Beruflich Dokumente
Kultur Dokumente
Editors
Energy Economics
and Financial Markets
123
Editors
Andr Dorsman
Department of Finance
VU University Amsterdam
Amsterdam
The Netherlands
Wim Westerman
Faculty of Economics and Business
Economics, Econometrics and Finance
University of Groningen
Groningen
The Netherlands
John L. Simpson
Curtin Business School
School of Economics and Finance
Curtin University
Perth, WA
Australia
ISBN 978-3-642-30600-6
DOI 10.1007/978-3-642-30601-3
ISBN 978-3-642-30601-3
(eBook)
Foreword
vi
Foreword
gain useful insights into the fundamental economic forces determining the global
energy supply and demand. It addresses transitional issues, and also explains the
connections of these to the related financial derivatives and primary markets.
GENs practice touches on many of the subjects addressed in this book: improving
the intelligibility of energy through market model analyses, process implications,
and hard-core performance indicators. Our tagline is Adding Value to Energy. For
this book specifically, I would like to stress the way its articles add value to the
evolution of the energy markets. Being a co-production between a wealth of scientists and practitioners, it supports the creation of a more uniform framework for the
different operating mechanisms in the economics and finance-related parts of the
energy sector. It is worth reading and holds an invitation from the authors, but
definitely also from me personally, to add to the insights offered and contributed to
building the firm knowledge base that will guide us to the future.
Sam Collot dEscury
CEO GEN
Contents
Part I
Part II
5
13
31
49
73
91
107
vii
viii
Contents
Part III
129
159
10
Part IV
175
11
197
12
215
13
235
Chapter 1
Abstract Energy issues feature frequently in the economic and financial press. It
is argued that the importance of energy production, consumption and trade and
raises fundamental economic issues that impact the global economy and financial
markets. Specific examples of daily energy issues stem from various countries and
can often be related to economics and finance. It is shown that energy economics
and financial market research issues can be grouped under the themes of supply
and demand, environmental impact and renewables, energy derivatives trading, as
well as finance and energy.
Energy
J. L. Simpson (&)
School of Economics and Finance, Curtin Business School,
Curtin University, Hayman Road, Bentley, Perth, WA 6845 Australia
e-mail: john.simpson@cbs.curtin.edu.au
W. Westerman
Faculty of Economics, Business, Econometrics and Finance,
University of Groningen, Nettelbosje 2, 9747 AE Groningen,
Groningen, The Netherlands
e-mail: w.westerman@rug.nl
A. Dorsman
Department of Finance, VU University, De Boelelaan 1105,
1081 HV Amsterdam, The Netherlands
e-mail: a.b.dorsman@vu.nl
J. L. Simpson et al.
1.1 Introduction
There are many critical areas of interest that relate either generally or specifically
to fossil fuels and alternatives, energy efficiency, energy independence and
security, energy safety issues, climate change, sustainability and renewables, the
transportation of energy resources, connecting energy suppliers and consumers,
electricity generation and so on. This book in some way touches all of those broad
issues explicitly or implicitly. The book cannot deal with all of the current energy
issues in detail, but it does represent a genuine effort to draw attention in applied
research in several important areas of energy economics and financial markets.
This chapter begins with a discussion of the importance of energy production,
consumption and trade and raises fundamental macro-economic issues that impact the
global economy and financial markets. The chapter then provides specific examples of
daily energy issues from various countries and relates them to economics and finance.
Energy economics and financial market research issues are related to the themes of
supply and demand, environmental impact and renewables, energy derivatives trading,
as well as finance and energy. A discussion of headlines of the several chapters of the
book shows the relevance of academic research in the area for energy economics and
financial market researchers, practitioners and policy makers alike.
For the data in this section see Global Energy Statistical Yearbook (2011), http://
yearbook.enerdata.net.
J. L. Simpson et al.
of electricity between 1998 and 2003 did not incentivize investment, but this
perhaps is changing. In 20032004 with greater demand, higher prices, and spare
capacity shortages blackouts actually occurred in the US and in Europe.
The investment problems need to be resolved given that electricity investment is
often not perceived to be attractive at the time. The investment will probably need
to be made more attractive with some form of government involvement. Electricity
investment is capital intensive, there is a long investment time frame, it has low
expected rates of return and such investment needs to compete with other areas of
financial markets which are more attractive in risk and return characteristics.
This brings on discussion about the role of the financial markets in the energy
sector under the current circumstances. The energy markets are now the place to be.
After the bond markets in the 1950s, the stock markets in the 1960s and the 1970s, the
option market in the 1980s and the 1990s of the last century it is now the turn of the
energy markets. Bonds, stocks, and options are products created in the minds of
people. The prices of one of these products should be universally more or less the
same. Deviations in prices of the same product are due to imperfections. That is not
true for energy and energy-related products. The price of these products depends on
time and location. On top of that, transport of energyfrom production place to
consumption placeexpends energy. Financial models that are true for imaginary
goods are too simple for energy. Armed with the knowledge of the markets of
imaginary goods it is important to expand knowledge of the energy markets.
Due to the fact that consumers and producers of energy are not equally distributed
over the world, it is logical that there is an apparent value shift from consuming
countries to producing countries. However, production of oil and gas is not unlimited.
The welfare obtained by oil and gas wells will disappear when the production of fossil
energy substantially decreases. Energy producing countries look at alternatives, such
as renewables (solar and wind energy) or creating financial institutions. After the
financial crisis 20072008 there were doubts about the solvency of American and
European banks. It is unusual to many, that, over the period of the global financial
crisis, there was little discussion about the viability of Russian Banks.
Such developments in the real world will drive research. The above events have
an important impact on oil supply at another point when issues relating to peak oil
keep coming to the fore-front of thinking in those markets. The events in the Gulf
of Mexico also triggered questions relating to the prices and price movements in
oil markets and also in individual oil companies and supply and demand aspects,
but overwhelmingly, questions arise on the real world issue of environmental
damage, sustainability, the desirability of fossil fuels in general in preference to a
more rigorous development of renewable energy resources.
Another timely real world issue may be one where there has not been a large
amount of focus. For example, again in the United States the issue of the regulation
and control of fracking wastewater has more recently arisen. The disposal of the
water from hydraulic fracturing in shale gas and coal-bed methane operations has
raised not only positive news about alternative cleaner burning fossil fuels and
methods of extraction that add to the supply of energy resources in an era of
rapidly increasing demand, but it is very important that the environmental impact
be considered and in this case also the impact on water resources.
Alternatives for fossil energy include renewables such as wind and solar
energy. However, these energy sources are also permeated with challenges. See for
example the political and economic effect of the recent financial problems of the
US solar firm Solyndra, which experienced problems despite a substantial
government guarantee from the Democratic Obama administration. Apart from the
blow to renewables advocates in practice as well as in research, the Republican
opposition appear to feel that they have gained some political advantage to help
them push hard for the desirability of an addiction to fossil fuels.
Energy is a commodity, which means that time and location matter. As a consequence
of location differences between producers and consumers of fossil energy, transport of
energy is a major economic activity. A hot topic refers to the economics of pipeline gas
when a specific issue is raised at a time in the global economy and certainly in the
developed economies where interest rates are low and stock markets are volatile, but weak.
Yet in taxation effective environments the attractiveness of low risk, low return pipeline
gas might be an attractive investment in some countries where it is appropriate. Still, this
would probably suggest a need for study in areas that are as yet not well exploited.
An issue for Australia and other fossil fuel exporters is the important need to
diversify exports away from an excessive reliance on only two or three importing
countries. On the importers side it is important for energy importers such as countries
in Western and Eastern Europe and also Turkey to consider the diversification of their
pipeline gas supplies from one or two countries that from all possible considerations
have to be considered as high political risk countries. Stability and security of energy
supply is an important issue that may interest energy economists amongst others.
For countries such as Australia, an increasingly important exporter of natural
gas,2 the issues considered by many are whether or not Australia should sell the
It is expected that by 2020 Australia will overtake Qatar as the largest global exporter of LNG.
See Forbes (2012).
J. L. Simpson et al.
Part III examines the dynamics of energy derivatives trading. The fourth part deals
with issues on the intersection of finance and energy. In this way the book itself is
organized into a total of twelve topical chapters.
Within Part I, which deals with supply and demand, one chapter is about energy
security in Asia, highlighting inter-country differences on specifically natural gas.
Helen Cabalu and Cristina Alfonso propose a composite gas supply security index
with four indicators: a gas consumption efficiency ratio, a gas import dependency
ratio, a gas production versus consumption ratio, and a geopolitical risk index. The
composite index is used to describe the situation in six important Asian countries.
China appears to be the least vulnerable country, whereas Thailand has become the
most vulnerable country.
Another chapter covers aspects of energy pricing and the pricing of finance for
buyers of energy resources, using country risk ratings. Author John L. Simpson
assumes that natural gas export returns represent the change in the amount of
export finance that might be required as buyer credit. Using country risk ratings, a
risk premium is ascribed to this buyer credit. Evidence suggests that Chinese buyer
credit from Australia is subject to a substantially larger credit risk than US buyer
credit from the same country. It is also shown that long-term equilibrium relationships exist in the early 2000s. The author therefore concludes that country risk
needs to be considered in loan pricing for gas exports.
The final chapter of the section deals with the drivers of energy demand in
developing countries. Here, Ayhan Kapusuzoglu and Mehmet Baha Karan suspect
a growing mismatch between energy demand and supply, which underlines the
need for research. They study 19712007 data for 30 developing countries. Energy
consumption is related to measures of not just gross domestic product (GDP), but
also rural population, total population, consumer prices (CPI) and CO2 emission.
The study signals common relationships in various directions between energy
consumption and the other factors studied.
The second part of the book deals with environmental issues and renewables.
Specifically investigated in the first chapter are renewable energy production
capacity and consumption, and their effects on economic growth and global
warming. Henk von Eije, Steven von Eije and Wim Westerman study global
relationships between gross domestic product (GDP) growth, CO2 emissions, fossil
fuel consumption, renewable energy consumption and also renewable energy
production variables. The authors show that renewable energy production reduces
both fossil fuel use and economic growth in the long run.
The following chapter is written by Tony Owen. It deals with carbon pricing
instruments aimed at reducing the demand for power and stimulating low-carbon
power generation technologies. Both emission trading schemes and carbon taxes
are theoretically least cost economic instruments if it comes to tackling climate
change. Factors that can in practice lead to significant levels for one instrument
relative to the other include transparency, operating costs, public acceptability,
dynamic efficiency, revenue and distributional issues, as well as international
harmonization.
J. L. Simpson et al.
The final chapter of the second part deals with emissions trading and stock returns,
with evidence taken from the European steel and construction industries. Jeroen
Bruggeman and Halit Gonenc show that price changes of European Union emission
allowances are only positively correlated with combustion industry stock returns in
the period 20052007 (Phase 1). In the years 20082010 (the first part of Phase 2), a
positive relationship is found in the steel industry as well. The authors also show that
emission allowances exposures are independent of firms characteristics.
Part III examines the trading dynamics of the energy derivatives. It deals in one
chapter with energy derivatives market dynamics. Specifically, Don Bredin,
amonn Ciagin and Cal Muckley study EU Emissions Trading Scheme (ETS)
options and futures market dynamics in the period 20052011. The authors show
that the EU ETS derivatives markets have only matured since the start of Phase 2.
As compared to the West Texas Intermediate crude oil market, spot/future
correlations, term structures and option volatility smiles and surfaces are behaving
quite similar in these markets over time.
Then the part moves to a chapter on the dynamics of the spot and futures
markets in crude oil by zgr Arslan-Ayaydin and Inna Khagleeva. The authors
find theoretical support for the view that future prices of crude oil are equal to the
spot price of crude oil, cost of carry and the convenience yield. They also show
theoretical modeling support for the view that future crude oil prices have no
predictive power on the spot markets. The authors support the latter view with an
empirical analysis of Western Texas Intermediate crude oil markets over the
period January 1986December 2011.
Lastly, the third part has a chapter on natural gas spot and futures markets. John
L. Simpson investigates the issue of decoupling of oil and gas or rather the
influences on these prices of global and domestic economic and market factors.
Liberalization progress has been made in both the US and the UK. Yet, the author
finds that US markets are more decoupled than UK markets, with domestic gas
price factors dominating global oil price factors in the determination of the future
spot gas price. He therefore supports the view that US deregulation policies have
been more effective than those in the UK.
The final part of the book has the theme of finance and energy. The first chapter
is written by Andr Dorsman, Andr Koch, Menno Jager and Andr Thibeault. It
covers the effects of the addition of oil to a portfolio of US stocks and bonds.
Indeed, when studying the period 19892010, inclusion of oil as an asset makes
the efficient frontier and the market portfolio change. While oil does also improve
the risk-return trade-off for investors, it only presents a hedge for bondholders and
not for stockholders. The authors therefore conclude that oil is not a safe haven for
stockholders and bondholders.
This is followed by a chapter on the imperfections in electricity networks for a
number of European countries, written by Andr Dorsman, Geert-Jan Franx and
Paul Pottuijt. Energy networks (grids) used to be nationally organized, but the
national grids are increasingly linked by market coupling. In this way, interconnector capacities are automatically allocated in a way that minimizes electricity
price differences. The authors show with evidence from Scandinavia (South), The
Netherlands, Belgium and France that market price differences indeed do diminish
during the period January 2005March 2011.
The final chapter by Bill Dimovski investigates the pricing of initial public
offerings of 158 energy companies in Australia from January 1994 to December
2010. Whilst the average underpricing returns are substantial (22 %), negative
returns are not rare. The underpricing returns negatively correlate with the amount
of equity capital raised and the engagement of underwriters. The use of underwriter
options increases the underpricing returns though. The author also finds that underpricing returns do not differ between before and after the global financial crisis
of 20072008.
The chapters selected for this book are genuine academic pieces of writing.
Such chapters investigate some of the major energy economics issues of the day.
The authors have identified topical research problems and have explained the
reasons why finding a solution to those problems is important. The authors have
followed academic practice in the provision of a relevant theoretical and literature
base from which they could derive their models and hypotheses. They have
employed conventional statistical and econometric analysis ranging from basic
descriptive analysis to advanced time series techniques to arrive at statistically
significant findings that add in either a minor or a major way to the body of
knowledge about energy economics and financial markets.
1.5 Conclusion
Various timely energy issues may attract interest. For example, inter alia, the
explosion aboard the Deepwater Horizon rig in 2010, the regulation and control
of fracking wastewater, subsidizing firms involved with renewable energy,
investments in gas pipelines, risks on energy supplier reliance, foreign equity
investment in the energy sector, energy source trade-offs in densely populated
areas and the timing of equity sales of energy firms. It is beyond the scope of this
book to cover all of these most interesting areas, yet the book directly or indirectly
touches upon these kinds of issues.
Energy economics and financial market research problems can be related to
various themes. This book embraces energy economics research questions from all
over the world, written up by a diverse team of authors. The book explores the
themes of supply and demand, environmental issues and renewables, energy
derivatives trading, as well as finance and energy. This discussion of headlines of
the several chapters of the book has been aimed to show the links between the
individual contributions as well as the relevance of academic research in the area
for energy economics and financial market researchers, practitioners and policy
makers alike. It is hoped that the readers both benefit from and enjoy the read.
10
J. L. Simpson et al.
References
Energy Insights. (2011). Energy Supply and Demand. Retrieved from: http://www.energyinsights.
net/content/articles/energysupplydemand.htm.
Forbes, A. (2012, February). The exciting future of LNGAnd how it will transform the global
gas market, European Energy Review. Retrieved from: http://www.europeanenergyreview.eu/
site/pagina.php?id=3497.
Global Energy Statistical Yearbook. (2011). Grenoble, France: Enerdata. Retrieved from: http://
yearbook.enerdata.net.
Part I
Chapter 2
Abstract Natural gas consumption in the future is expected to increase due to its
low environmental impact, ease of use and rise in the number of natural gas-fired
power plants. This chapter measures natural gas supply security in six Asian
economies including Japan, Korea, China, India, Singapore and Thailand from
1996 to 2009. Disruptions to long term security of supply can be caused by
inadequate investments in production and transmission infrastructure, lack of
supply diversity and import dependency. A composite gas supply security index is
derived from four indicators of security of gas supply, with a higher index indicating higher gas supply vulnerability. Results show that China and India are the
least vulnerable in terms of natural gas security because of their significant
domestic gas production and small share of gas in the energy mix. Thailand is the
most vulnerable among the countries studied due to its high reliance on natural gas
to power its electricity generation industry as well as its greater exposure to
geopolitical risks. With these analyses, governments can target possible sources of
supply disruptions and mitigate their effects. Diversification is highly encouraged
to spread the risk across different import and energy sources.
H. Cabalu (&)
School of Economics and Finance, Curtin Business School, Curtin University,
GPO Box U1987WA, Perth 6845, Australia
e-mail: h.cabalu@curtin.edu.au
C. Alfonso
Centre for Research in Energy and Minerals Economics (CREME), Curtin Business School,
Curtin University, GPO Box U1987 6845 Perth, WA, Australia
e-mail: c.alfonso@curtin.edu.au
13
14
2.1 Introduction
The oil shocks in the 1970s demonstrated how vulnerable the worlds economy
was to supply interruptions and price volatility. In addition, the recent increases in
energy prices, a steady rise in global energy demand, instability in energy producing regions and the threat of terrorist strikes against energy infrastructure have
significantly led to a growing concern over energy security. Any energy infrastructureoil, coal or natural gasis often vulnerable to disruption by insufficient
supply, accident or malice. Terrorism, technical mishap, or natural disasters that
damage the energy system could be nearly as devastating as a sizeable war.
Inadequate financial resources also increase vulnerability or insecurity by limiting
supply, transmission, and reliability while increasing prices of energy imports
adversely affect the macroeconomic balance of payments, contribute inflationary
pressures, and displace other consumption and investment because short-term
demand is inelastic. In the past, long term contracts between exporters and
importers have been an important element of security of supply. However, in
recent years long term contracts have not been an adequate assurance of uninterrupted deliveries.
Energy security has emerged as a major object of the energy policy agenda and
policy makers have engaged in a wide ranging debate over how best to address
future energy requirements. Along with this emergence, energy markets have
moved towards strengthening regional co-operation and energy supplies and
sources have become more diversified. There has also been a strong trend towards
shorter contract terms or a considerable decrease in the length of contracts caused
by either market-related or regulatory-related changes. Market changes due to
government regulatory initiatives and the creation of competitive markets have led
to this trend.
While many previous studies have focused on oil, this chapter provides
evidence on security of natural gas supply in selected Asian countries. With the
growing demand for gas, supply interruptions, increasing gas prices, transportation
and distribution bottlenecks, and a growing reliance on imports over longer
distances have rekindled a debate on gas security of supply. Extending the work by
Cabalu (2010), this chapter proposes a composite gas supply security index (GSSI)
which is derived as the root mean square of the scaled values of four security of
gas supply indicators, for the period 19962009. The four security of gas supply
indicators are interrelated and the GSSI derived from 1996 to 2009 provides a
trend in the composite quantitative measure of gas security by taking into account
the interactions and interdependences between the identified set of indicators. The
GSSI captures the sensitivity of the Asian economies to developments in the
international gas market, with a higher index indicating higher gas supply insecurity or vulnerability.
The existing literature does not identify a unique methodology that is factual,
objective, unbiased, transparent and accessible, to assess and quantify energy
security. However, it is important to provide metrics by evaluating a set of
15
parameters and indicators to assess overall natural gas supply security in the six
Asian economies of Japan, Korea, China, India, Singapore and Thailand, which
together account for almost 64 % of the total gas consumption in the AsiaPacific
region in 2010 (BP 2011). It is important for future policy making to benchmark
countries against quantified indicators and assess their gas security of supply
weakness. This chapter is divided into six sections. The sect. 2 provides a brief
background on the importance of energy security, particularly in natural gas and
includes a discussion of the vulnerabilities in the natural gas system. The sect. 3
reviews related literature on energy supply security, particularly focused on
identifying the various indicators used in the literature to indicate energy
vulnerability while Sect. 4 derives a composite gas supply security index for the
years 19962009 for the sample countries. Section 5 presents the results and
analysis and the final section concludes.
16
On the other hand, the Middle East (particularly Iran and Qatar) and Russia had
around 65 % of the worlds reserves and accounted for around 25 % of the demand
in 2010 (BP 2011).
In 2010, more than 11 % of the AsiaPacific primary energy consumption was
based on natural gas. Gas market requirements are mostly met through imports,
more than 85 % of which is LNG from Malaysia, Brunei, Indonesia, Australia and
the Middle East. Japan and Korea are almost entirely dependent on LNG imports
for their gas supplies. In Japan and Korea, imported gas exchanges are based on
long term contracts of 2025 years and indexation clauses where the gas price is
directly linked to the price of crude oil, including relatively strict clauses such as
take-or-pay clauses which require importers to pay for the gas even if their
deliveries are interrupted. In Australia and New Zealand, prices are set by gas-ongas or gas-on-coal competition (IAEE 2007; IEA 2007; BP 2011).
Short-term security of gas supply is the ability to maintain gas supply despite
exceptional demand and difficult supply conditions. Disruptions to supply may be
due to physical or economic factors. Physical disruptions can occur when gas supply
is exhausted or gas production is stopped. Economic disruptions can be caused by
dramatic gas price fluctuations which in turn, are due to physical disruptions or
unanticipated price changes associated with speculative reaction to potential
disruption. Long-term security of gas supply on the other hand, is the ability to ensure
that future gas demand can be met by a combination of domestic and imported gas
supplies. Disruptions to long term security of supply are caused by inadequate
investments in production and transmission infrastructure, lack of supply diversity
and risks associated with import dependency which are geopolitical in nature. Gasimporting countries have started to examine potential responses to disruptions to
ensure security of gas supply (Dolader 2003; Costantini et al. 2007).
17
imported energy sources, political stability in import sources, and the resource
base in import sources.
Similarly, Costantini et al. (2007) grouped indicators of supply security into
two categories: dependence, and vulnerability represented in physical and
economic terms. The distinction between dependence and vulnerability was made
and in their study, the physical dimension of dependence was represented by
indicators such as percentage share of net import of oil and gas in total primary
energy supply and the share of European oil and gas imports in world oil and gas
imports while the physical dimension of vulnerability was calculated in terms of
the degree of supply concentration in trade and production using the ShannonWeiner diversity index, percentage share of oil used in transportation, and
percentage share of electricity produced with gas. In terms of the economic
dimension of dependence and vulnerability, the value of oil and gas imports and
oil and gas consumption per dollar of GDP respectively, were estimated. These
indicators of the European energy system were analyzed under different energy
scenarios.
In a study by de Jong et al. (2007), a model was developed for reviewing and
assessing energy supply security in the European Union, on the basis of pre-agreed
criteria. It used two quantitative indicators and some qualitative considerations.
The first quantitative indicator, the Crisis Capability (CC) Index dealt with the risk
of sudden unforeseen short-term supply interruptions and the capability to manage
them. The second indicator, the Supply/Demand (S/D) Index covered present and
future energy supply and demand balances. Qualitative considerations included
multilateral measures for securing overall producer/consumer relations and safeguarding vulnerable transport routes for oil and gas.
A number of studies have focus on assessing energy vulnerability. Kendell
(1998) explores the meaning and value of measures of import vulnerability as
indicators of energy security, in particular, oil security in the United States. While
measures of oil import dependence showing the extent of a countrys imports may
be of interest, they offer a limited indication of energy security. Gupta (2008),
APERC (2007), and UNDP (2007) also examine the relative oil vulnerability of oilimporting countries on the basis of various factors. Using a principal component
technique, individual indicators such as domestic oil reserves relative to total oil
consumption, geopolitical oil risk, oil intensity, cost of oil in national income and
the ratio of oil consumption in total primary energy consumption are combined into
a composite index of oil vulnerability. Percebois (2007) clarifies the distinction
between vulnerability and energy dependence and presents a coherent set of indicators including import concentration, level of energy import value in output, risk
of blackout in the electricity sector, price volatility, exchange rates, and industrial
and technological factors that are used to analyze energy vulnerability. Gnansounou
(2008) defines a composite index of energy demand/supply weaknesses as a proxy
for energy vulnerability. The index is based on several indicators such as energy
intensity, oil and gas import dependency, CO2 content of primary energy supply,
electricity supply weaknesses and non-diversity in transport fuels. The assessment
of the composite index is applied to selected industrialized countries. In 2008, the
18
World Energy Council (2008) identified threats to the European economy which
could lead to potential energy crises and suggested solutions for facing related key
challenges. The study also developed a number of indicators to assess the level of
different types of vulnerability, as well as the overall vulnerability of a country or
region, including threats to physical disruption and higher energy prices.
The design of a composite index of energy security has been undertaken in
previous studies. A composite vulnerability index was developed by the World
Energy Council (2008) to benchmark and monitor European countries respective
efforts to cope with long-term energy vulnerability. Similarly, de Jong et al. (2007)
designed state-of-the-art indices of energy security risk (i.e., the Crisis Capability
Index and Supply/Demand Index) which are oriented towards a comprehensive
and analytical representation of the energy supply chain. However, the shortcoming of these approaches is the use of subjective-opinion-dominated weighting
systems and scoring rules where the weights and the rules are based on expert
judgments. In response to this shortcoming, Gnansounou (2008) proposes an
alternative method which is objective-value-oriented and statistics-based.
Gnansounou defines the composite index as the Euclidean distance to the best
energy security case represented by the zero point. The Euclidean distance is
standardized in order to get a value between 0 and 1. Following the more objective
methodology proposed by Gnansounou, Cabalu (2010) develops a composite gas
supply security index for selected Asian countries for the year 2008. This chapter
extends this previous study by calculating an annual gas supply security index for
the period 19962009 for net gas importing countries in Asia.
GCj
:
GDPj
The gas intensity of GDP of country j (G1j ) is measured as the ratio of total
natural gas consumed in country j (GCj ) to GDP of country j (GDPj ) and expressed
19
as cubic meters per unit of GDP or m3/GDP. The countrys output of goods and
services is measured by inflation-adjusted GDP.
The relative indicator for country j associated with G1 (u1j ) is estimated as:
u1j
G1j MinG1
:
MaxG1 MinG1
The relative indicator, u1j results in projection of G1j in the interval [0, 1].
A low value of u1j means that country j is less vulnerable or less insecure to supply
shocks compared to other countries in the study.
G2 is expressed as the ratio of net imported gas consumption to total primary
energy consumption. Net gas import dependency (G2 ) is calculated as:
G2j
GMj
:
TPECj
The gas import dependency of country j (G2j ) is represented by the ratio of net
imports of natural gas in country j (GMj ) to total primary energy consumption in
country j (TPECj ) expressed as a percentage.
Similarly, the relative indicator for country j associated with G2 (u2j ) is estimated as:
u2j
G2j MinG2
:
MaxG2 MinG2
The above adjustment transforms the indicator to the [0, 1] interval with the
value of 0 being assigned to the country with the lowest value of the selected
security of supply indicator and least vulnerable and the value 1 is assigned to the
country with the highest value of the selected indicator and hence most vulnerable.
G3 is measured as the ratio of domestic gas production to total domestic gas
consumption. Domestic production is a better indicator of the countrys capacity to
cope with short-term supply disruption than domestic reserves as production
excludes gas from stranded reserves which cannot be tapped immediately. Ratio of
domestic gas production to total domestic gas consumption (G3 ) is calculated as:
G3j
GPj
GCj
where GPj is domestic natural gas production in country j and GCj is total natural
gas consumed in country j.
This indicator, unlike the first two, is negatively related to gas supply vulnerability or security. A high value for G3 means that country j is less vulnerable or
less insecure to supply shocks compared to other countries in the study. To
accommodate this negative relationship, the relative indicator for country j associated with G3 (u3j ) is estimated as:
20
u3j
MaxG3 G3j
:
MaxG3 MinG3
The above adjustment transforms the indicator to the [0, 1] interval with the
value of 0 being assigned to the country with the highest value of the selected
security of supply indicator and least vulnerable and the value 1 is assigned to the
country with the lowest value of the selected indicator and hence most vulnerable.
G4 represents the exposure of an economy to political risk and is measured on
the basis of two factors: diversification of gas import sources and political stability
in gas-exporting countries. Geopolitical risk (G4 ) is largely determined by the
degree of diversification of gas import sources and the associated political s
tability of these sources. Jansen et al. (2004) suggests a methodology for quantifying such risk using the adjusted Shannon diversity index. The following formula
describes this index.
X
hi mi lnmi
S
i
where:
S = Shannon index of import flows of gas, adjusted for political stability in
exporting country i;
hi = extent of political stability in exporting country i, ranging from 0 (extremely
unstable) to 1 (extremely stable); and
mi = share of gas imports from country i in total gas imports.
The relative indicator for country j associated with G4 u4j is estimated as:
u4j
MaxG4 G4j
MaxG4 MinG4
21
High gas intensity of GDP results in larger adjustment costs and impacts on gas
supply security in the event of natural gas supply shocks. In addition, the higher
the share of imported gas in total energy demand the more vulnerable an economy
is to international gas developments. Diversification of supply sources, particularly
politically stable supply sources also reduces the risk and vulnerability to disruption. Dependence on domestically-sourced gas supply is preferred over
imported gas, as it avoids geopolitical uncertainties. In addition, the larger
domestic gas reserves relative to consumption or the larger domestic production
capabilities a country has, the smaller are the likely impacts on gas security.
It is difficult to quantify a countrys overall gas supply security using individual
indicators and it is even more difficult to synthesize different indicators. To
facilitate comparison or aggregation of several indicators, it may be better for these
to be expressed in the same units. To do this, for each of the four security indicators, a relative indicator ui , was estimated which was used to compute a composite indexthe gas supply security index (GSSI). The relative indicators are
estimated by using a scaling technique where the minimum value is set to 0 and the
maximum to 1. The value of 0 is assigned to the country with the least vulnerability or insecurity to supply disruptions and the value 1 is assigned to the country
with the most vulnerability to supply shocks. Table 2.2 presents calculations for
the relative indicators which are scaled values of the four security of supply
indicators.
The gas supply security index (GSSI) is derived as the root mean square of the
four relative indicators or scaled values of the four security of supply indicators.
s
P4
2
i1 /ij
GSSIj
4
The various relative indicators of gas security are interrelated and the GSSI
derived provides a composite quantitative measure of gas security by taking into
account the interactions and interdependences between the identified set of indicators. The GSSI captures the sensitivity of the Asian economies to developments
in the international gas market, with a higher index indicating higher gas supply
insecurity or vulnerability.
22
G3. China is rich in energy resources, particularly coal. Gas use in China is still
small and is significantly less than the use of other fossil fuels. Coal and oil
resources are utilized more extensively than natural gas for power generation and
industrial development purposes. Natural gas generally occupies a very small share
(4.0 % in 2010) in Chinas energy mix but is expected to double by 2030
(Komiyama et al. 2005; APERC 2008).
Chinas major gas fields are located in the western part of the country, making
transport to eastern demand centers difficult. The use of domestic gas production
was initially limited to areas near production sites such as in Sichuan, Liaoning
and Heilongjiang Provinces, where low cost gas is possible. However, with recent
increases in infrastructure investments on pipeline construction such as the West
East pipeline to transport inland domestic gas, demand for natural gas has
increased. Between 2005 and 2008, Chinas natural gas consumption increased by
23.8 % and it became one of the worlds top ten countries in terms of natural gas
consumption. This coincided with the period when China became relatively vulnerable. At the same time, LNG imports also started and its import dependence
increased rapidly due to a substantial rise in demand. This growth was driven
mainly by the increased use of gas for power generation, feed stock in chemical
fertilizer production and to operate oil and gas fields. Recent developments such as
the increased residential consumption due to penetration of city gas, together with
the urbanization of cities have also led to the significant increase in demand. In
addition, the Chinese government through policy and regulation has encouraged
the use of natural gas as a source of cleaner energy and a substitute for oil and
coal. While some of the rising demand will be fulfilled through increases in
domestic production, a large portion has come from pipeline and LNG imports.
Due to geographical accessibility, the small amount of imported LNG goes to
southern provinces along the coast like Guangdong and Fujian (Higashi 2009).
In 2010, China had 14 import sources compared to one import source in 2006.
However, most of the additional import sources are politically unstable which
explains Chinas relative poor performance on G4. China received its first-ever
LNG cargo in mid-2006 under a long-term contract with Australia. Australia
remains Chinas major source of LNG. Its second terminal in Fujian started
receiving cargoes from Indonesia in 2008. Another re-gasification terminal in the
Shanghai area started to import LNG from Malaysia in 2009. In the northern
inland areas of China, natural gas supply has been sourced from Qatar, Siberia,
Turkmenistan, Sakhalin and Sakha.
India ranks as the second less gas-vulnerable country in the sample. For the
period, 19962009, Indias natural gas security generally improved through time
as shown by a downward trend in its GSSI. The strength of this country is in G2
indicating a relatively low gas import dependency, and to a less extent in G1 for
having low gas intensity. In India, natural gas is a minor fuel in the overall energy
mix representing only 10.6 % of total primary energy consumption in 2010. In that
same year, Indias natural gas imports represent just over 2 % of its energy mix
and hence not reliant on imports. With coal as the major source of energy for
power generation, gas intensity of the economys GDP is low. However,
23
opportunities exist for gas in reducing regional air pollution and providing peaking
power. For the fertilizer sector, significant opportunities exist to import cheap
fertilizer; thereby reducing domestic gas demand, but political constraints will
likely buoy gas demand. Industrial consumers will benefit from increased supplies
of LNG to replace expensive liquid fuels, but cheap coal remains the dominant fuel
for many industrial applications (Jackson 2007).
However, Indias consumption of natural gas has risen faster than any other
fuel. The power and fertilizer industries are the key demand drivers for natural gas.
With domestic gas production only large enough to satisfy almost three-quarters of
its domestic gas consumption, Indias domestic natural gas supply is not likely to
keep pace with demand. Despite major new natural gas discoveries in recent years,
the country will have to import more, either via pipeline or as LNG. With an
increase in the demand for and supply of natural gas and with many new players
entering the market, the Indian governments Petroleum and Natural Gas Regulatory Board Act of 2006 has promoted competition among market players and
stabilized natural gas supply (Thacker 2006).
The bulk of Indias natural gas production comes from the western offshore
regions, especially the Mumbai High basin. The onshore fields in Assam, Andhra
Pradesh, and Gujarat states are also major producers of natural gas. In 2010, around
24 % of supply came from imported LNG. Currently, there are two re-gasification
terminals located on the Western coast of India, Dahej and Hazira. The Dahej
terminal is being supplied from Qatar under a long term contract, supplemented by
spot cargoes from other sources. A possible source of supply for the Hazira terminal
is Australias Gorgon LNG project. In 2012, India will have two more import
terminals, Dabhol-Ratnagiri and Kochi (EIA 2012).
One interesting result is Thailands natural gas supply vulnerability. Between
1996 and 1998, Thailand was ranked third least vulnerable country, after India.
However, between 1999 and 2009, Thailand became the most vulnerable among the
country sample. The sources of insecurity come from G1 to G4. Thailands heavy
reliance on natural gas to power 70 % of its electricity generation accounts for its
vulnerability in G1. Thailands high gas intensity is facilitated by a relatively wellestablished natural gas regulatory framework where third party access in gas transmission is quite developed and means the existence of non-discriminatory access to
the gas transmission system based on tariffs reflecting costs that provide a fair and
reasonable rate of return (Chandler and Padungkittimal 2008). In addition, despite
efforts to diversify sources of natural gas imports, a substantial amount comes from
Myanmar, increasing the countrys vulnerability due to its exposure to geopolitical
risks. The government aims to reduce Thailands dependency on natural gas for
power generation as stipulated in the Power Development Plan for 20072021 (EIU
2010). However, heavy government subsidy of electricity to residential users may
make this improbable in the next few years. Thailands strength lies in G3. Natural
gas production has improved due to several developments particularly at the Arthit
field in the Gulf of Thailand and the MalaysiaThailand Joint Development Area.
The construction of a third national gas pipeline in the Gulf of Thailand was finished
in 2007 further expanding natural gas production (EIU 2010).
24
India
Japan
Singapore
South Korea
Thailand
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
0.02
0.02
0.02
0.03
0.05
0.06
0.05
0.06
0.01
0.02
0.02
0.02
0.02
0.02
0.06
0.07
0.03
0.04
0.04
0.04
0.10
0.18
0.20
0.23
-0.15
-0.25
-0.11
0.15
0.00
0.00
0.69
2.42
11.17
12.75
13.41
16.96
4.27
4.48
13.32
14.31
6.77
9.02
11.92
12.81
0.00
2.35
8.07
7.86
G3 (%)
G4
108.79
111.01
104.51
96.02
100.00
100.00
91.75
75.68
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
100.00
92.21
74.83
78.81
0.49
0.00
0.00
0.67
0.06
0.00
0.02
0.71
0.99
1.09
1.13
1.27
1.04
0.74
0.62
0.34
0.47
0.85
0.94
1.10
0.00
0.00
0.00
0.00
Korea shows a relatively stable trend in its GSSI from 1996 to 2009. G1 and G4
are as its major strengths. To reduce the economys dependence on imported oil,
Korea introduced LNG in the 1980s to power its natural gas based city gas to the
residential sector. Since then, natural gas use has grown rapidly. Korea relies on
imported LNG for most of its natural gas, though it began producing a small
quantity from one offshore field in 2004. Korea is the second largest importer of
LNG worldwide accounting for 15 % of total LNG imports in 2010. The bulk of
Koreas LNG imports come from a much diversified group of sources which
explains its strength on G4. These 17 import sources include, among others, Qatar,
Indonesia, Malaysia, and Oman, with smaller volumes coming from Trinidad and
Tobago, Algeria, Nigeria, Belgium, Egypt, Brunei Darussalam, and Australia, and
occasional spot cargoes from elsewhere. Korean natural gas demand is shared
almost evenly between the electricity sector and the residential heating sector, with
a smaller amount consumed in petrochemical plants. With demand growing at an
average annual growth rate of 8.4 % between 2003 and 2010, Korea continues to
sign contracts for additional supplies, though most of the new LNG term contracts
25
Table 2.2 Relative indicators of security of supply in selected net gas-importing countries in
Asia
Country
Year
u1
u2
u3
u4
China
India
Japan
Singapore
South Korea
Thailand
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
1996
2000
2004
2009
0.09
0.03
0.04
0.06
0.46
0.26
0.21
0.20
0.00
0.00
0.00
0.00
0.08
0.02
0.25
0.24
0.14
0.12
0.16
0.13
1.00
1.00
1.00
1.00
0.00
0.00
0.00
0.00
0.01
0.02
0.06
0.14
1.00
1.00
1.00
1.00
0.39
0.36
0.99
0.84
0.61
0.71
0.89
0.75
0.01
0.20
0.61
0.46
0.00
0.00
0.00
0.00
0.08
0.10
0.12
0.21
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
0.08
0.17
0.28
0.18
0.53
1.00
1.00
0.47
0.94
1.00
0.98
0.44
0.05
0.00
0.00
0.00
0.00
0.32
0.45
0.73
0.55
0.22
0.17
0.13
1.00
1.00
1.00
1.00
in the past few years with Yemen, Malaysia and Russia include more flexibility for
the purchaser in terms of the ability to lower volumes if necessary. To ensure a
stable supply of gas, Korea is also increasing LNG storage capacity at its four
existing terminals (BP 2011).
Between 1996 and 2000, Singapores GSSI fluctuated but this trend was to
stabilize thereafter. Singapores gas security of supply profile is relatively weak on
G2 and G3 particularly during the tumultuous years. The absence of domestic gas
production combined with high domestic gas consumption makes Singapore relatively vulnerable to natural gas supply disruptions. Its consumption has risen
rapidly in recent years owing mostly to government programs aimed at reducing
carbon dioxide and sulphur emissions and encouraging the use of natural gas for
power generation and petrochemical production (EIA 2007). In 2008, natural gas
accounted for almost 15 % of Singapores total primary energy demand. Singapore
relies entirely on imports to meet its natural gas requirements which are mainly
26
India
Japan
Singapore
South Korea
Thailand
Fig. 2.1 Gas security of supply index of selected net gas-importing countries in Asia (19962009).
Source based on authors calculations
27
placed priority on the stable and secure supply of LNG, Japanese LNG buyers have
been in general paying a higher price than buyers in Europe or the United States
under the long-term take or pay contracts with rigid terms on volume and price.
Japan lacks a national pipeline network which could interconnect its consuming
areas. The possibility of a significant disruption at one LNG terminal in Japan
poses a potential supply vulnerability issue.
2.6 Conclusion
Many factors determine gas vulnerability of an economy. Domestic production,
gas efficiency usage, volume and sources of gas imports are very crucial in
determining an economys vulnerability. The analysis in this chapter highlights
inter-country differences in individual and overall indicators of gas security
which means that country differences exist with respect to vulnerability to natural gas supply disruptions. This implies that governments need to develop
policy responses that directly address individual countries weaknesses to enable
them to handle natural gas supply disruptions. Policy measures should reduce the
probability of supply disruptions occurring and the costs of disruptions. For
instance, India and China are relatively less vulnerable to supply disruptions
compared to other countries in the sample because of their significant domestic
gas production and small share of gas in their energy mix. This means that the
two countries do not have to rely on gas imports for energy generation.
Governments could implement various measures to better cope with supply
disruptions and significantly mitigate their effects. For instance, gas import
dependence has risks associated with price volatility, natural disaster, political
blackmail and terrorism. Imported gas supplies are either pipeline bound or sea
bound LNG. These transit options are both exposed to risks but it is the degree
of having viable alternative options that defines security of supply. When gas
imports depend dangerously on too few sources, it raises a concern whether this
is compatible with a sensible policy goal of gas supply security. This concern is
exacerbated when taking geopolitical considerations into account. Hence,
diversification of gas import sources is encouraged. Other diversification measures include fuel-switching and diversifying energy mix. Diversification in fuel
types and sources would reduce the costs of supply disruptions by spreading the
risks across different import and energy sources. As Percebois (2006) and
Reymond (2007) summed it up, a country which imports the majority of its gas
at a sustainable cost and ensures the security of supply by well-diversified and
politically-stable sources will not be vulnerable.
Governments also have the option of reducing overall gas dependence by
improving gas efficiency through research and development and adoption of
technologies that reduce gas consumption or increase the efficiency of gas use,
technologies that facilitate gas exploration and production, and alternative processing technologies such as gas to liquids plant. To enhance natural gas supply
28
Appendix
Table A.1 Political risk rating of selected gas-producing countries, selected years
Country
Political stability
Algeria
Australia
Bahamas
Bahrain
Belgium
Bermuda
Brunei
Cambodia
Canada
China
Denmark
Egypt
Equatorial Guinea
Finland
France
Germany
Hong Kong SAR, China
India
Indonesia
Iran
Ireland
Italy
Japan
Korea, North
Korea, South
Kuwait
Malaysia
Myanmar
Netherlands
Nigeria
Norway
1996
0
81
81
21
90
..
92
10
79
33
95
17
19
96
78
92
44
15
15
24
88
75
75
6
42
44
58
13
95
8
94
2000
4
90
86
49
82
69
91
22
85
36
96
34
41
100
74
91
75
25
6
32
97
78
83
41
48
64
52
9
99
10
97
2004
9
83
79
47
75
77
96
30
78
39
86
20
32
99
63
71
80
24
6
17
90
62
83
38
59
53
58
14
84
5
93
2009
13
76
78
41
74
72
95
25
85
30
86
25
43
96
66
77
82
13
24
8
84
65
83
35
52
59
47
7
83
4
92
(continued)
Political stability
Oman
Pakistan
Philippines
Qatar
Russia
Saudi Arabia
Singapore
Sweden
Switzerland
Thailand
Trinidad and Tobago
United Arab Emirates
United Kingdom
United States
Yemen
58
9
25
55
16
26
73
94
96
48
57
70
77
78
14
29
76
16
20
82
24
47
81
96
98
60
49
73
79
87
11
75
6
12
78
17
20
87
96
92
33
46
68
61
51
5
75
0
11
89
22
33
90
88
92
15
45
81
55
59
2
Source World Bank (2009). Political risk ratings range from 0 for high risk to 100 for low risk
References
APERC (Asia Pacific Energy Research Centre). (2007). A quest for energy security in the 21st
century: Resources and constraints. Tokyo, Japan: Institute of Energy Economics.
APERC (Asia Pacific Energy Research Centre). (2008). APEC energy overview 2007. Tokyo,
Japan: The Institute of Energy Economics.
British Petroleum (BP). (2010). BP statistical review of world energy June 2010. Retrieved from
http://www.bp.com/liveassets/bp_internet/globalbp/globalbp_uk_english/
reports_and_publications/statistical_energy_review_2008/STAGING/local_assets/
2010_downloads/statistical_review_of_world_energy_full_report_2010.pdf)
British Petroleum (BP). (2011). BP statistical review of world energy June 2011. Retrieved from
http://www.bp.com/statisticalreview
Cabalu, H. (2010). Indicators of security of natural gas supply in Asia. Energy Policy, 38, 218225.
Cabalu, H., & Manhutu, C. (2009). Vulnerability of natural gas supply in the Asian gas market.
Econ Anal Policy, 39(2), 255270.
