Study of Fundamental Analysis for Crude Oil Futures Prices

Xun Luo1

Apr 15, 2007


   Abstract:   This project studies the analysis of fundamental drivers for the prices of crude oil futures. The report first presents a summarization of the crude oil basics, including types, futures trading and modelling approaches. Then the NYMEX-traded crude oil futures data from 1983 to 2006 is used to undertake four types of fundamental analysis, in a basic-to-advanced order: supply-and-demand balances, geopolitical factors, inflation factors, and market expectations. Four quantitative models that use fundamental as parameters for crude oil price forecasting are subsequently evaluated. Following the individual sections for different factors the report conducts a case study of China Aviation Oil's trading loss in 2004, and the synergy of multiple fundamental drivers is investigated.

   Key words:   Crude Oil Futures Price, Fundamental Analysis, Case Study.

   Project Advisor:   Professor Stanley R. Pliska (Department of Finance, College of Business Administration, University of Illinois at Chicago).

   Revision History:  

1  Introduction

1.1  Motivation

Few will argue that the recent decades are interesting times for energy industry. This observation is true for both traders and investors. On the traders' side, for example, the deregulation of natural gas in the United States in the early 1990s is slowly but inexorably moving into Europe and Asia. Natural gas deregulation has strongly fostered competition, as well as called for needs for risk management. On the investors' side, a significant phenomenon is that for many of today's hedge funds, commodities were the hot tickets from 2000 to 2005, as their prices began to rocket, fuelled in large part by China's boom [9].

Although research in quantitative analysis has achieved undisputed success in financial markets, most models for pricing of energy derivatives (such as those in [10] and [4]) are still empirical and yet to be solidly verified. In the financial market, there are no arbitrage or term structure interest rate models which attempt to model the values of the interest rate securities as functions of one or a few variables, for example, spot interest rates, long-term interest rates, etc. The models must be consistent with the observed initial term structure and/or volatilities of the interest rates. A one-factor interest rate model uses short interest rates as a fundamental variable, so that the prices of interest rate derivatives depend only on short interest rates even though there are other factors that affect the prices, like tax effects, marketability, etc. Several one-factor interest rate models are widely used in the literature and practice, they include the Hull-White Model [7], Black-Derman-Toy model [5] and Black-Karasinski model [2], to name a few. Although there is no doubt that there are many similarities between the energy and financial markets, their difference are difficult to ignore that sometimes even experienced traders get deeply hurt.

A Connecticut hedge fund called Amaranth Advisors unluckily became such a victim and lost 5 billion in natural gas trade (it was thus listed as one of 2006's top 10 losing hedge fund, with regard to annual return). Below is the excerpt from the Wall Street Journal article [3] which attempted to gave reasons for Amaranth's loss:

"Unlike oil, gas can't readily be moved about the globe to fill local shortages or relieve local surpluses. Forecasts of freezing U.S. temperatures in winter or heat and hurricanes in summer can send prices jumping, while forecasts of mild weather can do the opposite. Last December, amid a cold snap, gas soared to a record 15.378 a million British thermal units on the New York Mercantile Exchange, or Nymex. This month, prices fell below 5 in the absence of major hurricanes and with forecasters talking about another warm winter. Yesterday, gas for October delivery settled at 4.942 a million BTUs on Nymex, off four cents."

The features of natural gas futures listed in the WSJ article, i.e., characteristics of supply patterns, susceptibility to seasonal weather conditions, as well as high volatility in price, are all non-significant factors in interest rate models but typical properties of energy derivatives. This example illustrates that how difference in fundamental drivers cause the uniqueness of energy derivative in their types, price evolution process, payoff calculation and risk management tools. To explore the effect of these fundamental drivers over real market data forms the main motivation for this project. Due to the market-to-market difference of commodities, limitation in report scope and data availability, this project focuses on the fundamental analysis approaches for the prices of a major kind of energy derivative: the crude oil futures.

1.2  Organization of This Report

This report is not strongly research-oriented, but more a literal reproduction of the processes of background information learning, get-hands-dirty practice and practical analytical problem solving on crude oil futures. To serve this purpose, it is organized as follows. Section 2 summarizes the background for crude oil futures, including crude oil types, futures trading at NYMEX, and a couple of modelling approaches. Section 3 illustrate the details about the data sources where the data sets for this project is obtained. In this section types of data providing agencies, data set scope and attributes are introduced. Two basic fundamental analysis - supply-and-demand balance and geopolitical factors are studied in Section 4. Section 5 drills further down to cover two more factors, namely inflation factor and market expectations. Following these, four quantitative oil price forecasting models which use fundamentals as parameters are evaluated in Section 6. In Section 7, the individual approaches are glued together towards a case study of China Aviation Oil's trading loss in 2004. Section 8 concludes the report. Appendix A lists a glossary of crude oil-related terminologies. At last, supplemental materials including data sets and literatures are available for download from the author's homepage [8].

2  Crude Oil Futures Basics

This section briefly describes the types of crude oil, futures trading at NYMEX and several price modelling approaches. Details of the modelling approaches mentioned are available in the corresponding references, while the definition of terminologies could be found in Appendix A.

2.1  Types of Crude Oil

There are literally hundreds of different crudes produced in the world, with there price differences not simply reflecting transportation costs to principle oil-consuming centers, but quality differences as well. In general, the two characteristics that are used to classify crude types are sulfur content and specific gravity (American Petroleum Institute (API) gravity). In a given refinery, higher API gravity crude oils inherently produce more "light" products, such as gasoline, jet fuel, and kerosene, while lower gravity oils will tend to produce more of the middle and "heavy" products, such as heating oil, diesel, and residual fuel. As a result, crudes that have a relatively high API gravity are referred to as "light" oils, and relatively low-gravity crudes are termed as "heavy". For example, West Text Intermediate crude, which has an average API gravity of 40, is a light oil, and Alaskan North Slope crude oil, which has an API gravity of 27, is a heavy crude.

The other general defining characteristic of crude oil, sulfur content, is considered a barometer of foreign materials contained in a given crude stream. When crude is produced, it contains a number of impurities, which are removed during the refining process, partly to prevent damage to the refining unit themselves. These materials include, among other things, heavy metals, waxes, dissolved gases and sulfur. Sulfur is the foreign material that garners the most attention because it is particularly difficult to remove during refining. Crude oils that have a sulfur content of 0.5 percent, or higher, by weight are referred to as "sour", while oils with a content under 0.5 percent by weight are termed "sweet". For example, Nigerian Bonny Light crude, which has a very low 0.1 percent sulfur content, is classified as a sweet oil, and Mexican Maya, which has a very high 3.3 percent of sulfur content, is categorized as a sour crude.

In many cases, light crude oils tend to be sweet. Conversely, heavy crudes tend to be sour. As a result, it is common within the oil industry to hear to term "sweet-sour spreads", referring to the price differential between heavy/sour oils and light/sweet crudes. In general, light/sweet oils almost always command a price premium to heavy/sour crudes, reflecting the inherently higher value of the refined products that are yielded from the refining process and the relatively lesser difficulty of refining these oils.

2.2  NYMEX Trading

Crude oil began futures trading on the NYMEX in 1983 and is the most heavily traded commodity (trading symbol: CL). The futures trade in units of 1,000 U.S. barrels (42,000 gallons). The trading months are 30 consecutive months plus long-dated futures initially listed 36, 48, 60, 72, and 84 months prior to delivery. Additionally, trading can be executed at an average differential to the previous day's settlement prices for periods of two to 30 consecutive months in a single transaction. These calendar strips are executed during open outcry trading hours. Crude Oil Futures are quoted in dollars and cents per barrel, with a minimum price fluctuation of 0.01 (1) per barrel (10 per contract). With regard to the maximum daily price fluctuation allowed, initial limits of 3.00 per barrel are in place in all but the first two months and rise to 6.00 per barrel if the previous day's settlement price in any back month is at the 3.00 limit. In the event of a 7.50 per barrel move in either of the first two contract months, limits on all months become 7.50 per barrel from the limit in place in the direction of the move following a one-hour trading halt.

