Estimates of The Social Cost of Carbon: Concepts and Results From The DICE-2013R Model and Alternative Approaches

Estimates of the Social Cost of Carbon:
Concepts and Results from the DICE-2013R

Model and Alternative Approaches
William Nordhaus
Abstract: The social cost of carbon (SCC) is an important concept for understanding and implementing climate change policies. This term represents the economic
cost caused by an additional ton of carbon dioxide emissions (or more succinctly
carbon) or its equivalent. The present study describes the development of the concept, provides examples of its use in current US regulator policies, examines its analytical background, and estimates the SCC using an updated integrated assessment
model, the DICE-2013R model. The study estimates that the SCC is $18.6 per ton
of CO2 in 2005 US dollars and international prices for the current period (2015).
For the central case, the real SCC grows at 3% per year over the period to 2050.
The major open issues concerning the SCC continue to be the appropriate discount
rate, the potential for catastrophic damages, the impact of incomplete harmonization
of abatement policies, and the effects of distortionary taxes.
JEL Codes: Q5, H23, Q54, H4, Q58
Keywords: Climate change, CO2, Externality, Social cost of carbon
A NEW A ND IM PO R T ANT concept that has taken center stage in economic and
policy discussions about global warming is the social cost of carbon, or SCC. This
term designates the economic cost caused by an additional ton of carbon dioxide
emissions (or more succinctly carbon) or its equivalent. In a more precise denition,
it is the change in the discounted value of the utility of consumption per unit of additional emissions, denominated in terms of current consumption. In the language of
William D. Nordhaus is at Yale University (e-mail: william.nordhaus@yale.edu). He is grateful to the editor, Lint Barrage, Elizabeth Kopits, Zhimin Li, Al McGartland, John Weyant,
and anonymous referees for useful comments. The research reported here was supported by
the Glaser Foundation, the US National Science Foundation, and the US Department of
Energy. The author declares that he has no relevant and material nancial conicts with the
research described in this article.
Received October 16, 2013; Accepted February 17, 2014; Published online June 5, 2014.
JAERE, volume 1, numbers 1/2. 2014 by The Association of Environmental and Resource Economists.
All rights reserved.
2333-5955/2014/0101-0010$10.00
http://dx.doi.org/10.1086/676035
273
This content downloaded from 036.083.077.113 on September 09, 2016 09:36:49 AM

All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
274
Journal of the Association of Environmental and Resource Economists
March/June 2014
mathematical programming, the SCC is the shadow price of carbon emissions along
a reference path of output, emissions, and climate change.
We can illustrate the concept in gure 1. This shows a base path of greenhouse
gas (GHG) emissions along with a base path of a comprehensive measure of economic welfare, such as generalized consumption. We show an increment of emissions in the second period, along with an alternative path of consumption. If we
take the difference in the value of consumption between the two paths, discount it
appropriately back to period 2, and then divide it by the increment in emissions,
that is the SCC in period 2.
With an optimized climate policy (abstracting away from complications due to
tax or regulatory distortions or inconsistent treatment in different sectors), the SCC
will equal the carbon price; this in turn is equal to the marginal cost of emissions
reduction and to the present value of the damages from a unit of emissions. In the
more realistic case where climate policy is not optimized, it is conventional to measure the SCC as the marginal damage of emissions along the actual path. There is
some inconsistency in the literature on the denition of the path along which the
Figure 1. Illustration of calculation of social cost of carbon. The gure illustrates an original (Base) path of emissions and consumption along with an alternative (Alt) path in
which emissions are increased by 1 unit in period 2. This leads to an alternative and lower
path of economic welfare (consumption), shown as Alt C. The SCC is calculated as the
present value of the difference in the consumption paths divided by the increment in emissions. A color version of this gure is available online.

Estimates of the Social Cost of Carbon
Nordhaus
275
SCC should be calculated. This paper will generally dene the SCC as the marginal
damages along the baseline path of emissions and output and not along the optimized emissions path. Comparisons are made between the two where that is useful.
I begin by providing a detailed analysis of a global integrated assessment climateeconomy model. For this purpose, I use an updated version, the DICE model (DICE
is an acronym for Dynamic Integrated model of Climate and the Economy). While
the DICE model is only a single model, the basic structure of the analysis is similar
in other integrated assessment models. The section also describes the major difference between the new and previous versions. The section presents the major results
of the model and compares them with other integrated assessment models (IAMs).
Section 2 then denes the SCC and describes several alternative scenarios for calculating it. The major scenarios are uncontrolled and optimized abatement, a model
calibrated to the 2C temperature target that governments and scientists have suggested, a run that uses the discounting assumed in the Stern Review (Stern Review
2007), and a high-discount-rate scenario. We show that the growth-corrected discount rate is crucial for understanding the way that discounting enters the analysis.
The section also calculates SCCs for major countries and regions. The central estimate
of the SCC in the baseline path is $18.6 per ton of CO2 in 2015 in 2005 US dollars.
Section 3 presents major alternative approaches to estimating the SCC. It nds
that virtually all estimates of the SCC have come from one of three major IAMs.
It describes the other IAMs briey. This section reviews the analysis of the US Interagency Working Group (IWG) that performed a harmonized comparison of the
three models. This section compares the IWGs output-based methodology with the
DICE models utility-based methodology. Finally, it shows the sensitivity of the analyses to alternative discount rate assumptions.
Section 4 examines several critiques of the concept of the SCC. One criticism
holds that the SCC is vastly understated because of the omission of catastrophic and
fat-tailed risks, while a second holds that the conceptual and empirical shortcomings
of IAMs are so great that the SCC concept is useless.
Section 5 examines major applications of the SCC. It focuses particularly on the
use in US regulatory policies and cost-benet analyses. While the SCC of around
$18.6 per ton of CO2 for 2015 would be an appropriate target for a harmonized
carbon tax in a rst-best world, it is likely to be higher than the appropriate number
when realistic factors are included. Appropriate adjustments are necessary to reect
national rather than global benets, tax distortions, the use of non-revenue-raising
regulations, and leakage. The analysis suggests that these adjustments would lead to a
lower implicit price than the standard SCC.
1. ANALYSIS OF THE SOCIAL COST OF CARBON
IN TH E DIC E-20 13 R M ODEL
Analyses of the SCC rely on IAMs for empirical estimates. All IAMs used for estimating the SCC have a similar structure, although they have varying emphases

276
March/June 2014
(different IAMs are reviewed in a later section). The discussion begins with the description of the model used to calculate the SCC in the present paper. Once the modeling details are developed, the precise denition of the SCC can be easily shown. I
then present numerical estimates of the SCC.
1.1. Background on the DICE and RICE Models: The Standard
DICE-2013R Model
The present discussion begins with a description of the DICE-2013R model. It will
be denoted the standard DICE-2013R model because we will also take variants
of the standard model to test for sensitivities. It is the latest version of a series of
models of the economics of global warming. The rst version of the global model
was Nordhaus (1992, 1994). The rst regional model was Nordhaus and Yang
(1996), with the most recent updated version Nordhaus (2010). This discussion will
present the major elements, and a more complete treatment is contained in Nordhaus
(2008, 2010) and Nordhaus and Sztorc (2013).
The current version of the DICE-2013R is available at dicemodel.net and http://
www.econ.yale/~nordhaus/homepage.Web-DICE-2013-April.htm. An extensive description of the 2013R model is available in the Users Manual online (Nordhaus
and Sztorc 2013).
The DICE model views climate change in the framework of economic growth
theory. In a standard neoclassical optimal growth model known as the Ramsey model,
society invests in capital goods, thereby reducing consumption today, in order to increase consumption in the future (Ramsey 1928; Koopmans 1965). The DICE model
modies the Ramsey model to include climate investments, which are analogous to
capital investments in the standard model. The model contains all elements from
economics through climate change to damages. The geophysical equations are simplied versions derived from large models or model experiments.
1.2. Equations of the DICE-2013R Model
I will list the major equations; for details readers are referred to Nordhaus and Sztorc
(2013). The model optimizes a social welfare function, W, which is the discounted
sum of the population-weighted utility of per capita consumption. The notation here
is that c(t) is per capita consumption, L(t) is population, and R(t) = (1 + )t is the
discount factor on utility or welfare, where is the pure rate of social time preference
or generational discount rate.
Tmax
W=
Uc(t); L(t)R(t).
o
=
t 1
The utility function is a constant elasticity with respect to consumption of the form
U(c) = c1 =(1 ). The parameter is interpreted as generational inequality aver-

Nordhaus
277
sion in this context. Net output, Q(t), is a function of gross output, Y(t). Net output
is gross output reduced by damages and mitigation costs:
Q(t) = (t)1 (t)Y(t) = C(t) + I(t).
In this specication, Q(t) is output net of damages and abatement; Y(t) is gross
output, which is a Cobb-Douglas function of capital, labor, and technology; C(t) is
consumption; and I(t) is gross investment. Labor is proportional to population, while
capital accumulates according to an optimized savings rate. The additional variables in
the production function are (t) and (t), which represent the damage function and
the abatement-cost function, respectively. The damage function is dened as (t) =
D(t)=1 + D(t), where
D(t) = 1 TAT (t) + 2 TAT (t)2 .
Equation (3) describes the economic impacts or damages of climate change, which is a
key component in calculating the SCC. The DICE-2013R model takes globally
averaged temperature change (TAT) as a sufcient statistic for damages. Equation (3)
assumes that damages can be reasonably well approximated by a quadratic function of
temperature change.
Uncontrolled industrial CO2 emissions are given by a level of carbon intensity,
(t), times gross output. Total CO2 emissions, E(t), are equal to uncontrolled emissions reduced by the emissions-reduction rate, (t) plus exogenous land use emissions.
E(t) = (t)1 (t)Y(t) + ELand (t).
The geophysical equations link greenhouse gas emissions to the carbon cycle, radiative
forcings, and climate change. Equation (5) represents the equations of the carbon cycle for three reservoirs.
Mj (t) = J0j E(t) +
J M (t 1).
o
=
ij
i 1
The three reservoirs are j = AT, UP, and LO, which are the atmosphere, the upper
oceans and biosphere, and the lower oceans, respectively. The parameters Jij represent the ow parameters between reservoirs per period. All emissions ow into the
atmosphere. As with many other components of the DICE model, the simplied
carbon cycle is a compromise between scientic accuracy and transparency.
The relationship between GHG accumulations and increased radiative forcing is
shown in equation (6).

