Technische Universität Berlin
School VII Economics and Management
Economics of the Global Steam Coal Market Modeling Trade, Competition and Climate Policies
vorgelegt von
Clemens Haftendorn
Von der Fakultät VII - Wirtschaft und Management
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Wirtschaftswissenschaften
Dr. rer. oec.
genehmigte Dissertation
Promotionsausschuss:
Vorsitzender:
Prof. Dr. Radosveta Ivanova-Stenzel
Berichter:
Prof. Dr. Christian von Hirschhausen (TU Berlin)
Prof. Dr. Claudia Kemfert (Hertie School of Governance)
Prof. Dr. Marco Runkel (TU Berlin)
Tag der wissenschaftlichen Aussprache: 17.10.2012
Berlin 2012
D 83
Abstract
Steam coal is one of the main pillars of global electricity generation and its consumption is increasing, mainly driven by the growing Asian economies. Since the early
2000s the global steam coal trade underwent some significant changes. The international
seaborne trade flows are growing rapidly and on the supply side a process of mergers and
acquisitions increased market concentration, fueling fears that an oligopolistic market
structure would develop similarly to the other fossil fuel markets (oil and natural gas).
Concerning environmental externalities, the continued use of coal poses a serious problem
especially because of the high carbon dioxide emissions that exacerbate climate change.
This thesis presents several numerical partial equilibrium models to analyze these
main issues affecting the market. In a first step the international seaborne trade of steam
coal in the years 2005 and 2006 is investigated using the COALMOD-Trade model. At the
time when market concentration on the supply side was the highest, we find no evidence of
non-cooperative strategic behavior. The international trade flows and prices are better
represented by a competitive model. However, a further analysis of the more mature
and liquid Atlantic sub-market between 2003 and 2006 finds some evidence of market
power exertion in this region. Yet we conclude that there is no structural market power
problem in the mid-term for the global steam coal market. The large scale COALMODWorld model is thus based on the assumption of perfect competition and computes yearly
trade flows until 2030 including both international trade and domestic markets as well as
endogenous investments in mining and transport infrastructure. This model replicates
actual trade flows for the base year 2006. One main finding is that, if global demand
continues to increase after 2015, global supply costs may rise due to the large need of
investments. COALMOD-World is also used to analyze the interactions between climate
policies and the global steam coal market. We find that the effectiveness of regional
policies reducing coal demand can be undermined by demand shifting to countries with
less stringent climate goals.
This thesis contributes to the energy economics literature in several respects. An
updated analysis of market power in the global steam coal market is performed. Furthermore, a contribution is made concerning the use of conjectural variations to represent
strategic interactions on a market, especially how conjectural variations models may fail
to represent Nash equilibria. Thus, an alternative for the modeling of oligopoly markets
with a competitive fringe is proposed. The COALMOD-World model contains novelties
such as an endogenous cost mechanism that have not yet been used in other energy and
resources market models. Finally, while the mechanisms of carbon leakage through energy markets are well understood in theory, we are able to provide a quantification for
the order of magnitude using the COALMOD-World model.
Keywords: global steam coal market, partial equilibrium numerical modeling, market
power modeling, climate policy
2
Zusammenfassung
Kesselkohle ist ein wichtiger Pfeiler der weltweiten Stromversorgung, dessen Verbrauchswachstum hauptsächlich von der wirtschaftlichen Entwicklung Asiens getrieben
wird. Seit den frühen 2000er Jahren befindet sich der globale Kesselkohlehandel im Umbruch. Die Handelsflüsse wachsen stetig und es hat ein Konzentrationsprozess auf der
Anbieterseite stattgefunden. Dies hat Befürchtungen über eine oligopolistische Marktstruktur geweckt. Die weitere Nutzung von Kesselkohle verursacht zudem erhebliche
Umweltexternalitäten, insbesondere auf das Klima durch den Ausstoß von Kohlendioxid.
Diese wichtigen Einflussfaktoren des Kesselkohlemarktes werden in der vorliegenden Dissertation mithilfe von numerischen partiellen Gleichgewichtsmodellen analysiert.
Zuerst wird der internationale Kesselkohlehandel in der Jahren 2005 und 2006 mit
dem COALMOD-Trade-Modell untersucht. Zum Zeitpunkt der höchsten Anbieterkonzentration in diesen Jahren kann kein Anhaltspunkt für strategisches Verhalten gefunden
werden. Die Handelsflüsse und Preise im globalen Markt werden durch ein wettbewerbliches Modell besser abgebildet. Im liquideren atlantischen Markt, der in einer genaueren
Untersuchung von 2003 bis 2006 analysiert wird, gibt es zwar Anzeichen einer punktuellen Marktmachtausübung. Dennoch kann für die folgende Modellentwicklung davon
ausgegangen werden, dass Marktmacht mittelfristig kein strukturelles Problem im globalen Kesselkohlemarkt darstellt. Das umfangreiche COALMOD-World-Modell simuliert
daher jährliche Handelsflüsse bis 2030 für den internationalen Handel und einheimische
Märkte mit endogenen Investitionen in Abbau- und Transportinfrastruktur. Das Modell
kann die Handelsflüsse für das Basisjahr 2006 nachbilden. Im Fall eines weiteren globalen
Nachfrageanstiegs jenseits von 2015 kann es zu Transportengpässen sowie steigenden Beschaffungskosten kommen. Dies ist auf den großen Investitionsbedarf zurückzuführen, der
nur schwer der Nachfrageentwicklung folgen kann. COALMOD-World wird im Weiteren
genutzt, um das Wechselspiel zwischen Klimapolitiken und globalem Kesselkohlemarkt
zu untersuchen. Eines der Ergebnisse ist, dass die Effektivität von regionalen Politiken
erheblich beeinträchtigt werden kann, falls es zur Verschiebung der Nachfrage in Länder,
die weniger oder keine klimapolitischen Anstrengungen tätigen, kommt.
Diese Dissertation trägt zur energieökonomischen Literatur in vielerlei Hinsicht bei.
Es wird eine aktuelle Marktmachtanalyse des internationalen Kesselkohlemarktes durchgeführt. Die Methodenforschung wird durch einen Beitrag zur Anwendung von “conjectural variations” zur Abbildung strategischen Verhaltens bereichert. Diese Modelle können
problematisch sein, da sie nicht in jedem Fall eine Nash-Lösung abbilden. Deswegen wird
ein Alternative für die Modellierung von Oligopolen mit einem wettbewerblichen Rand
vorgeschlagen. Des Weiteren beinhaltet COALMOD-World mehrere methodische Neuerungen, z.B. einen endogenen Kostenmechanismus. Zudem ermöglicht das Modell eine
Quantifizierung des theoretisch viel erforschten “Carbon Leakage” Effekts, hier über den
globalen Kesselkohlemarkt.
Schlüsselwörter: Globaler Kesselkohlemarkt, partielle numerische Gleichgewichtsmodellierung, Marktmachtmodellierung, Klimapolitik
3
Acknowledgements
First I would like to thank Professor Christian von Hirschhausen for giving me the opportunity to write this dissertation as well as for the motivation and incentives that helped
me finish this project successfully. Christian introduced me to the world of economic research and to the fascinating world of global energy markets. The knowledge and passion
for the energy markets he helped me gain during those research years will always be with
me.
I would like to give a very special thanks to Franziska Holz for her continuous support
and for the very fruitful cooperation on many research projects over the years. She
introduced me to numerical modeling, accompanied my dissertation project from the
very beginning and was always available for discussion. Franziska’s support and friendship
was fundamental for the successful completion of this dissertation. Another dear friend
and colleague I must thank is Daniel Huppmann for a lot of interesting discussions on
modeling and other topics and for his knowledge that he is always happy to share and
help with.
I am grateful to the DIW Graduate Center, Professor Claudia Kemfert and the entire department of Energy, Transportation and Environment for providing me with an
excellent research environment.
I would also like to thank all the wonderful people I have met in the various research
teams I have been in contact with over the years and that have contributed to the success
of this research project: the old Dresden team, the TU Berlin WIP group and of course the
team in the department of Energy, Transportation and Environment at DIW Berlin. I am
also indebted to Professor Steven Gabriel and to Ruud Egging for the many interesting
workshops that have helped improve my modeling skills. Many thanks also to Professor
Ben Hobbs for hosting me for a research stay at Johns Hopkins University in Baltimore.
Last but no least I would like to thank my friends, especially those in the 2009 DIW
GC cohort, my family and Catalina for their continuous moral support and for the fun
and laughter during those research years.
4
Workin’ in a coal mine
Goin’ down down down
Workin’ in a coal mine
Whop! about to slip down
Allen Toussaint (1966)
5
Contents
Abstract
2
Zusammenfassung
3
Acknowledgements
4
Table of Contents
6
List of Tables
9
List of Figures
11
1 Introduction
13
1.1
Trying not to “carry coal to Newcastle”
1.2
Market Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3
1.4
1.5
. . . . . . . . . . . . . . . . . . . 13
1.2.1
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.2
Numerical modeling of imperfect competition . . . . . . . . . . . . 15
1.2.3
Market concentration in the international steam coal trade . . . . . 19
1.2.4
Research questions and modeling challenges . . . . . . . . . . . . . 20
Trade and Investments Dynamics under Rising Demand . . . . . . . . . . 21
1.3.1
Rising demand in Asia . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.2
A complex supply side . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3.3
Research questions and modeling challenges . . . . . . . . . . . . . 23
Climate Change and Policy . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4.1
Stand or fall by coal . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4.2
Climate policy must not neglect global market dynamics . . . . . . 25
1.4.3
Research questions and modeling challenges . . . . . . . . . . . . . 25
Overview of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5.1
Chapter 2: Modeling and analysis of the international steam coal
trade: is the market competitive? . . . . . . . . . . . . . . . . . . . 27
1.5.2
Chapter 3: Atlantic steam coal market power: theory and application of oligopoly models with a competitive fringe . . . . . . . . . . 29
1.5.3
Chapter 4: A techno-economic analysis using the COALMODWorld model: the end of “cheap coal”? . . . . . . . . . . . . . . . . 30
6
Contents
1.5.4
Chapter 5: Climate policies and the global steam coal market:
interactions until 2030 . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6
Concluding Remarks and Outlook . . . . . . . . . . . . . . . . . . . . . . . 35
2 Modeling and Analysis of the International Steam Coal Trade
37
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2
State of the Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3
The COALMOD-Trade Model . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4
2.5
2.3.1
Description of the analytical model . . . . . . . . . . . . . . . . . . 40
2.3.2
Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.1
Market structure analysis . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.2
Pricing and export restrictions in a temporal perspective . . . . . . 48
2.4.3
Spatial price discrimination . . . . . . . . . . . . . . . . . . . . . . 50
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.A Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3 Atlantic Steam Coal Market Power
58
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2
Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3
Critique of Conjectural Variation Models . . . . . . . . . . . . . . . . . . . 60
3.3.1
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.2
Partial equilibrium modeling applications . . . . . . . . . . . . . . 61
3.4
Developments in the Atlantic Steam Coal Market in the Early 2000s . . . 65
3.5
Market Power Models with Dominant Players and a Competitive Fringe . 68
3.6
3.7
3.5.1
Modeling the competitive fringe . . . . . . . . . . . . . . . . . . . . 68
3.5.2
Cournot oligopoly as Stackelberg leader . . . . . . . . . . . . . . . 69
3.5.3
Stackelberg-Cartel model . . . . . . . . . . . . . . . . . . . . . . . 70
3.5.4
Nash-Bargaining model . . . . . . . . . . . . . . . . . . . . . . . . 70
Application to the Atlantic Steam Coal Market . . . . . . . . . . . . . . . 71
3.6.1
Model specification and data . . . . . . . . . . . . . . . . . . . . . 71
3.6.2
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.A Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.A.1 Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.A.2 Market data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4 A Techno-economic Analysis using the COALMOD-World Model
80
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2
Equilibrium Modeling of Energy Resource Markets . . . . . . . . . . . . . 83
4.3
The COALMOD-World Model . . . . . . . . . . . . . . . . . . . . . . . . 85
4.3.1
Model structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7
Contents
4.4
4.5
4.6
4.3.2
The producers’ problem . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.3
The exporters’ problem . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.4
Other model equations . . . . . . . . . . . . . . . . . . . . . . . . . 94
Model Specification and Data . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.4.1
Countries and nodes definition . . . . . . . . . . . . . . . . . . . . 96
4.4.2
Production, costs and reserves . . . . . . . . . . . . . . . . . . . . . 97
4.4.3
Land transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.4.4
Export ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.4.5
Freight rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4.6
Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4.7
Investments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.5.1
General assumptions and base year results . . . . . . . . . . . . . . 106
4.5.2
Increasing demand scenario . . . . . . . . . . . . . . . . . . . . . . 108
4.5.3
Stabilizing demand scenario . . . . . . . . . . . . . . . . . . . . . . 112
4.5.4
Results comparison with Hubbert-method based models . . . . . . 113
4.5.5
Model evaluation and criticism . . . . . . . . . . . . . . . . . . . . 115
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.A Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.A.1 Mathematical Formulation of the Model . . . . . . . . . . . . . . . 117
4.A.2 Nodes of COALMOD-World . . . . . . . . . . . . . . . . . . . . . . 119
4.A.3 Data of COALMOD-World . . . . . . . . . . . . . . . . . . . . . . 121
4.A.4 Results of COALMOD-World . . . . . . . . . . . . . . . . . . . . . 122
5 Climate Policies and the Global Steam Coal Market
125
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.2
Assessment of Positive Modeling Approaches . . . . . . . . . . . . . . . . 126
5.3
5.4
5.2.1
Overview of possible modeling approaches . . . . . . . . . . . . . . 126
5.2.2
Advantages of partial equilibrium models . . . . . . . . . . . . . . 128
5.2.3
Description of the COALMOD-World model . . . . . . . . . . . . . 129
Climate Policy Scenarios with the COALMOD-World Model . . . . . . . . 130
5.3.1
Worldwide climate policy . . . . . . . . . . . . . . . . . . . . . . . 132
5.3.2
Unilateral European climate policy . . . . . . . . . . . . . . . . . . 133
5.3.3
Yasuní-type supply-side policy in Indonesia . . . . . . . . . . . . . 134
5.3.4
CCS fast roll-out . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.3.5
Scenario combination: hedging of negative market adjustments . . 138
Conclusions and Policy Recommendations . . . . . . . . . . . . . . . . . . 139
5.A Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Bibliography
142
8
List of Tables
1.1
Models specifications in the chapters . . . . . . . . . . . . . . . . . . . . . 27
2.1
Share of imports in total consumption and share of imported steam coal
in total electricity generation of major steam coal consuming countries 2006 38
2.2
Units used for the COALMOD-Trade and COALMOD-Trade-Energy models 41
2.3
Countries in the COALMOD-Trade model . . . . . . . . . . . . . . . . . . 43
2.4
Coal quality by exporter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5
Average FOB prices of Australian steam coal exports to importing countries in 2005 and 2006, in USD per ton . . . . . . . . . . . . . . . . . . . . 51
2.6
Simulated import prices for the CMT model and reference CIF prices for
2005, in USD per ton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.7
Steam coal trade flows in the 2005 perfect competition scenario for the
CMT model, in million tons (Mt) . . . . . . . . . . . . . . . . . . . . . . . 54
2.8
Steam coal trade flows in the 2005 Cournot scenario for the CMT model,
in million tons (Mt) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.9
Actual steam coal trade flows in the base year 2005, in million tons (Mt)
54
2.10 Simulated import prices for the CMT model and reference CIF prices for
2006, in USD per ton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.11 Steam coal trade flows in the 2006 perfect competition scenario for the
CMT model, in million tons (Mt) . . . . . . . . . . . . . . . . . . . . . . . 55
2.12 Steam coal trade flows in the 2006 Cournot scenario for the CMT model,
in million tons (Mt) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.13 Actual steam coal trade flows in the base year 2006, in million tons (Mt)
55
2.14 Simulated import prices for the CMT-E model and converted reference
CIF prices for 2005, in USD per Gigajoules . . . . . . . . . . . . . . . . . 56
2.15 Steam coal trade flows in the 2005 perfect competition scenario for the
CMT-E model, in Petajoules (PJ)
. . . . . . . . . . . . . . . . . . . . . . 56
2.16 Steam coal trade flows in the 2005 Cournot scenario for the CMT-E model,
in Petajoules (PJ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.17 Actual steam coal trade flows in the base year 2005, converted in Petajoules
(PJ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
9
List of Tables
2.18 Simulated import prices for the CMT-E model and converted reference
CIF prices for 2006, in USD per Gigajoules . . . . . . . . . . . . . . . . . 57
2.19 Steam coal trade flows in the 2006 perfect competition scenario for the
CMT-E model, in Petajoules (PJ)
. . . . . . . . . . . . . . . . . . . . . . 57
2.20 Steam coal trade flows in the 2006 Cournot scenario for the CMT-E model,
in Petajoules (PJ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.21 Actual steam coal trade flows in the base year 2006, converted in Petajoules
(PJ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1
Producer data: marginal cost parameters (intercept and slope) in USD/t
and production capacities in million tons per year . . . . . . . . . . . . . . 71
3.2
Modeling results for the traded quantities in Mt and the prices in USD/t
in 2004 and 2005 for the four market structure scenarios.
3.3
. . . . . . . . . 72
Modeling results for the traded quantities in Mt and the prices in USD/t
in from 2003 to 2006 for the Cournot oligopoly and the perfect competition
cases.
3.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Steam coal trade flows to Europe in million tons and import market share
of South Africa and Colombia . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.5
Export volumes in million tons and export market share of the “Big Three” 79
4.1
Reserves of major countries in COALMOD-World in Million tons . . . . . 98
4.2
Energy content and quality κf of coal by production nodes . . . . . . . . . 99
4.3
World Energy Outlook demand projections for coal for power generation
in the reference scenario converted to Petajoules
. . . . . . . . . . . . . . 102
4.4
Nodes of the COALMOD-World Model . . . . . . . . . . . . . . . . . . . . 120
4.5
Data and assumptions for the endogenous cost mechanism . . . . . . . . . 121
4.6
Data assumptions for the per 5-years capacity expansion limitations in Mtpa121
4.7
Results of COALMOD-World: Domestic trade flows in Mtpa for the increasing demand and the stabilizing demand scenarios . . . . . . . . . . . 122
4.8
Results of COALMOD-World: International trade flows in Mtpa for the
increasing demand and the stabilizing demand scenarios (part 1/2) . . . . 123
4.9
Results of COALMOD-World: International trade flows in Mtpa for the
increasing demand and the stabilizing demand scenarios (part 2/2) . . . . 124
5.1
EU demand reduction in the Unilateral European Climate Policy scenario
compared to the reference scenarios in percentage . . . . . . . . . . . . . . 133
5.2
Assumed installed capacities of coal power plants with CCS for the CCS
scenario in GW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.3
Demand elasticities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10
List of Figures
2.1
Imported quantities in the perfect competition (PC), Cournot scenario
(CO), and reference data (RE) in 2005, in million tons (Mt) for the CMT
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2
Imported quantities in the perfect competition (PC), Cournot scenario
(CO), and reference data (RE) in 2005, in million tons (Mt) converted
from Petajoules for the CMT-E model . . . . . . . . . . . . . . . . . . . . 46
2.3
CIF Prices in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2005 for the CMT and CMT-E models, USD per ton 47
2.4
CIF Prices in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2006 for the CMT and CMT-E models, USD per ton 47
2.5
CIF Prices in the perfect competition (PC) and Cournot scenario (CO)
model results for different elasticity values, and reference data (RE) in
2005 and 2006 for the CMT-E model, USD per ton . . . . . . . . . . . . . 48
2.6
Historical steam coal prices: CIF(cost insurance freight) spot price in ARA
(Amsterdam-Rotterdam-Antwerpen) and price of delivered steam coal at
the German border . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.7
Imported quantities in the perfect competition (PC), Cournot scenario
(CO), and reference data (RE) in 2006, in million tons (Mt) for the CMT
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.8
Imported quantities in the perfect competition (PC), Cournot scenario
(CO), and reference data (RE) in 2006, in million tons (Mt) converted
from Petajoules for the CMT-E model . . . . . . . . . . . . . . . . . . . . 53
3.1
Price developments in the Atlantic steam coal market in the 2000s . . . . 66
3.2
Market structure of the coal market in South Africa and Colombia in the
early 2000s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3
Costs and merit-order of suppliers in the Atlantic steam coal market in
2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.1
Model players on the steam coal value-added chain . . . . . . . . . . . . . 86
4.2
COALMOD-World model structure . . . . . . . . . . . . . . . . . . . . . . 86
4.3
Endogenous cost mechanism in relation with short and long-run marginal
costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
11
List of Figures
4.4
Producer’s quality definition relative to its reserves . . . . . . . . . . . . . 94
4.5
Countries included in the COALMOD-World database . . . . . . . . . . . 96
4.6
FOB costs for all export countries implemented into COALMOD-World . 100
4.7
Linear regression of average freight rates between 2002 and 2009
4.8
Capacity and investment costs for all production nodes in the base year . 104
4.9
Capacity and investment costs for all export nodes in the base year . . . . 105
. . . . . 101
4.10 Increasing demand scenario results 2006: seaborne trade flows (in Mt) . . 107
4.11 Increasing demand scenario results 2010: seaborne trade flows (in Mt) . . 107
4.12 Increasing demand scenario results 2020: seaborne trade flows (in Mt) . . 107
4.13 Increasing demand scenario results 2030: seaborne trade flows (in Mt) . . 108
4.14 Increasing demand scenario: aggregated consumption and imports (in Mt) 110
4.15 Investments in additional mining capacity and capacity losses of producers
between 2006 and 2025 (in Mtpa) . . . . . . . . . . . . . . . . . . . . . . . 110
4.16 Increasing demand scenario: computed average prices representing the
marginal costs of supply of selected regions for all model years (in 2006
USD/t) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.17 Stabilizing demand scenario results 2020: seaborne trade flows (in Mt) . . 112
4.18 Stabilizing demand scenario results 2030: seaborne trade flows (in Mt) . . 113
4.19 Stabilizing demand scenario: aggregated consumption and imports (in Mt) 113
4.20 2020 FOB costs for all export countries calculated endogenously in the
stabilizing demand scenario . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.21 Comparison of production projections in COALMOD-World and Mohr
and Evans (2009) for bituminous coal (in Gt per year) . . . . . . . . . . . 114
5.1
Projected carbon emissions from fossil fuels and amounts in the ground . . 128
5.2
Scenario space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.3
Annual carbon dioxide emissions from steam coal consumption in the six
reference scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.4
Worldwide emissions reductions and adverse market adjustments in the
Unilateral European Climate Policy model scenario . . . . . . . . . . . . . 134
5.5
Worldwide emissions reduction in the Indonesian supply-side policy scenario135
5.6
Worldwide emissions in the CCS scenario . . . . . . . . . . . . . . . . . . 138
5.7
Unilateral European Climate policy results with and without additional
Indonesian supply-side policy . . . . . . . . . . . . . . . . . . . . . . . . . 139
12
Chapter 1
Introduction
1.1
Trying not to “carry coal to Newcastle”
Coal in its physical form has almost disappeared from our daily lives. We don’t use it to
cook or heat our homes anymore and the trains we take to travel are not pulled by steam
locomotives. However, coal remains present in our languages through idioms. “Carrying
coal to Newcastle” describes a pointless undertaking as Newcastle in Northern England
was the main coal export port during the industrial revolution. In French “aller au
charbon” means executing a cumbersome and laborious task while in colloquial German
“Kohle machen” means making money.
Before I started my research, my experience with coal was more on the historical and
literary level. On the one hand, it is easy to have a “romanticized” view of the age of
the industrial revolution as an age of great scientific discoveries and technical progress
where coal played a preeminent role. Indeed, coal was not only the fuel of the industrial
revolution but can also be seen as its trigger: the purpose of the first industrial steam
engine invented by Newcomen in 1712 was to pump water out of coal mines.1 On the
other hand, there was a much darker side to coal associated with the inhuman working
conditions of coal miners causing great social unrest as thematized in the novel Germinal
by Zola (1885).
This ambivalence of coal still exists today and increases as we are experiencing a
“renaissance” of steam coal2 , the coal type used for electricity generation that is the
main subject of investigation in this thesis. The use of steam coal is increasing in the fast
growing economies of Asia and helps millions of people to reach higher standards of living
through affordable electrification. International trade of steam coal is steadily growing
and this trend is expected to continue. But there are also some major environmental
concerns associated with the use of coal such as local pollution and, foremost, climate
change due to high intensity of carbon dioxide emissions that occur when burning coal.
In this thesis the global steam coal market is investigated using numerical modeling
1
For a complete account of the formidable history of coal I recommend reading the book: “Coal: A
Human History” by Freese (2004).
2
This expression of “renaissance of coal” has been used, for example, by Schernikau (2010).
13
Chapter 1. Introduction
methods that combine industrial organization with the applied mathematics of operations
research. I identify three main issues or subjects of investigation that I will develop
hereafter: the issue of market power, rising demand associated with trade and investments
dynamics, and climate change and policy. The outcomes of my modeling give insights
into the complexity and dynamics of the global steam coal market and make contributions
to numerical market modeling methods, and thus, I believe, reach the goal of creating
something original, or, in other words, not to “carry coal to Newcastle”.
1.2
1.2.1
Market Power
Overview
Market power is a common feature of our economies but at the same time it is complex
and difficult to grasp. Its technical definition is of utter simplicity as it is defined by
prices that are above marginal costs (Tirole, 1999). This, however, means that we fall
out of the paradigm of the perfect market and break the first fundamental theorem of
welfare economics: Pareto optimality. This situation is caused by the firms’ ability to
raise prices above marginal costs, which they often can. As Schmalensee (1982) puts it:
“absolutely perfect competition is rarely encountered outside textbooks. Almost all firms
have some market power” and what is relevant is the “importance” of market power.
Revealing and measuring market power holds an important place in economics due
to its relevance to antitrust proceedings and regulation. Kaplow and Shapiro (2007)
give an overview of the means of inferring market power. One can try to measure or
estimate the price-cost margin. Other methods have the goal to estimate factors that
may indicate or facilitate market power such as a low elasticity of demand, high market
shares or the rivals’ supply responses. These factors suggest that the industry structure
is relevant to understand market power and since we are often between the two well
understood extremes of perfect competition and monopoly there is a need for other
theoretical models of oligopolies. The distinctive feature of such oligopoly models is that
there are a few firms in the market and that these are strategically linked (Friedman,
1983). I describe some of these models in more detail in the next section.
One empirical method used to infer market power is based on econometrics. Bresnahan (1989) gives an overview of methods and studies of market power in industries. One
method described there, and widely used since, estimates a conduct parameter based on
the concept of conjectural variations. This model of conjectural variations, that will be
discussed in parts of this thesis, aims at describing a continuum of imperfectly competitive situations from perfect competition to collusion through the variation of a single
parameter. This thesis will highlight the theoretical and numerical problems associated
with this approach. In empirical studies, Corts (1998) argues that the use of conjectural
variation in the conduct parameter method is problematic because it is not valid to infer
market power with a method that is not based on a properly specified model of imperfect
competition.
14
1.2. Market Power
In this thesis, market power is dealt with in a different way, not using econometric
methods but numerical modeling methods with properly specified models of oligopoly.
This approach, initiated by Kolstad and Abbey (1984), uses numerical market modeling
based on market fundamentals such as production costs, capacities and demand and
compares different model specifications regarding competition and strategic behavior of
players with actual market outcomes.
1.2.2
Numerical modeling of imperfect competition
There exist various models of oligopoly. Competition can be based, for example, on
setting quantities like in the Cournot model or on setting prices like in the Bertrand or
Edgeworth models and the models can be static or dynamic. The starting point for the
numerical models in this thesis is the oldest model of imperfect competition: the model
by Cournot (1838). In this model, the firms decide independently what quantity they
supply to the market to maximize their profits given the output of the other firms. When
the output of any firm represents the best response to all the other firms’ output, a Nash
equilibrium is reached, the central concept in static game theory. The Cournot model
allows for the representation of market power in a non-cooperative way as well as perfect
competition (Vives, 2000).
Since the Cournot model is based on profit maximization we can use mathematical
programming which solves problems of choosing values within an allowed set in order to
optimize an objective function.3 In particular we can use the Karush-Kuhn-Tucker conditions to find an optimal solution.4 For example let us consider the following minimization
problem where we want to minimize an objective function f (x) subject to equality and
inequality constraints:
min
f (x)
x∈RN
gj (x) ≤ 0 ∀j = 1, . . . , m (λj )
s.t.
(1.1)
hi (x) = 0 ∀i = 1, . . . , n (µi )
The following Karush-Kuhn-Tucker (KKT) conditions are necessary for optimality:
∇x f (x) +
m
X
λj ∇x gj (x) +
j=1
n
X
µi ∇x hi (x) = 0
(1.2)
i=1
gj (x) ≤ 0,
λj ≥ 0,
gj (x) · λj = 0,
hi (x) = 0,
µi f ree,
∀i = 1, . . . , n
∀j = 1, . . . , m
(1.3)
(1.4)
If the objective function f and the inequality constraints gj are continuously differentiable convex functions and the equality constraints hi are affine functions, then the
3
As we will see later in this section there are other types of mathematical problems that do not
optimize functions but try to find an equilibrium state for a problem.
4
A complete description and proof of the Karush-Kuhn-Tucker theorem can be found in Dixit (1990).
15
Chapter 1. Introduction
necessary KKT conditions are sufficient for optimality. Equation (1.2) of the KKTs is
the stationarity condition and ensures that the gradient (or first-order derivative) of the
function is zero and thus an optimal point is reached. This equation is the gradient of
the Lagrange function which includes both the objective function and constraints shown
in (1.1). Equations (1.3) and (1.4) are feasibility conditions. The variables λj and µi are
the dual variables associated to each constraint.
In order to solve large scale problems or problems with multiple profit maximizing
players we can use the logic of the KKT to find an optimal solution but in a different
setting. We want to find not only the maximum profit for all the players involved but
also an equilibrium state of the market that results from the competition between the
firms. For that we need equilibrium programming that “refers to the modeling, analysis
and computation of equilibrium via the methodology of mathematical programming”.5
Since the beginning, the development of mathematical programming and economic theory have been interrelated and the method used here is called the variational inequality/complementarity approach to equilibrium programming. The variational inequality
problem is a general form that comprises the complementarity problems and can be
defined as follows (from Harker, 1993):
Let X be a nonempty subset of RN and F be a mapping from RN into itself. The
variational inequality problem, denoted by V I(X, F ), is to find a vector x∗ ∈ X such
that:
F (x∗ )T (y − x∗ ) ≥ 0 for all y ∈ X
(1.5)
A special case of the V I(X, F ) is the nonlinear complementarity problem N CP (F ):6
Let F be a mapping from RN into itself. The nonlinear complementarity problem, denoted
by N CP (F ), is to find a vector x∗ ∈ RN
+ such that:
∗ T ∗
F (x∗ ) ∈ RN
+ and F (x ) x = 0
(1.6)
When F (x) is an affine function of x, say F (x) = q + M x for some given vector q ∈ RN
and matrix M ∈ Rn×n , the problem N CP (F ) reduces to the linear complementarity
problem, which is denoted by LCP (q, M ):
x ≥ 0, q + M x ≥ 0, xT (q + M x) = 0
(1.7)
The name complementarity comes from the third condition in (1.7) which requires that
at least one of the expressions in the multiplication of xT with q + M x should be equal
to zero in the solution of the problem (Billups and Murty, 2000).7
It is important to recognize that the Karush-Kuhn-Tucker conditions of a quadratic
5
Harker (1993): p. 3
Harker (1993): p. 11
7
An alternative notation (used in the subsequent chapters) employs the perpendicular sign instead
of the multiplication. (1.7) would then read 0 ≤ q + M x ⊥ x ≥ 0
6
16
1.2. Market Power
program are a mixed LCP. This is also true for the Nonlinear Program (NLP) and the
Mixed Nonlinear Complementarity Problem (mixed NCP or MCP), a generalization of
the LCP, as shown by the following example.8 If we take problem (1.1), considering it
as a generic nonlinear program with KKTs (1.2)-(1.4) we get a mixed NCP as follows:

  
Pn
P
∇x f (x) + m
x
i=1 µi ∇x hi (x)
j=1 λj ∇x gj (x) +

  
F λ = 
−gj (x), ∀j = 1, . . . , m

hi , ∀i = 1, . . . , n
µ
(1.8)
This can be expressed as follows:
∇x f (x) +
m
X
λj ∇x gj (x) +
j=1
n
X
µi ∇x hi (x) = 0 x f ree
(1.9)
i=1
gj (x) ≤ 0,
λj ≥ 0,
gj (x) · λj = 0,
hi (x) = 0,
µi f ree,
∀i = 1, . . . , n
∀j = 1, . . . , m
(1.10)
(1.11)
We recognize the conditions (1.9)-(1.11) of the mixed NCP to be exactly the KKT conditions (1.2)-(1.4) of optimization problem (1.1). By solving the mixed NCP we obtain
the optimal solution of the optimization problem. The constraint qualifications (CQ)
ensure that the KKT solution is an optimal solution (Bazaraa et al., 2006). Typical
CQs are linearity and linear independence. In the case of a minimization problem with
convex objective function whose feasible region is defined by affine equality conditions
and convex inequalities, the KKT conditions are necessary and sufficient for an optimal
solution.
Complementarity problems help us to solve large scale equilibrium problems like market equilibria with multiple profit maximizing players. In order to add equality conditions
to a model like demand functions or market clearing conditions we have to formulate the
model in the form of a mixed nonlinear complementarity problem or simply mixed complementarity problem (MCP) as it is often referred to in the modeling literature. MCP is
a generalization of the NCP9 and thus comprises other problems like linear and nonlinear systems of equations, LCP, NCP and nonlinear programs. The profit maximization
problems of the different players included in an equilibrium model are nonlinear programs
(NLP), as we want to include nonlinear functions, for example for the cost functions. The
8
9
See Gabriel (2009): Slide 12
See Ferris and Kanzow (1999): p. 15
17
Chapter 1. Introduction
MCP is defined as:10
Given: f : RN → RN ,
l, u ∈ RN ,
Find:
z, w, v ∈ RN ,
s.t:
F (z) − w + v = 0,
l ≤ z ≤ u,
(1.12)
w ≥ 0,
wT (z − l) = 0,
v ≥ 0,
v T (u − z) = 0
with −∞ ≤ l ≤ u ≤ +∞ and F being continuously differentiable. We recognize the
complementarity relationship stated before, w and v being the dual variables of the
lower bound l and upper bound u, respectively.
The MCP format is also the format understood by the GAMS modeling language
and the PATH solver used for MCP.11 GAMS stands for General Algebraic Modeling
System and was developed in the 1980s to help economists at the World Bank to make
quantitative analyzes of economic policies. Since 1987 this system has been further developed by the GAMS Development Corporation. Today GAMS can resort to plenty
of different solvers to solve various mathematical programming problems such as linear
programming, non-linear programming, integer programming, and of course complementarity programming. A description of the PATH solver with an example of a GAMS
model code is provided by Ferris and Munson (2000).
So far I have considered one-stage representations of equilibria of imperfect markets
such as the Cournot model solved in the MCP format that represents the core of the
modeling work in this thesis. But two-stage games are also relevant for this thesis and
for a better understanding and representation of market power. One of the first examples
of two-stage games found in the literature is the model by Stackelberg (1934) with two
firms where one firm, the leader, chooses her optimal output first, knowing what the
follower firm will do in all possible cases. The Stackelberg equilibrium represents a
subgame perfect Nash equilibrium of a two-stage perfect information game (Vives, 2000).
This model can be expressed as a mathematical program with equilibrium constraints
(MPEC). An MPEC encompasses an optimization problem that can represent the leader
firm’s profit maximization and an equilibrium problem that can represent one or multiple
followers competing in a Cournot market. MPECs are generally hard to solve since
normal constraint qualifications do not hold due to the non-convexity of the feasible
region (Leyffer and Munson, 2010). However, it is possible to solve them in GAMS as
an MPEC class of models (Dirkse and Ferris, 1999), but being careful as the solution
delivered might only be local.
In order to achieve a more realistic market representation we may want to include
more than one player on the upper level problem. This problem is called an equilibrium
problem with equilibrium constraints (EPEC) since we have to consider an upper level
10
11
See Rutherford (1995): p. 1301 et seq.
See Ferris and Munson (2000)
18
1.2. Market Power
equilibrium, for example a Cournot model, where the players also need to take into
account the actions of the players of a subsequent lower level equilibrium. These model
are even harder to solve and are at the current research frontier. First methods for solving
EPECs are currently under development, for example by Leyffer and Munson (2010).
1.2.3
Market concentration in the international steam coal trade
To motivate the use of numerical models with market power representation we first have
to look at the market concentration in the international steam coal trade. Steam coal
used for electricity generation is primarily a domestically produced and consumed fuel
due to relatively high transport costs. However, in the wake of the oil crisis of the
1970s, and the subsequent move away from oil as main fuel for industry and electricity
generation, international seaborne steam coal trade flows started growing. What followed
in the next 30 years was a period of tremendous growth in the international steam coal
trade with a doubling of trade volumes every 10 years from 78Mt in 1980 to 597 Mt in
2010. The share of international seaborne trade in global steam coal consumption also
rose, from 3% in 1980 to 10% in 2000 and stabilized around 11% until 2010 (IEA, 2011a).
In this period of time the structure of trade also underwent several changes regarding
the actors involved.
In the early 1980s the main exporters were countries with a large domestic production
and consumption base such as the US, Poland and South Africa. From 1985 on, more
export driven players emerged such as Australia, Colombia and Indonesia. After the year
2000 high growth continued, especially of Indonesian exports and new players entered
the market such as China and Russia (EPRI, 2007).
The demand side of the international steam coal trade also underwent some significant
changes. Traditionally the international steam coal market can be separated into two
sub-markets: the Atlantic basin with Europe as main importing region and the Pacific
basin with the main East Asian economies as importers. The separation is due to the fact
that most of the trade flows occur between suppliers and importers of one basin due to
the long transport distances and associated freight costs. However, trade between basins
occurs and recent studies show that the regional markets are more and more integrated (Li
et al., 2010, Zaklan et al., 2009). In the early 1980s the volumes in the Atlantic steam coal
market represented around 90% of total seaborne trade but the Pacific steam coal trade
grew continuously and overtook the Atlantic market in the early 1990s. Until 2000 the
market shares were around 45% and 55% for the Atlantic and Pacific basins respectively.
In the early 2000s the Pacific market grew faster and expanded tremendously after 2005.
In 2010, the Pacific steam coal market made up for 73% of the international seaborne
trade (IEA, 2011a).
In the 1980s long-term contracts of up to 10 years were dominant in the international
steam coal trade. In the Atlantic steam coal market the nature of contracts changed
radically towards short term contracts which typically represent the delivery of a cargo
or series of cargoes in the next 3 months. These spot transactions grew from 14% of the
19
Chapter 1. Introduction
Atlantic market in 1983 to 80% in 2003 (Ekawan and Duchêne, 2006). In the Pacific
market there was a move towards long-term contracts with annual price negotiation or
annual contracts and some spot trade (Ekawan et al., 2006). By 2010 the Atlantic steam
coal market had developed into a highly liquid market with a significant amount of paper
trade whereas in the Pacific market long-term contracts are still dominant with shares
of more than 80% for most importers (IEA, 2011c).
In the 1980s and 1990s the international steam coal market was characterized by low
levels of market concentration on the supply side with a constant oversupply that lead
to low import prices around 35 USD/t with a slightly declining trend. Many relatively
small mining companies were active in the market as well as some oil majors that left
the market in the late 1990s, early 2000s. In the same time period, roughly from 1998
to 2002 a major consolidation of the industry took place in South Africa, Australia and
Colombia that led to the emergence of four multinational mining giants, the “Big Four”:
Anglo American, Rio Tinto, Xstrata/Glencore and BHP Billition. During that time
the share of these companies in the global steam coal seaborne trade grew from 29% to
33% (Couser and Goldsack, 2001). In 2005 this share was 32% and the position of the
Big Four has not grown stronger since; in 2007 it had fallen back to 27% (Kopal, 2007,
Rademacher, 2008). The position of the Big Four in the Atlantic market is stronger: in
the period 2002 to 2008 they were responsible for 80% of the exports from South Africa
and Colombia. However, the share of the Big Four in the total Atlantic market has been
steadily declining from 49% in 2002 to 43% in 2004, less than 40% after 2006 and 33%
in 2008.12
1.2.4
Research questions and modeling challenges
The market consolidation and the increased concentration on the international seaborne
steam coal market in the late 1990s and early 2000s lead to concerns about the use of
market power by the dominant suppliers (Couser and Goldsack, 2001). The fear about
collusion and the development of a cartel of coal suppliers, a “COALPEC” similarly to
the OPEC on the oil market was also present (EPRI, 2007, Haftendorn et al., 2008b).
