The Journal of Current Research in Global Business 14 (21), 2011

COINTEGRATION AND PRIORITY RELATIONSHIPS BETWEEN ENERGY STOCKS AND OIL PRICES Yongli Luo* Omar Esqueda College of Business Administration, University of Texas—Pan American This study investigates the cointegration and priority relationships between large-cap energy stocks and oil price changes over the last decade. The results reveal that energy stock prices have a long run linear relationship with oil price fluctuations. Asset prices are affected by the oil price deviations from the long run equilibrium as well as the short run dynamics. Moreover, some particular energy stocks have Granger causal impacts on oil market as well. The existence of this bivariate long run cointegration implies that employing portfolio diversification between these two markets would be efficient.

Introduction Empirical findings of the impacts of oil fluctuations on stock prices are controversial. Sadorsky (1999), Papapetrou (2001), and Park and Ratti (2008) report that oil price changes have a significant negative effect on stock prices, while Sadorsky (2001) and El-Sharif et al. (2005) claim that there is a positive association between oil prices and oil and gas stocks from oil-exporting countries. To detect the oil factor importance for stock prices, Pollet (2002), Driesprong et al. (2003) and Hammoudeh and Li (2004) find that stock market returns are predictable by oil price volatility, especially in oil-exporting economies. Furthermore, Gogineni (2007) and Yurtsever and Zahor (2007) confirm that stock prices respond to oil price changes asymmetrically. More recently, Bjornland (2008) and Nandha and Faff (2008) conclude that oil prices may affect stock prices directly though affecting economic activities as well as firm’s output. This paper attempts to detect the effects of oil price fluctuations on related capital assets pricing, particularly for energy sectors. Following the spirits of previous research, Jones and Kaul (1996) and Huang et al. (1996) provide evidence that there is a causal linkage between oil prices and stock market returns and that the reactions of stock prices to oil price adjustments are significant. Instead of investigating the impact of oil changes on the overall equity market, this study mainly focuses on large-cap energy stocks in the oil and gas sector. Previous empirical studies document that energy stock prices behave differently in response to oil fluctuations relative to the overall stock market. Huang et al. (1996) elicits that oil futures lead to some individual oil stock returns, but they do not have much impact on broad stock return index. El-Sharif et al. (2005) further find evidence in the U.K. that the price of crude oil affects positively oil and gas sector equity prices. Moreover, Nandha and Faff (2008) indicate that oil price increases have a negative impact on equity returns for all sectors except the mining and oil and gas industries. This paper contributes to previous research in the following ways. First, we test the long run linkage between the large-cap energy stock prices and oil price fluctuations to draw inferences on stock market efficiency and asset pricing theory. Second, this study analyzes the bivariate causal relationship between oil price fluctuations and stock price variations, which is quite different from previous research that it mainly examines the associations between these two markets. Third, unlike previous research that examines the oil price fluctuation effects on asset prices in an indirect way through the perceived effect on the macroeconomy, this paper applies the VECM approach to decompose the oil price changes into the long run equilibrium and the short run stochastic disturbance. In doing so, it is imperative to investigate the long run dynamic impacts of oil price fluctuations on the future cash flows of oil related firms. To detect the long run equilibrium between oil price fluctuations and large-cap energy stock prices, we collect 10-year daily prices for 17 large-cap energy stocks and the West Texas Intermediate (WTI) crude oil. Initially, we conduct the general diagnostic checking for normality and stationarity. Then we employ Johanson cointegration method to investigate the long run cointegration relationship between stock prices and oil prices. Further, we estimate the Vector Error Correction Model (VECM) to decompose the oil variation into the long run equilibrium and the short run

22

stochastic disturbance. Finally, we conduct the Granger causality test to detect the bivariate long run linkage between oil and stock prices. The empirical results reveal that 15 out of 17 energy stocks have long run linear relationship with crude oil prices, asset prices are affected by the changes in oil price from the long run equilibrium as well as the short run dynamics. Moreover, the oil market has much greater response to the previous period’s deviation from long run equilibrium. In addition, some particular energy stocks have Granger causal impacts on oil market as well. Therefore, the stock prices incorporate oil price information as well as other factors. In conclusion, the existence of this cointegrating relationship implies that employing portfolio diversification between these two markets would be efficient. The results have significant implication for energy stock market forecasting and investor benefits of portfolio diversification.

