Endogenous Technological Change in Strategies
for Mitigating Climate Change
vorgelegt von
Diplom-Systemwissenschaftler
Kai Lessmann
aus Detmold
von der Fakultät VI – Planen Bauen Umwelt
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Dokor der Naturwissenschaften
Dr. rer. nat.
genehmigte Dissertation
Gutachter:
Prof. Dr. Ottmar Edenhofer
Prof. Dr. Carlo Carraro
Promotionsausschuss:
Prof. Dr. Dieter Scherer (Vorsitz)
Prof. Dr. Ottmar Edenhofer
Prof. Dr. Volkmar Hartje (i.V.)
Tag der wissenschaftlichen Aussprache: 9.9.2009
Berlin 2009
D83
Contents
Summary 7
1 Introduction 11
1.1 Motivation 11
1.1.1 A Sense of Urgency 11
1.1.2 Economics of Climate Change 14
1.2 Thesis Objective 20
1.3 Thesis Outline 21
2 Mitigation Strategies and Costs of Climate Protection 23
2.1 Setting the Scene 25
2.2 The Model Structure of MIND 1.1 26
2.3 The Role of ETC in MIND 31
2.3.1 Local Sensitivity Analysis 31
2.3.2 Determinants of Opportunity Costs 33
2.3.3 Mitigation Strategies 36
2.4 Concluding Remarks 38
2.5 References 40
3 Implication of ITC for Atmospheric Stabilization 41
3.1 Introduction 44
3.2 Model Comparison in the Literature 45
3.3 Model Classification 47
3.3.1 Model Types in the IMCP 47
3.4 Method of Model Comparison 54
3.4.1 Scenario Definitions With and Without ITC 55
3.4.2 Model Output and Indicators 56
3
4 Contents
3.4.3 Concepts of Mitigation Costs 56
3.5 Results and Discussion 58
3.5.1 Mitigation Costs with Different Model Types 58
3.5.2 Mitigation Strategies for Different Stabilization Scenarios 68
3.5.3 Decomposition Analysis 69
3.5.4 Timing of Mitigation Options 76
3.5.5 Energy Mix 76
3.6 Conclusion 87
3.6.1 Baseline Effects 87
3.6.2 First Best or Second Best Assumptions 88
3.6.3 Model Structure 88
3.6.4 Long-term Decision Making: Foresight and Flexibilities 89
3.6.5 Backstop and End-of-pipe Technologies 90
3.6.6 Hints for a Future Research Agenda 90
3.7 References 92
4 Effects of Tariffs on Coalition Formation 95
4.1 Introduction and Motivation 97
4.2 Model Structure 100
4.2.1 Preferences 100
4.2.2 Technology 101
4.2.3 Climate Dynamics 101
4.2.4 Trade and Tariffs 102
4.3 Solving for a Nash Equilibrium 103
4.3.1 Finding a Nash Equilibrium 103
4.3.2 Numerical Verification of the Nash Equilibrium 104
4.3.3 Partial Agreement Nash Equilibria 104
4.4 Application to International Cooperation 105
4.4.1 Results 105
4.4.2 Sensitivity Analysis 109
4.5 Conclusion 111
4.6 Appendix: Parameter Choices 112
4.7 Bibliography 114
5 Research cooperation and international standards 117
5.1 Introduction 119
Contents 5
5.1.1 Issue Linking 120
5.1.2 Potential of Spillovers 120
5.1.3 Potential of International Standards 121
5.1.4 Coalition Formation 122
5.1.5 Novelty 124
5.2 The Model 124
5.2.1 Model Equations 125
5.2.2 Coalition Formation 127
5.3 Results 129
5.3.1 Participation in Environmental Cooperation 129
5.3.2 Environmental Effectiveness and Welfare Effects 134
5.3.3 Credibility of Exclusive R & D Cooperation 137
5.4 Sensitivity of Key Results 138
5.5 Summary and Conclusions 139
5.6 References 142
5.7 Appendix: Parameter Choices 145
6 Synthesis and Outlook 147
6.1 Induced technological change in integrated assessment modeling 148
6.1.1 Implications of ITC in the MIND model 148
6.1.2 Implications across models 149
6.1.3 Discussion 152
6.2 The Prospect of Issue Linking for Global Climate Policy 153
6.2.1 Trade Sanctions 153
6.2.2 Technology-oriented Agreements 154
6.2.3 Discussion 155
6.3 Outlook and Further Research 156
Bibliography 159
Statement of Contribution 169
Acknowledgements 171
Tools and Resources 173
6 Contents
Summary
This thesis suggests that induced technological change has potential to reduce the burden
that climate change mitigation puts on the economy. Furthermore, international coop-
eration on climate policy, which may trigger this induced technological change, may be
achieved by linking climate negotiations to other issues. The starting point of the research
presented here are the following two assumptions: first, action to mitigate climate change
is necessary, and second, technologies will play a key role in this effort because technol-
ogy and technological change facilitate the reduction of anthropogenic greenhouse gas
emissions. As a consequence, the way technological change is described in integrated as-
sessment models of climate change is of great importance, and a sound understanding of
such endogenous technological change and its interaction with climate policies is needed.
There is empirical evidence that technological change is induced by policies. How-
ever, previous assessments of such induced technological change (ITC), i.e. technological
progress triggered by policy, have been ambiguous about its responsiveness to climate
policies and its potential to reduce the costs of mitigating climate change. On the other
hand, a clear climate policy is required in order to induce the technological progress that
might facilitate emission abatement at low costs. Ideally, climate policy ought to be global
in order to prevent carbon leakage and to achieve efficiency. However, the literature on
international environmental agreements suggest that the prospect for global climate pol-
icy is not bright. This raises two broad research questions: First, what is the role of ITC
for climate change mitigation? And second, if there is a desirable contribution of ITC to
mitigation, how can we achieve a global policy that triggers this technological change?
The four papers presented in this thesis make contributions to these two questions.
The first paper focuses on the impact of ITC on the costs and strategies of mitigating cli-
mate change within a single integrated assessment model. I find that the impact of ITC
is significant. The analysis reveals two “directions” of technological change. First, there
is technological change that permeates the entire economy—this is reflected in a strong
impact on the overall macro-economic costs of mitigation. Second, there is technological
change whose impact is specific to the energy sector, as evident from strong changes in the
composition of mitigation options. ITC therefore proves to be an influential determinant
of mitigation costs and strategies. Costs may rise or fall due to ITC depending on whether
progress in low carbon energy or progress in the resource sectors prevails. The effect of
ITC on the competitiveness of mitigation options influences their contributions to overall
mitigation. Moreover, this stresses the importance of models that resolve important tech-
nological options, including their potential of ITC, and account for the economy-wide
impact of ITC.
7
8 Summary
The first paper used a single model with one specific formulation of ITC—but the question
how to incorporate ITC in models is far from trivial. On the contrary, among models that
include ITC there is a wide variety of approaches taken to describe ITC.
The second paper of this thesis compares ten state-of-the-art models that implement ITC.
It explores the resulting differences in their assessment of ITC, identifies the underlying
reasons for the differences, and draws conclusions that are robust across models. At
the heart of this comparison are ceteris paribus scenarios that aim to isolate and expose
the impact of ITC in the various models. The analysis reveals that ITC has potential to
reduce costs, in many models substantially. However, the magnitude of the impact of ITC
differs greatly among the models, ranging from 90 percent reduction of mitigation costs to
almost no effect. Numerous reasons for this were identified, including business-as-usual
emissions, differences in mitigation strategies, and modeling assumptions.
Business-as-usual emissions have a strong impact on mitigation costs because they deter-
mine the necessary emission reductions. Although an effort was made to harmonize the
business-as-usual scenarios of the different models, considerable differences remained
and need to be taken into account.
Mitigation strategies are explored on two levels of aggregation. First, abatement is de-
composed into the contributions of reductions of economic output, energy intensity of
output, and carbon intensity of energy. The analysis reveals that macro-economic models
without explicit representation of the energy sector tend to focus their abatement strategy
on reductions of energy intensity, whereas energy system models and models that fea-
ture an energy sector achieve the majority of their abatement through decarbonization.
Decarbonization becomes particularly important for large reductions of emissions. Sec-
ond, abating emissions through change in the composition of energy supply is considered.
The composition of the energy supply mirrors the trends of the decomposition analysis.
Models that focus their abatement strategy on reducing energy intensity and economic
output are those that lack options to decarbonize the energy system, or that simply did not
resolve the energy sector explicitly. Conversely, large reductions of carbon intensity are
implemented through large shares of carbon free energy.
Three key modeling assumptions were identified that explain some of the major differ-
ences in model results: first, when models include additional market distortions, i.e. they
describe a second-best world, climate policy may remove these distortions causing not
costs but a benefit of climate policy. Second, the choice of the model type is influential
because it often implies an equilibrium concept, which in turn implies different degrees of
flexibility to react to climate policy. Third, different assumptions about foresight of eco-
nomic agents determines their long-term investment behavior, which strongly influences
mitigation strategies and costs.
The first two papers looked at climate policies implemented as global policy targets taking
for granted that policies are agreed upon and implemented to achieve the targets, although
this is known to be difficult. The remaining papers look at the potential of issue linking
to help to build such agreements. To address issue linking, I develop a model of coalition
formation, which incorporates international trade and sanctions as well as knowledge
spillovers from research cooperation and international standards.
In the third paper, I show in numerical experiments that introducing trade sanctions pos-
itively affects international cooperation. Participation rises with the tariff rate, up to full
Summary 9
cooperation. How quickly participation rises depends on the ease with which taxed goods
are substituted with alternatives. Global welfare rises with participation despite the dis-
tortions caused by trade restriction. Tariffs therefore seem to be a feasible means of
increasing participation.
In the forth paper, I apply an extended version of the coalition model to issue linking of
environmental agreements and technology oriented agreements. It turns out that linking
the environmental agreement to cooperative research changes the incentive structure such
that more actors sign the agreement. The type of technological knowledge that spills over
makes a difference for the effectiveness of this type of issue linking: research cooperation
focusing on productivity is unambiguously more effective than cooperation on mitigation
technology in raising participation in the agreement, global welfare, and environmental
quality. International technology standards are also shown to have a positive effect on
coalition formation. While the existence of a separate standards agreement alone has
very little impact on environmental cooperation, it significantly increases participation in
a linked agreement on environmental and technological cooperation.
Overall, the studies reported in this thesis suggest that there is indeed potential that ITC
may reduce the burden that mitigation requirements will put on the economy. And while
there is no final conclusion on the magnitude of the impact of ITC due to the remaining
model uncertainty, this thesis advances the understanding of these uncertainties and the
underlying reasons for the variability in the results. To exploit a large potential of ITC,
a clear carbon price signal is required. This thesis suggests that linking the negotiations
on climate policy to trade sanctions or to research cooperations is a feasible way to create
incentives that make a cooperative global climate policy more likely. More research is
needed to determine the magnitude of the potential of issue linking, but its potential in
general has been shown and different issue linking proposals have been characterized with
respect to their advantages and disadvantages.
10 Summary
Chapter 1
Introduction
1.1 Motivation
This chapter sets out to motivate the main research questions addressed in this thesis on
the role of technological progress in integrated assessment modeling of greenhouse gas
mitigation.1The necessity to reduce emissions arises from the science of climate change
and its impacts. Therefore, the current state of knowledge on climate change and its
impacts is briefly summarized in this section before turning to an introduction of the eco-
nomic themes: first, the consequences of including the economic processes that cause
technological progress in models, thus making technological progress susceptible to (cli-
mate) policies. Second, building the international coalitions that are willing to implement
these policies, therefore setting the incentive for the technological change necessary for
mitigating climate change. This chapter closes with the statement of the main research
questions and an outline of the remaining chapters of this thesis.
1.1.1 A Sense of Urgency
The Scientific Basis
When describing the climate of the earth, climatologists distinguish the climate system
consisting of components such as atmosphere, oceans and land surface, and external fac-
tors that drive the dynamics of this system, so-called forcings. The earth’s climate will
change in response to variation in these forcings, which include solar radiation, the earth’s
albedo, and the greenhouse effect. The latter describes how a set of chemicals in the atmo-
sphere (the greenhouse gases, GHG) capture radiation from the earth that would otherwise
diffuse to outer space. The magnitude of this effect depends on the concentrations of the
GHG.
The earth’s climate has always been subject to changes due to variations in the natural
forcings, for example in solar radiation or volcanic eruptions. In recent earth history,
anthropogenic emission of GHG, mainly linked to fossil fuel combustion, have added to
the concentration of GHG in the atmosphere causing a trend of global warming.
1Integrated assessment models address a problem by combining knowledge across more than one disci-
pline to evaluates its whole cause-effect chain (see, for example, van der Sluijs, 2002).
11
12 Chapter 1 Introduction
Figure 1.1: Multi-model global averages of surface warming (relative to 1980-1999) for the SRES
scenarios A2, A1B and B1, shown as continuations of the 20th century simulations. Adapted from
IPCC (2007).
Working Group I of the Intergovernmental Panel on Climate Change (IPCC) has collected
the scientific evidence supporting this theory in their contribution to the IPCC assessment
reports (IPCC WG1, 2001, 2007). The reports show beyond reasonable doubt that we are
witnessing global warming, and that anthropogenic emissions are a major contribution
to it (Rahmstorf, 2008). The Fourth Assessment Report (IPCC WG1, 2007) attests that
0.7 ◦C warming relative to preindustrial levels have already occurred, and that even if
GHG concentration were not to increase any further, we are already committed to an
additional 0.5 ◦C of warming due to the inertia in the climate system (Figure 1.1.1). But
currently projected unmitigated GHG emissions will cause a much steeper increase in
GHG concentrations. If unabated, the projected increase in global mean temperature in
2100 is projected to be in the range of 1.7 ◦C and 7.0 ◦C.
Global warming in this order of magnitude is sufficient to disturb the dynamics of the earth
system. Due to its complexity and non-linearity, the earth system contains elements that
may switch to a qualitatively different behavior when climate change surpasses certain
thresholds, so-called tipping elements. Lenton et al. (2008) list 15 such tipping elements,
a prime example being Arctic sea ice. Sea ice cover reflects more solar radiation compared
to the darker ocean surface. Therefore melting sea ice has a positive feedback on warming,
and may be destabilized at low levels of global warming. In fact, the loss of Arctic summer
sea ice may already have been triggered. Other tipping elements are the ice sheets of
Greenland and West Antarctica, the Indian summer monsoon, the Amazon rainforest, and
the Boreal forests.
1.1 Motivation 13
Figure 1.2: Examples of impacts associated with global average temperature change. Adapted from
IPCC WG2 (2007).
Impacts
Besides these abrupt changes in the earth system, global warming affects ecosystems and
human societies in a variety of ways. Working Group II of the IPCC studies the impacts
of climate change. In IPCC WG2 (2007), the expected impacts are for the first time
presented scaled against climate change (Figure 1.2). Even at low levels of warming of
up to 2 ◦C relative to 1980-1999, adverse impacts on water availability, ecosystems, food
supply, coastal safety, and human health are expected for 2050, and increasingly so for
2100. When climate change proceeds unmitigated, impacts in 2100 at 4 to 5 ◦C warming
include billions of additional people subjected to increased water stress, extinction of
species around the world, millions additional people at risk of coastal flooding each year,
and increased mortality from heatwaves, floods, and droughts (Parry et al., 2008).
This increased pressure on natural systems and societies may give rise to a list of security
risks. The German Advisory Council on Global Change (WBGU) looked at connections
between climate change and international conflicts (Schubert et al., 2007). They conclude
that the consequences of unmitigated climate change for international conflicts are severe:
[...] climate change will draw ever-deeper lines of division and conflict in
international relations, triggering numerous conflicts between and within coun-
tries over the distribution of resources, especially water and land, over the man-
14 Chapter 1 Introduction
agement of migration, or over compensation payments between the countries
mainly responsible for climate change and those countries most affected by its
destructive effects.
1.1.2 Economics of Climate Change
Economists have attempted to monetize the impacts of climate change in so-called dam-
age functions (see, for example, Nordhaus and Boyer, 2000). This is a notoriously difficult
undertaking as it includes estimating the monetary value of ecosystem services, health,
and human life. Stern (2007) estimates the costs of business-as-usual climate change
to equate at least an average reduction of 5 percent global per capita consumption, now
and forever. When non-market impacts, high climate sensitivity, and the disproportionate
burden for poor regions are taken into account, his estimate rises to 20 percent.
A strong case for action against climate change would emerge if the costs of mitigat-
ing climate change are comparatively low—low compared to the impacts of unmitigated
climate change, and also low compared to adapting to changed climate. Therefore, the
economics of climate change need to address mitigation and adaptation.
Mitigation and Adaptation
It is now certain that mitigation and adaptation will have to complement each other. There
will be climate change even under the most stringent mitigation policy, and therefore there
will be need for at least some adaptation (Figure 1.2). On the other hand, the IPCC deems
it very likely that unmitigated climate change would exceed the world’s capacity to adapt
(IPCC, 2007, Topic 6.2). Hence there is need for at least some mitigation. Exactly where
to draw the line between between “avoiding the unmanageable” and “managing the un-
avoidable” is hard to tell. The aforementioned tipping points offer some guidance: the
“short list” of policy-relevant tipping elements in Lenton et al. (2008, Table 1) comprises
eight tipping elements for which a critical temperature range is given. The critical value
for six of them may be avoided by restraining global warming to 2 ◦C. Therefore, a pol-
icy goal like the European Union’s 2 ◦C target (EU Council, 2007) may serve as an ap-
proximation for the division of labor between mitigation and adaptation. The 2 ◦C target
requires an ambitious mitigation effort.
Technology is both part of the problem and part of the solution for the issue of climate
change mitigation. A majority of GHG emissions are of technological origins: 56.6 per-
cent of all GHG emissions are CO2emissions from fossil fuel combustion. In terms of the
corresponding activities, emissions from energy supply, industry, and transport amount to
58.4 percent of the global total. At the same time technology and technological change in
particular offer the main possibilities for reducing emissions (IPCC WG3, 2007, Ch. 3.4).
According to IPCC WG3 (2007), some of the main technological mitigation options are:
•Improving energy efficiency and energy conservation
•Reducing the carbon intensity of energy, e.g. by switching fuels like substituting gas
for coal
1.1 Motivation 15
N2O
7.9%
CH4
14.3%
CO2
(deforestation,
decay of
biomass, etc)
17.3% CO2 (other)
2.8%
CO2 fossil
fuel use
56.6%
F-gases
1.1%
Waste and wastewater
2.8%
Forestry
17.4%
Agriculture
13.5%
Industry
19.4%
Residential and
comercial buildungs
7.9%
Transport
13.1%
Energy supply
25.9%
Figure 1.3: Greenhouse gas (GHG) emissions. Share of different anthropogenic GHGs in total emis-
sions in 2004 in terms of CO2-equivalent (left), and share of different sectors in total anthropogenic
GHG emissions in 2004 in terms of CO2-equivalent (right). Adapted from IPCC (2007).
•Introducing carbon capture and storage technologies
•Energy from renewable energy sources
•Nuclear power
•Develop and diffuse new technologies and practices to reduce GHG from agriculture
and land use
Therefore, there is a strong link between mitigation and technological progress, and any
policy that aims at mitigating emissions will have to induce technological change, most
importantly the decarbonization of the energy sector. Hence, mitigation and technological
change are interwoven in at least two ways: first, technological progress is essential for
mitigation options. In particular, this refers to low carbon energy technology options,
and energy efficiency improvements. Second, mitigation policies need to set incentives
for technological progress, for example by establishing a price on GHG emissions. The
following sections explore these two aspects.
Mitigation Options: Technological Change
Technological progress does not happen automatically although early economic models
resorted to this assumption of so-called exogenous technological change (for example
Nordhaus, 1994; Nordhaus and Boyer, 2000), i.e. technological change is assumed to
happen independently of policy or other economic activities. On the contrary, it is the
result of actions of economic agents. The literature distinguishes three channels through
which endogenous technological change (ETC) occurs (IPCC WG3, 2007, Ch. 2.7):
•Research and Development (R&D), which refers to some entity (for example firms or
the government) spending resources on developing new technologies or improving
existing technology, for example research spent on improving fuel cell technology.
16 Chapter 1 Introduction
•Learning by Doing, which refers to advances made through production and use of
technologies. Examples include improving labor productivity in production of tech-
nologies which ultimately brings down production costs. As a result, unit costs of
the technology fall as a function of cumulated capacity.
•Spillovers, referring to the transfer of ideas and knowledge among firms, industries,
or other entities. The gas turbine technology transferred to electricity production
is one example, spillover of knowledge in-between countries due to foreign direct
investment is another.
There is empirical evidence for all three channels of technological change. For exam-
ple, there are econometric studies linking R&D expenditure to productivity increases (for
example Griliches, 1992), as well as statistical analyses on “learning curves” correlat-
ing increasing cumulative production volumes and technological advances (IPCC WG3,
2007, Ch. 3.4). These insights were originally taken up by two separate branches in the
modeling literature, macro-economic endogenous growth theory and the learning (or ex-
perience) curve literature (Köhler et al., 2006).
The endogenous (or “new”) growth theory focuses on R&D and spillover effects. In
these models, knowledge capital is accumulated through R&D investments, externalities
to physical capital accumulation, or other spillovers, leading to productivity improve-
ments (see, for example, Aghion and Howitt, 1998). The empirical evidence of learning
curves of individual technologies has been used by bottom-up energy system models.
As these models resolve technological detail, they can implement “learning curves” for
various energy technologies.
Although present in the literature, ETC was neglected in early policy models of climate
change. Even in 2002, Grubb and colleagues find that “most models of energy, economy,
and environment” use exogenous assumptions to describe technological change (Grubb
et al., 2002). The Third Assessment Report (IPCC WG3, 2001) included some new mod-
els that incorporated ETC, but still ETC was not prevalent. Surveying these existing ETC
models, Grubb and colleagues find “striking discrepancies in their basic conclusions.”
While they can cite several models where induced technological change is very respon-
sive to climate policy and hence has large effects, their survey includes models that show
only a modest response. Their conclusion is that there is neither agreement on how to
model ETC, nor are the results from ETC modeling consistent. Clearly, further research
on the impact of ETC is merited. This view is enforced by a subsequent survey (Sijm,
2004) on ETC in climate policy modeling. By this time, the list of models implementing
ETC had grown, but discrepancies among macro-economic models as well as between
top-down and bottom-up models were still large.
Given the importance of ETC for mitigation scenarios, more research is needed to, first,
identify robust conclusions about the likely effects of induced technological change in
climate policy models, and second, to understand and learn from the differences in model
predictions so far in order to improve this important feature.
1.1 Motivation 17
Mitigation Incentives: The Carbon Price
There are two sides to modeling induced technological change. On the one hand, as
discussed in the previous section, an endogenous formulation of technological progress
has to be part of the economic model. The model has to allow for technological change to
be induced. On the other hand, there has to be a policy (here: a climate policy) to induce
this change. In a world without a central authority that can imposes such a policy onto
all nations, achieving an efficient global climate policy requires voluntary cooperation of
sovereign states. In the climate policy models referred to in the previous section this issue
is simply assumed away: most of these models do not specify policies but global policy
goals, assuming their efficient implementations by nations. Furthermore, it is assumed
that all nations agree on the need to take action and on the extent of climate policy. In a
word, there is full cooperation concerning climate protection.
The following section sets the stage for investigating these assumptions. First, the theory
of externalities is introduced. In light of this theory, the fact that GHG emissions cause
climate change as an externality justifies policy intervention on the global level. More
specifically the realization of a global price on GHG emissions is justified—either by
means of a price policy such as a tax, or a quantity policy such as emission caps. Second,
the theory of international environmental agreements explores which incentive structures
qualify to foster international cooperation on such environmental policies.
Theory of externalities The emission of greenhouse gases poses an externality prob-
lem. Intuitively, these are situations where the economic decision of one actor directly or
indirectly affects a second actor who had no part in this decision. In the case of climate
change, GHG emissions are linked to economic decisions of the emitter, for example the
decision to burn fossil fuel to generate electricity. Other actors are then affected by cli-
mate change damages. Mathematically this means that a variable describing the economic
decision (emissions) is part of the utility functions of both players.
The theory of externalities investigates whether the existence of externalities has an ad-
verse effect on economic efficiency, i.e. whether the economy allocates goods and services
in a (Pareto-) efficient manner. In the institutional set-up of a competitive equilibrium,
achieving efficiency boils down to the existence markets. For example, if all externalities
are treated just like other commodities, i.e. there are markets for them, then according
to the first fundamental theorem of welfare, the resulting competitive equilibrium will be
efficient (see Cornes and Sandler, 1996, Chapter 3).
On the other hand, a rationale for policy intervention arises in the absence of such markets.
Then, the emerging allocation can be shown to be inefficient because social and private
costs of the externality diverge. In the case of climate change, the private costs of the
emitter are only the climate change impacts affecting the emitter herself, while the social
cost is the total of all climate change impacts.
An efficient equilibrium may be restored by adjusting the private costs to match the social
costs. One way of doing this is to impose a price on the externality, thus internalizing the
external costs—a tax in case the externality has a negative effect, or a subsidy in case of
a positive externality (Pigou, 1946, as cited in Cornes and Sandler, 1996, Chapter 4).
Alternatively, the conflicting interest of the emitting party and the damaged party could
18 Chapter 1 Introduction
be resolved by bargaining among them. The outcome of such a bargaining process would
depend on the initial property rights, but can be shown to be (Pareto-) efficient regardless
of the latter (Coase, 1960, as cited in Cornes and Sandler, 1996, Chapter 4). Suppose, for
example, that emission of GHG was completely unregulated. Implicitly, this amounts for
potential emitters to have the right for unlimited emissions granted to them. Any actors
who preferred lower GHG concentration level has an incentive to offer payment to the
emitters such that these reduce their emissions, and emitters would have an incentive to
accept payment. If, on the contrary, all parties had the right to a clean atmosphere, it
would be up to the emitters to offer payment for permission to emit. Coase argued that, if
transaction costs were low, such bargaining would take place due to the best interest of all
parties, and that this makes policy intervention such as Pigouvian taxes unnecessary. In
case of climate change, the considerable transaction costs for bilateral bargaining between
all affected parties may be reduced by establishing markets for emission permits.
In an undistorted competitive equilibrium, the price signals from either the emission tax
or the permit price will suffice to attain an allocation that is Pareto-efficient. And while
introducing additional features (for example uncertainty or an oligopolistic market struc-
ture) poses the question of “prices versus quantities” anew (see Hepburn, 2006 for an
overview of the extensive literature), it is undisputed that global cooperation to put a price
signal on GHG emissions is an approach to mitigating climate change that is, at least
potentially, Pareto-efficient.
International Environmental Agreements This is where the literature on international
environmental agreements picks up. This branch of the literature shifts the focus from the
question which policy instrument is preferable to the issue of how to build self-enforcing
coalitions of players that jointly implement a single environmental policy. Often, this
includes the application of game theoretic concepts to the question. Applied to climate
change, cooperation or non-cooperative behavior translates to abating GHG emissions or
not. In a world without central authority, such cooperation can only be voluntary, i.e. by
agreement.
A stable climate or a clean atmosphere has the properties of a public good: it is non-rival
and non-excludable in its use. When a good is non-rival, its provision has an externality:
once provided, it is available to others who were not part of the decision to create this
good. Therefore, as discussed previously in the section on externalities, provision of a
non-rival good constitutes a positive externality and is prone to undersupply and merits
policy intervention.
Non-excludability gives rise to a free-riding incentive. Since nobody can be excluded
from consumption of the good regardless of whether one contributed to its provision,
there is an incentive to let others provide the good and to enjoy its benefits for free.
This gives rise to a situation similar to the well known prisoners’ dilemma where two
prisoners are charged with a common crime (Table 1.1). Ideally, they would both deny
these charges and, in the absence of better evidence, be convicted for lesser crimes. For
if both confessed, they would face a more severe punishment. However, if only one
remained silent while the other confessed, the former will be incriminated while the latter
escapes punishment as a principal witness. The game theoretic analysis reveals that when
rational actors face this situation, both will try to incriminate the other. Therefore, in the
1.1 Motivation 19
Table 1.1: Prisoners’ Dilemma. The payoff structure of the Prisoners’ Dilemma.
Deny Confess
Deny probation, probation acquittal, jail
Confess jail, acquittal jail, jail
Deny Confess
Deny -2, -2 0, -6
Confess -6, 0 -4, -4
Table 1.2: Chicken Game. The payoff structure of the Chicken Game.
Swerve Straight
Swerve tie, tie coward, brave
Straight brave, coward crash, crash
Swerve Straight
Swerve 0, 0 -2, 1
Straight 1, -2 -10, -10
end, both will lose compared to the socially optimal outcome they could attain if they
cooperated.
The good news from early research in international environmental agreements is that
transboundary pollution problems do not fall into the category of prisoners’ dilemma
games (Carraro and Siniscalco, 1993). While players in a prisoners’ dilemma will al-
ways benefit from non-cooperative behavior, in transboundary pollution players may be
better off to abate their emissions even though other players do not. This game struc-
ture is known as a chicken game (Table 1.2). It refers to the situation of two cars racing
towards each other on a narrow lane. The drivers have a choice of avoiding a crash by
swerving but at the price of being a “chicken”, or coward. By not swerving, players show
bravery and win. But if neither driver swerves and the cars crash, the loss is far greater
than being ridiculed as a coward. Unlike in the prisoners’ dilemma, partial cooperation
is therefore preferable to no cooperation—however, both players prefer their opponent
plays “cooperatively.” Similarly in climate change, it may be rational to abate emissions
and thus prevent the worst from happening even though some nations do not cooperate,
i.e. participate in the abatement effort. Nevertheless, the situation where the others coop-
erate on abatement and oneself belongs to those enjoying the stable climate for free is still
preferable. Therefore, a strong incentive to free-ride remains.
Consequently, the bad news from the literature on international environmental agreements
is that stable coalitions tend to be small, in particular in cases where cooperation is needed
the most. That is, cooperation fails when the difference between cooperative and non-
cooperative behavior is large, and therefore much is to be gained by cooperating (for
example Barrett, 1994).
The above situation describes the incentives to sign an international environmental agree-
ment that restricts action to abatement or no abatement. But the “rules of the game” (or
the incentive structure of the treaty) change with the design of the agreement. Since the
early 1990s numerous suggestions have been made how to design international environ-
ment agreements in order to set the right incentives for voluntary participation if not by all
then at least by as many as possible. Suggestions include side payments or transfers, the
introduction of minimum participation clauses, financial penalties for non-participants,
trade sanctions, and linking the issue of environmental protection to other issues within
20 Chapter 1 Introduction
one agreement (see, for example, Wagner, 2001; Barrett and Stavins, 2003; Perez, 2005).
From the perspective of endogenous technological change modeling, this literature on
international environment agreements raises two fundamental questions. First, as men-
tioned above, there is no induced technological change without a corresponding policy.
For induced technological change to play a key role in mitigating GHG emissions, there
needs to be a price on carbon. Thus, the question of how to raise participation in inter-
national environmental agreements is an essential prerequisite to induced technological
change.
Second, technology itself is a potential incentive to broaden international environmental
agreements. Development and diffusion of technology as well as technology transfers
work on the international level. Linking technology oriented agreements to international
environmental agreements therefore has the potential to raise participation in international
environmental agreements (de Coninck et al., 2007).
1.2 Thesis Objective
The objective of this thesis is to explore the role of endogenous technological change
(ETC) for strategies to mitigate climate change. I address (a) the role of ETC for miti-
gation costs and options and (b) international cooperation as a necessary assumption for
inducing global technological change and the role of ETC in fostering this international
cooperation.
The following chapters of this thesis are guided by two sets of research questions corre-
sponding to these two broad topics. First, existing integrated assessment models of global
mitigation options are employed to address the following questions:
•What is the impact of ETC on mitigation policy scenarios? What is the role of
economy wide feedbacks concerning ETC? What are the implications of ETC in
particular for mitigation costs and mitigation strategies, i.e. the optimal composition
of mitigation options?
•How much do integrated assessment models differ in their analysis of ETC? What
are the underlying reasons for the differences? What conclusions are robust across
models despite the model uncertainty?
Second, a newly developed dynamic model of coalition stability is used to explore some
strong assumptions made in the previous chapters. These assumptions include global
agreement to take action in mitigating climate change, and to do so in a globally coordi-
nated, cooperative way, such as to yield prices on GHG emissions globally. The following
questions guide the research in these chapters:
•What is the prospect for international cooperation on climate change mitigation?
How can it be increased by the design of international environmental agreement?
What is the potential of trade sanctions to increase participation in international en-
vironmental agreements? What are the effects on environmental and global welfare
of trade sanctions on the one hand and increased cooperation on the other hand?
1.3 Thesis Outline 21
How can competitive equilibria be computed in models with emission externality,
international trade, and tariffs?
•How can ETC help to promote international cooperation on emission abatement?
What are the roles of different technology oriented agreements (TOA)? What is the
role of cooperative research and development and technological spillovers? In which
ways does the type of technology that spills over matter? What is the role of interna-
tional technology standards?
1.3 Thesis Outline
The research questions are addressed in four journal publications, which are reproduced
as Chapters 2 to 5. Chapter 6 summarizes and draws conclusions.
Chapter 2 explores the impact of endogenous technological change on the costs of climate
protection and on mitigation strategies in terms of the optimal mix of mitigation options. I
apply the integrated assessment model MIND in a numerical sensitivity analysis to assess
the implication of parameter uncertainty for conclusions concerning endogenous techno-
logical change. In extensive parameter studies of economic and technological parameters
these uncertainties are explored further, and insights are gained into feedbacks between
technological progress and macro-economic dynamics. This chapter has been published
in the Energy Journal.2
Chapter 3 compares ten state of the art integrated assessment models incorporating fea-
tures of endogenous technological change. The aim is to learn from the differences in the
effects of endogenous technological change in these models, and to identify conclusions
that are robust across models. In preparation of the model comparison exercise, all mod-
eling teams were invited to two workshops on the implementation of endogenous tech-
nological change within each model, and the implementation of the numerical scenarios
specific to this comparison. In particular, two sets of policy scenarios were run to analyse
the impact of technological change being endogenous under ceteris paribus conditions,
namely CO2concentration stabilization in presence and absence of endogenous techno-
logical change. The models’ business-as-usual projections were harmonized to minimize
so-called “baseline effects.” The analysis of model results focused on aggregated indices
of mitigation costs and strategies that could be obtained from all models despite the large
divergence in model design. Costs are evaluated as reductions in gross world product.
Mitigation strategies were analysed in two ways: first, by applying a decomposition anal-
ysis to carbon dioxide reductions along Kaya’s identity using the refined Laspeyres index
method, and second, by comparing mitigation strategies in terms of the mix of techno-
logical options in the energy sector. Furthermore, the carbon price and usage of carbon
sequestration and storage are assessed as indicators of the economies’ dependency on fos-
sil fuels and the importance of an end-of-pipe technology for carbon free energy. Close
cooperation with the participating modeling teams was necessary to ensure the accurate
interpretation of the numerical results. This chapter has been published in the Energy
2Edenhofer, O., K. Lessmann, N. Bauer (2006): Mitigation Strategies and Costs of Climate Protection:
The Effects of ETC in the Hybrid Model MIND. Energy Journal Special Issue Endogenous Technological
Change and the Economics of Atmospheric Stabilisation, 207–222.
22 Chapter 1 Introduction
Journal.3
In Chapter 4 I explore incentives to foster participation in an international environmen-
tal agreement that aims to mitigate GHG emissions. In particular, the prospect of trade
sanctions to stabilize coalitions are addressed. For this purpose, I develop an integrated
assessment model in the economic framework of multi-actor optimal growth models. The
model accounts for climate change as a stock pollutant (CO2concentration and global
mean temperature), damages from climate change, and international trade and tariffs.
The implementation includes an algorithm to solve for a competitive equilibrium despite
multiple externalities in the economy. In addition to the effect of tariffs on participation,
I analyse the impact on environmental effectiveness, global welfare, and credibility of
imposing the sanctions. This chapter is accepted for publication in Economic Modelling.4
Chapter 5 considers the scope of technology oriented agreements for fostering interna-
tional cooperation by examining the impact of cooperative research and development
(R & D) on the one hand and international technology standards on the other hand. The
basic model from Chapter 4 is extended for this paper to allow for knowledge spillovers
in two sectors: R & D aimed at augmenting labor productivity, and R & D targeting mit-
igation technology. In the analysis, R & D cooperations are compared in terms of their
effectiveness to raise participation, sustain environmental protection, and their effect on
global welfare. International technology standard are assessed as a complement to re-
search cooperation as well as by themselves. This chapter is submitted to Resource and
Energy Economics.5
Chapter 6 concludes.
3Edenhofer, O., Lessmann, K., Kemfert, C., Grubb, M., and Koehler, J. (2006): Induced Technological
Change: Exploring its Implications for the Economics of Atmospheric Stabilization: Synthesis Report from
the Innovation Modeling Comparison Project. Energy Journal Special Issue Endogenous Technological
Change and the Economics of Atmospheric Stabilisation, 57–107.
4Lessmann, K., R. Marschinski, and O. Edenhofer: The Effects of Tariffs on Coalition Formation in a
Dynamic Global Warming Game. Economic Modelling (2009), doi:10.1016/j.econmod.2009.01.005.
5Lessmann, K. and O. Edenhofer: Research cooperation and international standards in a model of coali-
tion stability. Resource and Energy Economics, submitted.
Chapter 2
Mitigation Strategies and Costs of Climate Protection
The Effects of ETC in the Hybrid Model MIND∗
Ottmar Edenhofer
Kai Lessmann
Nico Bauer
∗published in Energy Journal as Edenhofer, O., K. Lessmann, N. Bauer (2006), “Mitigation Strategies
and Costs of Climate Protection: The Effects of ETC in the Hybrid Model MIND.” Special Issue Endoge-
nous Technological Change and the Economics of Atmospheric Stabilisation, 207–222.
23
24 Chapter 2 Mitigation Strategies and Costs of Climate Protection
207
Mitigation Strategies and Costs of Climate Protection:
The Effects of ETC in the Hybrid Model MIND
Ottmar Edenhofer*, Kai Lessmann*, Nico Bauer**ttmar Edenhofer*, Kai Lessmann*, Nico Bauer**
MIND is a hybrid model incorporating several energy related sectors
in an endogenous growth model of the world economy. This model structure
allows a better understanding of the linkages between the energy sectors and the
macro-economic environment. We perform a sensitivity analysis and parameter
studies to improve the understanding of the economic mechanisms underlying
opportunity costs and the optimal mix of mitigation options. Parameters
representing technological change that permeates the entire economy have a
strong impact on both the opportunity costs of climate protection and on the
optimal mitigation strategies e.g. parameters in the macro-economic environment
and in the extraction sector. Sector-specific energy technology parameters change
the portfolio of mitigation options but have only modest effects on opportunity
costs e.g. learning rate of the renewable energy technologies. We conclude that
feedback loops between the macro-economy and the energy sectors are crucial
for the determination of opportunity costs and mitigation strategies.
1. SETTING THE SCENE
The Innovation Modeling Comparison Project (IMCP) explores the
consequences of endogenous technological change (ETC) for the economics of
stabilizing atmospheric carbon dioxide (CO2) concentration. This paper contributes
to the IMCP by presenting an analysis of technological change, both at different
levels and in different sectors of the Model of Investment and technological
Development (MIND). MIND combines an intertemporal endogenous growth
model of the macro-economy with sector-specific and technological details taken
The Energy Journal, Endogenous Technological Change and the Economics of Atmospheric
Stabilisation Special Issue. Copyright ©2006 by the IAEE. All rights reserved.
* Potsdam Institute for Climate Impact Research (PIK), P.O. Box 60 12 03, Germany, E-mail:
** Paul Scherrer Institute (PSI), 5232 Villigen PSI, Switzerland.
2.1 Setting the Scene 25
208 / The Energy Journal
from the field of energy system modeling. In particular, we explore the impact of
endogenous technological change on opportunity costs and mitigation strategies
within the framework of a social cost-effectiveness analysis.
We explore the impact of ETC in a social cost-effectiveness framework
because we want to understand how technological change is induced by climate
policy. Several studies have already incorporated aspects of ETC in this
framework (Buonanno et al, 2003; Chakravorty et al, 1997; Goulder and Mathai,
2002; Kypreos and Barreto, 2000; Nordhaus and Boyer, 2000; Nordhaus, 2002;
Popp, 2004a; 2004b). The added value of MIND arises mainly from two features.
First, we incorporate a wide spectrum of relevant mitigation options, including
improvement of energy efficiency, carbon capture and sequestration (CCS),
renewable energy technologies, and traditional non-fossil fuels (exogenous time
series for large hydropower and nuclear). Second, technological change in MIND
has an endogenous formulation with R&D investments in labor and energy
productivity, learning-by-doing, and vintage capital in the different energy sectors.
We believe that including these features of ETC is essential for the assessment of
macro-economic mitigation costs and the portfolio of mitigation options. MIND
is a hybrid model merging features from bottom-up and top-down models. It
resembles a bottom-up model because it comprises several energy sectors. However,
compared to energy system models, the technologies are represented at a more
aggregated level. In MIND, these sectors are embedded within a macro-economic
environment, in order to evaluate the feedbacks between the macro-economy and
the energy sector (see Manne et. al. 1995 for an example of a similar exercise).
We will show that these feedbacks are crucial for an understanding of opportunity
costs and mitigation strategies in an economy faced with climate policy.
The next section briefly introduces the model and its calibration,
highlighting the improved treatment of CCS in MIND 1.1. Section 3 discusses
technological change within MIND, forming the main part of this paper. Section
4 draws conclusions.
2. THE MODEL STRUCTURE OF MIND 1.1
The model equations of MIND are introduced and discussed in
Edenhofer, Bauer and Kriegler (2005). The model version 1.0 presented therein
has been extended by Bauer (2005), to replace exogenous scenarios of Carbon
Capture and Sequestration (CCS) with a technologically detailed, endogenous
treatment of the CCS option (model version 1.1). This study uses MIND 1.1,
adapted slightly to meet the requirements of the IMCP, and enhanced by a more
sophisticated carbon cycle (Hoos et al. 2001). The following section provides a
summary of the model structure and parameter calibrations. Model equations
are restricted to the parameters treated in the sensitivity analysis and parameter
studies in this article; for a comprehensive discussion of the model structure we
refer to Edenhofer et al. (2005) and Bauer (2005).
26 Chapter 2 Mitigation Strategies and Costs of Climate Protection
MIND is an integrated assessment model comprising a model of the world
economy drawing specific focus on the energy sector, and a climate module computing
global mean temperature changes. MIND therefore allows us to assess the impacts of
constraints to climatic change on the economy in cost-effectiveness analysis.
MIND models economic dynamics by adopting an endogenous growth
framework. It calculates time paths of investment and consumption decisions
that are intertemporally optimal. The objective is to maximize social welfare,
defined as the present value of utility (pure rate of time preferences is 1%), which
is a function of per capita consumption exhibiting diminishing marginal utility.
Most economic activity is subsumed in an aggregate CES production function
(equation 1), the output YA of which describes the gross world product (GWP).1
YA =
φ
A[
ξ
A
L(A * LA)–
ρ
A +
ξ
A
E(B * E)–
ρ
A +
ξ
A
K KA
–
ρ
A]–1/
ρ
A (1)
The income share related parameters
ξ
A are calibrated so that the actual income
shares of labor LA, energy E, and capital KA relate to each other at the ratio of
66:4:30. Total factor productivity
Φ
A is a fixed scalar calibrated to a value where
the historical output of 2000 is reproduced. The elasticity parameter
ρ
A determines
the elasticity of substitution
σ
A = (1+
ρ
A)-1. In some integrated assessment models,
the elasticity of substitution between capital and energy is 0.4 for developed
countries and 0.3 for developing countries (Manne et al, 1995). We have chosen
an overall elasticity of substitution for all three factors of
σ
A = 0.4. Labor LA
is described by an exogenous population scenario adopted from the commoncommon
POLES/IMAGEbaseline(CPI,Vuurenetal.2003).Capitalstock baseline (CPI, Vuurenetal.2003).CapitalstockVuuren et al. 2003). Capital stock. Capital stock KA is built up
through investments and depreciates at a rate of 5 %. The initial value of KA is
derived from YA and an estimated capital coefficient. Capital coefficients were
computed from the OECD database and from PWT6.1 for different countries.
Their values agglomerate around 2.5. Since energy sector capital is separate from
KA, we assume a lower capital coefficient of 2.0. Variables A and B denote the
productivities of labor and energy, respectively, and are stock variables determined
by R&D investments according to equation (2):
A
• RDA
—— =
α
A
———
γ
A
, with A(t =
τ
1) = A0 (2)
A Y
B
• RDB
—— =
α
B
———
γ
B
, with B(t =
τ
1) = B0 (3)
B Y
RDA and RDB are investment flows controlled by the central planner. The
parameters
γ
A and
γ
B (where 0<
γ
A <1. 0<
γ
B <1) model the decreasing marginal
productivity of R&D investments. They are assumed to take the values of 0.05
1. MIND is implemented in discrete time steps of 5 years. In the model equations of this text we
present the more intuitive continuous formulations, e.g. in case of derivatives.
Mitigation Strategies and Costs of Climate Protection / 209
2.2 The Model Structure of MIND 1.1 27
210 / The Energy Journal
and 0.1, respectively. Parameters
α
A and
α
B determine the productivity of R&D
investments. They are calibrated at a rate such that spending 1 % of the GWP on
energy R&D increases the energy efficiency parameter by 2.25 %; when 2.5 % of
GWP is spent on labor R&D, the labor efficiency parameter increases by 2 %.
The energy input to aggregate production, E, is an additive composite of
fossil energy, renewable energy, and traditional non-fossil energy, with the latter
given exogenously. Fossil energy is produced from energy conversion capital and
primary energy input in a CES production function. Fossil resources are converted
to primary energy using an exogenous assumption about the carbon/energy ratio
of the fossil fuel mix, its availability being described by a model of resource
extraction. Resource R is extracted by capital Kres, the average productivity of
which is subject to a scarcity effect (
κ
res,s) and a learning-by-doing effect (
κ
res,l):
R =
κ
res Kres (4)
κ
res =
κ
res,s
κ
res,l (5)
The initial resource extraction is R = 6.4 GtC (SRES), assumed to be produced
by a capital stock of Kres = 5 trillion $US. This determines
κ
res,l because
κ
res,s is
normalized to unity.
The scarcity effect
κ
res,s is determined by the marginal costs of resource
extraction Cres
mar:
χ
1
κ
res,s = ——— (6)
Cres
mar
In equation 6, parameter
χ
1 as well as the marginal costs in 2000 are set to $113.
During the simulation, marginal costs Cres
mar increase with cumulative resource
extraction CRres according to equations 7 and 8.
CRres
Cres
mar =
χ
1 +
χ
2
———
χ
4
(7)
χ
3
CRres(t) =
t
∫
τ
1
R(t')dt', with CRres (t =
τ
1) = 0 (8)
Parameter
χ
1 denotes initial costs of the fossil resource, the exponent
χ
4 captures
the curvature of the function (i.e. the timing of increasing costs), and
χ
2 gives
the marginal costs once the amount described by
χ
3 has been extracted. We
parameterize this function according to Rogner’s (1997) empirical assessment
of world hydrocarbon resources, and arrive at the values
χ
2 = 700,
χ
3 = 3500
and
χ
4 = 2.
28 Chapter 2 Mitigation Strategies and Costs of Climate Protection
The learning-by-doing effect of capital productivity
κ
res,l depends on the
ratio of actual resource extraction Eres,l to initial resource extraction E0
res,l.
κ
res,l Eres,l
κ
•
res,l = ———— (
κ
res,l
max –
κ
res,l)
———
β
res,l
– 1
(9)
τ
res,l
κ
res,l
max E0
res,l
with
κ
res,l (t =
τ
1 ) =
κ
0
res,l
The factor
β
res,l = 0.4 dampens the learning-by-doing effect: a rapid increase in
extraction induces a loss in productivity gains relative to the same increase in
extraction spread over a longer time period. Furthermore, productivity gains from
learning saturate when productivity approaches its maximum value
κ
res,l
max which is
set to twice its initial value. Parameter
τ
res,l determines the speed of learning and
is set to 100 years.
Renewable energy Eren is produced by capital Kapren which is employed
at FLHren = 2190 full load hours per year.
Eren (t) = FLHren * Kapren (t) (10)
Kapren (t) =
t
∫
t0
ω
(t – t')
κ
ren (t') Iren (t')dt' (11)
The available renewable energy capital stock in each point in time is determined
by summing over the investments into renewable energy Iren in preceding time
steps multiplied with the productivity of installed capital
κ
ren. Depreciation is
modeled by weights ω which determine the fraction of capital that still remains.
ω
1 to
ω
7 are set to 1.0, 0.9, 0.8, 0.7, 0.5, 0.15, 0.05, and
ω
i = 0 if i > 7. This allows
to model different capital productivities for different vintages of the capital stock.
Capital productivity
κ
ren indeed changes in time because the costs of renewable
energy equipment cren decrease, subject to learning-by-doing.
1
κ
ren = ——————— (12)
c
ren (t) + cfloor
The inverse of floor costs cfloor = 500 US$/kW constrains capital productivity from
above, while cren starts out at cren = 700 US$/kW and decreases with cumulative
installed capital CKapren:
CKapren =
t
∫
τ
0
Kapren (t')dt' (13)
The following equation describes the dynamics of learning-by-doing in the
renewable sector:
cren,t – cren,t– 1 = cren,0 CKap–μr
re
en
n,0
(CKap–μr
re
en
n,t – CKap –μr
re
en
n,t–1)
Mitigation Strategies and Costs of Climate Protection / 211
2.2 The Model Structure of MIND 1.1 29
212 / The Energy Journal
CKapren,t – 1
×
——————
β
ren
(14)
CKapren,t
with cren (t = 0) = c 0
ren,
The learning parameter μren determines the learning rate lr and reflects a learning
rate of 15 %, i.e. investment costs decrease by 15 % with every doubling of
cumulative installed capacity. Parameter
β
ren within the last factor of the right
hand side of the equation causes a dampening similar to
β
res,l in the learning-by-
doing equation of the fossil resource extraction (equation 9). Set to
β
ren = 0.4, it
prevents learning that is too fast.
