OIL BOOM AND BUST: What Determines Crude Oil Prices?

OIL BOOM AND BUST
What Determines Crude Oil Prices?
1988 – 2015 Study
Author
Andrean Anzhelov
Rashkov
Supervisors
Nina Lange
David Skovmand
18 March 2016
Master Thesis
MSc Economics and Business Administration (Cand. Merc.)
Accounting, Strategy and Control
Characters (with spaces): 162,333
Pages (bibliography, appendix excluded): 71.35

1OIL BOOM AND BUST: WHAT DETERMINES CRUDE OIL PRICES?
ABSTRACT
Crude oil constitutes one of the most important commodities which affect global economic
development. Variable nature of the oil marketplace represents a problem for policy makers,
consumers and businesses alike, even more so in the context of recent surges in volatility. Empirical
literature on the topic ascertains no factor to be single-handedly held accountable for unprecedented
levels of crude price fluctuations. Analysts’ opinions diverge and it remains unclear whether oil
price variances are driven by fundamental forces of supply and demand, by the OPEC cartel and
whether financial factors of growing importance play their part in commodity markets as well.
Current study has as its objectives to address such knowledge discrepancies and to uncover
contemporary factors influencing crude prices worldwide. The research examines Brent index price
dynamics over a period of almost three decades from 1988 to 2015 in an attempt to identify strong
influencers over time. Attention is focused on jointly analyzing factors previously reviewed in
isolation: traditional supply and demand fundamentals, significant geopolitical and economic
events, exchange rate fluctuations and measures of financial derivatives’ speculation in oil futures.
The analysis employs econometric technique Bayesian Model Averaging (BMA) as an effective
way of addressing model choice uncertainties of conventional empirical approaches. BMA results’
validity is then verified for its capacity to forecast observed historic data. The study concludes with
a relevant set of recommendations on effective risk management for market participants and policy
proposal for elected officials who aim to minimize negative economic effects of crude oil volatility.
Keywords:
Bayesian model averaging, Model uncertainty analysis, Crude Oil, Brent, Price Determinants,
Fundamental factors, Supply and Demand, OPEC role, Geopolitical events, Financial speculation

Table of Contents
1 Introduction ..............................................................................................................4
1.1 Research Objectives and Structure.............................................................................................6
1.2 Delimitations ..............................................................................................................................7
2 Literature Review.....................................................................................................8
2.1 Demand Factors..........................................................................................................................8
2.1.1 Oil Consumption.............................................................................................................................. 8
2.1.2 Monetary Policy............................................................................................................................... 9
2.1.3 Income Elasticity ........................................................................................................................... 10
2.1.4 Price Elasticity ............................................................................................................................... 11
2.2 Supply Factors..........................................................................................................................12
2.2.1 Hotelling’s Rule............................................................................................................................. 12
2.2.2 Peak Oil Hypothesis....................................................................................................................... 12
2.2.3 Oil Stocks ....................................................................................................................................... 13
2.2.4 Refining.......................................................................................................................................... 14
2.2.5 Oil Supply Inelasticity ................................................................................................................... 15
2.3 The Role of OPEC....................................................................................................................15
2.4 Financial Speculation ...............................................................................................................18
2.5 US Dollar Value.......................................................................................................................20
2.6 Geopolitical Factors .................................................................................................................21
2.7 Conclusions ..............................................................................................................................22
3 Methodology............................................................................................................23
3.1 Overview ..................................................................................................................................23
3.2 Philosophical Approach ...........................................................................................................23
3.3 Research Approach...................................................................................................................24
3.4 Research Strategy.....................................................................................................................24
3.5 Method of Analysis: Bayesian Model Averaging ....................................................................24
3.5.1 General overview ........................................................................................................................... 24
3.5.2 The Model ...................................................................................................................................... 25
3.5.3 Model Space Specification ............................................................................................................ 26
3.6 Method of Analysis: Black-Scholes Model..............................................................................28
3.7 Determinants and Hypotheses..................................................................................................29

3.8 Data Collection and Preparation...............................................................................................32
3.9 Reliability and Validity ............................................................................................................35
3.10 Data Limitations.....................................................................................................................36
4. Results and Analysis..............................................................................................40
4.1 Results from the Overall Model...............................................................................................40
4.1.1 General Considerations................................................................................................................. 40
4.1.2 Predictor Significance ................................................................................................................... 41
4.1.3 Coefficient Sign and Size of Effect ............................................................................................... 42
4.1.4 Coefficient Robustness .................................................................................................................. 44
4.1.5 Model Characteristics.................................................................................................................... 45
4.2 Results by Periods ....................................................................................................................46
4.2.1 Predictor Significance ................................................................................................................... 47
4.2.2 Coefficient Sign and Size of Effect ............................................................................................... 49
4.3 Discussion ................................................................................................................................52
4.4 Forecasting Properties..............................................................................................................56
5 Recommendations ..................................................................................................61
5.1 Risk Management Techniques .................................................................................................61
5.1.1 Full Exposure to Market Volatility ............................................................................................... 62
5.1.2 Hedging with a Straddle................................................................................................................ 62
5.1.3 Hedging with a Conditional Purchase of Options........................................................................ 63
5.1.4 Unconditional Hedging with Options ........................................................................................... 64
5.1.5 Speculation with Options............................................................................................................... 64
5.1.6 Overview of Risk Management Strategies .................................................................................... 65
5.2. Recommendations to Policy Makers.......................................................................................66
5.3 Areas for Future Research........................................................................................................68
6 Conclusions .............................................................................................................69
Bibliography...............................................................................................................71
Appendix .......................................................................................................................I

1 Introduction
Crude oil constitutes one of the most important commodities which affect global economic
development. Notably, world’s historical experiences from the Yom Kippur War in 1973 and the
Iranian Revolution in 1978, which resulted in first and second energy crisis respectively, have
incontestably demonstrated to what degree negative supply shocks (associated with significant
increments in oil prices) can derail market economies and cause prolonged episodes of
“stagflation”– a combination of economic recession and surging inflation. In stark contrast, the era
of solid economic growth during 1980s and 1990s coincided with a prolonged period of
uninterrupted commodity supply as well as stable and relatively low commodity prices.
A historical review of crude oil price dynamism in the new millennium uncovers potential
justification for the re-emergence of volatility. At first, the price per barrel of crude1
increased
almost twelve-fold from $11.11 in January 1999 to an all-time peak of $132.72 in July 2008. Such
an unprecedented increase in prices was quickly followed by a relentless 69.9% decline to $39.95 in
late 2008. The period between 2008 and 2011 was characterised with a significant recovery – crude
oil prices managed to reach three-digit values once again. Nonetheless, declines resumed in mid-
2014 and a barrel of crude is currently traded approximately 70% below 2011 peak levels (Energy
Information Administration, 2016a). It is worth noting that the recent price declines have developed
thus far without any prolific global economic downturns taking place as was the case in year 2008.
Analysing crude oil prices’ volatility is of paramount importance as any misinterpretation has the
capacity to induce substantial financial losses and poses multiple contingencies for a broad range of
economic agents: private and industrial consumers, producers and policy makers alike. Price
fluctuations play a key role not only in stimulating consumer price inflation, but they also generate
a higher degree of uncertainty in investment decisions made by organisations as well as tax revenue
planning conducted by national governments. These considerations are particularly relevant for oil
importing countries, for which a number of theoretical and empirical studies has confirmed the
negative role of price volatility through transmission channels such as production costs (supply
factors) and income transfers (demand factors) (Hamilton, 2003; Jimenez-Rodríguez, 2009). On the
contrary, scientific evidence has found the volatility of crude oil prices to have comparatively
mixed macro-economic effect in oil exporting countries (Bjornland, 2000; Abeysinghe, 2001).
1. Price references are made to Brent crude oil prices, more on choice of crude oil index in 1.3 Delimitations.

From a fundamental perspective, oil analysts are inclined to rationalize price fluctuations through
shifts in demand and supply. Often, however it is difficult to justify the magnitude of price swings
with traditionally slowly evolving demand and supply factors. Some of the newer propositions in
the field address these fluctuations with an increased responsiveness to changes in the global
business cycle. As a consequence, a marginal change in demand-related factors is likely to trigger a
significant move in spot pricings, thereby balancing crude oil markets and compensating for the
generally low responsiveness of demand and supply volumes in the short term (Lipsky, 2009).
In the early 2000s there has been a growing demand for commodity futures contracts as a viable
alternative asset classes to traditional investment instruments such as stocks and bonds. An interest
was expressed by both retail and institutional investors. Cheng and Xiong (2013) suggest that a total
of $200 billion were invested in commodity futures between 2000 and 2008. Both of these
processes combined have resulted in trends of increased market participation by predominantly
financial investors and increased levels of capitalisation (hence the term “financialisation”). These
new entrants in the oil market practice what is known as momentum-based investment strategies
(capturing gains on the continuance of existing trends), thus rendering the commodity field
susceptible to a higher degree of herding behaviour. It can then be argued that crude oil is purchased
even more in uptrends and sold even further once its price begins to decline. As a result, such
speculative behaviours have the potential to create short-term demand shocks, which increase price-
cycle extremities and translate into spot-market assessments far beyond (or below) crude oil’s
equilibrium price based on fundamental factors (Lipsky, 2009; Beidas-Strom and Pescatory, 2014).
The ascertained magnitude of crude oil price’s volatility, the challenges it presents to producers,
consumers and governments alike, combined with the existing dissimilar academic views on newly
postulated market phenomena have motivated the present research. It can be argued that the recent
trends toward crude oil market financialisation have challenged the validity of traditional demand
and supply factors and their explanatory power of price movements. Research in the field, however,
remains scarce and the few existing studies deliver inconclusive evidence for the destabilising role
of financial speculation (Fattouh et al., 2012; Büyükşahin and Robe, 2012; Kilian and Lee, 2013).
The remaining subsections of Chapter One focus on developing a more detailed research direction
as well as on presenting the theoretical and methodological delimitations of this paper.