278
March/June 2014
F(t) = flog 2 MAT (t)=MAT (1750)g + FEX (t);
where F(t) is the change in total radiative forcings of greenhouse gases from anthropogenic sources such as CO2 , FEX(t) is exogenous forcings, and the rst term is
the forcings due to atmospheric concentrations of CO2.
Forcings lead to warming according to a simplied two-level global climate model:
TAT (t) = TAT (t 1) + 1 fF(t) 2 TAT (t 1) 3 TAT (t 1) TLO (t 1)g. 7
TLO (t) = TLO (t 1) + 4 TAT (t 1) TLO (t 1).
In these equations, TAT(t) is the global mean surface temperature and T LO(t) is the
mean temperature of the lower oceans. Major revisions to the model are described
below.
1.3. Revisions from Earlier Versions of the DICE and RICE Models
There are several large and small changes in the DICE-2013R model compared to
earlier versions. The prior complete documented version of the DICE model is
Nordhaus (2008), while the last complete version of the regional (RICE, signifying
the Regional Integrated model of Climate and the Economy) model is in Nordhaus
(2010).
The rst revision is that the time step has been changed from 10 years to 5 years.
This change is taken because improvements in computational capacities allow the
model to be easily solved with a ner time resolution.
A second change is the projection of future output growth. Earlier versions of
the DICE and other IAMs tended to have a stagnationist bias, with the growth
rate of total factor productivity declining rapidly in the coming decades. The current version assumes continued rapid total factor productivity growth over the next
century, particularly for developing countries.
A third revision incorporates a less rapid decline in the CO2-output ratio in
several regions and for the world, which reects the last decades observations. Earlier
trends (through 2004) showed rapid global decarbonization, at a rate between 1.5%
and 2% per year. Data through 2010 indicate that decarbonization has been closer
to 1% per year. The new version assumes that, conditional on output growth, uncontrolled CO2 emissions will grow at 0.5% per year faster than earlier model assumptions.
A fourth assumption involves the damage function. The most recent versions
of both DICE and RICE used the impact estimates from the 2000 RICE model
(Nordhaus and Boyer 2000). There has been signicant further work on damages
since that time. The new model uses estimates of monetized damages from the Tol
(2009) survey as the starting point. Tols central estimate is that damages are about
3% of global output at a temperature increase of 3C. However, current studies

Nordhaus
279
generally omit several important factors (biodiversity, ocean acidication, and political reactions), extreme events (sea-level rise, changes in ocean circulation, and accelerated climate change), impacts that are inherently difcult to model (catastrophic
events and very long-term warming), and uncertainty (of virtually all components
from economic growth to damages). I have added an adjustment of 25% of the monetized damages to reect these nonmonetized impacts. While this is consistent with
the estimates from other studies (see Anthoff and Tol 2010; Hope 2011; FUND
2013), it is recognized that this is largely a judgmental adjustment. The current version assumes that damages are a quadratic function of temperature change and does
not include sharp thresholds or tipping points, but we return to this question in section 4.
A fth revision recalibrates the carbon cycle and climate models to recent earth system models. The equilibrium and transient temperature impacts of CO2 accumulation have been revised to include a wider range of estimates. Earlier versions relied
entirely on the estimates from general circulation models (e.g., the ensemble of models
used in the IPCC (Intergovernmental Panel on Climate Change) Fourth Assessment,
Science 2007). The present version uses estimates from sources such as the instrumental record and estimates based on the paleoclimatic reconstructions. The carbon
cycle has been adjusted to reect the saturation of ocean absorption with higher temperatures and carbon content. See the Users Manual for details.
A sixth set of changes are updates to incorporate the latest output, population,
and emissions data and projections. Output histories and projections come from the
International Monetary Funds World Economic Outlook database. Population projections through 2100 are from the United Nations. CO2 emissions are from the Carbon Dioxide Information Analysis Center (CDIAC). Non-CO2 radiative forcings for
2010 and projections to 2100 are also from projections prepared for the IPCC Fifth
Assessment. The denitions of regions (particularly the European Union and developing countries) have changed to reect changing compositions and reect the structure as of 2012.
A seventh question concerns the calibration of the model for rates of return on
capital. The philosophy behind the DICE model is that the capital structure and
rate of return should reect actual economic outcomes. This implies that the parameters should generate savings rates and rates of return on capital that are consistent
with observations (this is sometimes called the descriptive approach to discounting
after Arrow et al. [1996]). The data on rates of return used in the calibration are
as follows: (a) the risk-free real return, generally taken to be US or other prime sovereign debt, is in the range of 0%1% per year depending upon period, concept, and
tax status; (b) the rate of return on risky capital of large corporations in mature
markets, after company taxes but before individual taxes, is in the range of 5%8%
per year depending on period, concept, and tax status; (c) the rate of return on risky
investments in illiquid or immature markets, as well as for poorly capitalized individu-

280
March/June 2014
als, is generally much higher than for corporations and ranges from 0% to 100% per
year depending on the circumstances; (d) it is unclear how much of the difference between the return on risky capital and the risk-free return is compensation for nondiversiable risk (see Mehra 2008), but for the present study I assume that the equity
premium reects nondiversiable risks; (e) the extent to which climate investments
are correlated with systematic consumption risk is an open question, although preliminary work from Nordhaus (2008) and Gollier (2013) suggest a relatively high consumption beta. For the present study, I assume that the consumption beta on climate
investments is close to one.
Based on these considerations, I assume that the rate of return relevant for discounting the costs and benets of climate-sensitive investments and damages is 5%
per year in the near term and 4.5% per year over the period to 2100. This is the
global average of a lower gure for the United States and a higher gure for other
countries, and it is therefore consistent with estimates in other studies that use US
data, such as the US Interagency Working Group discussed later. With this calibration, we choose the pure rate of social time preference () to be 1.5% per year
and the consumption elasticity () to be 1.45.
An eighth revision is to change the units of measurement from tons of carbon to
tons of CO2 or CO2-equivalent, this being to reect the current conventions in most
price and economic data.
A ninth change is a redenition of the concept of the baseline for modeling
purposes. In earlier vintages, baseline meant no policies. In the current model,
baseline assumes that current policies as of 2013 are extended indenitely. This approach is standard for forecasting, say of government budgets, and is more appropriate for a world of evolving climate policies. Estimates from Nordhaus (2010) indicate that 2010 policies were the equivalent of $1 per ton of CO2 global emissions
reductions. Note that this approach requires calculating baseline emissions intensities
as reecting the current level of emissions reductions.
The major net effect of these revisions is to increase modestly the growth rate of
world output, the emissions and concentrations of CO2, the temperature trajectory,
climate damages, and the calculated social cost of carbon. Most of the revisions in
the data have been bad news for climate change.
1.4. Major Results for DICE-2013R
It will be useful to show some representative results from the revised model. We
also compare the results with a recent model comparison exercise, the EMF-22
study, surveyed under the aegis of the Energy Modeling Forum (EMF) at Stanford
University. This study compared the results of several IAMs from modeling teams
from around the world.
Figure 2 shows the industrial emissions of CO2 over the coming century. It
compares the present version of the model with the EMF projections. The gure

Nordhaus
281
Figure 2. Projected industrial CO2 emissions in baseline scenario. The heavy dashed line
with triangles is the average of the 11 models surveyed in the EMF-22 project. The heavy
line with squares is the DICE-2013R version. The light lines are the individual EMF-22
models. The EMF results are described in Clarke et al. (2009); detailed modeling results
were made available by Leon Clarke. A color version of this gure is available online.
shows both the average of the models in bold and 11 individual models as light
lines. The DICE-2013R version is near the top of the range, with the changes explained in the last section.
Figure 3 shows a comparison of projected temperatures of different models and
scenarios. For the temperature projections for 10 EMF-22 models, I have used the
CO2 concentrations for the different models and the DICE climate model and
non-CO2 forcings; this was necessary because most of the models did not calculate
temperature trends, and they used inconsistent non-CO2 forcings. The heavy lines
are the average of the EMF-22 models and the DICE-2013R model. Also shown
are four scenarios from the IPCC Fourth Assessment Reports that have been commonly cited. The DICE-2013R results are at the upper end primarily because of
the higher projected CO2 emissions, as just discussed.1 (Note that the Fifth Assessment Report uses a different methodology in its climate projections, and these
cannot be easily compared with economic models or with earlier estimates.)
1. Although some of the EMF-22 integrated assessment models provide temperature trajectories, they exclude short-lived greenhouse gases and aerosols and therefore do not provide a
comparable temperature projection. The runs shown in g. 3 take the industrial CO2 concentrations from the EMF-22 models. These are then combined with estimates of land-use CO2
emissions and the radiative forcings for other GHGs from the RICE-2010 model. We then

282
March/June 2014
Figure 3. Global mean temperature increase as projected by IPCC scenarios and integrated assessment economic models. The gure compares the projections of four scenarios
using IPCC scenarios with those of the DICE-2013R model and the average of 10 EMF-22
integrated economic models. The letters A1B, A2, B1, and B2 represent the results of four
IPCC standardized emissions and the ensemble of climate model projections from the IPCC
Fourth Assessment Report. A color version of this gure is available online.
These results provide the context for how the DICE-2013R model ts into
other model projections. They present a cautionary warning about both the difculties of projections as well as the differences that can arise from alternative model
assumptions, algorithms, output projections, and energy sectors. The spaghetti diagram of different forecasts has been a feature of emissions and other projections
since early surveys, such as Nordhaus and Yohe (1983), and it is a good indication
that forecast errors increase as we go further into future.
2 . ES T I M A T E S O F TH E S O C I A L C OS T O F C A R B O N
2.1. Denition of SCC

We now provide a precise denition of the social cost of carbon. Using equations (1)
(8), we can solve for the social welfare function, W, in terms of the various exogenous
and policy variables. We then dene the social cost of carbon at time t as SCC(t).
put all these into the climate module of the RICE-2010 model. The 10 models were ETSAPTIAM, FUND, GTEM, MERGE Optimistic, MERGE Pessimistic, MESSAGE, MiniCAMBASE, POLES, SGM, and WITCH. A full description of the models is contained in the
source at Clarke et al. (2009).

SCC(t)
Nordhaus

W
W
.
E(t) C(t)
283
The numerator is the marginal impact of emissions at time t on welfare, while the
denominator is the marginal welfare value of a unit of aggregate consumption in period
t. The ratio calculates the economic impact of a unit of emissions in terms of t-period
consumption as a numraire. In actual calculations, we take a discrete approximation
to (9). Note that the SCC is time-indexed. This indicates that the marginal cost of
emissions at time t (in terms of consumption at time t as a numraire) changes over
time.
2.2. Estimates of the SCC in the DICE-2013R Model
We have estimated the SCC in the DICE-2013R model for several alternative scenarios. These reect differing assumptions about policy, damages, and discounting.
We list the scenarios briey and describe them in more detail below.
1. The rst is the baseline scenario, which uses the standard DICE-2013R
model and assumes no changes in climate change policy from 2010 levels.
2. The second is the optimal climate policy scenario, which uses the
standard DICE-2013R model and optimizes the time path of emissions
reductions and investment.
3. A third run modies the damage function in the standard DICE model
so that the optimal path leads to a limit on temperature increase to 2C
above the 1900 level.
4. The fourth run is a variant of run 3 and assumes that the 2C limit is an
average rather than an annual or decadal maximum.
5. The fth run examines the impact of a near-zero discount rate on the
SCC.
6. The sixth run is a variant on the fth that calibrates other parameters to
keep real returns on capital in the lower-time-preference scenario equal to
the rate of return in the baseline scenario.
7. A nal run is a high-discount-rate sensitivity analysis that raises the pure
rate of social time preference to 3.5% per year.
More details on the assumptions in the different runs are provided as the analysis proceeds.
The methodology for estimating the SCC is straightforward using the GAMS
(General Algebraic Modeling System). We calculate the shadow price on CO2 emissions in each run and then divide that by the shadow price of consumption. This provides exactly the formula shown in equation (9) above. The units are 2005 US international dollars per metric ton of CO2 and are expressed in terms of consumption in
the given year. The results are shown in table 1, and the different scenarios are dis-

