1
Master Thesis—
Oil price sensitivity and oil risk premium in stock
returns of oil companies in United States
University of Amsterdam, Amsterdam Business School
Programme: MSc Business Economics, Finance track
Name: Xiaoqing Xu
Student number: 10840982
Supervisor: Dr. J.E. Ligterink
Statement of Originality
This document is written by Student Xiaoqing Xu who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
3
Acknowledgement
First and foremost, I would like to show my deepest gratitude to my supervisor, Dr. J.E. Ligterink. He is a really respectable, responsible and resourceful professor, who has provided me with valuable guidance in writing the thesis. I could not have made progress without his enlightening instruction and great patience. Besides, his
academic rigour, rich knowledge and clear thinking have impressed me a lot. I feel so lucky and happy to have an excellent supervisor.
I shall extend my thanks to Prof. Dr. F.C.J.M. (Frank) De Jong for all his kindness and help in the first stage of writing thesis. I am sorry to hear his illness and I hope he will recover as soon as possible. Last but not least, I want to thank to my parents who have supported me to study in such a good university.
Abstract
This thesis mainly researches the oil price sensitivity and assesses the oil risk
premium in the US market of oil firms in the oil and gas industry. First of all, I apply the multi- factor arbitrage pricing model to research the relationship between stock return and the changes of oil price. The first hypothesis is that the oil companies are positively exposed to oil price changes. And I find that the return of oil companies is significantly and positively related to the changes of crude oil prices over the period from 01-01-2002 to 31-12-2014. Then I apply Fama-Macbeth(1973) two-stage regression model to research the risk premium of oil price. The second hypothesis is that oil price risk is a priced factor for oil companies. I find that all the coefficients of oil price in step two are insignificant, which means that the risk of oil price is not priced for oil companies.
5
Table of Contents
Abstract ... 4
1. Introduction... 6
2. Literature Review ... 8
3. Data and methodology ... 18
3.1 Data and descriptive statistics ... 18
3.2 Methodology ... 24
4. Results ... 27
5. Robustness checks ... 40
5.1 Before financial crisis versus after financial crisis ... 40
5.2 A period of Iraq War versus a period of peace ... 40
5.3 The different functions of oil companies ... 42
5.4 The introduction of size factor ... 44
6. Conclusions... 45
1. Introduction
Oil price has changed significantly in history. For instance, in the middle to late of the 20th century, the oil crises have appeared three times in the world. The Fourth Middle East war, happened in Oct. 1973, caused the oil price increase from $3.011 per barrel to $10.651 per barrel. Similarly, Iranian Revolution in 1978 and Gulf War in 1990 also led the increase of crude oil price. Moreover, the crude oil price reached its peak on almost $143 per barrel in 2008.The changes of crude oil price are often regarded as an important factor affecting the world economy. Many previous studies have
researched the relationship between the oil price and the stock return. For instance, Kling (1985) has pointed out that the changes of oil price have negative and significant relationship with the stock market. Similarly, Jones & Kaul (1996) presented that the increases of stock return are accompanied by the decreases of oil prices. And Sadorsky (1999) also has pointed out that the stock return is significantly related to the changes of oil price.
The first part of this thesis mainly focused on the oil price sensitivity. After that, I try to find out whether there exists a risk premium for oil price to the listed oil
companies. Moreover, stock is often regarded as a barometer of the real economy. But the relation between the crude oil price and the stock return of oil companies still exist debates. A great number of existing studies such as (Jones & Kaul, 1996), (Sadorsky, 1999), (Scholtens & Wang, 2008) and (Wei, 2003) have acknowledged the
7
different opinions whether these two elements are related. Huang (2006) and Wei (2003) find no relationship between stock returns and changes in the price of oil futures. So, the first hypothesis is that the oil companies are positively exposed to oil price changes.
There exist different methodologies in existing studies to research this topic. The main methodologies are vector error correction (VECM) model and arbitrage pricing theory (APT) model. The main topic in this thesis is to research the oil risk exposure and oil risk premium, so I intend to apply arbitrage pricing theory (APT). In addition, I want to explain the different oil price sensitivities by macroeconomic factors and oil industry-specific factors. I also intend to follow the two-stage Fama-Macbeth (1973) regression model to estimate the premium rewarded to a particular risk factor
exposure for listed oil companies by the market.In the first step, the stock return of oil companies is regressed against four fundamental factors (crude oil price, market return, default risk and interest risk) time series to determine the factor exposure. The second step is to compute cross-sectional regressions of the returns on the estimates of the betas calculated from the first step to calculate the factor premiums for each factor. And the second hypothesis is that oil price is a priced factor to oil companies.
The whole test period is ranging from 01-01-2002 to 31-12-2014. There are 666 weeks in all within the test period. I select the oil companies which have unbroken series of historical prices and list in the U.S stock market during this period. There are 62 oil companies in accordance with these conditions. Moreover, all the returns are calculated by weekly data.
Contrary to existing studies, the contribution of this thesis is that this paper not only researches the fundamental factors to oil industry in multi- factor arbitrage pricing model but also analyzes the differences between the spot oil price and future oil price further. In addition, I also research the oil risk premium over the test period. In the robustness test, I have classified the oil companies by their main business to research whether the regression model have the same results, Besides, I also introduce the size factor which is similar to Fama & Macbeth (1973) to check whether the results will change.
The thesis is organized as follows. Section 1 introduces the whole thesis briefly. Section 2 gives the literature review about fundamental factors in stock returns of oil companies in United States. Section 3 presents the data used and the methodology. Section 4 shows the results and their interpretation as well as robustness tests. Section 5 points out the conclusions.
2. Literature Review
The changes of crude oil price are often regarded as an important factor explaining stock returns. Many previous studies have researched the relationship between the oil price and the stock return. For instance, Kling (1985) has pointed out that the changes of oil price have negative and significant relationship with the stock market. Similarly, Jones & Kaul (1996) presented that the increases of stock return are accompanied by the decreases of oil prices. And Sadorsky (1999) also has pointed out that the stock return is significantly related to the changes of oil price. In addition, Killian & Park
9
(2009) find that stock prices have negative response to oil price shocks. However, other studies have different opinions of this relation. For example, Huang et al. (1996) have stated that the increase of oil price fails to explain the decrease of S&P 500 index return from 1979 to 1990. And Wei (2003) also found that there is no negative relationship between crude oil price and stock return in U.S markets over the period from 1973 to 1974. But most of the related studies have proved the relationship between the oil price and aggregate stock return such as Kling (1985), Jones & Kaul (1996), Sadorsky (1999), Killian & Park (2009) and so on.