The abuse of market power could be one determining factor in the sudden rise of
import prices from around 35 USD/t before 2003 to almost 80 USD/t in 2004, remaining
on average higher than 60 USD/t in 2005 and 2006. Since the peak in the share of the
“Big Four” in the international steam coal market was around 2005 it is very relevant to
investigate the global market structure around that time to see with a partial equilibrium
model if the trade flows and price levels are in line with a competitive market or if market
power is present.
Furthermore, a closer look into the Atlantic steam coal market might reveal strategic
behavior. The market share of the “Big Four” on this market is significantly higher than
on the global market and the market is more liquid with a majority of short-term trades
which gives more room for strategic behavior than long-term contracts.
12
These figures were derived using the companies annual reports and the IEA (2011a) Coal Information.
20
1.3. Trade and Investments Dynamics under Rising Demand
From a technical modeling point of view the representation of non-cooperative Cournot
equilibria as well as competitive markets is well understood. However, for intermediary
cases, for example with a few dominant players, the concept of conjectural variations is
often used as a basis for the market power representation. These conjectural variations
need to be better understood as they can lead to counter-intuitive results. Better models
of dominant oligopolies with a competitive fringe need to be developed.
1.3
1.3.1
Trade and Investments Dynamics under Rising Demand
Rising demand in Asia
The last decade from the years 2000 to 2010 has been a “coal decade”. During that
period the incremental use of coal as primary energy was equivalent to that of gas, oil
and renewables, all put together, as highlighted in the latest World Energy Outlook by
the International Energy Agency (IEA, 2011c). The share of coal in primary energy grew
from 23% in 2000 to 28% in 2010. Most of this growth took place in Asia where steam
coal is the dominant fuel for electricity generation and electricity demand is strongly
correlated with GDP growth.
In 2010 Asia accounted for 72% of global steam coal consumption followed by America
with 17% and then Europe including Russia with 7%. China alone accounted for 52%
of global steam coal consumption. This rising demand in Asia also led to rising imports
and growth of the Pacific market as discussed before. In 2010 the first five importers of
steam coal are Asian countries: Japan and China with 129 Mt each, South Korea with
91 Mt, and then Taiwan and India around 59 Mt each. The Asia-Pacific region accounts
for 67% of world imports and the first European importers Germany and the United
Kingdom are far behind with import levels of 38 and 29 Mt respectively.
China has a particularly important influence on the world market as the size of its
internal market is more than three times the size of the global international trade. China,
a net exporter in the early 2000s, saw its exports steadily declining from a maximum
of 80 Mt in 2004 to less than 20 Mt in 2010. On the other hand, imports grew. After
2008 and a doubling of import levels China became a net importer of steam coal for the
first time and is expected to remain one. The reduction in exports can be explained
by the increase in internal demand that is hardly satisfied by internal supplies and the
increasing imports by the attractivity of imported steam coal, especially from Indonesia,
in the coastal areas of Southern China that are far away from the coal basins in the
North.
In India steam coal is also the dominant fuel for electricity generation and its consumption almost doubled since the year 2000 to reach the level of 560 Mt, the third
place after China and the USA. The development of imports was even more dramatic,
rising from 10 Mt in 2000 to almost 60 Mt in 2010 (IEA, 2011a). The rise of imports
can be explained by structural problems in the Indian coal mining sectors with a low
productivity, low quality of coal and and high transport costs which makes imported coal
21
Chapter 1. Introduction
especially attractive in the highly populated coastal areas (Haftendorn et al., 2008a).
In the latest World Energy Outlook by the International Energy Agency (IEA, 2011c)
more than two-thirds of the projected increased coal demand in the Current Policies and
New Policies scenarios will come from China and India. The first scenario represents the
business as usual case whereas in the second some degree of climate policy is implemented.
In the Current Policies scenario global coal demand is projected to rise by 64% from 2009
to 2035 and by 24% in the New Policies scenario. Most of the growth will come from
Asia and some coal-rich countries such as Russia and South Africa. Demand from the
OECD is projected to rise marginally in the Current Policies scenario and to decrease in
the New Policies scenario.
The trend observed in the decade 2000 to 2010 with an increase in coal consumption
in Asia is likely to continue. In certain projections this increase will be very strong and
this opens the question of the potential supply sources to serve that demand.
1.3.2
A complex supply side
Coal is the result of a biological and then geological process that transformed trees and
other plants over millions of years by trapping them under sediments in the earth’s
crust. Pressure, temperature and time increase the quality of coal towards higher carbon
contents. This quality parameter measured in energy unit per unit of mass is also the
main driver behind the economics of coal (Minchener, 2009). Higher quality coals are
called hard coal. They are traded internationally and can be transported over long
distances. The majority of produced hard coal is steam coal for electricity generation
and a smaller fraction is higher quality coking coal used for steel production. On the
other hand, lower quality brown coal is only used domestically for electricity generation
near the mines as it is not economical to transport it over large distances.13
Coal is the most abundant of all fossil resources. Measured in energy values the
reserves of hard coal with 18 ZJ14 are higher than the reserves of all conventional and
unconventional oil and gas resources of 16 ZJ (DERA, 2011). Reserves can by definition be extracted with current technologies at current market prices. Resources include
quantities that may be extracted in the future but are not yet economical or geologically
proven. The hard coal resource base is even more abundant with an estimated 426 ZJ
which is more than four times the resources of oil and gas and more than twenty times
the resources of conventional oil and gas. Hard coal resources are also more evenly distributed on the surface of the globe with large resource bases in North America, Eurasia
and the Asia-Pacific region. However, if we look at the distribution on a per country
basis we find that 82% of the reserves are found in only five countries: the US, China,
India, Russia and Australia (DERA, 2011).
Coal may be abundant but the the share of resource that is technically accessible and
13
One can also find the terms “metallurgical coal” for coking coal and “lignite” for brown coal in the
literature.
14
Zettajoule: 1 ZJ = 1021 Joules = 3.414 · 1010 tons of coal equivalent
22
1.3. Trade and Investments Dynamics under Rising Demand
can be economically brought to the market is significantly smaller since a capital-intensive
infrastructure is needed to bring the coal to the final users. The value-added chain of
steam coal starts at the mine where the coal is extracted. It is then transported overland
either directly to the final customer or to a coal export terminal where it is loaded on
bulk transport vessels that ship the coal overseas. On each step of the value chain,
different technologies might be used given the physical and geological characteristics of
the deposit and its location and this heterogeneity also affects the cost structure. For
example, mining can occur underground which is more labor and capital intensive. In
opencast mining large draglines are cheaper to remove the overburden and extract the
coal and the machinery can be powered by electricity. In less favorable terrain the truck
and shovel techniques uses a lot of explosives, tires and diesel (IEA, 2011b). The overland
transport can be done by barge, by truck over short distances or by train. Coal export
terminals require large investment sums and years of planning, even for the expansion of
ports already in place.
As mining and transport infrastructure investments may take up to four years to
come online, a certain commodity cycle could be observed in the past. Moreover, the
coordination of the expansion of the different supply chain components is difficult and can
lead to short to medium term bottlenecks (IEA, 2011b). An indicator for the tightness of
the market is the mine capacity utilization rate that has varied between 70% and 95% in
the past 20 years and remained higher than 85% from 2003 to 2009 due to the high Asian
demand (Rademacher and Braun, 2011). Investments remain strong as in 2008 and 2009
the 30 leading coal companies invested around 15 billion USD in mining projects and 16
billion USD in 2010 (IEA, 2011c).
1.3.3
Research questions and modeling challenges
In view of the expected increase in demand in Asia in the mid- and long-term and a
supply side that is operating close to its capacity limit, the main issue concerns the
investments. How much investments are needed in different demand scenarios? Can
certain capacity and investment restrictions affect the market adversely? On what stages
of the value-added chain will investments be necessary and specifically in what mining
basins, transport links or export terminals? Minchener (2009) gives an overview of
the possible future coal supply prospects. These are diverse and include for example
bringing coal to the market from new countries with high potential such as Mozambique
or connecting the Powder River Basin, the biggest coal mining basin of the US, to the
Pacific market by expanding railways and export terminals.
In this thesis I will investigate the future of the global steam coal market and answer
the above questions using a partial equilibrium numerical model. To do so the valueadded chain of steam coal has to be represented properly. Important characteristics
such as the quality of coal or the endogenous evolution of costs have to be represented.
This representation has to be detailed enough to be able to draw conclusions about the
components of the value added chain such as transport links or export terminals but also
23
Chapter 1. Introduction
aggregated enough so that the size of the model and data collection remain manageable.
Data for a numerical model is a challenge by itself. There exists no central database for
the coal market and data has to be collected from a wide range of heterogeneous sources
and aggregated, disaggregated, calculated or estimated to fit the specifications of the
model.
1.4
1.4.1
Climate Change and Policy
Stand or fall by coal
Unmanaged climate change represents a risk for humankind on an unprecedented scale.
As Stern and Rydge (2012) put it, a business as usual could lead to an increase of
carbon dioxide concentrations to a level of 750ppm and a warming of more than 5°C to
temperature levels never seen on earth since 30 million years and thus never experienced
by human societies. Disruptions to local habitats and economies through floods, droughts
and water scarcities would have far reaching consequences on a global level.
Climate change mitigation requires a response on a global level to decarbonize the
energy sector. This response is usually framed around two lines: the deployment of
renewable energy technologies and increased energy efficiency. However, another view is
possible: it is the imperative to move away from coal as a source of energy. Indeed, in
the different climate policy scenarios of the latest World Energy Outlook (IEA, 2011c)
the share of coal in world primary energy demand is the most sensitive to different levels
of climate policy compared to the other fossil fuels. In the most stringent climate policy
scenario that aims at stabilizing carbon dioxide level at 450ppm the consumption of coal
significantly drops in 2035 compared to 2008 levels whereas for oil and gas it is more or
less constant.
The criticality of coal to climate change mitigation is due to two factors: its high
carbon dioxide emission factor and the low efficiency of steam coal power plants. The
emission factor that describes how much carbon dioxide is released by burning one energy
unit of a fuel is twice as high for coal as for gas (DEHSt, 2005). Also the energy efficiency,
that is how much of the energy contained in the fuel can be transformed into electricity, is
relatively low for coal power plants. As of 2010, 75% of the worldwide installed capacity
were sub-critical coal power plants with energy efficiencies of less then 40% (IEA, 2011c).
In comparison combined cycle natural gas power plant can reach efficiencies of up to
60%.
If a stringent climate policy is implemented, coal will lose its competitiveness in the
power market. Sijm et al. (2005) show that natural gas power plants become more
competitive than coal in the merit-order curve at carbon dioxide prices as low as 18.5
Euro per ton. It then comes to no surprise that coal is only mentioned twice in the
Energy Roadmap 2050 of the European Commission (EC, 2011): first as a fuel that will
be substituted by the “transition fuel” natural gas and second with the remark that coal
could continue to play a role in Europe only with the use of the carbon capture and
24
1.5. Overview of the Thesis
storage technology. This technology to capture carbon dioxide at the power plant and
sequestrate it underground was once seen as the great hope for a sustainable use of coal
(IPCC, 2005). However, recent developments show that the promise of “clean coal” on
a large scale is fading. Numerous demonstrations projects have been put on hold or
canceled in Europe (Hirschhausen et al., 2012b).
1.4.2
Climate policy must not neglect global market dynamics
After the failure of the climate conference in Durban to reach a globally binding climate
agreement and the announcement there that China might only consider such an agreement after 2020 the outlook for climate policy on the global level is a quite unfavorable
in the mid-term. This means that for the time being, only national or regional climate
policies will be implemented. This creates a series of problems such as free-riding, since
the benefits of climate policies are shared but the costs are local. On top of that there
is the problem of leakage. This can happen in the goods market where investments may
be re-located to countries with no carbon costs and then the goods will be exported to
countries with climate policies. A second effect goes through the fossil fuel markets where
the reduced use in a country with climate policy incentivizes countries with no climate
policy to consume more due to a lower world price for the resource (Tirole, 2012).
While the first type of carbon leakage has been the most studied in the literature,
Böhringer et al. (2010) find that there is a significant share of leakage in energy markets
that can undermine policies targeting carbon leakage in goods such as border tax adjustments. Hence, there is an urgent need to better understand and quantify carbon leakage
in energy markets in a world where regional or national climate policies are likely to be
the only type of climate policies implemented in the mid-term.
1.4.3
Research questions and modeling challenges
The mechanism of carbon leakage in the global steam coal market is a reality derived
from the basic economic relation between price and demand. But what needs to be
assessed is the relevance and strength of this effect in the actual market context. The
effects of the size of the players expressed by their share in global exports or imports and
the magnitude of the quantity effects need to be taken into account. Regional policies
around the world should also be taken into consideration.
From a modeling perspective, the challenge is to find meaningful and realistic specifications for the global conditions as well as for the regional climate policies that might
induce carbon leakage in that context. The scenarios shall deliver valuable insights into
the effects of carbon leakage we can expect in the global steam coal market.
1.5
Overview of the Thesis
This thesis is based on four original research articles that are presented in Chapters 2
to 5. The following overview organizes the papers according to the three research issues
25
Chapter 1. Introduction
identified in the first part of this introduction and provides information regarding my
contribution and the state of publication. I am the main author of Chapters 2, 4 and 5
and the sole author of Chapter 3.
Market power:
Chapter 2: Modeling and analysis of the international steam coal trade: is
the market competitive?
◦ This chapter is joint work with Franziska Holz.
We jointly developed the
COALMOD-Trade model. I had the main responsibility for the data search and
writing.
◦ This work was published in The Energy Journal, Vol. 21, No. 4, 2010, pp. 205-229.
This is also the version of the paper used in the thesis.
Chapter 3: Atlantic steam coal market power: theory and application of
oligopoly models with a competitive fringe
◦ This chapter is my independent research.
◦ This work was published as DIW Discussion Paper 1185, 2012. This is also the
version of the paper used in the thesis. This paper was distinguished by the “Student
Paper Award” at the 35th IAEE International Conference 2012 in Perth, Australia
and accepted at the EEA 2012 conference in Málaga.
Trade and investment dynamics under rising demand:
Chapter 4:
A techno-economic analysis using the COALMOD-World
model: the end of “cheap coal”?
◦ This chapter is joint work with Franziska Holz and Christian von Hirschhausen. I
had the main responsibility for building the COALMOD-World model, for most of
the data search, calculations and for the writing.
◦ This work was published as DIW Discussion Paper 1067, 2010, and PESD Stanford
University Working Paper 96, 2010. An updated version has been accepted for
publication in the journal FUEL. This is the version of this chapter complemented
by additional sections that describe all the data sources.
Climate change and policy:
Chapter 5: Climate policies and the global steam coal market: interactions
until 2030
◦ This chapter is joint work with Claudia Kemfert and Franziska Holz. I had the
main responsibility for the scenario definition, the data search and model runs as
well as the writing.
◦ This work was published as DIW Discussion Paper 1146, 2011. An updated version
has been accepted for publication in the journal Energy Policy. It is the basis for
this chapter that also includes an additional scenario.
A major contribution of this thesis lies in the development of numerical partial equilibrium models to investigate the global steam coal sector. Table 1.1 gives an overview
26
1.5. Overview of the Thesis
Table 1.1: Models specifications in the chapters
Chapter
Model name
Time
Coverage
Market structure
Format/Solver
2
COALMOD
static
International
Perfect competition,
MCP/PATH
-Trade
2005-06
seaborne trade
Cournot competition
3
n/a
static
Atlantic market
Oligopoly as Cartel,
MPEC/MPEC
2003-06
seaborne trade
Cournot competitors,
MCP/PATH
Nash bargainers, with
NLP/CONOPT
a competitive fringe
4-5
COALMOD
Multi
International
-World
-period
seaborne and
2006
overland trade
-2030
Perfect competition
MCP/PATH
and major
domestic markets
over the different specifications and coverage of the models developped in this thesis. All
models are implemented in GAMS. The format is the type of model used from the theory
of operations research while the solver is the algorithm that solves the model numerically.
The modeling of the international steam coal after pioneering work by Kolstad and
Abbey (1984) has again found interest in the literature recently. Paulus and Trüby
(2011a) analyze the effects of Chinese infrastructure decision on the global trade flows
while Paulus and Trüby (2011b) and Paulus et al. (2011) focus their investigation on
strategic behavior and market power especially in the Pacific Market. The market power
analysis with the COALMOD-Trade model of Chapter 2 of this thesis predates those
works and the focus of Chapter 3 is on the Atlantic market and also on the theoretical approaches to market power modeling. Chapter 4 presents the COALMOD-World
model for the investigation of long-term developments and Chapter 5 concentrates on an
application of this model to climate policy issues.
1.5.1
Chapter 2: Modeling and analysis of the international steam coal
trade: is the market competitive?
In the early 2000s a movement of market concentration on the global steam coal market
led to the emergence of the “Big Four” multinational companies in Australia, South Africa
and Colombia. In the other major exporting countries such as Indonesia, Russia, China
and the US, only a few national champions dominate the export market. In this situation
market power may emerge. In Chapter 2, we analyze the international seaborne trade for
steam coal in a similar way as was done by Kolstad and Abbey (1984) for the market in
the early 1980s using partial equilibrium numerical modeling and countries as strategic
players.
While we are aware that companies are the active players on the global market, we
choose to use a country set-up. A country specification is easier to implement from
a data perspective and can give us first valuable insights into the market structure.
Company data, especially in the Pacific market, is hard to find and it would not have
27
Chapter 1. Introduction
been possible to properly specify the model. Also, nationalistic resource policies through
taxes and export restrictions are at work in certain countries such as Indonesia and China
as identified by Paulus et al. (2011). This also speaks in favor of a country set-up as a
first approach.
Chapter 2 develops a numerical partial equilibrium model of the international trade of
steam coal, the COALMOD-Trade model, based on the mixed-complementarity (MCP)
approach. The model consists of eight profit maximizing exporters constrained by production and export capacities that sell coal to 10 importers represented by an inverse
demand function specified for the years 2005 and 2006. The exporters can be modeled
as Cournot players or as competitive players. Furthermore, two different versions of the
model are tested, one where mass quantities of coal are traded and the other one where
it is the energy content of the coal expressed in energy units.
For both model specifications, mass and energy, the simulation of perfect competition
better fits the observed real market flows and prices. The prices are in the range of the
real prices while the Cournot prices are significantly higher. A sensitivity analysis with
respect to the price elasticity of demand also shows that the price results of the perfect
competition case are robust while the Cournot prices vary by more than 20%. The total
traded quantities are in line with the perfect competition case while in the Cournot case
there is more diversification as observed in reality. However, trade flow diversification by
itself is not an argument for market power as there are other reasons for diversification
such as security of supply or historical trade relations. This leads to the main finding of
this chapter that we reject a Cournot market structure on the international steam coal
market for the years 2005 and 2006.
Furthermore, we find that the energy specification of the model leads to a better
representation of the trade flows. The customers in fact want the energy in the coal and
this is also reflected in the market trade flows and prices. The energy specification of the
model also considers the mass of the coal through a conversion factor for the capacity
constraints and productions and transport costs.
Based on the modeling results, Chapter 2 also examines more closely the prices and
pricing mechanisms. While the modeled prices in the perfect competition case are in
the range of the real market prices, the variation of the real prices between countries is
significantly higher than the computed prices. Possible reasons for these higher variations
can be price discrimination, historical trade relations, the influence of domestic markets or
the nature of contracts and pricing that can negatively affect the flexibility and reactions
of the market.
In that context the timing issue of contracts is subject to scrutiny in this chapter.
Spot contracts are for deliveries three months ahead and long-term contracts with yearly
price adjustments are dominant in the Pacific market. Adding the relative intransparency
and lack of liquidity in the spot market as there are no central commodity exchanges but
just bi-weekly price indices based on quoted trades, a time-lag can arise in the pricing-in
of capacity constraints. This delays and weakens the signal that the price is sending to
28
1.5. Overview of the Thesis
incentivize new investment. This may have contributed to the run-up in prices observed
in 2007 and 2008 as the increased demand from China and India was not met by sufficient
global investments in production and export capacities.
1.5.2
Chapter 3: Atlantic steam coal market power: theory and application of oligopoly models with a competitive fringe
Having found in the previous chapter that the international market for steam coal in
the time of the peak in global market concentration in the years 2005 and 2006 is better
represented by a competitive model, we analyze the market conduct in the Atlantic steam
coal market where a higher market concentration and market liquidity may have favored
strategic behavior. The Atlantic steam coal market in the early 2000s was characterized
by three dominant companies active in two countries, South Africa and Colombia, and
by what we can describe as a competitive fringe composed of price-taking players from
countries such as Russia and the US. In order to represent such a market in a numerical
partial equilibrium model we cannot resort to the standard Cournot approach since we
assume to have some competitive price-taking players alongside a dominant oligopoly.
In order to represent imperfect competition in markets that lie between the Cournot
and perfect competition assumptions the literature has often resorted to the concept of
“conjectural variations” (Figuières, 2004). This concept is derived from the Nash-Cournot
model and describes the expected or “conjectured” reaction of players to the action of
other players. It is often stated that in a Nash-Cournot equilibrium players do not have
an incentive to deviate from their strategies, hence they will not react to changes of other
players and their conjectural variation is zero. This is an interpretation that may be valid
in equilibrium, however, the conjectural variation approach misinterprets and translates
this into a model that describes how to get to an equilibrium. Chapter 3 provides an
overview of the theoretical literature of conjectural variations that came to discredit the
approach described above due to the ambiguity of its equilibrium concept.
However, in the numerical applications in models of energy and resources, the conjectural variations concept is still widely used. For example, the gas market model of
the PRIMES model (E3MLab, 2011) that was used for the recent Energy Roadmap 2050
of the EU (EC, 2011) uses a conjectural variation approach to model competition between upstream producers. While the conjectural variation approach may be used as an
approximation when there is not enough information about the market structure of an
imperfectly competitive market, it must be used carefully as it can lead to counterintuitive results. In particular, a dominant player may in a conjectural variation model be
worse off than if he had exercised less market power, because he is assumed to have a
naive view regarding the reaction of the competitive players.
Chapter 3 strives to provide a better understanding of the concept of conjectural
variations from the theoretical basis to its implementation in numerical models. We
use an analytical duopoly model in different settings to show that the best response of
a player can be in conflict with the assumptions used in numerical partial equilibrium
29
Chapter 1. Introduction
models. In particular, we show that the use of mixed aggregated conjectural variations
can lead to outcomes that are inconsistent with the actions of rational profit-maximizing
players. Based on this, we claim that market power analysis with numerical partial
equilibrium models cannot be based on the conjectural variations approach.
We develop alternative models for the representation of markets with dominant players and a competitive fringe based on the Stackelberg model. This is a two-stage approach
where the dominant oligopoly takes fully into account the reaction of the competitive
fringe. We implement three different representations for the dominant oligopoly: a noncooperative Nash-Cournot competition, a cooperative Cartel with profit-sharing and a
cooperative Nash-bargaining approach without profit-sharing. The translation of these
models into numerical models solved in GAMS lies at the current research frontier of
operations research.
The numerical models of imperfect competition are applied to the Atlantic steam
coal market for the years 2003 to 2006, as well as a perfectly competitive benchmark
model. Before 2004 South Africa was the dominant steam coal exporter to the European
market. However, a new market situation with rising global demand and prices makes
room for a new entrant: Russia. The hypothesis investigated in Chapter 3 is that the
three incumbent dominant firms located in South Africa and Colombia reacted to that
new situation by exerting market power and withheld quantities from the market.
The model with a Cournot oligopoly as leader delivers the best reproduction of the
actual market situation in 2004 meaning that the dominant players exerted market power
in a non-cooperative way without profit-sharing. However, this use was only punctual
and in the year 2003 as well as in 2005 and 2006 the perfectly competitive model better
fits the real data of trade flows, which is in line with the results previously obtained in
Chapter 2.
The results of Chapters 2 and 3 show that the fears of structural market power
or even cartelization on the global steam coal market are unwarranted. The market
concentration has stabilized on a level significantly lower than in other fuel industries and
there exist a sustained investment activity and low barriers to market entry (IEA, 2011c).
Market power may arise punctually in regional markets but the market dynamics have
usually resolved this issue without the need for intervention of competition or regulation
authorities.
1.5.3
Chapter 4: A techno-economic analysis using the COALMODWorld model: the end of “cheap coal”?
The global steam coal market underwent some dramatic changes during the last decades
and further structural changes and shifts are expected in the future. This chapter presents
the COALMOD-World model that was developed to identify the main drivers of the market today and in the future. Taking into account all the factors that will affect the market
such as geology and costs, infrastructure and investment costs, demand projections and
energy and climate policy, the model gives insights into how the market might evolve
30
1.5. Overview of the Thesis
until 2030 and provides a tool for the counterfactual analysis of a wide range of scenarios. The main differences with the COALMOD-Trade model and the Atlantic steam
coal market models of the previous chapters are the multi-period time dimension and the
coverage of internal markets and overland trade in addition to the global seaborne trade.
The COALMOD-World model draws upon two significant findings from Chapters 2
and 3. First we found out that the global steam coal market may be better represented by
a competitive market model and that given current market trends the rise of structural
market power is unlikely. The assumption of perfect competition has several advantages
from a modeling perspective. Perfectly competitive numerical models are theoretically
well understood and easy to solve numerically. Also the perfect competition premise
allows us to make simplifying assumptions about the players. We can separate the
value-added chain vertically from the mine to the final customer and we can horizontally
aggregate players such as companies and this will not affect the outcomes since in a
perfectly competitive set-up demand will always be served in a way that minimizes costs.
The second finding from the previous chapters useful for our modeling is that a model
considering the energy and mass of the coal delivers a better reproduction of the market.
This is especially important for countries such as Indonesia that sell significant amounts
of lower-rank coal or for the modeling of internal markets where the quality of the coal
used is often inferior than seaborne traded coal.
The profit-maximizing players in the model are the producers and the exporters.
The model producers aggregate the companies active in one mining basin. They bear
the costs for mining the coal and transporting the coal overland. They can either sell the
coal directly to domestic demand regions or to the exporters. The exporters aggregate
the export capacity of a region and bear the port operating costs as well as the freight
costs for overseas transport. They can sell the coal to all the demand nodes with import
capabilities. Demand is represented by an inverse demand function. The players are
subject to various constraints such as reserve, production and transport capacities for
the producers and export capacity constraints for the exporters. The players maximize
their profit until 2030 using a net present value approach with perfect foresight about
future market situations. They can invest to increase the capacities and lift previous
constraints on production, inland transport and exports. Thus, the model shows how
future demand may be served optimally in a cost minimizing way.
One unique technical feature of the model is the endogenous production cost and
mine mortality mechanism that was developed for COALMOD-World. Based on cumulative extraction, investments and the specific geological and economical characteristics
of a mining basin, the short-term cost functions change endogenously in the model and
production capacity is reduced to simulate mine mortality. This allows for a more realistic depiction of the dynamics of a mining basin. To our knowledge it is the first time that
such a mechanism is implemented in a partial equilibrium numerical resource model.
The model specification we use consists of 25 producers, 15 exporters and 41 demand
nodes and covers virtually all steam coal production and consumption in the world with
31
Chapter 1. Introduction
the main exporters and importers as well as a detailed representation of the internal
markets of China, India, Russia and the US. The 1671 single data parameters the model
uses for all the cost and capacity components of the value-added chain and demand
have to be gathered from various sources or calculated and estimated and represent a
significant part of the work in this chapter.
The model calculates yearly market equilibria for the years 2006, 2010 and then up
to 2030 in five years steps. One first result and validation of the model is that the model
predicted the massive shift of trades flows from South Africa, traditionally the main
supplier in the Atlantic market to India. For the subsequent years two scenarios are
implemented based on the International Energy Agency’s 2010 World Energy Outlook
(IEA, 2010). In the “increasing demand” scenario, based on the IEA Current Policies
scenario, it is assumed that as of mid-2010 no change in the current policies will be
implemented and that the recently announced commitments are not acted upon. In
the “stabilizing demand” scenario, based on the IEA New Policies scenario, the recently
announced commitments and policies, for example the ones of the 2009 Copenhagen
Climate Conference, are fully implemented.
The most significant result is the shifting of trade flows towards the Asian markets.
We start in 2006 with a global integrated market where South Africa and Colombia are
the main suppliers to the Atlantic market and Indonesia and Australia to the Pacific
Market. Then a gradual shift eastwards occurs until 2020 with flows from South Africa
being directed toward Asia and, especially, India. Colombia replaces South Africa as
the key supplier to Europe. The second shift occurs in 2020. We see an additional shift
westwards with Colombia delivering to Japan and Korea, resource poor countries with a
high willingness to pay. By 2030, the overall picture on the global market has significantly
changed: Russia is the main supplier to Europe and South Africa, Europe’s traditional
supplier, is a major supplier to India and the Pacific market whereas Colombia becomes,
principally, a Pacific market supplier. This shift happens in both the “increasing demand”
and “stabilizing demand” scenario but in the latter scenario the volumes of the trade flows
are smaller and the shifts occur later in time.
From 2006 to 2030 global seaborne trade rises by 86% in the “increasing demand”
scenario and its share in total steam coal consumption rises from 16% to 21%. In the
“stabilizing demand” scenario trade rises by only 46% but its share in total consumption
remains high and reaches 18% in 2030. This is due to the fact that imports are expected
to remain an attractive source of coal for India and China. These two countries account
for half of global trade in 2030 in both scenarios and their imports are multiplied by 12.6
in the “increasing demand” scenario and by 9.5 in the “stabilizing demand” scenario in
the time period from 2006 to 2030.
The model includes constraints on the mining and export capacity that can be added
every five years based on recent historical data when investments were high but still
struggling to meet demand in time. These restrictions have an important effect in the
“increasing demand” scenario and lead to higher coal prices. However, contrary to some
32
1.5. Overview of the Thesis
recent papers discussed in this chapter, we do not see an “end of cheap coal” due to
reserve depletion. Coal will stay abundant for decades. However, in a scenario with high
demand increase, investments might not keep up leading to high coal prices and volatility
as experienced in recent years.
1.5.4
Chapter 5: Climate policies and the global steam coal market:
interactions until 2030
In the previous chapters we found that neither market power nor resource scarcity are
severe issues affecting the use of coal and the global steam coal market. The last big
remaining issue from those presented in the first part of this introductory chapter is the
problem of climate change. There is no doubt about the fact that the continued use of
coal is one of the major challenges to tackle in order to prevent dangerous climate change
as burning steam coal for electricity generation is one of the largest sources of carbon
dioxide emissions. In fact, the prerequisite for any serious global climate mitigation
scenario, as laid out by the International Energy Agency (IEA, 2011c), is a reduction in
coal utilization. The more ambitious the climate goals are, the less coal plays a role in
the global energy mix. In a world with a global climate policy and a global carbon price,
this price would provide the incentives to reduce the use of coal. However, the latest
failures of the international community to reach an agreement at the UNFCCC COP15
conferences in Copenhagen and Durban gives a bleak outlook on global climate policy
in the mid-term. For some time we will remain in a world of second-best heterogeneous
climate policies. Given this situation, we want to investigate how the global steam coal
market could affect the effectiveness of regional climate policies.
In this chapter we first discuss which methodology is the most appropriate to answer
this research question. We identify four main approaches to analyze resource markets
and climate policies and to assess their adequacy. Models based on the Hubbert curve
approach are based on historical production data only and thus fail to integrate paradigm
changes such as climate policies. Computable General Equilibrium models are too general
and lack a detailed representation of resource markets. Resource economics models based
on the work of Hotelling (1931) rely heavily on the mechanism of the scarcity rent that
drives up prices as the resource is depleted. Recent theoretical studies (Sinn, 2008) find
that in this framework climate policies will cause resource owners to deplete their resource
faster out of fear to lose their market in the future, thus causing a zero-sum game for the
climate. We argue that, due to the very large resource base of coal, a price formation
based on the geological scarcity rent is unlikely as well as the assumed reaction of the coal
producers. The remaining mid-term effects to be expected in the interaction between
climate policies and the global steam coal market are based on supply and demand
mechanisms that are called “market adjustments” in this chapter. Such effects are better
represented by partial equilibrium models such as the COALMOD-World model that was
introduced in the previous chapter of this thesis and is the main tool of analysis in this
15
United Nations Framework Convention on Climate Change Conferences of the Parties
33
Chapter 1. Introduction
chapter.
We investigate three different specific climate policy measures until 2030. The climate
policies are modeled as “policy shocks” inside a two-dimensional scenario space based on
the intensity of global climate policy and the market condition. There are three levels of
intensity of global climate policy: Current Policies (low), New Policies (medium) and 450
ppm (high) based on the scenarios of the 2011 World Energy Outlook (IEA, 2011c). They
affect steam coal demand in reverse order. The market situation can be constrained or
unconstrained. This is operationalized through a limit on additional production capacity
investments. This creates a 3x2 scenario matrix. The COALMOD-World model is then
calibrated and run for these six reference scenarios until 2030. The “policy shocks” are
then applied to these reference scenarios and the outcomes are compared. The three
“policy shocks” represent each a different type of climate policy. First, we investigate
a regional demand reducing policy with the unilateral European climate policy. Then
we analyze a supply-side policy with an export limitation for Indonesia. Finally we
implement a technology policy that supports a fast worldwide roll-out of the carbon
capture and storage technology.
The unilateral European climate policy scenario is the typical example of a regional
climate policy and can be seen as a continuation of the climate policies already in place.
This scenario is implemented as a steam coal demand reduction from all EU countries
based on the values of the Current Policies and New Policies scenario. The danger of such
a unilateral climate policy is a negative market adjustment, also called carbon leakage
through fossil fuel markets. The reduction in demand in one region depresses the global
price of steam coal and allows a region without stringent climate policy goals to consume
more. In our modeling this effect occurs in particular in the case of a low level of global
climate policy and of constrained investments in production capacity. The COALMODWorld model calculates that out of the emissions reduction in Europe, 66% in 2025 and
29% in 2030 are compensated by increased emissions from other regions, in particular in
Asia. This represents a serious problem for the effectiveness of regional climate policies.
The supply-side scenario in Indonesia is based on the idea of the Yasuní-ITT project.
This initiative proposed by the Ecuadorian government aims at protecting biodiversity,
indigenous people and the global climate by renouncing to exploit one fifth of the Ecuadorian oil reserves against financial compensation by an international consortium of donors.
Like Ecuador, Indonesia belongs to the group of “megadiverse” countries that host 70% of
the globe’s biodiversity.16 Coal mining in Indonesia is primarily opencast and takes place
in rainforest areas and thus threatens the biodiversity through deforestation and local
pollution. Thus, one could imagine a similar initiative as the Yasuní-ITT for Indonesia
not for oil but for coal. We implement this policy as an export limitation that is gradually
increased from a maximum of 50 million tons per year in 2020 to a complete export ban
in 2030. For comparison, in 2010, Indonesia’s steam coal exports amounted 160 million
16
A group of 17 like-minded megadiverse countries (LMMC) signed the Cancún declaration established in 2002 with the goal of protecting their biological patrimony
(http://es.wikisource.org/wiki/Declaración_de_Cancún)
34
1.6. Concluding Remarks and Outlook
tons. In the constrained investment cases in the years 2025 and 2030 this policy shock
reduces global emission from steam coal by around 4%. This is in the same range as the
reduction aim by the European unilateral climate policy. However, what is interesting
here is that 80% of the carbon emissions reduction comes from Asian countries.
For the CCS scenario we assume that there is already a medium or high level of
global climate policy and that with an eager policy and technological breakthroughs
the projected CCS capacities of the respective IEA scenarios are put in place five years
earlier. This additional CCS capacity combined with a lower efficiency of CCS power
plants create an additional demand for coal that increases the global price and in turn
reduces consumption and emissions of conventional coal power plants. This effect is
particularly clear in the case of a medium level of global climate policy intensity. In
the case of a very stringent climate policy, coal consumption is already so low that the
additional CCS capacity has only a limited impact on the market.
Our modeling results show that from a perspective of European climate policy, the
first priority to reduce global emissions should be to reach an international agreement
at least on the level of the targets discussed in the last climate conferences. When this
medium level of climate policy is reached, Europe can always go further in reducing its
coal usage without too much adverse effects from market adjustments. However, if this
is not the case, we can expect significant negative market adjustments that might undermine the effectiveness of the European climate policies. One strategy to avoid this would
be to use a supply-side climate policy such as the one proposed for Indonesia. We implement such a “hedging” against the market adjustment risk in the COALMOD-World
model by a combination of the European unilateral policy shock with the Indonesian
export restriction in the case of a low level of climate policy. As could be expected,
a simultaneous reduction of demand and supply leads to a net reduction in emissions,
showing that the hedging strategy works. Another advantage of the Indonesia export
limitation policy is that it targets directly the major steam coal consumers China and
India. Higher steam coal prices in these countries can give incentives to consumers to reduce their reliance on coal and motivate policymakers to implement policies transforming
the energy systems away from their fossil fuel dependency.
1.6
Concluding Remarks and Outlook
This thesis assesses various issues surrounding the global steam coal market using partial
equilibrium numerical modeling. The outcomes are reassuring regarding the issue of market power. The major players involved are not strategically withholding quantities from
the market or investments in new projects. Also the concentration process on the international market has stopped since 2005. This can be exemplified by the failed attempt
of BHP Billiton to take over Rio Tinto in the years 2007-08. However, a global economic
return to growth could again motivate mergers and takeovers. Moreover, nationalistic
policies and the nature and quantity of long-term contracts have to be watched as they
35
Chapter 1. Introduction
could remove quantities from the spot market and reduce liquidity. A smaller and more
illiquid market is more prone to strategic behavior. However, the main concerns regarding the use of steam coal are the environmental externalities. This thesis shows that
second-best climate policies on a regional level have to take into account the reactions of
the energy markets, even the most straightforward supply and demand effects.
This thesis provides the foundation for future work in several areas. The methodological part in Chapter 3 that proposes alternative models of oligopolies with a competitive
fringe can also be applied to other energy and resource markets such as the global crude
oil market. The crude oil market is characterized by an even more complex market structure and the methodology from Chapter 3 can shed some light on the functioning of this
market. Data availability, for example with respect to reserves and production costs, has
been a major challenge for my modeling endeavor. Aguilera (2008) proposes a methodology to assess petroleum reserves (crude oil, natural gas liquids) and Aguilera et al. (2009)
assess petroleum production costs in a bottom-up approach. This work should be applied to coal resources. Furthermore, understanding the interactions between fossil fuels
markets and climate policies is at the top of the future research agenda. In Hirschhausen
et al. (2012a) we identified that the absorption capacity of carbon dioxide in the atmosphere is a tighter resource than the fossil fuels available in the ground. Thus, one could
imagine to apply a “reverse Hotelling model” to this limited resource with the result of
steadily increasing carbon dioxide prices. Last but not least, the detailed numerical modeling approach developed in this thesis can be extended to cover all fossil fuel markets
(coal, natural gas and oil), including strategic behavior by the resource owners as well
as demand-side reactions. Such a model would help to better understand and assess the
whole range of possible interactions between fossil fuel markets and climate policies.
36
Chapter 2
Modeling and Analysis of the
International Steam Coal Trade: Is
the Market Competitive?
2.1
Introduction
This chapter analyzes the trade flows and prices on the international market for steam
coal17 by simulating the market for the years 2005 and 2006 using the complementarity
modeling technique.18 We develop two models: first, a trade model which is quantitybased and thus treats steam coal as a homogeneous good, and second, a model extension
that includes energy values and accounts for the energy contents of different coal qualities.
Both models can simulate a competitive market game or an oligopolistic market game
with imperfect competition à la Cournot. The models include capacity restrictions for
production and export activities and account for the spatial character of the market with
distance-based transport costs. Our major finding is that for both models the simulation
of perfect competition better fits the observed real market flows and prices in 2005 and
2006. We note, however, that it is necessary to use both energy and quantity values to
appropriately model coal markets.
The structure of international coal markets and the related supply security issues
have not been subject to much analysis recently despite the increased importance of
coal as a primary energy source. In fact, amid concerns about global warming and CO2
emission reductions, coal is currently experiencing a renaissance due to its abundance
and relatively low price compared to oil and natural gas. Electricity production based on
steam coal is receiving increasing attention due to the advent of clean-coal technologies
that target significant reductions in greenhouse gas emissions. Such technologies may
extend the viability of coal-based electricity generation despite the present climate change
17
Hard coal is distinguished by steam (thermal) and coking (metallurgic) coal, depending on its calorific
value and use. Steam coal, the subject of this analysis, is almost exclusively used for electricity production.
18
This chapter largely corresponds to the article published as Haftendorn and Holz (2010).
37
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Table 2.1: Share of imports in total consumption and share of imported steam coal in total
electricity generation of major steam coal consuming countries 2006
Import dependency rate
Japan
South Korea
Taiwan
UK
Germany
USA
China
Spain
Italy
99.50%
95.40%
100%
63.40%
69.20%
1.80%
11%
71%
99.50%
Share of imported steam coal
in electricity production
24.38%
33.49%
52.80%
21.37%
14.26%
0.86%
8.62%
16.69%
14.33%
Source: IEA (2007a,b)
concerns.
