scieee Science in your language
[en] (orig)
Investment Strategies
for Climate Change Mitigation
vorgelegt von
Diplom-Mathematikerin
Lavinia Baumstark
von der Fakultät VI – Planen Bauen Umwelt
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Wirtschaftswissenschaften
Dr. rer. oec.
genehmigte Dissertation
Gutachter:
Prof. Dr. Ottmar Edenhofer
Prof. Dr. Thomas Bruckner
Promotionsausschuss:
Prof. Dr. Volkmar Hartje (Vorsitz)
Prof. Dr. Ottmar Edenhofer
Prof. Dr. Thomas Bruckner
Tag der wissenschaftlichen Aussprache: 17.06.2011
Berlin 2011
D83
Contents
Abstract 7
Zusammenfassung 9
1 Introduction 11
1.1 The Problem of Anthropogenic Climate Change 11
1.2 Dimensions of Mitigation 16
1.3 Integrated Assessment Models 19
1.4 Technological Change 23
1.5 The Aim of the Thesis 25
1.6 Thesis Structure 26
2 Capital trade and technological spillovers 33
2.1 Introduction 35
2.2 Embodied Technological Spillover 36
2.3 MIND-RS 37
2.3.1 Technical Description 37
2.3.2 Set up of Experiments 39
2.3.3 Empirical Foundation 39
2.4 Scenario Definition 40
2.4.1 Baseline Scenarios 40
2.4.2 Policy Scenarios 41
2.5 Policy Analysis 41
2.5.1 Common Results 41
2.5.2 Spillovers vs. Non-spillovers 42
2.5.3 First-mover Advantage 43
2.5.4 Fragmented Policy Regime 44
3
4 Contents
2.5.5 Sensitivity Analysis 44
2.6 Conclusions 44
2.7 Appendix 45
2.8 References 49
3 Mitigation costs in a globalized world 51
3.1 Introduction 53
3.2 Model Description REMIND-R 55
3.2.1 Macro-economy Module 55
3.2.2 Energy System Module 56
3.2.3 Trade Module 58
3.2.4 Climate Module 59
3.3 Reference Scenario 59
3.4 Model Analysis of Climate Policy Regimes 60
3.4.1 Description of the Policy Regimes 60
3.4.2 Technology Development and Mitigation Strategies 62
3.4.3 Trade 64
3.4.4 Emissions and Emissions Trading 66
3.4.5 Mitigation Costs 67
3.4.6 Climate Sensitivity 68
3.5 Conclusions 69
3.6 References 70
4 The Role of Renewables in the Low-Carbon Transformation 73
4.1 Introduction 76
4.2 The REMIND-R Model 79
4.3 Scenarios 80
4.4 Results 82
4.4.1 First-best Solutions BASELINE and POLx82
4.4.2 Delayed Climate Policy and Early RET Deployment 84
4.4.3 Early Renewable Deployment and Immediate Climate Policy 86
4.5 Discussion and Further Research 88
4.6 References 91
4.7 Figures 93
5 Dimensions of Technological Change 99
Contents 5
5.1 Introduction 102
5.2 Model and Scenario Set-up 104
5.2.1 REMIND-RS 104
5.2.2 Default Scenarios 106
5.3 Dimensions of Technological Change - Mitigation Costs and Strategies 107
5.3.1 Dynamics of Technological Change 108
5.3.2 Direction of Technological Change 109
5.4 Different Worlds - Mitigation Costs and Strategies 115
5.4.1 Elasticity of Substitution 115
5.4.2 Technology Options 119
5.5 Conclusions 122
5.6 References 124
6 Endogenous R&D Investments into Energy Efficiency 129
6.1 Introduction 132
6.2 Endogenous Technological Change 133
6.2.1 Endogenous Growth 133
6.2.2 Energy Efficiency Improvements 135
6.3 The Model REMIND-RS 136
6.3.1 Overall Structure 136
6.3.2 Calibration and Data 139
6.4 Reference Scenarios 142
6.5 Policy Analysis 147
6.5.1 Energy Sector Experiments 147
6.5.2 Technology Experiments 149
6.6 Sensitivity Analysis 151
6.7 Conclusions 153
6.8 References 154
7 Synthesis and Outlook 159
7.1 Investment Strategies and Technological Spillover in an Optimal World 159
7.2 Second Best Worlds and Investment Strategies 161
7.3 Discussion and Further Research 164
Statement of Contribution 173
6 Contents
Acknowledgements 175
Tools and Resources 177
Abstract
The substantial threat of anthropogenic climate change implies the reducing of green-
house gas emissions. This thesis deals with the costs and strategies of climate change
mitigation. In particular, investment strategies for climate change mitigation are investi-
gated. The thesis is separated into five parts each focusing on subquestions of the overall
research question. After an introduction into the problem of climate change and the im-
portant macro-economic mechanisms for mitigation, these subquestions are answered in
separate chapters. For the analysis Integrated Assessment models are used.
First, the impacts of technological spillovers under climate policies are analyzed by means
of a multi-regional model with technological change in form of interregional spillovers.
Model results indicate that the higher the ratio between the spillover intensities for energy
and labour efficiency, the lower are mitigation costs. As well, first-mover advantages and
commitment incentives for climate policy scenarios are investigated. A multi-regional
hybrid model with a more complex energy system is used for studying investments into
energy technologies in detail. In climate policy scenarios the entire energy consumption
is reduced, while renewable energy and CCS technologies are expanded immediately.
Different regions follow quite different mitigation strategies. While ambitious climate
targets can be reached with moderate global costs, the regional costs show a high variance.
In addition, Integrated Assessment models are used to investigate what happens if the
world will not agree on a climate friendly policy within the next years. The impacts of
early investments into renewable energy technologies in first-best and second-best worlds
are analyzed. Mitigation costs increase significantly, if the climate policy implementation
is delayed. In contrast, early deployment of renewable energy technologies reduces the
global costs. Within a five-region hybrid model the impacts of dynamics and direction
of technological change under climate change mitigation are studied. It turns out that
mitigation costs and strategies are quite sensitive to these variables. Further experiments
indicate that the impacts depend on the set of available technologies. For studying the
role of endogenous technological change for climate change mitigation, this model is
extended by a new formulation of efficiency improvements. It turns out that investments
into the efficiency of some energy sectors play a crucial role for low mitigation costs. In
climate policy scenarios, the increased mitigation costs of technological restrictions can
be overcome by R&D investments into energy efficiencies.
However, the results of this thesis demonstrate the important role of investment strategies
for climate change mitigation costs. The world gains from early investments into both a
broad portfolio of technologies and energy efficiencies. Thereby the immediate support
and high diversity of investments mainly provide low mitigation costs.
7
8 Abstract
Zusammenfassung
Der anthropogene Klimawandel verlangt die Reduktion von Treibhausgasen. Diese Ar-
beit beschäftigt sich mit den Kosten und Strategien zur Vermeidung des Klimawan-
dels. Dabei werden vor Allem Investitionsstrategien der Vermeidung untersucht. Die
Arbeit is unterteilt in fünf Teile, die jeweils Unterfragen der allgemeinen Forschungs-
frage untersuchen. Nach einer Einleitung in das Problem des Klimawandels und Makro-
Ökonomischen Mechanismen der Vermeidung werden diese Unterfragen in einzelnen
Kapiteln beantwortet. Die Analyse basiert auf Integrated Assessment Modellen.
Zuerst werden die Auswirkungen von technologischem Spillover in einem Mehrregionen-
modell mit technologischem Wandel in Form von interregionalem Spillover analysiert.
Modellergebnisse zeigen, daß je größer der Quotient zwischen Arbeits- und Energieef-
fizienz steigernden Spilloverintensität ist, desto geringer sind die Vermeidungskosten.
Außerdem werden die Vorteile von Vorreitern und Anreize für eine Klimapolitik un-
tersucht. Ein mehrregionales Hybridmodell mit einem detailierten Energiesystem wird
benutzt, um die Investitionen in Energietechnologieen im Detail zu analysieren. In
Klimapolitikszenarien wird der gesammte Energiekonsum verringert, während erneuer-
bare Energie und CCS Technologieen sofort ausgebaut werden. Verschiedene Regio-
nen verfolgen grundsätzlich untershciedliche Vermeidungsstrategieen. Während ambi-
tionierte Klimaschutzschranken zu moderaten globalen Kosten erreicht werden können,
variieren die regionalen Kosten deutlich.
Des Weiteren werden Integrated Assessment Modelle genutzt, um herauszufinden, was
es bedeutet, wenn die Welt sich in den nächsten Jahren nicht auf eine klimafreundliche
Politik einigen kann. Die Auswirkungen von frühzeitigen Investitionen in erneuerbare
Energieen in erstbesten und zweitbesten Welten wird analysiert. Die Vermeidungskosten
steigen signifikant, wenn die Implementierung von Klimapolitik verzögert wird. Hinge-
gen verringert ein frühzeitiger Einsatz von erneuerbaren Energieen die globalen Kosten.
In einem Hybridmodell mit fünf Regionen werden die Auswirkungen von Dynamik und
Richtung des technologischen Wandels unter Klimapolitik untersucht. Es zeigt sich, daß
die Vermeidungskosten und -strategieen sensitiv auf diese Variablen reagieren. Weit-
ere Experimente deuten an, daß die Auswirkungen vom Spektrum der zur Verfügung
stehenden Technologieen abhängt. Um die Rolle des endogenen technologischen Wan-
dels für die Vermeidung des Klimawandels zu studieren, wird dieses Modell um eine
neue Formulierung von Effizienssteigerungen erweitert. Es zeigt sich, daß Investitio-
nen in die Effizienz von einigen Energiesektoren eine entscheidende Rolle für niedrige
Vermeidungskosten spielen. In Klimapolitikszenarien können die durch technologis-
che Einschränkungen erhöhten Vermeidungskosten durch F&E Investitionen in die En-
ergieeffiezienz reduziert werden.
9
10 Zusammenfassung
Wie auch immer, zeigen die Ergebnisse dieser Arbeit die wichtige Rolle von Investition-
sstrategien für die Vermeidungskosten von Klimawandel. Die Welt profitiert von frühzeit-
igen Investitionen in eine große Bandbreite von Technologieen und in Energieeffizienz.
Dabei erbringen vor allem die unmittelbare Förderung und die hohe Diversität der Investi-
tionen niedrige Vermeidungskosten.
Chapter 1
Introduction
The fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change
(IPCC) demonstrated the substantial threat of climate change. Therefore, one of the main
challenges of the world is to organize climate change adaptation and mitigation for the
next century. However, this thesis focuses on mitigation strategies. In order to reduce
carbon emissions significantly, the whole macro-economy and energy production have to
be adapted. In particular, different investment decisions are affected. This thesis deals
with the question: What are investment strategies for climate change mitigation? In the
following the importance of a stringent climate policy is motivated and three dimensions
or climate change mitigation are explained. Within this thesis, Integrated Assessment
models are used to study this dynamics. Such models calculate optimal investment strate-
gies into energy technologies, Research and Development (R&D) and/or macro-economic
capital. In addition, technological change due to investment decisions often plays a
key role for the costs of climate change mitigation. In the following, different mech-
anisms of technological change are discussed. However, the investment decisions for
climate change mitigation might be restricted due to technological or political restrictions.
The first Chapter is structured as follows: The problem of climate change is presented
in section 1.1. In section 1.2. three dimensions of mitigation: reduced economic
growth, factor substitution and adaptation of investment strategies are discussed. The
methodology of this thesis is based on Integrated Assessment models. This model
family is explained in section 1.3. The feature of technological change is described
in part 1.4. Sections 1.5 and 1.6 present the aim of this thesis and show the thesis structure.
1.1 The Problem of Anthropogenic Climate Change
During the last decades the problem of anthropogenic climate change was identified by
academics and largely acknowledged in the academic as well as political spheres. Emis-
sions of different greenhouse gases, associated mainly with industrial production increase
the global mean temperature via the greenhouse effect, affecting the whole Earth system.
These changes imply risks for human lives as well as natural ecosystems. In the following,
the impacts of different emissions on the greenhouse effect and consequently the global
11
12 Chapter 1 Introduction
mean temperature are illustrated. Several identified reasons for concern demonstrate why
advising a 2◦C temperature target deems necessary to avoid dramatical climate change.
The greenhouse effect was first described by Tyndall and Arrhenius in the end of the
19. century (Solomon [47]). Increased greenhouse gas (GHG) emissions from industrial
production raise the natural greenhouse effect. Thereby the forcing on the climate
system and the global mean temperature are increased as will be explained below. In
the AR4 of the Working Group I of the IPCC highlights that 0.7 ◦C global warming
against preindustrial levels has already occured. In addition, 0.5◦C future temperature
increase are expected due to the interactions in the climate system. Many GHG emissions
responsible for the global warming of the last decades result from the industrialization
process of developed countries. This induced anthropogenic climate change is analyzed
in detail since the nineties. Researchers investigate the drivers of global warming and
what happens if this effect will continue during the next century. In a first step, the origin
of the contributing emissions will be clarified.
The atmospheric concentrations of gases responsible for global warming, e.g. Car-
bon dioxide (CO2), Methane (CH4), Nitrous Oxide (N2O), F-gases and Aerosols are
influenced by human activity. CO2emissions predominately stem from burning fossil
fuels. Also the increase of CH4emissions can partly be explained by this aspect of
industrialization. The second part of CH4emissions and the majority of N2Oemissions
are emitted by land use change. F-gases and Aerosols result from industrial processes,
the use of fossil fuel, the use of traditional biomass and land use change. CO2is one
of the most important greenhouse gases because of the long lifetime in the atmosphere
(around 70 years) and the huge amount emitted in the last decades (more than 20 GtCO2
in 2000). In addition, these emissions are expected to be high for the future due to the
expected continuation of burning fossil fuels. The question of what are the contribu-
tions of the different GHG emissions to the global warming effect and how could they
be compared can be examined by calculating the radiative and integrated radiative forcing.
Radiative forcing is a measurement for changes in the energy balance of the Earth-
atmosphere system when factors that affect the climate are changed (Solomon [47]).
These factors influence the balance between incoming solar radiation and outgoing in-
frared radiation within the Earth’s atmosphere. The global radiative balance determines
the Earth’s surface temperature. A deviation from the normal level of radiative balance
is called radiative forcing. Higher GHG concentrations increase the atmospheric absorp-
tion of outgoing radiation. Increases in Aerosols reflect and absorb incoming solar ra-
diation and change cloud radiative properties. The radiative forcing of short-lived gases
and Aerosol depend significantly on both when and where they are emitted. However,
a positive or negative radiative forcing increases or decreases the global average surface
temperature.
Radiative forcing can be calculated for each year, integrating it over time results in the
integrated radiative forcing. To study the impact of a one-year “pulse” of global emis-
sions for future time horizons, the integrated radiative forcing is used. Jacobson [25], for
example used this approach for studying fossil fuel organic and black carbon Aerosols
compared to CO2. Figure 1.1 illustrates the integrated radiative forcing of different cur-
rent emissions in the year 2000. The forcing is calculated for a 20- and 100-years time
1.1 The Problem of Anthropogenic Climate Change 13
Figure 1.1: Integrated radiative forcing: Future climate impact of the current emissions (year
2000), 20- and 100-year time horizon. Shown are the following gases: carbon dioxide (CO2), ni-
trous oxide (N2O), methane (CH4), Chlorofluorocarbons(CfCs), sulphur hexafluoride(SF6), perfluo-
rocarbons (PFCs), hydrofluorocarbons (HFCs), carbon monoxide(CO), non-methane volatile organic
compound(NMVOC), nitrogen oxides(NOx), Organic carbon, Nitrate, sulphur dioxideSO2, Black car-
bon and Cloud albedo. Source: Forster [14].
14 Chapter 1 Introduction
Figure 1.2: Risks from climate change by reason for concern, appraised by the IPCC Third As-
sessment report compared with recently updated data. Climate change impacts are depicted against
increases in global mean temperature(◦C) after 1980. Source: Smith et al. [43].
horizon. This indicates the future impact of these emissions on climate change. The cool-
ing effect of Aerosols is nearly the same for both a 20-years and 100-years time horizon.
In a 100-year time step, especially the CO2emissions show a high integrated radiative
forcing representing the importance of this gas for the global warming effect. Moreover,
it indicates that for an adequate analysis of climate change long-term studies are neces-
sary.
For a more systematic comparison of the impacts of future emissions different metrics are
used. For example, Global Warming Potentials (GWP) are a common metric of different
short-lived and long-lived GHGs. They compare the integrated radiative forcing over a
specific time period, e.g. longer than 100 years. Thereby the potential climate change
impacts of different GHG emissions can be compared.
As discussed before, an increased concentration of some emissions, i.e. Aerosols reduce
the radiative forcing. However, in the sum across all GHGs the greenhouse effect
is intensified by increased emissions, especially in the long-run. Consequently the
global mean temperature is increased. This global warming effect results in significant
impacts for the Earth systems. These impacts, e.g. an expected increase in sea levels,
are analyzed to identify critical temperature levels. The area of Dangerous Anthro-
pogenic Interference (DAI) is defined. In the following, an illustration of the five reasons
for concern demonstrates why the politically discussed 2◦C temperature goal is necessary.
The Third Assessment Report of the IPCC identified five reasons for concern (McCarthy
[32]). Each of these categories includes different impacts linked to an increased global
mean temperature. The resulting so called burning embers diagram is shown in figure
1.2 (left). Each bar represents one reason for concern and the risk of it to become severe
associated with the respective increase in global mean temperature. The white regions
indicate neutral or low risks associated with an increase of the global mean temperature
1.1 The Problem of Anthropogenic Climate Change 15
above 1990s levels. Negative impacts or more significant risks are shown in yellow and
the red regions stand for substantial negative impacts or risks that are more widespread.
The first bar represents the ’Risks to Unique and Threatened Systems’. This reason
for concern focuses on potentials for increased damages or irreversible losses of unique
and threatened systems. Examples are coral reefs, tropical glaciers, endangered species,
unique ecosystems, biodiversity hotspots and small island states. The second reason for
concern is the ’Risk of Extreme Weather Events’ including extreme events with substan-
tial consequences for societies and natural systems. In particular, the increase of the
frequency, the intensity, or the consequences of heat waves, floods, droughts, wildfires
or tropical cyclones belong to this category. The ’Distribution of Impacts’ is the third
reason for concern. Regions, countries and populations are affected differently by climate
change. Some face greater harm, and some might be less harmed or would gain from a
temperature increase. The forth reason for concern are ’Aggregate Impacts’. Therefore
comprehensive metrics for impacts are collected. For example, monetary damages, lives
affected or lost lives are such measures. Finally, the last reason for concern focuses on
the ’Risks of Large-Scale Discontinuities’. This task analyzes the likelihood that certain
phenomena would occur. Such events as well are known as tipping points. Examples are
the deglaciation of the West Arctic or Greenland ice sheets or a substantial reduction of
the North Atlantic Meridorial Overturning Circulation.
Figure 1.2 (right) shows an updated burning embers diagram developed by Smith et al.
[43]. Here, new information about impacts and vulnerability and new data are taken into
account. For example an increased risk of species extinction and coral reef damages are
identified. As well, more extreme weather events are observed since the Third Assess-
ment report. New studies identify a higher vulnerability of specific populations, such as
the poor and elderly. In addition, it is likely that there will be higher damages for larger
magnitudes of increased global mean temperature. As well, the risk af additional contri-
butions to sea level rise from melting of both the Greenland and possible Arctic ice sheet
are now better understood. The updated reasons for concern show that smaller increases in
global mean temperature are estimated to lead to significant or substantial consequences.
The update of the burning embers diagram stresses the importance of a low emission
stabilization target. An temperature restriction of 2◦C global mean temperature increase
compared to the 1990 level is necessary. A significantly higher temperature increase
would most likely result in dangerous anthropogenic interference (red area). Avoiding
the necessity to deal with these impacts should rank very high on the global agenda of
interests. The European Union chose a climate policy goal of a 2◦C target in the EU
Council 2007 [13]. However, even under a stringent climate policy scenario, the world
will face some impacts of an increased global mean temperature. Therefore, at least some
adaptation to climate change has to be realized under mitigation policies.
The identified 2◦C target already implies significant changes of human life style and
behavior. The climate change problem is a challenge that can only be solved efficiently
and sustainable by a joint worldwide action. All long-lived GHG emissions increase
the global warming effect independently from the area of production. Consequently, a
global long-term cooperation is needed. The climate change problem could not be solved
by a single region or a few completely fragmented regions. Nevertheless, each world
region, each country and each small area has a different vulnerability of climate change
impacts and as well different possibilities for climate change adaptation and mitigation.
16 Chapter 1 Introduction
Adaptation strategies depend mainly on the geographical position of a region and include
water management, embankments and new crop seeds. A broad research area deals
with adaptation strategies, but this thesis focuses on mitigation. Mitigation strategies are
aimed at the reduction of GHG emissions and will be outlined in the following.
1.2 Dimensions of Mitigation
As described before, to avoid drastical damages due to climate change, all GHG emissions
have to be reduced within the next decades and kept at low levels until at least the end
of the century. The question of how emission reduction can be achieved can be analyzed
by identifying separate dimensions of mitigation. The operationalization of the latter
is described by mitigation options for carbon emission reduction. In the following, the
dimensions and options for mitigation are presented, as well as a definition of mitigation
costs and possible policy instruments for implementation.
The different dimensions of mitigation can be analyzed for example by a Kaya decompo-
sition (Kaya[26]). This is a common tool for studying emission dynamics (e.g. Rogner
[40]). Separating single reasons for emission reductions helps to identify mitigation op-
tions, which influence the amount of GHG emissions. Within the Kaya decomposition,
carbon emissions Fare split into four drivers: Population L, GDP per capita (Y/L), en-
ergy intensity of GDP (E/Y), carbon intensity of energy (F/E). The following equation
holds:
F=LY
LE
YF
E.(1.1)
If one assumes that population reduction is not a viable option, the equation demonstrates
that emission reduction can be realized in three dimensions: decreasing GDP growth, re-
ducing the energy intensity of GDP or reducing the carbon intensity. These dimensions of
mitigation can be influenced by the following mechanisms: (a) reduced economic growth,
(b) factor substitution and (c) adaptation of investment strategies. In the following it is ex-
plained how these options affect the drivers of carbon emissions. The interaction between
these the mechanisms for emission reductions determine the costs of climate change mit-
igation.
(a) Looking at historical data, during the last century economic growth was strongly
correlated with carbon emissions (Raupach [39]). The industrialized regions (e.g.
USA, Europe, Japan) are responsible for 86% of the cumulative historical carbon
emissions (Grübler and Fujii [20]). Fast growing regions like China increased their
emissions during the last decades. This trend is expected to hold on if no climate
change mitigation policy is introduced. In addition, developing countries claim the
right to grow without any external political restrictions for their economic growth
and choice of technologies leading to an even further increase in emissions. Figure
1.3 shows a Kaya decomposition of the CO2emissions of China between 1971 and
2005. Studying the reasons for emissions in this fast growing country underlines
the strong correlation of carbon emissions and economic growth in the past. The
1.2 Dimensions of Mitigation 17
Figure 1.3: Kaya decomposition of historical CO2emissions for China between 1971 and 2005.
Source: Steckel et al. [46]
increased carbon emissions between two years are allocated to population growth,
increased GDP per capita and changes in energy and carbon intensity. As become
evident by noticing the large orange shares in Figure 1.3, the fast GDP growth path
of China was mainly responsible for the steep rise of carbon emissions. Since 2000
the increased energy and carbon intensity have as been driving the carbon emissions.
The challenge of the next decades is to identify strategies to realize economic growth
independent of carbon emissions. Consequently, the other two mechanisms for mit-
igation play an important role, determining the future energy demand level and the
carbon intensity of the production processes.
(b) The production of each country is based on different input factors that can partly
be substituted by each other. Factor substitution to reduce carbon emissions can be
realized in two ways at different production levels: (i) primary energy fossil fuel
switch and (ii) substitution.
Carbon emissions from technologies based on coal or oil (e.g. conventional coal
power plants, diesel oil turbine) result in significantly higher emissions compared to
technologies based on natural gas like a gas turbine and natural gas combined cycle
power plants (Hirschberg [24]). A fuel switch towards natural gas instead of coal
and/or oil results in less carbon intensive energy production.
From the macro-economic point of view the whole energy intensity of the production
could be reduced for climate change mitigation. For example, a more labour and/or
capital intensive production can partly substitute energy use in the macro-economic
production. This implies structural changes in the production processes, for example
an extended use of the service sector for the macro-economic production. As well,
the increased use of electricity in the transport sector can substitute conventional
means of conveyance.
18 Chapter 1 Introduction
(c) The production structure of GDP in a country depends mainly on investment deci-
sions. The adaptation of investment strategies for climate change mitigation implies
three tasks: (i) capital investments driving macro-economic growth, (ii) investments
into technologies of the energy production and (iii) investments into technological
change and efficiency improvements.
Due to the macro-economic investment decisions capital accumulation and thereby
GDP growth can be influenced. As demonstrated above for China, in the past eco-
nomic growth also induced carbon emissions. A growing country needs more energy
for production processes and thereby as well carbon emissions may rise. In contrast,
if a country switches to a more capital intensive production, the development of the
capital stock becomes continuously more important for the economic growth. The
investments into the energy system of a region determine the technological mix of
the energy production. In 2005 25% of the total energy in the world is produced from
coal, 35% by the use of oil and 21% by technologies based on natural gas. Only 13%
of the world energy production was realized by renewable energy technologies (IEA
[23]). If the investments into the capacity of low carbon technologies like biomass,
wind, solar and nuclear are increased, the carbon intensity of the energy production
can be reduced. In addition, technologies to capture carbon emissions and store these
emissions in the ground (CCS technologies) might come into use to avoid drastical
climate change. Therefore investments into conventional fossil fuels would be re-
duced.
The third task of adopted investments for climate change mitigation relates to in-
vestments into efficiencies. Some studies indicate that especially energy efficiency
improvements due to technological change play a key role for climate change mit-
igation (Weyant [48], Löschel [30], Popp [36]). Efficiency improvements from a
macro-economic point of view give the possibility to produce the same GDP with
less energy. This partly decouples economic growth and energy production - the
energy intensity of a region decreases. The carbon intensity as well can be influ-
enced by technological change. New technologies with low carbon emissions have
to be developed. Existing, emerging carbon free technologies will be explored for
an adequate energy production under climate change mitigation policies. In the last
decades the investments into research and development in the energy sector were
quite small, e.g. 0.4% of the GDP in Europe in 2005. However, Nemet and Kammen
[34] analyzing data about research and development investments in the energy sector
in the U.S. find that a five to ten-fold increase in energy R&D investment is both
warranted and feasible for climate change mitigation.
All described mechanisms for mitigation result in either reduced GDP growth, lowered
energy intensity or less carbon intensity. These changes imply a deviation from a so
called business-as-usual development, in which climate change is completely ignored.
Thereby more investments into new technologies, less productive resources or more ex-
pensive production factors are used. The resulting differences between a climate policy
and a business-as-usual world represent the mitigation costs. Mitigation costs can be
measured as the increased energy system costs including the investments for the innova-
tion of new technologies, the installation of carbon free technologies and spendings for
energy resources. So the net present value of the difference of GDP or consumption in
a region describes the climate change mitigation costs. This thesis focuses on the lat-
1.3 Integrated Assessment Models 19
ter definition. Mitigation costs can vary significantly over time. While a climate policy
regime in the short-term might include macro-economic gains, there would be a time of
reconstruction with increased investment activities and lowered consumption. In addi-
tion, mitigation costs can vary between different countries and world regions. However,
climate change mitigation implies costly restructuring and investments especially in the
energy production and demand. Realization of such substantial deviation from the com-
mon development calls for a global policy regime. But in addition, adequate incentives
have to be identified.
The presented three dimensions of mitigation describe possibilities to reduce GHG emis-
sions. However, the challenge of political actors for the next century is to give such
incentives that emission reduction can be realized without dramatic economic growth re-
strictions. In policy studies different strategies to achieve GHG reductions are discussed.
Many policy studies are dealing with the question of what are the best policy instru-
ments for giving the right incentives to achieve climate change mitigation. In general, it
is distinguished between a carbon tax and an emission cap (see Newell and Pizer [45]).
Both instruments could be installed worldwide if a cooperative policy regime for the cli-
mate change problem is installed. As well, these instruments might be installed in single
regions or production sectors. A transition from a fragmented system to a cooperative
climate policy is discussed. In general, many policy instruments for climate change mit-
igation influence the three dimensions of mitigation: per capita GDP, energy and carbon
intensity.
1.3 Integrated Assessment Models
Integrates Assessment (IA) models are a valuable tool for studying the previously dis-
cussed three dimensions of mitigation and their interplay within one model. IA models
combine several disciplines: They comprise macro-economic production, a detailed en-
ergy sector and a climate system. In particular, they can represent all required quantities
(i.e. population, GDP, energy and carbon emissions), which is necessary for the calcu-
lation of costs and strategies for climate change mitigation. Studying climate change
mitigation includes natural science, political and macro-economic aspects and as well en-
gineering knowledge. Natural scientists identified the climate change threat and analyze
the impacts of different GHG emissions. The choice of political incentives necessary to
reduce such emissions is a crucial question of climate change mitigation research. As
presented in the former section, investment strategies of a region determine the energy
demand and the resulting carbon emissions. Due to higher energy efficiency and new low
carbon technologies in the energy production a carbon emission stabilization path can be
realized. This calls for engineering knowledge in a climate change mitigation analysis.
This thesis focuses on the macro-economic and energy system dimensions of cli-
mate change mitigation. Especially hybrid IA models include highly detailed energy
system modules. Consequently, the interaction between technological options and
macro-economic features can be studied. Within the applied models, political deci-
sions are mainly used for the definition of exogenous basic conditions, e.g. framing
the emission reduction target and the contributing regions. In this setting the question
of which investment strategies are necessary for carbon emission reduction is investigated.
20 Chapter 1 Introduction
Early IA models dealing with climate change emulated the world as one global area
without any barriers or differences between countries or world regions. However,
each region chooses different mitigation strategies, faces different costs and reacts in
its own way to climate change mitigation policies. Industrialized regions based their
historical growth on fossil fuels and have a lock-in to carbon intensive technologies. New
investments into carbon free technologies increase the costs of the energy production,
but these regions are also often technological leaders and might be very successful in
developing new technologies. The export of such new low carbon technologies might
be profitable for industrialized regions. Developing countries that are in the process of
installing their energy production sector call for information on which technology will be
most important in the future. Depending on the geographical determinants, the renewable
potential can vary significantly. These potentials determine for example the share and role
of wind and solar energy in the future energy production mix. Different types of potential
are considered: The theoretical, geographical, technical economic and implementation
potential. The regionally highest potential of onshore wind energy is found for the USA:
21PWh/y, while lowest figures are found for South East Asia, Southern and Western
Africa and Japan. Nevertheless, potentially high contributions of solar PV are expected
in North, East and West Africa and Australia. In Japan, OECD Europe and Eastern
Europe the relative potential is less (see Hoogwijk [22]). Moreover, the use of biomass
might induce conflicts between cheep carbon free energy production and regional food
demand. Beside the potential of renewable energy technologies, the regional endowment
with fossil reserves determine the decision of the investments into the future energy mix
and whether a country is willing to push climate change mitigation or not. However,
perhaps losses due to reduced fossil fuel exports might be compensated partly by biomass
exports in a climate policy. Because of these differences between regions influencing
the climate change mitigation costs and strategies significantly, for an adequate analysis
multi-regional IA models should be used. Nevertheless, a regionalization of an IA model
introduces new decision options like the emission permit allocation scheme.
If the world will agree on a global climate policy regime, this cooperative solution can
be realized by introducing allowances for CO2emissions. When looking at the regional
level of climate change mitigation costs, one of the key challenges is to solve the question
how to allocate emission permit rights in an equitable way. Especially under stringent
emission reduction targets, the price of a traded allowance for GHG emissions might grow
very fast. Regions with a surplus of emission rights might sell these permits profitably
to regions, that cannot reduce their emissions as fast as other countries. The literature
discusses different permit allocation schemes (Elzen [12], Chakravarty [7]). A common
scheme is for example the Contraction and Convergence principle (Meyer [33]), which
postulates that emission rights are allocated depending on the historical share of emis-
sions in the beginning and converge to equal per capita emission rights later in the century
(e.g. 2050). Other allocation schemes try not to claim economic growth and are oriented
on the GDP growth path via an intensity rule. A few studies argue that the historical GHG
emissions should be subtracted from the future emission rights, so that the cumulated
emissions over time are equal for each country. Figure 1.4 shows regional mitigation costs
for different emission permit allocation rules. Global mitigation costs are the same for all
schemes, but their impact on the regional mitigation cost distribution is high. A detailed
1.3 Integrated Assessment Models 21
Figure 1.4: Impact of different permit allocation schemes on regional mitigation costs. Each bar repre-
sents one of the following schemes: per-capita, per-GDP, C&C (contraction and convergence), C&C-
hist (contraction and convergence but historic emissions are taken into account), CDC (common but
differentiated convergence), GDR-stat (responsibility and capacity index), GDR-dyn (responsibility
and capacity index with dynamic adjustment of the capacity component), burden-per-GDP (mitigation
gap allocated according to GDP). Source: Knopf [28].
analysis of the impacts of different emission allocation schemes is postponed to chapter 3.
In the literature many IA models dealing with climate change mitigation are documented.
Different solution algorithms can be used including growth models and general equilib-
rium models. Examples for such models are RICE(Nordhaus [35]), MERGE (Manne
[31]), DEMETER (Gerlagh [16]), MIND (Edenhofer [8]) and WITCH (Bosetti [4]). The
models used in the analyses of this thesis are extended endogenous growth models of the
Ramsey type. These models maximize a global welfare-function and find an optimal so-
lution - optimal in any dimension: timing, region, technology, investment, etc. Climate
change mitigation studies often assume a fully cooperative world. In such a setting, the
investments, technologies and costs of emission reductions can be calculated in a first
best world. Such optimal mitigation behavior and strategies are studied in Chapter 2 and
Chapter 3 of this thesis. But if not all technologies installed in a hybrid model will be
available due to technological, political or social restrictions, an adaptation of the climate
change mitigation strategy is necessary.
Until now it is not clear, whether a completely carbon free technology will be developed
in the next century, capable of producing enough energy for the global demand. But
the energy system has to be rearranged for climate change mitigation even if no so
called backstop-technology will be available. Also, the use of some existing low carbon
technologies is strongly discussed. An extended or long-run use of nuclear power is
22 Chapter 1 Introduction
Figure 1.5: Mitigation costs for different models, stabilization scenarios and technological options.
’X’ indicates, where the target is not achieved. Source: Edenhofer et al. 2010.
declined by certain countries (e.g. Germany). The question of nuclear waste and the
possibility of military use of this technology scares many people. Also the installation
of CCS technologies is discussed. On the one hand it is not clear how to guarantee the
security of the depots and people get afraid of leakage. On the other hand it principally
seems to be difficult to measure the leakage of emissions from the ground (Bradshaw [6],
Friedmann [15]). As well, the intensity of the use of biomass for energy production is
debated strongly. An extended use of such technologies might get in conflict with the
food production and also might result in higher land-use-change emissions and increased
deforestation. The impacts of technological restrictions and how to overcome such
situations are analyzed in chapter 5 and 6 of this thesis.
Beside technological restrictions, the political situation may limit the area of possible
solutions. The debates of the last years showed that it seems to be difficult to install
a global climate change cooperation immediately. Perhaps different fragmented policy
regimes will be linked together later in the century. Nevertheless, some regions will not
join a climate change mitigation community anytime. Chapter 4 identifies the impacts
of a delayed global climate policy and investigates the possible advantage of early
deployment of renewable technologies in such a second-best world.
Beside studies based on one single IA model, model comparison projects dealing with cli-
mate change mitigation have been organized in the last years. Analyzing the vast amount
of results jointly enables the identification of robust effects and mechanisms for climate
policies. There are some comparison projects dealing with IA models and climate change
mitigation strategies and costs in both first-best worlds and under technological or polit-
ical restrictions. For example in the project ADAM ([11]) the results of five models for
different climate protection targets are compared. This study tried to figure out what are
the robust technological strategies for climate change mitigation. Figure 1.5 demonstrates
the costs of three models under six technological options for a 550ppm CO2eq. (left) sta-
bilization and 400ppm CO2eq. (right) stabilization scenario. All five energy-economic
models show that achieving low GHG concentration targets is technologically feasible and
economically viable. The ranking of the importance of individual technology options is
robust across the models. For all models, the aggregated costs up to 2100 are below 0.8%
global GDP for the 550ppmCO2eq stabilization, and below 2.5% for the 400ppmCO2eq
scenarios. The Report on Energy and Climate Policy in Europe (RECIPE, [10]) outlines
roadmaps towards a low-carbon world economy. This study uses three structurally differ-
1.4 Technological Change 23
ent energy-economic models to explore possible future development paths under different
athmospheric concentration targets. The results as well show that stabilizing atmospheric
CO2concentrations at 450ppm is technically feasible and economically affordable. Car-
bon capture and storage (CCS) and renewable technologies have the highest potential as
low-cost mitigation option. Very low stabilization requires advanced mitigation options
for generating negative emissions such as biomass in combination with CCS. The power
generation can be relatively easily decarbonized, while the transport sector would be more
difficult to decarbonize, especially, when electrification in this sector is not possible. In
addition, the results show that, if the world continues a business-as-usual scenario until
2020, global mitigation costs for reaching a 450ppm CO2stabilization by 2100 increases
significantly. An interesting result is that even if other regions delay climate policy until
2020, Europe will enjoy a first mover advantage when unilaterally implementing climate
policy.
However, this thesis deals with related research questions, but focuses more in detail on
some special effects of climate change mitigation. As described in the end of chapter
1, this includes a broad portfolio of technological options, the impacts of early deploy-
ments for renewable technologies, endogenous technological change and technological
spillovers.
1.4 Technological Change
A key factor for both reducing carbon and energy intensity is technological change. As
discussed in the section about the dimensions of mitigation, this can be realized by dif-
ferent strategies. Especially within IA models these different dimensions can be studied.
Beside the investments into the energy system for new technologies, the energy efficiency
development is important for the costs of climate change mitigation. Both areas can be an-
alyzed in IA models, which represent both the energy production and the macro-economy.
Many IA models include technological development by means of an exogenous formula-
tion. Implying that technological change appears to happen independently of the knowl-
edge level, research activities or policy decisions. Sanstad et.al [42] for example use a pa-
rameter named AEEI - the Autonomous Energy Efficiency Improvement. They highlight
the importance of closer attention to the empirical basis for modeling assumptions. For
example, they find substantial heterogeneity among both industries and countries, and a
number of cases of declining energy efficiency. Gerlagh and van der Zwaan [17] included
a parameter Autonomous Energy Service Efficiency Improvement (AESEI) in their sensi-
tivity analysis of mitigation costs. They find evidence that this exogenous parameter has
only a minor effect on the timing of emission reduction.
In contrast to the classical growth theory, new growth theory tries to explain technological
innovation endogenously. Examples in the literature for this theory are Romer [41] and
Grossman and Helpman [19]. They analyze the crucial question of the key drivers for
regionally differentiated economic growth and about an adequate design of endogenous
growth within a macro-economic model. The main challenge is to find a model for-
mulation that describes the impact of knowledge or human capital on the productivity
of a related economy. Within the literature it is differentiated between bottom-up and
top-down models as well as an exogenous vs. endogenous formulation for technological
24 Chapter 1 Introduction
change (see Löschel [30]). Endogenous technological change can be implemented in
three ways into models: (a) learning-by-doing (b) Research-and-Development (R&D)
and (c) technological spillovers. In the following these tools are described more in detail.
(a) When focusing on technological development, often learning curves are used for rep-
resenting technological change. First Arrow [2] described learning by doing as a
feature that decreases investment costs the more cumulative capacity of a technology
is installed. An advantage in technological change is assumed due to the produc-
tion and use of a technology. Especially in energy system models this approach is
used (see Grübler [21] and Rao [37]). Sometimes, learning curves are extended by a
learning by research mechanisms as described by Kypreos [29].
(b) R&D investments include all spendings into the development and improvements of
technologies. These investment can be realized by firms, institutions or the govern-
ment. Within models often a formulation is used that R&D investments increase pro-
duction factor efficiency. For example, the labour or energy efficiency is influenced
(e.g. MIND [8]). Alternatively, the stock of a production factor like knowledge is
influenced by R&D investments (e.g. WITCH [5]).
(c) Spillovers are expected to foster the diffusion of new technologies (Popp [36]). This
spillover can flow between production sectors, different technologies or regions. the
literature distinguishes between embodied and disembodied spillovers. In addition,
some models implement an absorptive capacity mechanism. Thereby the intensity of
the spillovers depend on basic conditions like the regional knowledge level. Perhaps
a spillovers receiving region can influence this parameter by some efforts - R&D
investments for example. Keller [27] provides a comprehensive overview on inter-
national technology diffusion and spillovers.
The presented ways for implementing endogenous technological change are adequate
and prefered by different types of models. Bottom-up models including a detailed energy
system with a large quantity of technologies mainly include endogenous technological
change by means of learning curves with learning-by-doing and/or learning-by-research
mechanisms. Top-down models, that look from a more general macro-economic perspec-
tive often implement R&D functions to influence technological improvements. Some
of these models also include technological spillovers. Within IA models, combining
a macro-economy and an energy system, all three tools for endogenous technological
change come into account. However, implementing endogenous technological change
always calls for empirical evidence and parameter calibration.
For all three channels of endogenous technological change empirical evidence can be
found. First, positive experience curves can be found for several production sectors, e.g.
manufacturing (Argotte and Epple [1]) and service sector (Rapping [38]). In energy sys-
tem models, the cumulative capacity of a technology usually determines the experience
and cost reductions (e.g. Messner [44]). Second, some empirical papers tried to identify
evidence for variables (e.g. trade liberalization) explaining growth. Therefore endoge-
nous productivity change is introduced into a model via defining a parameter Afor total
1.5 The Aim of the Thesis 25
factor productivity in the production function Y=A·F(L,T). The output Yis produced
by the production factors Land T, e.g. labour and land. Grossman and Helpman [18] con-
sider a two-sector, two-factor economy and interpret Aas the stock of knowledge capital.
Third, embodied technological spillover is analyzed in a theoretical and empirical way
by Coe and Helpman [3]. Romer (1990) deals with R&D investments and human capital
and describes disembodied technological spillover. These empirical studies find evidence
for the described tools of endogenous technological change. But how will endogenous
technological change interact with a climate policy regime?
Within the climate change mitigation analysis, often technological improvements due to
climate policies is in the interest of research. Under such a policy regime, technological
change is expected to help achieving emission reductions most efficiently. R&D invest-
ments can result in a less carbon and/or energy intensive production. For the analysis
of this induced technological change, models have to give the opportunity to influence
technological change endogenously. Depending on the technological level, possible spill-
overs and expected gains from R&D investments into renewable technologies, the energy
production mix, and the relation of all production factors are calculated endogenously.
If, in addition, a climate policy like an emission cap, technological subsidies or a carbon
tax are installed, the investment strategies might be adapted. Low carbon technologies
are fostered and the total energy efficiency is improved. However, even if the induced
technological change is small, each technological development determines both the
amount of emissions that have to be reduced and the chosen technologies for climate
change mitigation and thereby also the mitigation costs (see IMCP [9]). The chapters 2
and 6 of this thesis analyze endogenous technological change in a more detailed way,
focusing on technological spillovers and production sector efficiency improvements.
1.5 The Aim of the Thesis
This thesis concentrates on the field of climate change and especially on the costs of
mitigation. For the quantitative analysis, Integrated Assessment models are used, which
deal with different world regions. Because of the different endowments of fossil fuels,
biomass, solar and wind potentials, energy efficiencies, technological levels and macro-
economic growth paths, the costs and strategies to reduce carbon emissions vary over the
regions. However, the climate change mitigation costs depend on the investment strategies
which include R&D expenditures, investment into new technologies, capital, etc. and
which are chosen to keep a climate target. This thesis analyses these investment strategies
for climate change mitigation. The first part focuses on a world with optimal conditions
and free investment spendings and answers the following questions:
IHow do climate change mitigation costs depend on technological spillovers? What is
the effect of technological spillover bound to bilateral capital trade? Are there first
mover advantages or commitment incentives due to technological spillover?
II What are the optimal long-term investments in the energy system for climate change
mitigation? How differ mitigation costs and strategies between different world re-
gions? What are the mitigation costs under different climate policy regimes?
26 Chapter 1 Introduction
The second part of the thesis investigates technological or political restrictions for climate
change mitigation and deals with the following questions:
III How do mitigation costs increase due to delays of implementing emission caps at the
global level? Can near-term public support of renewable energy technologies contain
these increases? What are the effects of such a support on regional mitigation costs?
IV What are the impacts of different dynamics and directions of technological change on
climate change mitigation costs and strategies? How does the impact of technolog-
ical change vary under different elasticities of substitution between the production
factors? How does the impact of technological change depend on the availability of
energy technologies?
VWhen and where do R&D investments into production factor efficiency play an im-
portant role as climate change mitigation option? Can the increased mitigation costs
under limited technological options be compensated by a re-allocation of investments
into efficiency of energy related production factors?
1.6 Thesis Structure
This thesis is separated into five main chapters. Each chapter deals with one group of
research questions presented in the previous section. The first two chapters mostly as-
sume a perfect cooperative world with all technological options, while chapter 4 to 6
analyze climate change mitigation in a world with technological or political restrictions.
For the analysis different adjusted IA models are used. Chapter 2 describes a regionalized
growth model with technological spillover. Chapter 3 and 4 use a multi-regional hybrid
model with a detailed energy system part and nine respectively eleven world regions. The
model used in chapter 5 and 6 is based on the latter model but is scaled down to five
world regions. For chapter 6 a new formulation of endogenous technological change is
implemented. Chapter 7 is a synthesis chapter.
Chapter 2 focuses on technological spillovers bound to bilateral capital trade. An in-
tertemporal optimization model is designed to analyze climate policy scenarios
within a globalized world. The model MIND-RS analyses the behavior of four rep-
resentative world regions. The efficiency of energy use is increased due to importing
foreign capital. A technical description of the model and numerical results are pre-
sented in this chapter. The difference between mitigation policies that either take or
do not take technological spillovers into account are studied. In addition, first-mover
advantage and incentives for climate policy commitments are analyzed.
Leimbach, M., L. Baumstark (2010): The impact of capital trade and technological
spillovers on climate policies. Ecological Economics 69, 2341-2355
Chapter 3 uses the hybrid model REMIND-R, which links a macro-economic part and a
detailed energy system module. The calibration and structure of this multi-regional
IA model are presented. Long-term investment changes in the energy system to
attain climate protection targets are identified and compared over the modeled nine
1.6 Thesis Structure 27
world regions. The world regions are linked by global markets for goods, fossil
resources and emission permits. In addition, different policy regimes for climate
change mitigation are implemented and resulting differences of costs and trade flows
are studied.
Leimbach, M., N. Bauer, L. Baumstark, O. Edenhofer (2010): Mitigation costs in a
globalized world: climate policy analysis with REMIND-R. Environmental Modeling
and Assessment 15, 155-173
Chapter 4 also deals with the hybrid model REMIND-R. This chapter focuses on the
role of investments into renewable technologies. A set of first and second best sce-
narios are designed to analyze the impact of early deployment of renewable energy
technologies. It is figured out whether the additional costs of a delayed climate pol-
icy can be decreased by early investments into renewable technologies. As well, it
is analyzed how the emission path changes. Also the regional effects on mitigation
costs due to early renewable deployment are explored for the eleven world region of
REMIND-R.
Bauer, N., L. Baumstark, M. Leimbach (2011):The REMIND-R Model: The Role of
Renewables in the Low-Carbon Transformation. Revised to Climatic Change, Spe-
cial Issue
Chapter 5 analyses the impact of different dimensions of technological change on cli-
mate change mitigation. The dynamics and directions of technological change are
varied in the model REMIND-RS. This five world region model is based on the
hybrid model REMIND-R. The impacts on mitigation costs and strategies are dis-
cussed. The question of how the impacts of technological change are influenced by
different assumptions about the structural settings of the world is analyzed. These
structural conditions are emulated by a modified elasticity of substitution or techno-
logical restrictions.
Baumstark, L., M. Leimbach (2010): The Dimensions of Technological Change and
Their Impacts on Climate Change Mitigation, submitted to Energy Policy
Chapter 6 uses the model REMIND-RS as well, but extended by a new formulation for
investments into technological change of labour and different end-use energy sectors.
The calibration and simulation results of such an endogenous formulation are shown
in this chapter. The timing, direction and intensity of R&D investments as mitigation
option are explored. A restriction of technological options rise mitigation costs.
In addition, the role of investments into the efficiency of energy related production
factors is analyzed under such conditions and climate mitigation policy.
Baumstark, L., M. Leimbach (2011): Endogenous Sector-specific R&D Investments
into Energy Efficiency as Mitigation Option, submitted to Climatic Change
Chapter 7 synthesizes chapter 2 to 6. The results of these chapters and future research
questions are discussed.
28 Chapter 1 Introduction
Bibliography
[1] Argotte, L., D. Epple (1990). Learning curves in manufacturing. Science 247, 920-
924.
[2] Arrow, K., (1962). The economic implications of learning-bydoing. Review of Eco-
nomic Studies 29, 155/173.
[3] Coe, D.T., E. Helpman (1995). International R&D spillovers. European Economic
Review 39, 859-887.
[4] Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., Tavoni, M. (2006), WITCH: A
World Induced Technical Change Hybrid Model. The Energy Journal, Special Issue
2006, 13-38.
[5] Bosetti, Valentina, Carraro, Carlo, Massetti, Emanuele, Tavoni, Massimo (2007).
International Energy R&D Spillovers and the Economics of Greenhouse Gas Atmo-
spheric Stabilization. Energy Economics, Vol 30, Issue 6.
[6] Bradshaw J., C. Boreham, F. la Pedalina (2004) Storage retention time of CO2in
sedimentary basins: examples from petroleum systems. In: Proceedings of 7th inter-
national conference on greenhouse gas control technologies. Vol. 1: Peer-Reviewed
Papers and Plenary Presentations, IEA greenhouse gas programme, Cheltenham,
UK, 2004, IEA, Paris, France.
[7] Chakravarty, S., A. Chikkatur, H. de Coninck, S. Pacala, R. Socolow, M. Tavoni
(2009). Sharing global CO2 emission reductions among one billion high emitters.
Proc Natl Acad Sci USA, 106:11884-11888.
[8] Edenhofer, Ottmar, Bauer, Nico, Kriegler, Elmar (2004). The Impact of Technolog-
ical Change on Climate Protection and Welfare: Insights from the Model MIND.
Ecological Economics, Vol. 54, 277-292.
[9] Edenhofer, O., K. Lessmann, K. Kemfert, M. Grubb, J. Koehler (2006). Induced
Technological Change: Exploring its Implications for the Economica of Atmo-
spheric Stabilization: Synthesis Report from the Innovation Modeling Comparison
Project. The energy Journal (Special Issue: Endogenous Technological Change and
the Economicy of Atmospheric Stabilization): 57-107.
[10] Edenhofer, O., C. Carraro, J.-C. Hourcade, K. Neuhoff, G. Luderer, C. Flachsland,
M. Jakob, A. Popp, J. Steckel, J. Strohschein, N. Bauer, S. Brunner, M. Leimbach,
29
30 BIBLIOGRAPHY
H. Lotze-Campen, V. Bosetti, E. de Cian, M. Tavoni, O. Sassi, H. Waisman, R.
Crassous-Doerfler, S. Monjon, S. Dröge, H. van Essen, P. del Río, A. Türk (2009).
RECIPE - The Economics of Decarbonization. Synthesis Report.
[11] Edenhofer, O., Knopf, B., Barker, T., Baumstark, L., Bellevrat, E., Chateau, B.,
Criqui, P., Isaac, M., Kitous, A., Kypreos, S., Leimbach, M., Lessmann, K., Magné,
B., Scrieciu, S., Turton, H., van Vuuren, D.P. (2010): The economics of low stabili-
sation: exploring its implications for mitigation costs and strategies. The Energy
Journal, Volume 31 (Special Issue 1). The Economics of Low Stabilization, 2010.
[12] den Elzen, M., N. Höhne (2010). Sharing the reduction effort to limit global warm-
ing to 2◦C. Climate Policy, Volume 10, Number 3, pp. 247-260.
[13] EU Council, March 2007. Presidency conclusions. Tech. rep., Brussels European
Council.
[14] Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D.W. Fahey, J. Hay-
wood, J. Lean, D.C. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz, R.
Van Dorland, (2007). Changes in Atmospheric Constituents and in Radiative Forc-
ing. In: Climate Change 2007: The Physical Science Basis. Contribution of Working
Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change (Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt,
M.Tignor and H.L. Miller (eds.)). Cambridge University Press, Cambridge, United
Kingdom and New York, NY, USA
[15] Friedmann SJ, D. Nummedal (2003). Reassessing the geological risks of seals fail-
ure for saline aquifers and EOR projects. In: 2nd ann. carbon sequestration confer-
ence, DOE/NETL, Alexandria, VA; 2003.
[16] Gerlagh, R., van der Zwaan, B. (2003), Gross world product and consumption in a
global warming model with endogenous technological change. Resource and Energy
Economics 25, 35-57.
[17] Gerlagh, R., van der Zwaan, B. (2004). A Sensitivity Analysis of Timing and Costs
of Greenhouse Gas Emissions Reduction. Climatic Change 65, 39-71.
[18] Grossman,Gene, Helpman, Elhanan (1990). Trade, Innovation and Growth. The
American Review, Vol. 80, No. 2, 86-91.
[19] Grossman, G. M., E. Helpman (1991). Innovation and Growth in the Global Econ-
omy. MIT Press.
[20] Grübler, A., Y. Fujii (1991). Inter-generational and spatial equity issues of carbon
accounts. Energy 16, 1397–1416.
[21] Grübler, A., N. Nakicenovic, D.G. Victor (1999). Modeling technological change:
implications for the global environment. Annual Review of Energy and the Environ-
ment 24, 545/569.
[22] Hoogwijk, M. (2004). On the global and regional potential of renewable energy
sources. University of Utrecht, 256 pp. NWS-E-2004-2.
BIBLIOGRAPHY 31
[23] IEA: Key World Energy Statistics 2007.
[24] Hirschberg, S. (2005). Greenhouse gas emission reduction options: modeling and
implications. Energy 30, 2025–2041.
[25] Jacobson, M.Z., (2002). Control of fossil-fuel particulate black carbon and organic
matter, possibly the most effective method of slowing global warming. Journal of
Geophysical Research, 107(D19).
[26] Kaya, Y. (1990) Impact of Carbon Dioxide Emission Control on GNP Growth: Inter-
pretation of Proposed Scenarios, Paper presented to the IPCC Energy and Industry
subgroup, Responses strategies working group, Paris (mimeo).
[27] Keller, W. (2004). International technology Diffusion, Journal of economic Litera-
ture XLII, 752-782.
[28] Knopf, B, M. Kowarsch, M. Lüken, O. Edenhofer, G. Luderer (2010). A global
carbon market and the allocation of emission rights. In: Edenhofer, O.; Wallacher,
J.; Lotze-Campen, H.; Reder, M.; Knopf, B.; Müller, J. (eds.), Overcoming Injustice
in Climate Change - Linking Climate and Development Policy. Berlin, Germany,
Springer. under revision.
[29] Kypreos, Socrates (2005). Modeling experience curves in MERGE (model for eval-
uating regional and global effects). Energy 30, 2721-2737.
[30] Löschel, Andreas (2002). Technological change in economic models of environmen-
tal policy: a survey. Ecological Economicy 43, 105-126.
[31] Manne, A.S., Mendelsohn, M., Richels, R. (1995). MERGE - A Model for Evalu-
ating Regional and Global Effects of GHG Reduction Policies. Energy Policy 23,
17-34.
[32] McCarthy, J., O. Canziani, N. Leary, D. Dokken, K. White, eds (2001).Chlimate
Chagne 2001: Impacts, Adaptation, and Vulnerability. Contribution of Working
Group II to the Third Assessment Report of the Intergovernmental Panel on Climate
Change (Cambridge University Press, Cambridge, UK)
[33] Meyer, A. (2004). Briefing: Contraction and Convergence. Engineering Sustainabil-
ity (157). Issue 4, 189-192.
[34] Nemet, Gregory F., D.M. Kammen (2007). U.S. energy research and development:
Declining investment, increasing need, and the feasibility of expansion. Energy Pol-
icy, Vol. 35, pp. 746-755.
[35] Nordhaus, W.D., Yang, Z. (1996), A Regional Dynamic General-Equilibrium Model
of Alternative Climate-Change Strategies. American Economic Review 86, 741-765.
[36] Popp, D. (2006). Innovation in climate policy models: Implementing lessons from
the economics of R&D. Energy Economics 28, 596-609.
32 BIBLIOGRAPHY
[37] Rao, Shilpa, Ilkka Keppo, Keywan Riahi (2006). Importance of Technological
Change and Spillovers in Long-Term Climate Policy. The Energy Journal, Endoge-
nous Technological Change and the Economics of Atmospheric Stabilisation Special
Issue, 25-42.
[38] Rapping, L. (1965). Learning and World War II production functions. Review of
Economics and Statistics 47, 81-86.
[39] Raupach, M.R., G. Marland, P. Ciais, C. Le Quere, J.G. Canadell, G. Klepper, C.B.
Field (2007). Global and regional drivers of accelerating CO2 emissions. Proceed-
ings of the NAtional Academy of Sciences of the United States of America 104,
10288-10293.
[40] Rogner, H.-H., D. Zhou, R. Bradley. P. Crabbé, O. Edenhofer, B.Hare, L. Kuijpers,
M. Yamaguchi (2007). Introduction. In Climate Change 2007: Mitigation. Contribu-
tion of Working Group III to the Fourth Assessment Report of the Intergovernmen-
tal Panel on Climate Change [B. Metz, O.R. Davidson, P.R. Bosch, R. Dave, L.A.
Meyer (eds)], Cambridge University Press, Cambridge, United Kingdom and New
York, NY, USA.
[41] Romer, P.M. (1990). Endogenous Technological Change, J.P.E. 98, no.5, 71-102.
[42] Sanstad, A., J. Roy, J.A. Sathaye (2006). Estimating energy-augmenting technolog-
ical change in developing country industries. Energy Economicy 28, 720-729.
[43] Smith, J. B., S. H. Schneider, M. Oppenheimer, G. W. Yohe, W. Hare, M. D.
Mastrandrea, A. Patwardhan, I. Burton, J. Corfee-Morlot, C. H. D. Magadza, H.-
M. Füssel, A. Barrie Pittock, A. Rahman, A. Suarez and J.-P. van Ypersele et
al. (2009). Assessing dangerous climate change through an update of the Inter-
governmental Panel on Climate Change ( IPCC ) "reasons for concern. Proceed-
ings of the National Academy of Sciences of the United States of America, doi:
10.1073/pnas.0812355106.
[44] Messner, S. (1997). Endogenised Technological Learning in an Energy System
Model. Journal of Evolutionary Economics 7, 291-313.
[45] Newell, R. G., W. A. Pizer (2003). Regulating stock externalities under uncertainty.
Journal of Environmental Economics and Management 45, 416-432.
[46] Steckel, Jan, Michael Jakob, Robert Marschinski, Gunnar Luderer (submitted):
From carbonization to decarbonization? - Past trends and future scenarios for
China’s CO2 emissions. Submitted to Energy Policy
[47] Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tig-
nor, H.L. Miller (eds.) (2007). Technical Summary. In:Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment
Report of the Intergovernmental Panel on CLimate Change. Cambridge University
Press, Cambridge, United Kingdom and New York, NY, USA.
[48] Weyant, John P., T. Olavson (1999). Issues in modeling induced technological
change in energy, environmental, and climate policy. Environ Modeling and As-
sessment 4, 67-85.
Chapter 2
The impact of capital trade
and technological spillovers on climate policies∗
Marian Leimbach
Lavinia Baumstark
∗published as Leimbach, Marian, Lavinia Baumstark (2010): The impact of capital trade and techno-
logical spillovers on climate policies. Ecological Economics 69, 2341-2355.
33
34 Chapter 2 Capital trade and technological spillovers
Analysis
The impact of capital trade and technological spillovers on climate policies
Marian Leimbach ⁎, Lavinia Baumstark
Potsdam Institute for Climate Impact Research, Telegrafenberg, 14412 Potsdam, Germany
abstractarticle info
Article history:
Received 9 October 2009
Received in revised form 19 May 2010
Accepted 24 June 2010
Available online 3 August 2010
Keywords:
Climate policy
Multi-region model
Technological spillovers
International trade
In this paper, we present an intertemporal optimization model that is designed to analyze climate policy
scenarios within a globalized world which is characterized by the existence of technological spillovers. We
consider a type of technological spillovers that is bound to bilateral capital trade. Importing foreign capital
that increases the efficiency of energy use represents a mitigation option that extends the commonly
modeled portfolio. The technical details of the model are presented in this paper. The model is solved
numerically. First model applications highlight the differences between climate policy analyses which either
take or do not take technological spillovers into account. In the final part, we apply the model to investigate
first-mover advantages and commitment incentives in climate policy scenarios. The existence of both is
supported by simulation results.
© 2010 Elsevier B.V. All rights reserved.
1. Introduction
Recently, increasing attention in both research and policy-making
has been given to the interaction of international trade and climate
change (cf. Copeland and Taylor, 2005; Weber and Peters, 2009).
Climate policies are challenged by competitiveness and carbon
leakage concerns. A number of studies deal with these trade-related
issues (e.g. Böhringer and Rutherford, 2002; Peters and Hertwich,
2008). In climate policy modeling, less attention is paid to other
interregional effects like technological spillovers.
However, within the discussion about promising climate protec-
tion strategies, technological spillovers come to the fore. Spillovers are
expected to foster the diffusion of new technologies and hence
represent a component of endogenous technological change to be
included into climate policy models (cf. Löschel, 2002; Popp, 2006;
Gillingham et al., 2008).
The concept of technological spillover is based on the idea that
technological externalities, coming along with the process of capital
and knowledge accumulation, slow down the decrease of the
marginal returns on capital. Keller (2004) provides a comprehensive
overview on international technology diffusion and spillovers.
Principally, the literature distinguishes between embodied and
disembodied spillovers.
1
The former is rooted in the theoretical and
empirical work by Coe and Helpman (1995) and Grossman and
Helpman (1991), the latter is linked to the work on R&D investments
and human capital done e.g. by Romer (1990).
Disembodied spillovers represent a kind of technological change
that is driven by international diffusion of knowledge accumulated in
a freely available global pool. Embodied spillovers, in contrast,
represent technological change that is triggered by technological
know-how embodied in foreign products or directly transferred
innovations (patents).
While some empirical evidence is given (cf. Keller, 2004), the
processes of disembodied and embodied technology spillovers are far
from being fully understood. Therefore, only a few studies exist that
investigate climate policies in the presence of technological spillovers.
In Rao et al. (2006) the technological learning curves of the model
MESSAGE are subject to disembodied spillovers. Cross-sectoral
learning (“technological clusters”) is combined with inter-regional
learning. Verdolini and Galeotti (2009) and Dechezlepretre et al.
(2009) study the diffusion of energy-efficient and climate change
mitigation technologies based on patent data. The latter focus on a
direct form of embodied spillovers where innovators transfer their
inventions for the purpose of a commercial exploitation at a later
point in time.
Most prominently studied are R&D spillovers and their effects on
the stability of international environmental agreements (e.g. Carraro
and Siniscalco, 1997; Kemfert, 2004; Nagashima and Dellink, 2008).
Bosetti et al. (2008) model disembodied international energy R&D
spillovers in the WITCH model. Knowledge acquired from abroad is
combined with domestic R&D capital stock and thus contributes to the
production of new technologies at home. For a climate policy scenario
stabilizing CO
2
concentration on 450 ppm, Bosetti et al. report lower
optimal energy R&D investments and strong free-riding effects among
High Income countries, when knowledge spillovers are explicitly
modeled. However, due to spillovers, total knowledge stocks remain
unchanged and mitigation costs are slightly lower due to lower
expenditure in energy R&D.
Ecological Economics 69 (2010) 2341–2355
⁎Corresponding author. PIK—Potsdam Institute for Climate Impact Research, P.O.
Box 60 12 03, D-14412 Potsdam, Germany. Tel.: +49 331 288 2556.
E-mail address: [email protected] (M. Leimbach).
1
Another classification introduced by Jaffe (1998) distinguishes between knowl-
edge spillovers, market spillovers and network spillovers.
0921-8009/$ –see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2010.06.023
Contents lists available at ScienceDirect
Ecological Economics
journal homepage: www.elsevier.com/locate/ecolecon
2.1 Introduction 35
A few other studies exist that take disembodied R&D spillovers
into account when analyzing climate policies (e.g. Goulder and
Schneider, 1999; Buonanno et al., 2003), but there is hardly any
climate policy study that includes embodied technological spillovers.
The present study helps to fill this gap. It tries to answer the question
whether first-mover advantages and incentives to join international
climate agreements can be derived in a model framework that
endogenize technological spillovers on the level of world regions. In
this study, the first mover-advantage refers to benefits that regions
which push energy-efficient technologies can expect in a climate
policy setting.
The concept of embodied spillovers as followed in this study can be
conceived as a process of expanding technological know-how by
capital imports. With increasing economic integration through
international trade and foreign direct investments, a country's
productivity growth is likely to depend not only on knowledge
embedded in its own technology but also in the technology imported
from its trading partners.
This paper presents a multi-region growth model that allows
analysis of climate policy scenarios in the presence of capital trade and
technological spillovers. In the implemented model two spillover
channels are considered, leading either to an increase of labor
productivity or energy efficiency. While part of the real-world
heterogeneity is disregarded, this paper aims to investigate the
impacts of modeling embodied technological spillovers in an
Integrated Assessment framework built around a stylized Ramsey-
type economic growth model, similar to the approach of Bosetti et al.
(2008). By focusing on long-run transitional dynamics and going
beyond the common approach of studying spillover effects in a
reduced-form static model, this paper contributes to both the
literature on technological spillovers and on the role of endogenous
technological change in climate policy modeling.
The paper is structured as follows. In Section 2, we discuss the
concept of embodied spillovers, its empirical evidence and the way we
implemented this feature. In Section 3, the multi-region climate policy
model is presented. We apply this model in a setting of four generic
regions which are selected in order to address the issues of first-
mover advantages and commitment incentives. In Section 4,we
describe the construction of our baseline scenarios and the definition
of the climate policy scenarios. The basic co-operative policy scenario
aims at limiting the increase of global mean temperature to 2 °C above
preindustrial level and is based on an international CO
2
cap-and-trade
system. A comprehensive discussion of the results from different
model runs is given in Section 5. A sensitivity analysis shall help to
assess the robustness of the results. We end with some conclusions in
Section 6.
2. Embodied Technological Spillover
Embodied technological spillovers refer to situations where the
presence of physical capital, produced abroad and imported, affects
efficiency or productivity levels of the host economy.
While there are some similarities with the disembodied spillover
concept, significant differences exist. Disembodied technological
spillovers refer to international knowledge as a public good. Free
flow of knowledge fosters technological innovation in places different
from where they were originally conceived, thus favoring foreign
followers at the expense of domestic R&D investors. In contrast,
embodied spillovers are bound to foreign investments and imported
goods. This provides innovators with the possibility to appropriate
part of the social benefits from their R&D investments, e.g. by
additional export opportunities. Embodied spillovers could make the
difference that helps investors in new energy technologies to break
even (Barretto and Klaassen, 2004). From the macro-perspective
adopted in this study, it could thus pay off for single regions to become
forerunners in climate policy. Brandt and Svendsen (2006) distinguish
two types of first-mover advantages. The first type materializes in
exports to countries engaging in emission reductions. The second type
exists when newly developed technologies are competitive even in
situations where countries do not have reduction targets.
The body of empirical research on spillover externalities has
grown rapidly (e.g. Lumenga-Neso et al., 2005; Jordaan, 2005; Takii,
2004). Empirical evidence is indicated for technological spillovers
from capital trade and especially for spillovers from foreign direct
investments. Both types of transfer of physical capital are thought to
be nearly the same, since technological know-how is embodied in the
machinery that is built up abroad in either way. Therefore, both
support the concept of technological spillovers applied in this paper.
Keller (2004) analyzes empirical methods of measuring techno-
logical spillovers and distinguishes three types of econometric
studies: association studies, structure studies and the general
equilibrium approach. In association studies, the authors ask whether
a specific foreign activity leads to a particular domestic technology
outcome (e.g. Aitken and Harrison, 1999). Structure studies incorpo-
rate structural elements which include a foreign technology variable
and the specification of a spillover channel or diffusion mechanism
(e.g. Coe and Helpman, 1995). Empirical analyses that apply general
equilibrium models are important because instead of focusing on
reduced-form relationships within a subset of variables, they allow to
study general equilibrium effects (e.g. Eaton and Kortum, 1996).
Eaton and Kortum (2001) combine a technology diffusion model and
a Ricardian model of trade. In the resulting model, trade augments a
country's production possibilities for the classic Ricardian reason:
trade gives access to foreign goods or, implicitly, technologies.
A majority of studies indicate positive spillover effects from
foreign investments (e.g. Kokko, 1993; Blomström et al., 1999; Hejazi
and Safarian, 1999; Takii, 2004; Jordaan, 2005). Takii (2004)
demonstrated for several countries that foreign firms, resulting from
foreign direct investments, tend to have higher productivity than
domestic ones, hence improving the host country's aggregated
productivity. Likewise, empirical results presented by Coe and
Helpman (1995), Lee (1995), Xu and Wang (1999) and by Eaton
and Kortum (2001) indicate that imported capital goods imply
technological spillovers that account for significant parts of produc-
tivity changes. In contrast, the study of Keller (1998), referring to the
findings from Coe and Helpman (1995), casts some doubts on the
claim that patterns of international trade are important in driving R&D
spillovers.
Recent empirical results have seriously challenged former find-
ings. First, they suggest that positive externalities are less prevalent
than previously thought. Second, structural factors were identified
that affect the intensity of spillover effects, among them geographical
distance and technological proximity (MacGarvie, 2005). Most
commonly recognized is the concept of absorptive capacity indicating
that spillovers have a positive effect only when domestic firms or the
host region possess sufficient knowledge and skills to absorb positive
externalities from foreign investments. Jaffe (1998) found that firms
that do little R&D themselves suffer from competitive externalities
linked to technological spillovers. A recent study by Jordaan (2005)
explicitly analyzed existing indicators of absorptive capacity, notably
technological differences between the host and foreign economy. The
findings of Jordaan, however, are not in support of the notion of
absorptive capacity, but indicate again a strong correlation between
the extent of the technology gap and positive spillovers. Better proxies
are needed to capture the effect of absorptive capacity. Hence, we
include the concept of the technology gap but not that of the
absorptive capacity in the present framework.
In following the empirical findings, we assume productivity and
efficiency parameters as source and target of technological spillovers.
Potential spillover gains (spr
i,r
), i.e. productivity improvements, for
region idepend on the technological gap between the trading
partners iand r. We assume that the higher the productivity
2342 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
36 Chapter 2 Capital trade and technological spillovers
differential (A
r
−A
i
), the higher the potential spillover effect. The
realized potential of technological spillovers then increases with
intensity of capital trade (X
r,i
)between the regions rand i. Spillover
gains are, on the one hand, due to a direct productivity increase when
the more efficient foreign capital is aggregated with domestic capital.
On the other hand, additional know-how is generated in the process
of using the imported capital stock. Both effects refer to the ratio
between the imported capital and the domestic capital stock (X
r,i
/K
i
).
By introducing spillover elasticity ψb1 we assume a decreasing
marginal spillover effect of capital exports. With spillover intensity Ω,
this altogether yields:
spi;r=Xr;i
Ki

