Bachelor thesis
—
Oil risk exposure in stock returns of oil and
gas companies in Canada
University of Amsterdam, Amsterdam Business School
BAs Economics and Business, Finance and organization track
Name: Fei Wang
Student number: 10828591
Supervisor: Drs. Pepijn Trietsch
2
Statement of Originality
This document is written by Student Fei Wang, 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.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This thesis researches exposure of oil and gas companies’ stock returns to various risk factors. The relationship among stock returns of oil and gas firms with crude oil price, market return, exchange rate and interest rate in Canadian market is researched by employing the multi-factor arbitrage pricing model. The hypotheses are whether the above four mentioned factors are individually and jointly significant. I find that stock returns of oil and gas companies are positively related to oil prices and market return from 2005 to 2011. However, the oil and gas firms’ stock returns are negatively related to exchange rate and interest rate over the same period. By comparing the regression results of spot oil price and future oil price, it is clear that future oil price has a higher R-squared value than that of spot oil price. The p-values indicate that both types of oil price and market return are individually significant, while the exchange rate and interest rate are not significant. However, the F-statistic test shows that all these macroeconomic variables are jointly significant in explaining the changes of stock returns. Moreover, by separating the time period to before, during and after financial crisis, it is reasonable to suppose that financial crisis is a potential variable that would affect stock returns. This conjecture is confirmed by including the financial crisis as a dummy variable.
4
Table of Contents
1. Introduction... 5
2. Literature Review... 7
2.1 oil price...7
2.2 market return...8
2.3 exchange rate...8
2.4 interest rate...9
3. Methodology and data ...13
4. Results...17
6. Conclusion... 26
1. Introduction
The changes of oil price have fluctuated a lot from late of last century to recent years. In 1973, the oil price was $3.011 per barrel before the fourth Middle East war, increasing to $10.651 per barrel after the war. In last few years, oil price reached its peak at $143 per barrel in July of 2008. However, the price decreased to $115 in one month later and ended with only $45 per barrel at the end of the year.
Since oil price sensitivity is often regarded as an important factor that affects the world economy, a large amount of studies focus on the effects of oil price changes and other risk factors on oil and gas companies’ stock returns. The most well-know researchers in this field are Huang et al. (1996), Jones and Kaul (1996), Sadorsky (1999) and etc.
Sadorsky (2002) claims that “ the idea that macroeconomic variables can help to explain excess returns in equity and bond markets has recently been extended to commodity futures markets”. Specifically, most oil and gas companies’ stock value are driven by commodity prices. Looking through the literature, it is generally believe that oil price, market return, exchange rate and interest rate are fundamental determinants in excess stock returns. However, there are still opponents who find different results. Therefore, it is worthwhile to study the effects of macroeconomic variables on companies in oil and gas industry.
As one of the top 5 largest oil producing countries in the world, Canada (after the Russia, Saudi Arabia, United States and China) regards energy sector as an important sector, in the sense that 4.54% of the world’s oil output is obtained by Canada. Specifically, the energy sector is the fifth largest sector after the service-producing sector, goods-producing sector, real estate and rental and leasing sector and manufacturing sector. According to Statistics Canada (2009), the energy sector contributed approximately 7.98% to the Canada’s gross domestic product (GDP). Moreover, the importance of energy sector to Canada is not only evident in GDP, but also evident in stock market capitalization. As The Globe and Mail (2009) reports, the energy sector accounts for almost 10.54% of
6
stock market capitalization in Canada. Thus, it is also worthwhile to study the effects of oil risk exposure in Canadian market.
There are two main methodologies in existing studies to research variations of oil and gas firms’ stock returns in light of macroeconomic factors. These two main methods are the vector error correction (VECM) model and the multi-factor arbitrage pricing theory (APT) model. Since the hypotheses of this paper are whether oil risk determinants (crude oil price, market return, exchange rate and interest rate) are individually and jointly significant in Canadian oil and gas market, I apply the multi-factor APT model.
The test period is from 01-01-2005 to 31-12-2011, because oil price changed obviously during the financial crisis and I extent two years before and after the crisis to see whether crisis has an influence on stock returns. There are 84 months within this period. I select oil and gas companies, which have been listed on the Toronto Stock Exchange (TSE) market with unbroken series of historical prices. After deleting private, subsidiary companies and also companies have been merged or acquisitioned, there are 29 companies meet these requirements in the end.
Beyond to previous studies, the contribution of this paper is that I am not only compare the differences between the spot oil price and future oil price but also include the financial crisis as a dummy variable in the regression model. Moreover, the sample period is 2005-2011, which is more contemporary than previous studies focusing on Canadian oil and gas market.