This thesis puts more emphasis on the relationship between oil price and stock return of oil companies. Similarly, many existing studies have pointed out the
significant relationship between them. For instance, Al-Mudhaf and Goodwin (1993) have found out that the higher oil price can stimulate the higher stock return of oil companies. Sadorsky (2001) has found that the increase of oil price can explain the increase of stock return for oil companies in the Canadian markets. And Boyer & Fillion (2007) also found the positive and significant relationship between the oil price and stock return of oil companies in Canadian market over the period from 1995 to 2002. Additionally, Hammoudeh et al. (2004) have represented that the changes of oil price have a significant and positive relationship with the stock return of oil companies. Scholtens & Wang (2008) have focused on the U.S mark et and they have pointed out that returns of oil stocks are positively associated with the return of the market and the changes of spot crude oil price. Cong et al. (2008) have showed that the positive oil stock returns are influenced by the higher oil price. In summary, these
listed studies have showed the significant and direct relationship between the oil price and stock return of oil companies. So they provide me enough support to raise the first hypothesis. The first hypothesis is that the o il companies are positively exposed to oil price changes.
There exist two main methodologies in researching the relationship between the stock return of oil companies and oil price in prior studies. They are vector error correction (VECM) model and arbitrage pricing theory (APT) model.
First of all, vector error correction model is helpful to research plausible economic relations. And Maysami and Koh (2000) also have showed the feasibility in applying VECM model to research the relation between stock returns and macroeconomic factors in Singapore markets. Killian (2010) has applied VECM model to research the research the relationship between crude oil prices to gasoline price. Similarly,
Hammoudeh et al. (2004) found the stock price of oil companies, such as drilling companies and marketing&refining companies, are affected by the change of WTI oil price with the help of VECM model. And then Hammoudeh & Li (2005) also used VECM model to research and they pointed out the increase of oil price is related to the decrease of market return. Lanza et al. (2005) also applied VECM model to find out the fundamental factors to the sixth largest oil companies in the U.S. The VECM model is beneficial to estimating both short-term and long-term relationship between variables. The application of VECM model is beneficial to decrease the possibility of misspecification.
11
research this topic. For example, Al-Mudhaf & Goodwin (1993) used a multi- factor APT model to research return differences in 29 US oil companies in a period
surrounding the oil shock of 1973. Sadorsky (2001) applied the APT model to find out whether oil price can be explained by four other variables such as market return, crude oil price, exchange rate and interest rate. Boyer & Filion (2007) use APT model to measure the exposure of stock returns to different risk factor and investigate the determinants of stock returns of Canadian oil and gas companies. Scholtens & Wang (2008) also use APT model to assess the oil price sensitivities and oil risk premiums of NYSE listed oil & gas firms' returns. Moreover, contrary to a capital asset pricing model (CAPM) to explain the expected return of risky assets on the basis of the market return, multifactor arbitrage pricing theory is more appropriate to research this topic. First of all, some studies have presented the infeasibility to apply CAPM model. For instance, Morin (1980) shows that the market returns have not a significant power in Canadian markets through CAPM model. And Jorian & Swarts (1986) also use a CAPM model to point out that the North-American market returns cannot explain the reaction of stocks in Canadian markets. In addition, Gibbons (1982), Fema &French (1992) also argue the feasibility of applying CAPM model. Second, Bower, Bower, & Logue (1984) presents that arbitrage pricing theory (APT) leads to different and better estimates of expected return than the Capital Asset Pricing Model (CAPM), especially in the case of utility stock returns.
So I prefer applying multi- factor APT model to VECM model in this thesis. First of all, the main purpose of this thesis is to research oil price exposure and oil risk
premium, and the APT model is more direct to detect the long term oil price
sensitivity and risk premium. In addition, Brooks (2002) has presented the application of VECM model needs a much richer data structure. My sample is weekly data
ranging from 2002 to 2014. I am not sure that my data is sufficient enough to capture the dynamic properties by using VECM model.
The first hypothesis is that the oil companies are positively exposed to oil price changes. So the first step to build APT model is to choose suitable variables which may affect the return. Then I can get the result of price sensitivity through the
regression. Many existing studies have represented the introduction of determinates of the stock returns.
First of all, oil price is the most important factor which is introduced in the APT model to explain the stock return of oil companies. Addition to the listed studies before, Pincus & Rajgopal (2002) have stated that firms engaged in oil exploration and drilling are exposed to two kinds of risks that can cause earnings volatility: oil price risk and exploration risk. And Mohn (2008) have found out that an increase in the oil price will lead to more effort and efficiency of exploration. Furthermore, I intend to compare the differences between effects of crude oil future price and the crude oil spot price. Because Sadorsky (2001) argues that the futures oil price should be used in the model because spot prices are affected by short-run price fluctuations due to temporary shortages or surpluses. In other words, the oil companies have different proportions of spot contracts and future contracts in real life. I contend that it is more comprehensive to take all kinds of oil contracts into account compared to
13
previous studies.
Secondly, the market portfolio return is also considered as a significant role in the return of oil companies in other studies. Ferson & Harvey (1991) mention that the real interest rate and market return are the most important determinants for the return of American petroleum shares. Moreover, Kavussanos & Marcoulis (1997) study the Fema & French (1992) factors on the profitability of oil refining firms, and they find market return plays the most significant role in the share price of refining firms. Thus, I want to add the market return as another variable into the regression model.
In addition, Chen, Roll, & Ross (1986) have employed a set of macroeconomic variables on stock market returns and researched the relationship between stock returns to these variables. They find stock returns are exposed to these systematical variables such as industrial production, term structure and risk premia in the U.S. stock markets. Moreover, Koutoulas & Kryzanowski (1994) find the domestic
components of interest rate structure, lagged industrial production and interest rate of Euro deposits have a significant influence on the return of Canadian stocks. Mitto (1992) also mentions the 3- month T-bill interest rate explains the stock return. Furthermore, Abeysekera & Mahajan (1987) and Kryzanowski & To (1983) have noted a model with four or five variables might be enough to get a explanatory power. Taking their suggestions into account, I intend to mainly follow the basic
determinants of stock return in a multiple arbitrage pricing model from Chen Roll and Ross (1986). These variables include industrial production, risk premia and the term structure.
