CGE Modeling of Environmental Policy and Resource Management

blusharmenianΔιαχείριση

9 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

65 εμφανίσεις


1

2003
-
01
-
07



CGE Modeling of Environmental Policy and Resource Management


1

Prof. Lars Bergman and Prof. Magnus Henrekson)

Stockholm School of Economics

Department of Economics

Box 6501 (Sveavaegen 65)

SE
-
113 83 Stockholm, Sweden




1.
Introduction

Gene
ral equilibrium theory is a formalization of the simple but fundamental observation that
markets in real world economies are mutually interdependent. Theoretical general equilibrium
analyses have provided important insights about factors and mechanisms tha
t determine
relative prices and the allocation of resources within and between market economies. As
witnessed by, for instance, Debreu (1959) general equilibrium theory has reached a very high
level of rigor and elegance. However, most contributions to gen
eral equilibrium theory have
focused on the allocation of private goods and privately owned resources. The prime
exception is Mäler (1973) who, inspired by Ayres and Kneese (1969), extended the general
equilibrium framework to encompass externalities and e
nvironmental resources with public
goods characteristics.


Computable General Equilibrium (CGE)
2

modeling is an attempt to use general equilibrium
theory as an operational tool in empirically oriented analyses of resource allocation and
income distribution

issues in market economies. The first CGE model was presented in
Johansen (1960), and with the development of fast computers and suitable software a large
number of CGE models has been developed and used for policy analysis. The applications
include analy
ses of major tax reforms, changes in trade policy regimes, economic integration,
agricultural policies and energy policies. A number of CGE models have been designed to
elucidate various policy issues in developing countries
3
.


Since the beginning of the 1
990´s CGE modeling has also become a widely used tool for
analysis of environmental policy and natural resource management issues. The purpose of this
chapter is to review this branch of CGE modeling. The aim is to elucidate the modeling
approaches adopted

and the policy and resource management issues dealt with in
“environmental” CGE models. In addition some specific problems in environmental CGE
modeling will be discussed. A number of specific models will be referred to, but the chapter is



1

Professor of economics at the Stockholm School of Economics, Stockholm, Sweden. Financial support from
the National Swedish Energy Administration, as well as research assistance by Martin Hill and Charl
otte
Nilsson, is gratefully acknowledged. The author is grateful for comments by Eirik Amundsen, Martin Hill and
the editors of the
Handbook

on an earlier version, but is solely responsible for all remaining errors and mistakes.

2

Sometimes this class of n
umerical economic models is called Applied General Equilibrium (AGE) models.
However, as theoretical models of specific classes of economic problems, for instance international trade theory,
can be seen as applications of general equilibrium theory the lab
el “computable” seems more appropriate than
the label “applied”.

3

Robinson et.al. (1999) provide something like a handbook for building CGE models for policy analysis in
developing countries.


2

far from an exh
austive survey of all CGE models intended for environmental resource
management or policy analysis
4
.


The exposition is organized in the following way. In the ensuing section the distinguishing
features, and the potential usefulness, of CGE models are bri
efly discussed. Section 3 is
devoted to a short summary of the history of CGE modeling. In section 4 some general
modeling issues in relation to environmental CGE models are discussed. In section 5 global
models are discussed, while section 6 is devoted to

regional multi
-
country models. Then
single
-
country models are discussed in section 7 and 8, while some concluding remarks on
environmental CGE modeling are made in section 9.


2.
What is a CGE model


and what is it good for
?

There is no precise definitio
n of a CGE model, but whenever this particular label is used the
model in question tends to have certain specific features. A very basic one is that it is a multi
-
sector model based on real world data of one or several national economies. However, while
ge
neral equilibrium theory is concerned with the interactions of large numbers of individual
households and firms most CGE models are rather aggregated. Thus, in a typical CGE model
there is only one or possibly a few “households”, while the number of produc
tion sectors
generally is in the interval 5


50. It is a matter of taste whether numerical models with only a
couple of sectors should be denoted CGE models
5
.



In general the technology is assumed to exhibit constant returns to scale, and preferences are

assumed to be homothetic. Utility and profit maximization behavior on the part of households
and firms is generally assumed, and excess demand functions are homogenous of degree zero
in prices and satisfy Walras´ law. Moreover, product and factor markets
are assumed to be
competitive, and relative prices flexible enough to simultaneously clear all product and factor
markets. A key difference compared to Leontief´s input
-
output model is that in a typical CGE
model the technological coefficients are flexible

and determined by relative prices.


CGE models almost always are focused on the real side of the economy and thus do not
include markets for financial assets. This is one of several differences between CGE models
and the numerical models based on microeco
nomic theory that increasingly are used in
macroeconomics (see Ljungqvist and Sargent (2000))
6
. Consequently a typical CGE model
endogenously determines relative product and factor prices and the real exchange rate, but
cannot determine nominal prices and
the nominal exchange rate. This means that CGE models
generally are aimed at elucidating equilibrium resource allocations and growth paths rather
than business cycle or disequilibria phenomena. In particular CGE models are aimed at
quantifying the impact o
f specific policies on the equilibrium allocation of resources and
relative prices of goods and factors.


Categories of CGE models

In spite of these basic similarities there are also significant differences between individual
CGE models, and a number of d
istinct categories of CGE models can be distinguished. While
several classification alternatives can be envisaged the distinction between
static
and
dynamic




4

Se Conrad (1999) for a recent survey.

5
One
-
sector numerical m
odels such as the climate
-
change models DICE (Nordhaus (1994)) and RICE (Nordhaus
and Yang (1996)), however, should not be classified as CGE models. Yet there is reason to discuss these models
in a survey of environmental CGE modeling.

6

Other features fou
nd in these but not in standard CGE models are heterogeneous agents and incomplete
markets.


3

models seems to be appropriate for a broad classification of modeling approaches. However,
there is

a slight ambiguity with respect to the precise meaning of “dynamic” in this context. It
is obvious that models in which forward looking behavior on the part of households and firms
is assumed and stock accumulation relations are explicitly included should

be denoted
“dynamic”
7
. But several static CGE models are used for multi
-
period analyses. Thus solutions
are obtained for each one of a number of consecutive years, and the solution for an individual
year
t

is used to define the stock of capital and other
relevant assets available in year
t+1
. As
the model is static the agents are implicitly assumed have myopic expectations, i.e. to base
resource allocation decisions entirely on current conditions. In the following I will denote
these models “quasi
-
dynamic”
.


In addition to the static
-
dynamic dimension it is useful to distinguish between
single
-
country
,
multi
-
country
and
global

models. Single
-
country models tend to be more detailed in terms of
sectors and household types, and they are in general used for ana
lyses of country
-
specific
policy issues and proposals. Multi
-
country and global models, on the other hand, tend to have
less sector detail and to be designed for analysis of proposed multi
-
lateral policies such as
free
-
trade agreements. In the case of envi
ronmental CGE models the multi
-
country and global
models in most cases are designed for analysis of trans
-
boundary pollution problems.


Needless to say the models within each one of these categories can differ in many ways. In
particular they may differ w
ith respect to the number of production sectors, the number of
primary factors and the specification of international trade relations.


CGE models and environmental policy analysis

As an initial observation it should be noted that in general there is a cas
e for using a CGE
model if proposed policy measures, or other expected changes in exogenous conditions, are
likely to have general equilibrium effects. However, far from all policy measures related to
environmental and natural resource issues are likely to

have general equilibrium effects. Some
environmental problems are local and site
-
specific. This is the case for air quality and noise
problems in urban areas and in the vicinity of major industrial installations. Other
environmental problems are related t
o specific substances, such as CFCs, that are used only in
a few industrial processes or products, or relatively easily can be replaced by other
substances. Although measures to solve these environmental problems may be quite costly for
some firms and hous
eholds, the repercussions to the rest of the economy often are small or
close to non
-
existent.


However, there are indeed major environmental problems with a much wider geographic and
economic scope, and calling for measures with potentially quite signific
ant effects on the
allocation of resources in the entire national, or even global, economy. “Acid rain”, which is
related to emissions of sulfur and nitrogen oxides, is one example. The prime example,
however, is “climate change”, which is related to emiss
ions of carbon dioxide and other so
-
called greenhouse gases
8
. In both of these cases there is a strong link between the use of
energy and the emissions of pollutants. Moreover, in both cases very significant emission
reductions are considered to be necessa
ry in order to protect the environment. Not
surprisingly CGE models are widely used for evaluation of policies related to climate change
and acid rain issues.





7

A very nice introduction to this class of CGE models is given in Devarajan and Go (1998).

8

See the
Handbook

chapter by Charles Kolstad and Michael Toman.


4

CGE models focused on climate change or acid rain problems basically deal with externalities
and

policies aimed at internalizing externalities. However, environmental problems may also
reflect ill
-
defined property rights, badly functioning capital or insurance markets or some
other kind of market failure leading to poor management of natural resource
s and losses of
environmental amenities. Thus in economies highly dependent on natural resources like
forests, fisheries, agricultural land or grazing land changes in the natural resource
management regime may have economy
-
wide effects, and appropriately d
esigned CGE
models may be able to elucidate and quantify these effects. However, CGE models designed
for analysis of this type of natural resource management issues are likely to differ
substantially in many respects from CGE models designed for analysis o
f problems related to
externalities. The two types of CGE models are also likely to be used in quite different
settings.


In the following I will treat each category as a separate type of environmental CGE model.
Lacking a better terminology the first cat
egory will be called “Externality CGE Models”,
while the second will be called “Resource Management CGE Models”. It should immediately
be pointed out, however, that in terms of numbers the “Externality CGE Models” completely
dominate the field of environme
ntal CGE modeling.


What are CGE models good for?

Most authors in the field would probably claim that a CGE model is an appropriate substitute
for an analytical general equilibrium model whenever the size and complexity, in terms of the
number of household
s and production sectors or pre
-
existing taxes and other distortions, make
such a model mathematically intractable. They would probably also claim that a CGE model
is useful whenever the magnitude, and thus not only the sign, of the impact of changes in
ex
ogenous conditions on key economic variables are to be estimated. Needless to say most
evaluations of policy proposals have to be concerned about the magnitude of the impacts of
proposed policy measures, and the effects usually have to be estimated on a re
latively detailed
sector level. Thus in many instances there are a strong case for using multi
-
sector numerical
models for policy analysis. Whether a specific CGE model, or CGE models in general, can
satisfy this need is of course a slightly different issu
e.