Chandler, A., & Padungkittimal, N. (2008). Thai power sector. Retrieved from http://
www.ctlo.com/TPS-2.htm
Costantini, V., Gracceva, F., Markandya, A., & Vicini, G. (2007). Security of energy supply:
Comparing scenarios from a European perspective. Energy Policy, 35, 210226.
de Jong, J., Maters, H., Scheepers, M., & Seebregts, A. (2007). EU standards for energy security
of supply: Updates on the crisis capability index and the supply/demand index quantification
for EU-27. The Netherlands: Energy Research Centre of the Netherlands and Clingendael
International Energy Programme.
Dolader, J. (2003, June). Gas security of supply in a liberalised market. Paris: IEA.
Economist Intelligence Unit. (2010, November). Thailand: Energy report. Retrieved from http://
www.eiu.com/index.asp?layout=ib3Home&pubtypeid=1142462499
Energy Market Authority (EMA). (2010). LNG terminal development. Singapore: EMA.
Retrieved from http://www.ema.gov.sg/page/21/id:50/
Energy Information Administration (EIA) (2007). Energy Profile of Singapore, in Encyclopaedia
of Earth.Retrieved from http://www.eoearth.org/article/Energy_Profile_of_Singapore
30
Energy Information Administration (EIA). (2011). Annual energy outlook 2011: With projections to
2035. Washington, DC: EIA. Retrieved from http://www.eia.gov/forecasts/aeo/pdf/0383(2011).pdf/
Energy Information Administration (EIA). (2012). Energy Profile of India, in Encyclopaedia of
Earth. Retrieved from http://www.eoearth.org/article/Energy_profile_of_India
Gnansounou, E. (2008). Assessing the energy vulnerability: Case of industrialised countries.
Energy Policy, 36, 37343744.
Gupta, E. (2008). Oil vulnerability index of oil-importing countries. Energy Policy, 36, 11951211.
Higashi, N. (2009, June). Natural gas in China: Market evolution and strategy. (Working Paper
Series). Paris, France: International Energy Agency.
International Association for Energy Economics (IAEE). (2007). Natural gas: Is there a
decreasing trend? IAEE Newsletter, 16 (Third Quarter).
International Energy Agency (IEA). (2007). Natural gas market review 2007: Security in a
globalising market to 2015. Paris, France: International Energy Agency.
International Monetary Fund (IMF). (2010). World economic outlook database. Washington, DC:
IMF. Retrieved from http://www.imf.org/external/pubs/ft/weo/2009/01/weodata/index.aspx
Jackson, M. (2007, October). The future of natural gas in India: A study of major consuming
sectors (Working Paper #65). Stanford, CA: Program on Energy and Sustainable Development, Stanford University.
Jansen, J. C., van Arkel, W. G., & Boots, M. G. (2004). Designing indicators of long-term energy
supply security. Petten, The Netherlands: ECN.
Kendell, J. (1998). Measures of oil import dependence. Washington, DC: Energy Information
Administration, U.S. Department of Energy. Retrieved from http://www.eia.doe.gov/oiaf/
archive/issues98/oimport.html
Komiyama, R., Zhidong, L., & Ito, K. (2005). World energy outlook in 2020 focusing on Chinas
energy impacts on the world and Northeast Asia. International Journal of Global Energy
Issues, 24(34), 183210.
Percebois, J. (2006). Dpendance et vulnrabilit: deux faons connexes mais diffrentes
daborder les risques nergtiques. (Cahiers de recherch CREDEN, No. 06.03.64).
Montpellier, France: CREDEN.
Percebois, J. (2007). Energy vulnerability and its management. International Journal of Energy
Sector Management, 1, 5162.
Reymond, M. (2007). European key issues concerning natural gas: Dependence and vulnerability.
Energy Policy, 35, 41694176.
Thacker, S. (2006). Recent Indian petroleum, petroleum products and natural gas regulations.
Hydrocarbon World, 2324.
United Nations Development Programme (UNDP). (2007). Overcoming vulnerability to rising oil
prices: Options for Asia and the Pacific. Bangkok: UNDP Regional Centre.
World Bank. (2009). Worldwide governance indicators19962008. Washington, DC: World
Bank. Retrieved from http://info.worldbank.org/governance/wgi/index.asp
World Energy Council. (2008). Europes vulnerability to energy crises. London, United Kingdom.
Chapter 3
Keywords Export Pricing Buyer credit Risk Ratings Gas China United
States
J. L. Simpson (&)
School of Economics and Finance, Curtin Business School,
Curtin University, GPO Box U1987, Perth, WA 6845, Australia
e-mail: john.simpson@cbs.curtin.edu.au
31
32
J. L. Simpson
3.1 Introduction
Spot natural gas exports are largely paid for in cash, however, in some instances,
buyer credit is provided either by the exporter, or by the exporters bank or by a
syndicate of banks led by the exporters bank or by a government export, finance
and insurance agency in the exporters country. Such an agency (e.g., Export
Finance and Insurance Corporation of Australia or EFIC) may alternatively provide
direct loans to the gas buyer. Sometimes this latter arrangement is part of the
exporting contract, where the exporter has undertaken to, not only ship the goods,
but to find suitable finance for the buyer for the full amount of the export price. The
export finance might be provided through the Eurocurrency market at a margin
(spread) above the interbank offered rates. The margin is commensurate with
perceived country credit risk, but other factors will enter into calculations. For
example, interest rate policy by global economic powerhouses such as the US has a
significant bearing on debt spreads. For emerging countries with high debt to gross
national income levels, spreads have been found to increase beyond normal risk
premia by six to sixty-five basis points (Dailami et al. 2008).
This paper, for the purpose of a basic analysis, focuses on the basic spreads or
country risk premia derived from risk ratings. Economic and financial risk
components are accounted for in a general sense through composite country risk
ratings which of course include political risk ratings. The question is what size
credit risk premium (spread) should be applied to the finance? That is, what size of
credit risk premium should be added to a prime/reference interest rate? The latter
is assumed the same for all borrowers and assumed constant for the purposes of
this study. The amount of buyer credit equates to the total shipment export price,
assuming one hundred per cent of the sales contract is to be financed. The size
of the credit risk premium relates to the risk and return maxim originally applied in
the securities industry to portfolio theory expounded by Markowitz (1959).
As credit providers add more and more loan assets to their portfolios, whose
returns are less than perfectly positively correlated there is a greater opportunity to
diversify away the unsystematic risk of the portfolio. That is, as soon as the spread
is determined, credit providers also need to include an appropriately weighted
mixture of high and low credit risk loans in their portfolios to diversify their
unsystematic risks.
It should be noted that exports may be priced differently to buyer credit margins
or credit risk premia. Most natural gas exports from Australia are priced according
to a basic relationship with gas prices and the price of oil. This relationship is
discussed later in the chapter. However, the point is reinforced that a benchmark
gas export price equates to the amount of buyer credit needed and the amount of
credit requires the addition of a credit risk margin to the base interest rate.
The question then arises as to how the credit risk premium is to be quantified
assuming finance is available in the first place. Ford (2010), a senior economist
with EFIC, commenting on their behalf at a corporate and business banking forum,
made the point that whilst the global economic condition is worrying, banks and
33
3:1
As the returns of the gas price index equate to the changes in the amount of
export/buyer credit required (DCgt ), then.
DCgt Pgt Pgt1 =Pgt1
3:2
34
J. L. Simpson
Section 3.3 discusses the methodology and data. Section 3.4 explains the model
and reiterates the issues. This is followed by Sect. 3.5 which provides the results
and this is followed by a conclusion in Sect. 3.6.
3.2 Background
The fundamental position of the paper is that less credit risky countries, in accordance with the tenets of financial economics and the risk/return trade-off, should pay
less for buyer credit or finance for the purchases of all imported products including
natural gas. It is posited that country risk needs to be incorporated into buyer credit
premia for energy export finance. A search of the literature relating to buyer credit for
energy exports does not reveal extensive empirical work in this area. Of course, there
are other factors that influence loan pricing for all exported products, not only natural
gas. These factors relate, for example, to strategic market penetration, periods of
financing required and economies of scale.
Unusually, loan pricing mistakes as well as export pricing mistakes can be
made in times of excess demand for commodities such as natural gas (that is, the
exports or the export/buyer credit might be under-priced). Loan interest rate
benchmarks generally relate the price of credit to various credit risks over and
above costs of funds. These risk premia may be discounted deliberately in times of
excess demand so that the buyer can be induced to buy the product being financed.
The exported product may also be discounted in price. That is, export pricing as
well as loan pricing in an inefficient and imperfect market may be similarly
distorted.
For example, a large gas export deal negotiated in around 2002 by Australia
with China was under-priced in terms of the optimal relationship with oil prices.
However, the Australians involved negotiated a large export contract in a potentially large market. They thereby created an export momentum and thus may have
been conscious of an initial need to buy market share. If buyer credit was
required in this instance it is safe to assume that country credit risk premia
applicable to China would also have been significantly reduced to create less
expensive financing.
Nevertheless, the Chinese deal was negotiated during times of excess demand
when Australia should have been a loan price-maker and an export price-maker
and China, as an economy in transition with higher levels of country risk, should
have been the price- taker. Clearly the Australians bought the business to gain a
market toehold and a precedent was set. But, loyalty of buyers in the longer
term is not guaranteed and is also improbable in times of excess supply. It should
be stated that recent export pricing of Australian gas to China has been more in
line with the global market. It is argued that this should extend to buyer credit
where credit risk premia should be more in line with the market and in accordance
with the risk/return maxim.
35
Whilst export contract formulae or specific export contract prices are not in
general for public knowledge, natural gas pricing for importing countries is
receiving greater attention as more countries move over to gas energy. The method
of pricing of natural gas is known to differ from region to region. The pricing is
usually linked to oil prices, but which oil prices are they linked to? In Asia (e.g.,
Japan) the import price formula is thought to be based on a basket of crude oil
prices commonly referred to as the Japanese Crude Cocktail (JCC). The UK and
Europe may base their gas prices on Brent oil.
Another danger with the Chinese precedent from Australias view is that in
times of excess supply, other countries will want a similar deal to that written with
China during times of excess demand. For example, Eng (2006) attempts to build a
case for the adoption by New Zealand of the lower bound Chinese pricing model
for possible gas imports from Australia. These New Zealand imports may not
necessarily be sourced from Australia, but the distance from Australian natural gas
fields to New Zealand compares to that from Australia to China. It is, however,
understandable that New Zealand would want to negotiate a lower credit price,
if finance is required, due to its lower level of country risk. This study does not
examine the relationship of export prices and buyer credit to the price of oil. The
chapter merely seeks to test the relationship between buyer credit and risk ratings
of two countries to assist in the calculation of appropriate buyer credit risk premia.
The economic and financial components of country risk are measurable and
objectively assessed, but the overall assessment of country risk is difficult because
political risk is difficult to assess. Political risk is subjective in its assessment. It is
based on opinions relating to political outcomes. Risk rating agencies have
attempted to develop further the method of quantification of this component of
country risk. As mentioned this study examines the country risk ratings which
combine economic, financial and political risk ratings.
Significant progress in regulatory reform has been achieved in developed
economies (for example, the US) and generally in these countries, country risks are
lower. There is a greater choice of natural gas contracts through deregulation and
restructuring within many countries. For example, in the US since 1984 there has
been a separation of natural gas supply from interstate pipeline transportation,
deregulated natural gas production and the wholesale market, and competition has
been introduced in interstate pipeline transportation. Privatization is increasingly
being seen as a means of improving efficiency and increasing investment. Lower
country risks are therefore expected to be associated with developed economies
with deregulated natural gas markets (Juris 2005).
There is a connection between market integration and political risk. It follows
that there is also a connection between export prices (and thus the amount of buyer
credit if needed), loan pricing and the country risks in the buyer country. Asche
et al. (2000) find that cointegration tests show the different border prices for gas to
Germany move proportionally over time. This indicates integration of the German
gas market. Asche et al. also study whether or not there are large price differences
between gas from Norwegian, Dutch and Russian exporters. They find differences
36
J. L. Simpson
in mean prices and the reasons for the price differences are ascribed to differences
in volume flexibility and perceived political risk.
Hartley and Medlock (2005) commence a comprehensive study on the political
and economic influences on the future market for natural gas by recognizing that
required rates of return on investments in energy infrastructures vary geographically. They provide a base case, which assumes that the required rates of return
match those sought in similar US projects. It follows that higher risk borrowing
countries will also face higher required rates of return on buyer credit. Selected
scenarios (representing a range of political actions as well as economic outcomes)
are compared to the base case to ascertain the affect of such factors on the global
gas market.
When Hartley and Medlock (2005) calculate risk-adjusted returns for gas
projects they use data from the International Country Risk Guide (ICRG) as well
as a data series on the risk premium on lending provided by the World Bank. They
construct a gas investment risk index using pure political risk scores obtained
from the ICRG (2011). Hartley and Medlock (pp. 2728) find that increased
gas trade will enhance interactions between regional gas markets and promote
arbitrage and global pricing over time. For example, it is apparent that supply
infrastructure and demand growth in North East Asia will significantly influence
the developing global market for natural gas.
In the absence of an official global benchmark gas export price, the specification
of an interim export benchmark gas export price is proposed, that might be used by
a gas exporting country such as Australia. The change in export/buyer credit
required is derived from this price index. In deciding which benchmark to use as a
proxy the following questions are considered: Which is the most developed and
deregulated market and thus the most likely global leader in global gas market
integration? Which market is by far the greatest consumer of natural gas? Which
market supplies itself to the greatest degree? Which market is showing the greatest
transition from, for example, coal and oil fired power generation for industry?
Which market has an established benchmark for natural gas prices? Finally, is it
feasible for another gas trading country to use such a benchmark for its natural gas
export pricing?
For the sake of analysis, but for the above reasons and in the absence of a
suitable alternative, this chapter examines the use of the USs Henry Hub natural
gas price as the benchmark for export pricing of Australias natural gas. Such a
price series needs to be seasonally adjusted as Australia and the US have different
seasonality and storage factors operating. It also requires conversion into Australian dollars.
The natural gas price benchmark in the Americas is the Henry Hub (HH). This
price is determined at a physical location in Louisiana USA and has a greater
potential than the National Balancing Point (NBP) price in the UK as a candidate
to become a global benchmark price for natural gas (Mazighi 2005). This is
because more gas in the UK market is sold on long-term contracts with prices
indexed to oil. Even though the HH is said to show less normality in fluctuation
37
than the NBP price, the US is the largest market (estimated in 2005 at more than
660 billion cubic meters compared to around 100 billion cubic meters in the UK).
If the US market is by far the largest market in the world then it is conceivable
that the price mechanism in that market may be more representative, in a global
sense, than other pricing mechanisms such as the NBP. The HH prices are also
quoted on the New York Mercantile Exchange (NYMEX) on a daily basis. The
HH gas price index, for the sake of analysis, is selected in this chapter as the proxy
for a natural gas export price which could be adopted by a major natural gas
exporting country such as Australia.
38
J. L. Simpson
The composite risk ratings therefore represent country risk which combines
economic, financial and political factors to reflect very high, high, moderate, low
and very low riskiness in the rated countries. The ICRG ratings are preferred in
this study because there is a suitably weighted political risk component in the
composite scores. The method of composite risk calculation is described in ICRG
(2011). The categorization of risk is as follows: Very High Risk: 049.9; High
Risk: 5059.9; Moderate Risk: 6069.9; Low Risk: 7079.9; Very Low Risk;
80100.
The composite risk ratings in ICRG are rated monthly for each country from
zero to one hundred in terms of riskiness. In country risk ratings through ICRG, the
higher the risk score, the lower the risk. In this study the 10 % level of statistical
significance is selected over the full analysis due to the limitation of small sample
sizes.
3:3
In this case the regression intercept is agt , (representing the base spot gas export
returns or the base change in export/buyer credit required). The regression coefficient is bgt . This coefficient represents the contributions of the change in
composite risk in a particular country to the base export returns and the base
change in amount of export/buyer credit required. That is, it represents the contributions of the change in risk ratings to the changes in amounts required for
financing. The error term, egt represents the contribution to the inter-temporal
changes in export/buyer credit required not explained by an importing countrys
risk factors.
The most important part of this analysis is when Eq. 3.3 is re-specified by level
series variables into a Vector Auto Regressive Model (VAR) by optimally lagging
all of the variables on the right hand side of the equation and testing for cointegration. If cointegration is evident, testing is again undertaken in the framework of
a Vector Error Correction Model (VECM). If cointegration remains evident,
causality (exogeneity) tests may also be undertaken. The risk ratings in the United
States and China may be treated, as in a single multivariate model, as variables that
interact with the buyer credit variable.
39
HH
16
14
12
10
8
6
4
2
I
II
III
2002
IV
II
III
2003
IV
II
III
2004
IV
II
III
IV
2005
Note HH is the Henry Hub gas price, used for the sake of analysis, as thenatural gas exporter
country price.
Excess demand conditions are expected to prevail in the global gas industry for
at least the next decade as infrastructures and investment build up. In times of
excess supply, prices of gas will fall as more suppliers compete to sell their
product. In times of excess supply the prices will naturally fall and the consumer/
importer will, under normal conditions, select the lowest price. The exporter is
likely to become a price-taker under this market condition. This study provides for
an export price-maker scenario in times of excess demand where it is possible to
cost in a country risk premium to an export price and therefor the pricing of any
buyer credit.
3.5 Results
Preliminary results are shown in Figs. 3.1, 3.2 and 3.3. Trends of the level series
gas prices and the raw country risk ratings for the United States and China are
illustrated. Figure 3.1 shows that gas prices have demonstrated a degree of
stability with only a steady rise over the period of the study and only two outliers,
perhaps reflecting seasonal or storage factors.
Results show that over the study period the mean of gas prices is $6.09, the
median is $5.77, the minimum is $2.49, and the maximum is $15.00. The standard
40
J. L. Simpson
PRCHINA
65
64
63
62
61
60
59
58
I
II
III
IV
2002
II
III
IV
2003
II
III
IV
2004
II
III
IV
2005
PRUS
84
82
80
78
76
74
72
I
II
III
2002
IV
II
III
IV
2003
Note PRUS denotes raw country risk ratings for the US.
II
III
2004
IV
II
III
2005
IV
41
Statistic
Value
Adjusted R squared
Coefficient
Standard error
t-statistic
DW
0.0804
-3.9556
1.6373
-2.4160
2.2893
Statistic
Adjusted R squared
Coefficient
Standard error
t-statistic
DW
Value
0.0095
-1.4858
1.0206
-1.4559
2.1936
42
J. L. Simpson
-7.5979a
-7.7860a
-7.5403a
-7.6062a
-2.2249
-2.9239b
-2.0546
-3.2646b
Note a denotes significant at 1; b significance at 5 %. The ADF critical values are -3.5847, 2.9281 and -2.6022 for significance levels of 1, 5 and 10 % respectively
Tables 3.1 and 3.2 contain the results of the regression analysis for changes in
buyer credit required for China and for the US respectively. It is noted that the
explanatory power is not strong for each country model. However, the independent
variables in each case are statistically significant at the 10 % level. The DurbinWatson (DW) test statistic (Durbin and Watson 1971) reveals that serial correlation is not a problem in the errors and thus that the results are not spurious. White
tests reveal that heteroskedasticity is not persistent in the errors as the F statistics
of 0.3426 for China and 0.0438 for the US are not statistically significant at the
10 % level for China and the US respectively. The t-statistics and coefficients
show a negative relationship between country risk changes and gas export returns
in both the United States and in China.
Tables 3.1 and 3.2 show that the explanatory power of the models is lower in
the case of the US than that in China, thus indicating that country risk ratings
changes are more important in the case of China (a higher country risk country).
The results show that the change in country risk in China has a stronger relationship with the changes in the amount of Australian gas export credit required
than that with United States country risk changes (The adjusted R Squared value in
the case of China is approximately 8 % and that for the US is around 1 %).
The sign of the t-statistic indicates that as there is a reduction in the country risk
score changes (meaning higher risk), the changes in gas export credit required
(equating to export returns) increase. This is consistent with risk/return trade-off
theory (a lower risk rating reflects lower returns). However, these models are
unlagged and therefore do not provide an indication of long-term equilibrium
relationships nor do they confirm the exogeneity of the changes in risk ratings.
Augmented Dickey Fuller (ADF) unit root tests (Dickey and Fuller 1981)
results are included in Table 3.3.
The results show that each of the level series gas prices and country risk ratings
are non-stationary series and that the series in changes in buyer credit required and
changes in country risk ratings and the errors of those relationships are stationary.
The level series prices and raw ratings thus are integrated non-stationary
processes. A VAR and VECM are specified in a single model (inclusive of US as
well as Chinese data) in order to apply and confirm VAR based Johansen cointegration (Johansen 1988) and Granger causality tests (Granger 1988).
43
Table 3.4 Cointegration of export buyer credit required and country risk scores in the US and
China
Hypothesized number of cointegrating
Eigenvalue Statistic Critical
Probability
equations
value
Cointegration rank test (trace test)a
None
At most 1
At most 2
Cointegration rank test (maximum eigenvalue
None
At most 1
At most 2
0.460489
0.350097
0.148269
test)b
0.460489
0.350097
0.148269
53.17430 29.79707
26.02229 15.49471
7.061307 3.841466
0.0000
0.0009
0.0079
27.15201 21.13162
18.96098 14.26460
7.061307 3.841466
0.0063
0.0084
0.0079
Note A linear deterministic trend is assumed. The lag interval is 12 obtained from Akaike,
Schwartz and Hainan-Quinn information criteria.
a
Trace test indicates 3 cointegrating equations at the 0.05 level
b
Maximum eigenvalues indicate 3 cointegrating equations at the 0.05 level
Table 3.4 shows evidence of cointegration in this model. The variables in the
longer-term exhibit similar stochastic trends and together come to equilibrium on
an optimal lag of 12 months.
Table 3.5 contains evidence that Granger causality runs from country risks in
China to the amount of buyer credit required with significance at the 1 % level.
There is no evidence that causality runs from US country risks to the buyer credit
required variable.
There is evidence of joint causality of the US and Chinese variables at the 5 %
level of significance (however, as stated, considered individually the risks in China
are the statistically significant exogenous force at the 1 % level). There is no
statistically significant evidence of reverse causality from the buyer credit required
variable to the Chinese and US risk ratings variables.
Overall, the implications for loan pricing are that exporters of natural gas need
to examine the contribution of country risks to the variability of export returns
(changes in buyer credit required). These results may be supported when optimally
lagged models are also considered and when these models confirm long-term
equilibrium relationships and short term dynamics in exogeneity.
3.6 Discussion
This chapter has not explored other more specific factors that may widen the
spreads in international buyer credit financing, such as those related to country
economic and financial risk that affect overseas buyers ability to service their debt.
For example, high debt to gross national income levels as argued by Dailami et al.
(2008) cause an increase in risk that a country cannot service its debt. The
corollary is that the borrower, who may be financially sound, is unable to repay
44
J. L. Simpson
Chi-sq
US country risk
China country risk
All
0.957900
11.34101
12.12399
lag
Probability
2
2
4
0.6194
0.0034
0.0165
the debt because foreign exchange has been frozen. Thus economic and financial
risk components need to be combined with political risk components in order to
properly price buyer credit. The chapter deals with this issue by considering
country (composite) risk ratings which include economic, financial and political
risk components. Portfolio theory (Markowitz 1959) is also vital as credit
providers strive to diversify away the unsystematic or country specific component
of total country risk when lending across countries.
The importance of political risk in natural gas export markets has previously
been supported in continental Europe studies such as Asche et al. (2000) and,
Hartley and Medlock (2005). The latter also find relationships between investment
and bank lending returns and political risk. Support is found for a significant effect
of country risk on gas export returns (and thus on the amount of export credit
required) and lending and energy infrastructures investment including natural gas.
They also provided evidence for a wide range of political scenarios and economic
outcomes that could have an influence on a global gas market price.
3.7 Conclusion
The study reported in this chapter finds that changes in country risk scores for both
China and the United States are statistically significant variables in a gas export
returns/buyer credit model, which might be suitable for adoption by Australian gas
exporters and Australian providers of export/buyer credit. Risk ratings are
important inputs to assist in the pricing of buyer credit. Evidence is produced that
country risk changes in China have a substantially larger contribution to the gas
export-returns/buyer-credit variability than those in the US. This is an indication
that country risk factors may be an important consideration in ascribing credit
premia for China where the acceptable market rate of interest would attach a risk
premium at least eight times greater than that attached to the same reference
interest rate for the United States. This information is inferred from indications of
the comparative size of the single period unlagged regression parameters in
significant adjusted R2 values and, specifically, t-statistics.
In the context of this study involving gas exporters (and gas export credit
providers) and two large and powerful economies in the US and China as
importers, stronger evidence is provided in tests of cointegration and causality. It is
45
found that, in a single optimally lagged model, each of the series studied produce
similar stochastic trends and move to stability in the longer term. A long-term
equilibrium relationship exists. In addition, the US and Chinese country risk
variables, when interacting are exogenous in such a model, and, as such, these
variables lead gas export returns (and thus buyer credit) variables on a lag of
12 months with China risk ratings being the greater exogenous force.
Overall, the study suggests that country risk needs to be considered in loan
pricing for gas exports. Credit providers also need to focus on the goals of the firm
not only in shareholder wealth maximization, but also in the diversification away
of the unsystematic risks (country specific risks) associated with loan asset portfolios. This may suggest a direction for future buyer credit pricing for Australian
and other exporters of natural gas as the global market becomes more informationally efficient, as more countries deregulate their gas markets and import greater
volumes of this form of energy in partial replacement of oil imports. Global natural
gas market efficiency will mean substantially more of the natural gas export buyer
credit pricing will be decided by economic, financial and political risk factors.
Further research may be undertaken to compare the explanatory power of the
models in this chapter with those that include specific economic and financial
indicators that indicate the ability of countries and importing companies to service
debt along with pure political risk indicators.
3.8
Appendix
46
J. L. Simpson
47
has the strength and expertise to govern without major changes in policy or
interruptions in government services. That is, bureaucracies have a degree of
autonomy from political pressure with an established independent mechanism for
recruitment and training.
References
Asche, F., Osmundsen, P., & Tveteras, R. (2000). European market integration for gas? Volume
flexibility and political risk. (CESifo Working Paper No. 358). Munich, Germany: CESifo
Group.
Dailami, M., Masson, P. R., & Padou, J. J. (2008). Global monetary conditions versus country
specific factors in the determination of emerging market debt spreads. The Journal of
International Money and Finance, 27, 13251336.
Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series
within a unit root. Econometrica, 49, 10221057.
Durbin, J., & Watson, G. S. (1971). Testing for serial correlation in least squares regression. III.
Biometrica, 58, 119.
Eng, G. (2006). A formula for natural gas pricing. Ministry for Economic Development New
Zealand. Retrieved from http://www.med.govt.nz/Templates/multipageDocumentTOC___
23939.aspx.
Ford, B. ((2010, October)). International economic developments and trade finance. Sydney,
Australia: Background remarks for the trade finance panel at the AB +F corporate and
business banking forum.
Granger, C. W. J. (1988). Some recent developments in a concept of causality. Journal of
Econometrics, 39, 199211.
Hartley, P., & Medlock, K. B. (2005). Political and economic influences on the future world
market for natural gas (Geopolitics of Gas Working Paper). Retrieved from http://
www.rice.edu/energy/publications/docs/GAS_PoliticalEconomicInfluences.pdf.
ICRG. (2011). International country risk guide. East Syracuse NY: The PRS Group Inc.
Retrieved from www.prsgroup.com/ICRG_methodology.aspx.
Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics
& Control, 12, 231254.
Juris, A. (2005). The development of markets in the UK gas industry. (Policy Research Working
Paper). Washington, DC: World Bank, Private Sector Department.
Mazighi, A. E. H. (2005). Henry Hub and national balancing point prices: What will be the
international gas price reference? OPEC Review, 29, 219230.
Markowitz, H. (1959). Portfolio selection: Efficient diversification of investments. New Haven,
CT: Yale University Press.
Chapter 4
Abstract This paper examines the issues associated with the causal relationships
between the energy consumption and the factors (rural population, total population,
gross domestic product, consumer price index and carbon dioxide emission), with the
greatest impact on energy consumption as demonstrated in the literature, for 30
developing countries. Data for the period 19712007 are used with a Granger
causality test. In the light of obtained findings, the present study reveals common
relationships in various directions between energy consumption and the other factors. These results can be explained by factors such as energy markets, resources,
population etc. for individual countries. The findings of the study have significant
policy implications and are therefore of potential interest to policymakers.
Keywords Energy consumption
Causality Developing countries
4.1 Introduction
Energy, which has become one of the significant inputs in the economic development process, has gained further importance with the spread of globalization. As
a result, countries energy consumption (demand)and their dependence on energy
have rapidly increased Stern and Cleveland (2004).
A. Kapusuzoglu (&)
Yildirim Beyazit University, Ankara, Turkey
e-mail: ayhkap@gmail.com
M. B. Karan
Hacettepe University, Ankara, Turkey
e-mail: mbkaran@hacettepe.edu.tr
49
50
51
The presence of energy dependence in developing countries brings about the possibility of a mismatch between energy supply and demand. Such a possibility may well
have a potential to affect the level of economic activity and energy demand (consumption). In this context, it will contribute to the research predicting global oil demand
if the drivers of energy consumption in developing countries could be identified.
This study mainly aims to identify the fundamental drivers of energy consumption levels in developing countries and to reveal the causality relationships
between these drivers and energy consumption levels. The energy consumption
levels of the countries under study as well as other certain factors make up the
basic variables of the study, which will be analyzed separately for each country.
There is limited research available in the empirical literature that examines the
relationship between energy consumption and the main economic drivers for
developing countries, and most studies limit their analysis to a single variable. In
this respect the difference between this study and others, and its main contribution
to the literature, is that it examines a multitude of variables having potential
relationships with energy consumption levels. In addition, the study covers a large
number of developing countries and separately analyzes each country, which
allows more exhaustive and realistic results on the subject.
52
53
4.3 Literature
There is no consensus on the relationships between energy consumption and
economic growth or their mutual effects.The empirical evidence does not yet exist
54
to provide conclusive support for the claims of either the ecological or neoclassical
schools of thought. The trend in recent years suggests that the possibility of
decoupling energy use from economic growth has in fact been achieved largely by
a switch away from the direct use of low quality fuels such as coal to higher
quality fuels and energy inputs, electricity in particular (Cleveland et al. 1984;
Kaufmann 1992, 2004; Stern 1993; Stern and Cleveland 2004).
The causality relationship between energy consumption and economic growth
was first realized by Kraft and Kraft (1978), who detected the presence of a
causality relationship from economic growth to energy consumption. The causality
relationship between energy consumption and economic growth can be classified
under four groups.
The first group is the uni-directional causality relationship from energy consumption to economic growth. It is noted that in this relationship, an increase in
energy consumption positively contributes to economic growth. This is the most
commonly observed direction of the relationship in the literature. The second one
is the uni-directional causality relationship from economic growth to energy
consumption, which demonstrates that policies in favor of energy consumption
have a positive effect or at least do not have any negative effect on economic
growth. This type of relationship implies another most common relationship
between energy consumption and economic growth.
The third group is the bi-directional causality relationship between energy
consumption and economic growth, in which both variables influence each other.
Finally, there is no causality relationship between energy consumption and economic growth and the variables in question do not affect one another. On the other
hand, some studies have shown that certain factors, apart from economic growth,
directly or indirectly influence energy consumption in developing countries.
These studies examine the indirect impacts of factors such as rural population,
total population, consumer price index and CO2 emission upon the process of
energy consumption via their direct effects on energy consumption and economic
growth (Ang 2007; Akinlo 2008; Bentham and Romani 2009; Kebede et al. 2010;
Odhiambo 2010).
Tables 4.1 present the results of the empirical studies conducted on the presence and direction of causality relationships between energy consumption and
economic growth. It is clear that causality relationships exist in all but seven
developing country studies and in all but two developed country studies.
When Table 4.1 is considered, it contains findings in relation to studies conducted in developing countries. As is apparent in the field literature in this context,
there are differences in relation to existence and direction of the revealed causality
relations. While studies of Fatai et al. (2004); Narayan and Singh (2007); Odhiambo (2010) contain a causality relationship from energy consumption to GDP
and studies by Odhiambo (2010) and Kapusuzoglu and Karan (2010) contain a
causality relationship from GDP to energy consumption; in the studies of Murry
and Nan (1996), Wolde-Rufael (2006) and Kebede et al. (2010) it is revealed that
no causality relationship exists between the variables.
19501992
19501992
19701990
19521999
19611997
1965-2000
19702002
19602001
19602001
19602001
G-7
Korea
Shanghai
Canada
Sweden
Korea
G-11
Wolde-Rufael (2006)
19701990
19701990
19701990
19601999
19601999
19712001
19712001
19712001
19712002
19722006
19722006
1972-2006
19722006
19802004
19752006
India
Zambia
Philippines
Indonesia
India
Algeria
Congo Rep.
Kenya
Fiji Island
Congo
Kenya
S. Africa
Kenya
Sub-Africa
Turkey
(continued)
No causality
No causality
No causality
Energy consumption?GDP in Indonesia
Energy consumption?GDP in India
No causality
No causality
No causality
Electricity consumption?GDP in Fiji Island
GDP?Energy consumption in Congo
Energy consumption?GDP in Kenya
Energy consumption?GDP in S. Africa
Price?Energy consumption in Kenya
No causality in Central Africa
GDP?Electricity consumption in Turkey
Table 4.1 Empirical findings on the causality between energy (and electricity) consumption and economic growth
Author(s)
Country
Period
Direction of causality
Country
ASEAN
G-7
France
US
New Zealand
Yoo 2006
Zachariadis (2007)
Ang (2007)
Payne (2009)
Barleet and Gounder (2010)
19712002
19602004
19602000
19602000
19492006
19602004
Period
Electricity consumption$GDP in Singapore
No causality in Canada, Germany, UK and US
GDP?Energy consumption in France
GDP?CO2 in France
No causality in US
GDP?Energy consumption in New Zealand
Direction of causality
56
A. Kapusuzoglu and M. B. Karan
57
When Table 4.1 is considered, it contains findings in relation to studies conducted in developed countries and the findings of these studies also display differences. Studies of Soytas and Sari (2003); Hatemi and Irandoust (2005); Lee
(2006); Ang (2007), and Barleet and Gounder (2010) reveal a causality relationship from GDP to energy consumption, and studies of Oh and Lee (2004); WoldeRufael (2004); Lee (2006) and Yoo (2006) reveal a causality relationship from
energy consumption to GDP. No causality relationships were found between
variables in the studies of Zachariadis (2007) and Payne (2009).
These countries are;Algeria, Argentina, Bolivia, Cameroon, Congo, Costa Rica, Cote Dlvoire,
Egypt, El Salvador, Gabon, Ghana, Honduras, India, Korea, Malaysia, Mexico, Morocco,
Nigeria, Pakistan, Paraguay, Senegal, Sri Lanka, Sudan, Syria, Thailand, Togo, Trinidad Tobago,
Turkey, Uruguay and Venezuela.
58
dioxide emission (CO2kg/2000 US$ of GDP), which were obtained from World
Development Indicators (2011).
Since other countries, notably among developing countries, lacked data for the
period of analysis, these countries were excluded from the study. The main limitation of the study is the limited amount of data, which results from the very
comprehensive sample and variables of the study.
The econometric methods used in the present study bear a parallel to the
approaches used in the above previous studies. Before proceeding with the analysis, the natural logarithms of the data are first taken. Next, stationarity analysis is
performed, the first stage of which consists of the correlogram test. On the other
hand, in the decision-making process about the stationarity of the data, an Augmented Dickey-Fuller (ADF-1979) testthe most commonly used parametric test
and the Philips-Perron (PP-1988) testwhich takes into consideration structural
breaks and trends that possibly occur in time seriesare applied. Grangers (1969)
Causality test is performed to determine the presence and direction of the causality
relationship between the variables.
59
words, the statistics of this test indicate the Dickey-Fuller t-statistics are modified
by taking into account the less constraining nature of the error process (Enders
1995).
An examination of the results of the unit root tests performed for each country
and for each variable shows that the variables are not stationary of the same order
for all countries. Therefore, it was not possible to apply Johansen or EngleGranger cointegration tests to investigate the presence of any long-term relationship between the variables.
k
X
a1 xti
i1
yt b0
k
X
i1
k
X
a2 ytj u1t
4:1
b2 xtj u2t
4:2
j1
b1 yti
k
X
j1
In the equations, u1t and u2t denote error terms which do not exhibit zero means,
a finite covariance matrix and series correlation; and k denotes the lag number for
both variables. After validating the equation, the relationship is formulated as from
60
Algeria
Argentina
Bolivia
Cameroon
Congo
Costa Rica
Cote Dlvoire
Egypt
El Salvador
Gabon
Ghana
Honduras
India
Korea
Malaysia
Mexico
Morocco
Nigeria
Pakistan
Paraguay
Senegal
Sri Lanka
Sudan
Syria
Thailand
Togo
Trinidad Tobago
Turkey
Uruguay
Venezuela
Energy consumption-rural
population
Rural population-energy
consumption
Prob.
Causality
Prob.
Causality
3.187a
1.978
1.195
1.644
0.752
1.135
0.201
5.501c
0.577
1.815
5.998c
0.384
1.510
2.096
1.045
0.119
6.861c
0.094
5.063c
1.458
1.781
2.685a
1.050
1.129
2.211a
0.560
0.409
1.516
0.874
1.816
0.055
0.155
0.316
0.210
0.530
0.334
0.818
0.009
0.567
0.158
0.001
0.684
0.237
0.140
0.364
0.887
0.003
0.962
0.004
0.247
0.185
0.066
0.362
0.336
0.098
0.645
0.667
0.235
0.427
0.180
Yes
No
No
No
No
No
No
Yes
No
No
Yes
No
No
No
No
No
Yes
No
Yes
No
No
Yes
No
No
Yes
No
No
No
No
No
0.536
4.951b
1.552
0.414
0.122
0.355
4.714b
0.016
3.076b
7.542c
7.943c
0.345
2.231
8.675c
0.897
2.562a
1.952
2.975b
2.700a
4.911c
2.130
1.941
1.743
1.097
2.306a
2.026
0.086
1.183
2.611a
0.598
0.590
0.013
0.228
0.664
0.946
0.703
0.016
0.983
0.060
0.0004
0.0003
0.710
0.124
0.001
0.418
0.093
0.159
0.049
0.054
0.007
0.185
0.146
0.192
0.346
0.087
0.133
0.917
0.320
0.090
0.555
No
Yes
No
No
No
No
Yes
No
Yes
Yes
Yes
No
No
Yes
No
Yes
No
Yes
Yes
Yes
No
No
No
No
Yes
No
No
No
Yes
No
Direction
UD
UD
UD
UD
UD
UD
BD
UD
UD
UD
UD
BD
UD
UD
BD
UD
X to Y. The analyses reveal whether two variables influence each other with lag
effects, and if they do, whether this causality is uni-directional (from X to Y or
from Y to X) or bi-directional(both from X to Y and from Y to X).
In the process of applying the Granger causality test in the study, the causality
relationship between energy consumption and other variables is investigated for
each country in construct pairs. First, the causality relationship is examined
between energy consumption and rural population. The results of the analysis (see
Table 4.2) reveal that there is an inter-variable causality relationship in 16
61
Algeria
Argentina
Bolivia
Cameroon
Congo
Costa Rica
Cote Dlvoire
Egypt
El Salvador
Gabon
Ghana
Honduras
India
Korea
Malaysia
Mexico
Morocco
Nigeria
Pakistan
Paraguay
Senegal
Sri Lanka
Sudan
Syria
Thailand
Togo
Trinidad Tobago
Turkey
Uruguay
Venezuela
Energy consumption-total
population
Total population-energy
consumption
Prob.
Causality
Prob.
Causality
0.681
4.919c
0.103
6.982c
4.578c
0.392
0.483
9.121c
3.606b
0.950
2.450a
2.402a
1.197
4.452b
1.060
0.355
13.227c
3.346b
4.313c
3.028b
2.055
2.693a
0.259
0.227
4.145b
0.248
1.713
0.817
3.524b
1.475
0.611
0.004
0.980
0.0007
0.006
0.812
0.747
0.0001
0.019
0.452
0.073
0.077
0.316
0.042
0.397
0.703
0.000008
0.026
0.009
0.046
0.118
0.065
0.900
0.797
0.011
0.907
0.179
0.526
0.028
0.243
No
Yes
No
Yes
Yes
No
No
Yes
Yes
No
Yes
Yes
No
Yes
No
No
Yes
Yes
Yes
Yes
No
Yes
No
No
Yes
No
No
No
Yes
No
3.222b
2.423a
7.059c
0.455
0.322
5.921c
1.760
4.795c
0.398
1.977
5.509c
2.664a
2.202
1.155
1.529
1.643
1.481
3.740b
2.801b
2.516a
1.589
2.015
1.635
0.789
1.015
1.143
1.029
2.064
1.908
3.841b
0.029
0.076
0.0007
0.767
0.860
0.001
0.169
0.005
0.807
0.130
0.002
0.057
0.128
0.290
0.225
0.210
0.238
0.016
0.048
0.079
0.209
0.135
0.197
0.463
0.419
0.360
0.412
0.117
0.152
0.020
Yes
Yes
Yes
No
No
Yes
No
Yes
No
No
Yes
Yes
No
No
No
No
No
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
Yes
Direction
UD
BD
UD
UD
UD
UD
BD
UD
BD
BD
UD
UD
BD
BD
BD
UD
UD
UD
UD
countries, and there was no such relationship in the remaining 14. As for the
countries involving a causality relationship, there is a uni-directional relationship
from energy consumption to rural population in four countries (Algeria, Egypt,
Morocco and Sri Lanka), a uni-directional relationship from rural population to
energy consumption in nine countries (Argentina, Cote Dlvoire, El Salvador,
Gabon, Korea, Mexico, Nigeria, Paraguay and Uruguay), and a bi-directionalcausality relationship between energy consumption and rural population in three
countries (Ghana, Pakistan and Thailand).
62
Table 4.4 Granger causality test results (energy consumption-gross domestic product)
Countries
Independentdependent
Independentdependent
Direction
Algeria
Argentina
Bolivia
Cameroon
Congo
Costa Rica
Cote Dlvoire
Egypt
El Salvador
Gabon
Ghana
Honduras
India
Korea
Malaysia
Mexico
Morocco
Nigeria
Pakistan
Paraguay
Senegal
Sri Lanka
Sudan
Syria
Thailand
Togo
Trinidad Tobago
Turkey
Uruguay
Venezuela
Energy consumption-GDP
GDP-Energy consumption
Prob.
Causality
Prob.
Causality
0.348
1.845
2.505a
1.006
2.743a
3.568a
1.695
4.561b
2.124
9.879c
2.977a
2.540a
1.792
0.588
2.291
0.121
4.433b
1.378
1.594
1.107
1.594
0.147
2.736a
2.223
0.056
3.545a
1.093
0.761
1.775
2.551a
0.790
0.183
0.068
0.423
0.080
0.067
0.200
0.018
0.137
0.00007
0.066
0.095
0.183
0.448
0.139
0.729
0.042
0.248
0.208
0.343
0.215
0.703
0.080
0.145
0.945
0.068
0.348
0.389
0.175
0.094
No
No
Yes
No
Yes
Yes
No
Yes
No
Yes
Yes
Yes
No
No
No
No
Yes
No
No
No
No
No
Yes
No
No
Yes
No
No
No
Yes
3.669b
0.119
2.241a
0.805
2.027
0.365
1.847
2.973a
2.172
1.408
4.100b
1.565
2.416
2.170
7.500c
0.289
3.955a
0.855
3.669b
0.392
1.941
9.520c
0.336
4.119a
1.228
3.635a
0.917
0.864
2.300a
1.962
0.024
0.731
0.094
0.534
0.149
0.549
0.175
0.066
0.131
0.261
0.026
0.225
0.106
0.150
0.009
0.594
0.055
0.361
0.018
0.678
0.172
0.004
0.716
0.051
0.307
0.065
0.410
0.359
0.099
0.158
Yes
No
Yes
No
No
No
No
Yes
No
No
Yes
No
No
No
Yes
No
Yes
No
Yes
No
No
Yes
No
Yes
No
Yes
No
No
Yes
No
UD
BD
UD
UD
BD
UD
BD
UD
UD
BD
UD
UD
UD
UD
BD
UD
UD
63
Table 4.5 Granger causality test results (energy consumption-consumer price index)
Countries
Independentdependent
Independentdependent
Direction
Algeria
Argentina
Bolivia
Cameroon
Congo
Costa Rica
Cote Dlvoire
Egypt
El Salvador
Gabon
Ghana
Honduras
India
Korea
Malaysia
Mexico
Morocco
Nigeria
Pakistan
Paraguay
Senegal
Sri Lanka
Sudan
Syria
Thailand
Togo
Trinidad Tobago
Turkey
Uruguay
Venezuela
Energy consumption-CPI
CPI-Energy consumption
Prob.
Causality
Prob.