Literature Review and Hypothesis Formation The existing literature has devoted considerable attention to identifying the influences of international oil price volatility on certain macroeconomic variables, such as GDP growth, inflation, unemployment, and exchange rate fluctuation. However, fewer research attempts to detect the effects of oil price changes on related capital asset prices, such as corporate earnings or stock returns. Jones and Kaul (1996) are the initiators to observe the effect of oil price volatility on stock prices. They examine the extent to which stock prices adjust in response to oil price changes or whether changes in stock prices reflect current and future real cash flows. Using a cash flow/dividend valuation model, they find that international stock prices have significant reactions to oil price fluctuations; hence, oil prices can help predict stock returns and output. In the meantime, Huang et al. (1996) provide evidence supportive of causality effects from oil future prices to stock prices. Using a vector autoregression (VAR) model, they investigate the relationship between oil futures and stock returns and find that oil futures lead to some individual oil company stock returns, but they do not have much impact on a broader version of stock index. Similarly, Sadorsky (1999) also uses a VAR framework to establish a model for the U.S. context over the period 1947-1996. He finds that oil price fluctuations have significantly negative impact on stock returns. In addition, Papapetrou (2001) shows that oil price increases have a negative impact on stock prices by deteriorating the industrial output as well as employment. Similarly, Hong et al. (2002) identify a negative association between oil price changes and stock market returns. More recently, O’Neil et al. (2008) and Park and Ratti (2008) report that oil price fluctuations have a statistically significant negative effect on stock prices for an extended sample of 13 developed markets. On the contrary, earlier studies also document that there is a significant positive relationship between oil prices and stock returns. Allowing for the presence of several risk premiums, Sadorsky (2001) identifies a positive relationship between oil price changes and stock returns by using a multifactor market model. Through a sector-based analysis, El-Sharif et al. (2005) reports a significant positive association between oil prices and stock returns at the London Stock Exchange (LSE), albeit such an association varies extensively across sectors. The empirical findings of the oil factor for stock prices are controversial. From a global perspective, Pollet (2002) and Driesprong et al. (2003) find that oil price changes predict stock market returns. Further, Hammoudeh and Li (2004) report similar results in certain oil-sensitive economies. However, considering the political factors, Bittlingmayer (2005) documents that oil price shocks are associated with war risk and it exhibits an asymmetric effect on the behavior of stock prices. In addition, Gogineni (2007) and Yurtsever and Zahor (2007) confirm that stock prices respond asymmetrically to oil price changes, in other words, higher oil prices are associated with lower stock prices, while lower oil prices are not associated with higher stock prices. To examine whether there is an asymmetric effect of oil prices and to what extent it is related to stock returns, Guidi et al. (2006) investigate the impact of OPEC decisions in the context of U.K. and U.S. stock markets over the period 1986–2004, they conclude that asymmetry is only related to conflict regimes. By contrast, Nandha and Faff (2008) argue that oil price increases can have adverse effects on firm’s output as well as on firm’s profitability, especially for those firms where oil is used as an input. They find that oil price rises have a negative impact on equity returns for all sectors except the mining and oil and gas industries. This finding is consistent with El-Sharif et al.’s (2005) finding that the price of crude oil affects equity values positively in the oil and gas sector in the U.K. Recently, Bjornland (2008) shows that oil price fluctuations may affect stock prices, not through an indirect way, but by affecting economic activity, corporate earnings, inflation and monetary policy. In addition, Nicholas and Miller (2009) modify Kilian’s (2008) procedure and find that international stock market returns do not respond in a large way to oil market shocks. In general, the literature consistently supports that an increase in the oil prices has great implications for asset pricing as well as financial market decisions (Mussa, 2000).

Following the spirit of previous research, we hypothesize that the stock prices fully reflect all the relevant information, the dynamic impacts of oil price fluctuations on energy stocks are incorporated in the assets pricing process. Therefore, if the long run cointegration is detected, it is anticipated that the stock market would absorb all the information about the consequences of oil price fluctuations and reflect the present discounted value of the equity’s future cash flows. Further, we hypothesize that asset prices are determined on the stock market information about future prospects as well as current economic conditions faced by the firms. Numerous research concentrate on the notion that oil prices do not affect asset prices in isolation, but through the perceived effect on the macro economy. On the other hand, stock prices are affected by oil prices through the cash flows of oil related firms (Tobin’s Q effect). Therefore, asset prices act as an important transmission channel of wealth growth, both the current and the future impacts of oil price variations should be absorbed into the stock prices before those impacts actually occur. Hence, stock prices are affected by changes in oil prices from the long run equilibrium as well as the short run dynamics.