CGE models obviously rest upon strong assumptions about optimizing behavior, competitive
markets, and flexible relative prices. In addition lack of data usually prohibits econometric
estimation of key supply and demand parameters. In view of this the v
alidity and usefulness
for policy evaluation of the results generated by CGE models might be, and often is, seriously
questioned. However, there is no general answer to the question about what CGE models are
good for. The usefulness of a carefully designed

and implemented CGE model depends on
what it is intended for and what the alternatives are.


A CGE model of a complex real
-
world economy may be useful simply because it can help the
analyst to identify general equilibrium effects of changes in exogenous
conditions that
initially were not obvious. This is the case even if key parameters of the model are quite
uncertain. Moreover, even if uncertainty about the numerical values of key parameters makes
the magnitude of computed effects of policy changes uncer
tain, the analyst may be able to
safely conclude that the effects in question are “small” or “big”. This is particularly the case
as the computational capacity of modern computers has made it possible to carry out very
extensive sensitivity analyses, and t
hus to find out how uncertainty about parameter values

5

and structural aspects of the model affect the results and conclusions of the analysis
9
.
Sometimes CGE model results may seem counter
-
intuitive and in the process of explaining
such results the modeler

gains deeper insights into the interdependencies in the economy.


3.
The history of CGE modeling
10

The current literature on CGE modeling and economic analyses based on CGE models is vast.
It has developed from three quite distinct origins, each one associ
ated with the contributions
of a particular author. The three authors are Leif Johansen, Herbert Scarf and Dale W.
Jorgenson. In this section I will briefly discuss the contributions of these authors and how
they have influenced the development of CGE mode
ling. I will also briefly comment on the
impact of increasingly efficient computers and software on the development of CGE
modeling, and close the section with a brief account of the origins of environmental CGE
modeling.


Leif Johansen and the MSG model

I
n his dissertation
A
m
ulti
-
s
ectoral study of economic
g
rowth

(1960), the Norwegian
economist Leif Johansen presented a numerical model that soon became known as the “MSG
model”. This model, which is generally seen as the first CGE model, was primarily inte
nded
to be a tool for long term economic forecasting and economic policy evaluation. In the
original version there were 20 production sectors and one aggregated household sector. Public
consumption, net investments and exports were exogenously determined.
Johansen saw the
MSG
-
model as an extended version of an input
-
output model. Thus, keeping the fixed input
coefficients for intermediate inputs, Johansen added value
-
added production functions and
factor markets where market
-
clearing prices for labor and ca
pital were determined.


Although the MSG
-
model had an obvious flavor of Walrasian general equilibrium theory, it
also contained what seemed to be
ad hoc

assumptions about the determination of wages and
the rates of return on capital. Thus, although both la
bor and capital were entirely mobile
across sectors, there were equilibrium inter
-
sector differences in wages and rates of return on
capital. These deviations from Walrasian general equilibrium theory were motivated by the
existence of factors and conditio
ns not explicitly dealt with in the model but likely to have an
impact on the sectoral development of the economy. Among the factors and conditions
mentioned by Johansen were persistent inter
-
sector differences with respect to the
composition of the labor
force, working conditions, uncertainty and the degree of
monopolization of product markets. In view of these conditions a model entirely based on
Walrasian general equilibrium theory was not considered appropriate. Instead the MSG
-
model
was intended to be
an approximation of a complex but largely unknown “true” model.


The MSG model soon became a key instrument for long term economic planning and
forecasting in Norway, and it has been extended in several stages and directions (see Førsund,
Hoel and Longva (
1985). In particular a considerably more elaborated treatment of factor
substitution and energy demand has been incorporated, and a recent version, MSG
-
EE, is



9

Leif

Johansen, whose so
-
called MSG model is generally regarded as the first CGE model, commented upon the
usefulness of his model in the following way: “
The data and the quantitative analysis do serve the purpose of
illustrating the method and the model. But,
at the same time, if I were required to make decisions and take
actions in connection with relationships covered by this study, I would (in the absence of more reliable results,
and without doing more work) rely to a great extent on the data and the result
s presented in the following
chapters. Thus, the quantitative analysis does not solely serve the purpose of illustrating a method. I do believe
that the numerical results also give a rough description of some important economic relationships in the
Norwegi
an reality”
(Johansen (1960)).

10

This section is partly based on Bergman (1990).


6

especially designed for analysis of issues related to energy use and environmental pollution
(Alf
sen et.al. (1996)). It was also the role model for ORANI (see Dixon et.al.(1982)), which is
a very elaborated CGE model of the Australian economy and often referred to as a “Johansen
model”
11
. The MSG
-
model also had an influence on the design of CGE models
of developing
countries (see for instance Adelman and Robinson (1982)).


Herber Scarf and Scarf´s algorithm

Herbert Scarf´s famous algorithm for computing a Walrasian general equilibrium (Scarf
(1967)) was another point of departure for the development of
CGE modeling. Using Scarf´s
algorithm John Shoven and John Whalley proved the existence of and designed a
computational procedure for finding a general equilibrium with taxes (Shoven and Whalley
(1983)). Together with early work on a two
-
sector model by Ar
nold Harberger (Harberger
(1962)) this inspired a series of analyses of tax and trade policy issues within the frame of
Walrasian and Heckscher
-
Ohlin general equilibrium models. A contribution in the same spirit,
but focused on international trade and reso
urce allocation issues in a small open economy, is
Norman and Haaland (1987). An early survey is found in Shoven and Whalley (1984), and a
more textbook
-
like one in Shoven and Whalley (1992).


In contrast to Johansen´s MSG
-
model the models developed within

the Scarf
-
Shoven
-
Whalley tradition were firmly rooted in Walrasian general equilibrium theory. To some extent
the purpose of the modeling was to “put numbers on the theory”. While Johansen obviously
was very concerned about the ability of the model to app
roximately reflect real world
conditions, authors in the Scarf
-
Shoven
-
Whalley tradition have stressed transparency and
consistency with basic economic theory. Moreover, while Johansen focused on economic
growth and long
-
term structural change, most authors

in the Scarf
-
Shoven
-
Whalley tradition
have had a static welfare economic perspective and focused on the efficiency and
distributional effects of various economic policy measures.


Dale W. Jorgenson and econometric general equilibrium modeling

Dale W. Jorg
enson has made several contributions to CGE modeling, but the most unique of
these is the systematic use of econometric methods for parameter estimation. This is in sharp
contrast to most other CGE models where supply and demand function parameters are
est
imated with simple calibration techniques
12
. The development of econometric general
equilibrium modeling was made possible by significant contributions by Jorgenson (and co
-
authors) to production and utility analysis, econometrics and national accounting (s
ee
Jorgenson (1998)). An early user of Jorgenson’s approach to CGE modeling was Hazilla and
Kopp (1990).


Jorgenson´s approach to CGE modeling to some extent combines the Johansen tradition and
the Scarf
-
Shoven
-
Whalley tradition. Thus, as in Johansen´s wor
k there is a focus on capital
accumulation and economic growth. However, while Johansen could only compute the rates
of change of key economic variables at a specific point in time, Jorgenson´s analyses are
based on a fully dynamic model (of the US economy
). Like the models in the Scarf
-
Shoven
-
Whalley tradition Jorgenson´s models are firmly rooted in neoclassical economic theory and
have been used for analyses of the welfare effects of various forms of taxation. But while the
static models in the Scarf
-
Shov
en
-
Whalley tradition were focused on reallocation effects,
Jorgenson´s dynamic models were focused on growth effects of various tax policies.




11

It should be noted, however, that this label often was motivated by the fact that ORANI, like the MSG
-
model,
was solved on the basis of a linearization procedure.

12

For a di
scussion of calibration techniques, see Whalley and Mansur (1984).


7


Computers and software

Needless to say the development of CGE modeling would not have been possible without the
d
ramatic development of fast computers and suitable software. In the early days of CGE
modeling lack of sufficient computer capacity put serious constraints on the size and
specification of CGE models, and lack of user
-
friendly software made CGE modeling a
field
for specialists in numerical methods. Computer codes were model
-
specific and could not
easily be used by other modelers. Moreover, sensitivity analyses to evaluate the uncertainty
about parameter values were time
-
consuming.


A major change came with
the introduction of GAMS (
G
eneral
A
lgebraic
M
odeling
S
ystem,
Brooke et. al. (1988)), which allowed non
-
specialists in numerical methods to design and
solve Walrasian models. More efficient computers made it possible to solve models with
more sectors, and t
o take the first steps towards dynamic CGE modeling. It also made
extensive sensitivity analysis feasible. As a result the use of CGE models expanded rapidly.
The recent developments of GAMS/PATH (Ferris and Munson (2000)
13
) have made it easy to
solve dynam
ic models with a relatively large number of sectors at a low cost in terms of time
and money. This means that CGE modeling gradually has become an accessible tool for
applied economics, and “solution time” is no longer an issue for CGE modelers. Instead it

is a
typical feature of modern CGE studies that a very extensive sensitivity analysis, in which the
model is solved for several thousands of randomly selected combinations of values of the
uncertain parameters, is carried out.


Environmental CGE modeling

In connection with the Energy Policy Project in the early 1970´s, summarized in the volume
A Time to Choose (Ford Foundation (1974)), Hudson and Jorgenson developed an
econometric CGE model for energy policy analysis (see Hudson and Jorgenson (1975)). This

turned out to be the first of a large number of models designed for analysis of energy policy
issues in the wake of the oil price increases in 1973 and 1979. However, most of these models
were energy sector models in which the rest of the economy was repr
esented by an
exogenously determined rate of growth of energy demand. A well
-
known exception is Alan
Manne´s so called ETA
-
MACRO model in which a detailed energy technology assessment
model was linked to a neoclassical one
-
sector model of the rest of the e
conomy (see Manne
(1977)).


However, in the beginning of the 1990´s the focus shifted from problems associated with the
supply of energy to the external effects associated with the use of energy, particularly fossil
fuels. One concern was acid rain, but th
e prime concern was climate change caused by
emissions of carbon dioxide. Many of the energy models could easily be redesigned for
analysis of carbon taxation and other types of climate policies. In addition a new set of CGE
models, designed for climate po
licy analysis, was developed. One of the most well known is
the GREEN model developed at the OECD secretariat (see Burniaux et.al. (1992)) for
analysis of climate policy issues at a global scale.