There are three sources of carbon dioxide emissions: fossil fuel combustion,
leakage from sequestered CO2, and emissions from land-use and land-use change.
The latter are described by an exogenous time series. Since fossil resources are
measured in tons of carbon, resource use R and emissions Em coincide, except for
land-use emissions and Carbon Capturing and Sequestration (CCS):
Em(t) = R(t) + LULUC(t) – Rcap(t) + LEAK(t), (15)
where Rcap denotes the amount of CO2 captured in a given year and LEAK denotes
leakage.
CCS is modeled as a chain process distinguishing six steps: CO2
is captured at point sources (1) and transported via pipelines to sequestration
sites (2). There, the CO2 needs to be compressed (3) before it is injected into
the sequestration site (4). Then, it either remains in the site (5) or leaks into the
atmosphere (6). Processes 1-4 are capital intensive and are modeled as capital
stocks representing available capacities for the individual processes. Capacities
are built up by investments according to the following equation:
Kpq (t) =
t
∫
t0
ω
q
(t – t')
ι
–1
pq (t') Ipq (t')dt' (16)
Variables Kpq denote the capacities, index p denotes the process step, and the
index q denotes different investment alternatives such as one of five distinct
capture technologies or one of six distinct sequestration alternatives. Weighting
parameters
ω
introduce a depreciation scheme for different vintages of the
capital stocks, similar to equation (11) in case of renewable energy. Investments
are denoted Ipq and the investment costs are
ι
pq. Investment costs for capturing
capacity range from ~100 $US/tC to ~450 $US/tC depending on the specific
capture technology. When the productivity of CCS investments is varied in
parameter studies later on in this paper, the same relative change is applied to the
investment costs for each technology.
In addition to the limitation inflicted by the necessity to build up
capacity, the amount of carbon that may be captured is limited by a static and
a dynamic constraint. The static constraint limits the amount of carbon which
can be captured from a large power plant as a fraction of the resource use in
30 Chapter 2 Mitigation Strategies and Costs of Climate Protection
the business-as-usual scenario. The dynamic constraint defines an upper limit
of investments into the specific capture technologies in each period. The upper
limit is defined as a share of the investments in the power generation sector. The
rationale is that the capability of retrofit investments in large power plants depends
on the total amount of investments undertaken in the power generation sector.
The injection of CO2 into particular sequestration sites demands two
types of facilities: compressors and injection wells (steps 3 and 4). The modeling
approach takes into account that both facilities demand investments and secondary
energy. In steps 5 and 6, the modeling approach considers the capacity constraint
of each sequestration alternative j and leakage of sequestered carbon: Leakage is
described by a rate, and the capacity of each sequestration alternative is the upper
bound for the cumulative amount of CO2 that is injected into each sequestration
alternative.
3. THE ROLE OF ENDOGENOUS TECHNOLOGICAL CHANGE IN MIND
In what ways does endogenous technological change matter in policy
scenarios computed with MIND? In the following sections, we explore this
question using sensitivity analysis and miscellaneous parameter studies (see Bauer
et al, 2005 for initial parameter studies with MIND). In the sensitivity analysis, we
rank important technology-related model parameters according to their influence
on two model outputs: the opportunity costs of climate protection and the mix of
options used for CO2 mitigation. We then study the effect of parameter variations
on the same model outputs and analyze the underlying economic dynamics. All
model runs stabilize atmospheric CO2 concentration level at 450 ppm.
3.1 Local Sensitivity Analysis
Figure 1a and 1b show the influence of important parameters of
MIND on opportunity costs of climate policy (1a) and on the mix of mitigation
options (1b). The former are measured as losses of gross world product (GWP),
accumulated from 2000 to 2100 and discounted to present value at a rate of 5 %,
relative to the business-as-usual scenario. The latter is represented by the ratio of
the two dominant options, renewable energies and CCS, where a ratio of unity
implies that the same amount of CO2 reductions may be attributed to each of
the mitigation options. Parameter influence is measured by the response of the
model to a 5 % variation of the parameter. Taking the set of parameters from the
model calibration as the starting point, we vary one parameter at a time, hence the
effects reflect local sensitivity. As local sensitivity analysis assesses parameter
sensitivity at only one point in parameter space it neglects the fact that sensitivities
may vary tremendously at other points in parameter space. Using a measure of
global sensitivity, i.e. a measure that takes into account simultaneous variation of
several parameters, is preferable as it provides a remedy to this shortcoming.
Mitigation Strategies and Costs of Climate Protection / 213
2.3 The Role of ETC in MIND 31
214 / The Energy Journal
However, local sensitivity analysis is used in this paper for the following
two reasons. Firstly, the model response to a change in a single parameter, ceteris
paribus , is an intuitive measure. Secondly, the computational burden for a
local analysis is much lower. To emphasise, while this analysis sheds light on
Figure 1. Sensitivity Analysis
1a.
1b.
Figures 1a and 1b show the influence of important technological parameters on opportunity costs and
mix of mitigation options, respectively. Metric is the deviation of the output in response to an up to
5% increase (decrease) of the parameter. The parameter “e.o.s. production” refers to the elasticity of
substitution
σ
A in aggregate industrial production, i.e. production of the gross world product.
32 Chapter 2 Mitigation Strategies and Costs of Climate Protection
the influence of parameters and the potential influence of their uncertainties on
model results, we do not explicitly test parameter uncertainties. Therefore, we
make no statements about the relative importance of parameters in contributing
to the uncertainty of computed results, but rather, about the ir potential to impact
results themselves.
As Figure 1a indicates, the greatest influence on opportunity costs is
exerted by the elasticity of substitution σA, followed by the parameters describing
the availability of fossil resources, and the effectiveness of R&D investments in
labor productivity. The latter and the top three parameters have a positive effect
on costs, i.e. costs increase with the parameters, whereas the assumption of high
marginal future fossil resources costs have a negative effect. Productivity of energy
efficiency R&D and the learning rate of the renewable energy technologies rank
next, followed by two more sector specific parameters, the learning parameter in
fossil resource extraction and the efficiency of investments in CCS. Overall, the
relatively small responses of the model to parameter variations (less than 5%)
improves the confidence in the robustness of the computed opportunity costs.
In the next two sections we will explore the reasons for this observation, and
evaluate the role of technological change in deriving these results.
Figure 1b depicts the influence of parameters on the mix of mitigation
options. It is immediately evident from a comparison between Figure 1a and Figure
1b that the ranking of parameters has changed. Most notably, the elasticity of
substitution has dropped to the bottom rank, and two resource related parameters,
χ
2 and
χ
3, also emerge to fall in ranking. Conversely, the parameterization of
labor R&D, the learning rate of renewable technologies, and the efficiency of
CCS investments have risen in the hierarchy. Overall, the mitigation mix is
more sensitive (with variations up to 10 %) than the mitigation costs in Figure
1a. This result comes as no surprise. Since GWP losses are closely related to
social welfare, the maximization of which is the objective of MIND, GWP loss
is deliberately kept to a minimum. The mix of mitigation options, on the other
hand, is endogenously determined to minimize costs. It is intuitive that a change
in the parameter values alters the competitiveness of mitigation options, hence its
impact on the mitigation mix is significant.
3.2 Determinants of the Opportunity Costs
This section takes a closer look at the opportunity costs of climate
protection. We present parameter studies varying two parameters simultaneously.
This enables us to discuss the effects of varying these parameters, as well as
analyzing the interdependencies between them, hence taking a first step beyond
a local sensitivity analysis presented in Section 3.1. To an extent, this analysis
remains very much local in character since many parameters remain fixed at
their default levels. However, restricting the variation to two parameters at a time
enables an intuitive graphical presentation of the results, which provides deeper
and useful insights into the workings of MIND.
Mitigation Strategies and Costs of Climate Protection / 215
2.3 The Role of ETC in MIND 33
216 / The Energy Journal
We start out by taking a look at the engine of endogenous growth in
MIND: R&D investments that drive labor and energy efficiencies. Figure 2a
displays the productivity of these investments. While the two parameters are similar
with respect to the process they describe – accumulation of a knowledge stock
increasing the productivity of an input factor to aggregate production – their effects
on opportunity costs are contrary. An enhanced effectiveness of labor productivity
R&D raises costs, while better energy efficiency R&D reduces GWP losses. This is
due to opposite effects on the mitigation gap, i.e. the discrepancy of CO2 emissions
between business-as-usual and climate policy scenarios. More effective labor R&D
stimulates additional economic growth and implies higher CO2 emissions in the
baseline. More effective energy R&D investments, on the other hand, facilitate
much better energy efficiency in the baseline, and hence lowers CO2 emissions.
Figure 2. Parameter Studies of Mitigation Costs
Figures in this panel show discounted gross world product loss (discount rate is 5 %) for several
parameter studies. In figure 2a, energy R&D and labor R&D refer to the productivity of investment
into research that enhances the efficiency of the corresponding factor. In 2b, e.o.s. production refers to
the elasticity of substitution in the aggregate industrial production sector. Parameters
χ
3 and
χ
4 in figure
2b and 2c refer to the size of the fossil resource base and the exponent of the Rogner curve, respectively.
Figure 3d treats the learning rate of renewable technologies and the efficiency of investments in CCS
technology. The pairs of default parameter values are indicated with a bold cross.
34 Chapter 2 Mitigation Strategies and Costs of Climate Protection
The mitigation gap characterizes the challenge for the economy facing climate
protection goals and manifests itself in the opportunity costs.
Figure 2b compiles two parameters with an effect of the second type:
the elasticity of substitution in the aggregate production sector, and the estimated
size of the available fossil resources. Figure 2b shows that costs increase with the
elasticity of substitution. This too can be attributed to baseline effects: higher
elasticity of substitution implies a more flexible production technology which
induces higher economic growth in the business-as-usual scenario. Therefore,
achieving 450 ppm requires a substantial departure from the baseline and is
relatively costly. A variation of the resource base has a bigger impact on the
mitigation costs if the elasticity of substitution is relatively high. Low values of the
elasticity of substitution hinder economic growth and consequently imply a lower
demand for energy. At low energy demand, relaxing the scarcity of the resource
has a smaller effect. In general, a larger resource base allows higher economic
growth in the business-as-usual case. When climate policy constrains resource
use, it devaluates exhaustible resource as an economic asset and diminishes the
rent income of their owners. The loss of rent income increases with the resource
base because a relatively cheap and abundant resource can no longer be used as
input in production.
We take yet a closer look at the fossil resource base. Figure 2c studies
the variation of the size of the resource base
χ
3 and parameter
χ
4. Parameter
χ
4
as well as the resource base are proxy variables for the technological progress
in the extraction sector. Increasing
χ
3, i.e. assuming more abundant resources,
results in cheaper short to medium term supply of the fossil resource. Increasing
χ
4 trades a slow and steady increase of the marginal costs for a steeper increase
at a later time – thus making the resource cheaper and more easily available in
the short to medium term. High values of
χ
4 allow higher economic growth in
the business-as-usual case and induce a relatively large mitigation gap. For high
values of
χ
4 the marginal costs of extraction are essentially constant. Under this
condition, an increased resource base has moderate impact on macro-economic
mitigation costs. For low values of
χ
4, an increased resource base has a slightly
higher impact on the macro-economic costs because marginal improvements
in extraction already increase the shadow price of the resource. This parameter
study shows that climate protection becomes relatively costly if there is a high
rate of technological progress in the exploration and extraction of fossil fuels.
Accelerated technological progress in the extraction sector makes climate policy
more costly, because such policy devaluates assets (resources and capital stock
in the corresponding sectors). Therefore, special attention ought to be paid to
assumptions about resource availability and their uncertainties.
Contrary effects can be observed if technological progress decreases the
costs of mitigation technologies. The impact on opportunity costs is shown in
Figure 2d. We explore two parameters which are both closely related to mitigation
options: the efficiency of investments into Carbon Capture and Sequestration
technologies (CCS) and the learning rate of renewable energy technologies.
Mitigation Strategies and Costs of Climate Protection / 217
2.3 The Role of ETC in MIND 35
218 / The Energy Journal
Varying these two parameters shifts the competitive advantage between the two
mitigation options and, consequently, the extent to which they are used. It turns
out that the efficiency of CCS investments has no strong impact on the overall
opportunity costs if the learning rate of renewable energy technologies is relatively
high. The reason is that renewables are modeled as a backstop technology, i.e.
as a carbon-free energy source, and need no non-reproducible input for energy
production. In contrast to the renewables, CCS investments only bridge from the
fossil fuel age to a carbon-free era. CCS makes the transition of the energy system
smoother but has severe limitations if fossil fuels become more costly because of
increasing marginal extraction costs at the end of the 21st century. At the same time,
renewable energy becomes cheaper because of learning-by-doing. It is plausible
that this effect cannot be altered by high efficiencies of CCS investments. At low
learning rates of the backstop technology, CCS becomes more important.
3.3 Mitigation Strategies
In this section we analyze the impact of the same parameters explored
in the previous section on the option portfolio of an optimal mitigation strategy.
Mitigation options are compared on the basis of the amount of CO2 that they
enable the economy to reduce. For the CCS option, this is straightforward: it is
simply the amount of captured and sequestered CO2 (less the amount that leaks
from the sequestration site). In case of energy related mitigation options, i.e.
renewable energy and energy efficiency improvements, the corresponding amount
of “mitigated CO2 emissions” was derived from the equivalent amount of energy
from fossil fuels. In , the degree of efficiency on converting primary into final
energy is determined endogenously in the production function of the fossil sector.
In this ex post analysis, however, we estimate the “equivalent” amount of fossil
energy by assuming a fix coefficient. The remaining mitigation options, namely
energy savings by substitution of energy at the levels of energy transformation
and aggregate production, are visualized as the difference to the total reduction
of CO2.
Figure 3a shows that the amount of CCS within the portfolio of
mitigation options increases with the assumed resource base. The cumulative
amount of CO2 reduced by renewables within the next century decreases, energy
efficiency remains constant and energy savings increase. An increasing resource
base implies increasing rents for the owners. This increasing rent income makes
CCS a more profitable option. Due to high economic growth and relatively cheap
fossil fuels, the return on investment in renewables falls short of the returns on
CCS investments.
In figure 3b, energy savings (reduction of energy consumption by
substituting energy by capital in different sectors) become more profitable if the
elasticity of substitution increases; at the same time, the importance of energy
efficiency decreases.
36 Chapter 2 Mitigation Strategies and Costs of Climate Protection
Mitigation Strategies and Costs of Climate Protection / 219
Figure 3. Parameter Studies of Mix of Mitigation Options
Figures 3a-f show how the mix of mitigation options varies in parameter studies. CO2 reductions
caused by avoiding the use of fossil fuels (renewable energy, energy efficiency improvements, and
substitution) are estimated from the alternative use of fossil fuels. Dashed lines indicate the default
parameter value.
2.3 The Role of ETC in MIND 37
220 / The Energy Journal
A more surprising result is obtained in figure 3c and 3d. In figure 3c
an increasing productivity of R&D investment in labor enhancing activities also
increases the share of renewables in the mitigation portfolio. The explanation
is as follows: economic growth induces additional energy demand that is met
by carbon-free technologies. Due to high economic growth, marginal extraction
costs of fossil fuels increase sooner, and thus CCS is less competitive compared
to renewables. In contrast, when R&D investments in energy efficiency become
more productive, the mitigation gap shrinks, and the share of renewables within
the mitigation portfolio decreases (3d). Interestingly, changes in the productivity
of energy R&D investments affect the baseline rather than providing a more
attractive mitigation option. In this study, the energy efficiency parameter varies
from 63 to 245 % of its regular value in 2100 in the baseline, the latter implying
that energy use in 2100 is decreased by 60%. Climate policy, however, only
induces 0.4 to 2.7 % additional increases of the efficiency parameter. To sum,
higher energy efficiency and a lower baseline for economic growth reduce the
demand for renewables. The importance of the renewable energy option depends
heavily on the underlying economic growth path.
As figure 3e shows, high learning rates in the renewable energy sector
reduce the optimal amount of CCS substantially. In that sense CCS can be seen
as a joker-option if the learning rate of the renewables is relatively low. It is
also remarkable that energy savings are less important when the learning rate
is relatively high because the energy demand can be met by the carbon-free
renewables. Learning-by-doing reduces the price of electricity produced by
renewables and increases the demand for renewables which reduces their costs
further. This feedback loop makes CCS less important. As figure 3f indicates,
this effect can be counteracted by an increasing efficiency of CCS-investments.
4. CONCLUDING REMARKS
In what ways does technological change matter? Our analysis shows
that technological change works in two “directions”: we identify technological
progress that permeates the entire economy and technological progress that
is restricted in its effects to a single sector. Examples for such sector-specific
technological change are learning-by-doing effects associated with renewable
energy technologies and resource extraction, as well as technological progress in
CCS, here modeled via its investment efficiency. In , parameters associated with
such sector specific technological change have a significant impact2 on the optimal
mix of mitigation options. For example, an increased learning rate increases the
share of renewables, and improved investment efficiency in CCS increases the
share of CCS within the entire portfolio of mitigation options (Figures 1b and
2. We refer to the impact of a parameter in terms of a relatively large potential influence, i.e. a large
sensitivity of results to changes of this parameter. Recall, however, that the actual uncertainty about
parameters is not taken into account.
38 Chapter 2 Mitigation Strategies and Costs of Climate Protection
3ef). However, these parameters are less important in determining the overall
opportunity costs of climate protection which measure the impact on the overall
economy (Figure 1a).
In contrast, there is technological change with significant impact on the
macro-economic growth process, evident in its influence on opportunity costs.
Such technological change is described by parameters of the macro-economic
environment, like the elasticity of substitution, and the parameters characterizing
the effectiveness of labor- and energy R&D investments. Labor R&D investments
in particular have a strong influence on macro-economic growth as well as the
mix of mitigation options. Progress in resource extraction is an example of sector-
specific technological change with a macro-economic impact. This progress is
characterized by the parameters of Rogner’s scarcity curve and has been shown
to exert a significant influence on opportunity costs. The most prominent effect
of these parameters is their impact on the baseline.
We conclude that feedbacks between the macro-economy and the energy
system are crucial for determining mitigation costs and the development of the
mitigation portfolio in time. The case of technological change in resource extraction
shows how sector-specific processes may exert significant influence on the macro-
economy, while the impact of labor R&D productivity on the share of renewable
energy is an example of macro-economic influence on a distinct sector.
This has strong implication for policy. A sector-specific policy that
fosters technological change in the extraction sector induced by increasing
prices in the oil or gas market would increase the opportunity costs of climate
protection. A policy that increases the economy-wide energy efficiency in all
energy related sectors would reduce the costs of climate protection substantially.
Enhancing technological change in the extraction sector makes sense, if decision
makers intended only to increase energy security. Analysis here highlights that
the impact of such a policy on the opportunity costs of climate protection must
also be taken into account.
The results presented here indicate that partial-equilibrium models
omitting intertemporal and inter-sectoral aspects can be misleading for designing
a climate and energy policy. Thus, they stress the utility of hybrid models
incorporating endogenous technological change at the sector level as well as at
the macro-economic level. Moreover, hybrid models pose a coherent framework
not only for the assessment of the opportunity costs and portfolios of mitigation
strategies, but also for the design of climate and energy policy instruments.
ACKNOWLEDGEMENTS
We are grateful to Elmar Kriegler for productive discussions and
helpful comments on earlier version of this paper. This work was funded by the
Volkswagen Foundation, Project II/78470, which we gratefully acknowledge.
Mitigation Strategies and Costs of Climate Protection / 221
2.4 Concluding Remarks 39
222 / The Energy Journal
REFERENCES
Bauer, N. (2005). “Carbon Capturing and Sequestration. An Option to Buy Time?” Unpublished
Ph.D. Thesis at University Potsdam, Faculty of Economics and Social Sciences.
Buonanno, P., C. Carraro and M. Galeotti (2003). “Endogenous Induced Technical Change and the
Costs of Kyoto.” Resource and Energy Economics 25:11–34.
Chakravorty, U., J. Roumasset and K. Tse (1997). “Endogenous Substitution among Energy Sources
and Global Warming.” Journal of Political Economy 105:1201–1234.
Edenhofer, O., N. Bauer and E. Kriegler (2005). “The Impact of Technological Change on Climate
Protection and Welfare.” Ecological Economics 54:277-292.
Goulder, L. H. and K. Mathai (2002). “Optimal CO2 Abatement in the Presence of Induced
Technological Change.” In A. Grübler, N. Nakicenovic and W. D. Nordhaus, eds., Technological
Change and the Environment. Resources for the Future, Washington, USA, 210–250.
Hoos, G., R. Voss, K. Hasselmann, E. Maier-Reimer and F. Joos (2001). “A Nonlinear Impulse
Response Model of the Coupled Carbon Cycle Climate System (NICCS).” Climate Dynamics
18:189-202.
Kypreos, S. and L. Barreto (2000). “A Simple Global Electricity MARKAL Model with Endogenous
Learning.” Paul Scherrer Institute, General Energy Research Department-ENE, Energy Modelling
Group, Villigen, Switzerland.
Manne, A., R. Mendelsohn and R. Richels (1995). “MERGE. A Model for Evaluating Regional and
Global Effects of GHG Reduction Policies.” Energy Policy 23, 17–34.
Nordhaus, W. D. (2002). “Modelling Induced Innovation in Climate-Change Policy.” In A. Grübler,
N. Nakicenovic and W. D. Nordhaus, eds., Technological Change and the Environment. Resources
for the Future, Washington DC, USA, 182–209.
Nordhaus, W. D. and J. Boyer (2000). Warming the World. Models of Global Warming. MIT Press,
Cambridge, Massachusetts, USA.
Popp, D. (2004a). “ENTICE: Endogenous Technological Change in the DICE Model of Global
Warming.” Journal of Environmental Economics and Management 48:742–768.
Popp, D. (2004b). “ENTICE-BR: The Effects of Backstop Technology and R&D on Climate Policy
Models.” NBER Working Paper #10285.
Rogner, H.-H. (1997). “An Assessment of World Hydrocarbon Resources.” Annual Review of Energy
and Environment 22:217–262.
Vuuren, D. P. van, M. G. J. den Elzen, M. M. Berk, P. Lucas, B. Eickhout, H. Eerens, R. Oostenrijk
(2003). “Regional costs and benefits of alternative post-Kyoto climate regimes – Comparison of
variants of the Multi-stage and Per Capita Convergence regimes.” RIVM report 728001025/2003.
National Institute of Public Health and the Environment, Bilthoven, The Netherlands.
40 Chapter 2 Mitigation Strategies and Costs of Climate Protection
Chapter 3
Induced Technological Change: Exploring its Implications
for the Economics of Atmospheric Stabilization:
Synthesis Report from the Innovation Modeling
Comparison Project∗
Ottmar Edenhofer
Kai Lessmann
Claudia Kemfert
Michael Grubb
Jonathan Köhler
∗published in Energy Journal as Edenhofer, O., Lessmann, K., Kemfert, C., Grubb, M., and Koehler, J.
(2006) Induced Technological Change: Exploring its Implications for the Economics of Atmospheric Sta-
bilization: Synthesis Report from the Innovation Modeling Comparison Project. Special Issue Endogenous
Technological Change and the Economics of Atmospheric Stabilisation, 57–107.
41
42 Chapter 3 Implication of ITC for Atmospheric Stabilization
57
The Energy Journal, Endogenous Technological Change and the Economics of Atmospheric
Stabilisation Special Issue. Copyright ©2006 by the IAEE. All rights reserved.
* Potsdam Institute for Climate Impact Research, P.O. Box 60 12 03, Germany, E-mail: edenhofer@
pik-potsdam.de.
** DIW (German Institute for Economic Research) and Humboldt University Berlin, Germany, E-
mail: ckemfert@diw.de.
*** Imperial College and Faculty of Economics, Cambridge University, United Kingdom, E-Mail:
† Tyndall Centre and Faculty of Economics, University of Cambridge, Cambridge CB3 9DE,
Sidgwick Avenue, United Kingdom, E-Mail: J.K[email protected].
Induced Technological Change: Exploring its Implications
for the Economics of Atmospheric Stabilization:
Synthesis Report from the Innovation Modeling
Comparison Project
Ottmar Edenhofer*, Kai Lessmann*, Claudia Kemfert**,
Michael Grubb*** and Jonathan Köhler†
This paper summarizes results from ten global economy-energy-environment
models implementing mechanisms of endogenous technological change (ETC).
Climate policy goals represented as different CO2 stabilization levels are imposed,
and the contribution of induced technological change (ITC) to meeting the goals
is assessed. Findings indicate that climate policy induces additional technological
change, in some models substantially. Its effect is a reduction of abatement costs in
all participating models. The majority of models calculate abatement costs below 1
percent of present value aggregate gross world product for the period 2000-2100. The
models predict different dynamics for rising carbon costs, with some showing a decline
in carbon costs towards the end of the century. There are a number of reasons for
differences in results between models; however four major drivers of differences are
identified. First, the extent of the necessary CO2 reduction which depends mainly on
predicted baseline emissions, determines how much a model is challenged to comply
with climate policy. Second, when climate policy can offset market distortions, some
models show that not costs but benefits accrue from climate policy. Third, assumptions
about long-term investment behavior, e.g. foresight of actors and number of available
investment options, exert a major influence. Finally, whether and how options for
carbon-free energy are implemented (backstop and end-of-the-pipe technologies)
strongly affects both the mitigation strategy and the abatement costs.
43
58 / The Energy Journal
1. INTRODUCTION
The Innovation Modeling Comparison Project (IMCP) aims to look at
the impact of induced technological change (ITC) on the economics of stabiliz-
ing carbon dioxide emissions at different levels. The IMCP is motivated by the
conviction that endogenous technological change1 (ETC) is vital in modeling eco-
nomic dynamics over the lengthy time scales required in climate policy analysis.
Despite considerable progress in ETC research, significant discrepancies among
models as well as uncertainties of model results still remain. The IMCP advances
the understanding of ETC by assessing these discrepancies and analyzing their
potential causes. This paper summarizes a quantitative model comparison experi-
ment using a broad range of relevant models.
Two types of uncertainties contribute to the discrepancy of the results
from different models. First, there is parameter uncertainty, referring to a lack
of empirical knowledge to calibrate the parameters of a model to their “true”
values. Parameter uncertainty implies an uncertainty of the predictions of any
one model and discrepancies may result even in case of otherwise very similar
models. Parameter uncertainty is addressed in model specific uncertainty analy-
ses including sensitivity analysis and parameter studies, and modeling teams in
the IMCP were encouraged to explore parameter uncertainty in the individual
papers collected in this special issue. Second, there is structural uncertainty or
model uncertainty, defined as the uncertainty arising from having more than one
plausible model structure (Morgan and Henrion 1990, p. 67). In this paper, we
address model uncertainty.
In general, model uncertainty may be reduced by eliminating possible
model structures from the set of plausible models. One way of doing so is validat-
ing models against empirical evidence to discriminate “better” models and con-
sequently discard “bad” models. However, even “perfect validation” provides noHowever, even “perfect validation” provides no “perfect validation” provides noprovides no no
proof that a model best explains reality. Alternatively, “Ockham’s razor” proposesAlternatively, “Ockham’s razor” proposes“Ockham’s razor” proposesOckham’s razor” proposes’s razor” proposess razor” proposes” proposes proposes
that if another model explains the same empirical phenomena using less specific another model explains the same empirical phenomena using less specific
or more intuitive assumptions and parameters, then it can be deemed preferable.then it can be deemed preferable.it can be deemed preferable.can be deemed preferable. preferable..
Yet to this date, the theoretical and empirical foundation of technological change to this date, the theoretical and empirical foundation of technological change, the theoretical and empirical foundation of technological changethe theoretical and empirical foundation of technological change
within economics remains insufficient to allow for a sound evaluation of modelsfor a sound evaluation of modelsa sound evaluation of models
according to Ockham’s razor. In other words, the uncertainties about the appropri-
ate model structure remain.
Our approach to model uncertainty involves identifying discrepancies
in results of different models running the same scenarios, and investigating their
origins. The analysis follows four steps: First, we classify the models according to
their structure. Second, we assess discrepancies in a central model output, namely
the impact of climate policy on the economy, or the “costs” of climate policy.
1. We distinguish between endogenous and induced technological change: Technological change
is endogenous (ETC) if its course is an outcome of economic activity within the model. Given an
endogenous description, technological change in policy scenarios may exceed (or fall short of) its
extent in the baseline, i.e. policies induce additional technological change which we refer to as ITC.
44 Chapter 3 Implication of ITC for Atmospheric Stabilization
Third, we analyze the different model dynamics leading to the discrepancies us-
ing aggregated indicators of model behavior and drawing on structural informa-
tion about the models. We measure the impact of technological change on these
quantitative indicators,, ceteris paribus. Finally, we take a close look at the energy
system as a major contributor to possible climate change.
The objective of this comparison is improved understanding of how and improved understanding of how and
whether technological change matters. Technological change is a hotly debated
issue because its impact on mitigation costs and mitigation strategies has political
consequences. Recently, some models have been developed incorporating endog-
enous technological change. Examples of the papers which compare these models
in a qualitative way are Sijm (2004), Clarke and Weyant (2002), Löschel (2002), a qualitative way are Sijm (2004), Clarke and Weyant (2002), Löschel (2002), qualitative way are Sijm (2004), Clarke and Weyant (2002), Löschel (2002),
Weyant and Olavson (1999), Grubb, Köhler and Anderson (2002), and Köhler et
al. (2006), the latter includes an up to date survey of ETC in the literature.
The next section briefly summarizes the literature on modeling compari-
son; in the third section, the participating models are characterized and a taxon-
omy of models is provided. Section 4 outlines the method of comparison used in
the IMCP. In Section 5, we analyze the impact of ITC on mitigation costs, mitiga-
tion strategies, and energy mix. Section 6 offers some conclusions.
2. MODEL COMPARISONS IN THE LITERATURE
There is a broad literature on estimating the economic impact of climate
change mitigation policies using models of various types. The Assessment Reports
of the Intergovernmental Panel on Climate Change (IPCC) provide a comprehensive
overview (IPCC 1996, 2001). Moreover, the Second and Third Assessment Reports
(SAR and TAR) draw conclusions from comparative evaluations of these modeling
studies. Among the original studies of model comparison, those of the Stanford
Energy Modeling Forum (EMF) are particularly worth mentioning. This section
briefly summarizes some of the key findings of previous model comparisons.
The SAR differentiates top-down (economic) and bottom-up (engineer-
ing) models, further distinguishing Computable General Equilibrium models
(CGE), optimizing models, and econometric macroeconomic models among the
top-down approaches. Top-down and bottom-up models have been known to dif-
fer greatly in their estimates of the costs of mitigation policies. The authors of
SAR note that this classification is increasingly misleading as efforts are being
made to combine features from macro and CGE models, and to incorporate bot-
tom-up technological features in top-down models. Furthermore, they conclude
that different assumptions about the economic reality represented in the models,
e.g. about the nature of market barriers, have a far greater impact on the results
than the type of the model. In their extended discussion of results from SAR,
Hourcade and Robinson (1996) conclude that “there is no a-priori reason that
the two modeling approaches will give different results. Whether they [bottom-up
and top-down models] do or not depends largely on their respective input as-
sumptions”.
Induced Technological Change / 59
3.2 Model Comparison in the Literature 45
60 / The Energy Journal
Two Economics Reports of the PEW Center on Global Climate Change
summarize the economics of climate change policy and the role of technology
(see Weyant 2000, Edmonds et al. 2000). Both studies review why model results
differ. Weyant (2000) attributes the differences to variations mainly in the baselinevariations mainly in the baseline mainly in the baseline
emission scenarios, different flexibilities regarding where, when, and which GHGregarding where, when, and which GHGwhere, when, and which GHG
emissions are reduced, and whether or not benefits from avoided climate change
are taken into account. Once the effects of these differences are separated, the re-
sidual differences can be traced to substitution and technological change. EdmondsEdmonds
et al. (2000) emphasize Hourcade and Robinson’s (1996) finding of the importance (2000) emphasize Hourcade and Robinson’s (1996) finding of the importance(2000) emphasize Hourcade and Robinson’s (1996) finding of the importance
of assumptions underlying model design. Concerning the role of technological
change, they note that technological change mitigates costs and occurs over long, they note that technological change mitigates costs and occurs over long they note that technological change mitigates costs and occurs over long
time horizons. They stress that technological change can be induced by policies,
and that including induced technological change is important, however difficult.
On discussions about why studies differ, TAR revisits the top-down ver- discussions about why studies differ, TAR revisits the top-down ver-s about why studies differ, TAR revisits the top-down ver- why studies differ, TAR revisits the top-down ver-
sus bottom-up controversy. Top-down models are distinguished into CGE andto CGE and CGE and
time-series-based econometric models, and TAR points out that the former typetype
is arguably more suitable for describing long-run steady-state behavior, while the arguably more suitable for describing long-run steady-state behavior, while the
latter models are more suitable for forecasting in the short-run. TAR also notes thatmodels are more suitable for forecasting in the short-run. TAR also notes that more suitable for forecasting in the short-run. TAR also notes thatin the short-run. TAR also notes thatthe short-run. TAR also notes that
efforts are being made to eliminate these shortcomings (IPCC 2001, pp. 591).p. 591).. 591).
EMF 19 (2004) set out to understand how models being used for glob-
al climate change policy analyses represent current and potential future energy
technologies, and technological change. Weyant (2004) summarizes three main
insights from the study: developing and implementing new energy technology is
necessary for stabilizing atmospheric CO2 concentration; the required transition
will be costly to implement, and implementation will take many decades; but
costs may be moderated if it is possible to pursue many options, to phase in new
technologies gradually, and if supporting policies start soon.
In an extensive survey of the recent literature, Sijm (2004) focuses on
models that exhibit features of endogenous technological change.2 He separates
bottom-up and top-down studies and finds major similarities in the outcomes of
models in the former category, e.g. costs decline, the energy mix changes towards
fast learners, and total abatement costs decline. Modeling studies in the latter
category, however, show a wide diversity in outcomes with regard to the impact
of induced technological change. He identifies variations in the following modelHe identifies variations in the following model
features as possible explanations: ITC channels; optimization criteria; model: ITC channels; optimization criteria; model; optimization criteria; model optimization criteria; model; model model
functions; calibration; spillovers; and also aggregation; number and type of policy; calibration; spillovers; and also aggregation; number and type of policy calibration; spillovers; and also aggregation; number and type of policy; spillovers; and also aggregation; number and type of policy spillovers; and also aggregation; number and type of policy; and also aggregation; number and type of policy and also aggregation; number and type of policy; number and type of policy number and type of policy
instruments; and the time horizon.; and the time horizon. and the time horizon.
These modeling comparison exercises illuminate and outline reasonsilluminate and outline reasons reasons
why models differ in their cost estimates. Several studies list induced technologi-
cal change as a good candidate for explaining some of these differences. However,
the extent of its impact and the precise reasons as to how and why technological
change matters remain unclear in many cases. Focusing on the effects of ITC, all
2. For a recent collection of models incorporating ETC, see Vollebergh and Kemfert (2005).
46 Chapter 3 Implication of ITC for Atmospheric Stabilization
participating modeling teams of the IMCP deliver scenarios in which technologi-
cal change processes have been ‘switched off’ and ‘switched on’. A comparison
between these scenarios allows on the one hand, a quantitative assessment of tech-
nological change and on the other hand, a further explanation of the underlying
economic mechanisms that explain different model outputs.
3. MODEL CLASSIFICATION
The models considered in this comparative study have two common
aspects: they incorporate technological change in innovative ways and allow an: they incorporate technological change in innovative ways and allow anthey incorporate technological change in innovative ways and allow anhey incorporate technological change in innovative ways and allow an
assessment of costs of global carbon dioxide mitigation. At the same time, a widea wide
range of model types is represented in this project. Understanding the conceptions model types is represented in this project. Understanding the conceptions
underlying the designs of different model types is necessary when comparing
models within and across model types. In this section we give a summary of the
concepts on which we base our discussion. We start with a general classification,
which serves as a guideline for the brief introduction of the models that follows.
As the major motivation for the design of many models as well as a key question
in this study, we draw focus on the determination of the economic impact of cli-, we draw focus on the determination of the economic impact of cli- we draw focus on the determination of the economic impact of cli-e draw focus on the determination of the economic impact of cli- determination of the economic impact of cli-
mate policies in terms of social costs, and recapitulate different concepts of costs, and recapitulate different concepts of costsrecapitulate different concepts of costs
which are prominent in different model types..
3.1 Model Types in IMCP
In Table 1, we differentiate four models types, mainly characterized by
their calculus, i.e. the mathematical paradigm underlying the computation.
1. Optimal growth models – maximize social welfare intertemporally.
2. Energy system models – minimize costs in the energy sector.
3. Simulation models – solve initial value or boundary condition
problems (this includes econometric models, i.e. models which base
a subset of their relationships on historical time series).
4. General equilibrium market models – balance demand and supply
among multiple actors.
Many models in this study transcend the outlined categories. Whilst the
modeling paradigm that underlies a model is useful for understanding its dynam-
ics, we urge the reader to consult the individual papers for an in-depth discussion
of the models.
These papers also include discussions of the model calibration and sen-
sitivity analysis of crucial parameters. Model calibration is important to gauge the
parameter uncertainties going into the models, and sensitivity analysis assesses the
effect of these uncertainties. Model calibration includes equations of the basic mod-
el and the equations specifying how technological change behaves. That is the basic
model describing macroeconomic variables (such as gross world product, energy
demand, etc.) on the one hand, and how technological change affects the dynamics
of these main variables and is affected by them on the other hand. For this analysis,
Induced Technological Change / 61
3.3 Model Classification 47
62 / The Energy Journal
all models are calibrated such that the main variables show similar behavior during
the first twenty years of the projected time. Again, we refer the reader to the indi-
vidual model papers for details.
Model uncertainty, in particular structural differences in the description
of ETC is assessed in this report. For the purpose of model comparison, the di-, the di- the di-
versity of assumptions underlying the models (Table 2) becomes an asset to thisto this this
project as it allows for robust conclusions to be drawn.
3.1.1 Optimal Growth Models
Economic growth is a major driver for GHG emissions. Optimal growthfor GHG emissions. Optimal growth GHG emissions. Optimal growth
models are aimed at understanding growth dynamics over long term horizons. The
key property of neoclassical growth models is their social welfare maximizing be-
havior. Early growth models determined optimal capital accumulation. Endogenous
growth theory extends this framework to include economic forces that explain tech-
nological change. Among the growth models represented in this study a varying
degree of technological change is endogenous. In AIM/Dynamic-Global, growth, growth growth
accrues from autonomous energy efficiency improvements in addition to capital
accumulation (the later is of course present in all models). DEMETER-1CCS, EN-
TICE-BR and FEEM-RICE use exogenous total factor productivity (Table 2, last
column) hence ETC implemented in these models also contributes to economic
growth. In MIND, growth is fully endogenous. These models derive a first-best or
a second-best social optimum and may be used as intertemporal social cost benefit
analysis of mitigation strategies. First best models like MIND implicitly assume
perfect markets and the implementation of optimal policy tools. InIn second best mod- best mod-
Table 1. Classification of Models in the IMCP
Technological detail
Calculus Top Down Bottom Up
Welfare maximization Optimal growth models
ENTICE-BR
FEEM-RICE
DEMETER-1CCS
AIM/Dynamic-Global
MIND 1.1
Cost minimization Energy system models
MESSAGE-MACRO
GET-LFL
DNE21+
Initial value problems Simulation models
E3MG
Static equilibrium + Computational general equilibrium
recursive dynamics models (CGE)
IMACLIM-R
48 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 63
Table 2. Endogenous Technological Change (ETC) in the Participating Models
ETC related to energy intensity ETC related to carbon intensity Other ETC Exogenous TC
AIM/Dynamic- • Factor substitution in CES production • Carbon-free energy from backstop • AEEI for energy from
Global • Investments in energy conservation technology (nuclear/renewables) coal, oil, gas, and for
capital raises energy efficiency for electricity
coal, oil, gas, and electricity
DEMETER-1CCS • Factor substitution in CES production • Carbon-free energy from renewables • Learning-by-Doing • Overall productivity
and CCS for fossil fuels
• Learning-by-Doing for both
DNE21+ • Energy savings in end-use sectors • Carbon-free energy from backstop • Technological progress
modeled using the long-term price technologies (renewables/nuclear) energy technologies
elasticity. and CCS (other than wind,
• Learning curves for energy technologies photovoltaics, fuel
(wind, photovoltaic and fuel cell vehicle) cell vehicle)
E3MG • Cumulative investments and R&D • Learning curves for energy technologies • Cumulative investments
spending determine energy demand (electricity generation) and R&D spending
via a technology index determine exports via a
technology index
• Investments beyond
baseline levels trigger a
Keynesian multiplier effect
ENTICE-BR • Factor substitution in Cobb-Douglas • Carbon-free energy from generic • Total factor productivity
production backstop technology • Decarbonization
• R&D investments in energy • R&D investments lower price of accounting for e.g.
efficiency knowledge stock energy from backstop technology changing fuel mix
CONTINUED
3.3 Model Classification 49
64 / The Energy Journal
Table 2. Endogenous Technological Change (ETC) in the Participating Models (continued)
ETC related to energy intensity ETC related to carbon intensity Other ETC Exogenous TC
FEEM-RICE • Factor substitution in Cobb-Douglas • ETCI explicitly decreases carbon • Total factor productivity
production intensity (see ETCI in the energy • Decarbonization
• Energy technological change index intensity column) accounting for e.g.
(ETCI) increases elasticity of substitution changing fuel mix
• Learning-by-Doing in abatement raises
ETCI
• R&D investments raise ETCI
GET-LFL • Learning-by-Doing in energy conversion • Carbon-free energy from backstop
technologies (renewables) and CCS
• Learning curves for investment costs
• Spillovers in technology clusters
IMACLIM-R • Cumulative investments drive energy • Learning curves for energy • Endogenous labor
efficiency technologies (electricity generation) productivity, capital
• Fuel prices drive energy efficiency in deepening
transportation and residential sector
MESSAGE- • Factor substitution in CES production • Carbon-free energy from backstop • Declining costs in
MACRO in MACRO technologies (renewables, carbon extraction, production
scrubbing and sequestration)
• Learning curves for energy technologies
(electricity generation, renewable
hydrogen production)
MIND • R&D investments improve energy • Carbon-free energy from backstop • R&D investments • Technological progress
efficiency technologies (renewables) and CCS in labor productivity in resource extraction
• Factor substitution in CES production • Learning-by-Doing for renewable energy • Learning-by-Doing in
resource extraction
Note: This table provides an overview of the diverse implementations of ETC in this study. Features of ETC were loosely grouped according to their presumed
impact, relating them either to energy intensity reductions or carbon intensity reductions. Naturally, the exact effects of ETC in a complex model cannot be known
ex ante with certainty.
50 Chapter 3 Implication of ITC for Atmospheric Stabilization
els like FEEM-RICE market imperfections or sub-optimal policy tools are not re-are not re- not re-
movable or modifiable. Policy of non-reproducible input factors instruments would
be necessary. In other words, they may take so called no-regret options into account.
In this case, the opportunity costs of climate protection can be lower or sometimes or sometimes sometimes
even negative compared to the baseline, dependent on the design of climate policy.
In AIM/Dynamic-Global, ETC concerns energy efficiency (Masui et al.
2006). In addition to autonomous energy efficiency improvements, investments
in energy conservation capital raise macroeconomic3 energy efficiency in the
manufacturing sector, i.e. ETC affects the energy efficiency parameters in the
production function which increases if the energy conservation capital stock in-
creases faster than the output in the manufacturing sector. AIM/Dynamic-Global
divides the world into six regions and describes regions with nine sectors which
are mostly energy related.
FEEM-RICE (Bosetti et al. 2006) is modeled after Nordhaus’ regionalized
integrated assessment model, RICE 99 (Nordhaus and Boyer 2000). It differentiates
eight world regions and computes the global solution by solving a non-cooperative
Nash game. ETC in FEEM-RICE is represented by an energy technological change
index (ETCI) which is increased through R&D investments as well as by learn-
ing-by-doing in carbon abatement. Its impact is twofold: ETCI affects the partial
substitution coefficients in a Cobb-Douglas production function, shifting income
shares from energy to capital. Secondly, ETCI decreases the macroeconomic carbon
intensity. FEEM-RICE is presented in two parameterizations, FAST and SLOW,
reflecting different assumptions about the speed of technological progress, its effec-
tiveness and the crowding out effects between different types of investments.
ENTICE-BR (Popp 2006) is based on Nordhaus’ DICE model (Nordhaus
and Boyer 2000), hence it does not resolve regions. Among other modifications, Popp
incorporates in his model, an R&D sector with two knowledge stocks. They are built in his model, an R&D sector with two knowledge stocks. They are built, an R&D sector with two knowledge stocks. They are built an R&D sector with two knowledge stocks. They are built
up endogenously by R&D investments, one affecting macroeconomic energy effi-
ciency and the other lowering the price of a generic backstop technologyand the other lowering the price of a generic backstop technology the other lowering the price of a generic backstop technologying the price of a generic backstop technology the price of a generic backstop technology4. Energy is
produced either by this backstop technology, or from fossil fuels in a corresponding
sector. Both ENTICE-BR and FEEM-RICE derive a second-best social optimum by
simulating market behavior in an intertemporal optimization framework.
The model MIND (Edenhofer et al. 2006) is an intertemporal optimiza-
tion model with a macroeconomic sector and four different energy sectors: re-
source extraction, fossil-fuel based energy generation, a renewable energy source,
and carbon-capturing and sequestration (CCS). The growth engine in the macro-
economic sector is fueled by R&D investments in labor productivity and energy
efficiency. There is no autonomous total factor productivity improvement. The
investments in the different energy sectors are determined according to an inter-
temporal optimal investment time path. MIND derives a first-best social optimum
3. Here, we use the term macroeconomic to indicate an effect or process described at the macro
level, e.g. described by one parameter for the economy.
4. Backstop technologies provide carbon-free energy and are not subject to any scarcities.
Induced Technological Change / 65
3.3 Model Classification 51
66 / The Energy Journal
and therefore calculates the potential of ITC for reducing the costs of climate
protection if market failures and social traps at the international level are resolved
by appropriate policy measures.
DEMETER-1CCS models a dynamic economic system which is inter-
temporally optimal for the representative household. The firms solve a per-period
dynamic optimization problem, treating learning effects as external to the pro-
duction decision level (Gerlagh 2006). Moreover, it comprises a composite good
sector and different energy sectors for renewable energy sources (playing the role
of a backstop-technology) and for fossil fuels. In the energy sector the costs are
reduced through learning-by-doing.
3.1.2 Energy System Models
Energy system models usually derive a cost-minimum sequence of en-
ergy technologies for an exogenously given energy demand using linear program-
ming. In more advanced versions, the energy technologies are improved by learn-
ing-by-doing. The main advantages of this approach are the detailed depiction
of the energy sector and the possibility of basing technological change on an en-
gineering assessment of different technologies. Three energy system models are
participating: DNE21+, GET-LFL, and MESSAGE-MACRO.
DNE21+ differentiates eight primary energy sources in 77 world regions
(Sano et al. 2006). Technological change has an endogenous description for wind
power, photovoltaics, and fuel-cell vehicles; exogenous assumptions about tech-
nological change are made for other energy technologies. Energy demand in the
end-use sectors is modeled using long-term price elasticities; gross world product
(GWP) is exogenous to the model.
GET-LFL is a globally aggregated model differentiating eight primary
energy sources (Hedenus et al. 2006). It includes a carbon capturing and sequestra-
tion (CCS) option which is used with different fossil fuels as well as with biomass.
GET-LFL implements cost minimization with limited foresight in a partial equi-
librium (energy market), implying an elastic energy demand. ETC in GET-LFL is
implemented in learning curves for investment costs of carbon-free technologies as
well as energy conversion technologies, and spillovers in technology clusters.
MESSAGE-MACRO. The MESSAGE model describes the entire en-
ergy system from resource extraction, through imports and exports, to conversion,
transportation and end-use (Rao et al. 2006). Learning-by-doing is implemented
for energy technologies. MESSAGE is solved in an iterative process with the
economy model MACRO, allowing for some feedbacks between energy system
and the macroeconomic environment, such as an impact on GWP.
3.1.3 Simulation and Econometric Models
We use the term simulation model to refer to models that start at a given
state of the economy; then continue to calculate the next time step. In mathemati-
52 Chapter 3 Implication of ITC for Atmospheric Stabilization
cal terms, they solve initial value problems or boundary value problems given as
systems of differential equations. Econometric simulation models are additionally
based on time series data, i.e. the equations are estimated from data.
Econometric models are represented by the Tyndall Centre’s E3MG
model (Barker et al. 2006). It is based on a post-Keynesian disequilibrium macro-
economic structure with two sets of econometric equations (describing energy de-
mand and export demand) estimated using Engle-Granger cointegration. E3MG
differentiates 20 world regions modeled with input-output structures, 41 industrial
sectors, 27 consumption categories, twelve fuels, and 19 fuel users.