Last trading day of crude oil futures is at the close of business on the third business day prior to the 25th calendar day of the month preceding the delivery month. If the 25th calendar day of the month is a non-business day, trading shall cease on the third business day prior to the last business day preceding the 25th calendar day. Delivery is F.O.B. from seller's facility, Cushing, Oklahoma, at any pipeline or storage facility with pipeline access to TEPPCO, Cushing storage, or Equilon Pipeline Co., by in-tank transfer, in-line transfer, book-out, or inter-facility transfer (pumpover). All deliveries are rateable over the course of the month and must be initiated on or after the first calendar day and completed by the last calendar day of the delivery month. An Alternate Delivery Procedure is available to buyers and sellers who have been matched by the Exchange subsequent to the termination of trading in the spot month contract. If buyer and seller agree to consummate delivery under terms different from those prescribed in the contract specifications, they may proceed on that basis after submitting a notice of their intention to the Exchange. The commercial buyer or seller may exchange a futures position for a physical position of equal quantity by submitting a notice to the Exchange, called Exchange of Futures for, or in Connection with, Physicals (EFP). EFPs may be used to either initiate or liquidate a futures position. The deliverable grades include specific domestic crudes with 0.42% sulfur by weight or less, not less than 37 API gravity nor more than 42 API gravity. The following domestic crude streams are deliverable: West Texas Intermediate, Low Sweet Mix, New Mexican Sweet, North Texas Sweet, Oklahoma Sweet and South Texas Sweet. There are also foreign streams deliverable, including U.K. Brent and Forties, and Norwegian Oseberg Blend, for which the seller shall receive a 30-per-barrel discount below the final settlement price; Nigerian Bonny Light and Colombian Cusiana are delivered at 15 premiums; and Nigerian Qua Iboe is delivered at a 5 premium. Position limits in a month is capped at 20,000 net futures, but not to exceed 1,000 in the last three days of trading in the spot month. Margins are required for open futures or short options positions.

2.3  Several Modelling Approaches

In [10], Pilipovic asserted that price mean-reversion was the most appropriate quantitative model for energy markets. She described energy derivatives as exhibiting "split personality", i.e., discrepancy between short-term and long-term behaviors. She attributed the reason to different sets of fundamental drivers for short-term and long-term markets that caused term structures in convenience yields. These statements could be put as:

Cy µ (St - Lt) + K
(2.1)
and
Cy ® K as t ®
(2.2)

Where Cy is the convenience yield, St is the spot price, Lt is the equilibrium price, and K is a constant. Equation 2.1 and 2.2 manifest that the convenience yield is affected by supply-and-demand imbalance in the short-term. In the long run, the imbalance effect goes to zero, and the prices approaches equilibrium levels.

Some other researchers, such as Eydeland and Wolyniec, authors of [4], disapprove the models that incorporate convenience yield. Eydeland and Wolyniec argued that in energy markets, convenience yield was not observable, and the amount of historical data was not large enough to deduce a model for its evolution in a stable and reliable manner. They asserted that using liquidly traded products(such as forward contracts) for calibrating convenience yield models can be misleading and may result in mispricing of common structures such as recall options. The authors instead recommended several other models, including simple forward pricing models, continuous forward curve models, and market models. The debate over convenience yield, however, is beyond the scope of this report as it is more about quantitative analysis rather than fundamental analysis.

Schwager in [11] suggested that the equilibrium price was mainly determined by OPEC expectations and fell into a price band over a certain time period. he further listed three major modelling steps when deriving a crude oil price forecast:
  1. Construct the oil-balance projection using supply-and-demand analysis.
  2. Evaluate the relevant geopolitical factors.
  3. Gauging the market perception of fundamentals using crude oil time spread.
All the authors of [10,11,4] agree that seasonality is a significant fundamental driver of energy derivative prices. Pilipovic proposes that every spot price model should be treated as the synergy of two separate models, one for the "underlying price" where there is no seasonality, another for the seasonality models. Meanwhile, the two models should have a near-zero correlation. Eydeland and Wolyniec tended to construct multi-factor stochastic models that incorporate the seasonality factor. Schwager instead suggested a deseasonalization step before the analysis of other fundamental drivers.

3  Sources of Data

The data sets used by this report are mainly from two sources, namely, the Energy Information Administration(EIA) and the American Petroleum Institute(API).

3.1  The Energy Information Administration

The EIA is a statistical agency of the U.S. Department of Energy and was created by Congress in 1977. Its mission is to provide policy-independent data, forecasts, and analysis to promote sound policy making, efficient markets, and public understanding regarding energy and its interaction with the economy and the environment. Energy products covered by EIA are:
Petroleum
including crude oil, gasoline, heating oil, diesel, propane, jet fuel, and other petroleum based products.
Natural Gas
including crude oil, gasoline, heating oil, diesel, propane, jet fuel, and other petroleum based products.
Electricity
including sales, revenue and prices, power plants, fuel use, stocks, generation, trade, and demand & emissions.
Coal
including reserves, including production, prices, employment and productivity, distribution, stocks and imports and exports.
Nuclear
including Uranium fuel, including nuclear reactors, generation and spent fuel.
Renewable & Alternative Fuels
including hydropower, solar, wind, geothermal, biomass and ethanol.
[h]
Date Contract 1 Contract 2 Contract 3 Contract 4
12/11/2006 61.22 62.36 63.29 64.00
12/12/2006 61.02 61.99 62.92 63.63
12/13/2006 61.37 62.17 63.05 63.72
12/14/2006 62.51 63.33 64.22 64.87
12/15/2006 63.43 64.09 64.89 65.49
Table 1: EIA Data of NYMEX Crude Oil Futures Prices in the Week of 12/11/2006
 
For crude oil fundamental analysis, EIA data is a must-read, if not more emphasized. These data sets are published on daily, weekly, monthly and annual basis. The daily data include the spot prices of crude oil and petroleum products in the U.S. and selected international areas, as well as futures price at NYMEX. Table 1 illustrates crude oil futures (WTI at NYMEX) of a sample week. Contract n in the table refers to crude oil futures of the nth nearest month (contract 1 specifies January 2007 WTI, contract 2 specifies February 2007 WTI, and so on). EIA archived the historical data of these daily prices. For WTI, the historical data could be back-traced to March 30, 1983. The weekly publications include: This Week in Petroleum, which is generally released on Wednesdays and contains analysis, data, and charts of the latest weekly petroleum supply and price data; Weekly Petroleum Status Report, which reports the petroleum supply situation in the context of historical information and selected prices; and several other reports, mainly about price information. The monthly publications include: Company Level Imports, which is about imports data at the company level collected from the EIA-814 monthly imports report; Petroleum Marketing Monthly, which is of monthly price and volume statistics on crude oil and petroleum products at a national, regional and state levels; Petroleum Supply Monthly, which details supply and disposition of crude oil and petroleum products on a national and regional level. The data series describe production, imports and exports, movements and inventories; Prime Supplier Report which measures primary petroleum product deliveries into the U.S. where they are locally marketed and consumed. At last, the annual publications include: U.S. Crude Oil/Natural Gas/Natural Gas Liquids Reserves Annual Report, Petroleum Supply Annual, Petroleum Marketing Annual and Refinery Capacity Report.

What is really valuable of the EIA data repository is that not only raw market data are provided, a researcher could also get access to a compilation of frequently updated analyses and forecasts. Figure 1 is a graph excerpted from the "China" section of Country Analysis Briefs, an analysis of the worlds major oil producers and consumers. There are a great quantity of analyses of other economic fundamentals about crude oil. Due to the page limit they are not listed here. Interested reader could check the EIA website (http://www.eia.doe.gov/).

[tp]
Figure
Figure 1: EIA Statistics of China's Oil Production and Consumption, 1986-2006

3.2  The American Petroleum Institute

In contrast to the EIA's government background, the API is the main U.S. trade association for the oil and natural gas industry, representing about 400 corporate members involved in all aspects of the industry. API is involved in lobbying and government liaison on behalf of the American oil and natural gas industries. It takes positions on access, exploration, taxes, trade regulation, environmental regulation, fuels, industry security and climate change. API conducts or sponsors research ranging from economic analysis to toxicological testing. And it collects, maintains and publishes statistics and data on all aspects of U.S. industry operations, including supply and demand for various products, imports and exports, drilling activities and costs, and well completions. This data provides timely indicators of industry trends. API's Weekly Statistical Bulletin is the most recognized publication, widely reported by the media.

