Anthonie Nicolaas Bruijnis Wim Petersstraat 7 8019BZ Zwolle s1454331
Master thesis International Business Management, specialization International Financial Management August 2011, Rijksuniversiteit Groningen
First supervisor: Prof. Dr. L.J.R. Scholtens Second supervisor: Dr. N. Brunia ing.
“Does the changing oil price differential impact the share price of U.S . companies?”
Abstract:
In this thesis, the increasing oil price differential between West Texas Intermediate (WTI) and the Brent blend is investigated to check if this widening price gap has an impact on the stock return of U.S.
companies. The ordinary least squares method is used to calculate crude oil price exposures of U.S.
companies by both the WTI crude and the Brent blend. For comparison a similar set of calculations is
performed on European companies as well. Results show no increasing strength of the Brent blend as
source of oil price exposure in the United States at the expense of the West Texas Intermediate. Results
do show that the Brent crude is almost as valuable as a measure of oil price exposure in the United
States as WTI. Results of oil price exposure in Europe are found to be considerably lower than in the U.S.,
regardless the two crudes.
1. Introduction
“Brent and WTI have been trading increasingly as entirely separate commodities in recent weeks”
(Lawrence Eagles, head of research at JP Morgan in Financial Times, 27th January 2011)
This quote provides an important notion of the problem that I will investigate in this thesis. Recently, the price difference between West Texas Intermediate and Brent blend has been more and more increasing (see figure 1). Since the price of crude oil impacts on stock prices (Park and Ratti; 2008), I want to study if the increasing price difference of these two crude oils has influence on the stock markets as well.
The price of oil has been an interesting topic in many discussions for a long time. Ranging from its effects on the price of gasoline at a gas station to its effects on share prices of companies that have seemingly little to do with the oil market. Journalists of financial newspapers (see the paper clippings in appendix 1) and scientists are continuously reporting and investigating the effects of the oil price. Crude oil is
generally pumped up from large underground oil fields, transported and refined for further use. It can be used as a resource for gasoline, kerosene, chemicals and rubber. But even more far reaching, oil is an important energy source. This means that the price of oil seeps through in every layer of the entire economy.
Besides being a physical asset, crude oil is also a basis for a range of financial products and is thereby deeply rooted in the financial system as well. Contracts concerning delivery of crudes on a future date change ownership frequently. Buyers of these contracts are rarely interested in the underlying
commodity but participate in oil trading, in order to profit from supply and demand discrepancies and to hedge against uncertainty about future oil availability.
It should be clear that (the price of) crude oil has a lot of impact, in various ways and to various degrees, in our economy. The two most mentioned and most famous types of crude oil are the American based West Texas Intermediate (WTI) and the European Brent blend. WTI is used as a benchmark to price oil for American customers, where Brent is the benchmark for the EMEA
1region. These benchmarks are used both at the energy markets and in the academic literature. These benchmarks are then used in the academic literature to determine the effects of a price change of crude oil on, for instance, the stock
1 Europe, Middle East and Africa
price of a certain company. Benchmarks are used to price the many different types of crude oil that are present in the market. A lesser known crude is traded at a certain premium or discount (depending on, for example, the quality of this crude but also supply and demand forces) over the well-known
benchmarks.
When discussing oil in the academic finance literature, one is almost always talking about a benchmark instead of specific type of crude oil. Such a benchmark is used, for example, to determine the influence of crude oil on the share price of listed companies. For example the oil risk of listed U.S. company is usually examined by using the ‘American’ WTI benchmark. Since this is the oil that is produced, refined and consumed in the same region as where the American company is located. A European company, in contrast, is examined by using the ‘European’ Brent blend oil, with the same reasoning. Historically, the price of WTI and Brent were similar in following roughly the same pattern with just a few US dollars price difference per barrel
2based on difference of, for example, chemical properties. These chemical
properties are shortly explained in appendix 2. But the development of the oil price provides a clue that this de facto coupling between WTI and Brent might be broken, or at least is under stress. Academics also noticed the growing price gap between these two crude oils and are increasingly interested in the relationship between the prices of oil. They study whether the price of crude oil responds like it is one
‘global market’ (Adelman, 1984; Rodriguez and Williams, 1993) or that it the oil market is regionalized and not a homogenous market (Weiner, 1991).
Figure 1 shows how the oil price differential has been developing over the past 16 years. This graph shows an increasing price differential between the WTI and Brent crude. Also the sign of the differential seems to have changed in the most recent months. In light of Adelman’s (1984) statement of one global oil market, these price dynamics are at least interesting. Fattouh (2010) and Reboredo (2011) study this question by investigating the relationships of two or more crude oils. As will be shown in the literature section, most papers study the relationships between the crudes. I choose to start from a different angle.
I start from the company’s perspective and investigate if the effect of the oil price on the share prices of these companies is also changing together with the increasing price gap between WTI and Brent.
2 According to The National Institute of Standards and Technology (NIST), a barrel of crude oil contain 42 US Gallons, which matches 158.97 liters or about 35 UK Gallon. NIST is an agency of the U.S. Department of Commerce. http://www.nist.gov/pml/index.cfm, accessed August 4th 2011.
Figure 1. Oil price differential (spot price (in USD per barrel) WTI minus spot price (in USD per barrel) Brent January 1995 to December 2010. Largest premium and discount, 10.21$ and 8.65$ per barrel respectively.
In this paper a data analysis will be performed in order to study whether the oil price differential that arose the recently has impact on stock prices. By calculating the oil price exposure for both WTI and Brent blend on U.S. and European stocks, I obtain price exposures for these crudes. Since I calculate these exposures for 12 successive years (1999-2010) I will be able to compare results from years in which the WTI-Brent price differential seems normal with years that seem abnormal. I then compare the exposures and the corresponding adjusted R-squared values in order to examine if the increasing oil price differential can be seen in the oil price exposure. Results that I retrieve are limited, and no convincing arguments can be made regarding the oil price exposure. The increasing price gap between WTI and Brent blend is not visible in the stock returns of U.S. companies. This does not mean that no interesting results are found in this thesis. For example, the number of exposures of Brent blend is just slightly less than the number of exposures by West Texas Intermediate on U.S. companies. In 2009 more U.S. companies where significantly exposed to the European Brent blend than to their local WTI.
