crimmins, A., Cenacchi, N., Cole, J., Fukagawa, N.K., Mason-D'Croz, D., Myers,
S., Sarofim, M.C., Smith, M. and Ziska, L.H., 2019. Combining the effects of increased atmospheric
carbon dioxide on protein, iron, and zinc availability and projected climate change on global diets: a
modelling study. The Lancet Planetary Health, 3(7), pp.e307-e317.
Benveniste, H., Oppenheimer, M. and Fleurbaey, M., 2020. Effect of border policy on exposure and
vulnerability to climate change. Proceedings of the National Academy of Sciences, 117(43),
pp.26692-26702.
Boardman, A.E., Moore, M.A. and Vining, A.R., 2010. The social discount rate for Canada based on future
growth in consumption. Canadian Public Policy, 36(3), pp.325-343.
Brander, L.M., Rehdanz, K., Tol, R.S. and Van Beukering, P.J., 2012. The economic impact of ocean
acidification on coral reefs. Climate Change Economics, 3(01), p.1250002.
Bressler, R.D., 2021. The mortality cost of carbon. Nature Communications, 12(1), pp.1-12.
Brooks, W.R. and Newbold, S.C., 2014. An updated biodiversity nonuse value function for use in climate
change integrated assessment models. Ecological Economics, 105, pp.342-349.
Burke, M., Davis, W.M. and Diffenbaugh, N.S., 2018. Large potential reduction in economic damages
under UN mitigation targets. Nature, 557(7706), pp.549-553.
87
-------
EXTERNAL REVIEW DRAFT
Burke, M., Hsiang, S.M. and Miguel, E., 2015. Global non-linear effect of temperature on economic
production. Nature, 527(7577), pp.235-239.
Burke, M. and Tanutama, V., 2019. Climatic constraints on aggregate economic output (No. w25779).
National Bureau of Economic Research, https://www.nber.org/papers/w25779 .
Cai, Y., Lenton, T.M. and Lontzek, T.S., 2016. Risk of multiple interacting tipping points should encourage
rapid C02 emission reduction. Nature Climate Change, 6(5), pp.520-525.
Cai, Y. and Lontzek, T.S., 2019. The social cost of carbon with economic and climate risks. Journal of
Political Economy, 127(6), pp.2684-2734.
Cameron, T.A., 2010. Euthanizing the Value of a Statistical Life. Review of Environmental Economics and
Policy, 4(2), pp.161-78.
Carleton, T. and Greenstone, M., 2022. A guide to updating the U.S. government's social cost of carbon.
Review of Environmental Economics and Policy, 16(2), pp.196-218.
Carleton, T., Jina, A., Delgado, M., Greenstone, M., Houser, T., Hsiang, S., Hultgren, A., Kopp, R.E.,
McCusker, K.E., Nath, I., Rising, J., Ashwin, A., Seo, H., Viaene, A., Yaun, J., and Zhang, A., 2022.
Valuing the Global Mortality Consequences of Climate Change Accounting for Adaptation Costs and
Benefits. The Quarterly Journal of Economics, 137(4), pp. 2037-2105.
Cass, D., 1965. Optimum growth in an aggregative model of capital accumulation. The Review of
Economic Studies, 32(3), pp.233-240.
Center for Climate and Security (CCS), 2018. Military Expert Panel Report: Sea Level Rise and the U.S.
Military's Mission, Second Edition. https://perma.cc/QMR3-CDBZ.
Christensen, P., Gillingham, K. and Nordhaus, W., 2018. Uncertainty in forecasts of long-run economic
growth. Proceedings of the National Academy of Sciences, 115(21), pp.5409-5414.
Christensen, J.H. and Rudebusch, G.D., 2019. A new normal for interest rates? Evidence from inflation-
indexed debt. Review of Economics and Statistics, 101(5), pp.933-949.
Clarke, L., Edmonds, J., Krey, V., Richels, R., Rose, S. and Tavoni, M., 2009. International climate policy
architectures: Overview of the EMF 22 International Scenarios. Energy Economics, 31, pp.S64-S81.
Clarke, L., Eom, J., Marten, E.H., Horowitz, R., Kyle, P., Link, R., Mignone, B.K., Mundra, A. and Zhou, Y.,
2018. Effects of long-term climate change on global building energy expenditures. Energy
Economics, 72, pp.667-677.
Clarkson, R. and Deyes, K., 2002. Estimating the social cost of carbon emissions. London: HM Treasury.
Government Economic Service Working Paper 140.
Climate Impact Lab (CIL), 2022. Data-driven Spatial Climate Impact Model User Manual, Version 092022-
EPA. https://impactlab.org/research/dscim-user-manual-version-092022-epa
CNA Corporation Military Advisory Board (CNA), 2007. National Security and the Threat of Climate
Change. Alexandria, VA: CNA.
88
-------
EXTERNAL REVIEW DRAFT
https://www.cria.org/archive/CNA Files/pdf/national%20securitv%20arid%20the%20threat%20of%
20climate%20chaiige.pdf
CNA Corporation Military Advisory Board (CNA), 2014. National Security and the Accelerating Risks of
Climate Change. Alexandria, VA: CNA. https://www.cna.org/archive/CNA Files/pdf/mab 5-8-14.pdf
Colacito, R., Hoffmann, B. and Phan, T., 2019. Temperature and growth: A panel analysis of the United
States. Journal of Money, Credit and Banking, 51(2-3), pp.313-368.
Colon-Gonzalez, F.J., Sewe, M.O., Tompkins, A.M., Sjodin, H., Casallas, A., Rocklov, J., Caminade, C. and
Lowe, R., 2021. Projecting the risk of mosquito-borne diseases in a warmer and more populated
world: a multi-model, multi-scenario intercomparison modelling study. The Lancet Planetary
Health, 5(7), pp.e404-e414.
Colt, S.G. and Knapp, G.P., 2016. Economic effects of an ocean acidification catastrophe. American
Economic Review, 106(b), pp.615-19.
Congressional Budget Office (CBO), 2021a. The 2021 Long Term Budget Outlook.
https://www.cbo.gov/publication/56977
Congressional Budget Office (CBO), 2021b. An Update to the Budget and Economic Outlook: 2021 to
2031. https://www.cbo.gov/publication/57218
Cooley, S.R. and Doney, S.C., 2009. Anticipating ocean acidification's economic consequences for
commercial fisheries. Environmental Research Letters, 4(2) 024007, pp.1-8.
Cooley, S.R., Rheuban, J.E., Hart, D.R., Luu, V., Glover, D.M., Hare, J.A. and Doney, S.C., 2015. An
integrated assessment model for helping the United States sea scallop (Placopecten magellanicus)
fishery plan ahead for ocean acidification and warming. PLoS One, 10(5), p.e0124145.
Costinot, A., Donaldson, D. and Smith, C., 2016. Evolving comparative advantage and the impact of
climate change in agricultural markets: Evidence from 1.7 million fields around the world. Journal of
Political Economy, 124(1), pp.205-248.
Council of Economic Advisors (CEA). 2017. Discounting for Public Policy: Theory and Recent Evidence on
the Merits of Updating the Discount Rate. U.S. Government Council of Economic Advisors Issue Brief.
https://obamawhitehouse.archives.gov/sites/default/files/page/files/201701 cea discounting issu
e brief.pdf
Cromar K.R., Anenberg S.C., Balmes J.R., Fawcett A.A., Ghazipura M., Gohlke J.M., Hashizume M.,
Howard P., Lavigne E., Levy K., MadriganoJ., Martinich J.A., Mordecai E.A., Rice M.B., Saha S.,
Scovronick N.C., Sekercioglu F., Svendsen E.R., Zaitchik B.F., and Ewart G., 2022. Global Health
Impacts for Economic Models of Climate Change: A Systematic Review and Meta-Analysis. Annals of
the American Thoracic Society, 19(7), pp. 1203-1212.
Cropper, M.L., Freeman, M.C., Groom, B. and Pizer, W.A., 2014. Declining discount rates. American
Economic Review, 104(5), pp.538-43.
89
-------
EXTERNAL REVIEW DRAFT
Crost, B. and Traeger, C.P., 2014. Optimal C02 mitigation under damage risk valuation. Nature Climate
Change, 4(7), pp.631-636.
Daniel, K.D., Litterman, R.B. and Wagner, G., 2019. Declining C02 price paths. Proceedings of the
National Academy of Sciences, 116(42), pp.20886-20891.
Dasgupta, P., 2008. Discounting climate change. Journal of Risk and Uncertainty, 37(2), pp.141-169.
Dasgupta, P., 2020 (Summer). Ramsey and Intergenerational Welfare Economics. The Stanford
Encyclopedia of Philosophy. [Edward N. Zalta (ed.)].
https://plato.stanford.edu/archives/sum2020/entries/ramsev-economics/
Dell, M., Jones, B.F. and Olken, B.A., 2012. Temperature shocks and economic growth: Evidence from
the last half century. American Economic Journal: Macroeconomics, 4(3), pp.66-95.
Del Negro, M., Giannone, D., Giannoni, M.P. and Tambalotti, A., 2017. Safety, liquidity, and the natural
rate of interest. Brookings Papers on Economic Activity, 2017(1), pp.235-316.
Dennig, F., Budolfson, M.B., Fleurbaey, M., Siebert, A. and Socolow, R.H., 2015. Inequality, climate
impacts on the future poor, and carbon prices. Proceedings of the National Academy of
Sciences, 112(52), pp.15827-15832.
Depsky, N., Bolliger, I., Allen, D., Choi, J.H., Delgado, M., Greenstone, M., Hamidi, A., Houser, T., Kopp,
R.E. and Hsiang, S., 2022. DSCIM-Coastal vl. 0: An Open-Source Modeling Platform for Global
Impacts of Sea Level Rise. EGUsphere [preprint], https://doi.org/10.5194/egusphere-2022-198
Deryugina, T. and Hsiang, S., 2017. The marginal product of climate (No. w24072). National Bureau of
Economic Research, https://www.nber.org/papers/w24072
Diaz, D.B., 2016. Estimating global damages from sea level rise with the Coastal Impact and Adaptation
Model (CIAM). Climatic Change, 137(1), pp.143-156.
Dietz, S., Gollier, C. and Kessler, L., 2018. The climate beta. Journal of Environmental Economics and
Management, 87, pp.258-274.
Dietz, S., van der Ploeg, F., Rezai, A. and Venmans, F., 2021. Are economists getting climate dynamics
right and does it matter?. Journal of the Association of Environmental and Resource
Economists, 8(5), pp.895-921.
Dietz, S. and Venmans, F., 2019. Cumulative carbon emissions and economic policy: in search of general
principles. Journal of Environmental Economics and Management, 96, pp.108-129.
Dreze, J.H., 1974. Investment under private ownership: optimality, equilibrium and stability.
In Allocation under uncertainty: equilibrium and optimality (pp. 129-166). Palgrave Macmillan,
London.
Drupp, M.A., Freeman, M.C., Groom, B. and Nesje, F., 2018. Discounting disentangled. American
Economic Journal: Economic Policy, 10(4), pp.109-34.
90
-------
EXTERNAL REVIEW DRAFT
Edmonds, J. and Reilly, J., 1983. A long-term global energy-economic model of carbon dioxide release
from fossil fuel use. Energy Economics, 5(2), pp.74-88.
Executive Order (E.O.) 13783, March 28, 2017. Promoting energy independence and economic growth.
https://www.federalregister.gov/documents/2017/03/31/2017-06576/promoting-energy-
independence-and-economic-growth
Executive Order (E.O.) 13990, January 20, 2021. Protecting public health and the environment and
restoring science totackle the climate crisis.
https://www.federalregister.gov/documents/2021/01/25/2021-Q1765/protecting-public-health-
and-the-environment-and-restoring-science-to-tackle-the-climate-crisis
Executive Order (E.O.) 12866, October 4, 1993. Regulatory planning and review, Section 1(a).
https://www.archives.gov/files/federal-register/executive-orders/pdf/12866.pdf
Fankhauser, S., Tol, R.S. and Pearce, D.W., 1997. The aggregation of climate change damages: a welfare
theoretic approach. Environmental and Resource Economics, 10(3), pp.249-266.
Farrow, S., 1998. Environmental equity and sustainability: rejecting the Kaldor-Hicks criteria. Ecological
Economics, 27(2), pp.183-188.
Fawcett, A.A., Calvin, K.V., Francisco, C., Reilly, J.M. and Weyant, J.P., 2009. Overview of EMF 22 US
transition scenarios. Energy Economics, 31(Supplement 2), pp.S198-S211.
Fernandes, J.A., Papathanasopoulou, E., Hattam, C., Queiros, A.M., Cheung, W.W., Yool, A., Artioli, Y.,
Pope, E.C., Flynn, K.J., Merino, G. and Calosi, P., 2017. Estimating the ecological, economic and social
impacts of ocean acidification and warming on UK fisheries. Fish and Fisheries, 18(3), pp.389-411.
Freeman, A.M.III., 1979. The Benefits of Environmental Improvement. Baltimore: Johns Hopkins
University Press.
Freeman, A.M.III., 1982. Air and Water Pollution Control: A Benefit-Cost Assessment. New York: John
Wiley.
Freeman, J. and Guzman, A., 2009. Climate change and US interests. Columbia Law Review, 109,
pp.1531-1602.
Fuhrer, J.C., Olivei, G.P. and Tootell, G.M., 2012. Inflation dynamics when inflation is near zero. Journal
of Money, Credit and Banking, 44(si), pp.83-122.
Garner, G. G., T. Hermans, R. E. Kopp, A. B. A. Slangen, T. L. Edwards, A. Levermann, S. Nowikci, M. D.
Palmer, C. Smith, B. Fox-Kemper, H. T. Hewitt, C. Xiao, G. ASalgeirsdottir, S. S. Drijfhout, T. L.
Edwards, N. R. Golledge, M. Hemer, R. E. Kopp, G. Krinner, A. Mix, D. Notz, S. Nowicki, I. S. Nurhati,
L. Ruiz, J-B. Sallee, Y. Yu, L. Hua, T. Palmer, B. Pearson. 2021. IPCC AR6 Sea-Level Rise Projections.
Version 20210809. PO.DAAC, CA, USA. htt ps://podaac. jpl.nasa.gov/announcements/2Q21-Q8-Q9-
Sea-level-proiections-from-the-IPCC-6th-Assessment-Report
91
-------
EXTERNAL REVIEW DRAFT
Gasser, T., Crepin, L., Quilcaille, Y., Houghton, R.A., Ciais, P. and Obersteiner, M., 2020. Historical CO 2
emissions from land use and land cover change and their uncertainty. Biogeosciences, 17(15),
pp.4075-4101.
Geoffroy, O., Saint-Martin, D., Bellon, G., Voldoire, A., Olivie, D.J.L. and Tyteca, S., 2013. Transient
climate response in a two-layer energy-balance model. Part II: Representation of the efficacy of
deep-ocean heat uptake and validation for CMIP5 AOGCMs. Journal of Climate, 26(6), pp.1859-1876.
Glanemann, N., Willner, S.N. and Levermann, A., 2020. Paris Climate Agreement passes the cost-benefit
test. Nature Communications, 11(1), pp.1-11.
Gollier, C., 2014. Discounting and growth. American Economic Review, 104(5), pp.534-37.
Gollier, C. and Hammitt, J.K., 2014. The long-run discount rate controversy. Annual Review of Resource
Economics, 6, pp.273-295.
Gollier, C. and Weitzman, M.L., 2010. How should the distant future be discounted when discount rates
are uncertain?. Economics Letters, 107(3), pp.350-353.
Golosov, M., Hassler, J., Krusell, P. and Tsyvinski, A., 2014. Optimal taxes on fossil fuel in general
equilibrium. Econometrica, 82(1), pp.41-88.
