UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
March 21,2007
EPA-SAB-07-006
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subj ect: SAB Advisory on EPA 's Second Generation Model
Dear Administrator Johnson:
In 2004, EPA's Office of Atmospheric Programs (OAP) requested that the Science
Advisory Board (SAB) provide advice on a principal economics-based computer model used by
the EPA to perform analysis of potential U.S. climate change policies. This model, known as the
Second Generation Model (SGM), is a computer program that simulates the economic
interactions and contributions to greenhouse gas emissions of 14 regions of the world. The
model contains detail on the functioning of labor markets, energy fuels markets, and commodity
markets in each region. The SGM is a computable general equilibrium model. Its general-
equilibrium framework attempts to integrate consistently the behavior of these markets both
within and across regions. This type of framework is widely considered to be a very useful
approach for assessing the impacts of alternative climate policy options on the U.S. economy,
and the Panel applauds the Agency for having supported this approach. The SGM model applies
this framework to simulate such climate policies as carbon emissions fees, greenhouse gas
allowance trading, and incentives for accelerated energy conservation. For each policy
simulated, it indicates potential impacts on a range of economic and other variables, including
labor demand, investment, industrial output, GDP, energy use, emissions, and government
revenue. The model indicates policy impacts in both the near term and the long term, but with
greater uncertainty associated with longer time frames due to changing technology, energy
reserves and a host of other factors.
The Second Generation Model Advisory Panel met in its first face-to-face meeting on
February 4, 2005. Since that time, the SGM Advisory Panel has had several discussions with
EPA staff and other developers of the SGM, leading to the production of the enclosed Advisory.
-------
The Advisory contains the SGM Advisory Panel's recommendations for improving the
model. The SGM model has been a significant contributor to past analyses of climate policy.
However, the Panel believes it will not be satisfactory for future policy work without
modification along the lines suggested. The Panel believes that with the recommended
improvements, the SGM would be a significantly more useful tool for climate policy analysis.
The Panel believes that the Agency would be well advised to employ a portfolio of
models rather than relying on any single model, and that the SGM might well deserve a place in
that portfolio. The Panel is not prepared to make a definite recommendation regarding the future
use of the SGM. To be well-founded, such a recommendation would need to stem from a
comparison of net benefits from investments in the SGM and other models - a comparison
beyond the scope of the Panel's charge. The Panel's reluctance to make a recommendation
about future investments simply reflects its lack of information about alternative uses of funds
and is not meant to suggest limitations in the potential for the SGM to become an excellent
policy platform.
The Panel's recommendations pertain to the model's documentation, the empirical basis
and comprehensiveness of the model's data, the model's structure, and the reporting of the
model's output. The Panel's main recommendations include the following:
• Improve the documentation of the model's data, parameters, and structure. It is important
to make clear the empirical basis for the choice of parameter values that influence the
results of the model. It is also important to clarify major aspects of the model structure,
so that the internal consistency of the model can be evaluated.
• Update the model's data set. Much of the existing data set relates to stocks and flows of
economic variables dating back to 1990 or before. For the model to generate more
reliable policy assessments, it is important that the initial conditions or benchmark data
be closer to current conditions. In addition, some consideration should be given to
replacing the several current data sources with a different and more comprehensive data
source, the "GTAP" data set.
• Improve several aspects of the model's structure, as described below. Introducing each
of these model structure changes will substantially improve the model's ability to capture
the impacts of climate change policies.
o The model's current treatment of household behavior does not allow for
theoretically consistent assessments of the impacts of policies on human welfare.
Household behavior needs to be modeled in a way that allows for consistent
assessments of welfare impacts.
o The current specification for industry production opportunities is relatively
inflexible and should be replaced with a more flexible and realistic representation
of production. Without a more flexible specification, the model is likely to give
misleading predictions for the impact of climate policies on employment levels,
investment, and the prices and outputs of various commodities.
-------
o The model's current treatment of international trade is far too rigid in that U.S.
policies do not influence the pattern and volume of trade. A flexible, theoretically
consistent treatment of international trade should be included in the model.
Without these changes the model could give a very distorted picture of the
impacts of climate policies.
o Further detail is needed in the SGM's treatment of the electricity and forestry
sectors, so that users can capture important ways that climate policies can affect
these sectors.
o The model's treatment of greenhouse gas emissions should be improved. In the
current model, climate policies endogenously affect only the emissions of one
greenhouse gas - carbon dioxide. The model should be extended to capture
impacts on emissions of other greenhouse gases.
• Improve the reporting of the model's results. Previous applications of the model have not
sufficiently revealed the extent to which the results are sensitive to changes in various
data or parameter inputs. Such "sensitivity analysis" is crucial to evaluating various
policy options. In addition, the model needs to provide more information about the
uncertainties in the empirical estimates of parameters that drive the model, and about the
associated uncertainties in the simulated policy outcomes.
In summary, the Panel finds that although the SGM model has contributed importantly to
previous climate policy analyses, it requires significant improvements in order to be a fully
credible and effective policy evaluation tool. The Panel believes that the SGM will achieve its
potential as a policy tool only if the recommended changes are made.
Sincerely,
/Signed/ /Signed/
Dr. Granger Morgan, Chair Dr. Lawrence H. Goulder, Chair
EPA Science Advisory Board Second Generation Model Advisory Panel
-------
NOTICE
This report has been written as part of the activities of the EPA Science Advisory Board
(SAB), a public advisory group providing extramural scientific information and advice to the
Administrator and other officials of the Environmental Protection Agency. The SAB is
structured to provide balanced, expert assessment of scientific matters related to problems facing
the Agency. This report has not been reviewed for approval by the Agency and, hence, the
contents of this report do not necessarily represent the views and policies of the Environmental
Protection Agency, nor of other agencies in the Executive Branch of the Federal government, nor
does mention of trade names of commercial products constitute a recommendation for use.
Reports of the SAB are posted on the EPA website at http://www.epa.gov/sab.
IV
-------
U.S. Environmental Protection Agency
Science Advisory Board
Second Generation Model Advisory Panel
CHAIR
Dr. Lawrence Goulder, Shuzo Nishihara Professor of Environmental and Resource Economics,
Department of Economics, Stanford University, Stanford, CA
MEMBERS
Dr. Dallas Burtraw, Senior Fellow, Resources for the Future, Washington, DC
Dr. Carol Dahl, Professor, Division of Economics and Business, Colorado School of Mines,
Golden, CO
Dr. W. Michael Hanemann, Chancellor's Professor, Department of Agricultural and Resource
Economics, University of California, Berkeley, CA
Dr. Glenn W. Harrison, Professor of Economics, University of Central Florida, Orlando, FL
Dr. James Opaluch, Professor, Department of Environmental & Natural Resource Economics,
College of the Environment and Life Sciences, University of Rhode Island, Kingston, RI
Dr. Sergey Paltsev, Research Scientist, Joint Program on Science and Policy of Global Change,
MIT, Cambridge, MA
Dr. William Pizer, Fellow, Resources for the Future, Washington, DC
Dr. Adam Rose, Professor, Geography, Pennsylvania State University, State College, PA
Dr. James Shortle, Professor, Agricultural Economics and Rural Sociology, Pennsylvania State
University, University Park, PA
Dr. Ian Sue Wing, Assistant Professor, Geography, Boston University, Boston, MA
SCIENCE ADVISORY BOARD STAFF
Dr. Holly Stallworth, Designated Federal Officer, EPA Science Advisory Board Staff Office,
Washington, D.C.