Overall, this paper is organized as follows. Chapter 2 is a literature review about some macroeconomic factors that have an impact on stock returns of oil and gas companies and methodology used by researchers. Chapter 3 presents the regression model and data used in the test period. Then, chapter 4 shows the regression results and their interpretations. The thesis ends with a conclusion in chapter 5.
2. Literature Review
question of determining the accurate variables. Some authors, such as Koutoulas and Kryzanowski (1994) state that as many factors as possible should be included in the model to explain the entire stock return, but the fact is that only few stocks related to the same determinants. However, Kryzanowski and To (1983) and Abeysekera and Mahajan (1987) claim that four or five factors are sufficient to exhibit a credible explanatory power of the model. In sub-chapters, some macroeconomic variables are listed and are provided by reasons why it should or should not be included in the regression model.
2.1 oil price
Oil price sensitivity has always been regarded as a dominant variable in explaining the stock returns of oil and gas firms. Many researchers find there is a positive relationship between these two variables. For example, Sadorsky (2001) regards Toronto Stock Exchange (TSE) oil and gas index as a proxy of average oil and gas firms’ stock returns and includes oil price in the model to see whether it has an influence on stock returns. He finds that there is a positive relationship between them. Besides, Boyer and Filion (2007) follow the study of Sadorsky over the period 1995-2002. What they find is that oil and gas firms’ stock returns exhibit a strong and positive relationship with oil price. Scholtens and Wang (2008) investigate several macroeconomic variables’ behaviors, including oil price on stock returns in the firm level from 2002 to 2005. Their results also show that a higher oil price results in a higher stock price. In addition, Hammoudeh et al. (2004) and Cong et al. (2008) all point out that an increase in the oil stock is accompanied by an increase in oil price.
However, several authors find totally opposite results. For instance, Huang et al. (1996) points out that there is a negative relationship between oil price and stock returns of oil and gas firms. Wei (2003) also concludes with an increase in oil price fails to explain the changes in the stock returns. Although lots of literature exhibit that a positive relation between oil price and oil and gas companies’ stock
8
returns, few studies show a different relationship of these two variables. Therefore, it is worthwhile to investigate the relation between oil price and stock returns of oil and gas firms.
2.2 market return
In terms of market returns, a large number of studies prove that it plays an important role in explaining stocks returns. For instance, Ferson and Harvey (1991) analyze predictable variables of monthly bond and stock portfolios and they state that market return is one of the most significant variables for American petroleum shares’ returns. Kavussanos and Marcoulis (1997) deepen the study of Fama and French (1992) to profitability of oil refining companies. They point out that market return is positively related to the stock price of oil refining firms. Moreover, Sadorsky (2001) also finds that there is a positive relationship between market return and stock returns. Boyer and Filion (2007) point out that an increase in the market return leads to an increase in stock returns in their paper. In addition, a positive and significant relation has also been found by Scholtens and Wang (2008). Thus, given a lot of literature proving market return is another important factor, it should also be include in the model as well.
2.3 exchange rate
From the past few decades, a lot of attention has been given to the relation between exchange rate and stock returns in the literature. There are a couple of reasons why it is necessary to predict the relation between them. For individual investors, it allows them to assess their investment portfolios. For oil importing companies, oil price fluctuation will affect their foreign asset positions and trade balances. Thus, academic researchers study the relationship between these two variables.
Aggarwal (1981) investigates the relation between exchange rate and stock returns and he finds that these two variables are positively correlated. Besides,
Solnik (1987) also studies the relationship between stock returns and exchange rate, but in eight industrial countries. He believes that a positive but weak relation exists between them. However, others find different conclusions. Sadorsky (2001) claims that there is a negative relationship between exchange rate and stock returns. Moreover, Boyer and Filion (2007) point out that a depreciation of Canada dollars would result in an increase in stock returns. Given these contrasting arguments, exchange rate should also be included in the model.
2.4 interest rate
Since the oil and gas industry requires huge amount of capital investment to update the technology base, it is a very capital intensive industry. On the other hand, interest rate fluctuates with the movement of the monetary policy. Consequently, oil and gas industry is extremely sensitive to the interest rate.
The relation between stock returns and interest rate has also been examined by academic researchers. Sadorsky (2001) measures the short-term interest rate by annual premium between 1-month Treasury Bill and 3-month Treasury Bill. He concludes with interest rate is negatively correlated to oil and gas companies’ stock returns. Besides, the results of study of Boyer and Filion (2007) also reveal a negative but significant relation between oil and gas firms’ stock returns and interest rate. Abugri (2008) examines the relation between these two variables in four countries. He finds out that the response of stock returns on interest rate is negative and statistically significant. Moreover, Adam and Tweneboah (2008) investigate the relation between them in Ghana market and they claim that a negative relationship exists between interest rate and stock returns. Through what has been discussed above, interest rate then is included as an independent variable in the regression model.