Gouel, C. and Laborde, D., 2021. The crucial role of domestic and international market-mediated
adaptation to climate change. Journal of Environmental Economics and Management, 106,
p.102408.
Greenstone, M., 2016 (May). A new path forward for an empirical social cost of carbon. Presented at The
National Academies of Sciences, Engineering, and Medicine, Assessing Approaches to Updating the
Social Cost of Carbon.
https://sites.nationalacademies.org/cs/groups/dbassesite/documents/webpage/dbasse 172.599.pdf
Groom, B., Hepburn, C., Koundouri, P. and Pearce, D., 2005. Declining discount rates: the long and the
short of it. Environmental and Resource Economics, 32(4), pp.445-493.
Gundlach, J. and Livermore, M.A., 2022. Costs, confusion, and climate change. Yale Journal on
Regulation, 39(2), pp.564-594.
Guo, C. and Costello, C., 2013. The value of adaption: Climate change and timberland
management. Journal of Environmental Economics and Management, 65(3), pp.452-468.
Hammitt, J.K., and Treich, N., 2007. Statistical vs. identified lives in benefit-cost analysis. Journal of Risk
and Uncertainty, 35, pp.45-66.
Hanemann, W.M., 2008. What is the economic cost of climate change?. CUDARE Working Papers,
University of California, Berkeley, https://escholarship.org/uc/item/9gllz5cc
Hansel, M.C., Drupp, M.A., Johansson, D.J., Nesje, F., Azar, C., Freeman, M.C., Groom, B. and Sterner, T.,
2020. Climate economics support for the UN climate targets. Nature Climate Change, 10(8), pp.781-
789.
92
-------
EXTERNAL REVIEW DRAFT
Harberger, A.C., 1972. On measuring the social opportunity cost of labour. In Project Evaluation. London
Palgrave Macmillan, pp.157-183.
Hartin, C.A., Patel, P., Schwarber, A., Link, R.P. and Bond-Lamberty, B.P., 2015. A simple object-oriented
and open-source model for scientific and policy analyses of the global climate system-Hector vl.
0. Geoscientific Model Development, 8(4), pp.939-955.
Hasegawa, T., Fujimori, S., Takahashi, K., Yokohata, T. and Masui, T., 2016. Economic implications of
climate change impacts on human health through undernourishment. Climatic Change, 136(2),
pp.189-202.
Haushofer, J. and Fehr, E., 2014. On the psychology of poverty. Science, 344(6186), pp.862-867.
Heal, G. and Kristrom, B., 2002. Uncertainty and climate change. Environmental and Resource
Economics, 22(1), pp.3-39.
Henseler, M. and Schumacher, I., 2019. The impact of weather on economic growth and its production
factors. Climatic Change, 154(3), pp.417-433.
Hoffmann, R., Dimitrova, A., Muttarak, R., Crespo Cuaresma, J. and Peisker, J., 2020. A meta-analysis of
country-level studies on environmental change and migration. Nature Climate Change, 10(10),
pp.904-912.
Hope, C., 2006. The marginal impact of C02 from PAGE2002: An integrated assessment model
incorporating the IPCC's five reasons for concern. Integrated Assessment Journal, 6(1), pp. 19-56.
Hope, C.W., 2008. Optimal carbon emissions and the social cost of carbon over time under
uncertainty. Integrated Assessment Journal, 8(1), pp.107-122.
Hope, C., 2013. Critical issues for the calculation of the social cost of C02: why the estimates from
PAGE09 are higher than those from PAGE2002. Climatic Change, 117(3), pp.531-543.
Houser, T. and Larsen, K., 2021. Calculating the Climate Reciprocity Ratio for the US. Rhodium Group.
https://rhg.com/research/climate-reciprocity-ratio/
Howard, PH. and Livermore, M.A., 2021. Climate-society feedback effects: Be wary of unidentified
connections. International Review of Environmental and Resource Economics, 15, pp.1-61.
Howard, P.H. and Schwartz, J., 2017. Think global: International reciprocity as justification for a global
social cost of carbon. Columbia Journal of Environmental Law, 42(Symposium Issue), pp.203-294.
Howard, P.H. and Sterner, T., 2017. Few and not so far between: a meta-analysis of climate damage
estimates. Environmental and Resource Economics, 68(1), pp.197-225.
Howard, P.H. and Sylvan, D., 2015. Expert Consensus on the Economics of Climate Change. Institute for
Policy Integrity, New York: NYU School of Law.
https://www.edf.org/sites/default/files/expertconsensusreport.pdf
Howard, P.H. and Sylvan, D., 2020. Wisdom of the experts: using survey responses to address positive
and normative uncertainties in climate-economic models. Climatic Change, 162(2), pp.213-232.
-------
EXTERNAL REVIEW DRAFT
Howard, P.H. and Sylvan, D., 2021. Expert Elicitation and the Social Cost of Greenhouse Gases. Institute
for Policy Integrity.
https://policvintegrity.org/files/publications/Expert Elicitation and the Social Cost of Greenhous
e Gases.pdf
Hulshof, D. and Mulder, M., 2020. Willingness to pay for C02 emission reductions in passenger car
transport. Environmental and Resource Economics, 75(4), pp.899-929.
Hultgren, A., Carleton, T., Delgado, M., Gergel, D.R., Greenstone, M., Houser, T., Hsiang, S., Jina, A.,
Kopp, R.E., Malevich, S. B., McCusker, K., Mayer, T., Nath, I., Rising, J., Rode, A. and Yuan, J., 2022
(September). Estimating global impacts to agriculture from climate change accounting for
adaptation, https://papers.ssrn.com/sol3/papers.cfm7abstract id=4222020
Hsiang, S., Kopp, R., Jina, A., Rising, J., Delgado, M., Mohan, S., Rasmussen, D.J., Muir-Wood, R., Wilson,
P., Oppenheimer, M. and Larsen, K., 2017. Estimating economic damage from climate change in the
United States. Science, 356(6345), pp.1362-1369.
ICF International. 2011. Improving the assessment and valuation of climate change impacts for policy
and regulatory analysis. U.S. EPA and DOE Workshop Report.
https://www.epa.gov/sites/default/files/2017-09/documents/ee-0566 all.pdf
Interagency Working Group on Social Cost of Carbon (IWG). 2010 (FebruaryJ. Technical Support
Document: Social Cost of Carbon for Regulatory Impact Analysis Under Executive Order 12866.
United States Government, https://www.epa.gov/sites/default/files/2016-
12/documents/scc tsd 2010.pdf
Interagency Working Group on Social Cost of Carbon (IWG). 2013 (November). Technical Support
Document: Technical Update of the Social Cost of Carbon for Regulatory Impact Analysis Under
Executive Order 12866. United States Government.
https://obamawhitehouse.archives.gov/sites/default/files/omb/assets/inforeg/technical-update-
social-cost-of-carbon-for-regulator-impact-analysis.pdf
Interagency Working Group on Social Cost of Carbon (IWG). 2015 (July). Response to Comments: Social
Cost of Carbon for Regulatory Impact Analysis Under Executive Order 12866. United States
Government, https://obamawhitehouse.archives.gov/sites/default/files/omb/inforeg/scc-response-
to-comments-final-iuly-2015.pdf
Interagency Working Group on Social Cost of Greenhouse Gases (IWG). 2016a (August). Technical
Support Document: Technical Update of the Social Cost of Carbon for Regulatory Impact Analysis
Under 12866. United States Government, https://www.epa.gov/sites/default/files/2016-
12/documents/sc co2 tsd august 2016.pdf
Interagency Working Group on the Social Cost of Greenhouse Gases (IWG). 2016b (August). Addendum
to Technical Support Document on Social Cost of Carbon for Regulatory Impact Analysis under
Executive Order 12866: Application of the Methodology to Estimate the Social Cost of Methane and
the Social Cost of Nitrous Oxide. United States Government.
94
-------
EXTERNAL REVIEW DRAFT
https://www.epa.gov/sites/default/files/2016-12/documerits/adderidum to sc-
ene tsd august 2016.pdf
Interagency Working Group on Social Cost of Carbon (IWG). 2021 (February). Technical Support
Document: Social Cost of Carbon, Methane, and Nitrous Oxide: Interim Estimates under Executive
Order 13990. United States Government, https://www.whitehouse.gov/wp-
content/uploads/2021/02/TechnicalSupportDocument SocialCostofCarbonMethaneNitrousOxide.p
df
Intergovernmental Panel on Climate Change (IPCC). 2007a. Climate Change 2007: Synthesis Report.
Contribution of Working Groups I, II and III to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A.
(eds.)]. IPCC. https://www.ipcc.ch/site/assets/uploads/2018/02/ar4 svr full report.pdf
Intergovernmental Panel on Climate Change (IPCC). 2007b. Changes in Atmospheric Constituents and in
Radiative Forcing [Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D.W. Fahey, J.
Haywood, J. Lean, D.C. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz and R. Van Dorland],
In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M.
Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)]. Cambridge University
Press, https://www.ipcc.ch/site/assets/uploads/2018/02/ar4-wgl-chapter2-l.pdf
Intergovernmental Panel on Climate Change (IPCC). 2007c. Historical Overview of Climate Change [Le
Treut, H., R. Somerville, U. Cubasch, Y. Ding, C. Mauritzen, A. Mokssit, T. Peterson and M. Prather],
In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M.
Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)]. Cambridge University
Press, https://www.ipcc.ch/site/assets/uploads/2018/03/ar4-wgl-chapterl.pdf
Intergovernmental Panel on Climate Change (IPCC). 2013. Anthropogenic and Natural Radiative Forcing
[Myhre, G., D. Shindell, F.-M. Breon, W. Collins, J. Fuglestvedt, J. Huang, D. Koch, J.-F. Lamarque, D.
Lee, B. Mendoza, T. Nakajima, A. Robock, G. Stephens, T. Takemura and H. Zhang], In Climate
Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment
Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M.
Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge
University Press. https://www.ipcc.ch/site/assets/uploads/2018/02/WGlAR5 ChapterOS FINAL.pdf
Intergovernmental Panel on Climate Change (IPCC). 2014a. Climate Change 2014: Synthesis Report.
Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC
https://www.ipcc.ch/site/assets/uploads/2018/05/? i FINAL full wcover.pdf
Intergovernmental Panel on Climate Change (IPCC). 2014b. Social, Economic and Ethical Concepts and
Methods [Kolstad C., K. Urama, J. Broome, A. Bruvoll, M. Carino Olvera, D. Fullerton, C. Gollier, W.M.
Hanemann, R. Hassan, F. Jotzo, M.R. Khan, L. Meyer, and L. Mundaca], In Climate Change 2014:
Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of
the Intergovernmental Panel on Climate Change [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E.
-------
EXTERNAL REVIEW DRAFT
Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J.
Savolainen, S. Schlomer, C. von Stechow, T. Zwickel and J.C. Minx (eds.)]. Cambridge University
Press, https://www.ipcc.ch/site/assets/uploads/2018/02/ipcc wg3 ar5 chapter3.pdf
Intergovernmental Panel on Climate Change (IPCC). 2018. Global Warming of 1.5°C. An IPCC Special
Report on the impacts of global warming of 1.5°C above pre-industrial levels and related global
greenhouse gas emission pathways, in the context of strengthening the global response to the
threat of climate change, sustainable development, and efforts to eradicate poverty [Masson-
Delmotte, V., P. Zhai, H.-O. Portner, D. Roberts, J. Skea, P.R. Shukla, A. Pirani, W. Moufouma-Okia, C.
Pean, R. Pidcock, S. Connors, J.B.R. Matthews, Y. Chen, X. Zhou, M.I. Gomis, E. Lonnoy, T. Maycock,
M. Tignor, and T. Waterfield (eds.)].
Intergovernmental Panel on Climate Change (IPCC). 2019a. Climate Change and Land: an IPCC special
report on climate change, desertification, land degradation, sustainable land management, food
security, and greenhouse gas fluxes in terrestrial ecosystems [P.R. Shukla, J. Skea, E. Calvo Buendia,
V. Masson-Delmotte, H.-O. Portner, D. C. Roberts, P. Zhai, R. Slade, S. Connors, R. van Diemen, M.
Ferrat, E. Haughey, S. Luz, S. Neogi, M. Pathak, J. Petzold, J. Portugal Pereira, P. Vyas, E. Huntley, K.
Kissick, M. Belkacemi, J. Malley, (eds.)].
Intergovernmental Panel on Climate Change (IPCC). 2019b. IPCC Special Report on the Ocean and
Cryosphere in a Changing Climate [H.-O. Portner, D.C. Roberts, V. Masson-Delmotte, P. Zhai, M.
Tignor, E. Poloczanska, K. Mintenbeck, A. Alegrfa, M. Nicolai, A. Okem, J. Petzold, B. Rama, N.M.
Weyer (eds.)].
Intergovernmental Panel on Climate Change (IPCC). 2021a. Climate Change 2021: The Physical Science
Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental
Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Pean, S. Berger, N.
Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K.
Maycock, T. Waterfield, O. Yelekgi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press.
Intergovernmental Panel on Climate Change (IPCC). 2021b. The Earth's Energy Budget, Climate
Feedbacks, and Climate Sensitivity [Forster, P., T. Storelvmo, K. Armour, W. Collins, J.-L. Dufresne, D.
Frame, D.J. Lunt, T. Mauritsen, M.D. Palmer, M. Watanabe, M. Wild, and H. Zhang], In Climate
Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment
Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani,
S.L. Connors, C. Pean, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E.
Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekgi, R. Yu, and B. Zhou (eds.)].
Cambridge University Press.
https://www.ipcc.ch/report/ar6/wgl/downloads/report/IPCC AR6 WG1 Chapter07.pdf
Intergovernmental Panel on Climate Change (IPCC). 2021c. Ocean, Cryosphere and Sea Level Change
[Fox-Kemper, B., H. T. Hewitt, C. Xiao, G. ASalgeirsdottir, S. S. Drijfhout, T. L. Edwards, N. R.
Golledge, M. Hemer, R. E. Kopp, G. Krinner, A. Mix, D. Notz, S. Nowicki, I. S. Nurhati, L. Ruiz, J-B.
Sallee, A. B. A. Slangen, Y. Yu.]. In Climate Change 2021: The Physical Science Basis. Contribution of
Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change
[Masson-Delmotte, V., P. Zhai, A. Pirani, S. L. Connors, C. Pean, S. Berger, N. Caud, Y. Chen, L.
96
-------
EXTERNAL REVIEW DRAFT
Goldfarb, M. I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J. B. R. Matthews, T. K. Maycock, T.
Waterfield, O. Yelekgi, R. Yu and B. Zhou (eds.)]. Cambridge University Press
https://www.ipcc.ch/report/ar6/wgl/downloads/report/IPCC AR6 WG1 Chapter09.pdf
Intergovernmental Panel on Climate Change (IPCC). 2021d. Future Global Climate: Scenario-Based
Projections and Near-Term Information [Lee, J.-Y., J. Marotzke, G. Bala, L. Cao, S. Corti, J.P. Dunne, F.
Engelbrecht, E. Fischer, J.C. Fyfe, C. Jones, A. Maycock, J. Mutemi, O. Ndiaye, S. Panickal, and T.
Zhou], In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the
Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V.,
P. Zhai, A. Pirani, S.L. Connors, C. Pean, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M.
Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekgi, R. Yu, and B.
