- Email: [email protected]
The shift in US oil demand and its impact on OPEC's market share
Energy Economics 23 Ž2001. 659᎐666
The shift in US oil demand and its impact on OPEC’s market share Imad Jabir U Department of Economics, American Uni¨ ersity of Beirut, Beirut, Lebanon
Abstract Tremendous political pressure is being exerted on the US government by different political parties to diversify its sources of foreign oil supplies by switching from the reliance on OPEC’s oil to that originating from non-OPEC nations. Without a doubt, such a shift would adversely impact the market share of some OPEC members, particularly Saudi Arabia, Venezuela and Nigeria. These countries should therefore consider seriously the negative impact of this scenario and consequently formulate individual or joint production policies aiming at protecting their oil market share. To help OPEC achieve this objective, there is a need to estimate the demand function of US oil imports. This paper proffers an estimate of such a function, taking into account, among other variables, the impact of US Strategic Petroleum Reserve ŽSPR.. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Oil supplies; OPEC; Demand function; Strategic Petroleum Reserve ŽSR.
1. Introduction The United States consumes about one third of the World total oil production every year. This makes the US a significant customer for OPEC and other major oil producers. The considerable US role as an importer is expected to continue for the next two decades of the new millennium. The US has increased its crude oil imports by 40% during the period 1988᎐1998. While the US has increased its imports by 25% from the Arab Gulf states during the same period, its oil imports went up by 122% from the countries of the southern region of the Atlantic Ocean such as Venezuela, Nigeria, Angola and U
Corresponding author.
0140-9883r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 4 0 - 9 8 8 3 Ž 0 1 . 0 0 0 7 6 - 7
660
I. Jabir r Energy Economics 23 (2001) 659᎐666
Colombia. This indicates that the US has just shifted its oil supplies within OPEC members. Although the US government will continue to reduce its dependence on foreign oil, its imports from non-OPEC countries will increase. Hence, it is of paramount importance for OPEC nations to estimate the US long-run oil demand function. The estimation of such a function would help OPEC countries, particularly Saudi Arabia, Venezuela and Nigeria in formulating their production policies and making sound decisions regarding their respective productive capacity. The purpose of this short paper is to proffer an estimation of the US long-run oil demand function. The rest of the paper is made up of two sections. Section 2 discusses the econometric procedure, data, empirical results and implications for OPEC, and Section 3 concludes the paper.
2. Econometric methodology and results The performance of any economy is measured in terms of the growth rate of its gross domestic product ŽGDP.. In general, the US GDP appears to play a major role as a long-term determinant of its demand for crude oil, including foreign oil. In other words, GDP can be used together with oil price to form a long-run relationship to estimate the quantity of oil demand. Furthermore, the deterioration of crude oil prices in 1998 was attributed in part to the accumulated stocks of crude oil reserve throughout the world, and the subsequent price improvement in 1999 was due in part to the decline of these stocks. This linkage of oil prices to the size of oil inventory gives credence to the important role being played by the US Strategic Petroleum Reserve ŽSR.. Since its inception in 1977, the US government has resorted to SR to exert downward pressure on US oil demand and prices. Therefore, the change in SR will be used as an exogenous variable to determine the change in US oil imports. In this paper, we used the annual data for the period 1974᎐1998 to determine the US demand for crude oil imports. These data were extracted from different sources including several issues of the US Economic Report of the President, Annual Energy Review, Statistical Abstract of the US and Oil and Gas Journal. The variables included in the data are defined as follows: Z s the natural logarithm of the real price per barrel Žbbl. of imported crude oil. This price was measured by the US Energy Information Administration in 1992 dollars. Y s the natural logarithm of the US real GDP. This GDP was measured in 1992 dollars. X s the natural logarithm of the US crude oil imports measured for every year as a daily average of imported barrels. SR s The US annual strategic petroleum reserve, where every value of this reserve is reported at the end of the year. Furthermore, since this reserve
I. Jabir r Energy Economics 23 (2001) 659᎐666
661