3.1.4 General Equilibrium Models
General equilibrium models compute demand/supply equilibria in an
economy modeled in distinct, interdependent sectors. Implicitly, households and
firms within these sectors try independently to optimize their welfare and their
profits, respectively. Computable General Equilibrium models (CGE) are promi-
nent examples of this type. CGE models calculate static equilibria at each point in
time prescribing some growth dynamic in between time steps, i.e. they are recur-
sive dynamic. This guarantees not only that all markets are cleared but also that
a Pareto-optimum is achieved. Sectoral resolution and the dynamics of relative
prices are the main strengths of CGE models.
IMACLIM-R is solved recursively but includes an endogenous growth en-
gine that differs from standard CGE approaches (Crassous et al. 2006). The world is
disaggregated into five regions, each made up by ten economic sectors. Cumulative
investments drive both the energy efficiency and the labor efficiency at the same
time. IMACLIM-R represents formation of mobility needs through infrastructures
and technical progress in vehicles. Three transportation sectors (air, sea, and terres-
trial) are differentiated in which energy efficiency is driven by fuel prices. Addition-
ally, energy technologies in electricity generation improve via learning-by-doing.
3.1.5 A Comment on Model Types
Different modeling frameworks were created for different problems,
with each model design tailored to address a specific set of questions. The charac-
teristics of the modeling framework as well as the primary questions that guided
its designs must be kept in mind when comparing the model results. Repetto and
Austin (1997) note that macro and CGE models complement each other in pre-
dicting short-term and long-term responses to a climate policy. Making models
to predict century long economic behavior poses a great challenge in modeling
frameworks that rely on past data or the present structure of the economy. Growth
models using an optimizing framework allow endogenous savings and investment
decisions with unlimited foresight while many recursive dynamic CGE models
restrict optimizing behavior of its agents to a sequence of static equilibria. Hence,
the time path of emissions and investments derived by most CGEs are not inter-
Induced Technological Change / 67
3.3 Model Classification 53
68 / The Energy Journal
temporally cost-effective. This lack of optimality is not a shortcoming of these
models as they try to replicate the outcome of decentralized markets in which
market imperfections are inherent. In contrast to recursive CGE models, an opti-
mal economic growth model allows an understanding of transition paths and an
assessment of what decentralized markets could achieve if appropriate policy in-
struments were applied. On the other hand, most intertemporal economic growth
models lack economic detail and offer only limited insights into sectoral dynam-
ics. Energy system models focus on sectoral dynamics providing very detailed
predictions. When restricted to the energy sector, they neglect feedbacks with
the macroeconomic environment, e.g. the revaluation of capital. The integration
of energy system models with macroeconomic models is a topical subject under
scrutiny and a feature of several models in this study.
Three models, MIND, MESSAGE-MACRO and E3MG, adopt a hybrid
approach, i.e. they combine features from different model designs to address the
gap between them. MIND integrates technological detail similar to energy sys-
tem models in the framework of a growth model. MESSAGE-MACRO adds an
economic environment to an energy system model by iterating the models MES-
SAGE and MACRO. E3MG includes a cost minimizing energy system sector
within a Keynesian econometric model.
Finally, we note on the scope of the models. While all models are well
calibrated, some models make very specific assumptions to explore special sce-
narios. Three models in particular are explorative in character. First, IMACLIM-R
adopts a pessimistic view of technological change by assuming strong inertia and
by neglecting carbon-free energy sources from backstop technologies. Second,
AIM/Dynamic-Global focuses on the investment in energy-saving capital as a
mitigation option, and largely neglects other options. As a consequence, economic
growth cannot be decoupled from emissions. Third, FEEM-RICE is presented in
a FAST version where especially optimistic assumptions are made about learning
and the level of crowding-out.
4. METHODS OF MODEL COMPARISON
The following section outlines the IMCP approach of quantitative model
comparison, specifically which scenarios were run, and which model outputs were
reported. The effects of climate policies may be explored by comparing scenarios
of climate protection with a business-as-usual scenario (baseline). In accordance
with Article 2 of the UNFCCC which postulates stabilizing greenhouse gas con-
centrations, we investigate climate policy scenarios with the goal of stabilized CO2
concentration. We focus on carbon dioxide as the most influential GHG, defining
three policy scenarios stabilizing CO2 concentrations at levels of 450ppm, 500ppm,
and 550ppm, respectively. Where possible we also report results for a stabilization
level of 400ppm. For this stabilization level the probability to meet the 2°C target
is substantially increased (Hare and Meinshausen 2004). The 2°C target is per-
ceived by some scientists and influential politicians, CEOs (like Lord Browne) and
54 Chapter 3 Implication of ITC for Atmospheric Stabilization
governmental bodies (like the EU Commission) as an interpretation of Article 2 of
the UNFCCC. The concentration levels selected are somewhat arbitrary and serve
to explore model responses to increasingly ambitious policies. As we prescribe a
policy goal rather than a policy, model results represent a way of conforming to the
policy goal and may guide the design of actual climate policy measures.
To assess the model response to climate policies and in particular the role
of ITC, scenarios should ideally harmonize all other assumptions and also model
calibration in order to isolate the effects of different implementations of ITC. It is
known that the business-as-usual scenario has strong impact when evaluating the
consequences of climate policies: assuming lower economic growth and therefore
lower CO2 emissions implies that climate protection poses a lesser challenge to the
economy. Where models prescribe gross world product (GWP) and/or emissions ex-
ogenously, data from the Common POLES/IMAGE baseline (CPI) was used (Vuuren
et al. 2003). However, harmonizing economic output and emissions in models which
determine these numbers endogenously proves to be difficult if not impossible. Here,
modeling teams have made an effort to calibrate their models to the CPI baseline, but
there remain differences that must be taken in account when interpreting results.
Carbon dioxide concentration caps could not be imposed in models that
do not include a carbon cycle submodel to translate emissions into concentrations.
Such models either prescribe CO2 emission paths corresponding to the selected
concentration levels exogenously, or constrain the overall centennial carbon bud-
get. Differences in the implementation of carbon cycle models may imply that the
same concentration level requires more stringent emission paths. Care was taken
that the carbon cycle models showed good agreement.
4.1 Scenario Definitions With and Without ITC
To assess the impact of ETC model output, stabilization scenarios were
run with and without induced technological change. The baseline scenarios in IMCP
comprise all components of endogenous technological change potentially incorpo-
rated in the considered model. A policy scenario ‘with’ induced technological change
refers to a scenario in which additional endogenous technological change is induced
by climate policy. In contrast to this, a policy scenario ‘without’ induced techno-
logical change means that climate policy cannot induce endogenous technological
change beyond the baseline scenario. Therefore, in a policy scenario without ITC,
technological change simply follows the time path of the baseline scenario as if it
was given exogenously.5 A comparison between ‘with’ and ‘without’ induced tech-
nological change measures the extent to which climate policy induces technological
change in addition to baseline ETC. Table 3 summarizes these scenario definitions.
5. The time paths of ETC related variables in the baseline simulation are stored and then prescribed
as exogenous, fixed time series in this scenario.
Induced Technological Change / 69
3.4 Method of Model Comparison 55
70 / The Energy Journal
4.2 Model Output and Indicators
The broad range of models is a key asset of this comparison, naturally
comparable model outputs that are available in all models are of an aggregate
nature. More specific outputs might allow deeper insights into some models but
would exclude others. The selected model outputs (e.g. GWP, emissions, incre-
mental costs of carbon, energy use, and the fuel mix) and the derived indicators
(e.g. macroeconomic costs and sector costs, energy- and carbon intensity) reflect
this trade off.
Despite the effort to harmonize assumptions and scenarios among mod-
els, it remains a challenging task to determine why model results differ, i.e. to
disentangle the role of ITC from other assumptions. In addition to the analysis
offered in this paper, modelers were asked to elaborate on the calibration of their
model and its sensitivities in their paper contributions to this special issue, thus
providing a starting point to assess the assumptions underlying the model calibra-
tion and their implications.
4.3 Concepts of Mitigation Costs
The SAR distinguishes four types of mitigation costs (IPCC 1996, p.
269). This taxonomy of costs provides a useful guide for the interpretation of
results and is therefore recapitulated in the following:
1. Direct engineering costs of specific technical measures: These
numbers provide some information about the costs of a mitigation
measure or a specific technology. The cost estimates are mainly
derived from engineering process-based studies of specific
technologies. Examples include the costs of switching from coal to
gas. In this model comparison, they are presupposed in all models.
2. Economic costs for a specific sector are computed in sector-specific
models, which allow the integration of a multitude of mitigation
measures, often in a partial equilibrium framework. For example,
energy system models assess the sectoral costs of the energy sector.6
3. Macroeconomic costs reflect the impact of a given mitigation
strategy on the level of the gross domestic product (GDP) and its
components. At this level of analysis, feedbacks between sectors and
6. Note that MESSAGE-MACRO goes beyond this by linking with the MACRO model.
Table 3. Summary of IMCP Scenario Definitions
The baseline is a business-as-usual scenario. Technological change is determined endogenously.
Policy scenarios with ITC impose a policy goal of CO2 stabilization at three different levels (450,
500, 550ppm CO2) or comparable
Policy scenarios without ITC impose the same policy goal but restrict technological change to the
extent found in the baseline scenario
56 Chapter 3 Implication of ITC for Atmospheric Stabilization
the macroeconomic environment are accounted for. Such “general
equilibrium effects” can be calculated by models which encompass
either the whole economy, or coupled models of specific sectors and
macro-economy. Thus, macroeconomic costs include the effects of
engineering costs and sector-specific costs.
4. Welfare costs: The GDP variations, underlying the assessment
of macroeconomic costs, do not provide an adequate measure of
human welfare because the ultimate goal of economic activities
is not producing GDP but allowing consumption of private and/
or public goods and leisure. Mitigation policies, however, may
increase investments and thus GDP while at the same time reducing
consumption. Therefore, GDP is not a reasonable indicator for
human welfare. However, per capita consumption is also a flawed
indicator for welfare because human welfare is not always a linear
function of per capita consumption. Therefore, most intertemporal
optimization models assume in accordance with some empirical
evidence that the utility index is an increasing function of per capita
consumption, and marginal utility is decreasing with consumption.
This implies that costs measured in per capita consumption are
exaggerated or underestimated depending on the per capita
consumption level. Moreover, the utility index depends also on
the distributional issues and non-market traded goods and bads.
Economists who rely on welfare theory may argue that the utility
index could be modified according to fairness criteria and public
goods. Therefore, this index could be used as a reliable indicator
for human welfare.
Within IMCP, we analyze the impact of mitigation strategies on the sec-
ond and third types of costs. Welfare implications along the lines of item 4 are
not assessed explicitly because the models participating in IMCP do not share a
common measure of welfare.
It seems worthwhile to note that all these cost concepts leave room for
interpretation and may fuel a debate about the explanatory power of mitigation
cost estimations. When GWP losses and consumption losses per capita are report-
ed in absolute numbers, these are naturally large and may create the impression
that mitigation is a costly option. Put into perspective as relative percentage of the
net present value of the GWP in the business-as-usual scenario, mitigation may be
seen as only postponing economic growth for several months. A simple thought
experiment illustrates this point: Assume that GWP growth of 2% per year in the
business-as-usual scenario. If mitigation policy lowered growth to 1.97%, GWP
losses over the whole century discounted by 5 % would amount to 1%. In conse-
quence, the annual GWP that would have been achieved in 2100 is now reached
in 2101 (see Azar and Schneider 2002 for a similar argument). Does this imply
that mitigation costs nearly nothing for humankind? One could argue that with
Induced Technological Change / 71
3.4 Method of Model Comparison 57
72 / The Energy Journal
these trillions of dollars the lives of millions of poor people could be rescued, e.g.
by investing in clean water facilities. On the other hand, damages caused by non-
action may destroy the rural habitats of millions of people elsewhere which also
rarely count in terms of GWP. There is need for further investigation of the extent
to which rapid climate change affects the welfare of people. Whilst acknowledg-
ing that different social outcomes can be hidden behind an aggregated number like
GWP and the limitations of this approach, some useful insights about the impact
of ITC can be drawn using GWP. Clearly, a situation where GWP is increased
because of ITC is preferable to a situation where climate policy reduces the op-
portunities to invest in other desirable global projects.
In the context of IMCP we report GWP losses and consumption losses in
terms of relative net present value which means that we measure the net present
value losses between the business-as-usual scenario and the policy scenario and
relate them to the net present value of GDP in the business-as-usual scenario. This
allows a comparison of the cost estimations of different models.
When interpreting mitigation costs, it is necessary to recall that in the
IMCP we compare mitigation costs at given stabilization levels. Some models,
e.g. ENTICE-BR and FEEM-RICE estimate climate change impacts caused by
specific stabilization levels. Therefore, the benefits of avoiding such impacts are
reflected in the GWP losses in these models. In the IMCP, we inform the reader
only about the mitigation costs of achieving a certain stabilization level irrespec-
tive how much damages can be avoided by the predefined stabilization levels.
In the cases of ENTICE-BR and FEEM-RICE the mitigation costs are reduced
further by the damages caused at the specific stabilization level. Therefore, these
GWP losses can be interpreted as net mitigation costs. In the following section we
discuss the impact of technological change on these mitigation costs.
5. RESULTS AND DISCUSSION
This section presents the collected data as follows: First we outline and
analyze the costs of achieving specific stabilization targets. Second, we analyze
the necessary emission reductions in the different models in terms of their effect
on carbon intensity, energy intensity, and gross world product. Third, the transfor-
mation of the energy system which is a key challenge to meet the climate protec-
tion targets is described and evaluated.
5.1 Mitigation Costs within Different Model Types
In this section we refer simultaneously to two different representations of
mitigation costs. In both representations – Figure 1 and in Figure 2 – we show the
mitigation costs as a loss of gross world product (GWP). Figure 1a shows mitigation
costs from different models relative to the respective baseline GWP in the case when
technological change is switched on (cf. scenario definitions in Table 3). In Figure 1b
the cost estimations are reported when technological change is switched off, Figure
58 Chapter 3 Implication of ITC for Atmospheric Stabilization
1c indicates the additional mitigation costs for the scenarios without technological
change, i.e. the differences between Figure 1a and Figure 1b. Figure 1c shows the
potential to induce technological change in the different models: the larger the cost
increase when ITC is switched off, the lower the potential of endogenous technologi-
cal change incorporated in the implementation in that model. If a models incorpo-
rated no endogenous technological change, Figure 1c would indicate no additional
costs because costs with ITC would be the same as costs without ITC.
In Figure 2 the mitigation costs are shown as a function of the cumula-
tive CO2 reduction. The plotted data points correspond to the 550, 500 and 450
ppm stabilization scenario. The main purpose of Figure 2 is to relate costs to the
mitigation gap which has to be overcome by the different models. In some models
the costs are relatively low because of a small mitigation gap and not because of
a strong impact of ITC on the costs. In all but two models, mitigation costs are
computed as the difference in cumulated GWP (2000 to 2100) between baseline
and policy scenarios, discounted at a rate of 5% and relative to (discounted) base-
line GWP of the same time span.7 As there is no endogenous GWP in DNE21+
and GET-LFL, they present instead energy system costs and producer/consumer
surplus in the energy sector, respectively.8
By plotting the costs at different stabilization levels against the corre-
sponding cumulative CO2 reductions (also 2000 to 2100), the costs are put into
perspective of the mitigation challenge that each model is confronted with in the
policy scenarios.
The severity of the challenge is determined by the ‘mitigation gap‘, i.e.
the difference between predicted business-as-usual emissions and admissible
emissions in the policy scenario. Models tend to agree on the latter, which is a
property of the carbon cycle modules in the models, but advocate various pre-
dictions of business-as-usual GWP growth and CO2 emissions. Consequently, so
called baseline effects have a strong influence on the results. Figure 2a depicts re-
sults from scenarios with ITC; for the scenarios in Figure 2b, ITC was disabled.
With one exception (E3MG), the models agree about the trend of costs:
lower concentration targets imply larger costs. Also, costs rise disproportionately
with CO2 reductions.
In Figure 1a and Figure 2a, two models (E3MG and FEEM-RICE-FAST)
show negative costs, i.e. gains from implementing climate policies. In the case of
E3MG, this originates from the Keynesian treatment of demand-side long-term
7. We use a 5% rate to discount GWP reductions from all models to make numbers comparable
among models and to other studies in the literature. The rates of pure time preference used in models
that anticipate future development vary: ENTICE-BR and FEEM-RICE use a 3% rate initially which
declines over the course of the century; AIM/Dynamic-Global applies a 4% discount rate; the rates
of pure time preference are 3% and 1% in DEMETER-1CCS and MIND, respectively; the energy
system models (DNE21+, GET-LFL, and MESSAGE-MACRO) use a 5% discount rate. There is no
(macroeconomic) discounting in E3MG (except in the electricity sector) and IMACLIM-R.
8. Surplus and energy system costs are converted to the same metric as the GWP losses, i.e. their
difference between baseline and policy scenarios is presented relative to the present value of baseline GWP.
Induced Technological Change / 73
3.5 Results and Discussion 59
74 / The Energy Journal
Figure 1. Mitigation Costs
Figure 1a shows loss of gross world product, except for DNE21+, which reports the increase in en-
ergy system costs relative to the baseline, and GET-LFL, which reports the difference in producer
and consumer surplus. Figure 1b displays the corresponding data from the scenarios without ITC.
Figure 1c shows the difference between Figure 1a and Figure 1b.
(a) Mitigation costs with ITC
(b) Mitigation costs without ITC
c) Difference of mitigation costs with ITC and without ITC
60 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 75
Figure 2. Mitigation Costs as a Function of Cumulative CO2 Reduction
All models report loss of gross world product except the DNE21+ which reports the increase in
energy system costs relative to the baseline, and GET-LFL which reports the difference in producer
and consumer surplus. The plotted data points correspond to the 550, 500, and 450ppm stabilization
scenarios (with increasing CO2 reductions). In case of MESSAGE-MACRO, the presented scenario
is 500ppm stabilization. Not shown for scaling reasons are GWP losses from IMACLIM-R which
range from 2.5-6.2% in scenarios with ITC and 6.8-15.4% in scenarios without ITC.
(a) Mitigation costs with ITC relative to corresponding CO2 reductions
(b) Mitigation costs without ITC relative to corresponding CO2 reductions
3.5 Results and Discussion 61
76 / The Energy Journal
growth that assume increasing returns to production and under-employment of la-
bor resources in the global economy. In E3MG, policy-driven increases in carbon
prices lead to more investment and output. In the case of FEEM-RICE-FAST the
negative costs are the consequence of the optimistic assumptions on the effects of
R&D investments and of the role that stabilization targets have in inducing more
R&D investments. This reduces the inefficiencies in the global R&D market that
are calibrated in their second-best baseline scenario.
We now discuss these results in more detail by model design and by in-
dividual model. We start with cost estimates of energy system models, which are
relatively low, partially due to neglected general equilibrium effects. In a second
part we consider the results of general equilibrium market models and simulation
models which calculated relatively high mitigation costs because they are focused
on price effects and neglect intertemporal investment dynamics. Finally, the opti-
mal growth models within IMCP are discussed.
5.1.1 Energy System Models
Mitigation costs in the energy system models DNE21+, GET-LFL (Fig-
ure 1 and Figure 2) differ from those reported by other models in this exercise,
which measure the loss of GWP (or welfare). The opportunity costs of climate
protection are measured as the increase in energy system costs compared to the
baseline in DNE21+, and measured in terms of producer/consumer surplus rela-
tive to the baseline in the case of GET-LFL. We emphasize that using alternative
metrics in our comparisons is problematic. In fact, while macroeconomic models
are less adept to account for the system engineering costs in the energy sector,
some system engineering models do not report on the aggregated implications
of mitigation for total GWP. Thus, as the energy sector accounts for the partial
equilibrium effects, the mitigation costs appear relatively low in Figure 1 and
Figure 2. MESSAGE-MACRO adopts a hybrid approach, combining a systems
engineering and macroeconomic model, and thus calculates energy system costs
as well as GWP losses. However, it remains open to debate whether all intertem-
poral equilibrium conditions hold in this framework and thus all relevant compo-
nents of macro-economic mitigation costs are taken into account. For the sake of
consistency with the macroeconomic models, Figure 1 and Figure 2 reports loss
in terms of % GWP.
The main advantage of energy system models is their higher resolution
with respect to technology representation, emphasizing internal plausibility and
consistency of structural change in the energy system. They are hence better at ac-
counting for costs related to barriers of technology diffusion and adoption than mac-
roeconomic models, where technology is traditionally represented in a more stylized
and generic way. The downside of using purely systems engineering approaches
is that the reported energy system costs do not provide a comprehensive account
of potential welfare losses outside the energy sector. As discussed above, costs of
DNE21+ and GET-LFL presented in Figure 2 are thus relatively small compared
62 Chapter 3 Implication of ITC for Atmospheric Stabilization
to the majority of the macroeconomic models. The costs of mitigation depicted by
MESSAGE-MACRO are seen to be relatively low as well, but mainly because of the
small CO2 reductions required to meet the 500-ppm stabilization target.
From a methodological point of view, the three systems engineering
frameworks differ in particular with respect to representation of energy demand. In
DNE21+ demand is price inelastic, i.e. feedbacks from changes within and outside
the energy sector are not considered. GET-LFL takes into account price-elastic en-
ergy demand and therefore considers rebound effects in a partial equilibrium of the
energy market. In partial equilibrium models, producer and consumer rents may be
diminished by climate policy. Therefore, consumer and producer surpluses present
a better estimate of the mitigation costs than energy system costs in this model.
Both these estimates of energy system costs are relevant measures of the costs
imposed by climate policy, because the transformation of the energy system is one
of the greatest challenges posed by constraining CO2 emissions. In MESSAGE-
MACRO the price response of energy demand is estimated via its macroeconomic
module (MACRO), where the economy is viewed as a Ramsey-Solow model of
optimal long-term economic growth. In particular, feedbacks between energy and
non-energy sectors are determined by relative prices of the main production factors
capital stock, available labor, and energy inputs, subject to optimization.
Figure 1c compares the mitigation costs from Figure 1a (with ITC) and
Figure 1b (without ITC). It is apparent from the results of DNE21+ and GET-LFL
that ITC effects within the energy system are relatively small compared to those
given by macroeconomic models, which account also for GWP changes outside
the energy sector. Again, this might not come as a surprise because these energy
system models calculate only partial equilibrium effects. Another reason may be
that for the DNE21+ model, learning-by-doing to only selected technologies (wind,
photovoltaic, and fuel cell vehicle). GET-LFL, however extensively incorporates
learning-by-doing. In this case, climate policy does not induce significant progress
for two reasons: floor costs for carbon capturing and sequestration and biomass are
already nearly realized in the baseline scenario mainly because of spillover effects
in technology clusters. Additionally, abundant resources of natural gas help to close
the mitigation gap without further resorting to the carbon-free energy technologies
which lack learning potential in the scenario without ITC. Results of the latter mod-
el in particular illustrates that technological detail is needed to understand possible
compensation mechanisms that might limit inducement effects of climate policies
in the energy sector.
Figure 1 includes the GWP losses from MESSAGE-MACRO (for the
500ppm scenario only). In the scenario without ITC, mitigation costs are much
higher. However, comparability to the results from other models is limited, since
MESSAGE-MACRO ran a fixed cost “without ITC” scenario. In other words, the
structure of the energy system changes towards today’s best practice technologies
(given specific resource and environmental constraints). In contrast, the other models
have defined exogenous technological enhancements in the scenarios without ITC.
The effect of ITC in these and other macroeconomic models are discussed next.
Induced Technological Change / 77
3.5 Results and Discussion 63
78 / The Energy Journal
5.1.2 General Equilibrium Models
CGE models are represented in the IMCP by IMACLIM-R. CGE models
have been known to predict high costs and indeed, IMACLIM-R estimates GWP
losses for 550, 500, and 450ppm stabilization targets at 2.5, 4.6, and 6.2% (Figure
1). As expected, these numbers are the highest cost estimates in this and there are
reasons inherent to the model structure that explain this tendency.
Models like IMACLIM-R calculate a general equilibrium taking into ac-
count the relative price effects not only in the energy sectors but in all sectors. This
way, climate policy not only induces a transformation of the energy system but
also a revaluation of all capital stocks in the energy sectors and in turn in energy
demand sectors. It follows that resources within the economy need to be reallocat-
ed according to the changed equilibrium. Hence in a general equilibrium model,
climate policy has the potential to trigger a greater transformation than that of the
energy system alone. Pitted against the need for change throughout the economy
are potentially larger – economy wide – flexibilities to react to the restrictions of
climate policy. However, recursive dynamic CGE models lack foresight as well as
the flexibility of endogenous, sector specific investment decisions.
In particular, the IMACLIM-R model assumes that investments in the
composite good sector simultaneously enhance labor productivity and energy
productivity, i.e. investments in physical capital exhibit an externality. Addition-
ally, labor productivity is improved by learning-by-doing. Climate policy induc-
es increases and reallocations of investment in the energy sectors including the
corresponding learning-by-doing. Due to learning-by-doing energy prices de-
crease and cause an additional energy demand – a rebound effect. These invest-
ments in the energy and transport sectors crowd out investments in the composite
good sector and reduce economic growth. The reduction of investments in the
composite good sector also lowers the growth rate in labor productivity, which
reduces economic growth further. The double dividend of increasing investments
becomes a double burden if investments have to shrink. Among other things, the
crowding out effect and this double burden increase the opportunity costs of cli-
mate protection – an effect which is very pronounced in IMACLIM-R. Moreover,
the interplay between inertia in the transport sector, imperfect foresight and non-
optimal carbon tax profile induced further welfare losses. These welfare losses
can be considerably lowered by efficiency gains and technology diffusion.
Without induced technological change, costs increase further in IMA-
CLIM-R, demonstrating that the implementations of ETC endow the models with
additional flexibility (Figure 1c). In IMACLIM-R, mitigation costs for the 550,
500, and 450ppm scenarios climb to 6.8, 12.0, and 15.4%, respectively.
5.1.3 Simulation Models
In E3MG, CO2 permits and taxes are imposed on the economy in order to
achieve the required stabilization targets. In contrast to other long-term studies but
64 Chapter 3 Implication of ITC for Atmospheric Stabilization
consistent with many shorter-term studies (e.g. IPCC 2001, p. 516), climate policy in-
duces GWP gains. This result can be understood in comparison with the second-best
solutions of optimizing models. These try to reproduce the market behavior which in
general exhibits all sorts of market imperfections – like unemployment, postponed
price adjustments, etc. – by relaxing assumptions about perfect market clearing. A
crucial feature in E3MG is that although product markets clear, labor and other mar-
kets may not clear. Part of the effect of including ITC in the model is to raise growth
by more labor transfer from traditional to modern sectors in the world economy.
This effect of taxation in E3MG is due to the fact that investors are limited
in their foresight. In a perfect foresight model we would expect that investors adjust
their portfolio of investment according to long-term price and taxation expectations.
5.1.4 Optimal Growth Models
Four of the models in the IMCP are implemented in the framework of
growth models subject to intertemporal welfare maximization (MIND, ENTICE-
BR, AIM/Dynamic-Global, DEMETER-1CCS, and FEEM-RICE, the latter in
FAST and SLOW parameterizations). The large differences in CO2 reductions
necessary for stabilization between these models are caused by different baseline
projections of GWP and the corresponding emissions. These different projections
are a direct result of implementing ETC within these economy models. Whereas
optimal growth models without ETC make an assumption about GWP growth,
these models make assumptions about ETC which then contribute to overall GWP
growth. This makes GWP growth a result of how ETC is modeled rather than an
assumption. In most optimal growth models in the IMCP overall technological
change is determined by an exogenous total factor productivity in addition to an
implementation of ETC. MIND differs in this respect, describing technological
change fully endogenously. All models share a common starting point in 2000.
However, large differences result over the course of the century.
With the exception of AIM/Dynamic-Global, the cost predictions of the
growth models in Figure 2 are low (below 1% GWP up to the 450ppm scenario).
We have argued above that general equilibrium effects tend to raise the opportu-
nity costs of climate policy, but these models are endowed with perfect foresight.
In conjunction with endogenous investment possibilities this allows models to act
flexibly thus avoiding large mitigation costs.
AIM/Dynamic-Global incorporates perfect foresight but studies only a
single endogenous mitigation option. Energy efficiency depends on a stock of
energy conservation capital. Investment in energy conservation capital improves
energy efficiency and is a decision variable of the optimization. AIM/Dynamic-
Global also includes carbon-free energy from renewables and nuclear power, but
investments in these options cannot be induced by climate policy – only invest-
ments in energy conservation are a control variable. This demonstrates the impact
of flexibility on mitigation costs and how the exclusion of mitigation options in-
creases the costs substantially.
Induced Technological Change / 79
3.5 Results and Discussion 65
80 / The Energy Journal
In contrast, MIND includes investment decisions into capital stocks of
energy technologies, including the backstop technology in particular. We attribute
the low cost estimates of these models to this flexibility.
ENTICE-BR and FEEM-RICE-SLOW compute slightly higher costs
compared to MIND. ENTICE-BR incorporates a backstop technology which im-
proves through R&D investments. However, this effect is overcompensated by the
built-in crowding out effects caused by investments in the energy sector. In addi-
tion, the backstop technology displays most of its effects in the baseline scenario,
independent of stabilization targets. In FEEM-RICE-SLOW costs are low because
of the combined effect of learning-by-doing and R&D investments. An increase in
R&D investments induced by a stabilization target enhances learning-by-doing as
well. This makes R&D investments more profitable by oncreasing benefits from
climate change reductions. ENTICE-BR and FEEM-RICE GWP numbers include
benefits of climate policy, and that the gross numbers would be slightly higher.
In FEEM-RICE-FAST, there are negative mitigation costs, i.e. gains from
mitigating carbon. The FEEM-RICE model is a second-best model in the sense that
market imperfections occur in the baseline due to externalities in the R&D invest-
ments. Regions invest too little in R&D because of their non-cooperative behavior.
If faced with climate policy, they are induced to increase their R&D investments,
which get closer to cooperative levels. That is, an improvement of R&D investment
is a by-product of climate policy. Therefore, climate policy has a clear net benefit.
However, this net benefit changes to net costs if the learning-rate is slow and the
crowding out effect between different types of investments is large.
The DEMETER-1CCS model also computes a second-best solution
of the world economy accounting for independent actions of firms and house-
holds. DEMETER-1CCS’s cost estimates are among the lowest in this study, for
a number of reasons. In DEMETER-1CCS households are endowed with perfect
foresight, hence even though firms show a static profit maximizing behavior, the
model is at an advantage in averting mitigation costs. Moreover, the model makes
optimistic assumptions about substitution possibilities between fossil fuels and
carbon-free energy, and backstop technologies. The latter are assumed to exhibit
high learning rates (20% for renewables and 10% in case of CCS), and the share
of energy from these sources is not restricted, e.g. there is no sharp increase in
costs when the energy supply has to rise as it does in many energy system models.
Moreover, CO2 emissions are low in the baseline scenario, so that complying with
policy scenarios poses a smaller challenge than in other models.
If technological change is switched off (Figure 2b), costs increase. The
comparison of Figure 1a and Figure 1b in Figure 1c shows that the cost reduction
potential of ITC varies between different models: In FEEM-RICE-FAST as well
as in FEEM-RICE-SLOW, ITC shows a large potential for reducing the mitiga-
tion costs when low stabilization scenarios should be achieved. Both versions of
FEEM-RICE show remarkably similar behavior without ITC, in particular, GWP
gains in FEEM-RICE-FAST have turned into losses, hence the observed effect
can be attributed to “fast” technological change.
66 Chapter 3 Implication of ITC for Atmospheric Stabilization
In AIM/Dynamic-Global disabling energy conservation investments has
some influence on mitigation costs. The option of energy conservation invest-
ments is shown to have significant influence, but in comparison with options in
other models, this option is less important.
In MIND, mitigation costs increase sharply when ITC is switched off.
MIND demonstrates that removing backstop technologies when switching ITC
off has a significant impact.9 In scenarios without ITC, the MIND model exhibits
mitigation costs comparable to costs in CGE models.
In ENTICE-BR the net effect of ITC is small because of two effects:
first, investments in the energy sector are less productive than investments in the
rest of the economy. Therefore, less technological progress is induced in the poli-
cy scenario. Second, the exogenously determined total factor productivity further
reduces the impact of endogenous technological change on the model output.
5.1.5 Stricter Climate Policy (400ppm Stabilization)
Table 4 shows that a few models achieve a feasible solution when faced
with a stabilization target of 400ppm (DEMETER-1CCS, MIND, FEEM-RICE,
and GET-LFL). In general, the reason why many models cannot derive a feasible
solution can be found in the inflexibility of the energy system to manage the re-
quired cumulative emission reductions. The inflexibility comprises phenomena
like boundaries for the diffusion of backstop technologies, limited sets of mitiga-
tion options or myopic investment behavior.
5.1.6 Robust cost estimate
The IMCP set out not only to learn from the differences in model results,
but also to identify robust findings. Is it possible to identify a robust estimate of
9. In MIND, the availability of renewable energy sources and carbon capturing and sequestration
is considered an option of ETC because its use depends on the costs of carbon, consequently, in the
scenarios without ITC, the extent of renewables and CCS is restricted to the baseline. In all other
models, the availability of technologies is not considered as “ETC”, e.g. in DEMETER-1CCS’s
scenarios without ITC, renewables and CCS may be used; however there is no learning-by-doing for
these technologies in this scenario. Therefore, if endogenous technological change is switched off,
MIND can only reduce energy consumption and GWP.
Induced Technological Change / 81
Table 4. Mitigation Costs for 400ppm Stabilization
Mitigation costs [%GWP]
Model Name With ITC Without ITC
DEMETER-1CCS 0.07 0.17
FEEM-RICE-FAST 0.01 3.1
FEEM-RICE-SLOW 2.0 3.7
MIND 0.76 8.9
GET-LFL 0.62 0.67
3.5 Results and Discussion 67
82 / The Energy Journal
climate protection costs across models in the IMCP?
One might be hesitant to see robustness in the broad range of costs e.g. in
the case of 450ppm stabilization, ranging from benefits to costs greater than 6% of
aggregate GWP 2000-2100 (at present value). However, the range is reduced con-
siderably when we recognize that three models are of a predominantly exploratory
nature, i.e. their intent is not to give a best estimate but to explore an extreme sce-
nario. These are: IMACLIM-R, which explores the role of the transportation sector
under the assumption that energy sector and transportation sector are inflexible and
externalities of investments in physical capital are biased against energy efficiency;
AIM/Dynamic-Global limiting mitigation options to investments in energy conser-
vation capital, hence emissions cannot be decoupled from economic growth in the
long-run (these two models arrive at the highest costs in this study); FEEM-RICE-
FAST exploring the possibility of “fast” technological change, which then results
in benefits of climate protection rather than climate protection costs.
If we furthermore consider E3MG separately, because it is fundamentally
different with its Keynesian rather than neoclassical point of view, we are thus
left with a set of seven models and cost estimates that range from 0.04% to 0.66%
for 450ppm stabilization. Average climate protection costs among these remaining
models are 0.39, 0.16, and 0.1%, for 450ppm, 500ppm, and 550ppm stabilization,
respectively. Here, the MESSAGE-MACRO model is only included in the 500ppm
average because it did not run the other scenarios. If we exclude the two energy
system models that do not report costs in terms of GWP, the numbers only slightly
change to 0.41, 0.16, and 0.1 percent, for 450ppm, 500ppm, and 550ppm stabiliza-
tion, respectively. These last numbers average over 4, 5, and 4 models, respectively.
Table 5 summarizes these values along with average costs at alternative discount
rates, illustrating the influence of the discount rate on the cost estimate.
In view of this and with the considerable uncertainties about model
structure and other assumptions in mind, it seems a robust conclusion from the
presented energy system models and optimal growth models to expect climate
protection costs of up to one percent.
5.2 Mitigation Strategies for Different Stabilization Scenarios
In this section we identify the contributions of different carbon mitiga-
tion options towards achieving an overall mitigation target, and we assess the role
of technological change in the mitigation effort. Kaya’s identity10 provides a set of
indicators that pinpoint the different ways taken by models to meet a given target,
namely the attribution of total carbon dioxide emissions to global economic out-
put, energy intensity of GWP, and carbon intensity of the energy:
CO2 PE
CO2 = —— × —— × GWP (1)
10. Kaya’s identity originally also differentiates between income effect (GWP per capita) and a
population effect. As an exogenous population scenario is used in this study, we can neglect this factor.
68 Chapter 3 Implication of ITC for Atmospheric Stabilization
PE GWP
Here, CO2 denotes emissions, PE primary energy, and GWP is gross world
product. To facilitate interpretation and to help track down the features underlying
these aggregate effects in the models, we summarize endogenous and exogenous
technological change in the individual models in Table 2 and attribute the features
of technological change to their likely effects in terms of either energy intensity
or carbon intensity. Of course, the complex nature of the models does not allow a
definite classification. Still, these preliminary classifications may serve to structure
features of technological change and guide interpretation, for comprehensive
model descriptions we refer to the literature references in Section 3.
5.3 Decomposition Analysis
The indicators output, energy intensity and carbon intensity are chosen
because they provide information about fundamental differences in the mitigation
strategies pursued by the individual models. Yet because of their highly aggregate
nature, they abstract from the technological and implementational details in the
models, thus allowing quantitative comparison across models.
Reduction of carbon intensity makes it possible to maintain a high level
of energy use, putting relatively little stress on the economy as a whole (the climate
issue is ‘solved’ in the energy sector). If this solution is not feasible (this depends
largely on availability of carbon-free technologies), energy intensity must be de-
creased (implying a reduction of energy) to comply with the climate policy. Forcing
the economy to use drastically less energy can amount to ‘choking’ it, i.e. it may
lead to a reduction in output (gross world product). The decomposition analysis
allows quantification of the contribution of carbon intensity, energy intensity and
output reduction to the required effort of emission reduction. For the purpose of this
modeling comparison we use the refined Laspeyres index method (Sun 1998, Sun
Induced Technological Change / 83
Table 5. Average Discounted Abatement Costs
Concentration Declining
level 5% discount ratea 2% 1% undiscounted
[ppm CO2] [%GWP] [%GWP] [%GWP] [%GWP] [%GWP]
450 ppm 0.41 0.64 0.71 0.83 0.95
500 ppm 0.16 0.25 0.28 0.32 0.37
550 ppm 0.10 0.14 0.16 0.18 0.19
a. Declining discounting rates were adopted from the Green Book (HM Treasury 2003) starting at
3.5% for the first 30 years, then dropping to 3.0% until year 75, and 2.0 until year 125.
Table 5 shows abatement costs averaged over central models, i.e. we exclude models with a
predominant explorative nature and we restrict the average to GWP losses only ignoring the different
metrics from GET-LFL and DNE21+. That is, the above averages include ENTICE-BR. FEEM-
RICE-SLOW, DEMETER-1CCS, MIND, and MESSAGE.
3.5 Results and Discussion 69
84 / The Energy Journal
and Ang 2000). We apply the decomposition analysis to the differences of cumula-
tive values between baseline and policy scenario. Figure 3 displays the decomposi-
tion of the centennial CO2 reductions along Kaya’s identity for different models.
5.3.1 Mitigation Strategies to Comply with 550ppm Stabilization
The stacked bars in Figure 3 show the CO2 savings in the 550ppm policy
scenario from the baseline cumulated over the century. Additionally, shading
indicate how much reductions in carbon intensity, energy intensity, and output
(GWP) contribute to these savings.
The necessary carbon dioxide reductions differ widely between models.
The cumulative reductions necessary to comply with a 550ppm concentration cap
range from ~116GtC to ~987GtC (in FEEM-RICE and MIND, respectively), with
correspondingly great differences in the challenge that these reduction pose for
an economy.11 We stress that models tend to agree on the maximum cumulative
CO2 emissions for a given stabilization scenario: averages among models for
cumulative CO2 emissions are 589, 783, and 931 GtC for 450, 500, 550 ppm
stabilization scenarios, respectively. The corresponding standard deviations are
72, 77, and 92 GtC. The differences in Figure 3 stem mainly from different CO2
emission paths in the baseline: cumulative CO2 emissions in the baseline range
from 980 to 2000 GtC, mean 1430, with a standard deviation of 323 GtC. To
account for such baseline effects, we will base our analyses on measures that are
relative to this ‘mitigation effort’ as much as possible.
Note that baseline growth and CO2 emissions seem unrelated to model
types. This is not very surprising when growth and emissions are exogenous and
therefore arbitrary. In other models, it is possible to calibrate growth and emissions,
e.g. in recursive CGE models, by a variation of exogenous model parameters like
the total factor productivity. In the optimal growth models, total factor productiv-
ity, efficiency of R&D investments, and elasticity of substitution can be adjusted to
approximate a given baseline scenario. However, the baseline is not determined by
exogenous parameters alone but also by the endogenous features of technological
change. This implies that CO2 emissions of such models cannot be fully harmonized.
Nevertheless, there is no reason to assume that models with endogenous technologi-
cal change exhibit an inherent trend to particularly high or low emission scenarios.
A group of models (IMACLIM-R and AIM/Dynamic-Global) share
similar behavior. Here, the larger part of the CO2 reductions can be attributed to
lowered energy intensity and cut-backs in production. They also show the largest
cut-backs in production of all models. A possible explanation is that an inability
to provide enough carbon-free energy (which would show up as carbon intensity
reduction) forces economies to reduce the energy input (evident in the reduced
energy intensity) to an extent where it harms the economy (visible as GWP reduc-
11. An obvious corollary is that emission reductions are necessary to meet even the 550ppm
policy goal despite the presence of ETC in the baseline.
70 Chapter 3 Implication of ITC for Atmospheric Stabilization
tions). IMACLIM-R resorts to decreasing energy intensity and reducing GWP be-
cause it does not incorporate a backstop technology. Here, the increasing energy
price reduces energy demand and induces additional investments in the electric-
ity- and transport sectors which crowd out the overall investments in the com-
posite good sector which are needed to induce economic growth. An optimum,
cost-effective tax profile would probably lower costs compared to the exogenous
linearly increasing tax imposed in these scenarios.
The RICE/DICE models, FEEM-RICE and ENTICE-BR, show strikingly
similar behavior but this differs substantially from the remaining growth models.
Here, the predominant mitigation strategy is to increase the energy efficiency.
FEEM-RICE does allow explicitly for carbon intensity reduction as well as for
energy intensity reduction. However, both are driven by the same index of techno-
logical change. Hence the ratio of reductions in carbon- and energy intensities is
implied by model structure and calibration, and it is not a degree of freedom in the
model. Both FAST and SLOW versions of the FEEM-RICE rely more on energy
intensity reduction than on carbon intensity reduction. The FAST version shifts
the mitigation strategy towards carbon intensity reductions. ENTICE-BR explicitly
includes a backstop technology so one might expect a bigger carbon intensity ef-
fect. However, carbon-free energy is already strongly represented in the baseline
(the share of renewables rises from 4% in 2000 to 11% in 2100). The required CO2
abatement is therefore small and can be met by energy efficiency improvements via
R&D investment in a corresponding knowledge stock and factor substitution.
DEMETER-1CCS behaves differently. Here, energy intensity reductions
and carbon intensity reductions make equally large contributions, while produc-
Induced Technological Change / 85
Figure 3. Cumulative CO2 Reduction for the 550ppm Stabilization Scenario
CO2 reductions are attributed to reductions in carbon intensity, energy intensity, and gross world
product using decomposition analysis. Note that the 550ppm scenarios are not available from
MESSAGE-MACRO and we therefore display results from their 500ppm scenario using a separate
scale on the second y-axis.
3.5 Results and Discussion 71
86 / The Energy Journal
tion cut-backs are kept at a minimum. A low emissions baseline and optimistic
assumptions about substitution possibilities and carbon-free energy sources play a
key part in this and were discussed in detail in the preceding section.
In energy system models, the mitigation strategy relies heavily on carbon
intensity reduction, i.e. CO2 emissions are mitigated largely by switching to low car-
bon energy sources. Indeed, all these models include options to build up a backstop
technology providing carbon-free energy, and in each case learning curves are imple-
mented for some backstop technologies. At the same time, a significant share of the
CO2 reductions is attributed to reductions in energy intensity implying some sort of
energy conservation. In DNE21+, energy demand is exogenously given. However,
energy savings in end-use sectors in climate policy scenarios are modeled using long-
term price elasticities. GET-LFL implements learning-by-doing in energy conversion
technologies as well as a price dependent energy demand in a partial equilibrium. In
MESSAGE-MACRO runs, energy demand is determined in the MACRO economy
model, which allows energy to be substituted by other factors.
Remembering that MIND includes a reduced form energy sector that
borrows from bottom-up energy system models, the similar ratios of carbon and
energy intensity in MIND and in the energy system models is no surprise. Rather,
it indicates that energy system dynamics are successfully approximated by the re-
duced form model. Furthermore, MIND consistently describes the macroeconomic
environment taking into account general equilibrium effects. Hybrid models like
MIND therefore constitute an attempt to bridge the gap between top-down and bot-
tom-up models in order to assess the importance of the investment dynamics.
In E3MG most of the necessary reductions are attributed to reduced en-
ergy intensity. There are three routes by which carbon intensity and energy in-
tensity are affected: First, an increasing price of carbon induces a reduction in
energy demand, and second, a switch to carbon-free technologies within the power
and transport sectors. Finally, the share of fossil fuels in the overall energy mix is
slightly decreased because the elasticity of substitution in the energy and transport
sector is very low.
5.3.2 Effects of Enhanced Climate Policies
Figure 4 indicates the change of the portfolio of mitigation options, if
instead of 550ppm CO2 concentration, the more ambitious level of 450ppm has
to be achieved. How and in which way do the mitigation strategies change when
a more demanding climate protection goal is pursued? Bars in Figure 4 give the
change of the mitigation portfolio in terms of the contributions to overall CO2
reduction in Figure 3. They are symmetrical because an increased share of one
option is always balanced by a corresponding decrease in one or more other op-
tions. For example, a 20% increase of the carbon intensity effect accompanied by
the corresponding 20% decrease of the energy intensity effect in the case of DE-
METER-1CCS implies that the contribution of carbon intensity rises from 50% to
70% whereas the contribution of energy intensity drops to 30%.
72 Chapter 3 Implication of ITC for Atmospheric Stabilization
Figure 4 shows that lowering the stabilization level has different impacts
on the portfolio of mitigation options in the models. Whilst several models show
little change (e.g. MIND and E3MG), others show substantial changes. Large
changes may indicate that favorable mitigation options which contribute to CO2
abatement in laxer policy scenarios have been exhausted hence other options are
increasingly deployed for more stringent climate policies. Small changes suggest
that the greater challenge is addressed much the same way as the lesser challenge.
In DEMETER-1CCS, the contribution of carbon intensity reduction in-
creases by nearly 20% to a share of 70%. In other words, carbon free energy from
renewables and CCS now contribute to mitigation to a similar extent as they do in
energy system models. The reason lies in the fact that the 550ppm scenario in DE-
METER-1CCS is relatively close to the baseline, and a large share of the neces-
sary emission reductions can be accomplished by energy savings. In contrast, the
450ppm concentration target requires a much more substantial departure from the
baseline, and the option of factor substitution decreases in relative importance.
In many models (ENTICE-BR, AIM/Dynamic-Global, DEMETER-
1CCS, MIND, DNE21+, GET-LFL, E3MG) we observe a similar pattern of change
in the portfolio: to achieve 450ppm stabilization, a mitigation strategy is chosen that
incorporates a larger share of carbon intensity reduction than in case of the 550ppm
stabilization. In all of these cases, a carbon-free technology is implemented, and
this change can be attributed to a heavier use of carbon-free energy in the energy
mix. Exceptions to this pattern are FEEM-RICE and IMACLIM-R. FEEM-RICE
and IMACLIM-R have in common, the feature that they do not model a carbon-
Induced Technological Change / 87
Figure 4. Change of the Mitigation Strategy with More Ambitious
Climate Policy
The bars in this figure give the absolute differences between the percentages describing the
contributions of the options in the 550ppm and the 450ppm scenarios. There is no result for
MESSAGE-MACRO because only the 500ppm scenario was available.
3.5 Results and Discussion 73
88 / The Energy Journal
free energy technology. This seems to limit their potential to reduce carbon inten-
sity compared to models with a backstop technology. The difference is particularly
striking when FEEM-RICE is compared to ENTICE-BR. The two models share
the general model structure of Nordhaus’ DICE/RICE models, yet only the latter
incorporates a backstop technology with the consequence that it becomes possible
to increase the contribution of the carbon intensity effect.
In IMACLIM-R, most of the additional CO2 reductions are accomplished
by reducing GWP. The limited potential of carbon- and energy intensity reduc-
tion is largely exhausted at the 550ppm stabilization concentration. The reduction
potentials are limited due to capital inertia preventing the retirement of old capital.
As before in the 550ppm scenario, a rebound effect in the transportation sector
and crowding out of growth inducing investments in composite goods determine
the GWP losses.
5.3.3 Mitigation Strategies With and Without ITC
Figure 5 shows how the portfolio of mitigation options changes when fea-
tures of endogenous technological change are disabled, i.e. technological change
is restricted to the extent computed in the baseline. The bars give the change in
portfolio (cf. Figure 4). Large changes indicate that including the possibility for
ITC has a big impact on the mitigation strategy.
MIND, FEEM-RICE, and IMACLIM-R show relatively large changes.
In MIND, the modelers’ understanding of ITC plays an important part (see Foot-
note 9).12 When the common definition of ITC is applied, changes in MIND are
closest to the changes in DEMETER-1CCS, i.e. there are much smaller changes.
Four models show little change (AIM/Dynamic-Global, DNE21+, GET-LFL, and
ENTICE-BR) because model behavior with and without ITC is very similar.