Most of API's data need a paid subscription. However there are abundant free-licensed statistics on the API website (http://www.api.org) as well. These data are usually about company earning and spending, and end-consumer aspects of petroleum price, such as superimposed tax. Figure 2 shows an example graph taken from API website, which is about gasoline taxes rate at all the 50 U.S. states. A large portion of API data are gathered from sources of its industrial members. There are two main advantages of API's data and reports. The first is that they reflect to a large extent the downstream petroleum parties' views and concerns about the oil market (downstream parties are introduced in Section 4). Because of this, API data and reports are good reference for the U.S. domestic market players' expectations. The second is that as API is a liaison of industry to the general public, many basic concepts of oil and data are interpreted in ways easily understood by non-professionals, or novices in oil and natural gas research.

[tp]
Figure
Figure 2: API Statistics of U.S. Gasoline Taxes as of October 2006

4  Basic Fundamental Analysis

This section examines crude oil futures prices from the perspective of two fundamentals. One is supply-and-demand balances and the other is geopolitical factors. They are the basic fundamentals of crude oil futures. This fact is straightforward to understand: Supply-and-demand balance largely dictates virtually all commodity futures price. While the strategic role of oil for almost every country determines that oil price is bounded to be tied tightly with the geopolitical factors.

There is a good example for the effects of these two fundamentals on crude oil futures price: the 2003-2006 crude oil futures price curve. As illustrated by Figure 3, the price of WTI traded at NYMEX went from 25 to 38 per barrel in year 2003 to over 70 per barrel in the summer of 2006. If an analyst does not look at the fundamentals but believe that oil price has a long term trend, she is inclined to draw the conclusion that oil price has an increasing trend and explains it in the incorrect way. This false reasoning process could even be strengthened, if a chart comparison between oil futures price and the Dow Jones Industrial Index (Figure 4) is made. If Figure 3 and Figure 4 are put together, it will be found that trends of the two during 2003-2006 is quite similar. To make the illusion more realistic, a supporting metric: correlation between WTI future contract price and Dow Jones Index is as high as 0.7356!

However, any attempt to link the two and explain one with another will be deadly wrong. Dow Jones Index has a long-term trend of increasing, and its performance during 2003-2006 is the natural exhibit of this trend. On the contrary, crude oil futures price is more mean-reverting, globalized, and not necessarily associated with U.S. domestic economy. Actually, a long-term trend believer of crude oil prices, in the 2003-2006 data case, will be inclined to make the judgment that price will be keep increasing, or at least maintains at the same level as summer 2006, in winter 2006. This assertion is proved to be incorrect by the continuous decrease of WTI futures price after summer 2006. As of January 3, 2007, WTI future contract 1 price is 58.32 per barrel, a 24% drop compared with its peak of 77.03.

Using fundamental analysis, the author's opinion is that the price increase of crude oil during 2003-2006 could be attributed to: 1) limited surplus capacity, both upstream and downstream, 2) strong global demand growth, especially in Asia and the United States, and 3) geopolitical risks that have highlighted the need for more surplus capacity, both upstream and downstream. Below the analysis for each one is given more closely.
[tp]
Figure
Figure 3: Increasing WTI Future Contract 1 Price at NYMEX, 2003-2006
[tp]
Figure
Figure 4: Dow Jones Industrial Average Index, 2003-2006

4.1  Supply-and-Demand Balances

For the global oil industry, oil trade represents the close connection between two main centers of activity: upstream exploration and production, as well as downstream refining and marketing. The interactions between the upstream and the downstream largely determine crude oil supply-and-demand balancing dynamics. Mechanisms of such interactions are as following: Upstream parties are the major sellers of crude oil, and their productions are valued by downstream demand; While downstream parties are the major buyers of crude oil, and the cost of their feedstock is determined by the upstream supply. Operational decisions about combining output from various fields to create a specific crude oil export stream with certain characteristics are constantly tested in the market against the requirements of refiners for specific feedstock to meet final demand for a changing combination of products. The downstream marketing prices of the petroleum products, such as heating oil, gasoline, propane, aviation oil and kerosene are also determinants of crude oil price. Due to the extensive vertical integration of the oil industry until the early 1970s, these decisions used to be largely kept under the umbrella of major oil companies.

There have been several profound changes in the upstream-downstream structure since 1970s. Increased crude price volatility since the early 1970s in combination with other price-affecting factors, OPEC output quotas for example, signalled oil-importing developing countries such as South Korea, India, and Brazil to invest in refining capacity to mitigate both refined product volume and price risks. These same trends also created an incentive for governments in oil-exporting countries, notably Iran, Kuwait and Saudi Arabia, to build refineries in order to capture the value added in turning crude oil into refined products. Other global trends of oil companies include privatisation and large mergers among majors. These trends are finally challenging the long established dominance of big national oil companies in the top tiers of the international oil industry. While the largest state-owned companies are still playing a critically important role, the private sector companies are now becoming more important rivals. As of today, global upstream and downstream composition has been quite different from what it was in the 1970s. Figure 5 lists the top 10 crude oil exporters as of 2002, as well as the top 10 oil companies in 2002-2003.
[tp]
Figure
Figure 5: Top 10 Crude Oil Exporters and World's Largest Oil Companies, 2002-2003
 
Clearly, the extent to which upstream's or downstream's capacities are utilized during a certain time period greatly determines the crude oil price level and volatility of that period. In the case that the upstream has limited surplus in capacity, such as running tight on daily productions or lacking of new explorations when the supply-demand in market is barely balanced, a small portion of decrease in crude oil production would cause a significant price hike. For this reason, the market players will prefer to pay crude oil future contracts with higher premium. On the other hand, when the downstream has limited surplus in capability, such as fully-operating refineries, oil transport ports and storages, an instability factor in these facilities will trigger significant increases of the petroleum products price. Such increases will subsequently affect crude oil price in a indirect manner, making bulls in the futures market.

From 2003 to 2006, surplus global oil production capacity, which was as high as 5.6 million barrels per day in 2002, plummeted to 1.8 million barrels per day in 2003, and has been around 1 million barrels per day during most of 2004 to 2006. As demand has increased rapidly during the same period, the world has dipped into the surplus capacity that had been built up earlier. While some productive capacity has been brought online, it has been insufficient relative to demand growth. As a result, surplus capacity is extremely limited, dramatically reducing the ability to respond to any sudden surges in demand or disruptions in supply. The situation is similar downstream, where global refinery utilization has increased from an annual average of 85 percent in 2002 to 90 percent in 2005. This increase in refinery utilization has also reduced the system's flexibility to respond to any disruption in refinery production, either from hurricanes or other events. Increases in refinery utilization rates may also make crude oil markets more responsive to seasonal patterns for refined products. All these factors composed the first cluster of pulling-up forces for crude oil price during the 2003-2006 period.

This report also argues that strong growth in the world economy, and particularly in China and the United States, has fueled the need for more oil, thus putting upward pressure on prices. That is, strong global oil demands are the other cluster of factors causing oil prices to rise in recent years. As shown in Figure 6, Asia Pacific and the United States are world's largest oil consumption regions, and the main oil consumer in Asia Pacific are Japan and China. As of 2006, the U.S. ranks first in daily oil consumption, Japan ranks the second, and China the third. All these three counties' economy are in good shape from 2003-2006, with China being the particular. As a result, after averaging annual growth of just under 1 million barrels per day between 1991 and 2002 (under 0.9 million barrels per day for 2000-2002), world oil demand grew by 1.5 million barrels per day in 2003, 2.6 million barrels per day in 2004, and at least 1.1 million barrels per day in 2005. This greater-than-historical growth came even as oil prices more than doubled.
[tp]
Figure
Figure 6: Crude Oil Consumption Breakdown by Region, 2004
 
This report pays special attention is paid to the statistics about China, which is also the author's home country. China is the world's most populous country and has a rapidly growing economy. Its real gross domestic product (GDP) is estimated by EIA to have grown at 9.9 percent in 2005, down slightly from the 2004 rate of 10.1 percent. Economic forecasts remain strong for China, with real GDP expected by EIA to increase 9.9 percent in 2006. Inflows of foreign direct investment (FDI) into China totaled 86.1 billion in 2005, a new record and roughly double the level of 2001. China’s merchandise trade surplus soared to 102 billion in 2005, its largest surplus ever and roughly three times larger than the 2004 figure.