was formed first in 1977, the values of SR prior to this year were all considered to be zero. To extract the targeted demand equation, a vector error correction ŽVEC. model is used. This model requires testing for cointegration between the time series of the three endogenous variables X, Y and Z, which are in general non-stationary. It is statistically necessary to perform the cointegration test to discern a long-run relationship between X, Y and Z. The maximum likelihood procedure, developed in Johansen Ž1988., Johansen and Juselius Ž1990. and Johansen Ž1992., will be employed to test for cointegration. The test will be referred to hereafter as Johansen’s procedure. The computer package used to apply Johansen’s procedure is E-views, version 3.1. Following this procedure, we will initially test for normality of the residuals of ⌬ X, ⌬Y and ⌬ Z equations that form the VEC model. Considering the lag length k as k s 1, the normality assumption is tested by using the Jarque and Bera Ž1980. test. The results are reported in Table 1, which shows that all the residuals meet the normality assumption. This proves that Johansen’s procedure is quite applicable in finding a long-run relationship between the variables X, Y and Z. We discuss next how to determine jointly both the cointegration rank r for the three endogenous variables, and whether or not there is a linear trend in the model. In other words, which value of r should one use, and is Hr or HrU a better description of the data. The hypothesis Hr of at most r cointegrating relations implies the presence of linear trend, while the hypothesis HrU of at most r cointegrating relations implies the absence of linear trend. HrU is more restricted than Hr , and it is also considered as a sub-hypothesis of Hr . According to Johansen Ž1992., Hr can only be rejected if HrU is also rejected. Let Tr denote the likelihood ratio test statistic for the hypothesis Hr , and let TrU denote the likelihood ratio test statistic of the hypothesis HrU . The resulting values of TrU and Tr , as well as the 95% quantiles Žthe tabulated values. cUr Ž5%. and c r Ž5%., are shown in Table 2. Reading Table 2 from left to right and from top to bottom, it can be noted that H0U , H0 and H1U have to be rejected at the 5% level, but H1 cannot be rejected. Thus, it can be concluded that r s 1, and hence, there is only one cointegrating equation which represents the long-run relationship between the variables X, Y and Z. Additionally, the results show that there is a linear trend in the data of the three variables.
Table 1 Normality assumption test
⌬ Xt ⌬Yt ⌬Zt
2N Ž2.
Probability
Number of observations
Skewness
Kurtosis
2.226 0.567 1.120
0.328 0.753 0.571
23 23 23
0.752 0.384 y0.538
3.241 2.967 3.109
2N Ž2. is the Jarque and Bera normality test statistic.
I. Jabir r Energy Economics 23 (2001) 659᎐666
662
Table 2 The resulting values of TrU and Tr p-r
r
Tr
U
U cr Ž5%.
Tr
cr Ž5%.
3 2 1
0 1 2
58.90 24.32 9.79
34.91 19.96 9.24
44.39 11.09 0.08
29.68 15.41 3.76
p is the number of variables.
Johansen’s procedure is used here to test the hypothesis of the number of cointegrated equations, which will enable us to build the VEC model. The results of the cointegration test are given in Table 3. As shown in Table 3, there is at most one cointegration equation from which the estimated long-run US demand relationship for crude oil imports is obtained. This relationship is given in Eq. Ž1. below: X s 4.7343 q 0.80507 Y y 0.945298 Z Ž y2.003. Ž 7.1976.
Ž1.
The values of the t-statistics are reported in parentheses. The estimated Eq. Ž1. indicates that all the coefficients are statistically significant and have the correct signs. The long-run income and price elasticities of demand are 0.805 and y0.945, respectively. Clearly, the income elasticity is positive but its value is less than one. Compared to the newly industrialized countries such as the so-called Asian tigers, the US has a relatively low income elasticity. This implies two conclusions for OPEC nations to closely consider. First, the US has been increasing its dependence on some other sources of energy relative to crude oil imports. Second, the newly industrialized countries in Asia are expected to be a more promising market than the US for some OPEC members such as: the United Arab Emirates; Qatar; and Kuwait. The price elasticity indicated in Eq. Ž1. is relatively high. This can be attributed again to the reduction in US dependence on oil imports, which happened possibly because of technology enhancement that improved energy consumption efficiency in the US. The next step is to incorporate the resulting long-run relationship of Eq. Ž1. into Table 3 Cointegration test Eigenvalue
Likelihood ratio
5% Critical value
1% Critical value
Number of cointegrated equationŽs.
0.76493 0.38018 0.00365
44.386 11.086 0.084
29.68 15.41 3.76
35.65 20.04 6.65
Nonea At most 1 At most 2
a
Denotes rejection of the hypothesis at the 1% significance level.
I. Jabir r Energy Economics 23 (2001) 659᎐666
663
a vector autoregressive ŽVAR. model which will be an important ingredient in building the VEC model. As mentioned previously, the variables X, Y and Z are taken as endogenous variables while the stock of the US Strategic Petroleum Reserve SR, lagged one period, is treated as an exogenous variable. SR t is the stock of the reserve reported at the end of year t. Logically, the rise Žfall. in X during year t must follow the fall Žrise. in SR measured at the end of year t-1. Hence, the coefficient of SR ty1 is expected to have a negative sign in the equation of ⌬ X t of the VEC model. The estimated VEC model is specified in Table 4. The general formula for this model is: ⌬ X t s ⌫ ⌬ X ty1 q ⌸ X ty1 q q ␣ SR ty1 q t , Ž t s 1, . . . . . . ,25.