Zhou (eds.)]. Cambridge University Press.
https://www.ipcc.ch/report/ar6/wgl/downloads/report/IPCC AR6 WG1 Chapter04.pdf
Intergovernmental Panel on Climate Change (IPCC). 2021e. Short-Lived Climate Forcers [Szopa, S.,
V. Naik, B. Adhikary, P. Artaxo, T. Berntsen, W.D. Collins, S. Fuzzi, L. Gallardo, A. Kiendler-Scharr,
Z. Klimont, H. Liao, N. Unger, and P. Zanis], In Climate Change 2021: The Physical Science Basis.
Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on
Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Pean, S. Berger, N. Caud,
Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock,
T. Waterfield, O. Yelekgi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press.
https://www.ipcc.ch/report/ar6/wgl/downloads/report/IPCC AR6 WG1 Chapter06.pdf
Intergovernmental Panel on Climate Change (IPCC), 2022. Climate Change 2022: Impacts, Adaptation
and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the
Intergovernmental Panel on Climate Change [H.-O. Portner, D.C. Roberts, M. Tignor, E.S.
Poloczanska, K. Mintenbeck, A. Alegrfa, M. Craig, S. Langsdorf, S. Loschke, V. Moller, A. Okem, B.
Rama (eds.)]. Cambridge University Press.
https://www.ipcc.ch/report/ar6/wg2/downloads/report/IPCC AR6 WGII FullReport.pdf
Jensen, S. and Traeger, C.P., 2014. Optimal climate change mitigation under long-term growth
uncertainty: Stochastic integrated assessment and analytic findings. European Economic Review, 69,
pp.104-125.
Jensen, S. and Traeger, C., 2021. Pricing Climate Risk. CESifo Working Paper 9196.
Joos, F., Roth, R., Fuglestvedt, J.S., Peters, G.P., Enting, I.G., Von Bloh, W., Brovkin, V., Burke, E.J., Eby,
M., Edwards, N.R. and Friedrich, T., 2013. Carbon dioxide and climate impulse response functions for
the computation of greenhouse gas metrics: a multi-model analysis. Atmospheric Chemistry and
Physics, 13(5), pp.2793-2825.
Just, R.E., Hueth, D.L. and Schmitz, A., 2004. The welfare economics of public policy: a practical approach
to project and policy evaluation. Edward Elgar Publishing.
Kahn, M. E., Mohaddes, K., Ng, R. N., Pesaran, M. H., Raissi, M., & Yang, J. C. (2021). Long-term
macroeconomic effects of climate change: A cross-country analysis. Energy Economics, 104, 105624.
97
-------
EXTERNAL REVIEW DRAFT
Kalkuhl, M. and Wenz, L., 2020. The impact of climate conditions on economic production. Evidence
from a global panel of regions. Journal of Environmental Economics and Management, 103,
p.102360.
Kanter, D.R., Wagner-Riddle, C., Groffman, P.M., Davidson, E.A., Galloway, J.N., Gourevitch, J.D., van
Grinsven, H.J., Houlton, B.Z., Keeler, B.L., Ogle, S.M. and Pearen, H., 2021. Improving the social cost
of nitrous oxide. Nature Climate Change, 11(12), pp.1008-1010.
Kelleher, J.P. and Wagner, G. 2019. Ramsey discounting calls for subtracting climate damages from
economic growth rates. Applied Economics Letters, 26(1), pp.79-82.
Kikstra, J.S., Waidelich, P., Rising, J., Yumashev, D., Hope, C. and Brierley, C.M., 2021. The social cost of
carbon dioxide under climate-economy feedbacks and temperature variability. Environmental
Research Letters, 16(9), p.094037.
Kimball, M.S. 1990. Precautionary Saving in the Small and in the Large. Econometrica, 58(1), pp.53-73.
Kite-Powell, H.L., 2009. A global perspective on the economics of ocean acidification. Journal of Marine
Education, 25, pp.25-29.
Klein, S.G., Geraldi, N.R., Anton, A., Schmidt-Roach, S., Ziegler, M., Cziesielski, M.J., Martin, C., Radecker,
N., Frolicher, T.L., Mumby, P.J. and Pandolfi, J.M., 2022. Projecting coral responses to intensifying
marine heatwaves under ocean acidification. Global Change Biology, 28(5), pp.1753-1765.
Koopmans, T.C., 1963. On the Concept of Optimal Economic Growth (No. 163). Cowles Foundation for
Research in Economics, Yale University.
Kopits, E., Marten, A. and Wolverton, A., 2014. Incorporating 'catastrophic' climate change into policy
analysis. Climate Policy, 14(5). pp.637-664.
Kopp, R. E., DeConto, R. M., Bader, D. A., Hay, C. C., Horton, R. M., Kulp, S., Oppenheimer, M., Pollard,
D., & Strauss, B. H., 2017. Evolving understanding of Antarctic ice-sheet physics and ambiguity in
Probabilistic Sea-level projections. Earth's Future, 5(12), 1217-1233.
Kopp, R.E., Kemp, A.C., Bittermann, K., Horton, B.P., Donnelly, J.P., Gehrels, W.R., Hay, C.C., Mitrovica,
J.X., Morrow, E.D. and Rahmstorf, S., 2016. Temperature-driven global sea-level variability in the
Common Era. Proceedings of the National Academy of Sciences, 113(11), pp.E1434-E1441.
Kopp, R.E. and Mignone, B.K. 2013. Circumspection, reciprocity, and optimal carbon prices. Climatic
Change 120(4): pp.831-843.
Kotchen, M.J. 2018. Which Social Cost of Carbon? A Theoretical Perspective. Journal of the Association
of Environmental and Resource Economists, 5(3): pp.673-694.
Kotchen, M.J., 2021. Comment on Technical Support Document: Social Cost of Carbon, Methane, and
Nitrous Oxide Interim Estimates. Comment submitted on June 11, 2021.
https://www.regulations.gov/comment/OMB-2021-0006-0Q18
Kumar, S. and Khanna, M., 2019. Temperature and production efficiency growth: empirical
evidence. Climatic Change, 156(1), pp.209-229.
98
-------
EXTERNAL REVIEW DRAFT
Landrigan, Philip J, Richard Fuller, Nereus J R Acosta, Olusoji Adeyi, Robert Arnold, Niladri (Nil) Basu,
Abdoulaye Bibi Balde, Roberto Bertollini, Stephan Bose-O'Reilly, Jo Ivey Boufford, Patrick N Breysse,
Thomas Chiles, Chulabhorn Mahidol, Awa M Coll-Seck, Maureen L Cropper, Julius Fobil, Valentin
Fuster, Michael Greenstone, Andy Haines, David Hanrahan, David Hunter, Mukesh Khare, Alan
Krupnick, Bruce Lanphear, Bindu Lohani, Keith Martin, Karen V Mathiasen, Maureen A McTeer,
Christopher J L Murray, Johanita D Ndahimananjara, Frederica Perera, Janez Potocnik, Alexander S
Preker, Jairam Ramesh, Johan Rockstrom, Carlos Salinas, Leona D Samson, Karti Sandilya, Peter D
Sly, Kirk R Smith, Achim Steiner, Richard B Stewart, William A Suk, Onno C P van Schayck, Gautam N
Yadama, Kandeh Yumkella, and Ma Zhong., 2018. The Lancet Commission on pollution and health.
The Lancet, 391(10119), pp.462-512.
Lane, D.R., Ready, R.C., Buddemeier, R.W., Martinich, J.A., Shouse, K.C. and Wobus, C.W., 2013.
Quantifying and valuing potential climate change impacts on coral reefs in the United States:
Comparison of two scenarios. PLoS One, S(12), p.e82579.
Laxminarayan, R., Klein, E.Y., Dye, C., Floyd, K., Darley, S. and Adeyi, O., 2007. Economic Benefit of
Tuberculosis Control. World Bank Policy Research Working Paper, (4295).
https://openknowledge.worldbank.org/handle/10986/7483 License: CC BY 3.0 IGO."
Leach, N.J., Jenkins, S., NicholIs, Z., Smith, C.J., Lynch, J., Cain, M., Walsh, T., Wu, B., Tsutsui, J. and Allen,
M.R., 2021. FalRv2. 0.0: a generalized impulse response model for climate uncertainty and future
scenario exploration. Geoscientific Model Development, 14(5), pp.3007-3036.
Lemoine, D., 2021. The climate risk premium: how uncertainty affects the social cost of carbon. Journal
of the Association of Environmental and Resource Economists, 8(1), pp.27-57.
Levy, K., Woster, A.P., Goldstein, R.S. and Carlton, E.J., 2016. Untangling the impacts of climate change
on waterborne diseases: a systematic review of relationships between diarrheal diseases and
temperature, rainfall, flooding, and drought. Environmental Science & Technology, 50(10), pp.4905-
4922.
Li, Q. and Pizer, W.A., 2021. Use of the consumption discount rate for public policy over the distant
future. Journal of Environmental Economics and Management, 107, p.102428.
Lind, R.C., 1990. Reassessing the government's discount rate policy in light of new theory and data in a
world economy with a high degree of capital mobility. Journal of Environmental Economics and
Management, 18 (2): S8-S28.
Little, I.M.D., 2002. A critique of welfare economics. Oxford University Press.
Lusardi, A. and Mitchell, O.S., 2014. The economic importance of financial literacy: Theory and
evidence. Journal of Economic Literature, 52(1), pp. 5-44.
Lutz, M.A., 1995. Centering social economics on human dignity. Review of Social Economy, 53(2),
pp.171-194.
Lyon, R.M., 1990. Federal discount rate policy, the shadow price of capital, and challenges for
reforms. Journal of Environmental Economics and Management, 18(2), pp.S29-S50.
99
-------
EXTERNAL REVIEW DRAFT
Markandya, A., Sampedro, J., Smith, S.J., Van Dingenen, R., Pizarro-lrizar, C., Arto, I. and Gonzalez-
Eguino, M., 2018. Health co-benefits from air pollution and mitigation costs of the Paris Agreement:
a modelling study. The Lancet Planetary Health, 2(3), pp.el26-el33.
Marten, A.L. 2014. The role of scenario uncertainty in estimating the benefits of carbon mitigation.
Climate Change Economics 5(3), pp.1-29.
Marten, A. L., Kopits, E. A., Griffiths, C. W., Newbold, S. C., and Wolverton, A. 2015. Incremental CH4 and
N20 mitigation benefits consistent with the US Government's SC-C02 estimates. Climate Policy 15(2),
pp.272-298.
Masterman, C.J., and Viscusi, W.K., 2018. The income elasticity of global values of a statistical life: stated
preference evidence. Journal of Benefit-Cost Analysis 9(3), pp.407-434.
Matthews, H.D. and Caldeira, K., 2008. Stabilizing climate requires near-zero emissions. Geophysical
Research Letters, 35(4).
Meinshausen, M., Raper, S.C. and Wigley, T.M., 2011. Emulating coupled atmosphere-ocean and carbon
cycle models with a simpler model, MAGICC6-Part 1: Model description and
calibration. Atmospheric Chemistry and Physics, 11(4), pp.1417-1456.
Mendelsohn, R., 1980. An economic analysis of air pollution from coal-fired power plants. Journal of
Environmental Economics and Management, 7(1), pp.30-43.
Meyer, A. and Cooper, T., 1995. A recalculation of the social costs of climate change. The Ecologist.
Millar, R.J., Nicholls, Z.R., Friedlingstein, P. and Allen, M.R. 2017. A modified impulse-response
representation of the global near-surface air temperature and atmospheric concentration response
to carbon dioxide emissions. Atmospheric Chemistry and Physics, 17(11), pp.7213-7228.
Moore, C., 2015. Welfare estimates of avoided ocean acidification in the U.S. mollusk market. Journal of
Agricultural and Resource Economics, 40(1), pp.50-62.
Moore, M.A., Boardman, A.E., and Vining, A.R., 2013. More appropriate discounting: the rate of social
time preference and the value of the social discount rate. Journal of Benefit-Cost Analysis, 4(1), pp.l-
16.
Moore, F.C., Baldos, U., Hertel, T., and D.B. Diaz., 2017. New science of climate change impacts on
agriculture implies higher social cost of carbon. Nature Communications, 8(1), pp.1-9.
Mordecai, E.A., Ryan, S.J., Caldwell, J.M., Shah, M.M. and LaBeaud, A.D., 2020. Climate change could
shift disease burden from malaria to arboviruses in Africa. The Lancet Planetary Health, 4(9),
pp.e416-e423.
Muller, U.K., Stock, J.H. and Watson, M.W., 2022. An Econometric Model of International Growth
Dynamics for Long-Horizon Forecasting. Review of Economics and Statistics, 104(5), pp.857-876.
Muller, U.K., and M.W. Watson. 2016. Measuring Uncertainty about Long-Run Predictions. Review of
Economic Studies 83(4), pp.1711-40.
100
-------
EXTERNAL REVIEW DRAFT
Narita, D., Rehdanz, K. and Tol, R.S., 2012. Economic costs of ocean acidification: a look into the impacts
on global shellfish production. Climatic Change, 113(3), pp.1049-1063.
Narita, D. and Rehdanz, K., 2017. Economic impact of ocean acidification on shellfish production in
Europe. Journal of Environmental Planning and Management, 60(3), pp.500-518.
Narita, D., Tol, R.S. and Anthoff, D., 2010. Economic costs of extratropical storms under climate change:
an application of FUND. Journal of Environmental Planning and Management, 53(3), pp.371-384.
National Academies of Sciences, Engineering, and Medicine (National Academies). 2016a. Assessment of
Approaches to Updating the Social Cost of Carbon: Phase 1 Report on a Near-Term Update. National
Academies Press.
National Academies of Sciences, Engineering, and Medicine (National Academies). 2016b. Attribution of
Extreme Weather Events in the Context of Climate Change. National Academies Press.
National Academies of Sciences, Engineering, and Medicine (National Academies). 2017. Valuing Climate
Damages: Updating Estimation of the Social Cost of Carbon Dioxide. National Academies Press.
National Academies of Sciences, Engineering, and Medicine (National Academies). 2019. Climate Change
and Ecosystems. National Academies Press.
National Intelligence Council (NIC), 2021. National Intelligence Estimate: Climate Change and
International Responses Increasing Challenges to US National Security Through 2040. NIC-NIE-2021-
10030-A.
https://www.dni.gov/files/ODNI/documents/assessments/NIE Climate Change and National Secu
ritv.pdf.
National Research Council (NRC), 2011. America's climate choices. National Academies Press.
Newell, R.G., Pizer, W.A. and Prest, B.C., 2022. A Discounting Rule for the Social Cost of Carbon. Journal
of the Association of Environmental and Resource Economists, 9(5), pp. 1017-1046.
Newell, R.G. and Pizer, W.A., 2003. Discounting the distant future: how much do uncertain rates
increase valuations?. Journal of environmental economics and management, 46(1), pp.52-71.
Newell, R.G., Prest, B.C. and Sexton, S.E., 2021. The GDP-temperature relationship: implications for
climate change damages. Journal of Environmental Economics and Management, 108, p.102445
Nicholls Z., Meinshausen M., Lewis J., Gieseke R., Dommenget D., Dorheim K., Fan C., Fuglestvedt J.S.,
GasserT., Goliike U., Goodwin P., Hartin C., Hope A.P., Kriegler E., Leach N.J., Marchegiani D.,
McBride L.A., Quilcaille Y., RogeljJ., Salawitch R.J., Samset B.H., Sandstad M., Shiklomanov A.N.,
Skeie R.B., Smith C.J., Smith S., Tanaka K., Tsutsui J., and Xie Z., 2020. Reduced Complexity Model
Intercomparison Project Phase 1: introduction and evaluation of global-mean temperature
response. Geoscientific Model Development, 13(11), pp.5175-5190.
Nicholls Z., Meinshausen M., Lewis J., Corradi M.R., Dorheim K., Gasser T., Gieseke R., Hope A.P., Leach
N.J., McBride L.A., Quilcaille Y., RogeljJ., Salawitch R.J., Samset B.H., Sandstad M., Shiklomanov A.,
Skeie R.B., Smith C.J., Smith S.J., Su X., Tsutsui J., Vega-Westhoff B., and Woodard D.L., 2021.