Understanding the strong economic growth of China, it is consequently easy to see why China's demand for energy is surging rapidly. EIA forecasts that China’s oil consumption will increase by almost half a million barrels per day in 2006, or 38 percent of the total growth in world oil demand. (See both Figure 6 and Figure 1). With China's entry into the World Trade Organization (WTO) in November 2001, the Chinese government made a number of specific commitments to trade and investment liberalization which, if fully implemented, will substantially open the Chinese economy to foreign firms. In the energy sector, this will mean the lifting or sharp reduction of tariffs associated with imports of some classes of capital goods, and the eventual opening to foreign competition of some areas such as retail sales of petroleum products. The improvement in domestic markets openess will be surely be positive ingredients to China's oil demand. Readers who have further interest on this topic could check the EIA website, as well the cover-story article on the August 25, 2005 issue of The Economist [1].

4.2  Geopolitical Factors

Just as the lack of surplus capacity is related to the growth in global demand, the impact on prices due to geopolitical risks is related to the lack of surplus capacity. If surplus capacity was sufficient to make up for any reasonable likelihood of a loss in supply, then the risks would not have as great an impact on price. However, because there is very limited surplus capacity, concerns about potential or existing supply problems in Nigeria, Iran, Iraq, Venezuela, and elsewhere, have exacerbated price increases related to the supply-and-demand factor above. Or put another way, these risks to supply would not be putting as much upward pressure on prices if fundamentals were not tight to begin with.

The risks brought by geopolitical factors include instabilities of a nation's government and/or domestic economy, such nations do not necessarily to be an major crude oil exporter. For example, Singapore has strategic geographical location on the strait of Malacca, a main ocean waterway where 11.7 million barrels of crude oil passing by daily (2004 data). As a result, failure to crack pirate activities in the strait of Malacca by Singapore and other neighboring nations' law-enforcement departments will sometimes bring a up curve in the crude oil futures price. In another example for Venezuela, a disastrous two-month national oil strike, from December 2002 to February 2003, temporarily halted the whole nation's economic activity. Because Venezuela continues to be an important source of crude oil for the U.S. market. Both the instant effect of oil output volume collapse and aftermath effects as inflation and unemployment became fundamental drivers for a price hike of WTI future contact during the period, November 2002 to March 2003. That price hike is clearly reflected by Figure 7. Note also in Figure 7 that although Venezuela's national strike ended in February 2003, the following U.S. invasion to Iraq, started on March 20, 2003, kept the crude oil futures price at its local peak for another week.

During the period of 2003-2006, the forces exercised by geopolitical factors to global crude oil market are clearly pull-up ones. Besides the situation of Venezuela as described above, the U.S. invasion to Iraq successfully toppled the regime of Saddam Hussein in a short time frame, however the following insurgent activities in the war-torn country have been put the Middle East in long-time instability. During the same period, the conflicts between U.S. and Iran, the world's fourth largest oil exporter in 2004, have never really come to a rest. In year 2003-2004, some analysts even believed that a U.S. invasion to Iran had been planned and military actions of none-regular attacks, such as missile assaults might be taken. In Nigeria, it is not uncommon for oil producing and transporting facilities to be vandalized and result in sharp drop in oil output. In May 2005, Gasoline gushing from a ruptured pipeline exploded as villagers scavenged for fuel in Nigeria, killed up to 200 and caused a 50% drop in the nation's oil output for a week.
[tp]
Figure
Figure 7: Venezuela's Effects on WTI Future Contract 1 Price, 2002-2003
 
An interesting question is: how to determine the magnitude of a individual geopolitical factor's influence to global crude oil price? This question is not NP-hard, but a very difficult one if precise quantitative results are to be derived. In a gross level, a practical approach could be using short-term events of a specific geopolitical factor to gauge the corresponding factor's magnitude of influencing power. An example is shown below, about the world's biggest oil exporter: Saudi Arabia. On February 24, 2006, Islamic extremists took a bold daytime attack on the world's largest oil-processing facility, called Abqaiq close to Saudi Arabia's main export terminals on the Gulf coast. Although the attack was defeated at the security roadlock and did not affect the oil-processing facility's daily production at all, future contract 1 of WTI (to be delivered in March 2006) had a 3.4% price increase the next day, as illustrated by Figure 8. This is a terrific example of the magnitude of Saudi oil's influencing power to the global market. One hedge fund manager anticipated that the fall of the House of Saud would generate a 262 per barrel price in the year of 2006 [12].

[tp]
Figure
Figure 8: Saudi Arabia's Effect on WTI Future Contract 1 Price, Feb. 2006- Mar. 2006

5  Further Fundamental Analysis

In this section the use of two more fundamentals are studied: inflation factors and market expectations. These ones are not as major as the two basic fundamentals introduced in Section 4 but nevertheless inneglectable when analyzing crude oil future prices. Inflation factors determine view of the real prices of the crude oil, despite its nominal ones. While without knowledge of market expectations, even if all other fundamentals are reckoned correctly, one may still have great difficulty in fitting his/her explanation with the price behaviors.

5.1  Inflation Factors

A key difference between financial futures and commodity futures is that finance futures prices usually incorporate variation of its pricing currency, while for commodity futures this is not the case. Thus, sometimes changes of the nominal prices in commodity futures could be actually caused by the changes in the pricing currency. As the crude oil futures traded at NYMEX are measured by U.S. dollar, the factor of U.S. dollar inflation should be carefully considered when investigate historical data that span a wide variety of years. There are several indexes that can be used to measure inflation, such as gross domestic product (GDP) deflator, producer price index(PPI), or the consumer price index (CPI). Short descriptions of these indexes are given as follows.

The PPI measures average changes in prices received by U.S. domestic producers for their output. The PPI was known as the Wholesale Price Index, or WPI, up to 1978. The PPI is one of the oldest continuous systems of statistical data published by the Bureau of Labor Statistics, as well as one of the oldest economic time series compiled by the Federal Government. The origins of the index can be found in an 1891 U.S. Senate resolution authorizing the Senate Committee on Finance to investigate the effects of the tariff laws "upon the imports and exports, the growth, development, production, and prices of agricultural and manufactured articles at home and abroad."
[tp]
Figure
Figure 9: Historical Data of the U.S. Consumer Price Index (CPI-U), 1983-2006
 
The CPI or retail price index (RPI) is a statistical time-series measure of a weighted average of prices of a specified set of goods and services purchased by consumers. It is a price index that tracks the prices of a specified basket of consumer goods and services, providing a measure of inflation. The CPI is a fixed quantity price index and considered a cost-of-living index. In the U.S., CPI figures are prepared monthly by the Bureau of Labor Statistics of the United States Department of Labor. The CPI-U includes expenditures by all urban consumers and and is based upon a 1982 Base of 100. The CPI-W includes expenditures by consumer units with clerical workers, sales workers, craft workers, operative, service workers, or laborers. A CPI-U value of 158 indicates 58% inflation since 1982. The commonly quoted inflation rate of say 3% is actually the change in the CPI-U from a year earlier. The historical data of CPI-U from 1983-2006 (in align with the available period of WTI futures price data in this report) is plotted in Figure 9. The GDP deflator measures the change in prices in total GDP and for each of the GDP component. Though the CPI is a more closely watched inflation indicator, the GDP deflator is another key inflation measure. Unlike CPI, it has the advantage of not being a fixed basket of goods and services, so that changes in consumption patterns or the introduction of new goods and services will be reflected in the deflator.