Ž2.
where ⌬ X t , ⌬ X ty1 , X ty1 , , ␣ and t are Ž3 = 1. vectors, while ⌫ and ⌸ are Ž3 = 3. matrices. The column of ⌬ X t in Table 4 shows that there are some estimated coefficients that are not satisfactory. The estimated coefficient of ⌬Yty1 is not statistically significant, and the estimated coefficient of ⌬ Zty1 does not have the expected negative sign. Therefore, an ordinary least squares ŽOLS. model was built to come out with the final reliable demand equation, using the same error correction variable ŽECV., lagged one period, taken from Eq. Ž1.. This error correction variable is: ECVty1 s X ty1 y 0.805 Yty1 q 0.945 Zty1 y 4.734
Ž3.
Table 4 Estimated vector error correction model
Cointegration Equation ⌬ Xty 1 ⌬Yty 1 ⌬ Zty 1 Constant SRty 1 R-squared S.E. of regression
⌬ Xt
⌬Yt
⌬Zt
y0.7989 Žy5.398. y0.3324 Žy1.644. 0.1483 Ž0.179. 0.2140 Ž2.116. 0.4892 Ž5.102. y0.00114 Žy4.921. 0.768 0.0709
y0.0909 Žy2.524. y0.0557 Žy1.132. 0.0834 Ž0.414. 0.0177 Ž0.720. 0.0896 Ž3.840. y0.00016 Žy2.792. 0.415 0.0172
0.09796 Ž0.182. 0.5037 Ž0.685. y0.3691 Žy0.123. y0.0590 Žy0.161. y0.0194 Žy0.055. y0.00011 Žy0.130. 0.106 0.2577
The values of the t-statistics are given in parentheses. Sample Žadjusted.: 1976᎐1998. Included observations: 23 after adjusting end points.
I. Jabir r Energy Economics 23 (2001) 659᎐666
664
The resulting OLS model is: ⌬ X t s 0.238 y 0.451 Ž X ty1 y 0.805 Yty1 q 0.945 Zty1 y 4.734. Ž 4.974.
Ž y7.735.
q1.127⌬Yt y 0.174⌬ Zt y 0.000645 SR ty1 Ž 1.724.
Ž y3.067.
Ž y6.226.
Ž4.
R-squared s 0.818, S.E. of regression s 0.0609, D-W statistic s 2.388. The values of the t-statistics are reported in parentheses. It is shown that, with the OLS, there has been a considerable improvement in the values of R-squared and the standard error of regression in Eq. Ž4.. The non-zero estimate 0.451 of the adjustment parameter is highly significant, and hence, there is a quick adjustment by reverting to equilibrium between the levels of the variables. Theoretically speaking, even though unit roots caused a non-stationary problem to the data, transforming the variables to be IŽ0. has solved the problem and consequently generated a demand equation which is more plausible. IŽ0. means that the variables are integrated of order 0. Eq. Ž4. shows also that the estimate 0.000645 of the coefficient of SR is significant in determining the level of US oil imports. Yet, OPEC members should not rely on the potential reduction in the US SR to help them bolster the oil price. They should rather attach great importance to cooperation among themselves to restrict oil output. The output restriction should also be coordinated in concerted efforts with non-OPEC producers. The US is considered as a consumer and producer at the same time in the world crude oil market, hence, such coordination will be helpful to US oil and gas multinational corporations. A target price range of $23᎐26 a barrel would benefit OPEC members, as well as the US multinational corporations and consumers. OPEC members do not have vested interest in an increase in oil price beyond $26, as they will again start losing part of their market share. It will then be quite lucrative for non-OPEC nations to produce oil, and technology will feasibly be developed to replace oil and gas as sources of energy. According to Eq. Ž4., a one million barrel decrease in the SR stock reported by the end of year t y 1 causes an increase of 0.000645 in the natural logarithm of the ratio of oil imports in year t relative to that in year t y 1. In other words, there is about 0.1% increase in the level of oil imports of any year over the previous one as a result of the reduction of one million barrels in the SR already reported. Furthermore, as a result of a one million barrel increase in this stock of reserve of year t-1, the average price in the successive year is offset only by 0.4%. According to this analysis, the stock of SR is not a highly effective tool if it is to enable the US government to influence prices. Hence, the only two purposes for which it can be maintained is to reduce the macroeconomic damage that can occur in a crisis or to make oil easily available to some disadvantaged groups. Eq. Ž4. indicates that short-run income and price elasticities of demand are 1.127
I. Jabir r Energy Economics 23 (2001) 659᎐666
665
