ArticlePDF Available

Abstract and Figures

The complexity of the world oil market has increased dramatically in recent years and new approaches are needed to understand, model, and forecast oil prices today. Many models have been explored and most of the papers and modeling projects referenced in this paper identify their own short-comings and opportunities for future improvements. The reason to choose those models is they have been largely used due to their quality and contribution to their targeted objectives. Rather, this paper will focus on what remains to be done despite of the quality of research reviewed. Considering the complexity of the oil market including the roles of market actors, technical and engineering challenges, environmental issues, and political interests, the modeling challenge would be difficult enough in static conditions. In fact, the oil market is a constantly evolving system requiring the development of new theories and methods to address the questions of the day. The spirit of this paper is to present constructive approaches for future work rather than to provide formal assessment of models on any particular metric.
Content may be subject to copyright.
1
Critical Improvements in Oil Market Modeling
Michael Gucwa,a Ali Nouria, Hillard Huntingtona, and Saud M. Al-Fattahb
aEnergy Modeling Forum, Huang Engineering Center, Stanford University, 475 Via Ortega,
Stanford, CA 94305-4121.
bKing Abdullah Petroleum Studies and Research Center, P.O. Box 88550, Riyadh 11672, Saudi
Arabia; Email: saud.fattah@kapsarc.org
ABSTRACT
The complexity of the world oil market has increased dramatically in recent years and new
approaches are needed to understand, model, and forecast oil prices today. Many models have been
explored and most of the papers and modeling projects referenced in this paper identify their own
short-comings and opportunities for future improvements. The reason to choose those models is
they have been largely used due to their quality and contribution to their targeted objectives. Rather,
this paper will focus on what remains to be done despite of the quality of research reviewed.
Considering the complexity of the oil market including the roles of market actors, technical and
engineering challenges, environmental issues, and political interests, the modeling challenge would
be difficult enough in static conditions. In fact, the oil market is a constantly evolving system
requiring the development of new theories and methods to address the questions of the day. The
spirit of this paper is to present constructive approaches for future work rather than to provide
formal assessment of models on any particular metric.
Keywords: Oil market analysis, modeling oil market, financial markets
1. INTRODUCTION
The goal of this paper is to identify gaps in the oil price modeling literature. In providing a broad
overview, it is important to clarify that this document does not present a detailed extensive critique
of the existing models. Indeed, most of the papers and modeling projects referenced in this paper
identify their own short-comings and opportunities for future improvements. The selected models
have been chosen largely because they are notable for their quality and contribution to their targeted
objectives. Rather, this paper will focus on what remains to be done despite of the quality of research
reviewed. The oil market encompasses a highly complex set of market actors, technical and
engineering challenges, environmental issues, and political interests. The modeling challenge would
be difficult enough in static conditions, but in reality the oil market is a constantly evolving system
requiring the development of new theories and methods to address the questions of the day. The
spirit of the document is to present constructive approaches for future work rather than to provide
formal assessment of models on any particular metric.
2
The paper is organized in several major sections. The next section describes several objectives
for research in the oil market and how models could address these needs. Section 3 and Section 4
discuss specific topics in medium-term and long-term oil models, respectively. Section 5 concludes.
2. MODELING THE OIL MARKET
Oil computer models are simple representations of complex interactions found in the real markets.
Most models encompass oil consumers, oil producers from major exporters and oil producers from
other countries. Decisions by each set of actors interact to determine long-run oil-price paths and
the production and consumption levels in different regions of the world.
Decisions are influenced by a set of exogenous factors: world economic growth, vehicle stock,
cost and availability of oil resources, oil market structure, the costs of oil substitutes, and other
factors. Different models will determine which variables to simulate endogenously within the model,
and which parameters to leave as exogenous. Choosing which variables to include is an important
step in establishing the boundaries of the model and is influenced by the purpose of the model, the
desired level of model complexity, and data availability.
This section considers various questions that different modeling approaches answer and
presents our recommendations on the section of modeling techniques.
2.1 Uses of Market Models
Decision makers use models for many reasons:
Forecast future prices.
Understand how new conditions might influence future prices.
Understand how past conditions (growth, financialization) may have influenced previous
prices.
Provide a consistent framework for organizing new research findings in a coherent manner.
Conduct detailed policy studies that incorporate a complete set of energy-market effects.
Each of the objectives may require a different modeling approach. Most modeling teams begin
by identifying their goals: a good forecasting model may lack the details to provide reliable policy
analysis, where a long-run model may not be useful for evaluating short-run dynamics. The time-
frame for the model is a key consideration. Short-term models may be daily or monthly, where long-
term models may take substantially longer time steps (yearly or multiple years). Intermediate models
will have transitions between the two approaches featuring quarters of the year.
Numerous opportunities exist to add to the understanding of the oil market through the use of
models, with methodology and the type of model chosen depending on the final objectives of the
study. A few possible questions are listed below:
a. Making long-term decisions
An example of long-term decisions may be to choose capacity expansion or to select the right
strategy for OPEC (i.e. what capacity / quota system will lead to realizing the objective of stabilizing
3
prices). Building such a model relies on understanding the long-term drivers and forces moving the
oil market: demand behavior, resource availability, and production capacity.
b. Making medium-term decisions
What is the right level of production to choose? How does this decision interact with the futures
market and financial activities? To answer these questions a model needs details of the oil market
including production, transportation, and refining. Additionally, it may be important to understand
how the fundamental elements interact with the financial markets.
c. Making short-term decisions
What kind of market position should be taken in the oil market, either as a financial speculator or to
hedge risk? What kinds of actions or statements can be useful in stabilizing short-term price
volatilities? Creating a short-term model of this nature requires a detailed understanding of the
market players and their influences in the short-term market. Additionally, for understanding the
short-term and futures market the role of information and the formation of expectations need to be
considered more carefully. How do uncertainties and information asymmetries impact market
behaviors?
d. Understanding oil impacts on other sectors
How does future oil activity impact other economic or political choices? How should domestic
energy policy account for these spillover effects? How should economic programs such as technical
education and industry incentives be deployed? How do energy policies affect the nation’s GDP and
quality of life? The primary emphasis in such a model is to understand the interaction of oil prices
with other measures in the market such as consumer welfare. Typically the interaction between oil
prices and these measures is quite complex, and explicit price forecasting is rarely done
endogenously. Rather these models usually take a projected price path, and, based on model
outcomes, deviate in a controlled manner from the initial, pre-determined oil price paths.
e. Being able to describe previously unpredicted price movements
A model can be useful in validating if a certain theory proposed to explain historical price
movements is accurate. While the model may not have the ability to project future prices, it can
attempt to answer what forces have influenced prices in the past; potentially justifying the inclusion
of certain parameters in forecasting models or indicating area of concern to policy-makers. The
model can also provide explanations for previous price changes to the public or to politicians.
f. Scenario analysis
How will differential technological, energy, or environmental policies change the expectations for
the future? What capacity decisions are the best given the uncertainty surrounding these policies?
Scenario analysis could be used to consider very specific scenarios; for instance, what is the price
impact of losing the Suez Canal for oil transport? Such an analysis would incorporate more details
4
on the oil transportation infrastructure but may simplify other aspects. Scenario analysis is also
commonly used to check whether a certain policy/decision has a robust performance in different
cases. Usually in these cases, the relative (percent) change in the variables is more important than the
absolute values.
g. Providing common oil price forecasts to be used by many groups
In many net exporter countries oil plays a significant role across different sectors of the national
economy. One modeling goal may be to provide a reliable forecast to be shared across the different
sectors. Having a central agency to provide these forecasts helps other sectors’ planning efforts be
more consistent. It also reduces wasted effort from redundant modeling projects.
Based on these possible objectives, this paper focuses on the medium-term and long-term models
and presents research gaps in these models; the objectives require different modeling approaches.
While it may be possible to reconcile some of the goals into a single project with enough resources
and time, the approach most likely to be fruitful would be to define, up front and as clearly as
possible, the most critical objectives of the modeling effort. Not only will this lead to a better
selection process in terms of methodologies, data sets, and personnel, but it will also serve to
establish a set of metrics useful in evaluating the value of the model outputs. The choice of any of
the above topics could lead to a set of specific and productive research topics.
2.2 Modeling approaches
Markets are dynamic environments not only in the sense that the agents’ behaviors change
dynamically but also the market structure and agents themselves change. Models representing such
markets should also evolve through time. Periodically, the modeler studies the important drivers of
the market and incorporates them into a model. While existing model parameters are updated by
new data, modelers add new drivers to their model.
A five-step process can be established for keeping the models dynamic and consistent with the new
situations. Although based on the nature of the issue some steps have been emphasized more in the
literature, following all steps is a recommended approach.
1) Identify new issues
The first step is to search through observations to find the ones that do not match the model’s
expectations. These factors become candidates to be incorporated into the model. Based on data
availability, the relevant data with possible causal relationships are also considered to determine the
source of the observed anomalies. Examples of such breaks in the status quo in the oil markets
include: shifts in the OPEC collective decision-making process, price boom of 2008, and the
commencement of the financialization era.