-------
U.S. Environmental Protection Agency
Science Advisory Board
BOARD
CHAIR
Dr. M. Granger Morgan, Head and Professor, Department of Engineering and Public Policy,
Carnegie Mellon University, Pittsburgh, PA
SAB MEMBERS
Dr. Gregory Biddinger, Coordinator, Natural Land Management Programs, Toxicicology and
Environmental Sciences, ExxonMobil Biomedical Sciences, Inc, Houston, TX
Dr. James Bus, Director of External Technology, Toxicology and Environmental Research and
Consulting, The Dow Chemical Company, Midland, MI
Dr. Trudy Ann Cameron, Raymond F. Mikesell Professor of Environmental and Resource
Economics, Department of Economics, University of Oregon, Eugene, OR
Also Member: COUNCIL
Dr. Deborah Cory-Slechta, Director, Environmental and Occupational Health Sciences
Institute, Robert Wood Johnson Medical School, University of Medicine and Dentistry of New
Jersey and Rutgers University, Piscataway, NJ
Dr. Maureen L. Cropper, Professor, Department of Economics, University of Maryland,
College Park, MD
Dr. Virginia Dale, Corporate Fellow, Environmental Sciences Division, Oak Ridge National
Laboratory, Oak Ridge, TN
Dr. Kenneth Dickson, Professor, Institute of Applied Sciences, University of North Texas,
Denton, TX
Dr. Baruch Fischhoff, Howard Heinz University Professor, Department of Social and Decision
Sciences, Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh,
PA
Dr. James Galloway, Professor, Department of Environmental Sciences, University of Virginia,
Charlottesville, VA
Dr. Lawrence Goulder, Shuzo Nishihara Professor of Environmental and Resource Economics,
Department of Economics, Stanford University, Stanford, CA
VI
-------
Dr. James K. Hammitt, Professor of Economics and Decision Sciences, Harvard Center for
Risk Analysis, Harvard University, Boston, MA
Also Member: COUNCIL
Dr. Rogene Henderson, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
Also Member: CASAC
Dr. James H. Johnson, Professor and Dean, College of Engineering, Architecture & Computer
Sciences, Howard University, Washington, DC
Dr. Agnes Kane, Professor and Chair, Department of Pathology and Laboratory Medicine,
Brown University, Providence, RI
Dr. Meryl Karol, Professor Emerita, Graduate School of Public Health, University of
Pittsburgh, Pittsburgh, PA
Dr. Catherine Kling, Professor, Department of Economics, Iowa State University, Ames, IA
Dr. George Lambert, Associate Professor of Pediatrics, Director, Center for Childhood
Neurotoxicology, Robert Wood Johnson Medical School-UMDNJ, Belle Mead, NJ
Dr. Jill Lipoti, Director, Division of Environmental Safety and Health, New Jersey Department
of Environmental Protection, Trenton, NJ
Dr. Michael J. McFarland, Associate Professor, Department of Civil and Environmental
Engineering, Utah State University, Logan, UT
Dr. Judith L. Meyer, Distinguished Research Professor Emeritus, Institute of Ecology,
University of Georgia, Athens, GA
Dr. Jana Milford, Associate Professor, Department of Mechanical Engineering, University of
Colorado, Boulder, CO
Dr. Rebecca Parkin, Professor and Associate Dean, Environmental and Occupational Health,
School of Public Health and Health Services, The George Washington University Medical
Center, Washington, DC
Mr. David Rejeski, Director, Foresight and Governance Project, Woodrow Wilson
International Center for Scholars, Washington, DC
Dr. Stephen M. Roberts, Professor, Department of Physiological Sciences, Director, Center for
Environmental and Human Toxicology, University of Florida, Gainesville, FL
Dr. Joan B. Rose, Professor and Homer Nowlin Chair for Water Research, Department of
Fisheries and Wildlife, Michigan State University, East Lansing, MI
vn
-------
Dr. Jerald Schnoor, Allen S. Henry Chair Professor, Department of Civil and Environmental
Engineering, Co-Director, Center for Global and Regional Environmental Research, University
of Iowa, Iowa City, IA
Dr. Kathleen Segerson, Professor, Department of Economics, University of Connecticut, Storrs,
CT
Dr. Kristin Shrader-Frechette, O'Neil Professor of Philosophy, Department of Biological
Sciences and Philosophy Department, University of Notre Dame, Notre Dame, IN
Dr. Philip Singer, Professor, Department of Environmental Sciences and Engineering, School of
Public Health, University of North Carolina, Chapel Hill, NC
Dr. Robert Stavins, Albert Pratt Professor of Business and Government, Environment and
Natural Resources Program, John F. Kennedy School of Government, Harvard University,
Cambridge, MA
Dr. Deborah Swackhamer, Interim Director and Professor, Institute on the Environment,
University of Minnesota, Minneapolis, MN
Dr. Thomas L. Theis, Director, Institute for Environmental Science and Policy, University of
Illinois at Chicago, Chicago, IL
Dr. Valerie Thomas, Anderson Interface Associate Professor, School of Industrial and Systems
Engineering, Georgia Institute of Technology, Atlanta, GA
Dr. Barton H. (Buzz) Thompson, Jr., Robert E. Paradise Professor of Natural Resources Law,
Stanford Law School, and Director, Woods Institute for the Environment, Stanford University,
Stanford, CA
Dr. Robert Twiss, Professor and Emerita, University of California-Berkeley, Ross, CA
Dr. Terry F. Young, Consultant, Environmental Defense, Oakland, CA
Dr. Lauren Zeise, Chief, Reproductive and Cancer Hazard Assessment Branch, Office of
Environmental Health Hazard Assessment, California Environmental Protection Agency,
Oakland, CA
SCIENCE ADVISORY BOARD STAFF
Mr. Thomas Miller, Designated Federal Officer, 1200 Pennsylvania Avenue, NW
1400F, Washington, DC, Phone: 202-343-9982, Fax: 202-233-0643, (miller.tom@epa.gov)
Vlll
-------
TABLE OF ACRONYMS
CES Constant Elasticity of Substitution
GET Constant Elasticity of Transformation
CGE Computable General Equilibrium
CO2 Carbon Dioxide
CRP Conservation Reserve Program
EIA Energy Information Administration
EPA Environmental Protection Agency
ETE Everything Else Sector
GAMS General Algebraic Modeling System
GDP Gross Domestic Product
GHG Greenhouse Gas
GTAP Global Trade Analysis Project
GWP Global Warming Potential
IEA International Energy Administration
IO Input-Output
MACS Marginal Abatement Cost(s)
MCP Measure-Correlate-Predict
MIT EPPA ... Massachusetts Institute of Technology Emissions Prediction and Policy Analysis
OAP Office of Atmospheric Programs
PNNL Pacific Northwest National Laboratory
SAB Science Advisory Board
IX
-------
TABLE OF CONTENTS
Executive Summary 1
Introduction 3
Part I: Recommended Immediate Improvements 5
A. Improvements to Documentation 5
i. General 5
ii. Model Structure 5
Hi. Model Inputs - Data and Parameters 6
iv. Model Outputs and Reporting 8
v. Solution Method 8
B. Initial Improvements to Data and Parameters 9
C. Further Sensitivity Analysis 9
Part II: Recommended Subsequent Improvements 10
A. Further Improvements to the Data 10
i. Updating the Data Set 10
ii. Greater use of GTAP data 10
B. Model Structure Improvements 11
i. Household Utility Modeling and Welfare Calculations 11
ii. Production 12
Hi. International Trade 13
iv. Electricity and Agriculture/Forestry Sectors 17
v. Greenhouse Gas Emissions 19
C. Dealing with Uncertainty 20
i. Applying Monte Carlo Simulation in the Sensitivity Analysis 20
ii. Characterizing Uncertainty in the Presentation of Policy Results 22
D. Validation through "Backcasting" Exercises 22
E. Consideration of Additional Policy Details and Instruments 22
Appendix A: Charge Questions 24
Appendix B: Econometric Studies of Energy Demand Elasticities 27
Appendix C: Examples of Nested Production Structures 29
References 33
-------
Executive Summary
This Advisory contains the Second Generation Model Advisory Panel's
recommendations for improving the model. The Panel believes that the SGM Model is very
impressive and contains many features critical to evaluating U.S. climate change policies. At the
same time, the Panel feels that some important improvements are necessary to make the SGM a
fully credible tool for climate policy assessments.
The Panel's recommendations pertain to the model's documentation, the empirical basis
and comprehensiveness of the model's data, the model's structure, and the reporting of the
model's output. The Panel's main recommendations include the following:
• Improve the documentation of the model's data, parameters, and structure. It is important
to make clear the empirical basis for the choice of parameter values that influence the
results of the model. It is also important to clarify major aspects of the model structure,
so that the internal consistency of the model can be evaluated.
• Update the model's data set. Much of the existing data set relates to stocks and flows of
economic variables dating back to 1990 or before. For the model to generate more
reliable policy assessments, it is important that the initial conditions or benchmark data
be closer to current conditions. In addition, some consideration should be given to
replacing the several current data sources with a different and more comprehensive data
source, the "GTAP" data set.
• Improve several aspects of the model's structure:
o The model's current treatment of household behavior does not allow for
theoretically consistent assessments of the impacts of policies on human welfare.
Household behavior needs to be modeled in a way that allows for consistent
assessments of welfare impacts.
o The current specification for industry production opportunities is relatively
inflexible and should be replaced with a more flexible and realistic representation
of production. Without a more flexible specification, the model is likely to give
misleading predictions for the impact of climate policies on employment levels,
investment, and the prices and outputs of various commodities.
o The model's current treatment of international trade is far too rigid in that U.S.
policies do not influence the pattern and volume of trade. A flexible, theoretically
consistent treatment of international trade should be included in the model.
Without these changes the model could give a very distorted picture of the
impacts of climate policies.
o Further detail is needed in the SGM's treatment of the electricity and forestry
sectors, so that users can capture important ways that climate policies can affect
these sectors.
o The model's treatment of greenhouse gas emissions should be improved. In the
current model, climate policies endogenously affect only the emissions of one
-------
greenhouse gas - carbon dioxide. The model should be extended to capture
impacts on emissions of other greenhouse gases.
Introducing each of these model structure changes will substantially improve the
model's ability to capture the impacts of climate change policies.
• Improve the reporting of the model's results. Previous applications of the model have not
sufficiently revealed the extent to which the results are sensitive to changes in various
data or parameter inputs. Such "sensitivity analysis" is crucial to evaluating various
policy options. In addition, the model needs to provide more information about the
uncertainties in the empirical estimates of parameters that drive the model, and about the
associated uncertainties in the simulated policy outcomes.
In summary, the Panel finds that although the SGM model already has many impressive
features, the recommended improvements are necessary to make it a fully credible and effective
policy evaluation tool.
-------
Introduction
In 2004, EPA's Office of Atmospheric Programs (OAP) requested that the Science
Advisory Board (SAB) provide advice on a principal economics-based computer model used by
the EPA to perform analysis of potential U.S. climate change policies. This model, known as the
Second Generation Model (SGM), is a computer program that simulates the economic
interactions and contributions to greenhouse gas emissions of 14 regions of the world. The
model contains detail on the functioning of labor markets, energy/fuels markets, and commodity
markets in each region. Its general-equilibrium framework is geared toward integrating
consistently the behavior of these markets both within and across regions. The model is
designed to simulate such climate policies as carbon fees, greenhouse gas allowance trading, and
incentives for accelerated energy conservation. For each policy simulated, it indicates potential
impacts on a range of economic and other variables, including labor demand, investment,
industrial output, GDP, energy use, emissions, and government revenue. The model indicates
policy impacts not only in the near term but at various points in the future as well.
An extensive and detailed documentation of SGM's structure, parameters and
assumptions, as well as a shorter overview paper, may be found in the OAP section of the EPA's
web site at http://www.epa.gov/air/sgm-sab.html
Subsequent to OAP's request, the Science Advisory Board Staff Office solicited expertise
in a Federal Register Notice published July 9, 2004. The Second Generation Model Advisory
Panel (the Panel) was formed and met in its first face-to-face meeting on February 4, 2005.
Since that time, the SGM Advisory Panel has had several discussions with EPA personnel and
developers of the SGM. These discussions addressed the charge questions posed by the OAP (at
http://www.epa.gov/sab/pdf/sgm charge questions 111804.pdf, and as Appendix A to this
report) and have led to the production of the enclosed Advisory Report.
In response to initial comments and information requests from the Panel, the SGM team
provided the Panel with three documents:
The SGM: Comparison of SGM and GTAP Approaches to Data Development
The SGM: Data, Parameters, and Implementation
The SGM: Model Description in Theory
The Panel considers these documents to be very useful initial steps toward improving the SGM
modeling effort, particularly as regards model documentation. In this report, we refer to these
three documents respectively as papers 1, 2, and 3.