and y0.174, respectively. As expected, price elasticity is small and it gets higher after some time. In the short-run, the US oil imports tend to respond faster to changes of income than changes of price. This is due to the consumption nature of the US economy. Crude oil is more broadly defined than any petroleum product considered separately, such as gasoline. It is natural that gasoline has a higher income elasticity of demand in the long run than the short-run. But this does not have to be necessarily true for crude oil. As mentioned before, higher income can provide an access to some other sources of energy. Implicitly this means less dependence on crude oil, including the imported part, compared to the other sources. Therefore, with rising US GDP, more imported oil is demanded, but proportionately less over time. The prices of crude oil declined sharply throughout the whole world during 1986. Therefore, it is required to use this year to test Eq. Ž4. for stability. The breakpoint test by Chow Ž1960. is used in this paper to test for the stability of the estimated coefficients in the resulting demand equation, considering the 5% significance level. Chow test statistic is distributed here as an F distribution with 5 and 15 degrees of freedom under the null hypothesis that all the observations prior to and after 1986 belong to the same regression model, which means, in other words, that there is no structural change in the data. It was found that the calculated value of the test statistic is 2.306, which is less than the tabulated value of 2.9. This leads to the conclusion that the null hypothesis cannot be rejected.
3. Conclusions This paper uses a set of data for the already defined variables X, Y and Z. The time series of each of these variables was found non-stationary and IŽ1.. But differencing each series delivered a stationary one which is IŽ0.. In addition, the empirical results show that there is a linear trend in the data of X, Y and Z. It was also found that there is only one cointegration equation which represents the long-run US demand relationship for crude oil imports. Finally, a reliable stable demand equation was constructed through OLS using the long-run relationship between X, Y and Z. This demand model explained the growth rate in X in terms of: Ži. the past disequilibrium between the levels of X, Y and Z which is the long-run relationship lagged one period; Žii. the variable SR lagged one period; and Žiii. the growth in Y and Z. Investigating the US demand for crude oil imports in this paper has proved that OPEC’s ability to maintain prices depends on the coherence Žor cohesion. of its members to their assigned quotas and cooperation with non-OPEC producers in order to alleviate their competitive behavior. This is much more important than changes of the level of the strategic petroleum reserve possessed by the US. This paper concentrates also on the point that the US will remain a very important customer for OPEC even though there are recently some new emerging considerable customers.
666
I. Jabir r Energy Economics 23 (2001) 659᎐666
Energy Information Administration, Hendry and Juselius, 2000; Johansen, 1991; Oil Gas Journal Osterwald-Lenum, 1992; US Census Bureau, US Government, 1997; Wilson, 1998
Acknowledgements The author would like to thank anonymous reviewers of this journal for their helpful suggestions, and Professor Musa Essayyad, co-editor of this journal, for his editorial help. The author would also like to thank Professors Fabrizio Carlevaro, Jean-Paul Chaze, and Jaya Krishnakumar for their feedback during his summer 1999 visit to CUEPE of the University of Geneva, Switzerland.
References Chow, G.C., 1960. Tests of equality between sets of coefficients in two linear regressions. Econometrica 28 Ž3., 591᎐605. Jarque, C.M., Bera., A.K., 1980. Efficient tests for normality, homoscedasticity and serial independence of regression residuals. Econ. Lett. 6, 255᎐259. Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dyn. Control 12, 231᎐254. Johansen, S., Juselius., K., 1990. Maximum likelihood estimation and inference on cointegration ᎏ with applications to the demand for money. Oxf. Bull. Econ. Statistics 52 Ž2., 169᎐210. Johansen, S., 1992. Determination of cointegration rank in the presence of a linear trend. Oxf. Bull. Econ. Statistics 54 Ž3., 383᎐397.
Further reading Energy Information Administration, U.S.A. Annual Energy Review, several issues. Hendry, D.F., Juselius., K., 2000. Explaining cointegration analysis: Part 1. Energy J. 21 Ž1., 1᎐42. Johansen, S., 1991. Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica 59 Ž6., 1551᎐1580. Oil Gas J., U.S.A., several issues. Osterwald-Lenum, M., 1992. A note with quantiles of the asymptotic distribution of the maximum likelihood cointegration rank test statistics. Oxf. Bull. Econ. Statistics 54 Ž3., 461᎐472. US Census Bureau, the official statistics. Statistical Abstract of the United States, several issues. US Government, 1997. Economic Report of the President. Wilson, R., 1998. Economic Development in the Middle East. Routledge, New York.