2) Develop a theory
A next step is developing a theory of why such observations might have happened. At this step
economists make simple models (usually not even adjusted to real world data) to explain the
underlying economic forces that could possibly generate the observed outcomes. In some models
5
this step is skipped (especially in reduced form models), and a model is generated solely based on
econometric analysis.
3) Perform econometric studies
The next step is to analyze data statistically to check whether they match the theory or not.
Econometric models are also used to suggest possible causal relations that may have contributed to
the observed issue. Statistical tests on OPEC behavior or studies on the relationship between
investors’ positions and prices are examples of studies in this step. Since there is often not enough
available data, a firm conclusion is hard to attain at this step. Moreover, any finding of Granger-
causality between selected variables may not indicate the presence of structural or economic
causality.
4) Incorporate responses in a partial equilibrium model
Once a theory is developed and system parameters are adjusted accordingly, a new model is created
to analyze the original issue. This model may simplify many important drivers of the market and
highlight the specific issue considered. An example of a model at this step is one that tries to model
strategic interactions between producers by considering individual output adjustments in great detail
but that represents the whole world demand with a simple demand curve.
5) Incorporate responses in the main model
Finally, a module will be integrated into the main model if the phenomenon is considered of great
enough significance. It is important to note that it is not necessary to incorporate all issues into the
main model (step 5). Models trying to provide counter-factual analysis or stochastic simulations
could remain at step 4. Many issues are just studied through the first three steps in order to find out
whether some variable influences another variable. Statistical tests on OPEC behavior are examples
of such models. Some studies are done at step 2 to find out how current behaviors compare with
optimal strategies. Examples include models trying to determine the optimal levels of oil
import/export of major consumers/producers. As will be discussed later, financial models are also
mostly limited to the first three steps.
In this paper, important issues in today’s world oil market and their current state of
development are discussed. Some issues are developed enough to incorporate into a forecasting
model, and others are at the level of theory development and econometric studies. There are various
scientific tools that could be used in each step. Table 1, shown in the Appendix, summarizes
different modeling approaches that could be employed in developing the model. These approaches
could also serve as a framework to think about and model new issues.
2.3 Key Modeling Design Principles
The process taken for model development can be as important as the choice of objectives and
selection of relevant theories. Especially as the intended project grows in scope, it is important to
have a good process for designing, evaluating, and improving models.
6
2.3.1 Design
Models should be designed to be implemented quickly while producing useful results. Ideally,
models are based upon a relatively simple organizing framework and made more complicated over
time. Focusing on first building a minimum-viable product before modifying the model to contain
the full set of desired features has a few benefits. Building the simple model gives the researchers an
opportunity to understand the core mechanics and assumptions embedded in the modeling
approach. It allows the developer to evaluate the technical implementation earlier in the process. In
general, fast and minimalist iteration allows for earlier evaluation of assumptions, methods, and
technical choices resulting in adaptation before large investments in time and resources have been
made.
We strongly recommend that a supply-demand framework for balancing markets be applied. At
a minimum, this condition relates to a partial equilibrium in the oil market. It may be permissible to
allow temporary disequilibrium in short time-steps (if supply stockpiles are not explicitly considered,
for instance), but as a general principle partial equilibrium ensures the balance of material and
provides a well-grounded theory for the generation of prices. Initially, supply and demand
conditions might use simple parameters such as constant elasticities. As it becomes necessary (or
with the presence of greater data), supply and demand can be modeled in greater sophistication. For
example, world demand and non-OPEC supply can be modeled with elasticities while OPEC is
allowed to endogenously choose production to satisfy some objective function (profit, welfare,
revenue requirements).
Having an integrating model representing the major interactions and individual sub-modules for
evaluating detailed sub-systems can be a good way to manage model complexity. As models increase
in complexity, model software decomposition allows the complexity to remain manageable by the
modeler or team. Not only does a modular approach allow for the logical distinction of system
processes, but it also enables a diverse team to work together on a single project. Experts can create
sub-modules based on their unique knowledge without needing to understand other elements in the
system. Breaking code into sub-modules provides the added benefit of streamlining the revision
process. For instance, if new battery technology changes the choices available for personal
automobiles, only the ‘light-duty vehicle’ module may need to be updated leaving the other sub-
modules untouched.
2.3.2 Evaluation
Building a model is an ongoing balance between the desire for perfection and the cost of adding
additional features or modifying certain elements for greater fidelity. Models can be evaluated in a
variety of ways, and subjecting the model to open criticism will help ensure the required level of
quality.
The first method of evaluation starts within the team and entails a thorough project review. Is
the model able to obtain insight into the targeted question? Does it contain all the critical elements
from the initial project vision? In principle, such a review is no different from any project post-
7
mortem. The review serves as an opportunity to gather the full set of stakeholders involved in the
project and provides a context for looking critically at the project.
The next step in evaluation is to submit the model for discussion and review through
conferences and publications. With the attention of a greater number of experts and viewpoints,
many hidden assumptions, inappropriate data choices, or specification errors can be brought to the
surface. Pushing this process to the fullest extent, it can be beneficial to make the project completely
open source: sharing not only the conclusions and outputs of the model but also the very source
code used to compile it. Giving the community access to this code allows for the detection of
programming as well as model-logic errors and might lead to useful modifications to the code that
can be absorbed into the core development effort.
Finally the model output can be benchmarked against other models and against historical
results. Given a set of consistent inputs, how do the model’s outputs compare to existing well-
known models in use? Can differences be explained, and if so is it due to defensible choices in the
current modeling effort? If the purpose of the model is to provide forecasting, it is also useful to
view the model’s performance against historically observed data. As a regular publication the Energy
Information Administration (EIA) prepares a report of the performance of NEMS as compared to
historical effort in an effort to understand the disparities and to aid in modifying the model to
improve future performance (EIA, 2010). This useful exercise demonstrates that their model
produces far more accurate results for total U.S. petroleum consumption than for crude oil prices.
Absolute errors in projecting actual consumption ranged from a minimum of 2.0 percent to a
maximum of 9.2 percent, with an average for all years of 3.2 percent in annual outlooks prepared
between 1982 and 2009. Absolute errors in projecting actual crude oil prices were substantially
greater, ranging from a minimum of 6.3 percent to a maximum of 126.3 percent, with an average of
52.5 percent during the same years. The errors in forecasting oil prices are quite large and
underscore the difficulty of tracking this variable due to its high volatility relative to the smaller
fluctuations in observed oil consumption.
Another important process in developing a model’s credibility is participation in formal model-
comparison studies, such as those conducted by Stanford’s Energy Modeling Forum. Modeling
teams adopt similar assumptions that represent individual cases. After simulating these scenarios, the
teams report their results back to a central organizing unit that distributes charts comparing findings
across individual models. These comparisons become even more meaningful when they are
conducted in the presence of major model-using, private and public organizations, which will help to
establish communication between model developers and users and to emphasize how useful the
models are for addressing key issues, problems and decisions.
2.3.3 Improvement
Simplicity, flexibility, and expandability are three key issues in software development. Production
technologies and consumer choices could be approximated by many complicated functional forms,
but when problem dimensions increase, complicated descriptions make optimization problems less
tractable numerical convergence problems arise and run times increase exponentially. Similar
issues occur for data storage and management when more detailed data is required. It is strongly
8
recommended that software development experts are engaged before undertaking a major modeling
effort.
To conclude this section, two sides of modeling issues should be highlighted. On one hand it is
recommended to start with a simple modeling approach and on the other hand it is recommended
to have an image of the final (potentially complicated) design from the beginning. At first sight this
seems like contradicting advice, but the two frames of reference should coexist in a dynamic model
development program to ensure expandability of the model. Without a simple methodology the
model gets intractable very early and without a scalable software environment the model loses its
expandability quickly. Incorporating lots of detail will require a commercial model solver. Therefore
starting with the simple methodology and implementing it in a solver environment like General
Algebraic Modeling System (GAMS) could be an appropriate start in this way.
Last but not least, the preparation of comprehensive documentation is imperative from the
early stages of model development. A well prepared and up-to-date set of documentation is a key
factor for both the validation and sustainable use of models. It will also ease further modifications
and updates and facilitate the sharing of the software with other institutions.
3. MEDIUM-TERM AND FINANCIAL MODELS
For the purposes of this section, we categorize models based the time horizon of the modeling
effort. Although there are various points of views on terminology, here we define short-term time
horizons ranging from hours to at most one month. Medium-term horizons span from one month
to three years, and long-term models consider time horizons of three years or more. Each time
horizon is considered separately due to the different amounts of financial activity and market
liquidity observed in each range.