This Advisory contains the SGM Panel's recommendations for improving the model.
The two main parts of the Advisory separate the recommended improvements that the Panel
believes can be made in the near term (perhaps within the next 6-9 months) from the
improvements that would require more time to accomplish. Within each part, the
recommendations divide into those pertaining to documentation, model data, model structure,
and model output.
-------
The Panel felt that this particular organization of the Advisory would be most effective,
even though it does not address each charge question in the order these questions were originally
presented. The charge questions are addressed in the text below as follows:
Charge
Question
la
Ib
Ic
Id
le
If
2a
2b
2c
2d
2e
2f
3a
3b
3c
3d
4a
4b
4c
4d
4e
5a
6a
6b
Text Page(s)
Addressing the
Question
9
11-12, 15-16
16,20
10
11
17
3-4, 8-20
3-4, 8-20
9
3
4
3-4,9-11
10
4-5, 10
3-4,8
20
4-5,8
10, 15-16
10, 15-16
8
10
4-5,8
3-20
3-20
-------
Part I: Recommended Immediate Improvements
A. Improvements to Documentation
i. General
The recent documentation provided in response to the Panel's initial requests for
information (papers 1-3 listed in the introduction above) helped clarify the data, structure, and
outputs of the SGM model. However, the Panel believes that significant further improvements to
the documentation are warranted. In general, the documentation should be organized in a more
coherent way. One possible organization is as below:
• Model Structure — household behavior, producer behavior, energy sector
specification, international trade specification, technological change, government
behavior, dynamics, emissions modeling, agents' expectations, representation of
climate policies, disaggregation (of sectors, regions, resources)
• Model Inputs ~ data and parameters
• Model Outputs and Reporting ~ reporting of prices and quantities; measurement of
costs, welfare measures; treatment of uncertainties in outcomes; sensitivity analysis;
model validation
• Solution Method
ii. Model Structure
The Panel recommends improvements in the model's documentation of model structure
along the following lines:
1. Clarify how the various aspects of the model - production, household demand, trade,
government sector — are connected. Readers should be able to see all of the excess demand
equations. From there the reader should be able to trace back the equations determining each of
the elements on the supply and demand side of each of the excess demand equations. The
documentation should make clearer which prices are exogenous and which are endogenous. It
should be made clear to the reader that the number of endogenous prices matches the number of
excess demand equations.
2. Include a "Derivation of Behavioral Equations" section as an appendix to the SGM
documentation. This section should make clear the theoretical basis for the structural equations
determining producer and household behavior. If a given equation involves a departure from
accepted theory, the documentation should acknowledge the departure and explain the reason(s)
for it.
-------
3. Make clear the nature of the central case and indicate which of the many off/on features of the
model are off or on in the central case. When are prices in the "everything else" sector
exogenous, and when are they endogenous? When do land prices play a role, and when do they
not? Which production sectors use Leontief technology, and which use CES? What is the
central assumption about price-expectations? Which of the various technological change
parameters (related to labor, energy, etc.) actually are employed in the model?
4. Improve the nomenclature to make it more consistent. For example, make clear when a
variable is sector specific, and when it applies to all sectors. Omitting subscripts is appropriate so
long as the reader is informed of the omission.
5. Confirm that the model is set up to check that Walras's Law is satisfied at every iteration of
the solution algorithm. (If necessary, the model itself should be extended so that it indeed checks
for Walras's Law in every application.)
6. Clarify how the model treats the ETE "everything else" sector. In particular, it is important
to:
a. Make clear how this sector fits into the rest of the model, and which price is set to 1
for this sector. It is important to indicate what is in, and what is not in, the ETE sector by
region.
b. Clarify which emissions are attributed to the "everything else" sector in Table 3.2 by
defining the activities and their relation to the ETE. For example, what is activity
ODSSub and why does only the service sector emit HFCs from this activity? It seems
like many emissions ought to be tied to industrial production. Also, it is unclear whether
and how abatement costs and GHG prices feed back to higher prices for ETE goods.
c. Clarify the relationship between P, Pi, and Pr. PiETE seems to be the numeraire but
sometimes it is subscripted by the sector to which the good is sold, and at other times it is
not.
d. Clarify the consequences of using the ETE sector as the numeraire. To the extent
Walras's Law is verified, a change in this numeraire should have no effect on quantities.
e. Compare choices about sectoral detail to other Integrated Assessment Models.
iii. Model Inputs - Data and Parameters
1. Provide a detailed comparison of the SGM base year data with the GTAP data (Hertel, 1997).
Many researchers working on the issues related to climate change use the GTAP data set, and
virtually all researchers undertaking global trade policy modeling use it. The GTAP data
-------
includes detailed accounts of regional production and bilateral trade flows, currently covering 87
regions and 57 sectors in each country. While the GTAP dataset may or may not dominate the
SGM base year dataset, it does allow for comparisons to a body of existing work.
2. Indicate the extent to which the parameterized model can replicate the benchmark data.
We encourage the model developers to provide source links for all data and parameters in
the current version of SGM and to continue to do so as they move to update these inputs.1
3. Paper 1 (one of three papers listed in the introduction) provides a comparison of the SGM and
GTAP approaches to fitting the IO table and energy balances for China. It would be very useful
to add a similar comparison for other regions where IO data are not as questionable as in China.
For example, comparisons for regions such as the US or Europe could be provided as well.
4. Paper 1 does not provide any comparison of the SGM base year data with the GTAP energy
data. There is a discussion of GTAP expenditure (price times quantity) data, the IO table of
China (price times quantity), and IEA energy data (quantities), but the GTAP also provides
energy data derived from the IEA statistics.
5. A comparison between the SGM economic (i.e., not just energy) data and the GTAP
economic data at an aggregate level for all SGM regions should be provided, to see how they
balance globally. This international and global perspective is fundamental to the modeling of the
effects of major energy policies, and cannot be ignored.
6. The documentation should provide sources for the data on greenhouse emissions. It also
should report aggregate numbers for CO2 and other greenhouse gases in some form. In Paper 2
(page 12) there is a mention of the kind of data needed for non-CO2 gases tracked in the model.
The documentation lacks a reference to the database used.
7. The following additional documentation would be very useful:
a. A discussion of the specific EIA data and refinements needed to compile Table 2.5.
b. Documentation of the base year non-CO2 emission (or emission factor) values and
their sources.
c. The sources for Table 3.1
d. The source(s) for the MAC (marginal abatement cost) curves. Also, clarify their
assignment to sectors.
e. A reference for the derivation of equations 3a-6.
8. In response to initial recommendations by the Panel, PNNL recently provided a master list of
parameters in the theory sector as requested. This should be expanded to include benchmark
1 In response to initial recommendations by the Panel, PNNL has already made some very useful improvements to
its documentation of data and parameters, for example by providing links for input-output data sources. At the same
time, the Panel was unable to get any detail on data from outside the U.S. Documentation of such data is crucial to
the credibility of the model.
-------
parameter values, sources, and any refinements necessary to arrive at them with a cross listing to
model equations. A master list should be provided for all other data inputs.
9. The documentation could use improvement in its discussion of choices made regarding data
for hydroelectric and nuclear energy. In particular, it should explain why EIA rather than IEA
data were used.
10. The documentation should explain why the investment accelerator is set at 1.2 (page 28).
11. An inconsistency regarding the variable PN should be eliminated. In the theory chapter PN is
the rental price of capital; in the documentation it is the price of the numeraire.
iv. Model Outputs and Reporting
Existing documentation reveals almost no sensitivity analysis. This severely reduces the
user's ability to evaluate potential policy outcomes. To the extent that some sensitivity
experiments have already been performed, the results of these experiments should be displayed.
Further sensitivity analysis should be given high priority, as indicated in Part 1C below.
v. Solution Method
The documentation should refer to its chosen software and solution algorithm, and
compare its choices with other algorithmic tools for the CGE modeling (such as GAMS, or
GAMS-MPSGE software with an MCP algorithm).
-------
B. Initial Improvements to Data and Parameters
The three papers recently provided by the model developers (listed in the introduction)
indicate that the model developers have recently taken some significant steps toward improving
the model's data and parameters. However, some significant further improvements are called
for. Two of these can be accomplished in the near term. The recommended additional short-
term improvements are as follows.
1. The model developers should seriously consider making greater use of GTAP data. The
developers have indicated a preference for SGM data because it preserves physical units for
energy. However, it is not clear that GTAP could not be adjusted in this manner. Also, it is
possible that if the updating the SGM model data as often as GTAP would imply high (and
duplicative) costs.
2. Appendix B offers a list of studies that have generated econometric estimates of demand
elasticities. The Panel recommends that the model developers examine these estimates and
consider the extent to which the elasticities implied by SGM model's parameters are consistent
with these estimates. Parameters should be adjusted accordingly.
C. Further Sensitivity Analysis
Existing documentation reveals almost no sensitivity analysis. This severely reduces the
ability to evaluate potential policy outcomes. Further sensitivity analysis should be given high
priority. In the near term, simple sensitivity analysis could be conducted, in which policy
outcomes are generated under different values for key parameters. Section D of Part II indicates
the Panel's recommendations for a more systematic, extensive, and informative sensitivity
analysis procedure that generates probability distributions for policy outcomes.
-------
Part II: Recommended Subsequent Improvements
A. Further Improvements to the Data
i. Updating the Data Set
The Panel wholeheartedly supports the model developers' intention, stated on pp. 8-9 of
paper 2, to strengthen the empirical basis of the model and to automate the data updating process.
Such an update would reflect changes in economic conditions and in technologies in critical
sectors, and would allow comparability with other data sets. The base year of the data in the
SGM model is 1990. Many countries and regions have experienced substantial changes in
economic conditions and technology since 1990. The base year should be updated to reflect
these changes, particularly in critical sectors for the analysis of carbon policies.
ii. Greater use of GTAP data
As mentioned earlier, the model developers should seriously consider using the GTAP
data set. The SGM documentation (Paper 2) states that "the majority of time is spent obtaining
and processing the necessary data." The SGM developers should consider using the GTAP data
set to save the time spent in obtaining and processing the data. Many researchers working on
issues related to climate change use this data set, and virtually all researchers undertaking global
trade policy modeling use it. The GTAP data include detailed accounts of regional production
and bilateral trade flows, currently covering 87 regions and 57 sectors in each country. The
dataset also includes supplemental energy data in physical terms, which is linked to the
economic data. The base year for version 5 of GTAP is 1997, and for version 6 it is 2001. The
GTAP data set is available at extraordinarily low cost. Details on the GTAP data can be
obtained from http://www.gtap.agecon.purdue.edu, and extensive documentation of version 5 is
provided by Dimaranan and McDougall [2002].