This report makes use of the U.S. CPI-U data from 1983 to 2006 to study the effect of inflation factor on crude oil futures prices. The CPI-U data is incorporated for inflation adjustment in two approaches. In the first approach, the CPI-U value in 2006 is normalized to 1, and the normalization coefficient obtained is used to normalize the historical CPI-U values. Then the annual average prices of WTI contract 1 futures is multiplied by the normalized CPI-U values in the corresponding years. In this way the inflation-adjusted crude oil futures prices measured in the value of 2006 dollars are obtained. Figure 10 shows a graph of crude oil nominal and 2006-dollar-equivalent prices. In a similar way, the prices are measured in 1983 dollars in Figure 11.

The inflation-adjusted price curves tell more information of crude oil futures price than the nominal price curves, in terms of closeness to true market trends. Although the nominal price curves indicate that the WTI contract 1 price is at its historical high in 2006, it can be seen that this price level has actually been reached as dated back to 1983, when the inflation-adjusted WTI contract 1 price was about 60 a barrel in 2006 dollars (Figure 10). Meanwhile, it could be observed that the WTI contract 1 price is mean-reverting in this 20-year period, measured by 1983 dollars, it is mostly in a band between the 10 and 30 lines (Figure 11).
[tp]
Figure
Figure 10: WTI Contract 1 Annual Average, Nominal vs. Inflation-Adjusted (2006 Dollars), 1983-2006
[tp]
Figure
Figure 11: WTI Contract 1 Annual Average, Nominal vs. Inflation-Adjusted (1983 Dollars), 1983-2006
 
It can also be learned from Figure 10 and Figure 11 that inflation factors are long-term fundamentals. The difference between nominal price and inflation-adjusted price between 2003-2006, which is the time frame the basic fundamentals are studied in Section 4, is not huge and does not affect the conclusions much. However, if the inflation factor is ignored in a 10-year time frame, the resulted trend conclusions or price-forecasting models are bounded to be flawed. Due to limitations of data sources, this report is not able to associate WTI futures prices with inflation factors earlier than 1983. But a third-party plot of crude oil spot prices (Figure 12) could be a good example for this point. With this graph understood, it will be not totally absurd for some analysts' opinions in 2006 that crude oil could rocket as high as 100 a barrel.
[tp]
Figure
Figure 12: Monthly Average of Crude Oil Spot Prices, Nominal vs. Inflation-Adjusted (2006 Dollars), 1946-2006

5.2  Market Expectations

Market expectations could have radical influences on the price. Intuitively one of its mechanisms could be described as follows. Before the release of key actual statistics in each period, each player takes action to maximize his expected profit according to her expectation of price. When players' expectations are highly correlated, the collective action of these players can practically change the actual determinants of the price. Therefore neglecting the market expectation could lead to non-ignorable mistakes in a certain price prediction model. It seems reasonable that the market expectations during a given time period may be more relevant to determining prices during that period than are the actual statistics that only become know much later, and it may give us a more accurate model using past estimates rather than actual statistics as the price-explanatory variables.

There are quite a few literatures which introduce practical approaches about making use of market expectations. For example, three ways are described in [11] to incorporate expectation in a model:
 
It is important to realize that expectations for a coming season can often have a more strong price impact than do prevailing fundamentals. This is particularly true during the later half of a season when the fundamentals for the given season are well defined due to players' action results and not subject to significant variation. Players foresee the trend of the market in the next period clearly and take effective action to protect her next period's profit. In another word, under some circumstances expectation for the following period plays the dominant role in price-determination.

Some traders actually follow this concept in practical trading decision-making. An 2005 article of the Federal reserve bank of San FRANCISCO predicts the crude oil price by using "futures-spot spread" [13], which uses the spread between the current futures prices and the spot price to predict movements in the future price of WTI crude oil at NYMEX. The central idea of the article is that oil traders are knowledgeable about the industry, as a result, they are trying best to make sound investments, making the price-driving force of expectations a factor as strong as the spot price. The article compares the "futures-spot spread" model with three other models: random walk model which predicts that spot oil prices will stay at their current levels, cost-of-carry model which predicts that the future oil price will be the current spot price adjusted for the interest rate, and "futures" model which predicts an oil price level in the future identical to the current futures price level.
[tp]
Figure
Figure 13: Back-Test Accuracy of Expectation + Current Statistics Model on WTI Spot Price(1985-2005), with 3- and 12-month Estimation Horizons
 
In [13], the authors make estimation of WTI future spot price using the "futures-spot spread" model, over the WTI prices between 1985 and 2005, and calculate model forecasts for horizons that vary from one to twenty four months, they then compare the forecasts with the actual oil prices over these months. The results, shown in Figure 13 imply that this model outperforms than all the other three models: random-walk, cost-of-carry, and "futures". This study shows that expectation , as a fundamental for oil futures prices contain important information about price movements, especially for the near term.

Another example of market expectations is the WTI price variation during Spring 2003, when the U.S. invasion to Iraq took place. From Figure 14, it could be clearly observed that there is a significant difference between the future price and the spot price from March 2003 to May 2003. This two-month-lasting difference is a good illustration of the influence of market expectation on crude oil futures price. Before the war, many international observers announced that the war would be lengthy, and Iraq would become another swamp that the U.S. troops will have great difficulties to get out (ironically, this has quite become a reality as of 2006, but discussion of the topic is not within the scope of a master project report). People, at least those who made a living on the Street, widely believed so as well. The market thus reflect an expectation that since the Iraq War would seriously influence the relationships between U.S. and other important oil producing countries, the longer the war would go on the more seriously the crude oil supply would dampen. As a consequence of this expectation, the crude oil future price kept increasing, shown by the future prices' trend from January to early April, 2003. Till April 2003, the invasion turned out to be swift, with the collapse of the Iraq government and the military of Iraq in about three weeks. The oil infrastructure of Iraq was rapidly secured with limited damage in that time. Securing the oil infrastructure was considered of great importance to funding the rebuilding of Iraq after the invasion ended. In April, U.S. forces moved into Baghdad and soon after President George W. Bush announced "End of major combat operations". Initial plans which were for armored units to surround the city and gradually move in soon became unnecessary. All these news from Pentagon soon pushed down the crude oil futures price. Shown on Figure 13 that in April 2003, the crude oil spot price achieves its minimal point. The effect of the specific market expectation diminished in another month, and the large gap between the crude oil's spot price and future price disappeared.
[tp]
Figure
Figure 14: Effect of Market Expectation on WTI Prices, Jan.-June 2003

6  Evaluation of Quantitative Models That Use Fundamentals

This section presents and evaluates quantitative price-forecasting models that make use of fundamentals. The fundamental to be discussed is market expectations, as introduced in Section 5.2. When forecasting WTI price at NYMEX in a 3-month horizon, the market expectations fundamental is represented by the 3-month futures price, i.e. the price of WTI future contract 3 at NYMEX. Time series analysis techniques are used to estimate the model parameters. I pick the daily WTI price and contract 3 futures price at NYMEX, starting at January 2, 1986 (the first trading day of 1986) and ending in December 29, 2006 (the last trading day of 2006). One dataset to establish the models groups prices data within a trading month are averaged them. The other dataset to establish the models uses the price data of the last trading day as the representative data of the corresponding monthly price. Each year has 12 trading months, so each time series contain 12(month)*31(year)=252 data points.

There are four models to be addressed. Model 1 (Random Walk Model) and 2 (Interest Rate Model) are for benchmark only and use just WTI spot price and/or interest rates. Model 3 (Futures Model) uses WTI futures contract 3 price as the sole determinant of future WTI spot price. Model 4 (Future-Spot Spread Model) uses both the WTI spot price and futures contract 3 price to construct a spread in estimating the future WTI spot price. Below is a more detailed description of these models:
  1. Random walk model predicts that spot oil prices will stay at their current levels. Mathematically, the model could be presented as:

    Spot Pricet = a*Spot Pricet-3 + b
    (6.1)
  2. Interest rate model assumes that the opportunity cost of storing oil is the forgone interest rate. Mathematically, the model could be presented as:

    Spot Pricet = a*Spot Pricet-3b * eg*(1+Interest Rate)t-3
    (6.2)
    or:
    logSpot Pricet = loga + b*logSpot Pricet-3 + g* log(1+Interest Rate)t-3
    (6.3)
  3. Futures model predicts that an oil price level in the future is identical to the current futures price level. i.e., market expectations fundamental is the sole determinant for future oil spot price. Mathematically, the model could be presented as:

    Spot Pricet = a*Futures Pricet-3 + b
    (6.4)
  4. Future-Spot spread model assumes that future oil spot price is determined by a combination of current spot price and current futures price, and a properly constructed spread between these two current prices can predict the future oil spot price. Mathematically, the model could be presented as:

    Spot Pricet = a*Spot Pricet-3 + b*Futures Pricet-3 + g
    (6.5)
Performances of the different models are evaluated using two criteria. First, I estimate the model over the full sample (January 1986 to December 2006), calculate its forecasts for a 3-month horizon and then compare the forecasts with the actual oil prices over these weeks. The model with the smallest average prediction errors is said to have the best "in-sample" fit, since its parameters are estimated over the full sample.