Since Sadorsky (2001) and Boyer and Filion (2007) also pay their attention on the Canadian oil and gas market, their studies act as bases of this paper. Sadorsky (2001) includes four variables, which are future oil price, market return,
10
exchange rate and also short-term interest rate in the multi-factor APT model to explain the effects on oil and gas companies’ stock returns. He finds out that only the first two variables are significant to stock returns and have much larger power in explaining stock returns than the latter two factors. To be more specific, he observes that crude oil price and market return are positively correlated to stock returns whereas an increase in interest rate and a decrease in exchange rate are negatively correlated with oil and gas firms’ stock returns. Boyer and Filion (2007) include one more variable, which is natural gas price in their model as the fifth explanatory variable for oil and gas companies’ stock returns. Their regression results show that coefficients of market return, crude oil price and natural gas price are positive, while the estimated betas of exchange rate and interest rate are negative. However, natural gas price has been proven as a weak factor in the t-statistic test.
In summary, my regression model includes crude oil price, market return, exchange rate and interest rate as explanatory variables. Sadorsky (2001) argues that the NYMEX 1 month future oil price should be used in the model, because spot oil price is more easily affected by short-term price fluctuations due to temporary shortages or surpluses. So I intend to compare the differences between spot oil price and future oil price. In other words, oil and gas companies have two version contracts—spot oil contract and future oil contract. I believe that it would be more comprehensive to take both kinds of oil contracts into consideration.
The two main methodologies in researching the relationship between stock returns and various explanatory factors are the vector error correction (VECM) model and the multi-factor arbitrage pricing theory (APT) model.
The vector error correction model focuses on the long-term factors of each firms’ market value (Hammoudeh et al, 2004; Lanza et al, 2005). Including some error-correction terms to present deviations from long-run equilibrium, the VECM model can decrease the probability of misspecification (Hammoudeh and Li, 2005). Therefore, if there is a good integration between stock market and world market and both these markets’ returns are included as explanatory
variables in the model, then employing the VECM model would be a good method to research the relations between various independent factors and stock returns. There are some studies employ this method. For example, Maysami and Koh (2000) use the VECM method to research the macroeconomic factors’ effects on the stock returns in Singapore market.
The multi-factor arbitrage pricing theory (APT) model is another common methodology in researching this topic. The model regards oil price risk as a fundamental variable in addition to other macroeconomic variables, such as term premium and default premium. The assumptions behind this model are that there is no opportunity to arbitrage and all relevant factors, which will affect stock returns, are included in the model. Many researchers use this method to explore oil risk exposure in stock returns. For example, Al-Mudhaf & Goodwin (1993) employ the multi-factor APT model to study the differences of stock returns in 29 American oil companies in 1973. Scholtens & Wang (2008) also use the multi-factor APT model to evaluate oil price sensitivities on oil and gas firms’ stock returns.
As the purpose of this paper is to determinate some fundamental factors and investigate oil risk exposure on oil and gas companies’ stock returns, it is more appropriate to apply the multi-factor APT model. Moreover, Brooks (2002) claims that employing the VECM model needs a rich data structure. However, I am not sure about whether my sample data is sufficient enough to capture all the dynamic properties.
12
T
ab
le
1:
l
ist
in
g of
p
re
vi
ou
s s
tu
d
ie
s
A
ut
hor
(s
)
res
ea
rch pe
ri
od
num
ber
of
obs
er
vat
ions
m
et
hodol
ogy
m
ai
n r
es
ul
ts
F af f & Bra il sford 1983 -1999 24 indus tr ie s m u lt i-fa ct or A P T m ode l pos it iv e re la ti on : o il pri ce S adors ky 198 3 -1999 1 inde x m u lt i-fa ct or A P T m ode l pos it ive r el a ti on: oi l pri c e, m ark et re turn; n ega ti v e re la ti o n: i nt er es t r at e, exc ha nge r at e A le s-sa ndro e t a l. 1998 -2003 6 com pa n ie s v ec tor error c orre c ti on pos it iv e re la ti on: s pot oi l pri c e, m od el fut ur e oi l pr ic e, m ark et re turn ; n ega ti ve re la ti on : e xc h ange ra te Boye r & 1995 -2002 25 com pa n ie s m ul ti -fa c tor A P T m ode l pos it ive r el at ion: c rud e oi l pr ic e, F il ion na tura l ga s pri c e, m ar ke t re turn ; n ega ti ve re la ti on : i nt er es t ra te , exc ha nge r at e H am m oude h 1995 -2001 5 indi ce s ve ct or error corre ct ion pos it iv e re la ti on: s pot oi l pri c e, et a l. m ode l fut ure o il pri ce S adors kyauthor, research period, number of observations, methodologies and main results.