Besides, some studies also provide their opinions in researching the relationship between the industrial productions to the stock markets. For instance, Ahmed (1999) has found an absent or very weak relationship between stock returns and industrial production when he researches the major factors of the stock return and the
surrounding environment. In addition, Hamilton & Lin (1996) point out that the process followed by stock returns is completely unrelated to industrial production. Though Chang & Pinegar (1989) find that coefficients in industrial production for small firm are positive and significant in the presence of the stock market factor, most oil companies are very big companies owing to the particularity of oil industry. Thus, I don’t intend to add industrial production into the regression model.
In summary, my APT model is composed of four determinates. They are the changes of oil price, market portfolio return, risk premia and term structure. But the listed variables may not be sufficient to explain the risks of firm’s specialty. Norton (1994) states that unsystematic risks are mainly caused by factors specific to a firm or industry, such as labor strikes, raw material price movements, or advertising
campaigns. So it is also significant to choose determinates for presenting firm-specific risks. One of the most significant studies in this field is Fema & French (1992). This paper have pointed out that the stock return is also closely related to the two easily measured fundamental factors, size and book-to-market equity. Morover, Scholtens & Wang (2008) and Kavussanos & Marcoulis (1997) study the Fema and French (1992) factors because they can capture not only the systematic risk but also unsystematic risk through this method. Scholtens & Wang (2008) find that the return of oil stocks is
15
positively associated with the return of the market, the increase of changes of spot crude oil price, and negatively with the firm's book-to- market ratio. In addition, Basu (1983) shows that earnings-price ratios (E/P) help explain the cross-section of average returns on U.S. stocks. Apart from that, Osmundsen, Asche, Misund & Mohn (2006) state that many analysts put more emphasis on return on average capital employed (RoACE) to research the accounting return measures for oil companies. But I don’t intend to add this variable because the conclusions of this paper suggest that the general perception of RoACE is not an important valuation metric in the oil and gas industry. So I want to introduce another two variables to present the industry-specific risks. They are factor return of book-to- market ratio and factor return of
earnings-to-price ratio.
Furthermore, the previous studies also research whether the oil price is a priced factor for oil companies. For example, Al-Mudhaf &Goodwin (1993) have found an extra oil risk premium in the period of oil shock. And Lanza et al. (2005) have also pointed out the investors need a higher return for the risk of oil price. However, Scholten & Wang (2008) have found an oil risk premium in the integrated model and no oil risk premium in the macroeconomic model. However, Chen, Roll, & Ross (1986) have found no risk premium over the sample period. These studies lead to the second hypothesis in this thesis. The second one is that o il price risk is a priced factor for oil companies. But the APT model is not enough to explain the oil risk premium. I want to follow two-step regression model which is introduced by Fama & Macbeth (1973) to research the oil risk premium. I intend to compute cross-sectional
regressions of the returns on the estimates of the betas calculated from the first step to calculate the factor premiums for each factor.
But Ahn & Gadarowski (1999) have pointed out that the disadvantages of Fama & Macbeth (1973) two-step cross-sectional regressions are that they may lead to autocorrelation and heteroskedasticity problems.
Table 1 has presented the brief introductions to the main listed studies in the literature review. They mainly include results, methodologies and research markets.
17
Aut
hor
Yea
rRe
sea
rch
Pe
riod
Res
earc
h m
ark
et
Me
thod
olog
y
Res
ult
Ham
mou
deh
et a
l.
2004
199
5-2
001
U.S
oil
com
pan
ies
VE
CM
mo
del
the s
tock pr
ices
of oi
l com
pan
ies a
re a
ffec
ted by
the
cha
nge
of W
TI oi
l pr
ice
Ham
mou
deh
& L
i
2005
198
6-2
003
U.S
, M
exic
o an
d N
orw
ay o
il co
mpa
nies
VE
CM
mo
del
the i
ncre
ase
of oi
l pr
ice i
s re
late
d to t
he de
crea
se of
ma
rke
t re
turn
find ou
t the
fun
dam
enta
l fac
tors
to t
he s
ixth
larg
est oi
l com
pan
ies i
n th
e U
.S
find t
hat i
nve
stor
s ne
ed a
hig
her r
etur
n for
the
risk of
oil pr
ice
Al-Mu
dha
f &
Good
win
1993
1973
U.S
oil
com
pan
ies
AP
T m
ode
l
an e
xtra
oil r
isk pr
emi
um i
n th
e pe
riod o
f oil
shoc
k
Sador
sky
2001
198
3-1
999
Can
adia
n oi
l co
mpa
nies
AP
T m
ode
l
the i
ncre
ase
of oi
l pr
ice c
an e
xpla
in th
e in
crea
se of
stoc
k re
turn
for
oil c
ompa
nies
in t
he C
ana
da m
arke
ts
Boy
er &
Fill
ion
2007
199
5-2
002
Can
adia
n oi
l co
mpa
nies
AP
T m
ode
l
the pos
itive
and s
igni
fica
nt r
elat
ions
hip be
twe
en t
he oi
l pr
ice a
nd s
tock r
etur
n of
oil c
ompa
nies
in C
ana
da m
arke
t
retu
rns of
oil s
tocks
are
pos
itive
ly a
ssoc
iate
d w
ith t
he r
etur
n of
the
ma
rke
t
the c
han
ges
of s
pot
cru
de oi
l pr
ice a
nd f
oun
d an
oil r
isk pr
emi
um i
n th
e in
tegr
ated m
ode
l
Che
n, R
oll, &
Ros
s
1986
195
8-1
984
all t
he c
omp
anie
s in
U.
S m
ark
ets
AP
T m
ode
l
ther
e is
no oi
l ris
k pr
emi
um ov
er t
he s
ampl
e pe
riod
Kill
ian &
Pa
rk
2009
197
3-2
006
all t
he c
omp
anie
s in
U.
S m
ark
ets
VA
R m
ode
l
stoc
k pr
ices
hav
e ne
gati
ve r
espon
se t
o oi
l pr
ice s
hoc
ks.