In Figure 5, ENTICE-BR, FEEM-RICE, DEMETER-1CCS, and MIND
share the same sign for the change in the contribution of carbon intensity re-
duction. In these models, the carbon intensity effect decreases implying that the
induced technological change works more towards decarbonization rather than
reducing energy intensity. Naturally, this mirrors the fact that these models imple-
ment features of endogenous technological change that are related to decarbon-
ization, e.g. learning curves for backstop technologies. Two qualifications apply:
MIND also includes endogenous energy efficiency reduction. In this case, Figure
5 shows that induced carbon intensity reductions outweigh induced energy in-
tensity reductions. Secondly, in FEEM-RICE-SLOW the contribution of carbon
intensity decreases from an 11% contribution to -23% contribution. Here, the av-
erage global carbon intensity is higher in the policy scenario without ITC than in
the baseline because under climate policy, a larger share of global energy use is al-
12. A small carbon intensity effect remains, because the fixed amount of renewables represents
a greater share of the (reduced) total energy in the policy scenario without ITC than in the baseline,
which implies reduced carbon intensity for the energy mix.
74 Chapter 3 Implication of ITC for Atmospheric Stabilization
located to countries with relatively high carbon intensity (U.S., Europe, and other
high income countries), thus raising the global average relative to the baseline.
Conversely, in E3MG, MESSAGE-MACRO, and IMACLIM-R, the cli-
mate policy induces a larger contribution of energy intensity reduction, though
for differing reasons. In IMACLIM-R, stabilization levels without technological
change can only be achieved with a substantial reduction of GWP because of the
sunk costs in the energy system, the constant rate of exogenous technical change
and the absence of sequestration options. The carbon tax induces no additional
change in the pace of technological change. The economy only adapts to the im-
posed carbon tax through a changed energy mix (see the increasing carbon inten-
sity in Figure 5 if technological change is switched off). Therefore GWP has to be
reduced in order to compensate decreasing energy intensity.
In E3MG the key feature of the model underpinning the ITC results is
that GWP growth has been made endogenous, with technological change hav-
ing a major influence (via export equations). However, endogenous technological
change only has a small decarbonization effect on the global economy. Energy
demand and supply is very small in relation to the rest of the economy, around
3-4% of value added, and technological change is led by improvements in the
use of machinery and information technology and communications. These im-
provements allow long-term growth to proceed by decreasing energy-intensity
if technological change is switched on. The growth itself ultimately comes from
the demand by consumers for goods and services, promoted by technological and
marketing innovations.
Induced Technological Change / 89
Figure 5. Change in Mitigation Strategies when ITC is Disabled in the
550ppm Scenario
The bars in this figure give the absolute differences between the percentages describing the
contributions of the options in the scenarios with ITC and without ITC. For message-macro, the
500ppm scenario is used instead.
3.5 Results and Discussion 75
90 / The Energy Journal
Disabling ITC possibilities increases the contribution of GWP reduction
to mitigation in all cases. This comes as no surprise: Removing the flexibility of
inducing further technological change from the model makes it more difficult for
the models to reduce CO2 emissions without cutbacks in production.
5.4 Timing of Mitigation Options
Figure 6 depicts the timing of the mitigation options (adopted from Gerlagh
2006). We show the reduced carbon intensity in the 450ppm policy scenario relative
to the baseline versus the reduced energy intensity as a time trajectory, from 2000 un-
til 2100 with bullets set every 20 years. A trajectory where both options contributed
to the same extent would run along the bisector. Steeper or gentler slopes indicate a
preference for carbon intensity reduction or energy intensity reduction, respectively.
Interestingly, in a majority of models, the trajectory bends to the left with
time indicating that carbon intensity reduction becomes increasingly more impor-
tant. A plausible explanation is the widespread use of carbon-free technologies
that need to be built up gradually by investments, and often become increasingly
more productive through learning-by-doing. The trajectory of IMACLIM-R il-
lustrates well, how lack of a backstop technology prevents this change in the miti-
gation strategy: the model sticks to its mainly energy saving strategy over time.
FEEM-RICE-SLOW shows similar behavior: the reduction of energy intensity
dominates the reduction of carbon intensity (i.e. the slope of the trajectory is less
than unity) because of a missing backstop technology.
Similar to the other models, FEEM-RICE initially increases the reduction
of both energy intensity and carbon intensity. While FEEM-RICE-SLOW retains this
mitigation strategy, FEEM-RICE-FAST decreases reductions of carbon intensity. As
mentioned before, carbon intensity and the elasticity of substitution are driven by the
same endogenous index of technological change in FEEM-RICE, and the relation of
carbon intensity and energy intensity is therefore determined by model structure.
In GET-LFL energy demand is reduced by an increasing energy price,
which in latter periods is compensated by a stronger reduction of carbon intensity.
5.5 Energy Mix
In the previous section, we showed that the dynamics in the energy sec-
tor, e.g. the development of a carbon-free technology, have a key impact on carbon
abatement. In this section we take a close look at the projected development of the
energy system and the role of ITC.
Figure 7 shows the development of the energy system characterized by
the mix of energy sources at the beginning (2000), middle (2050) and end of
the century (2100). Five energy sources are distinguished, namely three fossil
energy sources (coal, gas, and oil) plus renewable energy sources, and nuclear
fission. If additional energy sources were implemented in a model which could
not be subsumed in these categories, or if a model does not differentiate between
76 Chapter 3 Implication of ITC for Atmospheric Stabilization
the categories, the data is presented in the categories of “aggregate fossil” and
“aggregate non-fossil” energy sources. Results are reported in three columns per
model giving the baseline energy mix, the 450ppm policy scenario with ITC, and
Induced Technological Change / 91
Figure 6. Trajectories in Energy Intensity/Carbon Intensity Space
Trajectories start at the origin and bullets are set 20 years apart. Figure 6a shows the 450ppm
scenario with ITC, Figure 6b the same scenario without ITC.
(a) Strategy trajectory with ITC
(b) Strategy trajectory without ITC
3.5 Results and Discussion 77
92 / The Energy Journal
the 450ppm scenario without ITC.13 In 2000, the three cases coincide. The models
FEEM-RICE and ENTICE-BR are not shown as these models do not compute
energy in Joules but incorporate “carbon services” to productions measured in
carbon instead. In the case of MESSAGE-MACRO, results from the 500ppm sce-
narios are displayed instead of the unavailable 450ppm scenarios.
5.5.1 Different Formulations of the Backstop
We have seen that implementing a backstop technology can make a great
difference in how models respond to climate policy goals. In accordance with the
literature, we define a backstop technology as a carbon-free technology whose
usage is not restricted by scarcity of non-reproducible production factors. What
makes backstop technologies so important in carbon abatement?
In Figure 8, we sketch model behavior given two different assumptions
about backstop technology. The price of energy from a fossil resource is indicated
in black, and an exogenously set price for energy from the backstop technology is
indicated in light gray. In contrast, the price of energy from a backstop technology
is plotted in dark gray for an endogenously determined backstop price. Solid time
paths indicate business as usual, and slashed curves are induced by a policy goal.
We assume that imposing a policy goal brings down the price of energy from the
backstop technology because larger investments in carbon-free energy sources
need to be made and therefore more learning occurs. The price of energy from
fossil resources rises due to the costs of the corresponding emissions, e.g. through
carbon taxes or emission permits.
Under climate policy, the price of non-backstop-technologies (like ex-
haustible resources) is rising sharply and intersecting the exogenous backstop
price, at which point the latter becomes economical and is used to an extent that
keeps the energy price at this same level (intersection 1).
For the backstop technology that is explicitly modeled, i.e. capacity is
being build up, and its price changes according to a learning curve, the backstop
technology is competitive much earlier and at a lower price (intersection 2). The
price of carbon-free energy declines from the beginning, indicating that invest-
ments are being made in anticipation of the later competitiveness. Intersection 3
illustrates that this may even be the case in the absence of a policy goal.
From these illustrations we conclude that the cost-decreasing potential of
backstop technologies is strengthened when lowering prices endogenously is an
option in the model, furthermore, if economic agents possess the foresight and the
possibilities to make early investments in order to use this option.
There are models in IMCP without a backstop technology (IMACLIM-R
and FEEM-RICE). As we have seen, these models mainly reduce energy intensity
13. Alternatively, the laxer scenarios could have been used to arrive at much the same conclusions.
We decided on the most stringent case because here the observed effects are more pronounced. The
alternative figures were omitted due to limited space.
78 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 93
Figure 7. Energy System Represented by the Contributions of Different
Energy Sources to the Overall Primary Energy Consumption
In 2050 and 2100, the three bars per model display the energy mix in the baseline scenario, 450ppm
policy scenario, and 450ppm policy scenario without ITC. In 2000, these three cases coincide. We use
darker shading for energy from fossil fuels and lighter shading for carbon free energy sources. Data
from the 500ppm scenario is shown in case of MESSAGE-MACRO. Also in case of this model, the
third bar represents a fixed costs scenario and not the usual scenario “without ITC.”
(a) Energy mix in 2000
(b) Energy mix in 2050
(c) Energy mix in 2100
3.5 Results and Discussion 79
94 / The Energy Journal
to achieve climate protection goals.
Those models that incorporate carbon-free energy from backstop tech-
nologies (i.e. rather than prescribing an exogenous price, the backstop technology
is endogenous to these model) are of the second type discussed above (ENTICE-
BR, AIM/Dynamic-Global, DEMETER-1CCS, MIND, GET-LFL, DNE21+,
MESSAGE-MACRO, and E3MG).
It is also interesting that especially in GET-LFL the investments in the
backstop technology are undertaken long before the break-even-point is achieved.
The reason is that intertemporal optimum decision-making anticipates the tem-
poral spillover effects (learning-by-doing or accumulation of knowledge through
R&D). The model GET-LFL is only a limited foresight model. Nevertheless, this
feature implies that temporal spill-overs are partially internalized. In GET-LFL
the impact of the backstop technology on the overall energy mix is very modest
because in both cases the backstop technology has gained a substantial propor-
tion of the energy mix in the business-as-usual scenario (Figure 7). In GET-LFL
enough cost reduction potential has already been realized in the business-as-usual
scenario. Moreover, the GET-LFL model assumes a high share of gas in the fossil
fuel mix, so that a modest reduction in the energy demand makes it possible to
achieve climate protection goals even without much ITC.
In DEMETER-1CCS, ITC has only a moderate impact on the energy mix
for two reasons: First, the business-as-usual scenario already assumes some learning
as the backstop technology is introduced as a technological option in 2025. Hence
the cost reduction potential in the policy scenario is limited. Second, the business-
as-usual scenario also assumes a decreasing fossil fuels price path, thus the marginal
effect of learning-by-doing is limited and the break-even point is changed little.
Figure 8 also helps to understand the role of technological change in the
resource extraction sector. Similar to technological change in the case with back-
stop technology, it could reduce the growth rate of the price of energy from fossil
fuels by making more fossil resources available at lower costs. If learning-by-do-
ing was assumed, the effect would be more pronounced in the baseline than in the
policy scenario, which would widen the gap between the resource price with and
without policy goal. Cost reductions of fossil fuels due to technological progress
decreases the competitiveness of the backstop technology and therefore increas-
es the opportunity costs of climate protection. Note, that sensitivity analysis in
MIND supports this qualitative insight – technological progress in the extraction
sector is one of the most sensitive parameters in determining the opportunity costs
of climate protection (Edenhofer et. al. 2006). Thus, it would be interesting to see
other model types including realistic representation of endogenous technological
change in resource extraction and its effects on resource availability into their
estimates of climate protection costs.
Another aspect is illustrated by Figure 7: as discussed above, some mod-
els will rather cut back on energy use relative to business-as-usual than provide
carbon-free (or low carbon) energy. This is evident in Figure 7 when overall en-
ergy consumption in the policy scenarios is much lower than in the baseline; ex-
80 Chapter 3 Implication of ITC for Atmospheric Stabilization
amples are IMACLIM-R, and E3MG. Other models manage to make almost as
much energy available as in the baseline by changing to low carbon or carbon-free
energy sources, e.g. MIND, DEMETER-1CCS and the energy system models.
This echoes the findings from the previous section, and is in fact one of the un-
derlying factors influencing whether a model implements a mitigation strategy of
carbon intensity reduction or energy intensity reduction.
5.5.2 Shadow Prices, Carbon Taxes and Path Dependency
The price of carbon plays a different role in different models (Figure
9 and Figure 10). First best models of the economy (e.g. MIND) make the im-
plicit assumption that all market imperfections may be cured. Hence, the result
of welfare maximization in these models is a Pareto-efficient solution without
any further restrictions. In these models, the shadow price of carbon represents
the social costs of carbon. Second best models, e.g. general equilibrium models,
simulate market behavior, i.e. the model incorporates distortions that cannot be
removed by policy instruments for institutional or political reasons. The carbon
tax in DEMETER-1CCS represents a second-best optimum in the sense that it
is imposed on the economy in order to guarantee the achievement of the stabilization
level and a minimum of welfare losses subject to the market distortions that cannot
be removed by policy instruments because of institutional or political inertia.
In the other models in Figure 9 (IMACLIM-R and E3MG) the imposed
tax does not represent a second best optimum because the carbon tax only allows
the achievement of a stabilization level irrespective of its welfare implications.
The carbon tax profiles in IMACLIM-R and E3MG are prescribed exogenously,
i.e. they are non-optimum.
Induced Technological Change / 95
Figure 8. Different Formulations of Backstop and Resource Scarcity
3.5 Results and Discussion 81
96 / The Energy Journal
In the class of optimal growth models, the carbon price is a dual variable
and represents the social costs of carbon (Figure 10). Moreover, the time path of
carbon follows an optimum path which could be interpreted as an ideal market for
carbon permits or as an imposed optimal carbon tax. In energy system models the
carbon price is also a dual variable in an optimization framework. However, the
carbon price does not necessarily represent the total social costs of carbon because
of the omitted feedback loops between the energy sector and the macro-economic
environment in that partial-equilibrium framework.
The carbon price also reflects the effect of ITC in some models. In nearly
all models the carbon price is higher in the scenarios without technological change.
However, in MIND the carbon price behaves differently: it increases exponentially
Figure 9. Carbon Tax
Figure 9 a shows the 450ppm CO2 stabilization scenario with ITC, Figure 9b shows the
corresponding scenario without ITC. Values greater than $800 per ton of C were cut off; the
corresponding maximum value is given.
(a) Carbon tax with ITC
(b) Carbon tax without ITC
82 Chapter 3 Implication of ITC for Atmospheric Stabilization
in the case without ITC but it peaks and decreases if ITC is switched on.
There is an interesting pattern in carbon price development in some mod-
els: towards the end of the century, the shadow price reaches a maximum and be-
gins to decline. This is true for all scenarios with ITC in MIND and in the 450ppm
scenario for DEMETER-1CCS. If the price of the backstop technology decreases
over time, even without an increasing shadow price of emissions (and fossil fuel
price), the backstop technology remains competitive with fossil fuels. In contrast
to a model with an exogenous price of the backstop technology, learning-by-do-
ing of the backstop technology creates a path dependency because its price is
determined endogenously by investments in learning-by-doing. There is no longer
an incentive for investors to promote fossil fuels after the energy system is trans-
formed because the price of the backstop technology also declines with the trans-
formation of the energy system. The shadow price in most energy system models
increases throughout the century indicating that the transformation of the energy
system is not completed before 2100. This may be in part because renewables or
nuclear power (as backstop technologies) are not able to substitute fossil fuels un-
til the end of the century, due to bounds on market share for renewables, moderate
price increases for fossil fuels that remain too low to trigger a transformation, and
relatively optimistic assumptions about CCS. The remaining share of fossil fuels
will turn carbon into a scarce factor in production with a positive price.
Path dependencies occur if the transformation to a carbon-free energy
system is irreversible in that the carbon-free technologies become the least cost
set of options.
5.5.3 The Specific Role of Carbon Capturing and Sequestration
Among the participating models, five explicitly incorporate the option
of capturing and storing CO2 emissions from combustion (DEMETER-1CCS,
MIND, DNE21+, GET-LFL, and MESSAGE-MACRO). Figure 11 shows how
much CO2 is captured in different scenarios, accumulated over the century. Fig-
ure 12 gives the corresponding time paths of carbon capturing and sequestration
(CCS) for one exemplary scenario (500ppm CO2 stabilization).
As one would expect, Figure 11 shows that the more challenging the
climate policy target, the more CO2 is captured and stored. There is no CCS in the
baseline, as capture and storage of CO2 is costly and hence only becomes econom-
ical in the presence of climate policy. DNE21+ is an exception, because the model
includes an option to use CCS in the context of enhanced oil recovery which
makes CCS economical in its own right. The contribution to overall abatement
(the difference of cumulative emissions between baseline and policy scenarios) is
substantial, in particular in MIND, DNE21+, and GET-LFL. However, nowhere
is CCS the dominant mitigation option but rather, it is always predicted to be one
among many (we conclude this from the fact that captured CO2 is only a small
proportion of the difference of emissions in baseline and policy scenario).
Induced Technological Change / 97
3.5 Results and Discussion 83
98 / The Energy Journal
Figure 10. Shadow Price of Carbon
Figure 10a shows the 450ppm scenario with ITC, Figure 10b shows the corresponding scenario
without ITC. In case of MESSAGE-MACRO, the figures show numbers from the 500ppm scenario
instead of the 450ppm scenario. Values greater than $800 per ton of C were cut off; the corresponding
maximum value is given.
(a) Shadow price with ITC
(b) Shadow price without ITC
84 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 99
Figure 11. Captured CO2 and Total CO2 Emissions
The figure summarizes usage of the CCS option in the baseline and two policy scenarios as a share
of total amount of CO2. CO2 that is not captured is emitted.
As mentioned before, the models show agreement on the allowable car-
bon budget in the policy scenarios, yet they predict divergent cumulative emis-
sions in the baseline. This affects the predicted extent of CCS. DEMETER-1CCS
and MESSAGE-MACRO, on the one hand show fairly low baseline emissions
and in turn low predictions for CCS. On the other hand the remaining three mod-
els are faced with a greater need to reduce emissions and resort to a stronger usage
of the CCS option. Both groups, DEMETER-1CCS and MESSAGE-MACRO as
well as MIND, DNE21+ and GET-LFL show good agreement in their predicted
utilization of the CCS option.
Figure 12 shows the development of CCS over the course of the century.
The five models show diverse behavior. In two of the linear-programming energy
system models (DNE21+ and GET-LFL) the capacity of CCS increases almost
linearly with time and is still rising at the end of the century. This suggests that the
rapidity of increasing this capacity is restricted, but no (anticipated) constraints
to the volume of CCS are effective yet. GET-LFL includes CCS in combination
with energy production from biomass. Thus in GET-LFL CCS is indeed not con-
strained by fossil fuel scarcity.
In contrast, CCS in DEMETER-1CCS levels off towards the end of the
century. Here, CCS activity has reached at least a temporary equilibrium. Possibly
the low emission profiles in the baseline allow these models to reach a CCS capac-
ity that is both sustainable and sufficient for the policy target.
MIND and MESSAGE-MACRO show yet another type of behavior. In
MIND, capacities for CCS are built up even faster than in the energy system models,
but after a peak around mid-century the usage of CCS declines. Similarly, in MES-
SAGE-MACRO CCS peaks in 2080 and declines. Both models respect the scarcity
3.5 Results and Discussion 85
100 / The Energy Journal
of fossil fuel resources increasing costs on the utilization of CCS in the long-run.
While CCS is at a competitive advantage over renewable energy technologies due
to cheap fossil fuels early on in MIND and MESSAGE-MACRO, this advantage is
lost as renewables become more economical due to learning-by-doing.
Two more features contribute to the temporary nature of CCS in MIND:
readily available storage sites are subject to scarcity14, and MIND includes leakage
from storage sites at a fixed rate (i.e. the same percentage leaks from the storage
site in each time period), implying that CCS does not prevent but only strongly
delays emissions into the atmosphere. The leakage rate is highly uncertain, but
it plays an important part in determining whether CCS constitutes a temporary
rather than a permanent solution. It would therefore be instructive to see whether
other models confirmed this result from MIND (Bauer et al. 2005), when leakage
is included.
Carbon capturing and sequestration (CCS) is different from backstop
technologies because it is dependent on non-reproducible inputs, e.g. fossil re-
sources15. Furthermore its extent is limited by the availability of storage sites.
If all relevant intertemporal social costs are taken into account, CCS is only a
temporary solution until the backstop technology becomes competitive. CCS is
an end-of-pipe technology allowing in the best case a welfare improving post-
ponement of the diffusion of the backstop technology. In a theoretical analysis,
14. In MIND, the assumption is that with the rising utilization of CCS, increasingly long pipelines
are needed to transport CO2 to the storage site. In general, spatial aggregation within the models and
limited knowledge about the location of suitable storage sites add to the uncertainties in modeling CCS.
15. GET-LFL also includes CCS in combination with energy production from biomass.
Figure 12. Carbon Capture and Sequestration Over the Course
of the Century
86 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 101
Edenhofer et al. (2005b) show that temporary welfare gains from CCS increase
when (a) the discount rate is increased, (b) the energy penalty is decreased, (c)
the operation and maintenance costs (O&M) are reduced, (c) the leakage rate of
deposits are lowered, (d) the capacity of deposits is increased and (e) the costs of
the fossil fuels are decreased. Gains are also higher when the price of the backstop
technology is high and/or when its learning rate is low.
The CGE model within IMCP has not incorporated CCS so far. In gen-
eral, CGE models could inform about the market potential of CCS under different
policy scenarios. However, CGE models allowing only for a recursive dynamic
are not appropriate for deriving realistic market behavior because they implicitly
assume purely myopic investment behavior which is arguably an exaggerated or
extreme behavior.
6. CONCLUSION
This model comparison aims to draw robust results on ETC by identify-
ing both the differences between and the underlying mechanisms of the multitude
of participating models. We find that the participating models describe a wide
range of possible futures, with and without climate policy. Although there is no
consensus on the potential role of induced technological change, we identify cru-
cial economic mechanisms that drive ITC. This modeling comparison exercise
demonstrates a large influence of the following determinants:
1. Baseline effects
2. First-best or second-best assumptions
3. Model structure
4. Long-term investment decisions
5. Backstop and end-of-the-pipe technologies
6.1 Baseline Effects
All models in the IMCP incorporate endogenous technological change
in their baseline, sometimes in addition to exogenous technological change. In
effect, baseline emissions are difficult to harmonize and vary widely. Both en-
dogenous and exogenous components contribute to this mitigation gap. In some
models optimistic assumptions about exogenous parameters result in relatively
low costs which are then due not to induced technological change, but mainly
to exogenous assumptions. In addition, if the baseline scenario already includes
many positive effects of technological change related to energy and carbon sav-
ings, then the introduction of stabilization targets does not induce much addtional
technological change. Consequently, the cost difference between scenarios with
and without ITC is small.
3.6 Conclusion 87
102 / The Energy Journal
6.2 First Best or Second Best Assumptions
It has important consequences whether a first best or a second best world
is modeled: First best models implicitly assume perfect markets and the imple-
mentation of optimum policy tools. In other words, first best models preclude so
called no-regret options. Therefore, they are inherently more pessimistic about the
costs of climate protection because climate protection reallocates scarce resources
which are utilized in an optimum way in the baseline to climate friendly invest-
ments. In contrast, second best models assume that climate policy can positively
affect market imperfections as a side effect. Compared to first best models the op-
portunity costs of climate protection in second best models can be lower and even
negative, depending on the design of policy.
6.3 Model Structure
Previous model comparison exercises have shown that CGE models
tend to calculate higher mitigation costs than energy system models or economic
growth models (Löschel 2002); we find that this result still holds. However, the
underlying reason is not necessarily the model type, but rather in assumptions
commonly made by “CGE modelers”, “energy system modelers”, and “economic
growth modelers”, e.g. about foresight and intertemporal behavior of the agents.
It turns out that energy system models calculate low mitigation costs
because they only assess the impact of mitigation strategies on energy system
costs. Yet partial equilibrium analysis explicitly omits general equilibrium effects
- partial equilibrium models by definition exclude feedback loops between the
energy sector and other sectors of the economy. In particular, energy system mod-
els implicitly assume that investments within the energy sector can be funded
by the economy at a constant rate of interest. However, this assumption is not
justified when an ambitious climate policy is imposed in the system. This would
depreciate capital stocks in various sectors and therefore also change the return on
investment in the energy sector. Consequently, the changed return on investment
induces a reallocation of investments across sectors. This investment dynamic is a
major determinant of macroeconomic costs of climate policy which is neglected
in partial equilibrium analyses. Moreover, most energy system models neglect
rebound effects and the crowding-out implications of investments. The impact of
these general equilibrium effects emerge to be significant.
In contrast, CGE models demonstrate the quantitative impact of general
equilibrium effects. However, recursive CGE models reduce the flexibility of long-
term investment behavior remarkably. By assumption, investment shares for dif-
ferent sectors are fixed even if an ambitious stabilization level is imposed on the
economy. Some CGE models include a backstop technology, however, its costs
are independent of the timing of investments. Mitigation costs are overestimated
because of the underlying assumptions that investors are myopic.
The econometric model in IMCP describe a second best world. Imper-
88 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 103
fections on the labor market and design of the carbon tax allow substantial welfare
improvements from climate policy. The policy implication is clear. Policy makers
can claim that climate policy is a free lunch. However, it should be emphasized
that second best do not claim that climate policy is the only way or the best way
to cure market failure. If better solutions exist, then climate policy is no longer a
free lunch but has positive opportunity costs. It seems promising to calculate these
opportunity costs based on the strength of both frameworks.
Optimal growth models allow greater flexibility. Some of the optimal
growth models are already designed as multi-sectoral and intertemporal optimiza-
tion models comprising a reduced form energy sector. These models demonstrate
the effect of full temporal and sectoral flexibility. In contrast to energy system
models they do not assume that the differences of the return on investments across
sectors can be ignored. It turns out that an appropriate timing of investments has
the potential to reduce the mitigation costs substantially. In particular, the opti-
mum timing of backstop technologies (like renewables) and end-of-pipe tech-
nologies (like CCS) has a great potential for cost reduction.
6.4 Long-term Decision Making: Foresight and Flexibilities
Assumptions about long-term investment decisions exert a major influ-
ence: The number and flexibility of mitigation options has been shown to have an
impact on mitigation costs (Edenhofer et al. 2005a). This observation is confirmed
in this study.
Perfect foresight enables investors to anticipate necessary long-term
changes and to control investment decisions accordingly, including possible ex-
ternalities such as learning-by-doing. The multi-sector optimal growth models in
this study demonstrate the potential of perfect foresight to reduce mitigation costs.
Models allowing for flexible and long-term investment decisions achieve an equi-
librium that can be characterized by low emissions and low macroeconomic costs.
Naturally, assuming perfect foresight is normative rather than descriptive, i.e. its
model results are motivation for policies rather than an exploration of its effects.
The assumption of intertemporal optimization may exaggerate the po-
tential of ITC to reduce mitigation costs because the rationality and foresight of
investors and entrepreneurs implicit in their intertemporal optimization behavior
represents an optimistic assumption. The assumption of great foresight of the ac-
tors in such models becomes more realistic when a macroeconomic policy ensures
credible expectations. Currently, the number of uncertainties for investors is large,
including uncertainty about emission targets, well-designed international tradable
permit schemes, subsidies for R&D investments, well-behaved capital markets
allowing for long-term investments, and competition and globalization on the en-
ergy market. A stable macro-economic environment and clear long-term emission
targets are crucial for the transformation of the energy system. Therefore, a focus
for post-Kyoto discussions beyond 2012 should be the design of policy instru-
ments allowing for long-term investments.
3.6 Conclusion 89
104 / The Energy Journal
6.5 Backstop and End-of-the-pipe Technologies
Finally, the results depend on the design of backstop and end-of-pipe
technologies: Whether and how a carbon-free energy source is implemented has an
essential impact on mitigation costs as well as on the mix of mitigation options.
If a model allows for endogenous long-term investments in backstop
technologies and/or end-of-pipe technologies, then mitigation costs are substan-
tially reduced and the stabilization targets can be met without drastic declines
in energy consumption. Moreover, available carbon-free energy sources shift the
abatement strategy towards decarbonization rather than energy saving.
Nearly all models conclude that more ambitious climate protection goals
increase the costs. It should be noted that this is not a trivial statement because
due to learning-by-doing, mitigation costs could be decreased if less ambitious
stabilization targets are imposed. However, modeling teams in IMCP assume that
learning-by-doing has its clear limits because of floor costs, barriers of diffusion
and other market imperfections like insufficient internalization of intertemporal
or interregional spillovers.
Over the past decade the debate has been focused mainly on the learning-
by-doing potential of backstop technologies. However, this study shows that this is
only one aspect. Another key factor determining the competitiveness of the back-
stop is technological progress in the fossil fuel sector. Assumptions about the fossil
fuel sector and its potential for technological change are crucial for determining
costs and strategies. Therefore, further modeling efforts should also focus on a more
realistic representation of technological progress within the fossil fuel sector.
Moreover, all models indicate carbon costs that rise with time in the ear-
ly years, and most maintain this across the century. However, some models which
incorporate backstop technologies and carbon capturing and sequestration show a
“hump” in the time path of carbon permit prices, i.e. carbon costs peak and decline
afterwards. This supports what some technical change analysts have supposed: expe-
rience from learning-by-doing or the reality of sunk costs introduce a path dependen-
cy scenario development, and thus the marginal costs of maintaining low emission
levels decrease in the long term due to cumulative learning effects and the usage of
a broad range of mitigation options like improvement of energy efficiency, the diffu-
sion of backstop technologies and the temporary use of end-of-pipe technologies.
6.6 Hints for a Future Research Agenda
This modeling comparison exercise takes a first step in assessing the quan-
titative impacts of ITC on mitigation costs and mitigation strategies. We assess the
impact of ITC is isolated by imposing ceteris paribus conditions, i.e. ITC is induced
by climate stabilization targets in a setting where boundary conditions and param-
eters remain unchanged.
Beyond the IMCP, we recommend research expansion two ways. First, fu-
ture model comparisons could refine the harmonization of the participating models
90 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 105
to a baseline of central variables (capital stock, investments, direction of technologi-
cal change) and parameters in order to minimize baseline effects. Second, more so-
phisticated ceteris paribus scenarios could be run, e.g. exploring the impact of single
ITC options rather than enabling and disabling all ITC as it was done here.
Not all important aspects of ITC could be addressed in this study. They
should be explored in future model comparisons, e.g. regional spillovers. More-
over, while this study restricted policy intervention to imposing stabilization levels
(i.e. represents only the targets approach to policy), the effects of different policy
instruments are neglected. An exercise comparing policy instruments across dif-
ferent model types could accelerate research on optimal climate policy design.
IMCP allows to set out a formulation of an agenda to improve model-
ing design. First, we have explored some reasons for the gaps between top-down
and bottom-up models and discussed several models that begin to bridge this gap.
These hybrid models seem a promising starting point from which to develop a
coherent framework incorporating intertemporal, intersectoral and interregional
effects of induced technological change. Second, as it has turned out in the IMCP,
assumptions about long-term investment behavior have a strong impact on mitiga-
tion costs and strategies. Therefore, experiments with different assumptions about
long-term expectations and long-term flexibility of investment behavior would be
highly valuable. Third, the way carbon-free energy is made available has turned out
to have a major influence on the response of the model to climate policy goals and
therefore deserves attention. This is explored by many models implementing back-
stop- and/or end-of-the-pipe technologies. We argue that endogenous technologi-
cal change in the extraction sector of fossil fuel is a complementary prerequisite for
a comprehensive understanding of ITC. Many modeling teams within IMCP have
incorporated learning-by-doing of the backstop technology. In contrast to this, en-
dogenous technological change in the exploration and extraction sector of fossil
fuels has not received as much attention. There is significant technological change
(e.g. in the resource extraction sector) with a potentially strong influence on the
opportunity costs of climate protection. A better understanding of the underlying
dynamics may therefore both satisfy scientific curiosity and also provide a prereq-
uisite for improving the design of climate policy.
ACKNOWLEDGEMENTS
We would like to thank all modeling teams for the productive coopera-
tion in this project, especially for repeatedly checking the interpretations and con-
clusions we derived about their models. While we were writing this synthesis re-
port we received many valuable comments and they are gratefully acknowledged
here. Beyond the IMCP modeling teams, we would like to thank three anony-
mous referees. We also would like to thank Nico Bauer and Marian Leimbach for
valuable comments and discussion. John Schellnhuber has supported our work in
many ways. Kai Lessmann was funded by the Volkswagen Foundation, Project
II/78470, which we gratefully acknowledge.
3.6 Conclusion 91
106 / The Energy Journal
REFERENCES
Azar, C. and S. H. Schneider (2002). “Are the economic costs of stabilising the atmosphere prohibi-
tive?” Ecological Economics 42:73-80.
Barker, T., H. Pan, J. Köhler, R. Warren and S. Winne (2006). “Decarbonizing the global economy
with induced technological change: scenarios to 2100 using E3MG”, The Energy Journal Special
Issue, Endogenous Technological Change and the Economics of Atmospheric Stabilization.
Bauer, N., O. Edenhofer, H. Held and E. Kriegler (2005). “Uncertainty of the Role of Carbon Capture
and Sequestration within Climate Change Mitigation Strategies.” In: E. S. Rubin, D. W. Keith and
C. F. Gilboy (Eds.), Proceedings of the 7th International Conference on Greenhouse Gas Control
Technologies. Volume 1: Peer-Reviewed Papers and Plenary Presentations, IEA Greenhouse Gas
Programme, Cheltenham, UK.
Bosetti, V., C. Carraro, M. Galeotti (2006). “The Dynamics of Carbon and Energy Intensity in a Model
of Endogenous Technical Change.” The Energy Journal Special Issue, Endogenous Technological
Change and the Economics of Atmospheric Stabilization.
Crassous, R., J.-C. Hourcade, O. Sassi (2006). “Endogenous Structural Change and Climate Targets in
the exploration and extraction sector of fossil fuels” The Energy Journal Special Issue, Endogenous
Technological Change and the Economics of Atmospheric Stabilization.
Clarke, LE. and JP Weyant (2002). “Modeling Induced Technical Change: An Overview”, in A.
Grübler, N. Nakicenovic, and WD. Nordhaus (eds.), Technological Change and the Environment,
Washington D.C.; Resources for the Future Press, 320-363.
Edenhofer, O., N. Bauer and E. Kriegler (2005a). “The Impact of Technological Change on Climate
Protection and Welfare: Insights from the Model MIND.” Ecological Economics 54:277-292.
Edenhofer, O., H. Held, and N. Bauer (2005b). “A Regulatory Framework for Carbon Capturing and
Sequestration with the Post-Kyoto Process.” In: E. S. Rubin, D. W. Keith and C. F. Gilboy (Eds.), Pro-
ceedings of the 7th International Conference on Greenhouse Gas Control Technologies. Volume 1: Peer-
Reviewed Papers and Plenary Presenatations, IEA Greenhouse Gas Programme, Cheltenham, UK.
Edenhofer, O., K. Lessmann, N. Bauer (2006). “Mitigation Strategies and Costs of Climate Protection:
The Effects of ETC in the Hybrid Model MIND.” The Energy Journal Special Issue, Endogenous
Technological Change and the Economics of Atmospheric Stabilization.
Edmonds, J., J. M. Roop, M. J. Scott (2000). “Technology and the economics of climate change
policy.” Pew Center on Global Climate Change, Batelle, Washington DC, USA.
EMF 19 (2004). “EMF 19 Alternative technology strategies for climate change policy.” J. P. Weyant
(Ed.), Special Issue of Energy Economics 26(4):501-755.
Gerlagh R. (2006). “ITC in a Global Growth-Climate Model with CCS: The Value of Induced Techni-
cal Change for Climate Stabilization.” The Energy Journal Special Issue, Endogenous Technological
Change and the Economics of Atmospheric Stabilization.
Grubb, M., J. Köhler and D. Anderson (2002). “Induced technical change in energy and environmental
modeling: Analytic Approaches and Policy Implications.” Annu. Rev. Energy Environ. 27:271-308.
Hare, B. and Meinshausen, M., (2004). “How much warming are we committed to and how much can
be avoided.” PIK Report No. 93.
Hedenus, F., C. Azar and K. Lindgren (2006). “Induced technological change in a limited foresight
optimization model.” The Energy Journal Special Issue, Endogenous Technological Change and the
Economics of Atmospheric Stabilization.
HM Treasury (2003). “Appraisal and Evaluation in Central Government (Green Book).” Website:
greenbook.treasury.gov.uk.
Hourcade, J-C. and J. Robinson (1996). “Mitigating factors – Assessing the costs of reducing GHG
emissions.” Energy Policy 24:863-873.
IPCC (1996). Climate Change 1995: Economic and Social Dimensions of Climate Change, Cam-
bridge University Press.
IPCC (2001). Climate Change 2001: Mitigation, Cambridge University Press.
92 Chapter 3 Implication of ITC for Atmospheric Stabilization
Induced Technological Change / 107
Köhler, J., M. Grubb, D. Popp, and O. Edenhofer (2006). “The Transition to Endogenous Technical
Change in Climate-Economy Models.” The Energy Journal Special Issue, Endogenous Technologi-
cal Change and the Economics of Atmospheric Stabilization.
Löschel, A. (2002). “Technological Change in Economic Models of Environmental
Policy: A Survey”, Ecological Economics, 43, 105-126.
Morgan, M. G. and M. Henrion (1990). Uncertainty – A Guide Dealing with Uncertainty in Quantita-
tive Risk and Policy Analysis. Cambridge University Press, UK.
Masui, T., T. Hanaoka, S. Hikita and M. Kainuma (2006). “Assessment of CO2 Reductions and Eco-
nomic Impacts Considering Energy-Saving Investments.” The Energy Journal Special Issue, Endog-
enous Technological Change and the Economics of Atmospheric Stabilization.
Nordhaus, W. D. and J. Boyer (2000). Warming the World. MIT Press, Cambridge.
Popp, D. (2006). “Comparison of Climate Policies in the ENTICE-BR Model.” The Energy Journal
Special Issue, Endogenous Technological Change and the Economics of Atmospheric Stabilization.
Sano, F., K. Akimoto, T. Homma, and T. Tomoda (2006). “Analysis of technological portofolios for
CO2 stabilizations and effects of technological changes.” The Energy Journal Special Issue, Endog-
enous Technological Change and the Economics of Atmospheric Stabilization.
Rao, S., I. Keppo and K. Riahi (2006). “Importance of Technological Change and Spillovers in Long-
Term Climate Policy.” The Energy Journal Special Issue, Endogenous Technological Change and
the Economics of Atmospheric Stabilization.
Sjim, J. P. M. (2004). “Induced Technological Change and Spillovers in Climate Policy Modeling.”
ECN report ECN-C--04-073, Energy Research Centre of the Netherlands.
Sun, J. W. (1998). “Changes in energy consumption and energy intensity: a complete decomposition
model.” Energy Economics 20(1):85-100.
Sun, J. W. and B. W. Ang (2000). “Some properties of an exact energy decomposition model.” Energy
25:1177-1188.
Repetto, R. and D. Austin (1997). “The Costs of Climate Protection: A Guide for the Perplexed.”
World Resources Institute, Washington DC, USA.
Vollebergh, H. R. J., and C. Kemfert (2005). “The role of technological change for sustainable devel-
opment.” Ecological Economics 54:133-147.
Vuuren, D.P. van, MGJ. den Elzen, MM. Berk, P. Lucas, B. Eickhout, H. Eerens and R. Oostenrijk
(2003). Regional costs and benefits of alternative post-Kyoto climate regimes – Comparison of vari-
ants of the Multi-stage and Per Capita Convergence regimes. RIVM report 728001025/2003. Na-
tional Institute of Public Health and the Environment. Bilthoven. The Netherlands.
Weyant, J. P. (2000). “An Introduction to the economics of climate change policy.” Pew Center on
Global Climate Change, Batelle, Washington DC, USA.
Weyant, J. P. (2004). “Introduction and overview.” Energy Economics 26:501-515.
Weyant, J. P. and T. Olavson (1999). “Issues in Modelling Induced Technological Change in Energy,
Environmental, and Climate Policy”, Environmental Modelling and Assessment, 4: 67-85.
3.7 References 93
94 Chapter 3 Implication of ITC for Atmospheric Stabilization
Chapter 4
The Effects of Tariffs on Coalition Formation in a
Dynamic Global Warming Game∗
Kai Lessmann
Robert Marschinski
Ottmar Edenhofer
∗accepted for publication in Economic Modelling as Lessmann, K., R. Marschinski, and O. Edenhofer:
“The Effects of Tariffs on Coalition Formation in a Dynamic Global Warming Game,” Economic Modelling
(2009), doi:10.1016/j.econmod.2009.01.005
95
96 Chapter 4 Effects of Tariffs on Coalition Formation
-<9::97HGC:-5F=::GCBC5@=H=CBCFA5H=CB=B5
MB5A=7 @C65@05FA=B; 5A9
$5=%9GGA5BB56+C69FH&5FG7<=BG?=5(HHA5F89B<C:9F5
,&:?>/,89>?4?@?01:=748,?08;,.?'0>0,=.3&%:C&:?>/,80=8,9D
-:==0>;:9/492,@?3:=08,4770>>8,99;46;:?>/,8/0;3:90
1,C
!" "
-<9DFCGD97HG:CF7CCD9F5H=CBCB7@=A5H9DFCH97H=CB69MCB85F97IFF9BH@MIB79FH5=B-<IGDC@=7M
=BGHFIA9BHGK<=7<:CGH9FD5FH=7=D5H=CB=B"BH9FB5H=CB5@BJ=FCBA9BH5@;F99A9BHG"5F9=B89A5B8
ACB;H<9=BGHFIA9BHGIB89F8=G7IGG=CB5F9HF589G5B7H=CBG&I@H=F9;=CBCDH=A5@;FCKH<AC89@G5F95
GH5H9C:H<95FHHCC@:CF=BH9;F5H985GG9GGA9BH6IH=BHFC8I7=B;HF589G5B7H=CBG8=GHCFHGH<97CAD9H=H=J9
9EI=@=6F=IAA5?=B;=H8=::=7I@HHC7CADIH9BIA9F=75@@M09=BHFC8I79HF5895B8HF589G5B7H=CBG=BHC5
AC89@C:7C5@=H=CBGH56=@=HMHC5GG9GGH<9DCH9BH=5@C:HF589G5B7H=CBGHCGIDDCFH5B"-F589=GAC89@98
6M<5J=B;5@@7CIBHF=9GDFC8I795;9B9F=7CIHDIH;CC86IH58CDH=B;B5H=CB5@DFC8I7H8=::9F9BH=5H=CB
FA=B;HCB5GGIADH=CBC5@=H=CBG5F9:F99HC=ADCG9H5F=::GCB=ADCFHG:FCABCB7CCD9F5H=B;7CIBHF=9G
09GC@J9H<9AC89@BIA9F=75@@MIG=B;5F9:=B98J9FG=CBC:'9;=G<=WG65G=75@;CF=H<A09H<9B
5DD@MH<9AC89@HC5B5@MN9H<9=B:@I9B79C:H5F=::GCB=BH9FB5H=CB5@7CCD9F5H=CB-<9AC89@GI;;9GHGH<5H
H<9F9=G=B8998G=;B=:=75BHDCH9BH=5@HCF5=G9D5FH=7=D5H=CBH<FCI;<HF589G5B7H=CBG9J9BK<9B;CC8G:FCA
8=::9F9BH7CIBHF=9G5F9B95F@MD9F:97HGI6GH=HIH9GIFH<9FACF9K9=BJ9GH=;5H9H<99::97HC:HF589G5B7H=CBG
CB;@C65@K9@:5F99BJ=FCBA9BH5@9::97H=J9B9GG5B8H<97F98=6=@=HMC:H<9H5F=::A97<5B=GA
".7,>>414.,?4:9!**
!0DB:=/> @=A5H9 7<5B;9 ,9@:9B:CF7=B; =BH9FB5H=CB5@ 9BJ=FCBA9BH5@ 5;F99A9BHG -F589 G5B7H=CBG
C5@=H=CBGH56=@=HM"BH9FB5H=CB5@7CCD9F5H=CB
" #""$"
CA6=B=B;9@9A9BHGC:H<997CBCA=7H<99B9F;M5B8H<97@=A5H9GMGH9A"BH9;F5H98
GG9GGA9BH&C89@G"&G<5J9697CA95B=B8=GD9BG56@9:CFA5@HCC@=BH<9F95@AC:
7@=A5H9DC@=7M5B5@MG=G-<9F95F9BIA9FCIG9L5AD@9GF5B;=B;:FCA'CF8<5IG
G9A=B5@"AC89@HCH<9@5H9GH;9B9F5H=CBC:F9;=CB5@=N98AC89@G:95HIF=B;<=;<@9J
9@GC:G97HCF5@5B8H97<BC@C;=75@89H5=@
DFCA=B9BH7@5GGK=H<=BH<9"&:5A=@M7CBG=GHGC:CDH=A5@;FCKH<AC89@GH<9G9
,99:CF9L5AD@9$MDF9CG5B85<B5F?9F9H5@F5GGCIG9H5@CG9HH=9H5@
4.1 Introduction and Motivation 97
"0>>8,990?,7.:9:84.#:/077492
6I=@8CB5HF58=H=CB;C=B;657?HC+5AG9M5B8J=9K577IAI@5H=CB5B897CBCA=7
;FCKH<5G8F=J9B6M5;9BHG=BH9FH9ADCF5@@MCDH=A=N98=BJ9GHA9BH897=G=CBGL5AD@9G
=B7@I89H<9+"":5A=@MC:AC89@G'CF8<5IG'CF8<5IG5B815B;
5B8=HGAC8=:=75H=CBGGI7<5G&+"CG9HH=9H5@CF'-")CDD
5GK9@@5GH<9&"'89B<C:9F9H5@5B8&-+ 9F@5;<
AC89@G
-KCA5=B5GD97HG>IGH=:MH<9IG9C:=BH9FH9ADCF5@CDH=A=N5H=CB=BH<97CBH9LHC:7@=
A5H9DC@=7M=FGH89B<C:9F9H5@5F;I9H<5HH<=G:F5A9KCF?=G5DDFCDF=5H9
K<9B9J9FH<9F9G95F7<EI9GH=CBF9EI=F9G5B97CBCA=7AC89@HC69FIBCJ9F@CB;H=A9
<CF=NCBG5B8HC75DHIF9GHFI7HIF5@7<5B;9G"B8998=B9FH=5=BH<97@=A5H9GMGH9AF9EI=F9G
HC58CDHH=A9<CF=NCBGC:ACF9H<5B579BHIFM,97CB8-IFBCJG?MDD5F;I
=B;:FCA5ACF9H<9CF9H=75@DC=BHC:J=9K657?GH<9=BH9FH9ADCF5@IH=@=HMA5L=A=N5H=CB
C:5F9DF9G9BH5H=J95;9BH5GH<9DF9:9FF98K5MHC;=J9A57FC97CBCA=7AC89@G5:=FA
A=7FC:CIB85H=CB5B8A5?9H<9AGI=H56@9:CFK9@:5F95B5@MG=G@H<CI;<7F=H=7GDC=BHHC
H<9:57HH<5H5GGIADH=CBGGI7<5GD9F:97H:CF9G=;<H5B8GHF=7HF5H=CB5@=HM5F957HI5@@M5H
C88GK=H<F95@=HMF9GI@HG:FCAGI7<AC89@GF9H5=BH<9=FIG9:I@B9GG5H@95GH=BH9FAGC:5
:=FGH69GH69B7<A5F?