Second, I conduct a more realistic "out-of-sample" forecasting experiment, where I still use the parameters estimated using the data up to December, 2006, but make forecasts for oil prices during January 2004 to December 2006, i.e. the price data of the most recent 36 months . The model with the smallest "out-of-sample" forecast errors has the most forecasting power, because, in practice, people are only able to undertake "out-of-sample" forecasts. However, both "in-sample fitness" and "out-of-sample fitness" criteria are used in evaluation of models, in order to obtain more robust conclusions.

6.1  In-Sample Forecasting Results

Table 2 depicts the overall characteristics of interest rate, monthly average spot price and monthly average futures price data. While Table 3 presents the linear regressed parameters of the four models. It could be observed that with respect to RMSE (Rooted Mean Square Error), Futures Model seems to have the best "in-sample" prediction. The rest three models have similar RMSE values. With regard to explaining power, all four models have similar R2values ( ³ 0.85), with Futures Model having the highest (0.911).
Table 2: Summary Statistics, Using Monthly Average Data
Variable Mean Std. Dev.
interest 4.959 2.175
spotp 25.919 13.491
futurep 25.705 13.883
N 252
Table 3: Fitted Parameters of Four Oil Spot Price Prediction Models, Using Monthly Average Data
Random Walk Hotelling Future Price Future & Spot Prices
Y=Spot PricetY=Log(spot pricet)Y=Spot PricetY=Spot Pricet
Spot Pricet-30.987^***-0.151
(0.0204) (0.274)
[1em] Log(Spot pricet-3)0.928^***
(0.0259)
[1em] Log(interestt-3+1)-1.245^*
(0.501)
[1em] Future Pricet-30.967^***1.115^***
(0.0193) (0.267)
[1em] Constant 0.851 0.301^**1.593^** 1.729^**
(0.584) (0.0938) (0.550) (0.603)
N 249 249 249 249
R2 0.905 0.867 0.911 0.911
adj. R20.904 0.866 0.911 0.910
RMSE 4.184 0.152^a 4.046 4.052
Standard errors in parentheses
^a Based on logarithm values
^* p < .05, ^** p < .01, ^*** p < .001
Table 4 depicts the overall characteristics of interest rate, last-day-of-month spot price and last-day-of-month futures price data. While Table 5 presents the linear regressed parameters of the four models. It could be observed that with respect to RMSE (Rooted Mean Square Error), Futures-Spot Spread Model seems to have the best "in-sample" prediction. The rest three models have similar RMSE values. With regard to explaining power, all four models have similar R2 values ( ³ 0.85), with Future-Spot Spread Model having the highest (0.904).
Table 4: Summary Statistics, Using Last-Day-of-Month Data
Variable Mean Std. Dev.
interest 4.959 2.175
spotp 25.931 13.737
futurep 25.752 14.115
N 252
Table 5: Comparision of Four Oil Spot Price Prediction Models, Using Last-Day-of-Month Data
Random Walk Hotelling Future Price Future & Spot Price
Y=Spot PricetY=Log(spot pricet)Y=Spot PricetY=Spot Pricet
Spot Pricet-30.982^***-0.351
(0.0212) (0.310)
[1em] Log(Spot pricet-3)0.919^***
(0.0264)
[1em] Log(interestt-3+1)-1.280^*
(0.516)
[1em] Future Pricet-30.964^***1.306^***
(0.0200) (0.304)
[1em] Constant 1.027 0.334^***1.698^** 1.988^**
(0.608) (0.096) (0.574) (0.629)
N 249 249 249 249
R2 0.897 0.861 0.904 0.904
adj. R20.897 0.860 0.903 0.904
RMSE 4.419 0.156^a 4.273 4.270
Standard errors in parentheses
^a Based on logarithm values
^* p < .05, ^** p < .01, ^*** p < .001

6.2  Out-of-Sample Forecasting Results

It is worth noting that my "Out-of-Sample" test is a partial one, i.e., training data and test data are not completely segregated.

6.2.1  Random Walk Model

Figure 15 illustrates how model 1 predicts the monthly averaged spot price. The average of absolute prediction error is 5.174699167. The standard error of prediction values is 6.141294508.
Figure 16 illustrates how model 1 predicts the spot price of the last day of each month. The average of absolute prediction error is 5.49767375. The standard error of prediction values is 6.514593256.
[tp]
Figure
Figure 15: Prediction of Monthly Average Price, by Model 1
[tp]
Figure
Figure 16: Prediction of Last-Day-of-Month Average Price, by Model 1

6.2.2  Interest Rate Model

Figure 17 illustrates how model 2 predicts the monthly averaged spot price. The average of absolute prediction error is 5.324511702. The standard error of prediction values is 5.951021091.
Figure 18 illustrates how model 2 predicts the spot price of the last day of each month. The average of absolute prediction error is 5.789853643. The standard error of prediction values is 6.694696508.
[tp]
Figure
Figure 17: Prediction of Monthly Average Price, by Model 2
[tp]
Figure
Figure 18: Prediction of Last-Day-of-Month Average Price, by Model 2

6.2.3  Futures Model

Figure 19 illustrates how model 3 predicts the monthly averaged spot price. The average of absolute prediction error is 4.946480417. The standard error of prediction values is 6.022400509.
Figure 20 illustrates how model 3 predicts the spot price of the last day of each month. The average of absolute prediction error is 5.254040833. The standard error of prediction values is 6.119488822.
[tp]
Figure
Figure 19: Prediction of Monthly Average Price, by Model 3
[tp]
Figure
Figure 20: Prediction of Last-Day-of-Month Average Price, by Model 3

6.2.4  Future-Spot Spread Model

Figure 21 illustrates how model 4 predicts the monthly averaged spot price. The average of absolute prediction error is 4.905905. The standard error of prediction values is 6.019941115.
Figure 22 illustrates how model 4 predicts the spot price of the last day of each month. The average of absolute prediction error is 5.230458958. The standard error of prediction values is 6.026120626.
[tp]
Figure
Figure 21: Prediction of Monthly Average Price, by Model 4
[tp]
Figure
Figure 22: Prediction of Last-Day-of-Month Average Price, by Model 1

6.3  Evaluation Conclusions

My evaluation of the four models shows slightly different result depending on which type of price is chosen as the monthly price representative: average or last-day-of-month. When average price is chosen as the representative, the Futures Model is the best performer in out-of-sample test; while when last-day-of-month price is chose the Futures-Spot Spread Model seems to perform better. On the contrary, Interest Rate Model and Random Walk Model seems to be weaker explanations for the test data, which spans from 2003 to 2006.
These evaluation result coincides with the result in [13], albeit the training dataset and testing dataset of mine and theirs are not exactly identical. The results strongly suggest that market expectations is fundamental that has significant effect on crude oil spot prices. There could also be further discussions of the conclusion because that the models I constructed is by no means comprehensive and sophiscated, and thus might not capture the effects of other fundamentals.

7  Case Study: Chinese Aviation Oil Trading Loss

In this section a real trading case of oil derivatives is presented and studies. One of the purposes of the case study is to examine synergies of the individual fundamental analysis approaches introduced earlier. Another purpose is to put the conceptual reasonings on paper into the practical environment, to get a sense of how people make use of fundamental analysis when solve real problems that involve serious capital investments.