1.1 Research Objectives and Structure
Based on considerations presented in section one, the guiding objectives of the current study are,
first, to reassess the roles of conventional fundamental oil-price determinants and, second, to
uncover and investigate any potential explanatory effects of momentum-based and speculative
market strategies. Put into this context, the principal research question is formulated as follows:
How can the significant fluctuations in crude oil prices be explained?
An exhaustive answer to this question requires a list of aims and objectives, which will thence serve
as a guiding framework for this research. All aims and objectives are formulated as questions and
are grouped in two categories: theoretical and empirical. There is a single theoretical question:
1. How can the existing theoretical and empirical literature be applied, in order
to develop a model of factors explaining fluctuations of crude oil market prices?
The second research category of aims and objectives is of empirical nature and reads as follows:
1. How do the roles and significance of fundamental oil-price determinants change over time?
2. How significant is the role of financial speculation?
3. How significant is the role of geopolitical and economic shocks?
4. How can the results be used to suggest relevant risk management tools
to policy makers and businesses, addressing the negative effects of oil price’s volatility?
From a structural perspective, this research applies strict academic framework directed by
abovementioned aims and objectives. Chapter Two is comprised of a literature review, in which
previous theoretical and empirical studies are scrutinized in search of factors affecting crude oil
prices historically. Based on its findings, a conceptual model and a list of relevant hypotheses for
testing are generated. Chapter Three, the methodological chapter, proceeds to clarify all elements
related to the applied research framework, data collection and interpretation. In Chapter Four, once
data are fully processed, a comprehensive list of findings is presented and further analysed with the
aim of validating the list of hypotheses. The model is then tested for its forecasting ability and
Chapter Five draws a relevant set of recommendations for consumers, producers and policy makers
alike. The study concludes with a summary section including reference points for further research.

1.2 Delimitations
The role of this delimitations chapter is to narrow the scope, define the boundaries and explicate
themes which have been intentionally omitted or not addressed. The current analysis of crude oil
price determinants has its delimitations primarily pertaining to the choice of research methodology.
To begin with, data pertaining to macroeconomic variables, such as gross domestic product (GDP),
inflation rate, etc. only include statistics from several large economies – namely the United States
(US), the Organization for Economic Cooperation and Development (OECD) countries (with a
combined measure) and China. These selected economies account for most of the global economic
activity and set the general sentiment for trends determining world’s demand for crude oil. Smaller
economies are thus inferred to have only have marginal influence on commodity prices and their
respective data are deliberately omitted. Furthermore, such a delimitation corners data availability
and reliability concerns – a number of the smaller countries have not published economic data
stretching back to 1980s or have annexed for uncertain consistency of methods used.
The choice of time period deliberately starts in late 1980s. A historical period of twenty-eight years,
spanning from 1988 to 2015 (both years included), is considered representative for the crude oil
markets and therefore relevant for identifying underlying trends and assessing the role of the crude
oil price determinants. Further research has found earlier data to often be incomplete or non-
available. Including low-quality measures may compromise the overall data reliability; hence
eliminating earlier periods strengthens the present study without undermining its outcomes validity.
Although crude oil is classified as a commodity, its quality and location play an important role for
transactions. In order to distinguish and thereby value different crude oil, markets have primarily
recognized three main benchmarks – Brent, Western Texas Intermediate (WTI) and Dubai/Oman.
The main benchmarks exhibit high correlation over the longer span of time; however prices have
significantly diverged within defined periods. Such divergences are likely to present deliberate
implications of the study’s results depending on which crude benchmark is selected.
The relative role of the WTI was most significant until 1980s; however, in years that followed the
so called “Energy Policy and Conservation Act of 1975” (effectively a trade barrier to U.S. oil
exports), WTI’s appeal lessened as it traded at a discount to Brent contracts (Opportune LLP, 2013).
Brent crude oil type, which designates crude oil extracted from fields located in the North Sea,
serves as a foundation for Brent index as price benchmark. More recently, according to reporting

agencies and oil producers, approximately two-thirds of the globally traded crude oil is priced off
the Brent benchmark nowadays, including West African, Mediterranean and Southeast Asia crudes
(Intercontinental Exchange, 2013, 2016; Energy and Capital, 2016). The global dominant role of the
Brent index (illustrated in Figure 1, Appendix) motivates its selection as crude oil price measure.
2 Literature Review
Chapter Two examines the existing theoretical and empirical literature for relevant oil-price
determinants. The secondary data evidence is extensive and readily available, allowing or an in-
depth analysis and design of a conceptual model of crude oil price determinants. The literature
review is organised in distinct sub-sections, which scrutinise a number of possible price
determinants, starting with three traditional fundamental categories – demand, supply and market
structure (the latter expressed through the role of the Organization of the Petroleum Exporting
Countries – OPEC). Supplementary to these, a new set of determinants, which attracted research
attention more recently, are added to the analysis. These include: the role of financial speculation as
a driver of oil prices; the fluctuations in the value of the US dollar – a standard currency
denominator for all commodity prices. Furthermore, the literature review includes a brief historical
analysis of multiple geopolitical and historical events which allegedly influenced commodity
markets worldwide. For readers’ convenience, all findings of the theoretical review are presented in
Section 2.7; in addition, Table 1 (Section 3.7) lists all relevant determinants included in the
conceptual model and their respective hypotheses which are empirically tested for in Chapter Four.
2.1 Demand Factors
2.1.1 Oil Consumption
The role of crude oil consumption as price determinant has been extensively reviewed in recent
academic literature (Breitenfellner et al., 2009; Hamilton, 2009; Kilian, 2009; Kilian and Hicks,
2009). Oil demand is prevailingly found to constitute a significant driving force in commodity
pricing. Breitenfellner et al. (2009) even argue that the growth in demand should be considered as a
more important factor than the size of the actual demand per se. Kilian and Hicks (2009) find that
emerging markets’ strong economic growth in the period 2003 – 2008 remained persistently higher
than market expectations and drove an unexpected surge in their demand for industrial commodities
(see Figure 2). In this context, significant oil consumption shocks in emerging economies such as
China, India and Latin American countries, should have translated into heightened crude oil prices.

Figure 2 – Asia’s oil demand growth (2000 – 2009; annual change)
Source: Petroleum Resources Branch (2010)
Breitenfellner et al. (2009) verify this assumption, illustrating that rapid demand increases in
emerging markets contributed to the surge of oil prices. Furthermore, they find the subsequent price
crash to be triggered by a sudden drop in oil demand. Nevertheless, the relationship between total
oil demand dynamics and prices is less conclusive in the case of the developed countries. Evidence
is only conclusive in the case of the US, where most of the economic recessions and associated
drops in oil demand are found to highly correlate with commodity price declines (Hamilton, 2009).
2.1.2 Monetary Policy
Barsky and Kilian (2002, 2004) explore the role of the monetary policy on commodity prices and
find it to both impact and respond to oil prices via two main transmission channels: inflation of
consumer prices and economic growth expectations. Financial literature, however, largely regards
these two channels as negatively correlated – an expansionary monetary policy leading to high
inflation is likely to have negative impact on economic growth and vice versa (Lucas, 2003; Issing,
2001). Building on previous research in the field, Frankel (2007) identifies the effects of monetary
policy on oil supply and demand factors. He verifies that real interest rates influence the opportunity
cost of carrying inventories. As a consequence of the positive income effect and decreased
inventory cost, low real interest rates facilitate demand growth. Conversely, lower rates exert a
negative pressure on supply, as the cost of holding inventories in the ground or in tanks diminishes.
Lastly, Frankel (2007) finds low interest rates to encourage commodity market speculation, because
of the lessened appeal of treasury bills’ returns, which result in higher commodity prices.
All of these theoretical considerations have been largely supported throughout the oil boom of
2003–2008. Researches by Krichene (2006) and Hamilton (2009) identify expansionary monetary
policy of the US as the main determinant for the prolonged upward price movement in mid-2000s.