The shift in US oil demand and its impact on OPEC’s market share Imad Jabir U Department of Economics, American Uni¨ ersity of Beirut, Beirut, Lebanon
Abstract Tremendous political pressure is being exerted on the US government by different political parties to diversify its sources of foreign oil supplies by switching from the reliance on OPEC’s oil to that originating from non-OPEC nations. Without a doubt, such a shift would adversely impact the market share of some OPEC members, particularly Saudi Arabia, Venezuela and Nigeria. These countries should therefore consider seriously the negative impact of this scenario and consequently formulate individual or joint production policies aiming at protecting their oil market share. To help OPEC achieve this objective, there is a need to estimate the demand function of US oil imports. This paper proffers an estimate of such a function, taking into account, among other variables, the impact of US Strategic Petroleum Reserve ŽSPR.. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Oil supplies; OPEC; Demand function; Strategic Petroleum Reserve ŽSR.
1. Introduction The United States consumes about one third of the World total oil production every year. This makes the US a significant customer for OPEC and other major oil producers. The considerable US role as an importer is expected to continue for the next two decades of the new millennium. The US has increased its crude oil imports by 40% during the period 1988᎐1998. While the US has increased its imports by 25% from the Arab Gulf states during the same period, its oil imports went up by 122% from the countries of the southern region of the Atlantic Ocean such as Venezuela, Nigeria, Angola and U
Corresponding author.
0140-9883r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 4 0 - 9 8 8 3 Ž 0 1 . 0 0 0 7 6 - 7
660
I. Jabir r Energy Economics 23 (2001) 659᎐666
Colombia. This indicates that the US has just shifted its oil supplies within OPEC members. Although the US government will continue to reduce its dependence on foreign oil, its imports from non-OPEC countries will increase. Hence, it is of paramount importance for OPEC nations to estimate the US long-run oil demand function. The estimation of such a function would help OPEC countries, particularly Saudi Arabia, Venezuela and Nigeria in formulating their production policies and making sound decisions regarding their respective productive capacity. The purpose of this short paper is to proffer an estimation of the US long-run oil demand function. The rest of the paper is made up of two sections. Section 2 discusses the econometric procedure, data, empirical results and implications for OPEC, and Section 3 concludes the paper.
2. Econometric methodology and results The performance of any economy is measured in terms of the growth rate of its gross domestic product ŽGDP.. In general, the US GDP appears to play a major role as a long-term determinant of its demand for crude oil, including foreign oil. In other words, GDP can be used together with oil price to form a long-run relationship to estimate the quantity of oil demand. Furthermore, the deterioration of crude oil prices in 1998 was attributed in part to the accumulated stocks of crude oil reserve throughout the world, and the subsequent price improvement in 1999 was due in part to the decline of these stocks. This linkage of oil prices to the size of oil inventory gives credence to the important role being played by the US Strategic Petroleum Reserve ŽSR.. Since its inception in 1977, the US government has resorted to SR to exert downward pressure on US oil demand and prices. Therefore, the change in SR will be used as an exogenous variable to determine the change in US oil imports. In this paper, we used the annual data for the period 1974᎐1998 to determine the US demand for crude oil imports. These data were extracted from different sources including several issues of the US Economic Report of the President, Annual Energy Review, Statistical Abstract of the US and Oil and Gas Journal. The variables included in the data are defined as follows: Z s the natural logarithm of the real price per barrel Žbbl. of imported crude oil. This price was measured by the US Energy Information Administration in 1992 dollars. Y s the natural logarithm of the US real GDP. This GDP was measured in 1992 dollars. X s the natural logarithm of the US crude oil imports measured for every year as a daily average of imported barrels. SR s The US annual strategic petroleum reserve, where every value of this reserve is reported at the end of the year. Furthermore, since this reserve
I. Jabir r Energy Economics 23 (2001) 659᎐666
661