Although such a year is an exception in my twelve year period, it is notable that the European oil price in
-10-5 0 5 10 15
jan 06, 1995
jan 06, 1996
jan 06, 1997
jan 06, 1998
jan 06, 1999
jan 06, 2000
jan 06, 2001
jan 06, 2002
jan 06, 2003
jan 06, 2004
jan 06, 2005
jan 06, 2006
jan 06, 2007
jan 06, 2008
jan 06, 2009
jan 06, 2010
US Dollar
one year affects more U.S. companies than the American oil price does. Another outstanding result is the adjusted R-squared values in my two-step regression for Brent blend seems more stable year on year than the adjusted R-squared values for the WTI with respect to the U.S. companies (see figure 6).
This thesis is structured as follows: Chapter 2 describes and discusses the current state of academic literature, focusing on two fields, the oil price exposure and oil price differentials. A subsection regarding sector influences in oil price exposure is included. Chapter 3 follows with the construction and
explanation of the hypotheses I wish to use in this study. Followed by chapter 4 in which the
methodology of the models and methods is explained which I employ for my data analysis in order to test the hypotheses and answer the questions posed in this paper. Chapter 5 describes the data I study, which choices I made when constructing my data set and collecting the data. This chapter provides descriptive statistics of the data and information about my data sources. Chapter 6 presents the results of the statistical analysis. This includes the results from both steps of the two-step regression. The discussion of these results can be found in chapter 7, as well as a conclusion and the limitations of this research.
2. Literature review
In this chapter, I give an overview of the current academic literature. This chapter will discuss two
separate fields of research. These two fields are discussed since both are used in this thesis. The first
field is about the oil price exposure of firms. Literature concerning oil price exposure is complemented
with literature regarding other commodities, such as gold, and exchange rate exposures of firms. Both
exchange rate exposure and non-oil commodity price exposure literature give useful support in the study
for oil price exposure. The second field of interest the crude oil price differentials. The price difference
between two types of crude oil is called the oil price differential. The existence of oil price differences for
different crude oils are discussed. Also, reasoning derived from the literature is given, why oil prices may
differ across the world. The literature regarding oil price exposure and oil price difference are both
utilized to investigate the question whether the exposure on returns of companies by the crude oil price
is shifting jointly with the increasing price gap of WTI and Brent oil. Table 1 provides an overview of the
literature discussed in section 2.1, ‘Oil price exposure’.
Author Location of interest
Subject of study Statistics used N Results/conclusions
Jorion (1990)
US Multinationals,
exchange rates
Regression/OLS/cross sectional
287 + with degree of foreign involvement and include market portfolio in regression
Tufano (1996)
North America Hedging, gold mining Multivariate/ univeriate tests of differences
48 Little/no evidence of risk management which creates share holder value. Risk aversion with manager who are holding shares
Tufano (1998)
North America 48
Bartram (2005)
Germany Nonfinancials/ various commodities
Linear regression 490 Generally exposure with more than 5% of the firms, but nothing too important. Often hedged.
Haushalter (2000)
US Oil and gas producers Pearsson correlation, Tobit, wilcoxon rank sum, probit (truncated)
100 Firms with more financial leverage hedge more
He and Ng (1998)
Japan Exchange rate Cross sectional regression
171 25% of firms show significant positive exposure.
Aloui and Jammazi (2009)
Japan, UK, France
Crude oil, Regime switching
Markow, Egarch 228 Oil price rises have sign role in determining volatility of stock return & probab of transition across regimes
Nandha and Faff (2008)
World Regresion,
autoregression
35 indices
CAPM confirmed, beta’s sign.. Only sign positive gamma’s for mining/oil&gas. Rest (except 5) are sign.
Negative.
Chen (2010)
US Higher oil, stock market go bear?
Time varying transition probab. Markow switch
500 Higher oil price, higher probability of a bear market
Miller and Ratti (2009)
SP500 and OECD
Oil Cointegration, error corr.
Matrix
6 OECD countries: there is relation between oil price and stock return. Stock market prices increases as oil price decreases, and vice versa.
Balabanoff (1995)
US WTI spot/WTI futures (NYMEX)
Cointegration - Results show support for the view of a relationship between futures and oil stocks
Park and Ratti (2008)
EU/US Oil price shock and stock markets
cointegration 13
countries
Norway stocks positive effect by increase oil price. Rest Europe negative, US not negative (and not positive) Oil price has more effect than interest rate. No asymmetric effect oil price, except in US and Norway.
Sadorsky (1999)
S&P500 Oil price shock and stock market
Vector Autoregression Proof assymetric effects on stock market. Both oil prices and oil price volatility play important role in stock return.
Scholtens and Wang (2008)
NYSE Oil and gas companies Two step regression, multifactor APT
96 Most oil/gas firms positively sensitive to oil price.
Sadorsky (2001)
TSE April 83 – April 99, oil price stock
APT model index Use future price instead of spot. Spot prices are more affected by short term fluctuations.
Table 1. Oil price (exposure) literature summary
2.1 Oil price exposure
The meaning of the concept of ‘oil price exposure’ is important to be clarified at the beginning of this chapter. Companies can be exposed to the price of a commodity mainly by one reason. Financial theory argues that the value of equity, represented in the stock price, consists of the discounted future cash flows.
Changes in the commodity prices have an effect on these cash flows. For instance, cash flows are directly hit when a company has to pay more for the same amount of the needed commodity. These companies are said to be negatively exposed to price of that commodity. Alternatively, companies that receive more cash from buyers when they sell a product when the commodity price goes up, are said to be positively exposed.
Bartram (2005, p.171) states, additionally, that the exposure is determined by the unexpected price change of the commodity. When a price change of a company is expected by the market, the efficient market hypothesis argues that this information is already incorporated in the stock price. Besides the direction of the exposure (negative or positive) on a share price, the size of the exposure plays an important role as well.
In resemblance to the capital assets pricing model, where the beta displays the volatility of the individual stock return relative to the return of market as a whole, the volatility of the stock return relative to the commodity price is expressed in a coefficient. In this thesis the coefficient is expressed with the greek letter gamma. Beta is in financial literature the coefficient for the exposure to the market return, where gamma is the coefficient to the commodity price. Put in a formula it shows like this.
(1)
Where R
i,trepresents the daily return of firm i in period t, R
m,tthe daily return of the market index in period
t, R
c,tis the commodity price which is under investigation. γ
iis the estimate of the commodity price
exposure of firm i. Adler and Dumas (1984) for instance, use the regression coefficient in order to study exchange rate exposure.