Krichene (2006) empirically addresses a key point in the interdependence of monetary policy and
oil prices. Remarkably, during demand shocks, monetary policy is found to lead oil-price
adjustments – an increase of real interest rates will contain the upturn of prices caused by
unexpected increases in demand (see 2.1.1); conversely a cut of interest rates is needed to curtail
deflationary pressures posed by acute fall of oil prices in response to unexpected drop of demand.
2.1.3 Income Elasticity
In order to fully understand existing changes in oil demand, one also needs to address the existing
income elasticity, i.e. the relationship between GDP changes and oil demand. Hamilton (2009)
shows that elasticity is a dynamic indicator by revealing the relevant historical developments in the
US in a time span of almost over six decades (Figure 3). Initially, the US oil demand has grown
faster than the GDP as the elasticity sloped to a value of 1.2 between 1949 and 1961. It has
remained well above 1 until the first oil crisis in 1973, however it also exhibited a tendency for
gradual decrease. Thanks to technological advancements and economic restructuring towards a
lower share of manufacturing, oil consumption in the US significantly decreased after 1985
irrespective of the dramatic price decline during 1980s and 1990s. Latest data indicate that the US
income elasticity has generally remained below 0.5 threshold during early 2000s (Hamilton, 2009).
Figure 3 – The United States income elasticity of oil demand (1949 – 2006)
Vertical axis: Cumulative change in natural logarithm of total oil supply to the US for the period.
Horizontal axis: Cumulative change in natural logarithm of the real GDP
Source: Hamilton (2009)

Academic researchers have investigated at length existing income elasticities across the globe and
substantial differences have been noticed between the developing and developed countries (e.g.
Ibrahim and Hurst, 1990; Dahl, 1993; Gately and Huntington, 2002; Brook et al., 2004). In general,
developing countries demonstrate significantly higher income elasticity in comparison to developed
ones. Subject to the time of measurement and the exact economic structure, developing countries’
elasticities of GDP growth in relation to changes of oil demand fluctuate between 0.6 and 1.2.
Corresponding elasticity measures for OECD countries vary within the range of 0.4 and 0.6. On a
further note, empirical analyses indicate a decreasing trend over time, thereby providing evidence
for the positive effects of technological advancement in ensuring less energy-dependent growth.
2.1.4 Price Elasticity
Another important aspect of crude oil demand is identified as its responsiveness to the commodity
price itself. Their correlation is explained in the context of price elasticity. A distinct number of
studies demonstrate the existence of short- and long-term price elasticity (e.g. Dahl, 1993; Gately
and Huntington, 2002; Brook et al., 2004; Krichene, 2006). The short-term response indicates
market participants’ immediate reactions to present shifts in oil prices. On the other hand, the long-
term price elasticity captures the effects of monetary policies, investments, conservation policies
and shifts to more energy efficient technologies which take place over a prolonged period of time.
In this context, Gately and Huntington (2002) theorize how these long-term developments can lead
to a tendency of lessened oil consumption, which cannot be reversed by commodity price discounts.
Both the short- and the long-term oil price elasticities have been extensively analysed. Most of the
evidence was collected during 1990s and 2000s. In both types of economies – developed and
developing – short-term price elasticity of demand varies within a small range between -0.02 and
-0.09. The long-term oil price elasticity exhibits values between -0.03 and -0.64 (Dahl, 1993; Gately
and Huntington, 2002; Cooper, 2003; Brook et al., 2004; Krichene, 2006).
In summary, previously reviewed empirical data on oil prices’ elasticity of demand allows for two
primary conclusions: first, short-term oil price swings have only a marginal effect on demand;
second, the long-term demand elasticity increases as a result of strategic actions towards energy
conservation and substitution. Nevertheless, the value of the long-term price elasticity of demand
remains relatively low, which suggest for the high degree of oil dependence to persist historically.

2.2 Supply Factors
2.2.1 Hotelling’s Rule
Early research attention emphasises on the specific status of oil as an exhaustible resource
associated with scarcity rents and continuously increasing prices. According to a pioneer study by
Hotelling (1931), the long-term price dynamics of an exhaustible resource follow a rate of increase
equal to the level of interest rates. Since commodity prices rise at a rate determined by the given
interest rate level, Hotelling’s work hypothesize that commodity producers are time indifferent.
The initial theoretical proposition by Hotelling (1931) has triggered considerable research attention,
which resulted in a mixed empirical support for his theory (Weinstein and Zeckhauser, 1975;
Fattouh, 2007). Frankel (2006, 2007) suggests that commodity prices are significantly influenced by
the level of real interest rates. The variable cost of money affects the return provided by alternative
investment vehicles, which in turn alters producers’ incentives to extract oil and their preferences
for inventory levels. Multiple empirical studies have confirmed the proposed negative relationship
between the level of real interest rates and crude oil prices (Krichene, 2008; Akram, 2009).
Yet, there is no general consensus with respect to Hotelling rule’s applicability in practice.
Nordhaus (1973) and Stiglitz (1974) suggest that commodities may experience a manifold of valid
price trajectories which fit in the requirements for asset-market equilibrium as outlined by the
Hotelling’s rule. As a consequence, any commodity-price dynamics can be theoretically explained
depending on the specific set of assumptions and resource-related terminal solutions. Nevertheless,
this remains of little real value for accurately recognising when prices are fundamentally justified.
According to Livernois (2009), the Hotelling’s rule has a limited empirical applicability because of
the multiple additional factors which could potentially trigger price changes far detached from their
fundamental values. According to Gronwald (2009), this is particularly valid in the case of crude oil
prices, which remain quite sensitive to the side effects of news and other major announcements.
2.2.2 Peak Oil Hypothesis
The extended trend of crude oil appreciation from 2003 to 2008 led to re-emergence of the peak oil
hypothesis. The peak oil theory was initially proposed by Hubbert (1956), who suggested that the
rate of oil production follows a specific bell-shaped curve. According to the theory, production peak
is observed when approximately half of the known commodity reserves have been extracted. In
Fattouh’s (2007) words, Hubbert’s peak oil hypothesis gained popularity in the 1970s because of its

relative accuracy in forecasting US oil-production rates. From a global perspective, statistical data
unambiguously demonstrate that the world’s crude oil production was largely stagnant in the period
between 2005 and 2008, failing to address growing commodity demand in the same period (Energy
Information Administration, 2016b, 2016c). This piece of evidence match the peak oil hypothesis,
suggesting stagnant or falling production levels to present upward pressures on commodity prices.
Nevertheless, the peak oil hypothesis received substantial criticism which is supported by solid
empirical evidence. A key shortcoming of the theory is its static approach in reviewing the size of
the world’s ultimately recoverable reserves (URR) (Watkins, 2006). In reality, the size of the URR
has been continuously expanding as a result of more efficient and modern technological methods. In
addition, analysis of previous predictions demonstrates that the effect of depleting availability is
overstated. Considering that oil discoveries are regularly made, Adelman (1990) argues that the
commodity status as an exhaustible resource is to be reconsidered. Instead, oil reserves are to be
categorised as a renewable inventory which changes its size as a result of two opposing processes:
depletion and new exploration and development (Adelman, 1990). In this context, the dynamics of
the two processes are significantly influenced by spot pricing. Higher oil prices create strong
incentives for new exploration and development (reserves accumulation), which in turn is likely to
cause future price declines and increased consumer demand. Once demand exceeds supply, prices
gravitate upwards, creating a specific perception cycle of scarcity and abundance (Mabro, 1991).
Hypothesizing a step further, Lynch (2002) proposes that the actual production curves are not bell-
shape symmetrical, as was initially proposed, but rather skewed to the right due to verifiable
reductions in the production scarcity effects; such reductions are realised through discoveries of
new oil reserves, introductions of more efficient technologies for extraction and implementations of
unconventional oil sources – oil sands, oil shales, Arctic oil, etc. Finally, the theoretical work of
Kilian and Murphy (2014) reviews oil price declines post 2008 peak and concludes that supply-side
factors have only had a marginal role in crude oil price formation, thus rejecting the peak oil theory.
2.2.3 Oil Stocks
Crude oil and its refined outputs are relatively easy to store, hence changes in the inventories of
these products are often examined as fundamental determinants not only of crude oil pricing, but
important components constituting other traditional demand and supply factors (Davig et al., 2015).
Breitenfellner et al. (2009) suggest that build-up of inventories allows for a greater degree of

flexibility to address short-term supply shortages. In essence, accumulated oil-stocks serve as a
cushion, which to a certain degree neutralises the adverse effects of supply shocks on prices.
In addition to other considerations, the size of oil stocks is likely to reflect the prevailing market
sentiment with respect to future supply (Fattouh, 2007). Market participants are keen to hold
inventories at a changing rate depending on future supply prospects. Expectations for lessened
future supply are likely to trigger “precautionary demand” measures, which are associated with the
build-up of oil stocks for future purposes. And vice versa: stakeholders are less willing to hold oil
stocks and incur storage costs, if abundant future supply is expected. Consequently, the recent
dramatic decline in crude oil prices is supposedly explained through shifts in precautionary demand,
rather than significant changes in the fundamental supply and demand factors (Davig et al., 2015).
2.2.4 Refining
Once crude oil is extracted, it must be further processed in refineries for the uses of its final
consumers. The absolute supply of end products thereby depends on the available oil refining
capacity and its respective capacity utilisation rates. As Breitenfellner et al. (2009) argue, these two
supply-side factors are likely to influence crude oil prices’ dynamics as well. In general, an
expanding refining capacity should be seen as a favourable factor for addressing possible demand-
driven bottlenecks and it tends to reduce crude pricing. On the contrary, market shortages usually
occur when the utilisation rates reach capacity limits, resulting in an upward pressure on oil prices.
Empirical evidence has shown support for the roles of refining capacity and utilisation rates as
possible price determinants. From a historical perspective, the oil refining capacity in the US has
been stagnant since 1981, despite a distinct the trend for soaring demand for the commodity (Dées
et al., 2008). In addition, two unrelated events led to supply shortfalls which exacerbated the trend
for rising prices between 2005 and 2008. In the first instance, hurricane Catrina led to disruptions in
refineries’ production cycles in the southern parts of the US (Cohen, 2007). Secondly, unscheduled
refinery maintenance has caused temporary refining capacity declines in 2007 (Dées et al., 2008).
Reviewing capacity utilisation rate, Cohen (2007) demonstrates that in the same period, levels
reached 92%, which was significantly higher than the historical average of 85%. Such a high
utilisation rate leaves little space for further improvements and cannot fully compensate the
negative effects of stagnant capacity. As Cohen (2007) suggests, the insufficient supply of refined
products resulted in increased price spreads for them, which fuelled the trend for higher oil prices.