was formed first in 1977, the values of SR prior to this year were all considered to be zero. To extract the targeted demand equation, a vector error correction ŽVEC. model is used. This model requires testing for cointegration between the time series of the three endogenous variables X, Y and Z, which are in general non-stationary. It is statistically necessary to perform the cointegration test to discern a long-run relationship between X, Y and Z. The maximum likelihood procedure, developed in Johansen Ž1988., Johansen and Juselius Ž1990. and Johansen Ž1992., will be employed to test for cointegration. The test will be referred to hereafter as Johansen’s procedure. The computer package used to apply Johansen’s procedure is E-views, version 3.1. Following this procedure, we will initially test for normality of the residuals of ⌬ X, ⌬Y and ⌬ Z equations that form the VEC model. Considering the lag length k as k s 1, the normality assumption is tested by using the Jarque and Bera Ž1980. test. The results are reported in Table 1, which shows that all the residuals meet the normality assumption. This proves that Johansen’s procedure is quite applicable in finding a long-run relationship between the variables X, Y and Z. We discuss next how to determine jointly both the cointegration rank r for the three endogenous variables, and whether or not there is a linear trend in the model. In other words, which value of r should one use, and is Hr or HrU a better description of the data. The hypothesis Hr of at most r cointegrating relations implies the presence of linear trend, while the hypothesis HrU of at most r cointegrating relations implies the absence of linear trend. HrU is more restricted than Hr , and it is also considered as a sub-hypothesis of Hr . According to Johansen Ž1992., Hr can only be rejected if HrU is also rejected. Let Tr denote the likelihood ratio test statistic for the hypothesis Hr , and let TrU denote the likelihood ratio test statistic of the hypothesis HrU . The resulting values of TrU and Tr , as well as the 95% quantiles Žthe tabulated values. cUr Ž5%. and c r Ž5%., are shown in Table 2. Reading Table 2 from left to right and from top to bottom, it can be noted that H0U , H0 and H1U have to be rejected at the 5% level, but H1 cannot be rejected. Thus, it can be concluded that r s 1, and hence, there is only one cointegrating equation which represents the long-run relationship between the variables X, Y and Z. Additionally, the results show that there is a linear trend in the data of the three variables.
Table 1 Normality assumption test
⌬ Xt ⌬Yt ⌬Zt
2N Ž2.
Probability
Number of observations
Skewness
Kurtosis
2.226 0.567 1.120
0.328 0.753 0.571
23 23 23
0.752 0.384 y0.538
3.241 2.967 3.109
2N Ž2. is the Jarque and Bera normality test statistic.
I. Jabir r Energy Economics 23 (2001) 659᎐666
662
Table 2 The resulting values of TrU and Tr p-r
r
Tr
U
U cr Ž5%.
Tr
cr Ž5%.
3 2 1
0 1 2
58.90 24.32 9.79
34.91 19.96 9.24
44.39 11.09 0.08
29.68 15.41 3.76
p is the number of variables.
Johansen’s procedure is used here to test the hypothesis of the number of cointegrated equations, which will enable us to build the VEC model. The results of the cointegration test are given in Table 3. As shown in Table 3, there is at most one cointegration equation from which the estimated long-run US demand relationship for crude oil imports is obtained. This relationship is given in Eq. Ž1. below: X s 4.7343 q 0.80507 Y y 0.945298 Z Ž y2.003. Ž 7.1976.
Ž1.
The values of the t-statistics are reported in parentheses. The estimated Eq. Ž1. indicates that all the coefficients are statistically significant and have the correct signs. The long-run income and price elasticities of demand are 0.805 and y0.945, respectively. Clearly, the income elasticity is positive but its value is less than one. Compared to the newly industrialized countries such as the so-called Asian tigers, the US has a relatively low income elasticity. This implies two conclusions for OPEC nations to closely consider. First, the US has been increasing its dependence on some other sources of energy relative to crude oil imports. Second, the newly industrialized countries in Asia are expected to be a more promising market than the US for some OPEC members such as: the United Arab Emirates; Qatar; and Kuwait. The price elasticity indicated in Eq. Ž1. is relatively high. This can be attributed again to the reduction in US dependence on oil imports, which happened possibly because of technology enhancement that improved energy consumption efficiency in the US. The next step is to incorporate the resulting long-run relationship of Eq. Ž1. into Table 3 Cointegration test Eigenvalue
Likelihood ratio
5% Critical value
1% Critical value
Number of cointegrated equationŽs.
0.76493 0.38018 0.00365
44.386 11.086 0.084
29.68 15.41 3.76
35.65 20.04 6.65
Nonea At most 1 At most 2
a
Denotes rejection of the hypothesis at the 1% significance level.
I. Jabir r Energy Economics 23 (2001) 659᎐666
663
a vector autoregressive ŽVAR. model which will be an important ingredient in building the VEC model. As mentioned previously, the variables X, Y and Z are taken as endogenous variables while the stock of the US Strategic Petroleum Reserve SR, lagged one period, is treated as an exogenous variable. SR t is the stock of the reserve reported at the end of year t. Logically, the rise Žfall. in X during year t must follow the fall Žrise. in SR measured at the end of year t-1. Hence, the coefficient of SR ty1 is expected to have a negative sign in the equation of ⌬ X t of the VEC model. The estimated VEC model is specified in Table 4. The general formula for this model is: ⌬ X t s ⌫ ⌬ X ty1 q ⌸ X ty1 q q ␣ SR ty1 q t , Ž t s 1, . . . . . . ,25.