Besides academic researchers, there are three major groups interested in modeling. Financial
players and specially hedge funds are particularly interested in short-term price movements, taking
part in commodities markets to manage their portfolios’ short-term risks or to gain benefit using
their private information about the market. A second group is comprised of the industries with
physical oil needs, such as airlines, who are more interested in medium-term price changes to
manage their risks and usually do not seek benefits by engaging in short-term trading. The third
group includes major consumers and producers of oil whose long-term prospects depend strongly
on even longer-term oil trends. This group is interested in medium term and long-term models that
analyze long-term trends affecting capacity decisions, profitability, and competition. Forecasting
medium-term price trends is interesting to producers not for the purpose of making physical
expansion decisions but to decide daily/monthly production levels. Managers need to explain the
medium-term price changes and determine whether the observed trends are expected to continue.
As discussed earlier, not all modeling approaches are intended for forecasting purposes. Many
models are developed to understand and/or exploit anomalies observed in the market. Financial
forces have recently been introduced to the market, and the current models are at the stage of
developing theories and performing statistical tests. Such models and analyses provide a better
normative and descriptive understanding of the medium-term market. In this section several
improvements in medium-term modeling are introduced. First the problem of modeling financial
9
activities is discussed. Following that, market volatility, medium-term hybrid models, information
signals, and high-impact events are considered.
3.1 Futures Market and Financial Activity
During the previous decade, there has been an immense growth in investment in commodities
markets, coinciding with a huge increase in oil prices. There have been numerous studies trying to
identify the role financial investors played in the 2008 price boom by studying investors’ positions.
Usually these studies are primarily limited by data availability. Hence, as more detailed data become
available, more precise results may be obtained. Although there is no unanimous agreement among
researchers on the effect of financial activities, some do find evidence that such activities may have
influenced the market to some extent.
3.1.1 Modeling the effect of financial activities
In developing a medium-term model, an interesting issue is whether financial activities should be
expected to influence the market in the future. Will these activities push the prices to higher levels
than the equilibrium? If true, assuming an efficient market, it is expected that investors knowing
such effects will take positions to benefit from the gap in prices, thus offsetting the previous effect
and pushing the prices back to their fundamental levels. Therefore, one could expect not to see such
effects in future. On the other hand, bubbles have been observed throughout the history of world’s
financial markets. Real world experience shows the possibility of bubbles recurring in the future, but
at unexpected intervals.
In fact, answering the question above and implementing it into an oil model is an interesting
and challenging task to accomplish. Considering the disagreement in the literature on this issue, it
seems that more studies are needed prior to adding such a module to oil models. The chronology of
modeling approaches suggests that first econometric studies should be performed on an issue and
that only later can the results be incorporated in a model. The state-of-the-art studies on this topic
employ financial econometric techniques and use high-frequency private data to analyze the
problem. Furthermore, to forecast the effect of financialization, one should first forecast the level of
future financial activity, which is itself a complicated problem. While this research gap presents a
large opportunity for future modeling efforts, it may be more prudent to first invest in more
econometric studies in this area.
3.1.2 The role of physical oil producers in financial markets
Another important issue in medium-term models is identifying the role of producers of the physical
product in the financial market. The producers have the option to sell oil directly to their consumers
(e.g. through a bilateral contract), to participate in the futures market, or to sell on the spot market.
Research in this area is limited by a lack of data, especially for bilateral contracts. However, from the
perspective of a single producer, one can model the effect of signing forward contracts outside the
market or participation directly in the futures market. Part of the complexity in modeling these
decisions arises from the fact that a large amount of supply in the futures market can change the
market price. Futures contract prices, as well as any other contract price, are controlled by supply
10
and demand. Increasing the supply of futures contract by a major exporter could decrease the
futures prices.
Although there is not enough public data to model the effect of real commodity producers’
active participation in the futures markets, Saudi Arabia may be able to model these effects using its
own private data. Controlling for all price drivers of futures prices, Saudi Arabia can find the effect
of its sold contracts (open interest) on the market. Such a model can help Saudi Arabia optimize the
extent to which they should participate in futures markets either for the objective of maximizing
profits or for stabilizing prices.
Physical commodity producers participate in the forward and futures markets mainly because of
strategic and hedging incentives, although many producers are also known to increase their hedging
volumes when they believe prices are “high” and vice versawhich is a form of speculation (Stulz,
1996). Although there is no consensus about the effect of taking positions in forward contracts,
scholars have modeled firms’ incentives under different market structures such as competitive or
oligopolistic, different levels of risk-aversion, and different time horizons. It has been generally
believed that, holding all else equal, expansion of forward contracts is followed by aggressive
behavior in spot market (i.e. firms produce more and thus market price decreases) because there is
less unmet demand to satisfy and less incentive to push prices upward. Allaz (1992) has shown that
depending on the relative strength of the hedging and strategic motives, a firm’s optimal position
taken in forward markets may be short or long. A producer going long (buying commodities in the
forward market) is rather a counter-intuitive result. However, such actions could increase the firm’s
market power in the spot market.
On the other hand, participating in forward contracts creates an additional stage of competition
for firms. Brandts, Pezanis-Christou and Schram (2008) consider a Cournot (quantity) competition
and argue that participating in forward contracts reduces the spot prices. Liski and Montero (2006)
considering an infinite-horizon problem, argue that participating in forward contracts may even help
firms to obtain collusive profits.
Besides the mentioned models that analyze the problem for a general commodity, there are also
many studies on the effect of forward contracts in specific industries. Kavussanos, Visvikis and
Batchelor (2004) study shipping industries and document that introducing forward contracts
(Forward Freight Agreement) has had a stabilizing impact on spot price volatility. Forward contracts
in electricity markets are also deeply studied. Wolak (2000) demonstrates that forward contract
holdings reduce incentives to bid to raise the spot market price. He argues that facing the same
residual demand curves, firms signing forward contracts can gain a less volatile and lower average
cost pattern of production throughout the day. Wolak (2003) also highlight the possible preventive
role that forward contracts could have played in the California energy crisis. Lien (2001) analyzes the
effect of forward contracts under various assumptions on market competitiveness.
Comprehensive analysis of the role of futures contracts in the oil market has been limited, and
the insights derived from other markets may have limited applicability to the oil market. For
example, the well-studied electricity market has a major difference with oil market; oil can be stored
whereas electricity is a largely non-storable commodity.
11
3.2 Modeling Volatility
Oil price volatility is of great importance both to producers and consumers. Price volatility increases
the revenue risk of exploration or capacity expansion projects for producers, and it increases the
cost risk of alternative technologies and efficiency projects for consumers. As a result, knowing the
effect of different market activities (either by the financial, industrial or production sectors) on price
volatility could be helpful in policy analysis for stabilizing the market.
Many studies performed on price levels could be extended to volatility levels. The volatility can
be calculated based on short-term historical data or estimated using volatility implied by oil
derivatives (Szakmary and et al. 2003). For its Short-term Energy Outlook, the Energy Information
Administration uses implied volatility to estimate future volatilities (Energy Information
Administration 2009). Considering the flat forward curves and the co-movement observed among
spot prices and near-dated futures contracts, the estimates of volatilities using the two methods are
expected to be close. Sanders and Irwin (2011), who consider both forward looking implied
volatilities and estimates based on historical data, find evidence that larger long positions by index
traders lead to lower market volatility in a Granger sense. This result is contrary to the popular belief
that index traders’ activities increase market volatility.
While estimating volatilities using either historical averages or through implied volatilities makes
sense for the purpose of forecasting in intermediate term models, the methods are less useful when
considering models with a longer time horizon or where agents in the model are attempting to
influence (stabilize) volatility. Since implied volatility relies on large volumes of information from the
financial market, using some form of implied volatility in a model where market actors are trying to
control volatility is a hopeless endeavor. Meanwhile, using historical averages is insufficient given
that, by construction, volatility must be allowed to change in such a model.
Another approach is to build some kind of stochastic structural model. In such a model, Monte
Carlo methods could be applied to simulate supply and demand shocks to the oil market. A player in
the market wishing to influence the volatility would choose its investments (in capacity or reserves
for instance) and its response strategy (how much production / reserve draw-down) as a response to
market shocks.
Indeed, even for models where volatility is not an explicit concern, a model might need to
capture how volatility and price shocks lead to changes in the behavior of consumers and producers.
The use and production of oil is heavily tied to existing capital stock and capacity investments, and
price shocks even over a relatively short time frame can have lasting impacts on demand and supply
for years to come. Beyond the oil market, the shocks may have impacts in terms of policy
interventions, investment in alternative energy sources, or on trade balances in the world economy.
When detection volatility patterns in the data is more important than providing economic
explanations, a powerful, easy to use, and promising approach is to apply artificial intelligence
technology with neural networks to model and forecast global oil market volatility. Artificial neural
networks can capture the factors and parameters influencing the market volatility more accurately
than conventional methods. The King Abdullah Petroleum Studies and Research Center
(KAPSARC) has recently initiated a project to apply artificial neural network approach for modeling
and forecasting oil and gas markets volatilities.
12
3.3 Medium-term Hybrid models
Medium-term hybrid models combine structural models and financial models with the intent to have
a model that provides both long-term and short-term forecasts. However these two types of models
are very different in their natures. Long-run models usually focus on supply and demand details and
provide equilibrium prices. On the other hand, short-term models usually focus on the time series of
the prices and try to forecast them based on the observed trends, which could be different from the
equilibrium levels. If it is desired to combine these two modeling efforts into a single model, one can
consider them as two different blocks working in parallel with minimal interaction.