ψ
ΩAr−Ai
ðÞ∀i;r:AibAr:ð1Þ
The resulting productivity gains are combined with technological
progress that is fueled by domestic R&D investments (see next section).
This linkage can be found in modified form in Bosetti et al. (2008), Coe
and Helpman (1995). Both elements of technological change are
combined additively. This means that domestic R&D investments are
not a prerequisite for spillovers to take effect and thus cannot be
interpreted as investments in building up absorptive capacities.
The spillover intensity Ωis a key parameter. Most authors
acknowledge its importance, but given the state of the literature, it
is difficult to provide sound empirical foundation. Variation of this
parameter (cf. Nagashima and Dellink, 2008) reveals sensitivity, yet
the selected values for spillover intensity cannot be compared across
different models. We provide our own sensitivity analysis in Section 5.
In an intertemporal optimizing framework, as it is the subject here,
spillovers provide an additional factor of technological change which
impacts regional growth dynamics. Commonly, spillovers are dealt
with as an externality. This notion involves that spillovers cannot be
anticipated and hence controlled by agents. However, while this view
may be true for disembodied international knowledge spillovers, it is
debatable in the context of embodied technological spillovers. If
empirical studies suggest a link between positive productivity gains
and capital trade, why should this not be taken into account by agents
in decision-making and why should foresighted agents not be more
proactive in attracting foreign direct investments and capital exports?
Dechezlepretre et al. (2009) show that firms are willing to invest
money in order to gain protection in foreign markets for their
inventions, thus indicating that they are aware of technological
spillovers when trying to market their superior technology there. We
apply an approach where technological spillovers are anticipated by
the regional actors, hence, influencing the dynamics of the control
variables.
3. MIND-RS
MIND-RS is a multi-region model based on the single-region global
Integrated Assessment model MIND (Edenhofer et al., 2005). MIND-
RS adopts from MIND the structure of the energy system (except for
the carbon capture and sequestration technology) and basic invest-
ment dynamics including R&D investments which represent a major
feature of endogenous technological change. As a new channel of
technological change, MIND-RS includes technological spillovers.
Unlike MIND, MIND-RS separates the aggregated industrial sector
into a consumption goods/service sector and an investment goods
sector. Moreover, MIND-RS takes trade interactions into account.
While market monopolies are excluded, trade flows represent control
variables which are bound to an intertemporal budget constraint.
Fig. 1 presents most of the new features of the macroeconomic system
of MIND-RS.
MIND-RS represents a dynamic trade model, but does not show
the sectoral detail of recursive dynamic computable general equilib-
rium models. By offering the feature of intertemporal investment
dynamics, MIND-RS is classified as an economic growth model suited
for long-term analysis. The way bilateral trade and technological
spillovers are handled as endogenous variables distinguishes MIND-
RS from models of a similar type like RICE (Nordhaus and Yang, 1996;
Nordhaus and Boyer, 2000) and MERGE (Manne et al., 1995; Kypreos
and Bahn, 2003).
3.1. Technical Description
We restrict the description here to the relevant macroeconomic
model parts. More details about the energy sector are described in
Appendix B. Furthermore, in addition to some explanation on
variables and parameters in this section, a compact list of them can
be found in Appendix A. The following indices are used throughout
the model presentation:
t1, 2, …,Ttime periods,
i,r1, 2, …,nregions,
jtradable goods and sectors,
mK,L,E,PE production factors.
With J={C,I,Q,P,f,ren,nf} and j∈Jthe following sectors and goods
are distinguished:
Cconsumption goods (tradable),
Iinvestment goods (tradable),
Qfossil energy resources (tradable) or extraction sector,
Pemission permits (tradable),
ffossil energy transformation sector,
ren renewable energy sector,
nf remaining energy sector.
Capitals from the above list simultaneously represent indices as
well as corresponding variables. This also applies to the production
factors labor (L), capital (K), final energy (E) and primary energy (PE).
We denote the sectoral index and the production factor index by a
subscript throughout the model presentation. For transparency
reasons, we use a continuous formulation for the time and region
index of the variables. Nevertheless, the model is implemented as a
discrete one.
In each region, a representative agent is assumed to summarize
households' consumption decisions and firms' investment and trade
decisions. The objective is to maximize the welfare W
WiðÞ=∫T
t¼1e−σt⋅Li;tðÞ⋅ln Ci;t
ðÞ
Li;t
ðÞ

dt ð2Þ
of nregions, where σis the pure rate of time preference and L
represents the regions' population which provides the exogenously
given production factor labor. Cdenotes consumption.
Aggregated output is the sum of the output of the consumption
goods/service sector and the investment goods sector. The production
function Y
j
of both sectors is specified as a CES function (with elasticity
of substitution parameter ρand input weight parameters ξ
m
):
Yji;t
ðÞ
=Φji
ðÞ
hξKKji;t
ðÞ
ρiðÞ +ξLθL;ji;t
ðÞA
Li;t
ðÞ
Li;t
ðÞ

ρiðÞ
+ξEθE;ji;tðÞA
Ei;tðÞEi;tðÞ

ρiðÞ
i
1
ρiðÞ ∀j∈fC;Ig;
ð3Þ
Φrepresents total factor productivity, Kthe capital stock, A
L
labor
efficiency, A
E
energy efficiency and Ethe energy input. θ
m,j
represents
the share of the respective production factors with
θm;C=1−θm;I∀m∈fL;Eg:ð4Þ
2343M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.3 MIND-RS 37
Factor market equilibrium is characterized by θ
L,j
=θ
E,j
.The
efficiency variables are subject to R&D investments rd
m
(cf. Edenhofer
et al., 2005) and technological spillovers sp
m
as described by equation
˙
Ami;tðÞ=ζiðÞ rdmi;t
ðÞ
YCi;tðÞ+YIi;tðÞ

αm
Ami;tðÞ+spmi;tðÞ∀m∈fL;Eg:
ð5Þ
Embodied technological spillovers increase labor efficiency and
energy efficiency. These spillover effects are induced by capital
exports X
I
{r,i} from region rto region i. As introduced in Section 2,
we define the spillover function for all m∈{L,E}:
spmi;t
ðÞ
=
∑r
XIr;i;tðÞ
KIi;tðÞ

ψ
ΩmAmr;tðÞ−Ami;tðÞðÞ

:Ami;tðÞbAmr;tðÞ
0:Ami;tðÞ≥Amr;tðÞ
8
>
>
<
>
>
:ð6Þ
where Ω
m
describes the spillover intensity and ψthe spillover
elasticity of foreign investments.
The budget constraint of the consumption goods and service
sector distributes the sectoral output to domestic consumption,
exports X
C
(i,r) and R&D investments:
YCi;tðÞ=Ci;tðÞ+∑
r
XCi;r;tðÞ−∑
r
XCr;i;tðÞ+rdLi;tðÞ+rdEi;tðÞ:
ð7Þ
Imports of goods X
C
(r,i) relax this constraint. For transparency
reasons we omit trading costs, which actually are assigned to all
import variables.
2
Output of the investment goods sector, added by capital imports
X
I
(r,i), is used for domestic investments I
j
into the industrial and
energy sectors, and for foreign investments X
I
(i,r):
YIi;tðÞ=ICi;tðÞ+IIi;tðÞ+Inf i;tðÞ+Ifi;tðÞ+IQi;tðÞ+Iren i;tðÞ
+∑
r
XIi;r;t
ðÞ
−∑
r
XIr;i;t
ðÞ
:ð8Þ
Capital accumulation in all sectors other than the renewable
energy sector follows the standard equation of capital stock formation
˙
Kji;tðÞ=Iji;tðÞ−δjiðÞ⋅Kji;tðÞ∀j∈fC;I;Q;fg:ð9Þ
In modeling the renewable energy sector, the concept of vintage
capital is applied (see Appendix B).
Within an international climate policy regime, we assume that
each region is allocated an amount of emission permits P. For each
unit of fossil resources converted into final energy, a permit is needed.
Emissions trading X
P
provides the opportunity to buy and sell them.
The resulting constraint for using fossil resources is given by
Qi;tðÞ+∑
r
XQr;i;tðÞ−XQi;r;tðÞ

≤PiðÞ+∑
r
XPr;i;tðÞ−XPi;r;tðÞðÞ
ð10Þ
2
Within the model implemented for numerical simulations, trade costs of 5% are
assumed for each traded good. This corresponds to the estimate of ad-valorem costs
for ocean shipping in 2000 by Hummels (2007) which amounts to 5.2%.
Fig. 1. Structure of the macroeconomic system of MIND-RS.
2344 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
38 Chapter 2 Capital trade and technological spillovers
where Qdenotes domestic fossil resource extraction and X
Q
denotes
the export and import of fossil resources.
The above system of equations forms a multi-region optimization
problem with a single objective function for eachregion. The investment
and trade variables represent control variables. In order to solve this
problem we apply the iterative algorithm developed by Leimbach and
Eisenack (2009). In assuming that the spillover effect is taken into
account when agents make investment and trade decisions, this
decentralized problem is solved as a co-operative game. Trade flows
are adjusted endogenously to find a pareto-optimum that provides
trade benefits for all regions. The applied trade algorithm iterates
between a decentralized model version where each region optimizes its
own welfare based on a given trade structure, and a Social Planner
model version where the regions' welfare functions are combined by a
set of welfare weights. The Social Planner model derives the optimal
trade structure for the given set of welfare weights, while being subject
to market clearance conditions. The welfare weights are adjusted
iteratively according to the intertemporal trade balance of each region
which has to be leveled off in the equilibrium point:
∑
r∑
j∑
t
pjtðÞXji;r;tðÞ−pjtðÞXjr;i;tðÞ