The GTAP data may be accessed either using the GEMPACK software provided with the
data package, or through GAMS using tools developed by Thomas Rutherford
(http://www.mpsge.org/gtap6). In either case the available software provides flexible aggregation
schemes, to allow the user to match the resolution of the GTAP data to their own needs,
removing the need to carry along all of the complete detail in the full data set. The GTAP data
set is illustrated in applications contained in Hertel [1997], although one does not need to use the
GTAP models in order to use the GTAP data set.
Whether the SGM's developers ultimately decide to use the GTAP data set, the Panel
urges the developers, at a minimum, to provide a comparison between the SGM data and the
GTAP data. For the energy data these comparisons should be in value terms and in physical
flows. The use of constrained optimization routines to facilitate the choice of data set has a
venerable tradition, and has become much more common in recent years (see Stone,
Champernowne and Meade [1942] and Harrison, Rutherford, Tarr and Gurgel [2004; p.297]).
10
-------
One disadvantage of the GTAP dataset for carbon policy analyses is that the electricity
sector is currently a single aggregated sector. Therefore, this sector would have to be
disaggregated further, to reflect alternative energy supply technologies such as coal, hydro-
power, nuclear, wind, biomass, etc. Such disaggregation would not be difficult (e.g., the TEA
provides detailed energy balances for many countries).
In summary, the Panel suggests two feasible scenarios for updating the SGM data set.
One possibility is that SGM developers retain the current procedures they use for data collection
and calibration, update the data to 1997 or 2001, and provide a detailed comparison with GTAP
data being used by other modelers. The other is that SGM developers use the GTAP data with
additional disaggregation of electricity sector. The Panel tends to prefer the latter option, but the
best path will depend on information obtained by the model developers as they examine and
compare the data sets.
B. Model Structure Improvements
i. Household Utility Modeling and Welfare Calculations
The Panel urges the model developers to improve the specification of individual
household behavior. Currently, such behavior is not derived from an explicit utility function.
This precludes the use of theoretically consistent measures of the welfare impacts of policies.
Without such measures, the model will not be capable of yielding estimates of the true
compliance costs of policies.
The discussion below focuses only on the general representation of consumer demand
functions, and on the use of demand functions to construct welfare measures. This section does
not include a discussion of specific issues relevant to inter-temporal decision making, such as
providing a utility theoretic basis for the allocation of income among current versus future
consumption through savings/borrowing, nor the allocation of time among labor/leisure. These
are significant topics involving specialized issues that are deserving of separate consideration,
but are not covered below.
The simplest approach for creating a utility theoretic basis for an aggregate model is to
apply the notion of the "representative consumer". Here, aggregate (or average) demand is
treated as if it were generated from a single utility maximizing individual (see, for example,
Deaton and Muellbauer, 1980, p 149-158). The representative consumer approach can be a
pragmatic way to assess welfare effects. This could be the first major improvement to SGM's
modeling of consumer behavior.
However, the representative consumer approach has been widely criticized by
economists. (See for example, Kirman, 1992; Stoker, 1993; Slesnick, 1998.) Constructing an
11
-------
aggregate demand function that is "rationalized" by a particular utility function implies either
placing strong restrictions on preferences or the use of market demand functions that are not
logically consistent with the aggregation of a set of disaggregate consumer demand functions.
As a potential subsequent step, the Panel recommends that the SGM be modified to
include multiple representative consumers, one representative consumer for each of several
socio-demographic groups. This would partly address the theoretical objections to the
representative consumer approach. The use of multiple representative consumers also has the
appeal of allowing one to estimate distributional effects of policies across various consumer
groups.
The challenge faced in extending the model to multiple representative consumers is to
identify data adequate to specify demand functions for separate representative consumers (e.g., a
representative low income vs. middle income vs. upper income individual). The Computable
General Equilibrium literature has several examples of models based on multiple representative
consumers, involving anywhere from small to very large numbers of separate representative
consumers (e.g., Piggott and Whalley, 1985; Cockburn, 2001; Cogneau and Robillard, 2000;
Harrison, Rutherford, Tarr and Gurgel, 2005).
ii. Production
General Recommendation
The Panel's main recommendation is that the model developers replace the flat (one-
level) CES production specification with a more flexible specification such as a nested CES
structure.
Specifically, the Panel recommends that the developers survey the recent literature (e.g.,
Burniaux and Truong 2002) and employ a nested-CES production structure more in line with
existing CGE models and parameterized based on empirical data. Model choice (as well as
documentation) should pay particular attention to issues, such as short-run complementarity
versus long-run substitutability of capital and energy, highlighted in the literature. In the future,
the modeling team should consider exploring other functional forms (maintaining global
regularity) and making their own empirical parameter estimates using updated data.
Background
The goal of production modeling is a flexible, parsimonious, practical representation
grounded in empirical data. A production system is flexible to the extent that it can capture the
full range of theoretically consistent, local substitution possibilities. It will be parsimonious if it
captures only the detail needed to assess the problem as hand. It is practical if it be applied to
and solve under a wide range of policy scenarios.
A fully-flexible representation is one that provides a second-order differential
approximation to an arbitrary twice continuous differentiable cost or production functions
(Diewert and Wales 1987). That is, it can accommodate any pattern of local substitutability /
12
-------
complementarity of inputs about the initial benchmark prices. Examples of such functions in the
literature include the translog and generalized Leontief, as well as a number of other less
common forms.
A key concern in these functions is regularity. That is, downward sloping input demand
curves for all inputs (and linear combination of inputs). Global regularity for all non-negative
input (and input combinations) is especially hard to guarantee when the second-derivatives are
complex functions of both parameters and inputs. For simulations to be admissable, regularity is
theoretically necessary only over the range of equilibrium prices and quantities—however
narrowly or widely they vary. In practice, however, most computational algorithms have
trouble with non-globally regular functions—in the course of finding the equilibrium, prices and
quantities can wander far beyond the eventual equilbrium. Therefore, local regularity about an
equilibrium (or range of equilibria) is not generally sufficient.
In response to this, Perroni and Rutherford (1995) propose a non-separable CES
functional form that can represent local second-order flexibility and remains globally regular.
Their formulation does not provide a unique representation (many representations match the
same second-order conditions), however, and has not been widely implemented.
More common approaches in the CGE modeling literature focus on more structured, less
flexible production models, in part because of the difficulty in parameterizing a fully flexible
model (which will have n x (n - 1) / 2 parameters, where n is the number of inputs). These
models typically employ nested CES functions, where the nests represent sets of inputs that are
separable from other inputs—in contrast to the above, non-separable model required for full
flexibility. In its simplest form, without any nests, the CES requires one parameter to describe
the common elasticity of substitution among all inputs. A few examples of nesting structures are
given at the end of this section. In particular, we see examples with materials separated from a
capital-labor-energy aggregate, versus all four groups together in one tier. We also see, within
the capital-labor-energy aggregate, either a capital-energy sub-tier or a capital-labor sub-tier.
The choice of nesting structure depends both on the questions being asked and empirical
data. Analysis of climate change policies, for example, requires considerable energy detail as all
of the referenced models demonstrate, and energy is typically in its own sub-tier. It should,
however, be an empirical question whether capital and labor are more likely separable, versus
energy and capital. Sources of empirical elasticity estimates are cited elsewhere in this report
(see, for example, Burniaux et al. 1992).
The Panel urges the SGM modelers to consider choose nested production structures that
are appropriate for the sector involved. Appendix C provides examples of nested structures for
production.
iii. International Trade
General Recommendation
13
-------
The panel feels that the model's treatment of international trade significantly
limits its ability to evaluate climate change policies. The treatment of international trade is
currently exogenous in that domestic policies do not influence the volume and composition of
international trade. An endogenous treatment of international trade is crucial for understanding
the principal impacts of climate change policy. This is obvious if the policy being evaluated
involves countries besides the United States, such as proposed multilateral or joint policies such
as Kyoto-type policies (e.g., Harrison and Rutherford [1999] and Pinto and Harrison [2003]). But
it is equally important if the policies are only "domestic" in orientation, since the effects may be
dramatically muted if trade offsets them.2The SGM modeling team has indicated that it uses the
SGM model in two ways. Some applications focus on impacts on the U.S., and employ a "USA-
only," stand-alone, version of SGM. Other applications make use of the full, global SGM model.
The Panel believes that improvements to the treatment of international trade are crucial for both
applications.
There are two ways in which the trade component of the model could be improved. The
first method is a near-term step, which is applicable to the stand-alone, "USA-only" version of
SGM. Our recommendation is to transform the current single-region closed-economy model into
a single-region open economy model that is "closed" with a trade sector that allows for
substitutability between domestic and foreign produced goods, but that treats the global terms of
trade as fixed.3
The second method is part of a longer-term strategy of model development, and applies
to the full-blown multi-region version of the SGM. Our recommendation is to extend the current
structure—which is currently little more than a collection of closed-economy models which can
engage in trade in emission rights—to be a truly global model, by explicitly including bilateral
trade in commodities between regions.
Both of these approaches have long traditions in the broader general equilibrium
modeling literature, and the strengths and weaknesses of each are well known. The second
approach is needed if one is to seriously consider modeling global policies: relying on other
models and modelers to fill in critical simulations is perilous, even if it sounds like the
diplomatically correct thing to do. This is particularly true if the other models are unavailable for
public scrutiny, as appears to be the case with the partners chosen by SGM. On the other hand,
building a global model may be a lot of work if the SGM team insists on constructing its own
database. A move to the GTAP database would dramatically reduce these costs.