3. Methodology and data
The multi-factor APT model is used by many researchers (e.g. Kryzanowski and To (1983); Chen et al., 1986; Abeysekera and Mahajan (1987); Koutoulas and Kryzanowski (1994); Sadorsky 2001; Boyer and Filion 2007) and could be written as below:
Ri = α+β1*WTI+β2*MKT+β3*EXC+β4*INT +ε.
Where Ri is the monthly excess return of ith oil and gas companies. The excess return is based on the return on the closing price of last trading day of the month to the next trading day of the month minus the 1-month Canadian Treasury Bill rate. WTI is the monthly growth rate of spot oil price and MKT is the monthly excess return on market index, which is return to market index minus the 1 month Canadian Treasury Bill rate. Moreover, EXC represents monthly growth rate of exchange rate (Canada-Us). Besides, INT means interest rate premium, which is the difference between annual yield of 3-month Canadian Treasury Bill and annual yield of 1-month Treasury Bill (Harvey, 1989; Sadorsky, 2001). The ε means the error term.
My target market is Canada, so I include all oil and gas companies listed in Toronto Stock Exchange (TSE). My test period is from 1-1-2005 to 31-12-2011, so there are 84 months in total. I select companies with fundamental information and unbroken historical prices during this period. By filtering those companies which are private, subsidiary, foreign companies as well as companies being merged, acquired, there are 29 oil and gas companies in the end and the total observations are 2436.
All these companies’ historical monthly prices are collected from Quandl. The source of West Texas Index spot oil price and NYMEX 1 month future oil price are all available on Quandl. The historical exchange rate is collected from Federal Reserve Statistical Database and the information of historical interest rate
14
is available on Bank of Canada Statistical Database.
Table 2: summarized descriptive statistics
Variable Mean Median Std. Dev. Min Max
Average stock returns -0.0006 0.0018 0.0826 -0.2443 0.2007 Spot oil price 0.0102 0.0149 0.0813 -0.2761 0.2442 Future oil price -0.0013 0.0029 0.0583 -0.1765 0.1698 Market return -0.0165 0.1004 0.0294 -0.1015 0.0549 Exchange rate -0.0019 -0.0035 0.0229 -0.0583 0.1195 Interest rate 0.0009 0.0007 0.0009 -0.0013 0.0031
Table 2 represents a summary of both dependent and independent variables and shows descriptive statistics in five aspects, which are mean, median, standard deviation, maximum and minimum.
The first column of this table lists dependent variable—average stock returns, and independent variables, which are spot oil price, future oil price, market return, exchange rate and information rate. The second to sixth columns present descriptive information and details will be explained in below.
The second row in the table above shows that average monthly excess stock returns of all sample companies during the test period is around -0.0006 and its standard deviation is 0.0826. The median number is 0.0018 and its minimum and maximum numbers are -0.2443 and 0.2007 respectively. I also find out that the excess stock returns reached its highest point in 2009 and its lowest point in 2008 when looking through the whole database.
The third row is about crude oil spot price, which is measured by West Texas Intermediate (WTI). The reason why I choose WTI crude oil spot price is because Hammoudeh, Ewing & Thompson (2008) point out that WTI and Brent are highly liquid, very actively traded and adjust to equilibrium in the long run. However, Sadorsky (2001) includes future oil price in the model, because he believes that spot oil price is more easily affected by short-run price fluctuations due to
temporary surpluses or shortages. Thus, I compare spot oil price and future oil price to be part of my contribution and see which one has a higher power in explaining the changes of stock returns. As Gülen (1999) reports, the largest-volume future contracts trading on future commodities in the world are NYMEX crude oil. Moreover, Fama and French (1987) argue that the nearby future contracts with at least 1 month to maturity on the first trading day of each month should be used in constructing continuous series. So I adopt NYMEX 1 month future oil price when measuring future oil price. Again, both WTI crude spot oil price and NYMEX I month future oil price are collected from Quandl.
Graphs 1 & 2: movements of oil price with time
S pot oi l pri c e F ut ure oi l pri ce
16
Graphs 1 and 2 represent movements of spot oil price and future oil price from 2005 to 2011 respectively.