=0 j∈fC;I;Q;Pg:ð11Þ
Market prices p
j
are derived from shadow prices of the
decentralized model version. The model provides a first-best
equilibrium solution in labor, capital and goods markets implying
full employment and converging return rates across regional and
sectoral investments.
3.2. Set Up of Experiments
We want to apply the described model in a set of experiments that
aims to assess climate policy implication of technological spillovers.
We try to analyse this in a setting of four world regions —two
developed and two developing ones.
Productivity enhancing technological know-how spills over mainly
from the developed to the developing world regions. However, the two
developed world regions are distinguished by different degrees of labor
productivity and energy efficiency. Consequently, the export of
investment goods contributes relatively more to an increase of labor
productivity growth in the one case and of energy efficiency growth in
the other case. The interesting question to be answered is than: Does the
region that takes the lead in producing energy-efficient technologies
benefit in a climate policy setting (first-mover advantage)? The
developing regions differ in its starting income level and its growth
dynamics with the initially poorer region exhibiting higher economic
growth. Assuming that by means of climate protocols the transfer of
technological know-how can be intensified or restricted, the question
arises: Are there incentives for developing world regions to join
international agreements in the presence of technological spillovers?
To simplify the analysis we make two further assumptions. First, the
developed world regions only export investment goods (this is in line
with findings by Eaton and Kortum, 2001). Second, the less dynamically
growing developing world region is the only exporter ofenergy resources.
In order to increase the clearness of our study and to avoid to get
lost in an anonymous framework of investigation, we selected
existing regions as representatives of the above described generic
regions: the USA and Europe as the developed world regions, with
higher labor productivity in the USA and higher energy efficiency in
Europe, China as the fast growing developing region, and finally Rest
of the World (ROW) as major resource supplier.
3.3. Empirical Foundation
The empirical foundation of MIND-RS starts from calibration
results of the global model MIND. Parts of the model parameters are
adopted directly (in particular those of the R&D investment
functions), others needed to be regionalized.
The calibration is done for the 4 world regions Europe, China, the
USA and rest of the world (ROW). This follows the intention of
providing a good benchmark for a group of generic region. The lack of
good regional data demands for restraining from conclusions that go
beyond the generic region level.
Major data sources of MIND-RS are:
WDI (“World Development Indicators”) database
CPI (“Common Poles Image”) database
GTAP6 (“Global Trade Analysis Project”) database
PWT (“Penn World Table”) database
IEA (“International Energy Agency”) database.
Main initial values are shown in Table 1. Deriving sound initial
values for the capital stocks in the different energy sectors is most
difficult because of a lack of appropriate data. In aggregating sectoral
information from GTAP6 we derived estimates that in sum were
significantly lower than the MIND values and would result in extreme
adjustments of capital stocks in the first simulation periods. Therefore,
we only use the regional shares in the global sectoral capital stocks as
derived from GTAP6 and adjust the absolute level in order to avoid
extreme model behavior.
With respect to the parameters of the production functions we
stick to the MIND values in general. However, following findings in
the literature (cf. Bernstein et al., 1999) we differentiate between a
somewhat higher elasticity of factor substitution (0.4) in the
aggregated industrial sectors of the developed world regions and a
somewhat lower value (0.3) for the other two regions. The
distribution parameters ξ
m
are initialized according to the factors'
usual aggregated income shares and assumed to be the same in the
production functions of the consumption goods sector and the
investment goods sector. Within the fossil energy sector, share
parameters of 0.5 for capital and energy and substitution elasticities
of 0.3 are assumed for all regions. With the elasticity and income share
parameters given, we are able to compute the initial efficiency and
productivity parameters (see Appendix C for a derivation of the
calibration formula). These values are shown in Table 2. China exhibits
Table 1
Initial values.
Parameter Europe USA China ROW
GDP in trill. $US 8.8 10.0 1.16 11.2
Investment share of GDP in percent 0.22 0.24 0.27 0.24
Industrial capital stock (trill. $US) 25.7 22.4 2.74 33.6
Cap. stock in fossil energy sector
(trill. $US)
1.4 1.6 0.27 2.8
Capital stock in extraction sector
(trill. $US)
1.1 1.4 0.22 1.8
Invest. cost renew. energy sector
($US/kW)
1320 1383 1400 1330
Learning rate (renewable energy sector) 0.15 0.13 0.1 0.11
Industrial CO
2
-emissions in GtC 1.14 1.54 0.87 2.81
Table 2
Calibrated initial values of efficiency and productivity parameters.
Parameter Europe USA China ROW
Total factor productivity (industrial sectors) 0.34 0.45 0.42 0.33
Labour efficiency 0.5 0.8 0.02 0.85
Energy efficiency 5.24 3.45 0.64 2.55
Total factor productivity (fossil energy sector) 3.12 3.82 13.0 3.55
2345M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.3 MIND-RS 39
a remarkably high productivity in the fossil energy sector which is
partly due to the fact that labor is not taken into account in this sector.
Nevertheless, most of this comparative advantage is consumed by the
low energy efficiency in both industrial sectors.
The CPI database provides CO
2
emissions for the base year (see
Table 1). Initial resource extraction is derived from the emissions data
by using the carbon content coefficient of MIND and taking trade in
fossil resources into account. Initial data for resource exports are
derived from IEA's World Energy Outlook (2006).
Particular attention was paid to the calibration of the para-
meters of the marginal extraction cost curve and the learning
curves in the renewable energy sector. As to the marginal
extraction cost curve (see Eq. (17) in Appendix B), we adjusted
the χ
3
parameter such that the regional values reflect expected
scarcities and sum up to 3500 GtC —the value of global carbon
reserves assumed within MIND (Edenhofer et al., 2005). The
distribution of reserves follows the IEA World Energy Outlook
(2002) from which we derive shares of around 10%, 20%, 10% and
60%forEurope,USA,ChinaandROW,respectively.
Given the difference in the learning rates (with highest learning
rates for solar and wind technologies) and the different composition
of the renewable energy sector in the four regions, we derived
regional differentiated initial investment costs and learning rates as
shown in Table 1. The learning rates which specify the percentage cost
reduction for a doubling of cumulated capacities are assumed to be
constant over time. The regional learning curves are not subject to
international spillovers.
Although the body of empirical research on spillover externalities
has grown rapidly (see Section 2), data are restricted to case studies
mostly on the level of firms. On a country level, Coe and Helpman
(1995, p. 874) estimated that around one quarter of benefits from
R&D investments (i.e. productivity increases) in G7 countries accrue
due to trade. We selected values for the spillover intensity and the
spillover-trade elasticity such that in the baseline scenario around 15%
of labor productivity growth in ROW is due to embodied technological
spillovers. That is in the range of 10–20% indicated by Kokko (1993,
p. 161). In the default model setting, we apply the same value for the
labor-productivity-enhancing and the energy-efficiency-improving
spillover intensity.
4. Scenario Definition
4.1. Baseline Scenarios
Based on initial data and calibration, we generate a baseline
scenario that represents economic dynamics in the presence of
technological spillovers. In a first step, we contrast this scenario with
another baseline scenario that is based on the same set of parameters
and initial values but neglects the spillover effect. Missing this aspect
of endogenous technological change results in fundamentally differ-
ent growth dynamics in the spillover scenario (BAU-S) and the non-
spillover scenario (BAU) as shown in Fig. 2. This Figure combines
historic developments and model projections. The growth difference
is particularly high in China. China benefits most from technological
spillovers with a strong accelerating impact on economic dynamics.
Simultaneously, energy consumption and emissions are much higher
in the presence of technological spillovers.
Due to this baseline growth effect and the expanded emission
mitigation gap, expected mitigation costs in climate policy scenarios
would be higher in the spillover than in the non-spillover scenario.
Due to the public good characteristics of the atmosphere as a sink of
greenhouse gases, increased emission reduction needs would also
Fig. 2. GDP trajectory (until 2000, empirical data from WDI database; thereafter model simulations).
2346 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
40 Chapter 2 Capital trade and technological spillovers
affect regions that not directly profit from the baseline growth effect.
In order to disentangle this baseline effect, we constructed a new non-
spillover baseline scenario. We compensate the non-spillover scenar-
io for the missing feature of endogenous technological change by
increasing energy efficiency and labor productivity exogenously by
the same amount as in the spillover scenario. Consequently, similar
GDP paths for the spillover baseline and the non-spillover baseline
scenario (BAU-ex) are simulated. Also the CO
2
emission baselines,
increasing up to 31.5 GtC in 2100, fit to each other.
However, there are still differences in the internal dynamics. In the
spillover scenario, additional growth is induced by capital exports
from the USA and Europe. This results in a higher degree of
specialization of both regions in investment goods production.
Globally, the overall share of investments on GDP is slightly higher
in the spillover scenario. Moreover, there are significant differences in
the trade structure. E.g., while Europe and USA are importers of the
consumption good in the spillover scenario, they become goods
exporters in the absence of technological spillovers.
4.2. Policy Scenarios
The baseline scenarios represent business-as-usual dynamics by
neglecting the climate change problem. Within the policy scenarios this
problem is taken into account. By adopting the target of the EU to limit
global mean temperature increase to 2 °C above preindustrial level, the
policy scenarios frame the search for optimal mitigation policies.
Technically, we used the optimal emission path that meets the 2 °C
target in a model run with the global model MIND (Edenhofer et al.,
2005). From this global emission path we derive the amount of emission
permits that can be allocated between the four regions. In allocating the
permits, we follow the contraction & convergence approach (Meyer,
2000; Leimbach, 2003). In the base year 2000, permits are allocated
according to the status-quo, providing the USA and Europe with a higher
per capita share. Per capita allocation of permits is assumed to converge
over time with equal per capita allocation achieved in 2050. The total
amount of permits contracts over time, thus requiring emission
reduction to keep the 2 °C temperature target.
In general, opposite spillover-related impacts affect the costs of
climate change mitigation. On the one hand, the growth effect,
induced by the spillover channel that increases labor productivity,
enlarges the mitigation gap between the baseline and the policy
scenario, hence, increases the costs. The energy efficiency effect, on
the other hand, helps to fill the mitigation gap more efficiently, hence,
reduces the costs. With the construction of the baseline scenario, we
disentangled the baseline growth effect, allowing to analyse differ-
ences in climate policy implications between spillover scenarios and
non-spillover scenarios more clearly.
While in the default setting we investigate climate policies in a co-
operative world, we formulate an alternative policy scenario
representing fragmented policy regimes. We consider the possibility
that a single region will not join the grand coalition. The region that is
not willing to accept binding emission reduction commitments is
assumed to run in a business-as-usual mode, while all other regions
are committed to the same amount of emission reduction as within
the full co-operative policy regime. Total emissions increase com-
pared to the co-operative policy regime. Emissions trading is allowed
only between partners within the coalition. The non-committed
region is partly excluded from technological spillovers. The spillover
channel that affects energy efficiency is closed completely and the
intensity of spillovers that increase labor efficiency is reduced up to
50%. This is a rather extreme scenario, but it reflects the idea of issue
linking by combining a climate policy regime with a technology
protocol and some kind of trade sanctions as it is analyzed in the
literature on the stability of international environmental agreements
(e.g. Carraro and Siniscalco, 1997; Barrett, 2003; Kemfert, 2004;
Lessmann et al., 2009).
A list of all scenarios is provided in Appendix A.
5. Policy Analysis
5.1. Common Results
By discussing the results from model simulations, we first focus on
results that are robust across spillover and non-spillover scenarios.
Given the policy target, the non-spillover policy scenario COP-ex and
the spillover policy scenario COP-S come up with the same optimal
emission trajectory (see Fig. 3) as part of the co-operative solution.
The gaps between the emission baseline trajectories and the optimal
policy paths are huge. Under climate policies, emissions have to be
reduced globally by around 50% in 2050 compared to the base year
(2000) and up to 85% compared to the baseline.
Fig. 4 shows the regional distribution of mitigation costs.
Mitigation costs are measured as the percentage loss of consumption
in the policy scenario compared to the corresponding baseline
scenario. Note that benefits due to avoided climate change damages
are not taken into account, but will probably shift consumption losses
into gains when the policy scenario is compared with a reference
scenario that includes climate damages.
Irrespective of the presence or absence of spillovers, Europe faces
the lowest costs (between 1% and 2% on average) in the co-operative
policy scenarios and China faces the highest costs (between 5% and 6%
on average). The high level of mitigation costs in China is mainly due
to the high economic growth path and a comparatively low level of
energy efficiency. Due to a fast contraction of globally available
Fig. 3. Optimal CO
2
emissions.
Fig. 4. Average mitigation costs in COP-ex and COP-S scenario.
2347M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.5 Policy Analysis 41
emission permits, China suffers from a shortage of permits under the
applied contraction & convergence allocation rule. Europe benefits
from a higher level of energy efficiency and a lower level of carbon
intensity compared to the USA. ROW is a quite heterogenous region
with slow and fast growing countries. Two opposite effects have a
major impact on the level of mitigation costs in ROW. Revenue losses
from fossil resources export for one part of countries are accompanied
by gains from emissions trading for other parts of countries.
Between 10% and 30% of the mitigation costs can be saved across
all regions by emissions trading. Absolute volumes of permits traded
in scenario COP-S are shown in Fig. 5. The USA buys permits over the
whole time horizon. ROW sells permits of around 1 GtC per year over
the whole time span. China starts as a seller of emission rights.
However, as the amount of available permits contracts soon, it quickly
becomes a buyer of emission permits. The opposite applies to Europe.
European permit exports represent, however, only a small volume.
Trade pattern differences between the baseline and the policy
scenarios are mainly represented by trade in emission permits and
differences in the trade of fossil resources. Beyond that, the trade
pattern is quite robust between the baseline and the policy scenarios,
though it is different between the spillover and the non-spillover
scenarios. On the resource market, reduced demand of Europe and
USA causes export losses of ROW.
5.2. Spillovers vs. Non-spillovers
Even though mitigation cost differences between the spillover and
the non-spillover scenario are moderate (cf. Fig. 4), they indicate the
impact of spillover-specific mechanisms. Understanding these
mechanisms may help in designing effective climate policies.
The major difference between the spillover and the non-spillover
scenario in responding to climate policies is the additional possibility
of intensifying and redirecting capital trade in order to employ the
energy-efficiency-enhancing spillover effect in the spillover scenario.
Figs. 6 and 7 show the differences in consumption goods trade and
capital trade between the baseline scenarios and the corresponding
policy scenarios. In general, there is less capital trade in the policy
scenarios. Both the USA and Europe withhold and redirect part of their
investments for restructuring their domestic energy systems. However,
capital trade reduction is less distinct in the presence of technological
spillovers. This is mainly due to the fact that, in contrast to the non-
spillover scenario, capital trade reductions have a negative impact on
the labor productivity and the energy efficiency in the spillover scenario.
Due to the attractiveness of foreign capital in the spillover scenario,
trade volumes are much higher in this scenario than in the non-spillover
scenario. Higher demands on the capital and permit market cause
higher relativecapital and carbon prices in the spilloverscenario. Higher
capital prices favor the capital exporters Europe and USA. Europe can
profit from the improved terms-of-trade. For theUSA, this positive effect
is dominated by higher expenditures on the carbon market.
In contrast to the capital market, an even qualitatively different
reaction can be observed on the goods market. While the USA finances
permit imports (cf. Fig. 5) by additional goods exports in the COP-ex
scenario, it forgoes goods imports in the COP-S scenario. In contrast,
Europe uses revenues from the permit market to reduce goods exports
in the non-spillover scenario but uses such revenues to increase imports
of goods in the spillover scenario. This indicates better terms-of-trade in
the spillover scenario.
Although ROW benefits from higher carbon prices in the spillover
scenario, mitigation costs for this region are somewhat higher in the
presence of spillovers. ROW suffers from a larger negative difference in
resource prices between the COP-S and the BAU-S scenario than
between the COP-ex and the BAU-ex scenario. In contrast to ROW, China
exhibits less mitigation costs in the spillover scenario than in the non-
spillover scenario, even though capital imports are reduced compared to
the baseline. Capital trade reduction happens in both policy scenarios
but has an efficiency-decreasing impact in the spillover scenario only.
The reason for this positive mitigation cost difference is a remaining
baseline growth effect. Based on the exogenously given productivity
improvements in the non-spillover scenario, China shifts investments
Fig. 5. Flow of emission permits in COP-S scenario.
Fig. 6. Trade flow differences between policy scenario (COP-ex) and baseline scenario (BAU-ex) —cumulated net present values.
2348 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
42 Chapter 2 Capital trade and technological spillovers
across sectors and time. This yields a slightly faster growing consump-
tion path in the non-spillover scenario.
5.3. First-mover Advantage
According to available data (see Table 2), Europe has currently a
higher energy efficiency than the USA. By interpreting high initial
energy efficiency as a result of a first-mover climate policy, we
investigate the model results for effects that represent first-mover
advantages bound to the presence of technological spillovers. The
question arises: Do the lower mitigation costs for Europe in the
scenario COP-S compared to scenario COP-ex (cf. Fig. 4) indicate first-
mover advantages?
Trade patterns are crucial. Comparing the differences in capital
trade in Figs. 6 and 7, we do not observe a major redirection of capital
trade flows. While this hardly provides an argument for the first-
mover advantage, the fact that Europe in contrast or proportion to all
other regions and in contrast to its own trade pattern in the non-
spillover scenario increases its share on international trade does so.
We run an additional experiment in order to demonstrate the
significance of the first-mover advantage and to exclude that Europe's
relative gains are due to price effects that relate to the endogenous
representation of technological change in general, but not to the energy-
efficiency-enhancing spillover channel in particular. The current model
design is characterized by the fact that the growth-enhancing and the
energy-efficiency-enhancing effect oftechnological spillovers are jointly
bound to capital trade. Each decision for capital imports, intended to
access energy-efficient capital and know-how, simultaneously implies
productivity growth which can counteract the former intention. The
ratio between the intensity of both spillover effects plays therefore an
important role.
In the additional experiment, we refrain from assuming a unique
spillover intensity parameter Ω, but use an energy-efficiency-related
spillover intensity that is tenfold higher than the original value while the
labor productivity related spillover intensity is held constant.
3
Based on
this single change, we simulate a baseline scenario (BAU-H) that in the
same way as described in Section 4 is used to create a corresponding
non-spillover baseline scenario (BAU-H-ex). Consequently, the latter
differs from the BAU-ex scenario mainly due to the increased level of
exogenous energy efficiency growth.
Fig. 8 shows the mitigation costs for the new set of policy scenarios
(COP-H and COP-H-ex). While mitigation costs compared to the
default scenario (see Fig. 4) decrease in general, cost reductions are
most remarkable in China. This clearly demonstrates the first order
effect of spillovers. China exhibits the highest technology gap in
energy efficiency and hence profits most from energy-efficiency-
improving technological spillovers. Moreover, it turns out that Europe
can furthermore reduce mitigation costs in the presence of spillovers.
Mitigation costs are reduced to 1.27% which is 0.44 percentage points
less than the mitigation costs in the non-spillover scenario.
4
In
particular, Europe's share both in capital exports and in goods imports
increases compared to the baseline scenario (see Fig. 9). Why? The
reason for this is the differentiated spillover channel.
The high labor efficiency of the USA, which is subject to spillovers,
made capital exports from the USA most attractive in general. Under the
condition of a carbon-constrained world, however, energy efficiency
becomes more important. Foreign investments from Europe become
more attractive because they embody technological know-how that
contributes to a higher extent to an increase in energy efficiency of
capital importing regions than technological spillovers from invest-
ments goods of the USA.
This causal relationship explains the higher capital export of Europe in
the COP-H scenario compared to the BAU-H scenario and indicates a first-
mover advantage in the presence of technological spillovers. It results in
substantially less mitigation costs for Europe in scenario COP-H com-
pared to scenario COP-H-ex. Continually improved terms-of-trade allow
Europe to increase their shares in global trade at the expense of the USA.
Fig. 7. Trade flow differences between policy scenario (COP-S) and baseline scenario (BAU-S) —cumulated net present values.
Fig. 8. Average mitigation costs of the scenarios COP-H-ex and COP-H.
3
In Section 5.5, parameter Ωis subject to a comprehensive sensitivity analysis.
4
The mitigation cost difference between the non-spillover and the spillover
scenario amounts to 0.13 and 0.14 percentage points for ROW and the USA,
respectively. Both regions benefit in the spillover scenario from a shift of capital
trade in time.
2349M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.5 Policy Analysis 43
5.4. Fragmented Policy Regime
We finally investigate fragmented policy scenarios where China does
not join the climate policy regime. Under these scenarios, trade pattern
and mitigation costs change substantially. China will suffer from
receiving less technological spillovers. Unless the labor-productivity-
enhancing spillover channel is completely opened (as in scenario NoCH-
0—see Fig. 10), this loss cannot be compensated by a relaxed emission
constraint which allows the use of cheap fossil fuels for a longer time.
Depending on the design of the restrictiveness of the trade or
technology protocol that come into force, the additional mitigation
costs can be huge. Closing the energy-efficiency increasing spillover
channel and restricting labor-productivity enhancing technological
spillovers by 25% (NoCH-25) results in mitigation costs of 14% for
China. With a 50% reduction of labor-productivity-increasing spillovers,
costs increase to around 27% (see Fig. 10).
For the other regions, losses compared to the co-operative solution
slightly change.
5
Nevertheless, Europe cannot fully employ its first-
mover advantage. To avoid this situation, it is important that
technology protocols and trade sanctions do not only affect energy-
efficient technologies but products with embodied technological
know-how in general. Under such circumstances, there would be clear
incentives for China to accept commitments.
5.5. Sensitivity Analysis
We carry out sensitivity analyses to test robustness of results. In
particular we investigate into the sensitivity of:
•consumption and mitigation costs to spillover intensity Ω
•consumption and mitigation costs to R&D efficiency ζ.
We selected those parameters that presumably have a huge impact,
but have a weak empirical foundation. The spillover intensity relates to the
extent of endogenous technical progress with a high impact on economic
growth. The same applies to the R&D parameter which represents the
efficiency of R&D investments into labor efficiency improvements. The
variation of parameter values is always the same for each region.
For all variations we see smooth and well-behaved changes.
6
This
indicates numerical robustness. Nevertheless, sensitivity is quite high in
some cases. With respect to the spillover intensity, significant impacts
can be demonstrated for all regions, in particular for China (see Fig. 11 in
Appendix D). However, while China faces drastic changes in consump-
tion aswell as in mitigation costs, for the USA we observe high sensitivity
of mitigation costs but moderate changes in consumption. For Europe
even changes in mitigation costs are moderate. Differences in sensitiv-
ities of mitigation costs are due to the growth effect of the labor-
efficiency-enhancing spillover channel. This spillover channel is more
intensively used in the baseline scenario than in the policy scenario.
Within the selected range of variation, economic growth is very
sensitive to the R&D efficiency parameter in all regions. The higher the
efficiency of R&D investments into labor efficiency improvements, the
higher the growth (see Fig. 12 in Appendix D). Consequently, a
climate-policy-induced shift from R&D investments (and capital
imports), that increase labor efficiency, into R&D investments (and
capital imports), that increase energy efficiency, results in higher
mitigation costs in all regions but Europe. Europe benefits in the
short-run from the redirection of the other regions' expenditures to
energy-efficiency-improving capital imports (for which Europe is
most attractive).
6. Conclusions
In this paper, we analyzed the implications of modeling embodied
technological spillovers. We presented a multi-region growth model
with endogenous technological change and discussed its application
in a climate policy context. In the presence of spillovers that enhance
labor efficiency and energy efficiency, two opposite spillover effects
impact mitigation costs. While a growth effect tends to increase
mitigation costs, energy efficiency improvements reduce mitigation
Fig. 9. Trade flow differences between policy scenario (COP-H) and baseline scenario (BAU-H) —cumulated net present values.
5
Partly lower mitigation costs for the committed regions are outweighed by higher
climate change risks due to the increased amount of emissions.
6
Changes of differential measures as mitigation costs are obviously not as smooth as
those of the basic variables. Fig. 10. Mitigation costs in fragmented policy regime.
2350 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
44 Chapter 2 Capital trade and technological spillovers
costs. The higher the ratio between the spillover intensities that either
increase energy efficiency or labor productivity, the lower are the
mitigation costs for all regions. Importing foreign capital that
increases the efficiency of energy use represents a mitigation option
that extends the commonly modeled portfolio. In consequence of this
option, associated terms-of-trade effects favor capital-exporting
regions in climate policy scenarios.
Advantages in energy saving technologies thus pay off in climate
policy scenarios. This finding gives support to the hypothesis that there
are some benefits for forerunners in climate policies. This relationship
would be more distinctive if we considered not only advantages in
energy-efficient technologies but also in carbon-free energy technolo-
gies. From the sensitivityanalysis, it furthermore turnsout that thisfirst-
mover advantage is the larger the higher the efficiency of domestic R&D
investments in labor productivity growth is, making labor-efficiency-
enhancing capital imports less attractive and energy-efficiency-enhanc-
ingcapital imports more attractive. Both results emanate from modeling
the embodied typeof technologicalspillovers, but cannot be derived in a
model with disembodied spillovers only.
Simulations of policy scenarios that include embodied technological
spillovers discover various incentives for single regions to take active
part in climate policies. In particular, it turns out that restrictions on
technology transfers will provide incentives for developing world
regions to join a climate policy regime; this is in line with findings from
the literature on the stability of international environmental agree-
ments which are mainly based on disembodied spillovers. In addition,
this study showed that it is important that restrictions on technology
transfers do not only affect energy-efficient technologies but products
with embodied technological know-how in general.
All model results are subject to a number of assumptions and
simplifications, in particular:
•limited empirical foundation (this applies for example to the
differentiation between the intensity parameters of the labor
productivity and the energy efficiency spillover function)
•neglecting the impact of absorptive capacities
•foreign and domestic goods are considered as perfect substitutes,
which results in undue specialization
•limited disaggregation of the energy sector; neglect of the option of
carbon capturing and sequestration (neglecting this option results
in higher average mitigation costs than those simulated by the
global reference model MIND).
This list gives rise to future research demand.
Appendix A. Parameters, Variables and Scenarios
Symbol Set/index
i,rRegion
jSector
mProduction factor
tTime
zTime step, five years
τVintages of renewable
Symbol Parameter
σDiscount rate
δDepreciation rate
ρ(i) Substitution elasticity in consumption and investment goods sector
ρ
f
Substitution elasticity in fossil energy sector
ξ
m
Weight parameter for factor min aggregated production function
ξ
m
f
Weight parameter for factor min fossil energy sector
Φ
j
(i) Total factor productivity in sector j in region i
D(i) Primary energy efficiency in region i
ζ
m
(i) Productivity of R&D investments in improving efficiency of factor m
in region i
α
m
Parameter of efficiency-augmenting R&D function
ψParameter of spillover function
Appendix A (continued)
Symbol Set/index
Ω
m
Spillover intensity
K
max
(i,t)Maximum productivity in extraction sector in region iat time t
k(i,t) Conversion coefficient in region iat time t
ν(i) Inverse learning rate in resource sector in region i
μLearning dampening factor
χ
1
(i) Parameter of marginal extraction cost curve in region i
χ
2
(i) Parameter of marginal extraction cost curve in region i
χ
3
(i) Parameter of marginal extraction cost curve in region i
χ
4
Parameter of marginal extraction cost curve
l(t) Load factors of vintages for renewable energy production
w(t) Weights for vintages for renewable energy production at time t
fC(i) Floor investment costs of vintages in region i
γ(i) Learning parameter in renewable energy sector in region i
Symbol Control variable
θ
m
,jShare of factor min sector j
rd
L
(i,t) R&D investments in labor efficiency in region iat time t
rd
E
(i,t) R&D investments in energy efficiency in region iat time t
I
j
(i,t) Investment in sector j in region iat time t
X
I
(i,r,t) Export of investment goods from region ito region rat time t
X
C
(i,r,t) Export of consumption goods from region ito region rat time t
X
Q
(i,r,t) Export of resources from region ito region rat time t
X
P
(i,r,t) Export of emission permits from region ito region rat time t
Symbol State variable
K
j
(i,t) Capital stock of sector jin region iat time t
A
L
(i,t) Labor efficiency in region iat time t
A
E
(i,t) Energy efficiency in region iat time t
cQ(i,t) Cumulative resource extraction in region iat time t
K(i,t)Production factor of extraction sector in region iat time t
V(i,t) Vintage of renewable energy capacities in region iat time t
κ(i,t) Variable investment costs of vintages in region iat time t
cN(i,t) Cumulative installed capacity in region iat time t
Symbol Other variable
WWelfare
C(i,t) Consumption in region iat time t
L(i,t) Labor in region iat time t—exogenous
E(i,t) Total energy in region iat time t
Y
C
(i,t) Output in consumption goods sector in region iat time t
Y
I
(i,t) Output in investment goods sector in region iat time t
I(i,t) Total investment in region iat time t
sp
L
(i,t) Spillover in labor efficiency to region iat time t
sp
E
(i,t) Spillover in energy efficiency to region iat time t
I
nf
(i,t) Investment in other energy sector in region iat time t—exogenous
E
f
(i,t) Fossil energy in region iat time t
E
ren
(i,t) Renewable energy in region iat time t
E
nf
(i,t) Energy from other energy sources in region iat time t
PE(i,t) Fossil primary energy in region iat time t
mC(i,t) Marginal extraction costs in region iat time t
Q(i,t) Resource extraction in region iat time t
EM(t) Global CO
2
emissions at time t
LU(t)CO
2
emissions from land-use change at time t
P(i,t) Emission permits in region iat time t—exogenous
p
j
(t) World market price (net present value) of good jrat time t
Symbol Scenario
BAU Business as usual without spillovers
BAU-S Business as usual with spillovers
BAU-ex Business as usual with exogenous technological change
BAU-H Business as usual with spillovers; high energy-efficiency spillovers
BAU-H-ex Business as usual with exog. techn. change; high energy-efficiency
spillovers
COP Co-operative policy scenario without spillovers
COP-S Co-operative policy scenario with spillovers
COP-ex Co-operative policy scenario with exogenous technological change
COP-H Co-operative policy scenario with spillovers; high energy-efficiency
spillovers
COP-H-ex Co-op. pol. scenario with exog. techn. change; high energy-efficiency
spillovers
NoCH-0 China is not part of the policy regime; no energy-efficiency-
enhancing spillovers
NoCH-25 Like NoCH-0; labor-productivity-enhancing spillovers reduced by 25%
NoCH-50 Like NoCH-0; labor-productivity-enhancing spillovers reduced by 50%
Parameter
2351M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.7 Appendix 45
Appendix B. Energy Sector Equations
B.1. Final Energy Sector
The fossil, renewable and remaining energy production sectors
deliver final energy
Ei;tðÞ=Efi;tðÞ+Eren i;tðÞ+Enf i;tðÞ:ð12Þ
In the fossil energy sector, final energy is generated according to
the following CES production function (with the production factors
capital K
f
and primary fossil energy PE):
Efi;tðÞ=ΦfiðÞ ξf
KKfi;tðÞ
ρf+ξf
PE DiðÞ⋅PE i;tðÞðÞ
ρf
hi
1
ρfð13Þ
In the renewable sector, final energy is produced based on the
active vintages Vand the respective load factors l:
Eren i;tðÞ=∑
τlt−τðÞ⋅Vi;t−τðÞ⋅wτðÞ:ð14Þ
wis a weighting factor that represents the still active part of the
vintages. Each vintage is a function of the investments I
ren
(see
Eq. (21)) and considered to exist over τtime steps.
The remaining energy sector provides energy E
nf
from nuclear
power, hydro power and traditional biomass sources. Its future supply
is given exogenously.
7
B.2. Fossil Resource Extraction Sector
Primary fossil energy is produced from energy resources Qand net
resource imports:
PE i;tðÞ=ki;tðÞ⋅Qi;tðÞ−∑
r
XQi;r;tðÞ−XQr;i;tðÞ


:ð15Þ
krepresents a conversion factor that converts carbon into Joule. The
extraction of fossil resources is restricted by the capacity constraint
Qi;tðÞ⋅mC i;tðÞ=Ki;tðÞ⋅KQi;tðÞ:ð16Þ
mC denotes the marginal cost of extraction (i.e. the price of
resources) and Krepresents the productivity of the capital stock
in the extraction sector. The marginal cost of extraction are derived
from the so-called Rogner curve
mC i;tðÞ=1+ χ2iðÞ
χ1iðÞ
cQ i;tðÞ
χ3iðÞ

χ4
:ð17Þ
The cumulative amount of extraction cQ is given by
cQ i;t+1ðÞ=cQ i;tðÞ+z⋅Qi;tðÞ:ð18Þ
The productivity of the capital stock in the extraction sector is
subject to learning-by-doing and evolves according to:
Kði;t+1Þ=Kði;tÞ1+ KiðÞ
max−Ki;tðÞ

z⋅νiðÞ
KiðÞ
max
Qi;tðÞ
Qi;0ðÞ

μ−1

:ð19Þ
Total anthropogenic CO
2
emissions sum up CO
2
emitted by
burning fossil fuels and emissions from land-use change:
EM tðÞ=∑
i
Qi;tðÞ+LU tðÞ:ð20Þ
Qrepresents the carbon content of extracted fossil fuels.
B.3. Renewable Energy Sector
Vintage capital Vis built up by investments and transformed into
capacity units by taking the floor costs fC and the variable investment
costs κinto account:
Vi;t+1ðÞ=z⋅Iren i;tðÞ
fC iðÞ+κi;tðÞ
:ð21Þ
Similar to the extraction sector, endogenous technological change
takes place in the renewable energy sector. Based on the cumulated
installed capacity cN, with
cN i;tðÞ=cN i;t−1ðÞ+Vi;tðÞ;ð22Þ
productivity of the renewable energy sector changes:
κi;tðÞ=κi;0ðÞ
⋅cN i;tðÞ
cN i;0ðÞ

−γiðÞ
:ð23Þ
Appendix C. Equations of Calibration
The first-order partial derivatives of the production function for
consumption goods Y
C
are
∂YC
∂L=Φρ
CξLθρ
L;CAρ
LLρ−1Y1−ρ
C;ð24Þ
∂YC
∂E=Φρ
CξEθρ
E;CAρ
EEρ−1Y1−ρ
Cð25Þ
and
∂YC
∂KC
=Φρ
CξKKρ−1
CY1−ρ
C:ð26Þ
So the income shares are
∂YC
∂L
L
YC
=Φρ
CξLθρ
L;CAρ
L
L
YC

ρ
∂YC
∂E
E
YC
=Φρ
CξEθρ
E;CAρ
E
E
YC

ρ
∂YC
∂KC
KC
YC
=Φρ
CξK
KC
YC

ρ
:
ð27Þ
This can be used for the calibration. Given the start values Y
C0
,L
0
,
E
0
,K
C0
, we can derive
∂YC
∂KC
KC
YC
=Φρ
CξK
KC
YC