Globally, the use of coal has increased, mainly due to high energy demand in China
and India (IEA, 2007c). Most large coal consumers satisfy a significant share of their
demand on the world market, often because while domestic reserves have declined, a
certain industry structure has remained “locked in” with coal use (e.g., Germany, the
UK). For these consumers imported steam coal becomes more attractive than exploiting
their own high cost reserves. In the last decades the global coal markets have provided a
relatively cheap supply which also attracted new consumers like China and more recently
India.
Table 2.1 reports the import share of total consumption for the major consumers of
steam coal. Several European countries rely on imports for about 70% of their steam coal
consumption, while for some resource-poor East Asian countries (Japan, Taiwan, South
Korea) this rate goes up to 100%. Table 2.1 also shows the contribution of imported steam
coal to total electricity generation. In the European countries this share ranges from 15%
to 20 % and in Asia, from 20% to more than 50% (Taiwan). Given this importance of
the international market for steam coal, we would like to better understand its supply
structure.
Virtually all major exporters can be considered as “safe” countries in geopolitical
terms and the supply risks on political grounds are low.19 Short-term supply disruptions
may occur due to natural disasters or social tensions leading to strikes. However, efficient
supply management with stock-keeping and supply diversification reduces the risk of
disruption for coal importers. Nevertheless, it is helpful to examine whether the few
exporters on the world market are able to exert market power vis-à-vis their customers,
many of whom depend heavily on imports. In fact, it is not clear whether steam coal’s
price increase since 2002 is due only to a demand shift (in particular in India and China)
or to the greater concentration on the supply side.
19
Major exporters are Australia, South Africa, Indonesia, USA, Russia, China and Colombia.
38
2.2. State of the Literature
Recent research finds that the traditional separation of the Pacific and the Atlantic
coal markets has faded (e.g., Ellerman, 1995; Warell, 2006; Li, 2008). In our modeling
effort we therefore consider the global market as one integrated market, albeit not neglecting the spatial aspect of the market where transport costs play a role in determining
trade relations. Simultaneously with the trend to global market integration, we also observe a trend to commoditization of coal in increasingly liquid market places. This means
that coal is being traded more as a homogeneous good which is reflected by the creation
of price indices with standardized coal qualities and an increasing volume of paper trade.
The trend to commoditization is a motivation to use a quantity-based model of the global
steam coal trade.
However, the energy content of steam coal sold on the international market varies
by producer. The differences in coal qualities on the international market are not as
large as between the coal types that are produced and sold domestically20 but there are
some significant variations. This provides the motivation to employ a second model that
incorporates different coal qualities and uses both types of information about quantity
and energy. Previous energy market modeling has also used energy values as discussed
in the following literature review.
2.2
State of the Literature
The modeling effort applied to international steam coal trade has been rather sparse in
the last decade, particularly when compared to other energy commodity markets like
natural gas. Often, coal trade is embedded in energy system models. For example, the
LIBEMOD model (Aune et al., 2001) is primarily a model of the West European natural
gas and electricity markets, but it also includes the West European and the world market
for coal. This model only includes energy values and assumes perfect competition on the
coal market.
The Coal Market Module (EIA, 2007) of the U.S. National Energy Modeling System
(NEMS) is an international trade model that produces a forecast of US coal imports for
the NEMS. It incorporates both quantity and quality information and the world coal trade
is modeled using linear programming minimizing costs for a fixed demand. The assumed
trade pattern is perfect competition with various exogenously imposed constraints, e.g.,
import diversification, and quality constraints that potentially pre-determine a number
of results.
A modeling approach widely used to study international commodity trade is spatial
equilibrium modeling, first initiated by Samuelson (1952). Starting from the question
of pricing in spatially separated markets, Samuelson developed a linear programming
model maximizing welfare to find an equilibrium in perfectly competitive markets. This
problem was reformulated by Takayama and Judge (1964). They present a quadratic
20
Typically, low-quality coal such as lignite is sold domestically since its relatively low-energy content
per ton compared to steam coal makes long-distance transport uneconomic.
39
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
programming formulation and an algorithm that solves Samuelson’s partial equilibrium
formulation and extend it to a multi-commodity set.
However, perfect competition models often delivered disappointing results since the
modeled trade flows and prices do not reflect the reality. This was the case for the
international steam coal market in the early 1980s and motivated Kolstad and Abbey
(1984) to model imperfect competition. Their model (described in Kolstad et al., 1983)
is one of the few that specifically represents international steam coal trade and also uses
quantity and energy values. However, it does not incorporate any capacity restrictions
which can significantly influence model outcomes. Kolstad and Abbey (1984) examine
whether market power exerted by some players could be responsible for the observed
trade patterns. They conclude that a supply duopoly (South Africa and Australia) and
an import monopsony (Japan) is very similar to the actual trade pattern of the 1980s.
A general formulation of how the Takayama-Judge spatial equilibrium model can collapse into a spatial Cournot model appears in Yang et al. (2002). Their spatial equilibrium Cournot model is solved using linear complementarity programming and is applied
to the US coal market. This model uses energy values and the results of the spatial
Cournot model as well as of a standard competitive spatial equilibrium model are compared to the observed trade pattern. In the case of the US coal market, the competitive
spatial equilibrium model yields more realistic results.
The situation on the international steam coal market has evolved since the 1980s
and it is the goal of this chapter to understand which market conduct influences today’s
trade patterns. In summary, given the spatial character of the international steam coal
market and our earlier experience with Cournot modeling of natural gas markets (e.g.,
Holz et al., 2008; Egging et al., 2008) we use a partial equilibrium Cournot model in a
spatial setting.
2.3
2.3.1
The COALMOD-Trade Model
Description of the analytical model
We adopt a complementarity approach which is regularly used in energy sector modeling.
The international steam coal market is modeled as a non-cooperative static game among
the suppliers (export countries). We assume they desire to maximize their individual
payoffs (profits). The exporters produce the steam coal, sell it and transport it to the
importers. We characterize importing countries by a demand function for imported steam
coal. The market is simulated as a Cournot model with the possibility for the export
countries to exert market power, or as a perfect competition model where the exporters
are price takers.
The exporters x maximize their profit Πx (yxm ), defined by the revenue net of costs of
production and of transport to each importing country m by choosing the optimal trade
flow yxm to sell to each importing country m, given a production and an export capacity
constraint. Thus, the trade flows yxm and the associated prices pm in the importing
40
2.3. The COALMOD-Trade Model
Table 2.2: Units used for the COALMOD-Trade and COALMOD-Trade-Energy models
CMT Model CMT-E Model
Million tons
Petajoules
USD/ton
USD/Gigajoule
Million tons
Petajoules
USD/ton
USD/ton
Million tons
Million tons
Demand
Prices (pm )
Trade flows (yxm )
Production costs
Transport costs
Production capacities
Export capacities
country m are endogenous model results. Exogenous data inputs are the parameters of
the demand functions, production costs, transport costs and the production and export
capacity constraints for each exporting country. We choose a linear demand function
defined around a reference point, a quadratic production cost function and use unit
transport costs (see 2.3.2 for details and the parameter input).
We specify two models that differ in the units assigned to the parameters and variables
(Table 2.2). In the following, we abbreviate the specification that relies only on mass
quantities with “COALMOD-Trade” (CMT) and the specification that also includes the
energy content of the coals with “COALMOD-Trade-Energy” (CMT-E). For the energy
specification of the model we use a conversion factor κx expressed in tons per Gigajoules.
This factor defines a quality of coal supplied by the exporter x and is employed to
express costs and capacities per energy value. The following equations describe the
energy specification of the model, in which the conversion factor κx is added. The model
structure for the quantities-only specification is the same without the energy conversion
factor.
For a linear demand function of the type pm = am + bm ym , a strategic player x with
the capacity to influence the demand function has the following optimization problem of
exports to importer m:
!
X
max Πx =
pm
yxm
m
X
yxm
!
· yxm − cx
X
κx · yxm
x
m
−
X
trans_cxm · κx · yxm (2.1)
m
such that production and export capacity of country x are respected and the decision
variable is non-negative:
prod_capx −
X
κx · yxm ≥ 0
(λx )
(2.2)
κx · yxm ≥ 0
(µx )
(2.3)
m
exp_capx −
X
m
yxm ≥ 0
41
(2.4)
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
We choose the functional forms of the profit function (demand and cost functions)
such that the first-order conditions (also known as Karush-Kuhn-Tucker conditions,
KKT) of the optimization problem are necessary conditions for the optimal solution
(see section 2.3.2 for the functions). Taking the KKTs of all players x simultaneously
gives a non-linear complementarity problem. We obtain the following KKT conditions
of the optimization problem (2.1) - (2.4):21
∂cx
0 ≤ − pm − bm · yxm +
+ trans_cx · κx + λx · κx + µx · κx
∂yxm
X
0 ≤prod_capx −
κx · yxm
⊥ yxm ≥ 0
(2.5)
⊥ λx ≥ 0
(2.6)
⊥ µx ≥ 0
(2.7)
m
0 ≤exp_capx −
X
κx · yxm
c
We consider a strategic player, accounting for its influence on the demand function
and whose derivative of the linear demand function is
∂pm (ym )
∂yxm
= bm . The term bm · yxm
gives the oligopolistic mark-up that the strategic player can obtain. A competitive player,
on the other hand, does not take into account the demand function but behaves as pricetaker. For such a player
∂pm
∂yxm
= 0. We can therefore introduce a market power parameter
αx for each player x that is multiplied with the term bm ·yxm and that is defined as αx = 0
for a competitive player x, and αx = 1 for a Cournot player. Indeed, αx is nothing else
than the conjectural variation of a player x reacting to its competitors −x.
Combining the KKT conditions (2.5) - (2.7) with a market-clearing condition for the
import market, we obtain a unique equilibrium solution for the market model. The following market-clearing condition determines the price given the demand function pm (ym ).
!
pm − pm
X
yxm
= 0,
pm
(free)
(2.8)
x
This complementarity model is programmed in GAMS, and solved with a standard
algorithm for MCP (mixed complementarity problems), PATH.22
2.3.2
Data
This section details the parameter input for the two specifications of the model with
respect to quantity and energy content. We use data for exports and imports at the
country level, and assume each country to be one player.23 Table 2.3 shows the countries
in the data set which are the main exporters and importers in the international steam
coal market.24
As mentioned in Section 2.3.1, we assume a linear inverse demand function of the type
pm = am + bm ym for each importer m. We construct a different linear inverse demand
function for each importing country m using their reference prices (pref
m ) and reference
21
Following the literature of complementarity modeling, the profit maximization problem is converted
42
2.3. The COALMOD-Trade Model
Table 2.3: Countries in the COALMOD-Trade model
Exporting Countries
Australia
Indonesia
South Africa
Russia West (Baltic Sea)
Russia East (Pacific)
China
Colombia
United States of America
ref
demand value (ym
=
Importing Countries
Japan
Taiwan
South Korea
United Kingdom
Germany
United States of America
Spain
Italy
India
China
P ref
yxm ) of each base year 2005 and 2006 and assumptions on the
x
demand elasticities (εm ). In particular, we define bm =
following the demand elasticity definition εm =
inverse demand function:
pm = pref
m
pref
m
ref
ym
ref
ym −ym
ref
pm −pm
·
ref
· ε1m and am = pref
m −bm ·ym ,
pm
ym .
This gives the following
P

yxm
1 ref  x
p
− 1
+
ref
εm m
ym
(2.9)
As shown in Table 2.2, the type of input needed for the two model specifications
differs due to the different units used. For the CMT model, the reference import quantities and CIF (cost insurance freight) prices for 2005 and 2006 are obtained from the
OECD’s International Energy Agency (IEA, 2007a). For the CMT-E model, we use coal
quality data in addition (Table 2.4). We convert the mass import flows obtained from
IEA (2007a) into energy flows and aggregate them to obtain a reference demand expressed in Petajoules (1015 Joules). We calculate the reference prices in USD/Gigajoules
(109 Joules) by dividing the total value of imports by the demand in Petajoules. Demand
elasticities εm are chosen in the calibration process and based on Dahl (1993). In the
benchmark specification we use εm = −0.3 for all countries.
In the CMT model, the production cost function of each exporter is assumed to be
P
quadratic of the type cx = (acx + bcx · Yx ) · Yx , with total production Yx = yxm . Thus,
m
the marginal cost function is mcx = acx + 2 · bcx · Yx . The cost functions are obtained
to a minimization problem before deriving the KKTs.
22
Cf. Ferris and Munson (2000) for an overview.
23
Due to Russia’s size, we assume two players, one on the Western (Baltic) shore and one on the
Eastern (Pacific) shore.
24
Since China and the US are both exporting and importing countries, they are introduced twice.
This is because our focus is on the international trade and we do not aim at a representation of domestic
markets. For model consistency, the exporter and importer of the same country are not allowed to trade
with each other. In this case, the importing and exporting port are different (US: Mobile (AL) and
Hampton Roads (VA); China: Shenzhen and Quinhuangdao).
43
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Table 2.4: Coal quality by exporter
Australia
Indonesia
South Africa
Russia West
Russia East
China
Colombia
US
kcal/kg
6400
5450
6260
6400
6300
6200
6375
12500[Btu/lb]
GJ/t
26.80
22.82
26.21
26.80
26.38
25.96
26.69
29.08
κx in t/GJ
0.03732
0.04382
0.03815
0.03732
0.03791
0.03852
0.03747
0.03439
Source: Platts (2008)
using data provided by Ritschel and Schiffer (2005) that gives lower and upper bounds
on average costs for each exporter. We use this information to construct linear average
cost curves. The intercept parameter acx corresponds to the lower bound. To determine
the slope bcx we use a second point defined by the maximum production capacity and the
upper bound of the average costs assuming a linear average cost function avcx = acx +
bcx · Yx . For the CMT-E model the conversion factor κx must be added to the equations.
Thus, we obtain the following marginal cost function mcx = κx ·acx +2·κ2x ·bcx ·Yx =
∂cx
∂yxm
that we can integrate in the KKT condition (2.5).
The unit seaborne transport costs trans_cxm expressed in USD/t are based on observations of selected freight rates for each base year 2005 and 2006 and derive from the
technical freight literature and IEA (2007a). We use these observations in a regression to
obtain linear functions of freight rates based on distance and then to obtain the specific
value of trans_cxm for every possible route between the exporters and the importers.
Finally, the data of production capacity that is available for exports (export mines)
is from Kopal (2007). In addition, we include export harbor capacity constraints based
on Ritschel and Schiffer (2005) and VDKI (2006). We implicitly assume that shipping
(boat) and import harbor capacity is available without capacity limitation.
2.4
2.4.1
Results
Market structure analysis
We want to determine which market scenario is more likely to explain the trade pattern
of the international steam coal market observed in 2005 and 2006, following Kolstad
and Abbey (1984). The result will give us an indication of whether import dependent
countries must pay a higher price than the competitive price level, and whether they are
subject to the exercise of market power by the exporters.
As discussed, we model two different market scenarios on the supply side: perfect
competition and Cournot competition. The scenarios are implemented via a modification
of the value of parameter αx . For the scenario of a perfectly competitive market, we set
44
2.4. Results
Figure 2.1: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2005, in million tons (Mt) for the CMT model
αx = 0 for all x; conversely, αx = 1 for all x in the Cournot scenario. We compare
the results of our simulations with the observed trade flows in 2005 and 2006. We
also compare the outputs from the two different model specifications to determine the
influence of the additional quality information on the results.
Figures 2.1 and 2.2 show the model results and actual trade flows in 2005 for our two
specifications CMT and CMT-E. The results of the CMT-E model are converted from
Petajoules to million tons using the coal quality data from Table 2.4. Total import quantities obtained in both models show a remarkable similarity of the perfect competition
results with the reference data for 2005 and 2006. This is also true for 2006 (Figures 2.7
and 2.8 in the Appendix 2.A). The Cournot scenario, on the other hand, gives smaller
quantities and considerably higher prices than observed in reality.
The detailed results for both years and both CMT and CMT-E, models are presented
in the Appendix 2.A. It can be seen that the number of flows (number of trading
relations) in the perfect competition simulation results is small and that most importing
countries rely on only one or two suppliers. Real-world flows in 2005 and 2006 showed
significantly more diversification of imports. The Cournot results present a more diverse
picture with each importer buying from virtually all exporters.
The results of the perfect competition scenario with little diversification are driven
by the cost-minimization mechanism that characterizes competitive markets. There is
no mark-up on the marginal cost price. Each country imports from the supplier that has
the lowest production and transport costs to deliver the coal to that market. In Cournot
markets, on the other hand, prices are above their marginal cost levels and attract more
45
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Figure 2.2: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2005, in million tons (Mt) converted from Petajoules for the CMT-E
model
suppliers, including those with higher costs. Although the more diversified trade flow
picture makes the Cournot scenario an attractive explanation of the real-world market,
we must discard it due to the very high prices and small total quantities obtained when
compared to the reference data.
Comparing the 2005 price results of both model specifications with the reference prices
(Figure 2.3), we see that the prices of the perfect competition simulations are closer to or
in the same range as the real observed prices. However, the relation of the prices between
the countries is similar in reality to the Cournot simulation. Cournot competition allows
for price discrimination between importing countries whereas the perfect competition
simulation does not. This issue will be addressed in Section 2.4.3. These two conclusions
also apply to the prices in 2006 (see Figure 2.4). However, the price levels in the model
results for 2006 are high (even perfect competition prices are above the real price level)
which can be explained by a certain lag in pricing-in capacity constraints (see Section
2.4.2).
When comparing the quantity and energy specification, we observe that the CMT
model’s perfect competition results show significantly less supply from Australia than in
reality (2005: 66 Mt vs. 110 Mt, 2006: 62 Mt vs. 105 Mt). On the other hand, the CMT
model results for supply from Indonesia are higher than in reality by approximatively
10 Mt for both years. One explanation is that a quantity model does not incorporate
information about the energy content of the coals. However, in the end, what is important for the customer is the energy contained in the coal. By not incorporating the
quality information a distortion is created that makes lower quality coal, like Indone46
2.4. Results
Figure 2.3: CIF Prices in the perfect competition (PC), Cournot scenario (CO), and reference
data (RE) in 2005 for the CMT and CMT-E models, USD per ton
Figure 2.4: CIF Prices in the perfect competition (PC), Cournot scenario (CO), and reference
data (RE) in 2006 for the CMT and CMT-E models, USD per ton
sian coal, more attractive: the associated transport costs that are mass dependent are
underevaluated in comparison to higher quality coals. Reciprocally, the same distortion
makes higher quality Australian coal less attractive. This is confirmed by the perfect
competition results of the CMT-E model which show higher Australian exports (75 Mt
in 2005 and 73 Mt in 2006) and lower Indonesian exports than in the CMT results.
Both model specifications provide evidence that rejects a Cournot market structure
for 2005 and 2006. Our perfect competition results of total exported and imported
quantities are closer to the real trade volumes. Thus, for the purpose of a market structure
analysis a quantity-based model may be sufficient. However, if one requires a better
representation of detailed trade flows, a model that incorporates both energy and mass
quantity information is needed. This will hold true in the next decades since we expect
an increase in the supply of lower quality coal to the world market.
To show the robustness of our results we perform a sensitivity analysis on the price
elasticity of the demand function. For this parameter, data sources are scarce and potentially outdated. For the base model data set, we choose εc = −0.3 following Dahl (1993)
who reported the short run elasticities for steam coal were between −0.3 and −0.55.
Aune et al. (2001) use −0.19 for their model. Using these numbers as boundaries for our
sensitivity analysis, we run the model with values between −0.2 and −0.6 for εm . As in
the other model runs, we use the same elasticity value for all countries m.
Figure 2.5 shows that the results of the perfect competition market appear to be more
47
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Figure 2.5: CIF Prices in the perfect competition (PC) and Cournot scenario (CO) model
results for different elasticity values, and reference data (RE) in 2005 and 2006 for the CMT-E
model, USD per ton
robust than the Cournot results. In 2005, the perfect competition results for the different
elasticity values are so close that they can hardly be distinguished visually. Also, the
higher the elasticity (in absolute values), the closer the Cournot price results are to the
real prices. However, they are still considerably above the observed prices. Given the
rather inflexible nature of short-term coal demand due to little switching possibilities
to other fuels, and previous elasticity estimations (Dahl, 1993), any value larger than
εm = −0.6 does not seem realistic. Hence, the sensitivity analysis confirms our conclusion
that the perfect competition model best represents the international market for steam
coal.
2.4.2
Pricing and export restrictions in a temporal perspective
The models introduced above are able to indicate whether there are physical bottlenecks
in the production and export capacity that hinder the coal trade. A positive value of the
dual prices λx and µx of the capacity constraints (2.2) and (2.3), respectively, points to
a capacity limitation for exporter x. Bottlenecks can potentially indicate where there is
need for investment in the producing and exporting infrastructure. The results of both
specifications allow a similar interpretation about pricing and export restriction. For this
reason the interpretations in this section refer only to the perfect competition results of
the CMT-E model specification.
The results of the CMT-E model show positive dual prices µx for South Africa, Russia
(Baltic and East), and Colombia in 2005 and for South Africa, Russia Baltic, China
and Colombia in 2006. This means that, in the model, these countries reached their
maximum export capacities. Production capacity constraints were not binding because
they are higher than the export capacity constraints for these countries. Also, the level
of the dual prices µx is significantly higher in 2006. This is the main reason why the
model calculates a higher average price of 73.31 USD/t in 2006, compared to 58.31 in
2005, despite a lower average of the reference demand prices in 2006 (61.12 UDS/t and
62.86 UDS/t in 2005).
Figure 2.6 sheds some light on the German import price and the European spot price
48
2.4. Results
Source: RWE (2008)
Figure 2.6: Historical steam coal prices: CIF(cost insurance freight) spot price in ARA
(Amsterdam-Rotterdam-Antwerpen) and price of delivered steam coal at the German border
around 2006. The darker curve represents the historical spot price to (Northwest) Europe
and the gray curve shows prices reported by the German Federal Office of Economics and
Export Control (BAFA) when the steam coal crosses the German border. We observe that
the BAFA price follows the CIF-ARA price with a few months’ time lag. Additionally,
there seems to be an asymmetric price adjustment in response to a rise of the spot price
that is higher than the response to a decline.25 The CIF-ARA spot price represents
the price for deliveries in the forward 90 days period. It is not a typical commodity
exchange price like for the major commodities (e.g., for crude oil), but a price index
based on a weekly survey of a limited number of deals submitted by industry participants.
Since this represents the individual views of the state of the market, the quality of this
price signal and its reaction time to market developments is not as reliable as for other
commodities. The prices from IEA (2007a) used in our model are the reported customs
prices (e.g., BAFA prices), which requires careful assessment of price developments, their
role as indicators of export capacity scarcity and their function as price signals for new
investments.
The rising prices in the global steam coal market since 2003 triggered investments
in export mine capacities starting in 2004 (Kopal, 2007) that explain the decreasing
CIF-ARA prices and the stabilization of the BAFA prices in 2005. Like the average
reference demand prices in our data input, the BAFA prices are slightly lower in 2006
than in 2005. But in the spot market the downward trend of 2005 does not persist and in
2006 the price increases, due to rising demand that is not met by sufficient investment.
25
This phenomenon is somewhat similar to the “rockets and feathers” situation in retail gasoline prices
which has been extensively studied.
49
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
This explains why the 2005 model prices are in the range of the observed prices whereas
the 2006 model prices are higher than the reference data. In 2005, the supplies were
hardly constrained and prices were on a downward trend. In 2006, a higher demand and
insufficient capacities are not fully incorporated in the real market pricing but the model
results already account for them. Thus, our simulated prices are consistent with the
upward trend in CIF-ARA spot prices for 2006 and signal the price explosion in 2007.
Sending investment signals in time is a problem because today’s international steam
coal market is neither transparent or liquid enough unlike other fuel commodity markets. Another possible reason why bottlenecks can occur is that players often withhold
capacity expansions in order to drive up prices. However, supply shortages can also occur when demand rises abruptly because investments to expand production and export
capacities cannot be carried out in the short-term. The observed lag in prices as well as
the poor quality of the price signals accentuates this effect. In other words, increased
demand primarily in China and India, and to a lesser extent in Europe and the US, along
with tighter supply due to insufficient global investment explain the run-up in prices in
international steam coal markets after 2007.
In a competitive market players should have incentives to remove the bottlenecks in
order to maximize profits. Between 2005 and 2006 some investment activity occurred,
e.g., Indonesia expanded its coal export terminals and Russia increased its export capacity. South Africa experienced technical problems in both 2005 and 2006 which affected
its export capacity but is currently expanding its export terminal at Richards Bay. This
activity somewhat supports the result of a more competitive supply structure. Kopal
(2007) also describes that bottlenecks in the coal sector have not been lasting in the last
decades and that the industry has been reactive to investment signals.
2.4.3
Spatial price discrimination
In Section 2.4.1 we noted that the price differences between importing countries for the
real prices and the Cournot simulation results are similar and significantly higher than in
the perfect competition simulation. In our Cournot framework exporters are allowed to
price discriminate, something they cannot do in a competitive framework. This indicates
that, in reality, exporters may be using price discrimination. Converting the model’s
reference prices in prices per energy unit does not reduce this variability but increases
it. The coefficient of variation26 for the prices expressed in USD/t is 0.1364 for 2005
and 0.1223 for 2006 and 0.1417 and 0.1282 for the prices converted to USD/GJ. This
means that the variation between CIF prices cannot be fully explained by the difference
in qualities of the imported coal.
Table 2.5 gives some evidence that Australia, one of the key players in the global
market and a price setter in the Pacific market, uses price discrimination and supplies
26
The coefficient of variation is as a normalized measure of dispersion defined as the ratio of the
standard deviation to the mean.
50
2.5. Conclusions
at different (FOB27 ) prices to different importers. The fact that Australia and possibly
other players price discriminate is a form of market power.
Table 2.5: Average FOB prices of Australian steam coal exports to importing countries in
2005 and 2006, in USD per ton
Importer
Weighted Average
Belgium
Denmark
France
Germany
Ireland
Japan
Netherlands
Spain
Sweden
United Kingdom
2005
48.86
n.a.
44.09
45.88
61.3
29.92
49.42
60.71
42.82
67.75
54.27
2006
47.05
67.08
55.39
46.28
42.03
n.a.
50.59
50.5
51.11
66.81
46.3
Source: IEA (2008a)
Seminal research by Hoover (1937) and Greenhut and Greenhut (1975) shows how
firms exploit their locational advantage (and that of their rivals) to drive prices above
marginal costs. Spatial price discrimination in the theoretical framework of Cournot and
Bertrand markets is analyzed by Hobbs (1986). However, these models use a two-firm
linear spatial setting. It is not clear whether a spatial price discrimination model can be
implemented in a two-dimensional space in a game theoretic framework with multiple
firms such that it can be applied in a numerical simulation model, i.e., the COALMODTrade model.
2.5
Conclusions
Two complementarity models of the international steam coal market were used for numerical simulations for 2005 and 2006. Our goal was to determine whether this market
is subject to the exercise of market power by its major players. We find evidence that
rejects a Cournot market structure for the two years for our model specifications based
only on quantities and based on the energy content. Our perfect competition simulation
results for total exported and imported quantities are closer to the real trade volumes.
This differs from Kolstad and Abbey (1984) who use a similar model and a smaller data
set to find that the market of the 1980s was subject to market power by oligopolistic
suppliers and a monopsonistic importer.
Our results show that using a model that is based only on quantities produces distortions in the individual trade flows. For transport costs per units, considering quantities
alone makes lower-quality coal more attractive to the importers. This distortion can be
corrected using an energy and quantity model, i.e., the COALMOD-Trade-Energy model
27
FOB: Free on board, price at the exporting harbor.
51
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
specification which is superior for analyzing the international steam coal trade. We conclude that finely-tuned models will be more relevant as increased volumes of low-grade
coals are expected to meet the growing demand in coming years.
When analyzing the pricing on the international steam coal markets, we find evidence
of spatial price discrimination, suggesting that exporters sell to different importers at
different prices. The difference is higher than the transport cost differential. In addition,
there is evidence of a time-lag in the pricing of export capacity constraints. The reaction
of the market prices to supply constraints is delayed and the poor quality of the price
indices makes it difficult to accurately represent the current market situation. This may
explain how prices fail to provide the proper investment signals or, in the shorter term,
fail to give incentives to consumers to find alternative supply sources quickly. These
inefficient and delayed adjustments can explain how the price spikes observed in 2007
and 2008 can occur.
We suggest to enlarge the COALMOD-Trade-Energy model to include internal markets that are connected to the global market in order to analyze the interactions between
them. Moreover, to account for the dynamics of the world coal market, we suggest a
multi-period model with endogenous investment decisions in production and export capacity. This model, the “COALMOD-World” model, is introduced in Chapter 4. Such
modeling, among other applications, would allow the impacts of increased carbon constraints and their interplay with the global steam coal market to be properly analyzed.
This is done in Chapter 5. But first, in the following Chapter 3, we have a closer look
at a sub-market of the global steam coal market, the Atlantic steam coal market, and
use more refined models of strategic interaction to see if we can find evidence of strategic
behavior in this more mature and liquid market.
52
2.A. Appendix
2.A
Appendix
Figure 2.7: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2006, in million tons (Mt) for the CMT model
Figure 2.8: Imported quantities in the perfect competition (PC), Cournot scenario (CO), and
reference data (RE) in 2006, in million tons (Mt) converted from Petajoules for the CMT-E
model
53
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Table 2.6: Simulated import prices for the CMT model and reference CIF prices for 2005, in
USD per ton
Japan
Taiwan
South Korea
UK
Germany
USA
Spain
Italy
India
China
Cournot simulation
88.30
89.05
84.12
94.06
95.14
82.18
90.06
94.91
92.69
84.92
Perfect competition simulation
60.21
56.76
58.26
62.37
62.14
53.71
61.74
61.03
59.04
57.01
2005 CIF Prices
62.73
(65e)
55.76
70.24
72.48
47.39
62.94
73.20
68.50
50.39
Table 2.7: Steam coal trade flows in the 2005 perfect competition scenario for the CMT
model, in million tons (Mt)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
66.06
29.68
Taiwan
S. Korea
UK
Germany
USA
Spain
Italy
India
16.97
18.27
57.83
5.48
9.11
6.18
13.02
1.84
China
20.39
54.81
22.06
21.12
19.87
0.65
Table 2.8: Steam coal trade flows in the 2005 Cournot scenario for the CMT model, in million
tons (Mt)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
17.08
20.03
13.03
5.63
6.17
16.32
17.37
0.30
Taiwan
8.76
10.93
6.87
3.22
2.79
8.36
8.49
0.10
S. Korea
8.38
10.43
6.29
2.06
2.64
8.67
8.61
UK
5.05
5.41
4.80
5.03
0.25
3.49
6.47
1.97
Germany
2.74
2.93
2.61
2.74
0.22
1.92
3.43
1.13
USA
3.20
3.02
2.56
1.39
Spain
2.84
3.07
2.69
2.61
1.61
5.38
1.86
3.74
0.95
Italy
2.35
2.51
2.20
2.15
0.29
1.68
2.79
0.74
India
2.79
3.23
2.54
1.44
0.44
2.04
2.79
0.41
China
0.99
1.29
0.77
0.29
0.20
0.98
Table 2.9: Actual steam coal trade flows in the base year 2005, in million tons (Mt)
Aus.
Indo.
R.S.A.
Rus.W.
Rus.E.
China
Col.
USA
Japan
67.38
19.51
0.05
Taiwan
15.82
19.13
0.34
S. Korea
19.21
15.38
7.16
15.17
0.75
19.66
3.02
17.58
0.37
UK
0.93
1.62
13.03
16.83
Germany
0.77
8.22
7.50
USA
0.07
2.24
0.07
0.36
Spain
1.43
3.78
8.74
4.24
0.02
19.25
0.05
1.94
0.23
Italy
0.68
6.80
4.40
1.08
India
0.84
11.66
2.36
China
2.45
2.40
0.84
0.13
3.30
0.30
2.94
0.13
54
2.59
3.00
0.20
0.07
0.01
2.A. Appendix
Table 2.10: Simulated import prices for the CMT model and reference CIF prices for 2006, in
USD per ton
Japan
Taiwan
South Korea
UK
Germany
USA
Spain
Italy
India
China
Cournot simulation
95.43
94.12
88.53
101.34
101.34
91.79
95.92
100.04
95.54
91.31
Perfect competition simulation
68.52
64.36
67.63
73.57
73.30
63.45
72.95
71.66
69.28
64.67
2006 CIF Prices
63.33
(62e)
51.73
69.91
70.12
50.55
60.66
69.16
63.70
50.00
Table 2.11: Steam coal trade flows in the 2006 perfect competition scenario for the CMT
model, in million tons (Mt)
Aus.
Indo.
R.S.A.
Rus.W.
Rus.E.
China
Col.
USA
Japan
62.40
36.65
Taiwan
S. Korea
56.20
0.01
UK
Germany
21.71
USA
Spain
Italy
India
China
17.05
11.48
9.39
6.21
9.53
21.59
7.79
45.00
20.44
24.88
5.60
3.10
Table 2.12: Steam coal trade flows in the 2006 Cournot scenario for the CMT model, in
million tons (Mt)
Aus.
Indo.
R.S.A.
Rus.W.
Rus.E.
China
Col.
USA
Japan
15.25
19.63
9.26
7.44
6.52
15.29
12.64
0.06
Taiwan
8.44
12.01
5.40
4.35
2.80
8.57
6.44
S. Korea
8.30
11.73
4.33
3.06
2.73
9.19
6.40
UK
5.10
6.32
4.75
7.92
Germany
2.61
3.23
2.43
4.06
USA
3.47
3.58
2.38
3.28
Spain
2.23
2.87
2.05
3.41
3.38
6.79
2.79
1.73
3.47
1.44
1.63
6.00
1.33
3.09
1.05
Italy
2.22
2.71
2.00
3.04
0.00
1.52
2.60
0.85
India
3.29
4.40
2.69
2.34
0.15
2.37
2.66
0.10
China
1.12
1.66
0.69
0.55
0.27
0.83
Table 2.13: Actual steam coal trade flows in the base year 2006, in million tons (Mt)
Aus.
Indo.
R.S.A.
Rus.W.
Rus.E.
China
Col.
USA
Japan
60.41
20.20
0.08
Taiwan
18.44
24.30
0.07
S. Korea
17.34
20.70
6.09
14.77
1.31
12.74
3.93
15.65
0.54
UK
0.15
2.15
13.08
22.55
Germany
0.70
8.52
8.21
USA
0.15
2.86
0.06
0.85
Spain
1.58
4.02
8.21
3.61
0.04
22.99
3.00
1.53
0.45
Italy
0.88
8.73
4.78
0.82
India
0.44
16.05
2.20
China
4.44
1.42
0.96
0.03
4.07
0.79
4.00
0.47
55
3.21
2.03
Chapter 2. Modeling and Analysis of the International Steam Coal Trade
Table 2.14: Simulated import prices for the CMT-E model and converted reference CIF prices
for 2005, in USD per Gigajoules
Japan
Taiwan
South Korea
UK
Germany
USA
Spain
Italy
India
China
Cournot simulation
3.40
3.46
3.27
3.61
3.64
3.17
3.48
3.71
3.67
3.32
Perfect competition simulation
2.38
2.23
2.31
2.36
2.36
2.04
2.34
2.31
2.24
2.24
2005 conv. CIF Prices
2.42
(2.58e)
2.19
2.66
2.73
1.80
2.44
2.93
2.86
2.01
Table 2.15: Steam coal trade flows in the 2005 perfect competition scenario for the CMT-E
model, in Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
2019
130
134
Taiwan
S. Korea
UK
Germany
USA
Spain
Italy
India
429
447
1458
138
152
172
343
222
China
540
1390
662
555
465
68
Table 2.16: Steam coal trade flows in the 2005 Cournot scenario for the CMT-E model, in
Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
463
443
343
152
158
423
469
39
Taiwan
228
240
175
87
73
209
221
24
S. Korea
226
225
166
61
71
223
231
UK
141
116
129
133
6
92
178
58
Germany
77
63
70
73
5
51
95
33
USA
92
55
70
38
Spain
78
63
71
68
43
148
49
100
29
Italy
61
52
56
54
10
43
72
23
India
69
66
61
37
14
50
69
18
China
26
28
20
9
6
26
0
Table 2.17: Actual steam coal trade flows in the base year 2005, converted in Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
1806
445
1
Taiwan
424
437
9
S. Korea
515
351
189
394
20
510
80
456
11
UK
25
37
342
451
Germany
21
215
201
USA
2
51
2
10
Spain
38
86
229
113
0
514
1
52
7
Italy
18
155
115
29
India
22
266
62
China
66
55
22
3
88
9
78
4
56
67
80
6
2
0
2.A. Appendix
Table 2.18: Simulated import prices for the CMT-E model and converted reference CIF prices
for 2006, in USD per Gigajoules
Japan
Taiwan
South Korea
UK
Germany
USA
Spain
Italy
India
China
Cournot simulation
3.66
3.67
3.44
3.87
3.87
3.53
3.70
3.92
3.85
3.51
Perfect competition simulation
2.74
2.56
2.69
2.94
2.93
2.56
2.92
2.87
2.78
2.57
2006 conv. CIF Prices
2.45
(2.49e)
2.06
2.64
2.64
1.92
2.35
2.81
2.69
1.93
Table 2.19: Steam coal trade flows in the 2006 perfect competition scenario for the CMT-E
model, in Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
1960
496
Taiwan
S. Korea
UK
Germany
USA
Spain
Italy
India
China
422
50
464
159
58
1404
597
76
563
161
1168
497
637
225
182
Table 2.20: Steam coal trade flows in the 2006 Cournot scenario for the CMT-E model, in
Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
411
432
242
213
166
399
337
35
Taiwan
220
261
139
124
75
216
169
10
S. Korea
223
247
117
98
74
236
173
UK
144
130
127
216
Germany
74
67
65
112
USA
100
62
65
94
Spain
62
56
54
91
91
185
81
46
95
42
46
163
36
82
32
Italy
58
54
50
76
4
39
65
27
India
81
86
66
61
13
59
67
18
China
31
37
18
16
7
23
Table 2.21: Actual steam coal trade flows in the base year 2006, converted in Petajoules (PJ)
Aus.
Indo.
R.S.A
Rus.W.
Rus.E.
China
Col.
USA
Japan
1619
461
2
Taiwan
494
554
2
S. Korea
465
472
161
383
34
331
104
406
16
UK
4
49
343
604
Germany
19
223
220
USA
4
65
1
23
Spain
42
92
215
97
1
614
0
41
13
Italy
24
199
125
22
India
12
366
58
China
119
32
25
1
109
23
107
14
57
83
54
Chapter 3
Atlantic Steam Coal Market Power:
Theory and Application of Oligopoly
Models with a Competitive Fringe
3.1
Introduction
The contribution of this chapter is twofold.28 Firstly some theoretical considerations
regarding the market power representation in partial equilibrium modeling are made.
Secondly, having found that certain approaches are problematic we propose and apply a
different approach of modeling dominant firms with a competitive fringe to the Atlantic
steam coal market. We show that in the years 2004 and 2005 the dominant firms in this
market may have exerted market power.
A widely used approach in energy and resource modeling to model market power is
the use of conjectural variations (CV). This chapter gives an overview of the theoretical
literature of CV and presents an economic historiography of its use in applied partial
equilibrium analysis. This chapter also shows that this approach to represent market
power may be problematic. In fact, we demonstrate that using the CV approach may lead
to results that are in fact not Nash equilibria and do not represent the behavior of rational
actors. This is especially the case when some actors have market power and others are
assumed to behave competitively and is due to the fact that in its implementation the
conjectural variation approach uses relative mark-ups between prices and marginal costs
to represent market power. These mark-ups are used as equilibrium conditions but in
certain cases these mark-ups may actually depreciate the profit of the players assumed
to exert market power. This also relates to the problem that in the CV approach players
exerting market power may be considered as “naive” as they may make false assumptions
about how other players will react to their own actions.
Considering the shortcomings of the CV approach we use an alternative approach to
model a dominant oligopoly with a competitive fringe based on the Stackelberg (1934)
28
This chapter largely corresponds to Haftendorn (2012).
58
3.2. Literature Review
model. We then use three different ways to represent the dominant oligopoly: a noncooperative Cournot oligopoly and two cooperative approaches, a cartel and a Nash
bargaining model. These structural models are then applied to the Atlantic steam coal
market in 2004 and 2005. Based on market observations we test the hypothesis that
the incumbent firms located in South Africa and Colombia exerted market power by
withholding quantities as a reaction to a new entrant, Russian, in order to keep the more
expensive fringe players such as the US as marginal price setters. This represents a novel
approach in partial equilibrium modeling of resource markets as we model companies
and not countries. Our results show that the expected reaction seen in the actual market
outcome can be best reproduced with the Cournot oligopoly model, suggesting that the
dominant firms exerted market power in a non-cooperative way.
3.2
Literature Review
Partial equilibrium models have been widely used to model energy and resource markets.
The subjects of these analyses are broad, ranging from market structure investigations
to infrastructure and policy analysis. Since the markets analyzed are often concentrated
markets with only a few firms or players involved, modeling market power is necessary.