The formula above, a linear regression analysis, is one of the commonly used methods in order to examine
the exposure of commodity prices on company value. Jorion (1990) suggested in his seminal paper to
include the market in the regression model, where he tested for exchange rate exposure on 287 US
multinationals. The model constructed by Jorion provides a two-level regression in order to test for
exchange rate exposure. He found, by using this model, that exchange rate exposure is positively related
with foreign involvement of the company. Jorion’s model has since then often be replicated and modified to
test for different types of exposures on company performance. Stulz and Williamson (1997) argue that the
model developed by Jorion (1990) using exchange rates can be used for “all of the exposures to financial risk to which a firm may be subject”. The exposure that is often studied in the academic literature is about the foreign exchange risk, also the exchange rate risk. For example, He and Ng (1998) use a cross-sectional analysis to study the exchange rate exposure of 171 Japanese multinationals. They found a positive exchange rate exposure with 25% of the firms in their sample. This means that a depreciation of the Japanese Yen has a positive effect on the returns of Japanese multinationals.
Even though commodities are a physical entity, in contrast to the abstractness of an exchange rate, their prices are not as stable as one might expect. Bartram (2005) proves, that on average the standard deviation of commodity prices exceed the standard deviation of other financial prices. The volatility of commodity prices, according to Bartram, is even higher than the volatilities of exchange rates and interest rates.
Bartram (2005) therefore argues that the impact of this volatility on “firm value is a potentially important issue for corporate risk management”(p.163). A multivariate tobit analysis by Tufano (1996 and 1998) provides little evidence that risk management creates value for the shareholders of the 48 gold mining companies in the sample. On the other hand, the papers by Tufano do show that risk aversion is present within the management of these gold mining companies. Managers who hold shares, manage risk more than managers that hold options. This may imply that management of gold mining companies are convinced that the return of a company is exposed by commodity price changes, in this particular case the gold price.
Sadorsky (1999) examines the influence of crude oil on the broad S&P 500 index using an autoregression model. Both oil price, and oil price volatility are important variables in explaining economic activity. In contrast, a change in economic activity only seems to have a little influence on the price of crude oil.
Sadorsky (1999) finds a negative relationship between the price of crude oil and stock returns. Aloui and Jammazi (2009) use a Markov-switching EGARCH model to investigate if oil price shocks demonstrates effects on the behavior of stock markets. They have 228 observations (19 years of monthly returns) in three stock indices (FTSE100, Nikkei225 and the CAC40) and come to the conclusion that the increase of the oil price has a significant effect on the volatility of the stock returns. Aloui and Jammazi (2009) also find
significant evidence that the increase in oil price has a regime switching effect on the stock markets. When a
regime switch occurs, it means that there is a structural break in series. In the paper by Aloui and Jammazi
two regimes are denoted, low mean – high volatility and high mean – low volatility. Further, interesting
research has been conducted by Miller and Ratti (2009). They analyzed the long-run relationship between
the oil price and international stock markets of six OECD countries. They employed a vector error correction
model to test several periods for this relationship. A remarkable event occurs in the final period (September
1999 to May 2008), when the prevailing relationship of the past, the periods January 1971 to May 1980 and
February 1988 to September 1999 (between May 1980 and February 1988 no significance) is disintegrated.
This final period (September 1999 – May 2008) shows substantially lower positive signs and even “wrong”
(negative signs where positive signs were expected) signs in the United States and Canada, of which the negative sign in Canada is significant. As defended elsewhere, this thesis will focus on this ‘special’ time frame. Chen (2010) chose a different starting point in order to connect the price of crude oil with the stock market. Instead of testing the impact of oil price changes on stock returns, he used a Markov switching model to test if higher oil price leads to a recession of the stock market (bear market). Chen found strong evidence supporting a switch from a bull to a bear market regime when crude oil price is higher. Besides this switch, a stock market remains longer in a bear market when oil prices are high.
2.1.1. Oil price exposures – Sector influences
Nandha and Faff (2008) use 35 industry indices to check, through a standard market model regression, if oil
price changes have an effect on the prices of equity. The stock returns in all sectors react negatively on an
increase of the oil price, except for the mining sector and the oil and gas sector. According to Nandha and
Faff (2008) 28 of the 33 sectors show negative exposure at a 10% significance level, and also most all of
these 28 sectors at a 1% level. These findings show the extensive impact the price of crude oil has on
virtually the entire real economy. A study by Papapetrou (2001) with respect to Greece shows a similar
notion. She claims, by using a multivariate vector auto regression model, that the change of oil price is an
important determinant to explain stock price movements. Where Nandha and Faff (2008) studied all
industrial sectors and concluded that most industries, except the mining, oil and gas sectors are negatively
exposed to the crude oil price, shows a study by Scholtens and Wang (2008) that most of the 96 oil and gas
companies are positively sensitive to oil price changes when using a two-step regression model. Sadorsky
(2001) used an ordinary least squares method to calculate his regression model concerning Canadian oil and
gas stock prices. As one might suspect, these Canadian oil and gas companies are indeed, positively exposed
to the crude oil price. Boyer and Fillion (2007) studied also Canadian oil and gas companies. They also found,
using panel data on 105 firms, that when both the oil and the gas price increased, this had a positive effect
on the stock return of these companies. I choose to specify my data at a sector level because sectors are
exposed on different levels. As mentioned in this section, some sectors are positively exposed, as where
most sectors are negatively exposed. According to Park and Ratti (2008) this results that even entire
countries are significantly exposed to the oil price, such as Norway. Norway’s economy benefits from rising oil prices since it possesses oil reserves and numerous Norwegian companies are active in the oil and gas industry. The fact that oil price influence differs between sectors, justifies my decision to take these sectors into account in this thesis.
2.2 Oil price differential
The second part of the literature review, after the oil price exposure, focuses on the oil price differential. An oil price differential is commonly the price difference between two types of crude oil at date t. A price differential can therefore be obtained for spot prices, but also for future contracts when the contracts of both crudes mature at the same date. Such a difference can be calculated by subtracting the spot price at date t of, for instance; Brent blend from the spot price of West Texas intermediate at date t but can also be applied using future contracts. A similar approach when using future price is also present in the academic literature. Milonas an Henker (2001) for example use the future prices of contract, both maturing at the same time, of two crudes in order to calculate the price spread.
Based on the chemical properties, one would assume that WTI is traded at a premium over Brent crude.
Milonas and Henker (2001) found an average price premium of WTI on Brent in the period 1-2-1991 – 31-1- 1996, (daily observations) of 1.29 USD per barrel (which contains almost 159 liters). Fattouh (2010) shows in his paper the crude oil price differentials (the price of the first named crude minus the price of the second named crude) of three sets of crude oil, WTI-Dubai, WTI-Brent and Brent-Dubai. He examines weekly data in the period January 1997 to January 2008. The differential applicable to this thesis, WTI-Brent shows a volatile course, but WTI usually traded against a premium over Brent oil. Figure 1 shows how the oil price differential between WTI and Brent has developed over de past 16 years, starting in January 1995 and ending in December 2010. Data for this graph is retrieved on a weekly base, in line with Fattouh (2010). The graph visualizes that the oil price differential between WTI and Brent in the most recent years seemed to have increased.