-C7CA97@CG9FHCH<9DC@=H=75@F95@=HMC:5KCF@87CBG=GH=B;C:G9@:=BH9F9GH985B8
GCJ9F9=;BB5H=CBGH5H9GCDH=A5@;FCKH<AC89@G>IGH@=?9CH<9F"&G<5J9CJ9FH=A9
D5GG98:FCA5IB=F9;=CB5@KCF@8F9DF9G9BH5H=CBHC58979BHF5@=N98AI@H=F9;=CB5@:CF
AI@5H=CB.B:CFHIB5H9@M9J9BH<9GC@9=BHFC8I7H=CBC:9A=GG=CBGHF5897CA9G5HH<97CGH
C:5GI6GH5BH=5@5;;F5J5H=CBC:H<9BIA9F=7GF9EI=F98HC7CADIH97CAD9H=H=J99EI=@=6F=5
-<975@7I@5H=CBC:HF589:@CKG5B8DF=79J97HCFGKCI@8=BDF=B7=D@969GHF5=;<H:CFK5F8
K=H<'9;=G<=G5@;CF=H<AIH=BH<9DF9G9B79C:5B9LH9FB5@=HM@=?9H<97@=A5H9
:998657?5B5DDFCDF=5H9AC8=:=75H=CBC:H<95@;CF=H<A=GF9EI=F98-<9588=H=CB5@9:
:CFH=GC:7CIFG9>IGH=:=986MH<9B998HC9GH=A5H9H<9F9;=CB5@8=GHF=6IH=CBC:7@=A5H9
85A5;9G5B8A=H=;5H=CB7CGHG5GK9@@5G6MH<9B9KDCGG=6=@=HMHC7CADIH9G79B5F=CG=B
K<=7<CB@M5;FCIDC:B5H=CBGY5]7@=A5H97C5@=H=CB^Y897=89GHC7CCD9F5H9CB7@=A5H9
7<5B;9
"BCIFKCF?K9:C@@CKH<9AI@H=F9;=CB5@AC89@=B;5DDFC57<5B8:CFA5@@M9LH9B8=H
=BHKCK5MG:=FGH=BH9FB5H=CB5@HF589=B;CC8G=G=BHFC8I7986M8FCDD=B;H<97CAACB
5GGIADH=CBH<5H5@@7CIBHF=9GDFC8I79H<9G5A9D9F:97H@MGI6GH=HIH56@9;CC8=BGH958K9
5GGIA9H<5H;CC8G5F98=::9F9BH=5H98577CF8=B;HCH<9=FD@579C:CF=;=B-<=G5DDFC57<Y
GCA9H=A9GF9:9FF98HC5GFA=B;HCB5GGIADH=CBY=GC:H9B9B7CIBH9F98=B AC89@
=B;5B85@@CKGHCF9DFC8I79 =BH9FB5H=CB5@ 7CGHGD=@@CJ9FG :FCAA=H=;5H=CBDC@=7=9G
;"'CF8<5IG5B8&"'89B<C:9F9H5@
;+"'CF8<5IG5B815B;5B80"-!CG9HH=9H5@
"AD@9A9BH=B; HF589 =B H<9G9 AC89@G =G 7<5@@9B;=B; 'CF8<5IG 5B8 15B; M7?A5BG 5B8
-I@?9BG'CF8<5IG5B815B;A9BH=CBH<5HZ5A5>CF75IG9C:H<9@CB;;9GH5H=CBD9F=C8
C:H<=GF9G95F7<<5G699BH<98=::=7I@HM=B:=B8=B;5G5H=G:57HCFM5@;CF=H<A:CFGC@J=B;H<9=BH9FH9ADCF5@
;9B9F5@9EI=@=6F=IA[
;=BH<9+"'CF8<5IG5B815B;5B80"-!CG9HH=9H5@AC89@G
-<=GAC89@C:=BH9FB5H=CB5@HF589=G8=G7IGG989;=B99BGHF59H5@
;9FBGH9=B9H5@$9A:9FH
"BAC89@GK=H<CIHHF589CB97CIBHFM^G75F6CB7CBGHF5=BH695FGBC97CBCA=77CBG9EI9B79G:CFCH<9F
7CIBHF=9G-<=GG99AG7CBHF58=7HCFMK<9BH<=B?=B;C:G<=:HG=B7CAD9H=H=J958J5BH5;95B8GD97=5@=N5
H=CB]75F6CB@95?5;9^5GK9@@5GC:H<9B9;5H=J97CBG9EI9B79G:CFGCA97CIBHF=9G=::CGG=@:I9@89
98 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
,97CB8K9=BHFC8I795BCH<9F:95HIF9H<5H=G=B7CAD5H=6@9K=H<H<965G=7'9;=G<=5D
DFC57<B5A9@M5H5L8=GHCFH=CB=B:CFAC:5DIB=H=J9H5F=::8IHM
-<9:=FGHD5FHC:H<9D5D9F9AD<5G=N9GH<9:CFA5@5GD97HGC:GC@J=B;GI7<5AC89@
GHFI7HIF9:CF57CAD9H=H=J99EI=@=6F=IA0989G7F=69CIFGC@IH=CB5DDFC57<H<5H8F5KG
CBKCF?6M$9<C99H5@5B8%9=A657<5B889B<C:9F5B8=@@IGHF5H9<CK
5J5@=85H=CBC:H<97CAD9H=H=J99EI=@=6F=IA=GC6H5=B98
-C89ACBGHF5H9H<9IG9:I@B9GGC:H<9AC89@G9HID5B5DD@=75H=CBHC57IFF9BH=GGI9
=B7@=A5H9DC@=7M=GDF9G9BH98=BH<9G97CB8D5FHC:H<9D5D9F'5A9@MK95B5@MN9H<9
G7CD9:CFF9;=CB5@7CCD9F5H=CBYH<5H=GH<9J=56=@=HMC:5]7@=A5H97C5@=H=CB^Y5B8=BJ9G
H=;5H9K<9H<9FH5F=::G75B<9@DHC=B7F95G9D5FH=7=D5H=CB=BGI7<57C5@=H=CB
-<=GEI9GH=CBG99AGH=A9@M=BJ=9KC:H<97IFF9BH@MA95;9FDFCGD97HG:CF:I@@=BH9F
B5H=CB5@7CCD9F5H=CB5:H9FH<99LD=FMC:H<9$MCHC)FCHC7C@=B"B89985@=J9@M
8965H9<5G9A9F;98CBH<9G7CD9:CFF9;=CB5@7CCD9F5H=CB5B8J5F=CIGGIDDCFH=J9DC@=7M
=BGHFIA9BHG<5J9699B6FCI;<HID=BH<9@=H9F5HIF9GI7<5G+DFCHC7C@G5FF9HH
5FF5FC 9H 5@ 5 H97<BC@C;M :IB8 9B98=7? 5 &5FG<5@@ )@5B
,7<9@@=B;5B8@5GH6IHBCH@95GHHF589G5B7H=CBG9;@8M(FGN5;5B8,H=;@=HN
-<9IG9C:HF589F9GHF=7H=B;H5F=::8IH=9G<5G699BDFCDCG98=BH<9:CFAC:9B9F;MCF
(6CF89FH5L58>IGHA9BHGK=H<H<98CI6@9C6>97H=J9HC89H9F:F99F=8=B;5B8HC95G9
H<9@CGGC:7CAD9H=H=J9B9GG:CF7C5@=H=CBA9A69FG-<98965H9<5GGC:5F:C7IG98CBH<9
EI9GH=CBC:K<9H<9FH5F=::G5F9:95G=6@9IB89F@9;5@=9FA5BB5B8FC<A5B8
=AD@9A9BH5H=CB "GA9F 5B8 '9I<C:: 5GD97HG !CK9J9F 5BCH<9F EI9GH=CB =G
K<9H<9FH<9=F9AD@CMA9BHKCI@8697F98=6@9;=J9BH<5HCFH<C8CL97CBCA=7H<9CFMGI;
;9GHGH<5HH<98=GHCFH=CB5FM9::97HGC:H5F=::GKCI@869K9@:5F989DF9GG=B;:CF5@@D5FH=9G
&CF9GD97=:=75@@M,H=;@=HNDFCDCG9GHCF5=G9D5FH=7=D5H=CB=B57@=A5H9HF95HM
6M=ADCG=B;HF589G5B7H=CBG5;5=BGHBCBG=;B5HCF=9G!95F;I9GH<5HH<=G=GDCGG=6@95B8
9J9BF9EI=F98=BH<9@9;5@:F5A9KCF?C:H<90CF@8-F589(F;5B=N5H=CB0-(DFC8I7HG
:FCA7CIBHF=9GH<5H5@@CKIB7CBGHF5=B989A=GG=CBG5F9=AD@=7=H@MGI6G=8=N98K<=7<K5F
F5BHGHCDFC<=6=HCFH5F=::H<9=ADCFHC:GI7<DFC8I7HG)9F9N;=J9G589H5=@98
5B5@MG=GC:H<9@9;5@=AD@=75H=CBGC:GI7<5DFCDCG5@7CB7@I8=B;H<5HF979BHDF97989BHG
H<9GC75@@98ZG<F=AD897=G=CB[GI;;9GHH<5HH<90-(K=@@BCH=BH9F:9F9K=H<GI7<
H5F=::G,=A=@5FHCH<9G9HF589G5B7H=CBG'CF8<5IGDFCDCG9G6CF89FH5L58>IGH
A9BHGHC9B:CF797CAD@=5B79K=H<<5FACB=N9875F6CBH5L9G
-<99::97HGC:HF589G5B7H=CBGCB7C5@=H=CB:CFA5H=CB<5J95@GC699B5B5@MN98K=H<=B
:CFA5@AC89@G5FF9HH=BIG5B8+IB8G<5;9B5@69=HHC@9GG9F9LH9BHG
A9BH=CB9869:CF9H<9K=89@MIG98CDH=A5@;FCKH<AC89@G8CBCHB5HIF5@@M577CAAC
85H9HF589=B;CC8GCH<9FH<5B9A=GG=CBGHF5895B85F9H<9F9:CF9BCFA5@@MIBGI=H56@9
:CF5B5B5@MG=GC:H<99::97HGC:H5F=::G-<IG9L=GH=B;:CFA5@GHI8=9GC:HF589G5B7H=CBG
5B8=BH9FB5H=CB5@7CCD9F5H=CB9=H<9FIH=@=N95GH5H=7AC89@=B;:F5A9KCF?5FF9HH
CFCADIH56@9 9B9F5@EI=@=6F=IA AC89@G$9A:9FH
CFH<9DIFDCG9C:H<=GD5D9FK95DD@MH<9AC89@=B5GHM@=N98YH<5H=GBCH9AD=F=
75@@M75@=6F5H98Y:CFA=BCF89FHC9LD@CF9H<9G7CD9:CFH5F=::G=B=BH9FB5H=CB5@7CCD9F
5H=CB09:=B8H<5HIB89FH<95GGIADH=CBC:DF=795GK9@@5GH5F=::H5?=B;69<5J=CFC:5@@
A5B8D@IB;9G
4.1 Introduction and Motivation 99
"0>>8,990?,7.:9:84.#:/077492
7CIBHF=9GH<9=ADCG=H=CBC:H5F=::GCBBCB7C5@=H=CBA9A69FGIB9EI=JC75@@MF5=G9GH<9
G7CD9 :CF =BH9FB5H=CB5@ 7CCD9F5H=CB !CK9J9F H<9 7C5@=H=CBG K9@:5F9 ;5=BG GH5FH HC
897@=B9CB79H<9H5F=::G;C69MCB8579FH5=BH<F9G<C@85B8Y5H5GH=@@<=;<9F@9J9@Y
H5F=::G57HI5@@M697CA9K9@:5F9897F95G=B;5B8H<IG@CG97F98=6=@=HM09=BH9FDF9HH<9
C6G9FJ989::97HG5G57CBG9EI9B79C:H<9AC89@GF9DF9G9BH5H=CBC:=BH9FB5H=CB5@HF589
K<9B957<7CIBHFMGF9DF9G9BH5H=J9CIHDIH;CC875BCB@M69=AD9F:97H@MGI6GH=HIH986M
;CC8G:FCACH<9F7CIBHF=9G6IH5@@7CIBHF=9GAIGH69<5J95GDF=79H5?9FGH<9BH<9H5F=::
7CBGH=HIH9G 5B =B8=F97H DF=79 G9HH=B; A97<5B=GA K<=7< <9@DG 7C5@=H=CB 7CIBHF=9G HC
75D=H5@=N9CBH<9=F=AD@=7=HA5F?9HDCK9F5B8=B7F95G9H<9=FH9FAGC:HF589!CK9J9F
G=A=@5FHC5BCDH=AIAH5F=::F5H9CFACBCDC@MDF=79H<969B9:=HG:FCAH<=G=B7F95G9GH5FH
HCJ5B=G<CB79H<9H5F=::9L7998G579FH5=B@9J9@
"B@=B9K=H<97CBCA=7H<9CFMCIFAC89@G<CKGH<5HH<9=BHFC8I7H=CBC:H5F=::G8=GHCFHG
H<9CH<9FK=G99::=7=9BHA5F?9HG5B8<9B79;@C65@K9@:5F9KCI@869<=;<9FK=H<CIHH5F
=::G09:=B8<CK9J9FH<5HH<9G9@CGG9G5F995G=@MC::G9H6MH<9;5=BGC:=B7F95G987CCD
9F5H=CBH<5H5F9=B8I7986MH<9G9H5F=::G0=H<F9GD97HHC9BJ=FCBA9BH5@9::97H=J9B9GGK9
:=B8H<5H=BCIFAC89@75F6CB@95?5;9=GGA5@@=99A=GG=CB=B7F95G9G=B:F99F=8=B;
7CIBHF=9G8CBCHCIHK9=;<H<9565H9A9BH9::CFHC:H<97C5@=H=CB
@H<CI;<K99AD@CMH<9AC89@5B8H<95@;CF=H<A=B5B9L9AD@5FMK5M=BCF89FHC
9LD@CF9H<9G7CD9:CFH5F=::G=B7C5@=H=CB:CFA5H=CB=H75B6995G=@M9LH9B898HCCH<9F
F9G95F7<EI9GH=CBG9;HC=BJ9GH=;5H9H<99::97HGC:8=::9F9BH=5H986CF89FH5L58>IGHA9BHG
-CB7C5@=H=CB:CFA5H=CBCFHC5B5@MN9H<9@CB;H9FAGHFI7HIF5@9::97HGC:8=::9F9BH
CDH=A5@BCBCDH=A5@75F6CBH5L9G
-<9F9A5=B=B;D5FHC:H<9D5D9F=GCF;5B=N985G:C@@CKG-<9B9LHG97H=CBDF9G9BHG
H<9AC89@,97H=CB9LD@5=BGH<9GC@IH=CB5@;CF=H<A"B,97H=CBK98=G7IGG=HG5DD@=
75H=CBHC7C5@=H=CBGH56=@=HM=B5AC89@K=H<=ADCFHH5F=::G5B8,97H=CB7CB7@I89G
" #"#
0969;=B6MGH5H=B;H<9DFC6@9AK9=BHFC8I795AI@H=57HCF;FCKH<AC89@K=H<7@=
A5H97<5B;985A5;9G5B8H5F=::GCBHF589:@CKG
&=010=09.0>
57<F9;=CB 4 =GAC89@98:C@@CK=B;+5AG9M=9H<9A5L=A=N5H=CBC:8=G
7CIBH98 IH=@=HM 9B8C;9BCIG@M 89H9FA=B9G H<9 =BH9FH9ADCF5@ 7CBGIADH=CB=BJ9GHA9BH
D5HH9FB
B071,=04
0?74? *
.4? 74?
/?
"BGH5BH5B9CIGIH=@=HM*=G5B=B7F95G=B;5B87CB75J9:IB7H=CBC:D9F75D=H57CBGIAD
H=CB.7"H=GK9=;<H98K=H<H<9F9;=CBGHCH5@DCDI@5H=CB75B88=G7CIBH98K=H<5F5H9C:
DIF9H=A9DF9:9F9B79
ρ
"B5KCF@8K<9F9;CC8G:FCA8=::9F9BH7CIBHF=9G5F9=AD9F:97HGI6GH=HIH9GIH=@=HM
89D9B8GCBH<97CBGIADH=CBC:6CH<8CA9GH=7 ./:8 5B8:CF9=;B;CC8G .1:=K<=7<5F9
7CA6=B98=BHC5GC75@@98FA=B;HCB5;;F9;5H9J=55,CBGH5BH@5GH=7=HMC:
100 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
,I6GH=HIH=CB:IB7H=CB
.4?
>/:8
.4?
/:8
5
>5
1:=
.45?
1:=
-<99@5GH=7=HM
σ
=G89H9FA=B986MH<9D5F5A9H9F
ρ
∈(0,1)577CF8=B;HC
σ
ρ
,<5F9D5F5A9H9FG>/:85B8>51:=7<5F57H9F=N9H<9F9@5H=J9DF9:9F9B79:CF8CA9GH=75B8
:CF9=;B;CC8G5B8588IDHCCB9
)0.39:7:2D
095GGIA95A57FC97CBCA=7DFC8I7H=CB:IB7H=CBC:H<9C66CI;@5G:CFAH<5H
89D9B8GCBHKC=BDIH:57HCFG75D=H5@GHC7?65B8@56CFGIDD@M7
64? 74?
64?
,4? 74?
!9B79H97<BC@C;M=G7CBGH5BHF9HIFBGHCG75@95B8K=H<897F95G=B;A5F;=B5@DFC8I7
H=J=HM=B6CH<:57HCFG-<9DFC8I7H=J=HMD5F5A9H9F,;FCKG9LC;9BCIG@M5HH<97CBGH5BH
F5H92=5B8H<IG=B7CFDCF5H9G@56CF5I;A9BH=B;H97<BC@C;=75@DFC;F9GG
/
/? ,4? 2=,4?
0<=@9@56CF=G;=J9B9LC;9BCIG@M75D=H5@75B69577IAI@5H986M=BJ9GHA9BH
/
/? 64? 494?
748,?0/D9,84.>
F99B<CIG9;5G9A=GG=CBG05F9;9B9F5H985G56MDFC8I7HC:DFC8I7H=CB-<95IHCB
CACIG897F95G9C:9A=GG=CB=BH9BG=HM5H57CBGH5BHF5H9
ν
A5M699B<5B7986M=B
J9GHA9BHG 48 =B565H9A9BH 75D=H5@ 68)5F5A9H9F 4068 89H9FA=B9GH<9 =BJ9GHA9BHG
9::=7=9B7M
04? 4? D4? 9LD
?
4?
684?
/
/? 684? 4068484?
-<97@=A5H9GMGH9A=GF9DF9G9BH98=B5GHM@=N98K5M65G98CB)9HG7<9@!9@89H5@
-<9HCH5@GHC7?C:5HACGD<9F=7;F99B<CIG9;5G9G.0;FCKG8I9HCH<9=BGH5BH5
B9CIG9A=GG=CBGC:5@@7CIBHF=9G
/
/? .0?
05?
5B8=G@=B?98HCH<9;F99B<CIG9;5G7CB79BHF5H=CB.:9.577CF8=B;HC
4.2 Model Structure 101
"0>>8,990?,7.:9:84.#:/077492
/
/? .:9.? .0&
05?&
.:9.?.:9.
-<97CB79BHF5H=CB=BHIFB89H9FA=B9GH<97<5B;9C:;@C65@A95BH9AD9F5HIF9?08;
6M
/
/? ?08;?@C;
.:9.?.:9.
&
?08;??08;
,=A=@5FHC'CF8<5IG5B815B;H9AD9F5HIF97<5B;9G75IG97@=A5H97<5B;9
85A5;9G89GHFCM=B;5:F57H=CB
Ω
C:97CBCA=7CIHDIH
4?
/,84
?08;?
/,84
D4?4?
64? 74?
)=,/0,9/?,=411>
09=ADCG95B=BH9FH9ADCF5@6I8;9H7CBGHF5=BH9B:CF7=B;H<5H9LDCFHJ5@I95B8=A
DCFHJ5@I95F9I@H=A5H9@M65@5B798
5
;45?
8845? /?
5
;45?
CC45? /?
"ADCFHGF979=J986M4:FCA55F989BCH986M8459LDCFHG:FCA4HC56MC45'5HIF5@@M
=ADCFHG5B89LDCFHGH<5H89G7F=69H<9G5A9HF589:@CKAIGH69H<9G5A9<9B79845?C54?
"ADCFHG697CA9:CF9=;B7CBGIADH=CB;CC8G5:H9F=ADCFHH5F=::GY=:5BMY<5J9699B
898I7H98=BH<9:CFAC:=7969F;7CGHG
.
45?
1:=
45
8
45?
?=45? 45 845?
-5F=::F9J9BI9G ?= 5F9F97M7@98K=H<CIHH<97CBGIA9FF95@=N=B;H<9CF=;=BC:H<9
F9J9BI9G 09 7@CG9 H<9 97CBCAM 6M GH5H=B; H<9 D<MG=75@ 6I8;9H 7CBGHF5=BH K<=7<
65@5B79GH<95J5=@56@997CBCA=7CIHDIHK=H<7CBGIADH=CB6CH<=BJ9GHA9BHCDH=CBG5B8
9LDCFHGHCH<9F9GHC:H<9KCF@8
D4? .4? 494? 484?
5
C45?
=B5@@MK9B998HCID85H9H<9FA=B;HCB9EI5H=CBEI5H=CBHC=B7CFDCF5H9H<9
H5F=::F9J9BI9?=
.4?
>/:8
.4?
/:8
5
>5
1:=
.45?
1:= ?=45?
102 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
$ !# #
-<9AC89@ :95HIF9G HKC8=GHCFH=CBGDF9J9BH=B;H<5H7CAD9H=H=J9 9EI=@=6F=IA5B8
GC7=5@D@5BB9FGC@IH=CB7C=B7=897@=A5H97<5B;985A5;9G75IG986M9A=GG=CBG5B8=A
DCFHH5F=::G"BH<=GG97H=CBK989G7F=695B5@;CF=H<AH<5H:=B8G5'5G<9EI=@=6F=IA:CF
GI7<AC89@G
(IF 5DDFC57< HC 7CADIH9 5 7CAD9H=H=J9 9EI=@=6F=IA 6I=@8G CB '9;=G<=
$9<C99H5@5B8%9=A657<5B889B<C:9FCF58=G7IGG=CBC:5@;CF=H<
A=75@H9FB5H=J9GK9F9:9FHC%9=A657<5B889B<C:9F'9;=G<=G<CKGH<5H
57CAD9H=H=J99EI=@=6F=IAA5L=A=N9G5D5FH=7I@5FGC7=5@K9@:5F9:IB7H=CBK<=7<=G5
K9=;<H98GIAC:H<9IH=@=HM:IB7H=CBGC:H<9=B8=J=8I5@7CBGIA9FG!9B79A5L=A=N5H=CB
C:GI7<5GC7=5@K9@:5F9:IB7H=CBA5M69IG98HC7CADIH957CAD9H=H=J99EI=@=6F=IA
,=A=@5F@M$9<C99H5@IG9>C=BHA5L=A=N5H=CBHC7CADIH97CAD9H=H=J99EI=@=6F=5
6IH9LH9B8H<9G7CD9HC97CBCA=9GK=H<9LH9FB5@=H=9G-<9M5B5@MH=75@@M89ACBGHF5H9H<9
9EI=J5@9B79C:5G9HC:CDH=A=N5H=CBDFC6@9AG5B8H<97CAD9H=H=J99EI=@=6F=IA
-C:=B85B9EI=@=6F=IAK9=H9F5H9=B8=J=8I5@K9@:5F9A5L=A=N5H=CB:CF5@@D@5M9FG=B
588=H=CBHC5A5L=A=N5H=CBC:5;;F9;5H9GC7=5@K9@:5F95B5DDFC57<G=A=@5FHCH<9CB9
DFCDCG986M%9=A657<5B889B<C:9F098CGC6M:=L=B;J5F=56@9GC:H<9CDH=
A=N5H=CBDFC6@9AG5HDF9J=CIG@M89H9FA=B98@9J9@G
49/492,$,>30<@474-=4@8
-CGC@J9CIFAC89@:CF5'5G<9EI=@=6F=IAK9F9D95HH<9:C@@CK=B;H<F99GH9DGIBH=@
7CBJ9F;9B79=GF957<98
(?0;
09GH5FH6M:=B8=B;5'5G<9EI=@=6F=IA=B9A=GG=CBG0O0?P0?0?0$?
K<=7<5F989H9FA=B986MH<9=BJ9GHA9BH897=G=CBG=BDFC8I7H=CB75D=H5@495B8
565H9A9BH75D=H5@48=9K9GC@J95:=LDC=BHDFC6@9A00K<9F9=GH<9
G9@:=BH9F9GH98F9GDCBG9C:D@5M9FGHCCH<9FD@5M9FG9A=GG=CBHF5>97HCF=9G09
7CADIH96MGC@J=B;
4A5L
494? 484?
B071,=04
GI6>97HHC EI5H=CBG
5B884
5
?84
5
?C
4
5
?C4
5
?0
6?06? :CF 64
K=H<HF589:@CKG 845? 5B8 C45? 5B8CH<9FD@5M9FG9A=GG=CBG 06? :=L98HCH<9=F
DF9J=CIG@9J9@G5G=B8=75H986MH<965FG
(?0;
'9LHK9G95F7<:CF57CAD9H=H=J99EI=@=6F=IA=BHF589:@CKG8CK=H<8
O8?P8?845?5B8COC?PC?C45?K<=@9?99D=B;H<99A=GG=CB9LH9FB5@=HM
:=L985HH<9@9J9@
0
:CIB8=B,H9D-<=G=G8CB96MGC@J=B;H<9:=LDC=BHDFC6
@9A ?=?=K=H< ?= O?=?P ?=??=45?5B8 H<9F9GDCBG9C:H<9GC7=5@
D@5BB9FHC5;=J9BH5F=::F9J9BI97CBGHF5=BH
?=
=G7CADIH986MGC@J=B;H<9
4.3 Solving for a Nash Equilibrium 103
"0>>8,990?,7.:9:84.#:/077492
>C=BHCDH=A=N5H=CB
A5L
494? 484? 845? C45?
4B071,=04
GI6>97HHC EI5H=CBG
5B804?04? ?=?=
-<9D5F5A9H9FG
δ
4 F9DF9G9BHH<9F9;=CBGK9=;<HGK=H<=BH<9>C=B98GC7=5@
K9@:5F9:IB7H=CB5B85F95@GCF9:9FF98HC5G)5F9HCCF'9;=G<=K9=;<HG
(?0;
MIG=B;DF=79=B:CFA5H=CB89F=J98:FCAH<9%5;F5B;9AI@H=D@=9FGC:H<9
A5L=A=N5H=CBDFC6@9AK989H9FA=B989:=7=HG5B8GIFD@IG9G=BH<9=BH9FH9ADCF5@
6I8;9H 7CBGHF5=BHG EI5H=CB 09 65@5B79 H<9 6I8;9HG 6M 58>IGH=B; H<9
K9@:5F9K9=;<HG
δ
45B8F9D95H=B;GH9DG
CBJ9F;9B79=GF957<98K<9BH<9=BH9FH9ADCF5@6I8;9H=G=B65@5B795B8H<9:=L
DC=BH9EI5H=CBG=BGH9DG5B85F9G5H=G:=98
$@80=4.,7A0=414.,?4:9:1?30$,>30<@474-=4@8
09J9F=:MH<9F9GI@H=B;75B8=85H9'5G<9EI=@=6F=IAGHF5H9;=9G=B9A=GG=CBG5B8
HF589BIA9F=75@@M6M7CAD5F=B;H<9AHCH<9F9GI@HGC:H<9:C@@CK=B;A5L=A=N5H=CB
DFC6@9AG
4A5L
494? 484? 845? C45?
B071,=04
GI6>97HHC EI5H=CBG
5B8DF=79G ;45?
8;45?
C
K<=7<=B7@I89H<96I8;9H9EI5H=CBK=H<A5F?9HDF=79G:FCAH<9:=B5@AC89@
GC@IH=CB9J=5H=CBGC:H<=GAC89@:FCACIFGC@IH=CBG<CI@869K=H<=BH<9CF89FC:
A5;B=HI89C:BIA9F=75@577IF57MCB@MK<=7<=GK<5HK9:=B8BCHG<CKB"BD5FH=7I
@5FG=AI@H5B9CIG7@95F5B79C:5@@=BH9FB5H=CB5@A5F?9HG7CB:=FAGH<9'5G<9EI=@=6F=IA=B
=BH9FB5H=CB5@HF589
&,=?4,72=00809?$,>3<@474-=4,
CFH<95DD@=75H=CBC:H<=G5@;CF=H<AHCG9@:9B:CF7=B;"BH9FB5H=CB5@BJ=FCBA9BH5@
;F99A9BHG"K9B998HC9LH9B8H<95@;CF=H<A:FCAD@5=B'5G<9EI=@=6F=IAHC
)5FH=5@;F99A9BH'5G<EI=@=6F=IA)'0<9F95G=BH<9'5G<9EI=@=6F=IAH<9F9=G
BC7CCD9F5H=CB)'89:=B9GD5FH=5@7CCD9F5H=CB5GGC7=5@@MCDH=A5@69<5J=CF5ACB;5
GI6G9HC:D@5M9FGH<97C5@=H=CB)'=G5'5G<9EI=@=6F=IAC:H<97C5@=H=CB57H=B;5G
CB9 D@5M9F 5B85@@BCBA9A69FG 0=H<=BH<97C5@=H=CB5 IH=@=H5F=5B GC7=5@ K9@:5F9
'CH9H<5HK98CBCH5HH9ADHHCG<CKIB=EI9B9GGC:H<9=89BH=:=989EI=@=6F=IA"B8998$9<C99H5@
89ACBGHF5H9<CK;9B9F5@9EI=@=6F=IAAC89@G5F9DFCB9HCAI@H=D@=7=HM=BH<9DF9G9B79C:9LH9F
B5@=H=9G!CK9J9FH<9M5@GCG<CKH<5HH<=GC77IFGK<9BH<99LH9FB5@=HM=GF5H<9F@5F;9"BCIF75G9
K<9F9H5F=::G5B87@=A5H985A5;9G5F9CBH<9G75@9C:D9F79BHG5B8H9BD9F79BHF9GD97H=J9@MK9
5GGIA9H<5HH<9=GGI9C:AI@H=D@99EI=@=6F=5=GGH=@@B9;@=;=6@9-<=G=G7CFFC6CF5H986MH<9:57HH<5HCIF
BIA9F=75@G=AI@5H=CBGDFC8I798FC6IGHF9GI@HGK=H<CIH=B8=75H=CBC:AI@H=D@99EI=@=6F=5
104 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
:IB7H=CB=9H<99EI5@@MK9=;<H98GIAC:5@@=B8=J=8I5@K9@:5F9:IB7H=CBG=GA5L=
A=N98
""" " ""
"BH<=GG97H=CBK95DD@MCIFAC89@HCH<95B5@MG=GC:=ADCFHH5F=::G5G5HF589G5B7
H=CB 5;5=BGH BCBG=;B5HCF=9G C: 5B "BH9FB5H=CB5@ BJ=FCBA9BH5@ ;F99A9BH "
C@@CK=B; H<9 @=H9F5HIF9 CB G9@:9B:CF7=B; " 9; 5FF5FC 5B8 ,=B=G75@7C
5FF9HH K9 7CBG=89F 7C5@=H=CBG H<5H 5F9 =BH9FB5@@M 5B8 9LH9FB5@@M GH56@9 =9
A9A69FGC:H<97C5@=H=CB75BBCH=ADFCJ9H<9=FG=HI5H=CB6M@95J=B;H<97C5@=H=CB5B8
>C=B=B; H<9 ;FCID C: BCBA9A69FG K<=7< :F99F=89 CB H<9 9::CFH C: H<9 F9A5=B=B;
7C5@=H=CB5B8B9=H<9F8CBCBA9A69FG<5J95B=B79BH=J9HC>C=BH<97C5@=H=CB
-C5JC=8H<96@57?6CL9::97H5B8HC:57=@=H5H95B=BH9FDF9H5H=CBC:H<9EI5@=H5H=J9
9::97HGDFC8I7986MH<9AC89@K9F9GHF=7HH<9:C@@CK=B;5B5@MG=GHCH<9GMAA9HF=775G9
C:B=B9D9F:97H@M=89BH=75@7CIBHF=9G
'0>@7?>
),=411>[email protected]:9&,=?4.4;,?4:9
(IFAC89@7CB:=FAGH<5HH5F=::G5F9DCH9BH=5@@M5B9::97H=J9=BGHFIA9BHHC=B7F95G9H<9
G7CD9:CF=BH9FB5H=CB5@7CCD9F5H=CBD5FH=7=D5H=CB=BH<97C5@=H=CB697CA9GIB5A6=;I
CIG@M<=;<9FK<9B5H5F=::CB=ADCFHG:FCABCBA9A69F7CIBHF=9G=G5DD@=98-<=GF9GI@H
=G=@@IGHF5H98=B=;IF9=BH<956G9B79C:H5F=::GH<9@5F;9GHGH56@97C5@=H=CB<5GCB@M
H<F99CF:CIFA9A69FGK<=@95H5F=::F5H969HK99BHCD9F79BH=GGI::=7=9BHHC=B8I79
:I@@7CCD9F5H=CB
-<=G9::97H75B69IB89FGHCC8=BH<9@=;<HC:H<9AC89@GF9DF9G9BH5H=CBC:=BH9F
B5H=CB5@HF589=BK<=7<957<F9;=CBDFC8I79G5B=AD9F:97H@MGI6GH=HIH56@9;CC85B8
<9B798=GDCG9GYBCH5HH<9:=FA6IH5HH<97CIBHFM@9J9@YC:GCA9A5F?9HDCK9F"B
"BH<=G7CBH9LHA5F?9HDCK9F=GHC69IB89FGHCC85G5B5;;F9;5H9DFCD9FHMC:K<C@97CIBHF=9G5B8=G
42@=0",=20>?>?,-70.:,74?4:9>1:=,24A09
?,=411
42@=0'07,?4A0;=4.0:1.:,74?4:92::/>
4.4 Application to International Cooperation 105
"0>>8,990?,7.:9:84.#:/077492
9::97H5GA5@@H5F=::CB=ADCFHG:FCABCB7C5@=H=CBA9A69FG9LD@C=HGH<=GA5F?9HDCK9F
5B8@958GHC5F=G9=BH<9F9@5H=J9DF=79C:;CC8GDFC8I7986M7C5@=H=CBA9A69FGG99
=;IF9-<9@5HH9FC6H5=B5B9H69B9:=H:FCAH<=GDCG=H=J9H9FAGC:HF5899::97HG=A=@5F
=B=HGA97<5B=7GHCK<5H=G?BCKB:FCAH<95B5@MG=GC:CDH=A5@H5F=::GCFACBCDC@=GH=7
DF=7=B;,=B796M5GGIADH=CBCB@M7C5@=H=CBA9A69FG75B5DD@MGI7<5H5F=::=H
7CBGH=HIH9G5B=B79BH=J9HC>C=BH<97C5@=H=CB
'CH9H<5HH<9F9@5H=J9DF=79C:7C5@=H=CB;CC8G5@GCF=G9G>IGH5G5:IB7H=CBC:H<9G=N9
C:H<97C5@=H=CB9J9B=BH<956G9B79C:5BMH5F=::=;IF95H
τ
-<=G<5DD9BG69
75IG9H<99A=GG=CB7IHGF95@=N986M7C5@=H=CB7CIBHF=9G8=A=B=G<H<9=FCIHDIH5B8<9B79
H<9F9=GYK=H<F9GD97HHCH<96IG=B9GG5GIGI5@Y5F98I798GIDD@MC:7C5@=H=CB;CC8G":
89A5B8=G=B9@5GH=7
σ
Α
H<9F9@5H=J9DF=79AIGH7CBG9EI9BH@M;CID"B:57HH<9
DCGG=6=@=HMHCD5GGCBA=H=;5H=CB7CGHGHC:F99F=89FGJ=5GI7<H9FAGC:HF5899::97HG5@GC
9LD@5=BG<CK@5F;9F7C5@=H=CBG75B69]GH56=@=N98^9J9BK=H<CIHH5F=::G6MG=AD@M89
7F95G=B;H<99@5GH=7=HMC:GI6GH=HIH=CBHC5GI::=7=9BH@M@CK@9J9@5GG99B=B=;IF95H
τ
-<9;F5D<=B=;IF95@GCG<CKGH<5HH<99::97H=J9B9GGC:H5F=::G=GF98I798=BH<9
DF9G9B79C:<=;<9F9@5GH=7=H=9GC:GI6GH=HIH=CBCF9L5AD@95H5F=::C:D9F79BH=B8I79G
5GH56@97C5@=H=CBK=H<G=LCIHC:B=B9A9A69F7CIBHF=9GK<9B
σ
Α
:=J9A9A69FG
K<9B
σ
Α
5B8:CIFA9A69FGK<9B
σ
Α
,=B795<=;<9F9@5GH=7=HM=AD@=9G<=;<9F
GI6GH=HIH56=@=HM5B8<9B79@CK9FA5F?9HDCK9FH<=G69<5J=CF=G:I@@M7CBG=GH9BHK=H<CIF
9LD@5B5H=CB"B8998=B75G95@@;CC8G5F9D9F:97HGI6GH=HIH9G
σ
Α
H<9H5F=::@CG9G=HG
7@CIH9BH=F9@M5G9LD97H98
9A4=:9809?,7110.?4A090>>:1::;0=,?4:9
7CAACB5F;IA9BH6FCI;<H:CFK5F85;5=BGH7@=A5H97C5@=H=CBGK=H<=B7CAD@9H9
A9A69FG<=D=GH<9@95?5;9DFC6@9AH<99::97H=J9B9GGC:5BM7C@@97H=J99::CFH6MH<9
7C5@=H=CB7CI@869IB89FA=B98=:BCH5BB=<=@5H986M:F99F=89FGK<C=B7F95G9H<9=F
9A=GG=CBG=BF9GDCBG9HCH<97C5@=H=CBGF98I7H=CBGG=;IF9=@@IGHF5H9GH<99LHF9A9
75G9C:D9F79BH@95?5;9F5H9=GBCHDF9G9BH=BCIFAC89@"BGH958K9C6G9FJ9H<5H5B
=B7F95G9=BH<97C5@=H=CBG=N9IB5A6=;ICIG@MF9GI@HG=B5F98I7H=CBC:7IAI@5H=J9;@C65@
9A=GG=CBGF99F=8=B;8C9G75IG9GCA9@95?5;96IHH<99LH9BH=G@=A=H985B8KCI@8BCH
K5FF5BHH<98=G7CIF5;9A9BHC:7CCD9F5H=CB69HK99B5GI6G9HC:7CIBHF=9G=;IF9
-<9A=GG=B;=B8=75H=CBC:H<9D5F5A9H9FJ5@I9G:CF
τ
5B8
σ
Α
=B=;IF9G5B8<=BHG
5H5BCH<9F69<5J=CF5@7<5F57H9F=GH=7C:H<9AC89@9A=GG=CBHF5>97HCF=9G5F9:I@@M89H9F
A=B986MH<97C5@=H=CBG=N95B88CBCH89D9B8CBH<9FA=B;HCB9@5GH=7=H=9GCFH<9H5F=::
F5H9 )9F<5DG 7CIBH9F=BHI=H=J9 H<=G C6G9FJ5H=CB =G57HI5@@M =B @=B9 K=H< H<9 AC89@
5GGIADH=CBGK989:=B98IH=@=HM5GH<9@C;5F=H<AC:5@=B95F@M<CAC;9B9CIG:IB7H=CB
K<=7<6MIG=B;H<9=B8=F97HIH=@=HM:IB7H=CB5B85B9L57HDF=79=B89L75B69F9KF=HH9B5G
5GIAC:HKCH9FAGH<9:=FGHF9@5H98HCH<9CIHDIH@9J9@5B8H<9G97CB8HCH<9F9@5H=J9
8I9HCH<9:57HH<5H957<7CIBHFM^GF9DF9G9BH5H=J9CIHDIH6IB8@9=GGCA9K<5H8=::9F9BH!CK9J9FH<9F9
=GBCACBCDC@=GH=7A5F?9HGHFI7HIF95GGI7<G=B79H<9:=FAGA5?=B;ID957<7CIBHFM^G97CBCAM
5@K5MG69<5J97CAD9H=H=J9@M"B:57H,77'5G<9EI=@=6F=5=BH<=GGHI8MF9DF9G9BH7CAD9H=H=J99EI=@=6F=5
65G98CBDF=79H5?=B;69<5J=CF
-<97C5@=H=CBG^GH56=@=HMC:7CIFG989D9B8GCBH<9=FJ5@I9
106 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
DF=79G5B8H<99@5GH=7=HMC:GI6GH=HIH=CB)F=797<5B;9G=B8I7986M5H5F=::CF57<5B;9=B
σ
Α
<5J95B=B:@I9B79CB@MCBH<9@5HH9F 6IH 8C BCH 7<5B;9 H<9 CDH=A5@ 75D=H5@
577IAI@5H=CB5B85G58=F97H7CBG9EI9B79CIHDIH@9J9@G5B89A=GG=CBGF9A5=BH<9
G5A9
=0/4-474?D:1),=411>
-<F95H9B=B;HC=ADCG9H5F=::G=GCB@M7F98=6@9=:H<97C5@=H=CB=G69HH9FC::K=H<H<5B
K=H<CIHH5F=::G 0=H<=BCIFAC89@7<5F57H9F=N986MB5H=CB5@DFC8I7H8=::9F9BH=5H=CB
H5F=::GDFCJ=895B=B8=F97HA95BG:CF7C5@=H=CB7CIBHF=9GHC9LD@C=HH<9=F=AD@=7=HA5F?9H
DCK9F-<IG5H5F=::G<CI@86969B9:=7=5@5G@CB;5G=H=GBCHHCC<=;<H<9@=A=H89D9B8
=B;CBH<99@5GH=7=HMC:GI6GH=HIH=CB-<=G=BHI=H=CB=G7CB:=FA98=B=;IF9K<=7<G<CKG
<CK57C5@=H=CBGK9@:5F97<5B;9GK=H<=B7F95G=B;H5F=::G
G9LD97H98K9@:5F9=B=H=5@@M=B7F95G9G6IHGH5FHGHC897@=B95:H9FF957<=B;5A5L=
AIAJ5@I95B89J9BHI5@@M8FCDG69@CKN9FC-<9H<F9G<C@8J5@I95HK<=7<H<9K9@:5F9
9::97H697CA9GB9;5H=J9A5F?GH<9A5L=AIAH5F=::F5H9H<5H=GGH=@@7F98=6@9
@H<CI;< H<9 C6G9FJ98 EI5@=H5H=J9 D5HH9FB =G FC6IGH K=H< F9GD97H HC D5F5A9H9F
7<5B;9GH<9GD97=:=7J5@I9C:H<9A5L=AIAH5F=::5GK9@@5GH<9DCH9BH=5@K9@:5F9;5=B
89D9B8CBH<99@5GH=7=HMC:GI6GH=HIH=CB
σ
Α
5B8CBH<97C5@=H=CBG=N96CH<=B7F95G9K=H<
@CK9F9@5GH=7=H=9G5B8GA5@@9F7C5@=H=CBG=N9GCF9L5AD@95H
σ
Α
H5F=::F5H9GC:@9GG
H<5BD9F79BH5F97F98=6@9:CF5BM7C5@=H=CBG=N9K<=@95H
σ
H<97IHC::=G
5@F958M5H56CIHD9F79BH-<=G89D9B89B79CB
σ
75B5;5=B699LD@5=B98=BH9FAGC:
H<9;F95H9FA5F?9H=B:@I9B79H<5H75B69F95@=N98K=H<5@CK9F9@5GH=7=HM-<9C6G9FJ56@9
<=;<9FK9@:5F9;5=B:CF>8,770=7C5@=H=CBG=G57CBG9EI9B79C:<=;<9FH5F=::F9J9BI9G=B
H<9DF9G9B79C:@5F;97C5@=H=CBGH<9F95F9CB@M:9K:F99F=89FG@9:HK<CG9;CC8G5F957HI
5@@MGI6>97HHCH5F=::8IH=9GK<=@9H<9F95F9D5MA9BHG:FCA5@ACGH5@@HF58=B;D5FHB9FG=:
H<97C5@=H=CB<5GCB@MHKCA9A69FG
-<=G7CB79DHC:7F98=6=@=HM=GF5H<9FG<CFHG=;<H98K<9B7CBG=89F=B;CB@MH<9K9@:5F99::97HGC:H5F=::G
CBH<9AG9@J9G7C5@=H=CBA9A69FG=;BCF9H<5HH5F=::GA5M=B7F95G9D5FH=7=D5H=CB5B8H<IG6F=B;56CIH
B9HDCG=H=J9K9@:5F99::97HG9J9BK<9B]=B7F98=6@9^577CF8=B;HCH<=G7CB79DH-<=GG<CFHG=;<H98B9GG=G
<CK9J9F7CBG=GH9BHK=H<H<99AD@CM98G<CFHG=;<H987CB79DHC:GH56=@=HM
42@=0110.?:1.:,74?4:91:=8,?4:9:9?:?,7
.@8@7,?4A0084>>4:9>
42@=0A0=,201=00=4/0=,9/.:,74?4:9808
-0=084>>4:9>,>1@9.?4:9:1.:,74?4:9>4E0
4.4 Application to International Cooperation 107
"0>>8,990?,7.:9:84.#:/077492
+071,=08;74.,?4:9>:1),=411>
-5F=::G<5J95B5A6=;ICIG9::97HCB;@C65@K9@:5F9CBH<9CB9<5B8H<9M75B=B
7F95G9;@C65@K9@:5F96975IG9H<9M9B<5B79H<9G7CD9:CF7CCD9F5H=CB(BH<9CH<9F
<5B8Y5G:F99HF58958JC75H9GA=;<HC6>97HYH<9M8=GHCFH:F99HF5895B8H<IGIB89FA=B9
;@C65@9::=7=9B7MK<=7<CI;<HHC75IG95@CGGC:K9@:5F9K<=7<7CI@8=BH<9KCFGH75G9
CIHK9=;<H<9;5=BG097CAD5F9H<9HKCCDDCG=B;9::97HG=B=;IF9G5B8
=;IF9G<CKG;5=BG=B8I7986MH5F=::GA95GIF985GH<98=::9F9B79=B;@C65@K9@:5F9
69HK99BH<9@5F;9GHGH56@97C5@=H=CBK=H<5;=J9BH5F=::F5H95B8H<9@5F;9GHGH56@97C5@=
H=CB=BH<956G9B79C:H5F=::GG75B69G99BH<9K9@:5F9;5=BG5F9EI=H9G=;B=:=75BH
5B8F957<IDHCHCD9F79BH:CF:I@@7CCD9F5H=CB89D9B8=B;CBH<97C5@=H=CBG=N9
5B87CFF9GDCB8=B;K9@:5F9@9J9@GK=H<CIHH5F=::GG99=;IF9
"B7CBHF5GHH<9K9@:5F9@CGG9G75IG986MH<98=GHCFH=CB5FM9::97HGC:H5F=::G5F9G<CKB
=B=;IF9-<9M5F9A95GIF986MH5?=B;H<9@5F;9GHGH56@97C5@=H=CB5H957<H5F=::F5H9
5B87CADIH=B;H<9=B7F95G9=B;@C65@K9@:5F957<=9J986M8FCDD=B;5@@H5F=::G=;BCF=B;
H<5HH<97C5@=H=CBA5MBCH69GH56@95BMACF9"B5;F99A9BHK=H<GH5B85F897CBCA=7
H<9CFMH<9;F5D<G<CKGK9@:5F9@CGG9GH<5H=B7F95G9GH958=@MK=H<H<9H5F=::F5H9!CK
9J9FH<9K9@:5F9@CGG9G8I9HCH<9HF5898=GHCFH=CB5F9CB9CF89FC:A5;B=HI89GA5@@9F
H<5BH<9;5=BG57<=9J986M:IFH<9F=B;7CCD9F5H=CB"BBCFA5H=J9H9FAGH<=GGI;;9GHGH<5H
H<9HF5898=GHCFH=B;9::97HC:H5F=::GG<CI@8695B5779DH56@9DF=79HCD5M=B9L7<5B;9:CF
ACF9=B7@IG=J97@=A5H97C5@=H=CBG
'CFA5@=N98=B6CH<:=;IF9GHCH<9G75@989:=B986MH<9K9@:5F9;5D69HK99BH<9'5G<9EI=@=6F=IA5B8
GC7=5@CDH=AIA
"HA=;<HG99A7CIBH9F=BHI=H=J9H<5HK9@:5F9@CGG9G=B=;IF95F9<=;<9FK<9B;CC8G5F969HH9FGI6
GH=HIH9G9GD97=5@@MG=B79=BH<9@=A=H75G9
σ
H5F=::G697CA9=B9::97H=J95B8<9B79K9@:5F9@CGG9G
8FCDHCN9FC-<9=BHI=H=CB69<=B8H<=G9::97H=G5G:C@@CKG-5F=::G<5J9HKC9::97HG5B=B7CA99::97H
5B85GI6GH=HIH=CB9::97H-<9=B7CA99::97H8I9HCH<9DF=79=B7F95G9C:7C5@=H=CB;CC8G=GDF98CA
=B5BH@MC:8=GHF=6IH=CB5@B5HIF9@95J=B;;@C65@K9@:5F9@5F;9@MIB5::97H98-<9GI6GH=HIH=CB9::97HCB
H<9CH<9F<5B875IG9G5897@=B9=BH<9HCH5@JC@IA9C:KCF@8HF589K<=7<695FGK9@:5F97CGHG:CF5@@
7CIBHF=9G-<=G89J=5H=CB:FCAH<9GC7=5@@MCDH=A5@HF589JC@IA9=B7F95G9GK=H<<=;<9F9@5GH=7=H=9GC:
GI6GH=HIH=CB5B8H<IG697CA9GACF9DFCBCIB798:CF@5F;9J5@I9GC:
σ
42@=0=0/4-474?D:148;:>492?,=411>
0 2 4 6 8 10 12 14 16
−1
0
1
coal. size 2
coal. size 8
σ
A
= 20
0 2 4 6 8 10 12 14 16
−1
0
1
Welfare gain for coalition in percent
σ
A
= 40
0 2 4 6 8 10 12 14 16
−1
0
1σ
A
= 100
Tariff rate in percent
108 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
(09>4?4A4?D9,7D>4>
79BHF5@F9GI@H=BH<9DF9J=CIGG97H=CBK5GH<5H5H5F=::@9J=98CB=ADCFHG:FCA:F99
F=89F7CIBHF=9G=BH<9CF89FC:A5;B=HI89C:5:9KD9F79BHGIGH5=BG:I@@7CCD9F5H=CBCB
9A=GG=CBGF98I7H=CB"BH<=GG97H=CBK99LD@CF9=B<CK:5FH<=GF9GI@H7CBH=BI9GHC<C@8
K<9BH<9J5@I9GC:H<9AC89@G?9M=BDIHD5F5A9H9FG5F9GMGH9A5H=75@@M7<5B;98HC3423
A,7@05B87:BA,7@09GH=A5H9G"BCF89FHC?99DH<97CADIH5H=CB5@7CGHGA5B5;956@9K9
GH=7?HC5B9LD@CF5H=CBC:@C75@G9BG=H=J=H=9G
=;IF9F9DCFHGG9BG=H=J=H=9GC6H5=B98:FCAH<9J5F=5H=CBC:B=B9D5F5A9H9FG"B8=75H
985F9H<9@CK9GHH5F=::F5H9GH<5HGH=@@GIDDCFH:I@@7CCD9F5H=CB:CFH<97<CG9BD5F5A9H9F
J5@I9G-<9BIA9F=75@J5@I9G:CF<=;<5B8@CK5F9F9DCFH98B9LHHCH<985H5DC=BHK<=@9
H<9D5F5A9H9FGB5A95B889:5I@HJ5@I9=G;=J9B5HH<96CHHCAC:H<9:=;IF9-<9F9GI@HG
G<CKH<5H:CF5@@D5F5A9H9FJ5F=5H=CBG:I@@7CCD9F5H=CB75BGH=@@6957<=9J986M58>IGH=B;
H<9H5F=::F5H9IFH<9FACF9H<9F9EI=F98H5F=::F5H98C9GBCH9L7998:=J9D9F79BH:CFCIF
G9@97H=CBC:@CK5B8<=;<J5@I9G
5FF9HH^G7CB7@IG=CBH<5H7CCD9F5H=CB=G<5F89FHC57<=9J9K<9B=H=GACGH
B99898<9@DGHCIB89FGH5B8H<9G9BG=H=J=H=9G-<9@5F;9GH=AD57H=G9L9FH986MH<9F5H9C:
DIF9H=A9DF9:9F9B79
ρ
, K<=7<=G?BCKBHC<5J95GHFCB;=AD57HCB;FCKH<5B8H<9
5GGC7=5H989A=GG=CBGD5H=9B796CCGHGG5J=B;G@958=B;HCACF9DFC8I7H=CB88=H=CB5@
@MH<9K9=;<HC::IHIF985A5;9G=G=B7F95G98/5FM=B;D5F5A9H9FGC:H<985A5;9:IB7
H=CB=AA98=5H9@M@9GG9BGCF9L579F65H9GH<9B998:CF7CCF8=B5H98A=H=;5H=CB@GCH<9
B9LH HKC ACGH G9BG=H=J9 D5F5A9H9FG H<9 9LC;9BCIG F5H9G C: 8975F6CB=N5H=CB
ν
5B8
DFC8I7H=J=HM;FCKH<2=5F95;5=B7@CG9@MF9@5H98HC9A=GG=CBG5B897CBCA=7;FCKH<5B8
H<9F9:CF9H<9IF;9B7MC:9BJ=FCBA9BH5@7CCD9F5H=CB
"B588=H=CBHCH<9@C75@G9BG=H=J=HM5B5@MG=GK95@GC9LD@CF9H<97CBG9EI9B79GC:5
GHFI7HIF5@7<5B;9=BH<9AC89@=BEI5H=CBK95GGIA989LC;9BCIGH97<BC@C;=75@
(IF5DDFC57<=GG=A=@5FX5@69=HAI7<ACF97CB7=G9XHCH<9G9BG=H=J=HM5B5@MG=GC:H<9"AC89@=B
'CF8<5IG<)5F5A9H9FJ5F=5H=CBG@958=B;HC'CF8<5IG^5@H9FB5H=J9<=;<J5@I9G5F97CA
D5F56@9HCCIFG&CF9CJ9F:=J9C:H<99=;<H=89BH=:=98ACGHG9BG=H=J9D5F5A9H9FG<5J97CIBH9FD5FHG=B
CIF5B5@MG=GGCB98=::9F9B79=BCIFGHI8MH<9IB79FH5=BHMC:7@=A5H98MB5A=7G=GGC@9@M5GG9GG986M
J5FM=B;H<985A5;9:IB7H=CB
42@=0,49>4927:-,7B071,=0/@0?:?,=411> 42@=0":>>0>4927:-,7B071,=0/@0?:
?,=411>
4.4 Application to International Cooperation 109
"0>>8,990?,7.:9:84.#:/077492
DFC;F9GG5HH<97CBGH5BHF5H92=@H9FB5H=J9@MK9A=;<H:C@@CKH<97CB79DHC:#CB9G
5B80=@@=5AG5B889D=7HH<9DFC8I7H=J=HMD5F5A9H9F,5G5?BCK@98;9GHC7?H<5H
9JC@J9G9B8C;9BCIG@M577CF8=B;HC
/
/? ,4? 2=
40,4,4?