Jon
es &
Ka
ul
1996
197
0-1
988
all t
he c
omp
anie
s in
U.
S m
ark
ets
AP
T m
ode
l
the i
ncre
ase
s of
stoc
k re
turn
are
acc
ompa
nied by
the
dec
reas
es of
oil pr
ices
.
Sch
olte
ns &
Wa
ng
2008
200
2-2
005
AP
T m
ode
l
U.S
oil
com
pan
ies
Tab
le 1
: li
stin
g o
f pr
evio
us l
iter
atu
res
Lan
za e
t al.
2005
199
8-2
003
VE
CM
mo
del
Six
larg
est
oil c
omp
anie
s in
U.
S
3. Data and methodology
3.1 Data and descriptive statistics
My sample consists of all the oil companies listed on New York Stock Exchange (NYSE). All the oil companies are collected from Yahoo Finance. The sample period is from 01-01-2002 to 31-12-2014, and I select the oil companies which have unbroken series of historical prices d uring this period. Moreover, all the returns are calculated by weekly data. There are 666 weeks in all during the testing period. Finally, there are 62 oil companies in my test sample. There are 28 oil companies engaged in equipment and services. And there are 20 oil companies engaged in drilling and exploration. The main business for other 14 oil companies are refining and marketing services. Most of the oil companies are native American oil companies. These oil companies are mainly located in Houston, Calgary and Oklahoma City. The sample size of Al-Mudhaf & Goodwin (1993) is 29 firms and Lanza et al. (2005) is 6 firms, which is much smaller than my research. However, the sample size of Boyer & Filion (2007) includes 105 firms, but their target market is Canada. Scholtens & Wang (2008) have 96 oil companies, but their sample period is smaller than my sample. Their period is from January 01-01-2002 to 31-12-2005.
In the first step, I have to collect five macroeconomic factors which affect the stock return of oil companies. They are stock return of oil companies, crude oil price, the market return, default premium and the term structure. Table 2 has presented the summary of descriptive statistics of these variables. And it shows descriptive statistics
19
in five aspects. They are mean, median, maximum, minimum and standard deviations. I intend to describe and analyze these variables one after another in further detail below.
First of all, I excerpt the ticker of oil companies listed on Yahoo Finance and look up their ticker on Research in Security Prices (CRSP) to get the data of the stock return of oil companies. The CRSP database can provide me the historical weekly stock price of all the oil companies during testing period. All individual stocks' historical prices are closing prices adjusted for capital gains. I select oil stocks with unbroken series of historical prices by comparing the ticker. I choose the oil
companies with the same ticker between 01-01-2002 and 31-12-2014. There exist some companies which have delisted during this period. And there are some companies which change its official company name. So I have to check all the oil companies by hand and delete the repeated companies. The stock return is based on the change in the closing price from the last trading day of the week to the next
Variable Mean Median Maximum Minimum Std. Dev.
Stock return of oil companies 0.002594 0.002582 8.883177 -0.593505 0.081582
Excess return of oil companies -0.0096 -0.01044 8.831777 -0.961052 0.094689
Return of spot oil price 0.001425 0.00427 0.243179 -0.332848 0.051878
Return of 1-month future oil price 0.001401 0.003857 0.241221 -0.287988 0.049801
Return of 3-month future oil price 0.001403 0.00441 0.211343 -0.243314 0.044694
Return of 6-month future oil price 0.001452 0.003366 0.185984 -0.210451 0.040427
Return of 9-month future oil price 0.001509 0.003149 0.169975 -0.193056 0.037659
Market portfolio return 0.001211 0.002442 0.128939 -0.195348 0.026044
Market portfolio excess return -0.01223 -0.01006 0.128739 -0.195948 0.030616
Risk premia -0.03235 -0.03456 0.063791 -0.098911 0.021101
Term structure 0.018147 0.0202 0.0354 -0.0075 0.009833
Factor return of B2M 0.027165 -0.01961 0.019417 -0.117911 0.030944
Factor return of EP 0.045116 0.028547 0.228576 0.0000713 0.046354
Factor return of size -0.07799 -0.04968 -0.00649 -0.298018 0.076401
trading day minus the risk free rate. The second row in Table 2 has showed that the average weekly stock return is about 0.26%, which includes 36895 observations over the sample period. And I also find that the value of average weekly stock return is the highest in 2013, which is nearly around 11%. Moreover, the third row in Table 2 has showed that average weekly stock excess return for oil companies is -0.096% during the test period. And the median of excess stock return is about -0.1%. The average standard deviation is around 0.094. The excess return is calculated by weekly stock return minus one- month treasury bills. And the excess return in 2013 reaches its peak which is similar to weekly stock return.
Secondly, to know about the returns of crude oil price in U.S, I regard WTI (West Texas Intermediate) as the best oil benchmark price for my thesis. First of all, my target market is United States. WTI is the underlying commodity of New York Merchantile Exchange's oil futures contracts, which is more suitable than other three main oil benchmarks in the world. Second, Hammoudeh, Ewing & Thompson (2008) point out WTI and Brent are highly liquid, very actively traded, and adjust to
equilibrium in the long run. Third, Gülen (1999) states that NYMEX Division light sweet crude oil is the world's largest-volume futures contract trading on a physical commodity.The historical data of West Texas Intermediate (WTI) is composed of two parts: they are the spot price and the future price. The crude oil spot price is available in FRED economic data. And the crude oil future prices are collected from Quandl database. I intend to compare the difference between spot price and future price. I want to test the crude oil future price which will expire in one month, three months,
21
six months and nine months, separately. Because Sadorsky (2001) argues that the futures oil price should be used in the model because spot prices are affected by short-run price fluctuations due to temporary shortages or surpluses. Graph 1 has showed the movement of crude oil spot price and 6-month oil future price over the sample period. I can see the crude oil price has experienced a dramatically change during the recent ten years. The crude oil price has gradually increased from 2002 to 2008. And it reached the peak to $145.31 per barrel in 2008. After that, the oil price decreased a lot from $140 per barrel to nearly $40 per barrel. In recent several years, the oil price also went a downward trend and it is almost $50 per barrel at the end of 2014. Through the graph 1, I also find that the movement of oil future price is roughly consistent with the movement of oil spot price.