However, a more interesting and useful hybrid model is a model in which the short-term and
long-term modules interact with each other. Such hybrid models could enhance the medium-term
forecasts by reconciling the data from futures markets with the projections by the long-term model.
Developing these types of models is one of the most important research gaps. A first approach in
this regard is considering both types of variables in an econometric model. Since the number of
variables increases and the interactions among them becomes more important, vector auto-
regressive models fit the problem’s nature better. Gaining popularity recently, structural auto-
regression (SVAR) models share the benefits of both the conventional structural models and the
reduced form VAR models. In SVAR models, either parametric or sign restrictions will be imposed
to take economic relationships into account. In this manner, Kaufmann and Ullman (2009) study
the co-integration relationship between spot prices and futures prices, and Kilian and Murphy (2010)
construct a SVAR model including oil production, inventory, price, and a proxy for real economic
activity.
Similar to the conventional econometric models, SVAR models can be used to represent a
specific section of the oil market in a simulation or optimization model. However their main
advantage is taking the available data into account and specifying the significant relations between
variables. The main drawback is the SVAR model’s reliance on, and sensitivity to, identifying
restrictions that are difficult to validate.
Another group of medium-term hybrid models that are able to provide an economic framework
for describing the observed price trends incorporates theories of commodities and storage. (See
Deaton and Laroque (1996) and Routledge, Seppi, and Spatt (2000)). These studies specifically
consider storage agents in the commodities markets and determine how the levels of inventories
change with uncertainty and how forward curves appear in such settings. Routledge, Seppi, and
Spatt interpret the concept of convenience yield as the value of an option that storage agents own
and may exercise whenever they wish.
One can also consider medium-term hybrid models that are calibrated to base case forecasts of
a long-term model and that adjust based on the flow of the new market information and short-term
economic responses. The new market information could include price observations from futures
markets, forecasts in demand growth, supply shocks (e.g. 2011 interruption of Libya’s production),
etc. This view to hybrid modeling requires both short-term and long-term estimates of response
parameters (e.g. elasticities) to simulate price responses. From an engineering point of view, the
13
method is similar to considering a model tuned to a specific working point. The model takes
information signals as input and generates price and quantity paths as the output.
Among the available financial information, financial derivatives such as options and futures
contract prices and volumes have captured significant amounts of attention. Since these variables
convey information about the future they have been considered as a first step of incorporating
financial data in oil models. However, Alquist and Kilian (2010) find that no-change forecasts are
preferred to those derived from the price of crude oil futures. This result is explained by noting that
the futures prices are not simply the expectation of the prices in future.
Convenience yield and risk premium both contribute to the deviation of futures prices from the
expected price that will prevail in the future. In theory, if one could model both these two factors,
when combined with market data one should be able to obtain accurate estimates of expected prices
in the future. Pagano and Pisani (2009) document significant time-varying risk premium and have
developed two such proxies. One is the degree of capacity utilization in US manufacturing to
represent the US business cycle. The other is the oil inventory to represent oil market tightness.
Pagano and Pisani study historical values of these proxies and prices. If a risk premium proxy can be
found that is easy to forecast, it is possible that this method can be employed to more accurately
forecast prices based on futures contracts.
3.4 Information Signals in Oil Markets
Not all announcements are credible or significant enough to move spot prices noticeably. It is
important to distinguish between relevant, credible announcements and ones that are ignored by the
market. An important step in conducting this analysis is to consider the incentives of the issuers of
information signals: specifically, whether those objectives are aligned with the truthful revelation of
information. For market actors that do have a stake in the market, the question is how to make
credible announcements to the market.
Announcements vs. Signaling
In economics, signaling is the name given to the actions whereby one party (termed the agent)
conveys some information about itself to another party (the principal). For example, in Spence's well
known job-market signaling model, employees send a signal about their ability level to the employer
by acquiring certain education credentials. In such a framework the case where different types of
agents can be reliably distinguished based on their signal is of particular interest. These models work
properly when the cost of signaling is less for the better agents. (For a more precise definition see
Spence (1973)). In contrast to an announcement, which can be viewed simply as a statement of
information, a signal has a built in cost that ensures agents will not desire to send a signal contrary to
their type. As a general rule, signals that are not costly are not credible.
In the oil market, the flow of information about the future changes the price today, and
announcements have the potential to affect production and capacity expansion decisions. If the
entity making the statements has interest in those decisions, it might be to their benefit to announce
false information. Being able to distinguish between true signals and ‘cheap talk’ is an important
14
criterion for evaluating the informational content of a given announcement. As an example, consider
the following hypothetical case.
Suppose a large oil-consuming nation announces that its oil demand will be ten percent more
than the previously forecasted values for the next year and this announcement is believed credibly by
the whole market. As a result, the spot prices increase, inventory levels will rise and producers will
choose to engage in more aggressive capacity expansion. Next year when the market observes that
the oil demand is still at the old level, spot prices will decrease below the level that would have
existed in the base case without the announcement. This is due to the elevated inventory levels and
increased capacity levels. As this simple example illustrates, one should check whether it is to the
benefit of a consumer to send a wrong signal to market.
Of course, the reverse case could be considered for a producer. With price and revenue
assumptions in national budgets, producers are signaling their expectation of market prices and their
production decisions to the market. The cost of overestimating these values is the internal budget
that result deficits. However, by overestimating the prices, market expectations of prices will
increase and higher prices are more probable to occur.
This misalignment of incentives between market players weakens the value of any information
release even when the intent is to release information truthfully. It is well known, for example, that
even in cooperative games there will be no communication between rivals whose interests are
diametrically opposed. By understanding the perception of incentives in the market, a producer can
develop signals that create market credibility for their announcements (e.g. signaling reserves, spare
capacity, etc). In such cases, it might be beneficial to the agent to verify its announcement by taking
an action that is too costly for a non-truthful agent to take. This action might be increasing
information transparency, providing technical details, submitting to external auditing, or operating
the unused facilities even when they are not really needed.
3.5 High Impact Events
A unique challenge for any modeling effort is to understand the influence that large and unexpected
events might have on the system. While most modeling efforts do not include such events and
instead focus on business as usual economic interactions, given the historical role of wars and
geopolitical tensions in creating shocks to the oil price system, it may be worth reconsidering how to
include such shocks. Indeed, in the past three years many such events have occurred: from the Deep
water Horizon oil spill in the Gulf of Mexico, to the oil supply disruptions in Libya, to the partial
meltdown of Fukushima Daiichi nuclear power plant in Japan. With such a handful of large events,
it raises the uncomfortable question of whether such energy shocks are truly rare on a global scale.
On April 20, 2010 an explosion occurred on the Deepwater Horizon drilling rig that was nearing
completion of the Macondo well in the deep water of the Gulf of Mexico off the coast of Louisiana.
As a result of the explosion and ensuing fire, eleven crew members died. In time, an estimated four
million barrels of oil spilled into the gulf, making the event the largest ever accidental oil spill.
Ultimately, the spill led to billions of dollars in economic damages and unknown environmental
consequences. The event sparked a moratorium on deep water drilling and a reconsideration of U.S.
15
energy policy (National Commission on the BP Deepwater Horizon Oil Spill and Offshore Drilling
2011).
The methods of obtaining liquid fuels are becoming increasingly reliant on advanced and capital
intensive technologies. From deep water drilling, to the processing of oil sands, to advances in
refining, and larger ocean-going oil tankers, the oil market is changing its risk profile. While
engineers are constantly working to enact better control systems and reduce the chance of any single
failure, the potential for damage from any single catastrophe is increasing.
It is still an open question what these increased engineering risks mean for the future price and
policy paths in the oil market. For instance, beyond the immediate damages and loss of supply from
the Deepwater Horizon explosion and leak, what impacts will the event have on future deep water
drilling? Will it increase the cost of deep water drilling? Will it change the strategy of oil companies
to look at other prospects? Will it lead to specific policies in the United States designed to limit
gasoline consumption?
Even where such large-scale economic and environmental disasters may not be a concern, it
remains unclear how the increasing level of technical difficulty in crude oil extraction will influence
the amount of volatility in energy prices. The ‘lumpiness’ of production is increasing with the trend
in production towards more complex megaprojects, increasing the supply impact of a single outage
(or addition) and potentially leading to more price volatility (Skinner 2006).
The consideration of rare shocks in an energy model emphasizes the contrasting objectives of
forecasting and planning. These events are by their very nature unexpected, and where they can be
predicted with any regularity such a predictive model would be better applied directly in averting
such disasters rather than in forecasting their influence on oil price movements. On the other hand,
for the purposes of planning capacity expansion, domestic policies, and regulations these disruptions
can be critical considerations. A planner can choose to build policies that are robust against large
shocks--for instance, the excess capacity maintained by Saudi Arabia has proved useful in the past in
stabilizing prices against large shocks to supply. The precise events cannot be predicted in a
forecasting sense, but models can help determine how to respond when the events do occur and to
build policies that enable a more robust system.