101
-------
EXTERNAL REVIEW DRAFT
Reduced complexity Model Intercomparison Project Phase 2: Synthesizing Earth system knowledge
for probabilistic climate projections. Earth's Future, 9(6), p.e2020EF001900.
Nordhaus, W.D., 1993a. Optimal greenhouse-gas reductions and tax policy in the" DICE"
model. American Economic Review, 83(2), pp.313-317.
Nordhaus, W.D., 1993b. Rolling the 'DICE': an optimal transition path for controlling greenhouse
gases. Resource and Energy Economics, 15(1), pp.27-50.
Nordhaus, W., 2007. Alternative measures of output in global economic-environmental models:
Purchasing power parity or market exchange rates?. Energy Economics, 29(3), pp.349-372.
Nordhaus, W., 2008. A Question of Balance: Weighing the Options on Global Warming Policies. Yale
University Press.
Nordhaus, W., 2010. Economic aspects of global warming in a post-Copenhagen environment.
Proceedings of the National Academy of Sciences, 107(26): 11721-11726.
Nordhaus, W., 2014. Estimates of the Social Cost of Carbon: Concepts and Results from the DICE-2013R
Model and Alternative Approaches. Journal of the Association of Environmental and Resource
Economists, 1(1/2), 273-312.
Nordhaus, W., 2015. Climate clubs: Overcoming free-riding in international climate policy. American
Economic Review, 105(4), pp.1339-70.
Nordhaus, W.D., 2017a. Evolution of Assessments of the Economics of Global Warming: Changes in the
DICE model, 1992-2017 (No. w23319). National Bureau of Economic Research.
Nordhaus, W.D., 2017b. Revisiting the social cost of carbon. Proceedings of the National Academy of
Sciences, 114(7), pp.1518-1523.
Nordhaus, W., 2018. Projections and uncertainties about climate change in an era of minimal climate
policies. American economic journal: economic policy, 10(3), pp.333-60.
Nordhaus, W., 2019. Climate change: The ultimate challenge for economics. American Economic
Review, 109(6), pp.1991-2014.
Nordhaus W., and Boyer, J. 2000. Warming the World: Economic Models of Global Warming. MIT Press.
Nordhaus, W.D. and Moffat, A., 2017. A survey of global impacts of climate change: replication, survey
methods, and a statistical analysis. National Bureau of Economic Research. Working Paper 23646.
http://www.nber.org/papers/w23646
Nordhaus, W. and Sztorc, P., 2013. DICE 2013R: introduction and user's manual. (Cowles Found, New
Haven, CT).
Organization for Economic Cooperation and Development (OECD). 2018. Cost-Benefit Analysis and the
Environment: Further Developments and Policy Use. OECD Publishing.
https://www.oecd.org/governance/cost-benefit-analvsis-and-the-environment-9789264085169-
en.htm
102
-------
EXTERNAL REVIEW DRAFT
Organization for Economic Cooperation and Development (OECD). 2016. The Economic Consequences of
Outdoor Air Pollution. Paris: OECD Publishing, http://www.oecd.org/env/the-economic-
consequences-of-outdoor-air-pollution-9789264257474- en.htm
Office of Management and Budget (OMB). 2003. Circular A-4, Regulatory Analysis. OMB.
https://obamawhitehouse.archives.gov/omb/circulars a004 a-4/
Office of Management and Budget (OMB). 1972. Circular A-94, Guidelines and Discount Rates for
Benefit-Cost Analysis of Federal Programs. OMB.
https://obamawhitehouse.archives.gov/omb/circulars_a094/
Okullo, S.J., 2020. Determining the social cost of carbon: Under damage and climate sensitivity
uncertainty. Environmental and Resource Economics, 75(1), pp.79-103.
Persky, J., 2001. Cost-benefit analysis and the classical creed. Journal of Economic Perspectives, 15(4),
pp.199-208.
Pindyck, R.S., 2013. Climate change policy: what do the models tell us?. Journal of Economic
Literature, 51(3), pp.860-72.
Pindyck, R.S., 2017. Comments on Proposed Rule and Regulatory Impact Analysis on the Delay and
Suspension of Certain Requirements for Waster Prevention and Resource Conservation. Comment
submitted on Nov. 6, 2017. https://downloads.regulations.gov/EPA-HQ-QAR-2018-0283-
6184/attachment 6.pdf
Pindyck, R.S., 2019. The social cost of carbon revisited. Journal of Environmental Economics and
Management, 94, pp.140-160.
Pindyck, R.S. 2021. Comments on "Technical Support Document: Social Cost of Carbon, Methane, and
Nitrous Oxide Interim Estimates Under Executive Order 13990". Comment submitted on June 15,
2021. https://downloads.regulations.gov/OMB-2021-0006-0Q12/attachment l.pdf
Pizer, W., Adler, M., Aldy, J., Anthoff, D., Cropper, M., Gillingham, K., Greenstone, M., Murray, B.,
Newell, R., Richels, R. and Rowell, A., 2014. Using and improving the social cost of
carbon. Science, 346(6214), pp. 1189-1190.
Portmann, R.W., Daniel, J.S. and Ravishankara, A.R., 2012. Stratospheric ozone depletion due to nitrous
oxide: Influences of other gases. Philosophical transactions of the Royal Society of London. Series B,
Biological sciences, 367(1593), 1256-1264.
Raftery, A.E., and H. Sevcfkova. 2021. Probabilistic population forecasting: Short to very long-term.
International Journal of Forecasting.
Ramsey, F.P., 1928. A mathematical theory of saving. The Economic Journal, 38(152), pp.543-559.
Ravishankara, A.R., Daniel, J.S. and Portmann, R.W., 2009. Nitrous oxide (N20): the dominant ozone-
depleting substance emitted in the 21st century. Science, 326(5949), pp.123-125.
Rennert, K., Prest, B.C., Pizer, W.A., Newell, R.G., Anthoff, D., Kingdon, C., Rennels, L., Cooke, R., Raftery,
A.E., Sevcfkova, H. and Errickson, F., 2022a. The social cost of carbon: Advances in long-term
103
-------
EXTERNAL REVIEW DRAFT
probabilistic projections of population, GDP, emissions, and discount rates. Brookings Papers on
Economic Activity. Fall 2021, pp.223-305.
Rennert, K., Errickson, F., Prest, B.C., Rennels, L., Newell, R., Pizer, W., Kingdon, C., Wingenroth, J.,
Cooke, R., Parthum, B., Smith, D., Cromar, K., Diaz, D., Moore, F., Muller, U., Plevin, R., Raftery, A.,
Sevcfkova, H., Sheets, H., Stock, J., Tan, T., Watson, M., Wong, T., and Anthoff, D., 2022b.
[Forthcoming], Comprehensive evidence implies a higher social cost of C02. Nature.
Revesz, R., Greenstone, M., Hanemann, M., Livermore, M., Sterner, T., Grab, D., Howard, P. and
Schwartz, J., 2017. Best cost estimate of greenhouse gases. Science, 357(6352), pp.655-655.
Riahi, K., van Vuuren, D. P., Kriegler, E., Edmonds, J., O'Neill, B. C., Fujimori, S., Bauer, N., Calvin, K.,
Dellink, R., Fricko, O., Lutz, W., Popp, A., Cuaresma, J. C., Kc, S., Leimbach, M., Jiang, L., Kram, T.,
Rao, S., Emmerling, J., Ebi, K., Hasegawa, T., Havlik, P., Humpenoder, F., Da Silva, L. A., Smith, S.,
Stehfest, E., Bosetti, V., Eom, J., Gernaat, D., Masui, T., Rogelj, J., Strefler, J., Drouet, L., Krey, V.,
Luderer, G., Harmsen, M., Takahashi, K., Baumstark, L., Doelman, J. C., Kainuma, M., Klimont, Z.,
Marangoni, G., Lotze-Campen, H., Obersteiner, M., Tabeau, A., Tavoni, M., 2017. The Shared
Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: An
overview. Global Environmental Change. 42, pp.153-168.
Ries, J.B., Cohen, A.L. and McCorkle, D.C., 2009. Marine calcifiers exhibit mixed responses to C02-
induced ocean acidification. Geology, 37(12), pp.1131-1134.
Rising, J. and Devineni, N., 2020. Crop switching reduces agricultural losses from climate change in the
United States by half under RCP 8.5. Nature Communications, 11{ 1), pp.1-7.
Robinson, L.A., Hammitt, J.K., Cecchini, M., Chalkidou, K., Claxton, K., Cropper, M., Eozenou, P.H., de
Ferranti, D., Deolalikar, A.B., Guanais, F., and D.T. Jamison. 2019a. Reference case guidelines for
benefit-cost analysis in global health and development. Cambridge, MA: Harvard University.
https://cdnl.sph.harvard.edu/wp-content/uploads/sites/2447/2019/05/BCA-Guidelines-Mav-
2019.pdf
Robinson, L.A., Hammitt, J.K., and O'Keeffe, L., 2019b. Valuing Mortality Risk Reductions in Global
Benefit-Cost Analysis. Journal of Benefit-Cost Analysis 10(S1), pp.15-50.
Robinson, L.A., Hammitt, J.K., and O'Keeffe, L., 2018. Valuing Mortality Risk Reductions in Global Benefit-
Cost Analysis. Guidelines for Benefit-Cost Analysis Project, Working Paper No. 7.
https://cdn2.sph.harvard.edu/wp-content/uploads/sites/94/2017/01/Robinson-Hammitt-QKeeffe-
VSL.2018.03.23.pdf
Rode, A., Baker, R. E., Carleton, T., D'Agostino, A., Delgado, M., Foreman, T., Gergel, D. R., Greenstone,
M., Houser, T., Hsiang, S., Hultgren, A., Jina, A., Kopp, R. E., Malevich, S. B., McCusker, K., Nath, I.,
Pecenco, M., Rising, J. and Yuan, J., 2022. Labor disutility in a warmer world: The impact of climate
change on the global workforce. September.
https://papers.ssrn.com/sol3/papers.cfm7abstract id=4221478
104
-------
EXTERNAL REVIEW DRAFT
Rode, A., Carleton, T., Delgado, M., Greenstone, M., Houser, T., Hsiang, S., Hultgren, A., Jina, A., Kopp,
R.E., McCusker, K.E. and Nath, I., 2021. Estimating a social cost of carbon for global energy
consumption. Nature, 598(7880), pp.308-314.
Roe, G.H. and Baker, M.B. 2007. Why is climate sensitivity so unpredictable? Science 318(5850), pp.629-
632.
Rose, S., Turner, D., Blanford, G., Bistline, J., de la Chesnaye, F. and Wilson, T., 2014. Understanding the
social cost of carbon: A technical assessment. EPRI technical update report. Electric Power Research
Inst, Palo Alto, CA.
Roy, R., 2016. The cost of air pollution in Africa. OECD Development Centre Working Papers, No. 333,
OECD Publishing, Paris, https://doi.org/10.1787/5ilazq77x6f8-en.
Roy, R. and Braathen N., 2017. The Rising Cost of Ambient Air Pollution thus far in the 21st Century:
Results from the BRIICS and the OECD Countries. OECD Environment Working Papers, No. 124, OECD
Publishing, Paris, https://doi.org/10.1787/dlb2b844-en.
Ryan, S.J., Carlson, C.J., Mordecai, E.A. and Johnson, L.R., 2019. Global expansion and redistribution of
Aedes-borne virus transmission risk with climate change. PLoS Neglected Tropical Diseases, 13(3),
pp.1-20.
Ryan, S.J., McNally, A., Johnson, L.R., Mordecai, E.A., Ben-Horin, T., Paaijmans, K. and Lafferty, K.D.,
2015. Mapping physiological suitability limits for malaria in Africa under climate change. Vector-
Borne and Zoonotic Diseases, 15(12), pp.718-725.
Sandmo, A. and Dreze, J.H., 1971. Discount rates for public investment in closed and open
economies. Economica, 38(152), pp.395-412.
Sarofim, M.C., Smith, J.B., St Juliana, A. and Hartin, C., 2021a. Improving reduced complexity model
assessment and usability. Nature Climate Change, 11(1), pp.1-3.
Sarofim, M.C., Martinich, J., Neumann, J.E., Willwerth, J., Kerrich, Z., Kolian, M., Fant, C. and Hartin, C.,
2021b. A temperature binning approach for multi-sector climate impact analysis. Climatic
Change, 165(1), pp.1-18.
Sarofim, M.C., Waldhoff, S.T. and Anenberg, S.C., 2017. Valuing the ozone-related health benefits of
methane emission controls. Environmental and Resource Economics, 66(1), pp.45-63.
Schwartz, J. 2021. Strategically Estimating Climate Pollution Costs in a Global Environment. Institute for
Policy Integrity at the New York University School of Law, New York, NY.
https://policvintegritv.org/files/publications/Strategically Estimating Climate Pollution Costs in a
Global Environment.pdf
Schwartz, J. and Howard, P., 2022. Valuing the Future: Legal and Economic Considerations for Updating
Discount Rates. Yale Journal on Regulation, 39(2), pp.595-657.
Schenker, O., 2013. Exchanging goods and damages: the role of trade on the distribution of climate
change costs. Environmental and Resource Economics, 54(2), pp.261-282.
105
-------
EXTERNAL REVIEW DRAFT
Scitovsky, T., 1951. The state of welfare economics. The American Economic Review, 41(3), pp.303-315.
Skeie, R.B., Peters, G.P., Fuglestvedt, J. and Andrew, R., 2021. A future perspective of historical
contributions to climate change. Climatic Change, 164(24), pp.1-13.
Slangen, A.B.A., Carson, M., Katsman, C.A., Van de Wal, R.S.W., Kohl, A., Vermeersen, L.L.A. and
Stammer, D., 2014. Projecting twenty-first century regional sea-level changes. Climatic
Change, 124(1), pp.317-332.
Smith, C.J., Forster, P.M., Allen, M., Leach, N., Millar, R.J., Passerello, G.A. and Regayre, L.A., 2018. FAIR
vl. 3: a simple emissions-based impulse response and carbon cycle model. Geoscientific Model
Development, 11(6), pp.2273-2297.
Social Security Administration (SSA), 2021. The 2021 Annual Report of the Board of Trustees of the
Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Fund.
https://www.ssa.gov/OACT/TR/2Q21/
Speers, A.E., Besedin, E.Y., Palardy, J.E. and Moore, C., 2016. Impacts of climate change and ocean
acidification on coral reef fisheries: an integrated ecological-economic model. Ecological
Economics, 128, pp.33-43.
Springmann, M., Godfray, H.C.J., Rayner, M. and Scarborough, P., 2016. Analysis and valuation of the
health and climate change cobenefits of dietary change. Proceedings of the National Academy of
Sciences, 113(15), pp.4146-4151.
Stern, N., 2006. The economics of climate change: Stern Review. Cambridge University Press, 712 pp.
Stern, N., Stiglitz, J.E. and Taylor, C., 2022. The economics of immense risk, urgent action and radical
change: towards new approaches to the economics of climate change. Journal of Economic
Methodology, 29(3), pp. 181-216.
Stevanovic, M., Popp, A., Lotze-Campen, H., Dietrich, J.P., Miiller, C., Bonsch, M., Schmitz, C., Bodirsky,
B.L., Humpenoder, F. and Weindl, I., 2016. The impact of high-end climate change on agricultural
welfare. Science advances, 2(8), p.el501452.
Tan-soo, J-S., 2021. A Cost-Benefit Analysis of Tamil Nadu Urban Sanitation Improvement Plans. Asian
Development Bank Institute (ADBI). Case Study No. 2021-2 (July).
Thompson, T.M., 2018. Modeling the climate and carbon systems to estimate the social cost of
carbon. Wiley Interdisciplinary Reviews: Climate Change, 9(5), p.e532.
Tol, R., 2002a. Estimates of the damage costs of climate change. Part I: benchmark estimates.
Environmental and Resource Economics 21, pp.47-73.