,4?
-<9B9K7CBHFC@J5F=56@94,F9DF9G9BHG+=BJ9GHA9BHG40,H<9=F9::=7=9B7M5B8
λ
5B8
φ
D5F5A9H9FG:CFZGH9DD=B;CBHC9G[5B8ZGH5B8=B;CBG<CI@89FG[9::97HGF9GD97
H=J9@M -CH9GHH<9=B:@I9B79C:9B8C;9BCIGH97<BC@C;=75@7<5B;9K97<CCG9 40,
0
λ
5B8
φ
K<=7<F9DFC8I79GH<95J9F5;9;FCKH<F5H9C:H<989:5I@H
AC89@K=H<9LC;9BCIGH97<BC@C;=75@7<5B;9-<9@5HH9F75G9=GF97CJ9F98:FCAEI5H=CB
6MG9HH=B;
λ
φ
-<9=AD57HC:H<=GGHFI7HIF5@7<5B;9=GBC@5F;9FH<5BH<9
D5F5A9H9FJ5F=5H=CBGG99@5GH7C@IAB=B=;IF9
"BH<9A5=BD5FHC:H<=GD5D9FK9F9GHF=7HH<95B5@MG=GHCGMAA9HF=7F9;=CBG-<=G
;F95H@MF98I79GH<9BIA69FC:7CADIH5H=CBGB99898HC89H9FA=B9H<9@5F;9GHGH56@97C5@=
H=CB:CF9GMAA9HF=7F9;=CBG9AC89@9J5@I5H=CBGGI::=79=BCIF75G9K<9F95G9
<9H9FC;9B9CIGF9;=CBGF9EI=F999AC89@FIBG=BCIF75G9"B-56@9K9H5?9CB9
GH9DHCK5F8G<9H9FC;9B9CIGF9;=CBG6M9LD@CF=B;H<9=AD57HC:ZGHM@=N98[<9H9FC;9B9=HM
-CH<=G9B8K989:=B9H<F998=::9F9BHG79B5F=CGK=H<<9H9FC;9B9CIGD5F5A9H9FG
=FGHG79B5F=CFCK=B7CFDCF5H9G<9H9FC;9B9=HM6M5GG=;B=B;957<F9;=CB58=:
:9F9BH5ACIBHC:=B=H=5@75D=H5@6G75B69G99B9J9BH<CI;<H<9DCCF9GH5B8F=7<9GH
F9;=CBG8=::9F6M5:57HCFH<99::97HCBH<9H5F=::F5H9B99898HC=B8I79:I@@7CCD9F5
H=CB=G5@@6IHB9;@=;=6@9"B89987CCD9F5H=CB697CA9G5@=HH@995G=9F
!9H9FC;9B9=HMG<CI@87CBGH=HIH95ACF9G9F=CIGC6GH57@9HC7CCD9F5H=CBK<9BH<9F9
(:7CIFG9H<9G9=BJ9GHA9BHGB998HC69898I7H98:FCAH<96I8;9H=BEI5H=CB
,99#CB9G5B80=@@=5AG:CF589H5=@988=G7IGG=CBC:H<99EI5H=CB
42@=0":.,7>09>4?4A4?D,9,7D>4>)30142@=0>3:B>3:B?30?,=411=,?090.0>>,=D?:
49/@.01@77.::;0=,?4:9.3,920>B30960D49;@?;,=,80?0=>,=0=0;7,.0/-D7:B0=:=
34230=A,7@0>
0
0.01
0.02
0.03
0.04
0.05
0.03
0.001
2
1
0.04
0.01
0.02
0.005
0.03
0.01
0.4
0.2
−0.1
−0.3
10
2.5
0.2
Default parameter values
0.01 1.5 0.02 0.01 0.023 0.3 −0.2 5 0
Tariff rate (τ)
ρdam2 dam1 νgr
export ratio γiekm λ
low value
high value
default
110 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
5F9GCA9F9;=CBGK=H<<=;<85A5;9G5B8<=;<A=H=;5H=CB7CGHG<=;<=BH9F9GH=B7CCD9F5
H=CB5B8GCA9K=H<@CK85A5;9G5B8@CKA=H=;5H=CB7CGHG@CK=BH9F9GH=B7CCD9F5H=CB
-<=G<MDCH<9G=G=GH9GH98=BG79B5F=CG5B8G<CKB=BFCKGAC89F5H9<9H9FC
;9B9=HM5B8GHFCB;<9H9FC;9B9=HMK<9F9H<985A5;95B8A=H=;5H=CB7CGHD5F5A9
H9FG<5J9K9F9G9H577CF8=B;@M
09:=B8H<5HH<=GHMD9C:<9H9FC;9B9=HM8C9GBCHDF9J9BH:I@@7CCD9F5H=CB9=H<9F9J9B
H<CI;<<=;<9FH5F=::F5H9G5F9B979GG5FM0<9H<9FH<9=B7F95G98@9J9@C:H5F=::G=G8I9HC
<9H9FC;9B9=HMF9A5=BG5BCD9BEI9GH=CB6CH<H<985A5;9G5B8A=H=;5H=CB7CGHG5F9
89H9FA=B98H<FCI;<BCB@=B95F:IB7H=CBG!9B799J9BH<CI;<K9J5F=98H<9D5F5A9H9FG
GI7<H<5HH<9=F5J9F5;9J5@I957FCGG5@@7CIBHF=9GF9A5=BGH<9G5A95J9F5;985A5;9G
5B85J9F5;9A=H=;5H=CB7CGHGA5MK9@@<5J97<5B;988I9HCH<9=BHFC8I7H=CBC:<9H9FC
;9B9=HM
#!!
-<=G GHI8M A5?9G 5 A9H<C8C@C;=75@ 5B8 5 DC@=7M7CBHF=6IH=CB HCH<9 =BH9;F5H98
5GG9GGA9BHAC89@=B;C:7@=A5H97<5B;909DF9G9BH5AC89@=BH<9HF58=H=CBC:AI@H=
F9;=CB5@ CDH=A5@ ;FCKH< AC89@G H<5H =B7@I89G HF589 F9@5H=CBG<=DG 69HK99B F9;=CBG
"B7@I8=B;7@=A5H985A5;9G5B8DIB=H=J9H5F=::G=BHFC8I79GHKC9LH9FB5@9::97HG=BHCH<9
AC89@-<IGH<97CAD9H=H=J99EI=@=6F=IAK=@@:5=@HC69GC7=5@@MCDH=A5@5B85ACF9
9@56CF5H95DDFC57<H<5BGC7=5@K9@:5F9A5L=A=N5H=CB=GB979GG5FMHC:=B85B9EI=@=6F=IA
GC@IH=CB
09588F9GGH<=G7<5@@9B;96MDF9G9BH=B;5B5@;CF=H<A=79LH9BG=CBHCH<95DDFC57<9G
6M'9;=G<=5B8$9<C99H5@09=@@IGHF5H9AC89@5B85@;CF=H<A6M5DD@M
=B;H<9AC89@HCH<97IFF9BH=GGI9C:HF589G5B7H=CBG5G5B=BGHFIA9BHHC:CGH9FD5FH=7=D5
H=CB=B5B=BH9FB5H=CB5@9BJ=FCBA9BH5@5;F99A9BH09:=B8
+9;=CB -5F=::
)5F5A9H9F,79B5F=C
τ
/,8
4068
6
/01,@7?
6
/,8
4068
/,8
4068
),-708;,.?:130?0=:20904?D)307,>?.:7@89>3:B>?30>8,770>??,=411=,?0
τ
?3,?4>
>@114.409??:49/@.01@77.::;0=,?4:9
τ
B,>A,=40/-0?B009,9/@>492,>?0;
>4E0:1%97D;,=,80?0=A,7@0>?3,?/4110=1=:8?304=/01,@7?>49=:B>,=074>?0/
40>.09,=4:>3:B>,A,=4,?4:9:1494?4,7.,;4?,76>.09,=4:>,9/>3:B
0C;0=4809?>?3,?0110.?84?42,?4:9.:>?>A4,4068,9/.748,?0.3,920/,8,20>A4,
/,8
4.5 Conclusion 111
"0>>8,990?,7.:9:84.#:/077492
0<9BH<97C5@=H=CB=ADCG9GH5F=::GCB=ADCFHG:FCA:F99F=8=B;F9;=CBGD5FH=7=
D5H=CB=BH<97C5@=H=CBF=G9G @C65@GC7=5@K9@:5F9F=G9G5@CB;K=H<D5FH=7=D5H=CB
89GD=H9GA5@@K9@:5F9@CGG9G8I9HCH<98=GHCFH=CB75IG986MH<9H5F=::=BGHFIA9BH
-CH<F95H9BBCBA9A69FGK=H<HF589G5B7H=CBG=G7F98=6@95G@CB;5GH<9H5F=::
F5H9=GGA5@@K<9F9GA5@@89D9B8GCBH<9FA=B;HCB9@5GH=7=HMCF@5F;9H5F=::
F5H9G7C5@=H=CBA9A69FGKCI@86969HH9FC::BCHHCG5B7H=CBHF589
'CBA9A69FGF9GDCB8HC9A=GG=CB7IHGCBH<9D5FHC:H<97C5@=H=CB6MF5=G=B;
H<9=FCKB9A=GG=CBG6IHK9:=B8H<=G@95?5;99::97HHC69GA5@@
-<9G9F9GI@HG5F97CADF9<9BG=6@9=B@=;<HC:H<9IB89F@M=B;H<9CF9H=75@AC89@C:
=BH9FB5H=CB5@HF589:C@@CK=B;H<97CB79DHC:B5H=CB5@DFC8I7H8=::9F9BH=5H=CB;CC8GDFC
8I7986M8=::9F9BHF9;=CBG5F95GGIA98HC69=AD9F:97HGI6GH=HIH9G5ACB;957<CH<9F
19H5@@7CIBHF=9G57H5GDF=79H5?9FG=B57CAD9H=H=J99EI=@=6F=IA"BHFC8I7=B;H5F=::G=B
H<=G7CBH9LH5@@CKG7C5@=H=CBA9A69FGHC75D=H5@=N9CBH<9=FDCH9BH=5@A5F?9HDCK9F-<9
9@5GH=7=HMC:GI6GH=HIH=CB69HK99B;CC8G89H9FA=B9GH<995G9K=H<K<=7<BCBA9A69FG
75B5JC=87C5@=H=CB;CC8G5B8<9B79DIHG5@=A=HCBH<9DCH9BH=5@7@CIHC:H<9H5F=::=B
GHFIA9BH
-<95DD@=75H=CBC:H<9AC89@B9J9FH<9@9GG=89BH=:=9GGCA9FC6IGHEI5@=H5H=J9F9@5H=CB
G<=DG5B87@95F@M89ACBGHF5H9GH<9IG9:I@B9GGC:H<95@;CF=H<A"B:57HH<9HF95HA9BHC:
9LH9FB5@=H=9GG?9H7<98=BH<=GD5D9F75B95G=@M69HF5BG:9FF98HCG=A=@5F8MB5A=7;5A9G
K=H<9LH9FB5@=H=9G=B5@@M=BCF89FHCDIHBIA69FGCBH<9=89BH=:=98EI5@=H5H=J99::97HG
<9H9FC;9B9CIGF9;=CBGG<CI@869=BHFC8I7985B86975@=6F5H98HCF95@KCF@8F9;=CBG
-<=GKCI@8:IFH<9F9B<5B79H<9DC@=7MF9@9J5B79C:H<9AC89@F9GI@HG
.69:B70/20809?>
09KCI@8@=?9HCH<5B?5F@C5FF5FC&=7<59@=BIG5FGH9B!9@A5B8F=HH5
-=9H>9B:CF8=G7IGG=CBGC:5B95F@=9F8F5:HC:H<=GD5D9F5B8HKC5BCBMACIGF9:9F99G:CF
H<9=F7CAA9BHG@@F9A5=B=B;9FFCFG5F9C:7CIFG9CIFCKB-<9AC89@9LD9F=A9BHG
A5899LH9BG=J9IG9C:H<9@97<G=;9H5@AI@H=FIBG=AI@5H=CB9BJ=FCB
A9BH $5= %9GGA5BB 5B8 +C69FH &5FG7<=BG?= F979=J98 :IB8=B; :FCA H<9 IFCD95B
CAA=GG=CBK=H<=BH<9&DFC>97HDFC>97H (
% " !
-56@9@=GHGCIF7<C=79C:D5F5A9H9FG09F9GHF=7HH<=GGHI8MHCH<975G9C:GMAA9H
F=7D@5M9FG<9B79575@=6F5H=CBHCF95@KCF@8F9;=CBG=GCIHC:EI9GH=CB'9J9FH<9@9GGK9
G9@97H985G9HC:D5F5A9H9FGGI7<5GHCDFC8I795G79B5F=CH<5H5DD95FGD@5IG=6@9-<=G
5DD9B8=L@=GHGH<95GGIADH=CBGK9A589
-<97<C=79C:H<9DIF9F5H9C:H=A9DF9:9F9B79<5GF979=J98AI7<5HH9BH=CBG=B79
,H9FBGI;;9GH985G=;B=:=75BH@M@CK9FJ5@I9H<5B95F@=9FGHI8=9G9;
=B'CF8<5IG09GHF=?9A=88@9;FCIB86MG9@97H=B;
ρ
6IH9LD@CF96CH<
,H9FBG5B8'CF8<5IG7<C=79G=BCIFG9BG=H=J=HM5B5@MG=G
097<CG9H<9F5H9C:9LC;9BCIG@56CF9B<5B7=B;H97<BC@C;=75@7<5B;92=GI7<H<5H
@CB;H9FA97CBCA=7;FCKH<5J9F5;9G5HD9F79BHD9FM95FK<=7<=GK=H<=BH<9F5B;9
112 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
C:H<9"),+,:5A=@MC:89J9@CDA9BHG79B5F=CG")
0=H<=B=H=5@@56CF5B8@56CFDFC8I7H=J=HM5HK97<CG9=B=H=5@75D=H5@GI7<H<5HH<9
G5J=B;GF5H9=G5DDFCL=A5H9@M7CBGH5BH5HD9F79BH8IF=B;H<9:=FGH897589G=9H<9
97CBCAM=GCB565@5B798;FCKH<D5H<-<=G:=;IF97CFF9GDCB8GHCH<9KCF@8G9AD=F=75@
5J9F5;9C:D9F79BH69HK99B5B85B?:CF"BH9FB5H=CB5@,9HH@9A9BHG
09:F9EI9BH@MJ5FMH<9FA=B;HCBD5F5A9H9F
σ
H<5H89H9FA=B9GH<99@5GH=7=HMC:
GI6GH=HIH=CB=BCIF9LD9F=A9BHGIG=B;J5@I9G69HK99B5B8097CAD5F9H<9G9
F9GI@HHCH<9@=A=H75G9C:5B=B:=B=H9
σ
5B89LD@CF9H<9HF5BG=H=CBHCH<9@=A=HIG=B;5
<=;<J5@I9C:
σ
"B75@=6F5H98F95@KCF@8AC89@GH<9G99@5GH=7=H=9GHMD=75@@M@=9
69HK99B5B89FBGH9=B9H5@-C9B<5B79H<97CAD5F56=@=HMC:75@7I@5H=CBG
K=H<8=::9F9BH
ρ
K9G9@97H98H<9G<5F9D5F5A9H9FG>/:85B8>1:=GI7<H<5H:CF5@@
ρ
H<9
9LDCFHF5H=C=G56CIHD9F79BH=BH<9'5G<9EI=@=6F=IACFH<90-(<5G9GH=
A5H98H<9F5H=CC:9LDCFHG=B;CC8G5B87CAA9F7=5@G9FJ=79GHC )5GD9F79BH
0-(
)5F5A9H9FG=BH<97@=A5H9AC8I@95F965G98CB@=H9F5HIF9J5@I9G;=J=B;IG5V
H9AD9F5HIF9=B7F95G96M5B85V=B7F95G96M=BH<96IG=B9GG5GIGI5@=9
K=H<CIH 7@=A5H9 7<5B;9 85A5;9G 5B8 K=H<CIH 5BM 7CCD9F5H=CB 69HK99B F9;=CBG
'CF8<5IG5B815B;9GH=A5H95G=A=@5FH9AD9F5HIF9=B7F95G9C:V=B
:CFH<9=FA5F?9HG79B5F=C
-<985A5;9:IB7H=CBK5G7<CG9BGI7<H<5H=B'5G<9EI=@=6F=IA85A5;9G=B5F9
D9F79BH097<CG9H<=GF9@5H=J9@M<=;<J5@I97CAD5F98HC85A5;9GF5B;=B;:FCAHC
56CIHD9F79BH57FCGGF9;=CBG=B+"K=H<5;@C65@5J9F5;9C:56CIHD9F79BHHC
577CIBH:CF,H9FB^G9GH=A5H=CBH<5HZ36IG=B9GG5GIGI5@47@=A5H97<5B;9K=@@
F98I79K9@:5F96M5B5ACIBH9EI=J5@9BHHC5F98I7H=CB=B7CBGIADH=CBD9F<958C:
69HK99B5B8D9F79BH[
0=H<=BH<9A=H=;5H=CBCDH=CBD5F5A9H9FG
ψ
5B84068K9F9G9@97H98GI7<H<5HCDH=A5@
565H9A9BHH<9GC7=5@D@5BB9FGC@IH=CBF98I79GH<9H9AD9F5HIF9=B7F95G9=B6M
V"B'CF8<5IG5B815B;7CCD9F5H=J969<5J=CFF98I79G;@C65@H9AD9F5HIF9
=B6MV
4.6 Appendix: Parameter Choices 113
"0>>8,990?,7.:9:84.#:/077492
)5F5A9H9F ,MA6C@ /5@I9
)IF9F5H9C:H=A9DF9:9F9B79
ρ
"B7CA9G<5F975D=H5@
β
%56CFDFC8I7H=J=HM;FCKH< 2=
+5H9C:5IHCBCACIG9A=GG=CB=BH9BG=HMF98I7H=CB
ν
"B=H=5@@56CF 7
"B=H=5@@56CFDFC8I7H=J=HM ,
"B=H=5@75D=H5@GHC7? 6
,<5F9D5F5A9H9F8CA9GH=7 >/:8 >00?0C?
,<5F9D5F5A9H9F:CF9=;B >1:= >00?0C?
FA=B;HCB9@5GH=7=HMC:GI6GH=HIH=CB
ρ
::97H=J9B9GGC:=BJ9GHA9BHG=B68 4068
65H9A9BH7CGH9LDCB9BH
ψ
(795B6=CGD<9F95G(GCIF79
β
&
HACGD<9F=7F9H9BH=CB:57HCF
Β
0
+58=5H=J9H9AD9F5HIF98F=J=B;:57HCF
μ
0
-9AD9F5HIF985AD=B;:57HCF
α
&0
(795B6=CGD<9F95G(G=B?
σ
&0
"B=H=5@7CB79BHF5H=CB .:9.
"B=H=5@H9AD9F5HIF9 ?08;
"B=H=5@7IAI@5H=J99A=GG=CBG .@80
5A5;9:IB7H=CB7C9::=7=9BH /,8
5A5;9:IB7H=CB9LDCB9BH /,8
),-70&,=,80?0=A,7@0>
&
@8M#(FGN5;)+,H=;@=HN#@=A5H9<5B;9B;9B85:CF @C65@C@@97H=J9
7H=CB)F9D5F98:CFH<97CB:9F9B79CB-<9-=A=B;C:@=A5H9<5B;9)C@=7=9G)9K9BH9FCB @C65@
@=A5H9<5B;9
5B? :CF "BH9FB5H=CB5@ ,9HH@9A9BHG BBI5@ +9DCFH
<HHDKKK6=GCF;DI6@5FD8:5F;D8:
5F?9F-!5CF5B)$U<@9F#05FF9B+0=BB9,975F6CB=G=B;H<9 @C65@7CBCAM
K=H<"B8I798-97<BC@C;=75@<5B;9,79B5F=CGHCIG=B;& -<9B9F;M#CIFB5@
5FF9HH,,9@:9B:CF7=B;"BH9FB5H=CB5@BJ=FCBA9BH5@;F99A9BHG(L:CF87CBCA=7)5D9FG
5FF9HH,-<9GHF5H9;MC:HF589G5B7H=CBG=B=BH9FB5H=CB5@9BJ=FCBA9BH5@5;F99A9BHG+9GCIF79
5B8B9F;M7CBCA=7G
5FF9HH,BJ=FCBA9BH5B8,H5H97F5:H-<9,HF5H9;MC:BJ=FCBA9BH5@-F95HM&5?=B;(L:CF8
.B=J9FG=HM)F9GG(L:CF8
9B98=7?+,HF=?=B;5'9K95@CB@=A5H9<5B;9"GGI9G=B,7=9B795B8-97<BC@C;M
114 Chapter 4 Effects of Tariffs on Coalition Formation
"0>>8,990?,7.:9:84.#:/077492
9FBGH9=B )& &CBH;CA9FM 0 +IH<9F:CF8 - @C65@ "AD57HG C: H<9 $MCHC
;F99A9BH+9GI@HG:FCAH<9&,&+-&C89@+9GCIF795B8B9F;M7CBCA=7G
=9FA5BBFC<A+"AD@9A9BH=B;H<9$MCHC)FCHC7C@K=H<CIHH<9.B=H98,H5H9G-<9
GHF5H9;=7FC@9C:9B9F;MH5L58>IGHA9BHG5HH<96CF89F@=A5H9)C@=7M
CG9HH=/5FF5FC 5@9CHH=&-<9MB5A=7GC:5F6CB5B8C:B9F;M"BH9BG=HM=B5
&C89@C:B8C;9BCIG-97<B=75@<5B;9CB85N=CB9B=BF=7C&5HH9=&5M
CG9HH=/5FF5FC 5@9CHH=&&5GG9HH=-5JCB=&0"-!0CF@8"B8I798
-97<B=75@ <5B;9 !M6F=8 &C89@ -<9 B9F;M #CIFB5@ ,D97=5@ "GGI9 !M6F=8 &C89@@=B; C: B9F;M
BJ=FCBA9BH)C@=7=9G+97CB7=@=B;CHHCAID5B8-CD8CKB
5FF5FC,=B=G75@7C,HF5H9;=9G:CFH<9=BH9FB5H=CB5@DFCH97H=CBC:H<99BJ=FCBA9BH
#CIFB5@C:)I6@=77CBCA=7G
5FF5FCI7<B9F9FGCG=AC"&5F7<=CF=57?HC$MCHC.,)5FH=7=D5H=CB5B8
H<9%=B?5;969HK99B+5B8@=A5H9CCD9F5H=CB,=:C0CF?=B;)5D9F,9F=9G'C&
0CF?=B;)5D9F'C
F5GGCIG+!CIF7589#,5GG=(B8C;9BCIGGHFI7HIF5@7<5B;95B87@=A5H9H5F;9HG
AC89@=B;9LD9F=A9BHGK=H<"A57@=A+B9F;M#CIFB5@
89B<C:9F ( 5I9F ' $F=9;@9F -<9 =AD57H C: H97<BC@C;=75@ 7<5B;9 CB 7@=A5H9
DFCH97H=CB5B8K9@:5F9"BG=;<HG:FCAH<9AC89@&"'7C@C;=75@7CBCA=7G
89B<C:9F(%9GGA5BB$$9A:9FH FI66&$C9<@9F# "B8I798-97<BC@C;=75@
<5B;9LD@CF=B;=HG"AD@=75H=CBG:CFH<97CBCA=7GC:HACGD<9F=7,H56=@=N5H=CB,MBH<9G=G+9DCFH
:FCA H<9 "BBCJ5H=CB &C89@=B; CAD5F=GCB )FC>97H B9F;M #CIFB5@ ,D97=5@ "GGI9 B8C;9BCIG
-97<BC@C;=75@<5B;95B8H<97CBCA=7GC:HACGD<9F=7,H56=@=G5H=CB
M7?A5BG#-I@?9BG!,=AI@5H=B;7C5@=H=CB5@@MGH56@96IF89BG<5F=B;5;F99A9BHG:CFH<9
7@=A5H97<5B;9DFC6@9A+9GCIF795B8B9F;M7CBCA=7G
99BGHF5+&5F?IG9B#+5B8+CG9$.G=B;H<9;F5J=HM9EI5H=CBHC8=::9F9BH=5H9
5ACB;5@H9FB5H=J9H<9CF=9GC:HF5895B58=5B#CIFB5@C:7CBCA=7G
=BIG&+IB8G<5;9B,HF5H9;=7%=B?G69HK99BBJ=FCBA9BH5@5B8-F589)C@=7=9G=:)@5BH
%C75H=CB"GB8C;9BCIG=G?IGG=CBG69=HF5;'F=G?IGG=CBG69=HF5;89G57<69F9=7<G0=FHG7<5:HG
K=GG9BG7<5:H9B89F9FB.B=J9FG=HQH!5;9B
@97<G=;&U<A.'C7?9-+57<=ACK-<9&I@H=+IB,=AI@5H=CBBJ=FCBA9BH
,=ABJ/9FG=CB,99<HHDKKKD=?DCHG85A89GC:HK5F9G=A9BJ
9F@5;<+"-=B5 @C65@ FCKH<@=A5H9&C89@K=H<,B9F;M#CIFB5@,D97=5@"GGI9
B8C;9BCIG-97<BC@C;=75@<5B;95B8H<97CBCA=7GC:HACGD<9F=7,H56=@=G5H=CB
"),D97=5@+9DCFHCBA=GG=CBG,79B5F=CG5GD97=5@F9DCFHC:0CF?=B; FCID"""C:H<9
"BH9F;CJ9FBA9BH5@)5B9@CB@=A5H9<5B;95A6F=8;9.B=J9FG=HM)F9GG
"GA9F+'9I<C::$CF89FH5L58>IGHA9BH5:95G=6@9K5MHCGIDDCFHGHF=B;9BH9A=GG=CB
HF58=B;IFCD95B#CIFB5@C:%5K5B87CBCA=7G
#CB9G"5B80=@@=5AG#&95GIF=B;H<9,C7=5@+9HIFBHC+*I5FH9F@M#CIFB5@C:
7CBCA=7G
$9<C9 -# %9J=B9 $ +CA9F )& (B 7<5F57H9F=N=B; 9EI=@=6F=5 C: 97CBCA=9G K=H<
9LH9FB5@=H=9G5B8H5L9G5GGC@IH=CBGHCCDH=A=N5H=CBDFC6@9AG7CBCA=7-<9CFM
$9A:9FH B "BH9;F5H98 GG9GGA9BH &C89@ C: 7CBCAMB9F;M@=A5H9 -<9 &C89@
0=5;9A"BH9;F5H98GG9GGA9BH
$9A:9FH@=A5H97C5@=H=CBG5B8=BH9FB5H=CB5@HF5895GG9GGA9BHC:7CCD9F5H=CB=B79BH=J9G
6M=GGI9@=B?5;9B9F;M)C@=7M
$MDF9CG,5<B(&+ AC89@K=H<9B8C;9BCIGH97<BC@C;=75@DFC;F9GG
BJ=FCBA9BH5@&C89@=B;5B8GG9GGA9BH
%9=A657<&89B<C:9F(-97<BC@C;=75@GD=@@CJ9FGK=H<=BAI@H=F9;=CBAC89@G"BH9FH9A
DCF5@CDH=A=N5H=CB69MCB8H<9'9;=G<=5DDFC57<7CBCA=7&C89@@=B;
'9;=G<=-09@:5F997CBCA=7G5B89L=GH9B79C:5B9EI=@=6F=IA:CF57CAD9H=H=J997CBCAM
&9HFC97CBCA=75
'CF8<5IG0&5B5;=B;H<9;@C65@7CAACBG&"-)F9GG5A6F=8;9&5GG57<IG9HHG
4.7 Bibliography 115
"0>>8,990?,7.:9:84.#:/077492
'CF8<5IG015B;2 +9;=CB5@MB5A=7 9B9F5@EI=@=6F=IA&C89@C:@H9FB5H=J9
@=A5H9<5B;9,HF5H9;=9G-<9A9F=75B7CBCA=7+9J=9K
'CF8<5IG0"GH<9$MCHC)FCHC7C@5958I7?F9-<9F9BM%=J9I7?GFCIB8
CAD5F=GCBC:@H9FB5H=J9 @C65@-F5856@9A=GG=CBG+9;=A9G7=H98=B5FF9HH,,H5J=BG+
"B7F95G=B; )5FH=7=D5H=CB 5B8 CAD@=5B79 =B "BH9FB5H=CB5@ @=A5H9 <5B;9 ;F99A9BHG "BH9FB5H=CB5@
BJ=FCBA9BH5@;F99A9BHG)C@=H=7G%5K5B87CBCA=7G
)9HG7<9@!9@8 ,7<9@@B<I69F!FI7?B9F--SH<%!5GG9@A5BB$-<9-C@9F56@9
0=B8CKGDDFC57<-<9CF9H=75@5B8&9H<C8C@C;=75@CIB85H=CBG@=A5H=7<5B;9
)9F9N(&I@H=D@9+9;=A9G"GGI9%=B?5;95B8"BH9FB5H=CB5@CCD9F5H=CBLD@CF=B;H<9+C@9
C:H<90-(.B=J9FG=HMC:)9BBGM@J5B=5#CIFB5@C:"BH9FB5H=CB5@7CBCA=7%5K
)CDD'-"9B8C;9BCIGH97<BC@C;=75@7<5B;9=BH<9"AC89@C:;@C65@K5FA=B;
#CIFB5@C:BJ=FCBA9BH5@7CBCA=7G5B8&5B5;9A9BH
+5AG9MA5H<9A5H=75@H<9CFMC:G5J=B;7CBCA=7#CIFB5@
,7<9@@=B;-0<5H&5?9G F99B<CIG9,9BG9-=A9HC+9H<=B?H<9$MCHC)FCHC7C@CF9=;B
::5=FG
,H9FB'-<97CBCA=7GC:@=A5H9<5B;9-<9,H9FB+9J=9K5A6F=8;9.B=J9FG=HM)F9GG
,H=;@=HN#&5?=B; @C65@=N5H=CB0CF?00'CFHCBCAD5BM'9K1CF?
-IFBCJG?M,#"BH9FB5H=CB5@A57FC97CBCA=78MB5A=7G&"-)F9GG
0-("BH9FB5H=CB5@-F589,H5H=GH=7G0-()I6@=75H=CBG 9B9J5,K=HN9F@5B8
116 Chapter 4 Effects of Tariffs on Coalition Formation
Chapter 5
Research cooperation and international standards
in a model of coalition stability∗
Kai Lessmann
Ottmar Edenhofer
∗submitted to Resource and Energy Economics as Lessmann, K. and O. Edenhofer, “Research coopera-
tion and international standards in a model of coalition stability’
117
118 Chapter 5 Research cooperation and international standards
Research cooperation and international standards in
a model of coalition stability
Kai Lessmann a,∗, Ottmar Edenhofer a
aPotsdam Institute for Climate Impact Research, PO Box 60 12 03, D-14412 Potsdam,
Germany
Abstract
Suggestions on international cooperation in climate policy beyond 2012 include substi-
tuting or complementing international environmental agreements (IEA) with technology
oriented agreements (TOA). We look at the impact of TOA on environmental cooperation
in the framework of coalition stability. Using a numerical model, we analyze the differ-
ences of several TOA and how they interact. We find that participation in and environmen-
tal effectiveness of the IEA are raised less effectively when the TOA focuses on mitigation
technology rather than augmenting productivity, which is due to the former having an effect
on all actors via emissions, whereas effects of the latter are exclusive to research partners.
For the same reason, we find that restricting the effects of R & D cooperation is credible
only in case of productivity. Technology standards may fail to foster participation when
they are restricting members and non-members alike, and may suffer from inefficiencies.
However, when implemented as a complementary instrument, these disadvantages did not
apply. Separately negotiated technology standards may hence facilitate participation in an
IEA without adding to its complexity.
Key words: Coalition Formation, International Environmental Agreements, Issue Linking,
Non-cooperative Game Theory, R & D Spillovers, Technology Standards
1 Introduction
Achieving full cooperation in a self-enforcing international environmental agree-
ment (IEA) is difficult when the underlying game presents the actors with a dilemma:
while global cooperation is socially optimal, it is often better for a number of play-
ers to act as free-riders, i.e. enjoying the benefits of other players’ abatement efforts
∗Corresponding author.
Email address: Kai.Lessmann @ pik-potsdam.de (Kai Lessmann).
Preprint submitted to Resource and Energy Economics November 05, 2008
5.1 Introduction 119
without reducing their own emissions. Consequently, it is a standard result in non-
cooperative game theoretic models that voluntary participation in environmental
cooperation alone tends to be low (see for example Carraro and Siniscalco, 1992,
and Barrett, 1994, or the more recent Finus et al., 2006, and Carraro et al., 2006).
By introducing additional incentives, the structure of the game may be changed
making cooperation easier to achieve. These incentives range from positive incen-
tives such as side payments, permit allocation, and issue linking to negative incen-
tives such as reciprocal measures, financial penalties, and trade restrictions (Barrett
and Stavins, 2003), see Wagner (2001) for an overview of incentives to stabilize
international environmental agreements. In this paper, we focus on linking envi-
ronmental cooperation to a technology oriented agreement (TOA).
1.1 Issue Linking
The unfavorable incentive structure in climate change mitigation is due to the public
good character of a stable climate. Enjoying a stable climate is non-rival, and there
are no means of excluding anybody from doing so, hence the possibility to free-
ride.
Issue linking attempts to improve the incentive structure by linking the provision
of the public good to an exclusive access to a club good (Carraro, 1999). When the
attractiveness of the club good outweighs the incentive to free-ride, the dilemma is
overcome. Possible candidates for such club goods are technology oriented agree-
ments. De Coninck et al. (2007) provide an overview of TOA stressing out the
potential role of TOA in addressing the free-riding incentives in climate protection
negotiations.
Among the TOA, in particular the spillovers from R & D agreements have the quali-
ties of a club good (non-rivalry and excludability). Previous issue linking modeling
studies have analyzed the potential of spillovers to raise participation in interna-
tional cooperation. In these studies, cooperative research and development creates
spillovers concerning production costs (Carraro and Siniscalco, 1997; Botteon and
Carraro, 1998), profit (Katsoulacos, 1997), energy efficiency (Kemfert, 2004), pro-
ductivity and emission intensity simultaneously (Buchner and Carraro, 2006), and
marginal abatement costs (Nagashima and Dellink, 2008).
1.2 Potential of spillovers
Research and development is known to have spillovers. Griliches (1992), for ex-
ample, reviews a number of empirical studies which estimate social and private
rates of return to R & D. Griliches concludes, “R & D spillovers are present, their
2
120 Chapter 5 Research cooperation and international standards
magnitude may be quite large, and social rates of return remain significantly above
private rates.”
Research partnerships may facilitate these spillovers. There are numerous reasons
for cooperative research, ranging from costs minimization to strategic considera-
tions. In particular, the list of reasons includes internalizing spillovers, e.g. learning
from partners, transfer of technology and technical knowledge, and increasing ef-
ficiency and synergies through network, as well as exploiting the non-rivalry of
knowledge, e.g. by sharing R & D costs (Hagedoorn et al., 2000). Of course, rais-
ing the spillover intensity is not a policy instrument at the disposal of governments.
But by encouraging research partnerships spillovers might be fostered indirectly.
Existing governmental policies aimed at encouraging cooperative R & D focus on
providing legal frameworks as well as financial support, noteworthy the EU Frame-
work Programmes on Research and Technological Development (FWP). Aimed at
industry as well as universities and research laboratories, the FWP offer financial
support of up to 50 percent of the total joint research costs but require the research
partnership to include members from at least two EU countries (Hagedoorn et al.,
2000), i.e. the FWP are a prime example of boosting international research cooper-
ation.
1.3 Potential of international standards
Contrary to R & D spillovers, international standards do not promise to lessen the
dilemma by raising payoffs in the participating regions. They are therefore not
suited to be issue-linked to an environmental agreement in the way of R & D, rather,
they could complement an environmental agreement. For example, Edmonds and
Wise (1999) propose a standard requiring carbon capture and storage for new elec-
tricity power plants and for synthetic fuels from coal. Barrett (2003) suggests com-
plementing technological R & D with an agreement on international technology
standards. These standards provide a market pull incentive to commercialize the
results of cooperative R & D. Participation in such a technology standard may be
spurred by the following incentives (Farrell and Saloner, 1987; David and Green-
stein, 1990; Barrett, 2003):
•Standards induce network externalities: the higher the number of participants in
a standard, the larger the benefits of adopting the standard.
•Standards protect their participants from lock-in in technologies that are then
abandoned.
•Standards help reduce costs when economies of scale can be exploited.
Additionally, a minimum participation clause and trade restrictions against non-
participants can be implemented to further strengthen these incentives (Barrett,
2003): Network externalities and scale effects increase with participation, hence
3
5.1 Introduction 121
minimum participation guarantees a minimum extent of these incentives. And com-
bined with a trade ban against non-compliant players, minimum participation en-
sures a growing market for the new technology as well as a shrinking market for
old technology, reinforcing the fear of being locked into an abandoned technol-
ogy. Barrett (2003) argues that these incentives make the adoption of technology
standards much easier than raising participation in an environmental agreement.
Barrett’s (2003) proposal is intended as a substitute for an emissions abatement
agreement with the advantage of a better incentive structure. Whether these in-
centives suffice to provide effective environmental protection has been challenged
(Philibert, 2003, 2004; de Coninck et al., 2007). The usefulness of technology stan-
dards as a complementary policy instrument is however undisputed.
1.4 Coalition formation
The formal analysis of self-enforcing international environmental agreements in
non-cooperative game theory was pioneered by Barrett (1991a,b) and Carraro and
Siniscalco (1992, 1993), and has recently been reviewed in Finus (2008).
The incentive of issue linking, which is the focus of this paper, has been studied
using conceptual and empirically calibrated models. Carraro and Siniscalco (1995,
1997) investigate linkage of environmental cooperation to cooperation on R & D in
a static three-stage game showing that linkage indeed furthers participation. Bot-
teon and Carraro (1998) extend this analysis adding heterogeneity based on empir-
ical data to this model. While this renders the model intractable, they confirm their
earlier findings numerically: participation in the IEA rises with spillover intensity
including full cooperation of five out of five players.
Katsoulacos (1997) questions the approach of having one entity decide upon both,
environmental and technological cooperation, arguing that the decision to cooper-
ate on technological R & D is taken by firms, not governments. Consequently, his
model distinguishes firms deciding on spillover levels and governments deciding on
R & D subsidies aimed at encouraging spillovers. The analysis is restricted to two
countries, which can be shown to enter joint cooperation on R & D and environment
if the gains from subsidies are large enough.
Kemfert (2004) explores the effects of issue linking in a CGE model calibrated to
the GTAP database (McDougall et al., 1998). The scenarios include cooperation on
energy efficiency R & D as well as trade barriers against non-cooperating countries.
In this model, introducing R & D cooperation has a strong effect on the incentive to
participate in an IEA. With R & D cooperation, all four of the negotiating countries
want to join the IEA, compared to none in the base case, i.e. full cooperation is
4
122 Chapter 5 Research cooperation and international standards
internally stable. 1
Buchner et al. (2005) apply a multi-actor optimal growth model to questions of
issue linkage. They limit their analysis to a selected set of coalitions, in particular
the coalition of Kyoto signatories plus the United States, and explore the effect of
linking R & D cooperation to environmental cooperation on the incentives for the
United States to join the Kyoto signatories. It turns out that, given sufficiently high
spillovers, such an agreement does indeed become stable. They note, however, that
making R & D cooperation dependant on environmental cooperation is not credible,
i.e. Kyoto signatories prefer to cooperate on R & D with the United States even if
the latter act non-cooperatively on emission abatement.
Nagashima and Dellink (2008) use the STACO model to explore the effects of
technology spillovers on the stability of coalitions. They focus on spillovers in mit-
igation technology, and model these through changes of the marginal abatement
cost curve. They observe that spillovers have a positive effect on the abatement ef-
fort, but the number of participating regions is only increased by one beyond the
business as usual maximum participation of six out of twelve regions. This finding
proves robust against a variation of the intensity of spillovers and the way spillovers
affect the marginal abatement costs curves, as well as the choice of the indicator
for the state of technology. Thus, the authors conclude that technology spillovers
do not substantially increase the success of IEA.
All issue linking studies cited above find that issue linking with spillovers has pos-
itive effects on participation in the IEA. But the extent of this positive effect varies,
ranging from complete success in stabilizing full cooperation (Botteon and Car-
raro, 1998; Kemfert, 2004), to merely marginal increases of the coalition size (Na-
gashima and Dellink, 2008). The models differ in a great number of ways and it is
unclear which modeling assumptions give rise to these differences in model results.
Most authors acknowledge that the intensity of spillovers is an important determi-
nant, but given the state of the literature, it is difficult to provide a sound empirical
basis for the choice of spillover intensity. Variation of this key parameter, as studied
in Botteon and Carraro (1998) and Nagashima and Dellink (2008), reveals the sen-
sitivity of this key assumption, yet the selected values for spillover intensity cannot
be compared across models. Furthermore, the sources of spillovers differ between
models. The implication of the kind of spillover, e.g. whether related to produc-
tivity as in Botteon and Carraro (1998), or related to mitigation technology as in
Nagashima and Dellink (2008) has not been studied.
1Coalitions are internally stable when no member has an incentive to leave. We define this
formally below.
5
5.1 Introduction 123
1.5 Novelty
We go beyond existing studies by comparing spillovers that arise from two different
research sectors, productivity and mitigation technology, showing that the effective-
ness of spillovers depends on the type of knowledge that spills over. The reason is
that, unlike in the case of productivity R & D, progress in mitigation technology
has an external effect via its impact on emissions, making it easier to achieve high
levels of cooperation by linking to productivity R & D. These results on participa-
tion carry over to similar conclusions about the impact of IEA on environmental
effectiveness and global welfare. In order to increase the comparability of spillover
intensity, we estimate the gains from spillovers in terms of additional consumption.
Furthermore, the effect of spillover from cooperative R & D has so far only been
investigated in isolation from international technology standards. We complement
spillovers by technology standards and explore the interdependence of the two, as
well as the scope of technology standards to stabilize coalitions by themselves. We
find that cooperative R & D and technology standards are mutually reinforcing in
their positive effect on international cooperation. By themselves, technology stan-
dards have almost no effect on participation in the IEA. The remainder of this paper
follows the usual three steps definition of the model (Section 2), results (Section 3)
including some sensitivity analysis (Section 4), and conclusions (Section 5).
2 The Model
We approach the assessment of coalition stability, research cooperation, and in-
ternational standards in a multi-actor optimal growth model, which is a common
modeling framework for the economy-climate stock pollutant problem in general
(e.g. Nordhaus and Yang, 1996; Kypreos and Bahn, 2003; Bosetti et al., 2006)
and also in coalition stability analyses (e.g. Eyckmans and Tulkens, 2003; Buch-
ner and Carraro, 2006). In particular, it is appropriate for the long economic time
horizon required for an integrated assessment of global warming (Edenhofer et al.,
2006). Furthermore, intertemporal utility maximization of a representative agent
gives macroeconomic models a firm micro-foundation and makes them suitable for
welfare analysis (Turnovsky, 2000, pp. 3).
6
124 Chapter 5 Research cooperation and international standards
2.1 Model Equations
Preferences
Within this framework, each region iis modeled following Ramsey (1928) as a
maximizer of its intertemporal welfare Wi. Here, we chose the utilitarian welfare
function with an instantaneous utility function U,U>0andU <0, and per
capita consumption cit /lit as an indicator of well-being. Parameter
ρ
denotes the
pure rate of time preference, −
η
is the elasticity of marginal utility, and lit the size
of the population.
Wi=
∞
0
lit U(cit /lit )e−
ρ
tdt (1)
U(cit /lit )=(cit /lit )1−
η
1−
η
(2)
Technology
Each region produces a single good using Cobb-Douglas technology Ffrom capital
kit and exogenously given labor supply lit , which is subject to labor enhancing
technological change ˜ait . Parameter
β
is the income share of capital.
F(˜ait lit ,kit )=(˜ait lit )1−
β
k
β
it (3)
Capital is made up from past investments, init . New ideas that contribute to labor
productivity ait in country iare a function of the funds invested in R & D, iait .
Parameters
λ
≤1andΦ≥0 describe effects of researchers “stepping on tows”
and “standing on shoulders,” respectively. Parameter
ξ
ais a scaling parameter. This
knowledge production function is proposed in an empirical study by Jones and
Williams (1998) and has been applied in integrated assessment in Edenhofer et al.
(2005, 2006).
d
dt kit =init (4)
d
dt ait =
ξ
a(iait )
λ
(ait )Φ(5)
Labor productivity ˜ait encompasses the accumulated knowledge of region i(ait )as
well as eventual spillovers from other regions. In the base case we assume no spill-
7
5.2 The Model 125
overs between regions and simply set ˜ait =ait . When R & D spillovers are modeled,
we use a weighted aggregate of labor productivity in all regions. This approach is
also used in the empirical literature on R & D spillovers, for example in Griliches
(1992).
˜ait =∑
j
ε
a
ijajt (6)
Griliches (1992) interprets
ε
ij as the “economic and technological distance” be-
tween iand jwhere large values of
ε
ij indicate “closeness”. We always set
ε
a
ii =1,
and in the base case
ε
a
ij =0fori=j.Valuesof
ε
a
ij >0 indicate spillovers and are
discussed below.
Climate Dynamics
We model greenhouse gas emissions eit as a by-product of economic activity (yit
below in Equation 16). Emission intensity of production decreases exogenously
with eitat an annual rate of dr but may be additionally decreased by investing in
a mitigation stock kmit . Mitigation kmit reduces emission intensity
σ
it with dimin-
ishing effectiveness described by
γ
<1.
eit =
σ
it eityit (7)
eit=exp(−
ν
t)(8)
σ
it =(1+˜
kmit )−
γ
(9)
d
dt kmit =
ξ
mimit (10)
Parameter
ξ
mdetermines the effectiveness of investments imit . As before in the case
of productivity, we allow for spillovers but set the spillover intensity
ε
m
ij to
ε
m
ij =0
(i=j) in the base case and
ε
ii =1.