Causality
1.410
0.627
2.141
0.333
2.532a
1.453
0.652
4.805c
0.822
7.315b
0.018
1.414
7.336c
5.928c
4.730b
3.131a
0.155
1.622
13.975c
2.916a
0.101
0.004
1.499
5.055b
4.780b
0.319
0.749
5.502c
3.169a
2.262
0.261
0.603
0.135
0.567
0.096
0.249
0.425
0.005
0.448
0.010
0.892
0.258
0.002
0.003
0.036
0.058
0.695
0.214
0.00005
0.069
0.751
0.944
0.239
0.012
0.036
0.575
0.393
0.009
0.056
0.121
No
No
No
No
Yes
No
No
Yes
No
Yes
No
No
Yes
Yes
Yes
Yes
No
No
Yes
Yes
No
No
No
Yes
Yes
No
No
No
Yes
No
1.356
0.404
2.107
2.527
0.665
1.843
0.377
1.261
1.325
0.343
1.626
0.784
1.325
2.141
4.070a
0.759
0.820
0.431
0.903
2.589a
0.823
4.024a
3.257a
1.420
0.115
2.641
1.830
1.825
2.089
1.404
0.277
0.750
0.139
0.121
0.521
0.163
0.543
0.312
0.280
0.561
0.210
0.465
0.280
0.118
0.051
0.476
0.371
0.653
0.415
0.091
0.370
0.053
0.052
0.257
0.736
0.113
0.185
0.178
0.141
0.261
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
No
No
No
No
Yes
No
Yes
Yes
No
No
No
No
No
No
No
UD
UD
UD
UD
UD
BD
UD
UD
BD
UD
UD
UD
UD
UD
bi-directional causality relationship between energy consumption and total population in seven countries (Argentina, Egypt, Ghana, Honduras, Nigeria, Pakistan
and Paraguay).
It is clear from the results of the causality relationship between energy consumption and GDP (see Table 4.4) that an inter-variable causality relationship
exists in 17 countries, while the remaining 13 countries lack such a relationship.
As for the countries involving a causality relationship, there is a uni-directional
relationship from energy consumption to GDP in six countries (Congo, Costa Rica,
Gabon, Honduras, Sudan and Venezuela), a uni-directional relationship from GDP
64
Direction
CO2-Energy consumption
Prob.
Causality F
Prob.
Causality
7.776c
0.681
1.105
0.031
1.216
3.190b
0.002
3.027a
2.139
0.328
7.874c
3.411a
14.932c
5.363b
8.821c
2.589a
8.342c
2.718
131.319c
1.552
0.034
0.292
1.096
8.490c
4.695b
10.062c
0.355
0.008
0.414
0.300
0.969
0.277
0.039
0.961
0.091
0.135
0.570
0.008
0.073
0.00003
0.026
0.005
0.062
0.006
0.108
0.0000000000005
0.221
0.854
0.592
0.302
0.006
0.016
0.003
0.555
Yes
No
No
No
No
Yes
No
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
No
No
No
No
Yes
Yes
Yes
No
0.00075
0.031
14.114c
1.810
2.537
0.739
1.328
0.00069
9.833b
4.395b
1.764
0.231
4.139b
1.516
2.416
3.568b
7.566c
0.000003
5.314b
0.951
0.314
0.035
13.236c
8.126c
1.163
0.484
0.309
0.978
0.860
0.0007
0.1810
0.120
0.538
0.257
0.979
0.0005
0.043
0.193
0.633
0.025
0.226
0.129
0.020
0.009
0.998
0.027
0.336
0.578
0.852
0.0009
0.007
0.326
0.491
0.581
No
No
Yes
No
No
No
No
No
Yes
Yes
No
No
Yes
No
No
Yes
Yes
No
Yes
No
No
No
Yes
Yes
No
No
No
UD
UD
UD
UD
UD
UD
UD
UD
BD
UD
UD
BD
BD
BD
UD
BD
UD
UD
0.752
0.489
3.622b
0.391
0.617
0.039
No
No
Yes
1.428
1.608
0.774
0.240
0.217
0.470
No
No
No
UD
65
Egypt, Gabon, India, Korea, Mexico, Pakistan, Syria, Thailand and Uruguay) have
a uni-directional relationship from energy consumption to CPI, two countries (Sri
Lanka and Sudan) have a uni-directional relationship from CPI to energy consumption, and two countries (Malaysia and Paraguay) have a bi-directional causality relationship between energy consumption and CPI.
As is clearly seen from the results of the causality relationship between energy
consumption and CO2 (see Table 4.6), 19 countries involve an inter-variable
causality relationship, which the remaining 11 countries lack. As for the countries
with a causality relationship, 10 countries (Algeria, Costa Rica, Egypt, Ghana,
Honduras, Korea, Malaysia, Thailand, Togo and Venezuela) have a uni-directional
relationship from energy consumption to CO2, four countries (Bolivia, El Salvador, Gabon and Sudan) have a uni-directional relationship from CO2 to energy
consumption, and five countries (India, Mexico, Morocco, Pakistan and Syria)
have a bi-directional causality relationship between energy consumption and CO2.
4.6 Conclusion
The present study investigates the causality relationship between energy consumption and the factors with the greatest impact on energy consumption as
demonstrated in the literature (rural population, total population, gross domestic
product, consumer price index and carbon dioxide emission) for 30 developing
countries between 1971 and 2007.
As suggested in the literature, quite complex and different causality relationships
are found among the variables in question (Soytas and Sari 2003; Fatai et al. 2004; Oh
and Lee 2004; Ghali and El-Sakka 2004; Yoo 2005; Hatemi and Irandoust 2005; Lee
2006; Yoo 2006; Ang 2007; Narayan and Singh 2007; Odhiambo 2010; Barleet and
Gounder 2010; Kapusuzoglu and Karan 2010) and there is no causality relationship
in certain circumstances (Murry and Nan 1996; Wolde-Rufael 2006; Zachariadis
2007; Payne 2009; Kebede et al. 2010). In the light of the results, it is clear that the
presence or absence of such relationships originate from the country specific characteristics rather than from the impact of variables. To put it differently, factors
including energy security of countries, their proximity to raw materials, potentials for
industrialization, energy production capacities, economic development levels,
population and its distribution are highly important in the process.
The results of the analyses revealed total population (19 countries) and CO2
emission (19 countries) as the variables with the highest causality relationships with
energy consumption for the developing countries under study. These results conform
to expectations as the population of countries also signifies the number of consumers
they have. Therefore, possible rises or declines in population are expected to have a
close relationship with energy consumption. The amount of CO2 emission is also
closely correlated with energy consumption. The amount of CO2 emission, which
could be explained as the harmful gases produced as a result of energy consumption,
varies with the changes in consumption levels. However, the change in the amount of
66
CO2 emission may not be directly proportional to energy consumption. Since the
same amount of output can be obtained by using smaller amounts of energy through
country energy efficiency and quality levels, less energy consumption needs less
energy production, thus also producing less CO2 emissions.
The consumer price index (CPI) was found to be the variable with the least
causality relationship with energy consumption for the countries under study. The
consumer price index is a variable used to indicate the impact of energy prices upon
the level of energy consumption. Moreover, it also provides insight into the flexibility
of the energy-demanding consumer units in these countries toward energy price. On
the basis of the results, it could be argued that the developing countries under study
have lower levels of flexibility toward energy prices, or in other words, energy
consumption levels are not strongly affected by the changes in energy prices.
By examining the analysis results in terms of GDP, a crucial indicator in the
process of economic growth, it is concluded that there exists a causality relationship
between energy consumption and GDP in 17 of the countries under study. Of the two
fundamental approaches examining the relationship between energy and economic
growth, the ecological approach, which considers energy as the main factor in the
process of economic growth, applies to the countries involving such a relationship;
while in the countries lacking such a relationship, the neoclassical approach is valid
as it does not regard energy as a crucial factor of production and simply focuses on
capital and labor as the main factors of production.
As a result, the present study reveals common relationships in various directions between energy consumption and the other factors. These results can be
explained by factors such as energy markets, resources, population for individual
countries. Therefore, empirical evidence on economic growth in the developing
countries over recent years can be explained by the ecological approach and it
seems at first sight to suggest some degree of decoupling. These results indicate
highly complicated relationships between the variables. Obviously, it is difficult to
make a generalization based on these results, and to arrive at a conclusion by
grouping the developing countries under study. Still, it may be possible to interpret
inter-variable relationships by considering the specific characteristics of individual
countries. It can be argued that the policy makers of developing countries should
also include the other socio-economic factors when considering the issue of energy
consumption by being aware that GDP is not the sole determining factor and
should take into consideration these other factors in any analytical models they
establish in relation to energy consumption.
References
Aghion, P., & Howitt, P. (1998). Endogenous Growth Theory. Cambridge, MA: MIT Press.
Akinlo, A. E. (2008). Energy consumption and economic growth: Evidence from 11 Sub-Sahara
African countries. Energy Economics, 30, 23912400.
Ang, J. B. (2007). CO2 emissions, energy consumption and output in France. Energy Policy, 35,
47724778.
67
Apergis, N. & Payne, J.E. (2010). Renewable energy consumption and ecnomic growth: Evidence
from a panel of OECD countries. Energy policy, 38, 656660.
Ayres, R. U., & Kneese, A. V. (1969). Production, consumption and externalities. American
Economic Review, 59, 282297.
Aytac, D. (2010). Enerji ve ekonomik byme iliskisinin ok degiskenli VAR yaklasm ile
tahmini.Maliye Dergisi, 158, 482495.
Barleet, M., & Gounder, R. (2010). Energy consumption and economic growth in New Zealand:
Results of trivariate and multivariate models. Energy Policy, 38, 35083517.
Barro, R. J. (1988). Governmentspending in a simple model of endogenous growth. (NBER
Working Paper, No. 2588). Washington, DC: NBER.
Bentham, A., & Romani, M. (2009). Fuelling growth: What drives energy demand in developing
countries. The Energy Journal, 30(3), 147170.
Burbridge, J., & Harrison, A. (1984). Testing for the effects of oil-price rises using vector auto
regression. International Economic Review, 25, 459484.
Chimobi, O. P., & Igwe, O. L. (2010). Budget deficit, money supply and inflation in Nigeria.
European Journal of Economics, Finance and Administrative Sciences, 19, 5260.
Cleveland, C. J., Costanza, R., Hall, C. A. S. & Kaufmann, R. (1984). Energy and the U.S.
economy: A biophysical perspective. Science, 225(4665), 890897.
Costanza, R. (1980). Embodied energy and economic valuation. Science, 210, 12191224.
Dickey, D., & Fuller, W. (1979). Distribution of the estimators for autoregressive time series with
a unit root. Journal of the American Statistical Association, 74(366), 427431.
Enders, W. (1995). Applied Econometric Time Series. New York, NY: John Wiley & Sons.
Fatai, K., Oxley, L., & Scrimgeour, F. G. (2004). Modelling the causality relationship between
energy consumption and GDP in New Zealand, Australia, India, Indonesia, The Philippines
and Thailand. Mathematics and Computers in Simulation, 64, 431445.
Gever, J., Kaufmann, R. K., Skole, D., & Vorosmarty, C. (1986). Beyond Oil: The Threat to Food
and Fuel in the Coming Decades. Cambridge, MA: Ballinger.
Ghali, K. H., & El-Sakka, M. I. T. (2004). Energy use and output in Canada: A multivariate
cointegration analysis. Energy Economics, 26, 225238.
Granger, C. W. J. (1969). Investigating causality relations by econometric models and crossspectral methods. Econometrica, 37(3), 424438.
_
Gujarati, D. N. (1999). Temel ekonometri (ev. U. Senesen ve G.G. Senesen). Istanbul,
Turkey:
Literatr Yaynlar.
_ kisi: OECD lkelerine
Guvenek, B., & Alptekin, V. (2010). Enerji Tketimi ve Byme Ilis
_ kin Bir Panel Veri Analizi. Enerji, Piyasa ve Dzenleme, 1(2), 172193.
Ilis
Hall, C. A. S., Cleveland, C. J., & Kaufmann, R. K. (1986). Energy and resource quality: the
ecology of the economic process. New York, NY: Wiley Interscience.
Hamilton, J. D. (1983). Oil and the macro economy Since World War II. Journal of Political
Economy, 91, 228248.
Hannon, B. (1973). The structure of the ecosystem. Journal of Theoretical Biology, 41, 535546.
Hatemi, A., & Irandoust, M. (2005). Energy consumption and economic growth in Sweden: A
leveraged bootstrap approach, (19652000). International Journal of Applied Econometrics
and Quantitative Studies, 2(4), 8798.
IMF (2010). International Monetary Fund emerging and developing economies list, World
economic outlook database. Washington, DC: IMF. Retrieved from http://www.imf.org.
Kaufmann, R. K. (1987). Biophysical and Marxist economics: Learning from each other.
Ecological Modelling, 38, 91105.
Kaufmann, R. K. (1992). A biophysical analysis of the energy/real GDP ratio: Implications for
substitution and technical change. Ecological Economics, 6, 3556.
Kaufmann, R. K. (2004). The mechanisms for autonomous energy efficiency increases: A
cointegration analysis of the US energy/GDP ratio. Energy Journal, 25(1), 6386.
Kapusuzoglu, A., & Karan, M. B. (2010). An analysis of the co-integration and causality
relationship between electricity consumption and gross domestic product (GDP) in the
68
69
Part II
Chapter 5
The authors are associate professor at the Faculty of Economics and Business of the University
of Groningen, energy analyst at the Groningen-based Energy Delta Institute, and assistant
professor of the Faculty of Economics and Business of the University of Groningen,
respectively.
H. von Eije (&) W. Westerman
University of Groningen, P.O. Box 800 9700 AV, Groningen, The Netherlands
e-mail: j.h.von.eije@rug.nl
S. von Eije
Energy Delta Institute, P.O. Box 11073 9700 CB, Groningen, The Netherlands
73
74
5.1 Introduction
Economic growth requires energy consumption. Because energy consumption still
mainly comes from the fossil fuels, carbon dioxide (CO2) emissions occur. Such
emissions are very likely to generate global warming effects.1 Global warming
threatens the fundamentals of living for many, through lower access to water, food,
health, land, and the environment as well as sudden climate changes (Stern 2006).
It is thus of utmost importance to reduce CO2 emissions by diminishing the
magnitude of fossil fuel consumption. However, humanity is assumed to benefit
from economic growth.2 It is therefore relevant to increase the use of substitutes
for fossil fuel consumption and the capacity to produce such substitutes in order to
reduce CO2 emissions while retaining economic growth.
Relationships between energy consumption and economic growth have been
studied for a long time by many authors. In economics the first attempts are based
on including energy in the production function together with labor, capital and
materials. Later various types of energy are distinguished (Fuss 1977) and dynamic
relationships are studied (Pindyck and Rotemberg 1983). The relationships are also
studied the other way around, namely when energy consumption is derived from
the demand for a firms produced output (Berndt and Wood 1975). Also energy
prices are considered to influence output (McMillin and Smyth 1994). The various
causal relationships suggest the idea that the relationships may not be unilateral,
but bi-directional. New techniques, like the Granger causality and Vector Error
Correction models, nowadays enable researchers to study such mutual interactions
too (Alam et al. 2011; Apergis and Payne 2009; Chontanawat et al. 2008). This
paper builds upon the existing literature and extends it by incorporating the consumption of renewable energy and the production capacity of renewable energy.3
It appears that, no academic literature exists that takes into account both renewable
energy consumption and renewable energy generating capacity in the context of
economic growth. This work is therefore in line with the call for new approaches
as suggested in the literature review of Ozturk (2010).
1
While it is a fact that most of the CO2 emissions are not man-made, it is assumed that the
equilibrium of CO2 generation and CO2 use may be disturbed if humanity adds the CO2
embedded in fossil fuels to the atmosphere.
2
Economic growth may also have negative externalities not related to CO2 emissions, like the
destruction of tropical forests and habitats, the extinction of species, and environmental pollution.
3
The term renewable energy is preferred to alternative energy. Alternative energy sources
have not been considered so previously. Historically the use of wood and other plant products was
the major source of energy, and it is still an important source of energy in many developing
countries. Later, wind energy and water power became important energy sources. It will be clear
that these sources would then not be considered to be alternative sources. Only now, after the
industrial revolution where the massive use of coal, gas and oil became the major sources of
worldwide energy production, one may call them alternative energy uses, because they do not use
the fossil fuels generated during the ages of the existence of the world. Nowadays, also solar
energy, tide and wave based electricity and waste based sources of energy production are also
included amongst the (alternative or) renewable sources.
75
76
1980
1985
1990
ECON
RENCON
1995
2000
2005
NUCON
FOSC
Fig. 5.1 Total energy consumption and its major constituents in ktoe, (World, 19712008).
Source Data downloaded from the World Data Bank from the World Bank (http://
databank.worldbank.org/ddp/home.do). ECON is total energy consumption, FOSC is fossil fuel
consumption, NUCON is nuclear energy consumption, RENCON is all renewable energy
consumption
transport, usually by using oil, (2) heat demand, usually provided by gas in
developed countries, and (3) the use of electric appliances, for which electricity is
provided by nuclear sources, gas, coal and to a lesser extent oil. Each form of
renewable energy substitutes for one or more of these fuels. Figure 5.1 indicates
that energy consumption is still based primarily on fossil fuels and that the use of
renewables shows an upward trend too, except for the drop in 1990.4
Biomass is the largest source of renewable energy used globally. It can serve as
a substitute for each type of end use demand, though it is mainly exploited to
substitute heat demand in the form of cooking and heating houses. This form of
renewable energy, in contrast to the others, is mostly utilized in developing
countries. However, the consumption of biomass in developing countries for
heating and cooking is not captured by global overviews of renewable energy
consumption. The reason is that, in contrast to the other renewable sources, it is not
traded, and certainly not in formal markets. The most important source for measured renewable energy consumption originates from hydropower.
Figure 5.2 shows the consumption of energy by fuel type. The total contribution
of renewable energy is only 7.8 % (rounded to 8 % in Fig. 5.2). This share in turn
The World Bank database reports a large decline in renewable energy consumption in 1990. It
is likely that this is caused by a different approach in measuring such consumption. After 1990 the
gradual trend in renewable energy is visible again.
77
30%
Oil
Natural Gas
5%
1%
8%
24%
6%
Coal
Nuclear Energy
Hydro electricity
34%
Renewables
Fig. 5.2 World consumption by fuel type, 2010. Source Data from BP (2011)
78
levelized costs of fossil fuel generating capacity. Only when for carbon intensive
technologies, a 3 % points increase in the cost of capital is added to represent the
projected carbon price, the levelized costs are comparable.
Until now, renewable electricity generation capacity has thus been more expensive to build than fossil electricity generation capacity. A shift from the traditional
fossil fuel electricity generation to renewable electricity generation may have made
the electricity use on a macro-economic level more expensive and it may therefore
have had a negative impact on economic growth. Moreover, many governments are
willing to subsidize renewable energy projects. This will influence decision making
by private actors and shift the focus to renewable energy projects, but on the macro
level, at which this analysis is performed, subsidies are only a reallocation of public
means that could have been used to make a positive contribution to economic growth,
instead of stimulating uneconomical decision making. It is therefore even possible
that subsidies for renewable production capacity options have aggravated any negative impact of renewables on economic growth.
Obviously there are many more determinants for the generation of renewable energy, like
population growth, technological development, global politics on the availability of oil, local
politics on the stimulation of renewable energy sources, and the availability of capital and of
natural resources needed to generate capacity for renewable energy production. Here a
simplification is made by equating the price of renewable energy to the price of fossil energy. Of
course subsidies on renewable energy and CO2 emission costs may influence the price of nonfossil fuels, while also autonomous demand changes may have their impact. Furthermore, fuel
switching other than from fossil fuel to renewable energy has been ignored. Finally, in the figure
the impact of nuclear energy is discarded.
79
Economic growth
Fossil fuel
reserves and
production
capacity
Fossil fuel
consumption
Renewable
energy
production
capacity
Renewable energy
consumption
CO2
Emissions
Fig. 5.3 The major relations between economic growth and CO2 emissions
5.4 Data
Annual data for the world is retrieved and we require that at least data are available
for 2008. A search is made for data that goes back as many years as possible. Data
from the Statistical Review of World Energy 2011 of British Petroleum (BP 2011)
on CO2 emissions and for fossil fuel prices are used.6 Data are gathered from the
World Bank on GDP (in fixed US dollars of the year 2000), energy consumption
and energy production (in ktoe), and the combustible renewables and waste use (in
The Data of the World Bank on CO2 emissions started in 1960, but ended in 2007. The value of
CO2 emissions presented by BP started in 1965 and ended in 2009. During overlapping years the
emissions reported by BP are on average 0.1 % larger than emissions recorded by the World
Bank. Energy prices are measured by real oil prices in dollars per barrel against US 2009 prices (a
series which goes back to 1861).
80
The complement of the fossil fuel consumption percentage thus includes nuclear energy. For
the combustible renewables and waste measured in kiloton oil equivalent there are no data
available for 2007 and 2008. The data for these years is extended by using the percentage of these
fuels in energy use for 2007 and 2008, also provided by the World Bank. Because the numbers
presented are not fully consistent with the derivation of the same number from the percentages
and energy use, the kiloton oil equivalent levels are used which are calculated from the
percentages and the kiloton levels directly presented by the World Bank for 2006 to make the
data for 2007 and 2008 consistent with the previous numbers.
81
Std. Dev.
JBP
LCO2
LGDP
LECON
LFOS
LNONFOS
LRENCON
LRENCAP
LPRICE
DLCO2
DLGDP
DLECON
DLFOS
DLNONFOS
DLRENCON
DLRENCAP
DLPRICE
0.177
0.327
0.198
0.186
0.270
0.188
0.341
0.520
0.019
0.013
0.017
0.020
0.026
0.029
0.016
0.306
0.564
0.347
0.518
0.478
0.156
0.287
0.452
0.700
0.810
0.917
0.658
0.790
0.000
0.000
0.041
0.000
10.010
30.792
15.934
15.724
14.265
14.045
14.058
3.671
0.019
0.031
0.021
0.020
0.026
0.020
0.036
0.057
10.028
30.821
15.964
15.753
14.345
14.088
14.109
3.570
0.020
0.033
0.021
0.021
0.028
0.023
0.032
0.026
10.359
31.333
16.292
16.082
14.627
14.351
14.655
4.574
0.055
0.063
0.058
0.062
0.066
0.045
0.080
1.155
9.695
30.226
15.570
15.404
13.692
13.633
13.395
2.543
-0.026
0.004
-0.010
-0.022
-0.100
-0.141
0.015
-0.665
The table presents the characteristics of 38 annual observations. Variables preceded by L are
logarithmic transformations, and variables preceded by DL are first differences of logarithmic
transformations (and thereby start in 1972). CO2 is the CO2 production in kiloton, GDP is gross
domestic product in dollars at constant prices of 2000, ECON is energy consumption in ktoe, FOS
is fossil fuel consumption in ktoe, NONFOS is non-fossil fuel consumption in ktoe, RENCON is
all renewable energy consumption, RENCAP is the installed renewable electricity capacity,
PRICE is the price of fossil fuel. JBP represents the probability of the JarqueBera test on
normality
for the preceding years.8 Next a study is made of the percentage of total renewable
electricity in total electricity production for the period 1980 until 2008. This ratio
is 23.5 % in 1980 and 22.8 % in 2008 and it has a marginal, but significant, annual
decline of 0.056 %. Next the renewable energy percentage in total electricity
capacity each year before 1980 is reduced by 0.056 %. Then, the total calculated
electricity capacity is multiplied by the outcomes in order to get the renewable
capacity for the years 19711980. When first differences are taken, the database
starts in 1972, implying that there is a maximum of 37 years of observations. The
characteristics of the natural logarithm and of the differences in the natural logarithm for the data are presented in Table 5.1.
Figure 5.4 shows the growth in CO2 emissions, in relation with the growth of
GDP and the growth in energy consumption. The figure indicates that economic
growth and primary energy consumption are strongly related, while the
Corrections are made for leap years. This procedure implies that the electrical capacity is
calculated based on the assumption that 1980 was a normal electrical production and
electricity capacity year, that the relation between production and capacity is not cyclical before
1980 and that the development in production is capable of measuring the development in
electrical capacity.
82
.06
.04
.02
.00
-.02
-.04
1975
1980
1985
DLCO2
1990
DLGDP
1995
2000
2005
DLFOS
Fig. 5.4 The development of the relative annual change in carbon dioxide emissions, GDP and
primary energy consumption (World, 19722008). DLCO2 is the annual change in the natural
logarithm of CO2 emissions, DLGDP is the annual change in the natural logarithm of GDP,
DLFOS is the annual change in the natural logarithm of fossil fuel consumption
consumption of fossil fuels and CO2 emissions are very strongly related.9 A
remarkable aspect of Fig. 5.4 is that the shifts in economic growth coexist with
larger fluctuations in primary energy consumption and in CO2 emissions. If there
would be a unilateral instantaneous causation from GDP growth to primary energy
consumption, it would imply that the energy elasticity of worldwide GDP is larger
than one. This effect is, however, not significant, as the standard error is even
larger than the difference between the coefficient and one. To a smaller extent this
is also the case for the relationship between primary energy consumption and CO2
emissions, but also here these effects are not significantly different from one.
This is caused by the fact that the use of each type of fossil fuels has its own standard global
average CO2 emission consequences. Per TJ (terajoule = 1012 joules) coal emits 94,600 kg CO2,
oil 73,300 kg CO2, and gas 56,100 kg CO2 (BP 2011).
0.189
(0.092)**
0.828
(0.065)***
0.102
(0.106)
0.003
(0.003)
-0.089
(0.069)
-0.015
(0.018)
-0.002
(0.003)
37
0.94
0.492
(0.077)***
0.091
(0.150)
-0.001
(0.003)
0.082
(0.109)
0.004
(0.026)
0.017
(0.005)***
37
0.67
-0.175
(0.169)
0.005
(0.005)
0.159
(0.190)
-0.005
(0.029)
-0.019
(0.006)***
37
0.68
1.152
(0.222)***
0.007
(0.007)
0.190
(0.102)*
-0.161
(0.003)***
0.015
(0.007)**
37
0.89
0.143
(0.232)
-0.118
(0.116)
-0.597
(3.885)
1.243
(1.253)
-0.097
(0.094)
37
0.09
-2.226
(5.114)
3.506
(2.401)
7.066
(7.902)
(5)
DLPRICE
0.046
(0.031)
0.014
(0.010)
37
0.31
0.268
(0.385)
0.221
(0.275)
0.391
(0.188)**
-0.001
(0.008)
(6)
DLRENCAP
The first letter (D) represents the annual change, the second letter (L) is the natural log transformation, CO2 is the fossil fuel CO2 production in kiloton, FOS
is fossil fuel consumption in ktoe, RENCON is renewable energy consumption in ktoe. GDP is gross domestic product in US dollars at constant prices of
2000. PRICE is the price of fossil fuel. RENCAP is the installed renewable electricity capacity in million kilowatts. HAC robust standard errors are provided
in parentheses below the coefficients. * , ** , and *** represent significance at 10, 5, and 1 % respectively
Observations
R-squared
Constant
DUM90
DLRENCAP
DLPRICE
DLRENCON
DLFOS
DLGDP
Table 5.2 The statistical dependence of the relative changes in the relevant variables, (World, 19712008)
(1)
(2)
(3)
(4)
DLGDP
DLFOS
DLRENCON
DLCO2
84
Use of panel models (as used for example by Mahadevan and Asufu-Adjaye (2007); Apergis
and Payne (2009), and Sadorsky (2009)) is not considered for this analysis; also because
equilibrium equations and reactions of the various variables may be very different for different
countries, even if one includes panel intercepts.
85
Table 5.3 Test on the validity and the type of error correction model (World, 19712008)
No constant no trend
Intercept only
LogL
LR
FPE
No Intercept
No Trend
Intercept
No Trend
AIC
SC
HQ
-10.908
-23.224
-23.374
-23.223
-25.075*
-10.684
-21.877*
-20.905
-19.632
-20.361
-10.831
-22.765
-22.532
-21.998
-23.467*
Intercept
No Trend
Intercept
Trend
Intercept
Trend
3
1
5
1
these relations are cointegrated. This requires that each of the variables should
have a unit root. Table 5.3, Panel A, shows the adjusted DickeyFuller tests and it
shows that the existence of a unit root can only be rejected for the natural log of
GDP if an intercept and a constant term is assumed.
Second, the lag structure is determined. Table 5.3, Panel B, shows various VAR
tests for different sized lags. While most of these tests conclude on the use of four
lags, the Schwartz information criterion suggests the use of one lag. Because the
interpretation of a set of equations with one lag is more economical and because it
is an annual dataset and only a relatively small number of observations, only one
lag is used. Finally, the Johanson cointegration test based on one lag is used to
study how many cointegration factors might be needed (Table 5.3, Panel C). Again
for economical reasons, though the results again differ, only one cointegration
factor is applied. The vector error correction results are presented in Table 5.4.
86
Table 5.4 Vector error correction results for the period (World, 19712008)
Panel A The contegrating equation
LFOS(-1)
LRENCON(-1)
1.000
0.666
[6.417]
LGDP(-1)
-2.831
[-12.378]
LPRICE(-1)
-0.024
[-3.179]
LRENCAP(-1)
-0.459
[-3.604]
TREND
0.071
[10.543]
C
67.215
Panel B Error correction model
DLFOS
bi
0.517
[4.230]c
DLFOS(-1)
0.303
[1.239]
DLRENCON(-1)
0.169
[-1.334]
DLGDP(-1)
-0.095
[-0.251]
DLPRICE(-1)
-0.015
[-2.010]*
DLRENCAP(-1)
-0.317
[-1.728]a
C
0.0305
[4.045]c
DUM90
0.029
[2.125]b
R-squared
0.647
Adj. R-squared
0.558
Akaike AIC
-5.608
Schwarz SC
-5.257
Log likelihood
Akaike information criterion
Schwarz criterion
DLRENCON
0.148
[1.782]a
-0.098
[0.586]
0.057
[0.665]
-0.016
[-0.064]
0.005
[0.867]
0.305
[2.437]b
0.017
[3.316]c
-0.164
[-17.690]c
0.924
0.905
-6.375
-6.023
486.300
-24.461
-22.438
DLGDP
0.395
[5.770]c
-0.013
[-0.094]
-0.138
[-1.940]a
0.297
[1.405]
-0.008
[-1.806]*
-0.175
[-1.705]
0.031
[7.297]c
0.002
[0.307]
0.719
0.649
-6.768
-6.416
DLPRICE
-3.078
[-1.172]
10.516
[2.002]*
4.310
[1.583]
-8.137
[-1.002]
-0.102
[-0.623]
6.319
[1.603]
-0.200
[-1.235]
0.140
[0.480]
0.320
0.149
0.528
0.879
DLRENCAP
0.276
[2.635]b
-0.225
[-1.075]
0.083
[-0.767]
0.073
[0.224]
0.007
[1.123]
0.418
[2.662]b
0.024
[3.694]c
-0.012
[-0.993]
0.528
0.410
-5.918
-5.566
The letter (D) represents the annual change, the letter (L) is the natural logarithmic transformation, FOS is fossil fuel consumption in ktoe, RENCON is renewable energy consumption in
ktoe. GDP is gross domestic product in US dollars at constant prices of 2000. PRICE is the price
of fossil fuel. RENCAP is the installed renewable electricity capacity in million kilowatts. (-1)
indicates that the variable is one year lagged. TREND is a time trend and C is the constant of the
equation. DUM90 is a dummy that takes the value of 1 in the year 1990, and zero otherwise. bi
represents the speed of adjustment of the variables back to equilibrium. The t-values are shown in
brackets below the coefficients. a , b , and c represent significance at 10, 5, and 1 % respectively
87
Being the first to test the relations with renewable capacity, a test is undertaken
to determine if the use of renewable capacity in this system of equations is significant. It is tested to see if it can be excluded from the cointegration relation. This
gives a Chi-square with a p-value of 0.026. This means that the renewable capacity
from the equilibrium system cannot be excluded. Next step is a test to see if the
adjustment speed of the renewable capacity can be left out. This gives a Chi-square
with a p-value just below 0.050, also implying that the adjustment speed coefficient should be incorporated.
For the error correction part, relatively high adjusted R-squareds for the growth
rates of fossil fuel consumption (0.558), renewable energy consumption (0.905) and
GDP (0.649) are found. The lowest explanatory power goes to the energy prices.
The t-value presented in the error correction part of the table can be used to
evaluate the significance of the ai. These coefficients indicate how the variables
react to each other in the short run. Critical 10 % significant t-values with 36
observations and 8 coefficients in the equation are 1.701, while the critical values
for 5 and 1 % are 2.048 and 2.763. Here, the significant coefficients are discussed.
It is found that the fossil fuel consumption reacts relatively fast to deviations of the
equilibrium relation (coefficient of 0.517). Fossil fuel consumption is, moreover,
negatively influenced by the energy price, though the coefficient is small (-0.015).
Fossil fuel consumption is also negatively influenced by growth in renewable energy
capacity (-0.317). Fossil fuel consumption has an autonomous growth of 3.050 %.
Finally, the dummy that corrects for a possible misspecification of the World Bank
data of the fossil fuel use percentage is significantly positive.
Renewable consumption is slowly reacting to deviations from the equilibrium
(0.148), while one percent production capacity growth results the next year in a
0.3 % increase of the consumption of renewables. Also, there is an autonomous
trend of 1.7 % per year. The 1990 dummy is negative and highly significant for
renewable energy consumption.
GDP adjustment to distortions of equilibrium (coefficient of 0.395) is somewhat
slower than the reaction of fossil fuel consumption. Contrary to expectations, it is
found that growth in renewables consumption negatively affects the growth in
GDP (coefficient -0.138). An increase in fossil fuel prices negatively affects GDP
growth. GDP growth is also more than 3 % autonomous.
For the price of fossil fuels, only a 10 % significant positive effect of the growth
in the use of fossil fuels is found, but this coefficient is large (10.516).
Renewable capacity is mildly (coefficient 0.276) reacting to disturbances from
the equilibrium. If renewable capacity growth has been large last year, it is likely
that this is also the case this year (coefficient of 0.418).11 Renewable capacity is,
finally, also growing autonomously at 2.4 % per year.
11
While Marquez and Fuinhas (2011) find a persistency in renewable energy consumption, such
a positive relationship between the current and last years change in renewable energy
consumption are not found. In the present model these effects are captured through renewable
energy generation capacity.
88
.004
.002
.000
-.002
-.004
1
10
10
Fig. 5.5 Generalized response graphs for fossil fuel consumption and GDP caused by impulses
of renewable capacity (World, 19712008)
89
90
References
Alam, M. J., Begum, I. A., Buysse, J., Rhaman, S., & Van Huylenbroeck, G. (2011). Dynamic
modeling of causal relationship between energy consumption, CO2 emissions and economic
growth in India. Renewable and Sustainable Energy Reviews, 15(6), 32433251.
Apergis, N., & Payne, J. E. (2009). Energy consumption and economic growth in Central
America: Evidence from a panel cointegration and error correction model. Energy Economics,
31, 211216.
BP. (2011). Statistical Review of World Energy 2011. Retrieved from http://www.bp.com/
sectionbodycopy.do?categoryId=7500&contentId=7068481.
Berndt, E. R., & Wood, D. O. (1975). Technology, prices, and the derived demand for energy.
Review of Economics and Statistics, 57(3), 259268.
Chontanawat, J., Hunt, L. C., & Pierse, R. (2008). Does energy consumption cause economic
growth? Evidence from a systematic study of over 100 countries. Journal of Policy Modeling,
30, 209220.
Energy Information Administration (EIA). (2011). Annual Energy Outlook 2011. Retrieved from
http://www.eia.gov/forecasts/aeo/pdf/0383(2011).pdf.
Fuss, M. A. (1977). The demand for energy in Canadian manufacturing: An example of the
estimation of production structures with many inputs. Journal of Econometrics, 5, 89116.
Mahadevan, R., & Asufu-Adjaye, J. (2007). Energy consumption, economic growth and prices: A
reassessment using panel VECM for developed and developing countries. Energy Policy
Volume, 35(4), 24812490.
Marquez, A. C., & Fuinhas, J. A. (2011). Drivers promoting renewable energy: A dynamic panel
approach. Renewable and Sustainable Energy Reviews, 15(3), 16011608.
McMillin, W. D., & Smyth, D. J. (1994). A multivariate time series analysis of the United States
aggregate production function. Empirical Economics, 19, 659673.
Ozturk, I. (2010). A literature survey on energy-growth nexus. Energy Policy, 38(1), 340349.
Pindyck, R. S., & Rotemberg, J. J. (1983). Dynamic factor demands and the effects of energy
price shocks. American Economic Review, 73, 10661079.
Sadorsky, P. (2009). Renewable energy consumption, CO2 emissions and oil prices in the G7
countries. Energy Economics, 31(3), 456462.
Stern, N. (2006). Stern Review: The economics of climate change. Cambridge, UK: Cambridge
University Press. Available from: http://www.hm-treasury.gov.uk/stern_review_report.htm.
Chapter 6
Abstract Contemporary energy policy issues are dominated, directly and indirectly, by major concerns at both local and global levels of environmental degradation arising from combustion of fossil fuels. The advent of carbon pricing
(either through an emissions trading scheme or a carbon tax) represents an attempt
to impose a cost on consumers that will limit such degradation (i.e. the deleterious
impacts of climate change) to scientifically-determined acceptable levels. The
resulting higher cost of fossil fuel combustion for power generation should induce
a reduction in the demand for power (the demand effect) whilst simultaneously
stimulating investment in competitively-priced low carbon power generation
technologies (the supply effect). At least in theory, the trading of emission
permits can be shown to be a least-cost economic instrument for meeting a
specified level of reduction of carbon dioxide. However, a carbon tax possesses the
same property. In this chapter the relative merits of these two instruments will be
assessed, paying particular attention to factors that could, in practice, lead to
significant levels of inefficiency for one instrument relative to the other.
91
92
A. D. Owen
6.1 Introduction
The twentieth century witnessed historically unprecedented rates of growth in
energy systems, supported by the widespread availability of fossil fuel resources.
During the second half of the century, however, concerns associated with high
levels of fossil fuel dependence began to surface. Two issues were of particular
significance: the impact of modern energy systems on the environment and
security issues associated with fuel supply lines.
Environmental concerns had been evident in more localized areas for many
hundreds of years. Ancient Rome burned wood and Emperor Neros tutor, Seneca,
complained of the bad effect that smoke had on his health and of smoke damage to
temples, whilst anecdotal evidence indicates that air pollution had been a concern
in England as early as 1352 when a ban was introduced on coal burning in London.
Today, local pollution from energy systems remains a threat to the health of the
living environment. However, in the latter decades of the twentieth century, pollution resulting from combustion of fossil fuels became a global concern, with the
publication of credible scientific evidence that the planets climate was changing
as the result of a buildup of so-called greenhouse gases in the atmosphere.
The proposal to impose taxes on pollution, whilst more recent, is also far from
new, having been advanced at the turn of the last century by the famous British
economist Arthur Cecil Pigou as a means of reducing Londons famous fogs (or
smogs). Pigou (1920) observed that pollution imposed uncovered costs on third
parties that were not included in ordinary market transactions. His proposal was to
tax pollution by means of a so-called externality tax1 in order to internalize within
ordinary market transactions the damages caused by pollution. At the time Pigous
proposal was regarded as an academic curiosity, but several generations later it
was rejuvenated as the core of the polluter pays principle.
Historically, regulatory instruments have been the basic mechanism for
enacting environmental policy throughout the industrialized world. Environmental
quality has been seen as a public good that the state must secure by preventing
private agents from damaging it. Direct regulation involves the imposition of
standards (or even bans) regarding emissions and discharges, product or process
characteristics, etc., through licensing and monitoring. Legislation usually forms
the basis for this form of control, and compliance is generally mandatory with
sanctions for non-compliance.
More recently, the use of market-based economic instruments has emerged as a
more flexible alternative to the traditional command-and-control regulatory
approach to controlling emission of pollutants in market-based economies. Such
instruments can generally be divided (generically) into taxes and emission permits,
although both categories contain a large array of distinctly different forms of
instruments depending on the ultimate intention of their application.
93
94
Fig. 6.1 The optimal level
of pollution for an economy
A. D. Owen
$
MAC
MD
A
*
D
C
0
MA
B
M*
MP
Emissions
However, a hybrid scheme is possible whereby if the price of permits reaches a predetermined
price ceiling the latter becomes the fixed permit price (effectively, therefore, a carbon tax). The
initial years of the Australian Clean Energy Future Plan has this arrangement in place to avoid
price spikes in the early years of the permit trading regime and, as a consequence, reduce price
volatility risks for potential investors in power generation assets.
95
96
A. D. Owen
MAC1, MAC2, and MAC3 are marginal abatement cost curves for three different plants producing the same product, with different technologies reflected by
the different curves. They slope upwards from left to right indicating that the cost
of the marginal unit of pollution abatement increases as the total required reduction increases. Clearly, Plant 3 has the lowest abatement costs and Plant 1 the
highest. For simplicity assume that:
S1 S2 S3 3S2 and S1 S2 S2 S3
One way of achieving a given standard of pollution abatement, say 3S2, is to
instruct each plant to abate pollution by an amount 0S2. Under such circumstances,
Plant 1 would go to point A, Plant 2 to point B, and Plant 3 to point C, thus
achieving a total reduction of 3S2.
However, clearly their costs of pollution are very different. By imposing a tax
equal to t* the same total pollution abatement result can be achieved, but at lower
overall cost of compliance. Plant 1 now goes to point X, Plant 2 to point B, and
Plant 3 to point Y. The overall desired level of pollution abatement has been
achieved, with plants having the cheapest abatement options reducing more than
those with higher cost options. Thus, to the right of S1, it is cheaper for Plant 1 to
pay the tax rather than abate pollution, whereas for Plants 2 and 3 abatement
remains cheaper (until points B and Y are passed, respectively).
Now both standards and tax have achieved the same overall standard of 3S2.
However, the total compliance cost differs.
Under standards the total compliance cost is 0AS2 ? 0BS2 ? 0CS2; whilst
under taxation the total compliance cost is 0XS1 ? 0BS2 ? 0YS3. Subtracting the
latter from the former gives S1XAS2-S2CYS3
97
which is always positive. Thus standards setting incurs greater total abatement
costs than taxation to achieve the same standard.
In practice, a problem arises because the tax is based on plant emissions rather
than the cost of the impact of the emissions on the affected population. For
example, emissions from a coal-fired power plant located by the coast may have
significantly less impact on local communities than one situated inland if the
prevailing winds blow the pollutants from the former offshore (although other sorts
of problems may arise, such as acid rain falling on neighboring countries). Thus
some form of differential tax based on location (as a surrogate for impact) must be
levied. While this will complicate the issue, the combination of standards and
taxes still retains its advantage of not requiring calculation of damage arising from
the emissions. However, in the context of GHG emissions, the problem is dealing
with a uniform mixing pollutant. Thus whilst damages will vary across countries,
that damage is independent of the source of the pollutant. It follows that a uniform
tax is appropriate for emissions of CO2 (and other GHGs).
In addition to its least-cost property, taxation has the advantage of flexibility in
dealing with a range of environmental externalities (as opposed to direct regulation). Further, once in place, tax rates can readily be varied to reflect changing
standards, whereas changes in regulatory requirements would frequently require
new legislation to be enacted. However, in practice, the use of taxation as an
instrument for environmental regulation has been very limited.
The question of setting the precise level of tax to achieve the desired environmental target or objective depends on a reasonably precise knowledge of the
magnitude of the relevant energy demand elasticities. Estimates of elasticities of
demand for energy (and individual fuels) in individual countries vary widely, and
are generally based on econometric models estimated using data which do not
undergo changes of the magnitude in energy prices suggested in the context of the
imposition of a carbon tax. Since a carbon tax will vary by carbon content of fuel,
it is also necessary to know inter-fuel substitution elasticities as well as the
standard income and price elasticities. Whether or not these objections represent a
serious flaw in the entire concept of modeling the economic impact of carbon taxes
is an important question. Even if such an exercise is thought to produce estimates
that are believed to be reasonably reliable, it is clear that the sheer magnitude of
the revenue that would be generated by such a tax requires a general (rather than
partial) equilibrium approach in the econometric modeling component of any
study of the ultimate impact of the tax and use of the revenue raised by it.
The fact that a carbon tax cannot yield a precise level of CO2 emission
reductions for a given rate of tax should not be viewed as an argument against its
implementation, since it is a stock and not a flow of pollutant. In other words,
damages are related to the atmospheric concentrations, not to the flow of emissions. Therefore if the carbon tax does not achieve the expected reduction in
emissions in one period, there will be time to adjust the tax in subsequent periods.
However, therein lays another problem!
Setting an appropriate level of carbon tax (either nationally or globally)
depends critically on accurate estimation of the level of world economic activity
98
A. D. Owen
and relative real incomes. Changes in energy prices have had (in the past) a
complex and asymmetric influence on the level of world economic activity. Since
the precise level of tax to meet a given emissions target cannot be achieved, it is
imperative that some form of partial adjustment mechanism be present to ensure
that changes in the rate of the carbon tax do not generate violent fluctuations in
GDP with a resulting loss of business confidence.