ρ
=ξK
⇒Φρ
C
KC
YC

ρ
=1→ΦC=YC0
KC0
ð28Þ
∂YC
∂L
L
YC
=Φρ
CξLθρ
L;CAρ
L
L
YC

ρ
=ξL
⇒Φρ
Cθρ
L;CAρ
L
L
YC

ρ
=1→AL0=YC0
θL;CL0ΦC
=KC0
θL;CL0
ð29Þ
∂YC
∂E
E
YC
=Φρ
Cθρ
E;CAρ
EξE
E
YC

ρ
=ξE
⇒Φρ
Cθρ
E;CAρ
E
E
YC

ρ
=1→AE0=YC0
θE;CE0ΦC
=KC0
θE;CE0
:
ð30Þ
The same applies for Y
I
.
7
The global value of the production of the remaining energy sector used in MIND is
distributed according to the regional shares of this kind of energy consumption in the
CPI baseline scenario.
2352 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
46 Chapter 2 Capital trade and technological spillovers
Fig. 11. Sensitivity of consumption and mitigation costs to spillover intensity.
Appendix D. Sensitivity Analysis
2353M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.7 Appendix 47
Fig. 12. Sensitivity of consumption and mitigation costs to R&D efficiency.
2354 M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
48 Chapter 2 Capital trade and technological spillovers
References
Aitken, B.J., Harrison, A.E., 1999. Do domestic firms benefit from direct foreign
investments? Evidence from Venezuela. American Economic Review 89,
605–618.
Barrett, S., 2003. Environment and Statecraft. The Strategy of Environmental Treaty-
Making. Oxford University Press.
Barretto, L., Klaassen, G., 2004. Emissions trading and the role of learning-by-doing in
the “bottom-up”energy-systems ERIS model. International Journal of Energy
Technology and Policy 2, 70–95.
Bernstein, P.M., Montgomery, W.D., Rutherford, T.F., 1999. Global impacts of the Kyoto
agreement: results from the MS-MRT model. Resource and Energy Economics 21,
375–413.
Blomström, M., Globerman, S., Kokko, A., 1999. The determinants of host country
spillovers from foreign direct investments: review and synthesis of the literature.
SSE/EFI Working Paper Series in Economics and Finance, No. 339. School of
Economics, Stockholm.
Böhringer, C., Rutherford, T., 2002. Carbon abatement and international spillovers.
Environmental and Resource Economics 22, 391–417.
Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., Tavoni, M., 2008. International energy
R&D spillovers and the economics of greenhouse gas atmospheric stabilization.
Energy Economics 30, 2912–2929.
Brandt, U.S., Svendsen, G.T., 2006. Climate change negotiations and first-mover
advantages: the case of the wind turbine industry. Energy Policy 34, 1175–1184.
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., Siniscalco, D., 1997. International Environmental Agreements —Strategic
Policy Issues. Edward Elgar Publishing, Cheltenham.
Coe, D.T., Helpman, E., 1995. International R&D spillovers. European Economic Review
39, 859–887.
Copeland, B.R., Taylor, M.S., 2005. Free trade and global warming: a trade theory view of
the Kyoto protocol. Journal of Environmental Economics and Management 49,
205–234.
Dechezlepretre, A., Glachant, M., Hascic, I., Johnstone, N., Meniere, Y., 2009. Invention
and transfer of climate change mitigation technologies on a global scale: a study
drawing on patent data. Working Paper. FEEM Nota di Lavoro 82.2009, Fondazione
Eni Enrico Mattei, Milan.
Eaton, J., Kortum, S., 1996. Trade in ideas: patenting and productivity in OECD. Journal
of International Economics 40, 251–278.
Eaton, J., Kortum, S., 2001. Trade in capital goods. European Economic Review 45,
1195–1235.
Edenhofer, O., Bauer, N., Kriegler, E., 2005. The impact of technological change on
climate protection and welfare. Ecological Economics 54, 277–292.
Gillingham, K., Newell, R., Pizer, W., 2008. Modeling endogenous technological change
for climate policy analysis. Energy Economics 30, 2734–2753.
Goulder, L.H., Schneider, S.H., 1999. Induced technological change and the attractive-
ness of CO
2
abatement policies. Resource and Energy Economics 21, 211–253.
Grossman, G.M., Helpman, E., 1991. Innovation and Growth in the Global Economy. MIT
Press, Cambridge, MA.
Hejazi, W., Safarian, A.E., 1999. Trade, foreign direct investment, and R&D spillovers.
Journal of International Business Studies 30, 491–511.
Hummels, D., 2007. Transportation costs and international trade in the second era of
globalization. Journal of Economic Perspectives 21, 131–154.
IEA, 2002. World Energy Outlook. International Energy Agency, Paris.
IEA, 2006. World Energy Outlook. International Energy Agency, Paris.
Jaffe, A.B., 1998. The importance of “spillovers”in the policy mission of the advanced
technology program. Journal of Technology Transfer 23, 11–19.
Jordaan, A.J., 2005. Determinants of FDI-induced externalities: new empirical evidence
for Mexican manufacturing industries. World Development 33, 2103–2118.
Keller, W., 1998. Are international R&D spillovers trade-related? Analyzing spillovers
among randomly matched trade partners. European Economic Review 42, 1469–1481.
Keller, W., 2004. International technology diffusion. Journal of Economic Literature XLII,
752–782.
Kemfert, C., 2004. Climate coalitions and international trade: assessment of cooperation
incentives by issue linkage. Energy Policy 32, 455–465.
Kokko, A. (1993). Foreign Direct Investments, Host Country Characteristics, and
Spillover. Dissertation, EFI/The Economic Research Institute, Stockholm School of
Economics, Stockholm.
Kypreos, S., Bahn, O., 2003. A MERGE model with endogenous technological progress.
Environmental Modelling and Assessment 19, 333–358.
Lee, J.-W., 1995. Capital goods import and long-run growth. Journal of Development
Economics 48, 91–110.
Leimbach, M., 2003. Equity and carbon emissions trading: a model analysis. Energy
Policy 31, 1033–1044.
Leimbach, M., Eisenack, K., 2009. A trade algorithm for multi-region models subject to
spillover externalities. Computational Economics 33, 107–130.
Lessmann, K., Marschinski, R., Edenhofer, O., 2009. The effects of tariffs on coalition
formation in a dynamic global warming game. Economic Modelling 26, 641–649.
Löschel, A., 2002. Technological change in economic models of environmental policy: a
survey. Ecological Economics 43, 105–126.
Lumenga-Neso, O., Olarreaga, M., Schiff, M., 2005. On ‘indirect’trade-related R&D
spillovers. European Economic Review 49, 1785–1798.
MacGarvie, M., 2005. The determinants of international knowledge diffusion as
measured by patent citations. Economics Letters 87, 121–126.
Manne, A.S., Mendelsohn, M., Richels, R., 1995. MERGE —a model for evaluating
regional and global effects of GHG reduction policies. Energy Policy 23, 17–34.
Meyer, A., 2000. Contraction & convergence: the global solution to climate change.
Schumacher Briefing, No. 5. Green Books Ltd., Foxhole.
Nagashima, M., Dellink, R., 2008. Technology spillovers and stability of international
climate coalitions. International Environmental Agreements: Politics, Law and
Economics 8, 343–365.
Nordhaus, W.D., Boyer, J., 2000. Warming the World: Economic Models of Global
Warming. MIT Press, Cambridge, MA.
Nordhaus, W.D., Yang, Y., 1996. A regional dynamic general-equilibrium model of
alternative climate-change strategies. American Economic Review 86, 741–765.
Peters, G.P., Hertwich, E.G., 2008. CO
2
embodied in international trade with implications
for global climate policy. Environmental Science and Technology 42, 1401–1407.
Popp, D., 2006. Innovation in climate policy models: Implementing lessons from the
economics of R&D. Energy Economics 28, 596–609.
Rao, S., Peppo, I., Riahi, K., 2006. Importance of technological change and spillovers in
long-term climate policy. Energy Journal Special Issue #1 123–139.
Romer, P.M., 1990. Endogenous technological change. Journal of Political Economy 98,
S71–S102.
Takii, S., 2004. Productivity differentials between local and foreign plants in Indonesian
manufacturing, 1995. World Development 32, 1957–1969.
Verdolini, E., Galeotti, M., 2009. At home and abroad: an empirical analysis of
innovation and diffusion in energy-efficient technologies. Working Paper. FEEM
Nota di Lavoro 123.2009, Fondazione Eni Enrico Mattei, Milan.
Weber, C.L., Peters, G.P., 2009. Climate change policy and international trade: policy
considerations in the US. Energy Policy 37, 432–440.
Xu, B., Wang, J., 1999. Capital goods trade and R&D spillovers in the OECD. Canadian
Journal of Economics 32, 1258–1274.
2355M. Leimbach, L. Baumstark / Ecological Economics 69 (2010) 2341–2355
2.8 References 49
50 Chapter 2 Capital trade and technological spillovers
Chapter 3
Mitigation costs in a globalized world:
climate policy analysis with REMIND-R∗
Marian Leimbach
Nico Bauer
Lavinia Baumstark
Ottmar Edenhofer
∗published as Leimbach, M., N. Bauer, L. Baumstark, O. Edenhofer (2010): Mitigation costs in a global-
ized world: climate policy analysis with REMIND-R. Environmental Modeling and Assessment 15, 155-173
51
52 Chapter 3 Mitigation costs in a globalized world
Environ Model Assess (2010) 15:155–173
DOI 10.1007/s10666-009-9204-8
Mitigation Costs in a Globalized World:
Climate Policy Analysis with REMIND-R
Marian Leimbach ·Nico Bauer ·Lavinia Baumstark ·
Ottmar Edenhofer
Received: 18 July 2008 / Accepted: 10 August 2009 / Published online: 24 October 2009
© Springer Science + Business Media B.V. 2009
Abstract Within this paper, we present the novel
hybrid model REMIND-R and its application in a cli-
mate policy context based on the EU target to avoid a
warming of the Earth’s atmosphere by more than 2◦C
compared to the pre-industrial level. This paper aims to
identify necessary long-term changes in the energy sys-
tem and the magnitude of costs to attain such a climate
protection target under different designs of the post-
2012 climate policy regime. The regional specification
of mitigation costs is analyzed in the context of global-
ization where regions are linked by global markets for
emission permits, goods, and several resources. From
simulation experiments with REMIND-R, it turns
out that quite different strategies of restructuring the
energy system are pursued by the regions. Further-
more, it is demonstrated that the variance of mitigation
costs is higher across regions than across policy regimes.
First-order impacts, in particular, reduced rents from
trade in fossil resources, prevail regardless of the design
of the policy regime.
Keywords Climate policy ·
Energy–economy–environment modeling ·
Energy system ·International trade
1 Introduction
Climate change is recognized as a major global threat
that the current and the next generations have to deal
M. Leimbach (B)·N. Bauer ·L. Baumstark ·O. Edenhofer
PIK—Potsdam Institute for Climate Impact Research,
P.O. Box 60 12 03, 14412 Potsdam, Germany
e-mail: [email protected]
with. Science is asked to provide evidence for climate
change, but also to help policy-making by exploring
options of adaptation and mitigation. Model-based
quantitative analyses are frequently used in cli-
mate policy decision-making. A number of energy–
economy–climate models were developed and applied
over the last decade—e.g., RICE [30], MERGE
[22,26], MiniCAM [8], IMAGE [1], G-Cubed [25],
MESSAGE-MACRO [29], POLES [20], AIM [19],
DEMETER [12], MIND [6], FAIR [9], E3MG
[2], WITCH [4], Imaclim-R [5]. For an overview, see
[36]and[18]. The survey given by [18] indicated that,
most recently, major progress was made in modeling
endogenous and induced technological change (see also
[7,24]).
The energy sector is a key sector for technological
change, as well as for promising mitigation strategies. A
portfolio of different technological options and a flexi-
ble investment dynamic are crucial in transforming the
energy system in a climate-friendly way. Technological
potentials differ between regions and mitigation costs
depend on regional interactions. Only a few models
take all these aspects into account. Bottom-up models
can provide a detailed description of energy technolo-
gies. However, they have been criticized for ignoring
economic feedbacks of different energy pathways [17].
Top-down models, in contrast, have addressed macro-
economic consequences of energy and climate policies.
While also representing some microeconomic realism
(e.g., captured in consumer preferences and substitu-
tion elasticities) in the absence of structural breaks
in the development and consumption styles, top-down
models are poor in technological explicitness.
We present the hybrid model—REMIND-R—that
couples a macroeconomic system module with a highly
3.1 Introduction 53
156 M. Leimbach et al.
disaggregated energy system module (cf. [3]). Hybrid
models bridge the gap between conventional top-down
and bottom-up modeling approaches [17] and become
the preferable tool in supporting policy-making. [11]
stress the importance of hybrid models and their dy-
namic formulation in providing consistent policy analy-
ses, in particular due to the endogenous formulation
of investment decisions, which allows for an explicit
description of evolving specific capital stocks and tech-
nology mixes. In REMIND-R, mitigation cost esti-
mates are based on technological opportunities and
constraints in the development of new energy tech-
nologies. Most essentially, technological change in the
energy sector (as represented in bottom-up models)
is embedded in a macroeconomic environment (as
represented by top-down models) that, by means of
investment and trade decisions, governs regional devel-
opment. Altogether, this provides a new level of climate
policy decision support and a basis for assessing future
climate policy regimes.
Based on the EU target to avoid a warming of the
Earth’s atmosphere by more than 2◦C compared to
the pre-industrial level, this paper aims to identify the
magnitude of costs to attain such a climate protection
target under different designs of the post-2012 climate
policy regime. The regional specification of mitigation
costs is analyzed in the context of globalization where
regions are linked by different global markets for emis-
sion permits, goods, and resources. Three alternative
scenarios of climate policy regimes, based on a different
initial allocation of emission rights, have been inves-
tigated: (1) contraction and convergence, (2) intensity
target, and (3) multi-stage approach. It turns out that
the ambitious climate target can be achieved at costs of
around 1.5% of global gross domestic product (GDP).
Differences between the regional costs are, however,
large. In contrast, mitigation costs across different pol-
icy regimes differ less. Nevertheless, for each region,
there is one policy regime which is more beneficial than
all others.
In contrast to previous policy regime analyses (e.g.,
[10,38]) and in following actual discussions on the
possibility of very low stabilization, this analysis con-
siders more advanced stabilization targets and provides
new technology scenarios. It comes up with a much
broader variation of regional mitigation costs based on
a detailed description of the regional energy systems
and trade linkages.
In Section 2, we present the model REMIND-R
and discuss its components in some detail, including
important assumptions and empirical foundation. Re-
sults from REMIND-R simulations for a reference (i.e.,
business-as-usual) scenario are given in Section 3.The
main focus of the paper is on the analysis of climate
policy scenarios, which are presented in Section 4.
Section 5shall provide some conclusions.
Fig. 1 Structure of
REMIND-R Welfare
LabourCapital
Energy system
costs
Output
ConsumptionInvestments
Final
energy
Energy transformation and
conversion technologies
Fuel
costs Investments Operation and
Maintenance
Labour
efficiency Emissions
Learning
by doing
Resource and
potential
constraints
Macroeconomic
module
Energy system
module
Hard link
Exogenous Data
Energy
efficiency
Trade
Trade
Trade
Climate
module
54 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 157
2 Model Description REMIND-R
REMIND-R is a novel multi-regional hybrid model,
which couples an economic growth model with a de-
tailed energy system model and a simple climate model
(see Fig. 1). The individual regions are coupled by
means of a trade module. Only a few other hybrid mod-
els exist that are based on economic growth models—
MERGE, Imaclim-R, and WITCH are the most well-
known. With MERGE and WITCH, REMIND-R
shares the same intertemporal structure, but is distin-
guished from both by a higher degree of technological
resolution in the energy sector. This feature expands
the range of mitigation options, which are mainly based
on a switch between energy technologies, and com-
pensates for a restricted representation of technolog-
ical learning. Whereas WITCH is more elaborated in
modeling R&D investments and knowledge spillovers,
REMIND-R is more advanced in addressing trade is-
sues. Moreover, the model has no exogenous restric-
tions that provide maximum growth rates or maximum
shares in the energy mix for energy sources or technolo-
gies. Such restrictions of the solution space can quite
often be found in energy system modeling but are not
justified from our point of view. Each restriction can be
surmounted by innovation and investment.
A complete technical description of REMIND-R is
beyond the scope of this paper. We restrict ourselves to
just a few equations within this section. For a detailed
documentation, we refer to our web site.1
The applied version—REMIND-R 1.0—includes
nine world regions:
1. UCA—USA, Canada, Australia
2. EUR—EU27
3. JAP—Japan
4. CHN—China
5. IND—India
6. RUS—Russia
7. AFR—Sub-Saharan Africa (including Republic of
South Africa)
8. MEA—Middle East and North Africa
9. ROW—Rest of the world (including Latin
America, Pacific Asia, and the rest of Europe)
The population development of all regions follows
an exogenous population scenario [39]. World popula-
1On http://www.pik-potsdam.de/research/research-domains/
sustainable-solutions/remind-code-1, the technical description
of REMIND-R and the whole set of input data are available.
REMIND-R is programmed in GAMS. The code is available
from the authors on request.
tion grows from 6.6 billion in 2005 to 9.0 and 10.0 billion
in 2050 and 2100, respectively.
2.1 Macro-economy Module
The world-economic dynamics over the time horizon
2005 to 2100 is simulated by means of the macro-
economy module in REMIND-R. The time step is
5 years. Each region is modeled as a representative
household with a utility function U(r)that depends
upon the per capita consumption. With assuming the
intertemporal elasticity of substitution of per capita
consumption to be close to one, it holds:
U(r)=
T
t=t0t·e−ζ(t−t0)L(t,r)·ln C(t,r)
L(t,r) ∀r.(1)
C(t,r)represents consumption in time-step tand region
r,L(t,r)represents labor (equivalent to population),
and ζrepresents the pure rate of time preference.2
It is the objective of REMIND-R to maximise global
welfare Wthat results as a weighted sum of the regional
utility functions:
W=
r
(w(r)·U(r)).(2)
REMIND-R is run in the cost-effectiveness mode
when it is used for climate policy simulations, i.e., cli-
mate policy targets are integrated into the model by an
additional constraint (e.g., upper bound for tempera-
ture increase).
Marco-economic output, i.e., GDP, is determined by
a “constant elasticity of substitution” (CES) function
of the production factors labor, capital, and end-use
energy. End-use energy is the outcome of a nested
tree with additional CES production functions (see
Fig. 2). Each production function calculates the amount
of output (intermediate outputs and GDP), V(t,r,v
out),
from the associated factor input amounts V(t,r,v
in)
according to the following quantities:
•Parameter ρ(r,v
out):ρis calculated from the elas-
ticity of substitution3σaccording to the relation
σ=1
1−ρ
2We assume a pure rate of time preference of 3% for the simula-
tion experiments presented in later sections.
3The assumed values for the substitution elasticities (see Fig. 2)
are comparable to the values assumed by [13, p. 49]. Regarding
the nesting structure of the energy composite, we tried to repli-
cate the basic structure of energy system services composed of
mobile and stationary energy uses. Both are combined by a very
low elasticity of substitution.
3.2 Model Description REMIND-R 55
158 M. Leimbach et al.
Fig. 2 CES production structure in the macro-economy module
•Efficiency parameter A(t,r,v
in): It is calculated as
the product of an calibration-based initial value and
a time-dependent growth rate parameter.
It holds:
V(t,r,v
out)
=⎛
⎝
MCES
(A(t,r,v
in)·V(t,r,v
in))ρ(r,vout)
⎞
⎠
1/ρ(r,vout)
∀t,r,v
out.
(3)
The list MCES assigns the correct input types vin to
each output vout.
The produced GDP of a region is used for the re-
gional consumption C(t,r), investments into the macro-
economic capital stock, I(t,r), all expenditures in the
energy system, and for the export of the composite
good XG. Energy system costs consist of fuel costs
GF(t,r), investment costs GI(t,r), and operation and
maintenance costs GO(t,r). Imports of the composite
good MGincrease the disposable gross product. This
yields the following macroeconomic balance equation:
Y(t,r)−XG(t,r)+MG(t,r)≥C(t,r)+I(t,r)
+GF(t,r)+GI(t,r)
+GO(t,r)∀t,r.(4)
Macroeconomic investments enter a conventional
capital stock equation. Changes in the efficiency
A(t,r,v
in)of the individual production factors are given
by exogenous scenarios. For all energy production fac-
tors, efficiency change rates are defined in relation to la-
bor productivity changes, assuming, e.g., that efficiency
improvement for the production factors hydrogen and
electricity is higher than labor productivity growth, but
for solids and heat, it is lower. The rate of labor produc-
tivity change itself is based on a time profile that starts
on a level that is in accordance with empirical data
[32] and ends at a predefined level, which amounts to
1.2% for the developed world regions, 2.0% for China,
Russia, and ROW, and 2.5% for Africa, India, and
MEA. The transition from the initial to the final growth
rate level also differs between regions. The resulting
pattern of economic growth (see Fig. 4in Section 3)
resembles that found in the literature (e.g., [14, p. 868];
[33, p. 901]). It is characterized by decreasing growth
rates and gradual convergence of per capita incomes.
2.2 Energy System Module
The energy system module (ESM) of REMIND-R
comprises detailed technical and economic aspects of
energy transformation. It is based on the energy system
structure as it is designed for the single-region model
REMIND-G. The ESM depends, on the one hand, on
the macroeconomic output, which is used for financing
investments into energy transformation capacities, fuel
costs spendings, and expenditures for operation and
maintenance (see Eq. 4). It provides, on the other hand,
final energy Pf(t,r,es,ef,c)that is used in the macro-
economy:
V(t,r,ef)=
Ms→f
Pf(t,r,es,ef,c)∀t,r,ef.(5)
The list Ms→fdescribes the possible combinations
of secondary energy types es, final energy types ef,and
technologies c. Similar energy balances that equate pro-
duction and demand exist for primary and secondary
energy. Leontief-type technologies with efficiency pa-
rameter η
d
η(t,r,c,d)·Dp(t,r,ep,es,c,d)
=Ps(t,r,ep,es,c)∀t,r,ep,es,c(6)
transform primary energy Dpinto secondary energy
Psbased on different vintages d. Analog equations
apply for the transformation of secondary energy into
56 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 159
secondary energy of higher value and of secondary
into final energy. More than 50 different transformation
technologies are represented in the ESM (see Table 1
for an overview on primary-energy-transforming tech-
nologies). Technologies are bound to capacities, which
constrain the production potential. New capacities can
be built up by investments. The efficiency of some
technologies is assumed to increase over time, but only
for new vintages.
Multiple primary energy sources are available in the
ESM. There are renewable primary energy sources that
can be used in each period without changing the costs
of utilization in subsequent periods. However, they
cannot be used unboundedly. Region-specific and en-
ergy source-specific potentials are defined here. In ad-
dition, the potentials are classified into different grades,
which, as a result of optimization, leads to a gradual
extension of the use of renewable energy sources.
Besides, there are exhaustible primary energy
sources where the costs rise with increasing cu-
mulative extraction region-specifically and energy
source-specifically. Our assumption on the scarcity
of exhaustible resources is based on data from
ENERDATA.4Figure 3shows the reserves of ex-
haustible primary energy carriers differentiated by
energy sources and regions. Table 2shows the cost pa-
rameters for the exhaustible resources. For all regions
and energy types, the respective extraction curve starts
at the initial extraction costs. The extraction costs at
reserve limit are exactly met, when extraction reaches
the reserve limit. The initial extraction costs and those
at the reserve margin are connected by a quadratically
increasing function. Extraction of the primary energy
types beyond the reserve limit can be extended, but
the extraction costs continue to follow the quadratic
increase. The assumptions on extraction costs of fossil
resources are at the bottom edge of estimations to be
found in the literature (e.g., [34]).
As for exhaustible primary energy sources, the use
of fossil energy leads to CO2emissions, while the ap-
plication of carbon capture technologies can contribute
to a strong decrease of CO2emissions. The model
considers that captured CO2needs to be transported
and compressed prior to injection. Storage is assumed
to be in geological formations only. However, space
in geological formations is generously measured for all
regions. There is leakage in the process of capturing,
but by assumption, no leakage occurs from sequestered
CO2. Transformation technologies that use biomass
4Most recent data are available on http://www.enerdata.fr/
enerdatauk/index.html.
Table 1 Overview on primary and secondary energy carriers and the available conversion technologies
Primary energy types
Exhaustible Renewable
Coal Crude oil Natural gas Uranium Solar, wind, hydro Geothermal Biomass
Secondary Electricity PCa,IGCC
a, CoalCHP DOT GT, NGCCa,GasCHP LWR SPV
b,WT
b, Hydro HDR BioCHP
energy Hydrogen C2H2aSMRaB2H2a
types Gases C2G GasTR B2G
Heat CoalHP, CoalCHP GasHP, GasCHP GeoHP BioHP, BioCHP
Transport fuels C2LaRefinery B2La, BioEthanol
Other liquids Refinery
Solids CoalTR BioTR
PC conventional coal power plant, IGCC integrated coal gasification combined cycle power plant, CoalCHP coal combined heat and power, C2H2 coal to hydrogen, C2G coal to
gas, CoalHP coal heating plant, C2L coal to liquids, CoalTR coal transformation, DOT diesel oil turbine, GT gas turbine, NGCC natural gas combined cycle power plant, GasCHP
gas combined heat and power, SMR steam methane reforming, GasTR gas transformation, GasHP gas heating plant, LWR light water reactor, SPV solar photovoltaics, WT wind
turbine, Hydro hydroelectric power plant, HDR hot dry rock, GeoHP heat pump, BioCHP biomass combined heat and power, B2H2 biomass to hydrogen, B2G biogas plant, BioHP
biomass heating plant, B2L biomass to liquid, BioEthanol biomass to ethanol, BioTR biomass transformation
aThis technology is also available with carbon capture and sequestration (CCS)
bThis technology is characterized by endogenous technological learning
3.2 Model Description REMIND-R 57
160 M. Leimbach et al.
Fig. 3 Overview on reserves
of exhaustible primary energy
carriers. aCountries.
bEnergy carriers. Source:
ENERDATA
0 2 4 6 8 10 12
UCA
JAP
EUR
RUA
MEA
CHN
IND
AFR
ROW
ZJ
Uranium
Coal
Gas
Oil
(a)
0 10 20 30
Oil
Gas
Coal
Uranium
ZJ
UCA
JAP
EUR
RUA
MEA
CHN
IND
AFR
ROW
(b)
Table 2 Overview on cost
parameters of exhaustible
primary energy carriers
Coal Oil Natural gas Uranium
Initial extraction costs [$US per GJ] 1.5 3.5 3.5 30$US
kg
Extraction costs at reserve limit [$US per GJ] 3.5 6 6 80$US
kg
can also be complemented by CO2capturing provided
that they are used to produce fuels or hydrogen. It is
assumed that the production potentials for biomass will
increase until 2050 to around 200 EJ, where the long-
term potential is reached. All these assumptions de-
mand for sensitivity analyses as part of future research.
The investment costs for each technology are the
same in each region, with the exception of two learning
technologies, which are characterized by the fact that
their investment costs decrease by a certain percentage
(the learning rate) with each doubling of the cumulated
regional capacities. We assume learning rates of 10%
and 20% for the wind turbine and the solar photo-
voltaics technology, respectively.
When transforming secondary energy into final
energy carriers, real transformation options are of
lesser interest, but the distribution infrastructure is of
particular importance. Except for hydrogen, which can
be used for transportation and stationary energy, each
secondary energy source will be transformed into ex-
actly one final energy carrier. Losses that occur in the
distribution of secondary energy are estimated based
on statistical data differentiated by region. In modeling
the transport sector, the current model version does not
take the use of electricity into account.
2.3 Trade Module
The model REMIND-R calculates a pareto-optimal
solution that corresponds with a global planner solution
and/or a cooperative solution5. With this approach, it is
guaranteed that the necessary emission reductions are
carried out cost-efficiently and that all trade interac-
tions are directed at increasing welfare in general and
lowering mitigation costs in particular.
Trade is modeled in the following goods:
•Coal
•Gas
•Oil
•Uranium
•Composite good (aggregated output of the macro-
economic system)
•Permits (emission rights)
With Xj(t,r)and Mj(t,r)as export and import of
good jof region rin period t, the following trade
balance equation holds:
r
(Xj(t,r)−Mj(t,r)) =0∀t,j.(7)
In order to coordinate the export and import deci-
sions of the individual regions, REMIND-R uses the
Negishi-approach (cf. [23,27]). In this iterative ap-
proach, the objective functions of the individual regions
5In cases without or with internalized externalities (applies to
the climate change externality and technological learning), the
pareto-optimal solution computed by REMIND-R corresponds
also to a market solution.
58 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 161
are merged to a global objective function by means of
welfare weights w(cf. Eq. 2).
A particular pareto-optimal solution can be ob-
tained by adjusting the welfare weights according to the
intertemporal trade balances Bi(r):
Bi(r)=
t
jpij(t)·[Xi
j(t,r)−Mij(t,r)]∀r,i
(8)
wi+1(r)=f(wi,Bi(r)) ∀r,i,(9)
where irepresents the iteration index, which is skipped
from the equations above and pij(t)represents world
market prices derived as shadow prices from Eq. 7.
The higher the intertemporal trade balance deficit of a
region, the more the welfare weight of this region needs
be lowered. A lower weight has the result that goods
exports into this region contribute less but exports
from this region contribute more to the global welfare
function. This mechanism ensures that regions reduce
their intertemporal trade balance deficits.
With the new set of weights, we compute a new
solution from which we derive Bi+1(r).Thewelfare
weights are iteratively adjusted in a way such that
r|Bi+1(r)|<
r|Bi(r)|∀i(10)
and
lim
i→∞ Bi(r)=0∀r,(11)
i.e., the intertemporal trade balance converges to zero
for each region.
The trade pattern that will result from model runs
is highly impacted by the intertemporal trade balance
constraint. Each export of composite goods qualifies
the exporting region for a future import (of the same
present value), but implies for the current period a loss
of consumption. Trade with emission permits works
similarly to goods trade. Emission rights are distributed
free of charge in the different policy regimes according
to different allocation rules. The revenues from the sale
of emission rights prove completely advantageous for
the selling regions in the way that it generates entitle-
ments for future re-exports of permits or goods. Each
unit of CO2emitted by combusting fossil fuels E(t,r,c)
using technology cneeds to be covered by emission cer-
tificates (either allocated Q(t,r)net of exports XP(t,r)
or imported MP(t,r)):
c
E(t,r,c)≤Q(t,r)−XP(t,r)+MP(t,r)∀t,r.
(12)
In REMIND-R, trade in financial assets, represented
by trade in the composite good, guarantees an intertem-
poral and interregional equilibrium. The carbon price
and the interest rate can be viewed as the outcome
of speculation in forward-looking asset markets. Mod-
els omitting trade in financial assets cannot derive a
full intertemporal equilibrium in capital and simulta-
neously in other markets. In many energy–economy–
climate models, trade in permits is the only feature
of international trade. Welfare improvements by the
reallocation of capital or the reallocation of mitigation
efforts over regions or time are possible in these models
when intertemporal efficiency is violated. In contrast to
this model design, REMIND-R derives a benchmark
for a first-best intertemporal optimum in all markets.
2.4 Climate Module
Within the REMIND-R framework, the climate mod-
ule is represented as a set of equations that restrict
the welfare optimization. This version of REMIND-R
integrates a simple climate model [31]. For basic model
equations, as well as for parameter values and initial
values, see [21].
The climate module considers the impact of green-
house gas emissions and sulphate aerosols on the level
of global mean temperature. The emission of sulphates
is directly linked to the combustion of fossil fuels in
the energy sector. The radiative forcing of both the
non-CO2greenhouse gases and the CO2emissions
from land use change is taken into account by ex-
ogenous scenarios. The former follows the SRES B2-
scenario (model AIM) and the latter combines the
same scenario type with the additional assumption of
frozen CH4and N2O emissions after 2005. The climate
sensitivity—as the most important parameter of the
climate module—is set to 2.8◦C. In Section 4.6,we
briefly discuss the sensitivity of mitigation policies on
this parameter.
3 Reference Scenario
In the reference scenario (“business-as-usual” sce-
nario), we simulate a development as if climate change
has no economically and socially important effects. The
3.3 Reference Scenario 59
162 M. Leimbach et al.
2020 2040 2060 2080 2100
0
2
4
6
8
10
Year
Growth rate [%]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
World
Fig. 4 GDP growth rate
world-wide GDP of about 47 trillion $US6in 2005 will
increase to 412 trillion $US in 2100. Figure 4shows
the growth rates of the GDP for each region. China
starts with a very high growth rate of 9.3%, which will,
however, decrease to 2.2% until 2100. India and Africa
have the largest growth rates of approximately 3.6%
and 2.9% at the end of the century, whereas the regions
UCA and EUR have a growth rate of 1.2%. While the
relative gap between poor and rich regions is becoming
shorter, significant differences in the per capita income
still exist in 2100.
The development of the energy system is shown in
Fig. 5. The primary energy consumption7is increasing
continuously in the next hundred years from almost
470 EJ in 2005 to more than 1,400 EJ in 2100. A
weakening annual increase in primary energy consump-
tion is due to the population scenario, the decreasing
growth of demand in the developed countries, and the
increasing cost of fossil energy sources.
The primary energy mix remains mostly based on
fossil energy sources. Whereas the use of oil and gas
remains almost constant, the use of coal is strongly
increasing.
The economic attractiveness of coal is due to its
lower costs, the assumptions of flexible trade, and that
the use of coal is not subject to any regulations. There
is, however, a continuous increase of extraction costs
6Throughout this report, all relevant economic figures (e.g.,
GDP) are measured in constant international $US 2003 (market
exchange rate).
7The primary energy consumption of the renewable energy
sources wind, solar, and hydro power is put on the same level
as the related secondary energy production.
which, around the middle of the century, makes the use
of other energy sources competitive. Hydro energy and
especially wind energy will increasingly be used. The
use of biomass will also increase after 2030, which is
due to its increasing availability. Solar energy sources
are not employed in the reference scenario; nuclear
energy will be used as a considerable supplement for
coal at the end of the century. However, as extraction
costs of uranium increase quickly, coal consumption is
increasing again at the end of the century. Actually, on
the regional level, there is a permanent increase in the
use of coal in several developing countries, while there
is a cutback of coal consumption in the middle of the
century and a stabilization on a lower level afterwards
in all industrialized countries.
Figure 5b shows which secondary energy sources are
produced. The secondary energy production will in-
crease to around 900 EJ in 2100; the share of electricity
in particular will increase from roughly 19% to 44%.
In contrast, the use of the low-value energy sources
“solids” and “other liquids” will decrease.
From the analysis so far, it inevitably results that
there will be an increase of emissions. This is mostly
due to the conversion of coal into electricity. The
world-wide emissions amount to approximately 21 GtC
(76 Gt CO2) in 2100 (see Fig. 6). The increase of emis-
sions is quite high in the early decades—with a doubling
of the emissions between 2005 and 2025. The temporary
decrease of the emissions around 2060 accompanies the
interim reduction of the use of coal. In 2100, approx.
75% of the emissions in the energy sector originate
from the combustion of coal. Shares of approx. 15%
and 10% are allotted to oil and gas, respectively.
Large regional differences in the per capita emissions
can be observed. While the industrialized countries
increase their per capita emissions until 2025 and keep
them on a high level (5–9 tC per year) thereafter, they
rise to approx. 2–3 tC in China, India, and MEA. Africa
remains on a consistently low level with less than 1 tC
per capita.
4 Model Analysis of Climate Policy Regimes
4.1 Description of the Policy Regimes
The following analyses are based on the 2◦CEU
climate policy target. While assuming a cooperative
world, within each policy scenario, a global emission
path has to be determined that meets the 2◦C target.
Within REMIND-R, the energy-related CO2emissions
are under the control of the decision-maker only. Ex-
ogenous scenarios are applied for the development of
60 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 163
Fig. 5 Global consumption
of primary and production of
secondary energy sources in
the reference scenario. a
Primary energy. bSecondary
energy
2020 2040 2060 2080 2100
0
500
1000
1500
Year
Primary energy consumption [EJ]
Oil
Nat. Gas
Coal
Uranium
Geotherm.
Hydro
Wind
Solar
Biomass
(a)
2020 2040 2060 2080 2100
0
200
400
600
800
1000
Year
Production of secondary energy [EJ]
Solids
Oth. Liqu.
Fuels
Heat
Gases
H
2
Electricity
(b)
other greenhouse gas emissions. In the current model
setting, drastic emission reductions would have been
provided by the energy sector. While particular tech-
nologies generate negative emissions (e.g., the use of
biomass in combination with CCS), we assume that, on
a regional level, emissions are positive. Global energy-
related CO2emissions have to be reduced by 50% until
2035. The atmospheric CO2concentration reaches its
maximum at around 415 ppm in 2030.
In the analysis of how and at which costs such a
reduction path can be achieved, we investigate three
different designs of an international cap and trade sys-
tem (cf. [16]). In such a system, tradable emission rights
will be allocated to the individual regions as of 2010.
The endogenously determined global emission reduc-
tion path represents the world-wide available amount
of emission rights.
4.1.1 Contraction and Convergence (Policy Scenario A)
As of 2050, the same per capita emission rights are allo-
cated in this scenario. By determining these allocations
between 2010 and 2050, there is a smooth transition of
the regional shares between grandfathering and equal
per capita emissions. 2000 is assumed to become the
reference year for grandfathering.
4.1.2 Intensity Target (Policy Scenario B)
In this policy scenario, the shares of the regions on the
globally available emission rights correspond to their
shares in the world-wide gross product, i.e., each region
receives the same emission rights per unit GDP. In this
policy scenario, the developed countries are apparently
provided with more emission rights than in the other
two policy scenarios.
4.1.3 Multi-stage Approach (Policy Scenario C)
We selected a form of multi-stage approach in which
the quantitative reduction obligations of the individual
regions depend upon their per capita incomes. The
following four stages are distinguished:
1st stage: up to 2,000 $US per capita and year
2nd stage: up to 4,000 $US per capita and year
3rd stage: up to 8,000 $US per capita and year
4th stage: more than 8,000 $US per capita and year
Regions of the first stage are practically not obliged
to any reductions. They can, however, participate in the
emission trade and will be provided with certificates
to the amount of their reference case emissions. This
is in contrast to alternative definitions of the multi-
stage approach (cf. [10]). Regions of the second stage
will be provided with emission rights to the amount of
2020 2040 2060 2080 2100
0
5
10
15
20
25
Year
CO2 emissions [GtC]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
Fig. 6 World-wide emissions (energy-related) in the reference
scenario
3.4 Model Analysis of Climate Policy Regimes 61
164 M. Leimbach et al.
0.15 GtC per one trillion $US gross product (GDP).
Since a growth of the GDP can be expected as a rule,
this stage comprehends an increase of emission rights
for the respective regions. Regions of the third stage
are obliged to stabilize their emissions, i.e., the certifi-
cate amount last allocated in stage 2 is frozen on its
level. Regions of the fourth stage have to contribute
significantly to the emissions reduction. Their share of
emission rights results from deducting the number of
certificates used for the regions of stages 1 to 3 from the
global amount of certificates. The internal allocation
between the regions of stage 4 again follows the above-
described contraction and convergence approach.
In the base year, the industrial countries UCA, EUR,
and Japan are in stage 4. They are presumably quite
promptly followed by Russia and China, while China is
initially only in stage 2. MEA is initially also in stage 2,
ROW is in stage 3, and India and Africa are in stage 1.
4.2 Technology Development and Mitigation
Strategies
Drastic changes in the energy system are induced by cli-
mate policy. The fundamental changes compared to the
reference scenario can be summarized in five options
for action:
1. Reduction of the entire energy consumption
2. Immediate expansion of renewable energy tech-
nologies for the production of high-value energy
sources; expansion of nuclear energy
3. Application of CO2capturing and sequestration
(CCS) for the conversion of gas and coal into elec-
tricity, as well as biomass into hydrogen and fuels
4. Reducing the production of fuels and gases, since
technical mitigation options are less efficient here
5. Reducing the production of low-value energy
sources solids and other liquids
The technological development is the same in all
policy scenarios. This is due to the separability of ef-
ficiency and distribution, which applies to models of
the general equilibrium type in the absence of market
imperfections. This separability can be derived from the
Second Fundamental Theorem of Welfare Economics
(cf. [28, p. 522]) or for the case of externalities from
the Coase Theorem (cf. [35]). In general, there exists
a market price and a trade opportunity that provide
an efficient outcome no matter how the property rights
are allocated. In our case, due to emissions trading,
all regions will be enabled to follow a unique region-
specific optimal technological development path. We
renounce to present repeatedly similar development
patterns and focus on the results from policy scenario
A. This also applies for the trade patterns in the next
section.
Figure 7shows the global consumption of primary
and the production of secondary energy. Both will
be reduced in relation to the reference scenario. The
primary energy consumption reaches approx. 1,250 EJ
at the end of the century, whereas 1,430 EJ was reached
in the reference scenario. Secondary energy production
increases to roughly 770 EJ in 2100 compared to around
910 EJ in the reference scenario. The most obvious
change in the primary energy mix (compared to the
reference scenario) is the strong restriction in the use
of fossil energy sources and the stronger and earlier
expansion in the use of renewable energy sources and
nuclear energy. As of 2040, solar energy will also play
a role now. The increase of coal consumption in the
second half of the century is based on the use of CCS
technologies.
Fig. 7 Global consumption
of primary and production of
secondary energy sources in
policy scenario A. aPrimary
energy. bSecondary energy
2020 2040 2060 2080 2100
0
200
400
600
800
1000
1200
1400
Year
Primary energy production [EJ]
Oil
Nat. Gas
Coal
Uranium
Geotherm.
Hydro
Wind
Solar
Biomass
(a)
2020 2040 2060 2080 2100
0
100
200
300
400
500
600
700
800
Year
Production of secondary energy [EJ]
Solids
Oth. Liqu.
Fuels
Heat
Gases
H2
Electricity
(b)
62 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 165
2020 2040 2060 2080 2100
0
100
200
300
400
500
Year
Electricity production [EJ]
Diesel
Nat. Gas
Nat. Gas, CCS
Coal
Coal, CCS
Nuclear
HDR (Geoth.)
Hydro
Wind turbine
Solar
Biomass
Fig. 8 Global electricity production in policy scenario A
Solids and other liquids will already have been taken
out in secondary energy production. Gas, heat, and
fuels will be produced to a minor degree. The pro-
duction of hydrogen and electricity, however, will even
increase compared to the reference scenario. Electricity
production will reach 480 EJ in 2100. Similar results
can be found in [40, p. 514]. The energy mix in global
power generation is shown in Fig. 8. Wind and nuclear
technologies are dominating in the mid term, while the
CCS technology options and renewable energies (in
particular solar energy) are dominating in the long run.
When focussing on the regional energy strategies, it
turns out that, in the short term, the increase in primary
energy consumption is lower in the developed regions,
and it is even followed by a decrease in Japan and
EUR. The developed regions have in common that the
share of fossil fuels is decreasing in the first half of
the century, while the share of renewables and nuclear
energy increases. While Japan relies on nuclear energy,
UCA substitutes nuclear energy technologies as of 2050
by converting coal into electricity with CO2capturing.
The decrease in the consumption of oil and natural gas
can partially be compensated in the developed regions
by the use of biomass. This is especially obvious in
UCA.
While there are some differences in the energy mix
of the developed regions, differences are more pro-
nounced when comparing China and India on the one
side and Russia, Africa, ROW, and MEA on the other
side. The latter have high potentials in renewable en-
ergy sources that they are going to exploit to a high
degree. The largest deviation from the global pattern
of energy consumption can be observed for Russia and
MEA (see Fig. 9). MEA will employ its huge potential
of solar energy. Russia has high potentials in biomass,
which actually allows them to discontinue the use of
natural gas and to export it instead. Biomass is also the
dominating primary energy source in Africa.
The development of the energy mix in China and
India resembles the global pattern by using nuclear
energy technologies and CO2capturing in coal-fired
power plants. CO2capturing will, in addition, already
be used in gas-fired power plants as of 2040 in China.
Substantial shares of necessary gas and coal resources
will be imported.
On a regional level, results on technological devel-
opment cannot be compared because there is hardly
any literature that provides insights in this detail. Nev-
ertheless, the huge differences in the regional energy
mixes are quite remarkable. Results on technological
development on a regional level are, in general, less
robust than those on the global level. Shifts in resource
trade patterns might change the energy mix (in particu-
lar in resource-importing regions) without a significant
change in regional welfare. If comparing the results on
Fig. 9 Regional primary
energy consumption in policy
scenario A. aRUS. bMEA
2020 2040 2060 2080 2100
0
20
40
60
80
100
Year
Primary energy production [EJ]
Oil
Nat. Gas
Coal
Uranium
Geotherm.
Hydro
Wind
Solar
Biomass
(a)
2020 2040 2060 2080 2100
0
20
40
60
80
100
120
Year
Primary energy production [EJ]
Oil
Nat. Gas
Coal
Uranium
Geotherm.
Hydro
Wind
Solar
Biomass
(b)
3.4 Model Analysis of Climate Policy Regimes 63
166 M. Leimbach et al.
Fig. 10 Current account in
Africa. aReference scenario.
bPolicy scenario A
2020 2040 2060 2080 2100
–120
–100
–80
–60
–40
–20
0
20
40
Year
Current account AFR [Bill.$US]
Oil
Nat. Gas
Coal
Uranium
Permits
Goods
(a)
2020 2040 2060 2080 2100
–150
–100
–50
0
50
Year
Current account AFR [Bill.$US]
Oil
Nat. Gas
Coal
Uranium
Permits
Goods
(b)
a global level, taking model results from the Innovation
Modeling Comparison Project [7] as benchmark, the
total energy consumption is in line with MESSAGE
results. Like MESSAGE, REMIND-R simulates less
reduction of energy consumption in the policy scenario
compared to the reference scenario than most other
models. While the share of renewables is comparable
with the share simulated by other models, the high
share of biomass is striking. This applies even in the
reference scenario and results from the representation
of second generation biomass technologies based on
the most recent findings (e.g., [15,37]). While there
are plenty of options to decarbonize the production
of electricity, few options exist for the production of
fuels and gases. Biomass becomes a serious alternative
in this field. Furthermore, REMIND-R exhibits higher
shares of nuclear in the short to mid term and of coal
(combined with CCS) in the long term. Gas and oil
are, however, used less compared to most of the other
models of the Comparison Project.
4.3 Trade
The overall trade structure changes only slightly com-
pared to the reference development in all regions.
However, significant changes occur on the energy re-
source market and the carbon market.
The developed countries use the option of emissions
trade and buy permits in considerable amounts. This
import, however, is, on a value basis, hardly visible in
the current account. The basis for the current accounts
is the present value price of permits, which is in a range
between 40$US/tC and 80$US/tC. The nominal values,
however, rise quite impressively to a level of more than
500$US/tC in 2050 and even more than 6,000$US/tC
in 2100.8This indicates a very restrictive carbon con-
straint. The macroeconomic effect of emissions trad-
ing is slightly higher for the big sellers of emission
rights—ROW and, above all, Africa. This is indicated
by Fig. 10, which compares the current accounts of
Africa in the reference scenario and the policy scenario
A. In return to the sale of permits, the import of goods
is expanded in Africa. In addition, Africa produces
significant export revenues from the trade of uranium.
In contrast to the prices of fossil fuels, the price of
uranium increases in the policy scenario compared to
the reference scenario (by more than 200% even in the
short run).
In Fig. 11, resource trade differences (in physical
units cumulated over the century) between the policy
scenario and the reference scenario are shown. Neg-
ative values represent less trade (either imports or
exports) in the policy scenario. In general, trade in fossil
resources decreases and trade in uranium increases.
Trade with oil decreases significantly (up to more than
30 EJ per year in 2050). Major importers like UCA,
EUR, China, and India reduce their demands to the
account of MEA’s exports. Likewise, the trade of coal
is substantially reduced. In the reference scenario, coal
trade increases quickly. As of 2040, trade volumes in
the coal market are even higher than in the oil market.
In contrast, there is no increase at all in coal trade in the
policy scenario until 2050. Exports from UCA decrease
drastically, while China, the major coal importer, shifts
some parts of its imports into the second half of the
8While for the year 2050, the carbon price here is of the same
order of magnitude as the carbon prices simulated for the less
ambitious 450 ppm CO2stabilization scenario within the Innova-
tion Modeling Comparison Project (cf. [7, p. 96]), it is significantly
higher for the year 2100.
64 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 167
Fig. 11 Differences in trade
between policy scenario A
and reference scenario. aOil.
bCoal. cGas. dUranium
–3000 –2000 –1000 0 1000
ROW
AFR
IND
CHN
MEA
RUS
EUR
JAP
UCA
trade differences [EJ]
Export
Import
(a)
–8000 –6000 –4000 –2000 0 2000
ROW
AFR
IND
CHN
MEA
RUS
EUR
JAP
UCA
trade differences [EJ]
Export
Import
(b)
–2000 –1000 0 1000 2000
ROW
AFR
IND
CHN
MEA
RUS
EUR
JAP
UCA
trade differences [EJ]
Export
Import
(c)
–2 0 2 4 6 8
ROW
AFR
IND
CHN
MEA
RUS
EUR
JAP
UCA
trade differences [MtUr]
Export
Import
(d)
century. The diffusion of CCS technologies in the policy
scenario revitalizes the use and the international trade
of coal as of 2050.
Due to the better CO2balance, compared to the
other fossil energy sources, the short-term downturn
and the overall decrease is significantly smaller in the
trade of gas. However, major shifts in the regional
shares can be found in the gas market. EUR, UCA, and,
above all, India import less, while China increases its
imports substantially in the long run. The overall net
reduction in gas trading is at the expense of Russia.
MEA exports less in the mid term but more in the long
term.
In line with demand changes on the resource mar-
kets, we see changes in prices. The oil price in the policy
scenario is significantly lower than in the reference
scenario (see Fig. 12). The difference is somewhat lower
for gas and somewhat higher for coal, while the price
difference is reversed for uranium.
Worsened terms-of-trade can clearly be expected for
the exporters of fossil fuels—MEA, UCA, and Russia.
Less export revenues have to be compensated by less
imports of goods, which limits consumption. MEA and
Russia are probably more strongly affected than UCA,
as resource exports bear a higher share in their current
accounts.
2020 2040 2060 2080 2100
1
1.5
2
2.5
3
3.5
4
4.5
Year
Oil price index
Reference scenario
Policy scenario A
Fig. 12 Oil price index in the reference and policy scenario A
(2005 =1)
3.4 Model Analysis of Climate Policy Regimes 65
168 M. Leimbach et al.
2020 2040 2060 2080 2100
–15
–10
–5
0
5
10
Year
CO
2
emissions [GtC]
Oil
Nat. Gas
Coal
Biomass
Fig. 13 CO2emissions in policy scenario A differentiated by the
use of primary energy sources
Gains and losses from emissions trading, changes in
the energy resource market, and price-induced terms-
of-trade effects have a substantial impact on the mit-
igation costs (see Subsection 4.5). In contrast to the
effects from emissions trading, changes in the resource
market represent first-order impacts that depend on the
stabilization target but not on the allocation of permits.
4.4 Emissions and Emissions Trading
The pursued stabilization scenario requires a fast and
drastic decrease of emissions. Figure 13 shows (ex-
emplarily for policy scenario A) the emissions on the
positive side and CO2capturing on the negative side.
It can quickly be seen that the share of oil in the
entire remaining emissions is highest. Total emissions
from fossil fuels stay above 3 GtC until the end of
the century. Most of these emissions are neutralized
by CCS technologies in combination with the use of
biomass (green area in Fig. 13). The emissions are most
rapidly decreasing in electricity production. In this area,
a lot of CO2capturing is done, especially when using
coal. More than 10 GtC would be captured in 2100.
4.4.1 Contraction and Convergence (Policy Scenario A)
Figure 14a shows the permit allocation. The global sum
corresponds to the global emission trajectory. Reduc-
tions are most drastic between 2025 and 2050. The
permit share of the developing world regions and ROW
increases drastically. In the case of a missing emissions
trading market, the industrialized world regions would