One approach that has been extensively employed is the use of conjectural variations to
model different market structures. This concept was first developed by Bowley (1924) and
the term conjectural variation (CV) was introduced by Frisch (1933). It is used in static
oligopoly models to describe that a player takes into account in his payoff maximizing
decision that the other players might react to his decision. A formal description of the CV
model as well as a literature overview and a critical assessment are provided in Section
3.3.
The application of CV to partial equilibrium models started in the 1980s with Kolstad
and Wolak (1983, 1985) and their application to the coal trade in the Western US. In
these models the players exerting market power are not companies using quantities as
strategic decision variables but US states using a tax rate as decision variable in order to
maximize their revenues. The CV component comes from the fact that the states have
to make assumptions about how the other states will change their tax rates as a reaction
to a their own tax rate change. A similar tax rate and CV approach has been used to
model the international steam coal trade by Kolstad et al. (1983) and Kolstad and Abbey
(1984). These coal market models use different assumptions about the CV values to see
which market structure better represents trade. More recently Chen et al. (2006a) used
a tax CV model for the international rice trade with an optimization approach to find
what CV values led to a better representation of trade flows. These tax CV models also
allow for the representation of market power on the buyers side through the simulation,
for example, of oligopsonies.
A second and recently more prolific stream of literature applies the traditional CV approach to model market power. Here, the CV represents an assumption about how other
59
Chapter 3. Atlantic Steam Coal Market Power
players will react in quantity changes to the players own quantity changes. This approach
was first proposed by Nelson and McCarl (1984) and Kolstad and Burris (1986) who did
the first application to the international wheat market. Various other applications followed in the agricultural field, for example Kawaguchi et al. (1997) for the Japanese milk
market or for the international coking coal market in Graham et al. (1999). But it is
really in the 2000s due to new advances in algorithms and solving methods that the use
of CV models in the field of energy and resource modeling expanded greatly. We find
applications to the electricity market in Bushnell (2003) and Chen et al. (2006b). Another field where the CV approach has been used is in gas market modeling, for example
by Zwart and Mulder (2006) and Egging et al. (2008, 2010). In the field of coal market
modeling, which is the focus of the applied part of this chapter, recent use of the CV
approach was done by Paulus and Trüby (2011b) and Paulus et al. (2011).
In order to model different market structure settings these papers use different values
of CV for different players regarding the conjecture of how all the other aggregated players
might react to a change in the players output. This is why we will call this approach
mixed aggregated conjectural variation (MACV). This approach can be problematic, as
we show in the next section.
3.3
3.3.1
Critique of Conjectural Variation Models
Theory
First we give a formal overview of the CV theory using the formulation of Figuières
(2004) with two symmetric firms i competing in quantities qi on a market for a homoP
geneous good. The inverse demand function and market price is p = b − qi and the
i
firms’ productions costs are given by C(qi ) = cqi with c strictly positive. The profit
maximization of one firm i is as follows:
max Πi (qi, qj (qi )) = pqi − cqi
qi
(3.1)
qj (qi ) is the assumption of firm i about how the other firm j will set its quantities
given firm i’s quantities. This is the core assumption of the CV model. When we take
the total derivative of Πi with respect to qi we obtain the following first order condition
(FOC):
b − (2qi + qj ) − c − qi qj0 (qi ) = 0
(3.2)
which can be transformed into:
p − c − (1 + ri )qi = 0
(3.3)
where ri = qj0 (qi ) is the CV, or firm i’s conjecture, about the other firm j’s reaction to
a small change in qi . Different values of the CV in Equation (3.3) yield different market
60
3.3. Critique of Conjectural Variation Models
outcomes. ri = 0 is the Cournot-Nash equilibrium, ri = −1, is the Bertrand perfect
competition outcome where prices equate marginal costs. Values higher than zero and
up to 1 are also possible to model collusion and monopoly equilibria but we are more
interested in non-cooperative market outcomes. Hence, it is possible by setting the value
of ri to any number between -1 and 0 to compute a range of equilibria that are thought
to represent various degrees of competition or market power as stated for example by
Egging et al. (2008). In the following we evaluate this statement carefully as what seems
to be an easy and straightforward way to deal with imperfect competition can lead to
counter-intuitive and unsatisfactory results.
But before we critically assess the application of the CV model we give a quick
overview of the theoretical CV literature. As pointed out in the previous section the
concept of CV is old and dates back to Bowley (1924). However, due to the fact that
in the traditional CV equilibrium (CVE) the conjectures are exogenous, any arbitrary
equilibrium can be computed and these equilibria are not grounded in refutable economic
theory (see Figuières, 2004). Laitner (1980) identified that when using CVE a consistency
problem arises as a large set of possible output combinations and conjectures can satisfy
the equilibrium conditions. Therefore, a new stream of literature arose with Laitner
(1980), Bresnahan (1981) and Perry (1982) where the conjectures are required to be
rational or consistent. These conditions require that the expectations expressed in the
CV be met by the actual behavior of the players in equilibrium or in the neighborhood
of equilibrium (see Figuières, 2004). However, this approach has also been criticized
as not representing the behavior of profit maximizing firms by Makowski (1987). Also,
Lindh (1992) finds that this concept leads to circular reasoning and that in the case
of rational behavior based on common knowledge Cournot or Stackelberg equilibria are
the only possible equilibria. The main critique of CV models comes from the time
component. As Vives (2000) puts it, the CV approach “attempts what seems impossible
- the consideration of dynamics in a static model” as a player takes into account possible
reactions of the other players, but the other players will actually not react. Also the
expectation need not and might not be correct (see Friedman, 1983).
Figuières (2004) view on the CV from a theoretical perspective is that it may provide useful “shortcuts” to capture more complicated models in the absence of complete
information, common knowledge or a proper dynamic formulation. However, this is not
how the applied partial equilibrium literature has used this approach.
3.3.2
Partial equilibrium modeling applications
As shown in the literature review the first application of CV dates back to the early
1980s when the consistent CV model was developed and first critiques (for example in
Friedman, 1983) of both CVE and consistent CVE arose. Regardless of those theoretical
considerations applied researchers started using two features of CV models that had
never really been studied by the theoretical literature. First, different values of ri for
each player are used since the setting is usually more complex than the duopoly models
61
Chapter 3. Atlantic Steam Coal Market Power
of the theoretical literature. Secondly, aggregated conjectural variations, that treat all
the other players as one reacting to the player’s quantity in a unique way, are introduced.
These features explain the name mixed aggregated conjectural variation (MACV) that
we introduced for these kind of models.
One especially common feature of MACV models is to use Equation (3.3) with a
value ri = 0 for the dominant players and ri = −1 for the competitive fringe. This
is problematic from point of view of rationality and complete information since this
represents a situation where the dominant players naively consider the actions of the
fringe as given (see Gabriel and Smeers, 2006), as if the fringe players were also Cournot
players. Smeers (2008) further argues that, given the ambiguous nature of the MACV
model, a simpler way to model imperfect competition would be to use a mark-up on the
perfect competition price which is as arbitrary as using CV. As a matter of fact using a
MACV model does nothing else than imposing a relative mark-up on marginal costs and
cannot be considered as a model of rational profit-maximizing players.
Let us consider the case of a player with the CV of ri = 0. In that case the first order
condition is the same as in the Cournot-Nash equilibrium and if all other players behave
also as Cournot-Nash players, his FOC yields the non-cooperative profit-maximization
solution. Reorganizing Equation (3.2) and introducing parameter a as the slope of the
inverse demand function as well as C 0 (qi ) as an increasing marginal cost function, we
obtain the following expression for the first order condition:
aqi = b − aqi − aqj − C 0 (qi )
(3.4)
Given that the price is p = b − aqi − aqj we can see that aqi represents the mark-up as
difference between price and marginal costs. This mark-up is relative as it depends on qi .
We can see it as a percentage mark-up that defines the mark-up to equal a percentage a
of the quantity supplied. It is at this point that, if not all the players are Cournot-Nash
players, we depart from the idea of profit maximizing players. What the player does
in this case according to Equation (3.4) is to maintain this percentage mark-up. By
inspection we can determine what happens when player j behaves competitively with
a CV of rj = −1, or a zero mark-up. In comparison to the Cournot-Nash case qj will
increase. As a reaction and to keep the percentage mark-up and keep the price above the
competitive level, player i will decrease its quantity qi . In the case of two symmetric firms
with the same cost functions, the competitive firm will equate price to marginal costs.
But the firm supposedly exerting market power and subject to the same market price will
further reduce it’s quantity to hold the mark-up objective. This lowered quantity causes
its profit to be actually lower than if it had behaved competitively with ri = −1. A more
formal and differentiated proof of this occurrence is given in Ulph and Folie (1980). Here
we have the first demonstration that the MACV model may lead to counter-intuitive
and unsatisfactory results as the player supposedly having market power would be better
off by exerting less market-power or requiring a lower mark-up. We can derive formally
what the optimal mark-up would be.
62
3.3. Critique of Conjectural Variation Models
In the spirit of Laitner (1980), Daughety (1985) and Dockner (1992) that use dynamic
formulations to analyze CV models, we use a sequential formulation of the MACV model
allowing the two players i and j, starting with the one that is worse off in comparison
with its expectations, to revise and choose their CV value ri optimally. The first order
condition of i, and similarly j, is:
b − 2qi − qj − C 0 (qi ) − ri qi = 0
(3.5)
Furthermore, we use the following proposition from Kolstad and Wolak (1986): “assuming a quadratic duopoly model with conjectural variations no smaller than −1, and
with identical cost structures for the two firms, the equilibrium output for each firm is
monotone with respect to each of the conjectural variations”:
∂qi∗ /∂ri < 0,
∂qi∗ /∂rj > 0
(3.6)
Proposition 1: Given that the players i and j are symmetric with constant marginal
costs, one player can always increase its profit by reducing its CV value ri . The only stable
solution is ri = rj = −1. The optimal value for ri given the other player’s CV is:
ri = −
1
2 + rj
(3.7)
The proof is in Appendix 3.A.1. This proposition shows that in the simplest setting
of the MACV model the “exercise of more market power” is detrimental to the player.
If the player is given the possibility to determine his CV in a profit maximizing way, he
will in fact want to “exert less market power” and have a lower mark-up.
Proposition 2: If we introduce capacity constraints to the above model, and given
that the capacity of player j is binding when ri = 0, then ri = 0 is the optimal profitmaximizing strategy for player i. In the case that the capacity constraints are not binding
the rule of proposition 1 applies.
Appendix 3.A.1 to this chapter supplies a proof. This case highlights a situation
where the application of the MACV model may be valid. If we consider that player i
exerts full market power with ri = 0 and player j acts as the competitive fringe with
rj = −1 and has a binding capacity constraint, this model is equivalent to the dominant
firm competitive fringe model where the dominant firm equates marginal revenues to
marginal costs and takes the actions of the fringe as given.
This also relates to the result of Ulph and Folie (1980) that show that the incentive
to behave as Cournot-Nash players depends on the relative steepness of the slope of the
supply or cost function of the fringe. In the case of a binding capacity constraint this
slope becomes infinite. In the next proposition we show how the slope of the competitor
influences the optimal CV value.
63
Chapter 3. Atlantic Steam Coal Market Power
Proposition 3: Introducing heterogeneous cost functions of the form C(qi ) = (Cinti +
1/2Cslpi qi )qi
with Cinti > 0 the marginal cost function’s intercept and Cslpi > 0 the
marginal cost function’s slope, the optimal profit maximizing choice for player i is given
by:
ri = −
1
2 + Cslpj + rj
(3.8)
for qj > 0. If qj = 0, player i should increase ri until p + = Cintj with an
infinitesimally small number. This latter strategy represents the limit-pricing strategy
where an incumbent prevents the entry of a competitor in the market.
The proofs are in Appendix 3.A.1. Equation (3.8) shows that “exerting more market
power” than the opponent, or ri > rj , can be the optimal strategy if the opponent’s slope
Cslpj is high enough. The steeper the slope, the higher ri can be. However, since the
value of ri = 0 can never be reached and very high marginal cost function’s slopes of the
fringe may not be realistic, this formulation of the MACV model remains problematic.
It is also interesting to note that the setting where the players can sequentially correct
their CV, the above rule of proposition 3 has no stable converging solution.
We have seen that apart from the case where the competitive fringe has a binding
capacity constraint using the MACV approach in a quantity setting to model a dominant
oligopoly with a competitive fringe cannot be done properly. This is due to the fact that
the MACV model leads to a representation of players that are not profit-maximizing
rational actors, but naive margin-constrained players. This can, for example, be seen in
the MACV application to the coking coal market by Graham et al. (1999), where the
MACV is implemented as an actual constraint in a revenue maximizing problem.
As seen in the literature review, the first applications of the CV approach were done
through tax models of revenue maximization. The players are assumed to be public
governmental authorities that maximize a revenue Ri = πi qi by choosing a tax rate πi .
The producers of the resource are assumed to be competitive price takers and the taxing
authorities choose the optimal tax rate given the CV about how the other tax authorities
might react to a change of πi . The first order conditions of this problem are:
qi + πi
dqi
=0
dπi
p − πi − C 0 (qi ) = 0
(3.9)
(3.10)
To actually solve the problem, we would have to incorporate the CV by taking the
total derivative of expression (3.10). For a formal description of the model see Kolstad
and Wolak (1983, 1985). Applying the same arguments as above for Equation (3.4) and
using the above conditions (3.9) and (3.10) shows that the tax MACV model does not
present the same problem as the quantity CV market model. If the player i makes a false
conjecture about player j and a higher quantity qj is supplied to the market causing price
64
3.4. Developments in the Atlantic Steam Coal Market in the Early 2000s
p to fall, player i has the possibility to lower the tax rate and increase qi so that condition
(3.9) is still respected. The tax rate here is nothing else than the mark-up and since the
players have a direct possibility to influence it, the tax MACV model leads to an optimal
revenue solution whereas the quantity MACV model does in most cases not lead to a
maximum profit solution. The tax MACV model also prevents unsatisfactory outcomes
where players exerting market power would earn less than if they behaved competitively.
This is due to the fact that for decreasing values of πi , condition (3.10) converges to the
solution where price equals marginal costs. This is also shown graphically in Kolstad and
Wolak (1986, Figure 2). However, the problem of rationality remains as players could be
better off in the case that their conjectures about the tax adjustment of the other players
to their tax change is false.
We have seen that even in their simplest formulation MACV models can be highly
problematic to represent rational actors in a market with imperfect competition. As
Ralph and Smeers (2006) put it, the CV approach is “discredited as a theoretical explanation of market power but it allows representing it”. Therefore the MACV approach is
valid if there is an uncertainty about the structure of a market and when CVs are used
in a calibration process to replicate the market. However, using the MACV approach for
multi-period models (therefore assuming the same unclear market structure as today) or
in order to investigate market structures seems to be a meander not to follow since the
CV approach is not grounded in economic theory. We show other models more suited to
model dominant players with a competitive fringe in Section 3.5.
3.4
Developments in the Atlantic Steam Coal Market in the
Early 2000s
The international trade of steam coal for electricity generation developed in the wake of
the two oil crisis in 1973 and 1979 when more coal power plants started to be built using
inexpensive imported coal. Until the mid-2000s the international steam coal trade was
separated in two relatively distinct markets: the Atlantic market and the Pacific market.
In the Atlantic market the main importer is Europe and the most important suppliers
are South Africa, Colombia, Russia and the US. The importers in the Pacific market are
the Asian developing or developed economies such as Japan, South Korea, Taiwan and
India, supplied mainly by Australia, Indonesia and China. The focus of this chapter is
on the Atlantic steam coal market and the market that we study is the European steam
coal import market. European steam coal imports grew from around 50 million tons
(Mt) in 1978 to more than 100 Mt in 1990 and more than 200 Mt in 2007 (IEA, 2011a).
One other characteristic in the Atlantic market is the evolution of the contract nature
from long-term contracts (up to 10 years) before the 1980s to cargo based short term
spot contracts that represented approx. 80% of trade in 2003 (see Ekawan and Duchêne,
2006). The price indices can be FOB (free on board) at the export port and CIF (cost
65
Chapter 3. Atlantic Steam Coal Market Power
insurance freight) at the import port.29 Figure 3.1 shows the development of the main
indices for the Atlantic market in the 2000s. The FOB prices are from South Africa
(Richards Bay) and Colombia (Puerto Bolivar). The CIF index is the main European
import marker ARA (Amsterdam, Rotterdam, Antwerpen).
USD/t
Comparison of FOB and CIF prices
Comparison of FOB plus freight rates and CIF prices
240
85
220
80
75
200
180
160
140
120
70
cif_ara_6000
95
fob_richards_bay_6000
90
fob_bolivar_6300
85
cif_ara_6000
fob_rich_bay+freight
fob_bolivar+freight
80
65
75
60
v
70
55
50
65
45
60
40
55
35
50
Jan-04
Apr-04 Aug-04 Nov-04 Feb-05 May-05 Sep-05 Dec-05
Jan-04
Apr-04
Jul-04
Oct-04
Feb-05 May-05 Aug-05 Dec-05
100
80
60
40
cif_ara_6000
20
fob_richards_bay_6000
fob_bolivar_6300
0
Feb-99
Dec-99
Oct-00
Jul-01
May-02
Notable price differences
Mar-03
Jan-04
Nov-04
Sep-05
Jul-06
Apr-07
Feb-08
Dec-08
Oct-09
Source: Platts data
Figure 3.1: Price developments in the Atlantic steam coal market in the 2000s
As we can see in Figure 3.1, the two FOB price indices from South Africa and Colombia are on a similar level most of the time with the Colombian coal priced slightly higher
since it has a higher energy content of approx. 6300 kcal/kg than the South African
coal. However, we can distinguish two periods where there is a notable price difference
(circled in Figure 3.1). The period in 2009 where South African coal is priced higher
than the Colombian can be explained by the fact that South Africa started exporting
significant amounts of coal to India (more than 20 Mt, see Table 3.4 in Appendix 3.A.2)
which influenced the price-setting upwards. The second divergence happened in 2004
and 2005 where Colombian coal was priced significantly higher. The upper left graph
included in Figure 3.1 shows a “zoom” of this period. It seems that the FOB Richards
Bay price follows the CIF ARA price more closely whereas the FOB Bolivar price lags
in the general price decline observed in 2004 and 2005. In the upper right graph we
compare the CIF prices to the FOB price plus the freight rate on that specific route. In
a well functioning market those prices should be in a close range like it is the case in the
first half of 2004 and the second half of 2005. However, in the period in-between we see
the counter-intuitive phenomenon where the FOB plus freight prices are higher than the
29
There is no central commodity exchange where steam coal is traded. The price indices are based on
reported deals (spot: delivery three months ahead) submitted by traders to different organizations such
as Platts that collect and publish this information.
66
3.4. Developments in the Atlantic Steam Coal Market in the Early 2000s
(in Ireland owned
by the Big 3)
33%
33%
33%
g
ketin
mar
sive
lu
c
ex
35%
PRODECO
35%
100%
Figure 3.2: Market structure of the coal market in South Africa and Colombia in the early
2000s
CIF ARA price. The South African prices normalize by the beginning of 2005 but the
divergence remains in the Colombian prices until mid-2005. These divergences may be a
sign of the market power abuse we want to investigate in this chapter.
Indeed, there is potential to exert market power in the Atlantic market as the market
is dominated by three mining giants, the “big three”: BHP Billiton, Anglo American and
Xstrata that are present in South Africa and Colombia. In Table 3.4 (in Appendix 3.A.2)
we see that the share of South Africa and Colombia in European imports is high, but
declining: 61% in 2002 and 52% in 2005. The share of the “big three” in the export of both
countries is also high, around 80% (see Table 3.5 in Appendix 3.A.2). This concentrated
market structure is reinforced by the fact that in Colombia the “big three” own the largest
coal mine Cerrejón and market the coal together. The other important coal company in
Colombia, Prodeco, is owned by the international trading company Glencore that owns
35% of Xstrata. This complex ownership structure is illustrated in Figure 3.2.
The focus of our analysis is on the years 2004 and 2005. As we can see in Table 3.5 in
Appendix 3.A.2 during those years South Africa reduced its exports to Europe whereas
Colombia increased them slightly and Russia expanded significantly. Indeed, we can see
in the merit-order of suppliers in Figure 3.3 that Russia supplies Europe at approx. 45
USD/t and it is after this mark in the CIF ARA price was passed for the first time during
the year 2003 that Russia expanded its exports dramatically.
The hypothesis we examine in this chapter is that the incumbent oligopoly formed
by the “big three” reacted to the entry of the new competitive player Russia by exerting
market power by withholding quantities of South Africa exports. South Africa has some
mines that have quite expensive production costs so the withholding strategy makes
sense there and not in Colombia. By withholding quantities the incumbent oligopoly
tried to maintain high prices in 2004 and 2005 by keeping expensive marginal fringe
players in the market, especially the US. This is the hypothesis we will examine in the
67
Chapter 3. Atlantic Steam Coal Market Power
Source: Ritschel and Schiffer (2007)
Figure 3.3: Costs and merit-order of suppliers in the Atlantic steam coal market in 2004/05
next modeling section by modeling the strategic behavior of the incumbent oligopoly
reacting to a change of the supply of an aggregated competitive fringe.
3.5
Market Power Models with Dominant Players and a
Competitive Fringe
3.5.1
Modeling the competitive fringe
An alternative way to model a dominant oligopoly with a competitive fringe that could
prove to be more satisfactory than MACV model discussed above are models of the
Stackelberg type. In a Stackelberg game the dominant player, or leader, is not naive
like in the MACV approach but fully informed about the decisions of the fringe, or
followers, and can take them into account in its own decision. The classical Stackelberg
(1934) model was developed for the case of one leader and one follower. Sherali (1984)
expanded it for the case of multiple leaders and followers but with the restriction that the
leaders are symmetrical. One problem that can arise with the multiple-leader formulation
are multiple equilbria. This is due to the fact that the reaction function of the followers
that enters the problem of the leader is nonsmooth as they produce either a quantity that
is a function of the leaders output or zero. Ehrenmann (2004) show that this problem
can arise even in a very simple setting with two symmetric leaders and one follower.
However, more recent work by Xu (2005) and DeMiguel and Xu (2009) shows that under
certain conditions unique solutions can be reached, in particular that the fringe players
68
3.5. Market Power Models with Dominant Players and a Competitive Fringe
do not have a binding capacity limit. In our analysis we model the competitive fringe as
a single player and ensure that the fringe always provides a positive quantity, hence we
get unique solutions.
The quantity of the fringe qf as a function of the quantity of the leaders Ql shown
in Equation (3.13) is obtained by transforming the first order conditions (3.12) of the
profit maximization problem (3.11) after removing the expression −a · qf from (3.12) so
that we model the fringe as competitive player that equates the price to its marginal
costs. Parameters b and a are respectively the demand intercept and slope and Cintf
and Cslpf the slope and intercept of the marginal cost function.
1
2
max Πf = {b − a (qf + QL )} qf − Cintf · qf + Cslpf · qf
qf
2
(3.11)
∂Πf
!
= b − a (qf + QL ) − (1 + rf )a · qf − Cintf − Cslpf · qf = 0
∂qf
(3.12)
Assuming that the fringe is a competitive player then rf = −1 and
qf =
b − a · QL − Cintf
a + Cslpf
(3.13)
In the following section we discuss possible alternatives to model the dominant
oligopoly so as to reflect the market structure of the Atlantic coal market with the
“big three” companies.
3.5.2
Cournot oligopoly as Stackelberg leader
We can now integrate the quantity of the competitive fringe into the optimization problem
(3.14) of the leaders l. The leaders can operate in different countries e, so they maximize
their profit by choosing their optimal quantities qle to produce and sell given countryspecific quadratic production costs and production capacity constraints Caple .
PP



b−a·
(qle ) − Cintf  X

XX
l e

max Πl = b − a 
(qle ) +
(qle )
qle


a
+ Cslpf
e
e
l
X
1
2
−
Cintle · qle + Cslple · qle
2
e
s.t.
Caple > qle
(3.14)
(λle )
The optimization problem (3.14) leads to the following Karush-Kuhn-Tucker (KKT)
conditions (3.15) after deriving the Lagrangian function with respect to the decisions
variables and dual variables.
69
Chapter 3. Atlantic Steam Coal Market Power
PP


b−a·
(qle ) − Cintf
X
X
∂Lle
l e

06
= −b + a 
(qle ) +
∂qle
a
+
Cslp
f
e
l
2
a
qle + Cintle + Cslple · qle + λle ⊥qle > 0
+ a−
a + Cslpf
∂Lle
06
= Caple − qle ⊥λle > 0
∂λle
(3.15)
This model in its complete form with the upper level equilibrium problem of the
leader and the lower-level equilibrium problem of the fringe is actually an equilibrium
problem with equilibrium constraints (EPEC). It can be numerically solved in the mixed
complementarity format (MCP) using the PATH solver (see Ferris and Munson, 2000).
This model formulation represents the case where the members of the “big three” dominant oligopoly act in a strategic non-cooperative manner knowing how the other strategic
players as well as the fringe react to their actions.
3.5.3
Stackelberg-Cartel model
The Stackelberg-Cartel model formulation is the closest to the original Stackelberg (1934)
formulation. The leader acts as a single profit maximizing unit incorporating all the
players and their production countries e that are member of the cartel. What is being
maximized is a single joint profit under the assumption that a financial transfer between
the cartel members in possible. This problem can be modeled as a mathematical problem
with equilibrium constraints (MPEC, see Dirkse and Ferris, 1999) because one maximizes
the profit of the cartel under the equilibrium constraints of the fringe and the market
clearing condition that defines the price p depending on the total quantity supplied to
the market Q.
max Πl =p ·
X
qe
(qe ) −
e
s.t.
X
1
Cinte · qe + Cslpe · qe2
2
e
Cape > qe
(3.16)
0 6 − p + Cintf + Cslpf · qf ⊥qf > 0
p =b − a · Q,
p(f ree)
This model can be solved in GAMS with the non-linear programming (NLP) solver
CONOPT as an MPEC class of model (see Ferris et al., 2002).
3.5.4
Nash-Bargaining model
The Nash-Bargaining model is also a cooperative equilibrium between players but without
the possibility of financial transfer. This may be more realistic in the case of the “big
three” since they operate as separate entities in South Africa. The players jointly choose
the totally supplied quantities that are optimal for each of them. A numerical solution
70
3.6. Application to the Atlantic Steam Coal Market
method including stability conditions was first developed in Harrington et al. (2005).
In their formulation of the Nash bargaining game, players maximize the product of the
difference between the profit of a new cooperative equilibrium and the profit of the noncooperative Cournot-Nash equilibrium ΠN
l so that all players are better off in the Nash
Bargaining solution. Similarly to Harrington et al. (2005) we use optimization under
constraint to obtain a numerical solution.
max Πl =
qle
Y l
b−a·
X X
b−a
(qle ) +
PP
l
e
(qle ) − Cintf X
a + Cslpf
e
!
X
1
2
Cintle · qle + Cslple · qle
−
− ΠN
l
2
e
s.t.
(qle )
e
(3.17)
Caple > qle
This model can be solved in GAMS as a NLP using the solver CONOPT.
3.6
3.6.1
Application to the Atlantic Steam Coal Market
Model specification and data
In this section we present the detailed data of our market structure analysis. In the
years 2004 and 2005 the Atlantic steam coal market was dominated by a multi-location
oligopoly composed of the “big three” mining companies Anglo American, BHP Billiton
and Xstrata active in South Africa and Colombia. We analyze how this oligopoly may
have used its market power by reacting strategically to the arrival of a new entrant in
the market: Russia.
Table 3.1: Producer data: marginal cost parameters (intercept and slope) in USD/t and
production capacities in million tons per year
S. Africa
AngloAm
BHPB
Xstrata
Colombia
AngloAm
BHPB
Xstrata
Fringe
Russia, US
MC Int.
2004
MC Slope
MC Int.
2005
MC Slope
36
36
36
1.100
0.880
1.467
0.360
0.360
0.360
P. Cap.
60
20
25
15
25
8.33
8.33
8.33
43
43
43
0.346
0.346
0.346
P. Cap.
60
20
25
15
26
8.66
8.66
8.66
36
36
36
1.100
0.880
1.467
43
43
43
29
0.850
∞
29
0.680
∞
Source: own after: Ritschel and Schiffer (2007); Baruya (2007); IEA (2011a) and company reports
Table 3.1 presents the producer data and the data sources. In South Africa production
costs and capacity did not change between 2004 and 2005 and Colombia added 1 million
71
Chapter 3. Atlantic Steam Coal Market Power
tons per year (Mtpa) of capacity in 2005. The fringe aggregates the competitive players,
mainly Russia and the US, but since the capacity is not limited it also represents potential
entrants from the Pacific coal market. The main change between 2003 and 2005 is the
annual addition of 10 Mtpa of capacity in Russia from 30 Mtpa in 2003 to 40 Mtpa
in 2004 and finally to 50 Mtpa in 2005. Since Russia is the cheapest producer in the
“merit-order” curve of the fringe this is represented by a lowering of the fringe’s marginal
cost slope.
We construct one linear demand curve for European steam coal imports using the
following reference data: reference price pref = 55 U SD/t, reference quantity Qref =
160 M t and a price elasticity of demand = −0.3. Price and quantities are taken from
IEA, 2011a, the elasticity as well as the methodology are the same as in Chapter 2. Since
the European demand was stable between 2004 and 2005 we take the same demand data
for both years.
3.6.2
Results
The results for the market structure scenarios introduced in Section 3.5 for the years
2004 and 2005 are shown in Table 3.2. Additionally to the Cournot oligopoly, the Cartel
and the Nash bargaining cases this table also presents the perfect competition case as a
benchmark (modeled as MCP, see Chapter 2).
Table 3.2: Modeling results for the traded quantities in Mt and the prices in USD/t in 2004
and 2005 for the four market structure scenarios.
Big 3 S.A. (Mt)
Big 3 Col. (Mt)
Fringe (Mt)
Price (USD/t)
Cournot
04
05
52.4 43.8
25
26
60.5 70.8
80.4 77.2
Cartel
04
05
43
40
25
26
65.8 73.2
85
78.8
Nash-Barg.
04
05
42.9
39.9
25
26
65.9
73.2
84.9
78.8
Perfect C.
04
05
60
60
25
26
56.1 60.7
76.7 70.3
The strategic reaction of the incumbent cartel to the entry of Russia can be seen
in all three market power cases where we see a reduction of supplies from South Africa
below the competitive benchmark in 2004 and 2005. In the perfect competition case
South Africa produces at full capacity. The results of the cartel and Nash bargaining
simulations are almost identical because the players have similar cost functions and the
capacity retrictions are not binding in South Africa and equally binding in Colombia.
Proposition 4: In the case of a duopoly with two symmetrical players i with zero
production costs the produced quantities qi will be the same if they behave as a cartel or
as Nash bargaining competitors: qiC = qiN B =
b
4a
with a and b respectively the demand
curve slope and intercept.
Appendix 3.A.1 provides a proof but the intuition is straightforward. Since the players
are identical, they have identical benefits from withholding their supplies and do not need
72
3.6. Application to the Atlantic Steam Coal Market
financial transfers between players. Hence, they can reach the same monopoly solution
of the cartel case in the Nash bargaining competition.
If we look again at the simulated quantities in Table 3.2, the magnitude of the supply
withholding in the Cournot oligopoly case is more in line with the observed quantities
(see Tables 3.4 and 3.5 in Appendix 3.A.2) than in the Cartel and Nash bargaining cases
where the supplied quantities are significantly smaller. This suggests that players may
have exerted market power in a non-cooperative way. Even if the “big three” actually
cooperate in Colombia by owning a mine and marketing its coal together, the strategic
decision to exert market power is to reduce supplies from South Africa where the “big
three” are independent players.
If we look at the quantities supplied by the “big three” in South Africa the simulated
results in 2004 of 52.4 Mt are quite similar to the actual quantity of 51 Mt shown in
Table 3.5. However, this is not true for 2005 where the simulated withholding is too
high and the companies in reality increased production again. This mismatch of results
and observations is due to the yearly time resolution of our analysis and the fact that
the time frame where that market power was supposedly used stretches from mid-2004
to mid-2005, with a stronger effect in 2004 (see 3.4, Figure 3.1). In the second half of
2005 the “big three” may have stopped their strategic withholding and this is why they
exported more. Additionally, timing is a difficult issue in this analysis as the benchmark
values can differ in time. The price indices are for price deliveries three months ahead
and there is a time difference between exports and imports on which our yearly analysis
is somewhat artificially imposed.
In order to see if this effect was only relevant in the years 2004 and 2005 or existed
before and after we implement some additional simulations for the years 2003 and 2006
for the Cournot oligopoly case and the perfect competition case. Regarding the data
we assume production costs in 2003 about 10 to 20 per cent lower than in 2004 and
used the following reference prices and quantities for 2003 and 2006: p2003
ref = 45 U SD/t,
2006
2006
Q2003
ref = 140 M t, pref = 60 U SD/t and Qref = 175 M t. Table 3.3 shows the results.
Table 3.3: Modeling results for the traded quantities in Mt and the prices in USD/t in from
2003 to 2006 for the Cournot oligopoly and the perfect competition cases.
Big 3 SA (Mt)
Big 3 Col (Mt)
Fringe (Mt)
Price (USD/t)
Cournot Oligopoly
03
04
05
06
48.4 52.4 43.8 42.9
25
25
26
28
44.1 60.5 70.8 87.7
69.1 80.4 77.2 78.7
Perfect Competition
03
04
05
06
60
60
60
60
25
25
26
28
42.9 56.1 60.7 76.3
57.9 76.7 70.3 72.2
In the years 2003 and 2006 the actual prices and quantities seem to be more in line
with perfect competition, whereas to quantities supplied by South Africa in the Cournot
simulations are significantly lower. Therefore, it seems that market power was exerted in
73
Chapter 3. Atlantic Steam Coal Market Power
a limited time period as a sort of “defense mechanism” 30 against the new entrant Russia
between mid-2004 to mid-2005. But when the optimal withholding became too great or
would have attracted more entrants the market became competitive again in the second
half of 2005. This is in line with the results of Chapter 2 where perfect competition
better represents the global trade flows in 2005 and 2006.
In this analysis we considered only the benefit the “big three” may have had by using
market power through quantity. However, as we have observed in Section 3.4 this quantity
effect may have been accompanied by an unusual increase of Colombian (and to some
extend South African) F.O.B. prices. Higher Colombian prices would make the quantity
effect stronger or require less of a quantity effect for the same amount of additional profit.
However, there is no theoretical model to our knowledge where such a divergence can be
reproduced without breaking some assumption of the perfect market, especially perfect
information and rationality of players.
3.7
Conclusions
Integrating market power in large scale numerical partial equilibrium models is a difficult
undertaking. We have shown that a widely used approach using the theory of conjectural variations (CV) that we called mixed aggregated conjectural variation (MACV)
modeling fails to properly represent rational players exerting market power and may
lead to counter-intuitive results where the players exerting market power are worse-off
than if they had behaved competitively. This is due to the equilibrium conditions of
the MACV model that constrain the player to a given relative mark-up. In the case
that all the players have market power these condition represent the profit maximizing
Cournot-Nash equilibrium. However, when heterogeneous strategic behavior is assumed,
the MACV model fails to represent profit maximizing strategies because rational players
would want to choose another relative mark-up value to maximize their profits.
Given these shortcomings we have presented alternative modeling approaches based
on the Stackelberg model that ensure profit maximizing strategies of the players exercising
market power because they are fully aware of the reaction of the fringe players. This
approach was applied to the Atlantic steam coal market in 2004 and 2005 where some
unusual price and quantity effects occurred that may have been due to the strategic
behavior of the “big three” dominant companies. Different market structure scenarios
were implemented to model the dominant oligopoly: a non-cooperative Cournot-Nash
model and two cooperative models (cartel and Nash bargaining competition). We find
evidence that the oligopoly composed of the three mining companies BHP Billiton, Anglo
American and Xstrata exerted market power in a non-cooperative way between in 2004
and 2005 by withholding supplies from South Africa in order to keep more expensive
fringe players as price setters. The motivation for this strategic behavior is induced by
30
An alternative strategy of the incumbent cartel could have been to flood the market to prevent the
entry of competitors. This was however not possible due to production and export capacity limitations.
74
3.7. Conclusions
the entry and the production capacity expansions of a new fringe producer, Russia.
We find this exercise of market power to be an occurrence limited in time as in
the years 2003 and 2006 a competitive model is a better representation of the actual
market outcome. Also this punctual strategic behavior does not seem to be an issue that
would require the action of competition authorities since entry in the market is relatively
easy. Static models can deliver some insights in market power issues but in our case
more complex models where capacities are not fixed could also be helpful to analyze the
investment and entry dynamics.
We found in Chapters 2 and 3 that market power is not a fundamental and structural
issue in the global steam coal market. Thus, in the following Chapter 4 we introduce,
based on the assumption of perfect competition, a model of the global steam coal market
with a higher dimension in two aspects. The “COALMOD-Word” model provides a
virtually complete coverage of demand as it also includes domestic markets, and the
time horizon is not static for one year anymore, but runs until 2030 with endogenous
investments.
75
Chapter 3. Atlantic Steam Coal Market Power
3.A
Appendix
3.A.1
Proofs
Proof of proposition 1
Given the first-order conditions: b − 2qi − qj − C 0 (qi ) − ri qi = 0 and
b − 2qj − qi − C 0 (qj ) − rj qj = 0, and symmetry (C 0 (qi ) = C 0 (qj ) = c) we obtain
following expressions:
qj =
b−qi −c
2+rj
and ri =
−b+2qi +qj +c
−qi
Replacing the expression of qj in ri yields following expression for ri :
ri = −2 +
1
2+rj
−
b
c
−b+c+ 2+r
− 2+r
j
j
qi
Now to obtain the obtain the profit maximizing value of qi we modify the optimization
problem of player i by integrating the above expression of qi , so that the firm can
choose its profit-maximizing quantity given the other firms CV rj :
h
i
i −c
max Πi = b − qi + b−q
qi − C (qi )
2+rj
qi
∂Πi
∂qi
qi =
= b − qi −
b−qi −c
2+rj
+ −1 +
1
2+rj
qi − c = 0
−b+c−brj +crj
−2−2j
Integrating this expression of qi in the above expression of ri , yields, after simplification:
1
ri = − 2+r
j
Proof of proposition 2
Assuming that the capacity constraint Kj > qj of player j is binding then Kj = qj .
Then expression ri from the proof of proposition 1 can be rewritten as:
ri = −2 −
−b+c+Kj
qi
Accordingly, the new optimization problem of player i is:
max Πi = [b − (qi + Kj )] qi − C (qi )
qi
∂Πi
∂qi
qi =
= b − 2qi − Kj − c = 0
b−Kj −c
2
Integrating this expression of qi in the above expression of ri , yields, after simplification:
76
3.A. Appendix
ri = 0
If the capacity is not binding when ri = 0, given any rj , then it will never be since
∂qj∗ /∂ri > 0. A reduction of ri leads to a lower qj∗ .
Proof of proposition 3
Assuming now that the firm are asymmetric with heterogeneous cost functions of the
form C(qi ) = (Cinti + 1/2Cslpi qi )qi with Cinti > 0 the marginal cost’s intercept and
Cslpi > 0 the marginal cost’s slope, we obtain:
qj =
b−qi −Cintj
2+rj +Cslpj
and ri =
b−2qi −qj −Cinti −Cslpi qi
−qi
Replacing the expression of qj in ri yields following expression for ri :
ri = −2 − Cslpi +
1
2+Cslpj +rj
−
b
b−Cinti − 2+Cslp
Cint
j +rj
− 2+Cslp j+r
j
j
qi
Now to obtain the obtain the profit maximizing value of qi modify the optimization
problem of player i by integrating the above expression of qi , so that the firm can
choose its profit-maximizing quantity given the other firms CV rj :
h
max Πi = b − qi +
qi
∂Πi
∂qi
qi =
= b − qi −
b−qi −Cintj
2+rj +Cslpj
b−qi −Cintj
2+rj +Cslpj
b−Cinti − 2+r
b
j +Cslpj
i
qi − (Cinti + 1/2Cslpi qi )qi
+ −1 +
Cintj
j +Cslpj
+ 2+r
1
2+rj +Cslpj
/
qi − Cinti − Cslpi qi = 0
2
j +Cslpj
2+Cslpi − +r
Integrating this expression of qi in the above expression of ri , yields, after simplification:
1
ri = − 2+Cslp
j +rj
If for the calculated value of ri , qj = 0 because p < Cintj , then since ∂qi∗ /∂ri < 0 and
∂qi∗ /∂rj > 0 and because Cslpi > 0, player i can increase it’s profit by reducing it’s
quantity and increasing the market price until p + = Cintj with an infinitesimally
small number, which represents the limit-pricing strategy.