˜
kmit =∑
j
ε
m
ijkmjt (11)
To account for the stock pollutant character of global warming, we include a styl-
ized model of the climate system (Petschel-Held et al., 1999). Parameters of the
climate system are defined in Appendix A. The total stock of atmospheric green-
house gases cetgrows due to the instantaneous emissions of all countries
8
126 Chapter 5 Research cooperation and international standards
d
dt cet=∑
j
ejt (12)
and is linked to the greenhouse gas concentration conctaccording to
d
dt conct=Bcet+
β
P∑
j
ejt −
σ
P(conct−conc0)(13)
The concentration, in turn, determines the change of global mean temperature temp
by
d
dttempt=
μ
log(conct/conc0)−
α
P(tempt−temp0)(14)
For a detailed description of the climate equations and their parameters we refer to
the original publication.
Adapted from Nordhaus and Yang (1996), temperature changes cause climate change
damages, destroying a fraction 1 −Ωit of economic output:
Ωit =1/(1+dam1i(tempt)dam2i)(15)
yit =Ωit F(kit ,lit )(16)
The physical budget constraint closes the economy.
yit =cit +init +iait +imit (17)
2.2 Coalition Formation
Coalition formation is modeled as a two stage game. In the first stage, a membership
game is played, i.e. regions choose whether to become members and henceforth
act cooperatively on emission abatement with the other coalition members, or to
remain individual entities as non-members. In the second stage, the emission game,
non-members and the coalition (acting as one player) determine their emissions
indirectly by deciding on their consumption and investment behavior.
9
5.2 The Model 127
Coalition Stability
Among all possible coalitions, we consider stable coalitions in the sense of internal
and external stability of D’Aspremont and Gabszewicz (1986). Coalitions are in-
ternally stable if no member has an incentive to leave the coalition (Wi|S≥Wi|S\{i}
for i∈S), and externally stable if no non-member has an incentive to join (WjS>
WjS∪{j}for j/∈S). The coalition is thus self-enforced by economic incentives.
R & D Cooperation and Issue Linking
When applied to the provision of a public good, the motivation for issue linking
is to offset the incentive to free-ride on the non-excludable benefits of the public
good by the incentive to gain access to an (excludable) club good (Perez, 2005).
We adopt this view for our paper by identifying the coalition of regions dedicated
to cooperation on emission reduction with a club of regions that shares spillovers
from R & D.
Spillovers become a club good of the coalition S(a subset of the set of all regions
N) via the spillover intensities
ε
ij in Equation 6 or Equation 11, which compute the
weighted sums of productivity and mitigation, respectively. We set only spillover
intensities
ε
ij for i,j∈Sto non-zero levels. This restricts spillovers to coalition
members. In contrast, if spillovers of cooperative R & D within the coalition are
apublic good, spillovers extend to all regions, in which case we can set
ε
ij for
i∈S,j∈Nand i=jto positive values. We use the public good case when we test
credibility of the club good assumption.
International Standards
As argued in the introduction, standards on the technology level exhibit incentives
that foster a broad adoption of such standards on their own right. In this study, we
are interested in the effects of an existing standard on participation and issue link-
ing. Therefore, we assume that the decision of adopting the standards has already
taken place, i.e. this decision is exogenous to our model. 2
We implement international standards by requiring a reduction of endogenous emis-
2Adoption of the international standards may be viewed as a third stage game of the coali-
tion formation game taking place before the membership game: Players meet to decide on
the adoption of standards first, then, based on the (possibly partial) standards agreement,
go on to decide upon membership in the environmental agreement, and finally decide upon
emission strategies. In this setting, our assumption is that the outcome of the first stage
is adoption of standards by all players. This is also a welcome reduction of the computa-
tional burden (i.e. we only explore two out of nine possible outcomes of the first stage: full
adoption and no adoption at all).
10
128 Chapter 5 Research cooperation and international standards
sion intensity
σ
it by a fraction
θ
of the business-as-usual emission intensity, i.e. the
non-cooperative equilibrium intensity
σ
NE
it .3
σ
it ≤(1−
θ
)
σ
NE
it (18)
In effect, this implements a performance standard, which we use to approximate the
effect of technology standards. 4The implicit assumption is that a broad adoption
of technological standards aimed at low emissions technologies will translate to low
emission intensity on the macro-economic level. While this is plausible, it is clearly
desirable to check this assumption in a model with the necessary technological
detail in the future.
3Results
For our analyzes, we run the following experiments: To assess the impact of spill-
over intensity and the stringency of standards on stable coalition size, environmen-
tal effectiveness, and welfare, we systematically vary
θ
as well as
ε
a
ij and
ε
m
ij for
i,j∈S, with the coalition Sranging from the empty set to the set of all players
(see Equations 6, 11, and 18). For exploring the credibility of threatening exclusive
access to spillovers we additionally need to vary
ε
a
ij and
ε
m
ij for i∈Sand j∈N.
3.1 Participation in Environmental Cooperation
The first experiment looks at the effect of spillovers on coalition formation. We
plot the size of the largest stable coalition (participation) for different spillover
intensities.
3To avoid numerical infeasibility of the model, we implement a smooth transition from
no standards to the full level of standards in early years of the simulation.
4The literature distinguishes equipment standards (particularly technology standards) and
performance standards. The positive effects of technology standards cited in the introduc-
tion (network externalities, no lock-ins, economies of scale) are often due to the ability
of these standards to enforce compatibility. Performance standards are technology-neutral.
This characteristic is likely to increase their cost-effectiveness when applied to emission
reduction but they lack much of the positive incentives of equipment standards (Barrett,
2003, Ch. 9).
11
5.3 Results 129
0 0.5 1 1.5
2
3
4
5
6
7
8
9
0.0
0.019
0.1
0.168
0.213
0.26
0.3
0.337
0.0
0.00015
0.0006
0.00085
0.00097
0.00103
0.00109
0.00115
Spillover intensity (induced consumption gain in percent)
Coalition size
Mitigation technology
Productivity
Figure 1. Participation (size of the largest stable coalition) as a function of spillover in-
tensity. Spillover intensity is measured as induced consumption gain, i.e. the increase due
to spillovers in discounted consumption for the respective coalition size, relative to the
no-spillovers case (see also Footnote 5). The values of spillover intensity parameters
ε
m
and
ε
aare given next to the data points.
Cooperative R & D
To make spillovers of knowledge in mitigation technology and productivity com-
parable, we use induced consumption gains on the x-axis. Spillovers are manna
from heaven compared to an economy without spillovers, and the additional payoff
due to the same parameter value of the spillover intensity of productivity,
ε
a,or
mitigation technology,
ε
m, may vary. The induced consumption gain is the addi-
tional consumption due to spillovers for the coalition under consideration and thus
a proxy for its intensity. 5
In Figure 1 we observe the following: First, participation is low in absence of spill-
overs. This is in line with the literature and confirms that players in this model are
indeed facing a dilemma, i.e. the incentive to free-ride is large enough for play-
ers to act non-cooperatively. Second, for both kinds of spillovers participation rises
with spillover intensity. Again, this is in line with the literature. For high spillovers,
full cooperation is supported. Third, participation rises more rapidly in the case of
productivity cooperation. This is the case in terms of parameter values, which are
smaller by a couple of orders of magnitude, as well as, more importantly, in terms
of induced consumption gains.
To understand why productivity R & D is more effective in raising participation,
we take a closer look at how spillovers raise participation, i.e. create incentives for
larger stable coalitions. In particular, we take a look at payoffs received inside and
5Technically, we take the difference of consumption paths with and without spillovers,
discounted using a 3 percent discount rate. We convert to percentages of discounted base
case consumption. We prefer a consumption based metric to a welfare metric to make the
order of magnitude of the necessary spillovers easier to grasp.
12
130 Chapter 5 Research cooperation and international standards
2 3 4 5 6 7 8 9
0
20
40
60
80
100
120
Coalition size
Payoff (percent)
Inside
Outside
Inside with productivity spillovers
Outside with productivity spillovers
2 3 4 5 6 7 8 9
0
50
100
150
Coalition size
Payoff (percent)
Inside
Outside
Inside with mitigation technology spillover
Outside with mitigation technology spillover
Figure 2. Payoff inside and outside of a given coalition. The figures compare the inside
payoff received by a member of a coalition of size n(on the x-axis) with the outside payoff
of a non-member free-riding on the effort of a coalition of size n−1. An inside payoff
larger than an outside payoff indicates a stable coalition. We show payoffs for the base
case without spillovers, and one exemplary case with spillovers, for productivity (left) and
mitigation spillovers spillovers (right). The corresponding data points for stable coalitions
are circled. Payoffs are given as percentage of the difference between full cooperation and
no cooperation without any spillovers.
outside a given coalition, i.e. the inside payoff of a player within the coalition of
size nversus the outside payoff of the same player should she abandon the coalition
and instead face the remaining coalition of n−1 players as a non-member.
The left graph in Figure 2 shows payoffs for introducing spillovers in productivity,
the right hand graph shows results of introducing spillovers in mitigation technol-
ogy. Both figures show the case of no spillovers and one exemplary level of spill-
overs to illustrate the discussion; the argument presented holds for all intensities of
spillovers considered in this study.
Without spillovers the payoffs both inside and outside any given coalition rise with
the size of the coalition. Inside the coalition the payoff rises because the emission
externality is increasingly internalized. Outside the coalition, players free-ride on
the abatement effort of the coalition, which becomes increasingly more ambitious
as participation rises and thus the benefit of free-riding increases. The curves of in-
side payoff and outside payoffs intersect before coalition size 3, marking a coalition
of 2 as the largest stable coalition.
What changes when spillovers are introduced? Spillovers are restricted to coali-
tion members only, therefore in case of productivity the outside payoff curve re-
mains unchanged. Member payoffs increase with spillovers, thus shifting the in-
side payoff curve upwards and tilting it to the left because spillovers affect larger
coalitions more strongly: there simply are more players benefiting from them. In
13
5.3 Results 131
00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
3
4
5
6
7
8
9
ε
m
= 0
ε
m
= 0.019
ε
m
= 0.1
ε
m
= 0.168
ε
a
= 0
ε
a
= 0.00015
ε
a
= 0.00085
ε
m
= 0
ε
m
= 0.019
ε
m
= 0.1
ε
m
= 0.168
ε
a
= 0
ε
a
= 0.00015
ε
a
= 0.0006
ε
a
= 0.00085
Standard stringency θ
Coalition size
Mitigation technology
Productivity
00.5 11.5
2
3
4
5
6
7
8
9
Coalition size
Mitigation technology
00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
2
3
4
5
6
7
8
9
Spillover intensity (induced consumption gain in percent)
Coalition size
Productivity
no θ
0.7
0.5
0.3
0.1
0
no θ
0.7
0.5
0.3
0.1
0
Figure 3. Effect of international standards. The left figure shows the effect of standards on
participation for selected spillover intensities. The figures on the right shows the effect of
standards on the spillover-participation relationship analogous to Figure 1.
effect, this moves the intersection of inside and outside payoff curves to the right—
participation increases.
Spillovers in mitigation technology shift and tilt the inside payoff curve in the
same way, upwards and to the left. However, in contrast to the case of produc-
tivity, the outside payoff curve is tilted counterclockwise, too. 6Whereas produc-
tivity of coalition members hardly affects non-members, spillovers in mitigation
technology lead to an increased abatement effort by reducing the abatement costs
for coalition members. Reduced global emissions, however, have an effect on all
players: non-members, too, enjoy these additional emission reductions in form of
reduced damages. Thus the positive effect on the inside payoff curve is partially
offset by the tilting outside payoff curve—participation is still increased, but less
effectively.
Technology Standards
In the following experiments, we combine spillovers with standards, i.e. we in-
troduce standards in a world where simultaneously research cooperation is imple-
mented.
Figure 3 shows participation as a function of the stringency of the technology stan-
dard (left). The stringency
θ
indicates the prescribed reduction of emission inten-
sity relative to emission intensity in non-cooperative equilibrium (Equation 18).
Technology standards by themselves (i.e. for
ε
=0) have very little impact. Partic-
ipation remains low with only a temporary increase by one member at
θ
=0.2and
6We do not observe an upward shift of the outside payoff curve the way to inside payoff
curve is shifted. It simply rotates around the fixed-point (2,0)because the outside payoff
of a coalition of 2 is simply the non-cooperative equilibrium where there are no spillovers
irrespective of the spillover intensity parameter. In contrast, the fixed-point of the tilting
inside payoff curve is (1,0), which we observe as a tilting and shift upwards in the range
of coalition from 2 to 9.
14
132 Chapter 5 Research cooperation and international standards
2 3 4 5 6 7 8 9
0
20
40
60
80
100
120
Coalition size
Payoff (percent)
Inside
Outside
Inside with productivity spillovers
Outside with productivity spillovers
2 3 4 5 6 7 8 9
0
50
100
150
Coalition size
Payoff (percent)
Inside
Outside
Inside with mitigation technology spillover
Outside with mitigation technology spillover
Figure 4. Inside payoff and outside payoff with standards. Analogous to Figure 2, the graphs
show inside payoff (received by a member of a coalition of size n), which is larger than
outside payoff (received by a non-member facing a coalition of size n−1) if the coalition
is stable. Spillover on the left are in productivity and in mitigation technology on the right.
For stable coalitions the corresponding data points are circled. Payoff is scaled to the gap
between no cooperation (0 percent) and full cooperation (100 percent) in the base model
without spillovers.
θ
=0.3. 7Only in combination with spillovers, standards raise participation sub-
stantially. Likewise, the positive effect of spillovers on coalition size is strengthened
by standards (Figure 3, right).
Why do standards hardly change participation by themselves, but they do enlarge
stable coalitions when combined with spillovers? Again, we take a look at payoffs
inside and outside the coalitions for mitigation spillovers and productivity spill-
overs (Figure 4).
Standards guarantee investment in abatement beyond the level of Nash equilibrium
without standards. Hence, ambitious standards reduce climate change damages and
give all players higher payoffs even in non-cooperative equilibrium or in presence
7Without standards, emission intensity is lower for coalition members compared to non-
members. With increasing standards stringency, non-members are forced to abate more.
The coalition benefits to the extent that coalitions of three instead of coalitions of two
become stable. However, due to (a) the small coalition size and (b) the absence of spillovers,
emission intensities within the coalition do not differ much from emission intensities of
non-members. Hence with the stringency of standards increasing furthermore, there soon
comes a point where standards also affect the abatement behavior of coalition members.
This lowers the coalition welfare enough to destabilize the coalition of three. Stability of
larger coalitions, or coalitions in calculations with non-zero spillovers are not affected in
this way, because the emission intensity within the coalition is lower to begin with and is
therefore not affected by standards.
15
5.3 Results 133
of small coalitions. Compared to the case without standards in Figure 2, payoff in
non-cooperative equilibrium (outside payoff for coalition size 2) is now lifted half
way towards payoff for fully cooperative behavior (50 percent of the gap between
no cooperation and full cooperation).
The distance between non-cooperative and fully cooperative solutions has therefore
been decreased. In the absence of spillover effects, however, this does not facilitate
more participation since the relative position of inside and outside payoff curves
is not affected, i.e. outside payoff grows more rapidly with coalition size than the
inside payoff and soon (at coalition size 3) exceeds it. Spillovers make a difference
because, as discussed above, they shift the curve of inside payoffs upwards, thus
delaying the interception of the two curves and hence increasing the size of the
largest stable coalition.
The argument holds for spillovers in mitigation technology as well as productivity.
Again, the latter is more effective in raising participation because here there outside
payoff is largely unaffected by spillovers.
3.2 Environmental Effectiveness and Welfare Effects
In the previous section we have seen under which circumstances cooperative R & D
and technology standards may raise participation. This section explores the impli-
cations of increased participation for environmental effectiveness and for global
welfare. We begin the analysis by turning to cooperative R & D.
Cooperative R & D
Figure 5 shows environmental effectiveness relative to socially optimal emission
levels in absence of spillovers as the reference point (i.e. 100 percent, whereas
emissions from non-cooperative behavior are scaled to 0 percent). Environmental
effectiveness increases with spillover intensity in a very similar way to participation
(Figure 1), indicating that coalition size is a major determinant and hence a good
proxy for environmental effectiveness in this model.
An interesting difference to participation is that environmental effectiveness is ex-
ceeded in case of mitigation technology spillovers but not for spillovers of produc-
tivity. The reason is that spillovers in mitigation technology decrease abatement
costs and therefore a cleaner environment becomes socially optimal.
The impact of mitigation spillovers on environmental effectiveness offsets some of
the drawbacks of mitigation spillovers in terms of participation: Figure 1 stressed
that achieving full cooperation required larger spillover intensities in case of miti-
gation technology. This is also true for environmental effectiveness. However, Fig-
16
134 Chapter 5 Research cooperation and international standards
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0
20
40
60
80
100
120
3
4
5
6
7
8
9
3
4
5
6
7
8
9
Environmental effectiveness (percent)
Spillover intensity (induced consumption gain in percent)
non−cooperative solution
fully cooperative solution
Mitigation technology
Productivity
Figure 5. Environmental effectiveness of cooperative R & D. This figure shows environ-
mental effectiveness for stable coalitions, where zero percent is the emission level in ab-
sence of spillovers and coalitions, and 100 percent describes socially optimal emissions in
an economy without spillovers. Spillover intensity is measured in consumption gain (see
Figure 1). We indicate the size of the respective stable coalitions next to the data points.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
4
6
8
Coalition size
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
4
6
8
Coalition size
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
4
6
8
Coalition size
Mitigation technology
Productivity
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
2
4
6
8
Coalition size
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0
50
100
150
3
4
5
6
7
8
9
3
4
5
6
7
8
9
Spillover intensity (induced consumption gain in percent)
Global welfare (percent of gap)
non−cooperative solution
fully cooperative solution
Figure 6. Welfare effects of cooperative R & D. In this figure we show how global welfare
of stable coalitions increases with spillover intensity (measured in consumption gain, see
Figure 1). Much of the effect is due to rising participation (see Figure 1), hence we indicate
coalition size next to the respective data points.
ure 5 shows that the difference in spillover intensity to achieve 100 percent environ-
mental effectiveness is less than the difference in achieving full cooperation. Still,
productivity cooperation remains the more effective incentive.
Figure 6 shows the welfare effect of stable coalitions. Welfare is normalized to
the non-cooperative behavior (0 percent) and full cooperation (100 percent) in an
economy without spillovers. Again, we find a similar picture to participation and
environmental effectiveness. Participation, or the degree of cooperation, is also a
strong determinant of global welfare. Global welfare exceeds 100 percent of wel-
fare without spillovers for both cases of R & D cooperation, highlighting the fact
that spillovers are manna from heaven, i.e. compared to an economy without spill-
17
5.3 Results 135
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
200
250
300
350
400
450
Cumulative Emissions [GtC]
εm = 0
εm = 0.019
εm = 0.1
εm = 0.168
εm = 0.213
εm = 0.26
εm = 0.3
Standard stringency θ
Social Optimum
Nash Equilibrium
Nash Equilibrium with standards
Stable coalition with standards
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
20
40
60
80
100
Welfare [normalized]
Standard stringency θ
εm = 0
εm = 0.019
εm = 0.1
εm = 0.168
εm = 0.213
εm = 0.26
εm = 0.3
Social Optimum
Nash Equilibrium
Nash Equilibrium with standards
Stable coalition with standards
Figure 7. Effects of standards on cumulative emissions (left) and welfare (right). Stable
coalitions are induced by spillovers of mitigation technology, the spillover parameter
ε
mis
included next to the corresponding curves.
overs they provide an additional free income.
Technology Standards
This section explores environmental effectiveness and welfare implications of im-
posing technology standards. When standards are stringent enough, they might
solve the environmental dilemma by themselves irrespective of any cooperation
agreements on environment or R & D. We look at the effectiveness and the welfare
implications of standards and assess the scope that cooperative agreements have in
this setting.
Figure 7 shows the effect of standards on emissions in Nash equilibrium and in case
of stable coalitions. When the stringency of the standard is increased, the effect
on cumulative global emissions is to bring them down towards their optimal level
and below. As cumulative global emissions approach their optimum levels, so does
global welfare (Figure 7). However, welfare does not reach its optimum level but
starts to fall before emissions reach the optimum. This is due the fact that the timing
of emission intensity reduction prescribed by the standards are not cost-effective.
It is the inefficiency of standards as a policy instrument manifesting in this figure.
This disadvantage of command and control instruments like standards compared
to market or price incentive based instruments is well known (see e.g. Requate,
2005). Indeed any of the levels of cumulative emissions in the previous figure could
likely be reached at lower costs and higher global welfare if the timing of emission
intensity reduction was not prescribed but chosen optimally.
Cooperative agreements on environment and R & D can bridge this gap: Figure 3
includes welfare levels for a number of coalitions that are stable at the given stan-
dard stringency due to including cooperative R & D (spillovers) in the agreement.
Standards that fall short of enforcing optimal emission levels and are inefficient to
begin with, may still be sufficient to induce full cooperation in combination with
some spillovers. We observe that often the standards that were necessary to sta-
18
136 Chapter 5 Research cooperation and international standards
0 0.1 0.2 0.3 0.4 0.5 0.6
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
C
re
dibili
ty
(
sta
bl
e coa
li
t
i
ons
)
Productivity
0.7
0.5
0.3
0.1
0
0 0.5 1 1.5
−0.16
−0.14
−0.12
−0.1
−0.08
−0.06
−0.04
−0.02
0
Credibility (stable coalitions)
Mitigation Technology
0.7
0.5
0.3
0.1
0
Figure 8. Credibility of restricting spillovers to coalition members. We compute credibility
as the difference of coalition members’ welfare with restricted spillovers minus the case of
spillovers to all regions, hence positive values imply credibility.
bilize full cooperation are exceeded under full cooperation—otherwise standards
would distort the optimal solution resulting in below-optimal welfare levels.
3.3 Credibility of Exclusive R & D Cooperation
Restricting spillovers to coalition members is only credible if coalition members are
not worse off compared to the case where knowledge is public, i.e. spillovers are
unrestricted. Hence we investigate the credibility of exclusive R & D cooperation by
comparing it to a scenario where R & D spillovers extend to all regions and not just
coalition members. We continue to assume that only coalition members participate
in R & D cooperation, i.e. spillovers extend to non-members but not vice versa.
Figure 8 shows whether restricting spillovers to the coalition is beneficial to its
members. Values are plotted for different stringencies of standards and only for sta-
ble coalitions. In case of productivity we find that threatening exclusiveness is cred-
ible for all stable coalitions, and all unstable coalitions as well (not shown). There
is no advantage for coalition members in boosting productivity for non-members.
Quite the contrary, the increased productivity would entice non-members to pro-
duce and pollute more.
Excluding non-members from spillovers of R & D in mitigation technology is al-
most always a non-credible threat for stable coalitions (Figure 8, right). Coalition
members benefit from letting spillovers extend to non-members, because the spill-
overs add to non-member abatement and further reduce the emission intensity and
actual emissions of the non-members. Coalition members then benefit from reduced
climate change damages. This is a crucial difference to productivity spillovers that
do not have this feedback onto the coalition.
For both kinds of spillovers, credibility approaches zero for small as well as for
large coalitions and exhibits a maximum for medium coalition sizes. This depends
19
5.3 Results 137
0
0.5
1
1.5
2
2.5
3
0.02
0.005
0.04
0.01
2
10.02
0.005
10
2.5 −0.5
−0.7 0.02
0.001
2
0.5 0.09
0.05
0.3
0.1
0.2
0.1
Productivity
0.02
0.005
0.04
0.01
2
10.02
0.005
10
2.5 −0.5
−0.7
0.02
0.001
2
0.5 0.09
0.05
0.3
0.1
0.2
0.1
Mitigation
Default parameter values
0.01 0.02 1.5 0.01 5 −0.6 0.01 1 0.07 0.2 0.15
Spillover intensity (induced consumption gain)
grl
dam1
dam2
νξ
m
γ
ρ
ηξ
a
Φ
λ
low value
high value
default value
Figure 9. Spillover intensities that support full cooperation. We show the impact of param-
eter variation on the effectiveness of research cooperation by giving the lowest spillover
intensity that induces full cooperation. Results for cooperation on productivity and miti-
gation are shown side by side for each parameter; numbers above and below data points
indicate the low and high values used.
on the extent of spillovers to nonmembers: the smaller coalitions are, the lower
the number of players generating spillovers. On the other end of the spectrum, the
larger the coalition, the lower the number of nonmembers receiving spillovers.
4 Sensitivity of Key Results
This section explores the sensitivity of key results towards variation of input pa-
rameters. Central results of the preceding sections are that TOA may sustain full
cooperation depending on the spillover intensity, and that linking to productivity
cooperation is generally more effective in raising environmental cooperation than
linking to cooperative research on mitigation.
We explore in how far these results continue to hold when parameter values change
by running high value and low value scenarios for key parameters. An assessment
of global sensitivities, i.e. a simultaneous variation of all parameters, would be
preferable because it accounts for the fact that sensitivity of the results for variation
of one parameter will in general depend on all other parameters. We stick with an
exploration of local sensitivities to limit the computational burden.
Figure 9 shows results from these low value/high value calculations. Using full co-
operation as a reference point, Figure 9 reports the spillover intensity necessary for
the grand coalition of all players to be stable. The first message from this figure
is that in all variations, either full cooperation was sustained by raising spillover
20
138 Chapter 5 Research cooperation and international standards
intensities or was even achieved at lower spillover intensities. Thus, R & D cooper-
ation proves to be a sufficiently strong incentive for all parameter values in these
variations. More importantly though, the spillover intensities necessary to achieve
full cooperation via cooperative mitigation research are always higher than in case
of the corresponding calculation featuring cooperation on productivity. Hence, our
this finding is also robust with respect to our parameter variations.
Table 1 summarizes our choice of high and low parameter values and also reports
the impact of these parameter variations on more key results. The difference be-
tween cooperation on productivity versus mitigation is measured by the “difference
in incentives” in columns 6-7, reported as the difference in spillover intensities that
are sufficient to stabilize full cooperation. In Figure 1 this is the distance between
the topmost data points of mitigation cooperation and productivity cooperation on
the x-axis. This difference is considerably affected by parameter changes, mostly
in the range of plus/minus thirty percent of the default, yet it is always positive and
larger than 0.5 percent, indicating that R & D cooperation on productivity remains
significantly more effective than cooperation on mitigation R & D.
Similarly, columns 8-9 show the difference in environmental effectiveness for the
same stable grand coalitions from Figure 9. We take the metric of environmental
effectiveness from Figure 5, i.e. the numbers in this table measure the difference
of the topmost data points in Figure 5 on the y-axis. Analogously, columns 10-11
show the impact of parameter variation on the difference in global welfare of grand
coalitions, i.e. the distance of the topmost data points in Figure 6 on the y-axis. The
values in columns 8-11 of Table 1 show that even though parameter variation has a
considerable impact, our conclusions remain intact.
5 Summary and Conclusions
We assessed different technology oriented agreements (TAO) in a conceptual model
and had to resort to numerical solutions. Naturally, any conclusions from these
result about the economy described by the model must be taken with a grain of
salt. Nevertheless, the model suggests some rather general differences between the
selected TOA, which we summarize in the following.
Cooperative R & D in mitigation technology is less effective because via emissions
reductions, spillovers of mitigation technology raise both, the coalition payoff and
the free-rider incentive. This feedback of mitigation reduces the positive incentive
of spillovers on coalition formation making cooperative R & D that is unrelated to
emission abatement a more attractive option for setting incentives for participation.
Contrary to R & D in productivity, R & D in mitigation technology has a positive
impact on the environment by reducing abatement costs. Indeed the same level of
21
5.5 Summary and Conclusions 139
Table 1: Parameter values and effectiveness difference. Columns 3-5 list the default, low, and high parameter values. Columns 6-7 report the
difference in spillover intensity between cooperation on mitigation and productivity to achieve full cooperation. For default parameters this value
is 0.85, the metric for spillover intensity is induced consumption gain, see Figure 1 or Footnote 5. Columns 8-9 and columns 10-11 report the
corresponding difference in environmental effectiveness and global welfare, respectively. Here, the default values are 30.0 and 41.5, respectively.
Parameter values Diff. in Incentive Env. Effect. Welfare Effect
Parameter Symbol Default Low High Low High Low High Low High
Pure rate of time preference
ρ
0.01 0.001 0.02 1.03 0.91 24.739.640.044.1
Mare rate of time preference
η
1.00.52.00.96 0.92 25.740.340.544.2
Growth rate of labor supply grl 0.01 0.005 0.02 0.61 1.62 27.141.842.640.2
Rate of decarbonization
ν
0.01 0.005 0.02 1.00 0.61 47.314.540.641.6
Effectiveness of investments in km
ξ
m5.02.510.01.11 0.63 39.622.542.139.9
Effectiveness of investments in a
ξ
a0.07 0.05 0.09 0.74 0.93 27.132.241.240.7
Stepping on toes effect
λ
0.15 0.10.20.79 0.91 28.731.340.341.3
Standing on shoulders effect Φ0.20.10.30.77 0.94 27.932.541.340.6
Abatement cost exponent −
γ
0.60.50.70.66 1.07 21.243.435.448.7
Damage function exponent dam21.51.02.00.57 1.09 42.623.242.340.0
Damage function coefficient dam10.02 0.01 0.04 0.56 1.23 39.522.641.340.8
22
140 Chapter 5 Research cooperation and international standards
environmental effectiveness can be reached with smaller coalitions using R & D co-
operation in mitigation technology rather than productivity. Nevertheless, the spill-
over intensity necessary to reach this same level of environmental effectiveness is
larger than in case of productivity.
Moreover, our model suggests that restricting spillovers exclusively to the coali-
tion is non-credible in case of mitigation technology. This is plausible because in
a world with a global warming problem, it is desirable to let advanced mitigation
technology diffuse as much as possible. Overcoming non-credibility due to eco-
nomic reasons may be possible by means exogenous to this model, for example
by commitment (e.g. Houba and Bolt, 2002, Ch. 7), which could be enforced by
reputation or eliminating the alternatives. Nevertheless, this is a complication that
is absent in productivity spillovers.
This impact of the source of spillovers could be one of the reasons why Nagashima
and Dellink (2008) only find small effects of spillovers related to marginal abate-
ment costs, whereas Botteon and Carraro (1998) observe a significant increase of
participation up to full cooperation due to spillovers that reduce production costs. 8
We argued that if technology standards are easier to agree upon than a cooperative
environmental agreement, then adopting an agreement on standards may be a help-
ful first step towards an international environmental agreement. Our model suggests
that this works when standards cause emission reductions for non-members but are
fulfilled voluntary by coalition members. Here, this is the case only when at least
some cooperative R & D is carried out, setting the abatement levels of members and
non-members far enough apart.
A combination of technology standards and cooperative R & D is also promising
for a second reason. International standards by themselves reduce emissions in a
way that is not cost efficient. Combined with cooperative R & D, however, they
may induce environmental cooperation to an extend beyond standards, therefore
making its inefficiency unimportant.
Limitations
This study aimed to identify general cause-effect relationships in the interplay of
TAO and IEA. The simplifying assumptions of (ex ante) identical regions and lack
of technological detail facilitated the analysis, but at the same time they reduce
8Of course, the models used in Botteon and Carraro (1998) and Nagashima and Dellink
(2008) differs in many respects from this model, among them are: a different modeling
framework, heterogeneity of players, and inclusion of transfers within the coalition. The
feedback of a stronger abatement effort (due to lower abatement costs) onto non-members
ought to be present in the model nonetheless. It is also not clear how the assumed spillover
intensities in the different models compare.
23
5.5 Summary and Conclusions 141
the scope of its conclusions for real world policy. Therefore, testing the lessons
learnt from this study in models with heterogeneous regions and explicit technology
choice would be a step to confirm them and elaborate on their implications.
In particular, we analyzed the interaction of standards and spillovers from a purely
macro-economic perspective, arguing that standards come into force due to incen-
tives that are exogenous to the model. Recent integrated assessment models (e.g.
Bosetti et al., 2006) resolve some technological detail providing the basis to imple-
ment standards on the technology level and allow to explore the scope of the results
of this paper in a less conceptual setting.
Moreover, we argued that the spillover extent could be fostered through govern-
mental programs, assuming that this is possible at no additional societal costs.
While this assumption is backed by the very idea of R & D spillovers, namely that
R & D generates particularly high returns, it does not account for crowding out of
other R & D.
Acknowledgements
Discussions of previous versions of this paper with Carlo Carraro, Robert Mar-
schinski, Michael Finus, and Carsten Helm greatly helped to sharpen our ideas,
which we gratefully acknowledge. All remaining mistakes are, of course, our own.
The model experiments made extensive use of the SimEnv (Flechsig et al., 2008)
multi-run simulation environment. Kai Lessmann received funding from the Euro-
pean Commission within the ADAM project (project 018476-GOCE).
References
Barrett, S., 1991a. Economic analysis of international environmental agreements:
lessons for a global warming treaty. In: OECD, Responding to Climate Change:
Selected Economic Issues. OECD, Paris, pp. 111–49.
Barrett, S., 1991b. Economic instruments for climate change policy. In: OECD,
Responding to Climate Change: Selected Economic Issues. OECD, Paris, pp.
53–108.
Barrett, S., 1994. Self-enforcing international environmental agreements. Oxford
Economic Papers 46, 878–894.
Barrett, S., 2003. Environment and Statecraft: The Strategy of Environmental
Treaty-Making. Oxford University Press.
Barrett, S., Stavins, R., 2003. Increasing participation and compliance in interna-
tional climate change agreements. International Environmental Agreements: Pol-
itics, Law and Economics 3, 349–376.
Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., Tavoni, M., 2006. WITCH: A
World Induced Technical Change Hybrid model. The Energy Journal Special
24
142 Chapter 5 Research cooperation and international standards
Issue Hybrid Modelling of Energy Environment Policies: Reconciling Bottom-
up and Top-down, 13–38.
Botteon, M., Carraro, C., 1998. Game Theory and the Global Environmental. Ed-
ward Elgar Publishing, Cheltenham, UK, Ch. Strategies for environmental nego-
tiations: issue linkage with heterogeneous countries, pp. 180–200.
Buchner, B., Carraro, C., 2006. US, China and the economics of climate negotia-
tions. International Environmental Agreements 6, 63–89.
Buchner, B., Carraro, C., Cersosimo, I., Marchiori, C., 2005. The Coupling of Cli-
mate and Economic Dynamics. Vol. 22 of Advances In Global Change Research.
Springer Netherlands, Ch. Back to Kyoto? US Participation and the Linkage be-
tween R&D and Climate Cooperation, pp. 173–204.
Carraro, C., 1999. International Environmental Agreements on Clime Change.
Kluwer Academic Publishers, Ch. The structure of international environmental
agreements, pp. 9–25.
Carraro, C., Eyckmans, J., Finus, M., 2006. Optimal transfers and participation de-
cisions in international environmental agreements. Review of International Or-
ganizations 1 (4), 379–96.
Carraro, C., Siniscalco, D., 1992. The international dimension of environmental-
policy. European Economic Review 36 (2-3), 379–387.
Carraro, C., Siniscalco, D., 1993. Strategies for the international protection of the
environment. Journal Of Public Economics 52 (3), 309–328.
Carraro, C., Siniscalco, D., 1995. The Economics of Sustainable Development.
Cambridge University Press, Ch. Policy Coordination for Sustainability: Com-
mitments, Transfers, and Linked Negotiations.
Carraro, C., Siniscalco, D., 1997. International Environmental Agreements –
Strategic Policy Issues. Edward Elgar Publishing, Ch. R&D Cooperation and
the Stability of International Environmental Agreements, pp. 71–95.
David, P., Greenstein, S., 1990. The economics of compatibility standards: An in-
troduction to recent research. Economics of Innovation and New Technology
1 (1), 3–41.
de Coninck, H., Fischer, C., Newell, R. G., Ueno, T., 2007. International
technology-oriented agreements to address climate change. Energy Policy, 335–
356.
D’Aspremont, C., Gabszewicz, J. J., 1986. New Developments in the analysis of
Market Structures. Macmillan, New York, Ch. On the stability of collusion, pp.
243–64.
Edenhofer, O., Bauer, N., Kriegler, E., 2005. The impact of technological change
on climate protection and welfare: Insights from the model MIND. Ecological
Economics 54 (2-3), 277–292.
Edenhofer, O., Lessmann, K., Kemfert, C., Grubb, M., Koehler, J., 2006. Induced
technological change: Exploring its implications for the economics of atmo-
spheric stabilization – synthesis report from the Innovation Modeling Compar-
ison Project. Energy Journal Special Issue Endogenous Technological Change
and the Economics of Atmospheric Stabilisation, 57–107.
Edmonds, J., Wise, M., 1999. International Environmental Agreements on Climate
25
5.6 References 143
Change. Kluwer Academic Publishers, Ch. Exploring a technology strategy for
stabilizing atmospheric CO2, pp. 131–154.
Eyckmans, J., Tulkens, H., 2003. Simulating coalitionally stable burden sharing
agreements for the climate change problem. Resource and Energy Economics
25, 299–327.
Farrell, J., Saloner, G., 1987. Product Standardization and Competetive Strategy.
North-Holland, New York, Ch. Competition, Compatibility, and Standards: the
Economics of Horses, Penguins, and Lemmings, pp. 1–22.
Finus, M., 2008. Game theoretic research on the design of international environ-
mental agreements: Insights, critical remarks, and future challenges. Interna-
tional Review of Environmental and Resource Economics 2, 29–67.
Finus, M., van Ierland, E., Dellink, R., 2006. Stability of climate coalitions in a
cartel formation game. Economics of Governance 7, 271–91.
Flechsig, M., Böhm, U., Nocke, T., Rachimow, C., 2008. The multi-run simulation
environment SimEnv. Http://www.pik-potsdam.de/software/simenv/.
URL http://www.pik-potsdam.de/software/simenv/
Griliches, Z., 1992. The search for R&D spillovers. Scandinavian Journal of Eco-
nomics 94 (0), 29–47.
Hagedoorn, J., Link, A. N., Vonortas, N. S., 2000. Research partnerships. Research
Policy 29 (4-5), 567–586.
Houba, H., Bolt, W., 2002. Credible Threats in Negotiations – A Game-theoretic
Approach. Kluwer Academic Publishers, Boston, Dordrecht, London.
Jones, C. I., Williams, J. C., 1998. Measuring the social return to R&D. Quarterly
Journal of Economics 113 (4), 1119–1135.
Katsoulacos, Y., 1997. International Environmental Agreements – Strategic Policy
Issues. Edward Elgar Publishing, Ch. R&D spillovers, cooperation, subsidies and
international agreements, pp. 97–109.
Kemfert, C., 2004. Climate coalitions and international trade: assessment of coop-
eration incentives by issue linkage. Energy Policy 32, 455–65.
Kypreos, S., Bahn, O., 2003. A MERGE model with endogenous technological
progress. Environmental Modeling and Assessment 8 (3), 249–259.
McDougall, R., Elbehri, A., Truong, T., 1998. Global trade assistance and protec-
tion: The GTAP 4 data base, center for Global Trade Analysis, Purdue University.
Nagashima, M., Dellink, R., 2008. Technology spillovers and stability of interna-
tional climate coalitions. International Environmental Agreements: Politics, Law
and EconomicsDOI 10.1007/s10784-008-9079-1.
Nordhaus, W. D., Yang, Z., 1996. A regional dynamic general-equilibrium model
of alternative climate-change strategies. The American Economic Review 86 (4),
741–765.
Perez, O., 2005. Multiple regimes, issue linkage, and international cooperation: Ex-
ploring the role of the WTO. University of Pennsylvania Journal of International
Economic Law 26 (4), 735–778.
Petschel-Held, G., Schellnhuber, H.-J., Bruckner, T., Tóth, F. L., Hasselmann, K.,
1999. The tolerable windows approach: Theoretical and methodological founda-
tions. Climatic Change 41 (3), 303–331.
26
144 Chapter 5 Research cooperation and international standards
Philibert, C., 2003. Technology innovation, development and diffusion. OECD
and IEA Information Paper COM/ENV/EPOC/IEA/SLT(2003)4, OECD Envi-
ronment Directorate and International Energy Agency.
Philibert, C., 2004. Lessons from the Kyoto protocol: Implications for the future.
International Review for Environmental Statistics 5 (1), 311–322.
Ramsey, F., 1928. A mathematical theory of saving. Economic Journal 38, 543–
559.
Requate, T., 2005. Dynamic incentives by environmental policy instruments – a
survey. Ecological Economics 54 (2-3), 175–195.
Turnovsky, S. J., 2000. Methods of Macroeconomic Dynamics. MIT Press.
Wagner, U. J., 2001. The design of stable international environmental agreements:
Economic theory and political economy. Journal of Economic Surveys 15 (3),
377–411.
A Parameter Choices
Table A.1 lists our choice of parameters. We restrict this study to the case of sym-
metric players, hence a calibration to real world regions is out of question. Nev-
ertheless we selected a set of parameters that is plausible in light of the empirical
literature. This appendix lists the assumptions we made.
Parameter
ξ
adrives endogenous growth. We chose its value such that economic
output shows a 2.5 percent annual growth in the first century.
Parameters in the climate module are based on literature values, giving us a 3◦C
temperature increase by 2100, and a 7.5◦C increase by 2200 in non-cooperative
equilibrium and business as usual, i.e. without climate change damages.
The damage function was chosen such that in non-cooperative equilibrium damages
in 2100 are 6 percent. Within the mitigation option, parameters
γ
and
ξ
mwere
selected such that optimal abatement (the social optimum solution) reduces the
temperature increase in 2100 to 2.4◦C.
27
5.7 Appendix: Parameter Choices 145
Table A.1
Parameter values
Parameter Symbol Value
Pure rate of time preference
ρ
0.01
Elasticity of marginal utility
η
1
Income share capital
β
0.35
Growth rate of labor supply grl 0.01
Exogenous rate of decarbonization
ν
0.01
Initial labor l01
Initial labor productivity a01
Initial capital stock k034
Effectiveness of investments in a
ξ
a0.023
Effectiveness of investments in km
ξ
m5.0
Abatement cost exponent
γ
0.2
Ocean biosphere as CO2source
β
P0.47
Atmospheric retention factor B1.51e-3
Radiative temperature driving factor
μ
8.7e-2
Temperature damping factor
α
P1.7e-2
Ocean biosphere as CO2sink
σ
P2.15e-2
Initial concentration conc0377
Initial temperature temp00.41
Initial cumulative emissions cume0501
Damage function coefficient dam1 0.02
Damage function exponent dam21.5
28
146 Chapter 5 Research cooperation and international standards
Chapter 6
Synthesis and Outlook
The starting point for this thesis is set by two assumptions: first, action to mitigate climate
change is necessary, and second, technologies will play a key role in this effort because
technology and technological change facilitate the reduction of anthropogenic greenhouse
gas emissions. Both assumptions are supported by the latest scientific findings reported
in IPCC WG3 (2007). As a consequence, the way technological change is described in
integrated assessment models of climate change is of great importance, and a sound un-
derstanding of endogenous technological change and its interaction with climate policies
is needed.
There is empirical evidence that technological change is induced by policies, but results
from previous modeling assessments of induced technological change (ITC) have been
ambiguous about the responsiveness of technological change to climate policies and its
potential to reduce the costs of mitigating climate change. On the other hand, inducing
this kind of technological progress requires a clear climate policy. Stern (2007) concludes:
“Without a ‘loud, legal and long’ carbon price signal, in addition to direct support for
R & D, the technologies will not emerge with sufficient impact.” The carbon price signal
ought to extend globally in order to prevent carbon leakage and achieve efficiency, but
according to the literature on international environmental agreements the prospect for
global cooperation on climate policy is not bright.
This raises two broad research questions: First, what is the role of ITC for climate change
mitigation? And second, if there is a desirable contribution of ITC to mitigation, how can
we achieve the global policy that triggers this technological change?
The four papers presented in this thesis contribute to these two questions. The first two
papers explore the role of ITC within a single integrated assessment model (Chapter 2)
and across a broad range of models in a model comparison excercise (Chapter 3). The
remaining two papers address the second question, i.e. achieving a global policy, by look-
ing at the prospect of achieving high participation in a self-enforcing international cli-
mate agreement by linking climate policy negotiations to trade sanctions (Chapter 4) and
technology-oriented agreements (Chapter 5).
This chapter synthesizes these four papers, proceeding as follows: First, I summarize
Chapters 2 and 3 and discuss the role of ITC for mitigation strategies. Then I summarize
Chapters 4 and 5 and discuss issue linking. The thesis concludes with an outlook on
possible extensions of this work.
147
148 Chapter 6 Synthesis and Outlook
6.1 Induced technological change in integrated
assessment modeling
In the introduction of this thesis, I broke down the two broad research questions con-
cerning the role of ITC for mitigation on the one hand, and the necessary climate policy
regime on the other hand, in four sets of questions. The following sections answer these
questions drawing on the insights from Chapters 2–5.
6.1.1 Implications of ITC in the MIND model
The first set of questions focuses on the impact of ITC on the costs and strategies of
mitigating climate change.
•What is the impact of ITC on mitigation policy scenarios?
•What is the role of economy wide feedbacks concerning ITC?
•What are the implications of ITC in particular for mitigation costs and mitigation
strategies, i.e. the optimal composition of mitigation options?
MIND is a model built for the integrated assessment of climate change and global eco-
nomic development (Edenhofer et al., 2005a). Its novelty is that it incorporates macro-
economy and energy system, albeit in a stylized way, and endogenous technological
change (ETC) throughout the economy.1That is, macro-economic growth is driven by
ETC, and ETC is also implemented in the energy system sectors. This makes an analysis
of MIND well suited to address these question. Indeed, I find that in MIND, ITC has
significant impact on both, costs and strategies of mitigation.
In particular, the analysis reveals two “directions” of technological change (Table 6.1).
First, there is technological change that permeates the entire economy—this is reflected
in a strong impact on the overall macro-economic costs of mitigation policy. This is the
case for R & D that augments overall labor productivity or energy efficiency, or ETC in
the resource extraction sector, which has impact on the entire economy because of the
strong effect of cheap fossil fuel on economic growth in the baseline. And then there is
technological change whose impact is specific to a single sector, the energy sector, as evi-
dent from a strong impact on the contribution to mitigation options. For example, learning
by doing effects for renewable energy and for resource extraction belong to this class of
ETC. The competitiveness of mitigation options has a strong impact on the strategy, but
their effect on mitigation costs is negligible.
ETC therefore proves to be an influential determinant of mitigation costs and strategies.
Costs may rise or fall due to ETC depending on whether “clean” progress (e.g. in re-
newable energy technology), or “dirty” progress (e.g. in resource extraction technology)
1The concepts of endogenous technological change (ETC) and induced technological change (ITC) are
closely related, in fact the two terms are often used synonymously. Here, I use ETC to emphasize the mod-
eling assumption of endogenous (versus exogenous) technological change, and ITC to stress technological
change being triggered by climate policy.
6.1 Induced technological change in integrated assessment modeling 149
Impact of ETC is . . .
macro-economic sectoral
Macro-economic ETC labor R & D
energy efficiency
Sectoral ETC resource extraction renewables
resource extraction
Table 6.1: The scope of ETC in MIND
prevails. The effect of ETC on the competitiveness of mitigation options influences their
contributions to overall mitigation.
Moreover, this reveals the importance of economy-wide effects of ETC beyond sector
boundaries, and stresses the importance of models that resolve important technological
options including their potential of ETC, and account for the economy-wide impact of
ETC.
The analysis in Chapter 2 highlighted the importance of including ETC in climate policy
models that explore mitigation costs and strategies. The numerical experiments relied on
parameter variation, therefore assessing impact of parameter uncertainty for one partic-
ular implementation of ETC—but the question how to incorporate ETC in models is far
from trivial. On the contrary, among models that include ETC there is a wide variety of
approaches taken to describe ETC. I now turn to Chapter 3, which explored the resulting
differences in the assessment of ITC.
6.1.2 Implications across models
The variety in ETC implementations and the corresponding variety in the findings about
the effects of ITC are addressed in the next set of question.
•How much do integrated assessment models differ in their analysis of ITC?
•What are the underlying reasons for the differences?
•What conclusions are robust across models despite the model uncertainty?
The above set of questions was addressed in a comparison exercise of ten state-of-the-
art models of energy, economy, and environment. At the heart of this comparison is the
definition of ceteris paribus scenarios that aim to isolate and expose the impact of ITC
in the various models: policy scenarios that use exogenous technological change (taken
from a separate business-as-usual scenario) are compared to scenarios that implement the
same policy target, but allow for additional technological change to be induced by the
policy. I refer to these scenarios as “without ITC” and “with ITC.”
At the most aggregate level, the impact of ITC becomes apparent in mitigation costs with
and without ITC. The analysis reveals that ITC has potential to reduce costs: compared
to the scenarios without ITC, mitigation costs are lower in scenarios with ITC, in many
models substantially. Average mitigation costs in the participating neoclassical models,
150 Chapter 6 Synthesis and Outlook
Figure 6.1: Loss of GDP in 2100 in percent. IMCP models use the Common POLES/IMAGE baseline,
which is based on SRES scenarios A1B and B2 and assumes “a strong dependence on fossil fuels” (van
Vuuren et al., 2003). GDP losses in 2100 for MIND and DEMETER are very close to zero. Source:
adapted from IPCC WG3 (2007, Figure 3.25)
excluding those models that explore an extreme scenario (such as energy conservation as
the sole abatement option), are below 1 percent discounted GDP (gross domestic product)
for the next century. This is low compared to numbers from other models. Figure 6.1
from IPCC WG3 (2007) shows that mitigation cost estimates from this comparison study
(denoted as “-IMCP”) are at the lower end of a set of state-of-the-art models.