In addition, the data in Table 2 has presented the descriptive statistics for different average weekly oil price return. The fourth row in Table 2 shows that average weekly return for the spot price is around 0.1425% and standard deviation is about 0.05. I also
find out that the return of crude oil spot price reaches its highest point in 2009, which is around 1.8%, and this return reaches its lowest point in 2008 to -1.9%. Additionally, the average weekly returns for oil future price in Table 2 are also ranging from 0.14% to 0.15% over the sample period. And the lowest point for standard deviations is around 0.37 for 9- month oil future price. I also find out that standard deviations become smaller when the oil future contracts become longer.
The factor of market return is calculated according to NYSE Composite Index because it represents market portfolio return. The historical weekly data of NYSE Composite Index is also collected from Yahoo Finance. I also collect the weekly closing price for NYSE Composite Index to measure the factor of market portfolio return. The ninth row in Table 2 shows that market portfolio return is 0.12%. And the tenth row in Table 2 has presented the statistics of weekly excess market return, which is calculated by return of market portfolio minus the rates of one- month treasury bills. Table 2 shows that market excess return is around -1.22%. The whole average
negative market return may due to the financial crisis happened in 2007. Because I also find that the return of market portfolio is the lowest in 2007, which is nearly -4.1%.
Here, I want to capture the effects on returns of unanticipated changes in default premium by calculating the different between the “BAA and under” bond portfolio return and long-term government bond return followed by Chen et al. (1986). And this variable is mean zero in a risk-neutral world and is directly measured of the degree of risk aversion in pricing. I use the difference between the monthly returns of corporate
23
bonds rated Baa by Moody and the return of 7-year US T-notes as the state variable representing default premium, which is similar to Scholtens & Wang (2008). The historical data of 10-year US T-notes is downloaded from Yahoo Finance. And the monthly returns of corporate bonds rated Baa by Moody is available from FRED economic data. In order to calculate the default premium, I expand the monthly data to the weekly data from 2002 to 2014 by hand. I find that the average return of weekly corporate bonds is about -0.062% and the standard error is about 0.001. And the eleventh row in Table 2 shows that average return of default risk is about -3.2% during the test period and standard deviation is about 0.021.
Moreover, the difference between the average yield to maturity of a 7- year T- note and a 4-week T-bill serves as a state variable capturing the change in Treasury yield curve (Ferson and Harvey, 1991). This variable is used to present term structure. And these two historical data are both collected from Yahoo Finance. The 4-week T-bills is also served as risk free rate. Table 2 shows that the average return of term structure is about 1.8% and the standard deviation is about 0.01. The low standard deviations have showed the stability about the rates of interest risks.
The data for calculating factor return of book-to- market ratio and factor return of earning-to-price ratio are mainly collected from Compustat database. First of all, I intend to calculate the difference between return of high book-to-market ratio companies and return of low book-to- market ratio companies. The market equity is measured by shares outstanding times stock price. These two data is downloaded from CRSP database. And the book equity is calculated by the sum of stockholders equity,
preferred stock and deferred tax. Book-to- market ratio is the book value of common equity plus balance-sheet deferred taxes for fiscal year t - 1, over market equity for December of year t - 1. Table 2 shows that the average factor return of book-to-market factor is around -2.7%, which is not accordance with the existing literature that high book-to-market firms are more likely to have higher return than low book-to-market firms. Secondly, the calculation of earnings is relatively complicated. I calculate the sum of deferred tax and income before extraordinary items at first. After that, I get the earnings by using this result to minus subsidiary preferred tax. And then, I divide earning to stock price per share to get the ratio. And Table 2 also shows that the average return of earning-to-price factor is about 4.5%, which also proves the existing studies that low earning-to-price ratio firms tend to generate more return than high earning-to-price ratio firms.
In the robustness test, I have introduced the factor return of size to the regression model, which is similar to Fama & Macbeth (1973). The last row in Table 2 presents that the average weekly factor return of size is nearly -7.7%. This result shows that the relatively small oil companies have higher stock return that the relatively large oil companies over the sample period. But most of oil companies are very large
companies compared to other companies in other industries. In addition, the coefficient of size factor in Table 12 is also insignificant.
3.2 Methodology
One of the main purposes of my thesis is to investigate fundamental factors in stock returns of oil companies in the United States. It seems more appropriate to build a
25
multi- factor APT model to list the sufficient fundamental factors which may explain the stock returns of oil companies. Then I intend to use the two stage Fama-Macbeth (1973) regression model to estimate the premium rewarded to a particular risk factor exposure for listed oil companies by the market. In the first step, the stock return of oil companies is regressed against four fundamental factors (crude oil price, market return, default risk and interest risk) time series to determine the factor exposure, which is similar to Scholtens and Wang (2008). In the first step, the factor exposure betas are obtained by calculating all the regressions, each company on four factors. After calculating the betas, I can learn about what extent each asset’s or portfolio’s return is affected by each factor. If the regression models show the significant results, I can distinguish whether this fundamental factor have positive or negative effect to the dependent variable.
= + + + + + (1)
Where is the return of the different kinds of ith oil companies for time t. The return is based on the change in the closing price from the last trading day of the week to the next trading day minus the risk free rate. is the excess return. is the return of crude oil price for time t. is the return of the market index for time. is the risk premia for time t, is the premium of changes in term structure.
is the error term. The betas of market return, risk premia and term structure all
represent the company’s sensitivity to corresponding factors. And the beta of crude oil price shows the firm’s sensitivity will lead to a 1% change to crude oil price if other variables are constant.
The second regression model shows the firm’s sensitivity which takes both the systematic risks and unsystematic risks into account. Fema and French (1972) has showed that the firms with high book-to- market ratio are more likely to earn less than the firms with low book-to-market ratio and that small firms are more likely to earn less than large firms. Similarly, Basu (1983) shows that earnings-price ratios (E/P) help explain the cross-section of average returns on U.S. stocks. So I introduce the factor of earnings-to-price ratio, which tends to show the firms with lower
earning-to-price ratio tend to earn more than the firms with higher earnings-to-price ratio. These unsystematic factors are calculated by the difference between the return on the two kinds of portfolios.