From the graphs above, it is clear that the trend of spot oil price and future oil price is very similar. Both of them increase until the mid of 2008, but follows with a sharp collapse in 2009, and then increasing again in the rest of the years. The average monthly return of spot oil price is 0.0102 and the standard deviation is 0.0813. Additionally, the monthly return of future oil price is -0.0013 and its standard deviation is 0.0583. Both kinds of oil price reached its peak in 2009 and its bottom in 2008, which was consistent with the extreme points of excess stock returns.
The second dependent variable is market return, which is measured by Toronto Stock Exchange (TSE) 300 index. The historical monthly stock price of TSE 300 is also available on Quandl. The fifth row in the table represents that the average monthly excess market return is around -0.0165 and its standard deviation is 0.0294. The maximum and minimum values are 0.0549 and -0.1015, which are found in 2009 and 2008 respectively.
The third explainable variable is exchange rate, which is measured by monthly growth rate of the Canadian-US exchange rate ($US/$C). However, this format $C/$US will be more relevant to Canadian oil and gas companies. The historical data is also downloaded from Quandl. From the sixth row of the table, the average monthly growth rate is -0.0019 and the standard deviation is 0.0229. However, the maximum number 0.1195 and minimum number -0.05832 are found in 2008 and 2009.
The last independent variable is interest rate, which is calculated by the premium between annual yield on 3-month Treasury bonds and annual yield on 1-month Treasury bond. This calculation method is different from Fama and French (1992). They use rit = [((Rate premium between the yield on 10 years Canadian corporate Bonds and the yield on 10 years Canadian Treasury Bonds) - 90-day commercial paper rate)t / (Rate premium between the yield on 10 years
Canadian corporate Bonds and the yield on 10 years Canadian Government Bonds) - 90-day commercial paper rate) t-1 ] - 1. However, the data of 10 years Canadian corporate Bonds are not available. So I apply the former method to measure interest rate. Both yields of 3-month and 1-month Treasury bonds are collected from Bank of Canada Statistical database. According to the table, its mean is 0.0009 and the standard deviation is around 0.0001. The maximum number 0.0031 and minimum number -0.0013 are found in 2009 and 2008 respectively.
Prior of the regression, it is expected that the crude oil price and market return are positively correlated to stock returns of oil and gas companies, while the exchange rate and the interest rate have a negative effect on oil and gas companies’ stock returns. It is also expected that the first two variables mentioned above are individually significant, while the later two variables are not significant. However, it is expected that all these four variables are jointly significant.
4. Results
Before importing data into STATA, it is necessary to do the residual diagnostic test to check whether the specification is heteroscedasticity or serial correlation. The corresponding solutions are Breusch-Pagan test and
18
Table 3：Results of diagnostic test
Residual diagnostic test (P value as reported)
5% level of significance Spot Future Breush-Pagan test 0.4099 0.6446 Breusch-Godfrey test Lag (1) 0.2372 0.7027 Lag (12) 0.6911 0.9558
The table 3 presents p-values of Breush-Pagan test and Breusch-Godfrey test.
The null hypothesis of Breush-Pagan test is H0 = homoscedaciicy, while the alternative hypothesis is H1 = hetroscedaciticy. According to table 3, the p-values of Breush-Pagan test are 0.4099 and 0.6446 respectively, so they are all larger than α = 0.05. Therefore, we cannot reject the null hypothesis with homoscedaciticy. Breusch-Godfrey test is used to detect whether the model is serial correlation. Lag (1) means a test to check whether there is a serial correlation between time t and t-1, and lag (12) means whether there is a serial correlation between time t and t-1, t and t-2, t and t-3, and so on, until t and t-12. Since all the data I use is monthly data, it is reasonable to suspect there is serial correlation within 12 months. So the null hypothesis is H0 = no serial correlation, while alternative one is H1 = serial correlation. The corresponding p-values for spot oil price are 0.2372 and 0.6911, so they are all larger than α = 0.05 and the null hypothesis should not be rejected. In regarding to the future oil price, p-values for lag (1) and lag (12) are 0.7027 and 0.9558 respectively. Again, the null hypothesis should not be rejected.
Table 4: The correlation matrix
Variable Ri Spot Future MKT EXC INT
Ri 1 0.4784 0.7225 0.4970 -0.3953 -0.0040 Spot 1 0.6722 0.3877 -0.4273 0.0371 Future 1 0.3530 -0.3506 0.0435 MKT 1 -0.4239 0.0982 EXC 1 0.0395 INT 1
Table 4 presents a correlation matrix between dependent and independent variables. Ri means the average stock returns of individual oil and gas companies. Spot and future represent spot oil price and future oil price respectively. MKT means the market return and EXC means the exchange rate. INT represents the interest rate.