284
March/June 2014
Table 1. Global Social Cost of Carbon under Different Assumptions

Scenario
Base parameters:
Baseline*
Optimal controls
2C limit damage function:
Maximum
Max of average
Stern Review discounting:
Uncalibrated*
Calibrated*
Alternative high discount*
2015
2020
2025
2030
2050
18.6
17.7
22.1
21.2
26.2
25.0
30.6
29.3
53.1
51.5
47.6
25.0
60.1
30.6
75.5
37.1
94.4
44.7
216.4
87.9
89.8
20.7
6.4
103.7
25.0
7.7
117.4
30.1
9.2
131.3
35.9
10.9
190.0
66.9
19.6
Note.The social cost of carbon is measured in 2005 international US dollars. The years at the
top refer to the date at which emissions take place. Therefore, $18.6 is the cost of emissions in 2015 in
terms of consumption in 2015.
* Calculation along the reference path with current policy.
Calculation along the optimized emissions path.
cussed in turn. (An alternative approach is needed when the EXCEL version of the
DICE model is used. For that, we perturb emissions by 1 million tons in a given
year. We calculate the change in the present value of utility scaled by consumption in
the given year to get the SCC. This gives the identical results, subject to computational precision.)
2.2.1. SCC for Standard DICE Model Parameters
The central cases for the SCC are shown in the rst two rows of table 1. The rst
row shows the estimate for the standard DICE model with baseline or current climate policy. The SCC gure here is $18.6 per ton of CO2 for emissions in 2015,
$22.1 for emissions in 2020, and so forth. Recall that these are time-dated SCCs
since the cost of emissions depend upon the year in which emissions take place.
(Sometimes estimates are given in dollars per ton of carbon. The carbon weight is
1/3.667 times the CO2 weight, so the carbon price is 3.667 times the CO2 price, in
this case $18.6 per ton of CO2 is $68 per ton of carbon.) This SCC rises at a real
rate of 3% per year over the period to 2050. The rate of change of the SCC depends
upon several factors, particularly the rate of growth of world output, the removal rate
of atmospheric carbon, and the discount rate.
The SCC along an optimized path, shown in row 2, is about 5% less than along
the baseline path. It is a moderate surprise that the SCC does not differ markedly
between the optimized case and the baseline case. The reason for this result is that

Nordhaus
285
the damage function is close to linear in the range between the two cases. In other
words, the marginal damage in early periods is only slightly affected by optimizing
emissions.
2.2.2. The Social Cost of Carbon in the Context of the 2C Target
The rst set of estimates is based on the standard parameters from the DICE
model. The optimized path yields temperatures that are above the limit that has
been agreed upon in international meetings and has been recommended by several
scientic groups. The current consensus is known as the Copenhagen Accord (see
United Nations 2009). The accord adopts a target of limiting the increase in global
mean temperature, recognizing the scientic view that the increase . . . should be
below 2 degrees Celsius. There was no specic scientic document to support this
statement, although the European Union has developed such a target in the context of a cost-benet analysis. This target has been referenced in further International Conferences of the Parties.
It will be useful to put the Copenhagen target in the context of the current study.
The approach taken here assumes that policy makers who negotiated the Copenhagen temperature target implicitly have a different damage function from that used
in the DICE (and most other) economic models.
How can we rationalize the 2C target? In other words, (i) what change in the
economic parameters would lead to an optimized temperature that corresponds to
that adopted in the Copenhagen Accord? And (ii) what is the SCC that would be
associated with this alternative set of parameters? One possibility is a lower discount rate, which I investigate in the next section. The other possibility is a damage function that assumes much higher damages than standard models. No other
change in parameters that was tested would come close to producing the result. (I
note as an aside that the higher-damage interpretation appears more consistent with
the expressed gravity of concerns about climate change as well as high discount rates
that are implicit in other decisions.)
I then proceeded to recalibrate the DICE-model damage function so that an optimal abatement plan would lead to the 2C target. More precisely, I assume that
the 2C target reects a cost-benet optimum with a different damage function,
which is denoted the 2-degree-consistent damage function. This is implemented
by increasing the rst-order damage coefcient (the coefcients in the damage function in eq. [3]). I did this for two different cost-benet solutions. In the rst, the
cost-benet optimum produces a maximum temperature increase for the 2050
2250 period that is 2C above the 1900 level. In the second, the cost-benet optimum produces average temperature increases for the 20502250 period that are
2C above the 1900 level. The rationale of the second approach is that many of the
damages (such as ice-sheet melting and sea-level rise) are better approximated by a
function of the average than the maximum temperature.

286
March/June 2014
With this procedure, the damage coefcients would need to be increased by a

factor 2.0 times larger than the DICE base estimates to produce a 2C average target
and by a factor of 4.4 times larger to be consistent with a 2C maximum temperature. These functions assume that damages are 2.4% (DICE standard), 4.9% (2 C average), and 10.6% (2% maximum) of output at 3C warming. As an interpretive
note, while the implicit damage coefcient associated with the Copenhagen maximum target is not impossibly high, no credible estimates in the literature exist to justify such high damage rates. However, the coefcient associated with the 2C average
is not outside the range of hypothesized damages estimates.
Table 1 shows estimates of the SCC for the two 2C cases, which are $47.6 and
$25.0 per ton of CO2 in 2015 for the two cases of maximum and average. These
are approximately 2.6 and 1.3 times the SCC with the standard parameters, respectively. Note that a policy that looks at the temperature average rather than the
temperature maximum has a substantially lower SCC.
The conclusion here is as follows: economic models may not incorporate all the
concerns of scientists and policy makers about the damages of climate change. The
current damage estimates may signicantly underestimate the damages because of
impacts that are difcult to monetize (such as ecosystem effects) or concerns about
catastrophic outcomes. If we increase the damage function so that the economic optimum coincides with the 2C target, we nd that the SCC rises sharply. The implicit damage function in the 2C maximum target is much larger than any estimate
from the damage literature.
2.2.3. The Role of Discounting: A Simplied Example
Estimates of the SCC differ greatly across different approaches. The wild card in
calculations of the SCC is the discount rate. This point is intuitively obvious in a
simple example. Assume that we linearize all the equations of the DICE model. Assume that all environmental variables have reached a stationary state where emissions, concentrations, population, temperature, and other physical variables are constant. Output, consumption, and damages are growing at constant rate g, and the
goods discount rate is r. If we perturb emissions by 1 unit, this will cause a path of
damages that is distributed over the distant future. For simplicity, assume that the
damages start immediately but that the damage-output ratio declines at a decay
rate of per year.
We can use the Ramsey equation to evaluate the SCC as a function of the key
variables. The Ramsey equation provides the equilibrium rate of return in an optimal growth model with constant growth in population and per capita consumption without risk or taxes. In this equilibrium, the real interest rate (r) equals the
pure rate of social time preference ( ) plus the rate of growth of per capita consumption (g ) times the consumption elasticity of the utility function (). In longrun equilibrium, we have the Ramsey equation r = + g. The key variable will be

Nordhaus
287
the growth-corrected discount rate, r g. Under our assumptions, r g = +

( 1)g. To simplify, assume that = 1, or that the utility function is logarithmic,
which implies that r g = . (These long-run growth and discounting are used in
the Stern Review and are approximately the case for the DICE model.)
Under these assumptions, the SCC is proportional to 1=( + ). Assume that
= 0.005 per year, which is an approximate value for the decay rate of damages.
Then, for alternative values of = 0.001 (Stern Review), 0.015 (DICE), or 0.035
(the high discounting alternative examined here), and normalizing so that the SCC
with DICE is $20, we get values of the SCC at $67, $20, and $10 per ton of emissions for the three discount rates. There is no mystery here. With long lags in the
physical system and with damages far in the future, differences in discounting will
lead to large differences in the SCC.
2.2.4. SCC with Alternative Discount Rates
We next present estimates of the SCC in the DICE-2013R model with alternative
discounting assumptions. Row 5 in table 1 takes the parameters of the model and
only changes the pure rate of social time preference ( ) from 0.015 to 0.001 per
year, the latter being the rate assumed in the Stern Review. The Stern uncalibrated
run has a SCC of $89.8 per ton of CO2almost ve times the SCC with baseline
discounting. (We do not show the SCC along the optimal path, but it is about
one-third smaller because of the larger damages and lower discounting.)
Calculating the impact of alternative discount rates using the calibration approach of the DICE model is slightly more complicated than the case just given. If
we just change the rate of time preference, then we will get rates of return on capital projections that do not match observed baseline market values. The sixth line
in table 1 (Stern Review, recalibrated) shows the results of recalibrating the model
to match market discount rates. To recalibrate with a rate of time preference using
the Stern Review value ( = 0.001), we need to raise the rate of inequality aversion
(from = 1.45 to = 2.1). The calibration keeps the rate of return at the baseline
average for the rst 30 years. Figure 4 shows the real return on capital in the four
discounting cases, where the differences induced by the uncalibrated Stern Review
run are evident.
The results of the recalibration are dramatic. If we maintain the super-low rate
of time preference (at = 0.001) but adjust inequality aversion to maintain baseline
interest rates, the SCC is virtually indistinguishable from the baseline SCC. The
reason is that the upward adjustment of the elasticity of inequality aversion offsets
most of the downward adjustment in the pure rate of time preference.
A nal case is row 7 in table 1 uses a high alternative discount rate. This scenario raises the pure rate of time preference to 3.5% per year. The rate of return is
approximately 2 percentage points higher than the baseline, and the SCC is approximately one-third the baseline case. The SCC is also much lower.

288
March/June 2014
Figure 4. Rate of return on capital for alternative discount-rate scenarios. This gure shows
the rates of return on capital in four scenarios. These are also the consumption discount
rates because there are no distortionary taxes or regulations in the DICE model. A color
version of this gure is available online.
Figure 5 is a scatter plot of the growth-corrected return on capital and the 2015
SCC. The growth-corrected discount rate, dened as the rate of return on capital
minus the growth rate of output, uses the period 20102100 for the calculation.
Figure 5 shows dramatically how the growth-corrected discount rate enters in a
way suggested by the little algebraic example at the beginning of this section. The
intuition is to recognize that damages tend to be proportional to the size of the
economy in most studies on damages. In effect, if the goods discount rate is equal to
the growth rate of output, this means that an increase in the damage-output ratio
will be undiscounted. This shows why very low growth-corrected discount rates have
very high SCCs.
2.2.5. Regional SCCs
Several models calculate the SCCs of different regions. These are the marginal
impact of emissions on the economic welfare of a particular country or region. These
estimates are important for understanding the problem of noncooperative behavior
as well as the issues of using the SCC in domestic regulations, both of which are
discussed later. Table 2 shows the global SCC for the baseline case (line 1 of table 1).