Since almost all economic activities involve, directly or indirectly, use of energy
derived from fossil fuels, reducing emissions of CO2 can be achieved in a relatively
simple manner by taxing the carbon content of the fuels. Since a carbon tax will vary by
the carbon content of each fossil fuel, inter-fuel substitution between both fossil and
non-fossil fuels will occur as will substitution of other (now relatively cheaper) factors
of production (e.g. labor and capital) for fossil fuels. This process will be enhanced if
investment in energy saving technologies, or non-carbon emitting energy sources, is
encouraged through the fiscal regime. In addition, since carbon taxes will always be
present as long as carbon-based fuels are used, there is a continuing financial incentive
to develop alternative energy technologies or carbon disposal technologies.
99
then command a market price. Sources with low (i.e. below the market price)
abatement costs will have an incentive to sell permits and abate their emissions.
Conversely, sources with unit abatement costs above the market price will have an
incentive to purchase permits in the market. Assuming sources minimize their total
production costs, and the market for permits is competitive, it can be shown that the
overall cost of achieving the emissions target will be minimized.
Notwithstanding the drawbacks and problems associated with regulation and taxation as instruments for pollution control, they are both founded on well-worked and
familiar structures. Regulation and taxation are familiar tools to both industry and the
general public. However, with tradable permits a new market is being established,
largely on the basis of a theoretical model, with little practical experience to date.
Major questions that surround the implementation of a tradable permits scheme are:
How should the permits be allocated?
Should the initial allocation be auctioned, leased, or given away?
Who should be permitted to trade in permits? For example, should an environmental group be permitted to purchase permits and effectively withdraw
them from the marketplace?
To what extent must monitoring be undertaken to ensure that emissions are
correctly recorded (and what should be the penalties for non-compliance)?
What will be the cost of the administrative infrastructure required to facilitate
trading and associated activities (including prevention of collusive activities in
the permits market)?
To what extent must details of permit transfers (i.e. quantity, price, ownership,
etc.) be made publicly available?
Should there be a grace period at the end of each year to allow for permit
transfers in order to achieve compliance (and how long should it be)?
100
A. D. Owen
Fig. 6.4 Hypothetical example of a cap-and-trade scheme showing gains from trade
101
as an option to permits, then they will effectively place a ceiling on the price at
which permits can be traded. The scheme is designed to guard against extreme
misjudgment of the optimal level of emissions set as part of the process of
determining the aggregate value of the issued permits.
The mechanics of the scheme are as follows: the agency issues a predetermined
number of tradable emission permits for which a market emerges and a permit price is
determined (which is assumed to be equivalent to a cost of P per unit of emissions). At
the same time, the agency charges an emissions fee equivalent to (say) F per unit of
emissions. The latter can be paid irrespective of whether or not the polluter has tradable
permits. In addition, the agency is willing to provide a subsidy of S per unit of emissions for any unused permits. It follows, that in equilibrium, it must be that: S B P B F.
This must hold because if P were greater than F, no permits would be purchased, rather it would be cheaper to pay the emissions tax, so P would be forced
to fall. Conversely, if S exceeded P, it would be prudent to purchase as many
permits as were available, yielding a profit of (SP) per unit; although it is very
unlikely that anyone would be willing to sell a permit at that price!
Note that if S = 0 and F = ?, both subsidy and emissions taxes are effectively
eliminated and the system reverts to a pure permits scheme. It also reduces to a
pure taxation system if S = F = T, were T is the magnitude of the tax, which is
the level to which the price of a permit will be automatically driven. Thus, if either
the pure permit or pure taxation scheme is optimal, the maximization calculation will automatically ensure the elimination of the mixed system.
The operation of the mixed system provides an effective insurance for dealing
with situations where, presumably unknown to the regulator, permits or taxes fair
badly. In circumstances where permits would have performed badly, it would pay the
polluter under the mixed scheme to act in a way that transformed it into a taxation
regime. For example, if the regulator issues permits (to a value of q*) based on an
estimated cost of pollution reduction which is too high, then q* could be far below the
optimum level of pollution reduction. Conversely, if the regulator underestimates the
cost of pollution, q* will be correspondingly excessive. Now if the actual marginal
cleanup cost is higher than F, it will pay polluters operating under the mixed
arrangement to emit more than the permits allow and to pay the tax (F) on all
emissions that exceed those covered by their permit holdings. If, however, cleanup
costs turn out to be lower than S, it will pay polluters to continue to reduce their
emissions and hold their excess permits unused in return for the subsidy payment.
102
A. D. Owen
permits (and vice versa) can be derived.3 Relaxing the assumption that the MD
curve is known does not change the relative merits of taxes and permits since they
are coincident in their effects on abatement.
The attractive property of tradable permits is that, in the absence of cheating, a
specified maximum level of pollution can be guaranteed irrespective of the cost
associated with reaching this level. Similarly, pollution charges or taxes guarantee
that the marginal cost of emissions control will be equivalent to the level of the tax
imposed, irrespective of the size of the resulting quantity of emissions. Thus, if a
government agency adopted a system of marketable permits it can be assured of
meeting the stipulated level of emissions, but the associated cost could be surprisingly high. In contrast, if the agency employs an emissions tax it can be certain
about the resulting marginal control cost (irrespective of the true cost function),
but it may be significantly adrift from its desired level of emissions.
Essentially the problem stems from the agencys ignorance of its marginal cost
curve. If it overestimates, then emissions reduction will generally be inadequate
under a system of permits and excessive under taxes if both are initially set at their
perceived optimum levels. The reverse will be true if the marginal cost curve is
underestimated. This is shown diagrammatically in Fig. 6.5.
It is assumed that the marginal damages (MD) of emissions curve is known with
certainty, but that the marginal abatement cost (MAC) curve is unknown. The
regulator anticipates that MAC2 is the true curve, when in fact it is MAC1; that is,
the regulator has overestimated the MAC curve. The regulator will therefore select
a tax of 0t1, or will issue permits to the value of 0S1 emissions. But the optimal
reduction in emissions is 0S* and hence too many permits have been issued,
equivalent to S*S1 emissions reduction. The corresponding reduction brought
about by the tax will be 0S2, thus leading to a shortfall of emissions reduction
measured by S2S*. Thus the efficiency loss in using taxes under this form of
uncertainty is equivalent to the area shown in light shading. The corresponding
loss in efficiency with permits is equivalent to the area shown in dark shading.
Note that the slope of the MD curve is less than the slope of the two MAC curves.
Figure 6.6 illustrates the same situation of an error in estimating the true MAC,
but in this example the slope of the MD curve is greater than the slope of the two
MAC curves. Now the efficiency losses have been reversed, with taxes being
preferable to permits in this case.
In general, it can be shown that when the MD and MAC curves are linear,
tradable permits and taxes will produce the same absolute distortion when the
regulator miscalculates the MAC curve if the absolute values of the slopes of the
two curves are equal. If the absolute value of the slope of the MAC curve is less
than that of the MD curve, permits will lead to a smaller distortion, and vice versa.
This explanation is based upon results derived by Weitzman (1974). A number of studies have
also addressed the relative merits of taxes as opposed to emissions trading schemes in the
presence of uncertainty; for example, Nordhaus (1994), Hoel and Karp (2002), and Newell and
Pizer (2003).
103
However, if one also allows for uncertainty over its slope then the preference of taxes over
permits is expressed in terms of unknown parameters: a result that has little practical use (see
Quirion (2005) for further details).
104
A. D. Owen
105
Dynamic efficiency
Encourages adoption of low-carbon
Encourages adoption of low-carbon
technologies, but requires adjustment of tax
technologies, but allocation criteria for new
rates as economy expands or contracts.
entrants may involve high set-up costs,
particularly if permits are grandfathered.
Revenue and distributional issues
If tradable permits are auctioned, then taxes and permits are equivalent in terms of revenue
raising potential, and hence there is no difference in their distributional impacts. If a
proportion (or all) of the permits are allocated to emitters free of charge, then revenue will fall
correspondingly. Emitters would receive windfall gains.
International harmonization
Difficult to impose globally and hidden
Quantitative caps permit transparency for
subsidies could offset its impact.
international harmonization, ideally
delivering a single carbon price.
Union emissions trading scheme has been operating for six years, with a third
phase about to be finalized for the period 20132020. Although some lessons from
European Union experience now appear reasonably clear, others are still veiled in
uncertainty and confusion. Of particular concern is the potential for volatile prices
for emissions permits and their dampening impact on long-term investment
decisions, a problem exacerbated by the Global Financial Crisis and its aftermath.
Taxation, however, also has its practical drawbacks, as noted in Table 6.1. In
particular, in the context of a global scheme for reducing CO2 emissions, carbon taxes
would lack the transparency of quantitative emissions targets and avoidance in the
form of hidden subsidies could threaten the integrity of any such global agreement.5
In fact the subsidies need not be hidden since a social welfare argument may be made for
protecting certain industries or sections of society from the impact of carbon pricing. The IEA
(2010) estimated that worldwide fossil fuel consumption subsidies amounted to US$557 billion in
2008, so this is not a trivial issue.
106
A. D. Owen
As mentioned earlier, the ultimate goal for any carbon pricing regime is to
make carbon an internationally traded commodity. For this to happen, all of the
worlds major carbon emitting nations must have mandatory domestic carbon
emission ceilings in place. Without such a global accord, carbon emission
reduction targets based upon domestic production emissions ignores the emissions
embodied in traded goods imported from non-participating countries. Thus, whilst
production of carbon in the European Union has fallen significantly since the
introduction of its emissions trading scheme, there is evidence that supports a
conclusion that no corresponding fall has occurred in its embodied
consumption!6
References
Helm, D., & Hepburn, C. (Eds.). (2009). The economics and politics of climate change. Oxford:
Oxford University Press.
Hoel, M., & Karp, L. (2002). Taxes versus quotas for a stock pollutant. Resource and Energy
Economics, 24, 367384.
International Energy Agency (IEA). (2010). World energy outlook. Paris: IEA/OECD.
Newell, R. G., & Pizer, A. W. (2003). Regulating stock externalities under uncertainty. Journal of
Environmental Economics and Management, 45, 416432.
Nordhaus, W. D. (1994). Managing the global commons. Cambridge: MIT Press.
Pearce, D. W., & Turner, R. K. (1990). Economics of natural resources and the environment.
London: Harvester Wheatsheaf.
Pigou, A. C. (1920). The economics of welfare. London: Macmillan.
Quirion, P. (2005). Prices vs. quantities in a second-best setting. Environmental and Resource
Economics, 29, 337360.
Weitzman, M. (1974). Prices vs. quantities. Review of Economic Studies, 41, 477491.
This issue is addressed, at least partially, in the context of emissions embodied in exports from
China in Chapter 8 of Helm and Hepburn (2009).
Chapter 7
Abstract This chapter examines the impact of price changes of European Union
Emission Allowances (EUAs) on stock returns of a sample of firms operating in
the European steel and combustion industries. After the introduction of an
EU-wide CO2 emissions trading system in 2005, the first phase of the emission
plans was completed in 2007 and the process of the second phase is still evolving.
The empirical analysis of this paper covers both the first and second phases to deal
with capturing time variances for the importance of the trading scheme. It is found
that EUA price changes are positively correlated with stock returns of firms
operating only in combustion industries over the first phase (20052007).
However, there is evidence of a significant impact of EUA price changes on stock
returns of firms from both combustion and steel industries during the sample
period in the second phase (20082010). Firm size, age and leverage are also
examined to explain the size of emission trading exposures across firms. The
finding is that neither of these variables appears to be a significant determinant.
This evidence indicates that EUA exposure is independent of firms characteristics.
Keywords EU emission allowance
intensive industries
107
108
7.1 Introduction
The European Union (EU) considers climate change to be one of the greatest
environmental, social and economic threats facing the planet. Therefore, the EU
plays a leading role in the negotiations for international action against climate
change, in particular for the Kyoto Protocol (Kemfert et al. 2006). Whereas
participant countries are committed to reducing the emissions of greenhouse gases
by about 5 % during the period 20082012, compared to that in 1990 as the first
target, the EU is committed to 8 % reduction (Lund 2007; Oberndorfer 2009;
Gronwald et al. 2011). The EU Emission Trading Scheme (ETS) has evolved from
the 2000 Green Paper1 to what is now regularly characterized as the flagship of the
European Climate Change Program (Ellerman & Joskow 2008). The EU ETS
covers 27 countries with about 12,000 installations, including energy-production
facilities such as power utilities and oil refineries as well as energy-intensive
industries such as iron, steel, paper and minerals (Hassan & Molho 2009).
The ETS allocates carbon emission allowances (in tons of CO2) to firms in a
specific industry. The system allows firms to buy or sell pre-allocated allowances
on climate exchanges. Over a five-year trading period (20082012), the estimation
is that the value of permits accruing to the top ten companies will rise to 3.2
billion. This value represents the market value of the total number of permits, and
exceeds by a third the total EU budget for the environment and is more than double
the funding announced in the European Energy Program for Recovery (EEPR) for
renewable and clean technologies over the same period (Pearson 2010). In 2007,
allowance trading exceeded 2.1 billion tons of CO2 worth 35 billion (IETA
2008). Even if the permits are not directly sold on for profit, the value will still
remain on the companys balance sheets. ArcelorMittal, the worlds biggest steel
maker, continues to dominate this ranking by far, with a surplus of 31 million EU
allowances in 2010. At the current EUA price of 17, this surplus represents a
selling value of more than 500 million. These developments indicate that as a
response to the rise in prices of EUAs in future years the values of assets will
possibly increase (Pearson 2010). However, whether the price changes in emission
trading increase firms market values is an empirical question.
This chapter investigates the reflection of the economic importance of the EU
and aims to find a sufficient answer for the following question: What is the
influence of trend of emissions trading prices on firms operating in the European
Combustion and Steel Industries with regard to their stock returns? The aim is to
show significant exposure of trading pricing on the firm value. As an extension, for
The 2000 Green Paper was adopted by the European Commission on 29 November 2000 to
determine policies for a long-term energy strategy. The main issue addressed in the Green Paper
is that the EU must rebalance its supply policy by clear action in favor of a demand policy. With
regard to demand, the Green Paper is calling for a real change in consumer behavior being more
respectful of the environment. With regard to supply, the development of new and renewable
energies is the key to change with regard to global warming.
109
the further effects of emissions trading, the relationships between price changes of
ETS and the production capability and competitiveness of sample firms are also
examined. Furthermore, size is used along with leverage ratio for possible firms
characteristics to explain the level of pricing exposure.
The EU ETS was implemented in 2005 with an introduction phase until the end
of 2007, which is defined as Phase 1 in this chapter. Phase 2 is planned for the
period 20082012. The empirical analysis is able to cover this phase only until 31
December 2010 because of data availability. This chapter extends the studies by
Oberndorfer (2009) and Veith et al. (2009), where the authors search for the
influence of the EU ETS on the energy sector (in this chapter also referred to as the
combustion sector) for Phase 1, by arguing that there would be differences intraphases. Veith et al. (2009) find a positive relationship between combustion firms
share price and EUA prices for the first phase. Oberndorfer (2009) supports their
view and adds additional insights to be considered for the second phase of the EU
ETS. In addition, this chapter carries such analysis from one industry, the
combustion industry, to another one, the steel industry. Both Oberndorfer (2009)
and Veith et al. (2009) only investigate the combustion sector, but the results may
not necessarily apply to all other sectors involved in emissions trading. The steel
industry is one of the most energy-intensive industries with an annual energy
consumption of about 5 % of the worlds total energy consumption. In addition,
the steel industry accounts for 34 % of total world greenhouse gas emissions
(Xu and Cang 2010). For industries within the ETS, the EU cap-and-trade scheme
imposes extra costs and may, as a worst case scenario, even affect the competitiveness of energy-intensive industries in Europe (Lund 2007). This makes the
steel industry extremely interesting because this sector should notice the effects
from the trading more than any other. By investigating the steel industry in both
the first and second phases of the EU ETS, this study aims to show if there are
differences compared to the electricity producing firms.
This chapter is structured as follows: Sect. 7.2 provides background information
on the ETS and the implementation periods with the changes they bring.
Section 7.3 outlines the hypotheses and the reasoning behind them. Section 7.4
highlights the methodological approach. Section 7.5 provides the results and
interpretations, and finally Sect. 7.6 concludes.
110
emit CO2 has now become a tradable commodity and is a factor of production that
is subject to price changes (Gronwald et al. 2011).
The basic idea of emissions trading is to limit the amount of emissions by
creating rights to emissions and to make these rights, called allowances, tradable.
The scarcity of emission allowances gives them a market value and those emitters
whose avoidance costs are lower than the market value of allowances will reduce
their emissions and buy fewer certificates or sell excess emissions rights, and vice
versa for other emitters (Kemfert et al. 2006; Mansanet Bataller et al. 2007;
Gronwald et al. 2011). Thus, a lack of allowances requires a company to either buy
a sufficient amount of EUAs or to invest in some plant-specific process
improvements (Gronwald et al. 2011).
111
wide carbon price was established, businesses began incorporating this price into
their decision-making, and the market infrastructure for a multi-national trading
program is in place (Ellerman and Joskow 2008).
For example, the percentages of cap auctions in the first and second phases are as follows for
the following member states: Denmark 5 and 0; Hungary 2.5 and 2.3; Lithuania 1.5 and 2.9;
Ireland 0.75 and 0.5; Austria 0 and 1.2; Belgium 0 and 0.3; Germany 0 and 8.8; Netherlands 0 and
4; UK 0 and 7 (Ellerman & Joskow 2008).
112
113
114
Phase 1
Phase 2
Firms
Phase 1
Phase 2
-7.03
-8.20
-3.12
-25.83
-43.18
-27.19
-13.75
10.39
4.98
-12.04
-29.79
12.05
0.81
20.53
-30.42
0.11
-12.44
0.44
-13.17
-33.54
-6.85
-25.70
-19.04
-48.51
-7.71
-19.63
-4.70
-13.08
2.72
6.75
-22.65
8.39
-9.02
-30.11
-26.58
-18.77
-10.13
-5.41
9.69
15.20
-55.84
-10.63
ArcelorMittal
ThyssenKrupp
Voestalpine
Salzgitter
Rautaruukki
SSAB Svenskt Stal
Vallourec
Outokumpu
Viohalco
Hoganas
Acerinox
Schmolz-Bickenbach
Sandvik
Tenaris
Sidenor
30.41
13.66
2.61
2.75
6.35
84.55
11.59
42.52
48.06
36.56
9.52
3.15
11.98
10.82
48.06
56.45
41.39
-2.16
36.02
18.32
120.61
104.57
80.61
72.40
45.89
63.46
40.22
-1.83
-71.22
72.40
Average
22.47
43.19
Source Carbonmarketdata.com
This table reports the difference between allocated and verified EUAs in percentages for both Phases
1 and 2. Negative values indicate the shortages, and positive values are for excess of EUAs
115
Hypothesis 2 Stock returns for the European steel and combustion companies
are positively correlated to price changes of emissions trading in Phase 2.
Phase 2 covers the period of the recent economic and financial crisis where
asset and commodity prices are dramatically affected. Therefore, sub-periods are
also created during Phase 2 to be able to capture the role of this unfortunate event.
The first period, Period 21, runs from 27 February 2008 to 31 July 2009, the
second, Period 22, from 01 August 2009 to 31 December 2010. Gronwald et al.
(2011) note that the relationship between EUA returns and other financial variables
is particularly strong during the financial crisis. They find a stronger dependence
between EUA futures returns and most of the considered variables during the
global financial crisis. This confirms general results on asset returns from financial
markets exhibiting higher dependence during periods of extreme economic or
market downturn.
Hypothesis 2a Stock returns for the European steel and combustion companies
are positively correlated to price changes of emissions trading in Period 21 of
Phase 2.
Initial over-allocation to these industries has since been compounded by the
negative effects of global recession on production. During this period, many
companies find themselves in a position, in which they have far more permits to
pollute than they require (Pearson 2010). This leads to expectation that the second
period of Phase 2 will play no significant role in price changes in emission trading.
Hypothesis 2b There is no significant relationship between stock returns for the
European steel and combustion companies and price changes of emissions trading
in Period 22.
116
mean less exposure and more certainty for the corporations. CER stands for
Certified Emission Reduction. It allows emission reduction projects conducted in
developing countries to generate carbon credits which can be used in the EU ETS.
Nance et al. (1993) argue that firm size, a proxy for economies of scale in hedging
costs, is related to hedging incentives. Larger firms that have access to risk
management expertise, or that have economies of scale in hedging costs, are more
likely to hedge than smaller firms (Haushalter 2000). Thus, hedging seems to be
driven by economies of scale, reflecting the high fixed costs of establishing risk
management programs (Jin and Jorion 2006). As a result, bigger corporations
should be less exposed to exchange-rate risk (He and Ng 1998). However, there
are circumstances where smaller firms have more incentive to hedge than larger
firms; for instance, smaller firms will hedge more, because they face greater
bankruptcy costs (Haushalter 2000). Thus, the effect of firm size on exchange-rate
exposure is ambiguous and shall be empirically determined (He and Ng 1998).
Thus, the following hypothesis is formulated:
Hypothesis 4 Steel and combustion companies with a relative large market
value face lower emission price exposure.
Smith and Stulz (1985) argue that hedging can reduce the probability that a firm
will go bankrupt and thereby reduce the expected costs of financial distress. They
employ a firms long-term debt ratio to measure its probability of financial distress.
Ceteris paribus, firms with a higher level of debt tend to face larger expected costs
of financial distress and hence have a greater desire to engage in hedging activities.
Therefore, the role of debt level on emission price exposure is also examined:
Hypothesis 5 More leveraged steel and combustion corporations are more
exposed to emissions price fluctuations than less leveraged steel firms.
In the analysis testing Hypotheses 4 and 5, age of the firm is controlled for its
role on the choice of hedging. On one hand, older firms are more likely to hedge
than younger firms because managers in older firms have more experience and
expertise in risk management. Moreover, larger firms have economies of scale in
hedging costs. On the contrary, younger firms are also likely to have more
incentive to hedge than older firms. Younger firms bear higher bankruptcy costs,
and therefore, may tend to hedge more.
117
Bartram 2005). As such, EUA price exposure is the elasticity of stock returns with
respect to unanticipated changes of the price of EUAs.
Building on earlier work of Jorion (1990), which is widely used in many later
studies such as Bartram (2005) and Oberndorfer (2009), a two level regression
model is used to determine the extent of commodity price exposure by steel and
combustion firms:
Step 1 is the standard methodology to capture the impact of price changes of
emission trading on stock returns as shown by the following regression model;
Rit ai b1 Rmt b2 RSct e
7:1
where:
Rit = daily return on the common stock of firm i in period t,
Rmt = daily return on the market portfolio in period t,
RSct = daily return on spot contracts emission allowances in period t,
e = the error term.
The coefficients b1 and b2 capture the market risk as well as the exposure
towards the trade in pollution rights. The market portfolio controls for the impact
of unexpected macroeconomic changes. In addition to these basic factors, changes
in the steel price for the steel industry and the oil and gas prices for the combustion
firms are added into Eq. 7.1.
Rit ai b1 Rmt b2 RSct b3 Steel e
7:2
7:3
where:
RSteel = daily price change on steel,
ROil = daily price change on oil,
RGas = daily price change on gas.
Step 2 captures the effects of firms characteristics with respect to size and
leverage on the coefficient of emissions allowances, namely (b2), which is
obtained from the coefficients of regressions in step one for each firm. Using
ordinary least squares cross-sectional regression analysis the determinants of the
estimated exposures are related by:
b1 ai a1 Sizei a2 Leveragei e
7:4
7.4.2 Data
Powernext (now Bluenext) accounted for almost 79 % of the spot market transactions in the EU ETS (Daskalakis et al. 2009). Data were collected for the prices
of EUAs from Bluenext as it is the most used and complete set of data for the spot
118
EUA market. Moreover, Mansanet Bataller et al. (2007) indicate that EUA prices
have developed very similarly in all marketplaces, so that the choice of marketplace should not be crucial for the analysis. Although future or forward prices are
less affected by very short-run demand and supply fluctuations and therefore less
noisy in comparison to spot prices (Oberndorfer 2009), in line with Oberndorfer,
the settlement price is used instead of an EUA future from the EEX as there is little
trade in futures in comparison to the spot market.
EUA spot pricing data from 27 June 2005 until 31 December 2010 was
collected. As explained in the section on the hypotheses, several sub-periods in
Phases 1 and 2 of the ETS are used: Phase 1 of the ETS is the first period as a
whole from 27 June 2005 to 31 August 2007. This Phase 1 period is split-up
because of the April 2006 crash in EUA prices into a pre, during and post the crisis
period. It commences in the last days of April and thus Period 11 runs from 27
June 2005 to 25 April 2006; the Period 12 runs from 26 April 2006 until 10-May
2006, and the Period 13 from 11 May 2006 to 31 August 2007. For Phase 2, data
from 27 February 2008 to 31 December 2010 is used. The starting date is the first
day the new EUAs. The period is split into two sub-periods to see the impact of the
global financial crisis. Period 2-1 starts on 27 February 2008 and ends on 31 July
2009, and Period 2-2 runs from 1 August 2009 to 31 December 2010. Panel A of
Table 7.2 provides a general view of all these sample periods and total number of
observations used in the analysis.
Panel B of Table 7.2 reports summary statistics for sample firms characteristics. According to mean and median values of size, leverage and age, the sample
firms in the combustion industry use more debt, and are younger than those in the
steel industry, while firms in both industries are similar in size.
The Community Independent Transaction Log (CITL) is the definitive EU ETS
resource created by the European Commission and contains all records of issuance, transfer, cancellation, retirement and banking of allowances that take place
in the registry. Additionally, data for companies in the carbon trading markets are
gathered from the carbon market data website, which is the official data provider
for this market. They provide free access to the official EU ETS database, a tool
designed for all carbon market players.
To create a sample of firms, first a search of the list of companies from
DataStream is made. The largest publically traded companies in the combustion
and steel industries in Europe are collected. The names of these firms are then
cross-referenced with emissions data. Daily stock prices are collected from Yahoo
Finance, Google Finance, and Datastream. For market returns, the local index from
where the company is headquartered is used. Steel prices are calculated by using
the Dow Jones steel index from Yahoo Finance. For oil prices, the London Brent
Crude Oil Index is used. Generally, UK and continental gas prices are closely
related due to arbitrage possibilities (Oberndorfer 2009). Natural gas prices are
collected from the ICE Natural Gas prices as one month forward contracts, in line
with Oberndorfer (2009). Monthly steel production data per country and region is
collected from the World Steel Association website.
119
Median
Median
25-4-2006
10-5-2006
31-8-2007
31-8-2007
31-7-2009
31-12-2010
31-12-2010
6.77
0.52
50
Firms size is the log of total assets. Financial leverage is the ratio of the end-of-year book value
of long-term debt to the end-of-year market value of the firm. Both variables are collected from
Datastream. Firms ages are collected from web sources
Firm size is measured by taking the log of total assets at the beginning of the
periods of Phases 1 and 2. Financial leverage is defined as the ratio of the end-of-year
book value of long-term debt to the end-of-year market value of the firm. Since book
value of equity is influenced by accounting conventions applied in different countries, market value of equity in is used calculating leverage. Moreover, a market value
leverage measure is commonly used in the corporate finance literature. Variables to
calculate those characteristics are collected from Datastream.
RSct
Rmt
RSteel
RSct
Rmt
RSteel
RSct
Rmt
RSteel
(0.01)c
(0.00)c
(0.01)c
(0.03)b
(0.00)c
(0.00)c
(0.35)
(0.97)
(0.13)
(0.00)c
(0.62)
(0.36)
Variables
(0.07)a
(0.00)c
(0.54)
(0.45)
p values
Steel industry
-0.002
0.701
0.271
0.016
0.766
0.32
0.067
0.913
1.128
0.004
0.779
0.216
(0.93)
(0.00)c
(0.00)c
(0.38)
(0.04)b
(0.01)c
(0.00)c
(0.00)c
(0.00)c
(0.77)
(0.00)c
(0.00)c
p values
This table presents the estimated coefficients and p values of coefficients in the following regressions: (1) Combustion industry, Rit = ai ? b1 Rmt ? b2
RSct ? b3 RSteel ? e, and (2) Steel industry, Rit = ai ? b1 Rmt ? b2 RSct ? b3 ROil ? b4 RGas ? e. Rit is daily return on the common stock of firm i in
period t, Rmt is daily return on the market portfolio in period t, RSct is daily return on spot contracts emission allowances in period t. Variables Roil, RGas,
and RSteel are the changes in prices of commodities of oil, gas and steel. a , b , c denote statistical significance at 1, 5, and 10 % levels, respectively
Panel
RSct
Rmt
Roil
RGas
Panel
RSct
Rmt
Roil
RGas
Panel
RSct
Rmt
Roil
RGas
Panel
RSct
Rmt
Roil
RGas
Variables
120
J. Bruggeman and H. Gonenc
121
The first period of Phase 1 (Panel B) is associated with a positive and significant
relationship between both industries stock returns and EUA price changes. This is
in line with expectations and can be explained by the over-allocation in the steel
industry and the pass through possibilities of the combustion industry due to their
market power. Following the previous research by Oberndorfer (2009), a positive
correlation is also expected for the combustion industry in this period. He finds
evidence for a particularly strong impact of EUA price changes on electricity stock
returns during the period of market shock in April/May 2006. His findings seem to
hold for the combustion and steel industries.
The results regarding Period 12 (Panel C) indicates no significant effect of
price changes of EUA on stock returns for firms in the steel industry, but this effect
is significant for firms in the combustion industry. Thus, the steel industry might
not be able to sell off any of its surplus EUAs rendering them worthless, especially
because they cannot be banked or used in the following phase causing the insignificance for this period.
The results from Period 13 (Panel D) are consistent with the hypothesis,
indicating that there should be no significant relationship between stock returns
and price changes of EUAs.
122
Steel industry
p values
Variables
p values
(0.01)c
(0.00)c
(0.00)c
(0.88)
RSct
Rmt
RSteel
0.049
0.976
0.109
(0.00)c
(0.00)c
(0.00)c
(0.04)b
(0.00)c
(0.00)c
(0.58)
RSct
Rmt
RSteel
0.060
0.998
0.106
(0.00)c
(0.00)c
(0.00)c
(0.86)
(0.00)c
(0.08)a
(0.31)
RSct
Rmt
RSteel
0.016
0.925
0.109
(0.34)
(0.00)c
(0.00)c
This table presents the estimated coefficients and p values of coefficients in the following
regressions: (1) Combustion industry, Rit = ai ? b1 Rmt ? b2 RSct ? b3 RSteel ? e, and (2)
Steel industry, Rit = ai ? b1 Rmt ? b2 RSct ? b3 ROil ? b4 RGas ? e. Rit is daily return on the
common stock of firm i in period t, Rmt is daily return on the market portfolio in period t, RSct is
daily return on spot contracts emission allowances in period t. Variables Roil, RGas, and RSteel
are the changes in prices of commodities of oil, gas and steel. a , b , c denote statistical significance at 1, 5, and 10 % levels, respectively
lot of companies are royally over allocated as a consequence of the crisis period
and do not need to buy emission rights for years. Or at least it has become a lot
cheaper than during the first period of Phase 2. The profits decline to a minimum
and the significance for the corporations disappears. In fact, for the ETS to generate real incentives for net cuts to emissions in the iron and steel sector, it has to
make up for the over-allocation to the sector thus far, and which has been compounded by the recent recession, by mandating much greater cuts in the period
from 2013 to 2020 (Pearson 2010). A number of improvements for the EU ETS
have already been agreed, such as increased auctioning from 2013 onwards. But
with the banking of permits between phases, there is likely to be a significant
hangover effect from Phase 2. This could weaken and undermine the effectiveness
of the ETS from 2013 onwards.
123
7.6 Conclusions
This chapter addresses the question of whether price changes of trading EUAs
have an effect on stock returns and whether this effect differs over time. The
empirical analysis is designed to be able to answer these two questions by creating
several separate sub-periods within the two phases of the EU ETS. The investigation considers a sample of firms from the European steel industry in addition to
firms in the combustion industry.
For sub-period one in Phase 1, the introduction phase of the EU ETS, it is found
that there is a statistically significant effect of price changes of emission trading on
stock returns for both the combustion and steel industries. During the EUA price
crisis, period two of Phase 1, the results provide mixed findings; the effect is
statistically significant for firms in the combustion industry, but not for firms in the
steel industry. The question needing to be answered is what the difference might be
between the steel and combustion industry that makes them different during this
crisis period? It is suggested that the explanation is that the steel companies were
unable to sell their EUAs in this period because of the large supply of EUAs on the
market, while the combustion industry still had to buy EUAs. The third sub-period
of Phase 1, the post-crisis period, is expected to hold no significance due to over
allocation and low EUA prices until the period ends.
Phase 2, which consists of two sub-periods, is also examined. In sub-period one,
price changes of emission trading have a statistically significant impact on stock
returns for both the combustion and steel industry. In this phase, the over-
124
allocation is building up, and EUA prices gradually fall. For sub-period two there
are no significant effects for both industries because of the huge amount of unused
EUAs that have built up during the financial crisis.
Production and EUAs do not seem to be related which could be explained by
the fact that many orders are placed in advance and will thus not react fast on
changing EUA prices. Furthermore, the insignificant estimated coefficients of
corporate size and leverage indicate that those variables are not related to the
exposures of trading price of EUA.
A positive effect of emissions trading on the stock value of companies has
important implications for policy makers. The ETS is supposed to make companies aware of their emissions and polluting activities by letting them pay for their
waste. The proposed changes in the third phase go a long way because combustion
companies will be forced to buy all their EUAs from an auction. However, it is
noted that they will be able to pass through the additional costs just as they have
done over recent years. Then there is the case of the energy intensive industries.
They have negotiated that they will receive free allocations for Phase 3 as well
thus ensuring windfall profits for years to come.
Period 13 shows contradictory evidence for the two industries. The difference
is likely to be caused by the fact that one industry is a net buyer and the other is a
net seller of EUAs. The combustion industry could buy from various sources
holding excess EUAs because of the over-allocation while the steel industry had a
hard time selling their excess EUAs to the few buyers left in the market. The fact is
that the combustion industry was still spending money while the steel industry was
not able to create additional income through their EUAs.
For future research, it would be interesting to redo the analyses with a complete
data set from Phase 2, which covers the period 20082012. Additional insight is
necessary in the second trading phase. Literature on the first phase is plentiful, but
the effects for the second phase are relatively unknown. As this chapter plays a
leading role for this phase, it would be interesting to have future research confirm
or share new insights into these industries and their relationships to the EU ETS for
the later period of the second phase.
As reported by Mansanet Bataller et al. (2007), EUA prices have developed
very similarly in all marketplaces. According to those authors, the choice of
marketplace should not be crucial while researching the effects of the EU ETS.
However, it is the only study claiming this fact and it would be interesting to see if
this holds for the second phase.
References
Anderson, B., & Di Maria, C. (2011). Abatement and allocation in the pilot phase of the EU ETS.
Environmental and Resource Economics, 48(1), 83103.
Bartram, S. M. (2005). The impact of commodity price risk on firm valuean empirical analysis
of corporate commodity price exposures. Multinational Finance Journal, 9, 161187.
125
Daskalakis, G., Psychoyios, D., & Markellos, R. (2009). Modeling CO2 emission allowance
prices and derivatives: Evidence from the European trading scheme. Journal of Banking and
Finance, 33, 12301241.
Ellerman, A. D., & Joskow, P. L. (2008). The European Unions emissions trading system in
perspective. Arlington: PEW Centre on Climate Change.
Gronwald, M., Ketterer, J., & Trueck, S. (2011). The dependence structure between carbon
emission allowances and financial marketsa copula analysis (Working paper No. 3418).
Munich, Germany: CESinfo.
Hassan, M., & Molho, N. (2009). Phase III of the EU emissions trading scheme: Your Q&A
quide. London: CMS Cameron McKenna LLP Publications.
Haushalter, D. G. (2000). Financing policy, basis risk, and corporate hedging: Evidence from oil
and gas producers. Journal of Finance, 55, 107152.
He, J., & Ng, L. K. (1998). The foreign exchange exposure of Japanese multinational
corporations. Journal of Finance, 53, 733753.
IETA. (2008). State and trends of the carbon market 2008. Washington DC: World Bank CF
Research Report.
Jin, Y., & Jorion, P. (2006). Firm value and hedging: Evidence from U.S. oil and gas producers.
Journal of Finance, 61, 893919.
Jorion, P. (1990). The exchange-rate exposure of U.S. multinationals. Journal of Business, 63,
331345.
Kara, M., Syri, S., Lehtila, A., Helynen, S., Kekkonen, V., Ruska, M., et al. (2008). The impacts
of EU CO2 emissions trading on electricity markets and electricity consumers in Finland.
Energy Economics, 30, 193211.
Kemfert, C., Kohlhaas, M., Truong, P. T., & Protsenko, A. (2006). The environmental and
economic effects of European emissions trading. Climate Policy, 6, 441455.
Lund, P. D. (2007). Impacts of EU carbon emission trade directive on energy-intensive
industriesIndicative micro-economic analyses. Ecological Economics, 63, 799806.
Mansanet Bataller, M., Pardo, A., & Valor, E. (2007). CO2 prices, energy, and weather. The
Energy Journal, 28, 7392.
Nance, D. R., Smith, C. W., & Smithson, C. W. (1993). On the determinants of corporate
hedging. Journal of Finance, 48, 267284.
Oberndorfer, U. (2009). EU emission allowances and the stock market: Evidence from the
electricity industry. Ecological Economics, 68, 11161126.
Pearson, A. (2010). The carbon rich list: The companies profiting from the EU Emissions Trading
Scheme. London, UK: Sandbag Climate Campaign CIC.
Reinaud, J. (2007). CO2 allowance and electricity price interaction, impact on industrys
electricity purchasing strategies in Europe. (IEA information paper), Paris, France: OECD/
International Energy Agency.
Sijm, J., Neuhoff, K., & Chen, Y. (2006). CO2 cost pass through and windfall profits in the power
sector. (Cambridge Working Papers in Economics No. 0639). Cambridge, UK: Faculty of
Economics, Cambridge University.
Smith, C. W., & Stulz, R. M. (1985). The determinants of firms hedging policies. Journal of
Financial and Quantitative Analysis, 20, 391405.
Veith, S., Werner, J. R., & Zimmermann, J. (2009). Capital market response to emission right
returns: Evidence from the European power sector. Energy Economics, 31, 605613.
Xu, C., & Cang, D. (2010). A brief overview of low CO2 emission technologies for iron and steel
making. Journal of Iron and Steel Research International, 17(3), 17.
Zachmann, G., & von Hirschhausen, C. (2008). First evidence of asymmetric cost pass-through of
EU emissions allowances: Examining wholesale electricity prices in Germany. Econonic
Letters, 99, 465469.
Part III
Chapter 8
Abstract This chapter examines the EU Emissions Trading Scheme options and
futures markets dynamics during the period 20052011. Observations on returns,
volatilities and volumes on derivative instruments are studied. In addition, spot/
future correlations, term structures and option implied volatility smiles and surfaces
are examined. The aim is to ascertain whether the behavior of the EU ETS derivatives
markets can be compared to that of commodity markets, specifically the developed
West Texas Intermediate (WTI) crude oil derivatives market. The results indicate
that the EU Emissions Trading Scheme derivatives markets have matured markedly
since the start of Phase 2 of the Scheme, with rising volumes and declining return
volatilities. Spot/future correlations, term structures and option volatility smiles and
surfaces exhibit comparable behavior over time, albeit with certain discrepancies,
with that found in the developed WTI crude oil derivatives market. These results are
valuable both for traders of EU allowances and for those policy makers seeking to
improve the design of the EU Emissions Trading Scheme.
129
130
D. Bredin et al.
8.1 Introduction
The European Union Emissions Trading Scheme (ETS) was established in 2005 as
the cornerstone of the EU effort to comply with the demands of the Kyoto protocol. The protocol, adopted in 1997, aims to reduce the worlds CO2 emissions to
pre-1990 levels by the year 2020. The idea behind the scheme is to incentivize
reduced carbon emissions by creating a Europe-wide market in CO2 emissions
where allowances can be traded between countries and companies. The pilot Phase
of the EU ETS ran from 2005 to 2007, while the Phase 2 (Kyoto) is running from
2008 to 2012 and the Phase 3 will run from 2013 to 2020. This chapter seeks to
analyze the rapidly growing CO2 derivatives markets that have developed as a
result of the creation of the ETS.
This study is the first to comprehensively examine the dynamics of the ETS
derivatives markets for both Phases 1 and 2. Although there is some research of this
nature examining Phase 1 of the EU ETS, the uncertainty, primarily as a result of the
limitations on inter-period banking, means that a Phase 2 analysis should be more
informative. The primary aim is to investigate whether the market has changed since
the end of Phase 1 and whether there is any evidence of the emergence of maturing
market dynamics. Apart from investigating evidence of a maturing market, another
question addressed is whether the EU emissions market can be seen to exhibit
commodity like behavior. This is a research question which has been alluded to in
much of the literature but there is no consensus on whether EU emissions contracts fit
in this category (see, Bredin and Muckley 2011a, b). The study adopts oil derivatives,
specifically West Texas Intermediate (WTI) crude oil futures and options, as
benchmarks to analyze the development of the ETS derivatives markets. ETS
derivatives, and futures in particular, have overtaken the spot market in terms of
trading volume, and so an analysis of derivative behaviors will be more informative.
The research finds that volatility declined dramatically in Phase 2 while the
correlation between spot and futures contracts stabilized at a high level. It also
finds the term structure of futures prices to indicate contango, contradicting previous findings for typical commodity markets which find evidence of backwardation (see Pindyck 2001; Considine and Larson 2001a, b; Milonas and Henker
2001). In contrast, the current analysis of the term structure of futures prices for the
WTI crude oil data indicates periods of contango and periods of backwardation.
This echoes the findings of Escobar et al. (2003) who also find very little evidence
of consistency in term structures of oil futures over time. Consistent with Samuelson (1965) there is evidence to indicate a declining term structure of both EUA
and oil volatility. In the options market a clear development in the volatility smiles
and surfaces in Phase 2 relative to Phase 1 is observed. The Phase 2 analysis
indicates consistent volatility smiles and a persistent forward skew, with the
number of contracts traded approaching levels observed in the WTI crude oil
options market. Overall the results are clearly indicative of an increased maturity
and stability in the EU ETS derivatives markets. With the exception of the future
term structure, the EUA derivatives behavior is generally consistent with the WTI
131
It is noteworthy that, according to Europol, the European law-enforcement agency, the marked
development in these statistics, for the EU ETS derivatives markets, occurred despite a backdrop
of Europe having lost about 7.4 billion euros in taxation revenue for the 18 months prior to
December 2009 because of CO2 VAT trading fraud.
132
D. Bredin et al.
with the volume of transactions in the market going from 252 million tons in 2005
to 1.5 billion tons in 2007 (Chevalier et al. 2009). Fusaro (2007) estimated,
assuming that commodities usually trade at between 6 and 20 times the underlying
market, that the value of the global carbon market could grow to over $3 trillion in
the future, clearly illustrating the growth potential in this market.
In recent years the ETS has continued to grow with total ETS volumes,
including spot, forward and option markets, reaching 7.025 billion tons of CO2 in
2010, a 7 % increase on 2009. In terms of market composition, spot volumes fell
54 % in 2010 to 603 million tons, representing 9 % of the total ETS volume. In
contrast to this, option volume grew to 14 % of the market to overtake spot for the
first time while EUA futures, forwards and options combined, accounted for over
75 % of the total market (Chestney 2011).2
Recent figures also illustrate the dominance of exchanges in the trading of EUAs
with exchange volumes jumping 23 % to 3.84 billion tons in 2010 to account for over
55 % of the market while over the counter (OTC) volumes fell 8 % on 2009 to make
up under 45 % of the market. Significantly for this study, ICE Futures Europe, the
exchange from which the data is taken, was the dominant exchange accounting for
3.4 billion tons or 89 % of total exchange volume (Chestney 2011).
133
Looking at the effectiveness of the first phase in achieving its goal of reduced
emissions it can be seen that it may not have been entirely successful. Bloch
(2010) described the results of the early stages of the ETS as lackluster.
The author argues that the ETS is subject to extreme price volatility as a result of the
complete inelasticity of supply of permits and high inelasticity of demand over the
short-term. Bloch (2010) also points out that the scheme has been possibly marred
by rent-seeking behavior with participants seeking to maximize revenue rather than
focusing on emissions reduction. Lowrey (2006) highlights the centrally important
element of the scheme being the establishment of a market determined price for
CO2 allowances. This market price should serve as an incentive to reduce emissions
and to invest in low carbon technologies. Closely linked to this idea, Paolella and
Taschini (2008) discuss how the primary aim of the scheme should be to produce a
scarcity of allowances leading to an upward price trend. Clearly there was no
scarcity of allowances in Phase 1 and the collapse of the spot price led Newbury
(2009) to conclude that the Phase 1 EU ETS was not delivering the stable carbon
price necessary for long-term, low-carbon investment decisions.
134
D. Bredin et al.
between these drivers and the price of CO2 emissions reflects evidence of reverse
causality with CO2 exerting a significant influence on the prices of power, gas and
several other emission related commodities and activities.