Tol, R., 2002b. Estimates of the damage costs of climate change. Part II: dynamic estimates.
Environmental and Resource Economics 21, pp.135-160.
Tol, R., 2009. An analysis of mitigation as a response to climate change. Copenhagen Consensus on
Climate. Copenhagen Consensus Center.
106
-------
EXTERNAL REVIEW DRAFT
Tol, R., 2019. A social cost of carbon for (almost) every country. Energy Economics, 83, pp.555-566.
Trinanes, J. and Martinez-Urtaza, J., 2021. Future scenarios of risk of Vibrio infections in a warming
planet: a global mapping study. The Lancet Planetary Health, 5(7), pp.e426-e435.
United Nations (UN), 2015. World population prospects: the 2015 revision. New York: United Nations.
United Nations Environment Programme (UNEP) and Climate and Clean Air Coalition, 2021. Global
Methane Assessment: Benefits and Costs of Mitigating Methane Emissions. United Nations
Environment Programme.
U.S. Department of Defense (DoD), 2014. 2014 Climate Change Adaptation Roadmap.
https://www.acq.osd.mil/eie/downloads/CCARprint wForward e.pdf
U.S. Department of Defense (DoD), 2019. Report on Effects of a Changing Climate to the Department of
Defense. Office of the Under Secretary of Defense for Acquisition and Sustainment
https://media.defense.gOv/2019/Jan/29/2002084200/-l/-l/l/CLIMATE-CHANGE-REPQRT-2019.PDF
U.S. Department of Defense (DoD), 2021. Department of Defense Climate Risk Analysis. Office of the
Undersecretary for Policy (Strategy, Plans, and Capabilities). Report Submitted to National Security
Council. https://media.defense.gOv/2021/Oct/21/2002877353/-l/-l/0/DQD-CLIMATE-RISK-
ANALYSIS-FINAL.PDF
U.S. Environmental Protection Agency (EPA) and U.S. Department of Transportation (DOT), 2009.
Proposed Rulemaking to Establish Light-Duty Vehicle Greenhouse Gas Emission Standards and
Corporate Average Fuel Economy Standards. https://www.govinfo.gov/content/pkg/FR-2009-Q9-
28/pdf ZE9-22516.pdf
U.S. Environmental Protection Agency (EPA) and U.S. Department of Transportation (DOT), 2015. Draft
Regulatory Impact Analysis: Proposed Rulemaking for Greenhouse Gas Emissions and Fuel Efficiency
Standards for Medium- and Heavy-Duty Engines and Vehicles-Phase 2. June.
https://nepis.epa.gov/Exe/ZvPDF.cgi/P10QM KYR.PDF?Dockev=P100MKYR. PDF
U.S. Environmental Protection Agency (EPA) and U.S. Department of Transportation (DOT), 2016.
Regulatory Impact Analysis of the Final Rulemaking for Greenhouse Gas Emissions and Fuel
Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles-Phase 2. EPA-420-R-16-900.
https://nepis.epa.gov/Exe/ZyPDF.cgi/P100P7NS. PDF?Dockey=P100P7NS. PDF
U.S. Environmental Protection Agency (EPA), 2010. Guidelines for Preparing Economic Analyses. EPA-
240-R-10-001. https://www.epa.gov/environmental-economics/guidelines-preparing-economic-
analvses.
U.S. Environmental Protection Agency (EPA), 2011. EPA Science Advisory Board (SAB) Letter to
Administrator Jackson on SAB's Review of Valuing Mortality Risk Reductions for Environmental
Policy: A White Paper. December, https://www.epa.gov/system/files/documents/2022-
03/86189901 O.pdf.
U.S. Environmental Protection Agency (EPA), 2012a. Regulatory Impact Analysis: Final New Source
Performance Standards and Amendments to the National Emissions Standards for Hazardous Air
107
-------
EXTERNAL REVIEW DRAFT
Pollutants for the Oil and Natural Gas Industry. April.
http://www.epa.gov/ttn/ecas/regdata/RIAs/oil natural gas final neshap nsps ria.pdf
U.S. Environmental Protection Agency (EPA), 2012b. Regulatory Impact Analysis: Final Rulemaking for
2017-2025 Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel
Economy Standards. EPA-420-R-12-016. August.
https://nepis.epa.gov/Exe/ZvPURL.cgi?Dockey=P100EZIl.TXT
U.S. Environmental Protection Agency (EPA), 2015a. Regulatory Impact Analysis for the Proposed
Revisions to the Emission Guidelines for Existing Sources and Supplemental Proposed New Source
Performance Standards in the Municipal Solid Waste Landfills Sector. August.
https://www.regulations.gov/document?D=EPA-HQ-OAR-2014-0451-0086
U.S. Environmental Protection Agency (EPA), 2015b. Regulatory Impact Analysis of the Proposed
Emission Standards for New and Modified Sources in the Oil and Natural Gas Sector. EPA-452/R-15-
002. August. https://www.regulations.gov/document/EPA-HQ-QAR-2010-0505-5258
U.S. Environmental Protection Agency (EPA), 2016a. Regulatory Impact Analysis of the Final Oil and
Natural Gas Sector: Emission Standards for New, Reconstructed, and Modified Sources. EPA-452/R-
16-002. https://www.epa.gov/sites/default/files/20 ocuments/oilgas ria nsps final 2016-
05.pdf
U.S. Environmental Protection Agency (EPA), 2016b. Regulatory Impact Analysis for the Final Revisions
to the Emission Guidelines for Existing Sources and the Final New Source Performance Standards in
the Municipal Solid Waste Landfills Sector. EPA-452/R-16-003.
https://www3.epa.gov/ttnecasl/docs/ria/landfills ria final-eg-nsps 2016-07.pdf
U.S. Environmental Protection Agency (EPA), 2020. Integrated Science Assessment for Ozone and
Related Photochemical Oxidants, EPA/600/R-20/012.
https://ordspub.epa.gov/ords/eims/eimscomm.getfile7p download id=540022
U.S. Environmental Protection Agency (EPA), 2021a. Regulatory Impact Analysis for Phasing Down
Production and Consumption of Hydrofluorocarbons (HFCs).
https://www.epa.gov/system/files/documents/2021-09/ria-w-works-cited-for-docket.pdf.
U.S. Environmental Protection Agency (EPA), 2021b. Regulatory Impact Analysis for the Proposed
Standards of Performance for New, Reconstructed, and Modified Sources and Emissions Guidelines
for Existing Sources: Oil and Natural Gas Sector Climate Review. EPA-452/R-21-003. October.
https://www.epa.gov/system/files/documents/2021-ll/proposal-ria-oil-and-gas-nsps-eg-climate-
review O.pdf
U.S. Environmental Protection Agency (EPA), 2021c. Revised 2023 and Later Model Year Light Duty
Vehicle GHG Emissions Standards: Regulatory Impact Analysis. EPA-420-R-21-028. December
https://nepis.epa.gov/Exe/ZyPDF.cgi?Dockey=P1013QRN.pdf
U.S. Environmental Protection Agency (EPA), 2021d. Technical Documentation on the Framework for
Evaluating Damages and Impacts (FrEDI). EPA 430-R-21-004. https://www.epa.gov/cira/fredi
108
-------
EXTERNAL REVIEW DRAFT
U.S. Environmental Protection Agency (EPA), 2021e. Climate Change and Social Vulnerability in the
United States: A Focus on Six Impacts, EPA 430-R-21-003. https://www.epa.gov/cira/social"
vulnerability-report
U.S. Environmental Protection Agency (EPA), 2021f. EPA Science Advisory Board (SAB) Letter to
Administrator Wheeler on SAB's Review of EPA's Revised Guidelines for Preparing Economic
Analysis. January, https://sab.epa.gov/ords/sab/f?p=100:18:2203987188456:::RP,18:P18_ID:2545
U.S. Global Change Research Program (USGCRP), 2016. The Impacts of Climate Change on Human Health
in the United States: A Scientific Assessment. [Crimmins, A., Balbus, J., Gamble, J.L., Beard, C.B., Bell,
J.E., Dodgen, D., Eisen, R.J., Fann, N., Hawkins, M.D., Herring, S.C., Jantarasami, L., Mills, D.M., Saha,
S., Sarofim, M.C., Trtanj, J., and Ziska, L., (eds.)] U.S. Global Change Research Program, Washington,
DC, 312 pp.
U.S. Global Change Research Program (USGCRP), 2018a. Impacts, Risks, and Adaptation in the United
States: Fourth National Climate Assessment, Volume II [Reidmiller, D.R., C.W. Avery, D.R. Easterling,
K.E. Kunkel, K.L.M. Lewis, T.K. Maycock, and B.C. Stewart (eds.)]. U.S. Global Change Research
Program, Washington, DC, 1515 pp.
U.S. Global Change Research Program (USGCRP), 2018b. Human Health. In Impacts, Risks, and
Adaptation in the United States: Fourth National Climate Assessment, Volume II [Ebi, K.L., Balbus,
J.M., Luber, G., Bole, A., Crimmins, A., Glass, G., Saha, S., Shimamoto, M.M., Trtanj, J. and White-
Newsome, J.L. (authors)], [Reidmiller, D.R., C.W. Avery, D.R. Easterling, K.E. Kunkel, K.L.M. Lewis, T.K.
Maycock, and B.C. Stewart (eds.)]. U.S. Global Change Research Program, Washington, DC, 33 pp.
U.S. Global Change Research Program (USGCRP), 2018c: Air Quality. In Impacts, Risks, and Adaptation in
the United States: Fourth National Climate Assessment, Volume II [Nolte, C.G., P.D. Dolwick, N. Fann,
L.W. Horowitz, V. Naik, R.W. Pinder, T.L. Spero, D.A. Winner, and L.H. Ziska], [Reidmiller, D.R., C.W.
Avery, D.R. Easterling, K.E. Kunkel, K.L.M. Lewis, T.K. Maycock, and B.C. Stewart (eds.)]. U.S. Global
Change Research Program, Washington, DC, USA, pp. 512-538.
U.S. Government Accountability Office (GAO), 2014. Regulatory Impact Analysis: Development of Social
Cost of Carbon Estimates. GAO-14-663. July. Available at: https://www.gao.gov/products/GAO-14-
663.
U.S. Government Accountability Office (GAO), 2020. Social Cost of Carbon: Identifying a Federal Entity to
Address the National Academies' Recommendations Could Strengthen Regulatory Analysis. GAO-20-
254. June, https://www.gao.gov/assets/gao-20-254.pdf
U.S. Millenium Challenge Corporation (MCC), 2021. Cost Benefit Analysis Guidelines.
https://www.mcc.gov/resources/doc-pdf/cost-benefit-analysis-guidelines.
Van den Bremer, T.S. and Van der Ploeg, F., 2021. The risk-adjusted carbon price. American Economic
Review, 111(9), pp.2782-2810.
Vega-Westhoff, B., Sriver, R.L., Hartin, C.A., Wong, T.E., and Keller, K., 2019. Impacts of Observational
Constraints Related to Sea Level on Estimates of Climate Sensitivity. Earth's Future, 7(6), pp.677-690.
109
-------
EXTERNAL REVIEW DRAFT
Viscusi, K. and Masterman, J., 2017a. Income Elasticities and Global Values of a Statistical Life. Journal of
Benefit Cost Analysis, 8(2), pp.226-250
Viscusi, K. and Masterman, J., 2017b. Anchoring Biases in International Estimates of the Value of a
Statistical Life. Journal of Risk and Uncertainty, 54(2), pp. 103-128.
Weijer, W., Cheng, W., Garuba, O.A., Hu, A. and Nadiga, B.T., 2020. CMIP6 Models Predict Significant
21st Century Decline of the Atlantic Meridional Overturning Circulation. Geophysical Research
Letters 47(12), p. e2019GL086075.
Weitzman, M.L., 2007. A review of the Stern Review on the economics of climate change. Journal of
Economic Literature, 45(3), pp.703-724.
Weitzman, M.L., 2012. GHG targets as insurance against catastrophic climate damages. Journal of Public
Economic Theory, 14(2), pp.221-244.
Wong, T. E., Bakker, A. M. R., Ruckert, K., Applegate, P., Slangen, A. B. A., and Keller, K. 2017. BRICK v0.2,
a simple, accessible, and transparent model framework for climate and regional sea-level
projections. Geoscientific Model Development, 10, pp.2741-2760.
Woodard, D. L., Shiklomanov, A. N., Kravitz, B., Hartin, C., and Bond-Lamberty, B., 2021. A permafrost
implementation in the simple carbon-climate model Hector v.2.3pf, Geoscientific Model
Development, 14, 4751-4767.
World Bank, 2021. World Development Indicators. World Bank.
World Bank and Institute for Health Metrics and Evaluation (IHME). 2016. The Cost of Air Pollution:
Strengthening the Economic Case for Action.
http://documents.worldbank.org/curated/en/781521473177Q13155/The-cost-of-air-
pollutionstrengthening-the-economic-case-for-action
World Meteorological Organization (WMO), 2018. Scientific Assessment of Ozone Depletion: 2018,
Global Ozone Research and Monitoring Project —Report No. 58.
Ziska, L. H., 2020. An overview of rising C02 and climatic change on aeroallergens and allergic diseases.
Allergy, Asthma & Immunology Research, 12(5), p.771-782.
110
-------
EXTERNAL REVIEW DRAFT
A. Appendix
A.l. Additional Discussion of Scientific Updates in IPCC's Sixth Assessment Report
Several updates to the science of greenhouse gas radiative efficiency134, atmospheric lifetimes, and
chemistry have been made since the IWG published its first set of recommended SC-GHG estimates in
2010. In this report projections of temperature change from a pulse of GHG emissions are based on the
FaIR climate model, version 1.6.2, rather than using the simplified lifetime and forcing equations from the
IPCC AR4 assessment that were embedded in the lAMs underlying the SC-GHG estimates used to date.
While FaIR is a more complex model that includes internal feedbacks and chemistry such that gas lifetimes
and interactions are not constant, it can be instructive to examine how the more simplistic equations have
been updated between AR4 (IPCC 2007b) and AR6 (IPCC 2021b) as FaIR 1.6.2 reflects many of the same
scientific advances in understanding.
The radiative efficiency of all gases has been updated, in part because of updates to the science and in
part because radiative efficiency is a function of background concentrations. The radiative efficiency of
C02 has decreased by 5% relative to AR4, while the radiative efficiencies of CH4 and N20 have both
increased by about 5%. AR6 also updated the indirect effects of CH4 and N20 that occur through
atmospheric chemistry. The indirect radiative effects of CH4 that occur through increases in ozone and
stratospheric water vapor decreased by about 6%. Meanwhile, the radiative effects of N20 now include
the impact of N20 on CH4 and stratospheric ozone, leading to a decrease in N20 radiative efficiency of
almost 13%. When accounting for all radiative changes, the effective radiative efficiency of CH4 has
increased by about 10%, while that of N20 has decreased by almost 8%, relative to AR4.
Separately, the AR6 estimate of lifetime of CH4 decreased by about 2%, and that of N20 by about 4%,
relative to AR4. The changes in the C02 lifetime are more complex, but over 100 years, the effective
lifetime of C02 increased by about 13%. AR6 also included the possibility of accounting for the C02
produced through the oxidation of CH4 of fossil origins in the atmosphere, using an oxidation factor of
0.75 to account for CH4 that does not oxidize to C02 but rather leaves the atmosphere through a
deposition process.135 136 AR6 also accounts for the climate-carbon feedbacks that result from non-C02
greenhouse gases warming the atmosphere and impacting the carbon cycle; in AR4, this effect was only
included for C02.