77
Chapter 3. Atlantic Steam Coal Market Power
Proof of proposition 4
Assuming two firms i = 1, 2 that are symmetric with zero production costs and a and b
respectively the demand curve slope and intercept. The profit function Π1 is:
Π1 = (b − a(q1 + q2 )) · q1
The first order condition is:
∂Π1
∂q1
= b − a(q1 + q2 ) − a · q1 = 0
Using the symmetry of the player q1 = q2 , then qiN =
b
3a
and ΠN
i =
b2
9a .
This is the
non-cooperative Cournot-Nash solution we will use for the following Nash Bargaining
game, still assuming symmetry:
b2
ΠN B = (b − 2 · a · qi ) · qi − 9a
· (b − 2 · a · qi ) · qi −
∂ΠN B
∂qi
= (b − 2 · a · qi ) · qi −
b2
9a
b2
9a
· (b − 2 · a · qi − 2 · a · qi ) · 2 = 0
The above equation has multiple solution. However, since we want the players to be
better off than in the Cournot case the first expression of the product has to be
positive, hence:
b − 2 · a · qi − 2 · a · qi = 0, therefore qiN B =
b
4a
.
Now the we calculate the results for the Cartel model with q1 + q2 = Q:
ΠC = (b − a · Q) · Q
∂ΠC
∂Q
=b−a·Q−a·Q=0
Thus, Q =
b
2a
and qiC =
Q
2
=
b
4a
.
We therefore proved that qiC = qiN B =
b
4a
.
78
3.A. Appendix
3.A.2
Market data
Table 3.4: Steam coal trade flows to Europe in million tons and import market share of South
Africa and Colombia
To Europe
South Africa
Colombia
Russia
US
Indonesia
Australia
Sum
S. Afr.+Col.
Share
S. Afr.+Col.
2002
2003
2004
2005
2006
2007
2008
2009
54.5
21.4
22.7
4.5
11.0
9.8
123.9
75.9
58.9
24.5
28.5
2.5
12.4
12.2
139.0
83.4
55.9
25.6
41.3
3.7
13.6
10.4
150.5
81.5
53.0
26.5
48.7
2.4
14.2
7.6
152.4
79.5
55.2
28.5
58.0
3.4
20.6
6.5
172.1
83.7
48.6
31.8
60.0
5.5
17.1
7.7
170.7
80.3
38.4
29.8
60.4
12.9
15.5
6.4
163.5
68.2
30.6
34.0
61.1
11.1
12.8
5.0
154.6
64.5
0.61
0.60
0.54
0.52
0.49
0.47
0.42
0.42
3.8
2.9
0.7
2.4
3.6
7.7
7.8
20.6
To India
South Africa
Source: IEA (2011a)
Table 3.5: Export volumes in million tons and export market share of the “Big Three”
South Africa
Anglo American
BHP Billition
Xstrata
Total S.A. Exports
Total Big 3
Share Big 3
Colombia
Anglo American
BHP Billition
Xstrata
Glencore
Prodeco (Glencore)
Total Col. Exports
Total Big 3
Share Big 3
2002
2003
2004
2005
2006
2007
2008
15.7
23.6
12.6
68.5
52
0.76
18.6
22.8
13.8
70.9
55
0.78
17.4
20.5
12.9
67.0
51
0.76
20.3
21.9
13.5
70.9
56
0.78
22.8
20.3
13.2
68.1
56
0.83
24.0
18.8
13.7
65.1
56
0.87
22.3
12.4
12.3
59.4
47
0.79
6.9
5.4
8.7
8.2
9.6
8.2
10.1
8.7
11.0
9.5
9.2
11.3
9.9
9.9
11.5
10.5
10.5
6.0
5.0
36.5
30.3
0.83
8.0
5.0
45.6
38.6
0.85
9.7
5.0
50.9
42.1
0.83
8.5
5.0
53.6
42.3
0.79
5.0
62.0
45.7
0.74
13.5
64.6
55.8
0.86
13.5
67.8
57.5
0.85
Source: company annual reports 2002-2008
79
Chapter 4
A Techno-economic Analysis using
the COALMOD-World Model: The
End of “Cheap Coal”?
4.1
Introduction
This chapter introduces a tool to analyze the future developments of the international
steam coal market, the “COALMOD-World” model.31 The model includes virtually all
producing and consuming regions in the world by modeling domestic markets along with
the globalized seaborne market to see their interaction. The time horizon is 2006 until
2030. COALMOD-World is a multi-period model that simulates yearly market outcomes,
trade flows and prices for the years 2006, 2010, 2015, 2020, 2025 and 2030, as well as
investments in the coal sector’s production capacity and transport infrastructure. Trade
flows and investments may be subject to various capacity or expansion constraints. We
assume profit maximizing players who optimize their expected and discounted profit over
the total model horizon. In the model we integrate a wide range of geological, technical
and economical data and mechanisms that aim at a more realistic depiction of the future
coal market than is realized in previous models. We include the main drivers of the
market such as future demand. Geological data is integrated in the form of reserves,
heterogeneous coal qualities and with an endogenous costs mechanism that depends on
cumulative production and investments. Technical constraints also influence the model
outcomes and the whole model framework is grounded in economic theory and gametheoretic concepts.
We apply the model to two scenarios: one that sees global demand of coal continuously
increasing and another where the demand stabilizes after 2015. As in both cases demand
increases in Asia, especially in India and China, one main result of our modeling exercise
is an increase of the international seaborne trade both in absolute terms and in relative
31
This chapter is an updated and modified version of Haftendorn et al. (2010) and a modified version
of the forthcoming article Haftendorn et al. (2012a).
80
4.1. Introduction
terms compared to global consumption. We also expect an increase in imports from Asia
as well as a shift of global trade flows toward that region. Another significant result is
that until 2030, the end of cheap coal will not be caused by geological reserve constraints
but rather by infrastructure constraints. Especially in the scenario of a continuously
increasing global coal demand driven by Asia and China, the market may not be able to
supply enough steam coal due to restrictions in the expansion of mining and transport
capacities. These restrictions affect not just domestic supply in India and China but
also the global seaborne suppliers such as South Africa. Stabilizing world coal demand
after 2015 will lead to a less tight future market situation. A stabilization of future coal
demand will be beneficial both to the climate and to the global energy supply costs by
keeping coal relatively cheap.
Our research is motivated by the fact that international trade and global demand for
steam coal is mainly driven by demand in Asia. Thus, we want to be able to identify how
the interplay between domestic supply, exports and imports driven by demand as well as
by supply costs and constraints will influence future trade flows and prices. Another area
where we hope to make a contribution is to the discussion about “the end of cheap coal”
outlined in the eponymous comment written by Heinberg and Fridley (2010) in Nature.
The authors argue that “useful coal may be less abundant than has been assumed” (p.
367). The authors cite three recent studies that predict a more or less imminent end of
cheap and available coal through decreasing global production levels and rapid reserve
depletion (Patzek and Croft, 2010; Höök et al., 2010; Mohr and Evans, 2009). These
studies are based on the concept of the Hubbert curve first described by M. King Hubbert
(1959). The core mathematical assumption of this model is that cumulative resource
production follows a logistic growth path that derived with respect to time yields the
well-known symmetrical bell shaped curve of yearly production output; the summit of the
curve representing the “peak” of the yearly production rate. Thus, it is mathematically
possible to estimate the shape of the curve and thus the peak year as well as the ultimately
recoverable reserves, defined as the surface under the curve, based solely on historical
production data.32 This simple technique is subject to controversy. Its proponents
claim that “the Hubbert curves are based on [...] production and not on ill-defined and
subjective [...] ‘reserves’ ” and that “historical production trends reflect the prevailing
economics prior to the time of production” (Patzek and Croft, 2010, p. 3111). However,
its opponents, such as Lynch (2003), state that the “work of the Hubbert modelers has
proven to be incorrect in theory, and based heavily on assumptions that the available
evidence shows to be wrong. They have repeatedly misinterpreted political and economic
effects as reflecting geological constraints, and misunderstood the causality underlying
exploration, discovery and production” (p. 30). We do not intend to settle the general
debate about the Hubbert curves but rather make a critical evaluation of the three
aforementioned papers using this method for coal. We give an overview of the papers
starting with the earliest predicted “peak coal”, finishing with the one with the latest
32
A description of the mathematical transformations is given in Claerbout and Muir (2008)
81
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
peak date.
Patzek and Croft (2010) predict the global coal peak as early as 2011. The methodology used is the closest to the one described by Hubbert. The authors use historical
production data to fit Hubbert curves for each coalfield and then sum the data up. Hence
the name “multi-Hubbert cycle analysis”. To be able to perform this summation, the authors assume that coal mines are “independent of each other” in terms of production.
We find this to be a rather strong assumption. For example a power plant located in
Southern China that is able to choose between domestic coal from northern Shanxi or
imported coal from Indonesia will make an economic calculation based on the extraction
and transport costs. Thus, if coal from Indonesia is cheaper to source, the higher production from Indonesia will cause lower production in Shanxi, ergo the mines are not
independent. Also there may be some problems with using only historical production
data. The imminence of the predicted peak in China, also 2011, may be due to the
fact that infrastructure constraints, reforms and price regulations that can slow down
production are interpreted as geological depletion. Indeed, Peng (2011) found that the
effect of the recent coal market deregulation, while power prices are kept regulated, is
that “under pressure of price increases from domestic coal suppliers the power sector
responded by going to overseas markets to purchase coal”. Interestingly, another study
by Lin and Liu (2010), also using the Hubbert curve methodology, predict a much later
time for the peak, between the late 2020s and the early 2030s. Patzek and Croft (2010)
also concede that “Hubbert cycle predictions almost always underpredict the true future
production rate of a resource” but argue only verbally why new production capacity will
be hard to put in place due to transport or environmental restrictions. This is essentially
a question of investment dynamic that the Hubbert curve model is not able to integrate.
Also, past production can only represent past economic situations. The simple Hubbert
model fails to integrate paradigm shift that will affect the future production pattern such
as the carbon constraints of climate policy or the high economic growth in Asia.
Höök et al. (2010) use a more refined model as they integrate past production and
reserve estimates for the Hubbert curve fitting. In the “standard case outlook” the authors
predict the peak around 2025-2030. The global production level at that time is in the
same range as other projections such as the IEA (2010) World Energy Outlook Current
Policies scenario. The authors discuss reserves estimates extensively and show how they
developed over time. The underlying reserve estimations of the “standard case outlook”
is more restrictive than the ones of national geological services such as the USGS or
the German BGR and does not account for the expansion of the reserves. Thus, the
authors also model a “high case outlook” with a doubled reserve base. Logically the peak
is further in the future, around 2040. But the estimated annual production values also
increase and after 2015 these values are significantly higher that the estimates of the IEA
(2010) World Energy Outlook Current Policies scenario that represents the worst case in
climate policy. The Hubbert method, by trying to fit a bell curve, overestimates future
production. This is caused by the fact that realistic demand projections are not included
82
4.2. Equilibrium Modeling of Energy Resource Markets
in the model by Höök et al. (2010). The higher reserves of the “high case outlook” would
then mean that significant amounts of coal are available for at least 100 years.
The most complex model is developed by Mohr and Evans (2009). They include
supply and demand reactions, some form of investment (mine upgrade) mechanism and
distance themselves from the classic Hubbert methodology because “there is no underlying
theory explaining why production ought to follow a symmetric bell curve”. Using their
model the authors predict a peak in 2034 in the “Best Guess scenario” and are able to
produce an extraction path with a plateau with realistic maximal yearly extraction rates.
However, the model also relies on the fitting of historical data, that comes with the issues
discussed above, and needs input values for constants that seem difficult to estimate. The
supply and demand interaction, as well as the mine upgrades, is not based on economic
theory but on a mathematical mechanism using “supply gap” values. The authors also
recognize that the iterative process used to compute the equilibria in the model takes
“several hours to run” and that “the number of constants makes application of the model
difficult and time consuming”.
In our chapter we would like to change the affirmation made by Heinberg and Fridley
(2010) about “the end of cheap coal” into a question and show that equilibrium modeling
may be the better way to answer that question because it is able to integrate market and
investment mechanisms as well geological and infrastructure constraints.
4.2
Equilibrium Modeling of Energy Resource Markets
An extensive review of the – rather sparse – coal-market specific modeling literature is
provided in Chapter 2. Altogether, there is little modeling effort applied to international
coal markets in general and in particular using modern modeling techniques provided by
equilibrium modeling. Kolstad and Abbey (1984) are a notable exception, but their static
analysis covers the 1980s. However, both the situation on the international steam coal
market as well as modeling techniques have since evolved. Paulus and Trüby (2011a) use
a spatial equilibrium model to show how a Chinese infrastructure decision, transporting
coal-based electricity over long distances to the demand centers rather than the coal
itself, could affect the global market positively through reduced Chinese imports.
We follow the stream of literature of detailed equilibrium (complementarity) models of
various resource markets.33 The development of the COALMOD-World model is rooted
in the previous static, one-period model “COALMOD-Trade” (see Chapter 2), as well as
in the multi-period modeling experience of other markets (e.g., Egging et al., 2008, 2010;
Huppmann and Holz, 2009). In the following we present a brief overview of the existing
literature of complementarity models with endogenous investment decisions of resource
markets.
Complementarity models are numerical models that provide solutions to optimization
33
In this section, we ignore the extensive literature on modeling of electricity markets. In this literature,
many modeling advances have been made including the formulation and solution of complex, multi-stage
equilibrium problems.
83
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
problems under constraints (e.g., Cottle et al., 1992). The complementarity format can
be used to model games, in particular non-cooperative market games such as a Cournot
game. The complementarity model gives the Nash equilibrium solution, which is why they
are also called equilibrium models. These are formulated using the optimality conditions
(called Karush-Kuhn-Tucker conditions, or KKT) of the optimization problems under
constraints.34 Often, the optimization problems are profit maximization problems of
representative player types, with some given economic and technical constraints. For
a tractable model, some assumptions such as perfect and complete information of all
players and over all model periods are generally made.
There is wide interest in modeling natural gas markets, both at the European and
the global levels. Huntington (2009) provides an overview of some of these models.
Similar to the path that coal market modeling is taking, there was a predominance
of optimization models of natural gas markets for a long time. However, natural gas
markets, in particular in Europe, are characterized by strategic behavior with market
power exercise. Hence, one must use modeling techniques that can represent strategic
players, such as equilibrium modeling. Equilibrium modeling of natural gas markets, first
with static models, was initiated by Mathiesen et al. (1987) and especially Boots et al.
(2004). Once this model technique was well developed in the static context, attention
turned to the inclusion of investment decisions in a multi-period framework (Lise and
Hobbs, 2008; Egging et al., 2010). Endogenous investment decisions in these models are
generally limited to the transport infrastructure of pipelines and liquefied natural gas
ports, with production capacities given exogenously.
None of the multi-period models of natural gas markets incorporate endogenously
changing short-term production cost curves. Lise and Hobbs (2008) use long-run marginal
costs to incorporate the opportunity costs of production and state that using short-run
costs would underestimate the full costs. While this may be true for the gas market,
we believe that short-term marginal costs curves are better at representing the yearly
market outcomes of the coal market. It is also very difficult to obtain long run marginal
cost data. In their model, Egging et al. (2010) use the same cost functions for every
model year. Hence, in these natural gas market models the cost curves do not vary over
time. One exception can be found in Hartley and Medlock (2006), where the long-run
production cost curves shift in the future according to an assumed rate of technological
innovation in exploration and development costs. However, these changes are exogenous
as they are not dependent on the change of other model variables.
There is less literature on equilibrium models of the international oil market. However, many problems are similar to natural gas or coal markets: in the short run, the
prevalent market structure is unclear, with the possible economic models ranging from
perfect competition to cartel. Moreover, in the long run, capacity expansion both in
production and transport infrastructure (ports, pipelines) is an important prerequisite in
this market (IEA, 2009) and needs better modeling. The FRISBEE model (Aune et al.,
34
Put simply, these are the first order conditions of the optimization problems.
84
4.3. The COALMOD-World Model
2005) is a model of the global oil market with a focus on the Organization of Oil Exporting Countries (OPEC) and its production economics. The Oilmod model (Huppmann
and Holz, 2009) includes the price pools in the international market that are reference
prices for all international oil sales (e.g., Brent, WTI).35 These models can include finite
resources (reserves) as a constraint to the optimization such that an optimal reserve extraction path (under constraints) is implicitly obtained as solution. We also adopt this
approach for the coal market where reserves are globally available for many more decades
but may be limited in the near future for some countries (BP, 2009).
4.3
The COALMOD-World Model
The model setup follows the organization of the value added chain of the steam coal
sector. The value chain is complex and there are various types of players involved at
each stage. Producers can be large national and sometimes state-owned companies.
There are a few large multinational coal companies as well as many smaller companies,
usually operating in one country only. Transport infrastructure can be built by the
mining company or by another entity. Often, it consists of rail infrastructure, but in
some countries trucks or river barges are used. Export ports can be exclusively used
by one company or used by multiple companies, with a variety of possible ownership
structures. Traders as intermediaries also play a role as they can be vertically integrated
or contractually connected to every stage of the industry.
In Chapter 2, we provide an analysis of the market structure for the global steam coal
trade and simplify the value chain for the modeling purpose. There is some evidence that,
contrary to the oil market, the international steam coal market tends to be competitive.
This result allows us to make some simplifying assumptions for the COALMOD-World
model: since in a competitive market prices equal marginal costs, we can simplify the
role of the players on the value added chain to obtain two types of model players, the
producers and the exporters, shown in Figure 4.1. The two model player types, producers
and exporters, will maximize their profit in a perfectly competitive way and thus act as
price takers.
In Figure 4.1 the steps of the real-world value-added chain that are included in the
model are represented by the small rectangles included in the larger producer and exporter boxes. We exclude the coal import terminals and the subsequent land transport
link to the final consumer because this capacity is assumed to be sufficient. De facto,
we situate demand that cannot be reached by land close to the import port. The second
type of demand node can be reached by a land link directly from the producer. The producer player includes the coal mining company and also the land transport links. The
exporter operates the export terminal and pays for the sea transport. These players are
aggregated at a national or regional (sub-national) level.
Research on international coal markets points out that the traditional separation of
35
This static version is currently being expanded to a dynamic model with investments.
85
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
Coal producing company
Coal Mine:
underground
or opencast
Raw coal
Producer
Ex
po
rt
er
Run of
mine
stockpile
Raw coal
Coal
preparation
plant
Washed coal
Conveyor belts
Trucks
Stockpiles
at rail
loader
Railroad / barge/ truck coal transport
Final customer:
Power Plant
Coal export terminal
Seaborne coal transport
Coal import terminal
Railroad / barge/ truck coal transport
Final customer:
Power Plant
Demand
Figure 4.1: Model players on the steam coal value-added chain
the Pacific and the Atlantic market has faded (e.g., Ellerman, 1995; Warell, 2006; Li,
2008; Zaklan et al., 2009). In our model, we therefore consider the global market as one
integrated market, albeit the spatial aspect of the market where transport costs play a
role in determining the trade relations is not neglected.
4.3.1
Model structure
The COALMOD-World model is a multi-period equilibrium model of the global steam
coal market with two types of players: producers, f , and exporters, e, facing consumers,
c, represented by a demand function. The COALMOD-World’s model producers and
exporters represent stylized players defined for aggregated production, export and consumption nodes primarily determined using geographical parameters. A production node
represents a geographically restricted area (mining basin) and aggregates the mining companies present in that area into one player called producer. In the model, production node
and producer are equivalent terms. Production nodes are defined based on the following
criteria: geography of reserves, type of coal, and production cost characteristics.
$/GJ
Quality factor:
k = t / GJ
Pe
E}
$/t
Q
P}
Q }
k=
s
E*
st
co
t
or
sp
an
r
T
Tra
nsp
sts
E
Pc
PJ
Port operation
costs
$/GJ
E*k
=Q
ort
co
tes
t ra
ig h
Q
Fre
*k=
C
}
PJ
C
Pc
P: Producers
E: Exporters
C: Consumption
}: Capacity restriction
Figure 4.2: COALMOD-World model structure
An export node represents the coal export terminal of one region and aggregates
86
4.3. The COALMOD-World Model
the real world coal export harbors present in that region into one model player called
exporter. Here again, export node and exporter are used as equivalent terms. The export
nodes are primarily defined based on geographic factors.
A demand node represents a geographic area where the coal is consumed. It aggregates the consumption by the coal-fired power plants in a region. It can have access to
seaborne coal through a port or not. The demand nodes are primarily defined based on
geographic factors, but other factors may come into play such as the connection to a port
or the presence of mine-mouth power plants.
Figure 4.2 represents the model structure and the relationships between producers,
exporters and demand. The model runs until 2040 and calculates yearly equilibria for
the energy quantities sold in the years 2006, 2010, 2015, 2020, 2025, 2030, 2035 and 2040,
which we call “model years”. Also, the players can make investments in each model year
that will be available in the next model year.36 Thus, the model not only calculates
an equilibrium within each model year but also over the total model horizon regarding
optimal investments. For the years between the model years we interpolate the produced
quantities since these are necessary in order to model the reserve depletion. We assume
that production and other capacities will be made gradually available in the years between
the model years to reach their new value in the following model year. Both producer
and exporter problems are profit maximization problems over the entire model horizon.
The players have perfect foresight, meaning that they choose the optimal quantities to
be supplied in each model period and the investments between model periods under the
assumption of perfect information both about current and future demand. Thus, the
model simulates how demand should be served optimally, given that the players behave
rationally using all the information that is available to them. In the following section,
we present and explain the optimization problems of the model.
It is important to note that the traded quantities xaf c , yaf e and zaf c are the energy
quantities contained in the coal, expressed in Petajoules. Whenever the model needs to
deal with mass quantities in million tons of coal (for the costs, capacities and investments)
these energy quantities are converted in mass using a conversion factor κ defined in tons
per Gigajoule that is different for every producer.
4.3.2
The producers’ problem
• The producer’s profit optimization problem
The producers maximize their profit ΠPf (xaf c ; yaf e ; P invaf ; T inv_caf c ; T inv_eaf e ) over
the total model horizon A for all model years a ∈ A. The producers extract and treat
(produce) the coal and can sell it either to local demand nodes (xaf c ) or to the exporters
(yaf e ). They bear the production and the inland transport costs. Further, they can
invest in additional production capacities (P invaf ) and in transport capacities to local
36
We only interpret the results until 2030 because there is a risk of distortion of the investment results
given the short payback period after 2030.
87
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
demand (T inv_caf c ) or to the exporter (T inv_eaf e ). These investments are subject to
constraints.
ΠPf (xaf c ; yaf e ; P invaf ; T inv_caf c ; T inv_eaf e )
max
xaf c ; yaf e ; P invaf ; T inv_caf c ; T inv_eaf e
a hX
X
=
a∈A
1
1 + rf
pac · xaf c +
·
X
pae · yaf e
e
c
!
P
− Caf
X
X
xaf c · κaf +
yaf e · κaf
c
−
X
e
trans_caf c · xaf c · κaf −
c
X
trans_eaf e · yaf e · κaf
e
− P invaf · CP invaf
− T inv_caf c · CT inv_caf c − T inv_eaf e · CT inv_eaf e
i
(4.1)
s.t.
Xh X
X
P capf −
xa0 f c · κa0 f +
ya0 f e · κa0 f
a0 <a
c
!
· mc_int_varf
i
e
!
+
X
P inva0 f −
X
X
xaf c · κaf +
yaf e · κaf
a0 <a
c
X hX
a∈A
xaf c · κaf +
c
P inv
αaf
(4.2)
(4.3)
X
yaf e · κaf
e
5i
X
X
+
xa−1f c · κa−1f +
ya−1f e · κa−1f ∗
2
c
e
T cap_cf c +
e
P maxinvaf − P invaf ≥ 0
Resf −
P
αaf
≥0
X
αfRes
≥0
T inv_caf c − xaf c · κaf ≥ 0
(4.4)
(4.5)
(4.6)
T cap_c
αaf c
a0 <a
T cap_ef e +
X
T inv_eef c − yaf e · κaf ≥ 0
T cape
αaf
e
a0 <a
T maxinv_caf c − T inv_caf c ≥ 0
T maxinv_eaf e − T inv_eaf e ≥ 0
88
T inv_c
αaf c
(4.7)
(4.8)
T inv_e
αaf e
4.3. The COALMOD-World Model
xaf c ≥ 0; yaf e ≥ 0; P invaf ≥ 0; T inv_caf c ≥ 0; T inv_eaf e ≥ 0
(4.9)
In the second line of the producers’ objective function (4.1), we see that the summation
of the yearly net revenues in the squared brackets over all model years with the associated
discount rate rf . The following two terms after the brackets are the revenues from sales
to local demand nodes and to exporters. The third line of (4.1) shows the production
cost function in an undefined form. The fourth line of (4.1) represents the transport
costs to local demand and exporters. Line five of (4.1) calculates the total investment
costs in production capacity and line six does the same for the investments in transport
capacities to local demand and exporters.
The constraints are valid for each model year, except the constraint on the reserves (4.4)
that must hold over the total model horizon. Equation (4.2) represents the production
capacity constraint for one year, which depends on the capacity in the starting year and
investments in subsequent periods prior to the model year. Equation (4.3) is a restriction
on the maximum investments in production capacity that can be build up during the
next five years (i.e., until the next model year). (4.4) is the reserve constraint of the
producer over the total model running time and includes reserve utilization from the
production of the years between the model years. On the domestic transport market we
have (4.5) and (4.6), which are the transport capacity constraints for each model year for
transport routes to local demand nodes and exporters, respectively. (4.7) and (4.8) are
the respective maximum investments in additional transport capacity similarly to (4.3).
The symbols in parentheses are the dual variables associated with the constraints and
(4.9) are the non-negativity constraints of the decision variables.
• The production cost function
In this subsection, we specify the production cost functions for each period that were
left undefined in the previous subsection. Since the cost functions appear in each period,
we also call them short-run cost functions. Generally, we assume a quadratic cost function
of the type:
Cf = (mc_intf +
1
· mc_slpf · qf ) · qf
2
(4.10)
This leads to the following linear marginal cost function:
mcf = mc_intf + mc_slpf · qf
(4.11)
Since we have an energy based model but mass dependent production costs, we use the
conversion factor κf explained in detail in Section 4.3.4 to obtain the following marginal
cost function depending on the quantity qf expressed in energy units:
κf · mcf = κf · mc_intf + κ2f · mc_slpf · qf
(4.12)
Some resource markets models use the same short-run costs for every model period
89
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
(e.g., Egging et al., 2010). This is not a realistic solution for a model of the coal market
since there are many potential factors influencing future costs and changing short run
costs. Other models only use the long run marginal costs (e.g., Lise et al., 2008). This
is also problematic for a model of the coal market since the short-term marginal costs
determine the prices in each period and, as we have seen in our previous static modeling
work in Chapter 2, enable us to represent the trade flows accurately. In the following,
we discuss the influential factors and their impact on the short-run cost functions.
Geological factors are the main driver and reason for variability between production
costs, as described in BGR (2009). First we can distinguish between opencast and underground mining. Furthermore, the geological structure of the deposit such as the thickness
and depths of the seams as well as their inclination and the nature of the geological formation that hosts the seams influence the mining costs. On the techno-economic side
Rogner (1997) identifies future rates of technology change as well as productivity gains
as critical drivers for potential future production costs. For our own assessment we primarily use the geological factors and to a lesser extend assumptions about the potential
for productivity gains.
mc
mc
q
q
a. Aggregation from the individual mines' marginal costs to
the model producers’ (mining basin) marginal cost curve.
b. Effect of cumulated production on the intercept of the
model producers’ short-run marginal cost curve.
mc
mc
Model producer type:
1
q
2
3
4
Q
c. Effect of new investments in production capacity on the
slope of the producers’ short-run marginal cost curve.
d. Long-run marginal cost curve and short-run marginal
cost curves over the lifetime of a mining basin.
Figure 4.3: Endogenous cost mechanism in relation with short and long-run marginal costs
At the highest level of aggregation Rogner (1997) found that the long-run production
cost curves for all fossil fuels (oil, natural gas and coal) over the total potential reserves
have an S-shaped form similar the one shown in Figure 4.3d. We assume that a mining
basin, because it also represents a high level of aggregation, has a similar cost development
as the cumulative production increases. The exact form of the curve may vary but it
is important to distinguish four types of situations that a mining basin will be in over
90
4.3. The COALMOD-World Model
its lifetime as shown in Figure 4.3d. First, a mining basin has some easily accessible
resources (often the cause of an accidental discovery). But since these resources are
limited, production costs increase rapidly. Second, the production costs reach a relative
plateau, as the bulk of the reserves are similar in nature. Third, when the bulk of the
reserves is completely mined, costs start to increase more or less proportionally with
the cumulative production. Fourth, and finally, for the last deposits that are hard to
reach, extraction costs rise rapidly. Each coal mining basin can be put in relation with
one of these four types. Consequently, we assign each producer to one such type.37
This determines how the short-run costs will develop between 2010 and 2030. Before
we categorize the producers, we explain the endogenous cost mechanism starting at the
individual mine level.
Figure 4.3a shows the logic of aggregation of individual mines in a mining basin to
form the model producers’ marginal cost curves. We assume that a specific mine in
a certain geological setting operates at constant marginal costs. The horizontal line,
together with the dashed line, represent the reserves of a mine. The horizontal line
represents the production capacity at a given point in time. Thus, in order to obtain the
aggregated cost curves in one period, we add the production capacities on the q-axis and
connect it with its respective marginal costs on the mc-axis.
After this static consideration, let us consider how this cost function might evolve
over time. We first consider the effect of cumulative production, as illustrated in Figure
4.3b. We follow the rules stated by Hotelling (1931) that for exhaustible resources, the
cheapest deposits are extracted first and go further by assuming that, even if all the mines
along the cost curve may produce coal in one period, the cheapest mines are depleted
first.38 The principal reason is that, generally, the cheap mines are the oldest ones in
operation. The effect of cumulative production from one model period to another makes
the cheapest producer in Figure 4.3a disappear from the cost curve. This causes the
intercept of the cost function to increase as shown in Figure 4.3b. This is the core of the
first endogenous cost mechanism that enters the model with the following equation:
mc_intaf = mc_int(a−1)f
!
+ mc_slp(a−1)f ·
X
x(a−1)f c · κ(a−1)f +
c
X
y(a−1)f e · κ(a−1)f
· mc_int_varf ,
e
mc_intaf (free)
(4.13)
Equation (4.13) states that the intercept in year a is equal to the previous period’s
37
In the current model setup until 2030 each model producer stays in the assigned type. However, for
longer-term simulations a dynamic setup where producers change types could also be implemented.
38
This may not always be true in reality as some old and cheap mines may still have decades of
life expectancy. We do not model individual mines but provide a reasonable approximation for the
developements on a mining basin basis. The existence of cheap mines that will operate for a long time
can be captured in our endogenous cost mechanism by a slow increase in the intercept and a decrease in
the slope of the marginal cost curve. This is for example the case in the Powder River Basin.
91
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
intercept plus the previous period’s slope multiplied by the production in that year
and the factor mc_int_varf ∈ [0, 1]. The factor mc_int_varf determines how fast
the cheapest mines are mined out. It determines the position on the cost curve of the
previous period to determine the new intercept. Graphically, this is the passage from
Figure 4.3a to 4.3b. If the factor is one it means that the cumulative production leads
to a complete depletion of all the mine capacity that produced in the last period. This
may be true for mature and old mining basins. On the contrary, a factor close to zero
means that the mines situated on the low cost segment of the basin’s cost curve still have
significant reserves and will only be depleted in the mid to long-term.
The second endogenous cost mechanism included in the model simulates the effect
of new investments in production capacity or the addition of new mines to the slope
of the marginal cost curve. Graphically, this is represented by the step that leads from
Figure 4.3a to Figure 4.3c. Mathematically, this mechanism is described by the following
equation:
mc_slpaf = mc_slp_startf + mc_slp_varf ·
X
P inva0 f , mc_slpaf (free) (4.14)
a0 <a
The factor mc_slp_varf ∈ R, in equation (4.14), represents the effect of the cumulative investments in production capacity on the slope of the marginal cost curve. A
value of zero is used in the case that there is no influence of the investments on the slope
(model producer type 3). A negative value of mc_slp_varf causes the slope to decrease
(model producer type 2) and a positive value increases the slope with new investments
(model producer types 1 and 4).
In order to implement this mechanism we add the two equality constraints (4.13 and
4.14) and their respective complementarity variables to the producer’s problem. The two
equations are affine; thus, the KKT conditions are sufficient conditions for optimality.
The overall problem remains convex.39
• Mine mortality mechanism
The logic behind the mine mortality mechanism is already included in the previous
section about endogenous costs where we explained that the factor mc_int_varf determines how fast the cheapest mine are mined out, and thus, also the mine mortality,
or how much of the existing capacity disappears relative to the cumulative production
over the years. The term that is subtracted every year and is included in the production
capacity restriction in equation(4.2) of the optimization problem
and equation (4.30) of
i
Ph P
P
the KKT conditions is −
xaf c · κaf + yaf e · κaf · mc_int_varf .
a0 <a
c
e
39
The only detail that must be watched is in the case of a negative parameter mc_slp_varf . If
this parameter is not chosen correctly in the calibration process and is set very low, there is a risk
that equation (4.14) calculates a negative value for the slope mc_slpaf . This would make the model
non-convex and infeasible to solve. A careful calibration based on geological and techno-economical
information wards off such a risk since in reality we do not expect changes in the slope to be too drastic.
92
4.3. The COALMOD-World Model
4.3.3
The exporters’ problem
The exporters maximize their profit, ΠE
e (zaec ; Einvae ). Each exporter is linked to a
maximum of one producer. The profit for each year shown in (4.15) inside the squared
brackets is defined by the revenue from sales net of the costs of purchasing the coal at
a FOB price pae from the producer in the second line, the costs of operating the export
terminal in the third line, the costs of transport (shipping) to the final market c in the
fourth line and finally in the last line the costs of investing in additional export capacity.
The yearly profits are summed over the total model years and discounted by a rate re .
The index c represents a demand node. An exporter can only sell to a demand node
with a port. The exporter’s decision is to choose the optimal quantity zaec to sell to each
importing country c in each year a and also to invest in export capacity Einvae .
max
zaec ; Einvae
ΠE
e (zaec ;
a
1
Einvae ) =
·
1 + re
a∈A
hX
X
pac · zaec −
pae · zaec
X
c
c
X
− zaec · Cportae · κae
c
X
− zaec · searateaec · κae
c
i
(4.15)
(4.16)
−Einvae · CEinvae
s.t.
Ecape +
X
Einva0 e −
X
zaec · κae ≥ 0
a0 <a
µE
ae
c
Emaxinvae − Einvae ≥ 0
Emaxcape − Ecape −
µEinv
ae
X
Einvae ≥ 0
(4.17)
µEmax
e
(4.18)
a
zec ≥ 0; Einvae ≥ 0
(4.19)
Constraint (4.16) represents the maximum export capacity in each model year, which
depends on the capacity in the starting year and investments in subsequent periods prior
to model year a. Equation (4.17) expresses the maximum investments in export capacity for one model year. (4.18) represents the maximum possible investments over the
total model horizon until 2040. This constraint allows the model to determine endogenously during which model year the port expansions should take place. The symbols in
93
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
parentheses are the dual variables associated with the constraint.
4.3.4
Other model equations
• Quality equation
κ af
[t/GJ]
Total
reserves in
mass
Total
reserves in
energy
κf
PJ
Decreasing quality of coal reserves
Figure 4.4: Producer’s quality definition relative to its reserves
Since each model producer represents an entire mining basin with various mines and
significant amounts of reserves, the quality of the produced coal may not be constant
over time. Figure 4.4 visualizes this fact and shows how it would affects the model.40
The x-axis represents the energy value of the reserves and the y-axis the quality factor
κ associated with the reserves. The areas in this graph represent million tons. For
one model producer, we have different reserve blocks represented by the black to gray
blocks. We assume that the higher quality coals are mined first, thus the reserve blocks
are ordered by decreasing coal quality. Using this information we obtain the increasing
line over the hatched area by using a linear regression. Equation (4.20) formulates this
relationship between reserves and coal quality mathematically.
!
κaf = κf + δf ·
X
X
a0 ≤a
c
xaf c +
X
yaf e
(4.20)
e
Since each model exporter has a dedicated model producer, the quality factor κae of
the exporter is equal to the quality factor of the producer that supplies him for any given
year.
• Final demand
Final demand is located at a consuming node c. The following market clearing condition determines the price given the demand function pac (xaf c , zaec ) at the demand
40
This feature is integrated in the model but has not been used for the simulation runs of this chapter
due to the lack of data to properly determine the parameter δf . Thus, this parameter is set to zero for
all producers in equation (4.20).
94
4.4. Model Specification and Data
node.

pac − pac 
X

X
xaf c ,
zaec  = 0 , pac (free)
(4.21)
e
f
The producers can be in indirect contact with the final demand through their exporter
or in direct contact with their domestic demand. The prices are expressed in USD per
GJ, because we concentrate on the demand for energy embodied in the coal.
We assume a linear inverse demand function of the type pac = aac + bac · qac for
each consumer, c, in each model year, a. We construct a different linear inverse demand
function for each demand node, c, using their reference prices (pref
ac ) and reference demand
value (qac ) for the model starting year 2006 and use projections for future years. We make
assumptions about the demand elasticities (εac ). In particular, we define bac =
ref
and aac = pref
ac − bac · qac , following the demand elasticity definition εc =
pref
ac
ref
qac
ref
qac −qac
pac −pref
ac
·
·
1
εac
pac
qc .
This gives the following inverse demand function depending on the consumed quantity
P
P
qac = xaf c + zaec :
e
f
1 ref qac
ref
pac = pac +
−1
(4.22)
p
ref
εac ac qac
• Market clearing
In addition, one must consider market clearing conditions ensuring that the coal sold
by the producer to the exporter in a node equals the coal sold by the exporter to all the
importing demand nodes. This condition also determines the price pae at the exporting
node.
0 = yaf e −
X
zaec
, pae (free)
(4.23)
c
• China’s export restriction
Modeling China’s export restriction requires the additional equation (4.24). The
Chinese coal exports are restricted by politically determined export licenses. Thus, we
put a constraint on all consumption nodes with a non-Chinese import port (i.e., countries
N oChina(c)) using equation (4.24). China_lica ECHN represents the level of Chinese
export licenses for a given year in million tons.
China_lica ECHN −
X
zaec · κae ≥ 0
(πa ECHN )
(4.24)
N oChina(c)
4.4
Model Specification and Data
In Section 4.3.1 we have introduced the concepts of nodes and model players. The model
simulates the market on an aggregated basis, that is we do not include individual mines
95
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
or coal-fired power plants. However, the spatial characteristics of the market and the
transport costs associated with that spatial aspect make it necessary to define aggregated
nodes in the different producing and consuming countries. Section 4.4.1 describes our
choice of model countries and nodes before providing a detailed overview of our data in
the subsequent sections.
4.4.1
Countries and nodes definition
Figure 4.5: Countries included in the COALMOD-World database
We include all countries that were either consuming at least 5 Mtpa41 or producing
and exporting at least 5 Mtpa in 2006. Some more countries that are expected to become
relevant players on the global market in the next decades are included, too (e.g., Mongolia,
Mozambique). The world map in Figure 4.5 shows the represented countries including
their role on the world coal market (importer, exporter, or both).
In our dataset, we distinguish production and consumption nodes. Hence, a country
that only produces for export is represented in the data set with a production node
from which it also exports (e.g., Colombia). A country that only imports and consumes
coal is included with a consumption node (e.g., Italy, Turkey). For a country in which
production takes place and that also consumes coal, we include at least one production
node and one consumption node. For larger countries, there can be more than just
one production/demand node; this is the case for the U.S., China, India and Australia.
The complete list of the countries and nodes in the model can be found in Table 4.4 in
Appendix 4.A.2.
Producing nodes are generally defined by mining basins which are restricted by geological realities. The location of power plants is more dispersed as it relates to human
settlements. This makes it more difficult to locate our consuming nodes. For the con41
Mtpa: million tons per annum.
96
4.4. Model Specification and Data
sumers that can only be reached through an importing port we define the demand as
being located close to the port. For consumers that can be reached by land we aggregate regional data on capacities to form the demand node and define an average for the
transport costs.
4.4.2
Production, costs and reserves
The data collection required a major effort since there is no central source available. We
collected data from publicly available sources, that are detailed in the following.42 Most
of the cost data is based on Baruya (2007): “Supply costs for internationally traded coal ”
(IEA Clean Coal Center). For each export country, Baruya (2007) provides estimates for
the low and high average costs. This information is used to construct the producers’ cost
functions of the base year. We assume that the average costs also represent unit costs
for the cheapest and the most expensive mine. Thus, in the short run we have the same
variable costs and marginal costs for one mine. We construct a marginal cost function
using the low estimate to determine the curve intercept. We place the second point at
the intersection of the high cost estimate and the maximum production capacity in order
to obtain a linear marginal cost curve as described by equation (4.11) in Section 4.3.2.
Theses costs can be seen in Figure 4.6 for some selected producers as a component of the
FOB supply costs.