The average standard deviation of the oil price differential of the total period (1995-2010) period is 1.69
USD. The average standard deviation of separate sub periods, 1995-1998; 1999-2002; 2003-2006 and 2007-
2010 are respectively, 0.53$, 1.01$, 1.39$ and 2.68$. The mean value of the oil price differential in those
periods are respectively, 1.50$, 1.48$, 2.10$ and 0.66$. In 2010 the mean value dropped to -0.09$, which means that the value of WTI was on average almost 10 dollar cents below the value of Brent. The relevance of this seemingly changing oil price differential will be explained in the following section.
Adelman (1984) stated that the crude oil market is “like the world ocean, one great pool”. According to Adelman, crude oil can flow to places where the supply is scarcest and subsequently, the price the highest.
The flow of crude oil, is not necessarily a physical flow, it also can be effectuated by, for instance, swaps.
The ‘one great pool’ claim has been cited numerous of times in the recent literature, as well as confirmed by data analysis. For example by Fattouh (2010), using a two-regime threshold autoregressive process (testing a set of sub-systems in one autoregressive model) to test the differentials amongst seven types of crude oil.
Fattouh (2010) finds strong evidence for stationarity in the price differentials between the crudes. Also, Fattouh finds no threshold in effects in the three (WTI –Brent, WTI –Dubai and WTI-Sahara) crude
differentials. The absence of threshold effects can be the result of two things. Firstly, that deviations from a long-run equilibrium are quickly restored through arbitraging via the futures markets and that no specific regimes a present. Secondly, similar crudes are less likely to face refining difficulties, since their (chemical) properties are alike. Refining bottlenecks occur mainly between crudes that are dissimilar in their
properties. Other crude differentials, with crudes which are not similar in quality, with the exception of the Maya – Lloyd blend differential (both are heavy crudes) do show threshold effects. Geman (2005) notes that sweet and light crudes are traded at a premium over more sour and/or heavier crudes. Some critical notes are placed by Fattouh (2010), that crude oil prices are seemingly linked. This means that, at a very general level the world crude oil market is ‘one great pool’. Also, “oil markets are not necessarily integrated in every time period (…)” (p. 334, Fattouh 2010). His critiques on the ‘one great pool’ notion are supported by a study performed by Weiner (1991). Weiner uses relatively basic methods to test if the claim by Adelman is correct. Using correlation and regression on the price adjustment between regions he concludes that the world crude oil is not a unified market, but has more characteristics that would suggest regionalization.
Gülen (1999) questions the conclusions drawn by Weiner (1991) based on his findings using cointegration testing. In contrast to Weiner (1991), who used monthly data, Gülen(1999) used weekly data to analyze the co movement in the crude oil market. A reason for the shift in oil price differential posed by Fattouh (2010) lies in the transportation and storage of crude oil. Oil inventories in Cushing, Oklahoma, have been
increasing for example. Cushing is the delivery point for future contracts and also a trading point for crude
oil. It was logistically known as a bottleneck when it comes to supplying enough oil to Cushing. According to
Fattouh (2010) supplying Cushing with oil is no problem anymore. The trouble lies in the distributing oil
further from Cushing. This causes large build-ups of inventories in that area. These inventories are large to
such a degree that the price of WTI can decouple from oil prices in the rest of the world, including Brent.
These large inventories can be linked to the finding of Choi and Hammoudeh (2010). They state that Brent has more volatility persistence than WTI since Brent inventories are much smaller. Also, the production of Brent has been limited in the autumn of 2000, after which the price of Brent exceeded WTI with 3 USD per barrel.
2.3 Connecting this thesis to existing literature
This thesis uses findings from both the field of commodity price exposures and oil price differential to build upon. The oil price differential is the fuel for the question I investigate in this thesis. Where the literature discussed in section 2.2 focuses on the price differences between several types of crude oils, this paper differs from this field in the academic literature in several ways. The authors of papers regarding the oil price difference are generally not interested in the effects of the price of crude oil on the stock markets.
This is in contrast to the literature discussed in section 2.1, which is about the exposure of the price of crude
oil on the stock markets. Scholars in this field are usually not that interested in the dynamics of crude oil
price differentials. In this thesis I try to combine the questions raised by the field of the crude oil price
differentials with the techniques used in the field of oil price exposures. Since I am interested in the oil price
exposure, I will employ the techniques discussed in the literature regarding commodity price exposure (see
chapter 4). This thesis differs mainly from the standard approach of the commodity price literature at one
point. I will not only test for exposures regarding the local crude oil, but also test for exposures that may
occur with respect to the not local crude (i.e. exposure of Brent blend in the United States and exposure of
WTI in Europe). This approach allows me to compare the exposure of the crude oil price on listed companies
in the United States and in Europe without using two separate types of crude oil. But also, using the non-
local crude for exposure calculations allows me to investigate if the price dynamics of the non-local crudes
reaches beyond borders. My research relies on the oil price differential literature by using the cointegration
technique, which is used in most papers regarding this subject. This technique provides me with insights
whether the two series of crude oil are decoupled in the time period(s) which are under investigation in this
thesis.
3. Hypotheses
The main purpose of this thesis is to investigate if the changing WTI-Brent blend price differential is visible in the oil price exposure of listed companies in the United States. I found the importance for such a investigation summarized in the conclusion of the paper by Choi and Hammoudeh (2010), stating:
“Brent shows more volatility persistence than WTI. This result underscores the stronger impacts of imported oil benchmarked to Brent can have on economic activity in the United States. WTI and Brent are not perfect substitutes when it comes to volatility in economic activity.”
Combining this with the notion that WTI and Brent seem to decouple (see the literature section 2.2) in the recent years, this asks for a study that examines the influence of Brent blend on the American stock market. Such a study is particular interesting when WTI and Brent blend are decoupled. Because when both crudes are following each other closely, a price change in one crude, will also be reflected in the other. This implies that the exposure of both crudes on the U.S. stock market is similar.