Including all these scientific updates to lifetimes, atmospheric chemistry interactions, and radiative
efficiency, the AR6 assessment estimates that the 100-year global warming potential (GWP) of CH4 has
increased by almost 9% relative to the estimates from AR4 (from 25 to 27.2), whereas the 100-year GWP
of N20 has decreased by about 8% (from 298 to 273). Between AR4 and AR6 there was also a discussion
134 Radiative efficiency is a measure of a gas' greenhouse gas strength, defined as the change in radiative forcing for
a unit change in the atmospheric concentration of a gas (in W/m2/ppb).
135 While FaIR 1.6.2 reflects the advances in understanding presented in AR6, the CH4 oxidization factor in FaIR 1.6.2
was still set to 0.60 (based on AR5) in the model runs conducted for this report. In corresponding with the FaIR model
developers, they have stated that it will be updated to the AR6 value in the next version of FaIR 2.0.
136 Note that inventories based on using GWPs often use the non-fossil value for all CH4 emissions because in some
cases there is a potential for CO2 double counting: for example, if complete combustion is assumed when calculating
C02 emissions from a natural gas turbine, then the carbon from any methane leakage has already been accounted
for.
Ill
-------
EXTERNAL REVIEW DRAFT
of climate-carbon feedbacks. Including the climate-carbon feedback means taking into account the effect
that a changing climate has on the carbon cycle. AR4 GWPs were calculated with climate-carbon feedbacks
included for C02, but not for non-C02 greenhouse gases. This inconsistent treatment of climate-carbon
feedbacks can lead to underweighting the non-C02 greenhouse gases relative to their actual impacts. The
publication of more studies using climate-carbon feedbacks for all gases, and the determination that a
consistent approach was superior, led AR6 to include the climate-carbon feedbacks for all gases in the
only GWP that was presented.
Another way of considering the impact of different greenhouse gases is to attribute the temperature
changes of the last decade (2010-2019) to historical emissions of each gas. According to the AR6
assessment, historical emissions of carbon dioxide have contributed almost 0.8 degrees of warming to
those temperatures, compared to about half a degree for historical emissions of CH4, and almost one
tenth of a degree for historical emissions of N20. These attributed temperature increases sum to more
than the observed temperature change of almost 1.1 degrees because some of the warming is masked by
various cooling influences, the most important of which is about half a degree of cooling resulting from
historical emissions of sulfur dioxide.
A.2. Consumption Rate of Interest and Integration into Benefit-Cost Analysis
When analyzing policies and programs that result in GHG emission reductions, it is important to account
for the difference between the social and private rate of return on any capital investment affected by the
action. Market distortions, such as taxes on capital income, cause private returns on capital investments
to be different from the social returns. In well-functioning capital markets, arbitrage opportunities will
be dissipated, and the cost of investments will equal the present value of future private returns on those
investments. Therefore, an individual forgoing consumption or investment of equal amounts as the
result of a regulation will face an equal private burden. However, because the social rate of return on
the investment is greater than the private rate of return, the overall social burden will be greater in the
case where investment is displaced. Thus, society is not indifferent between a regulation that displaces
consumption versus investment in equal amounts.
OMB's Circular A-4 points out that "the analytically preferred method of handling temporal differences
between benefits and costs is to adjust all the benefits and costs to reflect their value in equivalent units
of consumption and to discount them at the rate consumers and savers would normally use in discounting
future consumption benefits" (OMB 2003). The damage estimates developed for use in the SC-GHG are
already estimated in consumption-equivalentterms. Therefore, an application of this OMBguidance would
use the consumption discount rate to calculate the SC-GHG, while also developing a more complete
estimate of social costs to account for the difference in private and social rates of return on capital for
any investment displaced as a result of the action being analyzed. This more complete estimate of social
costs could be developed using either the shadow price of capital approach or by estimating costs in a
general equilibrium framework, for example by using a computable general equilibrium model. In both
cases, displaced investment would be converted into a flow of consumption equivalents that could be
discounted at the consumption rate.
112
-------
EXTERNAL REVIEW DRAFT
In cases where the costs are not adjusted to be in consumption-equivalent terms, OMB's Circular A-4
recommends that analysts provide a range of estimates for net benefits based on two approaches. The
first approach is based on using the consumption rate of interest to discount all costs and benefits. This
approach is consistent with the case where costs are primarily borne as reduced consumption. The second
approach, the opportunity cost of capital approach, focuses on the case where the main effect of an action
is to displace or alter the use of capital in the private sector (OMB 2003). When interpreting the
opportunity cost of capital approach from the point of view of whether to invest in a single government
project, it is asking whetherthe benefits from the project would at least match the returns from investing
the same resources in the private sector. Interpreting the approach from the standpoint of a benefit-
cost analysis of a regulation, the approach focuses on adjusting estimates of benefits downward by
discounting at a higher rate to offset additional social costs not reflected in the private value of displaced
investment used to develop the cost estimate (assuming the costs of the regulation are borne upfront).
Harberger (1972) derived a general version of the opportunity cost of capital approach, recognizing that
policies will most likely displace a mix of consumption and investment and therefore, a blended discount
rate would be needed to adjust the benefits to account for the omitted costs. In his partial equilibrium
approach, the blended discount rate is a weighted average of the consumption interest rate and rate of
return on capital, where the weights are the share of a policy's costs borne by consumption versus
investment. This general result has been applied to the general equilibrium context by Sandmo and Dreze
(1971) and Dreze (1974) and can be extended to account for changes in foreign direct investment (CEA
2017). This highlights that using the opportunity cost of capital to discount benefits and costs is, at best,
creating a lower bound on the estimate of net benefits that would only be met in an extreme case where
regulatory costs fully displace investment. If the beneficial impacts of the regulation induce private
investment whose returns have not been quantified and fully converted to consumption equivalents, then
this approach would not even be a lower bound, as the net benefits calculated using the opportunity cost
of capital would be even lower than the theoretically correct lower bound.
An important limitation of the opportunity cost of capital approach is that its correct application depends
heavily on the temporal patterns of the displaced capital returns and future benefits, including the lifetime
of the displaced capital investment versus the lifetime of the benefit stream being valued (Li and Pizer
2021). In fact, using the opportunity cost of capital approach is only an accurate approximation of the
correct shadow price of capital approach if these patterns are exactly the same. Li and Pizer (2021) show
that a rate lower than the rate of return to capital is appropriate when displaced investment is relatively
short-lived compared to the benefits stream and a higher rate is appropriate when displaced investment
is relatively long-lived compared to benefits.
In benefit-cost analysis of policy actions whose benefits and costs occur over a relatively short time frame,
the range of net benefits computed using the two discounting approaches may be relatively narrow. In
this case, there may not be much error in presenting the opportunity cost of capital discounting approach
side-by-side with consumption discounting as an effort to represent an uninformed prior over the share
of regulatory costs that will displace investment and using the potential bounding cases for net benefits.
However, for cases where the costs are borne early in the time horizon and benefits occur for decades
or even centuries, such as with GHG mitigation, the two estimates of net benefits will differ significantly.
Importantly, in this circumstance, the opportunity cost of capital approach will substantially
underestimate net benefits even for the case where the policy fully displaces investment. In this case,
113
-------
EXTERNAL REVIEW DRAFT
there is high risk of uninformative results from an analysis when using this two-discount-rate approach
to provide an uninformed prior over the share of regulatory costs borne by investment. The preferred
approach (OMB 2003, Li and Pizer 2021) is to develop more complete consumption-equivalent measure of
costs and benefits, accounting for any effects on investment either by using a shadow price of capital
approach or a general equilibrium framework, and then discounting those streams at the consumption
rate of interest alone.
The "shadow price of capital" approach, described below, provides a method of ensuring that any
additional social costs of displaced capital are accounted for in an analysis, as has been widely recognized
in the academic literature (Lind 1990; Lyon 1990; Moore et al. 2013; Li and Pizer 2021) and in domestic
and foreign government guidance documents (OMB 1972, 2003; EPA 2010; OECD 2018) as more
appropriate than using the opportunity cost of capital approach. The most straightforward, although
extreme, illustration of this approach is to consider the consumption value of a marginal dollar of
displaced investment that persists forever. A permanent loss of investment is a very strong assumption
because we would expect the displaced investment to be replaced eventually, but it is an instructive
example of the approach. If this dollar had been invested, it would have earned a return on capital, n,
every period into the future. If that yield was returned as consumption (or taxes that ultimately benefit
households), the infinite stream of r* should be discounted at the consumption rate of interest rc. The
present value of this infinite stream is r;/rc.137 Under this strong assumption of a permanent displacement
of capital, the shadow price of capital (SPC) would be calculated as the opportunity cost of capital divided
by the consumption rate of interest. Because rt > rc, the SPC is greater than one, reflecting the additional
cost of the displaced capital. Multiplying any portion of costs (and/or benefits) that affect investment in
this way, and then discounting using the consumption rate of interest would appropriately account for
the displaced investment.
However, rj/rc would only be the correct SPC to use in the extreme case where changes in the productive
capital stock persist in perpetuity. A more realistic version of the SPC accounts for how savings and
depreciation cause the impact of displaced capital to dissipate in the future. In particular, with a savings
(or reinvestment) rate of s from gross income and a depreciation rate of /i, an invested dollar returns (7 —
s)(ri +/0 in consumption in the first period. Each period after that, the amount of investment that
continues to be displaced is determined by the savings rate, assuming a closed economy. However, the
invested capital also declines according to the depreciation rate. This creates a stream of consumption
benefits equal to
137 An infinite stream of return is a type of annuity called a perpetuity. The present value of a perpetuity, r;, that
y . y . y . y.
begins in year 1 and is discounted at a rate of rc is PV = —l-—I * -I i—3 + ••• = —. That is, the present value
(7+rc) (1+rcY (1+rc)d rc
of a perpetuity is the annual return, rit divided by the rate of discount, rc.
t=0
(A 2.7)
114
-------
EXTERNAL REVIEW DRAFT
which is discounted at the consumption discount rate rc. Including constant savings and depreciation rates
yields a shadow price of capital138 equal to
(7 — s)(r; + ju)
SPC = — ^ . (A. 2.2)
rc + ii - s(rt +11)
Equation A.2.2 can be updated to include a capital tax rate that explicitly defines a difference between
Tj and rc, but the result of the analysis would not change if the tax revenue was used to benefit society.139
In the analysis, the portion of costs (and/or benefits) that displace investment would be multiplied by the
SPC to adjust for any missing social impacts and then all costs and benefits would be discounted at the
consumption rate of interest.
Estimates of the closed economy SPC in the academic literature are in the range of 1.1 to 2.2 (Groom et
al. 2005, Boardman et al. 2010, Moore et al. 2013, Li and Pizer 2021). In an open economy model the SPC
may be closer to 1.0 (Lind 1990). Implementing this approach in practice can be challenging because it
requires an assessment of the portion of costs (and/or benefits) that displace investment. However, even
in the absence of information as to the share of costs that displace consumption, multiplying the full cost
estimate by the SPC and discounting all costs and benefits at the consumption rate of interest likely
provides a more informative lower-end bounding case for net benefits than using the opportunity cost of
capital approach under the premise of full displacement.
A.3. Derivations of the SC-GHG Values for use in Analyses
This report presents SC-GHG estimates as certainty-equivalent values that account for the uncertainty (a
range of possible outcomes) in future consumption underlying the RFF-SP probabilistic growth scenarios.
To recover a discounted present value of climate damages from future emissions, analysts consider the
SC-GHG associated with future emissions and then discount that value to the year of their analysis. For
example, an analyst interested in the present value in the year 2022 of changes in future emissions in the
year 2030 would use the 2030 SC-GHG and discount back to recover a present value in the year 2022.
However, there is uncertainty in future consumption such that analysts should account for the range of
138 When including depreciation, f.i, the gross return on a capital stock ko will be (r,+n)ko, where n is the depreciation
rate. With a savings or reinvestment rate of s, a capital stock of ko in period 0 will return (l-s)(r,+n)ko as consumption
and s(r,+n)ko will be saved for reinvestment. In period 1, the capital stock will be the original capital less depreciation,
plus the amount reinvested, ki = {(l-n)k0}+{s(r,+^)ko} = [l+s(r,+n)-n]ko. This will return (l-s)(r,+n)[l+s(r,+n)-jj]ko as
consumption in period 1 and s(n+n)[l+s(r,+n)-n]ko will be reinvested. The capital stock in period 2 will be = {(1-
ii)[l +s(r,+[j.)~lJ.]ko}+{s(r,+[j.)[l+s(r,+[j.)-[!]ko} = [l+s(n+n)-n]2ko, which will return (l-s)(r,+[j.)[l+s(r,+[j.)-[!]2ko as
consumption. This creates an infinite consumption stream of C = (l-s)(n+n)ko + (l-s)(r,+n)[l+s(r,+n)~n]ko + (1-
s)(r,+n)[l+s(n+n)-n]2ko + ... This is a perpetuity of [(l-s)(r,+iJ.)ko] with a growth rate of [s(r,+n)-n], and should be
discounted at the consumption rate of discount rc. The present value of perpetuity A, growing at a rate of g, and
A A (1 "I- q ") A (1 A
discounted at rate r is PV = 1- -—-y + ——-r- + ••• = . So, the present value of the perpetuity described
(7+r) (7+r)2 (7+r)3 (r-g) K
above would be PV = = (1~s)(r;^\ k0.
(rc-|s(r;+/i)-/ij) rc+[A.—s(ri+[A.)
139 If a portion of the tax revenues affect investments, then it requires an analogous adjustment to account for the
fact that it creates a current period consumption value greater than one according to the "marginal value of public
funds," vg. In this case, the numerator in the SPC equation would be equal to (7 — s)(rj + //) + (vG — 1)Trit where
t is the tax rate on capital (Li and Pizer 2021).
115
-------
EXTERNAL REVIEW DRAFT
possible outcomes. This is because risk-averse agents value the costs of future emissions differently than
risk-neutral agents by accounting for the range of uncertain outcomes. There are several ways to account
for this uncertainty. The approach taken in this report provides certainty-equivalent SC-GHG values that
can be easily used by analysts with a conventional discounting approach, as described in Section 4.2. This
section describes the equations used to recover those certainty-equivalent SC-GHG estimates for an
emissions year t, denoted as scghgT.
To begin with a motivating example, imagine a hypothetical regulation that reduces x tons of greenhouse
gas emissions in year t, and the regulation will be in place for the years 2040 through 2050. An analyst
wants to calculate the present value pv of the regulation's benefits from future reductions in greenhouse
gas emissions in the year of analysis j, where j is some year between now and 2040. The analyst would
use the SC-GHG estimates found in this report for each of the years from 2040 through 2050, each
denoted as scghgT. In addition, the analyst would need the certainty-equivalent discount rate path
specific to the year of analysis, ft, from year j to year x (see Figure 2.4.1 for one example path). The analyst
then calculates the present value of the regulation's benefits as
2050
pvj = ^ xT ¦ scghgT ¦ e~rt^T~^ (A 3.7)
t=2040
The scghgT values presented in this report yield the present value when discounted using the certainty-
equivalent discount factor This discount factor was written as ST in Section 4.2 but is defined in
more detail below.
The remainder of this section describes the derivation of the certainty-equivalent SC-GHG scghgT. The
certainty-equivalent discount factor for the Ramsey framework is
e-^t = E [e-Zl=o(p+vgs)i (A 32)
where ft is the certainty-equivalent discount rate. This is the single, time-averaged discount rate that
produces the same discount factor over a specific time horizon as the distribution of uncertain discount
rates. This certainty-equivalent discount rate is defined as
ft = p — E = p + ^jE In
A
\c0)
(A 3.3)
and
-t-
e
p+ii)E[ln&]
= e~tp ¦ E
f(Tl
= E
Ac0/ .