Nordhaus
289
Figure 5. Social cost of carbon- and growth-corrected discount rate in DICE model. This
gure shows the central role of the growth-corrected discount rate in determining the SCC.
The growth-corrected discount rate equals the discount rate on goods minus the growth rate
of consumption for 20102100. A color version of this gure is available online.
The regional estimates use a hybrid of the DICE-2013R model presented above
and the regional estimates from the RICE-2010 model. More precisely, the estimates use the regional SCC from the 2010 model and adjust them proportionally
so that the sum equals the global total for the 2013R version. Because the externality is a pure global public good, the global SCC exactly equals the sum of the regional SCCs. While the two models are not completely harmonized, experience with
earlier versions of the models indicates that the ratios of the national SCCs to the
global average are relatively stable.
China is estimated to have the highest regional SCC primarily because it has
high future output. Several other regions are close to China in their regions SCCs.
Damages are roughly proportional to discounted GDPs, with some countries deviating from this rule because of different climate impact sensitivities. Alternative estimates from the FUND and PAGE models are also shown in table 2. These numbers indicate that there is little consensus on the distribution of the SCC by region

290
March/June 2014
Table 2. Region Social Costs of Carbon

Percent of Global SCC
Region
United States
European Union
Japan
Russia
Eurasia
China
India
Middle East
Africa
Latin America
OHI
Other
Weighted country
average
Global
Emissions (Billions of
Tons CO2 , 2005)
SCC
(2015)
RICE
2010 (U)
FUND
2013
PAGE
2011
6.11
4.14
1.28
1.54
.92
6.14
1.48
2.14
.69
1.54
1.93
1.38
1.94
2.32
.43
.18
.16
3.02
2.21
1.89
2.09
1.30
.74
2.29
10
12
2
1
1
16
12
10
11
7
4
12
17
24
3
10
NA
8
5
NA
6
NA
NA
NA
7
9
NA
NA
NA
11
22
NA
26
11
NA
NA
29.30
1.92
18.6
100
100
100
Note.This table distributes the global SCC by region. It uses the global estimate of the SCC from
DICE-2013R and the regional distribution from RICE-2010. The weighted country average is the average of the country-specic SCCs. The results from other studies indicate that the regional distribution is
poorly understood. FUND results are from Anthoff (2013) and PAGE results are from Hope (2011).
Note that the regions do not always conform exactly across the models. OHI = other high-income
countries; NA = not available.
except that no region dominates the total. The different estimates reect the poor
understanding of the impacts by region.
The estimated SCC for the United States is $1.94 per ton CO2 (2015 in 2005
prices), or 10% of the global SCC. We will discuss the implications of this estimate
below because it is a small fraction of the SCC used in regulatory impact analyses
for the United States.
2.2.6. Regional SCCs and the Noncooperative (Nash) Equilibrium Carbon Price
The regional numbers are useful in providing an estimate of what countries would
abate in a noncooperative equilibrium in which there was no reaction among different regions. For example, if each region sets its own regional carbon price equal to
its regional SCC, the weighted average price would be $1.9 per ton of CO2, or
about 10% of the global SCC (see the penultimate entry in col. 2 of table 2).
Similar results have been found in earlier studies (see Nordhaus and Yang 1996;
Bosetti, Massetti, and Tavoni 2007; Yang 2008).

Nordhaus
291
3. ALTERNATIVE ESTIMATES
There have been many estimates of the SCC in different models. Those who would
like to apply these to energy or climate policy face the daunting task of sorting through
the different estimates and deciding which seem most appropriate. I discuss alternatives and compare them with the DICE-model estimates in this section.
3.1. Back to Basics
Estimates of the SCC come in different varieties. The primary sources are integrated
assessment models. An examination of the literature nds that there are three models that are the basis of virtually all estimates. One is the DICE/RICE family of
models, discussed above. A second is the PAGE model, originating with the work of
Chris Hope of Cambridge University, which has several vintages. The third is the
FUND model, developed by Richard Tol of Sussex University, with David Anthoff
of the University of California, Berkeley, a coauthor of the current version.
Another set of SCC estimates come as variations on a theme of one of the three
models. Some studies take one of the models and change parameters or do sensitivity
analyses. For example, the Stern Review based its quantitative modeling and estimates
of the SCC on a version of the PAGE model that varied primarily the discount rate
in a fashion discussed in the last section. Some studies are reviews. Others are compilations of studies, which include primary models, variations of models, and reviews
of models. One of the most inuential of the estimates, prepared by the US Interagency Working Group and used in US regulatory analyses (discussed below), relied
on their own runs of the three models discussed in the last paragraph.
Yet further estimates come from stylized models that simplify the structure of
the economic-climate nexus. An excellent example of this is Golosov et al. (2014).
They present a dynamic stochastic model with a climate externality. There are several insights from the model about important determining factors. For example, they
derive a simple formula for the SCC that depends only on four factors: (i) the size
of the global economy, (ii) discounting, (iii) the damage elasticity, and (iv) carbon
depreciation in the atmosphere. While this is useful, the last three factors are likely
to be variables rather than constants; moreover, the damage elasticity is extremely
complicated to calculate. A similar but more complicated SCC formula is derived in
Li and Nordhaus (2013). These stylized approaches cannot capture important aspects, such as the lags in geophysical and economic systems, as well as declining discount rates or nonstationarities. Moreover, analysts want to understand the logic
behind a model and its correspondence to more detailed scientic models, and those
are hidden in a stylized model.
As an example of the small number of independent analyses of the SCC, Li and
Nordhaus (2013) did a systematic study of estimates of SCC from 1980 to 2012. They
found 27 studies that actually produced independent estimates. Of these, 19 were from

292
March/June 2014
different vintages of the three standard models listed above. Most of the eight other
estimates were reduced-form models that took one of the three standard models and
either simplied it or added specic features.
The conclusion from this short summary is that most of the estimates of the
SCC come from one of the three IAMs that have been developed and revised over
a period of at least a decade.
3.2. The FUND Model and PAGE Model Estimates
In light of the central importance of the three standard models, I will briey describe the other two and then provide a comparison of estimates of SCC from them.
FUND and PAGE share the basic integrated structure of the DICE/RICE models
in linking output, emissions, concentrations, temperature, and damages. The similarity stops there.
The FUND model (Climate Framework for Uncertainty, Negotiation, and Distribution) was developed primarily to assess the impacts of policies in an integrated
framework. It is a recursive model that takes major economic variables as exogenous. There are 16 major regions. Climate change impacts are monetized and include agriculture, forestry, sea-level rise, health impacts, energy, consumption, water
resources, unmanaged ecosystems, and storm impacts. Each impact sector has a different functional form and is calculated separately for 16 geographic regions. The
model runs from 1950 to 3000 in time steps of 1 year. The source code, data, and
a technical description of the model are public, and the model has been used by
other modeling teams (http://www.fund-model.org).
The PAGE model (Policy Analysis of the Greenhouse Effect) projects future
increases in global mean temperature, the economic costs of damages caused by
climate change, and the economic costs of mitigation policies. It has a relatively simple
economic structure, taking output and emissions as exogenous with many periods,
countries, and sectors. The major innovations are detailed inventories of greenhouse
gases; reduced-form treatment of the atmospheric chemistry of gases; simplied global
and regional climate models, including of aerosols; and detailed regional impacts. Moreover, the PAGE model makes uncertainty a central focus, with 31 uncertain variables
(such as climate sensitivity, carbon cycle dynamics, impacts, and discontinuous impacts).
The damage structure is highly developed, with catastrophic thresholds and sharp discontinuities introduced probabilistically. The model is proprietary but is available to
others with permission and credits.
Two features of FUND and PAGE are worth highlighting. First, the models are
output-based, not utility-based. The metrics for measuring monetary values of output, impacts, and abatement over space and time are in constant US dollars, and
they are evaluated using discounted output or consumption.
Second, the models have elaborate treatment of uncertainties. However, most of
the distributions of the uncertain variables are judgmental and difcult to assess. On

Nordhaus
293
occasion, the specication has produced extreme results, and the interaction of the
many uncertainties with highly nonlinear dynamic models is not fully understood.2
In sum, the three standard models used for calculating SCCs are very different in
their structures, assumptions, treatments of uncertainty, and economic modelings.
These differences mean that there is no easy way to judge the relative reliability of
the models or their results.
3.3. Meta-analysis of Different SCC Estimates
In deciding on the appropriate SCC, policy makers may want to examine systematic
surveys of different estimates. One approach is a meta-analysis of existing studies,
of which a leading example is Tol (2008). This study collected 211 estimates of the
SCC from 47 studies. The results were run through several statistical lters, with
the mean SCC of $34$42 per ton of CO2 and median of $5$25 per ton CO2,
depending upon the aggregation procedure.3
While it is valuable to examine alternative estimates, it is misleading to call this a
meta-analysis in the usual sense.4 The basic structure of a meta-analysis is that there
are several samples from a population, and those samples can be combined to estimate a parameter or other statistics more efciently. The standard example would
be a clinical trial with multiple sites to test the effect of a treatment. The effects of
the treatment can be more efciently estimated in a meta-analysis by combining the
observations from the different sites.
We can see from this description why a classical meta-analysis cannot be properly used to combine different estimates of the SCC. First, as we saw from the description of the sources of the SCC in the last section, the different studies are not
independent samples from some underlying distribution. For example, the Tol sample includes studies that take a model and change some parameters. An example is
the SCC estimate in Cline (1997), which used the DICE model and changed the
discount rate. Interpreting this estimate is particularly complicated because it was
primarily used to defend a particular view on discounting. A similar difculty is the
estimate from the Stern Review discussed above. Second, there is no clear mechanism
by which the data are generated. Third, there is no sense in which there is a sample
size applied to a given study. Yet a fourth problem arises when a meta-analysis includes surveys or other meta-analyses. For example, Tol includes Clarkson and Deyes
2. For example, an earlier version of the FUND model had an assumption that led to
one of the uncertain variables being a noncentral Cauchy distribution, so the theoretical
value of the SCC was unbounded (see Ackerman and Munitz 2012).
3. The year of emissions and price level are not provided, but in other work, Tol has
used 1995 dollars.
4. A useful reference for the social sciences is Hedges and Olkin (1985).

294
March/June 2014
(2002), which is a review of earlier estimates and included other studies included in
the Tol meta-analysis. So the meta-analysis has triple-counted some studies included
in the review.
The main conclusion on the use of meta-analyses for the SCC is that these denitely do not meet the standard requirements for a statistical meta-analysis and
should be treated with caution. They would be more accurately described as research syntheses or quantitative summaries of the literature that show some of the
important factors driving the estimates. The results are especially questionable because they double-count and even triple-count different model estimates.
3.4. The US Interagency Working Group Estimates
A particularly inuential and important set of estimates of the SCC was provided by
the Interagency Working Group on Social Cost of Carbon, US Government, hereafter Interagency Working Group or IWG (IWG 2010; Greenstone, Kopits,
and Wolverton 2013). This was updated in IWG (2013).
The original analysis was developed by technical experts from numerous agencies of the US federal government,5 who met on a regular basis to consider public
comments, explore the technical literature in relevant elds, and discuss key model
inputs and assumptions. The main objective of this process was to develop a range
of SCC values using a defensible set of input assumptions grounded in the existing
scientic and economic literatures (IWG 2010, 1). The 2013 estimates update the
models but use the same methodology as the 2010 estimates. The analysis has been
used for rule making by the US government, and the application of the SCC in
those rules is considered in a later section.
The procedure used by IWG was particularly interesting because it obtained
the three standard models used to calculate the SCC and developed a harmonized
analysis by standardizing two sets of input parameters. The equilibrium temperature climate sensitivity, or TSC (temperature sensitivity coefcient; see eq. [7] and
eq. [8]) was harmonized using the Roe-Baker analysis (Roe and Baker 2007) with
a mean equilibrium TSC of 3.5C.6
5. The participants were the Council of Economic Advisers, Council on Environmental
Quality, Department of Agriculture, Department of Commerce, Department of Energy, Department of Transportation, Environmental Protection Agency, National Economic Council,
Ofce of Energy and Climate Change, Ofce of Management and Budget, Ofce of Science
and Technology Policy, and Department of the Treasury.
6. One strange feature of the treatment of the TSC by the IWG should be noted. The
IWG assumed a most likely value of about 3C. The IWG harmonized the Roe-Baker
distribution by setting the median at a target level of 3.0C, and for the Roe-Baker distribution this implied that the mean TSC was 3.5C. This choice of median rather than mean
was casually thrown out in a spirit that it did not much matter whether the central tendency
was mean, median, or mode. This procedure is indefensible, as we will see below.