The literature is full of studies of these effects. For example, Harrison and Kristrom [1998a][1998b] consider the
effects of unilateral carbon tax increases in Sweden, and find that they could actually increase global carbon
emissions, which is the very opposite of the intended environmental objective. The logic is simple: increases in
carbon taxes in Sweden cause a substitution away from Swedish-produced goods towards foreign-produced goods,
and if foreigners are more carbon-intensive in their production processes then emissions increase. Since Sweden has
considerable nuclear and hydro power, and there are many countries that it trades with, such as Denmark, Poland
and China, that do not, this trade-induced effect is quite likely for Sweden.
3 The global terms of trade is a trade-weighted average of the relative prices of imports and exports to a given
country. The assumption of fixed global terms of trade means that no given country has sufficient market power to
affect the prices of its imports relative to its exports.
14
-------
Specific Elements of a First Approach
Therefore, we focus the bulk of our discussion on incorporating trade using the first
approach, which is to treat the United States as a small open economy.4
The specific structural changes involved are as follows:
1. Imports of each commodity should be specified as a constant-elasticity-of-substitution
(CES) function of the price of the particular commodity and the quantity of an import
composite. The dual of the import composite is the price of foreign exchange.
2. Aggregate exports should be specified as a constant-elasticity-of-transformation (GET)
aggregation of the quantities of exports of the individual commodities in the model. As in
point (1), the dual of aggregate exports is the price of foreign exchange.
3. The production of commodities in each traded sector should be specified as splitting
gross output between domestically-consumed and exported goods using a GET function.
4. All traded commodities should be represented as Armington (CES) composites of
imported and domestically-produced varieties. The associated dual variables are the
Armington goods prices, which serve as the prices of commodity inputs to intermediate
and final demand.
5. Aggregate imports and exports should be linked by a balance-of-payments constraint.
We emphasize that these five alterations can be implemented immediately, and the new
structure numerically calibrated using the existing social accounting matrix.5
A major consequence of explicitly representing trade is the issue of what trade elasticities
to specify. There is considerable uncertainty related to these elasticities. Empirical work
employing structural econometric models tends to generate relatively low trade elasticities, while
reduced form approaches that give more weight to the apparently low level of market power by
many countries yield higher elasticities. Harrison, Rutherford and Tarr [1996] discuss the range
of estimates and the surrounding debates as to the true values. We recommend that, to
acknowledge the uncertainties, the SGM model employ two sets of trade elasticities, representing
estimates at the high and low end of range of estimates.
The Panel has some concerns about the SGM modeling team's current use of calibrated
logistic share equations to simulate the partitioning of the aggregate consumption of each
4 This is actually a plausible assumption, despite the fact that the US economy is large. Moreover, if there is some
concern that US policies might influence global terms of trade, those effects can be estimated "outside of the SGM"
and evaluated parametrically within SGM. Harrison and Kristrom [1998b] illustrate how one can take changes in the
global terms of trade from some other model and evaluate domestic carbon tax policies with and without that global
context. This requires some modest efforts at pairing up sectoral aggregations across models, but is not as difficult
as it might seem a priori since the pairings do not need to be exact or one-to-one.
5 The detailed specification of such a structure is described in De Melo and Tarr [1992] and Rutherford, Rutstrom
and Tarr [1997]. Detailed specifications for comparable multi-region trade structures are available in Rutherford and
Paltsev [2000; pp. 10-17, 21-28].
15
-------
commodity into imported and domestically-produced varieties. We would recommend instead
an Armington trade specification. The Armington specification has three key benefits:
• Flexible demand functions for domestic output and imports
• An abundance of econometric estimates of the elasticity of substitution in different
industries—e.g., for North America, Stern, Francis, and Schumacher (1976), Shiells, Stern,
and Deardorff (1986), Reinert and Roland-Hoist (1992), Shiells and Reinert (1993), and
Gallaway, McDaniel and Rivera (2003).
• Global regularity, in the sense of Perroni and Rutherford (1995)
The alternative approach proposed by the modeling team would not only fail to capitalize
on these advantages, it would also introduce significant additional work on their part, which can
be avoided. To see this, assume that the aggregate use of good /', Af, is produced from imported
and domestic varieties Dt andM,, respectively, according to an aggregation technology,/:
4 =/(A, ,)
The aggregate price of i,pt, is determined by the dual cost function ct, denominated over the
prices of imported and domestic varieties, pf and p™ :
Presumably the share equation to be employed will be similar to the following:
FM( dc^
(1) SM =-
dc, , dc, exp{^+/Tln^+/
where ^ is the import share of the value of aggregate use of commodity /', pD andpM the prices
of domestic and imported varieties of/', A represents the (Armington) aggregate demand for /',
and a°, aM, ^D, ^M, fPM and f and f, are estimated parameters.
In the general, w-input case the logit share equation does not contain an analogue of the
elasticity of substitution. This threatens to complicate analysis of the sensitivity of traded
industries to climate change mitigation measures, as there is seemingly no way to evaluate the
responsiveness of these sectors' activity levels to different degrees of substitutability between
domestic and imported varieties. However, Considine (1989: 934-938) shows that in the 2-input
case, eq. (1) is very similar to the CES function in its properties and behavior. This result both
highlights the utility of our recommended approach and implies that the modeling team can
avoid reinventing the wheel by simply following our original guidance.
Finally, the Panel is not comfortable with the SGM modeling team's justification for
using the logistic approach, namely, that it will preserve energy quantities in the aggregation
16
-------
process. The premise of this justification seems problematic. In the case of quantity shares
rather than cost shares, i.e.,
SM=-
M
the degree to which the logistic model retains its global properties is uncertain, as there does not
appear to be a well-posed dual form which can be easily integrated into the price-endogenous
equilibrium structure of a CGE model. What is certain is that the analog of the elasticity of
substitution vanishes, with the result that sensitivity analysis becomes a complicated task.
In our judgment it is far better to improve the model using economically meaningful
approaches rather than sacrifice the economics for the sake of getting energy quantities right. We
feel strongly on this point, particularly given that other models (e.g., GTAP in GAMS, MIT
EPPA) are able to employ the Armington specification without it having much of an adverse
impact on the calculation of energy trade volumes.
iv. Electricity and Agriculture/Forestry Sectors
The Panel is impressed by the SGM model's high degree of sectoral as well as regional
disaggregation. However, it concludes that further refinements to modeling of the electricity and
agriculture/forestry sectors would significantly improve the model's ability to evaluate climate
policies.
Electricity Sector
The electricity sector represents an important aspect of the model because it is the source
of a large portion of greenhouse gases and because it is the sector that is expected to provide a
large portion of emission reductions under various climate policies. Three features differentiate
the electricity sector from the rest of the economy in ways that may be important to SGM. One
is that capital investments are very long-lived. Second, half the nation still uses cost-of-service
regulation to determine electricity prices and a large part of the nation that is ostensibly under
competition also has regulated aspects to the determination of price. Third, the sector is the target
of many other environmental and technology policies that affect its performance with respect to
GHG emission reductions and cost of those reductions.
Rate of capital turnover. The Panel believes that the assumed lifetimes of capital in the
SGM model are shorter than what is implied by the empirical evidence. In the model, the
lifetime of capital is 20 years. This implies overlapping generations of technology with
improvements for 25% of the capital stock every five years. Capital lifetimes in important
sectors (e.g., electricity) are significantly longer than specified by the model. As a result, the
model implies too much flexibility in capital which, other things equal, biases downward the
estimated cost of climate policy. The Panel urges the model developers to consider refining the
17
-------
capital lifetimes in the electricity sector and other sectors to make them more consistent with
empirical evidence.
Regulated prices. The long-run significance of economic regulation is partly to affect the
pace of technological change and partly to affect the role of risk in investment decisions. But for
SGM, the most important effect is the differentiation of price from marginal cost by time of day
and the effect this has on choice of technology for electricity generation. The current structure of
demand reflects prices that do not differ by time of day for most customers, thereby providing no
incentive to change the time of electricity consumption. If time of day pricing becomes common,
one would expect to see a shift away from peak to off-peak (baseload) consumption. This
suggests a smaller role for gas and a larger role for nuclear and coal-fired generation.
Policy capabilities. The electricity sector is a target of policies such as renewable energy
portfolio standards, benefit programs promoting end-use conservation, tax incentives favoring
one or another technology. These policies have important vintage effects. SGM needs to be able
to characterize technology choices that may differ from least cost choices according to predicted
market prices over time. Perhaps this can be done with a shadow price adder that reflects
calibration to current data.
Agriculture and Forestry Sectors
The agricultural and forestry sectors are essential sources and sinks of CCv Agriculture
is also a significant source of non-CC>2 GHGs. Agricultural and forest sectors of both developed
and developing nations are subject to extensive policy interventions that influence the amount of
land crops, pasture, and forests, the types of commodities produced, and production practices
that affect carbon fluxes and non-CC>2 GHG emissions. An example receiving much attention is
the Conservation Reserve Program (CRP) in the U.S. The CRP pays farmers to take croplands
out of production. The lands are left unplowed and planted with a cover crop. Because of the
CRP, GHGs from the use of farm machinery are eliminated, carbon releases that occur with
tillage are eliminated, and carbon sequestered by soils increases. In addition to affecting carbon
fluxes and non-CO2 emissions, agricultural and forest policies influence the marginal costs of
sequestration in agricultural and forest land, and thus the potential participation of and gains
from agriculture and forestry in carbon trading or other carbon policies.
The SGM model is without agricultural and forest policies. More importantly, the highly
simplistic specification of the agriculture/forest sector in the model does not facilitate the effects
of agricultural and forest policies, or policies directly addressing GHGs, on key variables
influencing carbon fluxes in these sectors or non-carbon GHGs. The overall utility of the model
for GHG policy analysis is correspondingly limited. The Panel urges the modeling team to
extend the model to incorporate important climate policies related to the agriculture and forest
sector, such as the CRP.
Recommendations elsewhere in this report call for revisions of the production structure of
the model to use nested CES production functions. The revisions of the production structure
would offer an opportunity as well for revisiting the specification of the combined agriculture
18
-------
and forestry sector to increase the capacity of the model to reflect the influences of agricultural
and forest policies on GHGs and the marginal costs of mitigation policies.