However, the magnitude of the impact of ITC differs greatly, ranging from 90 percent
reduction of mitigation costs to models where introducing ITC has almost no effect. Nu-
merous reasons for this variety in model results were identified. Here, I summarize these
reasons in three categories: first, baseline effects, second, differences in mitigation strate-
gies, and third, modeling assumptions.
Baseline Effects
The baseline (or business-as-usual scenario) has a strong impact on mitigation costs be-
cause it determines the necessary emission reductions. Emission profiles that are consis-
tent with a certain GHG concentration stabilization target are very similar across models
because the uncertainty about climate models is relatively small. Contrary, predictions
of economic output and associated emissions for the next century vary strongly between
economic models. Although an effort was made to harmonize the business-as-usual sce-
narios of the different models, the remaining differences need to be taken into account
when interpreting the results.
Mitigation Strategies
The choice of a mitigation strategy is closely related to the corresponding mitigation costs:
when the mitigation strategy constitutes avoiding emissions by reducing the economic
output, costs will be high compared to strategies that rely on switching from fossil fuel
6.1 Induced technological change in integrated assessment modeling 151
combustion to carbon free energy sources. Here, mitigation strategies are explored on two
levels of aggregation. First, abatement is decomposed along the Kaya identity, i.e. into
reductions of economic output, energy intensity of output, and carbon intensity of energy
(Kaya, 1990). Second, abating emissions through change in the composition of energy
supply are considered, e.g. the usage of energy from renewable sources or utilization of
carbon capture and storage (CCS).
Overall emissions reductions are attributed to the factors of Kaya’s identity, carbon inten-
sity, energy intensity, and output2
emissions =emissions
energy ·energy
output ·output
using the refined Laspeyres index method (Sun, 1998). Naturally, reducing output as a
means of mitigation is only a last resort as it translates directly into mitigation costs mea-
sured as loss of GDP. The analysis reveals that macro-economic models without explicit
representation of the energy sector tend to focus their abatement strategy on reductions of
energy intensity, whereas energy system models and models that feature an energy sector
achieve the majority of their abatement through decarbonization. Reducing carbon inten-
sity becomes particularly important for large reductions of emissions: while many models
raise the contribution of carbon intensity reductions for lower levels of GHG concentra-
tion stabilization, those models without decarbonization options in explicitly modeled
energy systems resort to reducing their GDP. The tendency that carbon intensity reduc-
tions become increasingly more important with the level of emission reductions is con-
firmed by their contribution over time: reduction of energy intensity dominates abatement
in early periods of simulation time, but the contribution of carbon intensity reduction is
more important in later time periods.
The composition of the energy supply mirrors these trends. Models that focused their
abatement strategy on reducing energy intensity and GDP are those that lack options to
decarbonize the energy system, or that simply did not resolve the energy sector explicitly.
Conversely, large reductions of carbon intensity are implemented through large shares of
carbon free energy, e.g. energy from renewable sources, use of CCS, and nuclear power.
This implies that mitigation costs as well as strategies ought to be sensitive towards the
assumptions about the availability of carbon free energy, e.g. from backstop technologies
or end-of-the-pipe technologies—a hypothesis that is confirmed by the analysis in Chap-
ter 2 of this thesis. Among the carbon free energy sources, CCS plays a special role. With
respect to the utilization of this option over time, those models that account for rising
fossil fuel prices due to resource scarcity show a peak and decline of CCS. This suggests
that the competitive advantage of CCS is lost when resource scarcity raises the price of
fossil fuels and at the same time alternative carbon free energy sources such as renewable
energy become cheaper due to learning effects. Thus, CCS would only be a temporary
abatement option.
Modeling Assumptions
Three key modeling assumptions were identified that explain some of the major differ-
ences in model results: first, whether a model describes a first-best or a second-best world.
2As population dynamics are exogenous to all participating models, I omit it as a factor in the identity.
152 Chapter 6 Synthesis and Outlook
Second, the choice of the model type, e.g. energy system model, computational general
equilibrium model (CGE), or optimal growth model. Third, assumptions about foresight
of economic agents in their investment decisions.
First-best models abstract from market distortions. Therefore, comparing first-best sce-
narios with and without climate policy exposes the opportunity costs of mitigation. In
contrast, in a second-best model a (climate) policy may remove imperfections that distort
markets in business-as-usual in addition to implementing climate protection. Second-best
assumptions explain the outliers in the comparison that show very low costs or even neg-
ative mitigation costs.
The model type often implies a choice of equilibrium concept. The participating energy
system models are implemented as partial equilibrium models. They show particularly
low costs, which can be due to neglected general equilibrium effects. The participating
CGE model is solved recursively dynamic which reduces its investment flexibility com-
pared to models with perfect foresight. This helps to explain the extraordinary high costs
in this model. Finally, models in the optimal growth framework demonstrate the effects
of full sectoral and temporal flexibility.
The long-term investment behavior of economic agents in the different models is driven
by their ability of foresight. Under perfect foresight, the necessary investments may be
undertaken early, thus reducing mitigation costs. This is another reason for the low costs
reported in optimal growth models and energy system models.
6.1.3 Discussion
The previous sections stressed the importance of technological change in the assessment
of mitigation costs and strategies. Endogenous technological change and the implementa-
tion of technological detail in the energy sector were found to make potentially important
contributions to mitigating climate change at low costs. This is a plausible result. When
introducing ETC and a variety of low carbon technologies increase the flexibility within
a model to decarbonize the economy, the impact on costs ought to be favorable. How-
ever, this raises the question, whether the flexibility of the real world energy system and
economy, or rather their inertia and inflexibilities, are sufficiently captured.
Indeed, several models used in this thesis rest on assumptions that potentially overestimate
the world economy’s flexibility to be decarbonized. These assumptions include modeling
the world as one aggregate rather than distinguishing world regions, and focusing on the
combustion of fossil fuels in the energy sector as the main driver of GHG concentrations.
The latter neglects that abatement of emissions from, for example, transport or consump-
tion behavior of households may be more difficult to achieve. Indeed, one modeling team
cited the explicit transport sector in their model as an important factor contributing to
the pessimistic prediction of mitigation costs (IMACLIM-R). Modeling the world econ-
omy without regional disaggregation, on the other hand, cannot resolve any inefficiencies
that arise from regional heterogeneity. Consequently, the model comparison study rec-
ommended extending integrated models towards “hybrid models” comprising detailed
energy system models in a macro-economic setting, and to explore the regional effects of
mitigation.
6.2 The Prospect of Issue Linking for Global Climate Policy 153
Since the publication of Edenhofer et al. (2006a) and Edenhofer et al. (2006b), model
development in integrated assessment of climate change has shifted towards such hybrid
models (see for example the Special Issue on Hybrid Modelling edited by Hourcade et al.,
2006), in particular in new model developments such as the WITCH and REMIND-R
models (Bosetti et al., 2006b; Leimbach et al., 2008). Resolving technological detail
allows these models to draw on technological and empirical data, giving these models a
solid, data based calibration, which before was only available to energy system models.
Moreover, both WITCH and REMIND-R feature multiple regions.
The assumption of perfect foresight has a large influence in bringing down costs. Of
course, this model assumption overstates the ability of individual economic actors. In
this sense, models that assume perfect foresight only conclude that mitigation costs are
potentially low. However, this result should also be a motivation for policy to implement
a stable long-term climate policy and investment environment, which in effect increases
the ability of economic actors to take longer time horizons into account. The design of
stable mitigation agreements is addressed in Chapters 4 and 5 of this thesis.
6.2 The Prospect of Issue Linking for Global Climate
Policy
Chapters 2 and 3 looked at climate policies implemented as global policy targets, in par-
ticular maximum carbon dioxide concentrations. They took for granted that policies need
to be agreed upon and implemented to achieve these targets by placing the necessary
price on carbon, e.g. by determining a distribution of emission allowances and setting up
a global carbon market. Establishing a carbon price this way requires global coordination
and cooperation but it is known from literature as well as political experience that negoti-
ating such an international environmental agreement is difficult. Chapters 4 and 5 looked
at the potential of issue linking to help to build such agreements.
6.2.1 Trade Sanctions
In the introductory chapter, the following questions were raised concerning international
cooperation:
•What is the prospect for international cooperation on climate change mitigation?
•How can it be increased by the design of international environmental agreement?
•What is the potential of trade sanctions to increase participation in international en-
vironmental agreements?
•What are the effects on environmental and global welfare of trade sanctions on the
one hand and increased cooperation on the other hand?
•How can competitive equilibria be computed in models with emission externality,
international trade, and tariffs?
154 Chapter 6 Synthesis and Outlook
To answer these questions, I developed an integrated assessment model of coalition for-
mation. The topic at hand, namely international cooperation to mitigate climate change,
requires a modeling framework apt to describe long-term economic development due to
the inertia of the climate system. To account for the stock pollutant character of climate
change, a dynamic model is needed. And any model discussing the prospect of cooper-
ation needs to describe many actors. Therefore, the model developed for this thesis is a
multi-actor optimal growth model. Including international trade and implementing trade
sanctions in a coalition model within this framework has to the best of my knowledge not
been done before.
The model is calibrated such that global totals of the model outputs like economic output,
savings rate, export ratio, emissions, are consistent with real world data. However, actors
within the model are assumed to be identical. This limits the applicability of the model to
real world negotiation, but facilitates studying the incentives for cooperation in isolation
from the distributional effects introduced via heterogeneity of actors.
I show in numerical experiments that introducing trade sanctions positively affects coop-
eration on international cooperation: When those actors that cooperate on climate pro-
tection additionally impose import tariffs on goods from non-members of the coalition,
more actors are inclined to participate in the joined international agreement of both, envi-
ronment and trade, than elsewise. Indeed, participation rises with the tariff rate, reaching
full cooperation at some point. How quickly participation rises and at which tariff rate
full cooperation is sustained depends on the ease with which taxed goods are substituted
with alternatives. Global welfare rises with participation despite the distortions caused by
trade restriction. These results proved robust against variation of key parameters. Tariffs
therefore seem to be a feasible means of increasing participation.
Moreover, trade sanctions turn out to be a credible incentive in the sense that imposing an
import tariff is beneficial to the coalition. This is true as long as the tariff rate is ‘small’
in relation to the substitutability of the taxed good: when goods are easily substituted the
losses of reduced trade will exceed the additional income in form of tariff revenues.
Standard approaches to find a competitive equilibrium for traded goods could not be ap-
plied due to the market distortions introduced via climate change on the one hand and
tariffs on the other hand. This was overcome by extending existing approaches.
6.2.2 Technology-oriented Agreements
The next set of questions explores potential interactions between endogenous technolog-
ical change and international cooperation. Specifically, it focuses on issue linking of
environmental agreements and technology oriented agreements.
•How does ETC help to promote international cooperation on emission abatement?
•What are the roles of different technology oriented agreements (TOA)?
•What is the role of cooperative research and development and technological spill-
overs?
•In which ways does the type of technology that spills over matter?
6.2 The Prospect of Issue Linking for Global Climate Policy 155
•What is the role of international technology standards?
I address these questions by applying the model of coalition formation developed for
this thesis. The model was extended by two concepts of technological cooperation: in-
ternational actors could chose to cooperate on R & D (augmenting either productivity or
mitigation) and by doing so benefit from increased technology spillovers, or actors could
jointly implement international technology standards.
It turns out that ETC provides a means to increase participation in an environmental co-
operation: issue linking of the environmental agreement to cooperative research, which
induces increased technology spillovers changes the incentive structure such that more
actors sign the linked agreement. The number of participants rises with the intensity of
spillovers up to and including full cooperation.
The type of technological knowledge that spills over (either related to productivity or
mitigation) makes a difference for the effectiveness of this type of issue linking: cooper-
ation on productivity R & D is unambiguously more effective in raising participation in
the agreement, global welfare, and environmental quality. The reason is that the bene-
fit of cooperation is only truly exclusive to the coalition in case of productivity. In case
of cooperation on mitigation some of the benefits of cooperation spill over indirectly to
non-members via reduced climate change damages. This impact of the source of spill-
overs helps to understand why previous investigations into this topic came up with mixed
results: in Botteon and Carraro (1998) where spillovers are related to productivity high
levels of cooperation are sustained through issue linking. Contrary, in Nagashima and
Dellink (2008) spillovers are related to mitigation technology and the impact on partici-
pation is only modest.
International technology standards are also shown to have a positive effect on coalition
formation. By assumption, standards are easier to agree on and implement, and “standards
agreements” are accomplished ad hoc. The existence of a separate standards agreement
alone has very little impact on environmental cooperation. It would, however, signifi-
cantly increase participation in a linked agreement on environmental and technological
cooperation.
This renders a stepwise approach of building coalitions possible: creating a global agree-
ment on technology standards would not (according to this model) solve the climate prob-
lem in the sense of inducing increased cooperation in an environmental agreement. It
does, however, prepare the ground for a linked agreement on environmental and techno-
logical cooperation.
6.2.3 Discussion
Chapters 4 and 5 assessed the impact of linking environmental cooperation to trade sanc-
tions or technology-oriented agreements on the incentives to participate in an interna-
tional environmental agreement using numeric model experiments. In principle, it would
be preferable to derive analytical solutions because of their greater generality. However,
showing the stability of coalitions in general is often not feasible, in particular when mod-
els include more complex interactions such as spillovers, trade, and tariffs. Relying on nu-
merical simulations enables me to explore these issues in the state-of-the-art framework of
156 Chapter 6 Synthesis and Outlook
optimal growth modeling. Additionally, I can avoid simplistic assumptions during model
development, for example, the stock pollutant characteristic of GHG concentrations can
adequately be accounted for by using a dynamic modeling setting. Still, it is important to
keep the limitations of numerical results in mind when drawing conclusions from model
results.
The modeling approach in this thesis takes middle ground between using stylized eco-
nomic actors and modeling real world actors by calibrating global totals of economic
output, GHG emissions, etc. to data. While using identical actors has advantages when
investigating the impact of issue linking in general, the analysis would benefit from cal-
ibrating all actors to real world regions to additionally estimate the magnitude of these
effects.
The model assumes national product differentiation in a single good world to describe
the driving forces of international trade. Consequently, when trade sanctions in form
of import tariffs are imposed by coalition members, all trade is affected. Realistically
though, a trade sanction under World Trade Organisation rules could only target carbon
intensive goods. So while the current description of trade suffices to access the scope
of trade sanctions in a very general setting, a real world analysis of the approximate
magnitudes of gains, losses, and impact on incentives of tariffs would require multiple
traded goods.
Furthermore, the exploration of technology oriented agreements includes a first numeri-
cal assessment of international technology standards. As discussed in the corresponding
chapter, an ad hoc agreement concerning the global implementation of these standards
is assumed because arguably, their incentive structure makes standards self-enforcing.
Naturally, the analysis would be more complete if the decision about an agreement on
standards was incorporated into the given model. The necessary model extension would
encompass introducing technological dynamics exhibiting the potential for increasing re-
turns to scale and lock-in effects for technologies.
6.3 Outlook and Further Research
The questions and concepts explored in this thesis can be extended in future research in
at least three ways: further model comparisons, advanced concepts of coalition stability,
and accounting for the large uncertainties, which are pervasive in integrated assessment
modeling of climate change.
Model comparisons are established as a tool to assess uncertainty in model structure. The
Stanford Energy Modeling Forum (EMF) has pioneered model comparisons for integrated
assessment models of climate change, and likewise, Chapter 3 of this thesis used a model
comparison to shed light on the impact of induced technological change. Section 3.6.6
suggests to move the comparison of mitigation strategies in IAM down to the level of tech-
nological options—this suggestion has been taken up in ADAM (2008, pp. 8–9). Game
theoretic models of climate change are not as numerous as their integrated assessment
counterparts. But new model developments from recent years, Finus et al. (2006), Bosetti
et al. (2006b) and the coalition model from this thesis (Lessmann et al., 2009), have led
to a critical mass in models and variability in their predictions (e.g. about the effects of
6.3 Outlook and Further Research 157
spillovers on participation in different models as discussed in Section 5.5), thus meriting
a comparison of these models and the driving forces of their results. Comparing the pre-
dicted stable coalition from these model would be a demanding task. Running comparable
scenarios in these different models is already a challenge for simple non-cooperative or
fully cooperative settings, and the game theoretic setting of stable coalitions adds another
layer of complexity: assumptions concerning the concept of stability, number of players,
choice of regions, intra-coalitional welfare transfers, etc. would require harmonization.
Nevertheless, valuable insights may be gained, e.g. concerning which assumptions are
decisive for the incentive to free-ride for different regions in different models.
The concept of coalitional stability used in this thesis, cartel stability, is frequently used in
applied analyses of international environmental agreements. It has great intuitive appeal
and is readily implemented, partially due to its myopic perspective, e.g. a player aban-
doning a coalition will not anticipate whether the remaining coalition perseveres, even
though this has strong implications for her payoff. In a setting where actors take their
economic decision with perfect foresight, this assumption on strategic foresight seems
rather limiting. More advanced coalition concepts are discussed in the game theoretic
literature, such as Faresightedly Stable Coalitions and the Coalition Proof Nash Equilib-
rium. To date, these concepts have rarely been applied in integrated assessment models.
Furthermore, in case of the cartel stability concept, international environmental agree-
ments are modeled as a one-shot, static game, even though the negotiations on a climate
agreement under the UNFCCC seeks to increase participation in future commitment peri-
ods. Dynamic games would capture this better than the static approach, indeed non-static
concepts have recently been employed: Weikard and Dellink (2008) allow to renegotiate
their agreement in several commitment periods, Rubio and Ulph (2007) analyse dynamic
membership. Still, modeling dynamic international environmental agreements is in its
infancy.
In economic models that span several centuries, uncertainties abound. This is especially
true for the estimates of climate change damages, which are an integral part of game the-
oretic models of climate change because the assumption on damage functions determine
the benefits of abatement. The analysis of coalition stability would therefore benefit from
additional research on damage functions.
Overall, the studies reported in this thesis suggest that there is indeed potential that ITC
may reduce the burden that mitigation requirements will put on the economy. And while
there is no final conclusion to the magnitude of the impact of ITC due to the model uncer-
tainty, which remains large, this thesis advanced the understanding of these uncertainties
and the underlying reasons for the variability in the results. To exploit a large potential
of ITC, a clear carbon price signal is required. This thesis suggests that linking the nego-
tiations on climate policy to trade sanctions or to research cooperations is a feasible way
to create incentives that make a cooperative global climate policy more likely. Again,
more research is needed to determine the magnitude of the potential of issue linking, but
its potential in general has been shown and different issue linking proposal have been
characterized with respect to their advantages and disadvantages.
158 Chapter 6 Synthesis and Outlook
Bibliography
ADAM, 2008. ADAM – Adaptation and Mitigation Strategies: supporting European cli-
mate policy. http://www.adamproject.eu/, ADAM Interim Results.
Aghion, P., Howitt, P., 1998. Endogenous Growth Theory. MIT Press.
Aldy, J. E., Orszag, P. R., Stiglitz, J. E., 2001. Climate change: An agenda for global col-
lective action, prepared for the conference on The Timing of Climate Change Policies,
Pew Center on Global Climate Change.
Azar, C., Schneider, S. H., 2002. Are the economic costs of stabilising the atmosphere
prohibitive? Ecological Economics 42, 73–80.
Bank for International Settlements, 2004. Annual report 2004/05. Tech. rep.,
http://www.bis.org/publ/arpdf/ar2005g2.pdf.
Barker, T., Haoran, P., Köhler, J., Warren, R., Winne, S., 2006. Decarbonising the global
economy with induced technological change: Scenarios to 2100 using E3MG. The
Energy Journal 27, 143–160.
Barrett, S., 1991a. Economic analysis of international environmental agreements: lessons
for a global warming treaty. In: OECD, Responding to Climate Change: Selected Eco-
nomic Issues. OECD, Paris, pp. 111–49.
Barrett, S., 1991b. Economic instruments for climate change policy. In: OECD, Respond-
ing to Climate Change: Selected Economic Issues. OECD, Paris, pp. 53–108.
Barrett, S., 1994. Self-enforcing international environmental agreements. Oxford Eco-
nomic Papers 46, 878–894.
Barrett, S., 1997. The strategy of trade sanctions in international environmental agree-
ments. Resource and Energy Economics 19, 345–361.
Barrett, S., 2003. Environment and Statecraft: The Strategy of Environmental Treaty-
Making. Oxford University Press.
Barrett, S., Stavins, R., 2003. Increasing participation and compliance in international
climate change agreements. International Environmental Agreements: Politics, Law
and Economics 3, 349–376.
Bauer, N., 2005. Carbon capturing and sequestration. an option to buy time? Ph.D. thesis,
University of Potsdam, Potsdam University, faculy of Economics and Social Sciences.
159
160 Bibliography
Bauer, N., Edenhofer, O., Held, H., Kriegler, E., 2005. Uncertainty of the role of carbon
capture and sequestration within climate change mitigation strategies. In: Rubin, E. S.,
Keith, D. W., Gilboy, C. F. (Eds.), Proceedings of the 7 International Conference on
Greenhouse Gas Control Technologies. Volume 1: Peer-Reviewed Papers AND Plenary
Presentations, IEA Greenhouse Gas Programme. Cheltenham, UK.
Benedick, R. E., 2001. Striking a new deal on climate change. Issues in Science and
Technology, 71–76.
Bernstein, P., Montgomery, W., Rutherford, T., 1999. Global impacts of the kyoto agree-
ment: Results from the MS-MRT model. Resource and Energy Economics 21 (3), 375–
413.
Biermann, F., Brohm, R., 2005. Implementing the kyoto protocol without the USA: the
strategic role of energy tax adjustments at the border. Climate Policy 4 (3), 289–302.
Bosetti, V., Carraro, C., Galeotti, M., 2004. The Dynamics of Carbon and Energy Intensity
in a Model of Endogenous Technical Change, fondazione Eni Enrico Mattei.
Bosetti, V., Carraro, C., Galeotti, M., 2006a. The dynamics of carbon and energy inten-
sity in a model of endogenous technical change. The Energy Journal Special Issue,
Endogenous Technological Change and the Economics of Atmosperic Stabilization.
Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., Tavoni, M., 2006b. WITCH: A World
Induced Technical Change Hybrid model. The Energy Journal Special Issue Hybrid
Modelling of Energy Environment Policies: Reconciling Bottom-up and Top-down,
13–38.
Botteon, M., Carraro, C., 1998. Game Theory and the Global Environmental. Edward El-
gar Publishing, Cheltenham, UK, Ch. Strategies for environmental negotiations: issue
linkage with heterogeneous countries, pp. 180–200.
Brooke, A., Kendrick, D., Meeraus, A., Rosenthal, R., 1988. GAMS: A User’s Guide.
The Scientific Press.
Buchner, B., Carraro, C., 2006. US, China and the economics of climate negotiations.
International Environmental Agreements 6, 63–89.
Buchner, B., Carraro, C., Cersosimo, I., Marchiori, C., 2005. The Coupling of Climate
and Economic Dynamics. Vol. 22 of Advances In Global Change Research. Springer
Netherlands, Ch. Back to Kyoto? US Participation and the Linkage between R&D and
Climate Cooperation, pp. 173–204.
Buonanno, P., Carraro, C., Galeotti, M., 2003. Endogenous induced technical change and
the costs of kyoto. Resource and Energy Economics 25, 11–34.
Carraro, C., 1999. International Environmental Agreements on Clime Change. Kluwer
Academic Publishers, Ch. The structure of international environmental agreements, pp.
9–25.
Carraro, C., Buchner, B., Cersosimo, I., Marchiori, C., 2002. Back to Kyoto? US partici-
pation and the linkage between R&D and climate cooperation.
Bibliography 161
Carraro, C., Eyckmans, J., Finus, M., 2006. Optimal transfers and participation decisions
in international environmental agreements. Review of International Organizations 1 (4),
379–96.
Carraro, C., Siniscalco, D., 1992. The international dimension of environmental-policy.
European Economic Review 36 (2-3), 379–387.
Carraro, C., Siniscalco, D., 1993. Strategies for the international protection of the envi-
ronment. Journal Of Public Economics 52 (3), 309–328.
Carraro, C., Siniscalco, D., 1995. The Economics of Sustainable Development. Cam-
bridge University Press, Ch. Policy Coordination for Sustainability: Commitments,
Transfers, and Linked Negotiations.
Carraro, C., Siniscalco, D., 1997. International Environmental Agreements – Strategic
Policy Issues. Edward Elgar Publishing, Ch. R&D Cooperation and the Stability of
International Environmental Agreements, pp. 71–95.
Chakravorty, U., Roumasset, J., Tse, K., 1997. Endogenous substitution among energy
sources and global warming. Journal of Political Economy 105, 1201–1234.
Clarke, L. E., Weyant, J., 2002. Technological Change and the Environment. Resources
for the Future Press, Washington D.C., Ch. Modeling Induced Technical Change: An
Overview, pp. 320–363.
Coase, R. H., 1960. The problem of social cost. Journal of Law and Economics 3, 1–44.
Cornes, R., Sandler, T., 1996. The Theory of Externalities, Public Goods, and Club
Goods. Cambridge University Press.
Crassous, R., Hourcade, J.-C., Sassi, O., 2006. Endogenous structural change and climate
targets. The Energy Journal Special Issue, Endogenous Technological Change AND
the Economics of Atmosperic Stabilization.
D’Aspremont, C., Gabszewicz, J. J., 1986. New Developments in the analysis of Market
Structures. Macmillan, New York, Ch. On the stability of collusion, pp. 243–64.
David, P., Greenstein, S., 1990. The economics of compatibility standards: An introduc-
tion to recent research. Economics of Innovation and New Technology 1 (1), 3–41.
de Coninck, H., Fischer, C., Newell, R. G., Ueno, T., 2007. International technology-
oriented agreements to address climate change. Energy Policy, 335–356.
Drud, A., 1994. CONOPT–A Large-Scale GRG Code. INFORMS Journal on Computing
6 (2), 207.
Edenhofer, O., Bauer, N., Kriegler, E., 2005a. The impact of technological change on
climate protection and welfare: Insights from the model mind. Ecological Economics
54, 277–292.
162 Bibliography
Edenhofer, O., Held, H., Bauer, N., 2005b. A regulatory framework for carbon capturing
and sequestration with the post-kyoto process. In: Rubin, E. S., Keith, D. W., Gilboy,
C. F. (Eds.), Proceedings of the 7th International Conference on Greenhouse Gas Con-
trol Technologies. Volume 1: Peer-Reviewed Papers AND Plenary Presenatations, IEA
Greenhouse Gas Programme. Cheltenham, UK.
Edenhofer, O., Lessmann, K., Bauer, N., 2006a. Mitigation strategies and costs of climate
protection: The effects of ETC in the hybrid model MIND. Energy Journal Special
Issue Endogenous Technological Change and the Economics of Atmospheric Stabili-
sation, 89–104.
Edenhofer, O., Lessmann, K., Kemfert, C., Grubb, M., Koehler, J., 2006b. Induced tech-
nological change: Exploring its implications for the economics of atmospheric stabi-
lization – synthesis report from the Innovation Modeling Comparison Project. Energy
Journal Special Issue Endogenous Technological Change and the Economics of Atmo-
spheric Stabilisation, 57–107.
Edmonds, J., Roop, J. M., Scott, M. J., 2000. Technology and the economics of climate
change policy, pew Center on Global Climate Change.
Edmonds, J., Wise, M., 1999. International Environmental Agreements on Climate
Change. Kluwer Academic Publishers, Ch. Exploring a technology strategy for sta-
bilizing atmospheric CO2, pp. 131–154.
EMF 19, 2004. EMF 19 Alternative technology strategies for climate change policy.
Vol. 26. Special Issue of Energy Economics, pp. 501–755.
EU Council, March 2007. Presidency conclusions. Tech. rep., Brussels European Council.
Eyckmans, J., Tulkens, H., 2003. Simulating coalitionally stable burden sharing agree-
ments for the climate change problem. Resource and Energy Economics 25, 299–327.
Farrell, J., Saloner, G., 1987. Product Standardization and Competetive Strategy. North-
Holland, New York, Ch. Competition, Compatibility, and Standards: the Economics of
Horses, Penguins, and Lemmings, pp. 1–22.
Feenstra, R., Markusen, J., Rose, A., 2001. Using the gravity equation to differentiate
among alternative theories of trade. Canadian Journal of Economics 34 (2), 430–447.
Finus, M., 2008. Game theoretic research on the design of international environmental
agreements: Insights, critical remarks, and future challenges. International Review of
Environmental and Resource Economics 2, 29–67.
Finus, M., Rundshagen, B., 2000. Strategic links between environmental and trade poli-
cies if plant location is endogenous, working Paper in Economics No. 283, University
of Hagen.
Finus, M., van Ierland, E., Dellink, R., 2006. Stability of climate coalitions in a cartel
formation game. Economics of Governance 7, 271–91.
Flechsig, M., Böhm, U., Nocke, T., Rachimow, C., 2008. The multi-run simulation envi-
ronment SimEnv. Http://www.pik-potsdam.de/software/simenv/.
Bibliography 163
Gerlagh, R., 2006. ITC in a Global Growth-Climate Model with CCS. The Energy Journal
Special Issue, Endogenous Technological Change and the Economics of Atmosperic
Stabilization.
Goulder, L. H., Mathai, K., 2002. Optimal CO2 Abatement in the Presence of Induced
Technological Change. Resources for the Future, Washington, USA.
Griliches, Z., 1992. The search for R&D spillovers. Scandinavian Journal of Economics
94 (0), 29–47.
Grubb, M., Koehler, J., Anderson, D., 2002. Induced Technical Change in Energy and En-
vironmental Modeling: Analytic Approaches and Policy Implications. Annual Reviews
in Energy and the Environment 27 (1), 271–308.
Hagedoorn, J., Link, A. N., Vonortas, N. S., 2000. Research partnerships. Research Policy
29 (4-5), 567–586.
Hare, B., Meinshausen, M., 2004. How much warming are we committed to and how
much can be avoided, PIK Report No. 93.
Hedenuns, F., Azar, C., Lindgren, K., 2006. Induced technological change in a limited
foresight optimization model. The Energy Journal Special Issue, Endogenous Techno-
logical Change and the Economics of Atmosperic Stabilization.
Hepburn, C., 2006. Regulating by prices, quantities or both: an update and an overview.
Oxford Review of Economic Policy 22 (2), 226–247.
HM Treasury, 2003. Appraisal and evaluation in central government (green book), web-
site: greenbook.treasury.gov.uk.
Hoos, G., R. Voss, K. Hasselmann, E. M.-R., Joos, F., 2001. A nonlinear impulse response
model of the coupled carbon cycle climate system NICCS. Climate Dynamics 18, 189–
202.
Houba, H., Bolt, W., 2002. Credible Threats in Negotiations – A Game-theoretic Ap-
proach. Kluwer Academic Publishers, Boston, Dordrecht, London.
Hourcade, J.-C., Jaccard, M., Bataille, C., Ghersi, F., 2006. Hybrid modeling: New an-
swers to old challenges – introduction to the special issue of the energy journal. Energy
Journal Special Issue Hybrid Modeling, 1–12.
Hourcade, J.-C., Robinson, J., 1996. Mitigating factors – assessing the costs of reducing
ghg emissions. Energy Policy 24, 863–873.
IPCC, 1996. Climate Change 1995: Economic and Social Dimensions of Climate Change.
Cambridge University Press.
IPCC, 2000. Special Report on Emissions Scenarios: a special report of Working Group
III of the Intergovernmental Panel on Climate Change. Cambridge University Press.
IPCC, 2001. Climate Change 2001: Mitigation. Cambridge University Press.
164 Bibliography
IPCC, 2007. Climate change 2007: Synthesis report. Tech. rep., Intergovernmental Panel
on Climate Change.
IPCC WG1, 2001. Climate Change 2001: The Scientific Basis. Cambridge University
Press.
IPCC WG1, 2007. Climate Change 2007: The Physical Science Basis: Contribution of
Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on
Climate Change. Cambridge University Press.
IPCC WG2, 2007. Climate change 2007: Impacts, Adaptation and Vulnerability: Contri-
bution of Working Group III to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press.
IPCC WG3, 2001. Climate Change 2001: Mitigation. Cambridge University Press.
IPCC WG3, 2007. Climate Change 2007 Mitigation: Contribution of Working Group III
to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change.
Cambridge Univ. Press.
Ismer, R., Neuhoff, K., 2007. Border tax adjustment: a feasible way to support stringent
emission trading. European Journal of Law and Economics 24 (2), 137–164.
Jones, C. I., Williams, J. C., 1998. Measuring the social return to R&D. Quarterly Journal
of Economics 113 (4), 1119–1135.
Katsoulacos, Y., 1997. International Environmental Agreements – Strategic Policy Issues.
Edward Elgar Publishing, Ch. R&D spillovers, cooperation, subsidies and international
agreements, pp. 97–109.
Kaya, Y., 1990. Impact of carbon dioxide emissions control on GNP growth: Interpre-
tation of proposed scenarios., paper presented to the IPCC Energy and Industry Sub-
group, Response Strategies Working Group, Paris, France.
Kehoe, T., Levine, D., Romer, P., 1992. On characterizing equilibria of economies with
externalities and taxes as solutions to optimization problems. Economic Theory 2 (1),
43–68.
Kemfert, C., 2002. An Integrated Assessment Model of Economy-Energy-Climate-The
Model Wiagem. Integrated Assessment 3 (4), 281–298.
Kemfert, C., 2004. Climate coalitions and international trade: assessment of cooperation
incentives by issue linkage. Energy Policy 32, 455–65.
Köhler, J., Grubb, M., Popp, D., Nakicenovic, N., 2006. Modelling endogenous technical
change in climate-economy models. The Energy Journal Special Issue, Endogenous
Technological Change and the Economics of Atmosperic Stabilization.
Kypreos, S., Bahn, O., 2003. A MERGE model with endogenous technological progress.
Environmental Modeling and Assessment 8 (3), 249–259.
Bibliography 165
Kypreos, S., Barreto, L., 2000. A simple global electricity MARKAL model with en-
dogenous learning, paul Scherrer Institute, General Energy Research Department-ENE,
Energy Modelling Group.
Lamport, L., 1994. LaTeX: A Document Preparation System User’s Guide and Reference
Manual. Reading, Mass.
Leimbach, M., Bauer, N., Baumstark, L., Edenhofer, O., 2008. Mitigation costs in a glob-
alized world: climate policy analysis with REMIND-R, working Paper available at
www.pik-potsdam.de.
Leimbach, M., Edenhofer, O., 2007. Technological spillovers within multi-region models:
Intertemporal optimization beyond the Negishi approach. Economic Modelling 24 (2),
272–294.
Lenton, T., Held, H., Kriegler, E., Hall, J., Lucht, W., Rahmstorf, S., Schellnhuber, H.,
2008. Tipping elements in the earth’s climate system. Proceedings of the National
Academy of Sciences 105 (6), 1786.
Lessmann, K., Edenhofer, O., submitted. Research cooperation and international stan-
dards in a model of coalition stability. Resource and Energy Economics.
Lessmann, K., Marschinski, R., Edenhofer, O., 2009. The effects of tariffs on coalition
formation in a dynamic global warming game. Economic Modeling, accepted.
Löschel, A., 2002. Technological change in economic models of environmental policy: A
survey. Ecological Economics, 105–126.
Manne, A., Mendelsohn, R., Richels, R., 1995. MERGE. a model for evaluating regional
and global effects of ghg reduction policies. Energy Policy 23, 17–34.
Masui, T., Hanaoka, T., Hikita, S., Kainuma, M., 2006. Assessment of CO2reductions and
economic impacts considering energy-saving investments. The Energy Journal Special
Issue, Endogenous Technological Change and the Economics of Atmosperic Stabiliza-
tion.
MATLAB, 1998. MATLAB – The Language of Technical Computing. The MathWorks.
Matthias, A., 2006. The pdfpages package. Tech. rep., CTAN,
CTAN://macros/latex/contrib/pdfpages.
McDougall, R., Elbehri, A., Truong, T., 1998. Global trade assistance and protection: The
GTAP 4 data base, center for Global Trade Analysis, Purdue University.
Morgan, M. G., Henrion, M., 1990. Uncertainty – A Guide Dealing with Uncertainty in
Quantitative Risk and Policy Analysis. Cambridge University Press, UK.
Nagashima, M., Dellink, R., 2008. Technology spillovers and stability of international
climate coalitions. International Environmental Agreements: Politics, Law and Eco-
nomics, DOI 10.1007/s10784-008-9079-1.
Negishi, T., 1960. Welfare economics and existence of an equilibrium for a competitive
economy. Metroeconomica 12 (2-3), 92–97.
166 Bibliography
Nordhaus, W. D., 1994. Managing the global commons. MIT Press, Cambridge, Mas-
sachusetts.
Nordhaus, W. D., 1998. Is the Kyoto Protocol a dead duck? Are there any live ducks
around? comparison of alternative global tradable emissions regimes, working Paper,
Department of Economics, Yale University.
Nordhaus, W. D., 2002. In Technological Change and the Environment. Resources for the
Future, Washington DC, USA, Ch. Modelling Induced Innovation in Climate-Change
Policy, pp. 182–209.
Nordhaus, W. D., Boyer, J., 2000. Warming the World. Models of Global Warming. MIT
Press, Cambridge, Massachusetts, USA.
Nordhaus, W. D., Yang, Z., 1996. A regional dynamic general-equilibrium model of al-
ternative climate-change strategies. The American Economic Review 86 (4), 741–765.
Parry, M., Palutikof, J., Hanson, C., Lowe, J., 2008. Squaring up to reality. Nature Reports
Climate Change, 68.
Perez, O., 2005. Multiple regimes, issue linkage, and international cooperation: Exploring
the role of the WTO. University of Pennsylvania Journal of International Economic
Law 26 (4), 735–778.
Petschel-Held, G., Schellnhuber, H.-J., Bruckner, T., Tóth, F. L., Hasselmann, K., 1999.
The tolerable windows approach: Theoretical and methodological foundations. Cli-
matic Change 41 (3), 303–331.
Philibert, C., 2003. Technology innovation, development and diffusion. OECD and IEA
Information Paper COM/ENV/EPOC/IEA/SLT(2003)4, OECD Environment Direc-
torate and International Energy Agency.
Philibert, C., 2004. Lessons from the Kyoto protocol: Implications for the future. Inter-
national Review for Environmental Statistics 5 (1), 311–322.
Pigou, A. C., 1946. The Economics of Welfare, 4th Edition. Macmillan.
Popp, D., 2004a. ENTICE-BR: The effects of backstop technology and r&d on climate
policy models, NBER Working Paper #10285.
Popp, D., 2004b. ENTICE: endogenous technological change in the DICE model of
global warming. Journal of Environmental Economics and Management 48 (1), 742–
768.
Popp, D., 2006. Comparison of climate policies in the ENTICE-BR model. The Energy
Journal Special Issue, Endogenous Technological Change AND the Economics of At-
mosperic Stabilization.
R Development Core Team, 2008. R: A Language and Environment for Statistical Com-
puting. R Foundation for Statistical Computing, Vienna, Austria, ISBN 3-900051-07-0.
URL http://www.R-project.org
Bibliography 167
Rahmstorf, S., 2008. Global Warming: Looking Beyond Kyoto. Brookings Institution
Press, Ch. Anthropogenic Climate Change: Revisiting the Facts, pp. 34–53.
Ramsey, F., 1928. A mathematical theory of saving. Economic Journal 38, 543–559.
Rao, S., Keppo, I., Riahi, K., 2006. Importance of technological change and spillovers
in effective long-term climate policy. The Energy Journal Special Issue, Endogenous
Technological Change AND the Economics of Atmosperic Stabilization.
Repetto, R., Austin, D., 1997. The costs of climate protection: A guide for the perplexed,
world Resources Institute.
Requate, T., 2005. Dynamic incentives by environmental policy instruments – a survey.
Ecological Economics 54 (2-3), 175–195.
Rogner, H.-H., 1997. An assessment of world hydrocarbon resources. Annual Review of
Energy and Environment 22, 217–262.
Rubio, S., Ulph, A., 2007. An infinite-horizon model of dynamic membership of interna-
tional environmental agreements. Journal of Environmental Economics and Manage-
ment 54 (3), 296–310.
Sano, F., Akimoto, K., Homma, T., , Tomoda, T., 2006. Analysis of technological porto-
folios for CO2stabilizations and effects of technological changes. The Energy Journal
Special Issue, Endogenous Technological Change AND the Economics of Atmosperic
Stabilization.
Schelling, T., 2002. What Makes Greenhouse Sense? Foreign Affairs 81 (3), 2–9.
Schubert, R., Schellnhuber, H. J., Buchmann, N., Epiney, A., Grießhammer, R., Kulessa,
M., Messner, D., Rahmstorf, S., Schmidt, J., 2007. Climate Change as a Security Risk.
WBGU Flagship Report. Earthscan, London.
Sijm, J. P. M., 2004. Induced technological change and spillovers in climate policy mod-
eling. Tech. rep., Energy Research Centre of the Netherlands, ECN Report ECN-C--
04-073.
Stern, N., 2007. The Economics of Climate Change: The Stern Review. Cambridge Uni-
versity Press.
Stiglitz, J. E., 2006. Making Globalization Work. Norton & Company.
Sun, J. W., 1998. Changes in energy consumption and energy intensity: a complete de-
composition model. Energy Economics 20 (1), 85–100.
Sun, J. W., Ang, B. W., 2000. Some properties of an exact energy decomposition model.
Energy 25, 1177–1188.
Turnovsky, S., 1997. International Macroeconomic Dynamics. MIT Press.
Turnovsky, S. J., 2000. Methods of Macroeconomic Dynamics. MIT Press.
168 Bibliography
van der Sluijs, J. P., 2002. Encyclopedia of Global Environmental Change. John Wiley &
Sons, Ltd, Chichester, Ch. Definition of Integrated Assessment, pp. 249–250.
van Vuuren, D. P., den Elzen, M. G. J., Berk, M. M., Lucas, P., Eickhout, B., H. Eerens,
R. O., 2003. Regional costs and benefits of alternative post-kyoto climate regimes –
comparison of variants of the multi-stage and per capita convergence regimes, RIVM
Report 728001025/2003.
Vollebergh, H. R. J., Kemfert, C., 2005. The role of technological change for sustainable
development. Ecological Economics 54, 133–147.
Wagner, U. J., 2001. The design of stable international environmental agreements: Eco-
nomic theory and political economy. Journal of Economic Surveys 15 (3), 377–411.
Weikard, H.-P., Dellink, R., 2008. Sticks and carrots for the design of international climate
agreements with renegotiations, presented at the Coalition Theory Network workshop.
Weyant, J. P., 2000. An introduction to the economics of climate change policy, Pew
Center on Global Climate Change.
Weyant, J. P., 2004. Introduction and overview. Energy Economics 26, 501–515.
Weyant, J. P., Olavson, T., 1999. Issues in modelling induced technological change in
energy, environmental, and climate policy. Environmental Modelling and Assessment
4, 67–85.
WTO, 2007. International trade statistics. Tech. rep., WTO, Geneva, Switzerland.
Statement of Contribution
The four core chapters of this thesis (Chapters 2 to 5) are the result of collaborations in this
PhD project between the author of this thesis and his advisor, Prof. Dr. Ottmar Edenhofer,
sometimes involving additional colleagues as indicated. The author of this thesis has
made extensive contributions to the contents of all four papers, from conceptual design
and technical development to writing. This section details the contribution of the author
to the four papers and acknowledges major contributions of others.
Chapter 2 This chapter uses the MIND model developed by Edenhofer, Bauer, and
Kriegler (2005a). The author’s contribution to this chapter is the conceptual design of the
research question and the numerical experiments, their implementation and execution,
and the subsequent analysis and processing of model results and their visualization in the
graphs of the chapter. Discussions and conclusions from these results were written in
close cooperation with O. Edenhofer.
Chapter 3 In preparation of this chapter, the author made decisive contributions to the
conception and design of the Innovation Modeling Comparison Project, in particular the
definition of the baseline and policy scenario. The author coordinated the quantitative
comparison of model results by preparing a questionnaire on the details of the model as
well as a reporting scheme for model output, and was in charge of correspondence with
the modeling teams, particularly for collecting and processing questionnaires and model
output. Consequently, all data analysis and visualization in graphs are due to the author.
The text of the chapter was written in close cooperation by the author and O. Edenhofer,
with revisions by all coauthors of the article.
Chapter 4 The conception of the dynamic model of coalition stability and its applica-
tion within the theme of issue linking was jointly developed by O. Edenhofer and the au-
thor; R. Marschinski contributed to the conception of how to include trade in the model.
The implementation of the model and development and implementation of the solution
algorithm in GAMS are due to the author. Likewise, all calculations, analysis of model
output and graphs for the article were done by the author. Interpretation, discussion,
conclusions and the writing of the article text were done in close collaboration with R.
Marschinski.
169
170 Statement of Contribution
Chapter 5 The research question for this chapter was developed in cooperation with O.
Edenhofer. Model development, numerical experiments, their interpretation and discus-
sion, and the manuscript are due to the author.
Acknowledgements
I would like to use this opportunity to express my gratitude towards colleagues, friends,
and family, whose support throughout these years enabled me to research and write this
thesis.
First and most importantly, I want to thank Ottmar Edenhofer, who supervised my dis-
sertation. Ottmar’s support of my work extends from providing funding and suggesting
research ideas to jointly working on these ideas and related economic concepts. Thank
you for this close cooperation, for granting me the liberty to pursue my own ideas, and
for your valuable advise in professional and personal affairs.
I am grateful to Hermann Held for productive cooperation, and for initially inviting me to
PIK to meet the (former) SMART working group.
I would like to thank all my colleagues in PIK’s Research Domain “Sustainable Solu-
tions” for making this research group a work environment to look forward to even on the
frustrating days of research. Special thanks to the colleagues whom I had the pleasure of
sharing an office with, Elmar, Bob, Alexander, and Matthias.
I am indebted to Robert whose advise helped me to stay in touch with reality, for inspiring
discussions, ideas for my research, and comments on my writings.
Christian frequently proofread my manuscripts and thesis chapters. Thanks for that, and
for your encouraging and positive feedback.
I wish to thank Anne for valuable and critical comments on chapters of this thesis, and
for initiating our reading group for economic theory.
Thanks to Nico, who patiently answered my questions about economics, GAMS, and
SimEnv in the early days of my doctoral project, which were the final days of his.
Marian also helped me out when economics puzzled me, especially when I was stuck with
problems of numerical optimization, or when I was in doubt about the correct proceedings
in the scientific community. Thanks for your help!
I would like to thank Elmar for his input during the innovation modeling comparison
project, in particular for his advise related to the climate science aspects of the project,
and for his detailed comments on our manuscripts for this project.
Lavinia, thanks for your help in set theory and getting those definitions of optimal trans-
fers sorted out.
I want to express my gratitude to Carlo Carraro, for detailed comments on my research at
various workshops, and for offering to review this thesis.
171
172 Acknowledgements
I thank my coauthors for the productive collaborations.
Special thanks are due to the administrators who run and maintain the computational
services at PIK. I benefited greatly from the high availability and reliability of the IT
infrastructure in general, and especially from the parallel computation cluster, without
which my research would have had to take a different course. I am thankful in particular
to Michael Flechsig for developing the SimEnv software and for prompt and in-depth help
when I failed to use it correctly, to Dietmar Giebitz, who among other things enabled me to
run time-consuming model experience comfortably from home, and to Roger Grzondziel
for supporting Linux desktops at PIK and for recovering my data from a crashed hard disc
drive.
I would like to thank Jutta for making this research domain run in all organizational mat-
ters, for organizing its social events, and for her supply of chocolate and cookies to PhD
students in need.
Horst and Erica, thank you for your last-minute comments on the summary of this thesis
and for polishing the language.
I wish to thank Theresa and Dominik for their moral support during all the years of my
thesis. Thanks for listening to me when I had reason to complain and celebrating with me
when I had good news.
I would like to express my deep thanks to my parents for supporting and understand-
ing me. The older I get, the more I realize how lucky I am that I was born their son.
Also, thanks to my sisters, Constanze and Nadine, for their interest in my work and their
supportive phone calls.
I thank Britta for her comments and thoughts on my research, and for frequently helping
me figure out why I was stuck in my thoughts or in coding. Often, our discussions helped
me see things differently. Myself aside, you probably carried the largest share of the
burden of this thesis; thank you for supporting me, and thanks for being there for me.
Finally, I would like to thank the executive committee of the International Max Planck
Research School on Earth System Modelling (IMPRS-ESM) for their patience with the
long time it took to write this thesis and for creating an environment for interdisciplinary
exchange among PhD students. Special thanks to Antje Weitz for her understanding and
support.
Financial support for the research presented in this thesis was provided by Volkswagen
Foundation and the European Commission under the Sixth Framework Programme, which
is gratefully acknowledged.
Tools and Resources
This dissertation relies heavily on numerical modeling. Naturally, a number of software
tools were used to create and run the models, and to process, analyze and visualize the
results. This section lists these tools.
Modeling All model experiments performed by the author made use of the General
Algebraic Modeling System (GAMS), version 22.7.2 (Brooke et al., 1988) and the
CONOPT3 solver, version 3.14S, for non-linear programs (Drud, 1994). The multi-run
environment SimEnv, versions 1.15–2.01, was used frequently (Flechsig et al., 2008).
Data Processing Model output was analyzed using The MathWorks’ MATLAB, ver-
sion 6.5 release 13 (MATLAB, 1998) and the NetCDF Toolbox for MATLAB by Charles
R. Denham, as well as the statistical software R, version 2.4.0 (R Development Core
Team, 2008).
Typesetting This document was prepared using L
A
T
EX2
ε(Lamport, 1994), particularly
the pdfpages package (Matthias, 2006), to include Chapters 2 to 5 in their given layouts.
173
174 Tools and Resources