Nordhaus
295
Second, the IWG used ve sets of output, population, and emissions trajectories
based on a model comparison study (the EMF-22 runs discussed above). It then
took the resulting changes in output and consumption and calculated the SCC using
discount rates of 2.5%, 3%, and 5% per year. This procedure is valuable because it is
to my knowledge the only SCC estimates that have undertaken partial harmonization of different models.
Third, the IWG used an output-based method, while the DICE model uses a
utility-based approach. The two will give equivalent results if the discount rate on
goods is the same. However, in virtually all models with endogenous discount rates,
the discount rate is declining over time, while the IWG assumes a constant discount
rate. The rationale for a declining discount rate is explained in Arrow et al. (2013).
Table 3 shows the results of the IWG estimate and compares them with the
DICE-model calculations presented above. Here are some of the highlights of table 3. All gures in table 3 are the global SCC for 2015 in 2005 US dollars.
Begin in panels A and B, which are the results of the IWG calculations
(2010, 2013). These use the IWG estimates and discounting methodology.
The 2013 estimates are revised upward substantially for the FUND and
PAGE models. For the preferred 3% constant discount rate, the SCC for
2015 is revised upward from $22.4 to $35.8 per ton of CO2.
Panel C adopts the IWG assumption of a constant discount rate but
imbeds this in the DICE model so that it is generated endogenously and
used consistently. In these calculations, the pure rate of time preference is
set equal to the goods discount rate, and the consumption elasticity is set
equal to zero. This approach assumes linear utility, that is, that there is no
inequality aversion. It is consistent with the approach of projecting output
and damages and not relying on utility calculations.
The rst line of panel C is most comparable to the IWGs 2013 estimates
in line 1 of panel B. The DICE-2010 constant-discount-rate-scenario is
virtually identical to the comparable IWG calculation, which indicates that
the DICE output and emissions projections are close to the average used
in the IWG scenarios. The second line of panel C uses the RICE-2013R
model with the IWG TSC average of 3.5. It is 40%50% higher than the
IWG estimate of the DICE model largely because of the upward revision
of output and emissions in the DICE-2013R model.
The last line of panel C uses the DICE-2013R model and the 2013R
TSC, which is 2.9C. The bottom line here is that the DICE-2013R
model estimates (including the DICE TSC) are somewhat higher than
the latest IWG estimates for the DICE model.
Panel D moves to the DICE-2013R models utility approach from the
output approach of the IWG. This uses the DICE model baseline

Table 3. Estimates of the Social Cost of Carbon for 2010 from US Interagency Working
Group and Comparison with Alternative Model Estimates
Constant Discount Rate on Goods
Model and Scenario
A. Estimates of 2015 SCC from
US Working Group, 2010:
DICE-2007
PAGE
FUND
Average
B. Estimates of 2015 SCC from
US Working Group, 2013:
DICE-2010
PAGE
FUND
Average
C. DICE model estimates: working group
methodology (constant discount rate):
DICE-2010 (TSC = 3.5)
DICE-2013R (TSC = 3.5)
DICE-2013R (TSC = 2.9)
D. DICE 2013R model estimates (DICE
model declining discount rate):
DICE-2013R (baseline)
5%
4.2%*
4%*
3%
2.5%
10.2
7.4
1.5
5.4
17.4
15.3
2.5
11.7
29.6
31.3
6.3
22.4
43.5
50.9
14.0
36.2
11.0
20.2
2.7
11.3
18.6
34.4
6.9
20.0
31.4
58.6
17.3
35.8
48.1
85.3
30.4
54.6
10.4
14.8
12.3
16.8
25.8
21.0
31.1
50.0
40.2
46.2
72.5
58.0
18.5
18.6
Note.Panel A shows estimates of the 2010 SCC from the Interagency Working Group. The
three models have harmonized outputs, emissions, populations, and temperature sensitivity coefcient
(TSC) distribution and use constant discount rates. Panel B shows the revised IWG estimates from the
2013 report. Panel C shows the results of the DICE-2013R model using a constant discount rate
and a consistent modeling framework as described in the text. The rst two lines use the Working
Group average TSC of 3.5, while the third line uses the DICE-2013R TSC of 2.9. Panel D uses the
standard endogenous declining discount rate of the DICE-2013R model. With an average discount rate
to 2100 of 4.2% per year, the estimates are the same as the estimates using the constant discount rate
approach in panel C.
* Note that the Interagency Working Group did not present estimates for a 4% or 4.2% discount
rate. These numbers for the Interagency Working Group are interpolated between the 3% and 5% estimates to compare with other estimates. The 2015 estimates are interpolated from the given gures in
the 2010 and 2013 reports.

Nordhaus
297
economic and geophysical estimates and TSC and the standard declining
discount rate on goods (seen in g. 4). The estimated SCC is the same as
the IWGs constant-discount-rate approach for a constant discount rate of
4.2% per year.
What is the bottom line? These results indicate that the revised IWG and
the revised DICE estimates are virtually identical conditional on the
discount rate. (Compare the average of the three models in the line 4 of
panel B with line 3 of panel C.)
Taking all these results together, ve major points stands out. First, the most
recent IWG estimates of the SCC are close to the 2013R DICE model conditional
on the discount rate. While there are numerous methodological and economic differences, the nal products are within 10% of each other.
Second, there remain substantial differences in panels A and B of table 3 among
the modelseven after harmonizing for emissions, output, population, and temperature sensitivity. While the IWG was unable to determine the source of the differences, it is likely that a major component is the damage function. For example, FUND
has negative damages up to 2C, which suggests why the FUND SCC is very low
(in the 2013 estimate) or negative (in the 2010 estimate) at high discount rates. The
PAGE model assumes catastrophic damages in the tails, which leads to substantially
higher estimates of the SCC in the most recent vintage. These results suggest that a
major remaining task is to compare models with a harmonized damage function.
Third, a remaining issue in the IWG methodology is the risk of making implicit
assumptions that would be questionable if made explicitly. One example was the
decision to make independent assumptions about economic growth and discounting. Virtually all economic models would link the two, but the IWG took them as
independent. One result is that the growth-corrected discount rates differ greatly
across scenarios. At the low end of the discount range (2.5%), three of the ve scenarios (IMAGE, MESSAGE, and 550 ppm [parts per million] average) have negative growth-corrected discount rates over the period 20002050; this assumption
can induce unbounded present values if extrapolated. Additionally, the IWG generally relied on US rates of return on investment but then calculated global welfare
and damage estimates. The underlying problem is that the harmonization undertaken by the IWG took place outside of the models rather than inside the models.
That is, the modications and harmonizations were made externally and were not
linked to structural parameters in the models. In future work, it would be better to
integrate the underlying assumptions.
Fourth, the IWG makes much of including an elaborate Monte Carlo exercise in
the temperature sensitivity coefcient (TSC). The IWG did not indicate whether uncertainty about the TSC makes any difference for estimates of the SCC. While that
question will differ across models, I have run a simple experiment for the DICE-

298
March/June 2014
2013R model estimating the impact of a mean-preserving spread of TSC on the

SCC. The result is that the value of the SCC is virtually unaffected by uncertainty
about the TSC. For example, a binomial distribution with a standard deviation of
1C has a SCC of 0.7% higher than the certainty equivalent. At least for the DICE
model, introducing uncertainty about the temperature sensitivity has essentially no
impact on the SCC.
Finally, the role of the discount rate is again central to the differences in the estimates of the SCC. All the alternative estimates show the continuing difculties
from lack of clarity on how to discount future damages.
4. CR IT IQ UE S A ND MA J O R O PE N I SS UE S FO R THE SC C
The concept and estimates of the SCC have been widely criticized on several grounds.
The criticisms fall into two general categories. First are those holding that the models exclude major dangerous or potentially catastrophic impacts of climate change. In
these, the SCC is thought to be underestimated, perhaps dramatically so. A second
category holds that the models and estimates are so awed that the SCC estimates
are worthless. I address each of these in turn.
4.1. Tipping Points, Fat Tails, and Potential Catastrophes
There have been several critiques of the SCC calculations in the DICE model (as
well as FUND and PAGE), arguing that the models omit signicant issues. For a
particularly sharp critique, see Ackerman and Stanton (2010, 2012).
One of the most vexing issues in climate change is the potential for abrupt, irreversible, or catastrophic climate change (see National Research Council, Committee on Abrupt Climate Change 2002; Lenton et al. 2008; IPCC Extremes 2012;
IPCC Fifth Assessment, Science 2013). Estimates for the economic costs of such scenarios are included conceptually in the damage estimates in the DICE-2013R model.
However, the model does not deal explicitly with tipping elements, primarily because
these have not been reliably determined. It must be emphasized that there is virtually
no basis for determining the size, timing, or probability of such events or the economic damages that would ensue. Work on investigating the impact on policy in the
DICE model of thresholds with and without irreversibilities is underway, but this
project involves many scientic, economic, and algorithmic obstacles.
Another difculty that has not been fully addressed in the literature is the potential for highly skewed distributions of uncertain variables and potentially unbounded values of the SCC. This issue was raised in Weitzman (2009) and has
been discussed extensively in the literature (see the symposium in Review of Environmental Economics and Policy, summer 2011, for several articles, as well as Weitzman
[2013]). A simplied way to state Weitzmans argument is that the combination of
fat tails and strong risk aversion may lead to large losses in expected welfare. As a
result, the SCC may be unbounded or extremely large.

Nordhaus
299
The impact of catastrophic damages or a fat-tailed distribution of important

parameters can have major impacts on the SCC. Take as a simple example a case
with catastrophic damages above a threshold temperature increase T*. To be concrete, assume that the catastrophic damage function is highly convex and takes the
form D/Y = 0.01(T/T*)8 (this being the eighth power, not a footnote). With a
threshold of T* = 2C, this leads to damages equal to 20% of output at 3C and
72% of output at 4C. While useful for illustrative purposes, currently no plausible
damage estimates would approach these numbers.
Table 4 shows the SCC with and without abatement policies for different
thresholds. An important and generally neglected feature of catastrophic damages is
that the catastrophic social cost of carbon depends critically on whether adequate
policies are taken. For the threshold of 2C and no climate change policies, the
SCC is $1,046 per ton CO2. However, when optimal policies are taken immediately, the social cost of carbon is only $54 per ton of CO2. This resultthat fat
tails or catastrophic damages generally have extreme results on the SCC only when
good policies are not takenis a warning for both policy and analysis. (For a more
detailed discussion and further examples, see Nordhaus [2012].)
4.2. Warming and Modeling Skeptics
The issues surrounding the debates about the existence, extent, and causes of climate change are beyond the scope of this article. They are fully addressed in the
IPCC Fifth Assessment, Science (2013), which presents a detailed review of the science and evidence. This review indicates that the basic science and trends are well
established but that there remain vast uncertainties about the scope, scale, and regional dimensions of climate change.
Criticisms aimed specically at IAMs and the SCC are exemplied by a recent
article by Pindyck (2013). He concludes that IAMs are of little or no value for
Table 4. Social Cost of Carbon with Catastrophic Threshold, with and without
Climate Policy
Social Cost Carbon (2015 in 2005$)
Threshold Temperature, T*
1.5C
2C
3C
4C
With Optimal Policy
With No Policies for 100 Years
125
54
24
19
1,495
1,046
197
33
Note.The cases without policy assume no abatement for a century. The cases with policy assume
immediate optimal abatement. The catastrophic damage function assumes that the damage-output ratio
is 0.01(T(t)/T*)8, where T* is the catastrophic threshold.