need to decrease their per capita emissions to around
5% of today’s level by 2050, MEA, China, and ROW
to 20–25%, while India and Africa could still increase
their per capita emissions. For both regions, it is, how-
ever, obviously more favorable not to increase their
own emissions but to sell the allocated emission rights
profitably. Taking emissions trading into consideration,
the reductions are lower in the developed regions. The
respective per capita emissions would need to be re-
duced by approx. 20–35% in 2025 and by approx. 70–
80% in 2050. Moreover, all regions need to reach per
capita emissions of less than 1.2 tC per year in 2050,
and even less than 0.2 tC in 2100.
The trade of emissions as presented in Fig. 14bdi-
vides the big sellers (ROW, India, China, and Africa)
and the big buyers (EUR, UCA, and MEA). Initially,
permit trading is concentrated on China, ROW, EUR,
and UCA. The entire trade volume increases to more
than 1.4 GtC until 2040 and decreases then to approx.
Fig. 14 Permit allocation and
permit trade in policy
scenario A differentiated by
regions. aPermit allocation.
bPermit trade
2020 2040 2060 2080 2100
0
1
2
3
4
5
6
7
8
Year
Emission permits [GtC]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
(a)
2020 2040 2060 2080 2100
–1.5
–1
–0.5
0
0.5
1
1.5
Year
Permit trade [GtC]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
(b)
66 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 169
Fig. 15 Permit trade in policy
scenarios B and C. aPolicy
scenario B. bPolicy
scenario C
2020 2040 2060 2080 2100
–1.5
–1
–0.5
0
0.5
1
1.5
Year
Permit trade [GtC]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
(a)
2020 2040 2060 2080 2100
–2
–1
0
1
2
Year
Permit trade [GtC]
UCA
JAP
EUR
RUS
MEA
CHN
IND
AFR
ROW
(b)
0.15 GtC until 2100. With a permit price of more than
300$US/tC, transfers in the order of nearly 500 billion
$US are simulated for the year 2040. As of 2030, the
developing regions will sell more than 50% of the emis-
sions rights allocated to them. This share will, in fact,
rise up to 100% in Africa, and later, also in India and
Russia. The developed regions will increasingly cover
their emissions by buying additional emission rights.
Already in 2030, in all industrialized countries, more
than half of the emissions will be covered by buying
additional permits (in UCA even more than 75%). In
the second half of the century, this share will even
further increase; the world-wide available amount of
emission rights, however, will decrease to a level of less
than 1 GtC, and thus, the entire trade volume will also
decrease.
4.4.2 Policy Scenarios B and C
In contrast to the other policy scenarios, the distribu-
tion of emission rights according to GDP enables the
industrialized countries not only to reduce the share of
imported carbon certificates but even to sell their emis-
sion rights in a significant magnitude. Initially, Japan,
EUR, and UCA represent big sellers in the permit
market (see Fig. 15a). In return, all developing regions
(in particular, MEA and China) and Russia are buying
permits. MEA remains the largest importer of emission
permits, while China and ROW become major sellers
of permits in the mid and long terms. The peak in
emissions trading of nearly 1.5 GtC already appears in
2010. The yearly trade volume decreases fast to 0.5 GtC
in 2030, and thereafter more slowly to 0.15 GtC in 2100.
In policy scenario C, the distribution of roles be-
tween emission right purchasers and sellers is similar
to policy scenario A. However, there is a considerable
shift of shares on the sellers’ side. In policy scenario
C, China’s export shares are negligible. Since Africa
will raise its per capita income quite slowly (true for
all scenarios), it will not reach the stage where substan-
tial emission reductions will become necessary. This is
also true for India until 2070. The resultant amount of
emission rights for India and Africa restricts, on the one
hand, the allocation of emission rights to other regions
and results, on the other hand, in a quasi-monopolistic
position of Africa in the sale of emission rights after
2070. In the short to mid term, India dominates the
export of emission rights. ROW plays its role as major
exporter of permits until 2050 only.
4.5 Mitigation Costs
All policy scenarios pursue the same stabilization tar-
get. Regarding ecological efficiency (i.e., its contribu-
tion to climate stabilization), they are almost equal.
They are also similar with respect to global mitigation
costs, which is due to the above-mentioned separability
of efficiency and allocation. Global average mitigation
costs, measured as consumption losses relating to the
reference scenario, are between 1.4% and 1.5%. Global
GDP losses are of the same magnitude.9
Regional mitigation costs, however, are quite dif-
ferent. Figure 16 provides an overview of the average
regional mitigation costs for the three investigated sce-
narios. Policy scenarios A and C have similar cost struc-
tures for UCA, JAP, EUR, MEA, and ROW. While
the contraction and convergence scenario is more ben-
eficial for Russia and China, Africa and India benefit
significantly from the multi-stage scenario. Policy sce-
nario B has the smallest range in regional mitigation
costs. However, at the same time, it is also a scenario
of extremes. For many regions, it is either the most
9In general, regional GDP losses differ from regional consump-
tion losses. This is due to the effects of international trade.
3.4 Model Analysis of Climate Policy Regimes 67
170 M. Leimbach et al.
–15 –10 –5 0 5 10 15
World
ROW
AFR
IND
CHN
MEA
RUS
EUR
JAP
UCA
mitigation costs (% of BAU consumption)
pol A
pol B
pol C
Fig. 16 Average mitigation costs in policy scenarios A, B and C
(pol A, pol B and pol C)
favorable or the worst scenario. It is most favorable
for industrialized countries. The developing regions,
on the other hand, need to bear significant mitigation
costs. In the light of the distribution of the historical
responsibility for the climate problem, this could be a
heavy burden in future climate negotiations.
As a robust result, it turns out that the variance of
mitigation costs is higher between the different regions
than between the different policy scenarios. Obviously,
first-order impacts prevail regardless of the design of
the policy regime. MEA has to bear the highest costs
in all scenarios (always more than 9%). The recon-
struction of the global energy system reduces part of
the possible rents of this region, whose revenues are,
to a large part, derived from selling fossil resources.
This is, in a slightly milder form, also true for Russia
(mitigation costs of always more than 5%). For the
three developed regions UCA, Japan, and EUR, the
costs over the different scenarios develop according to
a fixed pattern. The highest mitigation costs among this
group can be found in UCA, they are slightly lower
in Europe, and they are lowest in Japan. Besides the
different base level (highest per capita emissions in
UCA), the growth pattern is also reflected in this re-
lation. In general, it holds: the higher economic growth,
the higher the mitigation costs. For all three regions,
policy scenario B is the most favorable one (average
mitigation costs amount to 1% or less). For China, the
lowest costs arise in policy scenario A; however, vari-
ance of costs between the scenarios is relatively small.
The contrary holds for India, where all scenarios but
the multi-stage scenario C are quite expensive. Africa
benefits in all policy scenarios, most remarkably in
the multi-stage scenario (more than 10% consumption
gains), which is mainly due to the fact that, for a long
time span, Africa is provided with an amount of permits
according to its baseline emissions.
The cost differences between the policy scenarios are
clearly linked to the transfers on the carbon market.
Revenues increase consumption directly, but do not
change investment decisions. Emissions trading evens
out any changes in relative prices between the different
policy scenarios. The fact that mitigation cost differ-
ences are relatively low, given the huge differences in
permit allocation, may reduce conflicts in the interna-
tional negotiation process.
4.6 Climate Sensitivity
While a comprehensive sensitivity analysis is beyond
the scope of the paper, we provide some additional
insights with respect to the impact of the climate sensi-
tivity parameter. This parameter is considered as one of
the most uncertain parameters in integrated assessment
models. Departing from the default value of 2.8◦C,
we run model experiments (exemplarily for policy sce-
nario A) by assuming the climate sensitivity to amount
to 2.0◦C (scenario var 1) and 3.5◦C (scenario var 2).
Figure 17 shows the sensitivity of the global emis-
sions on this assumption. Whereas a low climate sen-
sitivity allows emissions to stay above the current level
until 2050, a high climate sensitivity demands for a more
drastic reduction of emissions than the default policy
scenario A (cf. Fig. 14). Within the current setting,
REMIND-R is not able to find a feasible solution with
a climate sensitivity of 3.6◦C and higher. Achieving a
feasible solution in some of these cases, which requests
for an even faster reduction of CO2emissions than in
2020 2040 2060 2080 2100
0
2
4
6
8
10
12
Year
Global Emissions [GtC]
polA
var1
var2
Fig. 17 Global emissions for policy scenario A with different
climate sensitivities
68 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 171
0 1 2 3 4
var 2
var 1
pol A
mitigation costs (% of BAU consumption)
Fig. 18 Average mitigation costs for policy scenario A with
different climate sensitivities
scenario var 1, will be possible if we allow for idle
capacities and negative emissions.
The drastic emission reduction in the 3.5◦C cli-
mate sensitivity scenario results in an increasing share
of solar energy and a complete fade out of coal
technologies—even with the CCS option—due to
the remaining emissions. Most significantly, the total
amount of primary energy consumption is reduced—
on a global level from around 1,250 EJ in 2100 in the
default policy scenario A to around 950 EJ in the high
climate sensitivity scenario.
The sensitivity of mitigation costs is shown in Fig. 18.
Mitigation costs more than halve for the low climate
sensitivity scenario and more than double for the high
climate sensitivity scenario. Both changes indicate a
dominant impact of this parameter. The carbon price
increases by an order of magnitude in the high climate
sensitivity scenario. This would extremely benefit the
permit seller and requests for an additional compensa-
tion scheme.
5 Conclusions
This study analyzes climate policy implications in the
context of globalization by means of the energy–
economy–climate model REMIND-R. In determining
regional mitigation costs and the technological devel-
opment in the energy sector, REMIND-R considers the
feedbacks of investment and trade decisions of regions
that are linked by global markets. The analyzed policy
regimes are primarily differentiated by their allocation
of emission rights. Moreover, they represent alternative
designs of an international cap and trade system that is
geared to meet the 2◦C climate target. The following
conclusions can be drawn:
•Ambitious climate targets that meet the 2◦C cli-
mate target with high likelihood can be reached
with costs amounting to approx. 1.5% of the global
gross product; this roughly confirms cost estimates
of low stabilization scenarios from earlier studies
based on global models [7]. This number, however,
can halve or double within quite a narrow range of
climate sensitivity variation.
•The regional burden of emission reductions consid-
erably varies with the particular designs of a post-
2012 climate policy regime; however, the variance
of mitigation costs between the regions is higher
than between the policy regimes.
•Regions with high shares in trade of fossil resources
(MEA and Russia) bear the highest costs, while
Africa can considerably benefit from an integration
into a global emissions trading system.
The present study analyzes ambitious climate pro-
tection scenarios that require drastic reduction policies
(reductions of 70–80% globally until 2050). Immediate
and multilateral action is needed in such scenarios.
Given the rather small variance of mitigation costs in
major regions like UCA, Europe, MEA, and China,
a policy regime should be chosen that provides high
incentives to join an international agreement for the
remaining regions. From this perspective, either the
contraction and convergence scenario (incentive for
Russia) or the multi-stage approach (incentive for
Africa and India) is preferable.
As usual, all results are only valid within the
framework of the assumptions made. In the current
context, we in particular assume perfect markets and
perfect intertemporal foresight. Both slightly tend to
decrease the mitigation costs by optimally investing in
most promising long-term mitigation measures based
on optimal trade flows. However, for the regions with
high shares in resource trade, mitigation costs could
be overestimated by the model due to the fact that the
reference scenario accounts for too optimistic trade
volumes. Trade losses in the fossil-constrained policy
scenario rise consequently. Additional experiments
furthermore show that, with the assumption of lower
fossil resource availability, mitigation costs decrease
significantly.
From the analysis of the technology development in
the energy sector, it turns out that the regions follow
quite different strategies. However, while the mitiga-
tion cost estimates are robust against variations of input
3.5 Conclusions 69
172 M. Leimbach et al.
parameters, the regional energy mix is sensitive. More
research is needed to integrate further technologies
(e.g., electric vehicles in the transport sector) and to
systematically investigate to which degree and which
costs major carbon-free technologies can be substituted
by each other. First experiments in this direction indi-
cate that doing without nuclear energy is not costly, but
forgoing the CCS option will increase the mitigation
costs substantially.
Acknowledgements For their help in implementing the
REMIND model, we are grateful to our colleagues Michael
Lüken and Markus Haller. We also thank the German Federal
Environmental Agency for the financial support of this research.
References
1. Alcamo, J., Leemans, R., & Kreileman, G. J. J. (1998). Global
change scenarios of the 21st century. Results from the IMAGE
2.1 model. London: Pergamon & Elsevier Science.
2. Barker, T., Köhler, J., Anderson, D., & Pan, H. (2006). Com-
bining energy technology dynamics and macroeconometrics:
The E3MG model for climate stabilization scenarios. The
Energy Journal, 113–133 (Special Issue 2006).
3. Bauer, N., Edenhofer, O., & Kypreos, S. (2008). Linking
energy system and macroeconomic growth models. Journal
of Computational Management Science, 5, 95–117.
4. Bosetti, V., Carraro, C., Galeotti, M., Massetti, E., & Tavoni,
M. (2006). WITCH: A world induced technical change hybrid
model. The Energy Journal, 13–38 (Special Issue 2006).
5. Crassous, R., Hourcade, J.-C., & Sassi, O. (2006). Endoge-
nous structural change and climate targets. Modeling ex-
periments with Imaclim-R. The Energy Journal, 259–276
(Special Issue 2006).
6. 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,
277–292.
7. Edenhofer, O., Lessmann, K., Kemfert, C., Grubb, M., &
Köhler, J. (2006). Induced technological change: Exploring
its implications for the economics of athmospheric stabiliza-
tion: Synthesis report from innovation modeling comparison
project. The Energy Journal, 57–107 (Special Issue 2006).
8. Edmonds, J., Wise, M., Pitcher, H., Richels, R., Wigley, T., &
MacCracken, C. (1997). An integrated assessment of climate
change and the accelerated introduction of advanced energy
technologies. Mitigation and Adaptation Strategies for Global
Change, 1, 311–339.
9. den Elzen, M. G. J, & Lucas, P. L. (2005). The FAIR model:
A tool to analyse environmental and costs implication of
regimes of future commitments. Environmental Modeling
and Assessment, 10, 115–134.
10. den Elzen, M., Lucas, P.-A., & van Vuuren, D. (2005). Abate-
ment costs of post-Kyoto climate regimes. Energy Policy, 33,
2138–2151.
11. Frei, C. W., Haldi, P., & Sarlos, G. (2003). Dynamic formu-
lation of a top-down and bottom-up merging energy policy
model. Energy Policy, 31, 1017–1031.
12. Gerlagh, R., & van der Zwaan, B. (2003). Gross world pro-
duct and consumption in a global warming model with
endogenous technological change. Resource and Energy Eco-
nomics, 25, 35–57.
13. Gerlagh, R., & van der Zwaan, B. (2004). Sensitivity analysis
of timing and costs of greenhouse gas emissions reduction.
Climatic Change, 65, 39–71.
14. Guest, R. S., & McDonald, I. M. (2007). Global GDP shares
in the 21st century—An equilibrium approach. Economic
Modelling, 24, 859–877.
15. Hamelinck, C. (2004). Outlook for advanced biofuels. Ph.D.
thesis, University of Utrecht, The Netherlands.
16. Höhne, N., Galleguillos, C., Blok, K., Harnisch, J., &
Phylipsen, D. (2003). Evolution of commitments under
the UNFCCC: Involving newly industrial economies and
developing countries. Research Report, UBA-FB 000412,
Umweltbundesamt.
17. Hourcade, J.-C., Jaccard, M., Bataille, C., & Ghersi, F. (2006).
Hybrid modeling: New answers to old challenges. The Energy
Journal, 1–12 (Special Issue 2006).
18. Kahouli-Brahmi, S. (2008). Technological learning in energy-
environment-economy modelling: A survey. Energy Policy,
36, 138–162.
19. Kainuma, M., Matsuoka, Y., & Morita, T. (2003). Climate
policy assessment: Asia-Pacific integrated modeling. Tokyo:
Springer.
20. Kouvaritakis, N., Soria, A., & Isoard, S. (2000). Endoge-
nous learning in world post-Kyoto scenarios: Applications of
the POLES model under adaptive expectations. International
Journal of Global Energy Issues, 14, 222–248.
21. Kriegler, E., & Bruckner, T. (2004). Sensitivity of emissions
corridors for the 21st century. Climatic Change, 66, 345–387.
22. Kypreos, S., & Bahn, O. (2003). A MERGE model with
endogenous technological progress. Environmental Modeling
and Assessment, 19, 333–358.
23. Leimbach, M., & Toth, F. (2003). Economic development
and emission control over the long term: The ICLIPS
aggregated economic model. Climatic Change, 56, 139–
165.
24. Löschel, A. (2002). Technnological change in economic
models of environmental policy: A survey. Ecological Eco-
nomics, 43, 105–126.
25. McKibbin, W. J., & Wilcoxen, P. J. (1999). The theoretical
and empirical structure of the G-cubed model. Economic
Modelling, 16, 123–148.
26. Manne, A. S., Mendelsohn, M., & Richels, R. (1995).
MERGE—A model for evaluating regional and global effects
of GHG reduction policies. Energy Policy, 23, 17–34.
27. Manne, A. S., & Rutherford, T. (1994). International trade,
capital flows and sectoral analysis: Formulation and solution
of intertemporal equilibrium models. In W. W. Cooper, &
A. B. Whinston (Eds.), New directions in computational eco-
nomics (pp. 191–205). Dordrecht: Kluwer.
28. Mas-Colell, A., Whinston, M. D., & Green, J. R. (1995). Mi-
croeconomic theory. New York: Oxford University Press.
29. Messner, S., & Schrattenholzer, L. (2000). MESSAGE-
MACRO: Linking an energy supply model with a macro-
economic model and solving it interactively. Energy, 25,
267–282.
30. Nordhaus, W. D., & Yang, Z. (1996). A regional dy-
namic general-equilibrium model of alternative climate-
change strategies. American Economic Review, 86, 741–765.
31. Petschel-Held, G., Schellnhuber, H. J., Bruckner, T.,
Toth, F. L., & Hasselmann, K. (1999). The tolerable windows
approach: Theoretical and methodological foundations. Cli-
matic Change, 41, 303–331.
32. Penn World Tables (2007). Version 6.2.http://pwt.econ.
upenn.edu/php-site/pwt-index.php.
33. Riahi, K., Grübler, A., & Nakicenovic, N. (2007). Scenarios of
long-term socio-economic and environmental development
70 Chapter 3 Mitigation costs in a globalized world
Climate policy analysis with REMIND-R 173
under climate stabilization. Technological Forecasting and
Social Change, 74, 887–935.
34. Rogner, H. H. (1997). An assessment of world hydrocarbon
resources. Annual Review of Energy and Environment, 22,
217–262.
35. Rose, A., & Stevens, B. (1993). The efficiency and equity of
marketable permits for CO2emissions. Resource and Energy
Economics, 15, 117–146.
36. Sanstad, A. H., & Greening, L. A. (1998). Economic models
for climate policy analysis: A critical discussion. Environmen-
tal Modeling and Assessment, 3, 3–18.
37. Takeshita, T., & Yamaij, K. (2008). Important roles of
Fischer-Tropsch synfuels in the global energy future. Energy
Policy, 36, 2773–2784.
38. Vaillancourt, K., Loulou, R., & Kanudia, A. (2008). The
role of abatement costs in GHG permit allocations: A global
stabilization scenario analysis. Environmental Modeling and
Assessment, 13, 169–179.
39. World Bank (2005). World development indicators.
Washington, DC: World Bank.
40. Weyant, J. P. (2004). Introduction and overview. Energy Eco-
nomics, 26, 501–515.
3.6 References 71
72 Chapter 3 Mitigation costs in a globalized world
Chapter 4
The REMIND-R Model:
The Role of Renewables in the Low-Carbon
Transformation ∗
Nico Bauer
Lavinia Baumstark
Marian Leimbach
∗revised to Bauer, N., L. Baumstark, M. Leimbach (2011):The REMIND-R Model: The Role of Re-
newables in the Low-Carbon Transformation. Climatic Change Special Issue.
73
74 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
DOI 10.1007/s10584-011-0129-2
The REMIND-R model: the role of renewables
in the low-carbon transformation—first-best
vs. second-best worlds
Nico Bauer·Lavinia Baumstark·Marian Leimbach
Received: 19 May 2010 / Accepted: 23 May 2011
© Springer Science+Business Media B.V. 2011
Abstract Can near-term public support of renewable energy technologies contain
the increase of mitigation costs due to delays of implementing emission caps at the
global level? To answer this question we design a set of first and second best scenarios
to analyze the impact of early deployment of renewable energy technologies on
welfare and emission timing to achieve atmospheric carbon stabilization by 2100.
We use the global multiregional energy–economy–climate hybrid model REMIND-
R as a tool for this analysis. An important design feature of the policy scenarios is
the timing of climate policy. Immediate climate policy contains the mitigation costs
at less than 1% even if the CO2concentration target is 410 ppm by 2100. Delayed
climate policy increases the costs significantly because the absence of a strong carbon
price signal continues the carbon intensive growth path. The additional costs can be
decreased by early technology policies supporting renewable energy technologies be-
cause emissions grow less, alternative energy technologies are increased in capacity
and their costs are reduced through learning by doing. The effects of early technology
policy are different in scenarios with immediate carbon pricing. In the case of delayed
climate policy, the emission path can be brought closer to the first-best solution,
whereas in the case of immediate climate policy additional technology policy would
lead to deviations from the optimal emission path. Hence, technology policy in the
delayed climate policy case reduces costs, but in the case of immediate climate policy
they increase. However, the near-term emission reductions are smaller in the case
of delayed climate policies. At the regional level the effects on mitigation costs are
heterogeneously distributed. For the USA and Europe early technology policy has a
The authors like to thank Enrica DeCian and Valentina Bosseti for valuable comments
on an earlier draft. The authors acknowledge financial support from the German
BMBF for the ALICE project. All remaining errors are ours.
N. Bauer (B)·L. Baumstark ·M. Leimbach
Potsdam Institute for Climate Impact Research, P.O. Box 60 12 03, 14412 Potsdam, Germany
e-mail: [email protected]
URL: http://www.pik-potsdam.de
75
Climatic Change
positive welfare effect for immediate and delayed climate policies. In contrast, India
looses in both cases. China loses in the case of immediate climate policy, but profits in
the delayed case. Early support of renewable energy technologies devalues the stock
of emission allowances, and this effect is considerable for delayed climate policies.
In combination with the initial allocation rule of contraction and convergence a
relatively well-endowed country like India loses and potential importers like the EU
gain from early renewable deployment.
1 Introduction
The transformation of the global energy system towards de-carbonization is iden-
tified as a key challenge for the 21st century. Following historical trends of de-
carbonization is not sufficient to meet stringent climate change mitigation targets
as well as major objectives related to environmental protection and economic devel-
opment, see e.g. Nakicenovic and Riahi (2002). Renewable energy technologies have
been identified as an essential option for the transformation of the energy system to
meet climate change mitigation. The main driver triggering the recent deployment of
renewable energy technologies has not been carbon pricing but dedicated support
for renewable energy technologies. May convert the logical relationship into the
opposite direction: high renewable deployment contains the costs for achieving
stringent climate policies and, hence, increases the social acceptability and economic
affordability of such long-term goals. The present paper aims at gaining insight on
the role of renewable energy technologies in the transformation of the global energy
system and on how they interact with global and regional mitigation costs, if climate
policy is delayed.
The issues of climate change mitigation and renewable energy technology (RET)
deployment are high on the political agenda. An international agreement on binding
caps on greenhouse gas (GHG) emissions is yet not implemented and it is expected
that it will take some years before such agreement will enter into force at the global
level. The fifteenth Conference of the Parties (COP) 2009 to the UN Framework
Convention on Climate Change (UNFCCC) in Copenhagen was expected to make
a great step into this direction, but it failed to meet these expectations. However,
the Copenhagen Accord called for long-term co-operative action, recognizing the
scientific view that the increase of global mean temperature should not exceed 2◦C.
This is an important outcome in making Art. 2 of the UNFCCC more operational
that formulated the ultimate objective of stabilizing GHG concentrations at a level
that prevents dangerous interference with the climate system. Though the interna-
tional negotiation process arrived at a more concrete long-term target, there is no
internationally binding agreement on how to deal with emissions in the short-term.
Rogelj et al. (2010) reviewed the pledges to the Copenhagen Accord and concluded
that according to the pessimistic interpretation the emission cap in 2020 “is nearly
equal to the business-as-usual [emissions].”
The growth of CO2emissions during the last decade was the highest ever reported,
which was mainly triggered by the increasing use of coal. This trend is expected to
continue over the next decades, if no effective climate policies are implemented. The
global economic crisis of 2008 is not expected to make a huge difference; see e.g.
IEA-WEO (2010).
76 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
Fig. 1 Global generation
capacities of RETs in the
electricity sector 1996 to 2009.
Sources: GWEC (2010), IEA
WEO various issues, REN 21
various issues, Jäger-Waldau
(2009)
At the same time, RET deployment is pushed forward by many national govern-
ments, for a number of reasons, including climate change mitigation. Feed-in tariffs,
renewable energy quotas and other measures triggered a boom of RET, especially
in the electricity sector. The recent recession did not interrupt the development.
Green Recovery Programs—see e.g. Edenhofer et al. (2009)—are accelerating the
development. Jäger-Waldau (2009, p. 10) notes that the major part of the various
national fiscal stimuli for renewable energies have not yet been spent, but are going
to become effective in 2010 and 2011.
Figure 1reports the global generation capacities of electricity producing RETs in
log-scale. The global cumulative capacity for Wind has been increasing at an annual
rate between 21% and 37%; for solar PV the cumulative capacity has been growing
at 20% p.a. in the late nineties and by more than 40% p.a. in the period 2005 to 2008.
Jäger-Waldau (2009, p. 103) concludes that “the photovoltaic industry is developing
into a fully-fledged mass-producing industry.” Hydro, geothermal, and biomass were
only slowly increasing, but started at relatively high levels in the 1990s. It should
be noted that the capacity additions of hydro in 2008 are reported at 35GW, which
is still larger than the sum of wind and solar, amounting to 32.5GW. The capacity
additions of wind and solar are rapidly growing and, hence, are expected to exceed
the new installations of the more traditional renewable electricity technologies within
the next few years.
The recent boom of RET over the first decade of the 21st century outpaced all
earlier expectations.1Policies already in place and recent observations of market
developments confirm the expectation that RET deployment will keep on rapidly
growing in the near-term future, see e.g. GWEC (2010, p. 15), Jäger-Waldau (2009,
p. 17).
1For example, IEA-WEO (2002, p. 412) expected the wind power capacity in 2010 at 55GW. The
realized value in 2009, however, is 159GW.
4.1 Introduction 77
Climatic Change
In summary, the present situation is ambivalent. The need for limiting climate
change is accepted but no emission limitations are implemented at the global level,
and renewable energy sources as well as coal use grow at very high rates. From an
economic point of view these ambivalent developments raise two sets of research
questions:
1. Assume the implementation of global climate policy is delayed until 2020. Is
the near-term RET deployment reducing or increasing the mitigation costs of
delayed climate change mitigation policies? How is this related to the global CO2
emissions in 2020? What are the mitigation cost impacts for different regions?
2. For comparison, assume that RET deployment is varied and a global cap-and-
trade system for CO2emissions is implemented immediately. How do deviations
from the optimal RET deployment scenario change the time path and the
regional distribution of mitigation cost? What is the impact on global CO2
emissions?
To answer the first set of questions is the main objective of the present study. The
complex interplay of climate and technology policies leads to various effects. To
provide a clear understanding of the forces at work, the second set of questions is
elaborated.
The methodological approach addressing these questions is to design and to
analyze a set of first- and second-best scenarios using the energy–economy–climate
model REMIND-R. This is in line with the general philosophy of the RECIPE
project; see Luderer et al. (2011, this issue). Previous contributions to the economics
of climate change mitigation have intensively applied this methodology for designing
scenarios to assess mitigation costs. Manne and Richels (2004)discusstheimpact
of endogenous technology learning-by-doing on the optimal emission path and the
costs for achieving a given stabilization target. They find that technology learning
has little impact on the optimal emission path, but significantly reduces the costs
for achieving the stabilization target. The International Model Comparison Project
assessed the contribution of induced technological change to meeting atmospheric
stabilization of GHGs at the lowest possible costs; see Edenhofer et al. (2006). The
EU ADAM project mainly focused on the significance of having available specific
low-carbon technologies; see Edenhofer et al. (2010). The 22nd round of the Stanford
Energy Modeling Forum focused on the impact of delayed mitigation policies and the
significance of temporary over-shooting of the stabilization targets; see Clarke et al.
(2009).
We extend the methodology in two directions focusing especially on the tim-
ing of climate and technology policies. The first set of questions suggests the
comparison of two different second-best scenarios: delayed climate policy and
early RET deployment are combined. Up to our knowledge the comparison of
two second-best scenarios is an innovation to the methodology and extends the
debate on the economics of climate change mitigation and technology deploy-
ment. The resulting impact of early technology deployment on the mitigation
costs is considerable. For improving our understanding we formulate the second
set of scenarios which analyze the impact of weaker versus stronger technology
development in combination with immediate climate policy. This is also different
to common technology second-best scenarios, where the availability of technolo-
78 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
gies is limited below the optimal case. In the present study the deployment of
RET is constrained to deviate negatively as well as positively from the first-best
solution.
The analysis of second-best scenarios is related to the discussion about the optimal
timing and coordination of policies in the context of climate change mitigation; see
Sorrell and Sijm (2003). Böhringer et al. (2009) evaluate the simultaneous application
of a cap-and-trade system and renewable penetration targets in the EU using a
computable general equilibrium model. They find that the additional costs of the
technology policy are small and the CO2permit price is decreased. Kverndokk
and Rosendahl (2007) discuss the optimal choice of emission taxes and technology
subsidies for limiting cumulative emissions in the presence of spill-over effects due
technology learning-by-doing. The highly stylized partial model covers the electricity
sector of a single region. The study analyzes delays in choosing carbon taxes and
technology subsidies. The paper finds that delaying the optimal carbon emissions
tax has little impact on welfare. In addition, its effect is smaller than delaying the
optimal technology subsidy. Similar studies on first- and second-best policies with
explicit representation of policy instruments to address multiple and interlinked
externality problems have been undertaken on the coordination of technology R&D
and emission taxing policies; see e.g. Gerlagh et al. (2009). Also other market failures
could be considered, but like the issue of technology R&D this is not the focus of this
paper.
The present study applies a multi-regional model covering the total energy system,
the macro-economy and the climate system. It computes first- and second-best
scenarios to quantify the economic, technological, and environmental effects of
climate and technology policy. The first- and second-best scenarios are implemented
by imposing constraints on environmental variables (i.e. atmospheric CO2concentra-
tion) and on technology related control variables (i.e. investments). The approach to
design second-best scenarios by constraining investment variables is different to the
second-best policy analysis as has been studied by Kverndokk and Rosendahl (2007),
who imposed constraints on policy variables that aim at changing the investment
decisions of autonomous private agents.
The remainder of the paper is organized as follows. Section 2introduces the
REMIND-R model that is the numerical tool for assessing the research questions.
Section 3introduces the design of the scenarios and how they are related to the
research questions. Section 4presents and discusses the results. Section 5concludes
and points to promising fields for future research.
2 The REMIND-R model
The Refined Model of Investment and Technological Development (REMIND) is
used as the numerical tool to address the research questions raised above. The model
is documented in literature; see Bauer et al. (2009) and Leimbach et al. (2010). In the
following the general structure of the model and the role of renewable energy is
elaborated as it is important for the sake of this paper.
Figure 2provides a generic overview on the model structure. REMIND-R is an
inter-temporal, general equilibrium, multi-regional energy–economy–climate model.
4.2 The REMIND-R Model 79
Climatic Change
Welfare
LabourCapital
Energy system
costs
Output
ConsumptionInvestments
Final
energy
Energy transformations and
conversion technologies
Fuel costs Investment
costs
Operation and
Maintenance costs
Labour
efficiency
Emissions
Learning
by doing
Resource and
potential
constraints
Macroeconomic
module Energy system-
module
Hard link
Exogenous Data
Energy
efficiency
Trade
Trade
Climate module
Trade
Fig. 2 Overview of the REMIND-R model framework. Blue boxes on the left are related to the
macroeconomic growth model, yellow boxes on the right denote elements of the energy system
model. The red arrows highlight the hard-links between models. The light-colored arrows indicate
trade relationships
In each of the eleven world regions2a Ramsey-type growth model represents the
macro-economy. The energy sector model is embedded into the macro-economic
growth model. Both models interact by financial and energy trade flows. A social
optimum for each region is computed by maximizing the inter-temporal, social
welfare subject to economic and technological constraints as well as prices for
internationally traded goods. In each region a hard-link between the macro-economy
and the energy system guarantees simultaneous equilibrium on all markets for final
goods, capital, labor and energy; see Bauer et al. (2008). The Negishi-approach is
applied to compute the Pareto-equilibrium of trade between the regions; see Manne
and Rutherford (1994) and Leimbach and Toth (2003). International trade comprises
the generic macro-economic good, coal, oil, gas, uranium and emission permits.
Climate policy is imposed by setting a constraint on the atmospheric CO2concen-
tration. The resulting CO2emission path minimizes the mitigation costs for the world
economy by fully exploring ‘when’- and ‘where’-flexibility of mitigation measures
given the full tradability of emission permits and inter-temporal equilibrium of the
international capital market. The emission permits are distributed to the regions
according to an allocation rule, based on the contraction and convergence scheme.
2In the present study we have focus on the US (USA), Europe (EUR), China (CHN), and India
(IND). The other countries are summarized in the two aggregates Rest of Annex-1 (RAI), which
comprise those countries committed to reduce emissions under the Kyoto Protocol and Rest of Non-
Annex-1 (RNAI), which are all the other countries that have not agreed on reducing emissions.
80 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
The allocation to the regions follows a two step approach. First, the global emission
path is optimized. Second, each region receives a share of this aggregate according
to the allocation scheme.
The energy sector represents the conversion of energy carriers by various linear
production technologies. Each technology is described by a set of techno-economic
characteristics. Energy conversion requires the availability of capacities that are
extended by specific investments and decreased by technical depreciation. Table 1
gives an overview of all available technologies and conversion routes from primary
to secondary energy. Primary energy is distinguished in exhaustible and renewable
energy carriers. The former are subject to extraction costs that increase in cumulative
extraction. The tradable endowments are owned by the region where they are
located. The latter are non-tradable and subject to potentials that are differentiated
by grades that capture the decreasing quality of different locations. Secondary energy
carriers are delivered to the macro-economic sector that uses them in combination
with capital and labor to generate macro-economic output using a nested CES
production function.
Renewable energy carriers are an option to supply the rapidly growing demand
for electricity. The renewable energy technologies solar PV and wind turbines are
learning technologies with learning rates of 20% and 12%, respectively. Learning-
by-doing can reduce the investment costs of wind turbines from 1200$US per kW
to a minimum of 883$US per kW. The assumptions for solar PV are 4900$US per
kW and 600$US per kW, respectively. The fluctuating nature of both primary energy
sources is addressed by imposing constraints that imply the need of storage facilities
and excess capacities, both, depending on the technologies’ share in the generation
mix; see Pietzcker et al. (2009) based on techno-economic assumptions from Chen
et al. (2009).
Biomass is notable for the supply of other secondary energy carriers. Currently
biomass is mainly used in solid form for satisfying basic needs using traditional
biomass technologies. With growing income and growing demand for modern energy
carriers traditional biomass utilization fades out of the system. The supply for
modern ligno-cellulosic biomass utilization is growing to a maximum of 200EJ p.a.
in 2050; see van Vuuren et al. (2009) for an assessment of this assumption. Biomass
can also be utilized in combination with carbon capture and sequestration (CCS)
for the production of electricity, hydrogen and transportation fuels. Hydrogen could
be produced from low-carbon electricity sources like renewables or nuclear vie
electrolysis (not shown in Table 1). More direct conversion routes for hydrogen
production from renewables and nuclear are discussed in Magné et al. (2010).
3 Scenarios
The development of the scenario set-up in this study follows a five step approach.
Table 2summarizes the scenario assumptions that are elaborated in more detailed
next. The first step is to compute a case without climate policy constraints, denoted
as the BASELINE scenario.
The second step is to compute three first-best climate change mitigation scenarios.
The scenarios are constrained to stabilize CO2concentrations at 410, 450, and
490 ppm by 2100 with temporary over-shoot. In each scenario the CO2concentration
4.3 Scenarios 81
Climatic Change
Table 1 Overview of primary energy carriers, secondary energy carriers and the technologies for conversionQ2
Secondary energy carriers Primary energy carriers
Exhaustible Renewable
Coal Oil Gas Uranium Solar, wind, hydro Geo-thermal Biomass
Electricity PCa,IGCC
a, CoalCHP DOT GT, NGCCa, GasCHP TNR, FNR SPV, WT, Hydro HDR BioCHP, BIGCCa
H2 C2H2 SMRaB2H2a
Gases C2G GasTR B2G
Heat CoalHP, CoalCHP GasHP, GasCHP GeoHP BioHP, BioCHP
Liquid fuels C2LaRefin. B2La, BioEthanol
Other liquids Refin.
Solids CoalTR BioTR
PC conventional coal power plant, IGCC integrated coal gasification combined cycle, CoalCHP coal combined heat power, C2H2coal to H2, C2Gcoal to gas,
CoalHP coal heating plant, C2Lcoal to liquids, CoalTR coal transformation, DOT diesel oil turbine, Refin. Refinery, GT gas turbine, NGCC natural gas combined
cycle, GasCHP Gas combined heat power, SMR steam methane reforming, GasTR gas transformation, GasHP gas heating plant, TNR thermal nuclear reactor,
FNR Fast nuclear reactor, SPV solar photovoltaic, WT wind turbine, Hydro hydro power, HDR hot-dry-rock, GeoHP heating pump, BioCHP biomass combined
heat and power, BIGCC Biomass IGCC, B2H2biomass to H2, B2Gbiogas, BioHP biomass heating plant, B2Lbiomass to liquids, BioEthanol biomass to ethanol,
BioTR biomass transformation
aThese technologies are also available with carbon capture
82 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
Table 2 Overview of scenarios
Climate policy (CO2only) RET deployment Comment
Baseline None Optimal
POLxStarting in 2010 ×indicates Optimal First-best scenarios
the stabilization at 410, 450,
490 ppm by 2100
POLDEL Delayed (DEL) until 2020 Optimal subject Second-best:
450 ppm by 2100 to delayed Delayed climate
climate policy policy
POLD&RDelayed (D) until 2020 RET deployment is Second-best: Delayed
(s/m/w) 450 ppm by 2100 strong (s), medium (m) climate policy and
or weak (w) Until 2020 exogenous RET
deployment
POLRET Starting in 2010 450 ppm RET support strong (s) Second-best:
(s/w,20/30) by 2100 or weak (w) Until 2020 Immediate climate
or 2030 policy with exogenous
RET deployment
is allowed to exceed the target by a maximum of 4.5%.3Non-energy CO2emissions
are assumed to follow an exogenous path; see Luderer et al. (2011, this issue).
The emission permits are consistent with the optimal emission path and distributed
among regions according to the convergence and contraction scheme achieving equal
per-capita allocation in 2050; see Den Elzen et al. (2008) and Leimbach et al. (2010).
The scenarios are denoted POLx, where x indicates the stabilization target.4
In the third step climate policy is delayed. The climate policy starts in 2020.
Until the beginning of climate policy all investment paths are constrained to the
BASELINE scenario. This scenario is denoted POLDEL.5
In the fourth step the delayed climate policy is combined with early RET deploy-
ment constraints. For this purpose in all regions the RET deployments of the three
POLxscenarios until 2050 are given as exogenous constraints in a scenario without
any climate policy. The resulting three scenarios provide different development paths
for the entire energy–economy system. These three developments until 2020 are used
as exogenous constraints for all stock variables, i.e. until the period, when climate
policy becomes active. Hence, the delayed climate policy is combined with weak,
medium and strong scenarios for early deployment renewables. These scenarios
are denoted POLD&R. The three deployment scenarios (weak, medium, strong) are
indicated in parentheses. The three scenarios are compared with the POL450 and the
POLDEL scenarios to address the first set of questions raised above.
In the fifth step immediate climate policy is combined with four non-optimal RET
deployment paths to achieve the 450 ppm stabilization level; i.e. the capacity values
3This degree of overshooting is the same as assumed in A1_CC_Overshoot in Luderer et al. (2011,
this issue). To maintain consistency across scenarios the same overshooting is allowed for the
scenarios POL410 and POL490.
4The scenario POL450 is the same as the scenario A1_CC_Overshoot in Luderer et al. (2011,
this issue). The scenario POL410, however, is stricter than the scenario A1_CC_410 because the
permissible overshoot is smaller.
5This scenario is the same as the scenario C8_DELAY2020 in Luderer et al. (2011, this issue).
4.4 Results 83
Climatic Change
of all technologies as given in the renewable columns of Table 1. For this purpose
s(trong) and w(eak) RET deployments are considered as exogenous scenarios. The
assumptions for the exogenous RET deployment pathways are taken from POL410
for the scenario strong scenario and from POL490 for the weak scenario. Further-
more, the deployment assumptions are fixed for two time horizons until (20)20
and (20)30. This set of second-best scenarios is denoted POLRET. A specification
in parentheses distinguishes the intensity and the duration of the exogenous RET
deployment assumptions. The benchmark for comparison is the first-best POL450
scenario. The analysis addresses the second set of questions raised above.
The design of the POLD&Rscenarios is based on model outcomes for the period
2010 to 2020 because this approach provides different levels of RET deployment.
The alternative would be to base the assumptions on deployment targets announced
by governments. It is, however, not the aim of the present study to assess the
various renewable policy targets, but to assess the significance of a range of RET
deployment assumptions. Deriving the assumptions from scenarios computed by the
same optimization model implies that the deployment paths are implicitly tailor-
made for the model. The technology policy assumptions are, thus, not subject to
the critique of failing to ‘picking the winner’, which means that the portfolio of the
technology policy is not a mis-allocation, but suits the needs of the energy sector
represented in the model.
4Results
4.1 First best solutions—BASELINE and POLx
The BASELINE and the POLxscenarios are the starting point for the analysis.
The POL450 scenario serves as a point of reference for the second-best scenarios
introduced in the following sub-sections. Furthermore, the POLxscenarios provide
RET deployment paths that are used as constraints for second-best scenarios below.
Figure 3shows the global CO2emissions from the energy sector for the four
scenarios. BASELINE emissions increase to 21GtC p.a. in 2050 and remain ap-
proximately at this level. The rapid increase of CO2emissions is mainly triggered
Fig. 3 Global CO2emissions
2005 to 2100 from the energy
sector for the scenarios
BASELINE and POLx
84 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
0
2000
4000
6000
8000
10000
12000
14000
16000
2020 2030 2050
Electricity Generation [TWh p.a.]
POL490
POL450
POL410
MTK10 Reference
MTK10 450ppm(eq)
WBGU-2˚C
Greenpeace Reference
Greenpeace Energy [R]evolution
IEA-WEO Reference
IEA-WEO 450ppm
EIA-IEO
IEA-ETP Baseline
IEA-ETP BLUE Map 450ppm
WETO-H2 Ref
WETO-H2 450ppm
3750019444
Fig. 4 Electricity generation from wind power turbines in the three POLxscenarios computed with
REMIND-R and results from other publications. Sources: WBGU (2003, p. 138), Magne et al. (2010,
p. 93), Greenpeace (2008, p. 190 and 191), IEA (2009, p. 229 and 623), IEA (2008 p. 85), EIA (2009,
p. 67), WETO is EC (2006, p. 120 and 129). IEA-WEO and EIA-IEO consider only a time horizon
until 2030; IEA-ETP reported numbers only for the year 2050
by the supply of huge amounts of cheap coal. The three POLxscenarios cover a
wide range of different emission pathways. To achieve the 450 ppm target, CO2
emissions deviate from the BASELINE immediately, peak in 2020 below 10GtC
p.a. and decrease to 3.6GtC p.a. in 2070 to stay at this level afterwards. The less
ambitious 490 ppm target allows the emissions to peak in 2025 at a much higher level
(12GtC p.a.). The emissions start to deviate from the BASELINE path after 2015. In
the longer run emissions stabilize at 5GtC p.a. To achieve the much more ambitious
410 ppm target emissions need to decrease immediately and stabilize at 2GtC p.a. in
2070.
The deployment of RET in the three POLxscenarios also varies over a large
range. In the following, the deployment of wind and solar for electricity production
are analyzed in depth and compared with scenarios from the literature. The choice
for these two technologies is due to the significant deployment computed with
REMIND-R and the rapid growth experienced in recent past.6Finally, remarks on
other RET that are subject to the technology deployment policies will be added.
Figure 4shows the electricity generation from wind power turbines in the years
2020, 2030 and 2050. The three POLxscenarios computed for this study show
significant sensitivity regarding timing of deployment of wind power turbines. The
largest relative differences are observed in 2020, but in 2050 all three paths converge
6It is worth to note that only few studies provide results on differentiated global renewable electricity
generation. It is common practice to report figures that contain an aggregate on “Other Renewables”,
which does not give insight into the contribution of wind, solar, etc.
4.4 Results 85
Climatic Change
to quite similar levels. For the scenario POL410 the growth is most significant in the
coming decades, whereas for the POL490 scenario growth accelerates significantly
after 2030.
The future wind electricity production of the three scenarios can be compared with
simple extrapolation of the historical time series given in Fig. 1above. Assuming a
constant growth rate of installed capacity until 2020 would result in a capacity of
3000GW. Assuming an annual average of 2000 full load hours, which is a relatively
low number, would result in an output of electricity of 6000TWh p.a. Comparing this
number with the results in Fig. 4indicates that for the scenario POL450 future growth
rates could even decrease from their historical levels. For the case POL410 the growth
rate of installed wind turbine capacity might need to increase above historical rates.
For the comparison with other studies it is useful to distinguish two groups. The
first group comprises modeling studies that are similar to the present REMIND-
R study. They apply inter-temporal energy-economy models with optimization
under perfect foresight. Magne et al. (2010) use the model MERGE-ETL (denoted
MTK10) and WBGU (2003) uses the model MESSAGE (denoted WBGU-2◦C).
Both studies show—like the REMIND-R scenarios—high wind power generation.
MTK10 provides two scenarios: the reference case does not consider any climate
policy and the 450 ppm scenario7considers all GHGs and, thus, is more stringent
than the POL450 scenario of the present study. In the near-term wind power genera-
tion is higher in the 450 ppm scenario than in the reference scenario, but in 2050 the
ranking is reversed. This can be explained with the absorptive capacity for fluctuating
electricity sources that decreases in the policy scenario because the total electricity
generation decreases. The WBGU scenario is supposed to limit the increase of global
mean temperature to not more than 2◦C above pre-industrial levels. Wind power
generation is the highest of all scenarios.8
The second group of scenarios is mainly motivated by energy issues and the
underlying models are not using inter-temporal optimization with perfect foresight.
The scenarios show much lower deployment of wind power. Even the Greenpeace
Energy [R]evolution scenario is much lower in 2050 than the scenarios belonging
to the first group.9The lowest scenario is provided by the US Energy Information
Administration (EIA) that does, however, not consider a climate stabilization target.
The smallest difference between the reference case and the climate policy case is
provided by the WETO-H2 study.
Figure 5shows the corresponding graph for electricity production from solar
sources in log-scale. For all scenarios solar electricity generation in 2030 is less
than for wind. In 2050 this ranking is reversed in some of the scenarios. Solar
electricity production increases by two orders of magnitude—and even more for
few scenarios—within three decades. Regarding the timing of deployment the three
7The scenario allows for over-shooting the concentration target.
8It should be noted that this scenario is subject to limited possibilities for using alternative tech-
nologies in the power sector. Nakicenovic and Riahi (2002) provide an in-depth analysis of the
MESSAGE model. For a broad range of scenarios wind power generation is significantly lower than
in the present scenario. However, the figures for electricity production are not reported.
9It should be noted that the Energy [R]evolution scenario assumes considerably higher increases of
energy efficiency than the other scenarios, hence, electricity demand is lower. Therefore, the share
of wind power in the overall generation mix is higher than for the other scenarios.
86 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
1
10
100
1000
10000
100000
2020 2030 2050
Electricity Generation [TWh p.a.]
POL490
POL450
POL410
MTK10 450ppm(eq)
MTK10 400ppm(eq)
WBGU-2˚C
Greenpeace Reference
Greenpeace Energy [R]evolution
IEA-WEO Reference
IEA-WEO 450ppm
IEA-ETP Baseline
IEA-ETP BLUE Map 450ppm
WETO-H2 Ref
WETO-H2 450ppm
Fig. 5 Electricity generation from solar technologies in the three POLxscenarios computed with
REMIND-R and results from other publications. Note the log-scale. Sources: WBGU (2003, p. 138),
Magne et al. (2010, p. 93), Greenpeace (2008, pp. 190 and 191), IEA (2009, pp. 229 and 623), IEA
(2008 p. 85 and 367), EIA (2009, p. 67), WETO is EC (2006, p. 120 and 129). IEA-WEO and EIA-
IEO consider only a time horizon until 2030; IEA-ETP do not report numbers only for the year
2020
POLxscenarios show a similar behavior as in the case for wind power production.
MTK10 shows lower deployment for solar power. Also the same sensitivity as for
the REMIND-R model can be observed regarding timing of deployment as the
stabilization target is tightened from 450 ppm(eq) to 400 ppm(eq). The WBGU-
2◦C scenario again shows much higher figures than REMIND-R. The scenarios of
the second group exhibit relatively small differences compared with the first group
in 2030. For example, the POL450 scenario shows nearly the same solar electricity
production as the WEO 450 ppm scenario and the ETP BLUE map scenario. Only
in the longer term in 2050 the POLxscenarios are much higher than the Greenpeace,
the ETP and the WETO-H2scenarios.