Data The data consists of 10-year daily prices, from January 1st 2000 to December 31st 2009, for 17 large cap (over 5 billion U.S. dollars in market capitalization) energy stocks1 listed on New York Stock Exchange (NYSE) and the West Texas Intermediate (WTI) crude oil. The daily adjusted closing stock prices are obtained from the Center for Research in Security Prices (CRSP). Originally, we have a sample size of 20 energy stocks. However, Ultra Petroleum Corporation (UPL) and Petrohawk Energy Corporation (HK) do not qualify for the minimum requirement of 10-year trading history, we exclude them. Further, we eliminate Continental Energy Corp. (CPPXF.OB) due to its stationary characteristics in levels. The daily oil price for the Cushing oil-WTI spot price (FOB U.S. dollars per Barrel) is drawn from the U.S. Energy Information Administration (EIA). Table 1: Descriptive Statistics for Large-Cap energy stocks and WTI index Series

Mean

Median

APC

36.45

CHK

19.11

COP

Min

32.64

78.81

13.2

14.25

0.39

0.65

2.60

194.61*

34.18

15.56

68.01

1.91

13.13

0.69

0.84

3.20

302.50*

14.58

41.3

37.59

89.57

13.84

19.89

0.48

0.54

2.05

214.05*

75.86

CVX

50.19

47.04

96.78

24.3

19.4

0.39

0.46

1.89

218.99*

149.48

DVN

46.54

38.3

122.04

14.72

25.36

0.54

0.77

2.74

255.28*

28.24

EOG

47.09

35.3

141.61

6.63

30.77

0.65

0.66

2.59

200.93*

22.63

MRO

23.99

17.62

61.61

7.75

14.64

0.61

0.72

2.26

277.23*

21.87

MUR

38.27

39.29

97.76

10.45

20.08

0.52

0.50

2.48

134.06*

10.21

NE

Std. Dev.

CV

Skewness

Kurtosis

JarqueBera

Max

Mkt Cap

28.4

24.17

67.55

10.75

12.28

0.43

0.88

2.91

322.41*

10.19

NFX

31

27.2

67.96

12.75

13.62

0.44

0.50

1.96

214.65*

6.33

OXY

33.93

26.57

94.12

6.05

23.39

0.69

0.59

2.11

229.57*

66.36

PXD

31.66

30.58

79.25

6.66

14.07

0.44

0.49

2.88

102.35*

6.11

RRC

20.25

13.61

74.48

0.94

18.53

0.92

0.78

2.51

280.64*

7.27

SWN

13.47

6.38

50.62

0.67

13.97

1.04

0.89

2.57

351.35*

13.05

VLO

26.25

17.28

73.6

4.19

20.66

0.79

0.76

2.07

335.19*

11.14

XOM

51.34

46.07

90.83

25.56

19.11

0.37

0.43

1.73

246.71*

314.15

XTO

23.49

19.34

71.93

1.1

17.52

0.75

0.47

2.07

183.81*

27.34

WTI

51.23

45.51

145.31

17.5

25.86

0.50

1.14

4.05

656.12

n.a

Notes: The total number of observations is 2509 over the sample period Jan.1, 2000 through Dec. 31,2009.

* denotes 1% significance level. Market cap is calculated as of March 26, 2010 in billion U.S. dollars. ―n.a‖ denotes the value is not available. CV is coefficient of variation. See footnotes 1 for the tickers’ names.

Methodology and Empirical Results Descriptive Statistics Table 1 presents the descriptive summary. Each mean stock price ranges from $13 to $52 with a positive skewness and an excess kurtosis. The mean price of WTI oil is $51.23, it also exhibits an excess kurtosis of 4.05, which is much greater than that of the corresponding energy stock prices. The coefficients of variation of most of the stocks are at least as high as that of the oil price (which is 25.86) except CVX, PXD, XOM, XTO, implying that the oil price is not as volatile as large-cap energy stocks. The significant Jarque-Bera statistics reject the null of normality at 1% level for all the series, indicating the presences of non-normality characteristics. As of March 26, 2010, all the energy stocks have market capitalizations over 5 billion U.S. dollars. Among them, XOM is the largest one with a market cap of 314.15 billion U.S. dollars, while the market cap of PDX is the smallest, only 6.11 billion U.S. dollars. Table 2: ADF and KPSS Unit Root Test for Energy Stock Prices and WTI indices ADF t-statistic

Series Level

KPSS LM-statistic

First difference

Level

First difference

APC

-3.013

(1)

-38.129

(1)

***

0.367

(40)

0.035

(13)

***

CHK

-1.853

(1)

-37.378

(1)

***

0.396

(40)

0.062

(6)

***

COP

-1.438

(1)

-39.623

(1)

***

0.540

(40)

0.089

(9)

***

CVX

-2.660

(3)

-30.155

(2)

***

0.609

(40)

0.047

(10)

***

DVN

-2.226

(2)

-39.631

(1)

***

0.466

(40)

0.060

(20)

***

EOG

-2.600

(1)

-54.147

0

***

0.390

(40)

0.045

(22)

***

MRO

-1.583

(3)

-29.467

(2)

***

0.560

(40)

0.101

(6)

***

MUR

-2.318

(2)

-39.609

(1)

***

0.560

(40)

0.046

(8)

***

NE

-2.269

(2)

-39.598

(1)