300
March/June 2014
evaluating alternative climate change policies and estimating the SCC (870). It will
be useful to review the Pindyck critiquem as it states, albeit in an exaggerated form,
many of the reservations about calculations of the SCC.
Pindycks attack on IAMs touches both empirical and conceptual issues. Beginning with the empirical questions, he highlights (1) the social preference function,
particularly the discount rate, (ii) the damage function, (iii) the potential for catastrophic changes, and (iv) the temperature sensitivity to greenhouse gas increases.
The centrality of these issues for both IA models and the SCC has been clear
for many years. Let us begin with the issue of discounting. Nordhaus (1991) and
Cline (1992) showed the sensitivity of the SCC to the discount rate. The dilemma
of the prescriptive versus the descriptive views was described in an excellent review
in an early IPCC chapter (Arrow et al. 1996). The role of discounting was carefully addressed in a series of studies in a volume edited by Portney and Weyant
(1999). I showed above the role of the discount rate in affecting the calculations in
the DICE-2013R model. Unsettled issues about discounting are not an excuse for
discounting the SCC.
A second important issue discussed by Pindyck is the issue of the catastrophic
impacts of climate change. Pindyck states that the models ignore the possibility of a
catastrophic climate outcome (869). This is simply wrong. The DICE model contains an adjustment for nonmonetized and potentially catastrophic damages; the
PAGE model has a specic treatment for catastrophic damages and for the uncertainties of those; the FUND model has elements of catastrophic damages, such as sealevel rise and agricultural losses. There have been many studies of the impact of catastrophic impacts and its implications for the SCC (as described earlier in this section).
Pindyck continues, IAMs cannot tell us anything about catastrophic outcomes
(869). Well, of course they cannot. The issue is not the models. The issue is
that scientic understanding about the likely trajectory and impact of catastrophic
impacts is still highly conjectural. For example, natural scientists are highly uncertain
about the threshold and timing of the disintegration of the Greenland ice sheet. The
latest IPCC review of the issue concluded that warming beyond a threshold greater
than 2C but less than 4C of global warming would lead to the near-complete
loss of the Greenland Ice Sheet over a millennium or longer (IPCC Fifth Assessment, Science 2013, TS-37). This provides little guidance for IAMs.
Another important uncertain factor is climate sensitivity to CO2 increases. The
recent IPCC review concludes that 1.5C4.5C is assessed to be the likely range of
equilibrium climate sensitivitythis being exactly the range found in the Charney
Report of 1979 (National Research Council 1979, TS-33).
Yet a further insight on catastrophic damages comes from studying the IPCC
Fifth Assessment Report. This report reviews several components of the earth system susceptible to abrupt changes. These include the strength of the Atlantic Meridional Overturning Circulation, clathrate methane release, tropical and boreal forest

Nordhaus
301
dieback, disappearance of summer sea ice in the Arctic Ocean, long-term drought,
and monsoonal circulation. The report concluded that in general there is low condence and little consensus on the likelihood of such events over the 21st century
(TS-35). Without reliable estimates of the geophysical aspects, economic models can
hardly make reliable estimates about the impact and costs of catastrophic, abrupt, or
irreversible events.
While Pindycks observations about the empirical weaknesses of IAMs or calculations of the SCC are worthy of careful study, the conclusion that IAMs are therefore
useless fundamentally misconceives the enterprise. IAMs and the SCC are conceptual
frameworks for dealing with highly complex, nonlinear, dynamic, and uncertain systems. The human mind is incapable of solving all the equations simultaneously, and
modeling allows making if . . . , then . . . analyses of the impacts of different factors.
The models have provided important insights into many aspects of climate change
policy. The idea of the social cost of carbon is a natural consequence of considering
the economics of a global externality as well as policies to correct for this market
failure.
IAMs have improved our understanding of the importance of cost effectiveness,
the value of market instruments as compared to command and control, the value of
information about new technologies and improved science, the importance of broad
participation, the potential volatility of cap-and-trade systems, and the costs of alternative approaches to reducing emissions. Perhaps the most important contribution
is the ability of systematic modeling to highlight the critical issues (such as discounting and damages) and to bring new ndings into account in an orderly fashion.
A useful analog is the concept of the scal multiplier in macroeconomics. Even
after decades of research, economists are unsure about the exact value of the scal
multiplier in recessions. But that does not imply that scal policy or macroeconomic
analysis is useless. Similarly, even though the IA models and estimation of the SCC
raise daunting scientic and economic issuessome unresolvedthis does not diminish their importance for climate change economics and policy.
5 . CU R R E N T A P PL I CA T IO N S O F T H E S C C I N CL IM A T E
CHA N G E P O L I C IE S
Estimates of the SCC are a critical ingredient in climate change policy. They currently play two roles. At the broadest level, they provide guidance as to the appropriate level of emissions controls or carbon pricing in policies to reduce emissions
either at a national or an international level. For example, policy makers can use
estimates of the SCC to determine the optimal carbon taxes or the target rate of
emissions reductions under a cap-and-trade regime.
A second application is for rule making where countries do not have comprehensive policies covering all GHGs. In this context, regulators might use the SCC
in a calculation of social costs and benets of policies involving energy or climate-

302
March/June 2014
affecting decisions. I will illustrate this for the United States, although parallel policies have been used in the United Kingdom and the European Union.
Under US law and practice, major regulations must be accompanied by a regulatory impact analysis, or RIA.7 While it is not generally necessary that regulations
pass a benet-cost test, regulators have paid close attention to the analyses. Prior to
2008, the valuations of greenhouse gas (GHG) emissions were not included in
monetary estimates of costs and benets. As a result of a 2007 Supreme Court decision, the US EPA (Environmental Protection Agency) was required to regulate
CO2 and other GHG emissions as air pollutants under the Clean Air Act. After
an initial set of uncoordinated estimates, the US federal government established the
Interagency Working Group that reviewed and determined a set of estimates of the
SCC to be used in federal regulations (IWG 2010). I discussed the methods and
results of the IWG in a previous section.
As of January 2014, there were 58 proposed or nal rules that have used the
SCC in calculating benets. Table 5 provides a list of proposed or nal regulations
for the United States through 2012.8 In addition to the agency, date, and name of
the rule, the table shows both the total benets and the share of total benets from
reducing CO2 emissions. The rules shown are estimated to be of great signicance,
with total benets in the rules over the 3 years totaling $520 billion. The CO2 benets totaled $92 billion. The last column shows the percent of total monetized benets that come from including climate benets. Particularly important are regulations
such as fuel economy standards and new stationary sources (primarily power plants).
It is useful to illustrate the calculations for a specic rule, the 2012 standard for
new cars and light trucks as estimated in the RIA. (For this calculation, we provide
EPA/DOTs preferred methodology, which includes fuel savings; see EPA, Vehicles
2010.) Over model years 201725, the regulation is estimated to cost $150 billion.
Fuel savings and other benets realized by owners are estimated to total $556 billion, so the regulation is estimated to benet owners even without any external considerations. Pollution benets total $52 billion, of which $47 billion are from reduced CO2 emissions at a SCC averaging $35 per ton of CO2 and $5 billion are
from reduced local air pollution. So for this regulation, it would pass a benet-cost
test primarily on the basis of behavioral benets (consumers overdiscounting fuel
7. The source of regulatory analyses dates to the 1970s. The current relevant rule is Executive Order 12866, which states that agencies must, to the extent allowed by statute, assess both the costs and the benets of the intended regulation and . . . propose or adopt a
regulation only upon a reasoned determination that the benets of the intended regulation
justify its costs (OMB 2003).
8. I am grateful to Elizabeth Kopits of the Environmental Protection Agency for providing the total list and the shorter list, which have benets estimates in table 5.

Nordhaus
303
savings). It would not pass a benet-cost test on the basis of external costs alone.
Roughly 90% of the external benets are CO2 emissions.
Regulations generally set the cost of externalities at the marginal damage of a pollutant. For example, in each of the cost-benet studies listed in table 5, the climate
change costs are calculated using the SCC without any adjustments. Three important issues that arise are the adjustments due to leakage from incomplete harmonization, the adjustments for existing distortions of the tax system, and the issue of
whether to apply global or domestic SCC in regulations.
5.1. Leakage and Incomplete Harmonization
The rst adjustment would need to recognize that the SCC is applied to only a
small part of the national economy and the global economy, and this fact will generally lead to a leakage discount. For example, a rough estimate is that the implicit carbon pricing through regulations that include the SCC in the United States
applies to less than one-tenth of new tangible investment and to virtually no existing capital. The importance of incomplete harmonization is recognized in the extensive literature on leakage, but the implications of leakage for policy applications
of the SCC in government analysis and regulation have been largely ignored. There
is no mention of greenhouse gas leakage in any of the major EPA rule makings or
in the US Working Group calculations of the SCC.
An analysis in Baylis, Fullerton, and Karney (2014) shows that the extent of leakage depends on the substitution properties in production and consumption as well
as on the carbon intensity of the different sectors. An example using the abatement
structure of the DICE model will illustrate the point.
Suppose that national output is produced with two goods, A and B, with equal
CO2 intensity and constant costs of production. The two goods have identical constant elasticities of substitution in demand (). Marginal damages are assumed to
be $20 per ton of CO2 and are invariant to the level of emissions because we are
dealing with a stock pollutant. The regulatory authority can apply a tax on sector
A but cannot tax sector B. Sector A might be automobiles, while sector B is air
travel. Assuming that the tax on CO2 in sector B is zero, we can calculate the optimal tax in sector A. With constant costs of production in each sector, the optimal
tax on sector A will be $20 per ton with no substitution ( = 0), while it would be
zero with perfect substitution ( = ). A calculation shows that the optimal tax
in sector A would be about $2 per ton (or 10% of the SCC) if the = 4, while it
would be about $15 per ton (or 75% of the SCC) if the = 1=2.
However, as Baylis et al. show, the result is sensitive to the assumptions. Take
an alternative where the emissions intensity is much higher in the regulated than
the nonregulated sector, as, for example, with power plants. Here, the optimal tax
would be very close to the SCC. All of this is quite intuitive, even if the exact magnitudes will depend upon the details of the technology and preferences. Unfortunately,


Final Cross-State Air Pollution Rule (CSAPR)

(vacated by courts, in review)
Final Light Duty Vehicle GHG Standards (201725)
Final Light Duty Vehicle GHG Standards (201216)
Final MATS Rule
Final Boiler MACT (CO2 disbenets)
Proposed GHG Standards for New Stationary
Source EGUs
Final Cement NESHAP/NSPS (CO2 disbenets)
(under reconsideration)
Final ECS for Residential Refrigerators and Freezers
Final Medium-Heavy Duty Vehicles GHG Standards
Final ECS for Fluorescent Lamp Ballasts
Final ECS for Residential Water Heaters
Proposed ECS for Battery Chargers and External
Power Supplies
Final ECS for Small Electric Motors
Final ECS for Residential Dishwashers
Final Sewage Sludge Incinerators NSPS/Emissions
Guidelines (CO2 disbenets)
Proposed (supplemental) NESHAP: Mercury Cell
Chlor-Alkali Plants Amendments
EPA
3/14/2011
4/9/2010
5/30/2012
3/21/2011
9/15/2011
9/15/2011
11/14/2011
4/16/2010
3/27/2012
9/9/2010
10/15/2012
5/7/2010
2/16/2012
12/20/2012
4/13/2012
8/8/2011
Date
.0
.1
.1
.0
9.9
7.3
1.3
.2
.1
11.5
126.0
58.4
57.7
42.5
21.4
183.3
.0
.1
.1
(.0)
9.0
5.7
1.3
.2
.1
.0
46.6
17.0
.4
(.0)
11.0
.6
Total CO2 Benets

(Billions of $)
21.1
95.2
96.6
.3
91.0
78.1
96.6
95.3
94.1
.2
37.0
29.1
.6
.1
51.3
.3
CO2 as Share of Total