The correlation matrix above presents correlation relations between variables. The correlation numbers of spot oil price and future oil price are 0.4784 and 0.7225 respectively, indicating that both spot oil price and future oil price are positively correlated with oil and gas firms’ stock returns. It might suppose that future oil price has a higher explaining power than spot oil price, because it has a higher correlation coefficient with stock returns. Market return is also positively correlated with stock returns, while exchange rate and interest rate represent an opposite results. They are negatively correlated with oil and gas companies’ stock returns, with correlation coefficients -0.3953 and -0.004 respectively. Overall, there is no
20
multicollinearity among these explanatory variables.
Table 5: Regression results with spot oil price
Information dependent variables
variable Ri Ri Ri Ri Intercept -0.0055 0.0130 0.0106 0.0144 (0.0081) (0.0090) (0.0091) (0.0123) Spot 0.4858 0.3414 0.3005 0.3019 (0.0991) (0.0998) (0.1048) (0.1054) MKT 1.0314 0.9200 0.9358 (0.2766) (0.2899) (0.2933) EXC -0.468826 -0.4520 (0.3781315) (0.3817) INT -3.8137 (8.1754) R2 R2 R2 R2 0.2288 0.343 0.3556 0.3574 Table 5 represents regression results with spot oil price for single and multi-factor models. Standard errors are in the parentheses below the parameter estimates. Column 1 lists intercept and all explanatory variables. Row 3 to 6 represent the regression results from spot oil price to interest rate.
Table 6: p-values of regression model
p-values dependent variables α=0.05
(spot oil price) Ri Ri Ri Ri Intercept 0.496 0.153 0.248 0.243 Spot 0.000 0.001 0.005 0.005 Market 0.000 0.002 0.002 Exchange 0.219 0.240 Interest 0.642 F-statistic 20.89 14.53 10.74
p-values dependent variables α=0.05 (future oil price) Ri Ri Ri Ri
Intercept 0.233 0.386 0.530 0.343 Future 0.000 0.000 0.000 0.000 Market 0.001 0.003 0.003 Exchange 0.353 0.393 Interest 0.468 F-statistic 57.29 38.43 28.78
Table 6 presents p-values of regression model with spot oil price and future oil price at the 5% significance level.
From the regression results in the tables above, we see that the coefficients of spot oil price and market return are positive, while the estimated betas of exchange rate and interest rate are negative. This result is the same as conjecture before the regression.
The Coefficient of spot oil price from single factor regression model is around 0.49 and its relative p-value is 0.000. So it is positive and statistically significant at
22
the 5% level of significance. In addition, the R-squared value is 0.2288, which means with only one variable —spot oil price, it can explain 22.88% changes of stock returns on oil and gas companies.
From column 2 of table 5, coefficients of spot oil price and market return are both positive. According to table 6, the p-value of 0.000 shows that market return is also individually significant at 5% level of significance. At the meanwhile, F-statistic value of 20.89 indicates that they are jointly significant. An interesting observation is that the market return has a highest coefficient in all independent variables. This is not surprising given a lot of previous studies in the literature. For instance, Kavussanos and Marcoulis (1997) point out that market return plays the most important role in refining firms’ share price. Moreover, the estimated beta of market return is around 1.03 and only decreases a little when including exchange rate and interest rate in columns 3 and 4. These results indicate that the oil and gas industry is in the similar risk class as the market. It means that when market return changes by 1 unit, the stock returns on oil and gas companies’ will also changes by 1 unit. However, Pring (1991) claims that if oil and gas companies’ stock is a good way of hedging, then its market beta should be negative. This statement is in accordance with the findings of Sadorsky (2001), Boner and Filion (2007) and Scholtens and Wang (2008), they all find a much smaller market beta in their studies. Accordingly, the R-squared value increases to 0.343, indicating that explaining power increases by including the market return in the model.
However, when including exchange rate in the model, the R-squared value only increases a little bit to 0.3556. From table 5, the estimated beta of exchange rate is negative, with around -0.47. The p-value in table 6 is 0.219, revealing that exchange rate is insignificant. While it is jointly significant with crude oil price, market return when looking at the F-statistic test result of 14.53. Sadorsky (2001) also finds a negative beta of exchange rate return, but with a much smaller number of -1.1. The negative coefficient means that the depreciation in Canada dollar must increase costs more than its revenues of firms. Khoo (1994) states that the consequence of depreciation on domestic currency is worsen off firms’ financial
position. For instance, importing resources for Canadian firms will be much more expensive with increasing exchange rate.