At the same time a number of single
-
country models for env
ironmental policy and resource
management analysis in different individual countries were developed. Thus, Hazilla and
Kopp (1990) estimated the social cost of environmental quality regulations using a CGE
model of the US economy. Bergman (1990) estimated
the social cost of phasing out nuclear



13

See also www.gams.com.


8

power in the presence of SO
2
, NO
x

and CO
2
emission constraints, using a CGE model of the
Swedish economy. In the following sections I will discuss the approaches adopted, the issues
addressed, and some of the conclusi
ons that have been drawn in environmental CGE
modeling.


4.
Some General Issues in Environmental CGE modeling

Most environmental CGE models are designed to elucidate various aspects of climate change
or, in some cases, acid rain policies. To a large extent

climate change and acid rain problems
are caused by emissions from the combustion of fossil fuels. In both cases the environmental
damage depends on the accumulated stock rather than the current flow of pollutants.
Moreover, the stocks of the pollutants i
n question accumulate slowly so there is a
considerable time lag, particularly in the case of climate change, between the emission of
pollutants and the resulting impact on the environment. These observations have several
implications for the design of CGE

models intended for policy analysis.


One obvious implication is that the model should have an elaborated treatment of the supply
and demand for energy. In particular it should have an elaborated treatment of the possibilities
to substitute other forms o
f energy, or other factors of production, for fossil fuels. It should
also have an explicit treatment of the relation between the use of fossil fuels and the emission
of various pollutants.


Another implication of the nature of the environmental problems
in question is that the model
should take stock accumulation over very long periods of time into account. While the time
horizon is one or two decades into the future in typical CGE model analyses of tax or trade
policies, the relevant time horizon in clim
ate change policy analysis is several decades or even
a century or two into the future. The key modeling problem with such a distant time horizon is
that the potential impact of new technologies is quite significant.


A third implication for CGE modeling i
s related to the fact that the benefits of environmental
policy measures are “non
-
economic”, i.e. that they come in the form of better environmental
quality. Thus a CGE model intended for cost
-
benefit analyses of environmental policies
should have an “envi
ronmental module”, i.e. a module in which the environmental benefits of
reduced pollution are quantified and converted into a monetary measure of environmental
benefits. The environmental module could also include “feed
-
back” mechanisms, i.e. a sub
-
model o
f the impact of environmental improvement (or deterioration) on factor productivity
and household utility of environmental services.


It is obvious that environmental CGE modeling is quite a demanding task, and that the
modeler is bound to encounter a numb
er of intricate modeling issues. It is also obvious that
environmental CGE models should be dynamic or at least quasi
-
dynamic. The purpose of this
section is to briefly discuss some commonly adopted modeling approaches in this particular
field of CGE model
ing.


Production sectors

CGE models intended to elucidate climate change or acid rain policies need to have an
elaborated treatment of the demand for fossil fuels. This has certain implications for the
specification of production functions, but also for t
he production sector division of the model.
In particular there is a case for treating the fossil fuel intensive sectors as separate production
sectors. For this reason a typical “externality” CGE model has separate production sectors for
electricity, tran
sportation, metals, pulp and paper, and chemicals, while the rest of the

9

economy may be aggregated into only a few production sectors. When an elaborated
environmental module is incorporated, however, sectors that are affected by climate change
(for instan
ce agriculture) or acid rain (for instance forestry) are treated as separate production
sectors (see Nordhaus (1994) and Hill (2001)).


However, production sectors that are fossil fuel intensive may consist of sub
-
sectors that
differ significantly from th
is point of view. This is clearly the case for the electricity sector
where the output can be produced both by fossil fuel intensive technologies such as coal and
oil power, and fossil fuel free technologies such as hydroelectric power and nuclear power.
T
hus part of the electricity sector response to climate policy measures is to change the mix of
different technologies used for power production. In order to capture these substitution
possibilities in a realistic way the technological constraints of the el
ectricity sector, or the
entire energy sector, is sometimes represented by a separate sub
-
model rather than by a
standard neoclassical production function.


The common origin of the energy sector sub
-
models is the linear activity models used for
planning a
nd technology assessment in the energy sector. An elaborated example of a global
model in this tradition is Nordhaus (1974). The key feature of these models, often called
“bottom
-
up” models, is that individual technologies for energy extraction, conversion

and
transportation are distinguished. Among other things this modeling approach makes it easy to
incorporate new technologies, such as wind power, with factor input proportions that radically
differ from existing technologies. On the other hand the linear

energy sector model has to be
integrated with “neoclassical” models of the non
-
energy sectors. An early example of an
integrated “bottom
-
up” energy sector model and a neoclassical “rest
-
of
-
the
-
economy” model
is Alan Manne´s above mentioned ETA
-
MACRO model
. Other examples are Jorgenson
(1982) and Lundgren (1985),


The transportation sector is similar to the electricity sector in the sense that there are several
different modes of transportation that exhibit very different properties with respect to the use

of fossil fuels per unit of output. The long run response to climate policy measures affecting
the transportation sector are likely to include adjustments and substitutions both on the supply
and the demand side. However, the modeling of the transportatio
n sector is usually not very
elaborated in CGE models intended for environmental policy analysis. The different modes of
transportation are often not explicitly distinguished, and there is no measure of the
transportation services produced by ordinary firm
s and households.


Moreover, while the demand for transportation services obviously reflects the location of
production and consumption activities, most CGE models do not have a spatial dimension.
And while the choice between different modes of transportat
ion to a large extent depends on
the amount of time that the user has to spend, time is usually not treated as a scarce factor in
CGE models. In addition the relative competitiveness of different modes of transportation to a
large extent depends on the tra
nsportation infrastructure, i.e. roads, railways, airports, etc.
Altogether this means that the CGE models developed so far have little to say about the
demand for and substitution between different modes of transportation. In order to account for
the rele
vant substitution opportunities some kind of “bottom up” approach might be needed.


Production functions

The sectoral production functions basically define substitution possibilities between explicitly
defined input factors. In CGE models focused on envir
onmental policies related to climate
change or acid rain it is important to distinguish not only between capital, labor, non
-
energy

10

intermediate inputs and energy, but also between fossil and non
-
fossil energy. Often it is also
convenient to distinguish be
tween fuels and electricity. Thus the production function of a
representative production sector
j

in such a CGE model can be written


)
,
,
,
,
(
j
j
j
j
j
j
j
E
F
M
L
K
f
X

;





(1)


where
X

is gross output,
K

capital,
L

labor,
M

non
-
energy intermediate in
puts,
F

fuels and
E

electricity. In most cases F is an aggregate of various fossil and non
-
fossil fuels. In the
following non
-
energy intermediate inputs are denoted “materials”.


In some CGE models the production function
f
j
(.)
, or rather its dual cost fun
ction, is assumed
to have a so called flexible form (translog or generalized Leontief) and the parameters are
econometrically estimated. The use of flexible functional forms is a way to circumvent the
strong assumptions about the elasticities of substituti
on between different pairs of inputs
implied by the standard production functions. To some extent these functional forms were
developed in order to properly deal with the substitutability of energy and other factors of
production in econometric general equ
ilibrium models (see Jorgenson (1998a)). However,
lack of data often prevents econometric estimation of the sector cost functions. Instead the
elasticities of substitution between different inputs generally are “guesstimated”. This means
that both the nest
ing structure of the production functions and the adopted numerical values
are based on literature surveys of relevant econometric studies.


Thus, based on available external information about elasticities of factor substitutions the
technology in most CG
E models is described by some kind of nested production function
structure in which CES (constant elasticity of substitution), Cobb
-
Douglas and Leontief
production functions are combined. The existing literature on econometric studies of
production does no
t lead to definite conclusions about the most appropriate nesting structure.
However, in most models fuels and electricity, i.e.
F

and
E

in the equation above, are
combined in a CES function with a relatively high elasticity of substitution. The input “fue
ls”
is often defined as a CES
-
aggregate of different types fossil and non
-
fossil fuels. The
elasticities of substitution between different types of fuels are usually taken to be relatively
high.


In the case of capital and energy the econometric evidence i
s conflicting. Some studies
indicate that capital and energy are substitutes at the relevant level of aggregation, while
others suggest that capital and energy are complements. However, most CGE models assume
that capital and energy are substitutes, althou
gh the elasticity of substitution between capital
and energy is generally taken to be quite low. The nesting structure may differ between
different models, but the structure of the sector production function (2) below can be found in
many CGE models intend
ed for climate change or acid rain policy analysis.


))
),
,...,
(
(
,
(
,
,
(
1
j
jn
j
j
j
j
j
j
j
j
j
E
F
F
F
H
K
Q
M
L
f
X





(2)


Thus fuels (
F
), which is an aggregate of
n
different types of fossil and non
-
fossil fuels, and
electricity (
E
) are combined in a CES aggregate that defines

a composite energy good (
H
).
The composite energy input is combined with capital in a CES aggregate of capital
-
energy.
Then the composite capital
-
energy input
Q

is combined with labor (
L
) and materials (
M
). In
some models, however, capital and labor rathe
r than capital and energy are combined.



11

Emissions and abatement

In general the emissions of pollutant per unit of output can be reduced if the input of one, or
several, of the other inputs is increased. Thus the possibility to emit pollutants into the
env
ironment can be seen as a kind of input in the production process, and it should be
possible to estimate the elasticity of substitution between emissions into the environment and
other input factors. Estimation of these substitution possibilities obviously

requires that the
factor inputs and the emissions of pollutants can be appropriately measured. However, the
emissions of pollutants such as sulfur and nitrogen oxides and carbon dioxide generally are
not measured directly, and in many cases direct measure
ment is difficult and costly. Instead
the emissions are estimated on the assumption that they are proportional to the use of various
types of fossil fuels
14
.


This assumption implies that emission reductions can be brought about only by reductions of
the c
onsumption of fossil fuels or by changes in the composition of fossil fuel consumption.
In practice inter
-
fuel substitutions can lead to quite significant emission reductions. For
instance, the combustion of natural gas gives rise to less emissions of carb
on dioxide per unit
of energy than coal. Thus substitution of natural gas for coal
ceteris paribus

reduces the
emissions of carbon dioxide at give output levels.


However, the emissions of sulfur and nitrogen oxides can be reduced not only by output
reduct
ions and by fuel switching. There are also direct abatement possibilities. In order to
capture abatement measures some environmental CGE models incorporate abatement cost
functions, usually estimated on the basis of generic rather than site
-
specific engine
ering data.
In representative CGE models the abatement activity is assumed to depend on economic
incentives so that abatement takes place whenever the marginal cost of abatement is less than
or equal to the cost to the firm, or household, of marginal emiss
ions. The marginal cost of
emission, in turn, is determined by charges on emissions or by the price of emission permits
(see for instance Hill (2001)). From an institutional point of view it is assumed that
specialized firms are supplying abatement service
s to industries obliged to comply with
emission constraints.