7.1  The Story

The full name of China Aviation Oil (CAO) is China Aviation Oil (Singapore) Corporation Ltd. CAO is a Singapore-listed public company whose core business is to ensure the procurement of jet fuel (kerosene) for the airports in China. Due to the high-volatility in jet fuel prices, a large portion of its tradings involve energy derivatives. Initially, CAO traded only swaps and futures to help optimising this procurement duty. Later in 2002, on behalf of client airline companies, it started back-to-back option trading. Since 2003, the company began its speculative trading in fuel options by writing call options and holding put options. The speculations of CAO ended up bringing the company with big trading losses in October 2004, when the jet fuel price kept an one-year increase and the company failed to satisfy exchange margin calls on its options positions. The losses totaled 55 million and almost caused the company to go out of business.

PricewaterhouseCoopers (PWC) issued a postmortem report of the China Aviation Oil trading loss. From PwC' report it is possible to summarize the CAO options trading records as follows:
 
The above is, of course, just a simplified illustration of the CAO story. The real case is much more complex and actually involves false accounted profit & loss book, poor high-level management and intentional avoidance of information disclosure to the public share holders. But the author's interest is only on the company's trading strategies which in turn reflect its belief on oil prices - CAO was obviously holing a bearish view on jet fuel market. It is known that as one kind of petroleum product, jet fuel has high volatility in prices. This fact could be illustrated by its price curve (yellow) during 2003-2004 in Figure 23. Thus, it is not strange at all if completely different conclusions might be drawn by using a variety of technical analysis approaches (e.g., curve pattern recognition and chart reading). But could fundamental analysis give some hints of the underlying price movement direction in this case?

7.2  Analysis

The analysis using fundamentals of crude oil for the CAO case are based on two assumptions. The first is that crude oil and jet fuel are subject to similar composition of fundamentals, the second is that the prices of these two products are closely correlated. Assumption 1 is straightforward. With a stable oil processing capability in 2003-2004, the fundamentals to drive jet fuel (a product of the downstream) were basically the same as those that drove crude oil (feedstock from the upstream). As for assumption 2, some empirical evidence need to be shown to make sure that it was sound for the 2003-2004 time frame.
[tp]
Figure
Figure 23: Crude Oil and Jet Fuel Prices, 2003- 2004
 
To prove assumption 2, the following data are collected from EIA: spot price for WTI at NYMEX, spot price for New York Harbor Kerosene-Type Jet Fuel at NYMEX, and future contract 1 price of WTI at NYMEX. The spot prices are of January 2003 to January 2004 period, while the future contract price is of December 2002 to December 2003 period, due to its one-month advance of spot prices. To smooth the market fluctuations of jet fuel due to its own supply-demand dynamics, a 60-day moving average of the jet fuel spot price is also calculated. All data are plotted in Figure 23. It could be seen that despite the trading decomposition of crude oil spot markets and futures markets, the spot price and the future contract price of WTI almost fully overlap with each other. In the mean time, although the spot price of jet fuel has lots of location variations that are different from crude oil, its average trend (here in two-month) agrees with crude oil pretty well. Thus, assertions can be made that the following equations holds for the 2003-2004 period and it is sufficient to use WTI futures price to study the CAO case:

WTISpotPrice µ WTIFutureContract1Price
(7.1)
and
JetFuelSpotPrice µ WTISpotPrice
(7.2)
 
After deriving equations 7.1 and 7.2, the attention is switched onto crude oil futures prices and examination of the fundamental drivers are conducted. To simplify the analysis process, only the fundamentals mentioned in this report are used, namely supply-and-demand balance, geopolitical factors, inflation factors and market expectation. The ultimate goal of the analysis is a mocked market view that CAO trader should take in year 2004 - 1 for strong bearish and 5 for strong bullish, and the values in between denote mild bearish, neutral and mild bullish, in ascending order. The scoring process is broken down to each fundamental on a weighted basis (Table 6). After the scores for each fundamental (each ranges from 1-5) are derived, the weighted sum is calculated as the overall score.
[h]
Fundamental Scoring Weight (Percent)
Supply-and-Demand Balance 40
Geopolitical Factors 40
Inflation Factors 5
Market Expectation 15
Table 6: Scoring Weights of Each Fundamental in CAO Case Study
 
The demand for crude oil was in an growing trend in 2004. At the end of 2003, both International Energy Agency (IEA) and EIA issued reports estimating crude oil demand in 2004 to increase over 1 million barrel a day. Different factors were responsible for this increasing trend. Economy of China was maintaining a strong up arrow and accounted for over one-third of the crude oil demand increase in 2003. Side-by-side to China's economical achievement was the wake up of U.S. economy. The U.S. GDP growth rate was 6.6% in 2003, a historical high since the bust of Internet bubble and kept an up momentum. Global air traffic, which had been hit heavily by 9-11 was recovering and brought a higher demand for jet fuel. At the same time, the supplies were tight. OPEC decided to decrease its production quota in November 2003, by 0.9 million barrels a day. This was a strong sign for decreased global crude oil production, and OPEC's resolution to pull-up oil price. For Iraq, reconstruction of the oil field and pipe infrastructures had been largely postponed. With all these factors considered, this report assigns the 2004 score of 4 for the supply-and-demand fundamental.

As far as the geopolitical factors were concerned, the concern of terrorist attack was major among OECD countries. With Osama Bin Laden still at large, Al-Qaeda kept spreading hoaxes about assaults to civil and government facilities in 2004. For example, an unusual amount of nuclear power plants were shut down in Japan. There were not planned maintenance shutdowns; the plants were taken out of operation because of fears relative to their safety. The dates at which they would be brought back on stream were repeatedly postponed, in February 2004, most had not yet been restored to service, so Japan had to rely more heavily on thermal power plants burning crude or fuel oil. Accounting for all these factors, this report assigns the 2004 score of 3.5 for the geopolitical factors fundamental.

As mentioned earlier in Section 5, market expectations could be partially observed from the difference between futures price and the spot price. The 2004 crude oil futures curve maintained to be closely overlapped with the spot price curve. Also, market commentaries about price direction had been neutral and unclear. Based on these observations, this report assigns the 2004 score of 2.5 for the market expectations fundamental. Because inflation factor is a long-term fundamental, its effect on crude oil futures price is relatively small in a single year and this report assigns the 2004 score of 2.5 to the inflation factors fundamental.

To sum all scores up, the 2004 score for crude oil futures market view is: (4 * 0.4 + 3.5 * 0.4 + 2.5 * 0.05 + 2.5 * 0.15 ) = 3.3, between mild- and strong-bullish. Surely this calculation is quite coarse, but still could serve as a warning sign for CAO's bearish bets of the market, especially when it was speculating with option derivative, as the lever effect would make any decision mistake very risky. This risky nature of CAO's trading strategy was exactly the root cause for its huge loss in the oil deal.

8  Conclusions

This project tries to use fundamental analysis approaches for in-depth study of crude oil futures price. Although quantitative models have been very successful in the financial derivatives market, fundamental analysis is still an indispensable element for energy derivatives, which crude oil futures is one kind of. At the beginning of this report, the basics of crude oil and its future contracts are introduced. With the help of two main data sources: EIA and API, as well as a compilation of news articles, analysis are subsequently conducted over basic and more advanced fundamentals.

In basic fundamental analysis section, the 4-year price increase of crude oil futures from 2003 to 2006 is analyzed. Two fundamentals are presented: supply-and-demand balances and geopolitical factors. Each fundamental are first introduced with concrete examples, then used to explain the price increase phenomenon addressed. The section comes up with the argument that strong growth in demands, geopolitical risks and limited surplus in both upstream and downstream capabilities were the main pull-up drivers for crude oil future price during the 2003-2006 period. The further fundamental analysis section first demonstrates the use of inflation factors, then the market expectations, the other important fundamentals. Concrete examples are also given for each of these two fundamentals.

An important fact about crude oil futures prices is that seasonality plays an important role as one determinant. However, seasonality analysis is not a topic of this report. One reason is that seasonality patterns are not fundamental themselves, but rather derived results of other fundamentals. This is why seasonality analysis is often listed in parallel with fundamental analysis and technical analysis as approaches to study crude oil futures prices. The other reason, as described in Chapter 9 of [11], is that pure random price variations are very probably to cause seemly obvious "seasonality" patterns and result in misleading conclusions. The last but not least reason is that seasonality analysis has already been largely quantified and standardized by many popular statistical software packages, such as SAS.