***

0.424

(40)

0.048

(8)

***

NFX

-2.471

0

-50.144

0

***

0.475

(40)

0.048

(20)

***

OXY

-2.996

(2)

-40.707

(1)

***

0.596

(40)

0.028

(23)

***

PXD

-2.276

0

-50.281

0

***

0.543

(40)

0.044

(7)

***

RRC

-3.110

0

-52.947

0

***

0.680

(40)

0.043

(24)

***

SWN

-2.524

(2)

-41.778

(1)

***

1.104

(40)

0.020

(44)

***

VLO

-0.486

0

-48.284

0

***

0.760

(40)

0.179

(6)

XOM

-2.127

(2)

-45.018

(1)

***

0.706

(40)

0.095

(17)

***

XTO

-2.530

(2)

-39.198

(1)

***

0.449

(40)

0.045

(15)

***

WTI

-2.209

(1)

-53.055

0

***

0.264

(40)

0.056

(19)

***

*

Notes: The null hypotheses for Augmented Dicky-Fuller(ADF) test is that the tested series has a unit root. The null hypothesis for Kwiatkowski-Phillips-Schmidt-Shin(KPSS) test is that the tested series is stationary. P-values denote the MacKinnon (1996) one-sided p-values. ADF lag length is selected automatically based on SIC criterion. Maximum lags is 36.KPSS bandwidth is selected based on Newey-West using Bartlett kernel. Intercept and linear trend are included for each test. Optimal lags for ADF and optimal bandwidths for KPSS are in parenthesis. . ***, **, * denotes 1% , 5% and 10% significance level respectively.

Unit Root Tests Granger and Newbold (1974) suggest that unit root tests should be imposed on most time series before any modeling procedures, otherwise, serious problems with spurious regression are involved. To test the existence of unit

roots,we adopt the augmented model by Dicky-Fuller (1979) and the KPSS model by Kwiatkowski et al. (1992). Specifically, the tested sequence {yt} is expressed as the following equation. q

yt  yt 1    i yt  i   t

(1)

i 1

where Δyt is the first difference of the {yt} sequence. q is the optimal lags selected by SIC criteria. εt is a white noise process. The null hypothesis is H 0: β=0. In empirical studies, a drift term or a time trend is generally selected. we perform the test in levels and the first difference for each sequence. Table 3: Johanson Cointegration Test (unrestricted) between Energy Stocks and WTI Hypothesized

Trace

Critical

Eigenvalue

Statistic

Value(5%)

p-value

r=0***

0.0076

21.228

15.495

0.006

r=0***

r≤1

0.0008

2.099

3.841

0.147

r≤1

r=0***

0.0087

26.750

15.495

0.001

r≤1**

0.0020

4.921

3.841

0.027

r=0**

0.0057

16.571

15.495

r≤1

0.0009

2.175

r=0*

0.0058

r≤1

No. of CE(s)

APC CHK COP CVX DVN EOG MRO

Hypothesized

Max-Eigen

Critical

No. of CE(s)

Statistic

Value(5%)