1+p
Ct\~V
Co
(A 3.4)
Here, as described in Section 3.4, rt is the consumption discount rate in year t, p is the pure rate of time
preference and r| is the elasticity of marginal utility with respect to consumption. ct and gt are the
representative agent's year t consumption and consumption growth rate, respectively. Importantly, ct is
consumption net of climate change damages. Also, p = ep — 7 is the discrete annual pure rate of time
preference.
116
-------
EXTERNAL REVIEW DRAFT
Consider a stream of marginal damages mdt from a single emissions year t. The scghg0 is the present
value of the social cost of GHG emissions for year t = 0 and is given by
scahgo = ^E
t=0
7
(7 + pYKco
—) V mdt
(A 3.5)
where mdt is the marginal damage in year t from a pulse of emissions in year t. Because mdt is the
marginal damage from a single emissions year t, mdt = Ofor t=0 to x-1. The scghgg is the SC-GHG in the
present year t = 0. This is not equal to scghgT, which is the SC-GHG in year t.
The present value scghgg for any emission year r should also be equal to the scghgT discounted back to
current period t = 0 using the certainty-equivalent discount rate
scghgo = scghgT ¦ E [777^7 (^)
(A. 3.6)
So
scghgT
scghg0
7
t=o
7
(7+ pyycp
(i)"
mdt
7
(D"
(A. 3.7)
L(7+p)TVc0; J ^ L(7+ p)TVc0
Assuming that consumption is certain in the present year (t=0), Cg can be canceled
7
V*
scghgT = ^ —
(7+ P)]
¦(ct) vmdt
0 E
7
L(7+ P)
tCct)""
(A. 3.8)
Simplifying this expression yields
scghgT
7
EKcJ-*]
t=o
7
0+ P)
t—(ct) vmdt
(A. 3.9)
Note that equation (A.3.9) is not the same as simply discounting the marginal damages back to the year
of emissions, which would be the expected value
scghgT = ^ j
t=T
7
(7+ P)
t-T
CA~Vmdt
(A. 3.10)
The scghgT estimates based on the GIVE model (Rennert et al. 2022b) and the Meta-Analysis (Howard
and Sterner 2017) are directly estimated using equation (A.3.9). The scghgT estimates under the DSCIM
damage module, however, are adjusted post-estimation to exactly equal equation (A.3.9). The remainder
117
-------
EXTERNAL REVIEW DRAFT
of this section describes this adjustment alongside its analogue for GIVE. Consider trial i, year t, emissions
year t, net consumption per capita Cjt, and marginal damages A trial i is a unique socioeconomic
pathway and FalRl.6.2 climate scenario pairing. For each trial GIVE estimates
n + Lt-r mdit
scghgiT = , (A 3.7 7)
eIct \
and the scghgT from equation (A.3.9) results from applying the expectation operator to equation (A.3.11).
In contrast to equation (A.3.11), DSCIM estimates
2300
(C- 7
' C(T \ '
scghg'tT = ^ (7 + g)t-T (A. 3.12)
. -it' (.1 + P)
t=T
Equations (A.3.11) and (A.3.12) can be equated by
7
scghgiT = scghgiT (A 3.13)
CiTEiCT \
The first expression on the right-hand side of Equation (A.3.13) is the adjustment factor that is used to
convert the values provided by DSCIM for use in the report. This adjustment equation is trial-specific, so
the values presented in this report are the means across trials (i.e., applying expectation operator to
equation (A.3.11)).
The full derivation of a certainty-equivalent discount rate path involves damage-module-specific net
consumption paths, damage-module-specific SC-GHG estimates, and a unique certainty-equivalent rate
path for each analysis year. However, as noted in Section 4.2, the error associated with using a constant
discount rate rather than the certainty-equivalent rate path (i.e., E \-—1-—r(—) l in equation A.3.6) to
L(7 + py \c0J J
calculate the present value of a future stream of monetized climate benefits is small for analyses with
moderate time frames (e.g., 30 years or less). In other words, for analyses with a moderate time frame,
the present value of the regulation's benefits can be calculated as
2050
pvj = ^ xT ¦ scghgT ¦ e~r^T~J'\ (A 3.74)
t=2040
where f is simply the near-term (2.5%, 2%, and 1.5%) corresponding to the SC-GHG value used. Figure
A.3.1 provides an illustration of the amount that climate benefits from reductions in future emissions will
be underestimated by using a constant discount rate relative to the more complicated certainty-
equivalent rate path.
118
-------
EXTERNAL REVIEW DRAFT
Figure A.3.1 The Difference Between using a Certainty-Equivalent Rate and Constant Discount Rate to
Discount Climate Benefits from Future Reductions in GHG Emissions Back to the Year of the Analysis
0%
-2%
{/)
+->
M=
CD
£ -4%
CO
c
-------
EXTERNAL REVIEW DRAFT
A.4. Annual Unrounded SC-CO2, SC-CH4, and SC-N2O Values, 2020-2080
Table 4.2.1: Unrounded SC-C02, SC-CH4/ andSC-N20 Values, 2020-2080
SC-GHG and Near-term Ramsey Discount Rate
SC-CO2
SC-CH4
SC-N2O
(2020 dollars per metric ton ofC02)
(2020 dollars per metric ton ofCH4)
(2020 dollars per metric ton of N20)
Emission
Year
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2.5%
2.0%
1.5%
2020
117
193
337
1,257
1,648
2,305
35,232
54,139
87,284
2021
119
197
341
1,324
1,723
2,391
36,180
55,364
88,869
2022
122
200
346
1,390
1,799
2,478
37,128
56,590
90,454
2023
125
204
351
1,457
1,874
2,564
38,076
57,816
92,040
2024
128
208
356
1,524
1,950
2,650
39,024
59,041
93,625
2025
130
212
360
1,590
2,025
2,737
39,972
60,267
95,210
2026
133
215
365
1,657
2,101
2,823
40,920
61,492
96,796
2027
136
219
370
1,724
2,176
2,910
41,868
62,718
98,381
2028
139
223
375
1,791
2,252
2,996
42,816
63,944
99,966
2029
141
226
380
1,857
2,327
3,083
43,764
65,169
101,552
2030
144
230
384
1,924
2,403
3,169
44,712
66,395
103,137
2031
147
234
389
2,002
2,490
3,270
45,693
67,645
104,727
2032
150
237
394
2,080
2,578
3,371
46,674
68,895
106,316
2033
153
241
398
2,157
2,666
3,471
47,655
70,145
107,906
2034
155
245
403
2,235
2,754
3,572
48,636
71,394
109,495
2035
158
248
408
2,313
2,842
3,673
49,617
72,644
111,085
2036
161
252
412
2,391
2,929
3,774
50,598
73,894
112,674
2037
164
256
417
2,468
3,017
3,875
51,578
75,144
114,264
2038
167
259
422
2,546
3,105
3,975
52,559
76,394
115,853
2039
170
263
426
2,624
3,193
4,076
53,540
77,644
117,443
2040
173
267
431
2,702
3,280
4,177
54,521
78,894
119,032
2041
176
271
436
2,786
3,375
4,285
55,632
80,304
120,809
2042
179
275
441
2,871
3,471
4,394
56,744
81,714
122,586
2043
182
279
446
2,955
3,566
4,502
57,855
83,124
124,362
2044
186
283
451
3,040
3,661
4,610
58,966
84,535
126,139
2045
189
287
456
3,124
3,756
4,718
60,078
85,945
127,916
2046
192
291
462
3,209
3,851
4,827
61,189
87,355
129,693
2047
195
296
467
3,293
3,946
4,935
62,301
88,765
131,469
2048
199
300
472
3,378
4,041
5,043
63,412
90,176
133,246
2049
202
304
477
3,462
4,136
5,151
64,523
91,586
135,023
2050
205
308
482
3,547
4,231
5,260
65,635
92,996
136,799
120
-------
EXTERNAL REVIEW DRAFT
Table 4.2.2: Unrounded SC-C02, SC-CH4, andSC-N20 Values, 2020-2080 (continued...)
SC-GHG and Near-term Ramsey Discount Rate
SC-CO2 SC-CH4 SC-N2O
(2020 dollars per metric ton of C02) (2020 dollars per metric ton of CH4) (2020 dollars per metric ton of N2Q)
2.5% 2.0% 1.5% 2.5% 2.0% 1.5% 2.5% 2.0% 1.5%
Year
2050
205
308
482
3,547
4,231
5,260
65,635
92,996
136,799
2051
208
312
487
3,624
4,320
5,363
66,673
94,319
138,479
2052
211
315
491
3,701
4,409
5,466
67,712
95,642
140,158
2053
214
319
496
3,779
4,497
5,569
68,750
96,965
141,838
2054
217
323
500
3,856
4,586
5,672
69,789
98,288
143,517
2055
220
326
505
3,933
4,675
5,774
70,827
99,612
145,196
2056
222
330
510
4,011
4,763
5,877
71,866
100,935
146,876
2057
225
334
514
4,088
4,852
5,980
72,904
102,258
148,555
2058
228
338
519
4,165
4,941
6,083
73,943
103,581
150,235
2059
231
341
523
4,243
5,029
6,186
74,981
104,904
151,914
2060
234
345
528
4,320
5,118
6,289
76,020
106,227
153,594
2061
236
348
532
4,389
5,199
6,385
76,920
107,385
155,085
2062
239
351
535
4,458
5,280
6,480
77,820
108,542
156,576
2063
241
354
539
4,527
5,361
6,576
78,720
109,700
158,066
2064
244
357
543
4,596
5,442
6,671
79,620
110,857
159,557
2065
246
360
547
4,666
5,523
6,767
80,520
112,015
161,048
2066
248
363
550
4,735
5,604
6,862
81,419
113,172
162,539
2067
251
366
554
4,804
5,685
6,958
82,319
114,330
164,030
2068
253
369
558
4,873
5,765
7,053
83,219
115,487
165,521
2069
256
372
562
4,942
5,846
7,149
84,119
116,645
167,012
2070
258
375
565
5,011
5,927
7,244
85,019
117,802
168,503
2071
261
378
569
5,085
6,013
7,344
86,012
119,027
170,013
2072
263
382
573
5,160
6,099
7,444
87,006
120,252
171,523
2073
266
385
576
5,234
6,184
7,545
87,999
121,477
173,033
2074
269
388
580
5,309
6,270
7,645
88,992
122,702
174,543
2075
271
391
583
5,383
6,355
7,745
89,985
123,926
176,053
2076
274
394
587
5,458
6,441
7,845
90,978
125,151
177,563
2077
276
398
591
5,532
6,527
7,946
91,971
126,376
179,073
2078
279
401
594
5,607
6,612
8,046
92,964
127,601
180,582
2079
282
404
598
5,681
6,698
8,146
93,958
128,826
182,092
2080
284
407
601
5,756
6,783
8,246
94,951
130,050
183,602
121
-------
EXTERNAL REVIEW DRAFT
A.5. Additional Figures, Tables, and Results
Figure A.5.1: Net Annual Global Emissions of Methane (CH4) under the RFF-SPs and the SSPs, 1900-2300
g 2000
1000
-------
EXTERNAL REVIEW DRAFT
Figure A. 5.3: Global Atmospheric Concentrations of Methane (CH4), 1900-2300
6000
5000
_Q
Q_
Q_
4000
£ 3000
o
u
X
u
15 2000
-Q
_o
U
1000
1900 200 0 2100 2200 2300
Year
Figure A.5.4: Global Atmospheric Concentrations of Nitrous Oxide (N20), 2020-2300
1000
0
1900 200 0 2100 2200 2300
Year
Historical and future concentrations of methane (CH4, top) and nitrous oxide (N20, bottom) are based on the range of emissions
from the sampled RFF-SP scenarios used as inputs into FaIR 1.6.2. FaIR 1.6.2 is run with the full, AR6 calibrated (constrained)
uncertainty distribution. Therefore, the uncertainty ranges in this figure represent both emissions and physical carbon cycle
uncertainty. Mean (solid) and median (dashed) lines along with 5th to 95th (dark) and 1st to 99th (light) percentile ranges.
123
-------
EXTERNAL REVIEW DRAFT
Figure A.5.5: Global Temperature Anomaly from a Pulse of Methane (lMtCH4) Emissions, 2020-2300
7e-05
o
8 6e-05
u 5e-05
p 4e-05
| 3e-05
C
<
a;
3 2e-05
ra
i_
Q_
I 1e-05
0e+00
0.00025
0.00020
0.00015
0.00010
0.00005
0.00000
~2020 2100 2200 2300
Year
— FaIR 1.6.2 — HECTOR 2.5 — MAGICC 7.5.3
The global temperature response resulting from a pulse of emissions of CH4 (top) and N20 (bottom) in 2030 as projected by
FalRl.6.2, Hector 2.5, and MAGICC 7.5.3. This represents the difference between a reference scenario (using SSP2-RCP4.5for the
figure) and the same scenario including the pulse of emissions. The emission pulse size is 1 GtCfor carbon dioxide. Mean (solid)
and median (dashed) lines are shown along with the 5th to 95th (dark shade) and 1st to 99th (light shade) percentile ranges.
o
m
O
E
o
c
<
-------
EXTERNAL REVIEW DRAFT
Figure A.5.7: Dynamic temperature response of 256 climate science models (the CMIP5 ensemble) and
seven lAMs
1 Deciles CMIP5 combinations
0
PAGE
Best fit CMIP5 ensemble
+
GHKT14
o DICE16
GL18
DICE13
A
LR17
V FUND
0 20 40 60 80 100 120 140 160 180 200
years
Source: Dietz et al. (2021). The figure displays the dynamic temperature response of 256 climate science models (the CMIP5
ensemble) and seven lAMs to an instantaneous 100 GtC emission impulse against a constant background atmospheric C02
concentration of389 ppm. The temperature response of the lAMs is much slower than the climate science models, except Golosov
et al. (2014). After 200 years, the temperature response of the lAMs is often well outside the range of the climate science models.
The CMIP5 model responses are emulated/fitted by combining the Joos et al. (2013) carbon cycle model and the Geoffroy et al.
(2013) warming model.
125
-------
EXTERNAL REVIEW DRAFT
Figure A.5.8: Distribution ofSC-CH4 Estimates for 2030, by Damage Module and Discount Rate
$1.100
2.5%
2,300
$2,300
-he:
$1.60C
$2,800
$2,800
DSCIM
GIVE
Meta-Analysis
£2.400
1.5%
$3,500
$0
$2,000 $4,000 $6,000 $8,000 $10,000
SC-CH4 for 2030 emissions (2020$ per metric ton of CH4)
$12,000
Figure A.5.9: Distribution ofSC-N20 Estimates for 2030, by Damage Module and Discount Rate
$40.000
-rf
2.5%
$47,000
$46,000
$63.000
~
$69,000
~
DSCIM
GIVE
Meta-Analysis
$98.000
1.5%
$100,000
$110,000
$0
$100,000 $200,000 $300,000
SC-N2O for 2030 emissions (2020$ per metric ton of N2O)
$400,000
Boxes span the inner quartile range (25th to 75th percentiles), whiskers extend to the 5th (left) and the 95th (right) quantiles. The
vertical lines inside of the boxes mark the median of each distribution, and the points inside of the boxes and dollar estimates on
top of the boxes mark the simple mean (average).
126
-------
EXTERNAL REVIEW DRAFT
A.6. Valuation Methodologies to Use in Estimating the Social Cost of GHGs
The EPA will continue to review developments in the literature, including new and robust methodologies
for estimating the magnitude of the various direct and indirect damages from climate impacts. EPA will
also continue to assess whether there are other parts of this literature or other methodologies to evaluate
for potential inclusion in SC-GHG estimation.
Both DSCIM and the GIVE model incorporate sector-specific damage functions published in the peer-
reviewed literature. One advantage of the modular approach used by these models is that new or
alternative damage functions can be incorporated in a relatively straightforward way, while maintaining
the state-of-the-science modules dealing with socioeconomic scenarios, emission trajectories,
discounting, and climate modeling used in this report.