The papers of Delarue et al. (2008) and Ellerman and Feilhauer (2008) show
that energy prices are the most important driver of CO2 prices with high energy
prices leading to an increase in carbon prices. However, Kanen (2006) highlights
the role of Brent crude as a driver of carbon prices (via its impact on natural gas
and power prices). Drawing on these studies, Brent crude is examined as the
benchmark to compare with the ETS market. Finally, while the early literature on
ETS price drivers focused on Phase 1, Bredin and Muckley (2011a) note that in
Phase 2, stable relationships formed between allowances and theoretically consistent determinants. There had been virtually no evidence of this stability using
Phase 1 samples.
135
dynamics developing in the relationship between spot and futures volatility.4 This
is not surprising given the short sample period and the considerable uncertainty
throughout Phase 2. Overall the results reported by Borak et al. (2006) contradict
the literature in this area and provided ambivalent results on the Samuelson effect.
Borak et al. (2006) conclude that EUA price behavior in spot and futures markets
is substantially different to that in commodities markets.
Borak et al. (2006) also examine the correlation between spot and futures prices
and find a very strong correlation between spot and Phase 1 futures prices, with
reduced correlations for Phase 2 futures. The correlations also decline as there is a
movement out of the maturity spectrum, indicating that investors opinions about
distant time periods are less affected by short-term spot movements. Finally, Borak
et al. (2006) also find that futures expiring in the same phases exhibit very strong
correlations.
Other papers which have looked into the volatility and price dynamics of the
ETS include Benz and Truck (2009) and Paolella and Tachini (Paolella and
Taschini 2008) which both examine the price behavior of EUAs indicating different models for the dynamics of short-term EUA spot price behavior. Benz and
Hengelbrock (2008) were the first to extend this type of analysis to the futures
market, examining the price behavior, liquidity and correlations in different EUA
contracts.5 This study, based around Phase 1 data finds that transaction costs and
spreads had fallen rapidly in 2007 while trading volumes and intensity had
increased markedly. Recent studies indicate the existence of a convenience yield in
the EU ETS market (Uhrig-Homburg and Wagner 2009; Rittler 2011). Both
studies examine the relationship between spot and futures prices in the EU ETS,
with Uhrig-Homburg and Wagner (2009) adopting daily data and Rittler (2011)
adopting both daily and transaction (tick) level data. The authors find evidence of
long-run (or cointegrating) relationships, a convenience yield and price discovery
via futures markets (rather than spot markets).
Borak et al. (2006) also find evidence of a weakly increasing term structure when examining
Phase 2 instruments.
5
Bredin, Hyde and Muckley (Bredin et al. 2009) examine the microstructure behaviour of
trading volume, return volatility and transaction duration (time between consecutive trades) for
Phase 1 and the initial contract of Phase 2 using transaction level data.
136
D. Bredin et al.
Trading scheme. The price and volume data for all the ECX contracts started in
April 2005 although many of the later maturing contracts were not actively traded
during the early stages of their existence.
137
110
1,197
2,555
5,433
7,044
6,484
24,568
22,457
11,254
2,649
9,021
45
365
135
865
5,711
4,766
28,221
47,428
83,365
94,620
85,862
137,594
173,644
197,534
99,370
0
20
0
10
241
59
281
623
3,856
7,448
6,054
7,827
15,215
34,821
108,897
339,461
340,192
290,623
178,391
Dec-10
0
0
0
0
0
10
57
35
571
3,127
2,327
3,414
6,376
7,461
12,074
29,320
44,849
47,395
231,813
489,128
740,989
413,504
292,319
92,772 959,525 1,334,019 2,324,769
Dec-11 Totals
0
9,784
0
13,782
0
18,701
10
34,169
0
48,602
0
30,164
0
66,874
0
70,543
499
99,545
1,573 109,417
860 104,124
2,196 151,031
6,848 202,083
5,836 245,652
8,335 228,676
19,021 387,802
18,861 403,902
16,4 78 354,496
40,051 450,255
58,422 547,550
124,468 865,457
79,509 493,013
246,261 538,580
629,228 5,474,202
futures.6 This provides evidence of the stability and maturity of the ECX market
since the start of Phase 2.
The summary statistics for WTI December futures returns are reported in the Appendix to this
chapter, see Table 5.8.
138
D. Bredin et al.
0.22
0.14
-0.04
0.26
0.13
0.38
0.20
-0.91
-0.4 1
-1.61
0.27
0.12
0.42
0.21
-0.91
-0.44
-1.87
-2.89
-2.59
-0.83
-4 .99
Dec-08
(%)
Dec-09
(%)
Dec-10
(%)
Dec-11
(%)
0.28
0.07
0.15
0.50
-0.71
-0.31
-0.06
0.18
0.13
0.22
-0.15
0.38
-0.02
-0.33
-0.75
0.28
0.07
0.15
0.53
-0.69
-0.30
-0.06
0.17
0.12
0.23
-0.15
0.38
0.01
-0.35
-0.88
0.24
0.11
0.00
0.00
0.28
0.06
0.15
0.55
-0.68
-0.30
-0.06
0.16
0.10
0.23
-0.15
0.38
0.02
-0.36
-0.88
0.27
0.08
0.00
-0.18
0.14
-0.07
0.11
0.28
0.06
0.15
0.57
-0.67
-0.29
-0.06
0.15
0.09
0.24
-0.14
0.38
0.04
-0.35
-0.89
0.29
0.09
-0.02
-0.19
0.13
-0.09
0.11
-0.10
Table 8.3 reports the summary statistics for the December 2007 and the
December 2010 futures. The December 2010 futures summary statistics are
reported both over the full sample (February 2006November 2010) as well as
individually for both Phases 1 and 2. For comparison the summary statistics for
WTI futures are also reported. As can be seen, the standard deviation (over 8 %
per day) of the December 2007 contracts for Phase 1 is extremely high compared
to both the WTI and the December 2010 contract. This compares to the WTI
standard deviation of 1.4 % and indicates the levels of volatility in the ECX
market in what was its most difficult and volatile period. Looking across the life of
the December 2010 contract it can be seen that the standard deviation of daily
returns on this contract fell to 2.67 % in comparison with 1.76 % for the WTI
December 2110 futures for the same period. The dramatic fall in standard deviations is due to the move from pilot to full implementation and the relaxation of
the no-banking rule for Phase 2.
Examining the December 10 futures behavior over the two phases yields an
interesting comparison, in particular in relation to the WTI comparison. The
standard deviation of the ETS futures for a Phase 1 sample is 3.02 %. This reflects
the fact that this contract was for delivery in Phase 2, meaning that it wasnt
139
WTI
Mean
-0.0131
0.0007
Median
0
0.0013
Mode
0
0
Standard Deviation
0.0818
0.0141
Skewness
0.1911
-0.1786
Kurtosis
12.2179
3.4955
Dec - 10 Futures Summary Stats (Feb-06 to
Nov-10)
ECX
WTI
Mean
-0.0048
0.001
Median
0
0.294
Mode
0
0
Standard Deviation
0.0267
0.0176
Skewness
-0.7356
-0.037
Kurtosis
15.7172
0.5206
ECX
WTI
Mean
-0.0019
0.0055
Median
0.0021
0.0003
Mode
0
0
Standard Deviation
0.0302
0.0115
Skewness
-1.2879
-0.1005
Kurtosis
21.4549
3.0201
Dec - 10 Futures Summary Stats (Feb-06 to
Nov-10)
ECX
WTI
Mean
-0.0067
-0.0019
Median
0
0.0008
Mode
0
0
Standard Deviation
0.0242
0.0205
Skewness
-0.0473
0.0019
Kurtosis
5.285
4.4034
impacted as strongly by the collapse in the EUA prices at the end of Phase 1. This
compared with a WTI standard deviation of 1.15 % in the same period. Since the
start of the second phase, the standard deviation of daily returns for December 10
futures fell to 2.67 % while the WTI contract had a standard deviation of 1.76 %.
This confirms that a more stable market developed with less volatility than the
erratic first phase. Another point of note is that the standard deviations have fallen
to a level much closer to those of the benchmark commodity.
In addition to these findings it can also be seen that large changes in the
skewness and kurtosis are exhibited by EUA returns since the early stages of
140
D. Bredin et al.
3.69
2.03
2.34
3.75
2.00
2.30
1.46
10.25
2.48
3.92
3.77
2.06
2.28
1.39
10.07
2.46
5.13
10.04
8.05
11.02
35.34
Dec-08
(%)
Dec-09
(%)
Dec-10
(%)
Dec-11
(%)
3.69
2.59
2.42
2.32
5.70
2.01
2.86
3.18
2.70
1.92
1.68
2.06
2.05
2.38
3.41
3.64
2.59
2.45
2.40
5.67
1.92
2.82
3.05
2.67
1.83
1.64
2.01
2.07
1.92
1.68
2.06
2.05
2.38
3.41
3.59
2.57
2.42
2.43
5.61
1.87
275
2.97
2.87
2.21
1.63
1.97
2.08
2.23
2.93
4.61
2.43
2.02
2.34
1.61
2.06
1.33
1.15
3.54
2.54
2.43
2.72
5.56
1.87
267
2.90
2.62
1.69
1.62
1.91
2.05
2.18
2.88
4.52
2.40
1.98
2.28
1.56
2.03
1.32
1.11
trading scheme. For example from Table 8.3, the December 2007 contract has a
kurtosis of over 12 compared to only 3.5 for the December 2007 WTI future. This
indicates that EUA futures have a relatively fat tailed, low distribution compared
to the more normally distributed WTI future. The implication is that EUA returns
have infrequent, but large deviations from the mean, as opposed to frequent
modest departures for WPI.
The EUA December 10 futures contract indicate similar behavior in Phase 1 of
the scheme with high kurtosis of 21 relative to the WTI future, while the EUA
returns were also very strongly negatively skewed during this period. However,
looking at the behavior of the EUA December 10 contract for Phase 2, there is
evidence of extremely mild negative skew and low kurtosis, which is consistent
with that of the WTI market. Again this clearly illustrates the relative maturity of
the EUA market in Phase 2 as its behavior move into line with that of the
benchmark commodity market.
Table 8.4 reports the standard deviations of daily returns on a quarterly basis
for all December contracts from 2005 to 2011. Evidence of large volatility in
Phase 1 of the scheme is again reported, with standard deviations as high as
35.34 % for the last three months of 2007. In Phase 2, there is persistent evidence
141
of a maturing market, with lower and more persistent standard deviations. Also a
clear disconnect between the volatility of Phase 1 and Phase 2 expiry contracts can
be seen.7
Overall the findings on volatility are unambiguous, with all indicators pointing
to a more stable market over the last few years. The market has become far more
stable in Phase 2 and there is evidence that the changes in market structure since
the end of Phase 1 are proving very effective. Comparison with the WTI Crude
market provides further evidence of EUA market development, with consistent
levels of volatility emerging in both markets.
7
See Bredin, Hyde and Muckley (Bredin et al. 2009) for an intra-day volatility and volume
analysis of Phase 1 and early Phase 2 EUA December expiry futures contracts.
8
Borak et al. (2006) used prices to conduct their correlation analysis.
9
Borak et al. (2006) also find that futures prices within the same Phase exhibited stronger
correlations.
10
The sample of data used here is considerably larger than that adopted by Borak et al. (2006).
142
D. Bredin et al.
0.600
1
0.808
0.980
1
0.212
0.964
0.980
1
0.014
0.809
0.522
0.090
1
2009
2010
2011
2012
0.018
0.798
0.513
0.088
0.981
1
0.019
0.798
0.507
0.086
0.967
0.989
1
0.019
0.844
0.500
0.085
0.948
0.971
0.969
1
0.018
0.768
0.492
0.083
0.932
0.956
0.955
0.988
1
correlations are explained by the fact that they are based on a longer sample period
than the 2007 future. This meant that the relationship between spot and futures
recovered in Phase 2 and has been far more stable in the last few years.
Comparing these results to the WTI Crude market a far more consistent and
high correlation across all oil contracts with correlation again decreasing with
maturity is evident. As is the case with volatility and returns ECX spot/futures
correlation behaviors seem to again be reflecting WTI Crude much more since the
beginning of Phase 2.
Qualitatively similar term structures are present in 2009. These are available from the authors
on request.
143
Mar-08
Mar-09
Mar-10
Mar-11
435
3334
0
68
1362
0
20
12
1795
0
0
0
0
847
3769
1430
1827
847
0
0
0
0
10
7
17
0
0
0
0
0
10
10
Dec-09
(%)
Dec-10
(%)
Dec-11
(%)
1.32
1.13
1.01
1.36
1.34
0.95
0.96
1.34
1.74
2.10
3.12
3.89
3.04
1.94
2.22
1.93
1.33
1.12
1.03
1.31
1.34
0.94
0.90
1.24
1.70
1.99
2.87
3.20
2.51
1.56
1.72
1.46
1.50
1.93
1.57
1.74
1.35
1.13
1.05
1.30
1.34
0.93
0.88
1.21
1.68
1.95
2.80
2.86
2.29
1.42
1.53
1.39
1.34
1.74
1.36
1.14
1.42
1.35
1.45
1.97
1.32
1.19
1.11
1.61
1.48
1.12
1.45
2.07
1.30
1.13
1.02
1.43
1.36
0.99
1.09
1.59
1.86
2.28
3.47
4.53
Qualitatively similar term structures are present in 2009. These are available from the authors on
request
phases with a flat term structure at the short end where the Phase 1 futures prices
had converged on zero. In 2008, 2009 and 2010 consistent increasing term
structures are observed which provides further evidence of the maturity of the
market in Phase 2. The term structures from April 2008, April 2009 and April 2010
look almost identical and the market is clearly in contango as can be seen from the
increasing futures prices for all maturity periods.
144
D. Bredin et al.
0.21
0.01
-0.41
-0.39
0.26
0.04
-0.22
-0.22
0.21
0.11
0.22
0.44
0.21
0.03
-0.15
-0.21
0.23
0.02
0.14
0.14
0.48
0.19
-0.98
-1.74
Dec-09
(%)
Dec-10
(%)
Dec-11
(%)
0.19
0.02
-0.13
-0.18
0.21
0.01
0.11
0.17
0.44
0.26
-0.85
-0.48
0.00
0.34
0.19
0.24
0.18
0.02
-0.12
-0.15
0.18
0.04
0.09
0.18
0.41
0.28
-0.73
-0.43
0.06
0.29
0.15
-0.12
0.17
-0.18
0.03
0.03
0.18
0.01
-0.13
-0.12
0.16
0.05
0.09
0.19
0.40
0.28
-0.67
-0.42
0.08
0.27
0.13
-0.09
0.15
-0.16
0.04
0.17
Qualitatively similar term structures are present in 2009. These are available from the authors on
request
The results support the findings of Borak et al. (2006) with evidence both of a
much more settled and consistent market in Phase 2 and also evidence of
increasing term structures in each period.
Figure 8.4 reports the term structure of futures volatility for all 3 month periods from
the years 2006 to 2009 for spot and futures prices for delivery from 2005 to 2012. Overall
the results in Fig. 8.4 are quite intuitive. The results for Phase 1 are quite dynamic, but
are generally consistent with those reported by Borak et al. (2006). However, for Phase 2
there is evidence of a weakly declining term structure in ECX futures volatility with
maturity. This result is in keeping with Samuelsons (1965) findings.
145
82,270
324,876
2,012,976
2,582,757
21,718
146,640
351,531
488,262
683,033
1,045,992
3,031,371
3,190,314
4,274
36,972
100,807
206,619
293,315
428,891
631,966
526,075
736,991
1,017,666
2,777,140
2,812,657
DEC-09
DEC-10
2,453
12,018
31,350
64,479
75,280
104,420
151,485
187,022
272,013
432,161
473,930
464,408
545,931
731,885
2,494,281
2,707,020
1,791
4,686
11,074
20,202
30,181
32,019
59,805
74,356
103,752
131,030
140,612
147,125
192,955
222,488
365,184
367,202
616,590
984,339
3,037,636
2,805,307
5,002,879 8,958,861 9,573,373 8,750,136 9,348,334
DEC-11
TOTALS
850
753
2,827
5,734
17,626
12,666
29,569
38,963
34,858
42,252
34,340
36,279
46,694
45,413
78,594
96,809
154,080
239,219
490,133
784,926
2,192,585
113,356
525,945
2,510,565
3,368,053
1,099,435
1,623,988
3,904,196
4,016,730
1,147,614
1,623,109
3,426,022
3,460,469
785,580
999,786
2,938,059
3,171,031
770,670
1,223,558
3,527,769
3,590,233
43,826,168
It is important to note that trading in the December WTI Futures contracts normally ceased in
mid-November in their years of maturity. As a result of this the standard deviations, returns and
volumes for these contracts in their final three months are based around a smaller sample
Table 8.10 Correlation between WTI futures returns
Delivery
2006
2007
2008
2006
2007
2008
2009
2010
2011
0.786
1
0.698
0.844
1
2009
2010
2011
0.671
0.778
0.898
1
0.657
0.738
0.826
0.863
1
0.656
0.722
0.788
0.798
0.929
1
periods in Phases 1 and 2 and comparing them with the considerably more mature
WTI options market.
146
D. Bredin et al.
April 2006
2
-1
ec
n-
Ju
-1
12
11
D
ec
n-
Ju
ec
-1
10
n-
Ju
ec
-0
09
8
-0
Ju
n-
08
ec
n-
Ju
ec
-0
07
6
-0
n-
Ju
ec
D
Ju
n-
06
Price
Delivery Period
April 2007
ct
-1
2
O
r-1
Ap
ct
-1
1
r-1
-1
ct
Ap
0
Ap
r-1
-0
ct
r-0
Ap
ct
-0
8
r-0
Ap
7
-0
ct
Ap
r-0
Price
Delivery Period
April 2008
-0
Ju 8
n09
D
ec
-0
9
Ju
n1
D 0
ec
-1
0
Ju
n11
D
ec
-1
1
Ju
n12
D
ec
-1
2
Ju
n1
D 3
ec
-1
3
Ju
n14
D
ec
-1
4
ec
Price
Delivery Period
Fig. 8.3 Term structures for EUA futures prices
147
April 2009
Ju
n11
D
ec
-1
1
Ju
n12
D
ec
-1
2
Ju
n13
D
ec
-1
3
Ju
n14
D
ec
-1
4
Ju
n09
D
ec
-0
9
Ju
n10
D
ec
-1
0
Price
Delivery Period
April 2010
4
ec
-1
14
D
n-
3
Ju
ec
-1
13
D
n-
2
Ju
-1
ec
D
Ju
n-
12
1
-1
ec
D
n-
11
0
Ju
-1
ec
D
Ju
n-
10
Price
Delivery Period
Fig. 8.3 continued
One common type of volatility smile is the reverse volatility skew. This is the case
where implied volatilities are higher for low strike prices (ITM calls/OTM puts) than
they are for high strike prices (ITM puts/OTM calls). This shape is common in both
equity and some commodity markets. An intuitive explanation for this pattern is that
investors are generally concerned about market crashes and expect more volatility
when the price is falling than when the price is rising. Another variant on the volatility
smile that is common among commodities is known as a forward skew. This is where
implied volatility is increasing with strike price. The forward skew pattern is common for options on commodities because in many commodity markets, unlike equity
markets, shocks generally lead to upward price trends, making investors more wary
of rising prices than price falls. This drives up demand for in the money calls creating
a premium for these options and leading to an upward skew.
8.4.1.1 WTI Crude Volatility Smiles
The first step in analyzing the volatility smiles is to examine the smiles on different
options in 2007. Is it the case that standard smiles are emerging during early Phase
1 and particularly are there any commodity like features? Before examining the
EUA markets, a sample of WTI volatility smiles is observed as an indication of
typical commodity option behavior.
148
D. Bredin et al.
Return Volatility
25%
Dec-Feb
20%
Mar-May
Jun-Aug
15%
Sep-Nov
10%
5%
0%
Spot
Delivery Period
Volatility Term Structure
4 Quarters 2007
Return Volatility
25%
Dec-Feb
20%
Mar-May
15%
Jun-Aug
Sep-Nov
10%
5%
0%
Spot
Delivery Period
Volatility Term Structure
4 Quarters 2008
Return Volatility
25%
Dec-Feb
20%
Mar-May
15%
Jun-Aug
Sep-Nov
10%
5%
0%
Spot
Dec-08
Dec-09
Dec-10
Dec-11
Delivery Period
Dec-12
149
4 Quarters 2009
25%
Return Volatility
Dec-Feb
20%
Mar-May
15%
Jun-Aug
Sep-Nov
10%
5%
0%
Spot
Dec-09
Dec-10
Dec-11
Dec-12
Delivery Period
Fig. 8.5 Implied volatility smile for WTI options in October 2007
Figure 8.5 displays the implied volatility smiles of WTI options with maturities in
December 2007, 2008, 2009 and 2010 based on data from October 2007. The figure
indicates that the WTI options are exhibiting exactly the type of reverse skew that is
discussed earlier, with higher implied volatilities for low strike options and a smooth
curve across all strikes. This is clearly a mature and liquid market, with options
actively traded across a wide variety of strikes. The smiles on the December 2008,
2009 and 2010 expiring options are all classic volatility smiles. Of note in the Fig. 5.5
is how the smile for the December 2007 option shows some inconsistency with the
other options. This is not an uncommon phenomenon and it is a result of the fact that,
being closer to maturity, this option would be subject to unusual market forces and
price distortions. Another point to consider when observing these WTI options is the
wide range of strikes which are actively traded in the market.
150
D. Bredin et al.
Fig. 8.6 Implied volatility smiles for ECX options in October 2007
151
Fig. 8.7 Implied volatility smiles for ECX options in October 2010
some mild evidence of volatility smiles in these markets. This is not surprising,
given the early stages of the EUA market and the options market in particular.
The thin range of strikes has the implication of displaying the sharp trough on the smile.
152
D. Bredin et al.
Fig. 8.8 WTI implied volatility scatter on the 11th of September 2007
Building volatility surfaces involves a large amount of interpolation and the surfaces would
primarily consist of prices and implied volatilities that are not available in the market.
14
The equivalent surfaces are reported in the Appendix, Figs. 8.118.14.
153
Fig. 8.9 ECX implied volatility scatter on the 11th of September 2007
Fig. 8.10 ECX implied volatility scatter on the 12th of July 2010
A very consistent surface with no major kinks and with many contracts of
differing maturities all behaving similarly is reported in Figure 8.8. In addition, the
near expiry options have a slightly higher implied volatility for all strikes which is
consistent with the discussion above. This smooth and consistent implied volatility
behavior illustrates the stability, maturity and liquidity of the WTI Crude option
market.
154
D. Bredin et al.
Fig. 8.11 WTI implied volatility scatter on the 4th of January 2011 (Excluding near maturity
options for clarity)
15
One interesting fact to note is how the number of strikes actively traded increases as the
maturity of the options increases.
155
Fig. 8.12 WTI implied volatility surface on the 11th of September 2007
Fig. 8.13 ECX implied volatility surface on the 11th of September 2007
does support the finding that the EUA options market has developed since its early
Phase 1 difficulties. Although it is nowhere near the WTI Crude market in terms of
depth or breadth, the EUA market would now appear to be exhibiting mature
option market behavior.
8.5 Conclusion
In this chapter, the dynamics of the EU ETS derivatives markets are examined.
Specifically, the behavior of the ETS futures and options markets using WTI crude
156
D. Bredin et al.
Fig. 8.14 ECX implied volatility surface on the 12th of July 2010
oil futures and options as a benchmark are assessed. The futures market saw
volumes increasing rapidly from the markets inception while volatility fell dramatically in Phase 2. Skewness and kurtosis of returns have also fallen dramatically, while the correlation between spot and futures prices has settled at a high
level in Phase 2, indicating an emerging stability in the market. The term structure
of futures prices is found to indicate contango, while the term structure of futures
volatility is declining and so is consistent with the Samuelson (1965) effect. The
term structures of futures volatility for WTI crude oil also declines. In contrast, the
term structure of futures prices for WTI crude oil exhibits periods of contango and
periods of backwardation. Most studies examining commodities indicate that the
term structure of futures prices tend to exhibit backwardation (Pindyck 2001;
Considine and Larson 2001a, b; Milonas and Henker 2001).
In the options market clear development in the volatility smiles and scatters in
Phase 2 relative to Phase 1 are observed. The Phase 2 analysis shows consistent
smiles and scatters and a persistent forward skew indicating that market participants are more fearful of sudden price increases than price falls. This is contrary to
the findings for the WTI options market which shows a reverse skew. In addition,
the volume of trading for the EUA option market has clearly developed rapidly
over the last three years.
Overall the results are indicative of an increased maturity and stability in the
EU ETS derivatives markets. Both the futures and options markets have developed
significantly since Phase 1. The price and volatility behavior of both markets is
generally consistent with other functioning commodity derivatives markets, such
as the WTI Crude derivatives markets. These findings indicate that the EU has
been successful in its establishment of these ETS derivatives markets despite the
disorder that marred the early stages of both the futures and options markets.
157
References
Alberola, E., Chevallier, J., & Cheze, B. (2009). Emissions compliances and carbon prices under
the EU ETS: A country specific analysis of industrial sectors. Journal of Policy Modelling, 31,
446462.
Benz, E., & Hengelbrock, J. (2008). Liquidity and price discovery in the European CO2 futures
market: An intraday analysis (Working Paper). Bonn: Bonn Graduate School of Business.
Benz, E., & Truck, S. (2009). Modelling the price dynamics of CO2 emissions allowances.
Energy Economics, 1, 415.
Bloch, D. (2010). A note on emissions trading: The pricing of carbon derivatives (Working
Paper). Paris: Universit Paris VI Pierre et Marie Curie.
Bohringer, C., Koschel, H., & Moslener, U. (2008). Efficiency losses from overlapping regulation
of EU carbon emissions. Journal of Regulatory Economics, 33, 299317.
Borak, S., Hardle, W., Truck, S., & Weron, R. (2006). Convenience yields for CO2 emissions
allowance futures contracts (Working Paper). Berlin: Humboldt-University of Berlin.
Bredin, D., Hyde, S., & Muckley, C. (2009). A microstructure analysis of the carbon finance
market. Paper presented at the Finsia/Melbourne Centre for Financial Studies Banking and
Finance Conference, Melbourne, Australia. Retrieved from http://www.melbournecentre.
com.au/Finsia_MCFS/2009/Papers for website/25_Hyde_Carbon_Futures_Trading.pdf.
Bredin, D., & Muckley, C. (2011a). An emerging equilibrium in the EU Emissions Trading
Scheme. Energy Economics, 33, 353362.
Bredin, D., & Muckley, C. (2011b). The price forming process in energy markets. In A. Dorsman,
W. Westerman, M. B. Karan, & . Arslan (Eds.), Financial Aspects in Energy: The European
Perspective (pp. 88107). Berlin: Springer-Verlag.
Bunn, D. W., & Fezzi, C. (2007). Interaction of European carbon trading and energy prices,
(Working Paper 63). Milan: Fondazione Eni Enrico Mattei.
Chestney, N. (2011). EU CO2 market strong in 2010, volume up 7 pctreport, Reuters, [Online 13/
01/2011]. Retrieved from http://in.reuters.com/article/2011/01/13/idINIndia-54143320110113.
Chevallier, J., & Sevi, B. (2009). On the realized volatility of the ECX CO2 emissions 2008
futures contract: Distribution, dynamics and forecasting (Working Paper). Paris: Universit
Paris X Nanterre.
Chevallier, J., Sevi, B., & Le Pen, Y. (2009). Options introduction and volatility in the EU ETS
(Working Paper). GRANEM.
Christiansen, A., Arvanitakis, A., Tangen, K., & Hasselknippe, H. (2005). Price determinants in
the EU emissions trading scheme. Climate Policy, 5, 1530.
Considine, T. J., & Larson, D. F. (2001a). Uncertainty and the convenience yield in crude oil
price backwardations. Energy Economics, 23, 533548.
Considine, T. J., & Larson, D. F. (2001b). Risk premium on inventory assets: The case of crude
oil and natural gas. Journal of Futures Markets, 21, 109126.
Daskalakis, G., Markellos, R. N. (2008). Are the European carbon markets efficient? Review of
Futures Markets, 17(2), 103128.
Daskalakis, G., Psychoyios, D., & Markellos, R. N. (2009). Modelling CO2 emission allowance
prices and derivatives: Evidence from the European trading scheme, Journal of Banking and
Finance, 33, 12301241.
Delarue, E. D., Ellerman, A. D., & Dhaeseleer, H. (2008). Short-term CO2 abatement in the
European power sector (Working Paper No. 08.008). Cambridge, MA: CEEPR, MIT.
Ellerman, A. D., & Feilhauer, S. (2008). A top-down and bottom-up look at the emissions
abatement in Germany in response to the EU ETS (Working Paper No. 08-017), Cambridge,
MA: CEEPR, MIT.
Escobar, M., Hernandes, N., & Seco, L. (2003). Term structure of commodities futures,
forecasting and pricing. Toronto, Canada: RiskLab, University of Toronto. Retrieved from
http://www.risklab.ca/seco/Commodities term structures.pdf.
158
D. Bredin et al.
Fusaro, P. C. (2007, September). Energy and environmental hedge funds, Commodities Now,
11(3), 6163. Retrieved from http://www.commodities-now.com/publications/commoditiesnow/publication/7.html.
Kanen, J. L. M. (2006). Carbon trading and pricing, London. UK: Environmental Finance.
Publications.
Kara, M., Syri, S., Lehtila, A., Helynen, S., Kekkonen, V., Ruska, M., et al. (2008). The impacts
of EU CO2 emission trading on electricity markets and electricity consumers in Finland.
Energy Economics, 30(2), 193211.
Lowrey, C. (2006). A changing environment. FOW Energy, Spring, 2426.
Mansanet-Bataller, M., & Pardo, A. (2008). What you should know about carbon markets.
Energies, 1, 120153.
Mansanet-Bataller, M., Pardo, A., & Valor, E. (2007). CO2 prices, energy and weather. The
Energy Journal, 28, 7392.
Mills, P., (2008). The greening of markets. Finance and Development, 45(1). Washington, DC: IMF.
Milonas, N., & Henker, T. (2001). Price spread and convenience yield behavior in the
international oil market. Applied Financial Economics, 11, 2336.
Newbury, D. (2009). The role of carbon markets in preventing dangerous climate change. House
of Commons: Environmental Audit Committee Publications.
Paolella, M. S., & Taschini, L. (2008). An econometric analysis of emission trading allowances.
Journal of Banking & Finance, 32, 20222032.
Pindyck, R. (2001). The dynamics of commodity spot and futures market: A primer. The Energy
Journal, 22, 129.
Rittler, D. (2011). Price discovery and volatility spillovers in the European Union Emissions
Trading Scheme: A high-frequency analysis. Journal of Banking & Finance, 36, 774785.
Samuelson, P. (1965). Proof that properly anticipated prices fluctuate randomly. Industrial
Management Review, 6, 4149.
Seifert, J., Uhrig-Homburg, M., & Wagner, M. (2008). Dynamic behavior of CO2 spot prices.
Journal of Environmental Economics and Management, 56, 180194.
Uhrig-Homburg, M., & Wagner, M. (2009). Futures price dynamics of CO2 emission allowances:
An empirical analysis of the trial period. Journal of Derivatives, 17, 7388.
Chapter 9
Abstract This chapter centers on the question of whether futures markets can be
used in the competitive price discovery in crude oil markets. On the one hand, the
survey in this chapter uncovers considerable evidence on the theoretical
perspective that future prices of crude oil is equal to the spot price of crude oil,
plus the cost of carry plus the endogenous convenience yield. On the other hand,
through the empirical findings built on the Alquist and Kilian (2010) model, this
chapter concurs with the previous studies documenting that futures crude oil prices
are uninformative for forecasting spot crude oil prices.
Keywords Crude oil market
Price discovery
9.1 Introduction
With a consumption level increasing steadily each year and reaching around
85 million barrels per day, crude oil is one of the most important commodities in
the world. Moreover, crude oil consumption represents a significant proportion of
. Arslan-Ayaydin (&)
Department of Finance, University of Illinois at Chicago, University Hall, 601 South
Morgan, Chicago, IL 60607, USA
e-mail: orslan@uic.edu
. Arslan-Ayaydin
Department of Finance, Hacettepe University, University Hall, 601 South Morgan, Chicago,
IL 60607, USA
I. Khagleeva
Department of Information and Decision Sciences, University of Illinois at Chicago,
University Hall, 601 South Morgan, Chicago, IL 60607, USA
e-mail: ikhagl2@uic.edu
159
160
161
time t of a contract maturing at time T is the expected spot price at time T. The
studies by Crowder and Hamed (1993) and Herbert (1993) are the leading works
investigating the properties of the spot and forward prices of commodities in the
long run, through cointegration. Moosa and Al-Loughani (1994) shows that future
prices of crude oil are efficient and unbiased predictors of spot prices. Similarly by
using 1-month, 3-month and 6-month contracts, Gulan (1998) concludes that spot
prices of crude oil can be efficiently predicted by the futures market.
However, these results are not confirmed in some recent findings, such as
(Chinn et al. 2005) and (Chernenko et al. 2004), both of which apply MincerZarnowitz regressions to assess the forecast accuracy of their models, and more
recently by (Huang et al. 2009). Similarly, (Kaufmann and Ullman 2009) find no
direct link between the WTI1 spot market and the New York Mercantile Exchange
(NYMEX) futures market.
This chapter centers on the unreached consensus in the previous literature on
the linkage between future price and spot price of crude oil. The second part of the
chapter questions if the futures prices of crude oil are equal to the spot price plus
the cost of carry (cost of storage and interest rate) plus the endogenous convenience yield. In Sect. 9.2, both theoretical models and empirical findings following
the evidence of Kaldor (1939) are discussed. In Sect. 9.3 the relationship between
future and spot prices of crude oil will be dealt with considering the theoretical
model of Alquist and Kilian (2010). Empirical testing of the results obtained by
(Alquist and Kilian 2010) is provided through extending the sample period of
crude oil prices for contracts having a 1-month horizon (Sect. 9.4). Finally
Sect. 9.5 concludes the chapter.
West Texas Intermediate (WTI) is a grading system for crude oil to be used as a benchmark in
crude oil pricing and it is the underlying commodity of crude oil futures contracts in the Chicago
Mercantile Exchange.
2
For a detailed discussion on the differences, see (Routledge et al. 2000).
3
See, among others, (Carlson et al. 2007).
162
the future price of crude oil is equal to the spot price of crude oil plus the
cost of carry (cost of storage and interest rate) plus the endogenous convenience yield.
It must be noted that, through a reduced-form model, Casassus and Collin-Dufresne (2005)
show that when the spot price is high, the convenience yield is high as well and hence pushes the
spot price back toward a long-term mean. However this conclusion is derived under the
assumption of risk neutrality.
5
s = T-t.
163
yield and spot price of oil follow a joint diffusion process and the spot price of
crude oil has a lognormal stationary distribution such as the following;
dS
ldt r1 dz1 ;
S
9:1
dd ka ddt r2 dz2
9:2
In the above expression, dz1 dz2 qdt in which q is the correlation coefficient
between these two Brownian motions, namely dz1 and dz2.
The above expression centers on the mean reversion pattern. With the further
assumption that the price of oil contingent claim B(S, d, s) is a two-times
continuously differentiable function of S and d. In order to identify the instant price
change, Gibson and Schwartz (1990) use Its Lemma to give:
dB BS dS Bd dd Bs ds 0:5BSS dS2 0:5Bdd dd2 BSd dSdd;
dB Bs 0:5BSS r21 S2 BSd Sqr1 r2 0:5Bdd r22 BS lS Bd ka ddt
r1 SBS dz1 r2 Bd dz2
9:3
With the elimination of market imperfections and uncertainty of interest rates
the price of this claim must satisfy the partial differential equation:
0:5BSS S2 r21 0:5Bdd r22 BSd Sqr1 r2 BS Sr d
Bd ka d kr2 Bs rB 0
9:4
In the above expression, market price per unit of convenience yield risk is
denoted as k. Given that in perfect market conditions there is no arbitrage possibility, the price of a futures contact F(S, d, s) on one barrel of crude oil deliverable
at time T will satisfy the partial differential equation:
0:5FSS S2 r21 0:5Fdd r22 FSd Sqr1 r2 FS Sr d
Fd ka d kr2 Fs 0
9:5
Therefore the equation indicated above will be subject to the initial condition:
F S; d; 0 S
9:6
However, from the perspective of convenience yield, the reason that the linear
relationship between future and spot prices for crude oil may not hold is pointed
out by Williams and Wright (1991). These authors provide evidence that contrary
to the general view, inventories might be held profitably even without the existence of convenience yield when the future prices are not high enough, compared
to the relative prices to justify the average holding costs. However the loss
incurred by holding costs are likely to be compensated by time differences in
desired quality of crude oil or expected future increases in transport and storage
costs, which are positive capital gains.
164
The other component explaining the relationship between future and spot prices
is the cost of carry, which is the cost associated with holding the commodity until
the delivery date. Such costs include; the opportunity cost of holding the crude oil,
the cost of funding, the cost of storing the crude oil in a tank, and the cost of
insuring the commodity while held. The general definition of the cost of carry is it
is equal to the cost of financing a commodity plus the storage cost minus income
earned on the commodity. Bekiros and Dicks (2007) highlight the role played by
the cost of carry with the following relationship that indicates the relationship
between spot prices and future prices.
F SecyT
9:7
In the above equation cost of carry and convenience yield are represented by
c and y respectively. While T is the time to maturity, and e is Eulers number.
According to this equation, because convenience yield and cost of carry is
constant, an increase in spot prices must in turn increase future prices. Otherwise
arbitrage takes place and hence equilibrium is restored eventually between spot
and future prices.
Finally, in the light of the Kaldor model, Caporale et al. (2010) study future
contracts with different maturities between 1990 and 2008. As shown by Fig. 9.2,
the null of no contribution is strongly rejected for the futures market, which
represents the dominant market in terms of price discovery, with the spot market
acting as a satellite trading venue in the terminology of Hasbrouck (1995). For this
reason, the authors provide evidence on the linkage between future and forward
contracts. However, they document that the price discovery decreases considerably
as the maturity of the futures contract increases.
165
Fig. 9.2 Time-varying price discovery measures. Source (Caporale et al. 2010). Notes Panel A
reports the time-varying spot markets contribution to price discovery (on the left) and the
associated p-value for the test of the null of no contribution of that market to price discovery (on
the right). Panel B reports the time-varying 1-month future markets contribution to price
discovery (on the left) and the associated p-value for the test of the null of no contribution of that
market to price discovery. The dashed line indicates the 10 % level threshold. The horizontal axis
reports daily observations from 2 January 1990 to 31 December 2008, for a total of 4,758 datapoints: observations #1,000, #2,000, #3,000, #4,000 correspond roughly to the end of December
1993, 1997, 2001 and 2005, respectively
results show that a no change model for prices of crude oil, which uses the current
price at time t to forecast for a particular horizon, outperforms a futures-based
model. Alquist and Kilian (2010) base their result on the variability between crude
oil futures prices and spot prices, which is caused by marginal convenience yield.
166
Fig. 9.3 Volatility of crude oil price. Note Volatility is estimated using exponentially weighted
moving average smoothing with the daily decay rate of d = 0.005. An exponentially weighted
2
^2t drt1
^t is the
moving average estimator ofvariance
is defined as; r
1 d^
r2t1 , where r
st1
st2
is the logarithmic return from the close of the day t-2 to the
between the spot and future crude oil market decreases as maturity becomes
longer.
As also emphasized by Ates and Huang (2011), especially after the year 2007,
crude oil prices exhibit high volatility, with monthly price in USD per barrel
increasing from $53.40 in January 2007 to $132.55 in July 2008, yet plummeting
to $41.76 in February 2009. Specifically, Fig. 9.3 plots the exponentially weighted
moving average (EWMA) of the volatility of the daily spot prices of crude oil.
Moreover, Fig. 9.4 displays the spread between spot and future prices of crude oil.
The sample period in both figures is between December 1986 and January 2010.
Both the high increase in volatility of spot prices after the year 2007 shown in
Fig. 9.3 and the escalation of the spread which becomes more emphasized after
2007, underline the importance of incorporating the recent data set into the
analyses.
All datasets are obtained from the webpage of the Energy Information
Administration (EIA) and the WTI price of crude oil available for delivery at
Cushing, OK6 is used. In the analysis, the beginning of the time series is when
crude oil futures were first traded on the NYMEX, namely 1 January 1986. The
data ends on 21 December, 2011, totaling 312 months.
As for the future prices, the price associated with a one-month futures contract
is determined as the value at the end of the month, which is the price at which the
contract was traded closest to the last trading day of the month. This way there is
the ability to match, as closely as possible, future prices with spot prices at the end
of the month. For crude oil, each contract expires on the third business day prior to
the 25th calendar day of the month preceding the delivery month. If the 25th
calendar day of the month is a non-business day, trading ceases on the third
business day prior to the business day preceding the 25th calendar day. After a
It is pointed out by Alquist et al. (2011) that it is possible to obtain qualitatively similar results
by using Brent spot and future prices. Vansteenkiste (2011) also indicates that modeling based on
prices of WTI and those of Brent should not substantially affect the analysis although these two
prices do not fluctuate in line with one another over time.
167
Fig. 9.4 Spread between spot and futures prices of crude oil, $ per barrel
contract expires, Contract 1 for the remainder of that calendar month is the second
following month.
Here the Alquist and Kilian (2010) estimation is extended and updated by the
h
most current data set through a forecast horizon of one month. In the models, Ft
is denoted as the current (at time t) nominal price of the futures contract maturing
in h periods. The current nominal price of oil is St and the expected future spot
price at date t ? h, conditional on information available at t, is Et Sth .
The random walk model, which implies that changes in spot price are unpredictable, establishes the benchmark for forecasts based on future prices. Therefore,
according to Model 1, as specified below, only the current spot price of crude oil is
the best forecast of its spot price.
^
Sthjt St
h 1 month
Model 1
However, according to Model 2, future oil prices are best predicted by oil
futures prices.
h
^
Sthjt Ft
h 1 month
Model 2
According to the Model 3, the spot price of oil can be forecasted through the
spread between the sport price and futures price given that this spread indicates the
probability of the direction of the price of oil. The spread is expected to be an
indicator of the expected change in the spot price if the futures price becomes
h
equal to the expected spot price. This can be obtained .by dividing Ft
h
St . Henceforth, the
168
9.4 Results
Which of the forecasting model is the most accurate may depend on the loss
function of the forecaster.7 The predictive performance of the models is evaluated
by using two approaches. The first approach compares forecasting ability of a
candidate model
to that of the
.random walk without drift model using a percent
b
error, Pthjt Sth S thjt Sth , as a loss function where ^Sthjt is the h-stepahead-forecast. For an overall accuracy measure, the Root Mean Squared Percent
Error (RMSPE) is computed,
v
u T
u1 X
p2th;t
RMSPE t
T
t1
For details see Elliott and Timmermann (2008) and Alquist et al. (2011).
169
St
1
Ft
Success ratio
9.03
0.998
n.a.
0.435
h
. i
1
St 1 ln Ft
St
(0.499)
0.998
(0.989)
0.435
h
. i
1
St 1 ^
St
a ln Ft
(0.499)
1.488
(0.989)
0.500
(0.506)
1.413
(0.714)
0.493
(0.504)
1.359
(0.544)
0.497
(0.505)
(0.516)
5
6
. i
^ Ft1 St
St 1 ^
a bln
. i
^ Ft1 St
St 1 ^
a bln
Notes All Root Mean Squared Percent Error (RMSPE) results are presented as ratios relative to
the benchmark no-change forecast model, for which the actual RMSPE is reported. i.e., Model 1
has RSMPE equal to 9.03 % and Model 2 has RMSPE equal to 0.998 9 9.03 % & 9.01 %. pvalues are reported in parentheses below the respective statistic. The forecast evaluation period is
January 1986December 2011. The initial estimation window for recursive regressions of Models
4 to 6 is January 1986December 1986. All p-values in the RSMPE column refer to pairwise tests
of the null of equal predictive accuracy with the no-change benchmark model and are based on
the DM-test of Diebold and Mariano (1995) using N(0,1) critical values. The success ratio
statistics of Pesaran and Timmermann (1992) in the last column are defined as the fraction of
forecasts that correctly predict the sign of the change in the price of oil. The null hypothesis for
this statistic assumes no association between realized and forecasted direction of changes in spot
prices. This test cannot be applied when there is no variability in the predicted sign. In such a case
the statistic is reported as n.a
specification, Models 2 and 3 are slightly better than the benchmark Model 1 given
that their RMSPE is 0.998 of that of Model 1. However, these tiny improvements
are not statistically significant.
Consequently, it cannot be concluded that a random walk forecast is more
biased than the futures forecast. Moreover, for all models the success ratio is not
statistically significant. The results for the extended and updated period coincide
with those of Alquist and Kilian (2010) since it is found that accuracy of prediction
by futures prices on crude oil is not statistically significantly higher than the simple
no change forecasts.
The second set of results presented in Rows 3 to 6, report on the oil future
spreads as predictors of future spot prices. Here, an alternative analysis is
conducted by using recursively estimated spread regressions to generate forecasts
of oil price. Again, the results show that there is no systematic difference between
the RMSPE of the random walk forecast and that of the spread-based forecasts.