Fattouh (2010) claims that in 2007 WTI decoupled from the oil prices in the rest of the world. Since WTI is then also decoupled from Brent blend, it makes sense to study the price implications of Brent on the U.S. stock market because the effects of both crudes can be unique. It is important to note that a physical oil trade from the Brent fields in the North Sea to the United States is virtually non-existing, nor is a flow from Cushing, Oklahoma, towards Europe. However, this does not dismiss the influence of Brent in the United States. For example, about 50% of the oil imports of the United States (more than 50% of the oil consumption of the U.S. is imported) has its origin in Africa or the Middle East
3and is therefore benchmarked against Brent blend, because Brent is the leading benchmark in Africa. But also oil
imported in the United States benchmarked against WTI is affected. An example of a crude benchmarked against WTI is, according to Hammoudeh et al. (2008) the heavy and sour Mexican Maya crude.
According to Fattouh (2010), Maya crude is also decoupled from WTI, even though WTI is the benchmark for Mexican Maya crude.
Brent blend oil wells are globally not the most productive oil wells, the two largest European oil
producing countries (United Kingdom and Norway), produce combined 3.5 million barrels per day, versus
3 http://www.eia.gov/energyexplained/index.cfm?page=oil_imports Accessed March 4th 2011
almost 8 million barrels produced in the United States
4. But the Brent benchmark is undeniably an import global benchmark. Cong et al. (2008) for example, use the price of Brent as a representation for the world crude oil price when they use a vector auto regression model to test the impact of oil price shocks on Chinese companies.
But not only academics give rise to the question if the WTI benchmark is losing its strength and value. In the appendix 1 some articles retrieved from recent business journals are presented (Het Financieele Dagblad, 24 feb 2011; Financial Times (Europe edition), 11 feb 2011; Financial Times (Europe edition), 27 january 2011).
The main question I want to answer is the following:
Is the changing WTI-Brent blend price differential are visible in the oil price exposure of listed companies in the United States?
As made clear in the literature review section of this thesis, I will test whether the increasing WTI-Brent blend price differential has an impact on the U.S. stock market by calculating the oil price exposure of U.S. forms. Besides the increase of the price differential of these two crudes, the sign of the differential has changed within my time frame. With WTI trading at premium over Brent blend during the better part of this time period to Brent blend trading at a premium over WTI during almost the entire year 2010.
I have chosen to use the Brent blend as a substitute of WTI for a couple of reasons. Firstly, the chemical properties of WTI and Brent blend are similar (Geman, 2005). Secondly, both WTI and Brent are linked to a highly liquid futures market (Fattouh, 2010; Hammoudeh et al. 2010)). And thirdly, WTI and Brent are both important global benchmarks in business and academic literature (Hammoudeh et al. 2008).
Although Fatthouh (2010) argues that WTI and Brent are decoupled, I will compare both crudes with each other. This raises the first hypothesis.
H1: The oil price differential between WTI and Brent is not constant over time.
Many authors (e.g. Hammoudeh et al, 2004; Hammoudeh and Aleisa, 2004; Henriques and Sadorsky, 2008; Scholtens and Wang, 2008; Sadorsky, 2001; Boyer and Filion, 2007) implicitly state by using WTI, that West Texas intermediate is the best crude to apply when studying the oil price exposure of North
4 International Energy Agency, oil report http://omrpublic.iea.org/omrarchive/12may11sup.pdf Accessed 20th June 2011
American companies, since this is their benchmark of choice in their American studies. Scholtens and Wang (2008) test the oil risk on oil and gas stocks of companies listed on the New York Stock Exchange using the WTI benchmark. Also, Boyer and Fillion (2007) use the WTI and the NYMEX natural gas prices to test for the effects on Canadian companies. Sadorsky (2001) also uses WTI in his study regarding Canadian oil and gas companies. I do not want to take these practices for granted and will investigate whether this fundamental assumption still holds in the volatile financial markets of today. This thesis studies if the increasing price differential between WTI and Brent blend impacts the returns of U.S.
stocks. To the best of my knowledge such a study has not been conducted before.
Therefore the following hypotheses are constructed.
H2a: The price exposure of WTI benchmark on US companies is constant over time.
This hypothesis does not mean that just the WTI benchmark as a measure for oil price exposure is weakening. It is for example possible that the oil price exposure on the real economy in general is diminishing. To balance this possibility, a second part is added to hypothesis 2, as shown below.
H2b: The price exposure of Brent benchmark on US companies is constant over time.
When Brent both H2a and H2b are rejected, one can argue that Brent is, at the expense of WTI,
becoming a more important benchmark when determining the oil price exposure for the U.S. market. To be able to draw more powerful conclusions, I will not only test for exposure on U.S. companies against the oil price of WTI and Brent, but also a similar test using European companies. This is implemented for reason of comparison. This provides insights, since not all countries respond similarly to oil price
changes. Park and Ratti (2008) show in their study that some economies react positively to an oil price increase. They study the effect of oil price on stock markets in the United States and 13 European countries by using a vector auto regression model. Norway shows a positive exposure to the oil price, this brings forth the following hypothesis.
H3a: Oil price exposure does not differ between the United States and the European countries H3b: Oil price exposure does not differ amongst European companies.
Based on the findings of Park and Ratti (2008), I expect H3a and H3b to be rejected.
For this study, I try to construct a large (with many companies, but also many years) sample. This is
important for several reasons. First, a long term investigation is according to Miller and Ratti (2009)
important, since many recent papers focus on short term relationships between the oil price and stock market returns. Miller and Ratti argue that sometime during 1999 a break point is present where the long-term negative relationship between the oil price and stock market returns. Based on their conclusions, I therefore start my sample collection in 1999. Secondly, available literature suggests that not all (types of) companies and not all countries are affected in the same way. Nandha and Faff (2008), Scholtens and Wang (2008), Sadorsky (2001) and Boyer and Fillion (2007) support this view. Some sectors/types of companies are showing a positive relationship between their returns and the price of oil. This applies for mining, oil and gas companies. Papapetrou (2001) on the contrary, does not make a distinction between sectors or companies in her study regarding Greek companies. Her conclusion is therefore that oil price is negatively related to stock prices. This is conflict with previously mentioned authors (Nandha and Faff (2008), Scholtens and Wang (2008), Sadorsky (2001), Boyer and Fillion (2007)) whom present positive relationships between crude oil price and stock returns. When (and if) a shift in the importance of WTI towards Brent blend regarding U.S. companies, can be exposed, it is likely that this shift is not similar for every industrial sector. Bearing this in mind, I emphatically choose to separate the sample into industries to come to, possibly , a more convincing and a more balanced conclusion.
H4: The exposure of the oil price on the returns of companies does not differ between sectors.
Just as H3a and H3b I expect H4 to be rejected, based on the academic literature available.