2.2.5 Oil Supply Inelasticity
In their research Dées et al. (2008) justify the high volatility of crude-oil prices with the presence of
strong non-linear relationship between supply of the commodity and it’s pricing. They hypothesise
the non-linear relationship to be particularly present in the context of extreme market events or
“shocks”. A number of empirical studies also contribute with estimates of the long-term price
elasticity of supply. According to Krichene (2006) elasticity of non-OPEC supplies has a very low
value of 0.08. Other assessments propose more moderate values within the range between 0.15 and
0.58 (Dahl and Duggan, 1996; Gately, 2004). Despite assessment differences, reviewed estimates
conceptually confirm existence of inelastic relationship between commodity supply and its pricing.
More specifically, Kaufmann (1991) and Kaufmann and Cleveland (2001) suggest commodity
prices to show a high degree of sensitivity to supply especially in cases when production
approaches its full capacity and additional supply reserves are scarce or inaccessible. Their evidence
demonstrates these non-linear response curves to be primarily revealed in supply shock events, such
as producer lags in construction of new refining capacity or developing alternative energy sources.
2.3 The Role of OPEC
Reviewing crude oil market dynamics, one should acknowledge the specific role of countries
participating in the Organisation of Petroleum Exporting Countries (OPEC). In 2016, OPEC
consists of 13 countries – Algeria, Angola, Ecuador, Indonesia, Iran, Iraq, Kuwait, Libya, Nigeria,
Qatar, Saudi Arabia, the United Arab Emirates and Venezuela – which combined account for 73%
of the global oil reserves, 40% of the total production and 55% of the world’s oil exports
(Breitenfellner et al., 2009; OPEC, 2016). OPEC members coordinate their output decisions by
implementing mutually negotiated production quotas. Such collusion among the largest oil
producers results in extensive market power with strong implications for global crude oil markets.
From a theoretical standpoint, there exist several cartel models which attempt to explain OPEC’s
behaviour. The dominant firm model identifies Saudi Arabia as leading producer due to its ample
excess capacity. The country’s privileged position enables it to initiate production changes, which
are then followed by other members – as most of the cartel power is delivered by Saudi Arabia, the
rest of the cartel generally mirrors Saudi Arabia’s decisions and actions (Griffin and Nielson, 1994).
According to de Santis (2003), the dominant firm position of Saudi Arabia is particularly evident in
the context of long-term price fluctuations, whereas short-term variations are influenced by

announcements of OPEC’s decisions on production quota. Wirl and Kujundzic (2004) document
that the short-term effects of output announcements shift crude-oil pricing in its respective predicted
direction, however these effects remain statistically insignificant. On the other hand, the expected
long-term role of Saudi Arabia as a cartel leader is demonstrated in Figure 4, which maps the
relationship between Saudi Arabia’s production levels and changes in oil prices. Visual analysis
leads to two important conclusions (although statistical significance remains unclear): first, changes
in Saudi Arabia’s levels of production are negatively correlated to crude oil prices (increase of
production leads to decrease in prices); and second, there is a time lag effect as prices do not adjust
immediately after production levels are altered (US Energy Information Administration, 2015).
Figure 4 – Saudi Arabia production changes and oil price dynamics
Source: US Energy Information Administration (2015)
There are other applicable cartel models which address OPEC’s behaviour – various alternative
cartel models examine OPEC as a typical market-sharing cartel, where participating members to a
greater or lesser extent act simultaneously as a wealth-maximising monopolist (Griffin, 1994).
Böckem and Schiller (2004) apply a price leadership model, where OPEC has a special role as a
leader and non-OPEC producers are viewed as price takers. This privileged market position is also

theoretically investigated by Hnyilicza and Pindyck (1976) who apply a two-part cartel model and
demonstrate that OPEC’s output decisions are mainly driven by profit maximisation objectives.
Another body of research analyse empirically whether OPEC executes a significant market power.
Dvir and Rogoff (2010) focus on the long-term price fluctuations and demonstrate the increasing
role of OPEC especially in the period post 1970s, which has led to higher and more volatile
commodity prices. The advance of a powerful oil-cartel is linked to the declining relative share of
US based producers, which enabled OPEC countries to capture larger part of the global oil market.
Kaufman et al (2004) examine a period between 1986 and 2000 by applying a vector error
correction model. They demonstrate that the cartel affects commodity prices with two instruments,
shaping end-supply in accordance: production quotas and the specific capacity utilisation rates.
Furthermore, OPEC is viewed as a loosely cooperating cartel where “quota cheating” often takes
place (Figure 5). In this context, the work of Kaufman et al. (2004) also verifies empirically that the
degree of cheating itself has an even further statistically significant impact on commodity prices.
Figure 5 – Output quotas and cheating
Source: Energy Economist (2013)
Lastly, a growing number studies suggest that the role of OPEC as a price setter varies over time;
however, evidence for determining the exact influence of the cartel over the last several decades

remains ambiguous. On the one hand, Fattouh (2007) demonstrates that the cartel and Saudi Arabia
in particular effectively influenced oil production and price levels during 1970s and 1980s. This
view is largely supported by Chevillonand Rifflart (2009) who suggest OPEC’s price-setting
effectiveness continued to be significant for much of the 1980s and early 1990s. Contrasting
evidence, however, as investigated by Wirl and Kujundzic (2004) uncovers the effect of OPEC
meetings on oil prices to be weak at best and if at all existent restricted to scheduled price increases.
2.4 Financial Speculation
Speculative activity in commodity markets is generally divided into two categories. The first type,
referred to as inventory demand shock is driven by sudden changes in expectations. In the context
of this study, an increase in the demand for oil inventories triggers a significant shift of the demand
curve along the upward-sloping supply (section 2.1.1), which translates into higher commodity
prices (Kilian and Murphy, 2014). The second type is called speculative shock and it is related to oil
futures trading. Futures contracts per construct are not associated with a physical delivery until their
expiry and are traded at a margin (Silvapulle and Moosa, 1999). Regardless, price fluctuations of
futures affect investment decisions which producers are facing. For instance, sizeable purchase of
futures signals higher spot price expectations at the date of settlement, thus incentivising producers
accumulate inventories aiming for higher profits and the vice versa (Hamilton. 2009).
Historic data indicates the growing role of financial speculation as commodities have been
recognised as an attractive asset class and a possible portfolio diversification tool, alternative to
stocks and bonds, especially during the boom in mid-2000s. Furthermore, as Silvapulle and Moosa
(1999) suggest, futures contracts can be bought with a very little cash upfront, giving rise to
opportunity for leveraged trading, which further contributed to the growing interest in commodities.
Responding to these developments, commodity trading has increased from $13 billion in 2004 to
$260 billion in early 2008 (Juvenal and Petrella, 2012). According to Tang and Xiong (2011), the
growing investment flows in commodities played a paramount role for price escalations during the
boom period of 2004 – 2008. Fattouh (2007) also identifies the growing role of futures markets in
forming oil prices, in his view, futures’ price-movements reflect the aggregate market expectations.
Changes in speculative activity in oil futures contracts are reflected in Figure 6 – based on historic
data, three main observations can be made. First, the volume of speculative activity has been
growing over time irrespective of oil price fluctuations. Second, speculative behaviour and WTI

spot-price fluctuations closely mirror each other over time, and especially so after late 2008. Lastly,
periods of spikes in the volume of short positions can be noted to precede or coincide with a strong
positive correction of WTI prices, an evidence which supports inventory-demand shocks’ rationale.
Figure 6 – Net future positions held by money managers
Data Sources: US Energy Information Administration (2016a)
US Commodity Futures Trading Commission (2016)
The role of the speculative activity has been thoroughly reviewed in the literature on oil-price
determinants (Breitenfellner et al., 2009; Kaufmann and Ullman, 2009; Ellen and Zwinkels, 2010).
Most researchers find that in addition to significant fundamental factors, which lead to the peak in
2008, speculative activity also played a decisive role in the formation of commodity price bubble.
Subsequent academic literature suggests the magnitude of speculative activity as price-determining
factor to fluctuate over history. Vansteenkiste (2011) find oil-price variations prior 2004 to be best
0
20
40
60
80
100
120
140
160
-400
-300
-200
-100
0
100
200
300
400
500
600
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
money managers long money managers short money managers net WTI Spot

explained by oil fundamental factors. In contrast, Beidas-Strom and Pescatory (2014) illustrate
speculative behaviour as having contributed to price increases only on two occasions: between 2005
and 2008, as well as in 2011 – 2012. Their findings are in line with observations in Figure 6, as both
cited periods designate oil price peaks and intensified speculation in the final stages of bull markets.
2.5 US Dollar Value
Significant research attention has focused on uncovering a possible relationship between the US
dollar value and oil prices (Chaudhuri and Daniel, 1998; Habib and Kalamova, 2007). There exist
two primary considerations from theoretical standpoint. First, commodity prices are denominated in
US dollars, which is a predisposition for correlation between the real effective exchange rates and
the real oil price, respectively. Second, Krugman (1983) emphasises the role of international trade
and the corresponding wealth effect. According to his model, changes in oil prices serve as a
mechanism for wealth transfer among oil exporters and importers. Put into perspective – higher oil
prices are expected to cause current account deficits and portfolio-allocation effects, which will
adversely affect currency value of oil importing countries and the vice versa. The exact exchange
rate effects are furthermore influenced by the degree of oil dependence and level of exports in.
Most of the studies reviewed are concerned with the first theoretical consideration and theorize for a
negative correlation between the two measures: a strong value of the US dollar presents downward
pressures on oil prices and the vice versa (Muñoz and Dickey, 2009 and Hošek et al., 2011). More
specifically, Novotny (2012) suggests for a 1% depreciation in the nominal effective US dollar
exchange rates to result in a 2.1% increase in the commodity price during the 2005 – 2011 period.
On different note, the value of the US dollar vis-à-vis the Euro has been extensively used as a
measurement of the strength of the US currency. Hošek et al., (2011), for example, discover a
strong negative correlation of -0.9 describing the movements in the USD/EUR pair and oil prices
for the period between 2000 and 2009. In the same context, research by Cuaresma and
Breitenfellner (2008) identified a negative correlation of -0.73 for the period from 1998 to 2006.
In a recent long-term study, Fratzscher et al. (2014) also confirm the existence of a strong
relationship between the two instruments – a 1% US dollar depreciation is found to trigger a 0.73%
increase in oil prices. Through the use of variance decomposition analysis, however, it is
established that the economic importance of exchange-rates as determinants of oil-price swings