Ž2.
where ⌬ X t , ⌬ X ty1 , X ty1 , , ␣ and t are Ž3 = 1. vectors, while ⌫ and ⌸ are Ž3 = 3. matrices. The column of ⌬ X t in Table 4 shows that there are some estimated coefficients that are not satisfactory. The estimated coefficient of ⌬Yty1 is not statistically significant, and the estimated coefficient of ⌬ Zty1 does not have the expected negative sign. Therefore, an ordinary least squares ŽOLS. model was built to come out with the final reliable demand equation, using the same error correction variable ŽECV., lagged one period, taken from Eq. Ž1.. This error correction variable is: ECVty1 s X ty1 y 0.805 Yty1 q 0.945 Zty1 y 4.734
Ž3.
Table 4 Estimated vector error correction model
Cointegration Equation ⌬ Xty 1 ⌬Yty 1 ⌬ Zty 1 Constant SRty 1 R-squared S.E. of regression
⌬ Xt
⌬Yt
⌬Zt
y0.7989 Žy5.398. y0.3324 Žy1.644. 0.1483 Ž0.179. 0.2140 Ž2.116. 0.4892 Ž5.102. y0.00114 Žy4.921. 0.768 0.0709
y0.0909 Žy2.524. y0.0557 Žy1.132. 0.0834 Ž0.414. 0.0177 Ž0.720. 0.0896 Ž3.840. y0.00016 Žy2.792. 0.415 0.0172
0.09796 Ž0.182. 0.5037 Ž0.685. y0.3691 Žy0.123. y0.0590 Žy0.161. y0.0194 Žy0.055. y0.00011 Žy0.130. 0.106 0.2577
The values of the t-statistics are given in parentheses. Sample Žadjusted.: 1976᎐1998. Included observations: 23 after adjusting end points.
I. Jabir r Energy Economics 23 (2001) 659᎐666
664
The resulting OLS model is: ⌬ X t s 0.238 y 0.451 Ž X ty1 y 0.805 Yty1 q 0.945 Zty1 y 4.734. Ž 4.974.
Ž y7.735.
q1.127⌬Yt y 0.174⌬ Zt y 0.000645 SR ty1 Ž 1.724.
Ž y3.067.
Ž y6.226.
Ž4.
R-squared s 0.818, S.E. of regression s 0.0609, D-W statistic s 2.388. The values of the t-statistics are reported in parentheses. It is shown that, with the OLS, there has been a considerable improvement in the values of R-squared and the standard error of regression in Eq. Ž4.. The non-zero estimate 0.451 of the adjustment parameter is highly significant, and hence, there is a quick adjustment by reverting to equilibrium between the levels of the variables. Theoretically speaking, even though unit roots caused a non-stationary problem to the data, transforming the variables to be IŽ0. has solved the problem and consequently generated a demand equation which is more plausible. IŽ0. means that the variables are integrated of order 0. Eq. Ž4. shows also that the estimate 0.000645 of the coefficient of SR is significant in determining the level of US oil imports. Yet, OPEC members should not rely on the potential reduction in the US SR to help them bolster the oil price. They should rather attach great importance to cooperation among themselves to restrict oil output. The output restriction should also be coordinated in concerted efforts with non-OPEC producers. The US is considered as a consumer and producer at the same time in the world crude oil market, hence, such coordination will be helpful to US oil and gas multinational corporations. A target price range of $23᎐26 a barrel would benefit OPEC members, as well as the US multinational corporations and consumers. OPEC members do not have vested interest in an increase in oil price beyond $26, as they will again start losing part of their market share. It will then be quite lucrative for non-OPEC nations to produce oil, and technology will feasibly be developed to replace oil and gas as sources of energy. According to Eq. Ž4., a one million barrel decrease in the SR stock reported by the end of year t y 1 causes an increase of 0.000645 in the natural logarithm of the ratio of oil imports in year t relative to that in year t y 1. In other words, there is about 0.1% increase in the level of oil imports of any year over the previous one as a result of the reduction of one million barrels in the SR already reported. Furthermore, as a result of a one million barrel increase in this stock of reserve of year t-1, the average price in the successive year is offset only by 0.4%. According to this analysis, the stock of SR is not a highly effective tool if it is to enable the US government to influence prices. Hence, the only two purposes for which it can be maintained is to reduce the macroeconomic damage that can occur in a crisis or to make oil easily available to some disadvantaged groups. Eq. Ž4. indicates that short-run income and price elasticities of demand are 1.127
I. Jabir r Energy Economics 23 (2001) 659᎐666
665
and y0.174, respectively. As expected, price elasticity is small and it gets higher after some time. In the short-run, the US oil imports tend to respond faster to changes of income than changes of price. This is due to the consumption nature of the US economy. Crude oil is more broadly defined than any petroleum product considered separately, such as gasoline. It is natural that gasoline has a higher income elasticity of demand in the long run than the short-run. But this does not have to be necessarily true for crude oil. As mentioned before, higher income can provide an access to some other sources of energy. Implicitly this means less dependence on crude oil, including the imported part, compared to the other sources. Therefore, with rising US GDP, more imported oil is demanded, but proportionately less over time. The prices of crude oil declined sharply throughout the whole world during 1986. Therefore, it is required to use this year to test Eq. Ž4. for stability. The breakpoint test by Chow Ž1960. is used in this paper to test for the stability of the estimated coefficients in the resulting demand equation, considering the 5% significance level. Chow test statistic is distributed here as an F distribution with 5 and 15 degrees of freedom under the null hypothesis that all the observations prior to and after 1986 belong to the same regression model, which means, in other words, that there is no structural change in the data. It was found that the calculated value of the test statistic is 2.306, which is less than the tabulated value of 2.9. This leads to the conclusion that the null hypothesis cannot be rejected.