4. LONG-TERM MODELS
Even though there is a deep and lengthy literature modeling the price of oil over the long run, many
gaps exist in this area, especially from the perspective of oil-producing nations. The global economy
is seeing dramatic changes, and the oil market is not isolated. Developing economies are
experiencing rapid growth, with China standing out in terms of increased automobile ownership and
gasoline demand. The questioning of traditional assumptions about OPEC behavior and the rise of
new significant oil-producing nations such as Brazil have led to reconsideration of the correct
methods to model producer competition and behavior. Production technologies are reaching into
new territories and expanding the amount of oil that can be recovered from existing reservoirs.
Finally, the threat of anthropogenic climate change looms large, challenging the world to balance the
benefits of services enabled by fossil fuels against the long-term consequences of their intensive use.
16
Based on these challenges and trends, we will discuss five major research gaps relevant to long-term
oil models: 1) Demand behavior, 2) Modern OPEC behavior, 3) Strategic behavior of non-OPEC
producers, 4) Producer welfare analysis, and 5) Resource depletion.
4.1 Demand Behavior
With the exception of the large computational models, most oil models do not have a very
sophisticated or detailed representation of the demand side of the market. Great effort has gone into
understanding the notion of supply security, but recent events have seen drastic changes in demand
as well: from the rapid rise in leading up to the summer of 2008 to the subsequent drop during the
global financial crisis. Global oil demand grew by 3.8% in 2004 and declined by 1.5% in 2009. It
declined by 0.5% in 2008. Understanding these dynamics can be useful not only in explaining recent
price dynamics, but also in exploring the impacts this demand insecurity has on oil-producing
nations. Volatility in demand will have impacts on the welfare for producers and will change the
nature of their capacity investment decisions. Three major topics in demand behavior stand out as
needing more exploration: the high degree of demand growth in developing countries, the
asymmetric response of oil demand to price changes, and the role of technology change in altering
the energy intensity of oil-consuming activities.
4.1.1 Growth in developing countries
Even during the economic downturn in 2008 and 2009 when world oil demand contracted, China’s
oil demand continued to grow. As the global economy recovers, the expectation is that demand
growth will continue in developing countries. In the NEMS reference case, while OECD demand is
expected to remain relatively constant, non-OECD consumption is expected to grow at an average
yearly rate of 1.9% through 2035--implying almost a doubling of consumption over the years 2008
to 2035 (Energy Information Administration 2011). While demand is expected to grow strongly,
capacity is expected to be largely constrained, at least in the short run. The production shortfall is
expected to be so severe that constraints on liquid fuel supply have been highlighted by the U.S.
military as being one of ten major trends influencing global security (United States Joint Forces
Command 2010).
Understanding how such steep global demand growth will influence oil prices, technology
investment decisions, energy policies, and the potential spillover from international conflict is an
important matter for both oil-consuming and producing nations. It is not enough to simply know
that resource demand is growing so fast that supply cannot keep up. The shortfall will have
consequences that will be manifest in the price of oil, resulting in differential effects across demand
sectors and markets. How will the prices for different crude products change? How will increased oil
prices affect GDP growth? Projections differ in the degree of demand expansion, and even the
previously referenced military document views the shortfall as a temporary matter that can be
alleviated with capacity expansion in time. Nevertheless, with strong demand growth expected to
prevail in the coming decades, greater detail is required to understand demand trends in growing
economies.
17
4.1.2 Asymmetric response
Even short-term fluctuations in price may have longer term impacts on oil demand. Empirical
evidence illustrates that consumers as well as oil-consuming nations respond differently to price
increases and decreases. Consumers are more sensitive to price increases than decreases (Gately and
Huntington 2002). Thus after a strong run up in prices and a subsequent return to the initial price
level, demand will actually end lower than if prices had remained constant. It also has been shown
that the elasticities of oil demand with respect to income are not symmetric with a stronger response
to income increases than decreases. Thus if a nation has an increase and subsequent decrease in
income levels, oil demand will remain higher than in the beginning. Both of these asymmetries could
possibly be explained by capital stock choices made by individual consumers, where high prices lead
to investment in more efficient technologies and elevated incomes lead to first-time vehicle
purchases. Given the high levels of volatility observed in oil prices, it can be expected that models
using smooth price paths or symmetric elasticities over the course of the projection horizon will
overestimate the resulting long-run level of demand (provided all other factors are modeled with
complete accuracy).
Asymmetric effects of oil price movements can be found at a macro level as well, with positive
price shocks having greater impacts on GDP than price drops (Hamilton 2003; Hamilton 2009). In
this light, models that include a smooth price path may be overestimating worldwide income growth,
and in turn oil demand. For consuming nations, incorporating such asymmetries can be important in
understanding the losses associated with such shocks and determining appropriate policies to
mitigate the effects. For exporting nations, using GDP shock asymmetries will enable measuring the
chilling effects on demand from price instability, helping to inform capacity (and excess capacity)
decisions.
4.1.3 Technology change and product substitution
Crude oil is refined to serve a variety of different demand needs: from plastics, to fuels, to
lubricants. Different crude products have distinct market dynamics, and understanding how these
product markets will change in the long run is vital to understanding the future of oil exploration,
refining, and crude prices. The end uses of oil can each be subject to their unique set of policies,
efficiency improvements, or technological disruptions.
Analyzing the effects of technological change on the energy intensity of the services industries
is best done with a model that integrates rich information on specific technologies within a unifying
economic framework. Having a partial or general equilibrium model is particularly important when
the affected technology represents a significant portion of oil consumption and, therefore, has the
potential to have noticeable price impact on other consumption sectors. A promising approach that
considers only the use of automobiles and trucks in the United States is taken by Schäfer and Jacoby
(2006). Their model combines MARKAL (a bottom-up technology modeling framework that will be
discussed further in the concluding section) and EPPA (a general equilibrium emissions model
developed at MIT) to create a system that allows for technical detail as well as full economic
interactions. Extending the model to a global scale or to a different consuming sectors are promising
research directions.
18
4.2 Modern OPEC Behavior
In late 1970s, many studies modeled OPEC as the monopolist in the world oil market. Market
power assumptions for OPEC in models developed in the late 1970s and early 1980s shows the
strong consensus about OPEC’s market power. In the 1980s, supply disruptions were dominated by
passive supply shocks (e.g. Iran-Iraq war) instead of active reduction of production caused by
exercising market power. Following Griffin’s (1985) work on testing OPEC behavior, Smith (2005),
reviewed OPEC behavior studies in the two decades following Griffin’s work. Smith argues that
most of these studies are not conclusive. OPEC behavior has evolved through time and now is
much more complicated than in the 1970s. Other market factors have also changed: non-OPEC
production has expanded, unconventional resources have become economically feasible, demand
relationships have become more complicated, new major consumers have emerged in Asia, and the
market for crude oil is now much more liquid.
In the previous two decades, the global view of OPEC has changed. OPEC is no longer
considered definitively as a cartel exercising market power as it was considered in 1970s and 1980s.
However, there is also now less research effort to model OPEC’s behavior either econometrically or
theoretically. There is a need for more complex and detailed theoretical models to describe OPEC
and these models should be tested by the detailed data that has not been available in previous years.
OPEC’s stated mission in it Statute is “to coordinate and unify the petroleum policies of its
Member Countries and ensure the stabilization of oil markets in order to secure an efficient,
economic and regular supply of petroleum to consumers, a steady income to producers and a fair
return on capital for those investing in the petroleum industry.” (Organization of the Petroleum
Exporting Countries 2008)
Key questions in modeling are how this coordination between OPEC’s members functions,
how successful OPEC has been coordinating policies of its members, and how effective OPEC is
today in coordination to achieve the mentioned goals.
Among the factors influencing OPEC’s decisions are its members’ financial needs. Figure 1
demonstrates how OPEC’s per capita revenue from oil was lower in 2008 when compared to 1981.
The other fundamental difference between these two years is the dependence of members’
economies on oil. In the 1970s the nations did not expect such high oil revenue, so the money
coming in was largely surplus, where today the nations are making sophisticated assumptions about
oil prices and sales volumes and are including that in their government budgets--thus any reduction
in prices could lead to critical revenue shortfalls.
Brander and Lewis (1986) study the effect of financial liability on behavior in an oligopolistic
product market. They develop a theoretical method in the context of financial debt, but it could be
generalized to apply to the role of presumed oil prices in oil producers’ national budget. They argue
that agents behave more aggressively in the product market (i.e. decide to produce more in a
Cournot setting) if they have financial liabilities.
The politics of OPEC members are also more complicated than before. The Iraq war of 2003,
general instability in the Middle East and North Africa in 2011, strained political relations between
some members, different levels of concern about global energy security, emerging major oil
19
consumers in Asia, and quota enforcing mechanisms are also factors that could be considered in a
model.
Many studies have modeled the conflict of interests between OPEC members. One source of
this conflict has traditionally been the different interest rates they use to discount their cash flow.