For the producers from Eurasia, Colombia, Venezuela, South Africa and Indonesia,
the cost data and the parameters of the marginal cost function are based on Baruya
(2007); the capacity data is based on Rademacher (2008), except for Eurasia where the
capacity data is based on actual production assuming a capacity utilization of 90%.
Countries with more than one production node require more detailed data on production capacities that was determined using the following sources: for the U.S., Energy
Information Administration (2008, Tables 1 and 12); for China, data from the National
Bureau of Statistics of China (2007); and for Australia, data from the New South Wales
Department of Primary Industries (2009) and the Queensland Department of Mines and
Energy (2009) with a capacity utilization assumption of 80% (Rademacher, 2008, p. 78).
For Vietnam, the production capacity is taken from Rademacher (2008). For India
the actual production data from Datanet India (2009) is used, assuming 90% capacity
utilization. This assumption is also made for Poland using IEA (2008a). The Indian cost
data is based on average cost data for each subsidiary company of Coal India Ltd.. Since
there was no cost data available for Vietnam, these were determined using relevant price
data. For Poland, the costs are based on Ritschel and Schiffer (2007).
In order to determine the cost functions in the long run, some assumptions on the
42
Overall there is some scarcity of data in the public domain and improvements could be provided
by using more detailed data. The model would especially benefit from better cost data since it is a
competitive, cost-driven model. With accurate cost data and projections, the model could even be used to
deliver forecasts of future prices to a certain extent. Despite the issues mentioned above the COALMODWorld database is able to provide realistic runs and give insights into the future developments of the
global steam coal market as is shown in Section 4.5.
97
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
mining basin types and the intercept increase pace had to made. Table 4.5 in Appendix
4.A.3 shows these assumptions and the values of mc_slp_intf and mc_slp_varf . The
assignment of the model producers to a producer type is based on informations about
geological factors of each basin, the age of mines, as well as the prospects of future
productivity improvements. The main sources for this assessment are Minchener (2009),
EPRI (2007) and Ritschel and Schiffer (2007). For the U.S. the report by Luppens et al.
(2009) that is part of the National Coal Resource Assessment Overview was used.
Table 4.1: Reserves of major countries in COALMOD-World in Million tons
China
U.S.
Eurasia
India
South Africa
270800
270718
93494
63968
48740
Australia
Poland
Indonesia
Colombia
Mongolia
38593
13997
10000
6229
1170
Source: own work based on Energy Information Administration (2008); Geological Survey of India
(2008) and National Bureau of Statistics of China (2007)
In Table 4.1 we show the distribution of global steam coal reserves by major global
producing regions used in the model. In order to use consistent reserves data, we base
ourselves primarily on one source: Energy Information Administration (2008, Table 8.2).
It follows the standard definition of reserves by the World Energy Council:
“proved recoverable reserves are the tonnage within the proved amount in
place that can be recovered (extracted from the earth in raw form) under
present and expected local economic conditions with existing available technology” (Energy Information Administration, 2008).
This data is aggregated on a national level, thus, to get the distribution on a sub-national
level other sources had to be used. For the U.S., Energy Information Administration
(2008, Table 15) was used. The reserve distribution to the Indian production nodes is
based on the Geological Survey of India (2008) and for the Chinese producers on the
National Bureau of Statistics of China (2007). For Indonesia we considered the reserve
number of 4967 Mt too restrictive given the relatively low exploration activity in large
portions of Kalimantan that are thought to be coal rich. Thus, the estimate of 10000
Mt, roughly the double of the present estimates.
The coal quality parameter κf expressed in t/GJ is the inverse of the data in GJ/t
shown in Table 4.2. It is based on Platts (2009) for the U.S., Colombia, Venezuela,
Poland, Russia, South Africa, Indonesia and Australia. For China, it is based on IEA
(2007c) and for India on Datanet India (2009)43 . For Vietnam the quality data is taken
from Ritschel and Schiffer (2007).
The Energy Watch Group (2007) provides evidence that coal quality is generally
decreasing over time as the reserves are mined. According to this study the decline in
43
Spreadsheet: Grade/Company-wise Production of Non-Coking Coal, 1999-2000
98
4.4. Model Specification and Data
Table 4.2: Energy content and quality κf of coal by production nodes
Node
P_USA_PRB
P_USA_Rocky
P_USA_ILL
P_USA_APP
P_COL
P_VEN
P_POL
P_UKR
P_KAZ
P_RUS
P_ZAF
P_IND_North
P_IND_Orissa
Energy content in
GJ/t
20.00
26.52
26.05
29.08
26.69
29.31
26.38
25.96
25.12
26.80
26.80
17.83
12.73
Node
P_IND_West
P_IND_South
P_VNM
P_IDN
P_CHN_SIS
P_CHN_Northeast
P_CHN_HSA
P_CHN_YG
P_AUS_QLD
P_AUS_NSW
P_MNG
P_MOZ
Energy content in
GJ/t
17.53
17.53
29.31
22.82
25.54
23.45
22.61
21.77
27.21
26.38
25.54
26.80
Source: own work based on Platts (2009); Datanet India (2009) and Ritschel and Schiffer (2007)
coal quality is not only due to a shift toward lower rank coals, like sub-bituminous coals,
but also to a quality decline within each class. This is the reason why we implement
this effect in the model through the factor δf introduced in Section 4.3.4. However, no
consistent data is available for all production nodes and we are using δf = 0, implying
no quality variation over time. The model still captures some of this effect through the
different coal qualities of the producers. For example, if the recent developments in the
U.S. continue with more (lower grade) coal from the Powder River basin being produced,
the overall quality of U.S. coal will decrease.
4.4.3
Land transport
The land transport costs (trans_caf c , trans_eaf e ) and capacities (T cap_cf c , T cap_ef e )
are associated with the transport from a producer to local demand or to an exporter.
This represents mainly transport by train but can also include road transport on trucks
and in certain cases river transport by barges. The transport costs are assumed to be
constant over time.
The transport costs for Colombia, Venezuela, South Africa, Indonesia, China and
Australia are based on Baruya (2007). For these countries, transport capacity data is
based on relevant production, consumption and export data. For the U.S., this data
is based on Energy Information Administration (2004) for the transport costs. The
transport capacities inside the U.S. are determined using actual flow data given in Energy
Information Administration (2007).44 The transport cost data for the Eurasia region is
from Crocker and Kovalchuk (2008) and the capacities are determined using relevant
production, consumption and export data. This method is also used to estimate the
transport capacities in Vietnam and India. The Vietnamese costs were based on relevant
44
Spreadsheet Domestic Distribution of U.S. Coal by Origin State, Consumer, Destination and Method
of Transportation
99
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
price data. The Indian transport cost data is based on Datanet India (2009).45
4.4.4
Export ports
The data for the export ports includes the export capacity in the starting year and the
port handling costs. For Colombia, the Eurasia region, South Africa and Australia, the
capacity data is taken from Ritschel and Schiffer (2007) and the cost data from Baruya
(2007). For Venezuela the costs are assumed to be similar to Colombia, the capacities
are determined using relevant export data. For the U.S., both cost and capacity data
are taken from Baruya (2007) as well as the Chinese port handling costs. Chinese port
capacities are provided by the National Bureau of Statistics of China (2007). The costs
for Poland are taken from Ritschel and Schiffer (2007) and the capacity is based on export
data.
90
80
70
USD/t
60
50
40
30
20
10
low cost in USD/t
diff=(high - low) in USD/t
C
hi
na
C
ol
om
M
bi
oz
a
am
bi
qu
e
lia
(N
Au
SW
st
ra
)
C
lia
an
(Q
ad
LD
a
(fr
)
om
PR
Ea
B)
st
R
us
W
si
a
es
tR
us
si
a
Bl
ac
k
Se
So
a
ut
h
Af
ric
a
Ve
ne
zu
el
a
In
do
ne
si
a
Au
st
ra
Po
la
n
U
SA
d
0
transport cost in USD/t
port fee in USD/t
Source: own work based on Baruya (2007); Ritschel and Schiffer (2007) and National Bureau of
Statistics of China (2007)
Figure 4.6: FOB costs for all export countries implemented into COALMOD-World
The Chinese political export restriction that is introduced in Section 4.3.4 is assigned
to the Chinese exporter. For 2006 we use a value close to the actual exports of 59 million
tons (Mt). In 2007 this quota was 70 Mt and dropped further to 47.7 Mt in 2008 and
was likely not higher than 45 Mt in 2009.46 Forecasting the level of future export licenses
is difficult, and there are no such projections available. For the base case we assume the
following values: 2006, 60 Mt; 2010: 60 Mt; 2015: 80 Mt; 2020: 90 Mt; 2025: 100 Mt;
2030, 110 Mt; 2035: 120 Mt and 2040: 130 Mt.
In sum the cost of a ton of exported coal adds up from production costs, land transport
costs and the export fee. This is shown in Figure 4.6 for each exporter. In this figure, we
also include the range of production costs in the respective production area. This range
45
46
Spreadsheet Railway Freight on Coal in India
Source: China Daily website
100
4.4. Model Specification and Data
is represented by a white bar in the figure; it is calculated by subtracting the lowest
average costs from the highest average costs.
4.4.5
Freight rates
Source: own work based on Platts newsletters 2002-2009
Figure 4.7: Linear regression of average freight rates between 2002 and 2009
Freight rates result from the supply-demand equilibrium in the dry bulk carrier market and have been very volatile in the past.47 In general, the freight market behaves
cyclically. This makes it difficult to predict future freight rates, which we need as transportation cost input for the model. For the same route there is also a difference between
Capesize and more expensive Panamax freight rates. The capacity of Capesize ships
is higher but Panamax vessels are used more often on shorter routes. In the model, we
assume the freight rate (transport cost) to be dependent on distance to reflect the spatial
character of the international coal market. Given historical information on weekly freight
rates on all available routes, we specify a linear regression using distance as explanatory
variable.48 The model transport costs between every export node and every import node
with import possibility are calculated for 2006 using this equation by plugging in the
corresponding distance x.49 In 2006, the freight rates were below their average values
between 2002 and 2009. The regression equation obtained for the average between 2002
and 2009 is y = 0.0014x + 13.97. The computed values y are used as model transport
costs which are set to be constant from 2010 until 2040. Figure 4.7 depicts the regression
results of the freight rate data points in dependency on distance.
4.4.6
Demand
For the specification of the demand function of each consumption node, we need the
“reference” price and “reference” quantity data for each model year. In order to have a
47
Dry bulks include commodities such as iron ore, coal or grain.
Sources for weekly freight rates are McCloskey and Platts newsletters 2002-2009.
49
Distance calculated using the PortWorld online distance calculator, http://www.portworld.com/
map/
48
101
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
consistent demand database for all countries in the model we use primarily data from
the International Energy Agency (especially IEA, 2008a,b) and from the World Energy
Outlook 2010 (IEA, 2010). We implement the two main scenarios. In the increasing
demand scenario, based on the IEA Current Policies scenario, it is assumed that as of
mid-2010 no change in the current policies will be implemented and that the recently
announced commitments are not acted upon. In the stabilizing demand scenario, based
on the New Policies scenario, the recently announced commitments and policies, for
example the ones of the 2009 Copenhagen Climate Conference, are fully implemented.
The IEA data is expressed both in mass and energy units and thus fits the purpose of
our modeling work since we model demand in energy values (Petajoules). However, the
IEA data is on a very aggregated level, so the demand projections of the IEA (2010)
must be allocated to the model’s demand nodes. To achieve this, we take a bottom-up
approach based on national data and ensure consistency by checking with the IEA data.
Table 4.3: World Energy Outlook demand projections for coal for power generation in the
reference scenario converted to Petajoules
Region
WORLD
OECD N.A.
U.S.
OECD Pacific
Japan
OECD Europe
EU
Eurasia
Russia
Non OECD Asia
India
China
Middle East
Africa
2006
86876
22064
20808
6406
2554
10132
10312
6322
3433
38812
7249
28973
335
2512
Current
2015
110155
22776
21227
6783
2680
9127
9043
6322
3768
61127
9043
47143
419
2847
Policies
2020
119031
22106
20892
6866
2470
8876
8415
6071
3475
70673
11807
52419
544
3182
2030
140216
21897
20683
6364
2219
8081
7243
6908
4271
91356
16287
64937
879
3894
New Policies
2020
2030
106931 107810
21101
18129
20139
17459
6029
4271
2219
1424
6866
5359
6490
4815
5568
5485
3224
3224
63556
70673
10844
12812
47018
50242
419
586
2721
2680
Source: own work based on IEA (2008c)
For Japan, Korea, the EU countries, Turkey, Israel, Ukraine, Kazakhstan, Morocco
and the other Asian countries, IEA data was used to determine the relevant quantities,
either directly or as a share in a world subregion. For Spain, Germany and the UK we
model only import demand by subtracting the local production from demand. For the
demand projection we assume that the production in these countries drops by half in
2010 and stops in 2015. Price data was taken from the IEA and regional/national data.
Canada: the quantity data for 2006 is based on Statistics Canada (2009) and the
distribution for the future is based on the share of Canada in the OECD North America
region. Price data was not available and was estimated using U.S. FOB price data.
U.S.: the quantity data for 2006 is based on Energy Information Administration
(2008, Table 26) and this repartition is used to estimate the future share of the U.S.
model consumers in the U.S. projection. The prices are based on Energy Information
102
4.4. Model Specification and Data
Administration (2008, Table 34), converted to USD/GJ.
Russia: the quantity data is based on Energy Forecasting Agency (2008) which provides data on installed and projected capacities for coal-fired power plants with detailed
geographic coverage. Assuming a capacity factor of 80% and an average thermal efficiency of 35% (Crocker and Kovalchuk, 2008, p. 30) we get similar coal consumption
levels expressed in energy units as the IEA. The regional breakdown is used to determine
the shares of projected Russian demand of the two demand nodes. Inland price data was
not found and is estimated using relevant cost and export price data.
China: to get a regional breakdown of the IEA data for 2006 and the future to the
model consumers, Chinese provinces’ coal consumption data from the National Bureau of
Statistics of China (2007) was used. Price data is based on the China Coal Transportation
and Distribution Association’s (CCTD) database.
India: The quantity data is based on Datanet India (2009) and is consistent with
the IEA data. The consumption values of India’s demand nodes for 2006 was used to
allocate the IEA projection for India proportionally. The price data is determined using
data from the Indian Ministry of Coal (2005, p. 58) that estimates the Indian delivered
steam coal price to be between 12 and 16 USD per million kcal for distances between
1000 and 2000 kilometers.
South Africa: the quantity data is based on the IEA (2008a) for 2006 and the share
of South Africa in the region Africa is used to estimate future demand. The local price is
determined using the value of local sales in 2006 divided by the volume of sales.50 This
gives an average price of 13.69 USD per ton that is converted to USD per Gigajoules
using the relevant quality factor.51
There are only a few studies that incorporate long term price forecasts for coal.
EWI/Prognos (2005) forecast quasi constant prices from 2010 until 2030 at approximately
1.5 Euro(2000)/GJ (p. XX). A more recent study by the European Commission (2008)
forecasts a price decrease in 2010 in comparison to 2005 and then a continuous but slow
increase until 2030 (p. 11). The assumption that prices in 2010 are lower than in 2005
can not realistically be made given the recent development of prices. Hence, in our data
base we set the 2010 prices at the same level as 2006 and then increase all prices by 0.2%
every five years, which is congruent with the price growth forecasted by the study of the
European Commission (2008).
As shown in Section 4.3.4, own-price elasticities of coal demand are part of our demand curve definition. However, empirical research on elasticities, especially for coal,
is scarce and the results are often not very satisfying. Dahl (1993) estimates short run
elasticities to be between -0.55 and -0.3. Aune et al. (2001) use a value of -0.19 for the
short run elasticity of coal demand in their model. The most recent study by Liu (2004)
yields a rather peculiar result of a zero elasticity that is, of course, of rather limited use
50
Chamber of Mines of South Africa (2008, p. 18). Conversion done using average historical exchange
rate for 2006 provided by http://www.oanda.com/convert/fxhistory
51
The quality of the coal sold to the local market is very low with about 19 GJ/t, therefore the price
per ton is low, too.
103
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
for defining demand functions for the model. We conclude that the price elasticity of coal
demand is rather inelastic and assign elasticities εc of −0.1, −0.2 or −0.3 to the model
consumers based on the percentage of coal use in the total power generation. The more
dependent a country is on steam coal use in its electricity sector, the less elastic demand
is assumed to be.
4.4.7
Investments
The investment costs are a major input to the multi-period model since they determine
the future investment decisions. For the value-added chain from production to the export terminal, the IEA estimates investments costs of 50 USD (2007) per ton of annual
capacity addition (USD/tpa) and for some new projects this number goes up as high as
80 USD/tpa (IEA, 2008c, p. 136). Rademacher (2008) finds average investment costs
of 62 USD/tpa with a wide range from 15 USD/tpa for some Australian opencast mines
to 130 USD/tpa for new underground mines in Ukraine and Mozambique (p. 75). But
investment costs in Australia can also exceed 100 USD/tpa if the project includes new
transport and washing facilities. We therefore assign values from 40 to 80 USD/tpa to
the different producers’ investment costs in production capacity based on informations
about the country and the prevalent type of mining.52 Unit investment costs and the
production capacity for the base year and every production node are shown in Figure
4.8.
90
80
1200
70
1000
60
800
50
40
600
30
400
20
200
Investment Cost in USD/t
Production Capacity 2006 in mtpa
1400
10
0
Australia
Production capacity
IDN MOZ POL Rest of RSA
Asia
P_VEN
P_USA_ILL
USA
P_USA_Rocky
P_USA_APP
P_ZAF
P_USA_PRB
P_VNM
P_POL
P_MNG
P_IDN
P_MOZ
P_IND_West
India
P_IND_South
P_IND_Orissa
P_RUS
COL Former CIS
P_IND_North
P_KAZ
P_UKR
P_COL
P_CHN_YG
China
P_CHN_Northeast
P_CHN_SIS
P_CHN_HSA
P_AUS_QLD
P_AUS_NSW
0
VEN
Investment costs
Source: own work based on IEA (2008c) and Rademacher (2008)
Figure 4.8: Capacity and investment costs for all production nodes in the base year
52
The assignment is based on factors such as the prevalent type of mining, geology and the state of
technology.
104
4.4. Model Specification and Data
Investments in additional overland transport capacity are set in a range between 10
and 55 USD/tpa depending on distance, landscape/relief and if the project is mostly
greenfield or not. Investment costs for additional export capacity are set between 10 and
30 USD/tpa depending on the country and the preexisting infrastructure. Figure 4.9
shows the unit costs of expanding export capacity together with the exporting harbor
capacity in the base year.
500
35
30
400
25
350
300
20
250
15
200
150
10
100
Investment Cost in USD/t
Export Capacity 2006 in mtpa
450
5
50
hi
na
Au
st
ra
lia
(N
ew
C
U
So
SA
ut
h
W
al
es
)
C
ol
o
m
Au
So
bi
st
a
ut
ra
h
lia
A
(Q
fri
ca
ue
en
sl
an
M
d
oz
am )
bi
R
qu
us
e
si
a
(W
es
t)
Po
R
la
us
nd
si
a
(E
as
t)
Bl
ac
k
Se
Ve
a
ne
zu
el
a
C
an
ad
a
0
In
do
ne
si
a
0
export capacity
investment costs
Source: own work based on IEA (2008b) and Rademacher (2008)
Figure 4.9: Capacity and investment costs for all export nodes in the base year
Another important parameter for a multi-period model is the discount rate that is
applied to the profit functions of the producers and exporters. We use the costs of
capital to determine the discount rate. The database of A. Damodaran at the New York
University’s Stern Business School provides estimates of the costs of capital. For the
coal industry, using data from 18 U.S. coal companies including major ones like Peabody,
Massey Energy or Arch coal, he reports an average cost of capital of 10.3%.53 For the
model, a discount rate of 10% is used for both producers and exporters.
We assume that there are restrictions on production and export capacity expansions
for various reasons that can be geological, technical and economical (financial restrictions,
lack of qualified labor force or equipment). These restrictions are based on historical
capacity data provided by the USGS in the country reports of the Mineral Yearbook54
on historical production and export capacity data as well as export capacity expansion
plans. The data for these restrictions can be found in Table 4.6 in Appendix 4.A.3.
53
Website http://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/wacc.htm, accessed
on January 27, 2010.
54
http://minerals.usgs.gov/minerals/pubs/country/
105
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
4.5
4.5.1
Results
General assumptions and base year results
For each model year, the COALMOD-World model delivers results for the inland and
seaborne trade flows, the prices, the level of investments and the value of the dual variables of the constraints that indicate if the constraint is binding and how strongly. The
results of the last two model years 2035 and 2040 are not presented as there is a risk
of distortion because there is less incentive to invest without any possible revenue after
2040. For convenience, we only present the results for the years 2006, 2010, 2020 and
2030 here.
Our results are based on the assumption of competitive and liberalized markets.55
We also assume that the markets are fully integrated, that is, when a demand node can
be reached by different producers or can import coal from overseas, it can fully substitute
between the different sources. The base case results can be called “ideal” results, as they
tell us how future demand should be served optimally and in which countries investments
should take place.
The results for 2006 show a good similarity with the actual observed trade pattern.
The direction and relative amounts of the trade flows correspond to actual trade flows.56
This is an important achievement given that we not only simulate the trade flows shown
on the maps in Figures 4.10 to 4.13 and 4.17 and 4.18 but also simulate internal markets.57
For the purpose of model validation we also computed the Mean Absolute Percentage
Errors for the level of exports and imports from every country. The values are 19.4%
for the exports and 12.4% for the imports. Modeling the interaction between imports
and domestic supply is a difficult task since for demand nodes with these two sourcing
possibilities we assume total substitutablility. However, this may not always be the
case; for example some power plants may be specifically designed for domestic coal or,
conversely, some coastal power plants do not have the infrastructure to receive domestic
coal.58
55
We are aware that not all countries currently have fully liberalized domestic markets (e.g., India and
China). However, we assume that the markets’ structure or outcomes will move toward competitiveness
in the future.
56
The only notable exception being the lower levels of exports from Australia to Japan. Here we
suspect a strong bilateral relationship and long-term contracts to play a role. In the subsequent years
when the markets get tighter the Australian exports return to normal levels at production capacity.
57
COALMOD-World is an energy-based model that calculates trade flows in Petajoules. For better
representation, the results shown in Figures 4.10 to 4.13 and 4.17 and 4.18 are aggregated and expressed
in million tons (Mt). These values are calculated using the relevant quality factors. Detailed flow results
are reported in Appendix 4.A.4.
58
An optimal modeling exercise would require a database at a power plant level, which is difficult
to obtain, especially for countries like India or China. Nevertheless, the COALMOD-World model is
specific enough to identify major trends and dynamics on the world market and the interaction with
domestic markets.
106
4.5. Results
23
43
18
16
34
15
60
70
21
53
31
116
17
55
2006
in Million tons
Figure 4.10: Increasing demand scenario results 2006: seaborne trade flows (in Mt)
21
50
18
16
15
30
22
95
19
19
75
184
45
29
2010
in Million tons
Figure 4.11: Increasing demand scenario results 2010: seaborne trade flows (in Mt)
23
50
45
15
33
25
40
32
171
15
77
104
225
13
2020
in Million tons
Figure 4.12: Increasing demand scenario results 2020: seaborne trade flows (in Mt)
107
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
35
87
60
14
48
45
143
200
139
230
8
2030
in Million tons
Figure 4.13: Increasing demand scenario results 2030: seaborne trade flows (in Mt)
4.5.2
Increasing demand scenario
Figures 4.10 to 4.13 show the evolution of the global seaborne trade flows from 2006 to
2030 for the increasing demand scenario.
Australia and Indonesia remain key players in the Pacific market. Their exports
increase significantly and reach the high levels of 200 Mtpa for Australia in 2030 and
230 Mtpa for Indonesia. Indonesia has been the most dynamic player between 2000 and
2007, and now has greater exports than Australia. Our model confirms this trend for
2010 and forecasts that Indonesia consolidates its role as the leading steam coal exporter,
ahead of Australia. Low production costs and flexible, low cost investments are the main
reasons for this development.
The third most important exporter is South Africa with an export level that doubles
between 2006 and 2030. South Africa is also the producer with the most potential
since the export capacity investment restriction of 15 Mtpa over five years is constantly
binding, meaning that South Africa would be willing to export significantly more steam
coal. This is due to an increase in import demand in Asia and especially in India that
opens new markets for the good quality South African coal. We can see the emergence
of a third market (in addition to the traditional Atlantic and Pacific markets) that could
be called Indian Ocean market and South Africa would become the key player in this
market.
In the Atlantic market there are various players that supply Europe and the key
players vary over time. After South Africa in 2006 and Colombia from 2010 to 2020,
Russia become the most important supplier to Europe in 2030, exporting nearly 95 Mt.
The U.S. play a relatively small role on the Atlantic market as an importer of Colombian
coal from 2006 to 2020 and are self-sufficient in the remaining years.
Before 2015 China is a swing supplier on the world market with a high variability of
exports. China becomes a net importer in 2010 and completely ceases to export after
2015 due to the high internal demand. China’s exports amount to 60 Mt in 2006 and 26
Mt in 2015. The reason for this high variability is the interaction between the domestic
108
4.5. Results
supplies and the imports to Southern China that are multiplied by more than three to
reach a level of 274 Mt in 2030.
From a global perspective, the most significant result of our modeling exercise is the
shifting of trade flows toward the Asian/Pacific markets which occurs in two marked
steps. We start today with a global integrated market where South Africa and Colombia
are the are the main suppliers to the Atlantic market and Indonesia and Australia to
the Atlantic Market. Then, we notice a gradual shift eastwards until 2020 with flows
from South Africa being directed toward Asia and, especially, India. Colombia replaces
South Africa as the key supplier to Europe. The second step in the shift starts in 2020.
We expect an additional shift westwards with Colombia delivering to Japan and Korea,
resource poor countries with a high willingness to pay.59 By 2030, the overall picture on
the global market has significantly changed: Russia and Poland are the only suppliers
to Europe; South Africa, Europe’s traditional supplier, is a major supplier to India and
the Pacific market; and Colombia becomes, principally, a Pacific market supplier. Asian
market demand is very strong, especially China, which is importing significant quantities
from Russia and Mongolia, 60 and 45 Mt respectively. Figure 4.14 sets the trade results
in relation to the locally produced and consumed quantities of steam coal as well as to
the imports and local supply results for India and China. The total surface of this graph
represents the total consumption and the different areas differentiate the consumption
by its origin, seaborne trade or local supply with a special focus on India and China.
Unsurprisingly, China’s steam coal consumption represents the biggest share of 36% to
48% of the worldwide consumption in every model year. The volume of the international
seaborne trade increases by 86% from 2006 to 2030 and its share in total consumption
increases from 16% to 21%. Also, China and India account for most of this trade increase
as their seaborne imports are multiplied by 11.6 from 2006 to 2030. In 2030, the Chinese
and Indian imports represent close to half of the international trade. Seaborne imports
of other countries amount to 478 Mt in 2006 and increase to a level of 520 Mt in 2015 to
then gradually return to the level of 2006 in 2030.
The model also gives us the amount of investments in mining, transport or export
capacity necessary to serve the high demand predicted by the increasing scenario. The
cumulative amount of investments in mining capacity from 2006 to 2025 for the main
model producers is shown with the black bars in Figure 4.15. This figure also shows the
losses of mining capacity that resulted from the mine mortality mechanism described
in Section 4.3.2 represented by the gray bars. Thus, the difference between those two
bars represents the net capacity addition (or loss) during the time between 2006 and
2025. The most important net capacity additions occur in Northern China, Northern
India, Indonesia, the Powder River Basin in the U.S., and to a lesser extend in Mongolia
and Queensland, Australia. The net capacity losses are due to the fact that in certain
59
There are no extra costs in the data for using canals like the Panama canal. It is not clear if such an
inter-basin trade flow would prevail with the incorporation of this cost component. However, the current
expansion of the Panama canal that will be completed in 2015 is thought to facilitate Colombian exports
to Asia
109
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
5000
overland trade
international seaborne
trade w/o China and
India imports
India seaborne imports
Million tons
4000
3000
China seaborne
imports
2000
India domestic supply
China domestic supply
1000
local consumption w/o
China and India
0
2006
2010
2015
2020
2025
2030
In mtpa from 2006 to 2025
Figure 4.14: Increasing demand scenario: aggregated consumption and imports (in Mt)
600
600
500
500
400
400
300
300
200
200
100
100
Total investments
P_AUS_NSW
P_COL
P_ZAF
P_AUS_QLD
P_RUS
P_IND_Orissa
P_MNG
P_CHN_YG
P_CHN_Northeast
P_USA_PRB
P_IDN
P_IND_North
0
P_CHN_SIS
0
Capacity losses
Figure 4.15: Investments in additional mining capacity and capacity losses of producers
between 2006 and 2025 (in Mtpa)
producing regions (more than represented in Figure 4.15) the reserves are getting closer
to exhaustion and extraction costs increase. In South Africa the loss of capacity is due
to the fact that in the model the South African coal producer can deliver coal to both
export and domestic market without any restriction or domestic obligation. This results
in a diversion of local supplies to the export market penalizing and reducing the supplies
to the domestic market (see Figure 4.16 for the price effect). A domestic obligation or
the dedication of production capacities to the domestic market could be implemented
to remedy this but would not affect the export and global results of our model as the
investments in production capacity would then be higher.
The increasing demand scenario represents the worst case for climate policy but also
for the coal market and the coal consuming countries. This is due to the potential
restrictions that may affect an expansion of the production and export capacities and
that we have tried to implement in the model. A few example for the rationale of these
110
4.5. Results
restriction is provided in the next paragraph.
In the U.S., production could be threatened by environmental regulation, expressed by
a substantial increase in production costs and probably a reduction in available reserves
in the Appalachia region because of a possible ban of the mountain top removal mining
technique.60 The Chinese coal industry is in a process of a difficult restructuring. The
small, dangerous and often illegal township and village enterprise (TVE) mines must
be closed to make room for more efficient larger firms (Minchener, 2007). As of 2009
TVEs still accounted for 38% of national coal output (Tu, 2011) but after the state
driven market restructuration is complete other factors could limit Chinese production
expansion and increase the need for coal imports. Investments cost are growing and will
continue to grow as larger firms require more upfront capital investments and there will
be a move of the production towards less attractive deposits that are deeper or further
away from the coastal demand centers (Tu, 2011; Rui et al., 2010). In India the reform
process from state run enterprises to efficient firms is even more cumbersome (Carl et al.,
2008). The increase in Indian production capacity might also be limited by political and
technical factors.
In the model a restriction that is binding has a positive dual variable. In the increasing demand scenario we see a lot of binding restrictions and positive dual variables for
the capacity constraints on production and exports but also on the expansion of those
capacities. This is especially true for India and China where more restrictive limitations
were imposed but also for the exporters that try to satisfy this growing demand such as
Indonesia, Colombia, Australia or South Africa. The effect of these restrictions can be
seen in the prices that are discussed in the next paragraph.
Figure 4.16: Increasing demand scenario: computed average prices representing the marginal
costs of supply of selected regions for all model years (in 2006 USD/t)
Figure 4.16 shows the price development for some major regions. It is important to
60
Mountaintop removal mining (MTR), sometimes referred to as mountaintop mining (MTM), is a
form of surface mining that involves the mining of the summit or summit ridge of a mountain. The
process involves blasting with explosives to remove up to 300 m of mountain to expose underlying coal
seams.
111
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
note that these prices represent marginal costs of supply computed by a model based
on the assumption of perfect competition. Other factors that may affect the prices such
as long-term contracts, short-term disruptions, local market power and other market
distortions are not included here. Globally, the computed prices show an upward trend
between 2006 and 2030. The lowest prices with the lowest increase over time are the
domestic prices in South Africa, the U.S. and Russia. These demand nodes are close to
large and cheap sources of supply. There is some effect of the global market such as in
South Africa, but it is limited. We can also see that for the Russia-Central and RussiaSiberia regions domestic prices increase after 2020. This is due to the increased exports
to Europe in these years. The highest rise in prices occur in the Indian and Chinese
prices. The Indian import prices as well as the Chinese domestic and import prices all
converge to the same level as that of European and other Asian importers. This is due
to the high rise in imports but also domestically because the very high production makes
it necessary to use more expensive reserves.
4.5.3
Stabilizing demand scenario
28
50
25
15
15
30
35
8
151
30
48
114
226
10
2020
in Million tons
Figure 4.17: Stabilizing demand scenario results 2020: seaborne trade flows (in Mt)
The results of the stabilizing demand scenario paint a less drastic future picture of
the global steam coal market as in the increasing demand scenario. In Figures 4.17 and
4.18 we see that the regional flow shifts are similar that in the Current Polices scenario
but they are less extreme and occur later in time. This can be seen for example in the
fact that Colombia still supplies Europe with 23 Mt in 2030 and has not diverted all
its supplies to Asia, as in the previous scenario. From a global perspective, we see in
Figure 4.19 that global consumption remains constant after 2015. Interestingly, global
trade remains important and rises from a share of 16% to 18% due to the fact that India
and China rely significantly on imports to satisfy their coal demand. Imports from these
two countries is multiplied by 9.5 whereas imports from other countries decrease by 30%
thus making the share of China and India in the global trade even higher than in the
increasing demand scenario after 2015 to reach 56% in 2030. India’s domestic production
112
4.5. Results
28
50
25
13
15
30
45
63
179
23
145
230
17
2030
in Million tons
Figure 4.18: Stabilizing demand scenario results 2030: seaborne trade flows (in Mt)
5000
overland trade
international seaborne
trade w/o China and
India imports
India seaborne imports
Million tons
4000
3000
China seaborne
imports
2000
India domestic supply
China domestic supply
1000
local consumption w/o
China and India
0
2006
2010
2015
2020
2025
2030
Figure 4.19: Stabilizing demand scenario: aggregated consumption and imports (in Mt)
increase constantly from 2006 to 2030 whereas in China domestic production reaches a
peak in 2020 an decrease after to be replaced by cheaper imports this again highlights
the importance of global market.
Prices also add to the picture of a less strained global coal market in the stabilizing
demand scenario. We still observe a very steep rise in domestic Indian prices but only to
reach a level of 78 USD/t instead of 117 USD/t in the previous scenario. Also the global
price level for imports is around 60 USD/t in 2030 more than 10 USD/t lower than in
the increasing demand scenario for the same year.
Another output of the model that can be analyzed is the evolution of the supply costs
influenced by production and investments, as described in Section 4.3.2. Figure 4.20
shows the resulting FOB supply costs in 2020. If we compare with the starting values
in Figure 4.6 we can see some significant increases in production costs in Queensland,
Australia, Russia, South Africa , Indonesia and China.
4.5.4
Results comparison with Hubbert-method based models
In the introduction we discussed three papers using methods derived from the Hubbert
model. Figure 4.21 provides a comparison of our modeling results (black curves) with the
113
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
90
80
70
USD/t
60
50
40
30
20
10
ue
bi
a
bi
q
am
M
oz
hi
na
ol
om
C
W
C
st
R
us
si
a
es
tR
us
si
a
Bl
ac
k
Se
So
a
ut
h
Af
r ic
a
Ve
ne
zu
el
a
In
do
ne
si
a
)
R
B)
P
om
(fr
Ea
LD
W
)
(Q
C
an
a
da
(N
S
lia
lia
st
ra
nd
st
ra
Au
Au
U
la
Po
SA
0
low cost in USD/t
diff=(high - low) in USD/t
transport cost in USD/t
port fee in USD/t
Figure 4.20: 2020 FOB costs for all export countries calculated endogenously in the
stabilizing demand scenario
Figure 4.21: Comparison of production projections in COALMOD-World and Mohr and
Evans (2009) for bituminous coal (in Gt per year)
projections of Mohr and Evans (2009) for bituminous coal (gray curves), a coal quality
that covers most of the steam coal used in COALMOD-World. Patzek and Croft (2010)
and Höök et al. (2010) only provide results for all coal types which makes a comparison
with our results more difficult. In Figure 4.21 the plain gray Hubbert Linearisation curve
illustrates the issues associated with the use of the standard Hubbert curve method based
only on historical production data: we see an overestimation of production around the
peak period that exceeds any demand projections and then a sharp decline. The dotted
and dashed gray line shows the results of a more refined model that takes into account
supply and demand as well as reserves and cumulated production for the dotted line
whereas the dashed line represent the authors best guess. These projections do not show
such a sharp decline of production as in the Hubbert linearisation case but production
still peaks before 2020 and declines to levels lower than our projections. The results of
Mohr and Evans (2009) are driven by historical production data, which we have shown
to be problematic in the introduction of this chapter, and to a certain extend reserve
114
4.6. Conclusions
estimations. Reserve estimates of coal are also problematic as discussed by Höök et al.
(2010) as they are not done as thoroughly and frequently as for other resources and
assessment methods and definitions vary between countries. This causes upward and
downward corrections and jumps in the estimates that should however not be interpreted
as signs of increasing scarcity. In fact we do not see any reason why the market should
be affected by a reserve constraint in the mid-term until 2030. The level of worldwide
reserves has been constant at high levels for the last 25 years.61 We also have to keep in
mind that the definition of reserves is dynamic and that coal is underexplored in several
world regions which increases the likelihood of substantial reserve additions in the next
decades (see Minchener, 2009).
4.5.5
Model evaluation and criticism
The “COALMOD-World” model uses 1671 single data inputs and is expressed by 6599
single equations that calculate values for 6599 single variables. It is programmed in
GAMS using the mixed complementarity (MCP) format and the solver PATH (Ferris
and Munson, 2000). The model solves in less than 10 seconds using a standard desktop
computer (Pentium®Dual-core with 2.50 GHz CPU and 2.91 GB RAM). This rapidity
allows for a very flexible use of the model for fast data update and test or for scenario or
sensitivity analysis. The main weakness is the important amount of data input needed
and the difficulty to find the appropriate data. Better access to experts and proprietary
data could remedy these data issues.
4.6
Conclusions
In this chapter, we present a tool for analysis of the future global steam coal market,
the “COALMOD-World” model. From a starting point in 2006, we are able to give
insights into how production and trade flows will develop until 2030, using two different
scenarios for demand projection: one with a continuously increasing demand and one with
a stabilizing demand. We are able to give a differentiated answer to the question about
“the end of cheap coal?”. Based on our model analysis, we find that it is not geological
reserve depletion but capacity constraints as well as slow expansions of production and
export capacities that could make coal more expensive in the future.
The resulting scarcity that starts appearing after 2015 can be measured in the model
as a demand gap expressed by the percentage of the reference demand that is not satisfied.
In the increasing demand scenario the gap represents 6.2% of the demand in 2025 and
9.3% in 2030. In the stabilizing demand scenario these values are 3% and 4.7% for 2025
and 2030 respectively. We also calculated that in the increasing demand scenario 18.6%
of the assumed reserves are depleted until 2030 and only 16.3% of the reserves in the
stabilizing demand scenario.62
61
62
Source: data compiled by the World Energy Council: http://www.worldenergy.org/
In both scenarios three mining basins are completely depleted: Vietnam, Mongolia and the Chinese
115
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
This makes “peak coal” a less imminent question than the issue of the investments
in production and export capacity needed to satisfy future demand. In the increasing
demand scenario we calculate that through 2025, 124.6 billion USD in production capacity investments are needed and, due to the high amount of imports, 6.6 billion USD
of investments are needed in export port capacities. In the stabilizing demand scenario
84.8 billion USD in production capacity investements are required and only 2.4 billion
USD for port capacity expansions.
In the mid-term the more pressing issue concerning the use of coal are the environmental externalities, especially the carbon dioxide emissions affecting climate change.
Coal is considered by many to be the number one climate enemy. If more restrictions
on carbon dioxide emissions or a higher carbon price were imposed, there would be a
direct effect on electricity generation from coal. However, in the mid-term this is more
likely to happen in the main importing developed countries like the EU or Japan than
in India or China where most of the future increase in demand and imports will come
from. Therefore in the following Chapter 5 we use the COALMOD-World model to see
the interactions between different climate policies and the global steam coal market.
Henan province. Additionally, Venezuela hits its reserve constraint in the increasing demand scenario.
116
4.A. Appendix
4.A
4.A.1
Appendix
Mathematical Formulation of the Model
The profit maximization problem described in Sections 4.3.2 to 4.3.4 has the following
Karush-Kuhn-Tucker conditions (KKTs) of optimality that are obtained after deriving
the Lagrangian function of each player type with respect to their decision variables and
dual variables of constraints.