Furthermore, it is observed that when based on the RMSPE, Models 4 to 6 are
actually worse than Models 1 to 3. The p values of both the RMSPE and the
170
Fig. 9.5 Percent errors of forecasts. Notes Percent error = 100 Sth ^
Sthjt Sth , where ^
Sthjt
is the h-step-ahead-forecast. Here, h = 1 month
success ratio are not statistically significant and hence there is no statistically
significant gain in predicting performance of the spread models either.
Altogether, the findings with the updated sample period are broadly consistent
with the empirical results in Alquist and Kilian (2010), in that no evidence is
obtained to indicate that future crude oil prices are more accurate predictors of the
nominal price of crude oil than simple no-change forecasts. Therefore, it is
concluded that in practice, the price of crude oil futures is not the most accurate
predictor of the spot price of crude oil.
To be more specific, Fig. 9.5 shows the time series of percent errors for Models 1
(benchmark), 3 and 5. These three models are representative of others as they are
similar in their predictive performance. The Figure supports the main message of
Table 9.1 in the sense that, visually, none of the models has lower percent errors
than the benchmark model. In addition, this plot gives a historical performance of
forecasts by depicting that the predictive power of all models worsens at times of
high levels of price instability in markets. For example, during the first Gulf War in
the early 1990s, the percent errors of all models rise and exceed almost 50 %
171
particularly in Model 5. Even larger forecast biases occur during the 20072008
financial crisis, when all models systematically fail to predict the plunge in oil prices
and show errors as large as 100 %particularly Model 5.
9.5 Conclusions
The behavior of crude oil prices in the last several decades has attracted wide
attention and created a debate on the predictability of prices. Armed with the
strategic importance of crude oil as a commodity, this chapter centers on the
questions of whether futures markets can be used in the competitive price
discovery in crude oil markets. Sadly, previous studies provide contrasting views
on the unbiasedness of future prices in predicting the spot prices in crude oil
markets.
On the one hand, the survey in this chapter uncovers considerable evidence on
the finding that future prices of crude oil are equal to the spot price of crude oil
plus the cost of carry plus endogenous convenience yield. On the other hand,
through a detailed explanation of the Alquist and Kilian (2010) model, this chapter
also concurs with previous studies documenting that, even with the possibility of
arbitrage, future crude oil prices do not have any predictive power for spot prices.
To clarify this issue from a practical perspective, empirical results on the
relationship between future and spot future prices of crude oil are provided by first
treating the current level of futures prices as the predictor and next by basing the
analysis on the spread. Using an extended and thus updated sample set, the results
reconcile with those of (Alquist and Kilian 2010), who document large and timevarying deviations of crude oil futures prices from the spot price of oil, as
measured by the futures spread. In parallel to this, no reliable evidence is found to
show that oil futures prices significantly lower the RMSPE relative to the
no-change forecast.
Overall this chapter provides an illustration of the forgoing conflicting
discussions on the price dynamics in the crude oil market. By showing theoretical
arguments on this issue and providing the empirical evidence based on current
market data, the chapter points out that futures should be used with caution in
predicting spot crude oil prices.
References
Abosedraa, S., & Baghestani, H. (2004). On the predictive accuracy of crude oil futures prices.
Energy Policy, 32, 13891393.
Alquist, R., & Kilian, L. (2010). What do we learn from the price of crude oil futures? Journal of
Applied Econometrics, 25, 539573.
Alquist, R., Kilian, L., & Vigfusson, R. J. (2011). Forecasting the price of oil. International
finance discussion papers 1022, Board of Governors of the Federal Reserve System (U.S)
172
Ates, A., & Huang, S. C. (2011). The evolving relationship between crude oil and natural gas prices:
Evidence from a dynamic cointegration analysis. Pennsylvania Economic Review, 18(1), 19.
Bekiros, S. D., & Dicks, C. G. H. (2007). The relationship between crude oil spot and future
prices: Cointegration, linear and nonlinear causality (CeNDEF Working Papers, 0711).
Amsterdam: University of Amsterdam.
Bessembinder, H., Coughenour, P., & Smeller, M. (1995). Mean reversion in equilibrium asset
prices: Evidence from the futures term structure. Journal of Finance, 50, 361375.
Bopp, A. E., & Lady, G. M. (1991). A comparison of petroleum futures versus spot prices as
predictors of prices in the future. Energy Economics, 13(4), 274282.
Brennan, M. J. (1991). The price of convenience and the valuation of commodity contingent
claims. In D. Lund & B. ksendal (Eds.), Stochastic models and option values (pp. 3371).
Amsterdam: Elsevier.
Caporale, G., Ciferri, D., & Girardi, A. (2010). Time-varying spot and futures oil price dynamics
(Working Paper), Uxbridge, United Kingdom: Center for Empirical Finance, Brunel
University.
Carlson, M., Khokher, Z., & Titman, S. (2007). Equilibrium exhaustible resource price dynamics.
Journal of Finance, 62, 16631703.
Casassus, J., & Collin-Dufresne, P. (2005). Stochastic convenience yield implied from
commodity futures and interest rates. Journal of Finance, 60, 22832331.
CGES (2011). http://www.cges.co.uk/
Chernenko, S.V., Schwarz, K. B., & Wright, J. (2004). The information content of forward and
futures prices: market expectation and the price of risk. (International Finance Discussion
Paper No: 808). Washington, DC: Board of Governors of the Federal Reserve System.
Chinn, M. D., LeBlanc, M., & Coibion, O. (2005). The predictive content of energy futures: An
update on petroleum, natural gas, heating oil and gasoline. (Working Paper 11033).
Washington, DC: NBER.
Coppola, A. (2007). Forecasting oil price movements: Exploiting the information in the futures
market. Journal of Futures Markets, 28(1), 3456.
Crowder, W., & Hamed, A. (1993). A cointegration test for oil futures market efficiency. Journal
of Futures Markets, 13(8), 933941.
Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business and
Economic Statistics, 13, 253263.
Elliott, G., & Timmermann, A. (2008). Economic forecasting. Journal of Economic Literature,
46, 356.
Gibson, J., & Schwartz, E. (1990). Stochastic convenience yield and the pricing of oil contingent
claims. Journal of Finance, 45, 959976.
Gulan, G. (1998). Efficiency in the crude oil futures market. Journal of Energy, Finance and
Development, 3(1), 1321.
Hamilton, J. D. (2009). Causes and consequences of the oil shock of 200708. Brookings Papers
on Economic Activity, 1, 215261.
Hasbrouck, J. (1995). One security, many markets: Determining the contributions to price
discovery. Journal of Finance, 50, 11751199.
Herbert, J. H. (1993). The relation of monthly spot to futures prices for natural gas. Energy, 18,
11191124.
Huang, B.-N., Yang, M. J., & Hwang, C. W. (2009). The dynamics of a nonlinear relationship
between crude oil spot and futures prices: A multivariate threshold regression approach.
Energy Economics, 31(1), 9198.
Kaldor, N. (1939). Speculation and economic stability. Review of Economic Studies, 7, 127.
Kaufmann, R. K., & Ullman, B. (2009). Oil prices, speculation, and fundamentals: Interpreting
causal relations among spot and futures prices. Energy Economics, 31, 550558.
Moosa, I. A., & Al-Loughani, N. E. (1994). Unbiasedness and time varying risk premia in crude
oil futures markets. Energy Journal, 16(2), 99105.
Pesaran, M. H., & Timmermann, A. (1992). A simple nonparametric test of predictive
performance. Journal of Business and Economic Statistics, 10, 461465.
173
Pindyck, R. S. (1993). The present value model of rational commodity pricing. Economic
Journal, 103, 501530.
Routledge, B., Seppi, D., & Spatt, C. (2000). Equilibrium forward curves for commodities.
Journal of Finance, 55, 12971338.
Vansteenkiste, I. (2011). What is driving oil futures prices? Fundamentals versus speculation
(Working Paper, No. 1371). Frankfurt, Germany: European Central Bank.
Williams, J., & Wright, B. (1991). Storage and commodity markets. Cambridge: Cambridge
University Press.
Working, H. (1949). The theory of price of storage. American Economic Review, 39, 12541262.
Chapter 10
Abstract The study reported in this chapter builds on previous studies of the
extent of decoupling of oil and gas markets and thus the degree of deregulation of
the gas sector in each country. It examines both UK and US oil and gas spot and
futures market data. Spot gas and gas futures data from the respective domestic
markets represent domestic factors and oil prices from global datasets represent
global factors. Cointegration and exogeneity tests indicate that US markets have
achieved a greater degree of decoupling with domestic gas price factors dominating global oil price factors in the determination of the future spot gas price.
Therefore, it can be concluded that whilst progress in liberalization has been made
in both markets, US deregulation policies have been more effective than those in
the UK.
10.1 Introduction
The theory of market liberalization implies that oil and gas prices should decouple
as deregulation of natural gas markets progresses. This chapter increases the scope
of past studies (e.g., the study for the UK by Panagiotidis and Rutledge 2007), by
making a comparison of the US and the UK markets and by examining both spot
J. L. Simpson (&)
School of Economics and Finance, Curtin Business School,
Curtin University, GPO Box U1987 Perth, WA 6845, Australia
e-mail: john.simpson@cbs.curtin.edu.au
175
176
J. L. Simpson
and future oil and gas market variables. Findings depend on whether the data is
lagged or unlagged; level series prices, changes in prices or returns; spot prices or
futures prices or a combination thereof. The study reported in this chapter
examines these combinations of techniques and variables in order to seek clarity in
the relationships.
The central research issues covered (and to be expanded in Sect. 10.3) are
whether or not US and UK gas markets have decoupled, or what is the degree of
decoupling, as an indication of whether or not deregulation policies are working
and therefore whether or not domestic factors dominate global factors in the
deregulation process. The results of the study reported in this chapter demonstrate
that, in any similar investigation in the future, it is deemed important to control for
global factors (in oil price futures) and domestic factors (in gas price futures
markets, as the embodiment of the influence of country specific economic forecasts, and seasonality and storage). Following this introduction (Sect. 10.1) the
chapter is planned to cover a background of theory and literature (Sect. 10.2) and
issues to be tested, a specification of the models and a description of the methodology (Sect. 10.3), a description of the data (Sect. 10.4), a report on the findings
(Sects. 10.5 and 10.6) and a conclusion (Sect. 10.7).
10.2 Background
Liberalization of gas markets implies the removal of the nexus between oil and gas
prices. Macro- and micro-economic reforms encourage gas on gas competition.
Gas and oil prices decouple and welfare advantages accrue to consumers as gas
prices fall. The Law of One Price should apply (as revisited by Asche 2000), where
prices of homogenous goods from different producers and suppliers move together.
Price differentials only indicate differences in transportation costs and quality in
that product. This implies that there should be no connection between oil and gas
prices. The strict assumption for the operation of this law is that the goods are:
global commodities; do not change their nature; that prices are determined by the
free flow of supply and demand; and that the markets are informationally efficient
so that the prices reflect all available information.
This is not the case in reality because of the differences in strength and structure
of oil and gas markets, but for the sake of analysis, this paper will assume that the
commodities oil and gas are global commodities, but that oil markets, being the
dominant source of power globally, are more global in their behavior than gas
markets and that gas markets remain more influenced by domestic factors, such as
seasonality and storage in the countries studied.
Of course, the influence of governments in the countries studied is greater on
the gas markets than on the oil market. It may be that in order to avoid price setting
by governments in gas markets that the prices of gas contracts are linked to oil
prices. It may also be that gas and oil, both being used for power and heating
would imply a natural nexus (assuming that heating equipment and power
10
177
machinery is readily exchangeable for both oil and gas). In that case the decoupling of the gas and oil markets could be caused by a change, for example, in the
price of gas. These are separate issues that could be addressed in future studies.
The fact remains that the liberalization of gas markets means that such linkages,
for any reason, will need to disappear.
Forward markets in any commodity, should reflect market expectations, which
are impacted by domestic economic forecasts as well as seasonality and storage
influences. In the context of a comparison of US and UK gas markets, the central
issue of concern in this chapter is whether or not oil and gas markets have
decoupled and thus whether or not global factors (embodied in spot and futures oil
prices) or domestic factors (embodied in gas futures markets) dominate.
The US and the UK are two important Western economies that have undertaken
positive steps to deregulate their markets. The US gas market is around six times
larger than the UK market in terms of volumes consumed. Deregulation of gas
markets in the US began in 1984 with a separation of natural gas supply from
interstate pipeline transportation, deregulated natural gas production and the
wholesale market, and competition was introduced in interstate pipeline transportation. In the UK in 1986, the British government privatized British Gas and
further reforms required the unbundling of supply and transportation and the
releasing of some gas supplies to competitors. However, this chapter does not go
into an in-depth analysis of the specifics of deregulation policies.
The connection between natural gas and oil markets, the degree of integration
and the corresponding volatility and similarity in volatility of these markets, has
been extensively studied. For example, Krichene (2002), in a supply and demand
model examined world markets for crude oil and natural gas and finds that both
markets became highly volatile following the oil shock of 1973. Ewing et al.
(2002) look at time varying volatility in oil and gas markets across markets and
find that common patterns of volatility emerge that might be of interest to financial
market participants.
Adeleman and Watkins (2005) find a degree of stochastic similarity of movement in oil and gas reserve prices for the period 19822003 in the US using market
transaction data. A study by Regnier (2007) of crude oil, petroleum, and gas prices
over a period from 1945 to 2005 finds that these prices are more volatile than
prices for 95 % of products sold by domestic producers with oil prices showing
greater volatility since the 1973 oil crisis.
In relation to the direct real world connection between oil and gas prices, a
party-to-party gas price bargaining model is expounded and partly proven by
Okogu (2002). Other more recent work, for example, Burger et al. (2008), finds
that long-term gas contracts in Japan and South Korea are linked to crude oil prices
and discuss price formulae using oil indexation for European gas markets. The
Okogu model, for example, posits that one of the principles of gas pricing is to
relate the price of gas to its value in the market for oil as the major competing fuel.
The implication of such a model is that market power by State or privately owned
monopolies can extract rent from consumers of gas when oil is the only other
source of energy.
178
J. L. Simpson
Eng (2006), in debating the price that New Zealand should pay for its natural
gas imports from Australia, alluded to the differences in the Japanese and Chinese
pricing models for Australian natural gas. Both pricing models show the accepted
relationship between oil and gas prices, based on data from such sources as the
AsiaPacific Energy Research Centre and the Institute of Energy Economics of
Japan. Again these export pricing models imply a strong connection between oil
and gas prices.
The evidence on decoupling of oil and gas markets is mixed. Serlitis and
Rangel-Ruiz (2004) explore common features in North American energy markets
in shared trends and cycles between oil and gas markets. The study examined
Henry Hub Gas prices and crude oil prices and finds decoupling of oil and gas
prices as a result of deregulation in the US. Silverstovs et al. (2005) investigate the
degree of integration of natural gas markets in Europe, North America and Japan
in the period early 1990s2004 using a principal components and a cointegration
approach where oil and gas markets interacted. They find high levels of gas market
integration within Europe, between Europe and Japan as well as within the North
American market.
Mazighi (2005) noted that the UKs National Balancing Point (NBP) gas price
was significantly related to oil prices. There is also evidence of a statistically
significant relationship between oil and gas prices and industrial stock prices.
Using regression analysis to test the long-term behavior of the UK BP gas prices
he also finds a relationship between the changes in the volume of manufactured
production. As oil is used as a source of industrial power it follows that there is a
relationship between industrial stock prices as well as alternative energy prices.
Mazighi (2005) finds that more than 80 % of gas price changes in the US market
were not driven by their fundamental values. Other factors such as oil price
changes need to be considered to account for gas price changes. However, Mazighi
(2005) suggests that, in the long term and in accordance with economic theory, the
evolution of prices of natural gas and any other homogenous commodity is guided
by its supply and demand.
Asche (2006) also examines whether or not decoupling of natural gas prices
from prices of other energy commodities (such as oil and electricity) had taken
place in the liberalized UK and in the regulated continental gas markets after the
Interconnector had integrated these markets after 1998. Asche finds that monthly
price data from 1995 to 1998 indicated a highly integrated market where wholesale
demand appeared to be for energy generally, rather than specifically for oil or gas.
By 2003 the UK gas market was highly liberalized, according to Panagiotidis and
Rutledge (2007), who investigated the relationship between UK wholesale gas
prices and Brent oil prices over the period 19962003 to test whether or not orthodox
liberalization theory applied and whether or not oil and gas prices had decoupled.
Using cointegration techniques and tests of exogeneity of oil prices through impulse
response functions, their findings generally do not support the assumption of
decoupling of prices in the relatively highly liberalized UK market. The results may
at least have indicated that progress in deregulation had been made.
10
179
Studies of the connection between oil and gas markets have suffered because of
the use of spot prices only, when a growing body of evidence impresses the need to
take into account gas price expectations embodied in futures prices and thus prices
that capture forecasts of macro-economic data, as well as seasonality and storage
factors. In doing so these studies also demonstrate whether global factors (that is,
oil prices and oil futures prices) or domestic factors dominate (that is, gas futures
prices).
For example, one of the early studies to examine the relationship of monthly
spot to futures prices for natural gas in the US is Herbert (1993). According to this
study, accurate forecasts of spot gas prices could be obtained by regressing the
spot price for a delivery month on the futures contract price for the same month.
Whilst the general conclusion is that the gas market was inefficient, it is clear that
deregulation in US gas markets is having an effect with gas price expectations thus
strongly affecting gas prices.
A later joint study by Herbert and Kreil (1996) finds inefficiency in US gas
markets, which was addressed in part by the establishment of a second futures
market. They note that there is an active unregulated derivatives market in which
options and swaps are traded. Herbert and Kreil feel that the market changes
enable better responses to changes in market conditions, but that there are still
concerns relating to the allocation of pipeline space. Prices for gas and transport
are not transparent and some industry practices impede further progress in liberalization. They acknowledge however, that the US market is large and diverse and
that the regulatory authorities are at least trying to craft rules to improve business
behavior and performance.
Root and Lien (2003) use hedge ratios (which determine the effectiveness of a
hedge) and examine the relationship between futures and spot prices. Model
specification is important and the study investigates the appropriateness of using a
threshold cointegrated model of the natural gas markets as the basis for hedging
and forecasting. They conclude that whilst the threshold model is appropriate for
longer-term futures contracts it does not offer much improvement in hedging or
forecasting efficiency. Modjtahedi and Movassagh (2005) find spot and futures gas
prices are non-stationary with trends due to positive drifts in the random walk
components of the prices. They find that market forecast errors are stationary and
that futures prices are less than expected future spot prices (implying futures prices
are backward dated). They also find that the bias in futures prices is time varying
and that futures have statistically significant market timing ability.
Wong-Parodi et al. (2006) compare the accuracy of forecasts for natural gas
prices as reported by the Energy Information Administrations short-term energy
outlook (STEO) and the futures market for the period 19982004. They find that
on average the Henry Hub is a better predictor of natural gas prices than the STEO.
Economic modelers are also advised to compare the accuracy of their models to
the futures market. Other studies have examined the influence of seasonality and
storage factors.
Mu (2007) examined how weather shocks impact asset price dynamics in the
US natural gas futures market revealing a significant weather effect on the
180
J. L. Simpson
conditional mean and volatility of gas futures returns. Marzo and Zagaglia (2007)
modeled the joint movements of daily returns on one month futures for crude oil
and natural gas using a multivariate GARCH1 with dynamic conditional correlations and elliptical distributions. They find that the conditional correlation between
the futures prices of natural gas and crude oil had risen over the preceding 5 years,
but the correlation was low on average over most of the sample suggesting that
futures markets do not have an established history of pricing natural gas as a
function of oil market developments. Geman and Ohana (2009) remind their
readers that it is central in the theory of storage, that there is a role for inventory in
explaining the shape of the forward curve and spot price volatility in commodity
markets. They find that the negative relationship between price volatility and
inventory is globally significant for crude oil and the negative correlation applies
only during periods of scarcity and increases for natural gas during winter months.
The present study thus compares price and price change relationships in each of
the US and the UK to include the period of the global financial crisis. The issues
covered in this chapter relate to US and UK market comparisons of progress in gas
market decoupling. More specifically;
1. Are unlagged spot oil price changes, the gas futures price changes and gas spot
price change relationships significant and positive?
2. In optimally lagged data, are the level series price relationships significant and
positive?
3. In lagged data are there significant long-term cointegrating relationships in spot
oil prices and spot and gas futures prices?
4. In short-term dynamics, which markets are the significant exogenous forces in
the lagged models?
5. In both long- and short-term relationships, do domestic factors (gas futures
prices) significantly dominate global factors (spot oil and oil futures prices) in
each gas market?
The focus remains as to whether or not natural gas market liberalization policies
and deregulation legislation in the US and the UK are working. If domestic factors
dominate (that is, if gas futures prices dominate global oil and oil futures prices)
great progress has been made in market liberalization.
10
181
For the UK, due to an incomplete dataset, the full period is from 2 June 2003 to
31 May 2010. The relative connectivity between oil, spot gas and gas futures price
changes is examined. The preliminary analysis provides information on the
strength of the relationships in each spot natural gas market.
Based on evidence from the studies of the nexus between oil and gas prices
(e.g., Krichene 2002; Ewing et al. 2002; Okogu 2002; Serlitis and Rangel-Ruiz
2004; Adeleman and Watkins 2005; Mazighi 2005; Asche 2006; Eng 2006;
Regnier 2007; Panagiotidis and Rutledge 2007) and taking into account gas futures
price interaction (e.g., Modjtahedi and Movassagh 2005; Mu 2007; Marzo and
Zagaglia 2007 and Geman and Ohana 2009), the following unlagged and lagged
models are proposed for testing. The unlagged model is as follows:
DPgst at b1t DPost b2t DPgft et
10:1
where:
DPgs t ; DPgf t and DPost are the changes in the spot gas price, the gas futures
price and the spot oil price respectively in each of the country markets.
In the main analysis, optimally lagged level series data are initially examined in
a vector autoregressive model2 (VAR) for each of the US and UK markets. Once it
is ascertained that the variables are I(1) and optimally lagged, a vector error
correction model (VECM) is used in order to confirm cointegration and test
causality. The VECM is a re-parameterized version of the unrestricted VAR and is
appropriate when the variables are I(1) and cointegrated. In the presence of I(1)
variables, but no cointegration, causality would be studied from the VAR model
specified in first differences.
Thus, based on Eq. 10.1, initially specified in level series, the following model
(Eq. 10.2) in functional form is tested. Note, all variables on the right hand side of
the equation are specified in both an unlagged and a lagged form from t 1 to
t n with n as the optimal lag deduced from lag exclusion tests and lag order
information criteria. The endogenous variable is also lagged on the right hand side
of the equation.
Pgst f Post ; Pgf
10:2
A prominent pioneer of research into autoregressive time series processes including exogeneity
and causal ordering is Christopher Sims. The example of related work published is Sims (1977).
182
J. L. Simpson
Movassagh 2005; Wong-Parodi et al. 2006; Mu 2007; Marzo and Zagaglia 2007;
and Geman and Ohana 2009).
Thus, the analysis now moves to Eq. 10.3. Optimally lagged level series data
are again examined in a VECM for each of the US and UK markets in order to
confirm tests of long-term equilibrium relationships and to test for short-run
dynamics and exogeneity. Based on the above-mentioned evidence the following
lagged model in functional form is proposed for testing:
Pgfst f Pgf ; Pof
10:3
where:
Pgfs ; Pgf and Pof represent level series prices of future spot gas, gas futures and
oil futures at times t lagged to t n where n represents the optimal lag based on
information criteria working from lags t 1 to t n.
10
183
The proxy for spot oil prices is the Organization of the Petroleum Exporting
Countries (OPEC) oil prices. The justification for the use of this proxy follows.
The OPEC cartel (cartel means a formal/explicit agreement among competing
firms) consists of net oil-exporting countries primarily made up of Algeria,
Angola, Ecuador, Iran, Iraq, Kuwait, Libya, Nigeria, Qatar, Saudi Arabia, the
United Arab Emirates, and Venezuela at the time of writing. The cartel has
maintained its headquarters in Vienna since 1965 and hosts regular meetings
among the oil ministers of its Member Countries. Indonesia withdrew in 2008 after
it became a net importer of oil, but stated it would likely return if it became a net
exporter in the world again. The OPEC prices are the logical drivers of all other oil
prices, because of the market power of the group of net oil-exporting countries that
collectively control the exports of around 40 % of the worlds oil requirements.
These oil prices, though formed through the market power induced by cartel
behavior, are considered the major driver of and therefore a suitable proxy for the
global crude oil price.
The proxy for the UK gas futures price is the ICE London or UK natural gas
futures prices for six months. It is a Reuters continuation series, which gives the
data for 6 months forward. The proxy for the US gas futures price is the NYMEX
natural gas futures prices for six months, which is a six month forward rate. It
starts at the sixth nearest contract month, which forms the first values for the
continuous series until the first business day of the nearest contract month when, at
this point, the next contract month is taken.
These gas futures prices are considered representative of UK and US gas futures
markets respectively. The gas futures prices selected are a proxy for gas market
price expectations as the data embody various global and country economic
forecast data and information that impact domestic price expectations such as,
exchange rates, inflation, interest rates, growth rates and also storage factors and
seasonal effects on gas demand in the US and in Europe and the UK.
In order to test Eq. 10.3, the following data are used. Daily price data for
4 February 200330 November 2009 are obtained from the DataStream database
for all of the variables in the models for each country market. The spot and futures
prices are described above, but the newly specified endogenous variable is the
future spot gas prices for both the US and the UK. To obtain the future spot gas
prices, spot gas prices (HH and NBP gas prices) are brought forward by six
calendar months to coincide with the commencements of the sample period for
both gas futures and oil futures. Specifically spot gas price data from 4 February
2003 to 4 August 2003 are deleted from the data, thus removing six calendar
months of data from the spot gas price data sets so that spot gas price data may be
brought forward by the above period to coincide with the commencement of the
gas futures price data on the 4 February 2003. The full dataset of prices used to test
Eq. 10.3 now run from 4 February 2003 to 30 November 2009.
The proxy for the global oil futures prices applicable to the US and the UK are
provided in the NYMEX light Sweet Crude Oil futures index and the Brent Crude
Oil futures index. The price calculation method for each index is near month
change at the beginning of the first of the month. Each is a continuous series and a
184
J. L. Simpson
perpetual series of oil futures prices starting at the nearest contract month until
either the contract reaches its expiry date or until the first business day of the actual
contract month. At this point the next contract month is taken.
Prior to reporting the findings of this study the issues addressed in this chapter
as stated previously in Sect. 10.2 are repeated here for the convenience of readers.
1. Are unlagged spot oil price changes, the gas futures price changes and gas spot
price change relationships significant and positive?
2. In optimally lagged data, are the level series price relationships significant and
positive?
3. In lagged data are there significant long-term cointegrating relationships in spot
oil prices and spot and gas futures prices?
4. In short-term dynamics, which markets are the significant exogenous forces in
the lagged models?
5. In both long- and short-term relationships, do domestic factors (gas futures
prices) significantly dominate global factors (spot oil and oil futures prices) in
each gas market?
Inflation rate and exchange rate differences between the US and the UK are not assumed to be
major influences over the full period of this study. Thus it is assumed that a reasonable degree of
purchasing power parity and interest rate parity exists between the US and the UK over the full
period of the study.
10
185
OILOPEC
160
140
120
100
80
60
40
20
0
01
02
03
04
05
06
07
08
09
Fig. 10.1 Oil prices. Note OILOPEC is the OPEC oil price index sourced from DataStream.
Prices on the vertical axis are quoted in US Dollars per barrel. The vertical axis denotes years
160
140
120
100
80
60
40
20
2003
2004
2005
2006
OFUK
2007
2008
2009
OFUS
storage effects in each country. These graphs provide initial evidence that the gas
futures markets in both countries do not track global economic indicators to the
extent that oil prices do and their respective peaks and troughs indicate that
domestic macro-economic and other factors in seasonality and storage are the
major influences on these prices.
With regard to Eq. 10.1, the findings of the analysis of unlagged differenced
data are reported as preliminary findings as follows: preliminary analysis of all
186
J. L. Simpson
WORLDS
1.800
1.600
1.400
1.200
1.000
800
600
2003
2004
2005
2006
2007
2008
2009
2004
2005
2006
2007
2008
2009
unlagged level series for both the US and the UK and in all periods of the study
indicates skewness and kurtosis problems, which indicate lack of normality and
uniformity (tests from Jarque and Bera 1987).
This is a violation of the assumptions of ordinary least squares (OLS) regressions and indicates, in turn, problems in serial correlation of the errors for each of
the regression relationships. First differencing removes the problems of serial
correlation in the errors of the first difference relationships (according to DW tests
(Durbin and Watson 1971), but White tests indicate that heteroskedasticity problems remain in the errors thus indicating model misspecification. An autoregressive conditional heteroskedasticity (ARCH) model is deemed more suitable for
analysis. The results of this analysis are reported in Table 10.1.
10
187
GFUK
12.000
10.000
8.000
6.000
4.000
2.000
0
2003
2004
2005
2006
2007
2008
2009
Durbin
Watson
statistic
Variance
equation
coefficient
ARCH/GARCH
0.0366***
10.4862***/
10.8608***
0.3572***
2.0778***
0.0039***
1.1086/
4.9908***
5.2515
2.2574***
Note Significance levels for spot oil z statistics and ARCH/GARCH terms are at 1 %, no asterisk
means no significance
***
denotes significant at 1 % level
Serial correlation problems are not evident and thus do not detract from the
reliability of the models parameters, as indicated by DW statistics. The z statistics
for each of spot oil and gas futures in the US are positive and statistically
significant at the 1 % level. In the UK, only the gas futures variable is significant.
In unlagged data, it is evident that the US gas market has not decoupled with the
oil market. However, the z statistics for US gas futures are also significant at the
1 % level. This indicates that economic forecasts, seasonality and storage factors
embodied in US gas futures price changes are important in US gas markets. This
evidence indicates that, though the nexus between gas and oil prices has not been
broken, sound progress has been made in deregulation of US gas markets as seen
in the connection between gas futures price changes and spot gas price changes.
Global as well as domestic factors appear to be at work in US gas markets when
unlagged data are considered.
188
Table 10.2 Unit root tests
Variable
Equation 10.1: US spot gas
Spot oil
US gas futures
US spot gas errors
Equation 10.1: UK spot gas
Spot oil
UK gas futures
UK spot gas errors
J. L. Simpson
t statistic
(level series prices)
t statistic
(first differences)
-3.2471**
-1.3347
-1.8914
-2.9969**
-4.6458***
-1.4947
-2.0604
-5.0251***
-30.8007***
-25.2201***
-50.5273***
-30.5471***
-23.7836***
-22.1242***
-40.8704***
-23.6352***
**
at 5 % and at
***
1%
10
189
Table 10.3 Optimal lags and cointegration and causality test results
Model
Number of cointegrating
Optimal
Granger causality
relationships according to Trace lag order (Chi square statistic)
and Eigenvalue tests
in days
Equation 2: 1**
US spot
gas
4***
Equation 2: 1**
UK spot
gas
3***
Note The Johansen cointegration tests take the assumption that there is a linear deterministic
trend in the data. Optimal lags are decided based on the majority significance of the Likelihood
Ratio, the Final Prediction Error, and the Akaike, Schwarz and Hannan-Quinn information criteria. The number of cointegrating equations is based on both maximum eigenvalues and trace
statistics. In the number of cointegrating relationships, no asterisk means no significance.
*
denotes significance at the 10 % level, ** at the 5 % level and *** at the 1 % level. The
causality tests show similar levels of Chi Square statistical significance
The findings are that the models are stable, with no root lying outside the relative
unit root circle. The results for the lag order and the cointegration and causality
tests are shown in Table 10.3.
In both markets in the long term there is evidence of cointegration over the full
periods of the studies. This represents evidence that whilst short-term causal
relationships may show decoupling evidence (where there appears no causal
relationship between spot oil and spot gas), there remains a long-term cointegrating relationship between oil and gas prices and gas futures prices, whereby
these variables move in a similar way and come together to stability. This
represents evidence that in the long term the markets have not fully decoupled and
that in both markets deregulation policies still have some distance to go in
achieving full gas market liberalization. However, in the US it is evident when
lagged data are considered that domestic factors in gas futures prices have a
greater impact on spot gas prices than in the UK.
There is also evidence within the models, of causality between gas futures and
spot oil for both markets. Overall, in the short-term there is evidence of decoupling
in both markets, but the nexus between oil and gas futures prices is not fully
broken.
190
J. L. Simpson
Table 10.4 Unit root tests for variables in future spot gas price and price change relationships
Variable
t statistic
t statistic
(level series prices)
(first differences)
US future spot gas:
US Oil futures
US gas futures
Errors US Future Spot Gas
UK future spot gas
UK gas futures
UK Oil futures
Errors UK Future Spot Gas
-2.500
-1.603
-2.080
-2.764*
-4.6334***
-1.975
-1.410
-5.301***
-40.037***
-44.428***
-42.147***
-40.848***
-22.659***
-38.839***
-44.611***
-22.583***
Note No asterisk means non-significance. ADF test results are shown. * Significance levels are at
10 %, *** 1 %. The critical values for the ADF unit root tests are at 1 % (-3.434), at 5 % (2.863) and 10 % (-2.568) levels of significance
With regard to the testing of Eq. 10.3 the following results are reported. It is
recalled that Eq. 10.3 treats the future spot gas price endogenously against the gas
futures price and the oil futures price. The level series of each price index, the first
differenced series and the respective errors of these relationships for each country
and each period under study are tested for unit roots. The results of these tests are
shown in Table 10.4.
Table 10.4 results indicates, that the level series and errors of the level series
relationships are non-stationary processes and the first differenced series and the
errors of the first differenced relationships are each stationary at the 1 % level of
significance.
This enables a conclusion that in each case, for each country, the processes are
integrated and non-stationary and this in turn enables a move to the second part of
the main analysis. That is, to apply all level series to a VAR. As with the testing of
Eq. 10.2, when Eq. 10.3 is considered it is noted that, if I(1) variables are found to
be cointegrated the VAR is re-specified into a VECM and tests run to confirm
cointegration and to test causality.
Again lag order tests are conducted to ascribe an optimal lag for cointegration
and causality tests. The models are initially tested for stability over the full period
of the study for both US and UK markets using stability condition tests. The
findings are that the Eq. 10.3 models are also stable, with no root lying outside the
relative unit root circle. The results for the lag order, cointegration and causality
tests are shown in Table 10.5.
On testing of Eq. 10.3 the findings are that there are no significant long-term
equilibrium (cointegrating) relationships between the variables in the US model.
This represents evidence of decoupling of oil and gas. When the future spot gas
price is treated endogenously, as specified, there is no evidence of dual or one-way
causality running between that variable and US gas futures and oil futures. This
too represents evidence of decoupling of oil and gas. Within the model, when the
US gas futures variable is treated endogenously, causality runs from future spot
gas US to gas futures US at the 10 % level of significance. In the latter case it is
10
191
Table 10.5 Results: Optimal lags and cointegration and causality tests for future spot gas price
relationships
Model Number of cointegrating
Optimal
Granger causality
relationships according to Trace lag order
and Eigenvalue tests
in days
US
UK
Note The future spot gas prices in the US and the UK respectively are treated endogenously. The
Johansen cointegration tests take the assumption that there is a linear deterministic trend in
the data. Optimal lags are decided based on the majority significance of the Likelihood Ratio, the
Final Prediction Error, the Akaike, Schwarz and Hannan-Quinn information criteria. The number
of cointegrating equations is based on both maximum eigenvalues and trace statistics. In the
number of cointegrating relationships, no asterisk means no significance. The causality tests show
similar levels of Chi Square statistical significance
evident that when the future spot gas US price changes the gas futures US price
changes in the same direction within two days. This too represents evidence of
decoupling of oil and gas in US markets.
When the future spot gas UK variable is treated endogenously, there is a longterm equilibrium (cointegrating) relationship between the variables in the model.
192
J. L. Simpson
This represents evidence that in the UK, oil and gas markets have not decoupled.
In addition, dual causality exists between oil futures and the future spot gas price
UK, but the greater causal influence is from the future spot gas price UK to the oil
futures price with Chi Square values of 8.019 compared to 5.669. Significance
levels are at 10 %. This too represents evidence that in the UK, oil and gas markets
have not fully decoupled.
Within this model, when the gas futures UK variable is treated endogenously,
the future spot gas price UK causes gas futures UK at the 5 % level of significance.
This represents evidence of decoupling and the interplay of domestic factors in the
determination of future spot gas prices, however, dual Granger causality exists
between oil futures UK and gas futures UK with the stronger Granger causality
running from oil futures to gas futures UK at the 1 % significance level (with
Chi Square values at 17.981 compared to 15.134). This again does not represent
evidence of full decoupling of the UK future spot gas price with global factors
(oil futures) influencing domestic future gas prices.
10.7 Conclusion
The theoretical base for this chapter lies in the re-visitation of market liberalization
theory by Asche (2000), but focuses on updating and expanding past studies. For
example, a UK cointegration and causality study by Panagiotidis and Rutledge
(2007). Just as mixed evidence is produced by the authors reviewed, mixed
evidence is provided in the study in this chapter. The findings depend on whether
the data are lagged or unlagged, level or differenced, inclusive of spot prices and
gas futures prices or of future spot gas prices, and oil futures and gas futures prices.
The chapter commences with the assumption that there exists a relationship
between oil and gas prices as implied, for example, by Okogu (2002), Burger,
Graeber and Schindlmayr (2008) and Eng (2006). Mazhigi (2005) finds the UK gas
price is significantly related to oil prices. Researchers such as Krichene (2002),
Ewing et al. (2002), Adeleman and Watkins (2005) and Regnier (2007) find that
oil and gas markets possess similar trends in stochastic volatility.
The importance of gas futures in their embodiment of price expectations based
on economic, seasonal and storage information is put forward by researchers such
as Herbert (1993), Herbert and Kreil (1996), Modjtahedi and Movassagh (2005),
Mu (2007), Marzo and Zagaglia (2007) and Geman and Ohana (2009). Serlitis and
Rangel-Ruiz (2004), find decoupling of oil and gas prices as a result of deregulation in the US. Siliverstovs et al. (2005) find high levels of gas market integration
within Europe and North America. Asche (2006) finds that monthly price data
from 1995 to 1998 in the UK indicate a highly integrated gas market. The findings
of Panagiotidis and Rutledge (2007) generally do not support the assumption of
decoupling of prices in the relatively highly liberalized UK market, but imply that
progress is being made in deregulation policies.
10
193
The results of the testing of unlagged data in the study for this chapter indicates
that, though the nexus between gas and oil prices has not been broken (oil price
changes also are a significant determinant), sound progress has been made in
deregulation of US gas markets as seen in the connection of gas futures price
changes to spot gas price changes. Global as well as domestic factors initially
appear to be at work in US gas markets. In the UK, at least in unlagged data, the
nexus between oil and gas appears to have been broken and gas price expectations
embodied in gas futures price changes are significantly related to UK spot gas over
the full period of the study. Domestic factors are also more important in UK gas
markets in unlagged data.
However, the analysis of unlagged data does not provide a clear picture of longterm equilibrium relationships and short-term dynamics. When lagged data are
examined, evidence is produced that, in the long term, the markets have not fully
decoupled (cointegrating relationships exist in both US and UK markets) and that
in both markets deregulation policies still have some distance to go in achieving
full gas market liberalization. However, in the US it is evident, that domestic
factors in gas futures prices have a greater impact on spot gas prices than in the
UK. Therefore, more clarity is required on oil and gas price relationships.
Consequently, the study finally takes into account lagged relationships between
future spot gas prices, oil futures prices and gas futures prices and the results have
achieved greater clarity. Also tested is whether or not gas futures prices are a good
predictor of future gas prices. Whilst there is no strong evidence for this, it is
evident through cointegration and exogeneity testing, that US gas markets have
largely decoupled and that domestic forces in gas price expectations (in other
words US macroeconomic forecasts, seasonality and storage factors), play a major
role in the relationship with the future spot gas price. There is again no strong
evidence that UK futures prices are good predictors of future spot prices. In the UK
cointegration and exogeneity testing show that the nexus between oil and gas
futures markets has not been fully severed. Thus global factors in the oil futures
market have much to do with the maintenance of that connection.
References
Adeleman, M. A., & Watkins, G. C. (2005). US Oil and natural gas reserve prices, 19822003.
Energy Economics, 27, 553571.
Asche, F. (2000). European market integration for gas? Volume flexibility and political risk
(Working Paper No. 358). Munich, Germany: CESinfo
Asche, F. (2006). The UK Market for Natural Gas, Oil and Electricity: Are the Prices Decoupled?
The Energy Journal, 27(2), 2740.
Burger, M., Graeber, M., & Schindlmayr, G. (2008). Managing energy risk: An integrated view
on power and other energy markets. Chichester: Wiley.
Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series
within a unit root. Econometrica, 49, 10221057.
Durbin, J., & Watson, G. S. (1971). Testing for serial correlation in least squares regression.III.
Biometrika, 58, 119.
194
J. L. Simpson
Eng, G. (2006). A formula for LNG pricing, Ministry for Economic Development New Zealand.
Retrieved from http://www.med.govt.nz/Templates/multipageDocumentTOC___23939.aspx.
Ewing, B. T., Malik, F., & Ozfidan, O. (2002). Volatility transmission in the oil and natural gas
markets. Energy Economics, 24, 525538.
Geman, H., & Ohana, S. (2009). Forward curves, scarcity and price volatility in oil and natural
gas markets. Energy Economics, 31, 576585.
Granger, C. W. J. (1988). Some recent developments in a concept of causality. Journal of
Econometrics, 39, 199211.
Herbert, J. H. (1993). The relation of monthly spot to futures prices for natural gas. Energy, 18,
11191124.
Herbert, J. H., & Kreil, E. (1996). US natural gas markets: How efficient are they? Energy Policy,
24, 15.
Jarque, C. M., & Bera, A. K. (1987). A test for normality of observations and regression residuals.
International Statistical Review, 55, 163172.
Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics
& Control, 12, 231254.
Krichene, N. (2002). World crude oil and natural gas: A demand and supply model. Energy
Economics, 24, 557576.
Marzo, M., & Zagaglia, P. (2007). A note on the conditional correlation between energy prices:
Evidence from future markets. Energy Economics, 30, 24542458.
Mazighi, A. E. H. (2005). Henry Hub and national balancing point prices: What will be the
international gas price reference? OPEC Review, 29, 219230.
Modjtahedi, B., & Movassagh, N. (2005). Natural gas futures: Bias, predictive performance, and
the theory of storage. Energy Economics, 27, 617637.
Mu, X. (2007). Weather, storage and natural gas price dynamics: Fundamentals and volatility.
Energy Economics, 29, 4663.
Okogu, B. E. (2002). Issues in global natural gas: A primer and analysis. (Working Paper No.
40). Washington, DC: IMF.
Panagiotidis, T., & Rutledge, E. (2007). Oil and gas markets in the UK: Evidence from a
cointegration approach. Energy Economics, 29, 329347.
Regnier, E. (2007). Oil and energy price volatility. Energy Economics, 29, 405427.
Root, T. H., & Lien, D. (2003). Can modeling the natural gas futures market as a threshold
cointegrated system improve hedging and forecasting performance? International Review of
Financial Analysis, 12, 117133.
Serlitis, A., & Rangel-Ruiz, R. (2004). Testing for common features in North American energy
markets. Energy Economics, 26, 401414.
Silverstovs, B., LHegaret, G., Neumann, A., & von Hirschhausen, C. (2005). International
market integration for natural gas? A cointegration analysis of prices in Europe, North
America and Japan. Energy Economics, 27, 603615.
Sims, C. A. (1977). Exogeneity and causal ordering in macroeconomic models. In C. A. Sims
(Ed.), New methods in business cycle research (pp. 2343). Federal Reserve Bank of
Minneapolis: Minneapolis.
Wong-Parodi, G., Dale, L., & Lekov, A. (2006). Comparing price forecast accuracy of natural gas
models and futures markets. Energy Policy, 34, 41154122.
Part IV
Chapter 11
Abstract The work of Markowitz in the early 1950s triggered a revolution in the
investment management world. The concept of efficient portfolios and efficient
frontier gave an important impulse to the development of modern finance. Ever
since, the concept of efficient portfolios has been widely applied in many environments. While originally restricted to stock markets, applications have been
developed in the field of e.g. the optimisation of energy distribution (Letzelter
2005). In the last decade, asset managers look at the opportunity to improve their
expected return-risk trade off by adding commodities to their portfolio of stocks
and bonds. In this chapter we look at the contribution of oil to such a portfolio.
The goal of this paper is to investigate if the addition of oil to an investment
portfolio can improve an efficient set of traditional investments in stocks and bonds.
We believe that given the counter cyclicality of oil returns compared to the stock
market, that the inclusion of such assets should improve the risk-return trade-off.
It appears that oil is not a safe haven for stockholders and bondholders. Oil is not a
hedge for stockholders, but it does present a hedge for bondholders. When adding
oil to the portfolio we see a change in efficient frontier and market portfolio.
Holders of portfolios of bonds and stocks can improve their risk-return trade off by
enlarging their portfolio with an investment in oil.