Technological change

In the short and medium term substitution between inputs is a key mechanism in the
adjustment to various environmental policy measures. This is why the elasticity of
substit
ution parameters of the sector production functions are so important in CGE models
intended for environmental policy analysis. However, as was mentioned above the time
horizon in environmental policy analyses often extends several decades or even a century

or
two into the future. Thus the development and implementation of new technologies might
affect emissions and other impacts on the environment much more than substitution between
currently existing technologies. Expectations about future relative prices,

taxes and
regulations clearly have an impact on the speed and direction of technological development.


The links between past and current conditions, the formation of expectations about the future
and the development and implementation of new technologie
s are not well understood.
Nordhaus (1997) discusses induced technical change in the context of the optimal timing of
abatement measures. Goulder and Schneider (1999) introduce a market for R&D services in a
CGE model. As the R&D services can be used as a
substitute for other factors of production,
this means that technological change in effect becomes an endogenous process.




14

Even if the emissions of sulfur and nitrogen oxides could be independently measured, the lack of uniform
prices (or emission charges) would cause estimation probl
ems.


12


However, in most CGE models technological change is an exogenous factor making the total
factor productivity an increasing function o
f time. In CGE models intended for energy or
environmental policy analysis it is quite common to incorporate specific assumptions about
“autonomous energy efficiency improvements” (AEEI). The AEEI
-
factor is assumed to be
exogenously determined and to refle
ct all factors, except current price
-
induced substitutions,
that make the input of energy in a given production sector grow slower than the output of that
sector. The numerical value of the AEEI
-
factor is often assumed to be in the interval 0
-
2
percent per

annum.


Needless to say an AEEI
-
factor at the level of one percent per annum or more has a very
significant impact on energy use, and thus on emissions, in a 50
-
100 years time perspective.
Thus the assumptions made about the numerical value of AEEI in ke
y production sectors may
have a very significant impact on the results of the whole modeling exercise. As the CGE
model is supposed to elucidate the impact of changes in relative prices on the allocation of
resources in the economy, it is of course somewha
t disturbing to be forced to treat
technological change as an exogenous factor. What is even more disturbing is that the
assumptions about positive AEEI
-
factors seem to rest on somewhat uncertain empirical
grounds. Thus Hogan and Jorgenson (1991) show that

there is no clear evidence of
autonomous energy efficiency increases if price
-
induced substitution effects are taken into
account.


In environmental CGE models with a “bottom
-
up” description of the technology of the energy
sector it is quite common to inc
orporate a “back
-
stop” technology that becomes available
some time in the future. The back
-
stop technology is typically based on non
-
exhaustible
resources. The date when the back
-
stop technology is available is exogenously determined,
but whether the back
-
stop technology will be used at that date, at some later date or not at all
is endogenously determined in the model.


Environmental benefits

One way of using an environmental CGE model is to focus on the cost of specific
environmental policy measures (such

as a ban on the use of nuclear power), or on the cost of
attaining a specific environmental policy goal (such as reducing the total sulfur emissions by
30 percent). However, if the model is to be used for evaluation of policies it should be capable
of qua
ntifying both the costs and the benefits of the policies in question. This means that the
CGE model needs to have an “environmental module” in which the environmental benefits of
reduced pollution are quantified and expressed in monetary units. From a pure
ly theoretical
point of view the development of such an environmental module is fairly straightforward. In
reality a lack of relevant and reliable data makes it an almost impossible task.


What is needed in order to construct a “benefit function” can be di
vided into two sets of
functional relationships. The first is a set of physical damage functions that convert emissions
and other environmental effects of production and consumption into measures of physical
environmental damage (in the case of increased e
missions etc.) or improvements (in the case
of reduced emissions etc.). The estimation of such functions is obviously outside the realm of
economics, and it does not seem to be a prime concern for natural scientists. The second is a
set of functions defini
ng the value, in monetary units, of changes in the physical
characteristics of the environment. From an economic point of view these changes can take
two different, but not mutually exclusive, forms (see the
Handbook

chapter by Nancy
Bockstael and A. Myric
k Freeman III).


13


One is that the physical changes in the environment affect the supply of environmental
services that are directly “consumed” by the households. In terms of a CGE model this means
that changes in environmental quality affect welfare direct
ly via the utility functions of the
household(s). Obvious examples of such services are clean air and water. In these and most
other cases the environmental services in question can be characterized as public goods. Thus
the relevant values cannot be deter
mined on the basis of regular market prices. Instead the
valuation has to be based on some estimate of the willingness to pay for the environmental
services in question.


The other alternative is that the changes in the physical characteristics of the envi
ronment
only affect the productivity in sectors producing “ordinary” goods and services that are traded
on regular markets (see the
Handbook

chapter by Kenneth McConnell and Nancy Bockstael).
In terms of a CGE model this means that changes in environmental

quality affect welfare
indirectly via the cost of producing ordinary goods and services. The impact of environmental
damage on the cost of ordinary goods and services is an example of what is sometimes called
“feed
-
back effects”
15
. One example of such a fe
edback effect is the reduction of factor
productivity in forestry that may be the result of acid deposition caused by emissions of
sulfur. In this case the cost of the environmental damage can be estimated on the basis of
regular market prices for forest p
roducts.


Many environmental CGE models lack a module for environmental benefit calculation, or
have an environmental module that is based on shaky data and/or very bold assumptions.
Basically two types of approaches have been adopted. One is to focus on f
eedback effects. An
elaborate example is Nordhaus (1994) in which an advanced climate model is used to estimate
feedback effects of the emissions of green
-
house gases. Examples of CGE models with
explicit feedback effects are Harrison et.al. (1989), Vennem
o (1995) and Hill (2001). Another
approach is to assume that politically determined environmental goals, or international
agreements on emission reductions, represent an efficient trade
-
off between the relevant costs
and benefits. Given this assumption the

parameters of an environmental benefit function can
be determined. The benefit function can then be used to evaluate other policy proposals. One
example of this approach is Whalley and Wigle (1992).


International trade in CGE models

The specification of
international trade relations is an important aspect of all open
-
economy
CGE models, but seems particularly important in environmental CGE models. The primary
reason for this is that international relocation of economic activity is a key potential response

to unilateral environmental policy measures. It is beyond the scope of this chapter to discuss
specification issues in any detail, but a few words should be said about the treatment of
international trade in CGE models of open economies.


The natural poi
nt of departure then is the Heckscher
-
Ohlin model of a small open economy in
which the technology exhibits constant returns to scale, and the domestic producers are price
-
takers on international markets for tradable goods. However, with
n

goods and
m

facto
rs, and
n
>
m
, the equilibrium output levels in such a model are positive in at most
m

sectors.
Moreover, a small change in a world market price, or a domestic tax rate, may reduce the
equilibrium output level in a given sector from a relatively large positi
ve value to zero, or
from zero to a relatively large positive value. As most CGE models have many more sectors



15

Another type of feedback effect is the change in the demand for ordinary goods and services caused by
changes in environmental quality.


14

than factors this feature of models in the Heckscher
-
Ohlin tradition tends to produce rather
extreme and unrealistic patterns of specialization.
This so
-
called overspecialization problem
has attracted a lot of attention in the CGE
-
modeling literature, and several “solutions” have
been proposed.


The most widely used approach is to adopt the “Armington assumption” (Armington (1969)),
which implies t
hat goods with the same statistical classification but different countries of
origin are treated as non
-
perfect substitutes. The application of this idea in CGE models
amounts to defining domestically consumed goods as CES
-
aggregates of domestically
produc
ed and imported goods with the same statistical classification. As a result
16

the import
of a given type of goods depends on the relation between the prices of imported and
domestically produced goods of that type. Moreover, if the same assumption is applie
d on the
rest of the world the producers of the small open economy will face relative
-
price dependent
export demand functions, and the terms of trade will depend on the volume of exports. This
means that the properties of CGE models based on the Armington
assumption
17

may differ
quite significantly from the properties of models based on standard Heckscher
-
Ohlin
assumptions.


Another widely adopted approach to the “overspecialization” problem in CGE models is to
retain the assumption about exogenously given
terms of trade, while relative
-
price dependent
export supply functions are added. These functions usually are derived from constant
elasticity of transformation (CET) functions defining the output of a given sector as a
revenue
-
maximizing aggregate of good
s for the domestic market and goods for foreign
markets. This means that if the price of, say, goods for foreign markets increase the
composition of domestic supply is shifted in the direction of more goods for foreign markets
and less for the domestic mar
ket. The magnitude of the response to changes in relative prices
depends on the elasticity of transformation between goods for the two types of markets.


Both the Armington assumption and the CET function approach prevent extreme
specialization patterns i
n CGE models with more tradable goods than factors. However, the
empirical basis for these approaches seems somewhat questionable. Product differentiation,
which is implied by both approaches, clearly is a real world fact, but the patterns of product
diffe
rentiation depend on market conditions and change over time. In CGE models employing
the Armington assumption or using the CET function approach, however, the current patterns
of product differentiation in effect are assumed to persist. This means that the

models are
likely to underestimate the structural effects of long run changes in relative prices. As will be
discussed in some detail below this feature might be particularly important in CGE models
intended for analysis of environmental policies.


5.
Glo
bal “Externality” CGE Models

During the 1990´s a number of global CGE models intended for analysis of climate change
policies were developed and used for policy analysis. The major field of application has been
evaluations of various aspects of the Kyoto p
rotocol, i.e. the (not yet ratified) agreement to



16

The Armington assumption implies that the price of the domestically consumed composite of a given type of
goods

is a linearly homogenous function of the prices of imported and domestically produced goods of that type.
By Shephard´s lemma the share of imports in the composite good is given by the partial derivative of the price
function with respect to the price of
imports.

17

From a microeconomic point of view an Armington model of a small open economy is somewhat questionable
in the sense that the firms in a given sector collectively face a downward sloping export demand function, but
refrain from forming an export
cartel and exploiting their market power on foreign markets. Harris (1984)
avoided this problem by modeling product markets as monopolistically competitive.