= + + + + + + + (2)
The second step for this methodology is to compute cross-sectional regressions of the returns on the estimates of the betas calculated from the first step to calculate the factor premiums for each factor. The regression models are listed as follows:
= + + + + (3)
= + + + + + + (4)
Where describes the excess return of oil companies, which has the same value to the first step. And other independent variables show the sensitivities of fundamental factors.
The main hypothesis for this thesis is that the investors need a higher stock return when they are faced with higher risks. I can observe the regression results to see whether they are not significantly equal to zero in step 2. If the regression model in
27
step 2 shows the significant result, it shows the investors need a higher stock return when they are faced with higher risks.
But Ahn and Gadarowski (1999) point out that the disadvantages of Fema-Macbeth(1973) two-step cross-sectional regressions are that they may lead to autocorrelation and heteroskedasticity problems.
4.
Results
First of all, I research the level of correlation between the stock return of oil companies to other fundamental factors. Table 3 has presented the summary of correlation results. During the whole test period from 2002-1-1 to 2014-12-31, the correlation between stock return of oil companies to oil price is 0.2544. The highest correlation is 0.4414, which comes from the relationship between stock return to market portfolio return. In addition, the correlations between stock return to risk premia and term structure are 0.132 and 0.16, respectively. However, the correlation between stock return to factor return of book-to-market ratio is negative, which equals to about -0.03. Scholtens & Wang also have found the negative relationship between stock prices to factor return of book-to- market ratio.
Table 4 has presented the averaged time series estimates of the regression models presented in the first equation listed in the methodology. The main sample includes 62 listed oil companies in US stock markets from 2002-01-01 to 2014-12-31.
From the results in Table 4, I find that the average estimates of stock returns of oil companies are positively related to the market return, crude oil price return and negatively related to the default premium and term structure for most oil companies.
The results in Table 4 have presented that the average coefficients of crude oil price and market return are both positive and significant. It means the stock returns of oil companies are sensitive to the oil price and market return. This result is in accordance to Rajgopal (2002) who state oil price risk for oil firms can cause earnings volatility. I find the average estimated coefficient for spot oil price in Table 4 is 0.25. And this coefficient is similar to Scholtens & Wang (2008), whose results are about 0.24. However, there also exist five oil companies whose coefficients of oil price return are
29
negative and insignificant. It means not all of the oil companies are sensitive to the changes of crude oil price. Moreover, I replace the oil spot price to oil future price in the test regression model. I find the coefficient for one month oil future price is 0.28. Similarly, coefficient for three months oil future price is 0.33, and coefficient for six month is 0.37. Finally, coefficient for nine months oil future price is 0.40. I can find that coefficients for crude oil price become bigger when the crude oil future contracts become longer. But the R-squared value of spot price is the 0.373, which is the highest in all of the samples. The R-squared value of 1 month oil future price is 0.3675. And that of 3 months oil future price is 0.3705. Both of the R-squared value of 6 months oil future price and 9 months oil future are around 0.3705. This means the application of oil spot price can explain more changes and do better to the movement of oil stock returns. This result may not support Sadorsky (2001), which state that the futures oil price is superior to spot price because spot prices are affected by short-run price fluctuations due to temporary shortages or surpluses. In addition, the
coefficients for different oil price are relatively similar. This condition is reasonable because oil future price is calculated by oil spot price. And most international oil enterprises intend to avoid the risks of the changes of oil prices in the long run. So these firms not only invest in spot market but future market as well. These firms tend to keep the stability of price to avoid the oil price risks.
In addition to the effects of oil price, it is clear that effect of the market return plays the most significant role in the stock return of oil companies. The estimated
coefficients of other fundamental factors. This result also proves some conclusions in the existing studies. For instance, Kavussanos & Marcoulis (1997) find that market return plays the most significant role in the share price of refining firms. And Ferson and Harvey (1991) mention that market return is one of the most important
determinants for the return of American petroleum shares. Moreover, the average estimate of markets betas is higher than 1, which is different from the previous studies. Because Sadorsky (2001), Scholtens and Wang (2008), and Boner and Filion (2007) all find the coefficient of market return smaller than 1. It means if market return change one unit, the return for oil companies will change 1.1 units, after controlling other independent variables. So it means when the market return decrease, it will lead to a stronger shock to the oil stocks. In other words, oil stocks are more risky than the stocks in the whole market, especially when the market is recession during the testing period. Besides, during the test period, the crude oil price has experienced a dramatic change from $ 35 per barrel to $145.31 per barrel, and volatility of oil price is much higher than any other time in history. So the return for oil companies is relatively instability during this period. So it is not surprising that the beta of market return is higher than 1.
The coefficients of risk premia and term structure are both negative and statistically insignificant in Table 4, respectively. This result presents that the oil companies are not sensitive to default risk and interest risk. The measurement of default risk premia is similar to Chen et al. (1986), but they find a positive sensitivity to the default premium. In addition, term structure reflects expectations of market participants about
31
future changes in interest rates. But the coefficients of term structure in the models are nearly -0.14 and insignificant, which are opposite to the Scholtens & Wang(2008). But this result is accordance to Boyer and Fillion (2007), which also find that the sensitivity of interest risks is negatively related to the stock return of oil companies in Canada. Then oil industry has its specificity because oil projects always need huge financial support. So the loans account for a large part for financial support to oil companies. The higher interest rates may increase the cost for the oil companies. So it is reasonable to understand the negative relationship between stocks return and interest risks.
Similarly, the regression results in Table 5 show the average estimates of stock returns of oil companies are positively related to the market return, crude oil price return, factor return of book-to- market ratio, factor return of earnings-to-price ratio and negatively related to the risk premia and term structure for most oil companies.