p-value

19.129

14.265

0.008

2.099

3.841

0.147

r=0***

21.829

14.265

0.003

r≤1**

4.921

3.841

0.027

0.034

r=0**

14.396

14.265

0.048

3.841

0.140

r≤1

2.175

3.841

0.140

14.952

15.495

0.060

r=0**

14.494

14.265

0.046

0.0002

0.458

3.841

0.498

r≤1

0.458

3.841

0.498

r=0***

0.0107

28.350

15.495

0.000

r=0***

26.947

14.265

0.000

r≤1

0.0006

1.404

3.841

0.236

r≤1

1.404

3.841

0.236

r=0***

0.0071

19.921

15.495

0.010

r=0**

17.844

14.265

0.013

r≤1

0.0008

2.077

3.841

0.150

r≤1

2.077

3.841

0.150

r=0**

0.0063

17.224

15.495

0.027

r=0**

15.874

14.265

0.028

r≤1

0.0005

1.350

3.841

0.245

r≤1

r=0***

0.0080

22.953

15.495

0.003

r=0***

r≤1

0.0011

2.850

3.841

0.091

r≤1*

r=0***

0.0118

31.601

15.495

0.000

r=0***

r≤1

0.0008

1.964

3.841

0.161

r≤1

r=0***

0.0146

40.031

15.495

0.000

r=0***

r≤1*

0.0013

3.249

3.841

0.072

r≤1*

r=0*

0.0055

14.139

13.429

0.079

r≤1

0.0002

0.442

2.706

r=0*

0.0043

14.395

r≤1

0.0015

r=0* r≤1

1.350

3.841

0.245

20.102

14.265

0.005

2.850

3.841

0.091

29.637

14.265

0.000

1.964

3.841

0.161

36.782

14.265

0.000

3.249

3.841

0.072

r=0*

13.698

12.297

0.061

0.506

r≤1

0.442

2.706

0.506

13.429

0.073

r=0*

10.665

12.297

0.072

3.730

2.706

0.153

r≤1

3.730

2.706

0.153

0.0054

14.666

13.429

0.066

r=0*

13.551

12.297

0.065

0.0004

1.116

2.706

0.291

r≤1

1.116

2.706

0.291

r=0

0.0041

10.652

15.495

0.234

r=0

10.326

14.265

0.191

r≤1

0.0001

0.326

3.841

0.568

r≤1

0.326

3.841

0.568

r=0**

0.0059

18.268

15.495

0.019

r=0**

14.939

14.265

0.039

r≤1*

0.0013

3.329

3.841

0.068

r≤1*

3.329

3.841

0.068

r=0

0.0047

12.669

15.495

0.128

r=0

11.905

14.265

0.114

r≤1

0.0003

0.763

3.841

0.382

r≤1

0.763

3.841

0.382

r=0**

0.0052

18.881

15.495

0.015

r=0*

13.097

12.297

0.076

r≤1 0.0023 5.784 3.841 0.116 r≤1 5.784 2.706 Notes: Lags interval (in first differences) is from 1 to 4. P-values denote the MacKinnon-Haug-Michelis (1999) p-values.

0.116

MUR NE NFX OXY PXD RRC SWN VLO XOM XTO

***, **, * denotes rejection of the hypothesis at 1% , 5% and 10% significance level respectively. Table 2 presents the unit roots statistic results. Applying ADF test at level, we cannot reject the null of nonstationarity, the corresponding p values are greater than 1%, indicating that they are not stationary at level. However, in first difference, all the stock series are statistically significant at 1% level, suggesting that the stock prices are stationary at first difference. Seemingly, KPSS tests reconcile with the ADF results, although the test result for VLO is slightly different which is significant at 5% level. Similar procedures are applied to test the unit roots for crude oil price (WTI) as well. The results reveal that all series are non-stationary at level but stationary at first differences, indicating that they are I(1) processes.

Cointegration Rank Tests Table 3 contains the unrestricted Johanson cointegration rank test (Johansen and Juselius, 1990) results between each stock series and the WTI series for both trace statistic and maximum eigenvalue tests. At 5% level, APC, CHK, COP, CVX, DVN, EGO, MRO, MUR, NE, NFX and VLO have at least one cointegrating equation with WTI oil. At 10% level, there is at least one cointegrating equation between the stock series of OXY, PXD, RRC, XTO and WTI oil. However, the cointegration tests of XOM, SWN and WTI are not supported by both the trace and maximum eigenvalue tests. In sum, 15 out of 17 stock series are found to be cointegrated with the WTI oil, indicating that there is a long run linkage between large-cap energy stocks and oil prices. Further, we construct two large-cap energy stock portfolios with different weighting approaches. Portfolio EW is an equally weighted portfolio and VW is a value weighted portfolio of 15 stocks. Similar to the previous procedures of diagnostic checking, we conduct the ADF test and the KPSS test for both series. In Table 4, ADF test cannot reject the null in level, but reject the null in first difference at 1% confidence interval. Seemingly, the KPSS results confirm the ADF tests. Therefore, both EW and VW series are stationary in first differences, but are non-stationary in levels, indicating that they are I(1) processes. Table 4: Unit Root Test for EW and VW ADF t-statistic Level EW VW Critcal Value

KPSS LM-statisic

First difference

-1.979

(2)

-39.224

(1)

***

-1.845

(2)

-39.765

(1)

***

1% level

-3.962

5% level 10% level

Level

First difference

0.454

(40)

0.065

(13)

***

0.525

(40)

0.079

(3)

***

1% level

0.216

-3.412

5% level

0.146

-3.128

10% level

0.119

Notes: The null hypothesis for Augmented Dicky Fuller(ADF) test is : the series has a unit root. The null hypothesis for Kwiatkowski-PhillipsSchmidt-Shin(KPSS,1992) test is : the series is stationary. Prob. denotes the MacKinnon (1996) one-sided p-values. ADF lag length is selected automatically based on SIC criterion, a maximum lag length 36 is imposed. KPSS bandwidth is selected based on Newey-West using Bartlett kernel. Constant and linear trend are included as exogenous variables in the test. Lags for ADF and bandwidth for KPSS are in parenthesis. ***, ** and * denotes significance at 1%,5% and 10%, respectively.

