Available online at www.sciencedirect.com ScienceDirect Energy Procedia 100 (2016) 480 – 483 3rd International Conference on Power and Energy Systems Engineering, CPESE 2016, 8-12 September 2016, Kitakyushu, Japan Relationships between Natural Gas Production in Persian Gulf States and Natural Gas Consumption in the European Union Gaolu Zoua,*, Dingsheng Fengb, K W Chauc a School of Tourism and Economic Management, Chengdu University, Chengdu 610106, China b Investment Company, Sichuan Normal University, Chengdu 610066, China c The Ronald Coase Center for Property Rights Research, Faculty of Architecture, The University of Hong Kong, Hong Kong Abstract Qatar, Saudi Arabia, United Arab Emirates, Iraq and Kuwait hold abundant natural gas reserves This study examines the longrun and short-run relationships between natural gas production in the five Gulf states and consumption in the European Union (EU) The data consist of yearly time series covering the period from 1970 to 2012 We tested for cointegration and the Granger causality in a first-differenced VAR The tests did not indicate a long-run equilibrium or any short-run dynamics We suggest that the Gulf states have a huge potential to increase and establish a stable gas supply to the EU © Published by by Elsevier Ltd.Ltd This is an open access article under the CC BY-NC-ND license © 2016 2016The TheAuthors Authors Published Elsevier (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the organizing committee of CPESE 2016 Peer-review under responsibility of the organizing committee of CPESE 2016 Keywords: consumption; Persian Gulf states; natural gas; production; reserves Introduction In 2012, proven natural gas reserves in Qatar accounted for 13.4% of the world’s total reserves (187.3 trillion cubic meters or tcm), ranking third after Iran (18.0%) and Russia (17.6%) [1] Proven gas reserves in Saudi Arabia, the United Arab Emirates, Iraq, and Kuwait constituted 4.4%, 3.3%, 1.9%, and 1.0% of the world’s total reserves, respectively Therefore, excluding Iran, these five Persian Gulf states held 24% of the world’s proven reserves With the deteriorating Ukraine crisis, the EU urgently expects to decrease its heavy reliance on Russian gas [2] It has been suggested that significant potential suppliers include Gulf states, e.g [3] This study investigates the long-run and short-run relationships between natural gas production in the five Persian Gulf states and consumption in the EU in order to provide more evidence for the substitution of gas suppliers Methodology We estimate the Johansen trace statistics and cointegrating vector(s) [4] The Phillips-Ouliaris test can provide clues to the cointegration [5] The study tests for unit root using both the Augmented Dickey-Fuller (ADF) and * Corresponding author Tel.: +86-28-84617900; fax: 020-28819702 ext.12191 E-mail address: zougaolu@vip.163.com 1876-6102 © 2016 The Authors Published by Elsevier Ltd This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the organizing committee of CPESE 2016 doi:10.1016/j.egypro.2016.10.206 481 Gaolu Zou et al / Energy Procedia 100 (2016) 480 – 483 Phillips-Perron (PP) techniques Moreover, Perron structural break tests are conducted using the mixed innovational outlier (IO) model C [6] If series are cointegrated, an error-correction model (ECM) can be used to represent the long-run relationship [7] An Engle-type ECM in first differences is formulated as follows: p1 'yt p2 (1) O ¦ D i 'yt i ¦ E j 'xt j I z t 1 H t , i j where z t 1 is a cointegrating vector or error-correction (EC) term However, if the series are integrated of order one but not cointegrated, we estimate a traditional vector autoregressive (VAR) model in first differences by removing the EC term from Equation (1) In the ECM or VAR model, short-run and/or long-run effects in terms of elasticity and Granger causality can be determined [8] Wald-Ȥ2 statistics are estimated for the null hypothesis of no Granger causality from x to y: H0 : ȕ j (2) Data The data were gathered from BP [1] Natural gas production (PRODUCTION) represents the total natural gas production of Qatar, Saudi Arabia, United Arab Emirates, Iraq, and Kuwait, while natural gas consumption (CONSUMPTION) represents the total natural gas consumption in the EU The data were transformed into natural logarithms before they were used to conduct empirical tests and were presented as yearly series covering the period from 1970 to 2012 Natural gas production and consumption seemed to trend together since the mid-1990s (Fig 1) LOG PRODUCTION LOG CONSUMPTION 1970 1975 1980 1985 1990 1995 2000 2005 2010 Fig Natural gas production and consumption in logarithms Empirical Results Table Unit root tests ADF PP Statistic (k, p-value) Statistic (k, p-value) Level -2.32(2, 0.41) -1.65(2, 0.75) First difference -3.45(2, 0.06) -6.42(2, 0.00) CONSUMPTION Level -1.22(2, 0.89) -4.58(3, 0.00) First difference -3.36(2, 0.07) Notes: For ADF tests, we selected the lag length k using SIC; for PP tests, the Newey-West method was used [9] However, the number of lags was set between two and eight on a general-to-specific principle [10] MacKinnon one-sided p-values were used [11] Log variable PRODUCTION For the variable PRODUCTION, ADF and PP tests consistently suggested one unit root For the variable CONSUMPTION, the ADF test suggested one unit root, but the PP test suggested no unit roots Moreover, we did not detect a break date in the data (Table 2) Hence, we considered these two variables as I(1) series based on the recommendation in [12] Table Structural break tests Log Variable Į t Į* p-value Lag length Tˆb PRODUCTION -0.02 -0.05 0.96 1990 CONSUMPTION 0.85 4.44 0.00 1984 Notes: We chose