4. Methodology
H1 will be tested using a specialized statistical approach, where the prices of several crude oils are tested with the Johansen technique for cointegration. Cointegration is a technique which allows to test if two or more time series are integrated. The advantage of the cointegration technique over simple linear
regression is the fact that linear regressions with non-stationary data can lead to spurious regressions (Brooks, 2008). Cointegration allows using non-stationary data and applying techniques in such a way that this data can be tested. For this paper, the WTI spot price and Brent blend spot price are tested for cointegration using the Johansen procedure.
To test the hypotheses proposed in the previous chapter, I will mainly use a two-step regression analysis.
This model became popular since the paper of Fama and MacBeth (1973) in which they introduced this
‘multiperiod interpretation of the two-parameter model’. The first step of this dual step model is a time series regression analysis by using the ordinary least squares method. This step determines the exposure of the crude oil price on the return of the companies under investigation. The first step conceptualized in a formula shows as follows, based on the exchange rate exposure study by Dominguez and Tesar (2006):
(2)
R
itis the total daily return of company i at time t, R
mtis the daily return of the market portfolio for time t.
ΔR
ctis the unexpected daily change in the price of the crude oil price for time t. Both R
itand R
mtare adjusted for stock splits and dividend payments.
The second step, also based on the views of Dominguez and Tesar (2006), will test for specific elements such as country of listing, sector in which the company is active and two control variables. These control variables are size and leverage and are discussed later on this chapter.
(3)
is the crude oil price exposure as estimated in equation (2). S
iis the control variable size of company i and L
iis the control variable leverage of company i. More on these control variables further on in this chapter. C
countryis a list of 17 dummy variables to account for country specific factors in the oil price exposure. S
sectorsis also a list of dummy variables accounting for industry specific propositions in the oil price exposure. An explanation regarding these dummy variables will be given in this chapter.
Stulz and Williamson (1996) explain in their theoretical paper three different ways for computing a firm’s exposure to a risk factor. A risk factor is for example a commodity price, such as crude oil in this thesis.
One of these ways is using the cash flow statements of the companies in the sample, and investigate
how these are affected when a risk factor (e.g. crude oil price) changes. Bartram (2007) argues that
accounting data is not a adequate proxy for economic, in his paper exchange rate, exposure. Cash flow
statements for large samples are not readily available, then stock returns can be used as a proxy for
these cash flows. Bartram also states that earnings statements and balance sheets, of which a cash flow
statement can be derived, are not suitable for studying exposures over longer time periods. This induces
me to use market data (stock returns as a proxy for future cash flows) and apply regression techniques to
estimate crude oil price exposure. Rather than studying cash flow statements of the companies in my
sample.
Jorion (1990) suggests in his study regarding the exposure of U.S. multinationals with respect to the exchange rate exposure, to incorporate the market portfolio in the regression analysis. Omitting this market variable means not accounting for economic shocks, this leads to a biased estimation of the exchange rate exposure, since these rates are jointly determined with macro economic activity. This two- step regression will help test H2a, H2b, H3a, H3b and H4.
4.1 First step regression
The first step in the two-step regression is a time series regression, formularized (2) as follows:
(4)
Variables Subject Source
Endogenous Return on company stocks Daily return of company stock as retrieved from Thomsons’ Datastream
Exogenous Unexpected change in the price of crude oil Difference between the price of a barrel of crude oil on t and t-1
Market Return of the market index Daily return of the market index as retrieved from Thomsons’ Datastream
Table 2. First step variables in two-step regression
4.1.1 Market index
Connolly et al. (2000) demonstrate clearly that international exposure differs greatly between countries.
This means that country A might be very sensitive for international economic activity and shocks, where
country B is much less affected by such events. This heterogeneity between countries poses a problem in
the use of the market variable, or the return of the market index in my model regarding the European
stock market. A priori it is impossible to say which European country is sensible for international shocks,
and which is not. This implies that the return of the market index in the European part of my model can
differ across countries and that one single market index might give a low explanatory power of the
regression model. To overcome this problem I will apply a robustness check in order to check which
market index is the most appropriate for each European country. This will either be the regional
Standards and Poor 350 Europe index, or the national stock market index. The results for this test, regarding the European market, will be presented in the data chapter of this thesis.
The time period I am interested in runs from January 1999 until December 2010. This time frame is motivated by Miller and Ratti (2009), since they argue that in 1999 a breakpoint has occurred regarding the exposure of the world crude oil price and international stock markets. Miller and Ratti argue that the stability of the long-run relationship that was existing in the decades before 1999 is failing to respond as one would expect. They also state, that their empirical evidence supports the view that the relationship between the price of crude oil and the stock prices in the first ten years of this millennium has changed compared to the decades preceding the turn of the century. The findings of Miller and Ratti provide me with a rather specific start date for my sample. From 1999 until 2010 gives me twelve years to study if West Texas intermediate benchmark has diminished in strength regarding the exposure of U.S.
companies. Since the findings of Miller and Ratti (2009) concerning the break in 1999 are unexpected, I choose to divide my total time frame (1999-2010) into smaller sub periods. This allows detecting smaller and brief deviations of the oil price exposure. My sample will therefore consist of twelve time series, each comprising one year of daily data. Each year in my sample will start on the first trading day and end on the last trading day of that year. Hence, each of the twelve yearly time series will contain
approximately 260 data points.
Haushalter et al. (2002) state that daily data can cause a problem when studying stocks that are not traded frequently. Since my sample consists of only the largest companies in the United States and Europe this should not pose a problem in this particular case. Stocks of the companies in my sample are traded in a liquid manner.
In this first step, I obtain twelve oil price coefficients (γ
i) per company (one for each year) per crude. I
run this time series regressions for the U.S. companies against both the West Texas intermediate and the
Brent blend. The 24 coefficients that result from these tests are then compared in two ways. Do the oil
price coefficients change over the twelve years? And does the Brent blend exposure at U.S. companies
increase in strength, at the expense of the WTI benchmark? As mentioned in the hypothesis section, it is
possible that the relationship of the price of crude oil, so both West Texas intermediate and Brent blend,
with the U.S. stock market has changed. Therefore, I do not only test the U.S. stock market, but run the
same test for the 350 largest firms in Europe.