Table 5: Johanson Cointegration Test for EW/WTI and VW/WTI Hypothesized

Trace

Critical

Eigenvalue

Statistic

Value

r=0***

0.0096

26.061

15.495

0.001

r=0***

r≤1

0.0007

1.87

3.841

0.172

r≤1

r=0**

0.0075

19.486

15.495

0.012

r=0***

r≤1

0.0003

0.741

3.841

0.389

r≤1

No. of CE(s) EW VW

p-value

Hypothesized

Max-Eigen

Critical

No. of CE(s)

Statistic

Value

p-value

24.192

14.265

0.001

1.87

3.841

0.172

18.746

14.265

0.009

0.741

3.841

0.389

Notes: Lags interval (in first differences) is from 1 to 4. P-values denote the MacKinnon-Haug-Michelis (1999) p-values.

***, **, * denotes rejection of the hypothesis at 1% , 5% and 10% significance level respectively. Table 5 reports the results of unrestricted Johansen Cointegration rank test between EW, VW and WTI oil. For the equally weighted series, under the hypotheses of no cointegration equation (r=0), both the Trace statistic and Maxeigenvalue statistic are greater than the critical values, denoting rejection of the null hypothesis at 5% level. However, under the hypotheses of at most 1 cointegration equation (r≤1), both rank tests failed to reject the null at 5%, indicating that long run cointegration is supported. Seemingly, for the value weighted series VW, both Trace and Max-eigenvalue statistic reveal that there is one cointegration equation significant at 5% level. Therefore, both formulated energy stock portofolios have long run cointegrated relationships with WTI oil. Table 6: Vector Error Correction Estimations for EW/WTI and VW/WTI The estimated equation is p

p

i 1

i 1

p

p

i 1

i 1

yt  10   y ( xt 1   * yt 1 )   11 (i )yt i   12 (i )xt i   yt xt   20   x ( xt 1   * yt 1 )    21 (i )yt i    22 (i )xt i   xt ΔEW β

αx,y ΔEW (-1) ΔEW (-2) ΔWTI (-1) ΔWTI (-2) αi0

ΔWTI

ΔVW

ΔWTI

-1.194

-1.092

[-15.912]

[-12.711]

0.002

0.015

0.002

0.013

[ 1.137]

[ 4.979]

[ 1.202]

[ 4.281]

-0.009

0.09

-0.034

0.137

[-0.452]

[ 3.135]

[-1.668]

[ 3.510]

-0.094

-0.044

-0.098

-0.044

[-4.514]

[-1.536]

[-4.717]

[-1.126]

-0.022

-0.038

-0.011

-0.038

[-1.446]

[-1.839]

[-1.050]

[-1.852]

0.006

-0.019

0.004

-0.021

[ 0.390]

[-0.907]

[ 0.385]

[-1.043]

0.001

0.001

0.001

0.001

[ 1.814]

[ 0.825]

[ 2.051]

[ 0.791]

R-squared

0.01

0.017

0.012

0.015

Adj. R-squared

0.008

0.015

0.01

0.013

F-statistic

5.209

8.795

5.898

7.763

6314.244

5503.814

7097.374

5501.269

-5.035

-4.388

-5.66

-4.386

Log likelihood Akaike AIC

Notes: t-statistics are reported in brackets. Δ denotes first difference operator. ●(-1) , ●(-2) denote lagged one and two periods respectively. xt denotes the energy stock portfolio EW/VW series, yt denotes the WTI crude oil prices.

Vector Error Correction Estimations The main feature of the Vector Error Correction Model (VECM) is a system of equations capturing the errors for any disequilibrium that may affect the system from time to time (Enders, 2003). A change in oil prices not only directly affects itself but also is transmitted to all of the energy stock prices through the dynamic structure of the VECM. Since the error correction representation necessitates that two variables be cointegrated of order CI(1,1), 2 the error correction term picks up such disequilibrium and guides the variables of the system back to equilibrium. Specifically, the estimated equation is expressed as follows. p

p

i 1

i 1

p

p

i 1

i 1

yt  10   y ( xt 1   * yt 1 )   11 (i )yt i   12 (i )xt i   yt

(2)

xt   20   x ( xt 1   * yt 1 )    21 (i )yt i    22 (i )xt i   xt where Δ is the lag operator. xt denotes the energy stock portfolio EW/VW price series, yt denotes the WTI crude oil price. Both xt and yt series are in natural logarithmic form. αx and αy are the speed of adjustment parameters, while β is the estimated coefficient from the long run equilibrium between each stock portfolio and WTI price. εyt and εxt are white noise disturbance terms. Table 6 reports the VECM estimations between EW, VW and the TWI oil. The stock price changes are decomposed into the stochastic component (represented by ε yt and εxt) and the previous period’s deviation from long-run equilibrium (expressed as xt-1 -β* yt-1) The cointegrating vector are (1, -1.194) and (1, -1.092) respectively. αx and αy measure the speed of adjustment. The adjustment coefficients of oil price are 0.013-0.015, while the speed for stock prices is 0.002. Apparently, the oil market has slightly greater response to the previous period’s deviation from the long run equilibrium. However, the coefficients are relatively small and it is unfeasible to make strong statements about their speed of adjustment. Table 7: Pairwise Granger Causality Tests for SWN/WTI and XOM/WTI Null Hypothesis F-Statistic