lag length between two and nine on a general-to-specific basis [10] The t-statistic in absolute value was above or equal to 1.8 Tˆ was the possible break date detected The break date was selected between 1984 and 2000 One-sided test critical values for the sample size of b 70 were -6.32, -5.59, and -5.29 at the 1%, 5%, and 10% significance levels, respectively [6] 482 Gaolu Zou et al / Energy Procedia 100 (2016) 480 – 483 We used the Akaike information criterion (AIC) to select the optimal model for the Johansen trace tests Fig indicates that the data contain the intercept and trend Hence, we focused on the examination of Model based on [13] When lags were two and four, we obtained lower AIC; however, when lags were four, the cointegrating equation was autocorrelated Hence, we selected Model with a lag length of two for the trace test Moreover, the Johansen-type ECM satisfied the criterion of multivariate normality The tests did not detect a cointegrating vector (Table 3) Furthermore, the residual-based tests suggested no cointegration (Table 4) Hence, we estimated a firstdifferenced VAR, in which the tests did not detect any short-run Granger causality between PRODUCTION and CONSUMPTION (Table 5) Table The Johansen multivariate cointegration trace tests k r Trace Asymptotical critical value* p-value** C&L*** 23.1 25.9 0.11 28.3 ̰1 3.7 12.5 0.78 13.7 Notes: *5% asymptotical critical value in [14] **p-value in [15] ***Cheung-Lai finite-sample critical value [16] Table Phillips-Ouliaris residual-based cointegration tests ZĮ-statistic Dependent variable p-value* Log CONSUMPTION -13.1 0.21 Log PRODUCTION -14.5 0.15 Notes: Null hypothesis was that series were not cointegrated *p-value in [11] Table Estimates of first-differenced VAR and Granger causality tests Dependent: Time CONSUMPTION CONSUMPTION PRODUCTION Constant Estimate (t-statistic) t-1 -3.23 (-1.09) t-2 0.63 (0.26) t-1 18.71 (1.05) t-2 -2.05 (-0.15) Multivariate normality (p-value)* 0.54 (0.97) Q-statistic (p-value, lags)** 4.25 (0.37, 3) Granger causality (p-value) For excluding PRODUCTION: 2.66 (0.26) For excluding CONSUMPTION: 2.53 (0.28) 0.01 (1.22) Notes: Data were converted to logarithmic form Appropriate lags depends on data [10] Tests made AIC as small as possible VAR satisfied multivariate normality and removed autocorrelations *Jarque-Bera statistic based on Cholesky factorization matrix **Portmanteau autocorrelation adjusted Q-statistic R-squared: 0.41 Adj R-squared: 0.34 F-statistic: 6.18 Akaike AIC: -3.64 Concluding remarks The five Persian Gulf states (Qatar, Saudi Arabia, United Arab Emirates, Iraq, and Kuwait) hold about a quarter of the world’s proven natural gas reserves The EU has shown an increasing reliance on external gas supplies, such as those from the Gulf states Hence, this study investigates the long-run and short-run relationships between natural gas production in the five Persian Gulf states and natural gas consumption in the EU However, the cointegration tests did not indicate a long-run equilibrium, and the granger causality tests did not detect any short-run dynamics Therefore, the gas production showed neither long-run nor short-run effects on the EU’s gas consumption We argue that gas imports in the EU from the Gulf states were small In particular, in 2012, Europe and Eurasia constituted only 29.5% of Qatar’s liquefied natural gas exports [1], which may undermine the potential long-run and short-run effects An enormous potential exists for the Gulf states to increase their gas supply to the EU References [1] BP Statistical review of world energy 2013 http://www.bp.com/statisticalreview [2] Söderbergh B, Jakobsson K, Aleklett K European energy security: An analysis of future russian natural gas production and exports Energ Policy 2010; 38: 7827-7843 [3] Bilgin M Geopolitics of european natural gas demand: Supplies from russia, caspian and the middle east Energ Policy 2009; 37: 4482-4492 Gaolu Zou et al / Energy Procedia 100 (2016) 480 – 483 [4] Johansen S, Juselius K Maximum likelihood estimation and inference on cointegration with applications to the demand for money Oxford Bull Econ Statist 1990; 52: 169-210 [5] Phillips PCB, Ouliaris S Asymptotic properties of residual based tests for cointegration Econometrica 1990; 58: 165-193 [6] Perron P Further evidence on breaking trend functions in macroeconomic variables J Econometrics 1997; 80: 355-385 [7] Engle RF, Granger CWJ Cointegration and error correction: Representation, estimation and testing Econometrica 1987; 55: 251-276 [8] Granger CWJ Investigating causal relations by econometric models and cross-spectral methods Econometrica 1969; 37: 424-438 [9] Newey WK, West KD A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix Econometrica 1987; 55: 703-708 [10] Ng S, Perron P Lag length selection and the construction of unit root tests with good size and power Econometrica 2001; 69: 1519-1554 [11] MacKinnon JG Numerical distribution functions for unit root and cointegration tests J Appl Econometrics 1996; 11: 601-618 [12] Hendry DF, Juselius K Explaining cointegration analysis: Part i Energy J 2000; 21: 1-42 [13] Hendry DF, Juselius K Explaining cointegration analysis: Part ii Energy J 2001; 22: 75-120 [14] Osterwald-Lenum M A note with quantiles of the asymptotic distribution of the maximum likelihood cointegration rank test statistics Oxford Bull Econ Statist 1992; 54: 461-472 [15] MacKinnon JG, Haug AA, Michelis L Numerical distribution functions of likelihood ratio tests for cointegration J Appl Econometrics 1999; 14: 563-577 [16] Cheung Y-W, Lai KS Finite-sample sizes of johansen's likelihood ratio tests for cointegration Oxford Bull Econ Statist 1993; 55: 313-328 483