15

reduce the emissions of carbon dioxide and other greenhouse gases. In fact the commitments
by industrial countries under the Kyoto protocol seem to be the primary reason so many
global environmental CGE mod
els were developed during the 1990´s. The purpose of this
section is to briefly present some of the most well known models, and to discuss some of the
results obtained from simulations with global “externality” models. In particular I will discuss
what we
can learn from the models about the so
-
called “leakage” problem, i.e. the alleged
international relocation of emission
-
intensive production induced by unilateral climate policy
measures.


The models

The key characteristics of the selected models are summar
ized in Table 1. This is not a
complete list of all existing global CGE models for environmental policy analysis. However,
the collection of models included in the list should give a fairly complete account of the
modeling approaches, in terms regional and

sector division and several other dimensions,
generally adopted in this field of CGE modeling. Moreover the list to some extent reflects the
continuing development of the several of the models. Thus MIT
-
EPPA is an upgraded and
extended version of GREEN. I
n the same way 12RT is an extended version of Global 2100,
and RICE is a regionalized version of DICE (Nordhaus (1994)).


A common feature of the models is that baseline GDP growth essentially is determined by the
assumptions made about aggregate savings,
technological change and the growth of the labor
force in different regions of the world. This means that the emissions of green
-
house gases to
a large extent also is determined by these assumptions. However, the Kyoto commitments are
defined in terms of e
mission reductions in relation to a historical benchmark. Thus the
stringency of the imposed emission constraints, and to a large extent the cost of complying
with these constraints, in effect is determined by the assumptions about baseline economic
growth
. For instance, in Manne and Richels (1992) the baseline growth assumptions imply
that China will grow faster than the world average, and increase its share of world GDP from
1.8 percent in 1988 to 22.1 percent in the year 2100. Under these conditions it t
urns out that
global emission reduction policies will be very costly, particularly if no emission reduction
measures are implemented in China. If China grows more slowly than the world average,
however, attaining the emission reduction targets will be cons
iderably less costly.


Another common feature is that the models are used for simulations over periods that are long
enough to make resource depletion effects important. In order to capture depletion effects
Global 2100 and 12RT distinguish between two cat
egories of oil and gas resources. Thus the
cost of extracting oil and gas from currently proven reserves is lower than the corresponding
cost for the remaining but still unproven stock of reserves. In GREEN coal reserves are
assumed to be (practically) inf
inite, while oil and gas are assumed to be exhaustible resources.
The cost of these resources is taken to depend on the initial levels of proven and unproven
reserves, the rate of reserve discovery, and the rate of extraction. Moreover the rate of reserve
discovery is assumed to depend on world oil and gas prices. Mechanisms that reflect
increasing cost of oil and gas as currently proven reserves are exhausted are also incorporated
in CRTM, IIAM and the G
-
Cubed model. However, in the models where a back
-
sto
p
technology is incorporated the long run cost of energy is capped by the cost of using the back
-
stop technology for energy production.


The “leakage” issue

According to the Kyoto protocol the so
-
called Annex I countries, i.e. the high
-
income OECD
countrie
s, should start reducing their carbon dioxide emissions before the countries in the rest

16

of the world. One possible effect of such a policy is “carbon leakage”, i.e. emission sources
migrate from abating to non
-
abating countries. The possibility of carbon
leakage is a matter of
great concern in several countries, and it seems to be a major obstacle to unilateral emission
reduction policies. From a theoretical point of view it is obvious that unilateral action will
induce some “leakage”. The question is whet
her the leakage is quantitatively significant or
not. A global CGE model should be suitable tool for assessing the magnitude of carbon
leakage due to unilateral emission reduction policies.


Pezzey (1992) used the WW
-
model to estimate the leakage effects o
f unilateral European
Community and OECD carbon dioxide emission reduction policies. Assuming that the
emission target was 20 percent below the baseline level, the leakage turned out to be 70
percent. Thus, if OECD reduced emissions by 100 tons of carbon d
ioxide, the resulting
reduction of global emissions would only be 30 tons. Rutherford (1992) also finds that the
leakage is significant. In particular he finds that if OECD increases the emission reduction
target from 4 percent to 5 percent of current emis
sions, there would be no reduction at all of
global emissions. Thus, according to this particular study the “marginal” leakage effect is 100
percent!


However, other CGE
-
based studies have come to very different results concerning the leakage
effect. McKib
bin and Wilcoxen (1995) studied unilateral emission abatement by the Annex I
countries in accordance with the Kyoto protocol, and found that the leakage effect was 6
percent. In experiments with GREEN the leakage effect was only 3.5 percent when the OECD
c
arbon dioxide emissions were stabilized at the 1990 level. In 12RT the leakage effect is
smaller than in the studies by Pezzey and Rutherford, but significantly bigger than in GREEN
and the study by McKibbon & Wilcoxen. Thus the estimated leakage effect in

12 RT when
Annex I countries act unilaterally is 35 percent.


In order to explain these big differences Manne and Martins (1994) made a systematic
comparison of 12RT and GREEN simulation results. They found that the leakage effect
reflected two main effe
cts of climate policy measures. The first was due to the fact that a tax
on carbon dioxide emissions raised the cost of energy intensive production in the abating
countries. Thus the relative competitiveness of tradable energy intensive production in the
n
on
-
abating countries increased, and part of the production decrease in the abating countries
was compensated by increased production and net export in the non
-
abating countries. This
effect can be called the “relocation effect”. The second effect was due t
o the fact that a
reduction of energy consumption in the abating countries tends to reduce the world market
prices of oil and coal. As a result of the lower prices the consumption of energy, and the
emissions of carbon dioxide, increases in the non
-
abating

countries. This effect can be called
the “rebound effect”.


The large difference between the two models in terms of the leakage effect turned out to
primarily depend on the relocation effect, which in turn depended on differences in the
treatment of inter
national trade. In GREEN international trade is modeled in accordance with
the Armington assumption. This means that similar goods with different country of origin are
treated as imperfect substitutes, which tends to dampen the relocation effect. In 12RT,
on the
other hand, non
-
energy goods with different country of origin are assumed to be homogenous,
which tend to make international trade flows sensitive to changes in relative cost conditions
between countries. These findings suggest that the specificatio
n of international trade in
global CGE models may have a significant impact on the results and thus deserves
considerable attention. However, as is demonstrated by McKibbin and Wilcoxen (1995) any

17

factor that makes structural change costly tends to reduce
the leakage effect of unilateral
climate policies. In their case international relocation of capital is hampered by capital
installation costs in the non
-
abating countries.


Concluding remarks

A lot more can of course be said about the global models and th
e studies in which such
models have been used for analyses of environmental policy proposals. However, it suffices
to conclude that these models and studies to a large extent have formed the “common
wisdom” about the economic consequences of the policies s
uggested by the Kyoto protocol. It
could be added that the Kyoto protocol has created an ideal case for using CGE models: the
time horizon is far into the future so that short term adjustment problems can be neglected. At
the same time the proposed emissio
n reductions are significant and clearly call for policy
measures that are likely to have general equilibrium effects both within and between countries
and regions.


6.
Regional Multi
-
Country “Externality” CGE Models

A model in this category typically cov
ers a region, such as the European Union, and consists
of sub
-
models of each one of the countries within that region. From an environmental policy
point of view regional models are suitable for analyses of regional environmental problems,
such as acid rain
. Regional multi
-
country CGE models are also used for analyses of policy
proposals that imply coordination of the national policies within the region. One example is
analyses of the implications of the Kyoto protocol where the European Union, rather than t
he
individual member states, is a signatory party. In this section three representative models and
a special problem associated with this category of CGE models will be briefly discussed.


The models

The main features of the three models are summarized in
Table 2. GEM
-
E3, where E3 stands
for
E
nergy,
E
nvironment and
E
conomy, is the result of a major project within the JOULE
program funded by the European Union. The GEM
-
E3 model has a quite detailed treatment
both of the energy sector and the emissions to the

environment. Thus, within the “top
-
down”
energy sector four types of energy, namely electricity, coal, oil and gas, are distinguished.
Among other things this allows for a relatively detailed link between the consumption of
energy and the emissions to the

environment. In GEM
-
E3 the emissions of five different
pollutants, CO
2
, SO
2,
NO
x

and PM (particulate matter), are distinguished, and abatement cost
functions for all pollutants except CO
2

are included. In addition to “end of pipe” abatement
options possib
ilities to substitute less polluting forms of energy, and other factors of
production, for polluting forms of energy are included in the model.


The GEM
-
E3 also has a module in which emissions to the environment are converted into
damage to the ecosystem a
nd to public health. In addition damage to materials is included.
However, no feedback
-
effects are included in GEM
-
E3. In the extended version of the HRW
-
model, on the other hand, feedback effects are included. Thus the environmental module
distinguishes t
he emissions of CO
2
, SO
2

and PM, and mortality and morbidity effects are
assumed to depend on the stocks and flows of these pollutants. Increased morbidity has both
direct welfare effects and feedback effects on the demand for “ordinary” goods and services
.
Increased mortality is modeled as separable and turns out to be the most significant welfare
effect of emissions to the environment. The BFR, model, finally, only treats CO
2

emissions,
and does not include environmental benefit functions or feedback effe
cts.



18

Specific features and problems

Regional multi
-
country CGE models share many of the features of global CGE models. Both
types of models are often used for evaluations of unilateral vs. internationally coordinated
policies. As in global CGE models an e
laborated treatment of international trade between the
individual countries is needed in a regional multi
-
country model. Data problems are usually
less severe in regional multi
-
country models than in global models, and problems associated
with aggregation
of data for individual countries can in general be avoided. However, there is
a modeling problem that is unique to the regional multi
-
country models, namely the treatment
of “the rest of the world” (ROW).


In global models there is by definition no ROW, so

the problem does not have to be
addressed. In single
-
country models there is clearly a ROW, but in general it is reasonable to
adopt the “small country assumption”. In other words it is assumed that world market prices
are given and unaffected by changes
in the export, import and factor prices of the single
country in question. In regional multi
-
country models, however, the small country assumption
often is untenable. For instance, the European Union accounts for a very significant share of
the world trade
, and it is likely that the changes the Union’s trade with the rest of the world
would influence world market prices. However, if the small country assumption is not
adopted some other type of “world closure” has to be included in the model.


The problem o
f “world closure”, i.e. the modeling of how ROW would act in response to the
actions of the countries explicitly included in the model, has been widely discussed in the
literature on CGE modeling. Early contributions are Whalley and Yeung (1982) and de Mel
o
and Robinson (1989). Koschel and Schmidt (1998) used the GEM
-
E3 model for an extensive
test of the closure rules suggested by Whalley & Yeung and de Melo & Robinson.