The four scenarios imply different investment costs in the year 2020 as they
depend on technology deployment according to the dynamics of learning by doing.
Table 3presents the investment costs in 2020 for the wind and solar PV technology.
The lower investment costs for the stricter climate policy targets are a direct
consequence of the social optimal deployment of technologies shown above. This
result is important for the analysis below.
Table 3 Investment costs of wind and solar PV technologies in 2020 for the four scenarios BAU and
POLx
Technology Unit Scenario
BAU POL490 POL450 POL410
Wind $US/kW 1144 960 917 901
Solar PV $US/kW 4900 3823 2548 1543
4.4 Results 87
Climatic Change
Fig. 6 Impact of delayed
climate policy and different
assumptions of near-term RET
deployment on annual global
CO2emissions from the
energy sector from 2005 to
2050. The lines indicate
absolute differences of the
POLDEL and the three
POLD&Rscenarios with
respect to the POL450 scenario
Biomass is the most important other RET that is used in the POLxscenarios. In all
three scenarios the maximum potential of 200EJ p.a. is utilized in 2050. It is mainly
converted into synthetic natural gas and transportation fuels—the latter partly with
CCS. In the scenario POL450 biomass with CCS is deployed from 2030 on.
Finally, the mitigation costs of the three POLxcases are assessed. Mitigation costs
are measured as the cumulative discounted consumption losses for the 21st century
relative to the BASELINE case. For discounting we used the interest rate that is
computed endogenously in the REMIND-R model.10 For the medium case POL450
the losses are 0.5%. For the less ambitious POL490 the losses are only 0.3%, but
increase to 0.8% as the stabilization target is tightened in the POL410 case.
4.2 Delayed climate policy and early RET deployment—POLDEL and POLD&R
The assumption of an immediately established global cap-and-trade system is flawed
because the international negotiations take more time. The POLDEL scenario analy-
ses delayed climate policies by freezing the development of all stock variables
until 2020 on BASELINE levels. Measures to support renewable energies are not
reflected in this near term development. This design feature is added in the POLD&R
scenarios, which combine delayed climate policies and early RET deployment. The
design of POLD&Rscenarios captures the current situation with large renewable
deployment initiatives but missing global climate policy.
Figure 6shows the impact on global CO2emissions from the energy sector.
The graph depicts the absolute differences compared with the POL450 case (see
Fig. 3above) until 2050. In the year 2020 the emissions in the POLDEL case are
3.5GtC p.a. higher. This difference can be significantly reduced by imposing the RET
10The interest rate in REMIND is time-variable. The steady state value is 5.5% that is approached
starting from 8%. It is common practice to apply a constant discount rate to compute net present
values of consumption and GDP differences. In the present case the differences between the policy
scenarios and the BAU scenario are non-trivial because of multiple intersections. This implies that a
fixed discount rate can lead to counter-intuitive rankings of the scenarios; i.e. a second best scenario
performs better than a first best scenario. If we apply the endogenous time variable interest rate, the
rankings of scenarios are sound.
88 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
deployment. If the deployment from the original POL450 scenario is assumed until
2020, the difference in emissions reduces to 2.5GtC p.a. During the following decades
the differences change sign because the 450 ppm concentration target has to be met in
all scenarios. In the scenario POLDEL the rapid and enormous decreases of emissions
after 2020 are mainly achieved by abandoning new fossil investments and heavily
investing into biomass with CCS. Higher deployments of RET reduce the emission
level in 2020 and therefore do not require such massive changes after 2020 because
the peak is lower and, hence, this reduces the need to go quickly below the optimal
emission path of the POL450. These smoothed changes reduce the mitigation costs.
The mitigation costs for the different scenarios are of particular interest because
a-priori it is unclear whether the costs of delayed climate policy increase or decrease,
if early RET deployment is additionally assumed. It is valid to say that the POL450
scenario implies the lowest mitigation costs of the five scenarios considered here.
The costs of the POLDEL scenario are expected to be higher, because the energy–
economy system can not choose the cost-minimal timing of mitigation measures. The
key question is, whether political measures for early deployment of renewables are
justified as long as the global climate cap-and-trade system has not yet entered into
forcetoachieveacommonCO
2concentration target.
Figure 7shows the global and regional mitigation costs measured as the cu-
mulative discounted consumption differences relative to the BASELINE scenario.
At the global level it turns out that early RET deployment reduces the additional
mitigation costs of delayed climate policy. The additional costs of the POLDEL case
can be reduced by 0.23%-points, if RET deployment that originally was optimal in
the POL450 scenario is triggered by support measures; i.e. the scenario POLD&R(m).
However, the cost reducing effect is bound by the mitigation costs of the POL450 case.
The distributional effects are heterogeneous among regions. For the US and the
EU the cost reducing effect of early RET deployment is in the same direction as
at the global level. The same is also true for China, if only the medium and strong
deployment scenarios are considered. For India and the RNAI region the effect is of
the opposite sign.
The net effect of the mitigation costs for the different regions can be explained
by analyzing the different components. The differences of mitigation costs are
Fig. 7 Global and regional
mitigation costs for the
POL450,thePOL
D&Rand the
POLDEL scenarios. Mitigation
costs are measured as the
cumulative discounted
consumption losses relative to
the BASELINE case for the
time horizon 2005 to 2100. The
indices in the parenthesis
indicate s(trong), m(edium)
and w(eak) of RET
deployment. The time varying
endogenously computed
interest rate is used as the
discounting rate
4.5 Discussion and Further Research 89
Climatic Change
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
3
USA EUR CHN IND USA EUR CHN IND
POLD&R(m) POLD&R(s)
Difference [%-points]
Permit trade (net trade revenue)
Fuel (domestic + net trade revenue)
Energy system (Invest.+O&M)
Macroeconomic (GDP-Invest)
Fig. 8 Impact of various effects on the mitigation costs of the US, Europe, China and India for the
POLD&R(m) and POLD&R(s) scenarios compared with the POLDEL scenario. Negative (positive)
values indicate mitigation cost reductions compared with the POLDEL scenario. Note: the relative
differences are relative to changes of GDP, whereas mitigation costs are relative to consumption
differences
explained by changes in (a) macroeconomic variables of GDP and investments in
the macroeconomic capital stock (b) non-fuel energy system costs consisting of
investment and O&M costs, (c) fuel costs including net export revenues and (d)
permit trade. The methodology for deriving the single components from the results
of the optimal inter-temporal solution is provided by Lüken et al. (2009).
Figure 8shows the differences with respect to the POLDEL scenario. Negative
(positive) values indicate cost reduction (increases) with respect to the POLDEL
scenario. The sum of positive and negative components equal the difference that can
be observed in Fig. 7above. By far the most prominent influence is the emission
permit component. For the net permit importers US, Europe and China early
deployment of renewables is profitable because the cost escalating influence of
emission permits in the delayed climate policy scenario can be reduced. The opposite
line of argumentation works for the net permit exporter India. The redistribution
effect is stronger the more ambitious the RET deployment scenario is.11
The other three components are negligible compared with the permit effect.
Hence, early RET deployment in case of delayed climate policy is mainly affecting
the value of tradable emission permits, which in turn affects the redistribution among
regions related to the permit effect proportionally. The reason is that early RET
11Moreover, the regional redistribution of the value of permits scales almost linearly with the total
value of permits. For example, the total permit value of the POLD&R(m) scenario is 40% less than
in the POLDEL scenario. This reduces the permit effect in each region by about 40%. This result is
independent of the sign of the permit effect.
90 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
deployment decreases the CO2permit price in 2020. In the case POLDEL the price
amounts to 92$US/tCO2, which is reduced to 53$US/tCO2for the medium RET
deployment scenario.
In summary, the increase of mitigation costs due to delayed climate policies can
be decreased by early RET deployment. The most important factor that decreases
the mitigation costs is the devaluation of emission permits that is explained by
three reasons. First near-term emissions are decreased and therefore less of the
cumulative emissions are consumed and the emissions must be decreased from
less than the baseline level; see Fig. 6. Second, early deployment of renewables
increases the capacity of carbon free energy technologies; see Figs. 4and 5.Third,
learning-by-doing decreases the costs of additional deployment of renewable energy
technologies; see Table 3. The regional distribution impact of the devaluation of
emission permits, however, is very different and depends on the difference between
the initial and the market allocation; i.e. the closer the initial allocation of permits
matches the efficient market allocation the less emphasized would be the permit
trade effect.
4.3 Early renewable deployment and immediate climate policy—POLRET
For understanding the factors better that led to the huge cost decrease in the previous
section, we also provide another set of scenarios. In these scenarios deviations
from the optimal RET deployment are assessed for achieving the 450 ppm CO2
stabilization target. The exogenous variation of RET deployment shed light on the
impact on emissions and mitigation costs, if the renewables penetration is weaker
or stronger relative to the first best POL450 case. The deviation from the optimal
deployment path a-priori implies higher global mitigation costs and changes in the
emission time path. The open questions are whether on the global level early or
deferred RET is worse, what the effect on the regional distribution of mitigation costs
is and how the time-path of mitigation costs varies as the renewable deployment is
changed?
Figure 9shows the impact on CO2emissions for the four POLRET scenarios with
strong and weak RET deployment until the years 2020 and 2030. The graph depicts
the absolute differences with respect to the POL450 scenario until 2050. The graph
shows that deviations from the optimal path of renewable deployment to achieve the
450 ppm target imply near and long-term changes in CO2emissions. The variation of
RET deployment leads to an intuitive temporal reallocation of CO2emissions. For
the scenarios with stronger than optimal RET deployment—i.e. POLRET(s,20) and
POLRET(s,30)—coal use in the electricity sector is partially replaced in the near term
that allows higher emissions in the longer term; et vice versa for the scenarios with
weaker than optimal RET deployment. The pattern of deviations is more distinct
for the short-term deviations of RET deployment and levels out for the longer-
term deviations, since the stabilization target is the same. In general, the deviations
from the optimal CO2emission path of the POL450 scenario are assessed to be small
compared with the changes in the effects observed in the POLD&Rscenarios. The
maximum deviation of −1GtC p.a. in 2020 for the scenario POLRET(s,20) is relatively
small compared with the differences of emissions reductions between the POL410 and
the POL450 cases; see Fig. 3above. Hence, high penetration levels of RET are not
expected to replace fossil energy carriers on a one-to-one basis in a climate policy
4.6 References 91
Climatic Change
Fig. 9 Impact of deviations
from the optimal RET
deployment path on annual
global CO2emissions from the
energy sector from 2005 to
2050. The lines indicate
absolute differences of the
four POLRET scenarios with
respect to the POL450 scenario
regime, which is due to an emission rebound effect in the energy sector, see also
Bauer et al. (2010).
The impact of non-optimal RET deployment on mitigation costs is shown in
Fig. 10. The graph depicts the mitigation costs in the optimal POL450 scenario as a
reference and the mitigation costs of the four POLRET scenarios. The results confirm
the a-priori expectation that deviations from the optimal RET deployment path
increase mitigation costs, though the differences are quite small. The penalty on
mitigation cost is little higher for the two cases with RET deployment stronger than
the optimal path. Thus, at least on the global level the timing of RET deployment is
of minor importance for the mitigation costs.
Figure 11 presents the components that explain the difference between the POL450
and the two second-best scenarios that fix the RET deployment until 2030. The
methodology is the same as for Fig. 8given in the previous sub-section.
In all regions the positive (or negative) effect of strong (or weak) renewable
deployment on the macroeconomic component is offset by higher (or lower) energy
system expenditures. In the scenario POLRET(s,30) the negative effect of non-fuel
energy system costs exceeds the positive macroeconomic effect. In the scenario
Fig. 10 Global and regional
mitigation costs for the POL450
and the POLRET scenarios.
Mitigation costs are measured
as the cumulative discounted
consumption losses relative to
the BASELINE case for the
time horizon 2005 to 2100. The
indices in the parenthesis
indicate the deviation from the
optimal RET deployment
path: the intensity s(trong) and
w(eak) and the duration
(20)20 and (20)30.Thetime
varying endogenously
computed interest rate is used
as the discounting rate
92 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
USA EUR CHN IND USA EUR CHN IND
POLRET(s,30) POLRET(w,30)
Difference [%-points]
Permit trade (net trade revenue)
Fuel (domestic + net trade revenue)
Energy system (Invest.+O&M)
Macroeconomic (GDP-Invest)
Fig. 11 Impact of various effects on the mitigation costs of the US, Europe, China and India for the
POLRET(s,30) and POLRET(w,30) scenarios compared with the POL450 scenario. Note: the relative
differences are relative to changes of GDP, whereas mitigation costs are relative to consumption
differences
POLRET(w,30) the signs of these two effects is reversed. In the US and Europe
the fuel cost effect in the scenario with stronger (or weaker) than optimal RET de-
ployment reduces (or increases) the mitigation costs because less fuels are produced
domestically and imported. The opposite holds for China and India.12 The permit
effect is most important in the case of India, where strong RET deployment leads
to smaller permit export revenues because of a smaller export at lower prices et vice
versa in the case of weak RET deployment. The other regions however profit from
this effect, though the importance is smaller as they have a larger overall GDP.
The small differences of mitigation costs between the first-best and the four
POLRET second-best scenarios are mainly the net effects of larger redistribution
between economic activity and de-valuation of emission permits as well as energy
trade. The de-valuation effect, however, is much smaller than in the case of delayed
climate policy. The macro-economic and non-fuel energy system effects are the same
for all regions because the economy is supplied with more energy, but the fuel
cost and permit trade effect vary between the regions. Whether a region suffers or
gains from deviations from the optimal RET deployment depends on the direction
and magnitude of the latter two effects. The differences in the permit trade effect,
which is much smaller in the POLRET than in the POLD&Rscenarios, is due to the
relatively smaller impacts on the CO2prices because the renewable mitigation option
substitutes alternative options leading to only small CO2price changes. Also the
impact on the CO2emission path is smaller in the POLRET than in the POLD&R
12We do not elaborate the energy trade effect more here, because the complex interplay of prices
and quantities for the various energy carriers is not the focus of this study.
4.7 Figures 93
Climatic Change
Fig. 12 Time paths of
consumption differences 2005
to 2100 relative to the
BASELINE case for the
POL450 and the POLRET
scenarios. The indices in the
parenthesis indicate s(trong)
and w(eak) and the duration
of RET deployment; i.e.
(20)20 and (20)30
scenarios shown in Figs. 6and 9. Hence, the price and the quantity effect lead to a
decrease of the overall permit trade effect.
Since the transformation of the energy system is a challenge for the present as
well as the following generations, the time paths of mitigation costs of the scenarios
are analyzed next. Figure 12 depicts the time paths of consumption differences of the
POL450 as well as the four POLRET scenarios relative to the BASELINE case. The
optimal case POL450 shows a path that increases until 2040 peaking at 1.24% p.a. and
decreases towards zero afterwards.13 The four POLRET scenarios show significant
deviations from this path. Strong RET deployment until 2030 (POLRET(s,30)) leads
to an even flatter time path that already starts at 0.4% p.a. but does never exceed
1% p.a.. The two low deployment scenarios instead exhibit a much more emphasized
peaking behavior. They start with some moderate gains, which are due to lower
overall investments allocated to the energy system, but the maximum is higher.
This is especially the case for the POLRET(w,30) scenario, in which the weaker than
optimal RET deployment lasts until 2030 peaking at 1.5% p.a. in 2035.
The comparison of the first-best climate mitigation scenario with the second-
best scenarios shows that deviations from the optimal RET deployment paths only
imply small changes in the optimal emission time paths and the global cumulative
discounted mitigation costs. The regional distribution of the mitigation costs is
heterogeneous and depends on the intensity of the RET deployment scenario as
well as the contribution of various economic influences affecting the regions. Also
the time path of mitigation costs is changed significantly for the different RET
deployment scenarios. The cumulative discounted mitigation costs do not reflect the
shape of these time paths.
5 Discussion and further research
Early deployment of renewable energy technologies reduces the global costs for
achieving a 450 ppm CO2concentration target, if climate policy measures are
13The shape is not unusual for climate change mitigation assessments with stabilization targets; see
Edenhofer et al. (2010, p. 32).
94 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
delayed. The cost reduction is due to the devaluation of emission permit that can
be explained by three effects. First, early RET deployment replaces some fossil
fuel utilization and leaves more emissions for the rest of the 21st century. Second,
if large capacities of RET are available in 2020 the negative effect of significantly
reducing new fossil fuel investments and increasing the utilization of biomass with
CCS is contained. Finally, additional investments in RET will be cheaper post-2020
as early deployment will reduce the investment costs due to learning-by-doing. The
global mitigation costs, however, cannot be reduced below the first-best scenario with
optimal timing of all mitigation measures.
Similar results can be expected for other energy–economy models since Jakob
et al. (2011, this issue) and Clarke et al. (2009, p. S77) report significant increases
of carbon prices and mitigation costs as climate policy is delayed. The effectiveness
of technology policies for reducing the emissions in the near term and triggering
improvements of low carbon technologies is the crucial link to reduce the costs
of delaying climate policies. Emission rebound effects can turn out to be serious
obstacles to the positive impact of technology policies. Bauer et al. (2010) quantified
the emission rebound effects of various mitigation options. The debate about the
Green Paradox is even more pessimistic about technology policies, because fossil
fuel extraction is expected to increase in the near term as fossil fuel owners anticipate
future devaluation of their resources; see e.g. Sinn (2008).
Deviations from the optimal RET deployment path in the case of immediate
climate policy increase the global mitigation costs compared with the first-best
solution. The impacts on global emissions and discounted global mitigation costs
are quite small, which confirms the finding of Böhringer et al. (2009) in the context
of European climate and energy policy. Hence, the optimal timing of renewable
investments is of minor importance from a global point of view. The impact on
the time path of mitigation costs over the entire century, however, is significant.
The optimal time path of mitigation costs follows an inverted U-shape with a peak
around the middle of the century. Higher than optimal deployment of RET flattens
the curve, but less than optimal deployment increases the peak of mitigation costs.
This inter-temporal reallocation is not reflected in the discounted mitigation costs.
The intergenerational re-distribution of mitigation costs due to different renewable
energy investments is not discussed so far in the scientific literature. This issue is not
addressed in the present paper and left to future research. An additional argument
for stronger RET deployment (scenario POLRET(s)) is that it serves as a hedging
strategy against the case that in the future it might become necessary to achieve a
lower stabilization target than initially chosen, e.g. decreasing the CO2stabilization
level from 450 to 410 ppm. The significance of technology policies for hedging against
climate risks has not been explored yet and seems an interesting field of future
research.14
The impact of variations of RET deployment on global mitigation costs is larger
in the case of delayed than in the case immediate climate policy. This result is in
contrast with the finding of Kverndokk and Rosendahl (2007), who stated that the
delay of carbon pricing is less significant than the delay of technology subsidies.
The difference of results can be related to a number of factors. Kverndokk and
14The authors would like to thank an anonymous reviewer for this suggestion.
4.7 Figures 95
Climatic Change
Rosendahl (2007) only reflect the electricity sector, in which learning technologies
are very important, but the present study deals with the total energy sector, where
learning technologies are less important. Moreover, Kverndokk and Rosendahl
(2007) allow for optimal adjustment of the subsidy in case of delayed climate policy,
but the present study only considers exogenous variations of RET deployment paths.
Moreover, in case of a delayed technology subsidy the carbon tax in Kverndokk
and Rosendahl (2007) would need to increase significantly to achieve any emission
reduction (beyond demand responses) because besides two carbon-free learning
technologies there is only one fossil generation technology considered. The RE-
MIND model instead considers a large variety of different electricity generating
technologies and renewables are differentiated by quality grades, which implies a
smoothed transition.
The interregional distribution of mitigation costs is heterogeneous in both cases
of immediate and delayed climate policies. At the regional level the US and Europe
would gain from strong worldwide deployment of RET in both climate policy
regimes. China would gain from strong and medium early deployment, if climate
policies are delayed, but in case of immediate climate policy China would lose from
strong RET deployment. India and all other non-Annex 1 regions lose from early
RET deployment in both climate policy regimes.
In case of delayed climate policy variations of early RET deployment have
significant impact on the global mitigation costs as well as the regional distribution.
The main factor is that the total value of emission permits allocated to the regions
decreases as early RET deployment is imposed on the system. This has a negative
effect on regions, which receive relatively plentiful assignments and export these
permits. Conversely, net importers gain from early RET deployment.
The main conclusion for policy making from this study is that early deployment of
renewable energy technologies can reduce additional global costs of delayed climate
policy. High income regions and China can reduce the costs of delayed climate policy
by inducing this transformation of the global energy system. Especially the US and
Europe would also profit from strong world-wide renewable deployment in case of
immediate climate policy, thus, making this option a robust strategy for these regions.
Low-income regions may not experience this cost reducing effect. In both cases of
immediate and delayed climate policy low-income countries are found to lose from
ambitious renewable deployment policies.
The present study only analyzed the influence of early RET deployment, but
also other technology related policies should be studied. Most promising to us
seem energy efficiency, fossil CCS, gas for coal substitution, and nuclear. The
present study also showed that it is important to study the factors that determine
the mitigation costs. As the value of carbon permits appears to have a significant
influence on the results, future research should aim at identifying the reasons for
changing mitigation costs as technology policy is imposed in case of delayed climate
policy. Furthermore, the scenario space may also be extended to the land-use sector
asking for scenario assumptions that reduce GHG emissions other than CO2from
the energy sector. Following this line of research would broaden the perspective on
alternative scenarios limiting and reducing GHG emissions by different policies.
The method of developing second-best scenarios is appropriate to explore the
range of alternative future developments and the implications on costs providing
guidance to policy makers and supporting international negotiations. However,
96 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Climatic Change
the major interest of policy makers is the assessment of international agreements,
optimal policy instruments and their coordination in a policy mix. The contributions
of Kverndokk and Rosendahl (2007) and Gerlagh et al. (2009) address this challenge,
but need much more improvement. The extension towards higher technological
resolution and multi-regional frameworks requires additional theoretical foundation
and more powerful numerical algorithms to solve these models.
References
Bauer N, Edenhofer O, Kypreos S (2008) Linking energy system and macroeconomic growth models.
J Compo Manage Science (Special Issue) 5(1–2):95–117
Bauer N, Edenhofer O, Leimbach M (2009) Low-stabilization scenarios and technologies for carbon
capture and sequestration. Ene Proced 1(1):4031–38
Bauer N et al (2010) Technologies, policies and economics of global reductions of energy related
CO2emissions. An analysis with ReMIND. Paper presented at the World Conference of Envi-
ronmental and Resource Economists, Montreal. http://www.webmeets.com/WCERE/2010/prog/
viewpaper.asp?pid=834 June 28–July 2
Böhringer C, Löschel A, Moslener U, Rutherfor TF (2009) EU climate policy to 2020: an economic
impact assessment. Energy Econ 31:S295–S305
Chen H, Cong TN, Yang W, Tan C, Li Y, Ding Y (2009) Progress in electrical energy storage system:
a critical review. Prog Nat Sci 19(3):291–312
Clarke L, Edmonds J, Krey V, Richels R, Rose S, Tavoni M (2009) International climate policy
architectures: overview of the EMF22 international scenarios. Energy Econ 31:S64–S81
Den Elzen MGJ, Lucas PL, van Vuurren DP (2008) Regional abatement action and costs under
allocation schemes for emission allowances for achieving low CO2-equivalent concentrations.
Clim Change 90:243–68
Edenhofer O, Lessmann K, Kemfert C, Grubb M, Koehler J (2006) Induced technological change:
exploring its implications for the economics of atmospheric stabilization: synthesis report from
the innovation modeling comparison project. Energy J (Special Issue on Endogenous Techno-
logical Change and the Economics of Atmospheric Stabilization), pp 57–107
Edenhofer O, Stern N, Bauer N, Brunner S, Jakob M, XXX (2009) Towards a global green recovery.
Recommendations for Immediate G20 Action. Report prepared on behalf of the German For-
eign Ministry. URL: http://www.pik-potsdam.de/members/edenh/publications-1/global-green-
recovery_pik_lse
Edenhofer O, Knopf B et al (2010) The economics of low stabilization: model comparison of
mitigation strategies and costs. Energy J (Special Issue on the Economics of Low Stabilization),
pp 11–48
Energy Information Administration EIA (2009) International energy outlook 2009. Washington, DC
European Commission EC (2006) World energy technology outlook—WETO H2. ISBN: 92-79-
01636-9. 168 pages
Gerlagh R, Kverndokk S, Rosendahl KE (2009) Optimal timing of climate change policy: interaction
between carbon taxes and innovation externalities. Environ Resour Econ 43:369–390
Global Wind Energy Council GWEC (2010) Global wind report. Various issues. http://www.gwec.
net/index.php?id=8
Greenpeace International (2008) Energy [R]evolution. A sustainable global energy outlook. ISBN
9789073361898. URL: http://www.energyblueprint.info/. Accessed 25 Mar 2010
International Energy Agency IEA (2002) World Energy Outlook 2002. Paris, France
International Energy Agency IEA (2008) Energy Technology Program 2008. Paris, France
International Energy Agency IEA (2009) World Energy Outlook 2009. Paris, France
International Energy Agency IEA (2010) World Energy Outlook 2010. Paris, France
Jäger-Waldau A (2009) Photovoltaics status 2009: research, solar cell production and market imple-
mentation of photovoltaics; Euro-Report EUR 24027 EN, 2008. Luxembourg: Office for Official
Publications of the European Union, 2009. ISBN 978-92-79-12800-4, p 116
Jakob M et al (2011) Time to act now? Assessing the costs of delaying climate measures. Clim Change
(this issue)
4.7 Figures 97
Climatic Change
Kverndokk S, Rosendahl KE (2007) Climate policies and learning by doing: impacts and timing of
technology subsidies. Resour Energy Econ 29:58–82
Leimbach M, Toth F (2003) Economic development and emission control over the long term: the
ICLIPS aggregated economic mode. Clim Change 56:139–165
Leimbach M, Bauer N, Baumstark L, Edenhofer O (2010) Mitigation costs in a globalized world:
climate policy analysis with ReMIND-R. Environ Model Assess 15:155–173
Luderer G et al (2011) The economics of GHG emissions reductions—results and insights from the
RECIPE model intercomparison. Clim Change (this issue)
Lüken M, Edenhofer O, Knopf B, Leimbach M, Luderer G, Bauer N (2009) The role of technological
availability for the distributive impacts of climate change mitigation policy. Energy Policy (under
revision)
Magne B, Turton H, Kypreos S (2010) Technology options for low stabilization pathways with
MERGE. Energy J (Special Issue on the Economics of Low Stabilization), pp 83–107
Manne A, Richels R (2004) The impact of learning-by-doing on the timing and costs of CO2
abatement. Energy Econ 26:603–619
Manne AS, Rutherford T (1994) International trade, capital ows and sectoral analysis: formulation
and solution of intertemporal equilibrium models. In Cooper WW, Whinston AB (eds) New
directions in computational economics. Kluwer, Dordrecht, pp 191–205
Nakicenovic N, Riahi K (2002) An assessment of technology change across selected energy scenarios.
International Institute for Applied System Analysis (IIASA), RR-02-005, Laxenburg, Austria
Pietzcker R, Manger S, Bauer N, Luderer G, Bruckner T (2009) The role of concentrating solar
power and photovoltaics for climate protection. Paper presented at the 10th IAEE European
Conference, Vienna, Austria
Rogelj J, Nabel J, Chen C, Hare W, Markmann K, Meinshausen, Schaeffer M, Macey K, Höhne N
(2010) Copenhagen accord pledges are paltry. Nature 464:1126–1128
Sinn HW (2008) Public policies against global warming: a supply side approach. Int Tax Public Financ
15:360–394
Sorrell S, Sijm J (2003) Carbon trading in the policy mix. Oxf Rev Econ Policy 19(3):420–437
van Vuuren DP, Bellevrat E, Kitous A, Issac M (2009) Bio-energy use and low stabilization scenarios.
Energy J (Special Issue on the Economics of Low Stabilization), pp 193–221
Wissenschaftlicher Beirat Globale Umweltveränderung WBGU (2003) Energiewende zur Nach-
haltigkeit. Springer, Berlin
98 Chapter 4 The Role of Renewables in the Low-Carbon Transformation
Chapter 5
The Dimensions of Technological Change and
Their Impacts on Climate Change Mitigation.∗
Lavinia Baumstark
Marian Leimbach
∗submitted to Energy Policy as Baumstark, L., M. Leimbach (2010): The Dimensions of Technological
Change and Their Impacts on Climate Change Mitigation.
99
100 Chapter 5 Dimensions of Technological Change
The Dimensions of Technological Change and
Their Impacts on Climate Change Mitigation
Lavinia Baumstark∗
, Marian Leimbach
March 29, 2011
Abstract
In climate policy analysis, mitigation costs and strategies are key vari-
ables. This study explores the impact of two dimensions (i.e. dynamics and
direction) of technological change on these variables based on the climate
policy model REMIND-RS that links a macroeconomic model and an en-
ergy system model. Results of a sensitivity analysis show that mitigation
costs and strategies are quite sensitive to the dynamics and especially the
direction of technological change represented by changes of production fac-
tors’ efficiency. For example the higher the labour efficiency, the higher are
the mitigation costs. Higher energy-related efficiencies can more than com-
pensate this increase of mitigation costs. Moreover, it turns out that energy
efficiency improvements in the electricity and transport sector have a much
higher impact on the mitigation costs than those in the stationary non-electric
sector. The question arises whether the impact of technological change varies
in a world with different structural assumptions. The latter is emulated by
modifying the elasticity of substitution and by restricting the use of some
technologies. It turns out that the sign and intensity of the impacts of techno-
logical change are influenced by the scenario world assumed.
JEL classification: E27, O11, O30, Q43, Q54
∗PIK - Potsdam Institute for Climate Impact Research, P.O. Box 60 12 03, D-14412 Potsdam,
Germany, Tel. ++49/331/288-2535, e-mail: [email protected]
1
101
1 Introduction
The IPCC Third Assessment Report (AR3) highlights the importance of baseline
assumptions of economic development like GDP growth, energy intensity and
emissions in determining the economic costs and strategies of climate policies.
However, there are only a few studies (Manne [22], B¨
ohringer [3]) that analyse
this issue in a systematic way. This paper contributes to fill this research gap. By
exploring different scenarios of efficiency growth, the impact of the dynamics and
direction of technological change on the costs and strategies of climate change
mitigation is qualified and quantified. This investigation is conducted within a sce-
nario framework of alternative assumptions about possibilities of substitution and
availability of technologies.
Technological change studied in this paper is characterized by two dimensions -
dynamics and direction. The dynamics of technological change represented by
overall efficiency improvements determine the economic growth path. While this
influences the baseline results of a model, the question arises whether the dynam-
ics also influence the costs and/or strategies of climate policies. This question also
applies with respect to the impact of the direction of technological change. In this
study, the sensitivity of mitigation costs and strategies is analysed with a particu-
lar focus on the impacts of the relationship between labour and energy efficiency
improvements.
Mitigation costs describe the costs of countries, regions and the whole world for
keeping a given climate stabilization goal and can be measured as consumption
losses compared to a baseline scenario without climate policy measures. In in-
tertemporal growth models, these costs are on the one hand driven by the general
growth path. If a region grows fast in a baseline scenario, it will have to reduce a
huge amount of emissions and might face high mitigation costs in a policy scenario
(see L¨
oschel [21]). On the other hand, mitigation costs depend on the innovation
capacity and flexibility of the energy system. Assumptions on the growth path
of production factors’ efficiency affect both drivers of mitigation costs: general
macroeconomic growth and technological evolution of the energy system.
Within this study, which analyses the sensitivity of mitigation strategies on the
2
102 Chapter 5 Dimensions of Technological Change
variation of the dimension of technological change, the primary energy mix of a
region is used as an indicator for the chosen strategy. If the whole economy of a
region is growing fast, the total energy demand will increase. It is expected that
more and also gradually more expensive carbon-free technologies show up in the
optimal energy mix of a climate policy scenario ( Dowlatabadi [8]). Such changes
in the mitigation strategy will be explored in some detail in this paper as well as
changes in response to technological change that result in a more labour-intensive
or energy-intensive production structure.
Regarding energy specific technological progress, this study explores the impact of
exogenous efficiency growth on the mitigation costs and strategies based on a cli-
mate policy model that links a macroeconomic model and an energy system model.
In climate policy models, improvements in energy efficiency are often represented
by aggregated parameters, for example AEEI - the Autonomous Energy Efficiency
Improvement (see for example Sanstad [30]). Gerlagh and van der Zwaan [14] in-
cluded the parameter autonomous energy service efficiency improvement (AESEI)
in their sensitivity analysis of mitigation costs. By going a step further, the present
paper analyses the influence of variations of exogenous growth and technologi-
cal change in a model framework that assigns efficiency parameters to each factor
of a nested CES production function. Some of these production factors represent
end-use energy sectors. Exogenous growth that affects each production factor in a
different way mimics biased and directed technological change. In this context, it
is of interest, whether an increase of factor productivity results in factor savings or
rather in an extended use of the production factor (Nordhaus [27]).
With the exogenous growth assumption explored in this paper, a more or a less en-
ergy efficient world can be simulated. This exogenous mechanism of technological
change is simple, transparent and ready for sensitivity analysis (Gillingham [15]).
All model results depend on basic assumptions about structural elements. The
question arises whether the sensitivity on parameter variations changes in a world
with different baseline assumptions. A general structural change can be emulated
by switching to a value of the elasticity of substitution that shifts the whole model
behaviour. This study investigates whether the efficiency of a production factor
will play a significant role for mitigation cost estimates and whether this influence
depends on the elasticity of substitution. It is also analysed how mitigation cost
3
5.1 Introduction 103
impacts of productivity changes vary and how the strategies to achieve a climate
target have to be adapted in scenario worlds that differ with respect to the available
portfolio of energy technologies.
The structure of this paper is as follows: Within section 2, the applied model
REMIND-RS is described. Section 3 explains the dynamics and directions of tech-
nological change. Numerical results of a sensitivity analysis on the dimensions
of technological change will be discussed. Section 4 highlights the findings from
a rerun sensitivity analysis within different scenario worlds. In the first part, the
interaction between elasticities of substitution and efficiency growth changes is
analysed, while the second part studies the role of technological restrictions. Short
conclusions and possible further model developments are discussed in section 5.
2 Model and Scenario Set-up
2.1 REMIND-RS
The model REMIND-RS, which is used for the analysis of this paper, is based
on the model REMIND-R. The multi-regional hybrid model REMIND-R1couples
a macro-economic model, an energy system model and a simple climate model.
By maximizing a global welfare function, REMIND-R computes a social planner
solution. For a detailed description of the model see Leimbach [20]. For this type
of analysis, a fast and flexible model named REMIND-RS has been developed; it
includes the complete energy system module and climate module of REMIND-R
but has been scaled down from 11 to 5 regions. These world regions are:
•USA - USA
•EUR - EU27
•CHN - China
•INA - Developing regions, less resources
•ROW - Mainly resource exporting regions
1The equations are explained at http://www.pik-potsdam.de/research/research-
domains/sustainable-solutions/esm-group/remind-code
4
104 Chapter 5 Dimensions of Technological Change
The region INA includes India, Sub-Saharan Africa, Latin America and Other Asia
(Central and South East Asia), while Middle East and North Africa, Japan, Russia
and the rest of the world (Canada, Australia, etc.) are part of the region ROW. The
parameterisation of REMIND-RS is based on the version REMIND-R 1.3 used in
the project RECIPE ([11]).
The production function of REMIND-RS is represented by a nested CES function
of the general structure
OUTt,r =ÃX
i²CES
(AINi,t,r ·INi,t,r)ρOUT,r !1/ρOUT,r
.(1)
The output OUT for each region rby each time step tis the sum of the produc-
tion factor inputs IN multiplied by the associated efficiency parameter AIN to the
power of ρOUT = (σOUT −1)/σOUT , where σOUT is the elasticity of substitution
to produce the output OUT. The list CES assigns the associated inputs IN to
each output OUT. The first CES-level combines capital, labour and total final en-
ergy. This total final energy is produced by stationary and transport energy which
in turn are the result of a CES production function. The whole CES production
tree is shown in Figure 1. The elasticities of substitution σOUT are lower than one
in the first two CES-levels, while the production factors of the lower CES-levels
are assumed to be complements (i.e. σOUT >1). These values are comparable
to the values assumed by Gerlagh ([14]). The default REMIND model applies
an exogenous growth path of labour efficiency and energy efficiency change rates
which are defined in relation to labour productivity changes. This general exoge-
nously driven growth assumption is in common with many other growth models
(for example Nordhaus[26], Manne [23]), thus providing additional relevance to
this analysis for the interpretation of other mitigation cost studies.
5
5.2 Model and Scenario Set-up 105
Figure 1: Nested CES structure of the macro-economic module of REMIND-RS
2.2 Default Scenarios
In running a climate policy scenario, REMIND-RS simulates a cap-and-trade cli-
mate policy regime that applies a cumulative emission budget as a basis for de-
termining regional caps. A global budget of 300 GtC for 2000-2100 is divided
between the regions by the Contraction and Convergence approach (Meyer [25]).
Following Meinshausen [24], a global carbon budget is always linked to a likeli-
hood to keep a specific temperature goal.
The regional emissions which sum up to an optimal emission path that holds the
emission budget of a baseline scenario without any climate mitigation policy are
shown in Figure 2(a), those of a policy scenario in Figure 2(b). This stabilization
scenario implies negative global emissions (black line in Figure 2(b)) from 2070
on. This negative emissions are mainly realized by the use of technologies with
biomass and CCS.
6
106 Chapter 5 Dimensions of Technological Change
2020 2040 2060 2080 2100
0
5
10
15
20
25
Year
Emissions [GtC]
ROW
INA
CHN
EUR
USA
(a) Baseline scenario
2020 2040 2060 2080 2100
−4
−2
0
2
4
6
8
10
Year
Emissions [GtC]
ROW
INA
CHN
EUR
USA
(b) Policy scenario
Figure 2: Emissions of the default baseline and policy scenario
The global GDP grows from 48 trill. $US 2in 2005 to 521 trill. $US in 2100,
which classifies the baseline scenario as a high growth scenario.
The applied emissions trading regime (Contraction and Convergence3) which im-
plies equal per capita permit allocation from 2050 on provides INA due to its high
share of global population with a huge amount of global emission rights. INA can
sell parts of these permits profitably to the developed regions.
3 Dimensions of Technology Change - Impacts on Mitiga-
tion Costs and Strategies
The model REMIND-RS reflects technological change in the macro-economic pro-
duction part mainly due to the exogenous growth of the efficiency parameters AIN .
In the first part of this section, the impact of the dynamics of overall technological
change is analysed. This is represented by a proportional change of all efficiency
parameters in the macro and energy demand sector. Within these experiments, the
relationship between the different efficiency parameters is kept constant.
2All macro economic values are measured in constant international $US 2005 (market exchange
rate)
3First, the emission permits are allocated relative to grandfather emissions and via a linear tran-
sition equal per capita emissions for each region assumed from 2050 on.
7
5.3 Dimensions of Technological Change - Mitigation Costs and Strategies 107
In contrast, if the efficiency of only one production factor is varied, the relationship
of the growth path will change and differently directed technological change can
be analysed. The sensitivity of mitigation costs and strategies on this dimension of
technological change is studied in the second part of this section.
3.1 Dynamics of Technological Change
To simulate an alternative dynamic of technological change with generally higher
or lower growth, all efficiency parameters but the efficiency of capital are mul-
tiplied by 0.6, 0.8, 1.0, 1.2, 1.4. Within all these experiments of the sensitivity
analysis, the ratio between all efficiency growth parameters is the same as in the
default scenario.
0 0.5 1 1.5 2 2.5
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
0.6
1.0
1.4
(a) Mitigation costs
0.6 1.0 1.4
0
5
10
15
20
25
Cumulative primary energy [ZJ]
Fossil Fuels w/o CCS
CCS Fossil
Renewables w/o Biomass
Biomass w/o CCS
CCS Biomass
Nuclear
(b) Mitigation strategies
Figure 3: Dynamics variation
Figure 3(a) shows the mitigation costs for the default scenarios (1.0) and an in-
creased/decreased level of technological change - varied by 40% (1.4/0.6). Miti-
gation costs within this study are defined as the percental consumption difference
between a baseline (BAU) and a policy scenario averaged over time. The higher
the overall growth, the lower are the costs for all regions to keep the climate target.
The cost difference, however, is moderate. The relationship between labour effi-
ciency and total energy efficiency, which remains unchanged, is crucial. The sign
of the change of mitigation costs indicate that it is more important for the regions
8
108 Chapter 5 Dimensions of Technological Change
to be efficient in the energy sectors than it is worse that an increased growth path
means a higher amount of emissions to be reduced in the policy scenario.
The higher the (efficiency) growth path, the more primary energy is used. Figure
3(b) shows the global cumulative (over time) primary energy mix for the default
policy scenario and the two extreme variations, i.e. 0.6 and 1.4. The high ef-
ficiency scenario is associated with a higher share of renewable energies in the
energy mix, while the total amount of fossil fuels, biomass and nuclear are nearly
the same in all scenarios. Due to the increased carbon price and the level of con-
sumption already reached, it becomes increasingly expensive to use more of these
technologies. However, the variation of the dynamics of technological change has
a moderate impact on mitigation strategies, mainly increasing the use of renewable
energy when assuming a faster growth path.
3.2 Direction of Technological Change
Within the sensitivity analysis, sub-sets of efficiency growth parameters are var-
ied to mimic different directions of technological change. In detail, the efficiency
parameters of the following production factors are multiplied by 0.6, 0.8, 1.0, 1.2,
1.4:
1. Labour
2. Total energy
3. Transport energy
4. Electricity
5. Stationary non-electric energy
The variation of the efficiency of any energy type, which is not at the lower end of
the CES-tree, is achieved by varying all related end-use energy types of the lowest
CES-level. For example, for the variation of the efficiency of transport energy, the
efficiencies of hydro, diesel and petrol are changed (see figure 1).
Since REMIND-RS is a hybrid model, the impact of efficiencies in the macro-
economy and some energy production sectors can be studied simultaneously. It
turns out that the variation of all parameters but the efficiency of stationary non-
electric energy show a huge impact on mitigation costs. The impact on mitigation
9
5.3 Dimensions of Technological Change - Mitigation Costs and Strategies 109
strategies depends on the analysed/changed sector.
3.2.1 Labour and Total Energy
This part focuses on the impacts of the variation of efficiency parameters from the
macro-economic part of REMIND-RS, i.e. the first CES-level.
0 1 2 3
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
0.6
1.0
1.4
(a) Labour efficiency variation
0 1 2 3
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
0.6
1.0
1.4
(b) Total energy efficiency variation
Figure 4: Mitigation costs for labour and total energy efficiency variation
The higher the labour efficiency growth, the higher are mitigation costs for the
whole world (see Figure 4(a)). The main reason for this effect is the higher base-
line growth and because of this the higher mitigation gap to keep a budget target.
When increasing labour efficiency growth, the use of all other production factors
is increasing, more capital and total energy (see figure 5(a)) are used. This results
in higher consumption, including higher consumption of energy and fossil fuels
and because of this in higher emissions in the baseline scenario. The emission
gap, which has to be reduced in the policy scenario, develops in correlation with
the efficiency of labour. All but the region INA faces additional mitigation costs
when multiplying the labour efficiency growth with 1.2 and 1.4 compared to the
default experiments. INA gains from the higher demand for emission permits from
the other regions and from a higher permit price. The default permit price starts
from 50 $US/tC in 2010 and increases to 460 $US/tC in 2050, while the permit
price for the high labour efficiency (1.4) scenario starts from 80 $US/tC in 2010
10
110 Chapter 5 Dimensions of Technological Change
and increases to 810 $US/tC in 2050.
Figure 4(b) shows the impact of the variation of total energy efficiency growth.
The higher the total energy efficiency, the lower are the mitigation costs. This is in
contrast to the results of labour efficiency variations but these findings are in line
with Edenhofer [10]. The impact of higher total energy efficiency can be explained
by less energy that is needed to produce the baseline GDP. Because of this and
the fact that in the baseline energy is mainly produced based on fossil fuels, the
amount of emissions to be reduced in the policy scenario is lower. The increase of
total energy efficiency does not induce a general production increasing effect but
reduces the use of the production factor total energy. This reaction occurs because
of the low elasticity of substitution in the first CES-level and is known as factor
saving technological change (Fellner [13]). Given an exogenous population (pro-
duction factor labour) path to get an optimal share of production factors, the total
energy use is reduced. This lower demand for total energy can be met by the use
of cheap technologies. In contrast, a reduced productivity of the production factor
total energy increases the pressure on the energy system when trying to keep the
budget target. The share of renewable and CCS technologies on the cumulative
energy mix is increased (see figure5(b)). Technology possibilities and capacities
become limited, and it is more efficient to reduce general growth before investing
into very expensive low-carbon technologies.
Although the impact of changes in labour efficiency on the emission gap is higher
than the variation based on the same percentage variation of total energy efficiency,
the latter parameter shows the highest impact on global mitigation costs. The effi-
ciency of energy use seems to be more important for the mitigation costs than the
mitigation gap of the emissions. Lowering the total energy efficiency to 60% re-
sults in a 52% increase of mitigation costs (see Table 1), while the same reduction
of labour efficiency lowers mitigation costs by 23%.
11
5.3 Dimensions of Technological Change - Mitigation Costs and Strategies 111
0.6 1.0 1.4
0
5
10
15
20
25
Cumulative primary energy [ZJ]
Fossil Fuels w/o CCS
CCS Fossil
Renewables w/o Biomass
Biomass w/o CCS
CCS Biomass
Nuclear
(a) Labour efficiency variation
(b) Total energy efficiency variation
Figure 5: Mitigation strategies of policy scenarios subject to labour and total energy
efficiency variation
3.2.2 Energy-related Production Factors
In a simpler way as demonstrated for the impact of efficiency variation of total en-
ergy, increasing efficiency of the energy-related production factors from the lower
CES-levels reduces the global mitigation costs. As shown in Table 1, only the vari-
ation of the efficiency of stationary non-electric energy has a minor impact on the
global mitigation costs. All other energy-related production factors are as impor-
tant as labour efficiency and total energy efficiency for the level of mitigation costs,
thus showing the prominent role of the energy efficiency growth path in many pro-
duction sectors for climate change mitigation.
The small impact of the efficiency parameter of stationary non-electric energy
can be explained by the fact that this production factor can be easily substituted
(σS= 1.5) by an increase of electricity, which can be produced less carbon in-
tensively. But the other way around, a reduced efficiency of electricity cannot
be compensated by a higher use of stationary non-electric energy because of the
carbon intensity of this sector. Consequently, it results in significantly higher miti-
gation costs compared to the default scenarios. Thus, the possibility to decarbonize
influences the intensity of the impact of efficiency variations of production factors
in the same CES-level on mitigation costs changes.
12
112 Chapter 5 Dimensions of Technological Change
Table 1: Percentage change of global mitigation costs depending on the variation
of the following efficiency parameter
Parameter 0.6 0.8 1.2 1.4
Dynamic +11.04 +4.26 -2.38 -4.71
Labour efficiency -23.10 -12.36 +13.14 +25.02
Total energy efficiency +52.65 +20.61 -13.64 -23.12
Efficiency of transport energy +27.01 +10.46 -6.37 -10.66
Efficiency of electricity +23.40 +9.68 -6.74 -11.97
Efficiency of stat. non-elec. energy +1.58 +0.68 -0.58 -1.13
The reactions of the regional mitigation costs on the parameter variation are in
general in line with the global results. Yet INA, in contrast to the other regions,
faces higher mitigation costs when increasing the efficiency of transport energy.
In REMIND-RS, the transport sector is most difficult to decarbonize. As a result,
INA loses its gains from permit exports when the use of transport energy is reduced
because of a higher efficiency in this sector.
Mitigation cost changes caused by the explored efficiency parameter variations