Disaggregation of the combined agriculture and forest sector into separate sectors, and
possible further dissagregation of the agricultural and forest sectors into subtypes useful for
policy analysis (e.g., crops, livestock) would increase the capacity of the model. Further,
because of the importance of land and land policies to both sectors, and the importance of land
cover to GHGs, inclusion of land as an input in the production of these products would increase
the utility of the model for agricultural and forest policy analysis.
v. Greenhouse Gas Emissions
The Panel recommends the following improvements in the treatment of non-CO2
greenhouse gases:
1. In the SGM, the CC>2 emissions mitigation options are endogenous (i.e., the model responds
to a carbon price via changes in demand, supply, technology change, different investment
decisions, etc.). However, for non-CO2 emissions the SGM uses the exogenous curves relating
percent reduction in non-CC>2 emissions to the carbon price; this limits the full range of general
equilibrium effects. We recommend endogenizing the non-CC>2 emissions as other models of a
similar type have done. One possible way to implement endogenous mitigation options is as
follows:
• incorporate non-CC>2 emissions mitigation into the production structure;
• incorporate non- CC>2 emissions mitigation into consumption;
• take the base year GHG and economic data, and generate activity-specific emissions
coefficients for each gas; and
• generate region- and sector-specific time trends in emissions coefficients.
GHG mitigation activity levels will differ according to flows of inputs (e.g., fossil fuel
combustion, fertilizer use), flows of outputs (e.g., rice cultivation, natural gas transmission), and
stocks of inputs (number of ruminant animals, landfill volume).
2. The existing documentation states that for the non-CC>2 emissions, there are more than a
dozen sources, which makes "the process modeling used for CC>2 impractical." However, in the
SGM all nitrogen sources share a common cost curve, as do all high global warming potential
(GWP) sources. In actuality, mitigation differs greatly across most of these sources. We
recommend that the SGM move toward incorporating different cost curves for the different
nitrogen sources and high GWP sources.
3. The SGM documentation should provide references to CC>2 and other GHG databases used in
the model. It also should provide aggregate numbers for all GHG gases for the base year for all
SGM regions."
19
-------
4. In the SGM, the "exchange rate" between carbon prices and other GHG prices is determined
by global warming potential (GWP). It should be noted that the use of GWP implies constant
rates of exchange through time, which some authors consider a problematic assumption. (See,
for example, Eckaus (1992), Reilly and Richards (1993), Schmalensee (1993), Reilly et al.
(1999).)
C. Dealing with Uncertainty
The Panel believes that the absence of sensitivity analysis seriously compromises the
model's ability to evaluate climate policies. Focusing alone on simulation results based on
central or best-guess values of parameters or favored structural assumptions gives no information
as to the fragility or robustness of predicted policy outcomes. In Part I we indicated an initial
approach to sensitivity analysis that could be introduced immediately. Here we offer and
recommend the use of Monte Carlo techniques to provide greater information as to the
robustness of predicted policy outcomes.
i. Applying Monte Carlo Simulation in the Sensitivity Analysis
Uncertainties about parameter values
Modern computing capabilities make it relatively easy to use Monte Carlo techniques.
We first discuss these techniques as applied to parameters. Under Monte Carlo simulation, the
user specifies a range of possible distributions that each parameter or modeling assumption can
take. In performing policy simulations, the model then randomly draws from the posited
probability distributions of each parameter, repeatedly solving the model under different
randomly drawn sets of parameters. The probability distribution of policy outcomes can then be
characterized by simple and well-known statistical procedures. The key insight is to move away
from ad hoc sensitivity analyses that only perturb one elasticity or set of elasticities at a time,
since they do not adequately convey a sense of the fragility of policy simulations from general
equilibrium models.
In principle, the probability distributions for the model parameters should be joint
distributions, allowing for covariances across parameters. However, it may be difficult to obtain
information on such covariances, in which case it is reasonable to assume the probability
distributions of parameters are independent.
The existing literature provides ready guidance for how one might set up these sensitivity
analyses for parameter estimates (e.g., Harrison and Kimbell [1985], Pagan and Shannon [1987],
Harrison and Vinod [1992] and DeVuyst and Preckel [1997]). For example, one might use an
elasticity of substitution with a point estimate provided by an econometric study, and typically
that study will also provide an estimate of the standard error. One can then assume a distribution
for the parameter estimate, assume that it has no covariance with other parameter estimates, and
20
-------
use this information to guide the random draws for the Monte Carlo simulations. Depending on
the parameter involved, the assumption of a normal, joint normal, or other distrubution will be
most appropriate. It may also be appropriate to constrain certain parameters to be strictly
positive or non-negative. By using this information related to the uncertainties in the estimates,
the model results stem from range of potential values for these estimates, while putting weights
on the different values in accordance with their likelihoods.
If no estimate of the standard error is available, one can be assumed a priori. If system-
wide estimates are available, either of demand systems or supply systems, then the econometric
study will also provide a covariance matrix that can be used to allow for the correlation between
estimates; utilities for multivariate random number generation are readily available. The SGM
model should contain a default set of distributional assumptions for all key parameters, and
perhaps a scalar that can be used to inflate or deflate sets of elasticities. This would allow
researchers to "turn off the uncertainty about trade elasticities, for example, and see what the
contribution is from uncertainty about other elasticities.
Although the literature has naturally focused on uncertainty about elasticities, since the
values of the elasticities elasticities, typically drive the intuition of economists and the policy
debates, one could readily extend these idea to uncertainty about other data used in the model
(e.g., perturbations in raw transactions data could be considered, providing one had a re-
balancing routine that ensures micro-consistency once accounts were not in balance, say by
solving for the nearest set of data that satisfies those micro-consistency constraints and
minimizes some metric of deviation from the initial data).
Monte Carlo sampling methods of Harrison and Vinod [1992] have been widely
employed in models that are solved in "level form," and they do not entail significant additional
programming. The Gaussian quadrature methods of DeVuyst and Preckel [1997] are likely to be
more efficient in terms of the number of solutions required for a given estimate of the
distribution of policy effects, but will require slightly more up-front programming. Neither is
onerous, in relation to the other demands of modeling.6
Uncertainties in model specification
Although less common, the literature also shows how one can extend these ideas to
include uncertainty about model specification (e.g., Harrison, Jones, Kimbell and Wigle [1993]).
The idea is to posit two or more model specifications, treat the choice of these specifications as
coming from a discrete distribution, and assign probability weights to each. An appropriately
diffuse distribution would be to simply assign equal weight to each alternative. Alternatively,
where model structures have familiar application in the literature, one could rely on expert
elicitation techniques to assign probability weights. Or one could ascertain what weight has to be
put on one alternative in order for the qualitative policy results to change. In any event, the
computational logic is the same.
6 Specialized methods exist for models solved in "difference form," as illustrated by Pagan and Shannon [1987],
although these are not applicable for SGM.
21
-------
ii. Characterizing Uncertainty in the Presentation of Policy Results
The results of a systematic sensitivity analysis can be presented in several ways that
would dramatically improve the plausibility of the policy analyses undertaken with SGM. To
display the stability of model results with respect to policy recommendations, one popular
method is merely to display a histogram of the distribution of key results, along with information
on the empirical 90% confidence intervals, or the probability that the sign of the policy variable
is positive or negative. Policy-makers appreciate having some sense of the confidence in the
predicted sign of a policy variable, just as one expects to see a p-value or t-statistic beside any
statistical estimate of a policy effect.
Another use of sensitivity analysis is to guide the allocation of resources in model
refinement. Results of sensitivity analysis could be used to identify those variables that have the
largest effect on propagating uncertainty in the outcome measures and policy recommendations.
In the CGE model one can use the analysis to identify "key elasticities" that drive the policy
results. Although it is true as a formal matter that every elasticity and parameter matters for the
numerical results, it is almost always the case that uncertainty over several key numbers can
generate widely divergent policy results. By highlighting those data that are relatively more
important, the modeler is alerted to where it would be efficient to allocate effort to improve data.
D. Validation through "Backcasting" Exercises
The results obtained from large-scale simulation models such as SGM rest on many
parameter estimates and model assumptions. To avoid these policy simulations becoming a
"black box," it is valuable to have a sense of their sensitivity to variations in estimates and
assumptions.
Beyond sensitivity analysis, it is possible to gain a better sense of the validity of the
model's structure and inputs through "backcasting." This involves running the model forward,
starting from some point in the past, and observing how well the model tracks past history. The
Panel urges the model developers to conduct this type of experiment.
Such validation exercises have been conducted, for example, by researchers at the Dutch
Central Planning Bureau. Henri Theil applied this approach in the 1960s using an annual input-
output model. These backcasting exercises will provide useful information to the model
developers — information that can guide improvements to the model's structure or data.
E. Consideration of Additional Policy Details and Instruments.
The Panel recommends some extensions to the SGM model that would enable it to
capture important climate policies related to the electricity, forest, and agriculture sectors. These
recommendations were provided in Section B iv above.
22
-------
The Panel also recommends extending the model to enable it to consider additional
policy instruments that might apply across a number of sectors. The model currently is set up to
consider only carbon taxes or equivalent simple cap and trade systems. The Panel recommends
that the model be extended to consider some important details of the cap and trade systems
incorporated within actual and proposed international policies. One important detail would be
allowing for limits placed on individual countries' trades.
The current model does not allow for close consideration of non-market policies such as
performance standards. As a long-term objective, the model developers might consider
extending the model to address such policy instruments.
23
-------
Appendix A: Charge Questions
I. Is the SGM appropriate and useful for answering questions on the economic effects of
climate policies?
la. Are the costs of compliance with a carbon policy, as represented in SGM, an appropriate
measure of the aggregate or economy-wide cost of climate policies?
Ib. Does the SGM adequately represent the costs of climate policies to different sectors of the
economy?
Ic. Does the SGM adequately represent the geographic dynamics of possible responses to climate
policies (i.e., "where" flexibility)?
Id. Does the SGM's adequately capture intertemporal aspects of policies (e.g., "when"
flexibility) and the intertemporal adjustments to climate policies?
le. Does the SGM adequately represent the sectoral dynamics of possible responses to climate
policies, e.g., multiple gases, multiple sectors, etc.?