In the standard version of GEM
-
E3 it is assumed that the ROW export prices are given a
nd
independent of the demand for ROW goods by the European Union. The ROW import
demand functions are modeled in accordance with the Armington assumption. Thus the ROW
demand for imports from the European Union countries depends on the ratio between
exogen
ously given ROW prices and endogenously determined European Union export prices,
and an exogenously given level of output in ROW. In various experiment versions of GEM
-
E3 alternative trade specifications were used. The key difference was that ROW prices we
re
assumed to depend on the quantity of ROW exports to the European Union. The conclusion
was that the assumptions about the behavior of ROW did influence the results for the
countries explicitly treated in the model. This is hardly surprising, and suggest
s that the only
satisfactory way to deal with the “world closure” problem is to extend the regional multi
-
country model to become a global model, albeit one with a less detailed treatment of ROW.


7.
Single
-
country “Externality” CGE models

A large number o
f single
-
country environmental CGE models were developed during the
1990´s. Most were designed to elucidate environmental problems or policies that are specific
to the country in question. The most commonly studied environmental problem is emissions
to air

that contribute to climate change or acid rain. Thus most single
-
country environmental
CGE models can be characterized as “externality” CGE models. However, there are also a
few single
-
country models that can be classified as “resource management CGE mode
ls”, i.e.
are designed to elucidate problems like depletion of natural resources and other natural
resource management issues. A selection of representative models is briefly described in
Table 3.



19

Single
-
country environmental CGE models can broadly be di
vided into three main categories.
The first consist of models that primarily are constructed and used for analyses of specific
theoretical issues. The prime example is CGE analyses aimed at testing the existence of a so
-
called “double dividend” (see the
Ha
ndbook
chapter by Anil Markandya for a detailed
discussion of this issue). Goulder et.al. (1997) and Bovenberg and de Moij (1994) both belong
to this category. The second category is CGE models that are constructed for testing of new
model features or mode
ling approaches, which, if the testing is successful, may be
incorporated in models intended for policy analysis. Vennemo (1995), where an approach to
incorporate feedback effects from the environment to the economy, belongs to this category.
Another examp
le is Abler et.al. (1999) where the impact of parameter uncertainty on
simulation results is studied. The third is “multi
-
purpose” CGE models that are designed for
analyses of a wide range of economic and environmental policies. Harrison (1997) belongs to
this category. In the following the use of single
-
country CGE models for analysis of the
“double dividend” issue and some natural resource management issues will be briefly
discussed.


The double dividend issue.

The idea that revenues generated by emission

taxes could be used to reduce distortionary
taxes, and thus produce benefits in addition to those resulting from reduced emissions, has
been widely discussed in relation to environmental policy in several countries. Goulder
(1995a) contributed to the disc
ussion by defining three types of “double dividends”. The most
interesting was the “strong” double dividend which refers to a case where a revenue
-
neutral
substitution of an environmental tax for a representative distortionary tax would lead to a non
-
posit
ive welfare cost. In other words emission taxes could be welfare improving even if the
environmental benefits were small or even zero!


The existence of a strong double dividend seems to have been taken for granted by many
politicians. In particular repla
cement of part of the labor income tax with emission taxes has
been seen as an environmentally attractive way of increasing the demand for labor, and such
tax reforms have been proposed in several and implemented in some countries. Economists,
however, hav
e been more skeptical and the issue has become subject to extensive research. To
a large extent this research has been based on CGE models.


The double dividend issue offers an ideal case for CGE modeling. It is not a matter of
studying the impact of an en
vironmental tax in an economy without taxes, but in an economy
with an extensive system of distortionary taxes. Thus the existence of the strong double
dividend depends both on how the environmental tax interacts with other taxes, and on how
the revenues a
re recycled. Moreover, even if the sign of the double dividend is a key issue its
magnitude is quite important from a policy point of view. What, then, can we learn about the
double dividend from CGE model based studies?


Using a static CGE model Bovenberg

and de Moij (1994) show that a strong double dividend
is possible only if the labor supply function is backward
-
bending, which is not consistent with
the findings in empirical studies of labor supply behavior. On the basis of this result they
concluded th
at the strong double dividend does not exist for realistic values of the relevant
elasticity parameters. However, a static model is not well suited for analyses of investments
and capital taxation. Thus there is a case for using a dynamic CGE model for the

analysis of
the double dividend issue.



20

Jorgenson and Wilcoxen (1993), used a dynamic model and they found that a strong double
dividend exists when the revenues from the environmental tax are used to reduce capital
taxes. If the revenues instead were use
d to reduce labor taxes, however, there was no strong
double dividend. However, neither Goulder (1995b) nor Bovenberg and Goulder (1997), who
also used dynamic models of the U.S. economy, found evidence of a strong double dividend.
One reason for this was
that both Goulder and Bovenberg & Goulder assumed that capital
was immobile across sectors, while Jorgenson and Wilcoxen assumed full inter
-
sector capital
mobility.


In contrast to the results by Jorgenson and Wilcoxen (1993) Bye (2000), who used a dynamic

model of the Norwegian economy, found that a revenue
-
neutral swap between an increased
environmental tax and a reduced tax on labor income was welfare increasing. According to
Bye the differences between Jorgenson´s and Wilcoxens´s results and her own res
ults depend
on the fact that the marginal excess burden is higher for capital taxation than for labor
taxation in the U.S., while the opposite holds in Norway. However, in Böhringer and Pahlke
(1997), who also used a dynamic model, no strong double dividen
d could be found. Thus the
conclusion that emerges from the CGE model analyses is that the existence of a strong double
dividend can neither be taken for granted nor entirely ruled out.


8.
CGE models of resource depletion and management

As developing coun
tries tend to be more dependent on natural resources than industrialized
high
-
income countries CGE models focused on natural resource management issues are
typically models of developing countries. However, the “externality” type of environmental
CGE model
s completely dominates the field. In fact very few models are focused on natural
resource management and policy issues.


Devarajan (1988) surveys the issues that have to be dealt with in a CGE model of a
developing country in which the economy to a large
extent depends on a depletable resource.
Three different perspectives are adopted, and the related CGE modeling issues are discussed.
In the first the natural resource is seen as an input to production. In the second, denoted the
“Dutch disease” perspectiv
e, it is seen as a source of revenue for the economy. In the third the
analysis is focused on the exhaustibility of the resource and the inter
-
temporal resource
allocation issues related to that. However, most models of developing countries are focused
eit
her on pollution problems or issues related to excessive exploitation of natural resources. A
few examples are given in the following.


Xie and Saltzman (2000) present both a general framework for CGE models of developing
countries and a specific model of
the Chinese economy. As a basis for the CGE model they
develop an environmentally extended social accounting matrix (ESAM) to serve as a
consistent data set for calibrating the model. The China model is used for an evaluation of
pollution control policies
focused on wastewater, smog dust and solid waste.


A recent model with an elaborated treatment of natural resources is Abler et.al. (1999). It is a
model of Costa Rica, and one of the distinguishing features of the model is that it includes
eight different

environmental indicators, including the degree of deforestation and the degree
of over
-
fishing. The impact of production and consumption activities is essentially modeled
as external effects, and the environmental indicators are in effect treated as publi
c goods.



21

Persson and Munasinghe (1995)
18

also use a model of Costa Rica and focuses on
deforestation. In the model deforestation is an endogenous result of ill
-
defined rights to
forestland. Thus, when property rights to forest land are not well defined an
d protected
loggers and squatters neglect the value of maintaining the land in question for forestry in the
future, and the incentives to deforestation are strong. When, on the other hand, property rights
to forestland are well defined the owners take the
value of maintaining the forest into
consideration, and the incentives for deforestation are much weaker.


Another example of a CGE model focused on natural resource management issues is Unemo
(1995). The purpose of the study is to analyze unintended side
effects of government policies
in Botswana. One key feature of the model is that “capital” in the livestock sector has the
form of cattle. Another key feature is a measure of land pressure, measuring the ratio between
the number of hectares of grazing land

and the number of cattle held on that land. Grazing
land is treated as a common property resource. The idea is that an increase in the number of
cattle per unit of grazing land has a negative impact on productivity in the livestock sector.
The land pressu
re variable can also been seen as an indicator of environmental quality. Using
the model Unemo finds some interesting relations between, on the one hand, economic policy
measures and changes in world market conditions and, on the other hand, environmental
quality.


Thus in one case it was assumed that there was a fall in the world market price of diamonds,
which is a major export product of Botswana. According to the model the lower world market
price of diamonds would lead to a deterioration of environmen
tal quality, in terms of land
pressure, in Botswana. The reason is that lower revenues from diamond export lead to lower
demand for manufactured goods and thus a lower rate of return of capital in the
manufacturing sector. As a result of that capital is re
allocated to the livestock sector, i.e. the
number of cattle increased. As a result the pressure on land increases. This result illustrates
that a CGE model can reveal indirect interdependencies in the economy that are not
immediately obvious to the analys
t.


9.
Concluding remarks

In terms of the number of models, and studies based on these models, CGE modeling has
expanded very significantly, particularly during the 1990´s. Currently CGE modeling is both a
field for specialized research, and an almost stan
dard part of the toolbox of economists
concerned with policy
-
oriented research. A major reason for the widespread use of CGE
modeling probably is that a CGE model is an ideal bridge between economic theory and
applied policy research. The “bridge” perspect
ive, however, suggests that CGE modeling is a
way of using rather than testing economic theory. Yet carefully designed and estimated CGE
models have a lot to say about real world economies.


CGE modeling has made significant progress in terms of the size a
nd complexity of the
models that can be solved. Thus, while the early CGE models were simple static Walrasian
models of a single economy, later CGE models to a large extent are dynamic, multi
-
country or
global. There are also models with imperfect competit
ion in one or several markets. In
environmental CGE modeling both the dynamic and the multi
-
country features have made
CGE models useful for in analyses of important policy issues such as climate change and acid
rain. However, in many cases the damage caus
ed by emission of pollutants is uncertain and
policy measures could in effect be seen as an insurance against future catastrophic damage.



18

Their model is described in detail in Persson’s PhD dissertation, Haksar (1997).


22

Thus a desirable further development of environmental CGE modeling is to incorporate
uncertainty.