P-value

F-Statistic

P-value

F-Statistic

P-value

SWN --/--> WTI

2.3696

0.0087

2.3392

0.0007

2.1876

0.0002

WTI --/--> SWN

1.3863

0.1801

1.3742

0.1235

1.2539

0.1618

XOM --/--> WTI

2.6861

0.0028

2.5748

0.0002

2.0266

0.0008

WTI --/--> XOM

1.9174

0.0386

1.4545

0.0873

1.2542

0.1616

Lags N

10

20

30

2499

2489

2479

Granger Causality Tests Table 7 reports the Granger causality tests (Granger, 1969) for both SWN/WTI and XOM/WTI relationships. At 5% level, the F-statistic is significant, rejecting the null that SWN does not Granger cause WTI. However, under the null hypothesis that WTI does not Granger cause SWN, the F statistic is less than the critical value, indicating that the null is supported. To check the long run dynamics, we increase the lag lengths to 20 and 30, the results are consistent. Similarly, we conduct the causality test on the XOM/WTI pair, we reject the null that XOM does not Granger cause WTI, and the null that WTI has no Granger causality effect on XOM is also rejected if the lags extend to 20 or 30. The result is conjectured with previous research because it shows that the oil market has little causal effects on these two energy stocks. Adversely, the results illustrate that the fluctuations from these two energy stocks have direct influences on oil price. The reason why XOM prices are so independent of oil price variations is plausible. The explicit intuition

relies on the fact that during our sample period XOM has had an exceptionally dominant position in the energy industry, leading itself to an implicit oligopolistic position. Historic market capitalization data indicates that, on average, during the first half of the sample period, XOM was four times larger than CVX and 12.5 times larger than COP. Nevertheless, the distances have diminished in the last few years. On the other hand, the rationale for the reverse Granger causality of SWN is plausible. We are not certain about the possible causes of this inconsistent relationship.

Conclusions This study investigates the cointegration and priority relationships between large-cap energy stock prices and oil market fluctuations over the last decade. By adopting Johanson Cointegration test, we find that apart from SWN and XOM, 15 out of 17 stock series are cointegrated with the WTI oil price, indicating that there is a long run linear cointegrated relationship between energy stocks and oil prices. We then formulate these 15 stocks into two large-cap energy stock portfolios named as EW and VW. EW is the equally weighted and VW is the value weighted portfolio. The empirical results illustrate that both EW and VW series have long run linear relationships with the WTI oil price. In conclusion, the existence of this long run cointegration implies that employing portfolio diversification between these two markets would be efficient. The results have significant implication for energy stock market forecasting and investor benefits of portfolio diversification. Further, we analyze the particular cases of SWN and XOM, the Granger causality tests indicate that the oil market has little causal effects on these two energy stock prices. Adversely, the changes in these two energy stocks have Granger casual effects on oil prices. The reason why XOM is so independent of the market might be interpreted by its large share of market capitalizations, leading itself to an implicit oligopolistic position, whereas a justification for the SWN independence is unknown. This paper mainly focuses on large-cap energy stocks, further study can be done on middle-cap or small-cap energy equities to check the size effects. In addition, researchers may extend the research by examining variations in the oil price combined with other variables such as monetary policy in an asset pricing model, or examining effects from the perspective of other less frequency data such as weekly and monthly data.

Footnotes * Corresponding author Email: [email protected]. The authors are grateful to Dr. Dave Jackson and two anonymous reviewers for helpful comments. 1. The 17 energy stocks are Anadarko Petroleum Corporation (APC), Chesapeake Energy Corporation (CHK), Conoco Phillips (COP), Chevron Corp. (CVX), Devon Energy Corporation (DVN), EOG Resources, Inc. (EOG), Marathon Oil Corporation (MRO), Murphy Oil Corporation (MUR), Noble Corp. (NE), Newfield Exploration Co. (NFX), Occidental Petroleum Corporation (OXY), Pioneer Natural Resources Co. (PXD), Range Resources Corp. (RRC), Southwestern Energy Co. (SWN), Valero Energy Corp. (VLO), Exxon Mobil Corp. (XOM) and XTO Energy Inc. (XTO). 2. This result is unchanged if we formulate a more general model by introducing more lags ( p) into both equations.