If. Is the SGM capable of representing and contrasting the types of climate policies that are the
focus of policy discussions (e.g., allowance trading systems, performance standards, etc.)?
II. Are the model's structure and fundamental assumptions reasonable and consistent with
economic theory?
2a. Are the number and type of agents in the model (firms, households, governments, and
regions) appropriate for the problem, and are they adequately represented?
2b. Is each agent's optimization problem appropriately specified, and are the implied behavioral
equations used correctly?
2c. Is the model's specification of production technologies reasonable and capable of providing
reliable insights for climate policy analysis? Is the model's specification of short- and long-run
producer responses appropriate for assessing the dynamics of producer behavior?
2d. Are the market-clearing equations appropriate?
2e. Does the model satisfy basic tests for consistency with general equilibrium theory?
2f. Is SGM's representation of the flow of goods and services consistent with theory and
internally consistent? Please consider this question with respect to:
• decisions regarding investment and savings, and the relationship to the interest rate
• decisions of producers with respect to factor demand and the rate of production
• decisions of households with respect to their demands for goods and services and their labor
supply.
III. Are the parameter values employed in the model (e.g., elasticities of substitution and of
demand, price and income) within the range of values in the literature?
3a. The model employs a non-nested CES structure for its production functions. Both a long-run
and short-run elasticity of substitution are specified for each production process, where the long-
run elasticity of substitution is used for new vintages, and the short-run elasticity of substitution is
used for old vintages. Are the long-run and short-run elasticities of substitution used in the model
within the range of values in the literature? Are the values used appropriate for this model given
the long-term horizon and problems the model is designed to address?
3b. The model contains a large set of parameters used to simulate non-neutral technical change.
Given this specification of technical change, are the values of these parameters well chosen?
3c. Are the price elasticities of demand and income elasticities of demand within the range of
values in the literature, and are the values chosen appropriate for the model?
24
-------
3d. Is the calibration process for determining values for the behavioral parameters adequate?
IV. Are the model's parameterizations of physical phenomena logical, and are its projections of
future energy use and efficiency reasonable, given fundamental physical constraints and
rates of technological change?
4a. One of the important features of the SGM is its ability to track energy balances throughout
the model's time horizon. This is accomplished, in large part, by the creation of hybrid input-
output tables, which combine energy balance data from the International Energy Agency (IEA)
with national economic input-output data. Was the merging of these two data sets reasonable to
create the hybrid input-output tables? Are these hybrid input-output tables logical and useful
tools?
4b. Given that the climate problem is a long term one, and that technologies and prices are likely
to change significantly over the next fifty years irrespective of the form, or even presence, of a
climate policy, is the model's structure capable of generating projections for future energy use
and implicit energy efficiency that are logical and reasonable?
4c. Maintenance of energy balances through the model's time horizon without the explicit or
implicit assumption of an unspecified "backstop" technology has ramifications for the
parameterizations of elasticities of substitution. Essentially, physical laws constrain the real
world substitutability of capital, labor, or materials for energy, and are thus likewise embedded in
the model's production functions. Is the manner in which this is accomplished reasonable?
4d. Very often, the SGM is run using energy quantity and price projections from the NEMS
outputs in the Department of Energy's Annual Energy Outlook (through 2025) as inputs to SGM
and then extended to 2050. Is the use of NEMS model outputs as SGM input data an appropriate
modeling strategy?
4e. As noted above, the model does not make use of any assumed "backstop" technologies. The
consequence of this choice is that the model's energy use projections are of necessity based on an
existing set of technologies (though non-commercially available technologies can be and are
modeled). Given the timeframe of the climate problem and SGM's time horizon, as well as the
model's feature of tracking energy balances, is this an appropriate modeling choice? Is it the best
modeling choice?
V. In what areas is the model in need of further development?
5a. There are several areas of planned improvement for the SGM, and another set of areas which
have been the subject of some discussion but no concrete model development plan or effort.
These areas include:
• Conversion to C++ platform to facilitate modeling (underway)
• Broader sensitivity analysis to account more fully for uncertainties
• Inclusion of renewable energy technologies (underway)
• Inclusion of advanced fossil technologies (IGCC, NGCC), including regional representations
of carbon sequestration costs (underway)
• Examination of elasticities (with RFF, underway)
• Inclusion of an Agriculture and Land-use sub-model (underway)
• Endogenizing the full economics of non-CO2 abatement options (planned)
• Introduction of endogenous technological change (under discussion)
• Examination of some degree of regionalization of the electricity sector (under discussion)
25
-------
• Improvements to the representation of the household sector (under discussion)
Please comment on the importance of these tasks, and how we might best prioritize them.
5b. If any, please suggest other avenues for improvement and model development efforts that we
ought to pursue.
26
-------
Appendix B: Econometric Studies of Energy Demand Elasticities
Atkins, Frank J. and S. M. Tayyebi Jazayeri (2004) "A Literature of Demand Studies in World
Oil Markets," Oil Demand Workshop, OPEC Secretariat, Vienna., Austria., June 7-8, pp. 2-44.
Atkinson, Jago and Neil Manning (1995) "Chapter 3. A Survey of International Energy
Elasticities," included in Global Warming and Energy Demand. Edited by Terry Barker, Paul
Ekins and Nick Johnstone, pp 47-105.
Dahl, Carol A. (1986) "Gasoline Demand Survey," The Energy Journal, Vol. 7(1), pp. 67-82.
Dahl, Carol A. (1993) "A Survey of Oil Demand Elasticities for Developing Countries," OPEC
Review, XVII (4), Winter, pp. 399-419.
Dahl, Carol A. (1994) "A Survey of Energy Demand Elasticities for the Developing World,"
Journal of Energy and Development, Vol 18 (I), Autumn, pp. 1-48.
Dahl, Carol A. (1994) "A Survey of Oil Product Demand Elasticities for Developing Countries,"
OPEC Review, XVIII(l), pp. 47-87.
Dahl, Carol A. and Sterner, Thomas (1991) "A Survey of Econometric Gasoline Demand
Elasticities," International Journal of Energy Systems, ll(2):53-76.
Dahl, Carol A. and Sterner, Thomas (1991) "Analyzing Gasoline Demand Elasticities: A
Survey," Energy Economics, July, Vol 13(3):203-210.
Dahl, Carol A. and Sterner, Thomas (1991) "Analyzing Gasoline Demand Elasticities: A
Survey," Energy Economics, July, Vol 13(3):203-210.
Dahl, Carol A. (2005) Bibliography of Energy Demand Studies, Draft, Mineral Economics
Program, Colorado School of Mines.
Dargay, Joyce (2004) "The Effect of Prices and Income on Car Travel in the UK," ESRC
Transport Studies Unit Centre for Transport Studies University College London February,
downloaded www. cts .ucl. ac.uk/tsu/papers/FESWCTR2004Final. pdf.
Goodwin, P. B., Joyce Dargay, and M. Hanly (2004) "Elasticities of Road Traffic And Fuel
Consumption with Respect to Price and Income: A Review," Transport Reviews, 24(3) May,
pp.375-292.
Graham, D. and S.Glaister. (2002) Review of Income and Price Elasticities of Demand For Road
Traffic, Report to the Department for Transport, Dft, London. Downloaded From
Http://Www.Cts.Cv.Ic.Ac.Uk/Html/Researchactivities/Publicationdetails.Asp?Publicationid=267
May 24, 2004.
Graham, Dan and Stephen Glaister (2002) "The Demand for Automobile Fuel: A Survey of
Elasticities," Journal of Transport Economics and Policy, Vol. 36, Part I, January, pp. 1-26.
Graham, Dan and Stephen Glaister (2004)" Road Traffic Demand Elasticity Estimates: A
Review and Fuel Consumption with Respect to Price and Income: A Review," Transport
Reviews, 24(3), May, pp. 261-274.
27
-------
Graham, Dan and Stephen Glaister. (forthcoming) "The Responses of Motorists to Fuel Price
Changes: A Survey," Journal Transport Economics and Policy
Greening, L.A., D. L. Greene, and C. Difiglio (2000) "Energy Efficiency, and Consumption -
The Rebound Effect - A Survey," Energy Policy, 28, 389-401.
Soderholm, Patrick (1998) "The Modelling of Fuel Use in the Power Sector: A Survey of
Econometric Analysis," Journal of'Energy Literature, IV (2), Dec. 98, pp. 1-27
28
-------
Appendix C: Examples of Nested Production Structures
,
-------
U6Ii%
Lewel 1
Level 2
Level 3
Cons nCDCB 1C ic 4" •j
Me jsenc-d ut" ity s 3 CES fur
plioTi ,"-ic e'S.ireti"
Ccisu'-p'ion is a CatD-Za^glai
cc -iposife cf the 161o •* * L ;• >is
Each corsjme" jcoc s a CE3
earnest 33' y prod^ cec'
Level 4
Level S
Regional
Cutput
Dcn-,es"ic gocds 3f& s CE3 composite
cf loca1 y wockose gcic= sn: gcoos
P';cuc&d h DtnerregarB of ihe U.S
Val'.e A
l51o'
C a
(RTI International 2004)
Mcs1 pr3cucer gocds jse f >ea p*i-
p3rion= cf i-vter'nec'ole npirs a"d s
cs^ra -Isbor-ensrgy 'I- _E- son^pcsils.
Tis -LE cct-'ocste is a CES *i,nc:icn
c* sre^gy a~d V3l je-aofed « ^ — i
Mt^r^eoist-e ^n^its are the 1 1 to 3"4 ^s:^e~e"9
, - f «tf
E~ergy is a CES cc ^po5>te cf c :c " f je s. The
sl*ucx'e of + s fLncnon '<3--es 3?rcss se^t
lLs acoec -s s Ca
rnocs te :* zairis- ;ne
30
-------
Node
i
•"i
2>
4
2>
5
™t
S
9
10
11
12
13
Mnemonic
O
E
M
MA
MM
MN
MS
TA
\V?
OS
FM
MC
EQ
Interpretation
Output
Energy
Materials & Services
Agricultural Producti
Metal Products
Nonmefallic Products
Services
Textiles & Apparel
Wood &: Paper Products
Other Services,
Pnmaiy & Fabricated Metals
Maclunerj'
Equipment
Components
K,L_E,M
3.4.16.3031
6,M4^B,!.^!X.MS
1.7.3.TA.W?