remains limited. Moreover, the observed significant negative correlation between oil prices and the
US dollar value is found as relatively recent phenomenon which developed since the early 2000s.
2.6 Geopolitical Factors
A historical review of the last half century provides critical insights and identify unexpected
geopolitical events which coincide and allegedly stimulate significant short-term escalation of crude
prices. In early 1970s, the exhaustion of the US refining capacity combined with the end of the
existing Bretton Woods system led to abrupt depreciation of the US dollar which fuelled an initial
increase in oil pricing. In the same period, historic data and Hamilton (2010) verify significant
production cuts equating to 7.5% of global output. Such drastic changes are linked to the Arab Oil
embargo, which some OPEC countries proclaimed and targeted towards selected Western countries
supporting Israel in the Yom Kippur War. The oil embargo combined with the Iranian revolution in
1978 – 1979 and the Iran – Iraq war of 1980 have triggered multi-fold spikes in oil prices. Later on,
during political tensions surrounding the first Gulf war in 1990 – 1991, oil production dropped with
8.8% of global output which doubled prices (Hamilton, 2003). More recently, the 9/11 terrorist
attacks can be identified as yet another shock with substantial implications for crude market pricing.
Figure 7 – Various shocks and oil price response
Source: US Energy Information Administration (2015)

Significant oil price developments are not necessary related only to geopolitical events. The
overview in Figure 7 demonstrates that commodity prices also react in relation to economic crises
of regional and global magnitude. For instance, the Asian financial crisis on 1997 – 1998 is shown
as an event exerting a strong negative impact on oil prices. The global financial and economic crisis
following the Lehman Brothers collapse in September 2008 is another high-profile event which
coincides with significant crude decline of almost 75% for period of just four months in late 2008.
Based on the alignment of geopolitical and economic events with significant fluctuations in oil
prices, they make for viable determinant candidates and their impact needs to be tested empirically.
2.7 Conclusions
As a result of the literature analysis, several major groups of possible determinants are outlined:
traditional duality of supply and demand factors, palpable market structure in place, foreign
exchange rates, degree of financial speculation and important geopolitical events. All factors
exemplify the intricate nature of oil price determination which cannot be viewed unequivocally.
More specifically, a vast number of previous empirical research elucidate in further detail the
specific role and magnitude of effect of each variable. Investigators of demand appear to agree on
the positive association of emerging economies’ growth rates and oil prices, however remain less
conclusive in the case of developed countries. Second, monetary policy is found to exert both direct
and indirect impact via transmission channels – consumer price inflation and growth expectations,
which in turn are closely linked to foreign exchange rates and variances in elasticities of demand as
price influencers per se. Supply theorists, on the other hand are concentrated on exhaustiveness of
the commodity, rigidness of production levels and derivative supply factors such as accumulated oil
stocks, available refining capacities and their respective utilisation. The bulk part of evidence
models for short-term inflexibility of supply and its negative association with crude pricing. Fourth,
market structuralism has found OPEC cartel’s power and loosely cooperative form to yield mixed
influence in total. Last, analysis of recent geopolitical events and growing financial speculation are
theorised as having positive association with prices. Overall, existing empirical evidence is
inconclusive as it uncovers for variety of possible predictors to not only exert diverging impacts on
prices, but also to vary regularly in degree of significance subject to selected periods of observation.
Despite its ambiguous results, presented literature review constitutes a sophisticated framework of
main historic findings. A leading objective of this study is to incorporate such existing knowledge

with relevant new concepts for the purposes of hypothesis testing with latest contemporary data.
Ensuing research promotes uncertainty robust analysis and flexible approach to factors assessment.
3 Methodology
3.1 Overview
The methodology chapter of current study aims to provide a detailed description and specification
of the selected research framework. Choices related to the research design and the applied
qualitative and quantitative procedures are scrutinized in extensive detail. From structural
perspective, the discussion of methodology follows logic presented in the so called research “onion”
(Figure 8, Appendix), which systemises the analysis of all research steps and considerations.
First, a selection of an appropriate philosophical approach is discussed and it sets fundamental
direction of the complete methodological framework. Once the choice of philosophical approach is
clarified, the relevant research approaches and strategies are discussed, respectively. The chapter
proceeds to review methods of data collection, the chosen time period of investigation, as well as
the relevant empirical techniques (and limitations) which are applied in the data analysis process.
3.2 Philosophical Approach
According to Saunders et al. (2009), there are four types of philosophical approach: positivism,
interpretivism, realism and pragmatism. The current study applies positivism as a leading
philosophical approach in analysis of the significance of the various crude-oil price determinants.
Most of modern social science studies rely on the positivist framework, due to its independent and
quantifiable characteristics, which facilitate validation of the research process and its results.
Positivism is chosen as the most appropriate philosophical framework in the current case of oil-
markets research. Founding arguments for this choice are approaches’ preconditions as it assumes
the existence of a reality which is external, objective and most importantly independent from social
actors. Data collection is based only on observable phenomena which are able to deliver objective
and credible data. Positivist approach is of further value for the current study as it assumes causality
between reviewed variables and analysis’ statistical results can be generalised on. Typical
characteristics of the positive studies are the existence of large samples, quantitative approach of
measurement and the overall highly structured methods of analysis (Saunders et al., 2009).

3.3 Research Approach
Selected research approach has been distinctly described as it significantly influences the worldview
(Saunders et al., 2009). The positivist philosophical stance is closely associated with the application
of a deductive worldview. Deduction is usually preferred when investigating potential links
between variables which aligns with the causal nature of the current study (Saunders et al., 2009).
Furthermore, the deductive approach emphasises on drawing hypotheses when studying the set of
related variables. In most of the cases, previously discovered relevant pieces of theoretical and
empirical knowledge are fully utilised in the analysis of the studied topic. It is important to note that
quantitative data are also extensively used as a part of the chosen worldview (Saunders et al., 2009).
3.4 Research Strategy
According to Saunders et al. (2009) social studies comprise of three distinct research types:
explanatory, descriptive and causal researches. The primary aim of explanatory research is to
investigate problems in areas with insufficient theoretical knowledge where no prior studies have
been conducted. Descriptive works deal with topics and have as primary goal to explain current or
past phenomena which might have been explored already. The causal research chiefly focuses on
critical exploration of the possible relationships between variables (Saunders et al., 2009).
In its nature, the current study should be pertained to the causal type of research since its leading
goal is to investigate relationships between several variables already conceptualised by previous
research (Saunders et al., 2009). More specifically, from positivist point of view, it is suggested that
crude-oil price determinants reviewed in Chapter Two (presented in Table 1) are likely to cause
changes in the dependent variable – namely, the actual market price of a barrel of crude oil.
3.5 Method of Analysis: Bayesian Model Averaging
3.5.1 General overview
Bayesian model averaging (BMA) constitutes a sophisticated framework and has recently gained
popularity in academic fields due to its benefits in addressing model uncertainties. BMA is used in
practice as a substitute to established statistical approaches including discrete graphical models,
linear regression, Cox regression models, as well as generalised linear models (Breitenfellner et al.,
2009). Although widely applied in econometric studies, BMA analysis has remained largely
underutilised in the context of oil-price research. The studies of Breitenfellner et al. (2009) and

Lammerding et al. (2012), who model oil convenience yields’ in relation to bond yields and FED
funds rate, are identified as the only existing attempts to analyse oil-price drivers with Bayesian
approaches, which presents a unique opportunity to deliver findings in a largely uncovered field.
Conventional statistical practices usually rely on selecting a given model from a certain class of
models with the conviction that it is able to produce the most accurate and relevant results. The
approach of selecting a particular model is prone to produce overconfidence in the perceived
significance of variables as it overlooks potential risks associated with selected model’s uncertainty.
BMA effectively addresses the issues of overconfidence and model uncertainty by applying an
averaging technique over a space of relevant models. As a result, BMA outcomes in general have an
enhanced predictive ability and their conclusions are more robust to errors (Hoeting et al., 1999).
3.5.2 The Model
Following from the above, the main strength of BMA is that the procedure constructs an overall
model-unbiased average estimates of currently reviewed crude-oil price determinants. Its objectives
are achieved by applying a probabilistic approach across a defined set of models by using observed
data as a weight factor (Breitenfellner et al., 2009; Hoeting et al., 1999). Summarising as follows:
𝐩𝐫(∆| 𝑫) = ∑ 𝐩𝐫(∆| 𝐌 𝒌, 𝑫) 𝐩𝐫( 𝐌 𝒌| 𝑫)𝑲
𝒌=𝟏 (1)
In the expression ∆ is the quality of interest – in this case oil-price determinant under consideration.
Generally speaking, the expression translates to averaging each determinant’s posterior (resulting)
distribution under a set of models M1, M2….Mk by the posterior probabilities of the respective
models themselves, both given data D. Consequently, BMA-based estimates are highly dependent
on 1) prior distribution used to derive each determinant’ distribution in a particular model (M1,
M2….Mk) and 2) posterior model probability pr(M 𝑘| 𝐷). Subsequently, the latter measure is also a
function of two components: 1) the inclusion probability of the model in question, calculated
against a predefined model space and 2) model’s marginal likelihood (i.e. the specific degree to
which model fits the data) (Breitenfellner et al., 2009; Hoeting et al., 1999, Fernández et al., 2001).
Focusing on selecting a prior distribution over the parameters under each model, practitioners have
chiefly applied what is known as “flat prior” – that is a prior distributed as widely spread as possible
over the region where determinants’ likelihoods are largest. A growing body of research, however
(Fernández et al. 1998; Ley and Steel, 2009), empirically demonstrate that Zellner’s g- and BRIC