According to table 5, it is clear that the estimated beta of interest rate is negative. The corresponding p-value is 0.642, which is larger than α and should not be rejected. However, the F-statistic test result of 10.74 reveals it is necessary to include interest rate in the model. Additionally, Sadorsky (2001) and Boyer and Fillion (2007) also find a negative relationship between interest rate and stock returns of oil and gas companies, but this result is opposite to the finding of Scholtens & Wang (2008). Since oil and gas companies always need to borrow huge amount of money to invest in equipments, higher interest rate increases the costs of borrowing, therefore lowing the stock returns of oil and gas firms. Moreover, Column 5 has the highest R-squared value of 0.3574, presenting that the model has the best explaining power when all four variables are included.
24
Table 7: Regression results with future oil price
Information dependent variables
Variable Ri Ri Ri Ri Intercept -0.0076 0.0061 0.0046 0.0093 (0.0063) (0.0070) (0.0072) (0.0098) Future 0.6915 0.5981 0.5814 0.5831 (0.0735) (0.0733) (0.0755) (0.0758) MKT 0.7777 0.7041 0.7237 (0.2155) (0.2296) (0.2318) EXC -0.2744 -0.2540 (0.2935) (0.2957) INT -4.7288 (6.4817) R2 R2 R2 R2 0.522 0.5889 0.5934 0.5961
Table 7 represents regression results with future oil price for single and multi-factor models. Standard errors are in the parentheses below the parameter estimates. Column 1 lists intercept and all explanatory variables. Row 3 to 6 represent the regression results from future oil price to interest rate.
According to table 7, it is clear that the coefficients of future oil price and market return are positive whereas the estimated beats of exchange rate and interest rate are negative. This finding is the same as the regression results with spot oil price. Moreover, table 6 shows that the p-values of future oil price and market return are around 0.000 and 0.001 respectively, so they are individually significant. While p-values of exchange rate and interest rate are all larger than α, with 0.353
and 0.393 respectively, so they are not significant. As for the F-statistic test, it shows that all four macroeconomic variables are jointly significant. An interesting observation is that the estimated betas of future oil price are larger than that of spot oil price in all columns. Besides, the R-squared values of future oil price are also consistently higher than that of spot oil price. Higher R-squared value means that future oil price does better in the movement with stock returns on oil and gas companies and could explain more changes of the stock returns. This finding supports statement of Sadorsky (2001), who argues that future oil price is more appropriate than spot oil price in explaining the changes of stock returns.
Table 8: regression results before, during and after financial
crisis
Variable before financial crisis during financial crisis after financial crisis 2005-2007 2007-2009 2009-2011 Intercept 0.0288 0.0276 -0.0150 (0.0354) (0.0149) (0.0147) Future 0.5045 0.5590 0.6627 (0.1721) (0.1212) (0.1226) MKT 2.1908 0.7413 0.6838 (0.9610) (0.3085) (0.6295) EXC -0.1194 -0.2033 -0.3099 (0.8726) (0.4004) (0.6863) INT 2.0067 -17.4582 7.78399
(15.0104) (10.9764) (13.0455)
This table separates test period to three sub-periods — before financial crisis, during financial crisis and after financial crisis to see the effects of crisis on each explanatory variable.
26
Since future oil price has been observed with a higher R-squared value, which means it could explain more changes of stock returns, future oil price should be used when running regression in three separate periods. Comparing these three periods, I find that the estimated betas of future oil price become larger, from 0.50 to 0.66, indicating that changes of future oil price plays an important role in stock returns of oil and gas firms. On contrast, coefficients of market returns become smaller over the three periods. Similarly, coefficients of exchange rate decrease slightly during the whole period. Another interesting observation is that premium of interest rate during the crisis is extremely low, but increases sharply after the crisis, which is reasonable since interest rate is very high during the crisis, thus the return would be even lower than the other two periods.
In the case that crisis might have an impact on stock returns, it is reasonable to include it as a dummy variable. So the null hypothesis is βcrisis = 0, and alternative hypothesis is βcrisis ≠0. After adding crisis in the regression model, the t-statistic value is 2.03, so the null hypothesis would be rejected at the 5% significance level. This result indicates that the financial crisis is a significant factor that would influence the stock returns. This finding could also help to explain why observations of the minimum numbers of both types of oil price returns and market returns are all found in 2008 and their maximum returns are all observed in 2009.