• Producers KKTs:
0≤
a h
· − pac
1
1 + rf
P
∂Caf
+
+ trans_caf c · κaf
∂xaf c
P
· κaf +
+ αaf
0≤
P
∂Caf
∂yaf e
+ trans_eaf e · κaf
P
· κaf +
+ αaf
0≤
0≤
0≤
1
1 + rf
1
1 + rf
a
1
1 + rf
a
·
5 Res
T cap_c
· κaf ⊥ xaf c ≥ 0
· κaf + αaf c
·α
2 f
(4.25)
a h
· − pae
1
1 + rf
+
i
a
·
i
5 Res
T cap_e
· κaf ⊥ yaf e ≥ 0
· αf · κaf + αaf e
2
(4.26)
X
CP invaf
P
P inv
−
αaf
+ αaf
⊥ P invaf ≥ 0
5
0
(4.27)
a >a
CT inv_caf c X T cap_c
T inv_c
−
αaf c
+ αaf c
⊥ T inv_caf c ≥ 0 (4.28)
5
0
a >a
·
CT inv_eaf e X T cap_e
T inv_e
−
αaf e
+ αaf e
⊥ T inv_eaf e ≥ 0 (4.29)
5
0
a >a
Xh X
X
0 ≤P capf −
xaf c · κaf +
yaf e · κaf
a0 <a
c
!
· mc_int_varf
i
e
!
+
X
a0 <a
P invaf −
X
xaf c · κaf
c
X
+
yaf e · κaf
e
117
P
⊥ αaf
≥0
(4.30)
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
(4.31)
P inv
0 ≤P maxinvaf − P invaf ⊥ αaf
≥0
0 ≤Resf −
X hX
xaf c · κaf +
c
a∈A
X
yaf e · κaf
e
5i
X
X
+
x(a−1)f c · κ(a−1)f +
y(a−1)f e · κ(a−1)f ∗
⊥ αfRes ≥ 0
2
c
e
0 ≤T cap_cf c +
X
(4.32)
T cap_c
≥0
(4.33)
T cap_e
≥0
(4.34)
T inv_caf c − xaf c · κaf ⊥ αaf c
a0 <a
0 ≤T cap_ef e +
X
T inv_eef c − yaf e · κaf ⊥ αaf e
a0 <a
T inv_c
≥0
(4.35)
T inv_e
≥0
(4.36)
0 ≤T maxinv_caf c − T inv_caf c ⊥ αaf c
0 ≤T maxinv_eaf e − T inv_eaf e ⊥ αaf e
0 ≤mc_intaf = mc_int(a−1)f
!
X
+ mc_slp(a−1)f ·
x(a−1)f c · κ(a−1)f
X
+
y(a−1)f e · κ(a−1)f
c
· mc_int_varf
e
, mc_intaf (free)
(4.37)
0 ≤mc_slpaf = mc_slp_startf + mc_slp_varf ·
X
P inva0 f ,
a0 <a
mc_slpaf (free)
(4.38)
• Exporters KKTs:
0≤
1
1 + re
a h
· − pac
+ pae + Cportae · κae + searateaec · κae
i
+ µE
ae · κae ⊥ zaec ≥ 0
0≤
1
1 + re
a
·
CEinvae X E
−
µae + µEinv
+ µEmax
⊥ Einvae ≥ 0
ae
e
5
0
a >a
118
(4.39)
(4.40)
4.A. Appendix
0 ≤ Ecape +
X
Einvae −
X
zaec · κae ⊥ µE
ae ≥ 0
a0 <a
(4.41)
c
0 ≤ Emaxinvae − Einvae ⊥ µEinv
≥0
ae
0 ≤ maxcape − Ecape −
X
Einvae ⊥ µEmax
≥0
e
(4.42)
(4.43)
a
• Producers Quality Factor:
!
κaf = κf + δf ·
X
X
X
xaf c +
yaf e
a0 ≤a
c
, κaf (free)
(4.44)
e
• Final Demand Equation:


X
X
pac − pac  xaf c ,
zaec  = 0 , pac (free)
(4.45)
e
f
• Market Clearing Condition:
0 = yaf e −
X
zaec
, pae (free)
(4.46)
X
zaec · κae ⊥ πa ECHN ≥ 0
(4.47)
c
• Chinese Export Restriction:
0 ≤ China_lica ECHN −
N oChina(c)
The KKT (Karush Kuhn Tucker) optimality conditions of each model player and
the additional final demand, market clearing and quality equations form a mathematical
equilibrium problem in the MCP format. This model is programmed in GAMS and it is
solved using the PATH solver (Ferris and Munson, 2000).
4.A.2
Nodes of COALMOD-World
119
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
Producers
P_USA_PRB
P_USA_APP
P_USA_Rocky
P_USA_ILL
P_COL
P_VEN
P_POL
P_RUS
Kemerovo/Kuznets
Chhattisgarh, Jharkhand, Madhya
Pradesh, Uttar Pradesh, West Bengal
Orissa
Maharashtra
Andhra Pradesh
Ukrainian/Russian Donets
Kazakhstan/Ekibastuz
P_IND_Orissa
P_IND_West
P_IND_South
Shanxi, Shaanxi, Inner Mongolia , Hebei
Liaoning, Jilin , Heilongjiang
Henan, Shandong, Jiangxi, Fujian, Jiangsu
Guizhou, Hunan, Chongqing , Sichuan
P_AUS_QLD
P_AUS_NSW
P_MNG
P_VNM
P_IDN
P_CHN_SIS
P_CHN_Northeast
P_CHN_HSA
P_CHN_YG
P_UKR
P_KAZ
P_ZAF
P_MOZ
P_IND_North
Powder River Basin
Appalachian
Rocky Mountains
Illinois Basin
Table 4.4: Nodes of the COALMOD-World Model
Country
Canada
U.S.
Colombia
Venezuela
Morocco
Portugal
Spain
UK
NL_F_BEL
Germany
Denmark
Finland
Italy
Poland
Turkey
Israel
Eurasia
South Africa
Mozambique
India
Thailand
Malaysia
Vietnam
Indonesia
China
Mongolia
Korea
Japan
Taiwan
Philippines
Australia
Exporters
E_CAN
E_USA_East
Vancouver
Hampton Roads
Puerto Bolivar
Maracaibo
Gdansk
E_COL
E_VEN
E_POL
Baltic/Riga
Vostochny
Mariupol
Dalrymple Bay
Newcastle
Qinhuangdao
Campha
Richards Bay
Maputo
E_RUS_West
E_RUS_East
E_Black_Sea
E_ZAF
E_MOZ
E_VNM
E_IDN
E_CHN
E_AUS_QLD
E_AUS_NSW
Consumers
C_CAN
C_USA_Rocky
C_USA_East
C_USA_Central
C_USA_South
C_USA_Gulf
Ontario
Port
No
No
Boston
No
No
Mobile
Mohammedia
Sines
Gijon
Immingham
Rotterdam
Rotterdam
Aabenraa
Kotka
Taranto
No
Mersin/Samsun
Ashdod
No
No
Netherlands, France, Belgium
C_MAR
C_PRT
C_ESP
C_GBR
C_NFB
C_DEU
C_DNK
C_FIN
C_ITA
C_POL
C_TUR
C_ISR
C_RUS_Sibiria
C_RUS_Central
No
Ulsan
Yokohama
Kaohsiung
Pagbilao
Guangzhou
No
Mundra
Chennai
Bangkok
Lumut
No
No
No
No
No
Shanghai/Ningbo
No
No
No
No
Shanxi, Shaaxi, Inner Mongolia
Heilongjiang, Jilin, Liaoning
Beijing, Tianjin, Hebei, Henan, Shandong
Jiangsu, Hubei, Chongqing, Shanghai,
Zhejiang
Jiangxi, Guizhou, Sichuan, Guangdong,
Fujian, Guangxi and Hunan
Bihar, Jharkhand, West Bengal, Orissa,
Chhattisgarh
Delhi, Punjab, Rajasthan, Uttar Pradesh
Gujarat, Maharashtra, Madhya Pradesh
Andhra Pradesh, Tamil Nadu, Karnataka
C_UKR
C_KAZ
C_ZAF
C_IND_East
C_IND_North
C_IND_West
C_IND_South
C_THA
C_MYS
C_VNM
C_IDN
C_CHN_SIS
C_CHN_Northeast
C_CHN_Main
C_CHN_Eastern
C_CHN_South
C_MNG
C_KOR
C_JPN
C_TWN
C_PHL
120
4.A. Appendix
4.A.3
Data of COALMOD-World
Table 4.5: Data and assumptions for the endogenous cost mechanism
Country
U.S.
Colombia
Venezuela
Poland
Ukraine
Kazakhstan
Russia
South Africa
India
Vietnam
Indonesia
China
Australia
Mongolia
Mozambique
Model Producers
P_USA_PRB
P_USA_Rocky
P_USA_ILL
P_USA_APP
P_COL
P_VEN
P_POL
P_UKR
P_KAZ
P_RUS
P_ZAF
P_IND_North
P_IND_Orissa
P_IND_West
P_IND_South
P_VNM
P_IDN
P_CHN_SIS
P_CHN_Northeast
P_CHN_HSA
P_CHN_YG
P_AUS_QLD
P_AUS_NSW
P_MNG
P_MOZ
Mining
Basin Type
2
2
3
3
2
1
3
3
2
2
3
2
3
3
3
4
2
2
3
3
3
2
2
1
1
Intercept
Increase
slow
moderate
moderate
high
slow
high
slow
moderate
moderate
slow
moderate
moderate
high
moderate
moderate
high
slow
moderate
moderate
moderate
moderate
slow
slow
high
high
mc_slp_var_f
−1 · 10−6
−1 · 10−6
0
0
−1 · 10−5
1 · 10−2
0
0
−1 · 10−4
−2 · 10−5
0
0
0
0
0
1 · 10−2
−5 · 10−6
−2 · 10−8
0
0
0
−2 · 10−3
−2 · 10−3
1 · 10−3
1 · 10−2
mc_int_varf
0.02
0.04
0.04
0.06
0.05
0.2
0.05
0.2
0.1
0.05
0.1
0.07
0.25
0.15
0.15
0.3
0.05
0.08
0.1
0.18
0.14
0.05
0.05
0.2
0.4
Table 4.6: Data assumptions for the per 5-years capacity expansion limitations in Mtpa
U.S.
Colombia
Venezuela
Poland
Ukraine
Russia
South Africa
India
Indonesia
China
Australia
Production capacity limitation
276
22
10
14
7
51
47
63
51
292
44
121
Export capacity limitation
20
20
10
5
10
40
15
10
10
30
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
4.A.4
Results of COALMOD-World
Table 4.7: Results of COALMOD-World: Domestic trade flows in Mtpa for the increasing
demand and the stabilizing demand scenarios
Scenario
From
To
P_USA_PRB
C_USA_Rocky
P_USA_PRB
C_USA_Central
P_USA_PRB
C_USA_Gulf
P_USA_Rocky
C_USA_Rocky
P_USA_ILL
C_USA_South
P_USA_ILL
C_USA_Gulf
P_USA_APP
C_USA_South
P_USA_APP
C_USA_East
P_USA_APP
C_USA_Gulf
P_POL
C_POL
P_UKR
C_UKR
P_KAZ
C_KAZ
P_RUS
C_RUS_Siberia
P_RUS
C_RUS_Central
P_ZAF
C_ZAF
P_IND_North
C_IND_East
P_IND_North
C_IND_North
P_IND_Orissa
C_IND_East
P_IND_Orissa
C_IND_South
P_IND_West
C_IND_West
P_IND_South
C_IND_South
P_VNM
C_VNM
P_IDN
C_IDN
P_CHN_SIS
C_CHN_Northeast
P_CHN_SIS
C_CHN_Main
P_CHN_SIS
C_CHN_Eastern
P_CHN_SIS
C_CHN_SIS
P_CHN_Northeast
C_CHN_Northeast
P_CHN_HSA
C_CHN_Eastern
P_CHN_YG
C_CHN_South
P_MNG
C_MNG
2006
63
358
70
40
43
70
62
58
79
67
32
42
70
23
91
110
142
0
60
60
21
3
23
339
110
228
148
100
248
0.15
Increasing
2010
2015
104
104
359
373
166
166
9
12
23
19
85
85
78
85
57
59
65
34
43
96
26
102
115
157
9
26
58
37
2
29
59
31
40
94
24
102
116
174
26
13
57
18
1
36
429
131
287
190
157
297
0.13
518
200
360
233
157
273
0.13
122
demand
2020
2025
104
104
367
362
167
167
10
10
10
6
90
90
92
94
58
57
0
13
55
51
28
31
38
41
88
92
25
12
114
126
97
110
208
208
84
41
2030
98
356
167
14
3
90
96
57
11
47
30
43
96
10
136
97
234
86
Stabilizing demand
2020
2030
104
102
348
312
163
178
7
30
32
70
62
70
55
56
48
43
27
35
79
16
97
107
198
60
32
25
34
78
16
92
116
214
55
56
38
55
29
54
36
40
27
33
30
42
13
553
200
386
232
193
267
0.20
48
24
593
200
405
240
193
257
0.25
53
35
626
200
415
244
193
252
0.28
38
4
492
111
349
222
252
267
0.20
41
505
250
351
235
247
0.34
4.A. Appendix
Table 4.8: Results of COALMOD-World: International trade flows in Mtpa for the increasing
demand and the stabilizing demand scenarios (part 1/2)
Scenario
From
To
P_USA_APP
C_CAN
P_COL
C_USA_Gulf
P_COL
C_MAR
P_COL
C_PRT
P_COL
C_ESP
P_COL
C_GBR
P_COL
C_DEU
P_COL
C_ITA
P_COL
C_TUR
P_COL
C_CHN_Eastern
P_COL
C_KOR
P_COL
C_JPN
P_COL
C_TWN
P_VEN
C_ESP
P_VEN
C_NFB
P_VEN
C_ITA
P_VEN
C_TUR
P_VEN
C_ISR
P_VEN
C_IND_West
P_VEN
C_CHN_Eastern
P_POL
C_GBR
P_POL
C_NFB
P_POL
C_DNK
P_KAZ
C_RUS_Siberia
P_KAZ
C_RUS_Central
P_RUS
C_MAR
P_RUS
C_PRT
P_RUS
C_ESP
P_RUS
C_GBR
P_RUS
C_NFB
P_RUS
C_DEU
P_RUS
C_DNK
P_RUS
C_FIN
P_RUS
C_ITA
P_RUS
C_TUR
P_RUS
C_ISR
P_RUS
C_UKR
P_RUS
C_IND_West
P_RUS
C_CHN_Eastern
P_RUS
C_KOR
P_RUS
C_JPN
P_ZAF
C_MAR
P_ZAF
C_PRT
P_ZAF
C_ESP
P_ZAF
C_GBR
2006
16
21
21
24
Increasing demand
2010
2015
2020
2025
16
17
16
16
19
27
15
1
5
6
7
5
5
4
4
7
17
15
14
29
24
20
2
23
17
12
6
12
5
32
11
87
2030
16
54
79
10
Stabilizing demand
2020
2030
15
13
31
5
5
3
12
3
12
3
2
6
8
10
54
9
1
3
7
28
24
10
9
10
4
4
8
8
23
21
22
23
24
14
2
35
4
5
30
30
30
25
43
14
11
8
10
22
10
8
10
21
13
7
9
19
16
7
8
13
17
26
6
7
15
15
17
5
12
21
16
27
25
45
12
48
4
15
5
5
19
7
18
4
123
48
7
3
12
0
16
23
6
7
13
14
33
1
5
60
30
30
2
9
2
15
22
5
6
11
16
0
5
7
15
15
25
25
Chapter 4. A Techno-economic Analysis using the COALMOD-World Model
Table 4.9: Results of COALMOD-World: International trade flows in Mtpa for the increasing
demand and the stabilizing demand scenarios (part 2/2)
Scenario
From
To
P_ZAF
C_ITA
P_ZAF
C_TUR
P_ZAF
C_ISR
P_ZAF
C_IND_West
P_ZAF
C_IND_South
P_ZAF
C_THA
P_ZAF
C_MYS
P_VNM
C_CHN_South
P_IDN
C_IND_West
P_IDN
C_IND_South
P_IDN
C_THA
P_IDN
C_MYS
P_IDN
C_CHN_Eastern
P_IDN
C_CHN_South
P_IDN
C_KOR
P_IDN
C_JPN
P_IDN
C_TWN
P_IDN
C_PHL
P_CHN_SIS
C_KOR
P_AUS_QLD
C_CHN_Eastern
P_AUS_QLD
C_CHN_South
P_AUS_QLD
C_KOR
P_AUS_QLD
C_JPN
P_AUS_QLD
C_TWN
P_AUS_NSW
C_CHN_Eastern
P_AUS_NSW
C_CHN_South
P_AUS_NSW
C_KOR
P_AUS_NSW
C_JPN
P_AUS_NSW
C_TWN
P_AUS_NSW
C_PHL
P_MNG
C_RUS_Siberia
P_MNG
C_CHN_Main
P_MNG
C_CHN_SIS
P_MOZ
C_MAR
P_MOZ
C_PRT
P_MOZ
C_ESP
P_MOZ
C_GBR
P_MOZ
C_ITA
P_MOZ
C_TUR
P_MOZ
C_ISR
P_MOZ
C_IND_West
P_MOZ
C_IND_South
P_MOZ
C_THA
P_MOZ
C_MYS
2006
19
4
13
20
16
15
7
13
11
20
58
8
60
Increasing demand
2010
2015
2020
2025
19
4
16
11
21
28
48
56
24
47
56
76
2030
Stabilizing demand
2020
2030
58
81
51
57
64
67
8
9
17
16
13
8
16
10
11
3
12
24
13
11
10
8
11
22
50
7
73
9
154
202
193
206
156
199
40
12
26
8
41
12
6
13
20
9
0
3
57
13
60
31
42
12
73
35
86
97
20
57
100
30
44
81
85
86
17
63
1
4
30
47
2
13
14
22
5
30
5
33
10
33
15
35
5
36
9
3
0.03
7
8
13
1
6
6
8
5
1
2
2
5
5
0.11
1
0.05
0
0
3
124
Chapter 5
Climate Policies and the Global
Steam Coal Market: Interactions
until 2030
5.1
Introduction
This chapter presents an analysis of the different possible interactions between climate
policies and the global steam coal market.63 Worldwide steam coal consumption is continuously growing, drawn by economic growth and increasing electricity demand in Asia.
The use of coal is also a major contributor to global carbon dioxide emissions that have
a potentially high adverse effect on the future climate. Hence, the relationship between
the coal market and climate policy is of prime importance for the effectiveness of these
policies and needs to be investigated further.
In the set of possible interactions we show that a shift of production towards the
present due to strategic behavior of coal resource owners as a reaction to future climate
policies because they would be able to sell less coal in the future is not to be expected.
We therefore concentrate our analysis on pure supply and demand market effects using
the COALMOD-World model described in Chapter 4. This multi-period model of the
global steam coal market calculates yearly future market equilibria. We implement three
different climate policy shocks in different climate policy and market environments already in place. The scenarios are: a unilateral European climate policy, an Indonesian
export-limiting supply-side policy and a carbon capture and storage (CCS) fast roll-out
policy.
Our main findings are that coal market adjustments effects can have a potentially
adverse effect on climate policy effectiveness in the case of an unilateral European climate
policy. However, this effect never overcompensates saved emissions; even a unilateral
European climate effort will always be beneficial to the global climate. In the case of the
63
This chapter is an updated and modified version of Haftendorn et al. (2011) and a modified version
of the forthcoming article Haftendorn et al. (2012b).
125
Chapter 5. Climate Policies and the Global Steam Coal Market
Indonesian supply-side scenario and the CCS fast-roll out policy we see positive market
adjustment effects that speak for the effectiveness of these less conventional measures.
After having evaluated the different modeling results, we draw conclusions and policy
recommendations for the climate policy of the European Union (EU).
5.2
5.2.1
Assessment of Positive Modeling Approaches
Overview of possible modeling approaches
There are various modeling approaches at our disposition to help us understand what
possible effects we can expect from the interaction between climate policies and the global
steam coal market. In essence, all these models are positive, meaning that they are
“constructed with the objective of better explaining (or reproducing) observed resource
prices and production levels”(Pindyck, 1981). In this section we determine if these models
can be applied to the global steam coal market and if they may be suited to answer our
research questions.
One type of modeling approach to model future production patters of resource markets already discussed in Chapter 4 is based on the concept of the Hubbert curve first
described by M. King Hubbert (1959). The core mathematical assumption of this model
is that cumulative resource production follows a logistic growth path that derived with
respect to time yields the well-known symmetrical bell shaped curve of yearly production
output; the summit of the curve representing the “peak” of the production rate. It is
mathematically possible to estimate the shape of the curve and, thus, the peak year as
well as the ultimately recoverable reserves, defined as the surface under the curve, based
solely on historical production data. This simple technique is subject to controversy.
Its proponents claim that “the Hubbert curves are based on [...] production and not on
ill-defined and subjective [...] ’reserves” ’ and that “historical production trends reflect the
prevailing economics prior to the time of production” (Patzek and Croft, 2010). However,
its opponents, such as Lynch (2003), state that the “work of the Hubbert modelers has
proven to be incorrect in theory, and based heavily on assumptions that the available
evidence shows to be wrong. They have repeatedly misinterpreted political and economic
effects as reflecting geological constraints, and misunderstood the causality underlying
exploration, discovery and production”. The main problem for our analysis of climate
policy is that past production can only represent past economical and political situations.
The Hubbert model fails to integrate paradigm shifts that will affect future production
patterns such as the carbon constraints of climate policy or the high economic growth in
Asia. It is thus not suitable for our purpose.
Computable General Equilibrium models (CGE) have been widely used for climate
policy analyses. For example, Böhringer and Rutherford (2000) analyze the economic
implications of the Kyoto Protocol at the cross-country level, while Böhringer et al.
(2010) assess the effects of regional climate policies on industrial sectors’ competitiveness
and the scope for carbon leakage, i.e., the increase in carbon emissions outside emissions126
5.2. Assessment of Positive Modeling Approaches
regulating regions, e.g., due to relocation. Bovenberg and Goulder (2005) place greater
focus on industries affected by carbon-emissions regulation. Kemfert (2002) assesses
potential leakage effects and compensation options by issue linkage. However, detailed
industry, sector or market-level investigations require modifications to traditional CGE
models which generally exhibit only coarse representations of the energy sector. One
possible modification is to integrate top-down CGE models with bottom-up engineering
models (e.g., Böhringer, 1998); another one is to utilize detailed partial equilibrium
models that focus specifically on a single market or sector (e.g., Demailly and Quirion,
2008). However, since the focus and advantage of CGE models is the multi-sectoral
analysis and carbon leakage through relocation effects, they would not be the best fit for
the aim of our present study that focuses solely on interactions between climate policy
and the steam coal market.
The classical model for the theory of exhaustible resource extraction was developed
by Hotelling (1931). The model is built on the assumption of a known finite stock
of a resource that will be depleted over time. The resource owner chooses the optimal
extraction path over time in order to maximize his profit. There has been a recent stream
of literature using Hotelling-type resource economics models to assess the reaction of
resource markets and resource owners to climate policies. The first and most prominent
paper by Hans-Werner Sinn (2008) also coined the term “green paradox” to describe the
potential adverse effect climate policies may have on total carbon emissions due to the
reactions of resource owners. According to Sinn (2008), if a demand reducing climate
policy is implemented with increasing intensity in the future, the resource owners will
change the extraction path such that more of the resource is extracted by lowering the
price to consumers in earlier years. This reaction is only possible if there is a strictly
positive scarcity rent. This, however, may not be the case when the resource stock is very
high. A recent paper by Hart and Spiro (2011) shows that historically as well as today
the effect of the scarcity is marginal or non-existent and that other factors will shape the
prices of oil and coal in the next decades. In the case of coal, as we can see in Figure 5.1,
the IPCC (2011) estimates that only about 30% of the coal in the ground will be used
by 2100. We can also assume that the entire stock will never be depleted due to climate
policies already in place. With this information it is hard to see how a positive scarcity
rent could arise in the time frame relevant to climate policy, if it ever arises. Without a
positive scarcity rent and assuming a competitive coal market (see Chapter 2) the “green
paradox” effect cannot occur on the steam coal market: prices equal the marginal costs of
the marginal supplier and supply follows demand. This is also the finding of Michielsen
(2011) that shows that in a model combining oil and coal, intertemporal carbon leakage,
the “green paradox”, is less of a concern than spatial leakage, that will be at the heart of
the analysis in this chapter.
More recently resource markets have been investigated using partial equilibrium modeling techniques with a main focus on market power issues (e.g., Holz et al., 2008). We
show in Section 5.2.2 that partial equilibrium model are also suitable for climate policy
127
Chapter 5. Climate Policies and the Global Steam Coal Market
Source: IPCC (2011)
Figure 5.1: Projected carbon emissions from fossil fuels and amounts in the ground
analysis.
5.2.2
Advantages of partial equilibrium models
Having ruled out the possibility of a large-scale reaction of coal resource owners to demand
reducing climate policies in the previous section and since we want to focus on the
interaction of the coal market with climate policies only (and not with carbon leakage in
general that would require CGE model as described in Section 5.2.1), we now concentrate
on the remaining market effects. The introduction of heterogeneous types of climate
policies, geographically as well as in the policy type, will affect the market supply and
demand balance both locally and globally. The future use of coal in the world will be
primarily influenced by climate policies, which will consequently affect the quantities of
steam coal demanded. Quantity effects on the supply or demand side have price effects
on the global market of steam coal and in turn, as there is an elastic demand for coal,
influence coal consumption. These “market adjustments” 64 can affect the effectiveness
of climate policy in a positive or negative way. Partial equilibrium models are the ideal
way to assess these effects since detailed market effects are at the core of these models.
Comparative static or multi-period scenario analyses have been widely used with
partial equilibrium models, especially in the natural gas sector. Surprisingly, they have
not been used much for climate policy scenarios yet. One main focus has been on market
power (e.g., Lise and Hobbs, 2008, Holz et al., 2008) and other scenarios include demand
64
This mechanism has also been described as being part of the carbon leakage mechanisms in the
literature (see Dröge, 2009) but is rarely analyzed. To avoid confusion we restrict the term leakage only
to industrial operations’ relocation and investment effects and introduce the term “market adjustments”
for pure market effects.
128
5.2. Assessment of Positive Modeling Approaches
scenarios, supply modifications, investment constraints and disruptions (e.g., Lise et al.,
2008, Huppmann et al., 2011). The numerical modeling literature for the coal market
has focused on market power issues (Chapters 2 and 3 of this thesis; Paulus and Trüby,
2011b) or infrastructure decisions (Paulus and Trüby, 2011a).
5.2.3
Description of the COALMOD-World model
The COALMOD-World model, described in the previous Chapter 4, is a multi-period
partial equilibrium model of the global steam coal market. It calculates yearly market
equilibria for traded prices and quantities for the years 2006, 2010, 2015, 2020, 2025
and 2030 as well as investments in production and transport capacities between those
years. The profit maximizing players are 25 producers and 14 exporters serving a total
of 41 demand centers. The market is assumed to be competitive following the results
of Chapter 2. Virtually all worldwide steam coal demand is included as we model both
domestic markets and the global seaborne market. The level of detail and disaggregation
of the COALMOD-World model allows for a differentiated analysis of potential market
adjustment effects as a reaction to climate policies.
The profit maximizing players in the model are the producers and the exporters.
The model producers aggregate the companies active in one mining basin. They bear
the costs for mining the coal and transporting the coal overland. They can either sell
directly the coal to domestic demand regions or to the exporters. The exporters aggregate the export capacity of a region and bear the port operating costs as well as the
freight costs for overseas transport. They can sell the call to all the demand nodes with
import capabilities. Demand is represented by an inverse linear demand function with
the quantity expressed in energy units and a price per energy unit.
The players are subject to various constraints such as reverve, production and transport capacities for the producers and export capacity constraints for the exporters. The
players maximize their profit until 2030 using a net present value approach with perfect
foresight about future market situations. They can invest to increase the capacities and
lift previous constraints on production inland transport and exports. Thus, the model
shows how future demand may be served optimally in the future in a cost minimizing way.
The equations of the constrained profit maximization problems of the model producers
and exporters are presented in the previous Chapter 4.
A reserve constraint is included for each producer for indicative and completeness
reasons but they are virtually never binding until the model horizon, so they are not
considered in the players decision.65 Only two countries, Vietnam and Mongolia, have
a binding reserve constraint but since they sell all their coal to China we do not see
any distortion in the global trade flows. In our model a player with a binding reserve
constraint would perform a Hotelling-like intertemporal optimization taking into account
the scarcity rent of the reserve constraint. We however ruled out such a behavior in
65
The actual model horizon is until 2040 but the results are only analyzed until 2030 to avoid end of
time biases, especially for the investments.
129
Chapter 5. Climate Policies and the Global Steam Coal Market
Section 5.2.1.
5.3
Climate Policy Scenarios with the COALMOD-World
Model
A partial equilibrium model such as the COALMOD-World model allows for a broad
range of possible scenario. Out of this set we have to choose the most relevant scenarios
and ensure scenario consistency.
As stated before, climate policy is expected to be the main driver of steam coal
demand in the future. Thus, we divide the scenario space into three possible futures
of global climate policy intensity as defined by the IEA (2010) World Energy Outlook
(WEO). These scenarios are ordered here from the less intense climate policy implementation to the most: the Current Policies scenario, the New Policies scenario and the 450
ppm scenario. In the IEA Current Policies scenario, it is assumed that as of mid-2010
no change in the current policies will be implemented and that the recently announced
commitments are not acted upon. In the New Policies scenario, the recently announced
commitments and policies, for example from the 2009 Copenhagen Climate Conference,
are fully implemented. Finally the 450 ppm scenario is named after the low carbon
dioxide concentration in the atmosphere that is reached in order to keep the increase in
global average temperatures below 2°C.
The second division of the scenario space is made by an exogenous market constraint.
We assume that there are restrictions on production capacity expansions for various
reasons that can be geological, technical and economical (financial restrictions, lack of
qualified labor force or equipment). For the constrained case, the level of these restrictions is based on historical capacity data provided by the USGS in the country reports
of the Mineral Yearbook66 and on historical production data. The historical capacity
expansion data as well as the production data allow us to assess how, in periods of high
demand and high prices, as in the last ten years, producers managed to expand their
production over time. The values implemented in the model range from 5 to 30 Mtpa
of additional capacity over a 5 year period depending on the production node. This is a
rather conservative assessment that can be regarded as the upper bound on the axis of
the possible levels of exogenous market constraints. At the other extreme, these restrictions are completely lifted in the unconstrained case and the model producers can thus
invest in production capacity as much as needed to maximize their profits.
Figure 5.2 shows all the scenarios implemented for this chapter as the little squares
in their respective scenario space (policy framework and market condition) and, where
it applies, with the additional policy shock implemented. We investigate the effects of
three policy shocks in this chapter: a unilateral climate policy by Europe, a restriction of
Indonesian coal production and exports, and a fast roll-out of CCS. They are described
in the following Sections 5.3.2 to 5.3.4. In the European and CCS scenarios the policy
66
http://minerals.usgs.gov/minerals/pubs/country/
130
5.3. Climate Policy Scenarios with the COALMOD-World Model
Figure 5.2: Scenario space
shocks are implemented by respectively decreasing or increasing the quantity demanded
of certain demand nodes by modifying the demand functions. In the Indonesian scenario
an additional restriction that affects the Indonesian model exporter is introduced.
To ensure scenario consistency and comparability the shocks are applied to the six
reference scenarios (plain-colored squares) and the results of the policy shock simulations
are compared to their respective reference case. The reference cases are calibrated such
that for most of the demand centers the consumed steam coal quantities are in a 10%
range above or below the quantities defined by the IEA (2010) WEO. As our demand
functions are constructed using a reference demand, a reference price and a demand
elasticity, we calibrate the reference prices to fit the quantities. We calibrate such that
at least 80% of the demand nodes for all model years are in the 10% range above and
below the WEO quantities of their respective scenario.
A difficult issue in partial equilibrium analysis is the price elasticity of demand due
to the lack of econometric studies. Paulus and Trüby (2011b) give an overview of the
results of econometric studies that estimate short-term price elasticities for coal. The
range is between -0.05 and -0.57. The elasticities for the base year 2006 are based on
our previous work (see Chapter 4). The elasticities of the following years are gradually
set higher as we assume that countries will have a more diverse energy mix and higher
flexibility in their power systems in the future (see Appendix 5.A).
131
Chapter 5. Climate Policies and the Global Steam Coal Market
Unconstrained investments in
production capacity
Constrained investments in
production capacity
CO2
emissions
from
domestic
use and
overland
trade
CO2
emissions
from
global
seaborne
trade
in million
tons CO2
Reduction from
Current Policies
Reduction from
New Policies
450 ppm
scenario
Figure 5.3: Annual carbon dioxide emissions from steam coal consumption in the six
reference scenarios
5.3.1
Worldwide climate policy
Before we describe the outcomes of the scenarios resulting from an additional policy
shock on an already implemented level of global climate policy, we analyze the model
outcomes from these global climate policies that represent our reference cases. We have
three different levels of global climate policy: the Current Policies scenario, the New
Policies scenario and the 450 ppm scenario based on the projections of the IEA (2010)
WEO. Additionally to that policy framework we have to consider the market conditions
as shown in Figure 5.2. In one case, investments in production capacity are constrained,
in the other case not in order to represent two extremes of a continuum of exogenous
market constraints.
Figure 5.3 shows the results from the different modeling runs of the reference cases in
million tons of carbon dioxide emissions. Since the emissions are proportionally linked to
the consumption of coal in energy units we can also infer the development of consumption
and trade from these figures. The light grey area represents the emissions in the 450
ppm scenario (the same as for the other scenarios in 2010), adding the dark gray area
represents the emissions from the New Policies scenario and all the areas together show
the emissions in the Current Policies scenario.
In the case of unconstrained investment possibilities in production capacity, coal
consumption is significantly higher in the Current Policies scenario and slightly higher in
the New Policies scenario. Global seaborne trade remains important and will continue
to grow. We see a reduction in global trade only in the 450 ppm scenario. In the
132
5.3. Climate Policy Scenarios with the COALMOD-World Model
case of a constrained market condition global seaborne trade is especially important to
help countries like China and India to meet their coal demand as they might experience
difficulties in expanding their domestic production base.
5.3.2
Unilateral European climate policy
The unilateral European climate policy scenario is implemented in two global climate
policy frameworks: the Current Policies and the New Policies framework as shown in
the scenario overview of Figure 5.2. In this scenario the European Union goes a step
further and aims at reducing CO2 emissions by 30% compared to the level of 1990 by
2020 with further reductions in the future. This goal is reached through a significantly
lower coal consumption in the European Union. In the IEA (2010) WEO scenarios this
is represented by the demand values of the 450 ppm scenario. The steam coal demand
reduction compared to the reference scenarios is shown in Table 5.1.
Table 5.1: EU demand reduction in the Unilateral European Climate Policy scenario
compared to the reference scenarios in percentage
Demand reduction from Current Policies
Demand reduction from New Policies
2020
-0.24
-0.02
2025
-0.43
-0.20
2030
-0.64
-0.45
Source: own after IEA (2010)
The modeling results are shown in Figure 5.4. The dark grey area represents the
actual emissions reduction in the EU due to the EU unilateral climate policy. The light
grey area are the targeted emissions and together with the black market adjustments
they represent the actual global emissions. We can see that, given certain conditions,
market adjustments can seriously undermine a unilateral European climate effort. This is
especially the case in the Current Policies framework with a constrained market condition.
In that case global coal demand is high and the market somewhat constrained we see that
a reduction in European coal demand allows the Asian countries to consume significantly
more. In 2025, the market adjustment nullifies 66% of the European reduction target
and 29% in 2030. In the unconstrained case the market adjustment is negligible but
global emissions are much higher.
In the New Policies framework the market adjustment is much lower. It is interesting
to note that the market condition has a high impact on the global level of emissions and
on the market adjustment mechanism. In 2030, in the constrained case the mechanism
works as described above for the Current Policies constrained case: the European reduction allows Asia to consume more through the global price mechanism. However, in
the unconstrained case, in 2025, the market adjustment with considerably lower prices
occurs because more quantities are potentially available when there is no constraint. The
lower European consumption has a significant impact on prices that it has not in the
constrained case where global demand remains slightly restricted.
133
Chapter 5. Climate Policies and the Global Steam Coal Market
CO2
emissions
based on
Current
Policies
scenario
Constrained investments in
production capacity
Unconstrained investments in
production capacity
12200
12200
11800
11800
11400
11400
11000
11000
10600
10600
10200
10200
2025
CO2
emissions
based on
New
Policies
scenario
2030
2025
9200
9200
9000
9000
8800
8800
8600
8600
2025
in million
tons CO2
2030
2030
Actual reduction from
unilateral EU policy
2025
Emissions from
market adjustment
2030
Remaining
emissions
Figure 5.4: Worldwide emissions reductions and adverse market adjustments in the Unilateral
European Climate Policy model scenario
We can conclude that market adjustments are very likely to occur but that their
adverse effect is generally low and will not overcompensate the emissions reductions
from Europe. This is due to the relatively small size of EU demand in the global steam
coal demand. However, in the case of a low level of global climate policy as in the Current
Policies scenario the adverse market adjustment effect can be very high. Thus, it is logical
for the European Commission to say that it will aim at a 30% emissions reduction goal
only if other countries take a binding commitment to higher reduction goals. In such a
case, that can be described by the New Policies scenario, Europe can always go an extra
mile without expecting too much adverse market adjustments.
5.3.3
Yasuní-type supply-side policy in Indonesia
The Yasuní-ITT initiative proposed by the Ecuadorian government aims at combating
global warming, protecting biodiversity and indigenous people as well as implementing
a sustainable social and energetic development by refraining indefinitely from exploiting
the oil reserves of the Ishpingo-Tambococha-Tiputini (ITT) oil field within the Yasuní
National Park (Larrea, 2010).67 This field represents 20% of the Ecuadorian oil reserves
and the initiative requires a capital contribution of at least half of the earnings Ecuador
would receive from exploitation. Valuated at 76.38 USD per barrel this represents a sum
of 3.635 billion USD supplied by the international community to a fund managed by the
United Nations Development Programme. The initiative represents 407 Mt CO2 saved
67
http://yasuni-itt.gob.ec/
134
5.3. Climate Policy Scenarios with the COALMOD-World Model
CO2
emissions
based on
Current
Policies
scenario
Constrained investments in
production capacity
Unconstrained investments in
production capacity
12200
12200
11800
11800
11400
11400
11000
11000
10600
10600
10200
10200
2025
CO2
emissions
based on
New
Policies
scenario
2025
9200
9200
9000
9000
8800
8800
8600
2030
8600
2025
in million
tons CO2
2030
2030
Reduction through Indonesian supply
side policy of export restrictions
2025
2030
Remaining
emissions
Figure 5.5: Worldwide emissions reduction in the Indonesian supply-side policy scenario
from not using the oil resource and an additional 820 Mt CO2 mitigation potential over
20 years from avoided deforestation and forest management (Larrea, 2010).
For the Indonesian scenario, we use the same idea and apply it to another geographic
area and to our fuel of interest in this thesis, steam coal. The bulk of coal exploitation
in Indonesia takes place on the island of Kalimantan (formerly known as Borneo). This
island is home to one of the greatest rainforests in the world and a treasure of biodiversity
that is endangered by coal mining through deforestation and local air and water pollution. Fatah (2008) points out that coal mining has little to no beneficial effects on the
local economy. The revenues and benefits go to private companies and the government.
Thus, one could imagine a supply-side climate policy mechanism similar to the YasuníITT Initiative to preserve the Indonesian forest and prevent the extraction, export and
carbon dioxide emissions from coal albeit still allowing a local use of steam coal for power
generation. While such a policy is not explicitly considered by the Indonesian government and would require an intensive cooperation with international stakeholders, there
are signs in the debate about coal in Indonesia that show that it could be highly compatible with Indonesia’s national interests. More specifically, given the limited nature of
Indonesian coal reserves, nationalistic resource policies are starting to be implemented.
For example the “domestic market obligation” established by the 2009 Mining Law gives
the possibility to mandate that up to 35% of the production of a mining company has
to be sold on the domestic market (Lucarelli, 2010). Thus, an export limitation with an
international financial transfert to compensate lost export revenues would be a complement for such a coal policy concentrated on the domestic needs. In our particular case
135
Chapter 5. Climate Policies and the Global Steam Coal Market
we modeled this policy as an export restriction (maximum quantity that can be exported
in a given year) for Indonesia as follows: 2006 to 2015, no restriction; 2020: 50 Mtpa;
2025: 25 Mtpa; 2030: 0 Mtpa (phase-out of export).
In our reference cases for all policy environments Indonesia is the most important
supplier to the global market with yearly export values that can be higher than 200
Mt. The results of the Indonesia scenario run with export restrictions are summarized in
Figure 5.5. We see that the reduction effect is the strongest in the case of a constrained
global market because it is hard to find alternative suppliers on the world market that
could replace the lacking Indonesian exports. In the unconstrained case the effect is lower
as Indonesian coal is substituted by other producers. In any case most of the emissions
reduction, around 80% of the reduction, takes place in Asia, whereas Europe only accounts for 10%. The reduction in coal consumption in Asia due to higher prices may have
additional benefits as consumers will become more aware of other alternatives for their
energy supply such as renewables, energy conservation and efficiency and governements
might enact policies towards those ends. However, we must be wary that such an effect
may limited in time as the supply gap may be covered by other producers over time.