prior (with added Bayesian and risk inflation criteria) to consistently outperform all previous
methods (Breitenfellner et al., 2009), which has motivated BRIC’s choice as prior in this study.
Naturally, parameters’ likelihoods can be then reviewed post factum (not a priori) averaging.
A key hindrance in calculating expression (1) is the posterior model probability pr(M 𝑘| 𝐷) or in
other words, how likely a particular set (model) of determinants is to occur weighted against other
alternative combinations (models). Hence, posterior model probability is a function of firstly – the
number of models to average across (choice of model space, which will be addressed in Section
3.5.3) and secondly – model prior distribution or how likely a model is compared to the rest a priori.
The literature review conducted has revealed a large degree of uncertainty not only with respect to
proposed crude-oil determinants’ magnitude of correlation, but also direction of effect. Keeping
these considerations in mind, a priori knowledge for purposes of model prior distribution is
considered unreliable if at all available. Following the Bayesian way, one should thereby assume for
all models (combinations of variables) to be equally likely a priori. In other words, the probability
of each determinant to be present in a given model is 50% and it is independent from the choices
made for other measures included in that specific model. These conclusions closely align with
recommendations by Breitenfellner et al. (2009) and Hoeting et al. (1999), hence uninformative
uniform model prior distribution is chosen – leaving observed data to lead the posterior distribution.
3.5.3 Model Space Specification
Performing BMA often constitutes a resource challenge as aggregations are quite complex with
multiple variables, when billions possible models exist to choose from (Hoeting et al., 1999). Three
distinct calculation approaches are applied in practice to narrow, approximate or explore the model
space: Occam’s window, Markov chain Monte Carlo model composition and full enumeration.
The logic behind the Occam’s window calculation principle suggests the application of two guiding
principles. Initially, the predictive ability of all models is assessed and those performing much
worse than the best one are discontinued from further aggregation of estimates. The performance
criterion, however, is subjective as it is set by the researcher (C in the following expression):
𝓐′
= { 𝑴 𝒌 :
𝒎𝒂𝒙 𝒍{𝒑𝒓(𝑴 𝒍 | 𝑫)}
𝒑𝒓(𝑴 𝒌 | 𝑫)
≤ 𝑪} (2)
The second principle behind the Occam’s window is the application of the so called “Occam’s
razor”, where a second round of models exclusion takes place. In this step, more complex models

are eliminated, unless they receive substantially higher data support than alternative simpler ones:
𝓑 = { 𝑴 𝒌: ∃ 𝑴𝒍 ∈ 𝓐′
, 𝑴𝒍 ⊂ 𝑴 𝒌,
𝐩𝐫( 𝑴𝒍| 𝑫)
𝐩𝐫( 𝑴 𝒌| 𝑫)
> 𝟏} (3)
Resulting from (2 & 3), (1) is modified and the new posterior distribution of price determinant is:
𝐩𝐫(∆| 𝑫) = ∑ 𝐩𝐫(∆| 𝐌 𝒌, 𝑫) 𝐩𝐫( 𝐌 𝒌| 𝑫)𝑴 𝒌 ∈ 𝓐 (4)
In this case 𝒜 = 𝒜′
ℬ, so all BMA posterior probabilities will be tacitly conditional on the group
of models which participate in 𝒜 (in other words after the exclusion of models has been carried
out). As Hoeting et al. (1999) suggest, the principles behind the Occam’s window are solid and
logically correct, however, the approach can lead to the elimination of too many models from the
model-space framework and hence oversimplification. Often, the Occam’s window (with subjective
C measure and simple model bias) greatly reduces the number of included models to a single digit
number. As a result, averaged calculations under the Occam’s window principle are prone to the
risk of eliminating Bayesian Model Averaging’s main advantage: namely, its ability to extract the
best outcomes from a vast range of models (combinations of determinants) (Hoeting et al., 1999).
Another calculation approach suggests data processing based on direct approximation (model
composition) and follows the Markov chain Monte Carlo Method (hence MC3
). This method does
not require any elimination of models as it is the case with the Occam’s window approach, thereby
tackling the oversimplification issue, however convergence issues can be problematic (Hoeting et
al., 1999). MC3
is considered to be an adequate substitution for a variety of linear models and it
performs reasonably well even when missing data needs to be incorporated (Hoeting et al., 1999).
The current study aims to identify crude-oil determinants by relying on linear models. In this
context, the choice of MC3
would be appropriate to average across a representative drawn sample
rather than the entirety of combinations. In practice, calculations are often based on 200,000
Markov chain draws, where 100,000 draws are discarded (burn-in). In practice, the number of chain
and burn-in draws can be additionally increased but is unlikely to deliver more precise outcomes
(Madigan and York, 1995; Hoeting et al., 1999; Fernández et al., 2001; Breitenfellner et al., 2009).
The global model space has 2k
distinct models in total, where K equals the number of explanatory
variables or 21, as it will be discussed in Section 3.7, which totals for 2,097,152 combinations.

Whilst under larger number of determinants an MC3
sampler will be preferred, current model space
allows for the enumeration approach to be applied – that is test and record posterior probabilities as
per (1). This will allow for robust results and avoid unnecessary subjective biases or computational
shortcomings. Overall, as per Breitenfellner et al. (2009) the following general formula will govern
study’s oil price analysis which will be conducted across the entire set of model possibilities:
𝑷𝒕 = 𝜶 + ∑ 𝜷𝒋
𝒌
𝒋=𝟏 𝑿𝒋,𝒕−𝟏 + 𝜺𝒕 (5)
Pt reflects oil price inflation, 𝑋𝑗𝑡 for 𝑗 = 1, … , 𝐾 denotes all measures with a possible effect on Pt.
3.6 Method of Analysis: Black-Scholes Model
A critical component of applying relatively new technique of statistical analysis is to communicate
and ensure its applicability for the purposes of active market participants and regulators. Addressing
these concerns, a special focus will be dedicated to Brent options traded on the ICE and their
applicability to successful hedging strategies using BMA’s forecasts as basis for decision making.
For the purposes of obtaining historical option prices, the Black-Scholes is employed. The method
is widely used in practice to calculate traded prices of European style options – derivate instruments
which can only be exercised at a predefined expiration date. The option buyers are assumed to have
free access to investment alternatives which yield risk-free interest rates (Black and Scholes, 1973).
The Black-Scholes options pricing model (as well as its implications therefrom) is based, however,
on several assumptions. First, the risk-free interest rate is known in advance and remains constant
over the entire period until given option’s expiration date. Furthermore, it is assumed for price
dynamics to follow a random walk and exhibit constant variance rate of return over time. Next, no
transaction costs are applicable when buying a given stock (index in this case) or option and
borrowing at the short-term interest rate is allowed for both of the two possible instruments of the
portfolio. Finally, short-selling is not associated with additional costs (Black and Scholes, 1973).
All of these assumptions combined create the necessary conditions to model for accurate options’
pricing in historical terms. The Black-Scholes model for call options’ price can be summarised as:
𝑪( 𝑺, 𝒕) = 𝑺 𝑵( 𝒅 𝟏) − 𝑲𝒆−𝒓(𝑻−𝒕)
𝑵( 𝒅 𝟐) (6)
𝒅 𝟏 =
𝟏
𝝈√ 𝑻 − 𝒕
[𝒍𝒏 (
𝑺
𝑲
) + (𝒓 +
𝝈 𝟐
𝟐
) ( 𝑻 − 𝒕)] 𝒅 𝟐 = 𝒅 𝟏 − 𝝈√𝑻 − 𝒕

Solving for the call–put parity requirement, the price of a put-option can therefrom be derived as:
𝑷( 𝑺, 𝒕) = 𝑪( 𝑺, 𝒕) − 𝑺 + 𝑲𝒆−𝒓(𝑻−𝒕)
= 𝑵(−𝒅 𝟐) 𝑲𝒆−𝒓(𝑻−𝒕)
− 𝑵(−𝒅 𝟏) 𝑺 (7)
Where:
N ( ) denotes the cumulative distribution function of the normal distribution;
T-t equals the time to maturity; S refers as the spot price of the asset; K is the option strike price;
r designates the risk-free rate of return; σ the asset returns volatility rate.
Critical review of Black-Scholes’ model assumptions reveals a strong dependence on observed
volatility – higher crude index volatility will lead to higher option prices and the vice versa. For the
purposes of obtaining historical annualised volatility, GARCH models on three distinct indexes
were identified: Oil Volatility Index (OVX as pushed by the Chicago Board Options Exchange),
S&P GSCI Brent Crude Oil and WTI Crude Oil Indexes (Figure 9, Appendix). In Section 4.4, a
conservative view is pursued and each European-style Brent option is calculated basis market’s
highest measure of expected 30 day volatility among the three as per data by V-Lab (2016).
3.7 Determinants and Hypotheses
Descriptive review of existing literature outlined multiple factors which were previously theorised
or empirically tested as possible determinants of oil prices. The constellation of factors is hereby
organised in several main groups: demand, supply, market structure, financial speculation, US
dollar exchange rate. This breakdown is applied in order to separate and explore for differences
between conventional fundamental factors such as demand and supply, and some of the more recent
perspectives on commodity market – financialisation and exchange rate fluctuations. In addition,
some propositions by Breitenfellner et al. (2009) are also investigated as potential determinants.
In the end, several dummy variables are added to reflect selected high-profile geopolitical and
economic events with alleged strong short-term effects on oil prices. Each of the determinants is
followed by a sign in parenthesis, which demonstrates the expected effect on crude pricing – an
assumption based on theoretical and empirical literature reviewed. “+” stands as an indicator of
positive relationship between the particular determinant and oil price and the vice versa (Table 1).