Yet complexity shou
ld never be an end in itself in CGE modeling. Much of the usefulness of
a CGE model stems from its solid foundation in basic economic theory. Thus, even if
simulation results from a standard CGE model sometimes may be surprising they can always
be explaine
d in terms of well known income and substitution effects in combination with
interdependencies between markets. The addition of non
-
standard features, such as imperfect
competition on product and factor markets, price rigidities and inter
-
temporal relation
s, may
make the model more “realistic”. But that may also imply that the transparency of the CGE
model is lost.


References


Abler, D., A.G. Rodriguez and J.S. Shortle (1999), “Parameter Uncertainty in CGE Modeling
of the Environmental Impacts of Economic
Policy”,
Environmental and
Resource Economics
, 14, 75
-
94.

Adelman, I. and S. Robinson (1982),
Income Distribution Policy in Developing Countries
.
Oxford: Oxford University Press.

Alfsen, K.H., T.A. Bye and E. Holmøy (1996), “MSG
-
EE: An Applied General Equi
librium
Model for Energy and Environmental Analyses”. Social and Economic Studies.
Oslo: Statistics Norway.

Armington, P. (1969), “A Theory of Demand for Products Distinguished by Place of
Production”,
IMF Staff Papers
, 16, 159
-
78.

Ayres, R.U. and A.V. Kne
ese (1969), “Production, Consumption and Externalities”,
American Economic Review
, vol. 59, no. 3, June, 282
-
297.

Babiker, M.H., K.E. Maskus and T.F. Rutherford (1997), “Carbon Taxes and the Global
Trading System”. Paper presented at the International Ener
gy Workshop and
Energy Modeling Forum Meeting, IIASA, June 23
-
25.

Bergman, L. (1990), “Energy and Environmental Constraints on Growth: A CGE Modeling
Approach”,
Journal of Policy Modeling
, 12(4):671
-
691.

Bergman, L. (1990), “The Development of Computable G
eneral Equilibrium Modeling”, in
Bergman, L., D.W. Jorgenson and E. Zalai (1990),
General Equilibrium
Modeling and Economic Policy Analysis
. Cambridge and Oxford: Basil
Blackwell.

Bernstein, P.M., W.D. Montgomery, T.F. Rutherford and Gui
-
Fang Yang (1999),
“Effects of
Restrictions on International Permit Trading: The MS
-
MRT Model”,
The
Energy Journal
, Special Issue: The Costs of the Kyoto Protocol: A Multi
-
Model
Evaluation.

Böhringer, C. and A. Pahlke (1997), “Environmental Tax Reforms and the Prospects for
a
Double Dividend”

Böhringer, C., M. Ferris and T.F. Rutherford (1998), “Alternative CO
2

Abatement Strategies
for the European Union”, in Braden, J. and S. Proost (eds.)
Climate Change,
Transport and Environmental Policy: Empirical Applications in a Federa
l
System
. London: Edward Elgar.

Bollen, J., A. Gielen and H. Timmer (1999), “Clubs, Ceilings and CDM: Macroeconomics of
Compliance with the Kyoto Protocol”,
The Energy Journal
, Special Issue: The
Costs of the Kyoto Protocol: A Multi
-
Model Evaluation.


23

Boven
berg, A.L. and R.A. de Moij (1994), “Environmental Levies and Distortionary Taxation
and Labor Market Distortions”,
European Journal of Political Economy
, 10(4),
655
-
683.

Bovenberg, A.L. and L.H. Goulder (1997), “Costs of Environmentally Motivated Taxes in

the
Presence of Other Taxes: General Equilibrium Analyses”,
National Tax Journal,

Brooke, A., D. Kendrick and A. Meeraus (1988), GAMS. A User´s Guide. Redwood City,
Ca.: Scientific Press.

Burniaux, J
-
M, J.P. Martin and J.O. Martins (1992), “GREEN: A Globa
l Model for
Quantifying the Cost of Policies to Curb CO
2

Emissions”,
OECD Economic
Studies
, no. 19.

Bye, B. (2000), “Environmental Tax Reform and Producer Foresight: An Intertemporal
Computable General Equilibrium Analysis”,
Journal of Policy Modeling
, Vol
.
22, No. 6, 719
-
752.

Capros, P., T. Georgakopoulos, D. Van Regemorter, S. Proost, K. Conrad, T. Schmidt,Y.
Smeers, N. Ladoux, M. Vielle and P McGregor (1995),
GEM
-
E3
.
Computable
General Equilibrium Model for Studying Economy
-
Energy
-
Environment
Interactions
. European Commission. EUR 16714 EN.

Conrad, K. (1999), “Computable General Equilibrium Models for Environmental Economics
and Policy Analysis”, in J.C.J.M. van den Bergh (ed.)
Handbook of
Environmental and Resource Economics
, Cheltenham: Edward Elgar.

De
Melo, J. and S. Robinson (1989), “Product Differentiation and the Treatment of Foreign
Trade in Computable General Equilibrium Models”,
Journal of International
Economics
, 27, 47
-
67.

Debreu, G. (1959),
Theory of Value
. Cowles Foundation. New York: John Wil
ey & Sons.

Devarajan, S. (1988), “Natural Resources and Taxation in Computable General Equilibrium
Models of Developing Countries”,
Journal of Policy Modeling
, 10(4), 505
-
528.

Devarajan, S. and D.S. Go (1998), “The Simplest Dynamic General
-
Equilibrium Mode
l of an
Open Economy”,
Journal of Policy Modeling
, 20(6):677
-
714.

Farmer, K. and K.W. Steininger (1999), “Reducing CO2 Emissions under Fiscal
Retrenchment”: A Multi
-
cohort CGE model for Austria”,
Environmental and
Resource Economics
, 13(3):309
-
40.

Ferris,
M.C. and T.S. Munson (2000), GAMS/PATH User Guide,
www.gams.com
.

Ford Foundation (1974), A Time to Choose: America´s Energy Future (1974). Final Report
by the Energy Policy Project of the Ford Foundation. Cambridge: Bal
linger.

Førsund, F.R., M. Hoel and S. Longva (1985),
Production, Multi
-
Sectoral Growth and
Planning
. Amsterdam: North
-
Holland.

Goulder, L.H. (1995a), “Environmental Taxation and the ‘Double Dividend’: A Reader´s
Guide”,
International Tax and Public Finance
, 2(2), 157
-
183.

Goulder, L.H. (1995b), “Effects of Carbon Taxes in an Economy with Prior Tax Distortions:
An Intertemporal General Equilibrium Analysis”,
Journal of Environmental
Economics and Management,

October.

Goulder, L.H., I.W.H Perry and D. Burtraw

(1997), “Revenue
-
Raising vs. Other Approaches
to Environmental Protection: The Critical Significance of Pre
-
Existing Tax
Distortions”,
RAND Journal of Economics.

Winter.

Goulder, L.H. and S.H. Schneider (1999), “Induced Tecnical Change and the Attractiven
ess
of CO
2

Abatement Policies”,
Resource and Energy Economics
, 21(2
-
3), 211
-
253.

Haksar, A. (1997),
Environmental Effects of Economywide Policies: Case Studies of Costa
Rica and Sri Lanka
. PhD dissertation, Stockholm School of Economics.


24

Harberger, A.C. (1
962), “The Incidence of the Corporate Income Tax”,
Journal of Political
Economy
, 70 (3), June, 215
-
40.

Harris, R. (1984), “Applied General Equilibrium Analysis of Small Open Economies with
Scale Economies and Imperfect Competition”,
American Economic Revie
w
, 74,
December, 1016
-
31.

Harrison, G.W., T.F. Rutherford and I. Wooton (1989), “The Economic Impact of the
European Community”,
American Economic Review
, Papers and Proceedings 79
(May) 288
-
294.

Harrison, G.W. and T.F. Rutherford (1997), “Burden Sharing,
Joint Implementation, and
Carbon Coalitions”, in C. Carraro (ed.)
International Environmental Agreements
on Climate Change
, Amsterdam: Kluwer Academic Press.

Harrison, G.W., J. Jensen, M.I. Lau and T.F. Rutherford (1997), “”Passing the Laugh Test:
Version
0 of the MobiDK Model”. Available at
www.gams.com/projects/dk/mobiDK.htm
.

Hazilla, M. and R.J. Kopp (1990), “Social Cost of Environmental Quality Regulations: A
General Equilibrium Analysis”,
Jour
nal of Political Economy
, vol. 98, no. 4,
853
-
873.

Hill, M. (2001), “Essays on Environmental Policy Analysis: Computable General Equilibrium
Approaches Applied to Sweden”, PhD dissertation, Stockholm School of
Economics.

Hogan, W.W. and D.J. Jorgenson (199
1), “Productivity Trends and the Cost of Reducing CO
2

Emissions”,
The Energy Journal
, 12(1), 67
-
85.

Hudson, E.A.
and D.W. Jorgenson (1975), “U.S. Energy Policy & Economic Growth 1975
-
2000”,
Bell Journal of Economics and Management Science
, 5(2), Autumn,
4
61
-
514.

Johansen, L. (1960),
A Multi
-
sectoral Study of Economic Growth
. Amsterdam: North
-
Holland.

Jorgenson, D.W. (1982), “Econometric and Process Analysis Models for Energy Policy
Assessments”, in R. Amit and M. Ariel (eds.),
Perspectives on Resource Poli
cy
Modeling: Energy and Minerals
. Cambridge: Ballinger.

Jorgenson, D.W. (1998a), Growth. Volume 1: Econometric General Equilibrium Modeling.
Cambridge: The MIT Press.

Jorgenson, D.W. (1998b), Growth. Volume 2: Energy, the Environment and Economic
Growth. C
ambridge: The MIT Press.

Jorgenson, D.W. and P.J. Wilcoxen (1993), ”Reducing U.S. Carbon Emissions: An
Econometric General Equilibrium Assessment”,
Resource and Energy
Economics
, 14, 243
-
268.

Kainuma, M., Y. Matsuoka and T. Morita (1999), “Analysis of Post
-
Kyoto Scenarios: The
Asian
-
Pacific Integrated Model”,

Ljungqvist, L. and T.J. Sargent (2000),
Recursive Macroeconomic Theory
. Cambridge: MIT
Press.

Lundgren, S. (1985), “Model Integration and the Economics of Nuclear Power”. PhD
dissertation, Stockholm S
chool of Economics.

Mäler, K.G. (1974),
Environmental Economics
. Baltimore: Johns Hopkins Press.