5.3.4
CCS fast roll-out
Carbon capture and storage (CCS) is a set of technologies that aim at a reduction of
carbon dioxide emissions into the atmosphere by separating and capturing the CO2 at
the power plant and transporting it to a geological sink where it will be compressed and
stored underground (see IPCC, 2005). The CCS technology is regarded by the IPCC
and by the IEA to be one of the major options for climate change mitigation. However,
as of 2011 there are only about 10 pilot CCS plants operating in the world and not a
single large scale operation (22 are planned to start operating between 2014 and 2020).68
In the IEA (2010) WEO scenarios, CCS plays a significant role in the 450 ppm scenario
after 2025 and a smaller role in the New Policies scenario but only after 2030. There are
various technological and political barriers to the implementation that explain this late
roll-out of the CCS technology (see Gibbins and Chalmers, 2008).
Table 5.2: Assumed installed capacities of coal power plants with CCS for the CCS scenario
in GW
World
OECD+ (incl. Europe, USA, Japan)
OME (incl. China, Russia, South Africa)
OC (incl. India, South-East Asia)
2020
150
72
74
4
2025
286
134
145
6
2030
423
197
216
9
Source: own calculations based on IEA (2010)
In our CCS fast roll-out scenario we assume that technological breakthroughs, a fa68
source:
Carbon
Capture
and
Sequestration
(http://sequestration.mit.edu/tools/projects/index.html).
136
Technologies
Program
at
MIT
5.3. Climate Policy Scenarios with the COALMOD-World Model
vorable regulatory framework as well as a strong political support create the conditions
for a fast CCS roll-out with significant capacities coming in as early as 2020. Such a
scenario makes only sense in an overall environment of ambitious climate policy, thus we
apply this additional policy shock in the New Policies and the 450 ppm policy framework
only, as shown in Figure 5.2. For this scenario we assume that the worldwide installed capacities of coal power plants with CCS projected by the IEA (2010) WEO in the 450 ppm
scenario are put in place five years earlier. We assume that half of this additional capacity
replaces existing older coal power plants, the other half is integrated in the power system
as additional capacity, successfully competing with other technologies. Furthermore, for
our coal demand calculations, we assume that CCS power plants have a 38% efficiency
and a capacity factor of 82%. Thus, we actually compute two additional demand shocks:
one coming from half of the CCS capacity that is added to the coal demand and the other
because the lower efficiency of CCS power plants requires additional coal to produce the
same amount of energy. The assumed capacities of CCS for our modeling runs are shown
in Table 5.2 divided into the following regions: OECD+, Other Major Economies and
Other Countries.69
The results of the CCS scenario are presented in Figure 5.6. Let us start analyzing
the scenario based on the New Policies climate policy framework. CCS is insignificant
in the reference scenario and therefore the additional CCS capacity, half of which leads
to new coal demand, has a strong effect on the market. We see a market adjustment
that is positive for the climate. The higher demand leads to higher prices that lead to
a reduction of demand from conventional power plants. This effect is very strong in a
constrained market environment and significant in the unconstrained environment.
In the scenario based on the 450 ppm climate policy environment the global coal
demand is so low that the market condition has very little effect on the scenario outcomes.
However, we observe different market adjustment effects. In 2020, there is very little CCS
in the reference case so that the additional demand for coal creates a market adjustment
effect with high prices similar to the one described in the New Policies case. In 2025,
we have an opposite “negative” market adjustment effect with more emissions. This is
due to an effect of decreasing demand with a capacity effect: in 2020 capacities are build
up to serve the additional demand and they create a slight oversupply situation in 2025.
This effect overcompensates the little CCS addition but is only temporary.
In the case of a relatively intense global level of climate policy, such as in the New
Policies framework, a faster implementation of CCS would be very beneficial. We expect
that, additionally to the captured quantities of CO2 , a positive market adjustment will
further reduce coal consumption and emissions. The picture is less clear in the 450 ppm
case. But in the case this very ambitious climate scenario becomes reality in the future,
some amount of market adjustment on the steam coal market has little relevance in the
overall required transformation of the global energy system anyways.
69
See IEA (2010) for an exact definition of these aggregates.
137
Chapter 5. Climate Policies and the Global Steam Coal Market
Figure 5.6: Worldwide emissions in the CCS scenario
5.3.5
Scenario combination: hedging of negative market adjustments
As shown in Section 5.3.2, negative market adjustments can have a significant impact
and partly render ineffective an emissions reduction effort. In the case of a unilateral
European climate effort the use of steam coal is reduced significantly. However, through
pure markets effects and price mechanisms, this lower consumption in Europe is compensated by a higher consumption in the rest of the world, especially Asia. One possibility
to counteract this effect would be to accompany the demand reduction in Europe by
another policy measure to induce beneficial market adjustments as a way to “hedge”
against adverse market adjustments effects. In the following modeling exercise we therefore present a combination of the Unilateral European Climate Policy scenario with the
Indonesian supply-side policy of export restriction.
Figure 5.7 shows the results of this combined model run in comparison to the results
of the European Unilateral Climate policy scenario from Section 5.3.2. We concentrate on
the case based on the Current Policies scenario with constrained investments in production capacity where we can expect the most drastic effects. The black area in Figure 5.7
represents the additional emissions from negative market adjustments which, as shown
above, can be very strong in the original scenario. The darker grey area represents
the avoided emissions in Europe and the white area shows positive market adjustments
that is avoided emissions in the rest of the world. The lower grey area in the columns
represents the remaining emissions.
The right-hand graph shows the results of the combined scenarios: the addition of
138
5.4. Conclusions and Policy Recommendations
Figure 5.7: Unilateral European Climate policy results with and without additional
Indonesian supply-side policy
the Indonesian supply-side policy cancels out (or “hedges against”) and even overcompensates the negative market adjustment effect in 2020 and 2025 and leads to an overall
emissions reduction. In 2020 and 2025, the Indonesian export restriction induces a lower
consumption in Asia. In 2030, however, it affects Europe more, as the higher supply
costs lead to an import reduction. In fact there is a beneficial market adjustment taking
place in Europe that leads to an overachievement of the climate target. However, by 2030
we see the resurgence of some negative market adjustment effect in Asia, induced by an
investment effect. The lacking supplies from Indonesia are compensated by investments
in new mines, especially in China, that become economically available on a larger scale
by 2030.
Despite the overall positive effect of this combined policy we must be aware that its
effects might be limited in time. Also the proposed hedging strategy may be politically
too difficult to put in place. However, other “hedging” strategies are possible to mitigate
adverse market effects. These could be demand reducing policies, e.g., through mechanisms already in place such as the Clean Development Mechanisms (see Kemfert, 2002).
The “hedging” effect will appear when the build-up of renewable energy capacity replaces
the use of steam coal for electricity generation.
5.4
Conclusions and Policy Recommendations
Using the COALMOD-World model we are able to make differentiated conclusions about
the effectiveness of different types of climate policy alternatives through their interplay
with the global steam coal market in various market conditions.
If we take a European and policy-oriented view of this results, several conclusions
and recommendations can be drawn to prioritize different climate policy options. In the
case of a European unilateral climate policy in a context of little global climate effort, we
can expect adverse market adjustment effects that can compensate up to two-thirds of
139
Chapter 5. Climate Policies and the Global Steam Coal Market
European emissions reductions. Thus, the first priority of the European climate policy
should be to reach a global level of climate policy that is at least at the level of the nonbinding agreements taken in recent climate conferences in Copenhagen in 2009 and in
Cancún in 2010. A global climate policy has the biggest effect on global carbon emissions
coming from the coal sector. If this is reached the EU can always go further in reducing
its steam coal consumption without risking too much adverse effects from the global
market.
A supply-side climate policy in Indonesia also has some significant CO2 emissions
reduction effects that are potentially in the same order of magnitude as the European
unilateral climate policy and help reduce emissions mostly in Asia countries. It is interesting to note that this supply-side policy has its best performance in the context where the
European unilateral climate policy sees the most important negative market adjustment
effects, i.e., when there is a low intensity of global climate policy and when the market is
constrained. Thus, the EU could try and pursue such an unconventional climate policy
as a way to hedge against adverse effects from its own domestic climate policy. Also
since the emissions reductions occur mainly in Asian countries, especially China, this
might give additional impulses for consumers to reduce their reliance on imported fossil
fuels and for policymakers to implement ambitious climate policies. In Indonesia such a
policy would also have additional beneficial effects for nature conservation, the protection
of biodiversity as well as avoided CO2 emission from deforestation.
The first advantage of a policy that aims at a faster roll-out of the CCS technology is the emissions reduction through the capture of CO2 . Climate beneficial market
adjustment effects can also occur. A significant impact of CCS can only be expected
if this technology is implemented globally. Thus, the strategy of the EU should be to
support the roll-out of this technology in Europe but also abroad through international
cooperation.
If we rank all the different policy options examined in this chapter using the COALMOD-World model with regard to their effectiveness in reducing carbon emissions from
steam coal use on a global level, we obtain the following order of priority for EU climate policy. First, the EU should aim at establishing a strong globally binding climate
agreement. Secondly, the EU should support a fast roll-out of CCS, both in the EU and
globally. Thirdly, on the same level, the EU can set a more stringent emissions reduction
goal for itself and be open to more unconventional climate policies such as the described
supply-side reduction in Indonesia through production and export limitations.
140
5.A. Appendix
5.A
Appendix
Table 5.3: Demand elasticities
Demand node
C_NFB
C_ITA
C_RUS_Siberia
C_RUS_Central
C_CAN
C_THA
C_VNM
C_ESP
C_FIN
C_JPN
C_TUR
C_DEU
C_PRT
C_PHL
C_MYS
C_MNG
C_UKR
C_GBR
C_KOR
C_IDN
C_USA_Rocky
C_USA_Central
C_USA_South
C_USA_Gulf
C_USA_East
C_DNK
C_TWN
C_MAR
C_IND_East
C_IND_North
C_IND_West
C_IND_South
C_ISR
C_KAZ
C_CHN_Northeast
C_CHN_SIS
C_CHN_Main
C_CHN_Eastern
C_CHN_South
C_POL
C_ZAF
2006
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
2010
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
141
2015
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
2020
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
-0.2
2025
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
-0.3
2030
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
-0.4
Bibliography
Aguilera, Roberto F. 2008. The Availability of Global Fossil Energy. Germany: VDM
Publishing House.
Aguilera, Roberto F., Roderick G. Eggert, Gustavo Lagos C.C. and John E. Tilton. 2009.
“Depletion and the Future Availability of Petroleum Resources.” The Energy Journal
30(1):141–174.
Aune, Finn Roar, Solveig Glomsrod, Lars Lindholt and Knut Einar Rosendahl. 2005.
“Are high oil prices profitable for OPEC in the long run?” Statistics Norway Discussion
Paper 416: Oslo.
Aune, Finn Roar, Rolf Golombek, Sverre A. C. Kittelsen, Knut Einar Rosendahl and Ove
Wolfgang. 2001. “LIBEMOD - LIBEralisation MODel for the European Energy Markets: A Technical Description.” Ragnar Frisch Centre for Economic Research Working
paper 1/2001.
Baruya, Paul. 2007. Supply Costs for Internationally Traded Coal. London: IEA Clean
Coal Centre.
Bazaraa, M.S., H.D. Sherali and C.M. Shetty. 2006. Nonlinear programming: theory and
algorithms. Wiley-Interscience.
BGR. 2009. “Energierohstoffe 2009 - Reserven, Ressourcen, Verfügbarkeit.” Technical
report, Bundesanstalt für Geowissenschaften und Rohstoffe (BGR) - Federal Institute
for Geosciences and Natural Resources.
Billups, Stephen C. and Katta G. Murty. 2000. “Complementarity problems.” Journal of
Computational and Applied Mathematics 124:303–318.
Böhringer, C. 1998. “The Synthesis of Bottom-up and Top-down in Energy Policy Modeling.” Energy Economics 20(3):233 – 248.
Böhringer, C., C. Fischer and K.E. Rosendahl. 2010. “The Global Effects of Subglobal
Climate Policies.” The BE Journal of Economic Analysis and Policy 10(2).
Böhringer, C. and T.F. Rutherford. 2000. “Decomposing the Cost of Kyoto: a Global
CGE Analysis of Multilateral Policy Impacts.” ZEW Discussion Papers 00-11.
142
Bibliography
Boots, Maroeska G., Fieke A.M. Rijkers and Benjamin F. Hobbs. 2004. “Trading in the
Downstream European Gas Market: A Successive Oligopoly Approach.” The Energy
Journal 25(3):73–102.
Bovenberg, A.L. and L.H. Goulder. 2005. “Efficiency Costs of Meeting IndustryDistributional Constraints under Environmental Permits and Taxes.” The RAND Journal of Economics 36(4):951 – 971.
Bowley, A. L. 1924. The Mathematical Groundwork of Economics : an Introductory
Treatise. Clarendon Press, Oxford.
BP. 2009. BP Statistical Review of World Energy 2009. London: BP.
Bresnahan, Timothy F. 1981. “Duopoly Models with Consistent Conjectures.” The American Economic Review 71(5):934–945.
Bresnahan, Timothy F. 1989. “Empirical studies of industries with market power.” Elsevier, volume 2 of Handbook of Industrial Organization, chapter 17, 1011–1057.
Bushnell, James. 2003. “A Mixed Complementarity Model of Hydrothermal Electricity
Competition in the Western United States.” Operations Research 51(1):80–93.
Carl, Jeremy, Varun Rai and David Victor. 2008. “Energy and India’s Foreign Policy.”
Program on Energy and Sustainable Development Working Paper #75.
Chamber of Mines of South Africa. 2008. “Facts and Figures 2007.” http://www.
bullion.org.za/Publications/Facts&Figures2006/Facts&Figures2007.pdf.
Chen, Chi-Chung, Bruce A. McCarl and Ching-Cheng Chang. 2006a. “Estimating the
Impacts of Government Interventions in the International Rice Market.” Canadian
Journal of Agricultural Economics/Revue canadienne d’agroéconomie 54(1):81–100.
Chen, Yishu H., Benjamin F. Hobbs, Todd S. Munson and Sven Leyffer. 2006b. “LeaderFollower Equilibria for Electric Power and NOx Allowances Markets.” Computational
Management Science 3(4):307–330.
Claerbout, Jon and Francis Muir. 2008. “Hubbert math.” http://sepwww.stanford.
edu/sep/jon/hubbert.pdf.
Corts, Kenneth S. 1998. “Conduct Parameters and the Measurement of Market Power.”
Journal of Econometrics 88(2):227–250.
Cottle, Richard W., Jong-Shi Pang and Richard E. Stone. 1992. The Linear Complementarity Problem. San Diego: Academic Press.
Cournot, Antoine-Augustin. 1838. Recherches sur les Principes Mathématiques de la
Théorie des Richesses. Paris: L. Hachette.
143
Bibliography
Couser, Sue and David Goldsack. 2001. “Konzentrationstendenzen in der Internationalen
Kohleindustrie.” Zeitschrift für Energiewirtschaft 25(1):43–52.
Crocker, Geoff and Alex Kovalchuk. 2008. Prospects for Coal and Clean Coal Technologies
in Russia. London: IEA Clean Coal Centre.
Dahl, Carol. 1993. “A Survey of Energy Demand Elasticities in Support of the Development of the NEMS.” Colorado School of Mines.
Datanet India. 2009. “Indiaenergystat.” Various spreadsheets on coal in India retrieved
at http://www.indiaenergystat.com in June 2009. Datanet India Pvt. Ltd.
Daughety, Andrew F. 1985. “Reconsidering Cournot: The Cournot Equilibrium is Consistent.” The RAND Journal of Economics 16(3):368–379.
DEHSt. 2005. “Emissionsfaktoren und Kohlenstoffgehalte (Stoffliste 2005).” http://www.
dehst.de, deutsche Emissionshandelsstelle (DEHSt), Umweltbundesamt, Berlin; Federal Ministry for Environment, Nature Conservation and Nuclear Safety.
Demailly, D. and P. Quirion. 2008. “European Emission Trading Scheme and Competitiveness: A Case Study on the Iron and Steel Industry.” Energy Economics 30(4):2009
– 2027.
DeMiguel, Victor and Huifu Xu. 2009. “A Stochastic Multiple-Leader Stackelberg Model:
Analysis, Computation, and Application.” Operations Research 57(5):1220–1235.
DERA. 2011. “DERA Rohstoffinformationen - Kurzstudie: Reserven, Resourcen und Verfügbarkeit von Energierohstoffen 2011.” Technical report, DERA Deutsche Rohstoffagentur - Bundesanstalt für Geowissenschaften und Rohstoffe (BGR).
Dirkse, Steven P. and Michael C. Ferris. 1999. “Modeling and Solution Environments for
MPEC: GAMS and MATLAB.” In: Fukushima, Masao and Liqun Qi (eds.), Reformulation: nonsmooth, piecewise smooth, semismooth, and smoothing methods, Kluwer
Academic Publishers, Applied Optimization, 127–147.
Dixit, A.K. 1990. Optimization in economic theory. Oxford University Press.
Dockner, Engelbert J. 1992. “A Dynamic Theory of Conjectural Variations.” The Journal
of Industrial Economics 40(4):377–395.
Dröge, Susanne. 2009. “Tackling Leakage in a World of Unequal Carbon Prices, Synthesis
Report.” Technical report, Climate Strategies.
E3MLab. 2011. “PRIMES Model Presentation for Peer Review.” http://www.e3mlab.
ntua.gr/e3mlab/PRIMESManual, E3MLab - National Technical University of Athens.
EC. 2011. “Energy Roadmap 2050.” European Commission, COM(2011) 885/2.
144
Bibliography
Egging, Ruud, Steven A. Gabriel, Franziska Holz and Jifang Zhuang. 2008. “A Complementarity Model for the European Natural Gas Market.” Energy Policy 36(7):2385–
2414.
Egging, Ruud, Franziska Holz and Steven A. Gabriel. 2010. “The World Gas Model:
A Multi-Period Mixed Complementarity Model for the Global Natural Gas Market.”
Energy 35(10):4016–4029.
Ehrenmann, Andreas. 2004. Equilibrium Problems with Equilibrium Constraints. Ph.D.
thesis, University of Cambridge.
EIA. 2007. “Coal Market Module of the National Energy Modeling System - Model Documentation 2007.”
Ekawan, Rudianto and Michel Duchêne. 2006. “The Evolution of Hard Coal trade in the
Atlantic market.” Energy Policy 34(13):1487–1498.
Ekawan, Rudianto, Michel Duchêne and Damien Goetz. 2006. “The Evolution of Hard
Coal trade in the Pacific market.” Energy Policy 34(14):1853 – 1866.
Ellerman, A. Denny. 1995. “The World Price of Coal.” Energy Policy 23(6):499–506.
Energy Forecasting Agency. 2008. “General scheme for the positioning of power plants
till 2020.” http://www.e-apbe.ru/scheme/gs.doc, only in Russian.
Energy Information Administration. 2004. “Coal Transportation: Rates and Trends in
the United States, 1979-2001 (with Supplementary Data to 2002).” http://www.eia.
doe.gov/cneaf/coal/page/trans/ratesntrends.html.
Energy Information Administration. 2007. “Coal Distribution - 2006.” http://www.eia.
doe.gov/cneaf/coal/page/trans/ratesntrends.html.
Energy Information Administration. 2008. “Annual Coal Report - 2007.” http://www.
eia.doe.gov/cneaf/coal/page/acr/acr_sum.html.
Energy Watch Group. 2007. “Coal: Resources and Future Production.” http://www.
energywatchgroup.org, EWG-Paper No. 1/07.
EPRI. 2007. “International Coal Market Analysis.” Technical report, Electric Power Research Institute.
European Commission. 2008. “European Energy and Transport Trends to 2030 - 2007
Update.” European Commission - Directorate-General for Energy and Transport.
EWI/Prognos. 2005. “Energiereport IV - Die Entwicklung der Energiemärkte bis zum
Jahr 2030 - Energiewirtschaftliche Referenzprognose - Kurzfassung.”
Fatah, Luthfi. 2008. “The Impacts of Coal Mining on the Economy and Environment of
South Kalimantan Province, Indonesia.” ASEAN Economic Bulletin 25(1).
145
Bibliography
Ferris, Michael C., Steven P. Dirkse and Alexander Meeraus. 2002. “Mathematical Programs with Equilibrium Constraints: Automatic Reformulation and Solution via Constrained Optimization.”
Ferris, Michael C. and Christian Kanzow. 1999. “Complementarity and Related Problems.” Technical report, University of Wisconsin - Madision and University of Hamburg.
Ferris, Michael C. and Todd S. Munson. 2000. “Complementarity Problems in GAMS
and the PATH Solver.” Journal of Economic Dynamics and Control 24(2):165–88.
Figuières, Charles (ed.). 2004. Theory of Conjectural Variations. Number 2 in Series on
mathematical economics and game theory, Singapore: World Scientifc.
Freese, Barbara. 2004. Coal: A Human History. Penguin (Non-Classics).
Friedman, James W. 1983. Oligopoly Theory. Cambridge: Cambridge University Press.
Frisch, Ragnar. 1933. “Monopole - Polypole - La Notion de Force dans l’Economie.”
Tillægshefte til Nationaløkonomisk Tidsskrift 71:241–259.
Gabriel, Steven A. 2009. “Lecture Slides: Operations Research 3: Day 4: Introduction to
Complementarity Theory.” TU Berlin and DIW Berlin Course, DIW Berlin, Germany,
29 June- 3 July 2009.
Gabriel, Steven. A. and Yves Smeers. 2006. “Complementarity Problems in Restructured
Natural Gas Markets.” In: Seeger, A. (ed.), Lecture Notes in Economics and Mathematical Systems, Berlin/Heidelberg: Springer, Recent Advances in Optimization, 343–373.
Geological Survey of India. 2008. “Inventory of geological resource of Indian coal.” http:
//www.gsi.gov.in/coalreserve_010408.pdf.
Gibbins, Jon and Hannah Chalmers. 2008. “Carbon Capture and Storage.” Energy Policy
36(12):4317 – 4322.
Graham, Paul, Sally Thorpe and Lindsay Hogan. 1999. “Non-competitive Market Behaviour in the International Coking Coal Market.” Energy Economics 21(3):195–212.
Greenhut, John G. and M. L. Greenhut. 1975. “Spatial Price Discrimination, Competition
and Locational Effects.” Economica 42(168):401–419.
Haftendorn, Clemens. 2012. “Evidence of Market Power in the Atlantic Steam Coal Market Using Oligopoly Models with a Competitive Fringe.” DIW Discussion Paper 1185:
Berlin.
Haftendorn, Clemens, Karen Freund, Christian von Hirschhausen and Franziska
Holz. 2008a. “Indien: Entwicklung auf Kosten des Klimas.” DIW Wochenbericht
75(51/52):818–824.
146
Bibliography
Haftendorn, Clemens and Franziska Holz. 2010. “Modeling and Analysis of the International Steam Coal Trade.” The Energy Journal 31(4):201–225.
Haftendorn,
Clemens,
Franziska Holz and Christian von Hirschhausen. 2010.
“COALMOD-World: A Model to Assess International Coal Markets until 2030.” DIW
Discussion Paper 1067. Berlin.
Haftendorn, Clemens, Franziska Holz and Christian von Hirschhausen. 2012a. “The End
of Cheap Coal? A Techno-economic Analysis Until 2030 using the COALMOD-World
Model.” FUEL in press.
Haftendorn, Clemens, Claudia Kemfert and Franziska Holz. 2011. “What about Coal?
Interactions between Climate Policies and the Global Steam Coal Market until 2030.”
DIW Discussion Paper 1146: Berlin.
Haftendorn, Clemens, Claudia Kemfert and Franziska Holz. 2012b. “What about Coal?
Interactions between Climate Policies and the Global Steam Coal Market until 2030.”
Energy Policy in press.
Haftendorn, Clemens, Christian von Hirschhausen and Franziska Holz. 2008b. “Moving
towards a "COAL-PEC"?” DIW Weekly Report (10):62–67.
Harker, Patrick T. 1993. Lectures on Computation of Equilibria with Equation-Based
Methods. CORE Lecture Series, Louvain-la-Neuve: Université Catholique de Louvain.
Harrington, J. E., Benjamin F. Hobbs, Jong-Shi Pang, A. Liu and G. Roch. 2005. “Collusive Game Solutions via Optimization.” Math. Program. 407–435.
Hart, Rob and Daniel Spiro. 2011. “The Elephant in Hotelling’s Room.” Energy Policy
39(12):7834 – 7838.
Hartley, Peter and Kenneth B. Medlock. 2006. “The Baker Institute World Gas Trade
Model.” In: Victor, David G., Amy M. Jaffe and Mark H. Hayes (eds.), Natural Gas
and Geopolitics: From 1970 to 2040, Cambridge, Mass.: Cambridge University Press,
357–406.
Heinberg, Richard and David Fridley. 2010. “The End of Cheap Coal.” Nature 468:367–
369.
Hirschhausen, Christian von, Clemens Haftendorn, Tim Winke and Anne Neumann.
2012a. “Fossil Fuel Availability until 2050. The Real Shortage will be the CO2-Budget.”
In: Glachant, Jean-Michel, Nicole Ahner and Leonardo Meuus (eds.), EU Energy
Innovation towards 2050, Claeys & Casteels.
Hirschhausen, Christian von, Johannes Herold and Pao-Yu Oei. 2012b. “How a “Low Carbon” Innovation Can Fail - Tales from a “Lost Decade” for Carbon Capture, Transport,
and Sequestration (CCTS).” Economics of Energy & Environmental Policy 1(2).
147
Bibliography
Hobbs, Benjamin F. 1986. “Mill Pricing versus Spatial Price Discrimination under
Bertrand and Cournot Spatial Competition.” The Journal of Industrial Economics
35(2):173–191.
Holz, Franziska, Christian von Hirschhausen and Claudia Kemfert. 2008. “A Strategic
Model of European Gas Supply (GASMOD).” Energy Economics 30(3):766–788.
Höök, Mikael, Werner Zittel, Jörg Schindler and Kjell Aleklett. 2010. “Global Coal Production Outlooks Based on a Logistic Model.” Fuel 89(11):3546–3558.
Hoover, Edgard M. 1937. “Spatial Price Discrimination.” The Review of Economic Studies
4(3):182–191.
Hotelling, Harold. 1931. “The Economics of Exhaustible Resources.” Journal of Political
Economy 39(2):137–175.
Hubbert, M. King. 1959. “Techniques of Prediction with Application to the Petroleum
Industry.” URL http://www.hubbertpeak.com/hubbert/TechniquesOfPrediction.
pdf.
Huntington, Hillard G. 2009. “World Natural Gas Markets and Trade: A Multi-Modeling
Perspective.” The Energy Journal Special Issue:218 p.
Huppmann, Daniel, Ruud Egging, Franziska Holz, Christian von Hirschhausen and
Sophia Ruster. 2011. “The World Gas market in 2030. Development Scenarios using
the World Gas Model.” International Journal of Global Energy Issues 35(1):64 – 84.
Huppmann, Daniel and Franziska Holz. 2009. “A Model for the Global Crude Oil Market
Using a Multi-Pool MCP Approach.” DIW Discussion Paper 869: Berlin.
IEA. 2007a. Coal Information 2007. Paris: OECD.
IEA. 2007b. Electricity Information 2007. Paris: OECD.
IEA. 2007c. World Energy Outlook 2007 - China and India Insights. Paris: OECD.
IEA. 2008a. Coal Information 2008. Paris: OECD.
IEA. 2008b. Energy Statistics of Non-OECD Countries. Paris: OECD.
IEA. 2008c. World Energy Outlook 2008. Paris: OECD.
IEA. 2009. World Energy Outlook 2009. Paris: OECD.
IEA. 2010. World Energy Outlook 2010. Paris: OECD.
IEA. 2011a. Coal Information 2011. Paris: OECD.
IEA. 2011b. “Coal Medium-Term Market Report 2011 - Market Trends and Projections
to 2016.” International Energy Agency (IEA). Paris: OECD.
148
Bibliography
IEA. 2011c. World Energy Outlook 2011. Paris: OECD.
IPCC. 2005. “IPCC Special Report on Carbon Dioxide Capture and Storage.” URL http:
//www.ipcc.ch/pdf/special-reports/srccs/srccs\_wholereport.pdf.
IPCC. 2011. “The IPCC Special Report on Renewable Energy Sources and Climate
Change Mitigation - SRREN Generic Presentation.” URL http://srren.ipcc-wg3.
de/ipcc-generic-presentation.
Kaplow, Louis and Carl Shapiro. 2007. “Chapter 15 Antitrust.” Elsevier, volume 2 of
Handbook of Law and Economics, 1073 – 1225.
Kawaguchi, Tsunemasa, Nobuhiro Suzuki and Harry M. Kaiser. 1997. “A Spatial Equilibrium Model for Imperfectly Competitive Milk Markets.” American Journal of Agricultural Economics 79(3):851–859.
Kemfert, Claudia. 2002. “Global Economic Implications of Alternative Climate Policy
Strategies.” Environmental Science & Policy 5(5):367 – 384.
Kolstad, Charles D. and David S. Abbey. 1984. “The Effect of Market Conduct on International Steam Coal Trade.” European Economic Review 24(1):39–59.
Kolstad, Charles D., David S. Abbey and Robert L. Bivins. 1983. “Modeling International
Steam Coal Trade.” Los Alamos National Laboratory.
Kolstad, Charles D. and Anthony E. Burris. 1986. “Imperfectly Competitive Equilibria
in International Commodity Markets.” American Journal of Agricultural Economics
68(1):27–36.
Kolstad, Charles D. and Frank A. Wolak. 1983. “Competition in Interregional Taxation:
The Case of Western Coal.” Journal of Political Economy 91(3):443–460.
Kolstad, Charles D. and Frank A. Wolak. 1985. “Strategy and Market Structure in Western Coal Taxation.” The Review of Economics and Statistics 67(2):239–249.
Kolstad, Charles D. and Frank A. Wolak. 1986. “Conjectural Variation and the Indeterminacy of Duopolistic Equilibria.” The Canadian Journal of Economics / Revue
canadienne d’Economique 19(4):656–677.
Kopal, Christopher. 2007. “Entwicklung und Perspektive von Angebot und Nachfrage am
Steinkohlenweltmarkt.” Zeitschrift für Energiewirtschaft 31(1):15–34.
Laitner, John. 1980. “"Rational" Duopoly Equilibria.” The Quarterly Journal of Economics 95(4):641–662.
Larrea,
Carlos. 2010. “Yasuni-ITT: An Initiative to Change History.”
http://yasuni-itt.gob.ec/wp-content/uploads/initiative\_change\
_history\_sep.pdf, Government of Ecuador.
149
URL
Bibliography
Leyffer, Sven and Todd Munson. 2010. “Solving Multi-Leader-Common-Follower Games.”
Optimization Methods and Software 25(4):601–623.
Li, Raymond. 2008. “International Steam Coal Market Integration.” Department of Economics, Macquarie University, Sidney, Australia.
Li, Raymond, Roselyne Joyeux and Ronald D. Ripple. 2010. “International Steam Coal
Market Integration.” The Energy Journal 31(3).
Lin, Bo-qiang and Jiang-hua Liu. 2010. “Estimating Coal Production Peak and Trends
of Coal Imports in China.” Energy Policy 38(1):512 – 519.
Lindh, Thomas. 1992. “The Inconsistency of Consistent Conjectures : Coming back to
Cournot.” Journal of Economic Behavior & Organization 18(1):69–90.
Lise, Wietze and Benjamin F. Hobbs. 2008. “Future Evolution of the Liberalised European Gas Market: Simulation Results with a Dynamic Model.” Energy 33(7):989–1004.
Lise, Wietze, Benjamin F. Hobbs and Frits van Oostvoorn. 2008. “Natural Gas Corridors Between the EU and its Main Suppliers: Simulation Results with the Dynamic
GASTALE model.” Energy Policy 36(6):1890 – 1906.
Liu, Gang. 2004. “Estimating Energy Demand Elasticities for OECD Countries. A Dynamic Panel Approach.” Discusssion Papers Statistics Norway 373: Oslo.
Lucarelli, Bart. 2010. “The History and Future of Indonesia’s Coal Industry: Impact of
Politics and Regulatory Framework on Industry Structure and Performance.” PESD
Stanford Working Paper #93.
Luppens, James A., Timothy J. Rohrbacher, Lee M. Osmonson and M. Devereux Carter.
2009. “Coal Resource Availability, Recoverability and Economic Evaluations in the
United States - A Summary.” In: Pierce, Brenda S. and Kristin O. Dennen (eds.), The
National Coal Resource Assessment Overview, USGS, U.S. Geological Survey Professional Paper 1625-F, chapter D.
Lynch, Michael C. 2003. “The New Pessimism about Petroleum Resources: Debunking
the Hubbert Model (and Hubbert Modelers).” Minerals and Energy 18(1):21–32.
Makowski, Louis. 1987. “Are ‘Rational Conjectures’ Rational?” The Journal of Industrial
Economics 36(1):35–47.
Mathiesen, Lars, Kjell Roland and Knut Thonstad. 1987. “The European Natural Gas
Market: Degrees of Market Power on the Selling Side.” In: Golombek, Rolf, Michael
Hoel and Jon Vislie (eds.), Natural Gas Markets and Contracts, North-Holland, Contributions to Economic Analysis, 27–58.
Michielsen, Thomas. 2011. “Brown Backstops versus the Green Paradox.” CentER Discussion Paper No. 2011-110: Tilburg University.
150
Bibliography
Minchener, Andrew. 2007. Coal supply challenges for China. London: IEA Clean Coal
Centre.
Minchener, Andrew. 2009. Future Coal Supply Prospects. London: IEA Clean Coal Centre.
Ministry of Coal. 2005. “Report from the Expert Committee on Road Map for Coal Sector
Reforms.” http://coal.nic.in/expertreport.pdf, Government of India, New Delhi.
Mohr, S.H. and G.M. Evans. 2009. “Forecasting Coal Production until 2100.” FUEL
88(11):2059–2067.
National Bureau of Statistics of China. 2007. “China Energy Statistical Yearbook 2007.”
China Statistics Press, Beijing.
Nelson, Carl H. and Bruce A. McCarl. 1984. “Including Imperfect Competition in Spatial
Equilibrium Models.” Canadian Journal of Agricultural Economics/Revue canadienne
d’agroeconomie 32(1):55–70.
New South Wales Department of Primary Industries. 2009. “Summary of NSW coal
statistics.” http://www.dpi.nsw.gov.au/minerals/resources/coal.
Patzek, Tadeusz W. and Gregory D. Croft. 2010. “A Global Coal Production Forecast
with Multi-Hubbert Cycle Analysis.” Energy 35(8):3109–3122.
Paulus, Moritz and Johannes Trüby. 2011a. “Coal Lumps vs. Electrons: How do Chinese
Bulk Energy Transport Decisions Affect the Global Steam Coal Market?” Energy
Economics 33(6).
Paulus, Moritz and Johannes Trüby. 2011b. “Market Structure Scenarios in International
Steam Coal Trade.” EWI Working Paper 2011/02.
Paulus, Moritz, Johannes Trüby and Christian Growitsch. 2011. “Nations as Strategic
Players in Global Commodity Markets: Evidence from World Coal Trade.” EWI Working Paper 2011/04.
Peng, Wuyuan. 2011. “Coal Sector Reform and its Implications for the Power Sector in
China.” Resources Policy 36(1):60–71.
Perry, Martin K. 1982. “Oligopoly and Consistent Conjectural Variations.” The Bell Journal of Economics 13(1):197–205.
Pindyck, Robert S. 1981. “Models of Resource Markets and the Explanation of Resource
Price Behaviour.” Energy Economics 3(3):130 – 139.
Platts. 2008. “Methodology and Specifications Guide - Coal.” http://www.platts.com.
Platts. 2009. “Methodology and Specifications Guide - Coal, Update May 2009.” http:
//www.platts.com.
151
Bibliography
Queensland Department of Mines and Energy. 2009. “Coal statistics.” http://www.dme.
qld.gov.au/mines/coal_statistics.cfm.
Rademacher, Maggi. 2008. “Development and Perspectives on Supply and Demand in the
Global Hard Coal Market.” Zeitschrift für Energiewirtschaft 32(2):67–87.
Rademacher, Maggi and Raphael Braun. 2011. “The Impact of the Financial Crisis on the
Global Seaborne Hard Coal Market: Are there Implications for the Future?” Zeitschrift
für Energiewirtschaft 35:89–104.
Ralph, Daniel and Yves Smeers. 2006. “EPECs as Models for Electricity Markets.” IEEE,
2006 IEEE PES Power Systems Conference and Exposition, 74–80.
Ritschel, Wolfgang and Hans-Wilhelm Schiffer. 2005. “World Market for Hard Coal, 2005
Edition.” RWE Power.
Ritschel, Wolfgang and Hans-Wilhelm Schiffer. 2007. “World Market for Hard Coal, 2007
Edition.” RWE Power.
Rogner, H.-H. 1997. “An Assessment of World Hydrocarbon Resources.” Review Energy
Environment 22:217–262.
Rui, Huaichuan, Richard K. Morse and Gang He. 2010. “Remaking the World’s Largest
Coal Market: the Quest to Develop Large Coal-Power Bases in China.” Stanford University PESD Working Paper #98.
Rutherford, Thomas F. 1995. “Extension of GAMS for Complementarity Problems Arising in Applied Economic Analysis.” Journal of Economic Dynamics and Control
19(8):1299–1324.
RWE. 2008. “Facts and Figures - Update May 2008.”
Samuelson, Paul A. 1952. “Spatial Price Equilibrium and Linear Programming.” American Economic Review 42(3):283–303.
Schernikau, Lars. 2010. Economics of the International Coal Trade: The Renaissance of
Steam Coal. Springer Netherlands.
Schmalensee, Richard. 1982. “Another Look at Market Power.” Harvard Law Review
95(8):1789–1816.
Sherali, Hanif D. 1984. “A Multiple Leader Stackelberg Model and Analysis.” Operations
Research 32(2):390–404.
Sijm, J.P.M., S.J.A. Bakker, Y. Chen, H.W. Harmsen and W. Lise. 2005. “CO2 Price
Dynamics: The Implications of EU Emissions Trading for the Price of Electricity.”
ECN Discussion Paper.
152
Bibliography
Sinn, Hans-Werner. 2008. “Public Policies Against Global Warming: a Supply Side Approach.” International Tax and Public Finance 15(4):360–394.
Smeers, Yves. 2008. “Gas Models and Three Difficult Objectives.” CORE Discussion
Papers 2008009, Université catholique de Louvain, Center for Operations Research
and Econometrics (CORE).
Stackelberg, Heinrich. 1934. Marktform und Gleichgewicht. Vienna: Julius Springer.
Statistics Canada. 2009. “Energy Statistics Handbook - Third quarter 2008.” http://
www.statcan.gc.ca/pub/57-601-x/57-601-x2008003-eng.pdf.
Stern, Nicholas and James Rydge. 2012. “The New Energy-industrial Revolution and
International Agreement on Climate Change.” Economics of Energy & Environmental
Policy 1(1).
Takayama, T. and G.G. Judge. 1964. “Equilibrium Among Spatially Separated Markets:
A Reformulation.” Econometrica 32(4):510–524.
Tirole, Jean. 1999. Industrieökonomik. Wolls Lehr- und Handbücher der Wirtschafts- und
Sozialwissenschaften, München, Wien: Oldenbourg, 2nd edition.
Tirole, Jean. 2012. “Some Political Economy of Global Warming.” Economics of Energy
& Environmental Policy 1(1).
Toussaint, Allen. 1966. “Working in the Coal Mine.” Lyrics and music. New Orleans.
Tu, Jianjun. 2011. “Industrial Organization of the Chinese Coal Industry.” Stanford University PESD Working Paper #103.
Ulph, A. M. and G. M. Folie. 1980. “Economic Implications of Stackelberg and NashCournot Equilibria.” Journal of Economics 40:343–354.
VDKI. 2006. Jahresbericht 2006. Verein der Kohlenimporteure.
Vives, Xavier. 2000. Oligopoly Pricing - Old Ideas and New Tools. Cambridge, Mass.:
MIT Press.
Warell, Linda. 2006. “Market Integration in the International Coal Industry: A Cointegration Approach.” Energy Journal 27(1):99–118.
Xu, Huifu. 2005. “An MPCC Approach for Stochastic Stackelberg-Nash-Cournot Equilibrium.” Optimization 54(1):27–57.
Yang, Chin W., Ming Y. Hwang and Soong N. Sohng. 2002. “The Cournot Competition
in the Spatial Equilibrium Model.” Energy Economics 24(2):139–154.
Zaklan, Aleksandar, Astrid Cullmann, Anne Neumann and Christian von Hirschhausen.
2009. “The Globalization of Steam Coal Markets and the Role of Logistics: An Empirical Analysis.” DIW Discussion Paper 956: Berlin.
153
Bibliography
Zola, Emile. 1885. Germinal. Paris: G. Charpentier.
Zwart, Gijsbert and Machiel Mulder. 2006. “NATGAS: A Model of the European Natural
Gas Market.” CPB Memorandum 144.
154