Table 1 – Crude oil price determinants and hypotheses
Category Determinants Hypotheses
Demand
Federal funds rate (-)
US inflation (CPI) (+)
GDP growth US (+)
GDP growth OECD (+)
GDP growth China (+)
Global oil consumption (+)
H0: Changes in demand factors do not influence oil prices.
H1: Changes in demand factors influence oil prices.
Supply
Total oil rigs count (-)
US refining capacity (-)
US capacity utilisation (-)
Global oil supply (-)
Global oil reserves (-)
OECD oil stocks (-)
H0: Changes in supply factors do not influence oil prices.
H1: Changes in supply factors influence oil prices.
OPEC
OPEC oil supply (+)
Saudi Arabia oil supply (+)
OPEC reserves (+)
H0: The market power of OPEC does not influence prices.
H1: The market power of OPEC influences oil prices.
Financial
Speculation
Net Futures Positions (+)
H0: Financial speculation does not influence oil prices.
H1: Financial speculation influences oil prices.
Exchange Rate Nominal US dollar index (-)
H0: US dollar fluctuations do not influence oil prices.
H1: US dollar fluctuations influence oil prices.
Geopolitical
and Economic
Events
First Gulf War (+)
Second Gulf War (+)
9/11 Attacks (+)
Lehman Brothers (-)
H0: Shocks do not influence oil prices.
H1: Shocks influence oil prices.
Factor
Persistence
H0: Oil price determinants do not change over time.
H1: Oil price determinants change over time.
As repetitive theme in reviewed literature, this study observed vast discrepancies in many of
determinants’ estimates. Multiple factors only remain significant in specific time-periods while
others vary in dependence of the selected period of observation. Furthermore, ambiguity is present
in identifying not only the magnitude, but also the precise direction of correlation. In this context,
testing the validity of previous hypotheses with the most recent factors’ data is expected to provide
important insights into their dynamics over time. The issue is of critical importance in a rapidly
changing global environment and current study is strongly benefited by the availability of latest
economic data which includes dramatic 2014 – 2015 developments in the crude-oil marketplace.

Summarising per category in Table 1, increases in the values of all demand factors are projected to
lead to higher oil prices. The Fedераl funds rate constitute the only exception as it remains the only
demand variable which is negatively correlated with commodity prices. The overall demand
correlation to prices remain mixed. Reviewing the supply factors, any increases in the value of the
determinants are expected to have a negative impact on prices. Similarly, a negative cause and
effect relationship is expected to take place when contemplating on the strength of the US dollar.
The strength of Saudi Arabia and advances of OPEC’s market power in setting prices – translated
through proxy channels: market share and/or respectively reserves – are expected to exhibit an
overall positive correlation. In line with literature, growing oil futures positions are expected to
positively affect oil prices and the vice versa. Finally, four distinct dummy variables are included in
the list: First Gulf war (August 1990–February 1991), the 9/11 terrorist attacks in 2001 and
subsequent 2003 invasion of Iraq (March – April 2003) are presumed to have caused short-term
hikes in prices. Finally, the collapse of Lehman Brothers, which triggered markets’ shock response
of September – October 2008, is examined as having exerted a negative effect on crude-oil pricing.
As it was touched upon earlier, several factors, such as (but not limited to) monetary policy, income
elasticity, financial speculation and the US dollar, fluctuate significantly depending on the period
reviewed (e.g. Fratzscher et al., 2014; Hamilton, 2009; Krichene, 2006). This instability of factors is
addressed with a final hypothesis, which will examine whether particular factor’s significance,
direction or magnitude oscillate in accordance with/contrary to the prevailing market trend. Should
a consistent pattern of persistence be observed, conclusions arising therefrom will have important
implications for future understanding of oil-prices structural components in the longer run.
For aforementioned purposes, after cross-checking with existing literature as well as visual and
descriptive analysis of Brent and WTI indices, the overall period of 28 years will be divided into
three time episodes which are with similar lengths: January 1988 – December 1998, January 1999 –
July 2008 and August 2008 – December 2015. The first episode corresponds to a prolonged period
of historically depressed crude markets, starting from $16.75 per barrel (Bbl) and finally reaching
bottom of $9.82/Bbl in December 1998. The second time episode includes almost a decade of
persistently rising commodity prices which reach an all-time peak of $132.72/Bbl in July 2008. The
last time-section, which might be still developing in 2016, includes the post-peak significant retreat
of prices to $38.01/Bbl in December 2015. Exact cut-points of these macro trends are provisional
by nature as some of the periods include significant price corrections (1991, 2009 and 2011).

3.8 Data Collection and Preparation
Sufficient and reliable secondary data is of primary importance for conducting an objective analysis
of crude-oil price determinants. Oil status as a key commodity has led to the establishment of
comprehensive databases with detailed historical information on oil market developments. Table 2
below summarises data sources, frequencies and links where all information can be directly
accessed. The data collection process has emphasised on selecting sources with proven reliability.
The vast majority of crude-oil determinants were obtained through U.S. Energy Information
Administration’s website as an information source. The U.S. Energy Information Administration
(EIA) track a vast set of energy commodities, and current study uses their data as a source for
determinants in the categories supply and market structure as well as demand determinant Oil
consumption and historic data for spot Brent pricing. At the time of this study, consolidated data for
November and December 2015 are unavailable for Global oil consumption, OECD oil stocks
OECD, US refining capacity and utilisation, Global oil supply and reserves as well as OPEC and
Saudi Arabia oil supply per day. In order to circumvent this challenge, data from EIA’s short-term
oil forecast in October 2015 was used. Previous EIA’s short-term forecasts on aforementioned
determinants have been cross-checked backwards against available consolidated data and no
significant deviations were observed. For the sake of consistency, error levels within two-standard
deviations of period’s forecast errors (roughly 95% confidence level) were then validated against
two standard deviations of previous periods’ yearly movements with satisfactory results of less than
8% or less than 1.61% using period’s absolute yearly change as a base. EIA’s short-term forecasts’
data reliability is thereby considered intact for the purposes of further testing in the current study.
Information related to the group of demand variables is obtained from a variety of sources. The
Federal Reserve database demonstrates the dynamics of the federal funds rate over the investigated
period, as well as, information regarding the changes in the US dollar index. The Bureau of Labor
Statistics provides inflation data for the US which are necessary to model for variance of the crude-
oil prices in real terms. Changes in the GDP rates in the US and China were obtained through the
World Bank database and validated with respectively Bureau of Economic Analysis and National
Bureau of Statistics of China. Lastly, dynamics of OECD countries’ GDP development are accessed
through the official website and used as a joint measure (contribution adjusted weighted average).

Three additional sources are also used in the data collection process: Baker Hughes as publishers of
detailed information on the number of active oil rigs worldwide, a constituent variable of world’s
crude oil supply. Data on the degree of financial speculation are accessed primarily from the
database of the U.S. Commodity Futures Trading Commission and Quandl. For this variable, an
important distinction needs to be made – the current research focuses attention only on speculative
trading, which by definition excludes commercial activity when “the trader uses futures contracts in
that particular commodity for hedging” (US Commodity Futures Tading Comission, 2016).
Table 2 – Data sources
Factor Frequency Base Unit Source
Brent crude price Monthly 1 USD/Bbl US EIA (2016a)
Demand Side
Federal funds rate Monthly 1 Percent Federal Reserve (2016a)
US inflation (CPI) Monthly 1 Percent Bureau of Labor Statistics
GDP growth US Annual 1 Percent The World Bank; Bureau of Economic Analysis
GDP growth OECD Annual 1 Percent OECD (2016)
GDP growth China Annual 1 Percent The World Bank
Global oil consumption Annual 1000 Bbl/day US EIA (2016c,2016d)
Supply Side
Total oil rigs count Monthly 1 oil rig Baker Hughes
US refining capacity Monthly 1000 Bbl/day US EIA (2016b,2016b1)
US capacity utilisation Monthly 1 Percent US EIA (2016b,2016b1)
Global oil supply Annual 1000 Bbl/day US EIA (2016e,2016d)
Global oil reserves Annual 1 Billion Bbl US EIA (2016f)
OECD oil stocks Monthly 1 Million Bbl US EIA (2016g,2016g1)
Market Structure
OPEC oil supply Annual 1000 Bbl/day US EIA (2016h,2016d)
Saudi Arabia oil supply Annual 1000 Bbl/day US EIA (2016h,2016d)
OPEC reserves Annual 1 Billion Bbl US EIA (2016i)
Financial Markets
Net futures positions
(non-commercial)
Weekly 1 Contract
US Commodity Futures Trading Commission;
Quandl (2016a)
Broad Dollar Index Monthly 1 Index unit Federal Reserve (2016b)
Visualisation of each determinant’s developments plotted against Brent is presented at the end of the Chapter.