Hnyilicza and Pindyck (1976) separate OPEC in two groups based on their discount rates and search
for the Nash Bargaining Solution. Adelman (1982) has provided a model studying the effect of
different interest rates on both production decisions and exploration decisions. Other research has
highlighted the different financial needs of countries. Daly et. al. (1982) consider three groups in
OPEC: profit maximizers (with market power), price maximizers (following a target revenue) and
output maximizers (with limited market power).
Figure 1 OPEC Per Capita Net Oil Export Revenues.
Last but not least, the decision on production levels is not the only aspect of OPEC’s behavior
that should be considered in modeling. Smith (2009) suggests that in recent years OPEC has been
more successful exercising market power through capacity expansion investment decisions than
production decisions. Considering the financial needs, it is much more difficult for a producer to cut
back its production levels. In the same time, it seems convenient to cut back from investments. This
helps the monetary needs of the day, though it harms future cash flow. The literature in industrial
organizations provides a framework for the multi-stage problems of competition in investment
periods and production periods. A well-known example of such studies is Gilbert and Harris (1984)
where the researchers model a multi-period investment decision where investments are indivisible
and irreversible. Leahy (1993) also considers an investment decision-making problem and shows
how myopic behavior, defined as not considering other firms’ strategic behavior, could be optimal in
a competitive equilibrium setting.
Considering all of the structural changes in OPEC’s behavior, more detailed data availability on
members, and the importance of political relationships between Saudi Arabia and other OPEC
20
members, developing a more accurate model of OPEC members’ strategic behavior seems to be a
major progressive step. Such a model should allow different production rules for each country and
incorporate more data on financial/industrial needs of these countries. The next section will
describe more on modeling the strategic agents of the oil market in general.
4.3 Strategic Behavior of non-OPEC Producers
A great deal of literature has explored the nature of OPEC market power and strategic behavior..
Given the effect of OPEC market power in the 1970s and 1980s, the subject was the relevant
question for modelers of the oil market.
Nevertheless, the question remains of how other market actors might be exercising market
power. Several non-OPEC nations have been increasingly important in global production including
Canada and Russia. Will Canada become a producer with large market power given its high level of
reserves in oil sands? How will its influence over the market change as more easily exploited
traditional reserves become depleted?
Market power may have influences in different parts of the production chain, not just at the
reserve level. How does market power interact with technically difficult exploration such as deep
water, oil sands, or oil shale? What is the role of drilling and oil field service companies such as
Transocean, Schlumberger, and Halliburton (to name a few) in terms of driving the cost structure
for non-OPEC production? Is the exploitation of deep water reserves limited by constraints on rigs
and key engineering personnel? If technology and firm specific knowledge is expected to play a
larger role in extracting oil from increasingly difficult reserves, understanding how actors operating
in these domains interact with the market will be increasingly important in models of the oil market.
A final concern involves the use of actors in the futures market exercising market power to
influence the price of oil and obtain profits while inducing market volatility. In a pending case, the
Commodity Futures Trading Commission has recently brought civil charges against a set of traders
for illegal profits obtained by manipulation of the price of WTI in 2008 (Jeffrey 2011). It is
important to note that the alleged price manipulation was allegedly achieved through stockpiling and
subsequent release of physical oil supplies, not through acquisition of futures contracts or other
financial investments. However, this incident illustrates how large holders of physical commodities
who are with offsetting positions in the futures market can influence perceptions about the relative
scarcity of the good and in turn shape expectations over the future.
4.4 Producer Welfare
Many of the market-power models treat OPEC production decisions as if they were made by a
profit maximizing firm or cartel. When trying to understand the impact of OPEC production
decisions on global oil prices and consuming nations, such a formulation may be an adequate
approximation of the decisions made by the organization. However, in reality as sovereign nations,
political as well as economic concerns drive decision making. Oil-producing nations may constrain
prices in order to maintain demand growth or may sell oil at a lower price in the domestic market
than the international market. It may make more sense to view the nations as maximizing welfare
rather than maximizing profit.
21
Unfortunately, when moving from models that consider profit to ones that try to measure
welfare, the modeling techniques increase in complexity and require greater information on the
national economy as a whole. While recent models in the literature have started to approach this
important question, a great deal of work remains to be done in this area.
Celta and Dahl (2000) model OPEC members as maximizing social welfare which is composed
of two elements: profit from oil exports and consumer surplus from domestic consumption. Their
model shows that OPEC members should charge significantly less to consumers in the domestic
market than the price in the export market. The model adds greater nuance to the objectives of
OPEC members, but still treats oil in isolation from the rest of the national economy.
De Santis (2003) focuses on Saudi Arabia and builds a computable general equilibrium model to
estimate the welfare impacts of oil price changes on the country. The model considers a greater
variety of market goods as well as the effect of international trade balances on welfare. The model is
able to provide insight into the welfare impacts of oil price shocks, by showing not just how the
profits of the country change, but also how non-energy sectors of the economy are affected.
Nevertheless, as the author acknowledges in the paper, the assumptions about the form of consumer
utility are based on limited information.
The computable general equilibrium approach taken by De Santis can be extended by modeling
the Saudi economy in greater detail. Indeed, general equilibrium analysis may be required for
domestic policy analysis, given that the relatively large fraction of national GDP produced by the oil
and gas industry in Saudi Arabia makes the isolation of the market segment in a partial equilibrium
analysis an unreasonable assumption.
4.5 Resource Depletion
Oil reserves are finite and production will eventually hit a ‘peak.’ It remains unclear, however, when
such a peak will occur and if it will be based on a lack of available resources as posited in the classic
theory (Hubbert 1962) or if it will be due to lowered crude demand driven by fuel switching. The
threat of peak oil has loomed over the horizon, but with consistent resource discoveries and
technological breakthroughs, current thinking generally does not view crude resource scarcity as a
constraining force in the foreseeable future.
Aguilera et al (2009) considers the availability of resource at a given price. The paper estimates a
cumulative availability curve and finds that enough quantities of conventional and unconventional
petroleum are available at low enough prices to meet demand growing at 5% for a full 70 years. If
the growth rate is only 2% commercially viable resources are expected to last for 132 years.
While the study presents a first pass estimation, by aggregating resources the analysis is unable
to address local and distributional issues. Further, the study does not build the resource availability
curve into a model, leaving demand growth as an exogenous parameter, and does not calculate a
price path. A significant contribution would be made by combining a more detailed and thorough
resource assessment (including non-petroleum based substitutes) with a structural model including
supply and demand equilibrium. The model would provide greater insight into how resource scarcity
will change both consumer fuel choices and the distribution of revenues to different producers in
the market.
22
5. CONCLUSION
As this paper has shown, there is no lack of opportunities to extend the understanding of the oil
market through the use of computer models. Given the great amount of effort needed to tackle any
one of the issues, the next critical step is to define a research agenda and implementation roadmap.
After having established research priorities, a road map for modeling efforts will need to be defined.
In addition to the technical requirements of building models, a research institution is faced with the
additional challenges and opportunities of building a new modeling team, governance, and feedback
loops. With this organizational view in mind, it is strongly recommended that an institution begins
with less ambitious projects while building the necessary competencies.
Building an entirely new model from the ground up takes considerable time and effort.
Fortunately, existing frameworks exist that can be adapted to explore certain kinds of problems in a
generic energy system. Rather than building a model completely from scratch it may make sense to
integrate with such a system. With this approach, even a newly-established research center can
quickly contribute with extensions to existing well-known models while internalizing knowledge of
the advantages and limitations of the existing packages.
If a research center chooses to undertake the development of its own built-from-scratch model,
the project will become a software development as well as a research task. The code platform,
database system, and development process will need to be selected carefully. Presenting an outline of
software development best practices is outside the scope of this paper, but making the right choices
with several critical decisions is essential to building a model that scales, supports parallel
development by multiple individuals, and that can be easily evaluated and modified as the modeling
environment and research objectives change.
Models are built to meet specific objectives, and that the appropriate methods and research
opportunities vary by the intended model use. We have divided models into three broad groups
based on the time horizon of the analysis: medium- and long-term models.
In the medium-term modeling section, a few key open areas of research were discussed: the role
of financial activity, price volatility, information signals, high impact events, and hybrid models. For
the long-term models we covered changing demand side behavior, modern OPEC behavior,
strategic behavior by non-OPEC players, welfare analysis for oil-producing nations, and resource
depletion.
We see some of these research topics as being more relevant to research within oil-producing
countries: modeling volatility, understanding the role of financialization in the markets, and
reassessing modern OPEC behavior stand out as particularly relevant objectives. Meanwhile, topics
such as the strategic behavior of non-OPEC firms may be more important for regulatory agencies in
consuming nations to consider.
A final topic not listed is the general topic of “forecasting.” As a general rule, being able to
accurately forecast future oil prices is unlikely for many modeling efforts. No method has been
developed that outperforms on a systematic basis a simple ‘no-change’ rule estimate (Alquist, Kilian
et al. 2010). Given the high degree of interaction between future expectations and decision making
by critical market players (producers, consumers, and traders), obtaining greater precision forecasts
23
may be an impossible task. Nevertheless, common information and market perceptions have to
originate from the investment in data collection and modeling efforts by entities in the market.