3. Conclusions This paper uses a set of data for the already defined variables X, Y and Z. The time series of each of these variables was found non-stationary and IŽ1.. But differencing each series delivered a stationary one which is IŽ0.. In addition, the empirical results show that there is a linear trend in the data of X, Y and Z. It was also found that there is only one cointegration equation which represents the long-run US demand relationship for crude oil imports. Finally, a reliable stable demand equation was constructed through OLS using the long-run relationship between X, Y and Z. This demand model explained the growth rate in X in terms of: Ži. the past disequilibrium between the levels of X, Y and Z which is the long-run relationship lagged one period; Žii. the variable SR lagged one period; and Žiii. the growth in Y and Z. Investigating the US demand for crude oil imports in this paper has proved that OPEC’s ability to maintain prices depends on the coherence Žor cohesion. of its members to their assigned quotas and cooperation with non-OPEC producers in order to alleviate their competitive behavior. This is much more important than changes of the level of the strategic petroleum reserve possessed by the US. This paper concentrates also on the point that the US will remain a very important customer for OPEC even though there are recently some new emerging considerable customers.
666
I. Jabir r Energy Economics 23 (2001) 659᎐666
Energy Information Administration, Hendry and Juselius, 2000; Johansen, 1991; Oil Gas Journal Osterwald-Lenum, 1992; US Census Bureau, US Government, 1997; Wilson, 1998
Acknowledgements The author would like to thank anonymous reviewers of this journal for their helpful suggestions, and Professor Musa Essayyad, co-editor of this journal, for his editorial help. The author would also like to thank Professors Fabrizio Carlevaro, Jean-Paul Chaze, and Jaya Krishnakumar for their feedback during his summer 1999 visit to CUEPE of the University of Geneva, Switzerland.
References Chow, G.C., 1960. Tests of equality between sets of coefficients in two linear regressions. Econometrica 28 Ž3., 591᎐605. Jarque, C.M., Bera., A.K., 1980. Efficient tests for normality, homoscedasticity and serial independence of regression residuals. Econ. Lett. 6, 255᎐259. Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dyn. Control 12, 231᎐254. Johansen, S., Juselius., K., 1990. Maximum likelihood estimation and inference on cointegration ᎏ with applications to the demand for money. Oxf. Bull. Econ. Statistics 52 Ž2., 169᎐210. Johansen, S., 1992. Determination of cointegration rank in the presence of a linear trend. Oxf. Bull. Econ. Statistics 54 Ž3., 383᎐397.
Further reading Energy Information Administration, U.S.A. Annual Energy Review, several issues. Hendry, D.F., Juselius., K., 2000. Explaining cointegration analysis: Part 1. Energy J. 21 Ž1., 1᎐42. Johansen, S., 1991. Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models. Econometrica 59 Ž6., 1551᎐1580. Oil Gas J., U.S.A., several issues. Osterwald-Lenum, M., 1992. A note with quantiles of the asymptotic distribution of the maximum likelihood cointegration rank test statistics. Oxf. Bull. Econ. Statistics 54 Ž3., 461᎐472. US Census Bureau, the official statistics. Statistical Abstract of the United States, several issues. US Government, 1997. Economic Report of the President. Wilson, R., 1998. Economic Development in the Middle East. Routledge, New York.