FM.MC.EQ
5.15:17.19.27
OS.2S.32.33. 34
9.10,18
11. 12.13.14
29.3 5.N
2.20.21
22.23
24.25,26
(Wilcoxen 1988)
-------
Acts'eity i.^
Energy-Capital-Labor [ inmacf S,e
Energy-Capital , ; Labor
Capital i Energy 1 Skilled L-abor ; f Unskilled Labor I
Electricity ', " Fuels
I Solid Fuel ! Non-solid Fuels
Coal _ Biofuel ' ' Gas '. i Oil
(Hill and Kirstrom 2002)
32
-------
References
Bohringer, C., and A. Loschel. 2004. Economic Impacts of Carbon Abatement Strategies. In
Controlling Global Warming - Perspectives from Economics, Game Theory and Public Choice,
edited by C. Bohringer, M. Finus and C. Vogt. Cheltenham: Edward Elgar.
Brenkert A., R. Sands, S. Kim, H. Pitcher. 2004. Model Documentation. The Second Generation
Model, Pacific Northwest National Laboratory (PNNL-14256), Joint Global Research Institute,
College Park, MD.
Burniaux, Jean Mark, John P. Martin, Giuseppe Nicoletti, and Joaquim Oliveira Martins. 1992.
Green: A multi-sector, multi-region, general equilibrium model for quantifying the costs of
curbing CO2 emissions: A technical manual. Paris: OECD Economics Department Working
Papers No. 116.
Burniaux, Jean-Marc, and Truong P. Truong. 2002. GTAP-E: An energy-environmental version
of the GTAP model. Lafayette, IN: Center for Global Trade Analysis, Purdue University.
Cockburn, John. 2001. Trade Liberalisation and Poverty in Nepal: A Computable General
Equilibrium Micro Simulation Analysis. CREFA Working Paper 01-18; Available at
www.crefa.ecn.ulaval.ca/cahier/0118.pdf
Cogneau, Denis, and Robilliard, Anne Sophie. 2000. Growth, Distribution and Poverty in
Madagascar: Learning from a Microsimulation Model in a General Equilibrium Framework.
IFPRI TMD Discussion Paper 61.
Considine, T.J. 1989. Estimating the demand for energy and natural resource inputs: Trade-offs
in global properties (partial text). Applied Economics. 21: 931-945.
Deaton, Angus and John Muellbauer, 1980. Economics and Consumer Behavior, London:
Cambridge University Press.
De Melo, Jaime, and Tarr, David. 1992. A General Equilibrium Analysis of US Foreign Trade
Policy . Cambridge, MA: MIT Press.
DeVuyst, Eric A., and Preckel, Paul V. 1997. Sensitivity Analysis Revisited: A Quadrature-
Based Approach," Journal of Policy Modelling. 19(2): 175-185.
Diewert, W.E., and T.J. Wales. 1987. Flexible Functional Forms and Global Curvature
Conditions. Econometrica. 55 (l):43-68.
Dimaranan, B.V. and McDougall, R.A. 2002. Global Trade, Assistance, and Production: The
GTAP 5 Data Base (Center for Global Trade Analysis, Purdue University.
33
-------
Jacoby, Henry D., John M. Reilly, James R. McFarland, and Sergey Paltsev. 2004. Technology
and Technical Change in the MIT EPPA Model. Cambridge: MIT Joint Program on the Science
and Policy of Global Change.
Eckaus R. 1992. Comaring the effects of greenhouse gas emissions on global warming. Energy
Journal. 13:25-34.
Edmonds J., H. Pitcher, R. Sands. 2004. Second Generation Model: An Overview, Pacific
Northwest National Laboratory (PNNL-14916), Joint Global Research Institute, College Park,
MD.
Gallaway, Michael P., Christine A. McDaniel and Sandra A. Rivera. 2003. Short-Run and
Long-Run Industry-Level Estimates of U.S. Armington Elasticities. North American Journal of
Economics and Finance, 14(1): 49-68.
Harrison, Glenn W. and Kristrom, Bengt. 1998a. Carbon Emissions and the Economic Costs of
Transport Policy in Sweden," in R. Roson and K. A. Small (eds.), Environment and Transport in
Economic Modelling. Amsterdam: Kluwer Academic Press.
Harrison, Glenn W. and Kristrom, Bengt. 1998b. General Equilibrium Effects of Increasing
Carbon Taxes in Sweden," in R. Brannlund and I. Gren (eds.), Green Taxes: Theory and
Practice. London: Elgar.
Harrison, Glenn W., and Rutherford, Thomas F. 1999. Burden Sharing, Joint Implementation,
and Carbon Coalitions. In C. Carraro (ed.), International Environmental Agreements on Climate
Change. Amsterdam: Kluwer.
Harrison, Glenn W.; Rutherford, Thomas F., and Tarr, David G. 1996. Quantifying the Uruguay
Round," in W. Martin and L.A. Winters (eds.), The Uruguay Round and the Developing
Countries. New York: Cambridge University Press. 216-252.
Harrison, Glenn W.; Rutherford, Thomas F.; Tarr, David G., and Gurgel, Antonio. 2005. Welfare
Implications for Brazil and the Americas of MERCOSUR Trade Policy Options. In D.
Lederman and K. Schmidt-Hebbel (eds.), The Future of Trade Liberalization in the Americas.,
Santiago: Central Bank of Chile.
Harrison, G.W.; Rutherford, T.F., Tarr, D.G. and Gurgel,A. 2004. Trade Policy and Poverty
Reduction in Brazil. World Bank Economic Review.l8(3): 289-317.
Harrison, Glenn W.; Jones, Richard; Kimbell, Larry J., and Wigle, Randall. 1993. How Robust
Is Applied General Equilibrium Analysis?" Journal of Policy Modelling. 5(1): 99-115.
34
-------
Harrison, Glenn W., and Kimbell, Larry J. 1995. Economic Interdependence in the Pacific
Basin: A General Equilibrium Approach. In J. Piggott and J. Whalley (eds.), New Developments
in Applied General Equilibrium Analysis. New York: Cambridge University Press.
Harrison, Glenn W., and H.D. Vinod. 1992. The Sensitivity Analysis of Applied General
Equilibrium Models: Completely Randomized Factorial Sampling Designs. The Review of
Economics & Statistics. 74: 357-362.
Hertel, Thomas W. 1997. Global Trade Analysis: Modeling and Applications, Cambridge
University Press, Cambridge.
Hill, Martin, and Bengt Kirstrom. 2002. Sectoral EU Trading and Other Climate Policy Options:
Impacts on the Swedish Economy. Stockholm: Stockholm School of Economics.
Kirman, Alan, 1992. Who or What Does the Representative Individual Represent. Journal of
Economic Perspectives. 6(2): 117-136.
Pagan, Adrian R., and Shannon, John H. 1987. How Reliable are ORANI Conclusions?
Economic Record. 33-45.
Perroni, Carlo, and Thomas F. Rutherford. 1995. Regular flexibility of nested CES functions.
European Economic Review. 39:335-343.
Piggott, John, and Whalley, John. 1985. UK Tax Policy and Applied General Equilibrium
Analysis. New York:Cambridge University Press.
Pinto, Ligia M. and Harrison, Glenn W. 2003. Multilateral Negotiations Over Climate Change
Policy. Journal of Policy Modeling. 25(9): 911-930.
Reilly J. and K. Richards. 1993. Climate change damage and the trace gas index issue. Journal of
Environmental and Resource Economics. 3: 41-61.
Reilly J., R. Prinn, J. Harnisch, J. Fitzmaurice, H. Jacoby, D. Kicklighter, J. Melillo, P. Stone, A.
Sokolov, and C. Wang. 1999. Multi-gas assessment of the Kyoto Protocol. Nature, 401: 549-
555.
Reinert, Kenneth A. and David W. Roland-Hoist, 1992. "Armington Elasticities for United
States Manufacturing Sectors," Journal of Policy Modeling, v!4, n5, pp. 631-39.
RTI International. 2004. EMPAX-CGE Model Documentation. Washington: US EPA.
Rutherford, Thomas F. and Paltsev, Sergey V. 2000. GTAP-Energy in GAMS: The Dataset and
Static Model. Working Paper 00-02, University of Colorado Boulder, Economics Department,
http ://debreu. Colorado, edu/papers/gtaptext.pdf
35
-------
Rutherford, Thomas F.; Rutstrom, E. Elisabet; and Tarr, David. 1997. Morocco's Free Trade
Agreement with the EU: A Quantitative Assessment," Economic Modelling. 14(9): 237-269.
Schmalensee R. 1993. Comparing greenhouse gases for policy purposes. Energy Journal. 14:
245-255.
Shi ells, Clinton R., Robert M. Stern and Alan V. Deardorff. 1986. Estimates of the Elasticities
of Substitution between Imports and Home Goods for the United States. Weltwirtschaftliches-
Archiv. 122(3): 497-519.
Shi ells, Clinton R. and Kenneth A. Reinert, 1993. Armington Models and Terms-of-Trade
Effects: Some Econometric Evidence for North America. Canadian Journal of Economics. 26
(2): 299-316.
Slesnick, D.T. , 1998. Empirical Approaches to the Measurement of Welfare. Journal of
Economic Literature. 36.
Stern, Robert M., Jonathan Francis, and Bruce Schumacher. 1976. Price Elasticities in
International Trade: An Annotated Bibliography. London: Macmillan.
Stoker, Thomas M. 1993. Empirical Approaches to the Problem of Aggregation over Individuals.
Journal of Economic Literature. 31(4):1827-1874..
Stone, R.; Champernowne, D.G., and Meade, J.E. 1942. The Precision of National Income
Estimates,. Review of Economic Studies. 9:111-125.
Varian, Hal. 1992. Microeconomic Theory. New York: W.W. Norton & Company.
Wilcoxen, Peter J. 1988. The Effects of Environmental Regulation and Energy Prices on U.S.
Economic Performance. Department of Economics, Harvard University, Cambridge.
36
------- |