range on the upper end of changes that literature identifies for other kinds of
changes in model assumptions and within scenario analysis, e.g. technology avail-
ability, climate target (cf. Edenhofer et al. [12]). Comparably high percentage
mitigation cost differences are shown by Riahi [28] between SRES scenarios that
represent a complex of baseline assumptions and between different climate stabi-
lization targets.
In analysing how strong the mitigation strategies are influenced by the variation of
the direction of technological change, a Kaya decomposition (Steckel[31], Kaya[18])
is used. It represents a common tool for studying emission dynamics (e.g. Rogner
[29]).
Within the Kaya decomposition, carbon emissions Fare split into four terms: Pop-
ulation L, GDP per capita (Y/L), energy intensity of GDP (E/Y ), carbon inten-
sity of energy (F/E). It holds the equation
F=LµY
LµE
YµF
E.(2)
13
5.3 Dimensions of Technological Change - Mitigation Costs and Strategies 113
low high low high
−10
−5
0
5
10
∆ CO2 [%]
(a) Left panel labour, right panel total energy
low high low high low high
−3
−2
−1
0
1
2
3
∆ CO2 [%]
GDP per capita
Energy intensity
Carbon intensity
(b) Left panel transport energy, middle panel sta-
tionary non-electric energy, right panel electricity
Figure 6: Kaya decomposition of emission differences between policy scenarios
with efficiency variation and default policy scenario
Each bar in figure 6 stands for the Kaya decomposition of the difference of cumula-
tive carbon emissions between a default policy scenario and a policy scenario with
efficiency variation. Because of the exogenous population path, which is always
the same in all scenarios, the contribution of the factor population Lis always zero.
In a world with high (1.4) labour efficiency mainly an increased GDP per capita
causes the increased carbon emissions. However, the carbon intensity is lowered
due to an increased use of renewable energies. The other way around, the GDP-
effect will mainly reduce emissions, if labour efficiency is lowered. For all the
efficiencies of energy-related sectors GDP per capita plays a minor role for the
amount of emissions but energy and carbon intensity are important. A low (0.6)
total energy efficiency results in an increased total energy use (grey bar) but also an
increased share of renewable energy and thereby a lower carbon intensity (green
negative bar). So there are two effects influencing the emissions in the opposite
direction. This indicates a change of the mitigation strategies: more but less car-
bon intensive energy. In principle, these results also hold for the variation of the
efficiency parameters of the lower CES-levels: the lower the energy related effi-
ciency, the higher the energy intensity and the lower the carbon intensity because
of a higher share of renewable technologies in the primary energy mix.
14
114 Chapter 5 Dimensions of Technological Change
4 DifferentWorlds- Impacts on MitigationCosts and Strate-
gies
Two other aspects that determine state and perspectives of technological change
will be analysed in the following. This applies first to the elasticity of substitution
and second to the availability of energy technologies. As these elements are re-
lated to basic model assumptions, changes in these aspects are denoted as different
worlds. If such basic assumptions change, what will this mean for the impacts of
dimensions of technological change on mitigation costs and strategies? In a world
with another elasticity of substitution between some production factors or with
reduced availability of some technologies, certain production factors might play a
different role in a policy scenario. Because of this, the impact of the variation of the
efficiency of production factors on mitigation costs and strategies might change.
4.1 Elasticity of Substitution
4.1.1 Reference Scenario Setting
This section focusses on the impact of the elasticity of substitution without chang-
ing the dimensions of technological change. The assumptions about these param-
eters influence the whole model behaviour - macro-economic and energy system
related. Within this paper, two elasticities of substitution in key CES-levels are
changed, i.e. σYand σS(see figure1).
The first example is the elasticity of substitution of the first CES-level σY. In litera-
ture, there are a lot of empirical analyses which try to estimate this parameter. Most
authors indicate an elasticity of substitution less than one (Kemfert [19]), what is
in line with the default assumption of REMIND-RS of σY= 0.5. Burniaux [6]
outlines that in the long-run the elasticity between capital and energy is ranging
from 0.4 to 1.6. Some authors (Chang [7], Hazilla [17], Watanabe [33]) tend to
assume that energy and capital/labour are very good substitutes with an elasticity
of substitution greater than one. To analyse the impact of a drastically change, ex-
periments with σY= 1.2are run with REMIND-RS.
15
5.4 Different Worlds - Mitigation Costs and Strategies 115
The elasticity of substitution of the first CES-level has a huge impact on mitigation
costs. The higher the elasticity of substitution between labour, capital and energy,
the higher the mitigation costs. For σY= 1.2, mitigation cost amount to 3.1%
compared to 1.3% for the default scenarios. With increased elasticity of substitu-
tion in the first CES-level capital, labour and energy become very good substitutes.
Because of this, the use of total energy (and capital) is increased to overcome the
fix labour growth path and create a faster growing economy. This results in the
baseline scenario in an increased amount of carbon which has to be reduced in the
policy scenario.
2020 2040 2060 2080 2100
0
200
400
600
800
1000
1200
Year
Primary energy consumption [EJ]
Biomass
Solar
Wind
Hydro
Geotherm.
Uranium
Coal
Nat. Gas
Oil
(a) σS= 1.5
2020 2040 2060 2080 2100
0
200
400
600
800
1000
1200
Year
Primary energy consumption [EJ]
Biomass
Solar
Wind
Hydro
Geotherm.
Uranium
Coal
Nat. Gas
Oil
(b) σS= 0.8
Figure 7: Global primary energy consumption for a default (σS= 1.5) and alter-
native (σS= 0.8) policy scenario
The second example for a changed elasticity of substitution focuses on the flexi-
bility of the energy demand sector. Literature (e.g. Acemoglu [1]) constitutes that
the impact of efficiency variations depends on the elasticity of substitution of all
related CES-levels. This effect gets more important when some sectors are based
on completely different technologies and energy sources and hence the possibility
to substitute production factors by each other is restricted. Even worse, if there are
lock-in-effects or path dependencies, fast changes of these technologies will not be
possible (Weyant[34], Arthur[2]). This yields in a low elasticity of substitution.
To analyse the impact of a lower flexibility of the energy demand sector due to a
reduced elasticity of substitution, additional experiments are run where the elastic-
ity of substitution between electricity and stationary non-electric energy σS= 0.8
16
116 Chapter 5 Dimensions of Technological Change
instead of σS= 1.5. A low elasticity of substitution between these production fac-
tors is in line with other hybrid models like WITCH (σS= 0.5) (see Bosetti[4]).
Especially in the stationary energy sector, lock-in-effects are a plausible scenario
because of the complex infrastructure of this sector (Unruh[32]).
The elasticity of substitution of the stationary energy sector shows a huge impact
on mitigation costs (2.1% instead of 1.3% for the default scenarios). This effect can
be explained due to significant changes in the energy mix. Figure 7(b) shows the
global primary energy consumption for the policy scenario with reduced elasticity
of substitution compared to the default policy scenario (see Figure 7(a)). Because
of the limited possibility of substituting stationary non-electric energy by electric-
ity, more expensive technologies like geothermal and natural gas with CCS are
used in a more intensive way for the production of low carbon energy in the sta-
tionary non-electric energy sector, if σS= 0.8.
4.1.2 Dimensions of Technological Change and Elasticity of Substitution
As shown, the elasticity of substitution plays an important role for the technology
choice and therefore the production factor use. The question arises what are the
impacts of the dimensions of technological change (dynamics, directions) within
an alternative scenario world which assumes different elasticities of substitution?
Within this subsection, the sensitivity analysis about the impact of efficiency pa-
rameters is rerun for the previous two examples of alternative elasticities of substi-
tution (σY= 1.2instead of 0.5and σS= 0.8instead of 1.5).
Table 2 shows the percentage change of mitigation costs between the reference
scenarios and the scenarios with efficiency variation of total energy, of labour and
of all production factors simultaneously (i.e. dynamics’ variation). In contrast to
the results with σY= 0.5, changes in the efficiency of labour and total energy
affect the mitigation costs in the same direction, if σY= 1.2: The more efficient
the production factors, the higher is the general growth path, the more emissions
have to be reduced and the higher are the mitigation costs. The role of total energy
efficiency changes because in a world where total energy, labour and capital are
17
5.4 Different Worlds - Mitigation Costs and Strategies 117
good substitutes, an increased total energy efficiency results in an increased use of
this production factor. The exogenous labour growth path is no longer a limiting
factor.
Table 2: Percentage change of global mitigation costs depending on the variation
of the following efficiency parameter if σY= 1.2
Parameter 0.6 0.8 1.2 1.4
Dynamics -23.20 -11.75 +11.42 +23.02
Total energy efficiency -16.58 -8.17 +6.38 +12.84
Labour efficiency -9.17 -4.10 +3.76 +8.18
Table 3 shows the percentage change of mitigation costs for the variation of the ef-
ficiency of electricity and stationary non-electric energy. The impact of efficiency
in electricity use is lowered while the formerly meaningless efficiency of station-
ary non-electric energy becomes nearly as important as the other energy related
efficiency variations. Because electricity is not intensively used for substituting
stationary non-electric energy, if σS= 0.8, the efficiency of electricity use is no
longer a key parameter for influencing mitigation costs. On the contrary, a de-
crease of the efficiency of stationary non-electric energy increases mitigation costs
substantially, e.g. by 21% for a 40% efficiency growth reduction.
Table 3: Percentage change of global mitigation costs depending on the variation
of the following efficiency parameter if σS= 0.8
Parameter 0.6 0.8 1.2 1.4
Efficiency of electricity +13.23 +5.47 -4.13 -7.37
Efficiency of stat. non-elec. energy +21.54 +8.76 -6.44 -11.25
These results show that in general the flexibility of the energy use is one of the most
important aspects for low mitigation costs. This flexibility might be realized due
to a high elasticity of substitution or high energy efficiencies, especially for sectors
which can hardly be decarbonized. To create the basic conditions for such a flex-
ible world with low mitigation costs is a challenge for climate change mitigation
policies.
18
118 Chapter 5 Dimensions of Technological Change
4.2 Technology Options
4.2.1 Reference Scenario Setting
The following part of this study looks at another kind of alternative worlds - policy
scenarios where the use of some technologies is restricted. The flexibility of the
energy system is represented by the availability of technologies.
First, the question is answered what are the most important technologies to meet
the emission budget constraint? A few experiments are run in which the supply
from the following technologies is restricted to the baseline use:
•nonuc - nuclear energy
•noccs - all technologies with CCS
•nobio - all biomass technologies
•noren - all renewable energies
Within the default parameter, setting the use of renewable energies seems to be the
most important technology in a stabilization scenario. The restriction of nuclear
energy results only in a small increase of mitigation costs. The whole ranking of
technologies (see figure 8(a)), beginning with the least important one, is: nuclear
energy, technologies with CCS, biomass technologies and renewable energy tech-
nologies.
−5 0 5 10
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
def.
nonuce
noccs
nobio
noren
(a) Mitigation costs
noren nobio noccs nonuc def.
0
5
10
15
20
25
Cumulative primary energy [ZJ]
Fossil Fuels w/o CCS
CCS Fossil
Renewables w/o Biomass
Biomass w/o CCS
CCS Biomass
Nuclear
(b) Mitigation strategies
Figure 8: Technology options default
19
5.4 Different Worlds - Mitigation Costs and Strategies 119
Figure 8(b) shows the cumulative primary energy mix for the default policy sce-
nario and the policy scenarios with restricted use of some technologies. The most
drastically impact on the mitigation strategies has the norenew scenario where a
big share of energy based on CCS technologies is needed to achieve the climate
target.
4.2.2 Dimensions of Technological Change and Technology Options
Within this subsection, the question is addressed whether the impact of the dimen-
sions of technological change will vary, if the use of technologies is restricted to
baseline use. If the production structure of a sector mainly depends on a single
carbon-free technology and the availability of this technology is restricted, the ef-
ficiency parameter of this sector will likely play a key role for climate policies.
The most important technologies for low mitigation costs are based on renewable
energies. Figure 9 shows the results for labour efficiency growth multiplied by 1.4
on the left side and multiplied by 0.6 on the right side. Apart from renewable tech-
nologies, technologies which use CCS or biomass are highly cost-relevant. This
also holds for the variation of the energy-related efficiency parameters.
−5 0 5 10 15
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
1.4
nonuc
noccs
nobio
noren
(a) Labour efficiency high (1.4)
−2 0 2 4 6
World
USA
EUR
CHN
INA
ROW
mitigation costs (% of BAU consumption)
0.6
nonuc
noccs
nobio
noren
(b) Labour efficiency low (0.6)
Figure 9: Technology options for labour efficiency variation
While it is shown that the world-wide ranking of technologies remains the same as
in the default scenario for the variation of the dimensions of technological change,
20
120 Chapter 5 Dimensions of Technological Change
the question will be answered in the following whether mitigation costs and strate-
gies change due to efficiency growth variation depending on the world scenario
assumed.
The percentage differences between the reference mitigation costs for the different
technology worlds (i.e. default, nonuc, noccs, nobio, norenew) and the mitigation
costs within these worlds with an additional variation in the dimensions of techno-
logical change are presented in table 4.
Table 4: Percentage change of global mitigation costs depending on the variation
of technological change dimensions compared to the corresponding reference sce-
narios of technological change
Dimension default nonuc noccs nobio norenew
1.) Dynamics high -4.71 -5.22 +1.10 -4.41 -5.22
2.) Dynamics low 11.04 12.01 -1.58 -1.49 +0.79
3.) Labour efficiency high +25.02 +24.65 +38.86 +25.78 +20.24
4.) Labour efficiency low -23.10 -21.79 -38.29 -40.70 -38.29
5.) Total energy efficiency high -23.12 -23.11 -27.68 -26.66 -25.29
6.) Total energy efficiency low +52.65 +52.53 +61.51 +50.44 45.14
7.) Eff. of transport energy high -10.66 -10.46 -19.57 -29.37 -25.96
8.) Eff. of transport energy low +27.01 +26.61 +54.84 +45.90 +32.93
9.) Eff. of electricity high -11.97 -11.68 -6.76 -6.42 -8.21
10.) Eff. of electricity low +23.40 +23.26 +14.78 +0.12 +3.34
Although the sign of the change of mitigation costs for each type of efficiency
variation stays the same in each technology world (one line in table 4), the impact
intensity of variation of technological change dimensions on mitigation costs de-
pends on the technology portfolio.
In a world without using of CCS technologies, the efficiency in the transport sector
plays a major role. A reduction of this efficiency to 60% results in an increase of
mitigation costs of +55% instead of 27% in a default reference world. If the use of
biomass is restricted, a reduced efficiency of electricity will have a minor impact
(+0.12%) on mitigation costs in contrast to the default reference world. There are
modest differences between default first best scenarios and nonuc scenarios for all
dimensions of technological change (first and second column). The impact of total
energy efficiency is huge but nearly constant over default and technology scenarios
21
5.4 Different Worlds - Mitigation Costs and Strategies 121
(line five and six).
5 Conclusions
This paper shows that mitigation costs and strategies are quite sensitive to the dy-
namics and especially the direction of technological change represented by changes
of production factors’ efficiency. Most essential is the ratio between growth as-
sumed for labour efficiency and some energy efficiencies. As the impact of both
types of efficiency is in opposite direction, a simultaneous increase/decrease of
both has a moderate effect only. Increasing/decreasing the efficiency parameter of
only one type of these production factors, however, changes the costs substantially.
A decrease of labour efficiency to 60% of the default value reduces the mitigation
costs by 23%, while a decrease of energy-related efficiency parameters reduces the
mitigation costs by 2-53%. This impact is at the upper range of what is reported in
literature for other variations in model assumptions ( e.g. variation of technology
availability or climate target).
The mitigation strategies are subject to moderate changes when varying the direc-
tion of technological change. Mainly in scenarios with an increased total energy
demand, the share of renewable technologies is increased in climate policy scenar-
ios.
In follow-up experiments, the sensitivity analysis was rerun within alternative sce-
nario worlds characterized by modified elasticities of substitution or restricted
availability of energy technologies. It turns out that the sign and intensity of the
impacts of technological change are influenced by the scenario world assumed. If
the macro-economic production factors of the first CES-level are very good sub-
stitutes, the impact of both total energy efficiency variation and the dynamics of
technological change will affect mitigation costs in the opposite direction as in the
default scenarios. The higher the labour efficiency and/or total energy efficiency,
the higher are the mitigation costs.
Mitigation cost estimates and mitigation strategies also depend on the elasticity of
substitution between the production factors in the energy end-use sector. In scenar-
ios with a lowered elasticity of substitution in the stationary energy sector, more
22
122 Chapter 5 Dimensions of Technological Change
expensive low carbon technologies are used. The rerun sensitivity analysis shows
that the formerly meaningless efficiency variation of stationary non-electric energy
becomes more important for mitigation costs. This demonstrates the prominent
role of flexibility of the energy demand sector, indicated either by high efficiency
or by high elasticity of substitution.
It is finally shown that the impact intensity of parameter variations on mitigation
costs depends on the set of available technologies. In a world with restricted use
of technologies which use biomass, the efficiency of electricity plays a minor role
than in a world without any technology restrictions. The efficiency of the transport
sector becomes more important when neglecting the use of CCS technologies.
For climate policies, given the findings of this study, it is essential to support in-
vestments in energy efficiency improvements and facilitate structural changes that
result in higher substitution possibilities between end-use energy technologies. To
avoid high mitigation costs, also a diversification of technologies in the energy
production sector is essential.
While this study focuses on the impact of exogenous technological change, its
findings detected further research questions. The identified huge sensitivity of mit-
igation costs estimates to the level and especially the direction of technology im-
provements requests for an endogenous formulation of technological change. First
results of modeling the impact of endogenous technological change on climate
policies are presented for example by Griffith [16], Edenhofer [9] and Bosetti [5].
With endogenous growth, for instance based on R&D investments, new framework
conditions are set up and new findings might be generated. In such a framework
applied to REMIND-RS, the costs for efficiency increase have to be paid by the
regions, while the social planner can decide where to increase productivity.
6 Acknowledgements
The authors would like to thank Michael H¨
ubler, Kai Lessmann and Ottmar Eden-
hofer for their contributing comments and discussions.
23
5.5 Conclusions 123
References
[1] Acemoglu, D. (2002). Directed Technical Change. Review of Economic Stud-
ies 69, 781-809.
[2] Arthur, B., (1989). Competing technologies, increasing returns and lock-in
by historic events. Economics Journal 99, 116-131.
[3] B¨
ohringer, C., J. Jensen, T. F. Rutherford, (1999). Energy Market Projections
and Differentiated Carbon Abatement in the European Union. ZEW Discus-
sion Paper No. 99-11.
[4] Bosetti, V., C. Carraro, M. Galeotti, E. Massetti, M. Tavoni, (2006). WITCH:
A World Induced Technical Change Hybrid Model. The Energy Journal, Spe-
cial Issue. Hybrid Modeling of Energy-Environment Policies: Reconciling
Bottom-up and Top-down, 13-38.
[5] Bosetti, V., C. Carraro, M. Galeotti (2006). The Dynamics of Carbon and
Energy Intensity in a Model of Endogenous TEchnological Change. The En-
ergy Journal, Endogenous Technological Change and teh Economics of At-
mospheric Stabilisation Special Issue, 191-205.
[6] Burniaux, J. et al. (1991). GREEN - A Multi-Region Dynamic General Equi-
librium Model for Quantifying the Costs of Curbing CO2 Emissions: A
Technical Manual, OECD Economics Department Working Papers, No. 104,
OECD Publishing.
[7] Chang, K.-P.(1994). Capital-Energy Substitution and the Multi-level CES
Production Function. Energy Economics 10, 22-26.
[8] Dowlatabadi, H. (1998). Sensitivity of climate change mitigation estimates to
assumptions abour technical change. Energy Economicy 20, 473-793.
[9] Edenhofer, O., N. Bauer, E. Kriegler (2004). The Impact of Technological
Change on Climate Protection and Welfare: Insights from the Model MIND.
Ecological Economics, Vol. 54, 277-292.
[10] Edenhofer, O., K. Lessmann, N. Bauer (2006). Mitigation Strategies and
Costs of Climate Protection: The Effects of ETX in the Hybrid Model MIND.
The Energy Journal, Endogenous Technological Change and the Economics
of Atmospheric Stabilisation Special Issue.
24
124 Chapter 5 Dimensions of Technological Change
[11] Edenhofer, O., C. Carraro, J.-C. Hourcade, K. Neuhoff, G. Luderer, C. Flach-
sland, M. Jakob, A. Popp, J. Steckel, J. Strohschein, N. Bauer, S. Brunner,
M. Leimbach, H. Lotze-Campen, V. Bosetti, E. de Cian, M. Tavoni, O. Sassi,
H. Waisman, R. Crassous-Doerfler, S. Monjon, S. Dr¨
oge, H. van Essen, P. del
R´
ıo, A. T¨
urk (2009). RECIPE - The Economics of Decarbonization. Synthe-
sis Report.
[12] Edenhofer, O., Knopf, B., Barker, T., Baumstark, L., Bellevrat, E., Chateau,
B., Criqui, P., Isaac, M., Kitous, A., Kypreos, S., Leimbach, M., Lessmann,
K., Magn´
e, B., Scrieciu, S., Turton, H., van Vuuren, D.P. (2010): The eco-
nomics of low stabilisation: exploring its implications for mitigation costs
and strategies. The Energy Journal, Volume 31 (Special Issue 1). The Eco-
nomics of Low Stabilization, 2010, in press.
[13] Fellner,W. (1962). Does the Market Direct the Relative Factor-saving Effects
of Technological Progress?. Princeton University Press, 171-194.
[14] Gerlagh, R., van der Zwaan, B. (2004). A Sensitivity Analysis of Timing and
Costs of Greenhouse Gas Emissions Reduction. Climatic Change 65, 39-71.
[15] Gillingham, K., R.G. Newell, W.A. Pizer, (2008). Modeling Endogenous
Technological Change for Climate Policy Analysis. Energy Economics 30,
2734-2753.
[16] Griffith, R., S. Redding, J. Van Reenen (2003). R&D and Absorptive capacity:
Theary and Empirical Evidence. Scandinavian Journal of Economics, 105 (1),
99-118.
[17] Hazilla, M., and Kopp, R.J. (1986). Testing for Separable Functional Struc-
ture Using Temporary Equilibrium Models. Journal of Econometrics 33, 119-
141.
[18] Kaya, Y. (1990) Impact of Carbon Dioxide Emission Control on GNP
Growth: Interpretation of Proposed Scenarios, Paper presented to the IPCC
Energy and Industry subgroup, Responses strategies working group, Paris
(mimeo).
[19] Kemfert, C., H. Welsch (2000) Energy-Capital-Labour substitution and the
economic effects of CO2abatement: Evidence for Germany. Journal of Pol-
icy Modeling 22(6), 641-660.
[20] Leimbach, M., N. Bauer, L. Baumstark, and O. Edenhofer (2010). Mitiga-
tion costs in a globalized world: climate policy analysis with REMIND-R.
Environmental Modeling and Assessment 15, 155-173.
25
5.6 References 125
[21] L¨
oschel, A. (2002). Technological change in economic models of environ-
mental policy: a survey. Ecological Economics 43, 105-126.
[22] Manne, A.S., R.G. Richels (1990). The Costs of Reducing U.S. CO2 Emis-
sions - Further Sensitivity Analyses. The Energy Journal 11, 69-78.
[23] Manne A., R. Mendelsohn, R. Richels (1995). MERGE. A model for evalu-
ating regional and global effects of GHG reduction policies. Energy policy,
Vol.23, 17-34.
[24] Meinshausen, Malte, N. Meinshausen, W. Hare, S. C.B. Raper, K. Frieler, R.
Knutti, D. J. Frame, M. R. Allen(2009). Greenhouse-gas emission targets for
limiting global warming to 2 ◦C. Nature 458, 1158-1162.
[25] Meyer, A. (2004). Briefing: Contraction and Convergence. Engineering Sus-
tainability (157). Issue 4, 189-192.
[26] Nordhaus, W. D. (1993), Optimal Greenhouse-Gas Reductions and Tax Pol-
icy in the ”DICE” Model.The American Economic Review, Vol. 83, No. 2,
Papers and Proceedings of the Hundred and Fifth Annual Meeting of the
American Economic Association (May, 1993), 313-317
[27] Nordhaus, W. D. (2002). Modelling Induced Innovation in Climate-Change
Policy. In: A. Gr¨
ubler, N. Nakicenovic and W. D. Nordhaus (eds), Techno-
logical Change and the Environment. Resources for the Future, Washington
DC, USA, 182-209.
[28] Riahi, K., A. Gr¨
ubler, N. Nakicenovic (2007). Scenarios of long-term socioe-
conomic and environmental development under climate stabilization. Tech-
nological Forecasting and Social Change 74, 887-935.
[29] Rogner, H.-H., D. Zhou, R. Bradley. P. Crabb´
e, O. Edenhofer, B.Hare, L.
Kuijpers, M. Yamaguchi (2007). Introduction. In Climate Change 2007: Mit-
igation. Contribution of Working Group III to the Fourth Assessment Report
of the Intergovernmental Panel on Climate Change [B. Metz, O.R. David-
son, P.R. Bosch, R. Dave, L.A. Meyer (eds)], Cambridge University Press,
Cambridge, United Kingdom and New York, NY, USA.
[30] Sanstad, A., J. Roy, J.A. Sathaye (2006). Estimating energy-augmenting tech-
nological change in developing country industries. Energy Economicy 28,
720-729.
26
126 Chapter 5 Dimensions of Technological Change
[31] Steckel, Jan, Michael Jakob, Robert Marschinski, Gunnar Luderer (2010):
From carbonization to decarbonization? - Past trends and future scenarios for
China’s CO2 emissions. Paper presented at the International Energy Work-
shop 2010 in Stockholm, Sweden
[32] Unruh, G.C., (2000). Understanding carbon lock-in. Energy Policy 28 (12),
817-830.
[33] Watanabe, C. (1991) Trends in the substitution of production facotr to tech-
nology - empirical analysis of the inducing impact of theenergy crisis on
Japanese industrial technology. Research Policy 21, 481-505.
[34] Weyant, J., T. Olavson (1999). Issues in Modeling Induced Technological
Change in Energy, Environmental, and Climate Policy. Environmental Mod-
eling and Assessment 4, 67-85.
27
5.6 References 127
128 Chapter 5 Dimensions of Technological Change
Chapter 6
Endogenous Sector-specific R&D Investments into
Energy Efficiency as Mitigation Option.∗
Lavinia Baumstark
Marian Leimbach
∗submitted to Climatic Change Baumstark, L., M. Leimbach (2011): Endogenous Sector-specific R&D
Investments into Energy Efficiency as Mitigation Option.
129
130 Chapter 6 Endogenous R&D Investments into Energy Efficiency
Endogenous Sector-specific R&D Investments into
Energy Efficiency as Mitigation Option
Lavinia Baumstark∗
, Marian Leimbach
March 30, 2011
Abstract
In climate change mitigation analysis the influence of technological change
is strongly discussed. This paper deals with the question, when and where
energy efficiency enhancing R&D investments play an important role as mit-
igation option. Therefor the hybrid model REMIND-RS is extended by a
new formulation of endogenous sector-specific technological change. Simu-
lation results show that mainly efficiency improvements in the transport and
electricity sector play an important role for climate change mitigation. Re-
allocation of investments form labour efficiency into the efficiency of energy
related production factors helps significantly to reduce climate change miti-
gation costs, especially if the amount of technological options is reduced.
JEL classification: E27, O11, O33, O41, Q43, Q54
keywords: climate change mitigation, endogenous efficiency improvements,
R&D investments, endogenous growth
∗PIK - Potsdam Institute for Climate Impact Research, P.O. Box 60 12 03, D-14412 Potsdam,
Germany, Tel. ++49/331/288-2535, e-mail: [email protected]
1
131
1 Introduction
Climate change is one of the main challenges to deal with in the next century. In
order to avoid dramatic damages due to significantly increased global mean tem-
perature, global carbon emissions have to be reduced. In the literature there are
discussed a few mitigation options to meet this task: (i) restructuring the energy
production, (ii) creating a carbon free backstop-technology, (iii) production factor
substitution, (iv) energy efficiency improvements. Many paper analyse the possi-
bilities and conditions of the first three options. Fischer ([12]) find that an optimal
portfolio of policies for reducing CO2emissions will include an emissions price
and subsidies for technology R&D and learning.
This paper focuses on the fourth option: efficiency improvements, in particular
due to R&D investments. As Himmelberg [23] and Weyant [42] indicate, R&D
investments have a large impact on energy efficiency development. A mechanism
for endogenous efficiency development is introduced in an Integrated Assessment
Model to analyse the impacts and constraints of investments into production factor
efficiencies, . Thus the question can be answered how to allocate R&D invest-
ment when carbon emissions have to be reduced in a climate policy regime. It
is expected to be helpful to redirect investments from total factor productivity or
labour efficiency to energy related efficiencies, because with a higher efficiency a
reduced energy input helps lowering carbon emissions. Gillingham [14] names it
as a feature of multi-sector models, to provide insights on the effects of interactions
between sectors, such as spillovers - or crowding out - from R&D.
Implementing R&D investments into labour productivity allows to model endoge-
nous economic growth. Most of the models of endogenous growth described in the
literature represent macro-economic models which ignore the production factor en-
ergy or use a simple formulation of it. For a detailed analysis of climate change
mitigation options it is important to look at the interaction between restructuring
the energy production and increasing production factor efficiency due to R&D in-
vestments. Therefore a model with a detailed energy system and macro-economic
part is necessary to answer the question if and when energy improvements trig-
gered by R&D investments represent a key climate change mitigation option. This
paper applies such a model - REMIND-RS - which provides the possibility of en-
2
132 Chapter 6 Endogenous R&D Investments into Energy Efficiency
dogenous R&D investments into the efficiency of labour and as well as different
end-use energy types.
This paper is structured as follows: Section 2 describes endogenous technological
change as growth factor and how energy efficiency is modeled in the literature.
The overall structure and the calibration of the R&D function of REMIND-RS are
explained in section 3, while the results of reference scenarios are shown in section
4. Experiments with restricted investments into energy sectors and technology
restrictions are discussed in section 5. After a sensitivity analysis in section 6,
section 7 concludes the paper.
2 Endogenous Technological Change
2.1 Endogenous Growth
In contrast to the classical growth theory, endogenous growth theory tries to ex-
plain technical innovation endogenously. Important literature like Romer [39] and
Grossman and Helpman [20] analyses the crucial question about the key drivers
for regionally differentiated economic growth and about an adequate design of en-
dogenous growth within a macro-economic model. The main challenge is to find a
functional form to describe the impact of knowledge/human capital on the produc-
tivity of a related economy. Within the literature, mainly two ways of implementing
endogenous growth into macroeconomic models are described. On the one hand
the increase of total factor productivity may be interpreted as a result of invest-
ments in the knowledge pool of a country. Translating this to the productivity of
one production factor means increasing factor efficiency due to R&D investments.
On the other hand an additional production factor ”knowledge stock/human cap-
ital” may be introduced. Then the investment decisions determine the amount of
this production factor (knowledge stock).
There are a few economic models, that implement endogenous growth and tech-
nological change by introducing an additional production factor that represents
knowledge. ENTICE [36] is an advanced version of the model DICE with the
3
6.2 Endogenous Technological Change 133
same production function. Technological change comes through changes in the
total factor productivity. In addition, effective energy input is modeled as a com-
bination of carbon based fossil fuels and energy related human capital. A scaling
factor determines the level of energy savings resulting from new energy knowl-
edge.
The paper from Bosetti et all. [5] about the hybrid model WITCH deals with inter-
national energy R&D spillover. Following Popp [36], technological advances are
captured by a stock of knowledge and combined with energy in a CES function.
This new production formulation simulates the production of energy services de-
manded by the final good production sector.
The model REMIND-RS, used for the analysis of this paper, is based on the al-
ternative way of modeling endogenous growth via production efficiency instead
of introducing a new production factor. Investments can be spent to improve the
efficiency of labour or end-use energy sectors. Due to policy interventions, these
investments may be directed into different sectors.
Some empirical paper tried to identify evidence for variables (e.g. trade liberaliza-
tion) explaining growth. Therefore endogenous productivity change is introduced
into a model via defining a parameter Afor total factor productivity in the produc-
tion function Y=A·F(L, T). The output Yis produced by the production factors
Land T, e.g. labour and land. Grossman and Helpman [19] consider a two-sector,
two-factor economy and interpret Aas the stock of knowledge capital.
In trying to explain why some low-income countries show convergence whereas
others do not, Greiner and Semmler [16] analysed a growth model with a Coub-
Douglas production function where the knowledge capital raises the labour pro-
ductivity and itself is affected by education efforts.
Within the Integrated Assessment model MIND [8] investments into a R&D sec-
tor increase efficiency parameters. In an additional paper Edenhofer et.al. [9] run
a sensitivity analysis of the labour and energy efficiency among other important
parameter. They find evidence for the influence of R&D investments on macro
economic growth as well as the mix of mitigation options.
Griffith et.al.[17] discuss the theoretical literature on Schumpeterian endogenous
growth and the empirical literature on R&D. Within a theoretical multi-regional
4
134 Chapter 6 Endogenous R&D Investments into Energy Efficiency
model with three sectors, the authors analyze the impact of absorptive capacity on
R&D investments. The production function of the final good is a Cobb-Douglas
function of intermediate inputs and sector specific capital. A multiplied parameter
denotes the productivity or quality of intermediate inputs. Productivity changes of
intermediate inputs depend on the ratio between the most advanced regional pro-
ductivity and the regional productivity in the initial time step.
Bosetti et. al. [4] extended a version of RICE 99 by a formulation of the re-
lationship between technological change and both Learning-by-Researching and
Learning-by-Doing named FEEM-RICE v.3. They introduce an Energy Technical
Change Index (ETCI) which is defined as a convex combination of the stocks of
knowledge and abatement.
2.2 Energy Efficiency Improvements
Looking at climate change mitigation from a perspective of an energy system
model, the focus lies on technological change. L¨
oschel [30] differentiates be-
tween bottom-up and top-down models and an exogenous vs. endogenous for-
mulation of technological change. The exogenous formulation is often represented
by aggregated parameters, for example AEEI - the Autonomous Energy Efficiency
Improvement- as described by Sanstad [40]. Gerlagh and van der Zwaan [13] in-
cluded the parameter autonomous energy service efficiency improvement (AESEI)
in their sensitivity analysis of mitigation costs.
There are only a few models that represent energy efficiency development endoge-
nously. Das Gopal and Powell [15] model energy efficiency improvements based
on embodied technological spillover in the CGE model GTAP. Therefor the authors
build on the production technology tree in the GTAP model. Each production level
is described as a CES function with a technological progress parameter.
Crassous et.al. [6] envisage endogenous technological change in the recursive dy-
namic general equilibrium model Imaclim-R to study how modeling induced tech-
nological change affects costs of CO2stabilization. In this model energy efficiency
is a function of historical investments as well as variations of relative prices.
When focussing on technological development, often learning curves are used for
5
6.2 Endogenous Technological Change 135
representing technological change. First Arrow [3] described learning by doing as
a feature that decreases investment costs due to increasing installed capacity of a
technology. Especially in energy system models this approach is used (see Gr¨
ubler
[21] and Rao [38]) and sometimes extended by a learning by research factor in the
learning curves as described by Kypreos [28] 1. Within this paper, the impact of
R&D investment into the efficiency of end-use energy efficiency is analysed based
on REMIND-RS. These investments do not represent the R&D investments into
a single conversion technology but include the whole portfolio of spending into
efficiency, which sum up to the efficiency increase of the end-use energy.
3 The Model REMIND-RS
3.1 Overall Structure
For this study the integrated assessment model REMIND-R is used and upgraded
by implementing new features of endogenous technological change. This hybrid
model couples a macro-economic model, a detailed energy system model and a
simple climate model and is based on the multi-regional model REMIND-R2. For
a detailed description of the model see Leimbach [29]. REMIND-RS includes the
complete energy system module and climate module of REMIND-R but has been
scaled down from 11 to 5 regions. These world regions are:
•USA - USA
•EUR - EU27
•CHN - China
•INA - Developing regions, less resources
•ROW - Mainly resource exporting regions.
The region INA includes India, Sub-Saharan Africa, Latin America and Other Asia
(Central and South East Asia), while Middle East and North Africa, Japan, Russia
and the rest of the world (Canada, Australia, etc.) are part of the region ROW.
1REMIND-RS assumes learning by doing for wind and solar.
2The equations are explained at http://www.pik-potsdam.de/research/research-
domains/sustainable-solutions/esm-group/remind-code.
6
136 Chapter 6 Endogenous R&D Investments into Energy Efficiency
REMIND-RS maximizes a global welfare function and computes an intertemporal
optimal regional social planner solution for the time span 2005-2100. The param-
eterisation of REMIND-RS is based on the version REMIND-R 1.3 used in the
project RECIPE ([10]), but with updated data for the solar and wind potential and
including adjustment costs for fast extension of technologies.
The production function of REMIND-RS is represented by a nested CES (constant
elasticity of substitution) function of the general structure
OUTt,r =ÃX
i²CES
(AINi,t,r ·INi,t,r)ρOUT,r !1/ρOUT,r
.(1)
The output OUT for each region rby each time step tis the sum of the produc-
tion factor inputs IN multiplied by the associated efficiency parameter AINito the
power of ρOUT = (σOUT −1)/σOUT , where σOUT is the elasticity of substitution
to produce the output OUT. The list CES assigns the associated inputs IN to
each output OUT. The whole nested CES tree is shown in Figure 1. All outputs
(intermediate outputs and GDP) in the CES-tree represent monetary values. In the
first level GDP is the output of a CES function of the production factors capital,
total energy and labour. The total energy itself is produced by a CES function,
whose input factors are CES function outputs again (i.e. stationary and transport
energy). Stationary energy is the result of a CES production function of electric-
ity and stationary non-electric energy while transport energy is a combination of
liquid fuels and hydrogen. Sector-specific final energy types represent the bottom
end of the ’CES-tree’.The elasticities of substitution σOUT are lower than one in
the first two CES-levels, while the production factors of the lower CES-levels are
assumed to be substitutes (i.e. σOUT >1). These values are comparable to the
values assumed by Gerlagh [13].
In the new formulation of REMIND-RS, used in this paper, the efficiency develop-
ment of the following production factors is modeled endogenously and depends on
the decision of R&D investments of the social planner: labour L, transport energy
T, electricity EL and stationary non-electric energy SNE. The productivities of
the other production factors are assumed to keep constant to the calibrated value in
7
6.3 The Model REMIND-RS 137
Figure 1: Nested CES structure of the macro-economic module of REMIND-RS
2005. These factor productivities in the first period of the model are calculated to
balance the input- and output-values from empirical data.
Beside the endogenous efficiency improvements, REMIND-RS has a second chan-
nel for endogenous technological change: global learning-by-doing. The more
capacity of solar and wind is installed, the lower are the investment costs.
3.1.1 The R&D Function used in REMIND-RS
The efficiency variables AIN,t,r are subject to R&D investments RDIN,t,r de-
scribed by
AIN,t+1,r =µ1 + γIN,r µRDIN,t,r
Yt,r αIN,r AIN,t,r (2)
where αIN,r represents the elasticity of productivity growth on R&D investments
and γIN,r is the efficiency of R&D investments for each region rand production
8
138 Chapter 6 Endogenous R&D Investments into Energy Efficiency
sector IN. This formulation is similar to the R&D function, described by Jones
and Williams [27]. For REMIND-RS, it is assumed, that there are no innovation
clusters but a complete intertemporal knowledge spillover. H¨
ubler [24] argues that
this formulation also follows the Schumpeterian view based on the description by
Aghion and Howitt [2]. In addition, the R&D investments in REMIND-RS are nor-
malized by the GDP. Such a R&D function is used in the global model MIND ([8])
for labour and total energy efficiency. This formulation guarantees in a regional-
ized model that the same share of R&D investments on GDP results in the same
productivity increase in each region for the same parameter set of γIN,r and αIN,r.
Zachariadis [45] refers to Aghion and Howitt [1], who suggest that R&D intensity
(per GDP) is the proper empirical measure for the R&D input of the innovation
function in the context of endogenous growth models without scale effects.
The sum of all investments into production factor efficiency is financed by the pro-
duced output Yt,r and hence part of the budget equation Yr+Mr=Cr+Ir+
ESMr+Pin RDin,r that holds for each time step t. The sum of the regional con-
sumption Cr, investments Irinto capital stock, energy system costs EMSrand all
R&D investments equal the output Yrplus the net import of the composite good
Mr.
3.2 Calibration and Data
The literature about the influence of R&D investments on GDP growth finds that
R&D investment has a significant positive effect on the growth rate of total factor
productivity (Griliches [18]). However, calibration and initialization of the param-
eter of the R&D function is a key challenge.
In the model REMIND-RS, it is assumed that the parameter αIN,r is not constant
over time as it is common in other models, but depends on the level of efficiency of
a region AIN,t,r. Popp [35] finds that the quality of existing knowledge have strong
effects on innovation and Doraszelski ([7]) argues, that in most cases the impact of
current R&D on future productivity depends crucially on current productivity. The
long-run elasticity of labour related R&D investments is assumed to be αL,r = 0.1
in REMIND-RS. This is in line with empirical findings, e.g. Guellec [22] gets a
long-term elasticity of business R&D of 0.132. But the elasticity is not constant
9
6.3 The Model REMIND-RS 139
over time. In a region, which is on a very low technological level, R&D invest-
ments beyond the optimal level are expected to have a low marginal value, i.e. the
absorbtion capacity to employ additional R&D investments has still to be build up.
To emulate this situation, for less developed regions the elasticity of R&D invest-
ments is assumed to be small and increases with the level of labour efficiency. This
is supported by Wieser [43], who finds indication that the elasticity of R&D in-
creases over time. The assumed functional form of αL,r in REMIND-RS is shown
in figure 2(a).
The long-run elasticity of energy related R&D investments in REMIND-RS is as-
sumed to be αE,r = 0.5. That is higher than the used/choosen elasticity of labour
related R&D investments. Jamasb [26] analysed the elasticity of learning by re-
search and gets values between 0.02 and 0.8, depending on the development stages
of technologies. Mainly capital intensive technologies like nuclear, wind and solar
show a high R&D elasticity. Popp [35] calculates a long-run energy elasticity of
0.42. A sensitivity analysis about the assumptions on the elasticity of R&D invest-
ments is presented in section 6.
The calibration of γIN,r is based on the assumed functional relationship for αIN,r
and is described by
γIN,t,r =³AIN,t+1,r
AIN,t,r ´−1
³RDIN,t,r
Yt,r ´αIN,r .(3)
γIN,t,r is calculated for the time span 1996-2006 based on empirical data for
AIN,t,r,Yt,r and RDIN,t,r.γIN,r is then the arithmetic mean of γIN,t,r. To calcu-
late γIN,t,r, the efficiency growth of all modeled sectors for all regions would be
used. Because of the definition of the efficiency of a sector as OUT/IN, the GDP
(Y), labour (L), transport energy (T), electricity (EL) and stationary non-electric
energy (SNE) path is needed for all regions . But there is a lack of data for sector
specific energy consumption for some regions of REMIND-RS, so the calibration
of the R&D function for transport energy, electricity and stationary non-electric
energy is based on total energy efficiency. GDP development, population (labour)
and total energy for the last decades are used from WDI2010 ([44]). General R&D
expenditures (used within the calibration as representation for labour efficiency
R&D) are also documented in WDI2010 ([44]) from 1996 on. Data on energy
10
140 Chapter 6 Endogenous R&D Investments into Energy Efficiency
related R&D investments are rare and mainly available for industrialized regions
(IEA2007 [25]).
0 10 20 30 40
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
GDP/Labour
coef. γ, exp. α
coef. γ
exp. α
(a) Labour
0 10 20 30 40
0
0.1
0.2
0.3
0.4
0.5
GDP/Total Energy
coef. γ, exp. α
coef. γ
exp. α
(b) Total energy
Figure 2: Parameter of the R&D function for labour and total energy efficiency
Calibration results suggest that like αIN,r,γIN,r is not constant over time, but can
be described for all regions by a function depending on the efficiency level. The
function for the efficiency of labour related R&D investments indicate, that the
region China is on a catching up process with a high efficiency parameter γIN,r,
while industrialized regions - like Europe and USA now - have a long-run R&D
investment efficiency of 0.022 for labour related R&D investments. The used func-
tion γL,r depending on the labour efficiency level is shown in figure 2(a). The cir-
cles stand for the calibrated initiative values γL,r for the five regions of REMIND-
RS based on the empirical data.
Two functions γIN,r are calculated: one for R&D investments in labour efficiency
γL,r, and one for all energy related R&D investments γT,r =γEL,r =γSNE,r
based on empirical total energy efficiency (i.e. Y/E). In contrast to the function
of the efficiency parameter for labour related R&D investments, the efficiency for
energy related R&D investments is increasing in the efficiency level and reaches
a value of 0.25 in the long-run for industrialized regions. The resulting functional
relationship γT/EL/SNE,r depending on the efficiency level is shown in figure 2(b).
In the model experiments the same function for the parameters for the energy re-
lated sectors are used, but the concrete value depends on the R&D investments of
11
6.3 The Model REMIND-RS 141
the regions into the respective sector.
In REMIND-RS the region INA plays a special role. It is a heterogenous region,
including very low developed countries. It is assumed that INA will not catch up
as fast as China do. Therefore all functions for the parameters of the R&D func-
tions are moved to a higher efficiency level. With this transformation, the low
R&D shares of the region INA - as shown by empirical data - are met. The result-
ing growth path in REMIND-RS and the R&D investments into production factor
efficiency are discussed in section 4.
4 Reference Scenarios
Within this section, the results of reference scenarios that include endogenous in-
vestments into production factor efficiency as described in the former section are
shown.
The first reference scenario is a business as usual (BAU) scenario, where the opti-
mal consumption path for all factors and the resulting emissions (see figure 3(a))
are calculated. In such a BAU scenario climate change damage costs are ignored.
2020 2040 2060 2080 2100
0
5
10
15
20
25
Year
Carbon Emissions [Gt]
ROW
CHN
INA
EUR
USA
(a) Emissions
2020 2040 2060 2080 2100
0
100
200
300
400
500
Year
GDP [trill.$US]
ROW
INA
CHN
EUR
USA
(b) GDP
Figure 3: Carbon emissions and GDP for all regions
Figure 3(b) shows the GDP path for each modeled region until the year 2100. The
global GDP increases from 45 trill.$US in 2005 to 470 trill. $US in 2100. The in-
12
142 Chapter 6 Endogenous R&D Investments into Energy Efficiency
dustrialized regions Europe and USA are growing with a slightly decreasing growth
rate o beginning with 3% and ending with 1% in 2100. China starts with a high
GDP growth rate of 10%, which falls until 2040 to the level of the industrialized
regions (see figure 4(b)). The region INA will realize such a catching up process
in the mid of the century. In 2045 the total GDP of INA reaches the GDP of Eu-
rope, in 2055 the one of USA. Nevertheless, in terms of per capita GDP, INA is
still far behind the industrialized regions. With a delayed increase of GDP growth,
it reaches China in 2080 (see figure 4(a)). In the literature about the future GDP
development similar growth path are discussed (PWC2011 [37]).
2020 2040 2060 2080 2100
0
50
100
150
200
Year
Per capita income GDP/L
EUR
USA
ROW
INA
CHN
(a) GDP per capita
2020 2040 2060 2080 2100
0
0.05
0.1
0.15
0.2
Year
GDP growth rate p.a.
EUR
USA
ROW
INA
CHN
(b) GDP growth rate
Figure 4: GDP per capita and GDP growth rate for all regions
The discontinous GDP growth path for some regions in REMIND-RS is induced
by the labour efficiency growth path. This efficiency development depends on the
R&D investments and are influenced by the formulation and parameterisation of
the R&D functions. Figure 5 shows the R&D investment share per GDP for the
regions Europe, USA, China and INA for each sector. The R&D share of the in-
dustrialized regions Europe and USA are around 2.5 % of GDP and in line with
empirical data (WDI2010 [44]). The R&D investments into the energy related sec-
tors are one-fifth of the investments into labour efficiency. Margolis and Kammen
[32] document that the energy sector’s R&D intensity is extremely low in compar-
ison to many other sectors. The developing region INA invests a very low R&D
share in 2005, similar to the sparse data for developing countries (SEI ([41])), but
13
6.4 Reference Scenarios 143
increases its energy R&D investments to a share of 2.5% in 2065. The R&D share
of China is slightly to high in the beginning, compared to empirical data, decreas-
ing until 2030 and increasing again until 2085.
2020 2040 2060 2080 2100
0
0.01
0.02
0.03
0.04
0.05
Year
R&D investments per GDP
Labor
Electricity
Transport energy
Stationary energy
(a) Europe
2020 2040 2060 2080 2100
0
0.01
0.02
0.03
0.04
0.05
Year
R&D investments per GDP
Labor
Electricity
Transport energy
Stationary energy
(b) USA
2020 2040 2060 2080 2100
0
0.01
0.02
0.03
0.04
0.05
Year
R&D investments per GDP
Labor
Electricity
Transport energy
Stationary energy
(c) China
2020 2040 2060 2080 2100
0
0.01
0.02
0.03
0.04
0.05
Year
R&D investments per GDP
Labor
Electricity
Transport energy
Stationary energy
(d) INA
Figure 5: R&D investments per GDP, BAU scenario
The second reference scenario is a scenario with a climate policy strategy (POL).
The cumulated emissions are not allowed to exceed 300 GtC between 2005 and
2100. Meinshausen [34] showed that each emission budget can be associated with
a specific probability to reach a temperature cap. The whole emission budget is
allocated in form of tradable permits to the regions via the Contraction and Con-
vergence scheme as described in Meyer [33]. REMIND-RS calculates the optimal
use of production factors - especially the regional energy production. Thereby it is
14
144 Chapter 6 Endogenous R&D Investments into Energy Efficiency
guaranteed that CO2emissions are reduced first in those regions, where it can be
realized at lowest costs. The calculated carbon emissions in the POL scenario are
demonstrated in figure 6(a). Keeping the climate target means a steep reduction of
emissions as of 2030 and from 2070 on negative emissions.
The costs of climate change mitigation measured as the average percentage dif-
2020 2040 2060 2080 2100
−2
0
2
4
6
8
10
Year
Carbon Emissions [Gt]
ROW
CHN
INA
EUR
USA
(a) Emissions
2020 2040 2060 2080 2100
−2
0
2
4
6
8
Year
USA
EUR
CHN
INA
ROW
World
(b) Mitigation costs
Figure 6: Carbon emissions and mitigation costs for all regions
ference of (regional) consumption between these scenarios amount to 2% of global
consumption differences. The costs for the whole time horizon for each region
are presented in figure 6(b). China faces the highest mitigation costs of 6.2% con-
sumption losses in 2050, while INA gains from selling emission permits and gets
the lowest mitigation costs.
Within a climate policy scenario, energy efficiency plays an important role. To
reduce carbon emissions, the energy production has to be restructured based on low
carbon/carbon free technologies or the use of final energy has to be optimized by
efficiency improvements. In REMIND-RS, climate policy induces investments into
renewable, biomass, CCS or nuclear technologies. Moreover, R&D investments