4.2 Second step regression
The second step in the regression analysis is cross-sectional containing either one or two types of dummy exogenous variables and two control variables. The reason for the inclusion of sector variables is discussed on page 10-11. The U.S. sample has just one type of dummy variable, that is ‘sector’. The European sample has, besides this ‘sector’ dummy, also the type ‘country’ dummy. Generally, the price of crude oil has a negative relationship with the stock markets (eg. Papapetrou, 2001; Chen, 2010). But in some specific countries a positive can be found, for example in Canada (e.g. Sadorsky (2001); Boyer and Fillion (2007)). These positive exposures, and the findings of Park and Ratti (2008) where Norwegian stock market show a positive relationship with the price of crude oil, necessitate to make a distinction between countries as well. Since, by definition, all companies in the U.S. sample are listed in the United States, a dummy variable is of no use in this sample. There are seventeen countries in the European sample. Companies from these seventeen countries compose the S&P Europe 350 index.
The dummy variable for ‘sector’ is applicable for both the U.S. and the European sample. The companies are appointed a sector by Standard and Poor’s
5. This is based on the Global Industry Classification Standard (GICS)
6, which is a collaboration between Standard and Poor’s and Morgan Stanley Capital International (MSCI). The classification used by Standard and Poor’s consists of ten sectors; Consumer Discretionary, Consumer Staples, Energy, Financials, Health Care, Industrials, Information Technology, Materials, Telecommunication Services and Utilities. Intuitively, it might be somewhat confusing that oil (and gas) drilling, mining, exploration, production, refining, transportation and storage companies are in the ‘energy’ sector. Companies that distribute gas and electricity to business and households are ranked in the ‘utilities’ sector.
Dummy variables (both country and sector) are employed as follows. In equation (3) every country (if European sample) and every sector is included. A dummy can either take value 0, or 1. A dummy will take value 1 when the sector and country matches the company of which the exposure to the oil price is being regressed. For example, the company Investor AB, only the dummy that is connected to Country- Sweden and Sector-Financials are value at 1. All other dummies take value 0. This follows because Investor AB is a Swedish Financial company. Every European company has two, and only two, dummies
5 http://www.standardandpoors.com/indices/main/en/eu (retrieved at December 14th 2010)
6 http://www.standardandpoors.com/indices/gics/en/us (retrieved at February 20th 2011)
that take value 1. Every American company has one, and only one, dummy that takes value 1. This is inflicted by the notion that a company can only be assigned to only one sector, and only one country.
The control variables used in this thesis are company size and company leverage ratio. The natural logarithm of the market value is used as a proxy for company size. The market value is retrieved annually from DataStream. Leverage is calculated by dividing total debt over equity. The inclusion of leverage as a control variable is based on Haushalter (2000) and Boyer and Fillion (2007). In the study by Haushalter, companies with higher financial leverage (i.e. relatively much debt on the balance sheet) tend to hedge more against the exposure of the oil and gas price in order to reduce their uncertainty and risk they are running. Boyer and Fillion (2007) state that controlling for leverage is essential in such a regression model. According to Bartram and Bodnar (2007) size is negatively related to the exchange rate exposure.
He and Ng (1998), in contrast, state that size is positively related with foreign exposure. They argue that large companies are less likely to face financial distress, and for this, have less incentive to hedge against foreign exposure.
4.2.1 Control Variables
Bartram (2005) states that the volatility of commodity prices are higher than the volatility of exchange rates and interest rates. This higher volatility should trigger caution in companies that are affected by commodity, since commodity price risk can cause risk to these companies. Bartram’s study shows on the other hand that commodity price risk, irrespectively the high volatilities, is not significantly higher than other financial risks. Bartram finds this to be consistent with the limited cash flows in corporations that are affected by the commodity price, and the extensive possibilities for hedging strategies. However, hedging against commodity price risk is not a free option for every company that is involved in
commodities. Haushalter (2000) shows, using a tobit model, that there is a relationship between hedging
participation and commodity price exposure. Companies with higher indebtedness tend to hedge more
actively against financial uncertainty. This is in line with the argument of Haushalter (2000) that
financing costs are reduced by risk management. Companies with smaller financial room to maneuver,
measured by outstanding debt to cash holdings, are more likely to use hedging in order to protect
themselves against commodity price risks. The notion by Haushalter is supported by He and Ng (1998),
they argue that larger firms have less incentive to hedge. In their study regarding exchange rate
exposure, they concluded that smaller firms face lower exposure, as well as firms with low liquidity and high leverage. It is rather safe to conclude that the financial default setting of a company regarding price exposure is not to hedge automatically. Hedging occurs, according to recent literature with a reason, be it, for instance, scarcity of liquidity or high leverage. This implies that companies with an abundance of cash and/or lower leverage ratio are still subject to commodity price exposure. The importance and relevance of this thesis lies in realizing this.
5. Data description
The samples in this thesis are constructed according to a reasoning and selection process which is explained next. The choice for my focus on the United States stock market was fueled by existing literature. Many authors studied, one way or the other, the effects of the price of West Texas
intermediate on the U.S. stock market (amongst others: Balabanoff, 1995; Sadorsky, 1999; Haushalter, 2000; Scholtens and Wang, 2008; Chen, 2010). This means that the literature and knowledge regarding this subject is rather extensive. Besides, the West Texas intermediate is world renowned and the historical benchmark when studying the North American market. The S&P 500 was selected for multiple reasons. It is believed that this index is a good and solid representation of the large U.S. companies. And compared to, for example the Dow Jones index (only 30 companies) it houses also a wide variety of different stocks. A sample that contains 500 stocks allows me to use rather strict selection methods when it comes to missing data. The period I am focusing on, runs from January 1
st1999 until December 31
st2010. The stock listed in the S&P500 on December 31
st2010 were not all listed on the first of
January 1999. During this twelve year time frame companies were founded, merged or seized to exist. As a result of this, stock returns of 69 of the 500 companies could not be retrieved. Of a further 35
companies data to calculate control variable were missing for one or more years in the 1999-2010 time span. This leaves me with 396 companies that have a full and complete data set to work with.
Brent blend was selected as opponent with several points in mind. As highlighted in the literature section of this thesis, the chemical properties of Brent blend crude is quite similar to those of West Texas
intermediate crude. Both are relatively light and sweet and therefore chemically comparable. Also, both
Brent and WTI have a highly liquid futures market, which makes both crudes not only a commodity, but
also an adequate financial product. One of the main reasons to start this study was also the fact that the
slightly inferior Brent is recently trading at an premium over WTI. Historically speaking a new
phenomenon. For comparing the U.S. samples, I use the European equivalent of the S&P 500, the S&P Europe 350 index. Firms from seventeen countries are listed in this index. These seventeen countries are: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United. About 70% of the market capitalization of this region is covered in this index. Both euro and non-euro countries participant, but Thomson’s Datastream provides returns calculated in US dollar.