Monetized Benets
(%)
Note.The estimates in this table generally exclude fuel or energy savings from benets. They are based on the governments central SCC value from the 2010 study of $22
per ton for 2015 reductions in 2005$. EPA = Environmental Protection Agency; DOT = Department of Transportation; DOE = Department of Energy; GHG = greenhouse
gases; MATS = Mercury and Air Toxics Standards; NESHAP = National Emission Standards for Hazardous Air Pollutants; NSPS = New Source Performance Standards;
ECS = Energy Conservation Standards; NESHAPS = National Emission Standards for Hazardous Air Pollutants. Data were provided by the staff of the US Environmental
Protection Agency.
EPA
DOE
DOE
EPA
DOE
EPA/DOT
DOE
DOE
DOE
EPA
EPA/DOT
EPA/DOT
EPA
EPA
EPA
Rule
Agency
Total Benets
(Billions of $)
Table 5. Data for Major US Regulations Including a Social Cost of Carbon in Estimating Benets
Nordhaus
305
estimating the optimal second-best carbon price in the face of incomplete harmonization is extremely complex.9 This is an important piece of unnished business.
5.2. Existing Tax Distortions
A second qualication concerns existing economic distortions, particularly those involving taxes. As has been extensively discussed in the environmental literature (e.g.,
see Bovenberg and Goulder 1996), the presence of existing tax distortions will generally lead to an optimal pollution tax below the ideal Pigovian tax; for conciseness, I
will call this the distortion discount. In the present context, this implies that an
optimal carbon tax would be below the SCC. The reason is that imposing a carbon
tax raises the prices of goods and increases the tax wedge between taxed and nontaxed
sectors. The extent of the divergence is extremely complicated, as has been shown
in Barrage (2013). Using the economic and climate parameters of the DICE model
combined with a simple tax system, Barrage nds a range of distortion discounts up
to one-third, with the size of the discount depending upon the structure of existing
tax distortions as well as the way the revenues are used or recycled (also see Jorgenson
et al. 2013).
The distortion discount will be even larger when climate policies are implemented
with regulatory emissions limitations rather than taxes. Or, more generally, as Fullerton and Metcalf (2001) show, a discount would apply when policies create rents
that are not appropriated by the government. This implies that a distortion discount
should apply to cap-and-trade regulations in which allowances are allocated. Indeed,
a distortion discount would appear to apply to each of the regulations shown in table 5. The reasoning behind this result is that rent-creating regulations increase the
wedge between the taxed and nontaxed sectors just as pollution taxes do, but, raising
no revenues, they cannot reduce distortions by reducing other distortionary taxes. In
an analogous environmental market, Goulder, Parry, and Burtraw (1997) showed
that the costs of reducing sulfur dioxide emissions under Title IV of the 1990 Clean
Air Act could have been reduced by about 25% if the tradable permits had been
auctioned (and revenues recycled) rather than given out for free. Under some conditions, the distortions are greater than the environmental benets.
At present, there are no reliable estimates of the distortion discounts that
should be applied to regulations that apply the SCC, but they are clearly larger
than the tax distortion discounts, such as calculated by Barrage cited above.10 The
EPA has reviewed the issue of tax distortions and concluded that it was prema9. See Copeland and Taylor (2000) for an analysis showing how international trade introduces additional complexities. Felder and Rutherford (1993) show that leakage can be negative if carbon taxes affect fuel switching to high-carbon-intensity fuels.
10. This is discussed in Goulder, Parry, and Burtraw (1997), Goulder et al. (1999),
Parry, Williams, and Goulder (1999), Fullerton and Metcalf (2001), and Murray, Keeler,
and Thurman (2005).

306
March/June 2014
ture to make any adjustment given the current state of knowledge and existing
doubts about current results and their broad applicability (Murray, Thurman, and
Keeler 2000). Here, apparently, the EPA has decided that no number is better than
an uncertain number, an approach seldom followed in its regulatory philosophy.
5.3. National versus Global Damages in a Regulatory Framework?
In most regulatory proceedings, a benet-cost analysis would compare the national
benets and the national costs. For global externalities, the question arises as to the
domain of the economic analysis. Should the costs, benets, and discount rates refer
to the United States alone or to the entire globe?
The US regulatory framework is clear that the focus should be domestic. In the
Ofce of Management and Budgets (OMBs) guidance to federal agencies on the
development of regulatory analysis, it states that the analysis should focus on benets and costs that accrue to citizens and residents of the United States. Where you
[i.e., the regulator] choose to evaluate a regulation that is likely to have effects beyond the borders of the United States, these effects should be reported separately
(OMB 2003).
This question arose in the determination of whether CO2 endangered public
health or welfare and should therefore be regulated under the Clean Air Act (the
endangerment nding). The EPA stated that it considered foreign impacts only
insofar as they had an effect on domestic health and welfare: EPAs consideration
of international effects for purposes of determining endangerment is limited to how
those international effects impact the health and welfare of the US population
(EPA, Endangerment 2009).
The reasoning behind the endangerment nding was inconsistent with the application of the SCC in regulations. In applying the SCC, EPA has used a domestic
measure of costs and a global measure of benet; benets include foreign as well
as domestic benets of emissions reductions. This practice was based on the recommendations of the US Interagency Working Group:
The interagency group concluded that a global measure of the benets from
reducing US emissions is preferable because the climate change problem is
highly unusual in at least two respects. First, it involves a global externality. . . .
Consequently, to address the global nature of the problem, the SCC should
incorporate the full (global) damages caused by GHG emissions. Second, climate change presents a problem that the United States alone cannot solve.
Even if the United States were to reduce its greenhouse gas emissions to zero,
that step would be far from enough to avoid substantial climate change. (IWG
2010)
These reasons do not address the legal or economic grounds for using global
benets in a cost-benet analysis. The rst reason is a simple statement about the

Nordhaus
307
global nature of climate change. The second is a factual statement about the
inability of a single country acting alone to solve the problem.
The legal question of whether global benets can or should be used in US regulatory framework is unsettled. The answer depends in part on the statutory framework because the requirements for balancing costs and benets differ across statutes.
From a practical point of view, the use of global rather than national valuations
would be a major departure for most countries. In almost all areas, countries design
policies that benet their own residents and (except where there are treaties or
major harmful spillovers) ignore the impacts on other nations. Whether we examine trade, agriculture, military, tax, or macroeconomic policies, there are few cases
where the welfare in other regions weighs as heavily as that of citizens and residents. To take an extreme example, income transfers to low-income persons in the
United States average $3,000, while those outside the United States average about
$3 per person.11 The proposed use of a global SCC would weigh domestic and foreign impacts equally.
Are there economic or strategic reasons to use global benets for domestic regulatory purposes? From an ethical point of view, taking a global welfare perspective is a
welcome departure from the usual norm of nationalistic policies. It is not clear
whether this stance would long survive legislative scrutiny unless there are other
grounds.
A more convincing rationale would be strategic. There might be grounds for using
a global SCC as a signaling device, wherein it would serve as a golden rule for regulators that would lead to a globally efcient solution if universally followed. However,
doing that through uncoordinated actions would appear an inefcient device compared to explicit international agreements on harmonizing regulatory SCCs.
It should be emphasized that the decision to use global rather than national SCC
is of great empirical importance. The regional DICE model has a SCC for the
United States that is about 10% of the global SCC, and that number is approximately the same for the two other major regional models (FUND and PAGE).
This is clearly a key issue for climate change policy.
5.4. Conclusion on Qualications
While the SCC of slightly below $20 per ton of CO2 would be an appropriate target for a harmonized carbon tax in a rst-best world, it is likely to be higher than
the appropriate number when realistic factors are included. Using national rather
than global benets would reduce the number by an order of magnitude. Tax dis-
11. OMB (2013), table 3.2Outlays By Function and Subfunction, FY 2011. The
domestic budget includes functions 550, Health (excluding Medicare), and 600, Income
Security. The foreign budget includes function 151, International Development and Humanitarian Assistance.

308
March/June 2014
tortions, the use of non-revenue-raising regulations, and leakage would also lead to
the use of a smaller regulatory price. While the sign of these changes is clear, because of complexities of the regulatory, tax, and energy systems, the size of the
adjustment is highly uncertain.
6. CONC L U SION
The present study presents a new set of estimates of the social cost of carbon. The
distinguishing features of the present modeling are the following. First, it presents
an updated version of the DICE model, including revisions of all economic and
geophysical variables and modules. Second, it is a general equilibrium approach in
which important variables (such as the real interest rate, the economic growth rate,
consumption, and climatic variables) are determined endogenously rather than as exogenous assumptions.
The most important results are as follows. First, the estimated social cost of carbon for the current time (2015) is $18.6 per ton of CO2 in 2005 US international
prices. Second, the DICE model results are lower than one of the other two major modeling estimates (the PAGE model) but higher than the other (the FUND
model). Third, the major open issue concerning the SCC continues to be the appropriate discount rate. Fourth, use of the SCC in regulatory policies in the energy sector is increasingly important, but analyses of its application have not sufciently considered issues such as leakage, distortionary taxes, and the question of whether to
use global or domestic SCCs.
REFERENCES
Ackerman, Frank, and Charles Munitz. 2012. Climate damages in the FUND model: A disaggregated
analysis. Ecological Economics 77:21924.
Ackerman, Frank, and Elizabeth A. Stanton. 2010. The social cost of carbon: A report for the Economics
for Equity and the Environment Network. Stockholm Environment Institute, US Center (April 1).
http://www.climate-economics.org/papers/SocialCostOf Carbon_SEI_20100401.pdf.
. 2012. Climate risks and carbon prices: Revising the social cost of carbon. Economics: The OpenAccess, Open-Assessment E-Journal 6, no. 201210.
Anthoff, David. 2013. Personal communication with current estimates of regional SCC in the FUND
model.
Anthoff, David, and Richard S. J. Tol. 2010. On international equity weights and national decision
making on climate change. Journal of Environmental Economics and Management 60, no. 1:1420.
Arrow, K. J., W. R. Cline, K-G. Maler, M. Munasinghe, R. Squitieri, and J. E. Stiglitz. 1996.
Intertemporal equity, discounting, and economic efciency. In Climate change, 1995: Economic and social dimensions of climate change, contribution of Working Group III to the Second Assessment Report of
the Intergovernmental Panel on Climate Change, ed. James P. Bruce, Hoe-song Yi, and Erik F. Haites,
12544. Cambridge: Cambridge University Press.

Nordhaus
309
Arrow, K. J. J., M. Cropper, C. Gollier, B. Groom, G. Heal, R. Newell, W. Nordhaus, R. Pindyck, W.