5. Conclusion
The purpose of this thesis is to find accurate factors that would have an impact on stock returns on oil and gas companies. The hypotheses are whether they are individually and jointly significant. The model I apply to study the relationship between fundamental factors and oil and gas firms’ stock returns is the multi-factor APT model. The test sample consists of all oil and gas companies listed in the Toronto Stock Exchange with unbroken time series from 1-1-2005 to 31-12-2011. So in general, there are 84 months and 2436 observations in the sample.
the stock returns of oil and gas firms are positively and significantly related to returns of both kinds of oil price and market return, however, it is negatively correlated to exchange rate returns and interest rate premium. This result is consistent with findings of Sardosky (2001) and Boyer & Filion (2007). Besides, the p-values indicate that spot oil price, future oil price and market returns are individually significant, while the p-values of exchange rate and interest rate are all larger than α, so they are not significant individually. As for the F-statistic test, either regression results with spot oil price or future oil price shows all these four variables are jointly significant. In addition, when replacing spot oil price to future oil price, the estimated beta is relatively higher than spot one and the R-squared value is also larger than that of spot oil price. It means future oil price does better in the movement with stock returns on oil and gas companies and could explain more changes of the stock returns. This finding supports statement of Sardosky (2001), who believes that future oil price is more appropriate than spot oil price to be an explanatory variable in the regression model. Moreover, by separating the test period to before, during and after financial crisis to see whether the financial crisis would affect the macroeconomic variables. What I find is that future oil price return, market return and exchange rate return do not fluctuate a lot during three separate periods, but the interest rate return is much lower during the crisis than other two periods. Apparently, higher interest rate during the time period of 2007-2009 increases the cost of oil and gas firms and lowering the stock returns of these companies to some extent. Since financial crisis might be a factor that would also influence the stock returns, it is reasonable to include financial crisis as a dummy variable in the model. The corresponding regression result proves that financial crisis is a significant factor that would also affect the oil and gas companies’ stock returns.
There are some limitations of this paper. First, crude oil price is mainly depends on oil demand and supply. However, in last few years, the quota system has been adopted by the OPEC to achieve a stable level of income of oil exporting
28
countries. Since supply of oil is mainly controlled by the OPEC, the oil price will not change a lot. Therefore, return on crude oil price will not fluctuate a lot. Moreover, I only include four fundamental factors and one dummy variable in the model, but it should be admitted that there are other variables may also affect the stock returns of oil and gas companies. The R-squared value is around 0.6 in the regression model with future oil price, so it could be improved by introducing other variables that could explain changes of stock returns.
References
Abdullah, A. Dewan and Haywoth, C. Steven: “Macroeconomics of Stock Price Fluctuations”, Quarterly Journal of Business and Economics, 32 (1), pp. 49-63, 1993 Al-Mudhaf, A., & Goodwin, T. H. (1993). Oil shocks and oil stocks: evidence from the 1970s.Applied Economics, 25(3), 181-190.
Adam, A. M. and Tweneboah, G.: Macroeconomic Factors and Stock Market Movement: Evidence from Ghana. Social Science Research Network, 2008.
Boyer, M. M., & Filion, D. (2004). Common and Fundamental Factors in Stock Returns of Canadian Oil and Gas Companies, Energy Economics 29 (2007), pp. 428-453.
Brooks, C. (2002). Introductory Econometrics for Finance. Cambridge: Cambridge University Brooks, C. (2002). Introductory Econometrics for Finance. Cambridge: Cambridge University
Chen, N., Roll, R., & Ross, S. A. (1986). Economic Forces and the Stock Market. Journal of Business, 59(3), 383-403.
Fama, E., & French, K., 1989. Business conditions and expected returns on stocks and bonds. Journal of Financial Economics, 25 (1989), pp. 23-49.
Faff, R., & Chan, H.,1998.A multifactor model of gold industry stock returns: evidence from the Australian equity market, Applied Financial Economics, 8:1, pp. 21-28.
Ferson, W. E., & Harvey, C. R. (1991). The Variation of Economic Risk
Hammoudeh, S. S. Dibooglu, and E. Aleisa (2004). "Relationships among U.S. Oil Prices and Oil Industry Equity Indices." International Review of Economics and Finance 13: 427-453.
Hammoudeh, S., and H. Li (2005). "Oil Sensitivity and Systematic Risk in Oil-sensitive Stock Indi- ces." Journal of Economics and Business 57: 1-21.
Lanza, A., M. Manera, M. Grasso, and M. Giovannini (2003). "Long-run Models of Oil Stock Prices." Environmental Modelling and Software 20: 1423-1430
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.
Sadorsky, P. (1999). Oil price shocks and stock market activity, Energy Economics, 21 (5), pp. 449-469.
Sadorsky, P. (2000). The empirical relationship between energy futures prices and exchange rates, Energy Economics, 22 (2000), pp.253-266.
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.
Sadorsky, P., 2002. Time-varying risk premium in petroleum futures prices. Energy Economics 24, 539-556.
Scholtens, B., & Wang, L. (2008). Oil Risk in Oil Stocks. The Energy Journal, 29(1), 89-111.
Solnik, B., 1987. Using financial prices to test exchange rate models: a note. Journal of Finance 42, 141-149.