As it can be deduced from Table 2 obtained data was released in different time frequencies, forming
a significant challenge for further regression-based analysis. Most of the variables follow monthly
and annual time frequencies, whereas net future positions are released on weekly basis. In general,
data analysis requires all variables to follow a unified time frequency. One of the possible
approaches to address this issue is aggregating the data obtained to annual basis. The application is
largely straight-forward and suggests that all weekly and monthly variables should be converted to
approximate average annual values prior to conducting further analysis. Despite its easy
implementation, data aggregation approaches lead to a substantial reduction in the number of
observations, which in turn increases standard error and poses concerns on the validity of the
obtained results. The aggregation approach is applied in the case of net futures positions where the
weekly time frequency is transformed to the preferred monthly time frequency in this study.
From fundamental perspective, these aggregated net futures positions “lead” other data with at least
one month (i.e. March futures positions are commitments contracted for April or later). Information
on the precise split of net-commitments per specific futures rolling-month contract is unfortunately
unavailable and statistical analysis on strength of observed associations was conducted in order to
select appropriate lead period for further analysis (Figure 10, Appendix). Scatterplots of changes in
Brent and the three net financial variables (“spot”, “lead by one month” and “lead by two months”)
reveal loose linear relationship (0.01 and 99 percentile) with a few outliers, but certainly monotonic
in nature (large change in Brent is associated with strong change in net positions). As a next step,
Pearson’s and Spearman’s correlations were produced in or in order to compare associations of
respective variables to Brent. A cross-check of both is enacted as Spearman’s measure is less
sensitive to outliers and avoids the assumption of normal distribution of the data (Laerd, 2016).
Both tests found statistically significant association across all three measures, with strongest one
being net financial positions variable leading spot oil pricing by one month (0.721 and 0.684).
These results are further backed by the comparable almost double trade-volume on front month
contract (Figure 11, Appendix) and have led to its selection as best candidate for further analysis.
The last obstacle of data collection is presented by annually aggregated data with few observations
(in comparison with 336 cases on a monthly frequency for the period Jan 1988 – December 2015).
Addressing this issue, temporal disaggregation (TD) calculations have been widely applied for
transforming low- into high-frequency data. In this way, it is possible to maintain a relatively large
number of observations (and assure results are not invalidated by large error terms). From a

technical point of view, there are several available techniques for performing data disaggregation,
such as Chow-Lin, Litterman, Fernandez, Denton and Denton-Cholette (Sax and Steiner, 2013).
In the specific case of current research, temporal data disaggregation was performed following the
Denton-Cholette procedure without an indicator series – an operation which effectively interpolates
the data with temporal additivity constraint – averaged monthly movements will match annual ones.
Despite its weaknesses as reviewed by Sax and Steiner (2013), it is chosen as best method to protect
independence of the variables prior further analysis – usage of an indicator (for example monthly
Brent) will assume correlation and result in misleading data. An alternative method would be the
widely accepted Chow-Lin approach, which only transfers indicators movement if it correlates with
the variable on annual basis (Chow and Lin, 1973; Sax and Steiner, 2013). Such an approach is,
however, redundant as Brent and temporally-disaggregated data’s correlations will be addressed by
the BMA analysis. Lastly, Denton-Cholette method is chosen over the original Denton method, in
light of its correction to the otherwise observed transitory movement at the beginning of the new
series (disaggregated series won’t build up from unrealistic null level to first observed annual data).
3.9 Reliability and Validity
Controlling for high degree of data reliability and validity of results are of paramount importance
for achieving outcomes which contribute to better understanding of oil prices’ dynamics. Reliability
is concerned data collection and analysis procedures’ ability to deliver consistent results. For this
purpose, it is important to obtain identical results irrespective of situation specifics and the number
of observers (Saunders et al., 2009). As information on crude oil price determinants and the actual
price data are publicly released by transparent and objective institutions with a high level of
reputation, there exist no risk for damaged reliability. In this context, reliability issues are primarily
associated with studies relying on primary data in contrast to secondary as is the current scenario.
Validity on the other hand investigates whether research results are applicable in reality. In this
case, valid results are expected to empirically confirm the proposed causal relationships among
variables (Saunders et al., 2009). By relying on extensive topic-specific literature the risks of facing
internal validity issues are greatly reduced as the conceptual model captures determinants,
supported by deductive and empirical research, which are considered to be good measures of
reality, i.e. the real factors influencing oil prices. Furthermore, external validity is also preserved as
the expected results are fully generalizable and do not depend on research settings (Saunders et al.,

2009). The ultimate validity test is verification of results’ forecasting abilities basis statistically
significant oil-price predictors. In ensuing section, resulting model’s forecasting properties are
examined and the latter manage to demonstrate satisfactory prediction of oil-price movements.
3.10 Data Limitations
Several soft elements related to aspects of the data and consequently model choice can be identified.
These originate primarily from instances pertaining to insufficient data or its technical aspects,
rather than being inherited weaknesses of the overall methodological framework set in motion.
First, US exploration costs, a widely applied supply variable in multiple studies on the topic, have
been deliberately omitted. The rationale for disregarding it as a possible determinant is motivated
by inadequate availability of secondary data. Official reports on the size of the US exploration costs
are not available after 2007, which constitutes a substantial time gap. Unfortunately similar and
reliable information for other major oil producers, such as OPEC countries and Russia, is absent.
On a critical note, reviewed time period covers historical developments after 1988. Specified
timeframe is quite significant in length per se, nevertheless including earlier data would be of high
value for the purposes of identifying additional insights on price predictors’ changing magnitudes of
effect. For instance, the actions of the OPEC cartel during the oil crises from 1970s and early 1980s
constitute an intriguing case to analyse with respect to the specifics of the existing market structure.
Insufficient and inadequate public data for the chosen Brent oil type, however, effectively prevents
examination of earlier periods. Thus, a deliberate choice is made to limit the length of analysis as a
trade-off, which is believed to maintain reliability and validity even at the expense of prior data.
Lack of data in monthly frequency has dictated the use of temporal disaggregation, which can result
in smoother intra year movements. Despite protecting the effect of yearly changes, some unique
variances in the month-to-month continuum are lost. Moreover, proprietary monthly ICE data on
EU-style Brent Option and its underlying Brent Bullet future necessitates the use of Black Scholes
Model with its intrinsic assumptions in order to estimate historic European options for Chapter Six.
Finally, conducting Bayesian model averaging across all models with uniform likelihoods could
nevertheless be interpreted as a model choice (Hoeting et. al, 1999). Some further data elicitation on
priors’ distributions need to be achieved, in order to prudently weight the existing model space.
Review of existing literature found no consistent elicitation and hence uniform priors were selected.

0%
2%
4%
6%
8%
10%
12%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) Federal Funds Rate (right axis)
55
65
75
85
95
105
115
125
135
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) Dollar Index (right axis)
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
7%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) US Inflation (CPI) (right axis)
-5%
-4%
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) GDP growth US (right axis)
2%
4%
6%
8%
10%
12%
14%
16%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) GDP growth China (right axis)
-6%
-4%
-2%
0%
2%
4%
6%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) GDP growth OECD (right axis)
Potential Oil Price Determinants
Historic Overview 1988 – 2015

60
65
70
75
80
85
90
95
100
0
20
40
60
80
100
120
140 1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
MillionBarrel/Day
USD/Barrel
Brent (left axis) Global Oil Consumption (right axis)
60
65
70
75
80
85
90
95
100
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
MillionBarrel/Day
USD/Barrel
Brent (left axis) Global Oil Supply (right axis)
20
22
24
26
28
30
32
34
36
38
40
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
MillionBarrel/Day
USD/Barrel
Brent (left axis) OPEC Supply Share (right axis)
5
6
7
8
9
10
11
12
13
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
MillionBarrel/Day
USD/Barrel
Brent (left axis) SA Supply Share (right axis)
800
900
1,000
1,100
1,200
1,300
1,400
1,500
1,600
1,700
1,800
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
BillionBarrels
USD/Barrel
Brent (left axis) Global Oil Reserves (right axis)
600
700
800
900
1,000
1,100
1,200
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
BillionBarrels
USD/Barrel
Brent (left axis) OPEC reserves (right axis)

2
3
4
0
20
40
60
80
100
120
140 1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
BillionBarrels
USD/Barrel
Brent (left axis) Oil Stocks OECD Industry (right axis)
1,000
1,500
2,000
2,500
3,000
3,500
4,000
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) Total oil rigs (right axis)
15
16
16
17
17
18
18
19
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
MillionBarrel/Day
USD/Barrel
Brent (left axis) US refining capacity (right axis)
75%
80%
85%
90%
95%
100%
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
USD/Barrel
Brent (left axis) US refining utilisation (right axis)
-100
0
100
200
300
400
500
0
20
40
60
80
100
120
140
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
Thousands
USD/Barrel
Brent (left axis) Net Futures positions (right axis)

OIL BOOM AND BUST: What Determines Crude Oil Prices?

OIL BOOM AND BUST: What Determines Crude Oil Prices?

Recommended

Recommended

More Related Content

Similar to OIL BOOM AND BUST: What Determines Crude Oil Prices?

Similar to OIL BOOM AND BUST: What Determines Crude Oil Prices? (20)

Recently uploaded

Recently uploaded (20)

OIL BOOM AND BUST: What Determines Crude Oil Prices?