The Journal of Current Research in Global Business 14 (21), 2011 References Bittlingmayer, George (2005), ―Oil and Stocks: Is It War Risk?‖ Working Paper Series. University of Kansas. Bjornland, Hilde C. (2008), ―Oil Price Shocks and Stock Market Booms in an Oil Exporting Country.‖ Working Paper 16. Norges Bank, Norway. Dickey, David A., and Wayne A. Fuller (1979), ―Distribution of the Estimators for Autoregressive Time Series with a Unit Root. Journal of American Statistical Association, V74, p.427–431. Driesprong, Gerben, Ben Jacobsen, and Maat Benjiman (2003), ―Striking Oil: Another Puzzle?‖ Working Paper, Erasmus University Rotterdam. El-Sharif, Idris, Dick Brown, Bruce Burton, Bill Nixon. and Alex Russell (2005), ―Evidence on the Nature and Extent of the Relationship between Oil Prices and Equity Values in the UK.‖ Energy Economics, V27, N6, p. 819-830. Enders, Walter, ―Applied Econometric Time Series ( 2nd edition).‖ John Wiley & Sons Pte Ltd, 2003. Gogineni, Sridhar (2007), ―The Stock Market Reaction to Oil Price Changes.‖ Working Paper. University of Oklahoma. Granger, Clive W. J. (1969), ―Investigating Causal Relations by Econometric Models and Cross-spectral Methods.‖ Econometrica, V37, N3, p.424–438. Granger Clive W. J., and P. Newbold (1974), ―Spurious Regressions in Econometrics,‖ Journal of Econometrics, V2, N2, p.111–120. Guidi, Marco G. D., Alexander Russell, and Heather Tarbert (2006), ―The Effect of OPEC Policy Decisions on Oil and Stock Prices.‖ OPEC Review, V30, p.1–18. Hammoudeh, Shawkat, and Huimin Li (2004), ―Risk-return Relationships in Oil-sensitive Stock Markets.‖ Finance Letters, V2, p. 10–15. Hong, Harrison G., Walter N. Torous, and Rossen I. Valkanov (2002), ―Do Industries Lead the Stock Market? Gradual Diffusion of Information and Cross-asset Return Predictability.‖ Working Paper, Stanford University and UCLA. Huang, Roger D., Ronald W. Masulis, and Hans R. Stoll (1996), ―Energy Shocks and Financial Markets.‖ Journal of Futures Markets, V16, N1, p. 1–27. Johansen, Soren, and Katarina, Juselius (1990), ―Maximum Likelihood Estimation and Inference on Cointegration with an Application to the Demand for Mmoney.‖ Oxford Bulletin of Economics and Statistics V52, p. 169–210. Jones, Charles M., and Gautam Kaul (1996), ―Oil and the stock markets.‖ Journal of Finance, V51, N2, p. 463-491. Kilian, Lutz (2008), ―Exogenous Oil Supply Shocks: How Big Are They and How Much Do They Matter for the U.S. economy?‖ The Review of Economics and Statistics, V90, N2, p. 216-240. Kwiatkowski, D., P. C. B. Phillips, P. Schmidt, and Y. Shin (1992), ―Testing the Null Hypothesis of Stationarity against the Alternative of a Unit Root.‖ Journal of Econometric, V54, p. 159–178. Mussa, Michael (2000), ―The Impact of Higher Oil Prices on the Global Economy.‖ International Monetory Fund 2000. 8 December, http://www.imf.org/external/pubs/ft/oil/2000/. Nandha, Mohan, and Robert Faff (2008), ― Does Oil Move Equity Prices? A Global View.‖ Energy Journal, V30, N3, p. 986-997.

31

Nicholas, Apergis, and Stephen M. Miller (2009), ―Do Structural Oil-market Shocks Affect Stock Prices?‖ Energy Economics, V31, N4, p. 569–575 O'Neil, T. J., J. Penm, and R. D. Terrell (2008), ―The Role of Higher Oil Prices: A Case of Major Developed Countries.‖ Research in Finance, V24, p. 287–299. Papapetrou, Evangelia (2001), ―Oil Prices, Stock Market, Economic Activity and Employment in Greece.‖ Energy Economics, V23, N5, p. 511–532. Park, Jungwook, and Ronald A. Ratti (2008), ―Oil price shocks and stock markets in the U.S. and 13 European countries.‖ Energy Economics, V30, p. 2587–2608. Pollet, Joshua (2002), ―Predicting Asset Returns with Expected Oil Price Changes.‖ Working Paper. Harvard University. Sadorsky, Perry (1999), ―Oil Price Shocks and Stock Market Activity.‖ Energy Economics, V21, N5, p. 449–469. Sadorsky, Perry (2001), ―Risk Factors in Stock Returns of Canadian Oil and Gas Companies.‖ Energy Economics, V23, N1, p. 17–28. Yurtsever, Cenk, and Talib Zahor (2007), ―Oil Price Shocks and Stock Market in the Netherlands.‖ Working Paper Series. University of Groningen.