A great deal of theoretical work has gone into econometric time series analysis and forecasting
of prices. A common problem confronting such models is the availability of data. For example, an
important hindrance in constructing reliable world oil models rests with the lack of consistent time
series on petroleum product prices that reflect existing subsidies in many developing countries. In
general high quality accessible data only exist for OECD nations. Rather than applying or
developing new theories towards the forecasting of oil, we believe the greatest gap in this area is the
richness of the data drawn on by the models. Having a vast international resource of global oil data
can make a very substantial contribution.
ACKNOWLEDGEMENTS
This paper is based on research project supported fully by King Abdullah Petroleum Studies and
Research Center (KAPSARC), Riyadh, Saudi Arabia. The authors wish to acknowledge helpful
comments and discussions with Stephen Brown, Albert Huang and John Weyant. All responsibility
for the contents of this paper belongs to the principal authors, and none of the views and
conclusions can be attributed to any of the above individuals or the King Abdullah Petroleum
Studies and Research Center (KAPSARC).
REFERENCES
Adelman, M. A. (1982). OPEC as a Cartel. OPEC Behavior and World Oil Prices. J. M. Griffin, D. J.
Teece , ed. London; Boston, Allen & Unwin: 37-57.
Aguilera, R., F., R. Eggert, G., et al. (2009). Depletion and the Future Availability of Petroleum
Resources. The Energy Journal, 30(1), 141-174.
Allaz, B. (1992). Oligopoly, uncertainty and strategic forward transactions. International Journal of
Industrial Organization, 10(2), 297-308.
Alquist, R., L. Kilian, et al. (2010). Forecasting the Price of Oil. Handbook of Economic Forecasting. G.
Elliott and A. Timmermann.
Brander, J. A. and T. R. Lewis. (1986). Oligopoly and Financial Structure: The Limited Liability
Effect. The American Economic Review, 76(5), 956-970.
Brandts, J., P. Pezanis-Christou, et al. (2008). Competition with forward contracts: a laboratory
analysis motivated by electricity market design*. The Economic Journal, 118(525), 192-214.
Celta, D. J. and C. A. Dahl. (2000). OPEC as a Social Welfare Maximizer. GREEN, Cahiers de
recherche.
Daly, G., J. M. Griffin, et al. (1982). Recent Oil Price Escalations: Implications for OPEC Stability.
OPEC Behavior and World Oil Prices. J. M. Griffin, D. J. Teece , ed. London; Boston, Allen &
Unwin: 231.
24
De Santis, R. A. (2003). Crude Oil Price Fluctuations and Saudi Arabia's Behaviour. Energy Economics,
25(2), 155-173.
Deaton, A. and G. Laroque. (1996). Competitive Storage and Commodity Price Dynamics. The
Journal of Political Economy, 104(5), 896-923.
Energy Information Administration. (2009). Energy Price Volatility and Forecast Uncertainty. Washington
DC, Energy Information Administratrion.
Energy Information Administration. (2010). Annual Energy Outlook Retrospective Review: Evaluation of
Projections in Past Editions (1982-2009).
Energy Information Administration. (2011). Annual Energy Outlook 2011. Washington, D.C., Energy
Information Administration.
Energy Modeling Forum. (2010). Inventory of Research and Analysis on Overall Oil Price Determinants.
Stanford, CA, Energy Modeling Forum.
Energy Modeling Forum. (2011a). Review and Assessment of Computer Modeling Approaches to Oil Market
and Price Analysis. Stanford, Energy Modeling Forum.
Energy Modeling Forum. (2011b). Identification of a Potential Future Research Agenda. Stanford, Energy
Modeling Forum.
Gately, D. and H. G. Huntington. (2002). The Asymmetric Effects of Changes in Price and Income
on Energy and Oil Demand. Energy Journal, 23(1), 19-55.
Gilbert, R. J. and R. G. Harris. (1984). Competition with Lumpy Investment. The RAND Journal of
Economics, 15(2), 197-212.
Griffin, J. M. (1985). OPEC Behavior: A Test of Alternative Hypotheses. American Economic Review,
75(5), 954.
Hamilton, J. D. (2003). What Is an Oil Shock? Journal of Econometrics, 113(2), 363-398.
Hamilton, J. D. (2009). Causes and Consequences of the Oil Shock of 2007-08. National Bureau of
Economic Research Working Paper Series No. 15002.
Hnyilicza, E. and R. S. Pindyck. (1976). Pricing Policies for a Two-Part Exhaustible Resource Cartel:
The Case of OPEC. European Economic Review, 8(2), 139-154.
Hubbert, M. K. (1962). Energy resources.
Jeffrey, D. (2011). Parnon, Arcadia Manipulated Oil Prices, U.S. Agency Says. Retrieved June 8,
2011, from http://www.businessweek.com/news/2011-05-24/parnon-arcadia-manipulated-
oil-prices-u-s-agency-says.html.
Kaufmann, R. K. and B. Ullman. (2009). Oil Prices, Speculation, and Fundamentals: Interpreting
Causal Relations among Spot and Futures Prices. Energy Economics, 31(4), 550-558.
Kavussanos, M. G. and I. D. Visvikis. (2004). Market interactions in returns and volatilities between
spot and forward shipping freight markets. Journal of Banking & Finance, 28(8), 2015-2049.
Kilian, L. and D. Murphy. (2010). The Role of Inventories and Speculative Trading in the Global Market for
Crude Oil, C.E.P.R. Discussion Papers.
Leahy, J. V. (1993). Investment in Competitive Equilibrium: The Optimality of Myopic Behavior.
The Quarterly Journal of Economics, 108(4), 1105-1133.
Lien, J. S. (2001). The strategic use of forward contracts: Applications in power markets.
25
Liski, M. and J.P. Montero. (2006). Forward trading and collusion in oligopoly. Journal of Economic
Theory, 131(1), 212-230.
National Commission on the BP Deepwater Horizon Oil Spill and Offshore Drilling. (2011). Deep
Water: The Guild Oil Disaster and the Future of Offshore Drilling. Washington, National
Commission on the BP Deepwater Horizon Oil Spill and Offshore Drilling.
Organization of the Petroleum Exporting Countries. (2008). OPEC Statute.
Pagano, P. P., Massimiliano. (2009). Risk-Adjusted Forecasts of Oil Prices. The B.E. Journal of
Macroeconomics, 9(1).
Routledge, B. R., D. J. Seppi, et al. (2000). Equilibrium Forward Curves for Commodities. The Journal
of Finance, 55(3), 1297-1338.
Sanders, D. R. I., Scott H. (2011). The Impact of Index Funds in Commodity Futures Markets: A Systems
Approach.
Schäfer, A. and H. D. Jacoby. (2006). Vehicle technology under CO2 constraint: a general
equilibrium analysis. Energy Policy, 34(9), 975-985.
Skinner, R. (2006). Strategies for Greater Energy Security and Resource Security: Background Notes.
G20 Energy and Resources Seminar. Banff, Alberta, Oxford Institute for Energy Studies.
Smith, J. L. (2005). Inscrutable OPEC? Behavioral Tests of the Cartel Hypothesis. Energy Journal,
26(1), 51-82.
Smith, J. L. (2009). World Oil: Market or Mayhem? The Journal of Economic Perspectives, 23, 145-164.
Spence, A. M. (1973). Job Market Signalling. Quarterly Journal of Economics, 87(3), 355-374.
Stulz, R. M. (1996). Rethinking Risk Management. Journal of Applied Corporate Finance, 9(3), 8-25.
Szakmary, A., et al. (2003). The Predictive Power of Implied Volatility: Evidence from 35 Futures
Markets. Journal of Banking and Finance, 27(11), 2151-2175.
United States Joint Forces Command. (2010). The Joint Operating Environment. Norfolk, United States
Joint Forces Command.
Wolak, F. A. (2000). An Empirical Analysis of the Impact of Hedge Contracts on Bidding Behavior
in a Competitive Electricity Market. International Economic Journal, 14(2), 40.
Wolak, F. A. (2003). Measuring Unilateral Market Power in Wholesale Electricity Markets: The
California Market, 1998-2000. The American Economic Review, 93(2), 425-430.
26
APPENDIX: Table 1 – Summary of Different Modeling Approaches (Example of Reference Listed in Parentheses)
Approach (ref)
Explanation
Advantages
Disadvantages
Structural
Econometric Model
(Kilian & Murphy
2010)
Historical data establishes behavioral
responses of oil consumers and producers
when oil prices and external (exogenous)
variables like economic growth and
resource costs change.
Understand how taxes and
other policies as well as
government efforts influence
oil prices.
Not necessarily better than reduced
form method for short-term
projections where the structural
supply-demand relationships do not
change dramatically over the period.
Reduced form
Econometric Model
(Kaufmann et al 2004)
Historical data establishes the relationship
be
tween oil prices and external
(exogenous) variables without directly
considering the oil quantity effects.