As explained in Section 2.3, the damage module component of SC-GHG estimation translates changes in
temperature and other physical impacts of climate change into monetized estimates of net economic
damages based on the willingness to pay of individuals to avoid those damages. The developers of the
damage functions used in this report applied valuation methods that are consistent with the theoretical
underpinning of EPA's benefit-cost analysis (BCA) - the Kaldor-Hicks criterion.140 For example, in DSCIM
and GIVE, changes in agricultural output due to climate change are valued using expected market prices
for key agricultural commodities. Use of prices to value commodities traded in markets is generally
consistent with the Kaldor-Hicks criterion, sometimes called an economic efficiency test. For damage
categories that involve non-market impacts (commodities or services not traded in the market, like
changes in mortality risks) there is no readily observed price information and there are challenges in
capturing the value of something as precious as changes in life expectancy. However, economists have
developed a robust literature to infer values for these non-market commodities using methods that are
consistent with the economic efficiency test. Because of data limitations and other constraints to
performing original research to develop location- and context- specific values to assign to each non-
market impact, analysts regularly need to draw upon existing value estimates for use in benefits analysis
140 The Pareto criterion maintains that if an economic change does not harm any individual and makes at least one
individual better off, there is an increase in social welfare. The Kaldor-Hicks criterion captures the intuition of the
Pareto criterion, but allows for the identification of potential improvements in social welfare under conditions where
some may be made worse off by the economic change. For a potential increase in social welfare, there needs to be
a "potential" Pareto improvement, which occurs when those who gain from the economic change would be willing
to fully compensate those made worse off from the economic change. From this criterion, the rules of BCA as an
economic efficiency test follow, including the use of the consumer sovereignty principle whereby BCA must value
benefits and costs based on individuals' willingness to pay. If the impacts to individuals are measured using a value
other than their willingness to pay, the results of the BCA will be unable to identify potential Pareto improvements
under the Kaldor-Hick criterion and their interpretation may be unclear. The discipline of the private market to
allocate resources cannot work for pollution, so the BCA helps provide this information as one input, amongst many,
in the decision-making process. As in a private market, the price in the simulated market test should equal the
willingness to pay of individuals on the margin, as any other valuation would cause the test to fail in answering its
question. See EPA (2010) for more discussion.
127
-------
EXTERNAL REVIEW DRAFT
through "benefits transfer."141 The benefits transfer methods used by the developers of the DSCIM and
GIVE damage functions used in this report are also consistent with the economic efficiency test.
The challenge of valuing climate-related mortality risks provides an illuminating application of these
methods. As shown in Section 3.1, net costs of expected premature mortality associated with climate
change driven changes in hot- and cold weather comprise the largest share of the DSCIM and GIVE based
SC-GHG estimates presented in this report.142 It is worth noting that valuing premature mortality risks in
EPA BCAs is a routine occurrence. Particulate matter, ozone, lead, and many other environmental
contaminants can increase mortality risks through various modes of action including, increased
cardiovascular disease, cancer, and respiratory disease. To value changes in these mortality risks, EPA uses
published research that estimates individuals' willingness to pay to reduce mortality risks in their own
lives - a number that is inaptly termed the "Value of Statistical Life" (VSL)143 - and then transfers these
willingness to pay (WTP) estimates to the risk reductions expected from EPA policy options.144,145
EPA's benefit transfer also recognizes that as per capita income increases, willingness to pay for mortality
risk reductions also increases. This parallels the fact that as their income increases individuals are willing
to pay more for most goods and services.146 EPA increases the willingness to pay estimate over time to
reflect projected per capita income growth (i.e., by applying a positive income elasticity) as a way to
capture that the wealthier we are, the greater our willingness to pay to avoid mortality risks consistent
with the empirical evidence. For example, applying an income elasticity of one implies that for every one
percent increase in per capita income, the value of mortality risk reductions increases by one percent,
such that the willingness to pay for mortality risk reductions remains a constant share of people's income.
EPA's VSL methodology is peer reviewed by its Science Advisory Board (SAB). EPA periodically engages in
a consultation with the SAB on the appropriate range of income elasticities.
In estimating the SC-GHG, the question becomes what VSL to use to monetize expected mortality risk
reductions occurring in other countries. Given the small number of high-quality VSL studies in many
countries, the vast majority of countries do not have their own official recommended VSL estimates or
141 Benefits transfer is the process of applying values estimated in previous studies to a new context. See EPA (2010)
for an overview of current EPA guidance on best practices in benefits transfer.
142 Mortality risk changes are also partially captured in the coastal damage category in each model. See Section 2.3
for more discussion.
143 As noted by the SAB, "the conventional term used to describe the value of risk reduction (the "value of a statistical
life," or VSL) is easily misinterpreted, leading to confusion about key concepts" (EPA 2011). As explained in OMB
Circular A-4 the "phrase can be misleading because it suggests erroneously that the monetization exercise tries to
place a "value" on individual lives"; these terms refer to the measurement of willingness to pay for reductions
in only small risks of premature death. They have no application to an identifiable individual or to very large
reductions in individual risks. They do not suggest that any individual's life can be expressed in monetary terms. Their
sole purpose is to help describe better the likely benefits of a regulatory action" (OMB 2003). Put another way, the
VSL "represents the rate at which an individual views a change in the money he or she has available for spending as
equivalent to a small change in his or her own mortality risk within a specific time period, such as one year" (Robinson
et al. 2019b).
144 For more details on the derivation of EPA's values for mortality risk reductions, see EPA's Guidelines for Preparing
Economic Analyses (2010), p. 7-8. https://www.epa.gov/sites/default/files/2017-09/documents/ee-0568-07.pdf
145 A willingness to pay to reduce mortality risk is a ratio, where the numerator reflects the marginal disutility of
(usually small) increases in probability of experiencing premature mortality, usually within the next year, and the
denominator is the marginal utility associated with additional income/consumption.
146 In economics, goods for which individuals increase their demand as their income rises, signifying an increased
willingness to pay, are called normal goods.
128
-------
EXTERNAL REVIEW DRAFT
estimates from the empirical literature that can be readily adopted (Robinson et al. 2019a). Therefore,
analysts must rely on benefits transfer techniques to develop VSL estimates for other countries that are
extrapolated from existing estimates in the U.S. or other countries with robust empirical estimates.
With respect to this report, both the GIVE and DSCIM based damage modules explicitly model changes in
the risk of premature mortality due to GHG emissions driven climate change and monetize these climate-
related mortality risks consistent with the economic efficiency paradigm. Specifically, as described in
Section 2.3, projected changes in premature mortality in the U.S. are monetized using the same value of
mortality risk reduction as in the EPA's regulatory analyses ($4.8 million in 1990 (1990USD)) and adjusted
for income growth and inflation following current EPA guidelines and practice (EPA 2010) and consistent
with SAB advice (see e.g., EPA 2011, OMB 2003), resulting in a 2020 value of $10.05 million (2020USD).
Valuation of mortality risk changes outside the U.S. is based on an extrapolation of the EPA value that
equalizes willingness-to-pay as a percentage of per capita income across all countries (i.e., using an
assumed income elasticity of 1). The use of a benefits transfer approach based on a positive income
elasticity is consistent with the approach used in the default version of the damage functions and
published studies used in this report (e.g., Rennert et al. 2022b, Carleton et al. 2022, and Diaz 2016), other
academic literature (e.g., Hasegawa et al. 2016, Springmann et al. 2016, Sarofim et al. 2017, Markandya
et al. 2018, and the Lancet Commission on pollution and health (Landrigan et al. 2018)), advice given to
the IWG by experts at the 2011 U.S. EPA and U.S. DOE Workshop on Improving the Assessment and
Valuation of Climate Change Impacts for Policy and Regulatory Analysis (ICF International 2011), and other
prominent domestic and international guidance documents that speak to international mortality risk
reduction valuation. See, for example, the 2019 Gates Foundation Reference Case Guidelines for Benefit-
Cost Analysis in Global Health and Development Guidelines (Robinson et al. 2019a) and literature cited
therein (e.g., Robinson et al. 2018, 2019b, OECD 2016, World Bank and IHME 2016, Viscusi and Masterman
2017a, 2017b, Masterman and Viscusi 2018), and the U.S. Millennium Challenge Corporation guidance for
conducting benefit-cost analysis (MCC 2021). Many international organizations also regularly use country-
level measures of the willingness-to-pay for mortality risk reductions based on a positive income elasticity
in cross country analyses (see, for example, Tan-Soo 2021, Roy and Braathen 2017, Roy 2016,
Laxminarayan et al. 2007).
Given that the methodology in this report is grounded in a willingness to pay concept and the empirical
evidence shows a positive relationship between income and the willingness to pay for mortality risk
reductions, the willingness to pay for mortality risk reductions in countries with lower average incomes is
less than the willingness to pay for mortality risk reductions in higher income countries. It is important to
stress that this metric does not reflect the "value" that this approach places on mortality risks in different
parts of the world. Rather, it reflects an estimate of the willingness to pay for mortality risk reductions by
the average resident of countries or regions conditional on their income. EPA's Science Advisory Board,
while reviewing our methodology to assign monetized estimates to mortality risk reductions also
recognized this challenge:
"While it is clear from economic theory that individual WTP may vary with individual and risk
characteristics, the SAB acknowledges that the objectives, methods, and principles underlying benefit cost
analysis and particularly the values of mortality risk reductions and other non-market goods are often
misunderstood or rejected as inappropriate by many participants and commentators on the policymaking
process. In the past, for example, the Agency was criticized for considering VRRs [VSL] that differ by
129
-------
EXTERNAL REVIEW DRAFT
individuals' age. However, as acknowledged in the White Paper, values for health risk reductions are not
"one size fits all." Applying a willingness to pay value to a targeted population (such as low income or
elderly) that exceeds that group's willingness to pay for reduced risk could result in decisions that
ultimately reduce the well-being of the targeted group. The proposed change of terminology and
application of VRRs [VSL] that differ with individual and risk characteristics provide an opportunity for
constructive engagement with the public and other interested parties concerning these topics."147 (EPA
2011).
It is important to note that EPA's BCAs, based on the economic efficiency criterion, is one of several
economic analyses done to inform decision making and the public. Notably, distributional considerations
are also paramount. In general, when a BCA is undertaken, EPA also conducts an environmental justice
analysis, examining the incidence of environmental impacts both in the baseline and those that would
result from the policy options under review.148 This is in addition to economic impact analyses that are
conducted by EPA to examines how different populations are affected by other expected outcomes of the
policy options.
There is also a separate literature that argues that equity and other concerns should be addressed directly
throughout all elements of a BCA (e.g., Scitovsky 1951, Lutz 1995, Farrow 1998, Persky 2001, Little 2002).
This issue comes up with regard to climate change, since the impacts of climate change are not
manifesting uniformly across space and populations, as highlighted in Section 3.2, with some of the most
vulnerable populations living in locations that will experience some of the most severe effects. These
facets of climate change have led some analysts (e.g., Azar and Sterner 1996; Fankhauser et al. 1997; Azar
1999; Anthoff et al. 2009; Anthoff and Tol 2010; Dennig et al. 2015, Anthoff and Emmerling 2019) to
employ "equity weighting" to incorporate distributional equity objectives into estimates of the SC-GHG.
As noted by Anthoff and Emmerling (2019), "[ejxisting equity weighting studies assume a social welfare
147 In that same review, the SAB opined more specifically on whether EPA should use a country-wide average VSL or
more granular VSL estimates. While this SAB review was addressing how mortality risks for domestic EPA regulations
should be valued, the insight is easily extended to how the mortality risks in other countries are valued in this report.
"Recognizing that VRR [VSL] is a metric that can vary with both individual and risk characteristics, the conceptually
appropriate method to estimate the benefits to the U.S. population of a change in mortality risk that results from
environmental policy is to estimate the risk changes faced by each individual over time, value these changes using
the appropriate individual VRRs [VSLs], and sum the results over the population. In contrast, an alternative "short-
cut" approach is conventionally applied. The short-cut approach is to multiply the number of people in the
population by the population-mean risk reduction (yielding the number of "lives saved") and multiply that by the
population-mean VRR [VSL], The short-cut approach yields an approximation to the conceptually appropriate
method. It requires information on only the average VRR [VSL] and risk reduction, not on how VRR [VSL] and risk
reduction vary across individuals. The approximation is exact when any of three conditions hold: (a) all individuals
face the same risk reduction; (b) all individuals have the same VRR; or (c) individual risk reductions and VRRs
[VSLs]are uncorrelated in the population. If none of these conditions holds, the short-cut approach introduces bias
as a result of "premature aggregation" (Cameron 2010, Hammitt and Treich 2007)" EPA (2011).
148 EPA has detailed technical guidance on conducting environmental justice analyses. See Technical Guidance for
Assessing Environmental Justice in Regulatory Analysis, EPA 2015.
httpsi//www. epa.gov/environmentaliustice/technical-guidance-assessing-environmental-justice-regulatory-
analysis
130
-------
EXTERNAL REVIEW DRAFT
function (SWF) that exhibits inequality aversion over per capita consumption levels." As defined by EPA's
SAB "[a] social welfare function essentially involves two stages. In the first stage, each group has its own
definition of welfare, which is impacted by the various effects set out in this chapter. In the second stage,
the groups are weighted to account for distributional concerns" (EPA 2021f). The argument for equity
weighting in this strand of literature is "that a given (say one dollar) cost which affects a poor person (in
a poor country) should be valued as a higher welfare cost than an equivalent cost affecting an average
[high income country] citizen" to reflect a decreasing marginal utility of income (Azar and Sterner 1996).
The degree to which the valuations differ across those individuals will, in part, be dependent upon the
degree of society's intra-temporal inequality aversion specified within the SWF.
In place of directly incorporating distributional equity objectives through the specification of a SWF, a
couple of studies have explored the impact of alternative VSL assumptions within the analysis of mortality
impacts of climate change. Bressler (2021), in an effort to reflect distributional concerns, considered the
use of a constant VSL across all countries in place of an income adjusted VSL designed to reflect willingness
to pay. This approach weights the value of mortality risk changes to residents of lower income countries
such that it is higher than their willingness to pay and weights mortality risk changes to higher income
countries such that they are valued less than their willingness to pay. Carleton et al. (2022) included an
empirical exploration in sensitivity analyses of how climate-related mortality damages change under a
variety of valuations. They found net damages from climate change mortality risk changes of $15-$65 per
ton C02 when using a WTP-based VSL (similar to the approach used in this report) and damages of $46-
$144 per ton C02 when using a global average VSL, where the range is across the socioeconomic-emissions
scenario modeled.149
While EPA will continue to assess the broader literature on BCA, social welfare, and equity as it seeks to
apply the best available science in its analyses, this report develops SC-GHG estimates that are consistent
with the Kaldor-Hicks criterion that underlies all the other elements of the EPA's BCAs. In addition, this
approach is consistent with the benefits transfer approaches used in the default versions of the damage
functions and published studies used in this report. This approach also ensures that U.S. mortality risks
from climate impacts are valued consistently with how EPA values U.S. mortality risks from other causes.
In addition to conducting a Kaldor-Hicks based BCA, EPA has and will continueto conduct detailed analyses
of environmental justice concerns of climate change in its rulemakings as required and appropriate150 and
the distributional outcomes of climate change in detailed quantitative analyses,151 so as to ensure that
decision-makers and the public have robust information as to the damages of climate change and their
distributional effects.
149 These values were calculated using a constant 2% discount rate and only reflect damages from net changes in
mortality risks from climate change using a different scenarios and climate modeling than was applied in this report.
150 For example, https://www.epa.gov/regulations-emissions-vehicles-and-engines/final-rule-revise-existing-
national-ghg-emissions.
151 For example, 2021 Climate Change and Social Vulnerability report (EPA 2021e).
131
------- |