Applying the same strict selection methods as I used for the S&P 500 index results in the loss of 62 European companies which lack stock return data for the complete time period (1999 to 2010). Further, sufficient information for 13 companies regarding the control variable data is not available. This leaves 275 European companies in my sample with complete data set.
395 useable companies from the S&P 500 and 275 from the S&P Europe 350 index means that 79% and 79,1% respectively of the both the indices are available in the separate samples. My samples are generally larger than studies that focus on companies from a specific industries. Tufano (1998) has 48 companies in his gold mining sample, Haushalter (2000) has 100 companies in his oil and gas producers sample, where Scholtens and Wang (2008) have 96 companies in the oil and gas companies industry.
Other studies investigate not on a firm level, but at index level the effects of commodity prices. Sadorsky (1999) uses the S&P 500 and Sadorsky (2001) the Toronto Stock Exchange (1 index per paper), Park and Ratti (2008) study 13 indices, and Nandha and Faff (2008) investigate even 35 indices. Compared to studies using companies from a specific industry (Tufano, 1998; Haushalter, 2000; Scholtens and Wang, 2008) and studies using indices (Sadorsky, 1999; Sadorsky, 2001; Park an Ratti, 2008; Nandha and Faff (2008) my sample of respectively 395 and 277 companies seems very large. But it fits right between studies that focus on individual companies when studying exposure. Such as Jorion (1990) when studying exchange rate exposure on 287 US multinationals. Or the 171 Japanese companies in the paper of He and Ng (1998), also reviewing the exchange rate exposure. Aloui and Jammazi (2009). Bartram (2005) studies 490 firms regarding a range of commodity price exposures.
As already mentioned in section 4.1.1, I will test the European companies for the best market index. This will either be the national index, or the S&P Europe 350 index. Calculations show that there is little difference between the national index and the regional S&P Europe index. Intuitively this makes sense.
The 350 largest European companies, with a long and strong track record are in this index. These
companies act not only domestically. It is likely that they rely more heavily on foreign than domestic
activities. I therefore choose to use the S&P Europe 350 index as the market index for the all European companies, and not to use local, national indices for the different European countries.
5.1 Data sources
Stock price returns, market index returns, company data and crude oil prices are all retrieved from Thomson Reuters Datastream. The spot prices of crude oil (both WTI and Brent) were also retrieved, as an extra check, from the Energy Information Administration of the U.S. department of Energy
7. Crude oil prices from Datastream matched exactly the prices listed by the Energy Information Administration.
When Datastream failed to present stock prices returns of certain companies in both the U.S. and the European sample, a check was performed via Yahoo Finance to confirm that this company indeed was not listed in the 1999-2010 time frame. No stock price returns could be retrieved via Yahoo Finance when Datastream failed. As already shortly mentioned, not all data for the 500 companies listed in the S&P 500 index could be retrieved, as well as the data of the 350 companies in the S&P Europe 350. The main reason for this was that companies that are in the S&P 500 or the S&P Europe 350 today, where not in that index in one or more of the years in the sample period.
5.2 Descriptive statistics
The crude oil price time series are checked for stationarity. Using the augmented Dickey-Fuller test, in line with Hammoudeh et al. (2004), I check for a unit root in the series. Results for all series are that time series are stationary. The test statistic for the WTI spot price series is -5.731 which is more negative than the 1% critical value of -3.432, hence the null hypothesis of a unit root in the crude WTI spot price is rejected. For the Brent blend series the test statistic is -5.543, which is also more negative than the 1%
critical value of -3.432. Also the null hypothesis of a unit root in the crude Brent spot price is therefore rejected. No unit root implies stationary data, where non-stationary data can lead to spurious
regressions according to Brooks (2008).
7 http://www.eia.doe.gov/ (accessed on December 4th 2010)
Both crudes are also tested for cointegration using the Johansen technique. Results show cointegration relations between the West Texas Intermediate and Brent Blend. Based on the literature discussed this is to be expected. Fatthouh (2010) for instance, found cointegration relations between WTI and Brent blend.
My two (S&P 500 and S&P Europe 350) samples are diverse in their industry background. Table 3 (below) shows in which of the ten industries the companies of the two samples are classified. As can be seen in table 3, three industries, Consumer Discretionary, Financials and Industrials, are notably larger than the other industries. And one industry has a low representation in absolute numbers in both samples, Telecommunication Services. Energy and Materials have a low representation in the European sample. Evidently, a company can only be appointed to one industry.
S&P 500 S&P Europe 350 Total
Consumer Discretionary 59 41 100
Consumer Staples 35 26 61
Energy 33 11 44
Financials 60 61 121
Health Care 43 16 59
Industrials 51 55 106
Materials 29 11 40
Information Technology 45 31 76
Telecommunication Services 7 12 19
Utilities 33 13 46
Total 395 277
Table 3. Number of companies in each industry (GICS).
Table 4 shows the distribution of companies between countries in the European sample. Countries can
be divided in roughly three categories when comparing the number of companies they supply for the
S&P Europe 350 index. Countries with the most companies in their territory are France, Germany and
Great Britain. These three countries together house about 56% of the companies in my European
sample. Italy, the Netherlands, Spain, Sweden and Switzerland are countries with an average number of
companies. Companies from these five countries account for almost 31% of the European sample. The
remaining nine countries house a relative small portion (14%) of the European companies in my sample.
Country Number
Austria 3
Belgium 9
Denmark 4
Finland 7
France 40
Germany 30
Greece 4
Ireland 4
Italy 15
Luxemburg 1
Netherlands 15
Norway 3
Portugal 6
Spain 12
Sweden 21
Switzerland 21 Great Britain 82
Total 277
Table 4. Number of companies in each European country in the S&P Europe 350 sample
Descriptive statistics regarding the two series of crude oil prices, West Texas Intermediate and Brent blend, are presented in table 5. Besides the descriptive statistics regarding the spot prices also descriptive statistics regarding the WTI-Brent blend price differential is presented in table 5.
West Texas
Intermediate spot price
Brent blend spot price WTI – Brent blend price differential Mean
50.85 49.44 1.42
Standard Error
0.48 0.48 0.04
Median
45.32 43.21 1.64
Modus
29.51 25.3 0.75
Standard Deviation
26.67 26.88 2.11
Kurtosis
0.35 0.18 5.30
Skewness
3.86 3.82 2.51
Minimum
11.38 9.76 -9.93
Maximum
145.31 144.07 21.85
Table 5. Descriptive statistics regarding the spot price of WTI and Brent blend (ins USD per barrel) and the WTI- Brent blend price differential (spot price WTI at date t minus spot price Brent blend at date t)