Short-term projections where
the structural supply-demand
relationships do not change
dramatically over the period.
Less reliable for understanding how
taxes and other policies influence oil
prices.
Structural Simulation
(Parametric) Model
(Gately 1983)
Assumed parameters establish separate
supply and demand relationships as a
function of prices and external
(exogenous) variables like resource costs
and economic growth.
May provide better long-run
explanations and projections
of oil prices, particularly when
there is structural change in
the relationship between
supply and demand factors.
Many parameters are determined
judgmentally rather than estimated
from actual historical data.
Optimization Model
(Kalymon 1975)
Use formal mathematical programming
techniques to select the best outcome for
one agent who can influence price. Other
agents do not optimize their decisions.
Their behaviors are usually represented by
simulation rules.
Improves understanding of
the incentives for certain types
of agents with market control.
May be better at telling us what an
agent should do rather than what it
will do; may produce inferior
projections.
27
Game-Theoretic
Model (Salant 1982)
Use formal mathematical programming
techniques to select the best outcome for
several agents from some set of available
alternatives.
Improves understanding
about how the choices by one
agent depend on the strategies
of other influential agents.
Multiple optimizing decisions make
the model much more difficult to
solve. Solutions reveal what agents
should do rather than what they will
do.
Process
(Programming) Model
(Deam et al 1973)
Estimates the least cost option for meeting
a given
energy demand level under
competitive conditions. (Demand level can
also change with price.)
Attractive for representing
technologies with very
different costs and
performances.
Solutions require lots of data and are
often very sensitive to small changes
in input cost assumptions.
Computable General
Equilibrium Model
(Massetti & Sferra
2010)
Represents interactions between the oil
market and other energy and economic
sectors. May include multiple agents and
will link the inputs and outputs of
different
sectors. Each sector will
substitute between inputs as prices change.
Provides information about
how energy production and
use influences the broader
economy.
Difficult to parameterize and solve.
Uses inflexible equations that may
not adequately incorporat
e key
energy sectors (e.g., electric
generation). Often fits the data
poorly.
Partial Equilibrium
Model#
Focuses exclusively on the oil markets and
excludes other energy markets and non-
energy sectors in the economy.
Allows model to focus
explicitly on important oil-
market features like the
organization of OPEC
producers.
Ignores important changes in other
energy and economic sectors that
could have important implications
for oil markets.
Top-down Model*
Provides links between aggregate oil
variables without providing much detail on
different technologies, agents and crude
types.
Relatively easy to develop
long-
term forecasts or
outlooks with a minimum of
data requirements.
Aggregation often misses important
issues for making decisions.
28
Bottom-up Model**
Provides considerable detail on different
technologies, agents and crude types
without incorporating many key links
between aggregate oil variables.
Allows greater depth on issues
critical for decisions.
Difficult to develop long-term
forecasts or outlooks. Often requires
detailed data that is missing.
Hybrid Model
Combines two or more of the above
approaches. Example: process-economic
models combine process with CGE or
other macroeconomic model.
Depends upon which model
type is used.
Depends upon which model type is
used. Implementation challenges of
combining distinct modeling
philosophies.
# Partial equilibrium models are frequently econometric or simulation structural models.
* Top-down approaches take many forms, frequently as structural econometric or CGE models.
** Bottom-up approaches are often process models without any linkages to an economic model.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This paper explores similarities and differences between the run-up of oil prices in 2007-08 and earlier oil price shocks, looking at what caused the price increase and what effects it had on the economy. Whereas historical oil price shocks were primarily caused by physical disruptions of supply, the price run-up of 2007-08 was caused by strong demand confronting stagnating world production. Although the causes were different, the consequences for the economy appear to have been very similar to those observed in earlier episodes, with significant effects on overall consumption spending and purchases of domestic automobiles in particular. In the absence of those declines, it is unlikely that we would have characterized the period 2007:Q4 to 2008:Q3 as one of economic recession for the U.S. The experience of 2007-08 should thus be added to the list of recessions to which oil prices appear to have made a material contribution.
Chapter
This book addresses the economics of exhaustible resources under monopoly and competition. The conventional wisdom - that oil prices, after hesitation in the early 80's will move upward faster than world inflation - is questioned along with the models supporting these predictions. The prospect that world oil prices may collapse in the near future is treated as a definite possibility. The findings are optimistic for consumers and should serve to reorient energy policy.
Article
Although OPEC is commonly viewed as a syndicate of producers engaged in cooperative efforts to restrict production and raise price, to date there is a surprising dearth of supporting statistical evidence to that effect. I show that standard statistical tests of OPEC behavior have very low power across a wide range of alternative hypotheses regarding market structure. Consequently, it is difficult, given the current availability and precision of data on demand and costs, to distinguish collusive from competitive behavior in the world oil market. I apply a new, production-based approach for examining alternative hypotheses and find strong evidence of cooperative behavior among OPEC members. My results also suggest that OPEC's formal quota mechanism, introduced in 1982 to replace a system based on posted prices, increased transactions costs within the organization. Statistical evidence is mixed on the question of whether Saudi Arabia and other core producers have played a special role within the cartel.
Article
This dissertation develops three theoretical models that analyze forward trading by firms with market power. The models are discussed in the context of recently restructured power markets, but the results can be applied more generally. The first model considers the profitability of large firms in markets with limited economies of scale and free entry. When large firms apply their market power, small firms benefit from the high prices without incurring the costs of restricted output. When entry is considered, and profit opportunity is determined by the cost of entry, this asymmetry creates the "curse of market power;" the long-run profits of large firms are reduced because of their market power. I suggest ways that large power producers can cope with the curse of market power, including the sale of long-term forward contracts. Past research has shown that forward contracts can demonstrate commitment to aggressive behavior to a competing duopolist. I add explicitly modeled entry to this literature, and make the potential entrants the audience of the forward sale. The existence of a forward market decreases equilibrium entry, increases the profits of large firms, and enhances economic efficiency. In the second model, a consumer representative, such as a state government or regulated distribution utility, bargains in the forward market on behalf of end-consumers who cannot organize together in the spot market. The ability to organize in forward markets allows consumers to encourage economic efficiency. When multiple producers are considered, I find that the ability to offer contracts also increases consumer surplus by decreasing the producers' profits. In some specifications of the model, consumers are able to capture the full gains from trade. The third model of this dissertation considers the ability of a large producer to take advantage of anonymity by randomly alternating between forward sales and forward purchases. The large producer uses its market power to always obtain favorable settlement on its forward transactions. Since other participants in the market cannot anticipate the large producer's eventual spot market behavior they cannot effectively arbitrage between markets. I find that forward transaction anonymity leads to spot price destabilization and cost inefficiency.
Article
We develop an equilibrium model of the term structure of forward prices for storable commodities. As a consequence of a nonnegativity constraint on inventory, the spot commodity has an embedded timing option that is absent in forward contracts. This option's value changes over time due to both endogenous inventory and exogenous transitory shocks to supply and demand. Our model makes predictions about volatilities of forward prices at different horizons and shows how conditional violations of the ‘Samuelson effect’ occur. We extend the model to incorporate a permanent second factor and calibrate the model to crude oil futures data.
This paper presents a theory of corporate risk management that attempts to go beyond the "variance-minimization" model that dominates most academic discussions of the subject. It argues that the primary goal of risk management is not to dampen swings in corporate cash flows or value, but rather to provide protection against the possibility of "costly lower-tail outcomes"-situations that would cause financial distress or make a company unable to carry out its investment strategy. (In the jargon of finance specialists, risk management can be viewed as the purchase of well-out-of-the-money put options designed to limit downside risk.) 1996 Morgan Stanley.
Article
This paper estimates the effects on energy and oil demand of changes in income and oil prices, for 96 of the world's largest countries, in per-capita terms. We examine three important issues: the asymmetric effects on demand of increases and decreases in oil prices; the asymmetric effects on demand of increases and decreases in income; and the different speeds of demand adjustment to changes in price and in income. Our main conclusions are the following: (1) OECD demand responds much more to increases in oil prices than to decreases; ignoring this asymmetric price response will bias downward the estimated response to income changes; (2) demand's response to income decreases in many Non-OECD countries is not necessarily symmetric to its response to income increases; ignoring this asymmetric income response will bias the estimated response to income changes; (3) the speed of demand adjustment is faster to changes in income than to changes in price; ignoring this difference will bias upward the estimated response to income changes. Using correctly specified equations for energy and oil demand, the longrun response in demand for income growth is about 1.0 for Non-OECD Oil Exporters, Income Growers and perhaps all Non-OECD countries, and about 0.55 For OECD countries. These estimates for developing countries are significantly higher than current estimates used by the US Department of Energy. Our estimates for the OECD countries are also higher than those estimated recently by Schmalensee-Stoker-Judson (1998) and Holtz-Eakin and Selden (1995), who ignore the (asymmetric) effects of prices on demand. Higher responses to income changes, of course, will increase projections of energy and oil demand, and of carbon dioxide emissions.