After basic fundamental analysis and further fundamental analysis sections, four quantitative models that use fundamental as parameters for crude oil price prediction are presented and evaluated. The evaluation results using NYMEX data from 2003 to 2006 show that the model that incorporates the spread between spot price and futures price to predict future spot price is relatively more accurate. These results also partially reflected the quantified influential power of futures price in determination of future spot prices.

A case study about China Aviation Oil's trading loss in 2004 is presented and studied in Section 7, using the fundamentals listed in the previous sections for post-mortem analysis. The study is led by explanation of the interchangeability between crude oil future contract and jet fuel spot prices during 2003-2004, then followed by drilling down the contexts of several key fundamentals during that period. The case study shows that fundamental analysis could help to identify market directions if there are strong signals present, and control trading risks when properly used.

To identify the fundamental drivers for crude oil during a specific time period and make correct analysis of their effects are both non-trivial tasks, and require lots of experience and common sense. This report, by no means, claims itself to be error-free and a trading guide that could be followed. As stated in Section 1.2, its main purpose is to literally reproduce the processes of learning, practicing and analyzing. The author benefited from these processes a lot and sincerely wish this report could deliver them to the readers. Feedbacks in all forms about errors and improvement opportunities are greatly welcome.

Acknowledgements

The author owes sincere thank to Professor Stanley R. Pliska, who is his MISI project advisor and a renown scholar in finance. Without his substantial help in weekly meetings, data set collection and literature review, the completion of this project would have been significantly more difficult and slower. The author would also like to thank Jianhong Zhou for many useful discussions and suggestions. This report is prepared with a nicely edited LATEX template made by Matthias K. Gobbert [6].

A  Glossary of Terms

API
American Petroleum Institute.

API gravity
An arbitrary scale expressing the gravity or density of liquid petroleum products devised jointly by the American Petroleum Institute and the National Bureau of Standards. The measuring scale is calibrated in terms of degrees API. Oil with the least specific gravity has the highest API gravity. The formula for determining API Gravity is as follows: API Gravity = (141.5/Specific Gravity at 60°F) - 131.5.

Brent Crude
Brent Crude is one of the major classifications of oil consisting of Brent Crude, Brent Sweet Light Crude, Oseberg and Forties. Brent Crude is sourced from the North Sea. The name 'Brent' comes from the formation layers - Broom, Rannoch, Etieve, Ness and Tarbat. Oil production from Europe, Africa and the Middle East flowing West tends to be priced relative to this oil, i.e. it forms a benchmark. Brent blend is a light crude oil, though not as light as West Texas Intermediate (WTI). It contains approximately 0.37% of sulfur, classifying it as sweet crude, yet again not as sweet as WTI. Brent is ideal for production of gasoline and middle distillates. It is typically refined in Northwest Europe, but when the market prices are favorable for export, it can be refined also in East or Gulf Coast of the United States or the Mediterranean region. Typical price difference per barrel is about 1 less than WTI, and 1 more than OPEC Basket. Brent Crude has an API gravity of around 38.6.

bunker
Originally, a coal-bin on a steamship. Later, by analogy, a shipboard fuel tank for marine diesel fuel.

bunker oil
A residual fuel oil, used for marine diesel engines, power generators, and industrial boilers and furnaces. Examples: Bunker C, Singapore 380.

centistoke
A unit of viscosity, used for defining crude oil grades.

CIF price
The CIF price includes cost, insurance, and freight charges, whereas the FOB price excludes them.

crack spread
The price spread between the crude oil price and some specific amount of a product (e.g., gasoline, heating oil, or fuel oil) price.

crude oil time spread
The price differential between the first and second nearby NYMEX crude oil futures contracts.

CTS
The acronym for centistokes.

EIA
The acronym for the Energy Information Administration. Created by Congress in 1977, the EIA is a statistical agency of the U.S. Department of Energy. The EIA's mission is to provide policy-independent data, forecasts, and analysis to promote sound policy making, efficient markets, and public understanding regarding energy and its interaction with the economy and the environment.

F.O.B.
The acronym for Free on Board. The F.O.B. price excludes cost, insurance, and freight charges, whereas the CIF price includes them.

NYMEX
The acronym for the New York Mercantile Exchange.

OECD
The acronym for Organization for Economic Co-operation and Development. OECD is an international organization of those developed countries that accept the principles of representative democracy and a free market economy. It originated in 1948 as the Organization for European Economic Co-operation (OEEC), led by Frenchman Robert Marjolin, to help administer the Marshall Plan for the reconstruction of Europe after World War II. Later its membership was extended to non-European states, and in 1961 it was reformed into the Organization for Economic Co-operation and Development.

OPEC
The acronym for Organization of the Petroleum Exporting Countries. OPEC is an international organization made up of Algeria, Angola, Indonesia, Iran, Iraq, Kuwait, Libya, Nigeria, Qatar, Saudi Arabia, the United Arab Emirates, and Venezuela. Since 1965 its international headquarters have been in Vienna, Austria.

paleotempestology
Examining the physical record to gather data on hurricanes that predate the historical record.

residual
What remains from crude oil after removing the more valuable products, such as diesel, gasoline, kerosene, and naphtha.

sour crude
Crude oil that has a high sulfur content. The minimum level varies, depending on who's defining, but 2.5% sulfur by weight is high enough by any standard.

spark spread
A long position in electrical power and short position in fuel (typically natural gas) that simulates the profit from operating a power plant (e.g., a gas turbine generator). The heat rate determines the size of the short position in fuel.

spark spread option
An option on a specific spark spread.

sweet crude
Crude oil that has a low sulfur content. The maximum level varies, depending on who's defining, but 0.42% sulfur by weight is the cutoff for the NYMEX Light, Sweet Crude Oil Contract.

swing option
An option that contains an embedded quantity option.

West Texas Intermediate
A grade of crude oil that has its main delivery point in Cushing, OK. The spot price for WTI delivered to Cushing is the ultimate settlement price for the NYMEX oil futures contract.

WTI
The acronym for West Texas Intermediate.

References

[1]
The oiloholics. The Economist, August 25, 2005.
[2]
F. Black and P. Karasinski, Bond and option pricing when short-term rates are lognormal, Financial Analysts Jounal, June.-Aug. (1991), pp. 52-59.
[3]
A. Davis, How giant bets on natural gas sank brash hedge-fund trader. Wall Street Journal, September 19, 2006, Page A1.
[4]
A. Eydeland and K. Wolyniec, Engery and Power Risk Management, John Wiley & Sons, Inc., 2003.
[5]
E. D. Fischer Black and W. Toy, A one-factor model of interest rates and its application to treasury bond options, Financial Analysts Journal, Jan.-Feb. (1990), pp. 33-39.
[6]
M. K. Gobbert, Homepage with LATEX Introduction. http://www.math.umbc.edu/~gobbert.
[7]
J. Hull and A. White, Price interest rate derivative securities, Review of Financial Studies, 3 (1990), pp. 573-592.
[8]
X. Luo, Homepage with materials related to this report. http://www.cs.uic.edu/~xluo.
[9]
T. Petruno, 2006 takes a soft bounce. The Baltimore Sun, December 26, 2006.
[10]
D. Pilipovic, Energy Risk, McGraw-Hill, 1997.
[11]
J. D. Schwager, Schwager on Futures: Fundamental Analysis, John Wiley & Sons, Inc., 1995.
[12]
N. Schwartz, Ready for $262 a barrel oil? Fortune, April 11, 2006.
[13]
T. Wu and A. McCallum, Do oil futures prices help predict future oil prices?, FRBSF Economic Letter. Number 2005-38, December 30, 2005.

Footnotes:

1 Department of Computer Science, University of Illinois at Chicago. 851 S. Morgan Street, Chicago, IL 60607 (xluo1@uic.edu).


File translated from TEX by TTH, version 3.77.
On 15 Apr 2007, 19:07.