33
The results in Table 5 are similar to the results in Table 4 except the two new introduced variables. The coefficient for the oil price is about 0.258. Both the coefficients for oil price in Table 4 and Table 5 are positive and significant. These results are enough to answer the first hypothesis. The oil price is sensitive to the change of stock return for oil companies. And the oil price is negatively exposed to the oil companies through the regression results. Besides, the positive relationship between stock return of oil firms and factor return of book-to- market is accordance to the previous studies. The average coefficient is about 0.084, but it is not significant. Additionally, the average factor return of earnings-to-price is also positive and insignificant. The average coefficient of earning- to-price ratio is about 0.0435. These two results represent that the return of oil firms are not highly affected by two factors. Moreover, the value of R-squared in Table 1 is about 37%, and the value of R-squared in Table 2 is nearly 37.5%. Most of the value of R-squared in Table 2 is a little higher than that in Table 1. But the difference is negligible. This means the introduct ion of these two variables based on Fema and French(1972) is not very helpful to explain the return of oil companies. However, Scholtens & Wang (2008) have found that stock returns of oil companies are sensitive to these two introduced factors
35
The results in Table 6 have represented the risk premium of stock return for oil companies. The results in Table 3 and Table 4 have showed that the firms’ return is sensitive to crude oil price and market return. But when I apply the Fama & Macbeth (1973) two step regression and compute cross-sectional regressions of the returns on the estimates of the betas calculated from the first step to calculate the factor
premiums. I have found that none of the new coefficients are significant. The result of T-test for crude oil price is about -0.86 and that for market portfolio return is about -0.9. Both of them are insignificant at 1%, 5% and 10% level of confidence. This result can answer the second hypothesis in this thesis. The oil risk is not a priced factor for oil companies. And these results also mean that the investors don’t need to demand a higher return for oil companies when they are faced with a higher oil price risk. Similarly, Scholtens and Wang (2008) also found no risk premium in the model which only included the macroeconomic factors.
37
Moreover, the results in Table 7 are also insignificant. The result of T-test for crude oil price is about -0.35 and that for market portfolio return is about -0.04. Owing to the test of Fema and Macbeth (1973) is a two-sided test, the results are both
insignificant. This means even if I include the return factor of book-to- market and earning-to-price, oil price risk is also not a priced factor to oil companies. And the investors also don’t need a higher return for investing in oil stocks with higher oil risks. This result is in accordance to the results in Table 6. However, Scholtens & Wang (2008) have found a significant result to the crude oil price. The result of T-test is about -2.59 in their model.
Though the regression result in Table 6 and Table 7 is not satisfied with my conjecture, this result can be explained by several reasons. First of all, in the history of oil price boom, it is always closely related to the battle in East-Asia such as Fourth Middle East war in 1973, Iranian Revolution in 1978 and Gulf War in 1990. During my test period, there happened the well-known Iraq war. But Iraq is not one of the main oil suppliers compared to Iran and Sauti Arabia. Though Iraq is the fourth largest country whose oil reserves accounts for nearly 12% of the oil reserves in the world. But the production data from OPEC shows that Iraq is not one of the ten largest oil producer. And the OPEC applies quota system, so this shock to oil supply can be easily replaced by other countries. The influence of Iraq battle to oil risk is temporary and limited, which fail to lead to a huge change to return of oil firms. Secondly, after the oil crisis happened in 1973, the basic target for organization of petroleum
and to avoid other energy from replacing petroleum. Moreover, the quota system used by OPEC is beneficial to control the changes of oil price in a certain degree. Because the worldwide production and supply of oil is controlled by OPEC and the demand for the oil is always huge and stable. So the oil price is relatively stable in a range. For the oil firms, their main income comes from selling oil and other oil products. If the sale price of oil is stable in a range, their future return is considerable. In summary, it is understandable that investors don’t need a higher stock return with the higher oil price risks.
5. Robustness checks
5.1 Before financial crisis versus after financial crisis
Table 8 and Table 9 have showed the two sub-periods of results to research whether the financial crisis has influence to the risk premium of oil price. Compared to the first two rows in Table 5, I find that the coefficients of oil price become smaller, which decrease from 0.28 to 0.24. It seem that the change of oil price play a less significant role to the return of oil companies during the financial crisis. This result is satisfied with my expectation because the influence of financial crisis may change the return of oil firms in a certain degree. Moreover, the results in Table 6 in step 2 are both insignificant, which is similar to the main tests. This means that oil risk premiums are not closely related to both of two periods. In other words, the influences of financial crisis do not increase investors’ risk premium to the oil price. Oil price is also not a priced factor for oil companies
5.2 A period of Iraq War versus a period of peace
The third and fourth rows in Table 8 and Table 9 classify the test periods by the effect of the Iraq War. I find that the coefficients of crude oil price have become bigger, which have changed from 0.25 to 0.36. This result shows the oil firm’s return become more sensitive to the oil price. But the results in Table 6 also show that the investors do not need a higher return with higher oil risks in the test period. It means that the influence of Iraq war do not increase investors’ risk premium to the oil price.
5.3 The different functions of oil companies
Different oil companies may have different sensitivity to the oil price because of their special functions. I classify all of the oil firms into three categories, and they are drilling oil companies, equipment&service oil companies and refine&market oil companies. This classification is followed by the industry of Yahoo Finance. From the results in the Table 10, I find out that the coefficients of crude oil price for drilling companies are the biggest among the three kinds of oil companies. The coefficients
43
are about 0.33, which is both positive and significant. This result is considerable because most of the incomes for drilling companies come from selling oil. So the change of oil price plays more significant role in the return of oil companies. Moreover, the coefficients for refine&market oil firms are about 0.11. Because refine&market oil firms are regarded as midstream or downstream oil firms. Their main incomes come from the price difference between raw material and oil refined products. So the results of coefficients are less than the drilling companies. From the results in Table 11, I find that all the coefficients for three kinds of companies are insignificant, which means that investors do not need higher return for a specific oil industry with high oil price risks.
5.4 The introduction of size factor
Fama & Macbeth (1973) have employed the factor return of book-to- market ratio and size factor to research the firm-specific risks. But most of oil companies are large companies because of the character of oil industry, so I intend to introduce the size factor in the robustness test. I get the size factor by calculating the difference between return of market equity companies and return of low market equity companies.
45
Table 12 has presented that the introduction of size factor will not affect the positive and significant relationship between stock return of oil companies and the changes of crude oil spot price. The coefficient for oil price is about 0.25. And the coefficient for size factor is around 0.0024, which is insignificant. Besides, the results in Table 13 also show that the coefficient is also insignificant over the sample period. This means that oil price is also not a priced factor for oil companies with the introduction of size factor.