Manne, A. (1977), “ETA
-
MACRO: A Model of Energy
-
Economy Interactions”, in Hitch, C.J.
(ed.)
Modeling energy
-
economy interactions: Five approaches
.Washington DC
:
Resources for the Future.

Manne, A. and R. Richels (1990), “Global 2100: Model Formulation”. Stanford University.

Manne, A. and R. Richels (1992), Buying Greenhouse Insurance: The Economic Cost of CO
2

Emission Limits. Cambridge: MIT Press.


25

Manne, A. and
J.O. Martins (1994), “Comparison of Model Structure and Policy Scenarios:
GREEN and 12RT”,
OECD Economics Department Working Paper
. No. 146.

McKibbin, W.J. and P.J. Wilcoxen (1995), “The Theoretical and Empirical Structure of the
G
-
Cubed Model”,
Brookings
Discussion Papers in International Economics
#
18. The Brookings Institution. Washington D.C.

Nordhaus, W.D. (1974), “The Allocation of Energy Resources”,
Brookings Papers on
Economic Activity
, 4:1974, 529
-
577.

Nordhaus, W.D. (1994),
Managing the Global Co
mmons: The Economics of Climate Change
.
Cambridge: The MIT Press.

Nordhaus, W.D. and Z. Yang (1996), “A Regional Dynamic General
-
Equilibrium Model of
Alternative Climate Change Strategies”,
American Economic Review
, vol. 86,
No. 4, September, 741
-
765.

Nord
haus, W.D. (1997). Modeling Induced Innovation in Climate Change Policy. Paper
presented at the Workshop on Induced Technological Change, June 1997,
IIASA, Austria.

Norman, V.D. and J. Haaland (1987), “VEMOD


a Ricardo
-
Viner
-
Heckscher
-
Ohlin
-
Jones
Model of

Factor
-
Price Determination”,
Scandinavian Journal of Economics
, 89
(3), 251
-
70.

Parry, I.W.H, R. Williams and L.H. Goulder (1998), “When Can Carbon Abatement Policies
Increase Welfare? The Fundamental Role of Distorted Factor Markets”,
Journal
of Environm
ental Economics and Management
,

Persson, A. and M. Munasinghe (1995), “Natural Resource Management and Economic
Policies in Costa Rica: A Computable General Equilibrium (CGE) Approach”,
The World Bank Economic Review
, 9(2), 259
-
85.

Pezzey, J. (1992), “Anal
ysis of Unilateral CO
2

Control in the European Community and
OECD”,
The Energy Journal
, 13(3), 159
-
171.

Pohjola, J. (1999), “Economy
-
wide effects of reducing CO
2

emissions: A comparison
between net and gross emissions”,
Journal of Forest Economy
, 5:1, 139
-
168.

Robinson, S., A. Yúnez
-
Naude, R. Hinojosa
-
Ojeda, J.D. Lewis and S. Devarajan (1999),
“From stylized to applied models: Building multisector CGE models for policy
analysis”,
North American Journal of Economics and Finance
, 10, 5
-
38.

Rutherford, T.F. (1
992), “The Welfare Effects of Fossil Carbon Restrictions: Results from a
Recursively Dynamic Trade Model”,
OECD Economics Department Working
Papers
, No. 112.

Scarf, H.E. (1967), “On the Computation of Equilibrium Prices”, in Feliner, W.J. (ed),
Ten
Economi
c Studies in the Tradition of Irving Fisher
. New York: John Wiley &
Sons.

Shoven, J.B. and J. Whalley (1984), “Applied General Equilibrium Models of Taxation and
Trade: An Introduction and Survey”,
Journal of Economics Literature
, 22,
September, 1007
-
51.

S
hoven, J.B. and J. Whalley (1992),
Applying General Equilibrium
. Cambridge: Cambridge
University Press.

Unemo, L. (1995), “Environmental Impact of Governmental Policies and External Shocks in
Botswana: A Computable General Equilibrium Approach”, in Perring
er, C.A.
et.al. (eds.),
Biodiversity Conservation
. Amsterdam: Kluwer.

Vennemo, H. (1995), “A Dynamic Applied General Equilibrium Model with Environmental
Feedbacks”,
Economic Modeling
, 14, 99
-
114.

Whalley, J. and A. Mansur (1984), “Numerical Specification
of Applied General Equilibrium
Models: Estimation, Calibration and Data”, in H.E. Scarf and J.B. Shoven (eds),

26

Applied General Equilibrium Analysis
. Cambridge: Cambridge University Press,
69
-
127.

Whalley, J. and R. Wigle (1991), “The International Incidenc
e of Carbon Taxes”, in R.
Dornbush and J.M. Poterba,
Economic Policy Responses to Global Warming
.
Cambridge: MIT Press.

Whalley, J. and R. Wigle (1992), “International Linkages and Carbon Reduction Initiatives”,
in K. Anderson and R. Blackhurst (eds.),
The

Greening of World Trade Issues
.
Harvester, Wheatsheaf.

Whalley, J. and B. Yeung (1982), “External Sector “Closing” Rules in Applied General
Equilibrium Models”,
Journal of International Economics
, 16, 123
-
138.

Xie, J. and S. Saltzman (2000), “Environmenta
l Policy Analysis: An Environmental
Computable General Equilibrium Model for Developing Countries”,
Journal of
Policy Modeling
, 22(4):453
-
489.

Yang, Z., R.S. Eckaus, A.D. Ellerman and H.D. Jacoby (1996), “The MIT Emissions
Predictions and Policy Analysis (
EPPA) Model”. Report 6. MIT Joint Program
on the Science of Global Change.



27


Table 1. Key characteristics of selected global CGE models


Model

Reference

Regions

Sectors per
region

Dynamics

Energy sector

Backstop
technology

Technologi
-
cal change

Environmen
-
tal benefits

WW

Whalley and Wigle
(1991)

6

3

Static

Top
-
down

No

None

Yes

GREEN

Burniaux et.al. (1992)


12

11

Quasi
-
dynamic

Top
-
down

Yes

AEEI

No

Global 2100

Manne and Richels
(1992)

5

2

Fully
dynamic

Bottom
-
up

Yes

AEEI

No

12RT

Manne (1993)


12

2

Fully
dynamic

Bottom
-
up

Yes

AEEI

No

CRTM

Rutherford (1992)


5

3

Quasi
-
dynamic

Bottom
-
up

Yes

AEEI

No

G
-
Cubed

McKibbin et.al. (1995)


8

12

Fully
dynamic

Top
-
down

No

None

No

MIT
-
EPPA

Yang et.al. (1996)

12

8

Quasi
-
dynamic

Top
-
down

Yes

AEEI

No

RICE

Nordhaus and Y
ang
(1996)

13

1

Fully
dynamic

**

Yes

AEEI

Yes

IIAM

Harrison and Ruther
-
ford (1997)

5

2

Fully
dynamic

Top down

No

None

No

UR

Babiker et.al. (1997)


26

13

Static

Top down

No

None

No

MS
-
MRT

Bernstein et.al. (1999)

10

6

Fully
dynamic

Top down

Yes

AEEI

No

A
IM

Kainuma et.al. (1999)


21

11

Quasi
-
dynamic

Top
-
down

No

AEEI

No

WorldScan

Bollen et.al. (1999)

13

11

Quasi
-
dynamic

Top
-
down

No

None

No

* No trade between regions. In some cases carbon permits can be traded against the aggregate good.

** The energy se
ctor is a part of the single aggregated production sector.



28

Table 2. Key characteristics of selected regional multi
-
country CGE models


Model

Reference

Regions

Sectors per
region

Dynamics

Energy sector

Backstop
technology

Technological
change

Environmenta
l
benefits

GEM
-
E3

Capros et.al
(1995)

All EU Member States
and ROW

18

Quasi
-
dynamic

Top
-
down

No

AEEI

Yes

BFR

Böhringer et.al.
(1998).

Germany, France, UK,
Italy Spain Denmark
and ROW

23

Static

Top
-
down

No

-

No

HRW

Harrison et.al.
(1989)

US, Japan, Franc
e,
Italy, UK, Ireland
Germany, Netherlands,
Belgium, Denmark,
and ROW

6

Static

Top
-
down

No

-

Yes



29

Table 3. Key characteristics of selected single
-
country CGE models


Reference

Country

Number of
sectors

Dynamics

Energy goods

Emissions

Special features

Ha
zilla and Kopp
(1990)

USA

36

Quasi
-
dynamic

Electricity, coal, natural
gas, oil

-

Econometrically estimated parameters;
technology
-
based environmental
regulations

Bergman (1990)

Sweden

7

Static

Electricity, fuels

SO
2
, NO
x
, CO
2

Tradable emission quotas

Bov
enberg and
Goulder (1997)


13

Dynamic

Electricity, coal, natural
gas, oil

-

Synthetic fuel as a backstop resource

Parry et.al. (1998)



6

Static

Natural gas, coal, oil

CO
2

Tradable CO
2
quotas and taxes

Jorgenson and
Wilcoxen (1993)

USA

35

Dynamic

Electri
city, coal, natural
gas, oil

CO
2

Econometrically estimated parameters;
endogenous productivity growth

Alfsen et.al.
(1996)

Norway

33

Dynamic

Electricity, natural gas, oil

SO
2
, NO
x
, VOC, O
3
,
CO, CO
2
, CH
4
, N
2
O,

Damage functions defining damage to
public he
alth, forests, lakes and building
materials

Vennemo (1995)

Norway

6

Dynamic

Electricity, oil

SO
2
, NO
x
, CO, PM

Feedback effects.

Harrison et.al.
(1997)

Denmark

117

Static and
dynamic versions

Electricity, natural gas,
coal, oil

CO
2
,


Pohjola (1999)

Finla
nd

18

Quasi
-
dynamic

Coal, natural gas, peat,
wood, heating fuels,
gasoline

CO
2
,

Carbon sinks

Abler et.al. (1999)

Costa
Rica

15

Static

Electricity, oil


Eight environmental quality indicators

Farmer and
Steininger (1999)

Austria

8

Overlapping
generations

Electricity, fossil fuels

CO
2
,

Different cohorts

Hill (2001)

Sweden

17

Dynamic

Electricity, gas, coal, oil

SO
2
, NO
x
, CO
2

Inter
-
temporal emissions trading; feed
-

back effects

Xie and Saltzman
(2000)

China

7

Static

Aggregated energy

Waste water, smog
dust,

solid waste

Pollutant
-
specific abatement sectors