Pizer, P. Portney, T. Sterner, R. S. J. Tol, and M. Weitzman. 2013. Determining benets and costs
for future generations. Science 341, no. 6144:34950.
Barrage, Lint. 2013. Optimal dynamic carbon taxes in a climate-economy model with distortionary scal
policy. PhD diss., Yale University.
Baylis, Kathy, Don Fullerton, and Daniel H. Karney. 2014. Negative leakage. Journal of the Association
of Environmental and Resource Economists 1:5173.
Bosetti, Valentina, Emanuele Massetti, and Massimo Tavoni. 2007. The WITCH Model: Structure,
baseline, solutions. Working Paper no. 010.2007. FEEM (Fondazione Eni Enrico Mattei), Milan.
Bovenberg, A. Lans, and Lawrence H. Goulder. 1996. Optimal environmental taxation in the presence
of other taxes: General equilibrium analyses. American Economic Review 86, no. 4:9851000.
Clarke, Leon, Jae Edmonds, Volker Krey, Richard Richels, Steven Rose, and Massimo Tavoni. 2009.
International climate policy architectures: Overview of the EMF 22 International Scenarios. Energy
Economics 31, suppl. 2 (December 2009): S64S81.
Clarkson, Richard, and Kathryn Deyes. 2002. Estimating the social cost of carbon emissions. Government Economic Service Working Paper 140, Enquiry Unit, HM Treasury, London.
Cline, William. 1992. The economics of global warming. Washington, DC: Institute of International
Economics.
. 1997. Modeling economically efcient abatement of greenhouse gases. In Environment, energy,
and economy, ed. Yoichi Kaya and Keiichi Yokobori, 99122. Tokyo: United Nations University
Press.
Copeland, Brian, and R. M. Scott Taylor. 2000. Free trade and global warming: A trade theory view of
the Kyoto Protocol. NBER Working Paper 7657 (April), National Bureau of Economic Research,
Cambridge, MA.
EPA (Environment Protection Agency), Endangerment. 2009. Environmental Protection Agency, 40
CFR, chapter 1: Endangerment and cause or contribute ndings for greenhouse gases under Section
202(a) of the Clean Air Act; nal rule, Federal Register, Tuesday, December 15, pt. 5, 66514.
EPA, Vehicles. 2010. US Environmental Protection Agency nal rulemaking to establish light-duty
vehicle greenhouse gas emission standards and corporate average fuel economy standards: Regulatory
impact analysis, EPA-420-R-10009, April.
Felder, Stefan, and Thomas F. Rutherford. 1993. Unilateral CO2 reductions and carbon leakage: The
consequences of international trade in oil and basic materials. Journal of Environmental Economics and
Management 25, no. 2:16276.
Fullerton, Don, and Gilbert E. Metcalf. 2001. Environmental controls, scarcity rents, and pre-existing
distortions. Journal of Public Economics 80, no. 2:24967.
FUND. 2013. The Climate Framework for Uncertainty, Negotiation and Distribution (FUND). Documentation at www.fund-model.org.
Gollier, Christian. 2013. Pricing the planets future: The economics of discounting in an uncertain world.
Princeton, NJ: Princeton University Press.
Golosov, Mikhail, John Hassler, Per Krusell, and Aleh Tsyvinski. 2014. Optimal taxes on fossil fuel in
general equilibrium. Econometrica 82, no. 1:4188.
Goulder, Lawrence H., Ian W. H. Parry, and Dallas Burtraw. 1997. Revenue-raising versus other approaches to environmental protection: The critical signicance of preexisting tax distortions. RAND
Journal of Economics 28, no. 4:70831.

310
March/June 2014
Goulder, Lawrence H., Ian W. H. Parry, Roberton C. Williams III, and Dallas Burtraw. 1999. The
cost-effectiveness of alternative instruments for environmental protection in a second-best setting.
Journal of Public Economics, 72, no. 3:32960.
Greenstone, Michael, Elizabeth Kopits, and Ann Wolverton. 2011. Estimating the social cost of carbon
for use in U.S. federal rulemakings: A summary and interpretation. NBER Working Paper 16913,
National Bureau of Economic Research, Cambridge, MA.
. 2013. Developing a social cost of carbon for US regulatory analysis: A methodology and interpretation. Review of Environmental Economics and Policy 7, no. 1:2346.
Hedges, Larry V., and Ingram Olkin. 1985. Statistical methods for meta-analysis, San Diego, CA: Academic Press.
Hope, Chris. 2011. The social cost of CO2 from the PAGE09 Model. Economics Discussion Papers,
no. 2011-39 (September), Kiel Institute for the World Economy.
IPCC (Intergovernmental Panel on Climate Change) Extremes. 2012. Managing the risks of extreme events
and disasters to advance climate change adaptation: Special report of the intergovernmental panel on climate
change, ed. Christopher B. Field, Vicente Barros, Thomas F. Stocker, and Qin Dahe. Cambridge:
Cambridge University Press. http://www.ipcc.ch/pdf/special-reports/srex/SREX_Full_Report.pdf.
IPCC Fifth Assessment, Science. 2013. Intergovernmental Panel on Climate Change, climate change, 2013:
The physical science basis, contribution of Working Group I to the Fifth Assessment Report of the IPCC.
http://www.ipcc.ch/report/ar5/wg1/#.Ukn99hCBm71.
IPCC Fourth Assessment, Science. 2007. Intergovernmental Panel on Climate Change, climate change,
2007: The physical science basis, contribution of Working Group I to the Fourth Assessment Report of the
IPCC, Cambridge: Cambridge University Press. http://ipcc-wg1.ucar.edu/wg1/wg1-report.html.
IWG (Interagency Working Group). 2010. Interagency Working Group on Social Cost of Carbon, US
Government, technical support document: Social cost of carbon for regulatory impact analysis under
Executive Order 12866. http://www.epa.gov/oms/climate/regulations/scc-tsd.pdf.
. 2013. Interagency Working Group on Social Cost of Carbon, US Government, technical support document: Technical update of the social cost of carbon for regulatory impact analysis under
Executive Order 12866, May 2013. http://www.whitehouse.gov/sites/default/les/omb/ inforeg/social
_cost_of_carbon_for_ria_2013_update.pdf.
. 2013. Technical support document: Technical update of the social cost of carbon for regulatory
impact analysis. US Government, Washington, DC.
Johnson, Laurie T., and Chris Hope. 2012. The social cost of carbon in U.S. regulatory impact analyses:
An introduction and critique. Journal of Environmental Studies and Sciences 2, no. 3:20521.
Jorgenson, Dale W., Richard J. Goettle, Mun S. Ho, and Peter J. Wilcoxen. 2013. Double dividend:
Environmental taxes and scal reform in the United States. Cambridge, MA: MIT Press.
Koopmans, Tjalling C. 1965. On the concept of optimal economic growth. Academiae Scientiarum Scripta
Varia 28, no. 1:175. Available for download at http://cowles.econ.yale.edu/P/au/p_koopmans.htm.
Lenton, Timothy M., et al. 2008. Tipping elements in the Earths climate system. Nature 105, no. 66:
178693.
Li, Zhimin, and William Nordhaus. 2013. The social cost of carbon: Methods and a survey of estimates.
Unpublished manuscript, University of California, Berkeley.
Mehra, Rajnish. 2008. The handbook of the equity risk premium. Amsterdam: Elsevier.
Murray, Brian C., Andrew Keeler, and Walter N. Thurman. 2005. Tax interaction effects, environmental
regulation, and rule of thumb adjustments to social cost. Environmental and Resource Economics 30:7392.

Nordhaus
311
Murray, Brian C., Walter N. Thurman, and Andrew Keeler. 2000. Adjusting for tax interaction effects
in the economic analysis of environmental regulation: Some practical considerations. White paper,
US Environmental Protection Agency, prepared for US Environmental Protection Agency, Ofce of
Air Quality Planning and Standards, Innovative Strategies and Economics Group.
National Research Council. 1979. Carbon dioxide and climate: A scientic assessment. Washington, DC:
National Academies Press.
National Research Council, Committee on Abrupt Climate Change. 2002. Abrupt climate change:
Inevitable surprises. Washington, DC: National Academy Press.
Nordhaus, William D. 1991. To slow or not to slow: The economics of the greenhouse effect. Economic
Journal 101, no. 407:92037.
. 1992. An optimal transition path for controlling greenhouse gases. Science 258 (November 20):
131519.
. 1994. Managing the global commons: The economics of climate change. Cambridge, MA: MIT Press.
. 2007. A review of the Stern review on the economics of climate change. Journal of Economic
Literature 45, no. 3:686702.
. 2008. A question of balance: Weighing the options on global warming policies. New Haven, CT: Yale
University Press.
. 2010. Economic aspects of global warming in a post-Copenhagen environment. Proceedings of
the U.S. National Academy of Sciences 107, no. 26:1172126.
. 2012. Integrated economic and climate modeling. In Handbook of computable general equilibrium
modeling, ed. Peter B. Dixon and Dale W. Jorgenson. Amsterdam: North-Holland.
Nordhaus, William, and Joseph Boyer. 2000. Warming the world: Economic modeling of global warming.
Cambridge, MA: MIT Press.
Nordhaus, William, and Paul Sztorc. 2013. DICE 2013R: Introduction and Users Manual, available
from at http://www.econ.yale.edu/~nordhaus/homepage/, October 2013.
Nordhaus, William, and Zili Yang. 1996. A regional dynamic general-equilibrium model of alternative
climate-change strategies. American Economic Review 86, no. 4:74165.
Nordhaus, William, and Gary Yohe. 1983. Future carbon dioxide emissions from fossil fuels. In Changing climate: Report of the Carbon Dioxide Assessment Committee, 87153. Washington, DC: National
Research Council, National Academy Press.
OMB (Ofce of Management and Budget). 2003. Ofce of Management and Budget: Circular A-4,
September 17. http://www.whitehouse.gov/omb/circulars_a004_a-4.
. 2013. The budget of the United States for scal year 2013. OMB: Washington, DC. Available at
http://www.whitehouse.gov/omb/budget/Historicals.
Parry, Ian W. H., Roberton C. Williams III, and Lawrence H. Goulder. 1999. When can carbon
abatement policies increase welfare? The fundamental role of pre-existing factor market distortions.
Journal of Environmental Economics and Management 37:5284.
Pindyck, Robert S. 2013. Climate change policy: What do the models tell us? Journal of Economic Literature 51, no. 3:86072.
Portney, Paul R., and John Weyant, eds. 1999. Discounting and intergenerational equity. Washington,
DC: Resources for the Future.
Ramsey Frank P. 1928. A mathematical theory of saving. Economic Journal 38:54359.
Roe, Gerard, and Marcia Baker. 2007. Why is climate sensitivity so unpredictable? Science 318, no. 5850:
62932.

312
March/June 2014
Stern Review. 2007. The economics of climate change: The Stern review. Cambridge: Cambridge University
Press.
Stern, Nicholas. 2008. The economics of climate change. American Economic Review 98, no. 2:137.
. 2013. The structure of economic modeling of the potential impacts of climate change: Grafting
gross underestimation of risk onto already narrow science models. Journal of Economic Literature 51,
no. 3:83859.
Tol, Richard S. J. 2008. The social cost of carbon: Trends, outliers and catastrophes. Open-Access,
Open-Assessment E-Journal 2, no. 25:122.
. 2009. The economic impact of climate change. Journal of Economic Perspectives 23, no. 2:2951.
United Nations. 2009. Copenhagen Accord. http://unfccc.int/resource/docs/2009/cop15/eng/l07.pdf.
Weitzman, Martin L. 1998. Why the far-distant future should be discounted at its lowest possible rate.
Journal of Environmental Economics and Management 36, no. EE981052:20108.
. 2009. On modeling and interpreting the economics of catastrophic climate change. Review of
Economics and Statistics 91, no. 1:119.
. 2013. Tail-hedge discounting and the social cost of carbon. Journal of Economic Literature 51,
no. 3:87382.
Yang, Zili. 2008. Strategic bargaining and cooperation in greenhouse gas mitigations: An integrated assessment
modeling approach. Cambridge, MA: MIT Press.


Estimates of The Social Cost of Carbon: Concepts and Results From The DICE-2013R Model and Alternative Approaches

Uploaded by

Copyright:

Available Formats

Estimates of The Social Cost of Carbon: Concepts and Results From The DICE-2013R Model and Alternative Approaches

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Estimates of The Social Cost of Carbon: Concepts and Results From The DICE-2013R Model and Alternative Approaches

Uploaded by

Copyright:

Available Formats

Estimates of the Social Cost of Carbon:

Concepts and Results from the DICE-2013R

This content downloaded from 036.083.077.113 on September 09, 2016 09:36:49 AM

Journal of the Association of Environmental and Resource Economists

This content downloaded from 036.083.077.113 on September 09, 2016 09:36:49 AM