6. Conclusions
In this thesis, I try to find out oil price exposure in stock returns of oil companies in United States and assess the risk premium of oil price. I apply the APT model and Fama & Macbeth (1973) two-step regression model to research this topic. My
samples consist of all the oil companies listed on New York Stock Exchange (NYSE). The sample period is from 01-01-2002 to 31-12-2014. There are 666 weeks in all
during the testing period. And there are 62 oil companies in my test sample. There are 28 oil companies engaged in equipment and services. And there are 20 oil companies engaged in drilling and exploration. The main business for other 14 oil companies are refining and marketing services.
Through the regression model, I have found that the stock return of oil companies is significantly and positively related to the changes of crude oil prices and market portfolio return, which is in accordance to Scholtens &Wang (2008), Sardosky (2001) and Boyer & Filion (2004). But the returns of oil companies are not sensitive to interest risk, default risk, book-to-market ratio and earning-to-price ratio. This result is different from Scholtens & Wang (2008), which find that the returns of oil firms are negatively and significantly related to book-to-market ratio. But my results point out that the factor return of book-to- market fails to explain the stock return for oil companies in a longer period. Moreover, I have found that the results in Table 6 and Table 7 represent an insignificant oil risk premium under all of the regression models, which means that investors do not need higher returns with a higher risk of oil price. This result is also different from the existing studies. Because Scholtens & Wang (2008) have pointed out a significant and positive oil risk premium over the period of 2002 to 2005 under the multi- factor APT model. In addition, when I replace the crude oil future price to crude oil spot price, I have found that coefficients for crude oil price become bigger when the crude oil future contracts become longer. But the result of spot oil price shows the highest value of R-squared and it means that the spot price does better than oil future price in explaining the change of stock return for oil
47
companies. This result may not support Sadorsky (2001), which state that the futures oil price is superior to spot price because spot prices are affected by short-run price fluctuations due to temporary shortages or surpluses.
In the robustness test, I have found that the coefficients of crude oil price for drilling companies are the biggest among the three kinds of oil companies, which is classified by their main business. This finding means that the return of drilling oil companies is relatively closely related to the changes of oil price. In addition, I classify the test period into two periods according to financial crisis. The results in Table 5 and Table 6 show that change of oil price plays a less significant role to the return of oil companies during the financial crisis. Moreover, I also check the influence of Iraq war, and I have noticed that the oil firm’s return become more sensitive to the oil price after the Iraq war. But all of the robustness tests have represented an insignificant oil risk premium. Besides, I introduce the size factor which is similar to Fama & Macbeth (1973) to research the risk premium of oil price. The results in Table 13 also prove that oil price is also not a priced factor for oil companies.
This thesis also has its limitations. First of all, this thesis mainly focuses on the changes of crude oil price. But the main factors which affect the crude oil price are the relationship between oil supply and demand. In the recent forty years, OPEC has applied the quota system, which is beneficial to ensure stable income for oil exporting countries and to avoid other energy from replacing petroleum. So the supply of oil is controlled by the OPEC. The oil price will not likely to have a dramatically change,
especially when the countries tend to have a peaceful development. So risk of oil price is controllable. Second, there exist other fundamental factors which can affect the return of oil companies. Because the values of R-squared are about 38% in most of the regression models. I can’t reject the possibility to introduce other unknown variables, which can significantly beneficial to explain the stock return of oil firms.
49
References
Al-Mudhaf, A., & Goodwin, T. H. (1993). Oil shocks and oil stocks: evidence from the 1970s.Applied Economics, 25(3), 181-190.
Bower, D. H., Bower, R. S., & Logue, D. E. (1984). Arbitrage Pricing Theory and Utility Stock Returns. Journal of Finance, 39(4), 1041-1054.
Boyer, M. M., & Filion, D. (2004). Common and Fundamental Factors in Stock Returns of Canadian Oil and Gas Companies.
Chan, K. C., & Chen, N. (1988). An Unconditional Asset-Pricing Test and the Role of Firm Size as an Instrumental Variable for Risk. Journal of Finance, 43(1), 309-325.
Chen, N., Roll, R., & Ross, S. A. (1986). Economic Forces and the Stock Market. Journal of Business, 59(3), 383-403.
Elwood, S. K. (2001). Oil-Price Shocks: Beyond Standard Aggregate
Demand/Aggregate Supply Analysis. Journal of Economic Education, 32, 381-386. Faff, R. W., & Brailsford, T. J. (1999). Oil price risk and the Australian stock market. Journal of Energy Finance & Development, 4(1), 69-87.
FAMA, E. F., & FRENCH, K. R. (1992). The Cross-Section of Expected Stock Returns.American Finance Association, 47(2), 427-465.
Fama, E. F., & MacBeth, J. D. (1973). Risk, Return, and Equilibrium: Empirical Tests. Journal of Political Economy, 81(3), 607-636.
Ferderer, J. P. (1996). Oil price volatility and the macroeconomy. Journal of Macroeconomics, 18(1), 1-26.
Ferson, W. E., & Harvey, C. R. (1991). The Variation of Economic Risk Premiums. Journal of Political Economy, 17(3), 245-262.
Jones, C. M., & Kaul, G. (1996). Oil and the Stock Markets. American Finance Association, 51(2), 463-491.
Koutoulas, G., & Kryzanowski, L. (1996). Macrofactor Conditional Volatilities, Time-Varying Risk Premia and Stock Return Behavior. The Financial Review, 31(1), 169-195.
Maghyereh, A. (2004). Oil Price Shocks and Emerging Stock Markets: A Generalized VAR Approach.
Pincus, M., & Rajgopal, S. (2002). The Interaction between Accrual Management and Hedging: Evidence from Oil and Gas Firms. Accounting Review, 77(1), 127-160. Sadorsky, P. (1999). Oil price shocks and stock market activity. Energy
Economics, 21(5), 449-469.
Sadorsky, P. (2001). Risk factors in stock returns of Canadian oil and gas companies. Energy Economics, 23(2), 17-28.
Sadorsky, P., & Henriques, I. (2001). Multifactor risk and the stock returns of Canadian paper and forest products companies. Forest Policy and Economics, 3(2), 199-205.
Scholtens, B., & Wang, L. (2008). Oil Risk in Oil Stocks. The Energy Journal, 29(1), 89-111.
Smith, R. T., Bradley, M., & Jarrell, G. (1986). Studying Firm-Specific Effects of Regulation with Stock Market Data: An Application to Oil Price Regulation. Rand Journal of Economics, 17(4), 467-479.