into the efficiency of energy related production factors are increased compared to
the BAU scenario. Figure 7 shows the difference of R&D investment shares for the
regions Europe, USA, China and INA for all production sectors with endogenous
efficiency development in the POL scenario. Investments into the efficiency of
15
6.4 Reference Scenarios 145
2020 2040 2060 2080 2100
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(a) Europe
2020 2040 2060 2080 2100
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(b) USA
2020 2040 2060 2080 2100
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(c) China
2020 2040 2060 2080 2100
−0.2
0
0.2
0.4
0.6
0.8
1
1.2
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(d) INA
Figure 7: Difference of R&D investment share per GDP in percentage points, POL
scenario
electricity and transport energy use are increased in all regions, and especially in
INA during the mid of the century. In 2065 the R&D share of these sectors amount
to 1.8% compared to 1.2% in the BAU scenario. In the USA around 2020 the
investment into the efficiency of electricity consumption is increased and later in
the century the R&D investments into the efficiency of the transport sector are
intensified.
16
146 Chapter 6 Endogenous R&D Investments into Energy Efficiency
5 Policy Analysis
5.1 Energy Sector Experiments
Within the hybrid model REMIND-RS, it is possible to run sector specific experi-
ments. Due to the nested CES production function structure, R&D investments can
be directed into different sectors. The question arises, within which sectors R&D
investments are most important for climate change mitigation. To figure this out,
some experiments are run, where the R&D investments into one sector are fixed to
baseline investments within a climate policy scenario. The increase of the mitiga-
tion costs denotes the importance of the efficiency of this sector. The R&D sector
experiments are labeled as following:
•POLT- Investments into the efficiency of transport sector are fixed to BAU
•POLEL- Investments into the efficiency of electricity sector are fixed to
BAU
•POLSNE - Investments into the efficiency of stationary non-electric energy
sector are fixed to BAU
•POLL- Investments into labour efficiency are fixed to BAU
•POLE- Investments into all energy sector related efficiencies are fixed to
BAU.
The results of the experiments with REMIND-RS show that mainly the R&D in-
vestments into the transport sector are important for low mitigation costs. A re-
striction of the R&D investments into the transport sector increases global costs
from 1.99% to 2.11%. The transport sector is the sector, which in REMIND-RS is
most difficult to decarbonize by technology options. So the R&D investments in
this sector overcome the lack of carbon free technologies.
The R&D investments into the electricity sector also play a key role for low costs of
climate change mitigation. In this sector the energy production can be easily gener-
ated by low carbon technologies, but electricity is also used to substitute stationary
non-electric energy. Therefore the efficiency of the electricity sector is important
for climate change mitigation and fixing the investments into this sector increases
the costs to 2.12% global consumption losses. Restricting the R&D investments
17
6.5 Policy Analysis 147
2020 2040 2060 2080 2100
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
1.2
Year
USA
EUR
CHN
INA
ROW
World
(a) Transport
2020 2040 2060 2080 2100
−0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Year
USA
EUR
CHN
INA
ROW
World
(b) Electricity
2020 2040 2060 2080 2100
−0.03
−0.02
−0.01
0
0.01
0.02
0.03
Year
USA
EUR
CHN
INA
ROW
World
(c) Stat. non-elec.
2020 2040 2060 2080 2100
−0.3
−0.25
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
Year
USA
EUR
CHN
INA
ROW
World
(d) Labour
Figure 8: mitigation costs for the R&D sector experiments
in the stationary non-electric energy sector increases mitigation costs only by 0.01
percentage points. Restricted investments into labour efficiency yield consumption
losses of 1.95%.
Figure 8 shows the percentage consumption difference between the reference pol-
icy scenario POL and the sector scenarios POLT,POLEL,POLSNE and P OLL
over time. Positive values indicate more consumption in the POL scenario, where
all R&D investments are spent in optimal intensity and timing to keep the climate
target. If R&D investments in the transport sector are fixed to the investments of
the BAU scenario, especially from 2065 on, climate change mitigation policy is
more cost-intensive and increase the costs by 1 percent point in 2100 (see figure
8(a)). Restricting the R&D investments into the electricity sector also increases
18
148 Chapter 6 Endogenous R&D Investments into Energy Efficiency
mitigation costs in the second half of the century. In the first half of the century
in a POLEL and POLTscenario the utility is higher than in the POL scenario,
because of less investments into R&D , but in the long-run these early investments
are helpful to keep the carbon budget.
Fixing the investments into labour efficiency underlines the substitution effect of
these spendings between a BAU and POL scenario. The consumption difference
over time is in the opposite direction for labour and energy related sector experi-
ments. In the end of the century the consumption is higher in the POLLscenario
compared to the POL scenario, because in the latter scenario the R&D from labour
efficiency is re-allocated into energy related efficiency improvements.
5.2 Technology Experiments
Former studies (Edenhofer [10] [11]) indicated that a broad portfolio of technologi-
cal options are necessary for moderate mitigation costs especially for low emission
stabilization scenarios. In the following the question is analysed, whether R&D
investments might compensate the increased mitigation costs if the use of some
technologies is restricted. Therefor the following experiments are run:
•noCCS - no technologies with CCS are available
•noCCS −fixRD - no technologies with CCS are available and all invest-
ments into efficiency are fixed to BAU
•noRenew - use of renewable technologies restricted to BAU use
•noRenew −fixRD - use of renewable technologies restricted to BAU use
and all investments into efficiency are fixed to BAU
Figure 9 concentrates the results of the technological experiments. For most re-
gions R&D investments play an important role for climate change mitigation. This
importance is intensified, if the technological options for carbon free energy pro-
duction are reduced. The global mitigation costs rise by 0.5-3 percentage points
if in addition to a technological restriction, the R&D investments are constrained.
The noCCS scenario results in global mitigation costs of 3% consumption differ-
ences and this is increased to 3.5% for the noCCS-fixRD scenario. Especially
restricting the use of all renewable technologies including biomass, wind and solar
19
6.5 Policy Analysis 149
−5 0 5 10 15 20
World
ROW
INA
CHN
EUR
USA
Consumption change [%]
POL
POL−fixRD
noCCS
noCCS−fixRD
noRenew
noRenew−fixRD
Figure 9: Mitigation costs with restricted technological use and restricted invest-
ments into efficiency
results in high mitigation costs. If the R&D investments are fixed to BAU use, the
mitigation costs amount to 9.1% global consumption losses. But these costs are
lowered by 3 percentage points if endogenous re-allocation of the investments into
production factor efficiencies is allowed. These results demonstrate the important
role of investments into efficiency in a climate change scenario. The meaning of
R&D investments into the efficiency of energy related production sectors increases
the more the technological options of the energy system are restricted.
This general pattern holds for all regions but INA. INA is affected in the opposite
direction: the higher global mitigation costs, the better for INA. Because of the
Contraction and Convergence scheme of the permit allocation, INA is the main
permit exporter and gains from a higher permit price due to reduced options in
keeping a climate target.
Figure 10 shows the difference of the global R&D investment shares of GDP be-
tween the POL and the noCCS and noRenew scenarios. Mainly the investments
into the efficiency of the transport sector are increased if technological options are
20
150 Chapter 6 Endogenous R&D Investments into Energy Efficiency
2020 2040 2060 2080 2100
−0.5
0
0.5
1
1.5
2
2.5
3
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(a) noCCS
2020 2040 2060 2080 2100
−2
0
2
4
6
8
10
Year
Percentage Points
Labor
Electricity
Transport energy
Stationary energy
(b) noRenew
Figure 10: Differenceof the global R&D investments per GDP between technology
scenarios and the reference POL scenario
restricted. The share of energy efficiency related investments amount to 4% of the
global GDP in 2055 for the noRenew scenario. Such steep increase is feasible and
plausible as Nemet and Kammen ([31]) figured out. They analyse data about re-
search and development investments in the energy sector in the U.S. and indicate
that a five to ten-fold increase in energy R&D investment is both warranted and
feasible. However, the experiments with REMIND-RS simultanously showed that
they are also necessary to reduce climate change mitigation costs if technological
options are restricted.
6 Sensitivity Analysis
The calibration routine (see section 4) indicates the long-run elasticity as a crucial
paprameter of high uncertainty. Within this section the assumption on the long-run
elasticity of the R&D function is varied. The effect of a fifty percentage increase
and decrease is analysed. When changing this parameter, also the calibrated func-
tion for the second parameter γIN,r has to be adopted. The results indicate that
(i) the mitigation costs depend on the parameterisation of the R&D function, (ii)
the mitigation costs reducing effect of R&D investments depend on the elasticity
of R&D investments and (iii) the R&D share depends on the elasticity of the R&D
21
6.6 Sensitivity Analysis 151
investments.
Figure 11(a) shows the consumption difference between the respective BAU and
POL scenarios with following changes:
•M50 −en - long-run elasticity of energy related R&D investments lowered
by fifty percent
•M50 −lab - long-run elasticity of labour related R&D investments lowered
by fifty percent
•M50 - long-run elasticity of labour and energy related R&D investments
lowered by fifty percent
•P50 −en - long-run elasticity of energy related R&D investments increased
by fifty percent
•P50 −lab - long-run elasticity of labour related R&D investments increased
by fifty percent
•P50 - long-run elasticity of labour and energy related R&D investments in-
creased by fifty percent
0 2 4 6 8
World
ROW
INA
CHN
EUR
USA
Consumption change [%]
M50
M50−en
M50−lab
P50
P50−en
P50−lab
(a) Reference
−10 0 10 20 30
World
ROW
INA
CHN
EUR
USA
Consumption change [%]
M50−en
M50−en−fixRD P50−en
noRenew
noRenew−fixRD
M50−en−fixRD P50−en
P50−en−fixRD
(b) noRenew
Figure 11: Sensitivity analysis
The higher the elasticity of energy related R&D investments, the lower are the miti-
gation cost, while the opposite holds for the R&D function of labour efficiency. The
experiments with restricted technological use and fixed R&D investments to BAU
use are rerun for a variation of the energy related R&D functions. As demonstrated
in Figure 11(b) the effect that due to redirection of investments into efficiency the
22
152 Chapter 6 Endogenous R&D Investments into Energy Efficiency
high mitigation costs can be lowered is smaller. However, R&D investments reduce
the global mitigation costs for all parameterizations of the R&D function.
Table 1: R&D share in 2005 in Europe
R&D investments data BAU M50-en M50-lab P50-en P50-lab
labour 1.8 1.9 2.0 1.2 1.9 3.0
energy related 0.4 0.35 0.28 0.38 0.45 0.34
Table 1 compares the R&D shares of Europe in 2005 for the scenario of the sensi-
tivity analysis and the default scenarios with empirical data. The parameterization
of the R&D function used for this study are the best fit of the empirical data.
7 Conclusions
This paper uses an Integrated Assessment Model to assess the importance of effi-
ciency improvements induced by R&D investments for climate change mitigation
policies. The question is answered, how R&D investments are redirected under a
climate change policy regime compared to a BAU scenario.
It is shown that the model of endogenous technological change based on R&D in-
vestments in labour efficiency development simulates a plausible scenario of eco-
nomic growth for industrialized regions, fast growing region like China and de-
veloping regions. In scenarios with climate change mitigation policies, aming at
keeping a global carbon budget of 300GtC, the R&D investments are re-allocated
to further improve the efficiencies of the energy related production factors.
Within sector experiments, where the R&D investments into single sectors are fixed
to BAU use, it is figured out, which energy sector gains most from R&D invest-
ments. It is found that mainly investments into the efficiency of the transport and
electricity sector play a crucial role for lowering mitigation costs, especially in the
second part of the century. Without this option the climate change mitigation costs
are increased by 0.5-1 percentage points in the end of the century.
If some carbon free/low carbon technologies will not be available in a climate
change mitigation scenario, the costs could increase dramatically. If no CCS is
available the costs amount to 3% and restricting the use of all renewable technolo-
23
6.7 Conclusions 153
gies means mitigation costs of 6.2% consumption losses. The technology experi-
ments of this paper indicate that this cost increase is partly compensated by invest-
ments into end-use energy efficiency improvements. In comparison with a scenario
that fixes R&D investments to BAU scenario use, mitigation costs are reduced by
0.5-3 percentage points.
8 Acknowledgements
The authors would like to thank Michael H¨
ubler, Nico Bauer and Ottmar Edenhofer
for their contributing comments and discussions.
References
[1] Aghion, Philippe, P. Howitt (1998). Endogenous Growth, Chapter 12, Cam-
bridge, Massachusetts, MIT Press.
[2] Aghion, Philippe, P.Howitt (2009). The Economics of Growth. The MIT
Press, Cambridge, Massachussetts, USA and London, England.
[3] Arrow, K., (1962). The economic implications of learning-bydoing. Review
of Economic Studies 29, 155/173.
[4] Bosetti, Valentina, Carraro, Carlo, Galeotti, Marzio (2006). The Dynamics
of Carbon and Energy Intensity in a Model of Endogenous Technological
Change. The Energy Journal, Endogenous Technological Change and the
Economics of Atmospheric Stabilisation Special Issue, 191-205.
[5] Bosetti, Valentina, Carraro, Carlo, Massetti, Emanuele, Tavoni, Massimo
(2007). International Energy R&D Spillovers and the Economics of Green-
house Gas Atmospheric Stabilization. Energy Economics, Vol 30, Issue 6.
[6] Crassous, Renaud, Hourcade, Jean-Charles, Sassi, Oliver (2006). Endoge-
nous Structural Change and Climate Targets Modeling Experiments with
Imaclim-R. The Energy Journal, Endogenous Technological Change and the
Economics of Atmospheric Stabilisation Special Issue, 259-276.
24
154 Chapter 6 Endogenous R&D Investments into Energy Efficiency
[7] Doraszelski, Ulrich, J. Jaumandreu (2006). R&D and productivity: Estimat-
ing production functions when productivity is endogenous. MPRA Paper No.
1246.
[8] Edenhofer, Ottmar, Bauer, Nico, Kriegler, Elmar (2004). The Impact of
Technological Change on Climate Protection and Welfare: Insights from the
Model MIND. Ecological Econimics, Vol. 54, 277-292.
[9] Edenhofer, Ottmar, Lessmann, Kai, Bauer, Nico (2006). Mitigation Strate-
gies and Costs of Climate Protection: The Effects of ETC in the Hybrid
Model MIND. The Energy Journal, Endogenous Technological Change and
the Economics of Atmospheric Stabilisation Special Issue, 207-222.
[10] Edenhofer, O., C. Carraro, J.-C. Hourcade, K. Neuhoff, G. Luderer, C. Flach-
sland, M. Jakob, A. Popp, J. Steckel, J. Strohschein, N. Bauer, S. Brunner,
M. Leimbach, H. Lotze-Campen, V. Bosetti, E. de Cian, M. Tavoni, O. Sassi,
H. Waisman, R. Crassous-Doerfler, S. Monjon, S. Dr¨
oge, H. van Essen, P. del
R´
ıo, A. T¨
urk (2009). RECIPE - The Economics of Decarbonization. Synthe-
sis Report.
[11] Edenhofer, O., Knopf, B., Barker, T., Baumstark, L., Bellevrat, E., Chateau,
B., Criqui, P., Isaac, M., Kitous, A., Kypreos, S., Leimbach, M., Lessmann,
K., Magn´
e, B., Scrieciu, S., Turton, H., van Vuuren, D.P. (2010): The eco-
nomics of low stabilisation: exploring its implications for mitigation costs
and strategies. The Energy Journal, Volume 31 (Special Issue 1). The Eco-
nomics of Low Stabilization.
[12] Fischer, Carolyn, R. G. Newell (2007). Environmental and technology poli-
cies for climate mitigation. Journal of Environmental Economics and Man-
agement 55, 142-162.
[13] Gerlagh, R., van der Zwaan, B. (2004). A Sensitivity Analysis of Timing and
Costs of Greenhouse Gas Emissions Reduction. Climatic Change 65, 39-71.
[14] Gilligham, Kenneth, R.G. Newell, W.A. Pizer (2008). Modeling Endoge-
nous Technological Change for Climate Policy Analysis. Energy Economicy,
Vol.30, Issue 6, pp 2734-2753.
[15] Gopal Das, Gouranga, Powell, Alan A. (2000). Absorptive Capacity, Struc-
tural Similarity and Embodied Technology Spillovers in a ’Macro’ Model:
An Implementation within the GTAP Framework. Centre of Policy Studies
and the Impact Project, Preliminary Working Paper No. IP-77.
25
6.8 References 155
[16] Greiner,Alfred, Semmler, Willi (2001). Externalities of investment, education
and economic growth. Economic Modelling, 19, 709-724.
[17] Griffith, Rachel, Redding, Stephen, Van Reenen, John (2003). R&D and Ab-
sorptive Capacity: Theory and Empirical Evidence. Scandinavian Journal of
Economics, 105(1), 99-118.
[18] Griliches, Zvi(1998). Productivity Growth and R&D at the Business Level:
Results from the PIMS Data Base. from R&D and Productivity: The Eco-
nomic Evidence. University of Chicago Press, pp 134-156.
[19] Grossman,Gene, Helpman, Elhanan (1990). Trade, Innovation and Growth.
The American Review, Vol. 80, No. 2, 86-91.
[20] Grossman, G. M., E. Helpman (1991). Innovation and Growth in the Global
Economy. MIT Press.
[21] Gr¨
ubler, A., N. Nakicenovic, D.G. Victor (1999). Modeling technological
change: implications for the global environment. Annual Review of Energy
and the Environment 24, 545/569.
[22] Guellec,D., B. van Pottelsberghe de la Potterie (2001). R&D and Productiv-
ity Growth: Panel Data Analysis of 16 OECD Countries. OECD Economic
Studies, No. 33, 2001/II.
[23] Himmelberg, Charles P., B. C. Petersen (1994). R&D and internal finance: A
panel study of small firms in high-tech industires, The Review of Economics
and Statistics, Vol. 76, No.1, pp. 38-51.
[24] H¨
ubler, M., L. Baumstark, M. Leimbach, N. Bauer, O. Edenhofer (2011). A
Model of Endogenous Growth for Climate and Development Policy Assess-
ment. Working Paper, 34pp., submitted to Macroeconomic Dynamics, Cam-
bridge Journals.
[25] IEA (2010). Data Services. International Energy Agency, Paris, France. http
: //data.iea.org/IEASTORE/DEFAULT.ASP.
[26] Jamasb, Tooraj (2007). Technical Change Theory and Learning Curves: Pat-
terns of Progress in Electricity Generation Technologies. The Energy Journal
28, No. 3, 51-72.
[27] Jones, C.I., J.C. Williams (2000). Too Much of a Good Thing? The Eco-
nomics of Investment in R&D. Journal of Economic Growth 5, 65-85.
26
156 Chapter 6 Endogenous R&D Investments into Energy Efficiency
[28] Kypreos, Socrates (2005). Modeling experience curves in MERGE (model
for evaluating regional and global effects). Energy 30, 2721-2737.
[29] Leimbach, M., N. Bauer, L. Baumstark, and O. Edenhofer (2010). Mitiga-
tion costs in a globalized world: climate policy analysis with REMIND-R.
Environmental Modeling and Assessment 15, 155-173.
[30] L¨
oschel, Andreas (2002). Technological chagne in economic models of envi-
ronmental policy: a survey. Ecological Economicy 43, 105-126.
[31] Nemet, Gregory F., D.M. Kammen (2007) U.S. energy research and develop-
ment: Declining investment, increasing need, and the feasibility of expansion.
Energy Policy, Vol. 35, pp. 746-755.
[32] Margolis, Robert M., D. M. Kammen (1999). Underinvestmetn: The Energy
TEchnology and R&D Policy Challenge. Science 285, 690-692.
[33] Meyer, A. (2004). Briefing: Contraction and Convergence. Engineering Sus-
tainability (157). Issue 4, 189-192.
[34] Meinshausen, Malte, N. Meinshausen, W. Hare, S. C.B. Raper, K. Frieler, R.
Knutti, D. J. Frame, M. R. Allen(2009). Greenhouse-gas emission targets for
limiting global warming to 2 ◦C. Nature 458, 1158-1162.
[35] Popp, David (2002). Induced Innovation and Energy Prices. The American
Economic Review, Vol. 92, No.1, pp. 160-180.
[36] Popp, David (2004). ENTICE: endogenous technological change in the DICE
model of global warming. Journal of Environmental Economics and Manage-
ment, Vol. 48, 742-768.
[37] PWC (2011). The World in 2050. http://www.pwc.com.
[38] Rao, Shilpa, Ilkka Keppo, Keywan Riahi (2006). Importance of Technologi-
cal Change and Spillovers in Long-Term Climate Policy. The Energy Journal,
Endogenous Technological Change and the Economics of Atmospheric Sta-
bilisation Special Issue, 25-42.
[39] Romer, P.M. (1990). Endogenous Technological Change, J.P.E. 98, no.5, 71-
102.
[40] Sanstad, A., J. Roy, J.A. Sathaye (2006). Estimating energy-augmenting tech-
nological change in developing country industries. Energy Economicy 28,
720-729.
27
6.8 References 157
[41] SEI (2006). Science and Engineering Indicators. National Science Founda-
tion, USA. http : //www.nsf.gov/statistics/seind06/c4/c4s6.htm.
[42] Weyant, John P., T. Olavson (1999). Issues in modeling induced technological
change in energy, environmental, and climate policy. Environ Modeling and
Assessment 4, 67-85.
[43] Wieser, Robert (2004). Research and develpoment productivity and
spillovers: empirical evidence at the firm level. Journal of Economic Surveys,
Vol. 19, No. 4.
[44] WDI (2010). World Development Indicators. The World Bank, Washington
DC, USA. http : //data.worldbank.org/data - catalog/world - development -
indicators.
[45] Zachariadis, M. (2003). R&D, Innovation, and Technological Progress: A
Test of the Schumpeterian Framework without Scale Effects. The Canadian
Journal of Economics 36(3), 566-586.
28
158 Chapter 6 Endogenous R&D Investments into Energy Efficiency
Chapter 7
Synthesis and Outlook
This thesis deals with the costs and strategies of climate change mitigation. It investigates
the overall research question of what are investment strategies for climate change
mitigation. The chapters 2 to 6 answer the questions raised in the introduction of the
dissertation (chapter 1.5). Within this chapter the key results are presented and discussed
in a comprehensive way. Furthermore, an outlook for further research questions is drawn.
The first part of this synthesis chapter 7.1 presents the results of chapter 2 and 3 of this
thesis, which are mainly dealing with investment strategies in an optimal world. Chapter 4
to 6 in addition investigate the investment strategies to reduce GHG emissions in a world
with technological or political restrictions. These results are summarized in the second
part of this chapter 7.2. In 7.3 all results form chapter 2 to 6 are discussed and an outlook
and further research questions are presented.
7.1 Investment Strategies and Technological Spillover in
an Optimal World
Technological Spillovers
Chapter 2 focuses on the impacts of technological spillovers and investigates the fol-
lowing questions:
•How do climate change mitigation costs depend on technological spillovers?
•What is the effect of technological spillover bound to bilateral capital trade?
•Are there first-mover advantages or commitment incentives due to technological
spillover?
To answer these questions, a multi-regional growth model named MIND-RS is developed.
This model adopts a multi-regional extension of MIND with separated investment good
sector. It the structure, the energy system and endogenous technological change by R&D
159
160 Chapter 7 Synthesis and Outlook
investments from the global Integrated Assessment model MIND (Edenhofer [4]). MIND-
RS as well includes technological change by interregional spillovers. These spillovers are
bound to bilateral capital trade. Within an iterative algorithm a cooperative solution is
calculated.
The model experiments show that in the presence of spillovers that affect labour and
energy efficiency, two opposite effects impact climate change mitigation costs. On the one
hand, a growth effect increases mitigation costs. On the other hand, mitigation costs are
reduced by energy efficiency improvements. The ratio between the spillover intensities
that either increase energy or labour efficiency determines the mitigation costs. The higher
this ratio, the lower are the costs.
As well, the results indicate that due to the link of energy technological spillover to foreign
capital imports, a new climate change mitigation option is introduced. The trade volumes
are much higher in spillover scenarios as compared to non-spillover scenarios. Moreover,
this results in higher relative capital and carbon prices in a scenario with endogenous
technological change. Capital exporting regions like USA and Europe profit from these
improved terms-of-trade.
In MIND-RS experiments, energy efficiency advantages pay off in climate policy scenar-
ios, supporting the hypothesis of benefits for forerunners. It turns out that the first-mover
advantage increases with an increase in the efficiency of domestic R&D investments into
labour efficiency. Thereby, labour-efficiency enhancing capital imports become less at-
tractive, while energy-efficiency enhancing capital imports appear more attractive.
The model results from scenarios with embodied technological spillovers indicate an in-
centive for single regions to join a climate policy coalition. It turns out that restrictions
on technology transfers induce developing regions to take part in a climate policy regime.
MIND-RS is a model with a small energy system part, but models the technological
change including spillover in detail. For climate change mitigation, also the investment
strategy into single technologies is adopted. Therefore a more complex IA model is de-
veloped in chapter 3.
Investment Strategies into Technologies and Impacts of Policy Regimes
The detailed investment decisions in the energy system are the focus of chapter 3 and
the following questions are answered:
•What are the optimal long-term investments in the energy system for climate change
mitigation?
•How differ mitigation costs and strategies between different world regions?
•What are the mitigation costs under different climate policy regimes?
For the exploration of these questions a multi-regional hybrid model called REMIND-
R is developed. It couples a macro-economic growth model, a detailed energy system
model and a simple climate model. The nine world regions are linked by trade in goods,
7.2 Second Best Worlds and Investment Strategies 161
resources and emission permits. REMIND-R is run in a cost-effectiveness mode, maxi-
mizing global consumption under climate policy restrictions. The energy system module
involves more than 50 different transformation technologies. Technologies are bound to
capacities, which constrain the production potentials. Primary energy technologies pro-
duce secondary energy types, which are transformed into final energy.
The numerical analysis demonstrates that investments into the energy system are adopted
for climate change mitigation. In general, the entire energy consumption is reduced while
renewable energy technologies are expanded immediately. CCS technologies are used in
combination with the conversion of gas and coal into electricity. The production of fuels
and gases is lowered and the use of low-value energy sources like solids and other liquids
is reduced.
As well, the results show that the different regions follow quite different strategies to re-
duce carbon emissions. In the short-term, the increase in primary energy consumption is
lower in the industrialized regions. In the first half of the century, the share of fossil fuels
is decreasing in all developed regions, while the share of renewable and nuclear energy in-
creases. Japan relies on nuclear energy and USA substitutes nuclear energy technologies
as of 2050 by converting coal into electricity with CCS. The regions with high potentials
in renewable energy sources exploiting them to a high degree. The middle-eastern re-
gions will employ their huge potential of solar energy while Russia has high potentials in
biomass. China and India base their future energy production on using nuclear technolo-
gies and CCS in coal-fired power plants. Ambitious climate targets can be reached with
global costs amounting to approximately 1.5% of the global gross product. Regions with
high shares in exports of fossil fuels bear the highest costs. The region named MEA for
example faces more than 9% consumption losses. Developing regions like Africa might
benefit from a global climate policy regime due to in net benefits under climate regimes
compared to buiseness-as-usual scenarios.
In addition, it is underlined that the regional mitigation costs significantly depend on the
amount of emissions that have to be reduced in a world region. The regional burden of
emission reductions is influenced by the design of the policy regime, i.e. the emission
permit allocation scheme. For example, a contraction and convergence scenario gives
incentives for Russia to join a climate regime while the developing regions Africa and
India gain from a multi-stage approach. For the rest of the modeled regions the variance
of mitigation costs is rather small. Moreover, the variance of mitigation costs between the
regions is higher than between the policy regimes.
Chapter 2 and 3 analyze investment strategies mostly in first best worlds. Chapter 4 to 6
in addition deal with worlds with political or technological restrictions.
7.2 Second Best Worlds and Investment Strategies
Early Investments into Renewable Energy Technologies
The impacts of early investments into renewable energy technologies in first-best and
second-best worlds are studied in chapter 4 and the following questions are answered:
162 Chapter 7 Synthesis and Outlook
•How do mitigation costs increase due to delays of implementing emission caps at the
global level?
•Can near-term public support of renewable energy technologies contain these in-
creases?
•What are the effects of an early support of renewable energy technologies on regional
mitigation costs?
To answer these questions, again the multi-regional hybrid model REMIND-R is used. It
is extended with some small features. The regional solution in chapter 4 includes eleven
world regions. For the analysis a broad portfolio of different policy scenarios is run. In-
vestment strategies into renewable energy technologies are calculated in first-best climate
change mitigation scenarios that stabilize at 410, 450 and 490 ppm by 2100. These in-
vestment strategies are used in delayed climate policy scenarios to emulate early or late
support of renewable energy technologies.
The model experiments show that climate change mitigation costs increase significantly if
the climate policy implementation is delayed. After 2020 a rapid and enormous reduction
of emissions has to be realized. Therefore, new fossil fuel investments are shifted towards
biomass technologies with CCS.
In addition, it is demonstrated that under delayed climate policy, early deployment of
renewable energy technologies reduces the global costs for achieving a 450ppmCO2con-
centration target. This can be explained by the devaluation of emission permits due to
three effects: (i) some fossil fuels are replaced by renewable energy technologies, (ii)
the significant and negative shift of investments from fossil fuels to biomass technolo-
gies with CCS in 2020 is contained and (iii) additional investments into renewable energy
technologies will get cheaper because of a learning-by-doing effect. However, the global
mitigation costs can not be reduced below the first-best scenario.
In a scenario with immediate climate policy, a deviation from the optimal investment
strategy into renewable energy technologies increases the global mitigation costs slightly.
But the results show a significant effect on the timing of mitigation costs. The less than
the optimal deployment of renewable energy technologies, the higher the mitigation costs
peak around the mid of the century.
The results indicate that the variation of mitigation costs between regions is high in both
cases in immediate and delayed climate policies. The US and Europe always gain from
a strong deployment of renewable energy technologies. If climate policies are delayed,
China gains from some early deployment of renewable energy technologies. But if imme-
diately a climate policy is introduced, China loses from early investments. The developing
regions, e.g. India, Africa, lose from an early renewable energy technologies deployment
in both immediate and delayed climate policy scenarios.
The analysis of this section is based on an eleven-region IA model with a very complex
energy system module. In the following research study, a numerically faster model ver-
sion is prefered.
7.2 Second Best Worlds and Investment Strategies 163
Dimensions of Technological Change
Chapter 5 deals with dynamics and directions of technological change and investigates
the following questions:
•What are the impacts of different dynamics and directions of technological change
on climate change mitigation costs and strategies?
•How does the impact of technological change vary under different elasticities of
substitution between the production factors?
•How does the impact of technological change depend on the availability of energy
technologies?
The study presented in chapter 5 is based on the hybrid model REMIND-RS. This is a five-
regions version of the hybrid model REMIND-R. It adopts the macro-economic structure,
the detailed energy system and the simple climate module. The impacts of the exogenous
parameter of efficiency development are analyzed in a comprehensive sensitivity analysis.
The numerical experiments indicate that mitigation costs and strategies are quite sensitive
to the dynamics and especially the direction of technological change. The ratio between
the growth of labour and some energy efficiencies plays a key role. These efficiency de-
velopments influence the mitigation costs in the opposite direction. Simultaneous changes
of labour and energy efficiencies have only a moderate effect. But a decrease of the effi-
ciency of the single production factor labour to 60% decreases mitigation costs by 23%.
A decrease of energy related efficiency parameters result in a mitigation cost increase by
2 to 53%.
As well, it turns out that the sign and the intensity of the impacts of technological change
are influenced by the scenario world assumed. Experiments in which the macro-economic
production factors of the first CES-level are very good substitutes in contrast to the as-
sumed default low elasticity of substitution are studied. There is evidence that the dy-
namics of technological change affect mitigation costs in the opposite direction, i.e. the
higher the labour efficiency and/or total energy efficiency, the higher are climate change
mitigation costs. Additional experiments with a lowered elasticity of substitution in the
stationary energy sector are analyzed. It turns out that this results in more expensive low
carbon technologies. Thereby, the meaning of the efficiency of the stationary non-electric
energy sector for mitigation cost is increased.
Further experiments indicate that the impact of different dynamics and directions of tech-
nological change depend on the set of available technologies. For example the efficiency
of electricity plays a minor role in a world with restricted use of biomass technologies
compared to a world without any technological restrictions. The efficiency of the trans-
port sector becomes more important if no use of technologies which use CCS are allowed.
These results highlight the importance of the efficiency development of production factors
for climate change mitigation costs and strategies rising further research questions on
endogenous efficiency improvements.
164 Chapter 7 Synthesis and Outlook
Endogenous Technological Change
The role of endogenous technological change for climate change mitigation and the
following questions are analyzed in chapter 6:
•When and where do R&D investments into production factor efficiency play an im-
portant role as climate change mitigation option?
•Can the increased mitigation costs under limited technological options be compen-
sated by a re-allocation of investments into efficiency of energy related production
factors?
For the analysis of this questions, an extended version of the REMIND-RS model is used.
Into this five-region hybrid model endogenous technological change due to production
factor efficiency improvement is introduced. For comparing the impacts of R&D invest-
ments into different production factors, the possibility of direct investments into labour
efficiency or some end-use energy efficiency sectors is implemented. The change of effi-
ciency depends on the efficiency level of the investing region.
The model experiments demonstrate that endogenous R&D investments into labour ef-
ficiency mainly determine the macro-economic growth path. In scenarios with climate
change mitigation policies, the R&D investments are re-allocated to further improve the
efficiencies of the energy related production factors. Climate policy experiments, in which
the R&D investments into labour efficiency or some end-use energy production factors
are fixed to baseline spendings show that these investments affect mitigation costs in the
opposite direction over time. It is found that mainly investments into the efficiency of
the transport and electricity sector play a crucial role for lowering mitigation costs, es-
pecially in the second part of the century. The mitigation costs would increase by 0.5 to
1 percentage points, if no re-allocation of R&D investments into end-use energy related
productivities in a policy scenario is allowed for.
In a climate change policy scenario, the restricted use of some low carbon technologies
increases the mitigation costs significantly. If no CCS is available the costs amount to 3%
global consumption differences. Restricting the use of all renewable energy technologies
results in global mitigation costs of 6.2%. The experiments in chapter 6 show that these
dramatical mitigation costs can be lowered by the re-allocation of R&D investments into
end-use energy sector efficiencies. In comparison with a scenario that fixes R&D invest-
ments to BAU scenario use, mitigation costs are reduced by 0.5 to 3 percentage points.
All questions identified in the introduction of this thesis are answered within the chapter
2 to 6. The following section summarizes these results and discusses them in a compre-
hensive way.
7.3 Discussion and Further Research
Within this section, the results of chapter 2 to 6 are discussed. The common features
and methods will be highlighted and some limitations of the numerical approach will be
presented. Thereby future research question are identified.
7.3 Discussion and Further Research 165
Discussion
The answers on the questions investigated in chapter 2 to 6 give a broad picture for the
answer on the overall research question: What are possible investment strategies for cli-
mate change mitigation? The results of chapter 3, 5 and 6 demonstrate that a rich portfolio
of technologies is necessary for low emission reduction costs. Regional mitigation strate-
gies are significantly different and depend on the advantages of technologies and resources
of some regions. In addition, regional mitigation costs show drastical differences depend-
ing on the assumed permit allocation rule. Chapter 4 highlights that the investments into
low carbon technologies, respectively renewable energy technologies like wind, solar and
biomass should start early in the century. Especially if no global climate change mitiga-
tion regime is installed immediately, early investments into renewable energy technolo-
gies avoid dramatic mitigation costs in the second part of the century. Furthermore the
development and ratio of labour efficiency and energy efficiency improvements have a
high impact on mitigation mitigation costs and strategies. This is pointed out in chapter 5.
Chapter 6 shows how important R&D investments into production factor efficiencies are.
Especially sectors which are hardly to be decarbonized gain from investments into tech-
nological change and thereby climate change mitigation costs are reduced. In addition,
this chapter demonstrates that a re-direction of R&D investments from labour efficiency
to energy related technological improvements helps to overcome the increased mitiga-
tion costs under technological restrictions. Another feature of endogenous technological
change is analyzed in chapter 2 of this thesis. Technological spillovers bound to bilateral
capital trade shows a significant impact on climate change mitigation costs. As well, this
feature provides first mover advantages for some industrialized regions and incentives for
developing regions to join a climate policy regime.
Jointly, the findings of this thesis demonstrate the important role of investment strategies
for climate change mitigation costs. The world gains from early investments into both a
broad portfolio of technologies and into energy efficiencies. Thereby the immediate sup-
port and high diversity of investments mainly provide decreased costs of GHG emission
reductions.
All models used for the analysis are regional optimization models that maximize a global
welfare solution. This type of model provides some advantages and also some limita-
tions. The different models presented in chapter 2, chapter 3 and chapter 5 and 6 focus on
different aspects of climate change modeling. Therefore, they are specialized and partly
ignore features for answering questions from other chapters. The best model might be
one that includes all advantages of the presented models. Yet, until now such a big and
heavy model would exceed the numerical capacities. However, the specialized models are
constructed to answer the raised research questions. REMIND-R for example is a hybrid
model with a highly detailed energy system module. This is important for the analysis of
the questions answered in chapter 3 and 4. REMIND-RS includes less world regions than
REMIND-R but provides an endogenous formulation of technological change. Neverthe-
less, an extension of the existing models is part of future research.
The presented models assume perfect markets and perfect intertemporal foresight.
Thereby a benchmark solution with benchmark climate change mitigation costs are cre-
ated. All decisions - including investments into technologies, production factor use, etc. -
incorporate the optimal behavior of all regions at each time step. This approach tends to
166 Chapter 7 Synthesis and Outlook
decrease the mitigation costs by optimally investing in most promising long-term mitiga-
tion mechanisms. Chapter 4 to 6 partly leave this perfect world by including exogenous
restrictions. However, these experiments still assume perfect foresight and react to the
exogenous restrictions in an optimal way. In addition, all used models ignore any strate-
gic behavior of regions, firms or investors. For such investigations usually smaller models
based on game theory are used. The hybrid models of this thesis include too many vari-
ables, because of the detailed energy system and high regional solution for the analysis of
strategic behavior.
As for all models holds that the empirical foundation is sometimes problematic. Key data
about capital stocks, investments and capacities of technologies can easily be found for in-
dustrialized regions. But for a more detailed regional resolution, sometime different data
sources have to be used. Moreover, the empirical foundation of some specific parameters
is small. The elasticity of substitution between labour, capital and energy for example has
a big impact on mitigation costs but quite impossible to estimate from data. As well, only
space data exist for parameters needed for implementations of endogenous technological
change ( Bosetti [2], Keller [7], Verdolini [13]).
The atmospheric climate change problem is always facing some uncertainties - uncertain-
ties about the radiative forcing, climate sensitivity, and tipping points. The models used in
this thesis are run in a cost-effectiveness mode, which ignores all costs of climate change.
So the model results provide a benchmark world. The exogenous bound on emissions or
a temperature cap is introduced to avoid dramatical climate change. However, a complex
climate uncertainty would exceed this thesis. First results about the implications of mid-
century targets keeping long-term climate policy targets are presented by O’Neill [10].
Held et.al. [6] investigate the impacts of uncertainty in four key model parameters.
The discussed advantages and limitations indicate further questions for future research
analysis. A few examples are presented in the following section.
Outlook and Further Research
The results of this thesis answer different aspects of the overall research question, but
as well further questions are identified. Some research questions call for extensions of the
used models or need alternative solution algorithms. In the following, a few examples are
discussed.
The demonstrated models use a simple transport module but the results indicate that espe-
cially this sector plays an important role for the level of mitigation cost. Further research
with an extended transport sector including electricity transport is needed. Then, the in-
vestment strategies in this sector and the importance of R&D investments into efficiency
improvements of this sector can be identified.
As well, a more detailed analysis of the interactions between regional investment strate-
gies into technologies and R&D investments into factor efficiency developments is
needed. Therefore, a model with endogenous technological change and a high regional
solution should be developed. Research questions dealing with the impacts of high R&D
investments for developing regions could be analyzed with such a model. Perhaps climate
change mitigation investments are the key variables that result in economic divergence of
industrialized and developing regions. Thereby also technological spillovers may show
7.3 Discussion and Further Research 167
important effects on climate change mitigation costs. Some IA models for climate change
analysis (Bosetti [1], Crassous [3], Kypreos [8]) use a more detailed formulation of the
endogenous technological change including absorptive capacities or spendings separated
for innovation and imitation. More empirical research is needed for a better foundation of
the model formulations for endogenous technological change.
Beside the possibility for installing new features of describing effects in a more detailed
formulation, also the solution algorithm might be changed to investigate further ques-
tions. Especially if spillovers or other externalities are taken into account in a climate
change analysis, an alternative solution algorithms are needed. Further research ques-
tions are: When will a region join a climate coalition? What is the regional amount or
emission reductions, if no global climate regime is installed? How can a free-rider-effect
be overcome? Answering these research questions implies to leave the assumption of a
benchmark solution under optimal conditions. The strategic behavior of a country can be
calculated if the welfare of each country is maximized independently of global welfare
effects. A few studies are dealing with the identified research questions (Lessmann [9],
Flachsland [5]).
In addition there is a research domain analyzing uncertainty in climate change models.
These experiments might be run in a cost-benefit mode, where the costs of climate change
mitigation and climate change damages are compared and the cheapest mitigation level is
calculated (e.g. Pizer [11], Tol [12]). Alternatively, a Monte Carlo Analysis can be used
to study uncertainty implications. However, in numerical experiments with different so-
lution algorithms, endogenous technological change might play a different role. Further
research for example might investigate the question, whether the optimal share of R&D
investments depends on the uncertainty about climate sensitivity.
168 Chapter 7 Synthesis and Outlook
Bibliography
[1] Bosetti, V., C. Carraro, M. Galeotti, E. Massetti, M. Tavoni (2006). WITCH: A
World Induced Technical Change Hybrid Model. The Energy Journal, Special Issue
2006, 13-38.
[2] Bosetti, V., C. Carraro, M. Galeotti, E. Massetti, M. Tavoni (2008). International
energy R&D spillovers and the economics of greenhouse gas atmospheric stabiliza-
tion. Energy Economics 30, 2912-2929.
[3] Crassous, R., J.-C. Hourcade, O. Sassi (2006). Endogenous Structural Change and
Climate Targets. Modeling experiments with Imaclim-R. The Energy Journal, Spe-
cial Issue 2006, 259-276.
[4] Edenhofer, Ottmar, Nico Bauer, Elmar Kriegler (2004). The Impact of Technolog-
ical Change on Climate Protection and Welfare: Insights from the Model MIND.
Ecological Economics, Vol. 54, 277-292.
[5] Flachsland, C., R. Marschinski, O. Edenhofer (2009). Global Trading versus Link-
ing. Architectures for International Emissions Trading. Energy Policy 37, 1637-
1647.
[6] Held, H., E. Kriegler, K. Lessmann, O. Edenhofer (2009). Efficient climate policies
under technology and climate uncertainty. Energy Economics 31, S50-S61.
[7] Keller, W. (1998). Are international R&D spillovers trade-related? Analyzing spill-
overs among randomly matched trade partners. European Economic Review 42,
1469-1481.
[8] Kypreos, S., O. Bahn (2003). A MERGE model with endogenous technological
progress. Environmental Modeling and Assessment 19, 333-358.
[9] Lessmann, K., R. Marschinski, O. Edenhofer (2009). The effects of tariffs on coali-
tion formation in a dynamic global warming game Economic Modelling 26(3):641-
649.
[10] O’Neill, B.C., K. Riahib, I. Keppo (2009). Mitigation implications of midcentury
targets that preserve long-term climate policy options. PNAS 107, no. 3, 1011-1016.
[11] Pizer, W. A. (1997). Optimal Choice of Policy Instrument and Stringency under
Uncertainty: The Case of Climate Change. Resources for the Future, Discussion
Paper 97-17.
169
170 BIBLIOGRAPHY
[12] Tol, R.S.J. (1999). Safe policies in an uncertain climate: an application of FUND.
Global Environmental Change 9, 221-232.
[13] Verdolini, E., M. Galeotti (2009). At Home and Abroad: An Empirical Analysis of
Innovation and Diffusion in Energy-Efficient Technologies. Working Paper. FEEM
Nota di Lavoro 123.2009, Fondazione Eni Enrico Mattei, Milan.
171
172
Statement of Contribution
The chapters of this thesis are the result of collaborations between the author of this thesis
and her advisor, Prof. Dr. Ottmar Edenhofer, involving additional colleagues as indicated.
In particular, the chapters dealing with technological change (Chapters 2, 5 and 6) were
developed in close collaboration with Marian Leimbach.
The author of this thesis has made significant contributions to the contents of all five pa-
pers, from conceptual design over technical development to writing. This section details
the contribution of the author to the five papers and acknowledges major contributions of
others.
Chapter 2 The conceptional design and writing of the article has been undertaken by
Marian Leimbach. The author of this thesis contributed to the numerical experiments,
their implementation and execution, and processing of model results and their visualiza-
tion in the graphs of the chapter.
Chapter 3 The model REMIND-R used in this chapter is based on the model MIND,
mainly developed by Nico Bauer. Marian Leimbach was responsible for the conceptional
design and writing of this article. The author of this thesis contributed to the development
of the model, the numerical experiments, their visualization and interpretation. Ottmar
Edenhofer contributed by framing the research question and in extensive discussions.
Chapter 4 The conceptual design and writing has been undertaken by Nico Bauer. The
author of this thesis contributed to framing the research question and specifying the sce-
narios. She was responsible for most of the numerical experiments, their implementation
and visualization. In addition, the author gave inputs to some sections of the paper. Mar-
ian Leimbach was involved in the definition of the research question as well as the editing
process.
Chapter 5 The author was responsible for the conceptual design and writing of the
article. As well, she developed the used model REMIND-RS based on the REMIND-R
model. The model experiments are run by the author. Frequently therefor the multi-run
environment SimEnv was used. The author was responsible for the visualization and
interpretation of the results. Marian Leimbach contributed to the conceptual design of the
article, designing the numerical experiments and provided editing as well as some text
inputs.
173
174
Chapter 6 The extension of the model was implemented by the author. As well, she was
responsible for framing the research question, constructing the numerical experiments,
their visualization and writing the article. Marian Leimbach contributed to the conceptual
design of the article, designing the numerical experiments and provided editing as well as
some text inputs.
Acknowledgements
I would like to thank colleagues, friends, and family, whose support throughout these
years enabled me to research and write this thesis.
First, I am indebted to Ottmar Edenhofer’s support and want to thank for the open
discussions of the macro-economic aspects of climate change. I want to thank Marian
Leimbach for his support, advice and scrutiny in the collaborations that were essential in
researching and writing this dissertation.
I would like to thank all my colleagues from whom I learned a lot and that contributed
to providing the stimulating and collegial working atmosphere at PIK, in particular Kai
Lessmann, Nico Bauer, Brigitte Knopf, Alexander Lorenz, Michael Jakob, Annegret Kuh-
nigk, Wibke Avenhaus, Markus Haller, Gunnar Luderer, Robert Pietzker, Eva Schmid, Jan
Steckel, MichaelLücken, Michael Pahle, SteffenBrunner, Falco Ückert, David Klein, Jes-
sica Streffler, Sylvie Ludig, Robert Marschinski, Michael Hübler, Elmar Kriegler. Special
thanks to Christian Flachsland for sharing the office for so many years.
Jutta Neuhaus, Kristyiana Neumann, Nicole Reinhardt and Dorothe Ilskens provided
the efficient management that made administrative life at PIK relatively easy - thank you.
Many thanks to Kai Lessmann for patient support in editing the thesis manuscript,
and Eva Schmid, Alexander Lorenz, Michael Pahle, and Ansgar Snethlage for helpful
comments on the Introduction chapter.
Finally, I want to thank my parents for their trust and support, especially my husband
Ingo, for his encouragement and everything else and last Odo, for getting me out for a
walk periodical and defending the office.
175
176
Tools and Resources
This thesis relies mainly on numerical modeling and a number of software tools were
used to create and run the model experiments. In addition, tools for the analysis and
visualization of the results are used. 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 1and the CONOPT3 solver, version
3.14S 2, for non-linear programs. The multi-run environment SimEnv, versions 1.15–2.01
3, was used partly.
Data processing Model output was analyzed using The MathWorks’ MATLAB, ver-
sion 2007b 4and the NetCDF Toolbox for MATLAB by Charles R. Denham.
Typesetting This document was prepared using L
A
TEX5, particularly the pdfpages pack-
age 6to include Chapters 2 to 6 in their given layouts.
1Brooke, A., Kendrick, D., Meeraus, A., Rosenthal, R., 1988. GAMS: A User’s Guide. The Scientific
Press.
2Drud, A., 1994. CONOPT–A Large-Scale GRG Code. INFORMS Journal on Computing 6 (2), 207.
3Flechsig, M., Böhm, U., Nocke, T., Rachimow, C., 2008. The multi-run simula-
tion environment SimEnv. Http://www.pik-potsdam.de/software/simenv/. URL http://www.pik-
potsdam.de/software/simenv/
4MATLAB, 1998. MATLAB – The Language of Technical Computing. The MathWorks.
5Lamport, L. (1994): LaTeX: A Document Preparation System User’s Guide and Reference Manual.
Reading, Massachusets
6Matthias, A. (2006): The pdfpages package. Tech. rep. CTAN, CTAN://macros/latex/contrib/pdfpages.
177
178