Keywords Safe haven
A. Dorsman (&)
VU University Amsterdam, Amsterdam, The Netherlands
e-mail: a.b.dorsman@vu.nl
A. Koch M. Jager
Stachanov, Amsterdam, The Netherlands
A. Thibeault
Vlerick, Management School, Ghent, Belgium
197
198
A. Dorsman et al.
11.1 Introduction
Mandelbrot (1963, 1966) showed that stocks are neither normally nor log-normally
distributed. The condition of (log-) normality has become more and more
restrictive. The 2007/2008 global financial crisis and the stock price developments
after the 2011 earthquake and tsunami in Japan show that outliers are more frequent than one might expect under the (log-) normal condition. In other words,
there is obesity in the tail. It is also possible that an asset is not or negatively
correlated with another asset whenever market developments are difficult. Holders
of the second asset see, in that case, the first asset as a safe haven. There is some
evidence (Baur and Lucey 2009; Baur and Dermott 2010) that gold is a safe haven
for some stockholders, but not for bondholders.
The goal of this chapter is to investigate whether the addition of oil to a
traditional portfolio of stocks and bonds can improve the risk-return trade-off.
Adding oil to a portfolio of stocks and bonds can be interesting for portfolio
holders if it improves the expected risk-return trade-off.
Thus, our research investigates the impact of adding oil to a portfolio of bonds
and stocks. We are more specifically interested to look at oil as a safe haven, at oil
as a hedge for stockholders and bondholders and at the impact of oil on the
efficient frontier of a portfolio made of stocks, bonds, and oil. To conduct the
empirical tests, we use indices made of risk free U.S. government bonds, of
common stocks from Standard and Poor and of an oil index.
The statistical analysis is performed with the use of Oracle Crystal Ball software. The software is used to characterise the distribution of returns for our three
indices, to estimate the correlations between these indices, and to derive the
efficient frontier for our portfolios. In our analysis, we also consider the consequences of obesity in the tails.
The structure of this chapter is as follows. We start with a review overview in
Sect. 11.2. In Sect. 11.3 the data is described. The empirical results are presented
in Sect. 11.4. Section 11.5 contains a short summary and our conclusions.
11
199
portfolios. However, fat tails, tail dependency and micro-correlations reduce the
effect of diversification in an insurance portfolio.
For over a decade now, institutional investors try to reduce risk by diversifying
their portfolios with commodities. For example, gold can be a safe haven asset to
this purpose.1 Baur and Lucey (2009) make a distinction between a hedge, a
diversifier and a safe haven asset. A hedge is an asset that is uncorrelated or even
negatively correlated with another asset or portfolio. A diversifier is an asset that is
positively (but, not perfectly) correlated with another asset or portfolio, and a safe
haven is an asset that is uncorrelated or negatively correlated with another asset or
portfolio in times of market stress and turmoil. In the case of a safe haven,
correlations are different at times of large price falls on stock and bond markets.
Baur and Lucey (2009) analyze the role of gold in combination with stocks and
bonds for the markets of the US, the UK and Germany. They find evidence that
gold is a safe haven for stockholders, but not for bondholders. Baur and McDermott (2010) enlarged the study of Baur and Lucey (2009) to include other markets.
However, they only looked at the relationship between gold and stocks, and not to
the interactions between gold and bonds. They find that gold is a safe haven for the
well-developed European countries and the US, but not for Japan, Australia,
Canada and in the countries of Brazil, Russia, India and China (BRIC group),
which are all deemed to be at a similar stage of newly advanced economic
development. The acronym has come into widespread use as a symbol of the shift
in global economic power away from the developed G7 economies towards the
developing world.2 Also oil can be seen as a safe haven. In times of a substantial
price decreases on the stock markets not only gold, but also oil may increase in
price.
Arouri and Nguyen (2010) examined oilstock market relationships over the
last turbulent decade. Steering clear previous empirical investigations, which have
largely focused on broad-based national and regional market indices, they investigate short-term linkages on an aggregate level as well as on the sector by sector
level in Europe. Their main finding is that the responses of stock returns to oil
price changes differ greatly depending on the industry.
Geman and Karroubi (2008) look at the diversification effect brought by crude
oil futures contracts into a portfolio of stocks. They prefer oil futures because it is
the most liquid of commodity futures. However, introducing futures into the
database introduces new problems. Firstly, the maturity of a future is limited. At a
Another commodity is real estate. For example Chua (1999) studied the role of international
real estate in a mixed-asset portfolio while attempting to control for higher taxes, transaction
costs and asset management fees incurred when investing in real estate, as well as the appraisal
smoothing in real estate return indices.
2
In 2005 Goldman Sachs defined The Next Eleven (or N-11). The N-11 are eleven countries
Bangladesh, Egypt, Indonesia, Iran, Mexico, Nigeria, Pakistan, Philippines, South Korea, Turkey
and Vietnamidentified by Goldman Sachs investment bank as having a high potential of
becoming, along with the BRICs, the worlds largest economies in the twenty first century.
ONeill (2001)
200
A. Dorsman et al.
certain moment, one has to switch to a future with a longer maturity. Their finding
is that, in the case of distant maturities futures (e.g., 18 months), the negative
correlation effect is more pronounced regardless whether stock prices increase or
decrease. This property has the merit to avoid the hurdles of a frequent roll-over
while being quite desirable in the current trendless equity markets.
Our bond and stock indices are corrected to include interest and dividend
payments. All data are priced in dollars. Therefore we have no currency problem.
There are several types of crude oil. For example, light, sweet crude is of
greater use in production of gasoline, naphtha, propane and butane. Heavy sour
crude is used mainly to produce heavy heating oil, asphalt and bitumen.
Accordingly, the different types of crude require their own specific refineries and
refining processes. Heavy sour crude needs more refinery processing than does the
lightest and sweetest form, meaning that people are unwilling to pay as much for
heavy sour crude as for light sweet crude. As a result, prices of the various types of
oil differ. For that reason the price of West Texas Intermediate (WTI) crude will
most of the time differ from, for instance, Dubai crude. It is understood that this
difference does not remain steady over time, but varies as a result of many factors
such as available refining capacity and reserves. Our choice to select the WTI Oil
is arbitrary. Therefore, at times we will also check with the Brent Crude Oil.
Most oil contracts are bilateral between demanders and suppliers. Only a
limited part is traded on the spot market. Therefore the price on the spot market is
not a good indicator of the real oil price. As an alternative we use the WTI Oil
futures prices. However, by using futures we are also introducing the problems of
11
201
202
A. Dorsman et al.
graph of the bond, Fig. 11.1a, shows an upward trend for bonds caused by a global
decreasing of the interest rate. During the period 19892010 the development of
stocks (Fig. 11.1b) and oil (Fig. 11.1c) differ from the development of bonds. The
graph of the stocks shows the effect of the internet crisis in 20012002 and the
financial crisis in 20072008. The internet crisis in 20012002 had no influence on
the price of the oil, but the financial crisis in 20072008 had. The time series
presented in Fig. 11.1a (bonds), b (stocks) and c (oil) show substantial volatility.
By comparing these figures we see sometimes a co-movement (financial crisis) in
stocks and oil) and sometimes an independent development of the stock price and
the oil price. Our general conclusion is that specific cyclical or counter cyclical
patterns cannot be identified and neither can the potential for diversification.
However, the individual evolution of each time series can be explained by market
developments. Perusal of the price development of stocks (Fig. 11.1b) reveals large
prices decrease in 2000/2002 and again in 2007/2008. These periods refer to the
internet crisis and the financial crisis respectively. During the internet crisis, oil
prices did not move significantly, while during the financial crisis the oil price
dropped dramatically. The bond prices in Fig. 11.1a exhibit a steady growth during
the observed period. Only during the financial crisis the bond prices got hit substantially, but not as severely as the stock prices and the oil prices. As previously
inferred, these figures lead to a mixed interpretation. It is not clear from inspection of
these graphs, whether oil is a safe haven or a hedge for stock and/or bondholders.
The returns on oil prices, stock prices and bond prices are not normally distributed, because in all cases the minimum values of oil, stocks and bonds are zero.
Therefore we usein line with other researchersthe log-returns instead of
normal returns to get distributions that are not capped. Table 11.1 contains a
summary of the descriptive analysis of the log-returns on oil prices, stock prices
and bond prices. In Appendix we give some of these statistics for each of the years
of the observed periods.
Table 11.1 shows that during the observed period the mean values of bonds,
stocks and oil do not differ much. However, the standard deviation of bonds is
substantially lower than the standard deviation of stocks, which is again substantially
lower than the standard deviation of oil. This ranking makes sense if one considers
the relative riskiness of each instrument. The Hill estimator is used to determine
whether or not there is obesity in the tails. The Hill estimator estimates the aparameter of a Pareto distribution (see Kousky and Cooke 2009). Based on the logreturn distributions, it becomes clear that neither stocks nor oil exhibit fat tails.
Resnick (2007) finds that only when the log-returns are Pareto-distributed the Hillestimator works well. In the cases of other distributions, Resnick finds that the results
of the Hill-estimator are unstable. We find a Hill-estimator of a for bonds, stock and
oil of 3.6, 3.2 and 2.7 respectively, which indicates that there is obesity in the tails.
The bond index shows a distribution close to the normal distribution. The
kurtosis of the bond distribution is with 2.26 relatively low and the skewness, 0.20, differs not much from zero. Also for the stock index we see a small skewness,
-0.23. However, the stock index exhibits a significant kurtosis of 9.46, more than
three times that of a normal distribution. The departure from normality is even
11
(a)
203
2000
1800
Index bonds
1600
1400
1200
1000
800
600
400
200
0
12/31/88
12/31/92
12/31/96
12/31/00
12/31/04
12/31/08
12/31/04
12/31/08
12/31/04
12/31/08
Date
(b)
3000
Index stocks
2500
2000
1500
1000
500
0
12/31/88
12/31/92
12/31/96
12/31/00
Date
(c)
160
140
Index oil
120
100
80
60
40
20
0
12/31/88
12/31/92
12/31/96
12/31/00
Date
Fig. 11.1 a The price development of bonds during the period 19892010. b The price
development of stocks during the period 19892010. c The price development of oil during the
period 19892010
204
A. Dorsman et al.
Table 11.1 Descriptive statistics of the log returns of oil, stocks and bonds
Parameter
Bond
Stock
Oil
Number of observations
Mean (%)
Standard error of mean (%)
Median (%)
Standard deviation (%)
Variance (%)
Skewness
Kurtosis
Jarque-Bera
Hill estimator
Sum (%)
Minimum (%)
Maximum (%)
5741
0.029
0.032
0.000
2.448
0.060
-0.94
17.23
71822.21
2.7
166.779
-40.048
16.410
5741
0.032
0.008
0.039
0.581
0.003
-0.20
2.26
1255.81
3.6
181.891
-3.173
3.810
5741
0.035
0.015
0.033
1.138
0.013
-0.23
9.46
21462.71
3.2
199.310
-9.460
10.958
more significant for the oil index with a skewness of -0.94 and a kurtosis of 17.23.
These departures from normality necessitate a further investigation for the
appropriate distribution. We conclude that the bond distribution does not deviate
much from the normal distribution, while the oil price cannot be described by a
normal distribution. The stock index is somewhere in between.
In order to determine the efficient frontier and the optimal portfolio composition, conventional portfolio theory has been applied. The data have not been
modeled with a probability density function (PDF), but the real data are used in the
model. An analysis of the data shows that neither the stock, nor the bonds, and oil
returns contained fat tails. The as which have been determined for the various
asset classes are well and above two, which indicates that there are no undefined
first and second statistical moments. This proves that both the mean and the
variance of the data are defined. For real fat tails these moments are not defined
and the statistics change as more data are added. In case a would have been
smaller than two, no stable variance or standard variation would be available. The
consequence of an undefined variance would be that the portfolio does not apply.
Diversification and portfolio theory are based on the idea taking advantage of the
difference in variance. Fat tails with undefined second moments (variance) smash
the foundation on which the portfolio theory is built.
Using Oracle Crystal Ball software we learn that the log-distributions are not
normal, but follow a student t-distribution. In the Markowitz portfolio theory the
standard deviation of a portfolio is a function of the standard deviation, weights
and correlations of the components of the portfolio. Also, in the case of a student tdistribution, this equation holds. In Table 11.2 we present the correlations between
oil, stocks and bonds.
Perusal of the results for the entire period 19892010 reveals that the correlation between bonds and stocks is negative (-0.101). Also the correlation
between bonds and oil is negative and even smaller. However, during the whole
period the correlation between stocks and oil is positive (0.055).
11
205
Table 11.2 The correlations between oil, stock and bonds during the period 19892010
Years
Bond-stock
Bond-oil
Stock-oil
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1989-2010
0.237
0.489
0.395
0.206
0.375
0.616
0.497
0.612
0.236
-0.273
0.272
-0.056
-0.124
-0.532
-0.330
-0.037
0.010
0.101
-0.385
-0.392
-0.298
-0.486
-0.101
0.009
-0.289
-0.267
-0.045
-0.055
-0.200
0.052
0.007
0.013
-0.043
-0.087
-0.152
-0.060
-0.183
0.141
0.073
-0.022
0.022
-0.032
-0.300
-0.412
-0.368
-0.143
-0.014
-0.347
-0.243
0.060
-0.033
-0.181
0.027
-0.048
-0.105
0.080
-0.032
-0.061
-0.065
0.147
-0.255
-0.099
-0.044
0.006
0.062
0.258
0.458
0.651
0.055
The correlations are not constant during the observed period. In the cases of
bonds and stocks, the correlation is negative in 10 years and positive in 12 years
and moves between -0.532 in 2002 and +0.616 in 1994 respectively. In 15 of the
22 yearly observations the correlation between bonds and oil is negative and
fluctuates from a minimum in 2009 of -0.412, to a maximum in 2003 of +0.141.
The correlation between stocks and oil in 13 years is negative with a minimum of
-0.347 in 1990 and in 11 years, positive with a maximum of +0.651 in 2010.
Next, in an optimisation exercise the ideal portfolio is determined in terms of
returns for a given risk level. Different portfolio compositions with varying percentages of stocks, bonds, and oil are considered. For each composition the return
and standard deviation is calculated according to portfolio theory. The total return
is given by:
X
r
w i ri
11:1
in which ri is the return of component i and wi the percentage of component i with
respect to the total portfolio. The variance of the total portfolio is
X
X
2 2
r2
w
r
2
wrwrq
11:2
i
i
i
i6j i i j j ij
206
A. Dorsman et al.
Parameter
Mean
t-value
a
b0
b1
b2
b3
b4
b5
b6
b7
Durbin-Watson
0.001
-0.066
-1.003
1.691
-0.129
-0.363
-0.657
2.244
-4.843
2.04
1.802
-1.551
-0.984
1.443
-0.105
-4.694
-0.333
1.021
-1.549
in which ri is the standard deviation of component i and the correlation qij between
components i and j. The return is plotted against the standard deviation in a cluster of
points in which each point is a certain composition. The efficient frontier follows
from the upper line of this cloud of points. The market portfolio follows from the
composition with the highest expected return-risk ratio in which the gain is the
difference between the return of the portfolio and the return of US Treasury bonds.
After the quick scan of the individual graphs of bonds, stocks, and oil and the
descriptive statistics of the individual variables, we now look at advanced modelling techniques that allow for the inclusion of fat tail phenomena, tail dependence, and micro-correlations are required. Due to the fact that we look at the
attributive value of oil to a portfolio of bonds and stocks we take the oil price as
response variable (dependent variable) and the bonds price and stock price as
explanatory variables (independent variables). We also add to the equation
quantile variables to test the influence of extreme variables.
To test hypothesis H1 we estimate the parameters of the following equation:
Roil a b0 Rstock; q100 b1 Rstock; q10 b2 Rstock; q5
b3 R stock; q1 b4 Rbond; q100 b5 Rbond; q10
b6 Rbond; q5 b7 R bond; q1 e
11:3
Where:
R (oil) = the log-return of oil
R (stock, qx) = log return of stocks that are in the x % lower quantile
R (bond, qy) = log return of bonds that are in the y % lower quantile
e = error term
In case of a safe haven all the parameters b0,, b7 have to be negative. A
negative value means that the associated parameter is a hedge for oil.
In Table 11.3 we present the test results of Eq. (11.3). Only b4 is significantly
different from zero.3 Therefore we do not reject hypothesis H1 (that oil is not a
3
The values found for R2 are irrelevant for Eq. (11.3) since the function is only locally linear
and not globally. Calculating the adjusted R only makes sense when the function is linear over the
whole domain of variables.
11
207
Parameter
Mean
t-value
a
b1
b2
Durbin-Watson
Adjusted R2
0.000
0.089
-0.585
2.02
0.0222
1.390
3.147
-10.597
safe haven for stockholders and bondholders). When we apply the Brent Oil future
(code: OILBREN) instead of the WTI future we obtain the same conclusions when
testing hypothesis H1.
We use the following regression equationn to test our hypothesis H2 (that oil is
not a hedge for stockholders and bondholders).
R oil a b1 R stock b2 R bond e
11:4
Where:
R (oil) = the log-return of oil
R (stock) = log return of stocks
R (bond) = log return of bonds
e = error term
If oil is a hedge for stockholders and bondholders, the variables b1 and b2 are
\= 0.
Based on these results we reject hypotheses H2 (that oil is not a hedge for
stockholders and bondholders) in case of stocks as well as bonds. Also in testing
hypothesis H2 we used also Brent Oil instead of WTI. This replacement had no
influence on our conclusions. In Table 11.4, the adjusted R2 is close to zero which
means that the uncertainty of the coefficients a, b1 and b2 is high. Consequently,
the conclusions regarding H2 are not reliable.
In Fig. 11.2 we present the efficient frontier in the cases of stocks and bonds and
stocks, bonds and oil respectively. Tables 11.5 and 11.6 present for the various
sub-periods the weights of the components in the market portfolio in case this
portfolio does not contain respectively contains an investment in oil.
Figure 11.2 shows that adding oil to a portfolio of stocks and bonds means that
the efficient frontier moves upwards. From Table 11.5 we see that the optimal
portfolio without oil has 24 % stocks, 76 % bonds, an expected return of 0.0032, a
standard deviation of 0.000025 and an expected return-risk ratio of 11.9. The
optimal portfolio with oil, Table 11.6, has 6 % oil, 21 % stocks, 73 % bonds, an
average return of 0.00032, a standard deviation of 0.000022 and an expected
return-risk ratio of 14.0. Comparing the two portfolios, we see a small reduction of
risk and a small (nearly zero) increment of return when we add oil to the portfolio.
When we apply oil Brent future (code: OILBREN) instead of WTI (see appendix
Tables A4) we see that the optimal portfolio with oil has 8 % oil, 15 % stocks,
77 % bonds, an average return of 0.000189 a standard deviation of 0.0000254 and
an expected return-risk ratio of 7.1. The addition of Oil Brent futures exhibits a
greater change in the efficient frontier.
208
A. Dorsman et al.
Fig. 11.2 The efficient frontier for portfolios of stocks and bonds (lower graph) and the efficient
frontier for portfolios of oil, stocks and bonds (top graph) over period 1989 till 2010
Table 11.5 The weights of stock and bonds in the market portfolio without oil for each of the
sub-periods
Period
Stock (%) Bond (%) Mean
Standard deviation Expected return-risk ratio
19891998
19901999
19912000
19922001
19932002
19942003
19952004
19962005
19972006
19982007
19992008
20002009
20012010
19892010
27
31
25
24
20
24
23
23
22
22
14
16
21
24
73
69
75
76
80
76
77
77
78
78
86
84
79
76
4.9E-04
4.2E-04
4.3E-04
3.5E-04
3.4E-04
3.2E-04
3.7E-04
2.9E-04
2.9E-04
2.6E-04
2.5E-04
2.3E-04
2.0E-04
3.2E-04
2.4E-05
2.5E-05
2.3E-05
2.4E-05
2.4E-05
2.5E-05
2.4E-05
2.3E-05
2.1E-05
2.0E-05
2.4E-05
2.6E-05
2.6E-05
2.5E-05
19.8
15.7
17.4
13.7
13.5
12.0
14.9
11.7
13.1
12.2
10.0
8.4
7.5
12.9
Our third hypothesis is H3 (that the efficient frontier will not change when we add
oil as an alternative investment opportunity for the components stocks and bonds).
Based on our results we accept this hypothesis. Holders of portfolios of stocks and
bonds can improve their expected return-risk ratio by adding oil to their portfolio.
To test our fourth hypothesis, (that the market portfolio of oil, bonds and stocks
is constant during the observed period), we examine the weights of the portfolios
for every sub-period of 10 years. We started with the period 19891998, then
19901999, etc. till 20012010. In Tables 11.5 and 11.6 we present also the
weights of oil, stock and bonds for these sub-periods. The weight of oil moves
from 4 % (several sub-periods) to 8 % (20012010). We do not reject the fourth
hypothesis. Holders of portfolios of stocks and bonds who want to diversify their
portfolio with oil could opt for a 6 % oil, 21 % stock and 73 % bond split. This
distribution is more or less stable during the observed period.
11
209
Table 11.6 The weights of oil, stock and bonds in the market portfolio for each of the subperiods
Period
Oil (%)
Stock (%)
Bond (%)
Mean
Standard
Expected
deviation
return-risk ratio
19891998
19901999
19912000
19922001
19932002
19942003
19952004
19962005
19972006
19982007
19992008
20002009
20012010
19892010
4
5
4
4
5
6
5
6
5
7
7
7
8
6
27
30
24
23
18
22
22
22
21
20
12
14
17
21
69
65
72
73
77
72
73
72
74
73
81
79
75
73
4.6E-04
4.0E-04
4.1E-04
3.4E-04
3.4E-04
3.2E-04
3.7E-04
2.9E-04
2.9E-04
2.8E-04
2.7E-04
2.5E-04
2.3E-04
3.2E-04
2.2E-05
2.3E-05
2.2E-05
2.3E-05
2.2E-05
2.3E-05
2.2E-05
2.2E-05
2.0E-05
1.9E-05
2.3E-05
2.4E-05
2.5E-05
2.2E-05
20.6
16.8
18.1
14.2
14.2
13.0
15.8
12.8
14.0
14.0
11.5
9.7
8.9
14.0
210
A. Dorsman et al.
During the observed period 19892011 the markets were confronted with a
serious price falls. These occurred in the option markets in 1989, the internet crisis
in 20012002 and the financial crisis in 20072008 and the euro-crisis in
20102011. The impact of the first two crises on the real world was limited, while
the last two crises seriously damaged global markets and therefore affected also the
oil price.
Also we see that emerging markets (for example the BRIC and next-11
countries) and commodity countries like Australia show a different economic
development than the mature countries in Europe and the US. This study was limit
to US-based data. Further study has to show whether the results found in this
chapter also can be found for other periods and other countries.
A.1 Appendix
Table A.1 Descriptive statistics of the daily log returns of oil for every year during the observed
period 19892010
Year
Mean (%) Median (%) Standard deviation (%) Minimum (%) Maximum (%)
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
19892010
0.0906
0.1015
-0.1521
0.0075
-0.1223
0.0869
0.0369
0.1076
-0.1475
-0.1460
0.2887
0.0176
-0.1152
0.1735
0.0159
0.1106
0.1307
0.0001
0.1734
-0.2925
0.2208
0.0540
0.0291
0.1555
0.0000
0.0000
0.0000
-0.1450
0.0841
0.0000
0.2204
-0.1136
-0.1789
0.3224
0.1779
0.0000
0.0767
0.0000
0.1609
0.0707
0.0474
0.1233
-0.0707
0.1269
0.0000
0.0000
2.1807
3.7504
3.5109
1.2312
1.5353
1.7942
1.2575
2.4417
1.7758
2.8957
2.1848
2.6968
2.6915
2.1525
2.4389
2.2517
1.9894
1.7194
1.9171
3.8357
3.3682
1.7110
2.4480
-14.5131
-17.4480
-40.0478
-7.1345
-6.7555
-7.1924
-6.2365
-9.1199
-4.5261
-11.5463
-7.1541
-12.9400
-16.5445
-6.2753
-11.5404
-7.6977
-4.8965
-4.3478
-4.7942
-12.5952
-13.0654
-5.1170
-40.0478
8.6385
13.5724
12.6819
4.9381
4.7982
6.5426
3.3114
9.4076
5.3060
14.2309
6.5372
8.1129
8.0748
6.1330
6.3004
5.9621
6.7362
5.2189
7.3689
16.4097
13.1363
4.1633
16.4097
11
211
Table A.2 Descriptive statistics of the daily log returns of stocks for every year during the
observed period 19892010
Year
Mean (%) Median (%) Standard deviation (%) Minimum (%) Maximum (%)
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
19892010
0.1059
-0.0121
0.1019
0.0280
0.0368
0.0050
0.1227
0.0789
0.1103
0.0963
0.0732
-0.0367
-0.0485
-0.0957
0.0966
0.0394
0.0184
0.0564
0.0205
-0.1763
0.0900
0.0538
0.0347
0.1212
0.0812
0.0000
0.0073
0.0038
0.0159
0.0864
0.0381
0.1062
0.1090
0.0000
-0.0134
0.0000
-0.1406
0.1008
0.0474
0.0496
0.0789
0.0590
0.0000
0.1428
0.0741
0.0336
0.8132
0.9976
0.8850
0.6008
0.5341
0.6095
0.4852
0.7315
1.1271
1.2595
1.1178
1.3780
1.3233
1.6073
1.0552
0.6855
0.6378
0.6211
0.9901
2.5385
1.6869
1.1184
1.1382
-6.3115
-3.0432
-3.7257
-1.8685
-2.4129
-2.2425
-1.5499
-3.1307
-7.1130
-7.0419
-2.8456
-6.0044
-5.0114
-4.2408
-3.5859
-1.6416
-1.6857
-1.8489
-3.5255
-9.4595
-5.4254
-3.9657
-9.4595
2.7386
3.1761
3.6641
1.5541
1.9198
2.1409
1.8609
1.9289
4.9894
4.9708
3.4830
4.6673
4.9007
5.5754
3.4849
1.6229
1.9557
2.1379
2.9009
10.9582
6.8575
4.3064
10.9582
Table A.3 Descriptive statistics of the daily log returns of bonds for every year during the
observed period 19892010
Year
Mean (%) Median (%) Standard deviation (%) Minimum (%) Maximum (%)
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
0.0666
0.0240
0.0648
0.0292
0.0609
-0.0297
0.1029
-0.0038
0.0534
0.0487
-0.0345
0.0707
0.0158
0.0594
0.0094
0.0283
0.0626
0.0000
0.0478
0.0232
0.0460
0.0000
0.0584
0.0000
0.0526
0.0395
-0.0291
0.0877
0.0326
0.0951
0.0812
0.0455
0.4691
0.5317
0.4391
0.4097
0.4592
0.6101
0.4956
0.5941
0.4714
0.5241
0.5348
0.4809
0.6475
0.6000
0.6627
0.5329
-1.6727
-2.1669
-1.3497
-0.9646
-1.2375
-3.1343
-1.6156
-2.8071
-1.6538
-2.1514
-1.8114
-1.2402
-2.1747
-1.7950
-2.0644
-2.1188
2.0513
1.6311
1.8120
1.3717
1.2663
2.0846
1.8163
1.4998
1.7770
1.4138
1.3261
1.4957
2.1312
1.7745
1.7125
1.8362
(continued)
212
A. Dorsman et al.
0.0444
0.0422
0.0146
0.0475
0.0138
0.0182
0.0385
0.4551
0.3809
0.4960
0.8467
0.9465
0.8117
0.5815
-1.4505
-1.0720
-1.4191
-3.1728
-2.7796
-2.4081
-3.1728
1.0420
1.1992
1.7360
2.8990
3.8105
2.6871
3.8105
Table A.4 The weights of oil, stock and bonds in the market portfolio for each of the subperiods
Period
Oil (%) Stock (%) Bond (%) Mean
Standard
Expected return-risk
deviation ratio
19891998
19901999
19912000
19922001
19932002
19942003
19952004
19962005
19972006
19982007
19992008
20002009
20012010
19892010
4
5
3
4
5
5
4
5
5
7
7
8
10
6
27
30
25
23
18
23
22
22
21
20
13
14
17
22
69
65
72
73
77
72
74
73
74
73
80
78
73
72
4.6E-04
4.0E-04
4.1E-04
3.4E-04
3.4E-04
3.2E-04
3.7E-04
2.9E-04
2.9E-04
2.8E-04
2.7E-04
2.5E-04
2.4E-04
3.2E-04
2.1E-05
2.3E-05
2.2E-05
2.3E-05
2.3E-05
2.4E-05
2.3E-05
2.2E-05
2.0E-05
1.9E-05
2.3E-05
2.4E-05
2.5E-05
2.2E-05
20.7
16.8
17.9
14.2
14.1
12.8
15.7
12.7
14.0
14.1
11.3
9.7
9.3
14.1
References
Arouri, M. E. H., & Nguyen, D. G. (2010). Oil prices, stock markets and portfolio investment:
Evidence from sector analysis in Europe over the last decade. Energy Policy, 38(8),
45284539.
Baur, D.G., & Lucey, B.M. (2009). Is gold a hedge or a safe haven? An analysis of stocks, bonds
and gold, working paper
Baur, D. G., & Dermott, T. K. (2010). Is gold a safe haven? Journal of Banking and Finance, 34,
18861898.
Chua, A. (1999). The role of international real estate in global mixed-asset investment portfolios.
Journal of Real Estate Portfolio Management, 3(2), 129137.
Geman, H., & Kharoubi, C. (2008). WTI crude oil futures in portfolio diversification: The timeto-maturity effect. Journal of Banking and Finance, 32(12), 25532559.
Kousky, C., & Cooke, R.M. (2009). The unholy trinity: fat tails, tail dependence, and microcorrelations. RFF Discussion paper, November (rev.) 2009. Washington DC, Resources for
the Future
11
213
Letzelter, J.C. (2005). Finding the efficient frontier: Power plant portfolio assessment.
Proceedings of the 2005 Crystal Ball User Conference, June 13, 2005
Mandelbrot, B. (1963). The variation of certain speculative prices. Journal of Business, 36,
394419.
Mandelbrot, B. (1966). Forecasts of future prices, unbiased markets and martingale models.
Journal of Business, 39, 242255.
ONeill, J. (2001). Building better global economic BRICs, Global Economic papers no: 66,
Goldman Sachs
Resnick, S. (2007). Heavy tailed phenomenal: Probabilistic and statistical modelling. New York:
Springer.
Umutlu, G., Dorsman, A., & Telatar, E. (2011). Day-ahead market and futures market. In A.B.
Dorsman et al. (Eds.), Financial Aspects in Energy (pp. 109128). London: Springer
Chapter 12
Abstract In the past, energy networks (grids) were nationally organized. The
grids were linked by interconnectors. The capacities of the interconnectors were
limited and only used to counter an imbalance in one of the grids. Governments
fixed the prices and there was no energy price risk. Liberalization of the market
introduced prices that fluctuate every moment; with the liberalization, energy price
risk was introduced. The more volatile the energy prices, the larger the risk for
market participants. Market coupling links the former nationally organized
markets, which may cause a reduction in the volatility of the energy prices. At first
the TSOs (Transmission System Operators) sold connector capacity by so called
explicit auction, separate from the electricity auction. With the mechanism of
explicit auction it was relatively easy to realize a market based allocation of scarce
limited interconnector capacity on adjacent borders. Explicit auctions however do
not realize the optimal result. In due time, they are replaced by so-called implicit
auctions where the interconnectors capacities are automatically allocated in such
a way that electricity price differences between countries are minimized. This
implicit mechanism is also referred to as market coupling. In this chapter the effect
of market coupling on market prices is investigated in the observed period,
1 January 200531 March 2011, for Scandinavia (South), The Netherlands,
Belgium and France. It is found that due to market coupling the price differences
between the markets diminish.
Keywords Electricity market
215
216
A. Dorsman et al.
12.1 Introduction
In a perfect market all buyers and sellers have access to the same price information
and they can exercise their transactions orders against the same conditions (Porter
1980). To realize the best position for the buyers and sellers all trade has to be
concentrated as much as possible. Pagano (1989) argues that since the depth of
liquidity of a market depends on the entry decisions of all potential participants,
each trader assesses them according to conjectures about entry by others. He
argues that if trade is equally costly across markets, this externality leads to the
concentration of trade in one market. Stoll (2001) discusses the problem of market
fragmentation. This topic arose in the USA where several option markets developed. He wrote that heads of Goldman Sachs, Merrill Lynch and other companies
testified on the need for a central order book. Also in Europe a need is felt to
concentrate market transactions. For example, in The Netherlands in the sixties of
the last century the law determined that all transactions have to be traded on the
Amsterdam stock exchange, now Euronext Amsterdam. However, by doing so not
only the liquidity improves, but also monopolies are being created. A monopoly
does not feel the necessity to invest in new developments.
After the merger in 2000 between the stock exchanges of Amsterdam, Brussels
and Paris there was a need for one law for the new entity. The Dutch law was
adjusted and became in line with the French law. One substantial change was that
the condition that every buyer and seller in The Netherlands has to trade on the
Amsterdam Stock Exchange (Euronext Amsterdam) was skipped. On 1 November
2007 The European Union (EU) introduced the Markets in Financial Instruments
Directive (MIFID). The effect of MIFID was that new markets were allowed.
Alternative markets make use of this change in the law. In 2006 ChI-X started
and in 2011 TOM. By using cheaper ICT systems these new markets could
compete with Euronext. The condition of Pagano that trade was equally costly
across markets was no longer met. Cost-based competition between markets leads
to better prices for the market traders. In 2010, Chi-X was the second stock
exchange after the London Stock Exchange in terms of volume of trades in Europe
(Menkveld 2011).
The European electricity markets are more or less in the same situation as the
stock markets fifty years ago. Creating one market where all the bids and asks are
concentrated gives the best prices and contributes to market transparency.
Historically the electricity networks (grids) were nationally organized. The general
policy today is to link the several electricity markets to get the best prices.
Liberalization of the European electricity market started in Norway in 1991.1 In
that year the Norwegian Parliaments decision to deregulate the market for trading
of electricity became operational. In 1993 Statnett Marked AS was established as
an independent company. After a joint venture with Sweden, the company was
1
12
217
renamed as Nord Pool ASA. Finland joined in 1998. In 2004 and 2007 respectively, East Denmark and West Denmark joined Nord Pool. In this chapter
Scandinavia is referred to instead of the countries Norway, Sweden, Finland and
Denmark. Other European countries followed Scandinavia in the process of
liberalization. Among the first followers was The Netherlands. The Dutch electricity market has started with the modification of the Electricity Act in 1999. The
approach of liberalizing the Dutch electricity market was a phased approach,
where the first step was to open the wholesale markets for competition, followed
by a second phase focusing on smaller scale industries/large business, and finalized
by a third stage of opening of the market for small businesses and domestic
customers. Looking at the power exchanges, it can now be said that Nord Pool
(Scandinavia) and APX (The Netherlands) are also among the front-runners. The
advantage of APX is that it is situated more in the geographical center of Europe.
In the past the location of the power market was not important. Every country
had its own network (grid). However, with the liberalization the wish to create an
efficient electricity market in Europe became stronger. To realize this goal it was
necessary to reduce the hurdles (imperfections). However, electricity has some
special physical characteristics impacting the trading and price formation
(Shahidehpour and Alomoush 2001).
Compared to stocks and bonds, the law of one price does not hold. The price of
electricity depends on time and location. On top of that, electricity is not easy to
store. Geographical barriers can influence the building of the grid. In flat countries
like (large parts of Belgium) and The Netherlands the construction of the grid is
easier than in countries like Italy, where it is difficult to cross the Apennines with
cables. The grids in flat countries are therefore more comparable to the web of a
spider, while in countries like Italy the physical form of the grid is more comparable to a lattice. The construction of a spider web is less sensitive to local
power problems than the construction of a lattice. It is not surprising that there was
an increase of interconnectors mainly between flat countries, such as the Southern
part of Norway, The Netherlands, Belgium, France and Germany. From a technical
perspective they are easier to realize than in mountainous areas.
The TSO of every country is responsible for the power balance in the grid at
any moment. Two developments make this responsibility more difficult; namely
the process of connecting the grids and the increase in solar and wind energy.
In the past, interconnectors were only used to avoid imbalances in one of the
grids. With the liberalization of the electricity markets, the capacity of interconnectors became available for market participants by explicit auction. The sale of
the capacity of interconnectors to market participants reduces the opportunities for
the TSOs to get their grid in balance. The supply of solar and wind energy is not
constant over time. If the sun shines, the supply of solar energy is high, otherwise
the supply of solar power is much lower or zero. It is similar for wind energy.
The TSO has to manage this volatility in supply. If solar and wind energy
represent only a small part of the total supply of electricity, this volatility is easy to
manage. However, the larger the contribution of solar and wind energy in total, the
more difficult the task of the TSOs. It is easier to manage the fluctuations in wind
218
A. Dorsman et al.
and solar energy in a large market than in several relatively small markets. From
this perspective, the physical integration of networks is important too. For
example, if wind production creates overproduction (compared to the demand) in
one grid, the characteristic of electricity that it cannot be stored, leads to the
situation that the overproduction needs to be transported outside that particular
grid. In the case that that cannot be done, the imbalance in supply in that grid leads
to grid-instability risk or even black-outs.
Interconnectors between grids can function as an exit point in such situations of
overproduction due to renewable energy (e.g. the situation where there is too much
wind/solar energy). Ideally, the interconnectors are of such capacity that they can
always make sure a surplus of generated energy can be transported away from the
grid(s) where there is overproduction.
A very interesting example in this respect is the plan of Belgian and German TSOs
to construct the first interconnector between Belgium and Germany. Historically,
there has not been such an interconnector (possibly due to a geographical hurdle: The
Ardennes), but the actual development of increasing grid integration makes it timely
for this newly-planned interconnector. On the one hand, it can contribute to enhanced
grid security/stability (see above) but it can also contribute to enhanced trading
market integration (more possibilities to trade between trading zones). With this
development, the traditional nationally orientated networks are now becoming interlinked grids where TSOs have a shared responsibility for the European network.
The goal of the liberalization of the electricity markets is to get good working
electricity markets with efficient electricity prices. In such a market, all relevant
information is direct and fully absorbed in the prices. The efficiency of markets can
be promoted by reducing hurdles (imperfections) to bring demand and supply
together. One of the hurdles is that market participants have a substantial influence
on the price. These market participants can cause a deviation from the efficient
price. In other words, liberalization of the power markets does not only mean the
start of a power exchange, but also means that enough participants individually can
have no influence on the price; they are price takers. Another hurdle is the
interconnector capacity between the grids. If their capacity is limited, the price
development of the several grids can also deviate from optimal price development.
Before proceeding to the next section, a distinction must be made between two
methods of auctions for interconnector capacity, namely the methods of explicit and
implicit auction. A more detailed description between these two auction types can be
found in Dorsman et al. (2011). In short these mechanisms comprise the following:
12
219
capacity. For this chapter the focus is only on day-ahead capacity since this is by
far the most liquid, transparent and representative market.
After acquiring the transmission capacity via the explicit day-ahead auction, the
market participant can start scheduling the import or export energy flows that they
want to execute. Explicit auctioning mechanisms produce sub-optimal outcomes
(see Dorsman et al. 2011).
12.2 Data
Day-ahead prices are obtained from APX-Endex for the Dutch market, from
Belpex for the Belgian Market, from Epexspot for the French market and from
Nord Pool for the Scandinavian market. The observed period is 1 January 2005
220
A. Dorsman et al.
until 31 March 2011. This period has been chosen because the following events
could be analyzed.
1. 21 November 2006: Market coupling of Belgium, France and The
Netherlands
2. 5 May 2008: NorNed cable
3. 9 November 2009: Implicit coupling2 on cables of Germany and the Nordic
region.
4. 9 November 2010: implicit coupling of Belgium, France, The Netherlands,
Germany and Luxembourg. Interim implicit coupling between regions CWENordic.
5. 1 April 2011: BritNed cable.
1. 21 November 2006: Market coupling of Belgium, France
and The Netherlands
This event is important because individual energy markets already existed in
the separate regions. These markets were, however, connected via the
sub-optimal mechanism of explicit auctions and also the maturity of the
markets was not as it is today. With the start of the market coupling between
these markets, the worlds first implicit market coupling became a fact,
realizing an integrated and increased trading region.
2. 5 May 2008: NorNed cable
The study of this event is important, since a new connection (that is, the
580 km underwater NorNed cable) enables direct trading of electricity
between the Nordic region and the Dutch markets. The impact of this cable on
the both markets is significant; on the one hand the cable allows the Dutch
market to import cheap hydro-power during peak-hours. On the other hand,
the cable allows the surplus of Dutch coal-based electricity during nighttime
hours to be used in the Nordic market, which saves usage of hydro-power
electricity during those hours. This saved hydro-power electricity can be used
at moments with higher power demands. Although started as an explicit
auction, the coupling of the Nordic market with the energy markets in continental Europe has already led to very good results in optimizing price formation and usage of scarce interconnection capacity between the participating
CWE and Nordic market regions were coupled based on an interim implicit coupling
mechanism since a quick solution for implicit market coupling of these regions needed to be
available at the time of starting CWE MC. This interim solution produces better results than the
other available solution at that time for coupling the regions concerned, which was not sufficient
to be continued when CWE MC would GO-live. The interim variant for coupling CWE and
Nordic regions is planned to be replaced by an enduring solution (NWE Enduring market
coupling) that will realize the final, optimal results for coupling CWE, Nordic (and also UK)
markets.
12
221
222
A. Dorsman et al.
continental European market and between the continental European and the Nordic
regions.
In this chapter the following issues are examined. It is important to consider
whether or not:
1. Market coupling reduces the difference in daily price volatility between the
grids.
2. Market coupling reduces the price difference between the grids.
The supply of electricity in The Netherlands is very inflexible, which causes
high peak and low off-peak prices. In Scandinavia, electricity is generated by
means of hydro-power. Therefore, the storage of electricity is easier and cheaper in
Norway than in The Netherlands. Linking the Scandinavian and the Dutch systems
will have an equalizing influence on the price of electricity in The Netherlands,
which means that the gap between peak and off-peak prices will decrease. The
hours 8:0020:00 are peak hours and 0:008:00 h plus 20:0024:00 h are off-peak
hours.
The off-peak prices in The Netherlands are lower than in Scandinavia. During
night-time, electricity flows from The Netherlands to Scandinavia. Since on
average the electricity price in Scandinavia is substantially lower, it can be seen
that, in The Netherlands during daytime, the electricity flows from Scandinavia to
The Netherlands and during nighttime the flow is in reverse.
The first hypothesis to test is:
H1 Market coupling does not influence the difference in daily price volatility
between the grids.
In this study, daily price volatility is defined as the sample standard deviation
over a 24 h period (24 prices).
Also the influence of market coupling on off-peak and peak prices is investigated. In the case where market coupling has an effect on prices, it is expected that
the electricity prices of the relevant grids show a larger co-movement and a
smaller difference during off-peak and peak hours.
H2 Market coupling does not influence the price difference during off-peak hours
between the grids.
H3 Market coupling does not influence the price difference during peak hours
between the grids.
Note that every coupling has its own dynamics. For example, the coupling
between Scandinavia and The Netherlands will differ from the coupling between
The Netherlands, Belgium and France. Differences in price formation and also
differences in power generation in each country/region make every coupling
unique. Therefore, the hypotheses must be tested for every specific coupling
moment.
12
(a)
223
400,00
EUR
300,00
200,00
100,00
0,00
-100,00
-200,00
-300,00
-400,00
-500,00
-600,00
(b)
EUR
100,00
50,00
0,00
-50,00
-100,00
-150,00
-200,00
-250,00
Fig. A.1 The difference in volatility (a), off-peak prices (b) and peak prices (c) between The
Netherlands and France in the period 21 November 200521 November 2007
224
A. Dorsman et al.
(c)
200,00
EUR
150,00
100,00
50,00
0,00
-50,00
-100,00
12
225
Table 12.1 a Kolmogorov tests and b Wilcoxon tests on the difference in daily volatility,
difference in prices during off-peak and peak hours between The Netherlands and France, one
year before and one year after 21 November 2006
Daily Volatility
Off peak
Peak
Test statistics (NL-FR)
Most extreme differences
Absolute
Positive
Negative
KolmogorovSmirnov Z
Asymp. Sig. (2-tailed)
MannWhitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
0.441
0.044
-0.441
5.959
0.000
0.364
0.099
-0.364
4.923
0.000
0.463
0.049
-0.463
6.255
0.000
34,634.000
101,429.000
-11.225
0.000
48,353.000
115,148.000
-6.410
0.000
34115.000
100,910.000
-11.407
0.000
KolmogorovSmirnov Z
Asymp. Sig. (2-tailed)
MannWhitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
Absolute
Positive
Negative
Off peak
Peak
0.129
0.066
-0.129
1.745
0.005
0.108
0.035
-0.108
1.455
0.029
0.181
0.063
-0.181
2.450
0.000
64,933.000
131,728.000
-0.652
0.514
60,479.500
127,274.500
-2.212
0.027
61,018.500
127,813.500
-2.024
0.043
226
A. Dorsman et al.
Difference in Daily Volatility Netherlands-Norway (KRS)
(a)
400,00
350,00
300,00
EUR
250,00
200,00
150,00
100,00
50,00
0,00
(b)
100,00
80,00
EUR
60,00
40,00
20,00
0,00
20,00-
40,00-
Fig. A.2 The difference in volatility (a), off-peak prices (b) and peak prices (c) between The
Netherlands and Norway in the period 5 May 20075 May 2009
market coupling, also a harmonization of price volatility can be observed. Especially the harmonization of price volatility confirms that one integrated market has
12
(c)
227
500,00
EUR
400,00
300,00
200,00
100,00
0,00
-100,00
been realized after this first market coupling (harmonized volatility levels mean
that markets seem to mimic each other).
228
A. Dorsman et al.
(a)
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
-100,00
-75,00
-50,00
-25,00
0,00
25,00
50,00
75,00
100,00
(b)
Empirical Cumulative Distributions of the Price Difference NL-FR during Off Peak Hours
21 november 2005 -22 november 2006
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
-25,00
-20,00
-15,00
-10,00
-5,00
0,00
5,00
10,00
15,00
20,00
25,00
Fig. A.3 Cumulative distributions difference between The Netherlands (NL) and France (FR) in
daily volatility (a), off-peak hours (b) and peak hours (c) one year before and one year after the
moment of coupling 21 November 2006
12
(c)
229
Empirical Cumulative Distributions of the Price Difference NL-FR during Peak Hours
22 november 2005 - 21 november 2006
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
-100,00
-75,00
-50,00
-25,00
0,00
25,00
50,00
75,00
100,00
smaller due to the market coupling. Looking at Fig. A.4a, a clear-cut switch to the
left cannot be observed for the differences in volatilities between The Netherlands
and Norway (apart from the upper tail). Therefore a closer examination was made
of the individual price developments in off-peak and peak hours for The Netherlands and Norway (See Fig. A.5). From a perusal of the Dutch figures it can be
seen that, except from some extremes/outliers, the connection of the NorNed cable
on 5 May 2008 had little impact on the volatility in the Netherlands. However, the
volatility in Norway increased significantly.
In Table 12.3 the KolmogorovSmirnov test and also the Wilcoxon test
(Table 12.3) show that this increase in volatility for Norway is significant. In
Table 12.3, the two sample Wilcoxon test gives the same picture.
The increase in volatility in Norway is in line with expectations. During peak
hours the flow goes from Norway to The Netherlands, which means that in Norway
the supply of electricity decreases when the price is high, resulting in higher prices
during peak hours. On the other hand, during off-peak hours the flow goes from
The Netherlands to Norway. In other words, when the electricity price in Norway
is low, the supply increases by the cable, resulting in lower prices during off-peak
hours. Due to the NorNed cable, the Norwegian volatility in electricity prices
increased substantially, because the price difference between peak and off-peak
hours increased significantly. This was not the case in The Netherlands, where
only the extreme volatilities disappeared after 5 May 2008.
Before the implicit coupling in Event 1 (Belgium, France and The Netherlands)
there was already an exchange of electricity between the countries (explicit
coupling). Therefore the impact of the introduction of implicit coupling was less
than in Event 2 (Norway and The Netherlands) where before the implicit coupling
230
A. Dorsman et al.
(a)
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
-25,00
0,00
25,00
50,00
75,00
100,00
125,00
150,00
(b)
Empirical Cumulative Distributions of the Price Difference NL-NO during Off Peak Hours
6 may 2007 - 5 may 2008
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
-20,00 -15,00 -10,00 -5,00
0,00
5,00
10,00 15,00
20,00
25,00 30,00
35,00
40,00
45,00
50,00
Fig. A.4 Cumulative distributions difference between The Netherlands (NL) and Norway (NO)
in daily volatility (a), off-peak hours (b) and peak hours (c) one year before and one year after the
moment of coupling 21 November 2006
12
(c)
231
Empirical Cumulative Distributions of the Price Difference NL-NO during Peak Hours
6 may 2007 - 5 may 2008
50,00
100,00
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
-25,00
0,00
25,00
75,00
125,00
150,00
175,00
200,00
KolmogorovSmirnov Z
Asymp. Sig. (2-tailed)
MannWhitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
Absolute
Positive
Negative
0.097
0.097
-0.063
1.312
0.064
0.218
0.218
0.000
2.949
0.000
64,446.000
131,607.000
-0.823
0.411
46,684.000
113,845.000
-7.045
0.000
between Norway and The Netherlands, both after the introduction of the NorNed
cable. The observed period for this event study is one year before the event and
one year after the event (T-1, T ? 1). Due to the effect of specifically choosing
the (T-1, T ? 1) approach for this chapter, the market learning effect after
introduction of such a fundamental change in the North-Western energy markets is
also included in the analysis/outcomes. Especially the first half-year after introduction of the NorNed cable could be qualified as being new to the market,
meaning that market participants still needed to find out the optimal usage and
value for them of the cable. If an event study had been set up for one year periods
232
A. Dorsman et al.
(a)
1,00
0,90
0,80
0,70
0,60
0,50
0,40
0,30
0,20
0,10
0,00
0,00
25,00
50,00
75,00
100,00
125,00
150,00
Daily Volatility NL
(b)
1,00
0,90
0,80
0,70
Title
0,60
0,50
0,40
0,30
0,20
0,10
0,00
0,00
5,00
10,00
15,00
Title
Fig. A.5 The volatility in Dutch electricity prices (a) and Norwegian electricity prices (b) during
period 5 May 20075 May 2009
before and after the introduction of the NorNed cable, but running in the periods
[T-1.5, T-0.5] and [T ? 0.5, T ? 1.5], the results would show even lower
volatility for the Dutch market and also an even lower price difference (peakprices) between Norway and The Netherlands. This strengthens the conclusions
drawn in Event 2.
12
233
12.5 Conclusion
In the past, the capacities of interconnectors were limited. After the liberalization
of the electricity markets these limitations became hurdles (imperfections) that
may cause inefficient electricity prices. In Europe it can be seen that the capacity
of existing interconnectors is enlarged and that new interconnectors are built. Also,
the way of auctioning the capacity of interconnectors has changed from explicit to
implicit auction which has led to more integrated markets. Five major events have
taken place during the last five years: The market coupling in Belgium, France and
The Netherlands; the NorNed cable; the implicit coupling on cables between
Germany and the Nordic region; the implicit coupling of Belgium, France, The
Netherlands, Germany and Luxembourg and the BritNed cable.
In this chapter the influence of Events 1 and 2, are analyzed: Market coupling
between Belgium, France and The Netherlands and the NorNed cable. The first
event reduces the differences in prices between The Netherlands and France during
off-peak and peak hours as well as the volatility in the price differences. Also, the
price differences between The Netherlands and Norway decreased significantly
after opening the NorNed cable. However due to the NorNed cable, Norwegian
prices during peak hours increased and during off-peak hours decreased, causing
the volatility to increase. Overall the NorNed cable creates value in Europe, but the
increasing daily volatility does not benefit the electricity consumers in Norway.
Overall it can be seen that an increasing integration of the European electricity
networks, reduces hurdles and creates a market where price inefficiencies are
decreasing.
References
Dorsman, A. B., van Montfort, K., & Pottuijt, P. (2011). Market perfection in a changing energy
environment. In A. Dorsman, W. Westerman, M. B. Karan, & . Arslan (Eds.), Financial
aspects in energy: A European perspective (pp. 7184). Heidelberg: Springer.
Feller, W. (1948). On the Kolmogorov-Smirnov limit theorems for empirical distributions.
Annals of Mathematical Statistics, 19, 177189.
Kolmogorov, A. N. (1933). Sulla determinazione empirica di una legge di distribuzione. Giornale
dellInstituto Itaiano degli Attuari, 4, 8391.
Mann, H. B., & Whitney, D. R. (1947). On a test of whether one of two random variables is
stochastically larger than the other. Annals of Mathematical Statistics, 18(1), 5060.
Massey, F. J, Jr. (1951). The Kolmogorov-Smirnov test of goodness of fit. Journal of the
American Statistical Association, 46, 6878.
Menkveld, A. J. (2011). High frequency trading and the new-market makers. Unpublished
manuscript. Retreived from http://ssrn.com/abstract=1722924.
Pagano, M. (1989). Trading volume and asset liquidity. Quarterly Journal of Economics., 104(2),
255274.
Porter, M. E. (1980). Competitive strategytechniques for analyzing industry and competitors.
New York: The Free Press.
234
A. Dorsman et al.
Shahidehpour, M., & Alomoush, M. (2001). Restructured electrical power systems: Operation,
trading, and volatility. New York: Marcel Dekker.
Smirnov, V. I. (1939). On the estimation of the discrepancy between empirical curves of
distribution for two independent samples. Bulletin Mathmatique de lUniversit de Moscou,
2, fasc. 2, 316.
Stoll, H. R. (2001). Market fragmentation. Financial Analysts Journal, 57(4), 1620.
Wilcoxon, F. (1945). Individual comparisons by ranking methods. Biometrics Bulletin, 1(6),
8083.
Chapter 13
Abstract This study analyses 158 energy company initial public offerings (IPOs) in
Australia from January 1994 to December 2010, including the period of the global
financial crisis (GFC). The study finds that energy company IPOs had an average
22.0 % underpricing and that those IPOs that sought to raise more equity capital and
engaged underwriters had lower underpricing. There is also evidence that suggests
energy company IPOs that offered options to their underwriters had higher underpricing returns, effectively cancelling the lower underpricing effect of the underwriting itself. The energy IPOs that raised equity capital after the 2007/8 global
financial crisis do not appear to have offered on average, significantly different
underpricing returns to their investors compared to those energy IPOs that raised
capital prior to this GFC period. The findings of this study offer insights for issuers
who seek to lower underpricing, for underwriters involved in the capital raising and
for investors who are looking to invest in Australian energy company IPOs.
Keywords Underpricing
Energy IPOs
13.1 Introduction
Initial public offerings (IPOs) of equity capital are a common occurrence in
financial markets around the world. Companies wanting equity capital and seeking
to list on a stock exchange may sell their shares to institutional and individual
investors. These investors in turn will seek to earn a reasonable rate of return on
B. Dimovski (&)
School of Accounting, Economics and Finance, Deakin University,
Geelong, Victoria 3216, Australia
e-mail: wd@deakin.edu.au
235
236
B. Dimovski
their investment. This return will be able to be realized by the investors from the
very first moment of listing. This financial transaction seems simple enough and
normal enough, except that the returns achieved by investors in IPOs historically
have been substantial and in some cases breathtaking. What makes the returns look
even more remarkable is the fact that they are achieved on the first day of the
companys shares being listed on the stock exchange.
IPOs have been discussed in the finance literature for over 30 years. The major
focus has often been on the anomaly of IPOs demonstrating this common and
persistent finding of underpricing. Underpricing is the term used when the issue
price of the shares of a company, raising equity capital from the public and seeking
to list on a stock exchange, is below the closing price of the shares on the first day
of listing. As such, underpricing theoretically allows subscribing investors the
opportunity of making this return on the first day of listing. The international
evidence as examined in Ritter (2003) and Loughran et al. (1994) (but regularly
updated with the most recent update being November 8, 2011) has documented
that subscribing investors made handsome double-digit (for example US IPOs
16.8 %, UK IPOs16.2 %, Turkish IPOs10.6 %, Greek IPOs50.8 %) or
even triple-digit (for example Chinese IPOs137.4 %) statistically significant
positive first day returns, on average. These studies are generally however, of
industrial company IPOs.
The main purpose of this paper is to investigate the underpricing of energy
company IPOs in Australia from January 1994 to December 2010, noting that the
Australian energy sector is important to Australia and to the world. The Australian
energy sector was capitalized at A$176 billion and comprised 12 % of the market
capitalization of the entire Australian Stock Exchange (2012). Australia is the
worlds largest coal exporter and a major supplier of liquefied natural gas and
uranium to world markets (see www.asx.com.au for more detail). While the World
Federation of Exchanges (2009 and 2010) ranked the ASX slightly outside the top
ten largest stock exchanges in the world by market capitalisation, the ASX was the
third largest in the world by capital raised by initial and secondary capital raisings
during 2009, behind the New York Stock Exchange (NYSE) and the London Stock
Exchange (LSE) and the fourth largest in the world in 2008 behind the NYSE, LSE
and the Hong Kong Stock Exchange. Even though the ASX moved behind China
and India in terms of initial and secondary capital raisings (due to their rapid
industrial growth) in 2010, the World Federation of Exchanges consistently ranks
the ASX among the global leaders in capital raisings. The ASX (2010) reports that
energy company IPOs constituted 24 % of all ASX IPOs in 2008 and 20 % of all
ASX IPOs in 2009. The importance of the Australian market in terms of equity
capital raising and the importance of the Australian energy sector more broadly,
makes the Australian market an extremely useful market to investigate the initial
public offerings of energy company IPOs.
A second purpose of the study is to utilize a sufficiently large number of
observations so as to draw useful conclusions. Two previous studies into natural
resource IPOs in Australia only made fleeting mention of energy IPOs because of
the small sample sizes. How (2000) identified 2 solid fuel IPOs and 13 oil and gas
13
237
IPOs (amongst 130 other natural resource IPOs) during 1979 to 1990 and advised
average underpricing returns of 106.5 and 47.3 % respectively for investors
subscribing to these IPOs. Using a very small sample size, the study suggested the
underpricing of energy IPOs was much larger than industrial company IPOs.
Dimovski and Brooks (2004) investigated 19 energy IPOs (amongst 96 other
natural resource IPOs) from 1994 to 1999 and reported an average underpricing
return of 8.3 %. Given a later but also small sample size, the second study
suggested the underpricing of energy company IPOs was much lower than
industrial company IPOs.
The data set of 158 energy company IPOs used in this study is significantly
greater than these previous Australian studies. In brief, these 158 energy IPOs
raised over $2.6 billion of public equity capital from January 1994 to December
2010.
There are three parties involved in the IPO transaction that may be particularly
interested in the results of this study. The three parties are the initial subscribers,
the company issuing the equity and the underwriters engaged in guaranteeing the
success of the capital raising. Given the mean underpricing returns reported from
energy IPOs in Australia, this study suggests that if theoretically, the initial
subscribers were able to buy into every IPO, they could theoretically, have earned
significant returns in a short space of time by investing in these IPOs. (It is worth
noting here, however, that Rock (1986) suggests a winners curse where more
informed investors are able to acquire more of the more profitable, higher
underpriced issues than the less informed investors who may well be left with the
lower underpriced or even overpriced issues.) The results also suggest that issuers
could consider some of the characteristics of the offer and the parties involved in
the IPO process that may be related to lower levels of underpricing. Finally, while
underwriters are not mandatory for capital raising in Australia, they are often used
by issuers to guarantee a certain amount of capital will be raised. Not only will
underwriters want to maximize the amount of wealth retained by their issuing
company client but they will also want to sell appropriately priced shares to their
investor clients. This study may help underwriters provide better advice to both
parties.
This study also follows a highly influential paper in the IPO literature by Beatty
and Ritter (1986). They argue that the lower the uncertainty about the value of an
IPO, the lower the underpricing needed to attract subscribers. Given the linkage
between uncertainty and underpricing, this study seeks to identify the factors that
might influence uncertainty and hence underpricing. A more recent paper by Brau
and Fawcett (2006) uses survey evidence from 336 chief financial officers who
concur with Beatty and Ritter (1986) that underpricing compensates investors for
taking risk.
The third purpose of this study is to investigate whether energy IPOs are seen to
be riskier and more uncertain about their cash flows than other Australian IPOs.
The energy IPOs in this study show a 22 % mean underpricing to subscribing
investors. This is broadly similar but slightly less than the underpricing of 29 %
reported in Dimovski et al. (2011) for Australian industrial company IPOs from
238
B. Dimovski
1994 to 2004. It is also interesting to note the results from underpricing studies
focusing on other sectors. In the real estate investment trust (REIT) sector in
Australia, Dimovski and Brooks (2006) report a mean 2.6 % underpricing of REIT
IPOs, while with Australian infrastructure IPOs, Dimovski (2011) reports a mean
underpricing of 3.5 %but both are not statistically significantly different to a
zero underpricing. These latter two sectors appear to be seen as less risky IPO
investment sectors than energy IPOs. Additionally, Dimovski and Brooks (2008)
investigate the underpricing of gold mining IPOs in Australia and report a mean
underpricing of 13.3 %interestingly quite a bit lower than the underpricing of
industrial company IPOs in Australia and significantly lower than the underpricing
returns of gold mining IPOs prior to 1991.
The results of this study suggest that if the issue is underwritten, underpricing is
lower and that energy IPOs that seek to raise more equity capital have lower
underpricing. There is also some evidence that suggests energy company IPOs that
offered options to their underwriters had higher underpricing, effectively cancelling the lower underpricing effect of the underwriting itself. Interestingly, the
energy IPOs that raised equity capital after the 2007/8 global financial crisis (GFC)
do not appear to have offered significantly different underpricing returns, on
average, to the subscribers compared to those energy IPOs that raised capital prior
to this GFC period.
The plan of this paper is as follows. In Sect. 13.2 we briefly summarise some of
the underpricing literature. Section 13.3 presents the data and methods.
Section 13.4 reports the results. Section 13.5 makes some concluding comments.
13
239
Rock (1986) suggests there are two categories of investors that seek shares in
IPOsthe informed and the uninformed. He argues that the informed (and likely
more influential) investors crowd out the uninformed (and likely less influential)
leaving the uninformed buying more of the less profitable issues. In order to
compensate the uninformed for this winners curse and to induce subscribers to
future IPOs, issuers underprice. The third explanation is by Allen and Faulhaber
(1989) and by Welch (1989). They argue that underpricing encourages subscribing
investors to see the quality of the IPO firm which later allows the firm to make
subsequent equity issues at a higher price. As such, these companies recoup some
of that underpricing.
The next three explanations suggest an underwriter monopsony power because
underwriters have significant control over the price at which the IPOs shares are
offered. Tinics (1988) insurance hypothesis argues that underpricing is like an
insurance policy protecting the underwriters and the issuing firm from lawsuits.
Chalk and Peavy (1987) suggest that underwriters might issue shares to preferred
clients but then recoup this favor by charging higher fees later for services to such
clients. Benveniste and Spindt (1989) argue that underwriters allow new issues to
be underpriced to encourage investors to subscribe to the IPO to fill the new issue.
Investors otherwise will simply wait until listing to purchase the shares.
Ruud (1993) suggests that underpricing may not be a deliberate decision prior
to the listing. She suggests that underwriters actually price support the issue after it
is listed. This is unlikely in Australia because price support activities by underwriters are illegal under the Corporations Law of Australia.
Except for Ruud (1993), all of the explanations broadly suggest that uncertainty, issue price and underpricing are related. However, it was Beatty and
Ritters (1986) paper that more formally argued that reducing the uncertainty
about an IPOs valuation reduces the need for underpricing. Since that study
researchers have found that lower underpricing is associated in IPO firms:
240
B. Dimovski
13
241
the shares (plus any options) on the first day of listing minus the issue price, the
result of which is then divided by the issue price. Closing prices were obtained
from the FinAnalysis database.
The regression model with underpricing return as the dependent variable is:
RETURN b0 b1ISSUEPRI b2LNTOTAL b3TIMETOLIST b4UWRITTEN
b5UOPTIONS b6INDEPACC b7SHOPTIONS b8POST2007 e
13:1
where all the variables are as defined previously, the bs are unknown
parameters to be estimated and e is assumed * N (0, r2).
The ISSUEPRI and LNTOTAL variables are commonly used in underpricing
studies. They are both expected to be negatively related to RETURN. In regard
issue price, Chalk and Peavy (1987) examined 649 US IPOs from 1975 to 1982.
They found that the group of IPOs priced at US$1 or less, had an underpricing
return (of 56.4 %) almost five times higher than for the next pricing group (of
11.95 %). Tinic (1988) also argued that the offering price per share is a proxy for
ex ante uncertainty.
In regard the size of the issue, Beatty and Ritter (1986) use data from 1028 U.S.
IPOs during 1977 to 1982 and argued that the size of the issue was a proxy for
ex ante uncertainty about the value of the firm. They confirm that smaller offerings
are more speculative and more highly underpriced than larger offerings. Chalk and
Peavy (1987) similarly report a negative relationship between underpricing and
size. Michaely and Shaw (1994), in their study of 947 IPOs in the United States
between 1984 and 1988 also identify that underpricing was lower in larger
companies. The study by Ibbotson, Sindelar and Ritter (1994) with U.S. IPO data
dating back to 1960 confirmed that on average, larger IPOs were less underpriced
than smaller IPOs. In the Australian context, How et al. (1995) in their study of
340 Australian industrial IPOs from 1980 to 1990, similarly found that larger firms
were significantly less underpriced.
The UWRITTEN and INDEPACC variables test whether the use of an
underwriter or a top 5 independent accountant respectively, is useful in reducing
the level of uncertainty about the IPOs value and hence its underpricing. Carter
and Manaster (1990) confirmed that lower first day returns were associated with
higher reputation underwriters (using U.S. data collected between 1979 and 1984).
They suggest this was because the higher reputation underwriters accepted lower
risk issues. Carter et al. (1998) reaffirm this lower underpricing, higher underwriter
reputation relationship in an expanded study of U.S. IPOs over 1979 to 1991.
Michaely and Shaw (1994) in their study of 947 U.S. IPOs also found that those
managed by higher reputation underwriters have lower initial returns. In Australia,
IPOs do not need to be underwritten to list, nor do underwriters have a reputation
index. The UWRITTEN variable does however test whether the simple fact of
being underwritten does indeed influence underpricing returns.
In regard the quality of the auditor/accountant, Beatty (1989) suggested that the
costs of hiring a prestigious auditor (defined then as one of the Big 8 accounting
242
B. Dimovski
and audit firmsPrice Waterhouse, Coopers and Lybrand, Ernst and Whinney,
Arthur Anderson, Arthur Young & Co., Peat Marwick Mitchell, Deloitte Haskins
and Sells and Touche Ross) were justified by having lower underpricing. Michaely
and Shaw (1995) followed with a study that concluded that IPOs associated with
more prestigious auditors (the Big 8 accounting/audit firms) were less risky
because those auditors sought to protect their reputation capital. They argued that
good firms were willing to pay the higher fees charged by the Big 8 firms, so that
the issuing companys quality was more accurately revealed to the market.
The Big 8 were reduced to a Big 5 by 2001 due to mergers amongst the Big 8.
The Big 5 were PwC (Pricewaterhouse Coopers), KPMG, Deloitte Touche
Tohmatsu, Ernst and Young and Arthur Anderson (and then to a Big 4 when Arthur
Anderson folded in 2002 following the Enron collapse). For the purposes of this
study, the Big 5 nominated accounting firms are utilized since they most adequately
reflect those used by energy IPOs during the period of the study.
The TIMETOLIST variable is expected to have a negative coefficient as in Lee
et al. (1996b). This suggests more highly underpriced issues are subscribed to,
more quickly.
The SHOPTIONS variable tests the Schultz (1993) hypothesis that these so
called package IPOs minimize agency costs so that if the firms forecasted
performance is not up to expectations, the future expected equity capital from
those options does not flow into the firm. As such, SHOPTIONS is expected to be
negatively related to RETURN. Interestingly, How (2000), in her study of
Australian mining IPOs over the period 1979 to 1990 did not find that the type of
IPO, whether packaged or share only, was significantly related to the level of
underpricing.
Dunbar (1995) using data on 182 U.S. IPOs that offered underwriter warrants
(options) found that underwriters compensated with warrants do not increase
underpricing. It is expected that if underwriters are willing to accept options
(UOPTIONS) to buy more shares they are likely to be more certain about the value
of the IPO before listing. As such, UOPTIONS is expected to be negatively related
to RETURN.
This study treats all the explanatory variables as exogenous. The issue price, the
size of the capital raising, whether the issue is underwritten (not the choice of a
specific underwriter), whether underwriter options are offered, whether a big 5
accountant is employed (not the choice of a specific big 5 accountant) and whether
share options are offered to subscribers are all identified in the prospectus and as
such, predetermined. Australian IPOs are generally fixed price IPOs meaning that
the price is known at the outset and since the issue takes around 57 days on
average from the date of the prospectus to the date of listing we would expect these
variables to be exogenous. Additionally, we also know whether the issue is before
the period of the global financial crisis.
Ljundqvist and Wilhelm (2002) and Lowry and Shu (2002) examine the
endogeneity of some right hand side variables in book build environments where
the issue price is not finalized until the investment order book is filled and the IPO
ready to list. The only variable that may cause some endogeneity concern in this
13
243
present study, is the time to list variable. However even with this variable, the
issue needs to fill first and then the stock lists, often 1014 days later. Locating a
suitable instrumental variable is likely to prove difficult but the time to list variable
is not significant in any case.
Two specific hypotheses might be tested. Given that a great many of these
energy IPOs are involved in exploration, we can formally suggest a first hypothesis
that Australian energy IPOs are more risky than industrial company IPOs and as
such should provide higher underpricing. Additionally POST2007 is a new
variable that is included to test if energy IPOs that raised equity in the GFC period
offered significantly higher underpricing. This can be formalized into a second
hypothesisthat the mean underpricing return for energy IPOs is significantly
higher after 2007.
13.4 Results
Table 13.1 reports the summary statistics for the data set. The mean underpricing
return was 22.0 % while the median was 6.5 %. Because this is a sample, a t test
testing if the true mean underpricing return could be zero, is undertaken. The
resultant t-statistic is 5.64 (much larger than 2 for a 95 % confidence level)
suggesting that the mean underpricing return of Australian energy IPOs is statistically significantly greater than zero. It is worth restating, however, that these
mean and median return measures include the returns theoretically able to be made
from selling the newly listed shares and the subscriber options (if any) for each of
the IPOs, at the closing price of the shares and the options on the first day of
listing. This method is consistent with Schultz (1993), Jain (1994) and How and
Howe (2001).
One IPO was underpriced 295 % while one was overpriced by 35 %. Overpricing occurs when the subscription price paid by investors is higher than the
closing price of the stock on the first day of trading. While overpricing is not a
happy occasion for investors and may be considered an uncommon occurrence, it
was not at all uncommon amongst these energy IPO offerings. The company still
lists on the stock exchange and given the amount of capital raised is what it sought
to raise, the company can still do what it proposed to do in its prospectus. A total
of 47 of the 158 energy IPO observations were overpriced, 11 showed a 0 % return
and 100 were underpriced. As such, underpricing occurred in nearly 2/3rds of the
sample. The standard deviation of returns was 49.0 % reflecting quite a considerable measure of dispersion.
The issue price ranged from 20 cents to $5.96 with the median being 20 cents.
In Australia, since 1 July 1998 shares in companies are simply offered at an
issue price, no longer using the terms of par or nominal values and no
longer allowing accounting entities to record a share premium reserve account
reflecting amounts paid for the shares over and above the old par or nominal value.
244
B. Dimovski
Table 13.1 Summary statistics for the underpricing of energy IPOs in Australia Jan 1994 to
December 2010
Variable
158 observations
Median
Std Dev
Min
Max
Panel A
Mean
Return
Issue Price ($)
Total Raised ($m)
Time to list (days)
Underwritten (0 or 1)
Uoptions (0 or 1)
Indep Account (0 or 1)
Share options (0 or 1)
0.220
0.401
16.690
57.278
0.342
0.108
0.336
0.304
0.065
0.200
7.000
50.000
0.000
0.000
0.000
0.000
0.490
0.581
3.826
30.663
0.476
0.311
0.474
0.461
-0.350
0.200
0.300
3.000
0.000
0.000
0.000
0.000
2.950
5.960
400.000
282.000
1.000
1.000
1.000
1.000
Panel B
Return
Issue Price ($)
Total Raised ($m)
Time to list (days)
Underwritten (0 or 1)
Uoptions (0 or 1)
Indep Account (0 or 1)
Share options (0 or 1)
3 outliers removed
0.175
0.400
16.842
57.458
0.342
0.108
0.342
0.310
0.050
0.200
7.000
50.000
0.000
0.000
0.000
0.000
0.371
0.586
3.861
30.906
0.476
0.314
0.476
0.464
-0.350
0.200
0.300
3.000
0.000
0.000
0.000
0.000
1.475
5.960
400.000
282.000
1.000
1.000
1.000
1.000
Any old (pre 1998) share premium reserve accounts were transferred to the share
capital account.
This study uses issue price to mean the price at which the new IPO shares are
offered to the public. The vast majority of these energy IPO shares are offered at
20 and 25 cents (over 100 of the 158), which is a typical resource stock issue price in
Australia. The $5.96 price is a highly unusual issue price for a resource stock but that
is the price that Aston Resources Limited priced the shares in their IPO in their $400
million IPO capital raising. This also happened to be the highest individual capital
raising in this data set and was managed by Goldman Sachs, Macquarie Capital
Advisors, Credit Suisse and BKK Partners. Aston Resources Ltd happened to be
overpriced by 4.7 %. Other 100 or more million dollar capital raisings include
Novus Petroleum Ltd, Roc Oil Co Ltd, Excel Coal Ltd, and ERM Power Ltd. Their
underpricing was 10, 10, -2 (overpriced) and 3.4 % respectively.
The mean average capital raising by these energy IPOs was $16.69 million with
the smallest being $300,000 to the largest being $400 million. Around 86 % of the
energy IPO entities raised $5 million or less. Around 34 % of the IPOs (54 in
number) used underwriters to guarantee the success of the capital raising and
around 34 % of the firms (53 in number) used a top 5 accounting firm. They were
not the same firms. Only 10.8 % of the IPOs (17 in number) offered options to
underwriters to subscribe for more shares while 30.4 % (48 in number) offered
options to subscribers. There were only 24 IPOs that sought IPO equity capital
during 2008 to 2010 (with only six energy IPO offerings in 2008).
13
245
Table 13.2 Regression results for the underpricing of energy IPOs in Australia Jan 1994 to Dec
2010
Outliers Removed #
158 IPOs
155 IPOs
C
ISSUEPRI
LNTOTAL
TIMETOLIST
UWRITTEN
UOPTIONS
INDEPACC
SHOPTIONS
POST2007
R Squared
Adj R Squared
Jarque Bera
White Test
Reset Test
Coef.
Pr.
Coef.
Pr.
1.698
0.076
-0.078
-0.002
-0.150
0.158
-0.170
-0.150
-0.153
0.103
0.055
622.601
19.208
2.582
0.038
0.412
0.136
0.185
0.153
0.269
0.064a
0.086a
0.171
1.470
0.086
-0.072
-0.001
-0.192
0.225
-0.086
-0.068
-0.077
0.120
0.072
83.938
50.017
3.142
0.005
0.142
0.035b
0.540
0.001c
0.032b
0.144
0.302
0.326
0.000
0.995
0.070
0.000
0.091
0.055
Panel B of Table 13.1 excludes the outliers. The mean underpricing return is
now 17.5 % for the remaining 155 observations and the median is 5.0 %. The
standard deviation of returns falls to 37.1 %. The summary statistics for the other
variables remain broadly the same.
Tables 13.2, 13.3 and 13.4 report the ordinary least squares (OLS) regression
results. There were three observations that were over 3.5 standard deviations from
the mean return. These outlier observations were removed from the model and
modified regression results reported. This identification of outliers over 3.5 standard deviations is consistent with How (2000). These three observations showed a
224, 240 and 295 % first day return. Interestingly, there have been more highly
underpriced IPOs than these amongst other Australian IPOsDiamond Rose and
Western Diamond Corporation reported 545 and 500 % first day returns respectively while Australian industrial companies, CDS Technologies and Advanced
Engine Components reported 370 and 340 % first day returns respectively. Even
excluding the three energy IPO outliers in this data set, the mean average
underpricing is still over 17.5 % for the remaining sample of 155 energy IPOs.
A variety of standard regression diagnostics are reported. In testing for
non-normal errors, a JarqueBera statistic is reported, although the relatively large
sample size suggests normality is unlikely to be problematic. In testing for
heteroskedasticity, a White (1980) test is applied and robust results reported where
necessary using White corrected standard errors and the resultant p-values. The
affected heteroskedasticity-consistent affected column of results are identified with
246
B. Dimovski
Table 13.3 Regression results for the underpricing of energy IPOs in Australia Jan 1994 to Dec
2010fewer variables
Outliers Removed #
158 IPOs
155 IPOs
C
ISSUEPRI
TIMETOLIST
UWRITTEN
UOPTIONS
INDEPACC
SHOPTIONS
POST2007
R Squared
Adj R Squared
JarqueBera
White Test
Reset Test
Coef.
Pr.
Coef.
Pr.
0.489
0.001
-0.002
-0.192
0.158
-0.180
-0.136
-0.134
0.090
0.047
612.696
15.983
3.414
0.000
0.998
0.193
0.055a
0.274
0.050b
0.119
0.229
0.354
0.016
-0.001
-0.231
0.223
-0.096
-0.056
-0.060
0.100
0.057
81.155
4.290
7.349
0.010
0.667
0.557
0.000c
0.034b
0.101
0.403
0.439
0.000
0.975
0.040
0.000
0.015
0.013
Table 13.4 Regression results for the underpricing of energy IPOs in Australia Jan 1994 to Dec
2010fewer variables
Outliers Removed #
158 IPOs
155 IPOs
C
LNTOTAL
TIMETOLIST
UWRITTEN
UOPTIONS
INDEPACC
SHOPTIONS
POST2007
R Squared
Adj R Squared
Jarque Bera
White Test
Reset Test
a
Coef.
Pr.
Coef.
Pr.
1.352
-0.055
-0.002
-0.145
0.139
-0.153
-0.149
-0.132
0.102
439.005
616.355
18.001
0.3.237
0.053
0.212
0.149
0.165
0.326
0.087a
0.088a
0.224
1.077
-0.046
-0.001
-0.185
0.202
-0.066
-0.067
-0.054
0.111
0.069
89.463
50.626
5.364
0.012
0.073a
0.497
0.001c
0.053a
0.256
0.309
0.471
0.000
0.944
0.023
0.000
0.007
0.012
13
247
the hash (#) symbol at the top of the column in each of the three regression model
tables. The econometric software used to compute these robust results was EViews 7.
In testing for omitted variables or model misspecification, a Ramsey Reset test is
applied and reported. While the Durbin-Watson statistic is a useful diagnostic test
when an independent variable is a lagged value of the dependent variable, it is not
required here. This data set does not need to be ordered and the error term is not
correlated across observations; the independent and dependent variables are distinct.
For the overall model in Table 13.2, the results of the regression analysis
suggest that the LNTOTAL, UWRITTEN and UOPTIONS variables have
explanatory power with regard to the amount of underpricing return. The
ISSUEPRI and LNTOTMIL variables are however fairly highly correlated at
0.646 and multicollinearity may be a problem in our analysis hence the models are
run again with either one of the ISSUEPRI (in Table 13.3) or LNTOTAL (in
Table 13.4) variable.
Table 13.3 is without only the LNTOTAL variable and reports the UWRITTEN
and UOPTIONs variables as useful. Table 13.4 is without only the ISSUEPRI
variable and reports the UWRITTEN, OPTIONS and LNTOTAL variables as
useful. It appears that those Australian energy IPOs that are underwritten, benefit
substantially in terms of lower underpricing by about 20 %. What is interesting,
however, is that those energy company IPOs that offer options to the underwriters
appear to counteract the lower underpricing by about 20 %. There is also some
evidence to suggest that the greater the amount of capital sought, the lower the
underpricing.
13.5 Conclusion
This study examined 158 energy IPOs in Australia for the period January 1994 to
December 2010. What it found is that the mean underpricing return for these IPOs
is 22.0 % and statistically significant. The overall results of this study support the
findings of previous studies in that, IPOs on average, underprice, including energy
IPOs. While the implications suggest that investors will theoretically earn profits
by investing in these IPOs and selling on the first day of listing, recall that in
47 cases investors paid more by subscribing for shares through the prospectus than
they could buy the shares for at the closing price on the first day and in 11 cases,
their underpricing return was zero. Clearly, not all subscribers are going to be
making substantial first day returns on every issue and as Rock (1986) suggested,
nor would every subscriber be able to invest an equivalent amount into every issue
so as to take advantage of these averages. Also clear is that pricing new issues is
not easy and precise.
The results are broadly in line with recent Australian resource and industrial
company IPO studies [Brailsford et al. (2001), Dimovski and Brooks (2004) and
Dimovski et al. (2011)] and broadly in line with other international industrial
company IPO studies summarized in Ritter (2003) and Loughran et al. (1994)
248
B. Dimovski
updated to Nov 8, 2011. To the authors knowledge, however, this is the first large
sample energy IPO underpricing study reported.
The model used to investigate variables that might help explain the level of
underpricing in this industry sector is also particularly useful. An important finding
in the study for new issuers, underwriters and subscribing investors is that those
energy IPO firms that used underwriters had substantially lower underpricing, but
that the use of underwriter options increases underpricing. While some US studies
have suggested that the use of higher reputation underwriters is negatively related
to underpricing, just the use of underwriters having a negative influence on
underpricing is an interesting and useful finding. The UOPTIONS variable is
unexpectedly positive. This is contrary to the initial hypothesis and to Dunbars
(1995) US evidence, the allocation of options to underwriters appears to increase
the underpricing return amongst energy IPOs.
The other finding that larger issues are likely to have lower underpricing is
consistent with prior industrial company IPO studies. It is also interesting to note
that those energy company IPOs that sought to list after the global financial crisis
do not appear to have offered, on average, significantly different underpricing
returns to their investors compared to those energy IPOs that raised capital prior to
this global financial crisis period.
The underpricing returns made by subscribers to energy company IPOs, from
the issuing companys viewpoint, can be considered as an indirect cost to the
company itself. That is, the IPO entity has foregone raising the capital that the
subscribing investors have theoretically profited from (given the closing price on
the first day of trading). The expression sometimes used is money left on the
table by the IPO.
In terms of future research, it would be useful to consider factors influencing the
direct costs of raising the IPO equity capitalthe costs of underwriting,
accounting, legal fees and other expenses. This strand of research could follow Lee
et al. (1996a) with their broad US study on the costs of raising capital.
A second useful future research area would be to examine the longer-term
return performance of these energy company IPOs following their listing. It could
investigate the magnitude of returns an investor may have made from the IPO
investment in the longer term after having made (on average) initial day underpricing returns. Alternatively, if the investor had not received a sufficient allocation (or any allocation at all), in the IPO, whether an investment on the first day is
an appropriate and profitable strategy. Curiously, the consistent broad observation
(around the world) regarding IPOs is that their relative returns compared to the
market or matching firms (in size and industry) in the longer term is poor. The
energy sector, however, has not been specifically reported. A range of early
evidence on the longer-term return performance of IPOs generally is discussed in
Loughran et al. (1994) and more recently in Bradley et al. (2008).
249
References
Allen, F., & Faulhaber, G. R. (1989). Signaling by underpricing in the IPO market. Journal of
Financial Economics, 23, 303324.
ASX (2012). An international centre for resources capital (n.d.). Retrieved March 14, 2012 from
http://www.asx.com.au/images/resources/resources_brochure.pdf.
Australian stock exchange (2010). ASX: An international centre for resources capital, pp. 18.
Baron, D. P. (1982). A model of the demand for investment banking advising and distribution
services for new issues. Journal of Finance, 37(4), 955976.
Beatty, R. P. (1989). Auditor reputation and the pricing of initial public offerings. The Accounting
Review, 64, 693709.
Beatty, R. P., & Ritter, J. R. (1986). Investment banking, reputation, and the under pricing of
initial public offerings. Journal of Financial Economics, 15(12), 213232.
Benveniste, L. M., & Spindt, P. A. (1989). How investment bankers determine the offer price and
allocation of new issues. Journal of Financial Economics, 24, 343362.
Bradley, D., Chan, K., Kim, J., & Singh, A. (2008). Are there long run implications of analyst
coverage for IPOs? Journal of Banking and Finance, 32, 11201132.
Brailsford, T. J., Heaney, R. A., & Shi, J. (2001). The cyclical behaviour of the IPO market in
Australia. Accounting Research Journal, 14(1), 1734.
Brau, J. C., & Fawcett, S. E. (2006). Initial public offerings: An analysis of theory and practice. ,
61(1), 399436.
Carter, R. B., Dark, F. H., & Singh, A. K. (1998). Underwriter reputation, initial returns, and the
long run performance of IPO stocks. Journal of Finance, 53, 285311.
Carter, R., & Manaster, S. (1990). Initial public offerings and underwriter reputation. Journal of
Finance, 45(4), 10451067.
Chalk, A. J., &Peavy, J. W. III (1987). Initial public offerings: Daily returns, offering types and
the price effect. Financial Analysts Journal, 43(5), 6569.
Dimovski, W. (2011). The pricing of infrastructure initial public offerings: Evidence from
Australia. Journal of Property Research, 28(1), 314.
Dimovski, W., & Brooks, R. (2004). Initial public offerings in Australia 1994 to 1999. Recent
Evidence of Underpricing and Underperformance, Review of Quantitative Finance and
Accounting, 22, 179198.
Dimovski, W., & Brooks, R. (2006). The pricing of property trust IPOs in Australia. Journal of
Real Estate Finance and Economics, 32(2), 185199.
Dimovski, W., & Brooks, R. (2008). The underpricing of gold mining initial public offerings.
Research in International Business and Finance, 22, 116.
Dimovski, W., Philavanh, S. and R. Brooks, R. (2011). Underwriter reputation and underpricing:
Evidence from the Australian IPO market, review of quantitative finance and accounting, 37,
409-426.
Dunbar, C. G. (1995). The use of warrants as underwriter compensation in initial public offerings.
Journal of Financial Economics, 38, 5978.
How, J. C. Y. (2000). The initial and long run performances of mining IPOs in Australia.
Australian Journal of Management, 25, 95118.
How, J. C. Y., & Howe, J. S. (2001). Warrants in initial public offerings: Empirical evidence.
Journal of Business, 74(3), 433458.
How, J. C. Y., Izan, H. Y., & Monroe, G. S. (1995). Differential information and the underpricing
of initial public offerings: Australian evidence. Accounting and Finance, 35, 87105.
Ibbotson, R. G., Sindelar, J. L., & Ritter, J. R. (1994). The markets problems with the pricing of
initial public offerings. Journal of Applied Corporate Finance, 7(1), 6674.
Jain, B. A. (1994). The underpricing of unit initial public offerings. Quarterly Reviews of
Economics and Finance, 34, 309325.
James, C., & Wier, P. (1990). Borrowing relationships. Intermediation and the cost of Issuing
public securities, Journal of Financial Economics, 29, 149172.
250
B. Dimovski
Koop, G., & Li, K. (2001). The valuation of IPO and SEO firms. Journal of Empirical Finance, 8,
375401.
Lee, I., Lochhead, S., Ritter, J., & Zhao, Q. (1996a). The costs of raising capital. Journal of
Financial Research, 19(1), 5974.
Lee, P. J., Taylor, S. L., & Walter, T. S. (1996b). Australian IPO pricing in the short and long run.
Journal of Banking & Finance, 20(7), 11891210.
Liu, X., & Ritter, J. R. (2011). Local underwriter oligopolies and IPO underpricing. Journal of
Financial Economics, 102, 579601.
Ljundqvist, A. P., & Wilhelm, W. J. (2002). IPO allocations: Discriminatory or discretionary?
Journal of Financial Economics, 65, 167201.
Loughran, T., Ritter, J.R. and Rydqvist, K. (1994) Initial public offerings: International insights.
Pacific-Basin finance Journal, 2(23), 165199. Updated November 8, 2011.
Lowry, M., & Shu, S. (2002). Litigation risk and IPO underpricing. Journal of Financial
Economics, 65, 305339.
Michaely, R., & Shaw, W. H. (1994). The pricing of initial public offerings: Tests of adverseselection and signaling theories. Review of Financial Studies, 7(2), 279319.
Michaely, R., & Shaw, W. H. (1995). Does the choice of auditor convey quality in an initial
public offering. Financial Management, 24(4), 1530.
Ritter, J. R. (2003). Differences between European and American IPO markets. European
Financial Management, 9, 421434.
Rock, K. (1986). Why new issues are underpriced. Journal of Financial Economics, 15(12),
187212.
Ruud, J. S. (1993). Underwriter price support and the IPO under pricing puzzle. Journal of
Financial Economics, 34(2), 135151.
Schultz, P. (1993). Unit initial public offerings: A form of staged financing. Journal of Financial
Economics, 34, 199229.
Tinic, S. M. (1988). Anatomy of initial public offerings of common stock. Journal of Finance,
43(4), 789822.
Valentine, T., & Gordon, C. (2009). Economics in focus: The global financial crisis. Pearson
Education Australia.
Welch, I. (1989). Seasoned offerings, imitation costs, and the underpricing of initial public
offerings. Journal of Finance, 44, 421450.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity. Econometrica, 48(4), 817838.
World federation of exchanges (2009). Market highlights (2009) Retrieved March 14, 2012, from
http://www.google.com.au/search?q=WFE+market+highlights+2009&ie=utf-8&oe=utf8&aq=t&rls=org.mozilla:en-US:official&client=firefox-a.
World federation of exchanges (2010). Market highlights (2010) Retrieved March 14, 2012, from
http://www.scribd.com/doc/53724659/2010-WFE-Market-Highlights.