Energy-Output Linkages in Australia: Implications for Emissions Reduction Policies

Size: px
Start display at page:

Download "Energy-Output Linkages in Australia: Implications for Emissions Reduction Policies"

Transcription

1 Enegy-Output Linkages in Austalia: Implications fo Emissions Reduction Policies Md. Shahiduzzaman Univesity of Southen Queensland Khoshed Alam Univesity of Southen Queensland The study investigates the long-un and dynamic elationships between enegy consumption and output in Austalia using a multivaiate cointegation and causality famewok. Using both Engle-Gange and Johansen cointegation appoaches, the study finds that enegy consumption and eal Goss Domestic Poduct ae cointegated. The Gange causality tests suggest bidiectional Gange causality between enegy consumption and eal GDP, and Gange endogeineity in the system. Since the enegy secto lagely contibutes to cabon emissions in Austalia, we suggest that diect measues to educe cabon by putting constaints on the enegy consumption would pose significant economic costs fo the Austalian economy. INTRODUCTION In ecent yeas, the issues on the elationships between enegy consumption and output have ganeed inteest due to the consensus on educing geenhouse gases (GHGs) emissions to combat global waming. Among the vaious GHGs implicated, cabon dioxide emissions account fo moe than thee quates of total anthopogenic GHGs emissions. These have gown at a apid ate since pe-industial times, leading to an incease in global aveage ai and ocean tempeatues and a ise in sea levels (IPCC, 2007a, 2007b; McKibbin & Wilcoxen, 2002). These exacebations in climatic conditions ae linked to population and plants by a vaiety of negative consequences (Ganaut, 2008; Pickeing & Owen, 1997; Sten, 2008; Wills, 2006). In fact, GHGs emissions ae lagely attibuted to the use of enegy necessay to poduce goods and sevices, as well as to act as a final good fo end-uses. Any measues to educe GHGs, theefoe, pose implications fo economic gowth, via the linkages with enegy consumption. In Austalia, the enegy secto, compising of stationay enegy, tanspot and fugitive emissions, contibuted to about 70% of total emissions in 2006, ising fom about 52% in 1990 (Depatment of Climate Change, 2008). Enegy elated emissions in the county inceased by about 40% fom 1990 to 2006, wheeas a substantial decline (about 71%) in non-enegy emissions like land use, land use change and foesty occued duing this peiod (Depatment of Climate Change, 2008). Austalia is a top pe capita GHGs emitte in the wold as attibuted to its high-enegy consumption, as well as its eliance upon the fossil fuel as enegy souce. Its pe capita enegy use inceased fom, on aveage, kilotonnes (kt) of oil equivalent duing to kt of oil equivalent duing (Table 1). The 48 Jounal of Applied Business and Economics vol. 12(3) 2011

2 popotion of fossil fuel enegy consumption inceased gadually fom 89.8% in to 94.1% in , and thee was a eduction of the atio of clean enegy poduction to total enegy use duing fom the pevious peiods (Table 1). TABLE 1 ENERGY CONSUMPTION IN AUSTRALIA Enegy use pe capita kg of oil Fossil fuel enegy consumption Clean enegy poduction (% of total Electicity poduction fom coal souces equivalent (% of total) enegy use) Souce: Wold Bank (2009) (% of total) On the othe hand, the histoical alignment of enegy and output/income is vey high in Austalia. Figue 1 shows the five-yea aveages of the gowth ate of its Goss Domestic Poduct (GDP) and enegy use ove the peiod 1965 to As seen in Figue 1, both GDP and enegy use gowth emained pefectly aligned fom 1965 to 1980, along with the veacity that gowth ate enegy use outpaced the gowth of the GDP in the 1980s. Since 1980, while gowth in enegy consumption has educed, it moved with the changes in economic activities quite closely. FIGURE 1 GROWTH RATE OF GDP AND ENERGY CONSUMPTION GDP Gowth ate Souce: Wold Bank, 2009 GDP gowth (annual %) Enegy use gowth (annual aveage %) The cabon exposue of the enegy souces and the enegy intensity of the economy is a geat concen fo Austalia in the context of fomulating domestic policies on emissions eduction. This is because any Jounal of Applied Business and Economics vol. 12(3)

3 policy to educe emissions would hut the ongoing pace of economic gowth though its impact on enegy consumption because of the cabon intensity of the enegy secto. Nonetheless, the potential impact of emissions eduction policies can be diffeent depending on the diection of causality and dynamics between enegy and output in shot- and long-un. As fo example, in the case whee enegy consumption is found to stimulate economic gowth, it can be agued that measues to educe emissions that diectly affect enegy consumption (consevation) would hut economic gowth. Convesely, if the diection of causality uns fom GDP to enegy consumption, then diect measues of emissions eduction though enegy consevation may be implemented with little o no advese impacts on economic gowth. In essence, it is impotant to go beyond the simple knowledge that enegy and output ae inteelated but to undestand the diection of causality between them in the both shot- and long-un. The policy implication of such analysis is that its helps to undestand the pedicted impacts of the vaious enegy consevation and emission eductions policies on economic activities. The pupose of this study is to investigate long-un and dynamic causal elationship in Austalia, using the multivaiate cointegation appoach. Given the gowing concens ove the negative impacts of GHGs emissions and global consensus on the issue, it has now become a pioity agenda fo leading economies to implement domestic tagets on emissions eduction. The findings of the study will be elevant to a numbe of counties, which ae developing emissions eduction policies, but ae concened about the tadeoffs between enegy and output. elationship between enegy consumption and GDP/GNP (Goss National Poduct), and povided a ich set of pespectives and insights; but they failed to povide unanimous esults (Belloumi, 2009; Jobet & Kaanfil, 2007). We povide a moe detailed eview of the liteatue fo diffeent counties in the next section. In the case of Austalia, only thee studies boadly coveed the issues, howeve, they failed to povide unanimous esults. Fatai, Oxley and Scimgeou (2004) suggest the existence of unidiectional causality unning fom eal GDP to enegy consumptions. Naayan and Smyth (2005) find the evidence of a longun (cointegated) elationship between electicity consumption, employment, and eal income in Austalia, whee a long-un causality uns fom employment and eal income to electicity consumption. On the othe hand, Naayan and Pasad (2008) find the evidence of unidiectional causality fom electicity consumption to eal GDP in Austalia. While the studies in the context of Austalia mainly ely on a bivaiate famewok to study the long-un elationship and Gange causality, this study applies a multivaiate pocedue to educe the poblems of omitted vaiables (Akinlo, 2008; Ghali & El-Sakka, 2004; Sten & Cleveland, 2004). It is suggested that bivaiate tests of causality can poduce misleading esults because of the substitution effects that may take place between enegy and othe inputs of poduction, such as capital (Ghali & El-Sakka, 2004; Sten, 2000; Sten & Cleveland, 2004). Moeove, capital investment seems to play an impotant ole duing the pocess of economic upliftment (Jobet & Kaanfil, 2007). Enegy consevative policies may also stimulate fixed investment though the installation of enegy efficient machiney (Thompson & Taylo, 1995). Accodingly, this study investigates the causal elationship between enegy consumption and eal GDP, contolling fo possible affects of goss fixed capital fomation as a poxy of capital accumulation (as used by Soytas & Sai, 2009; Soytas, Sai, & Ewing, 2007), and to captue the effects of omitted vaiables. Theefoe, given the contadictoy esults on the causal elationship between enegy consumption and income in Austalia, this study can be consideed as complementay, whee the new pat of the study intoduces capital accumulation in the model. In addition, this study added data fo the ecent peiods, which would be beneficial in espect of new enegy/envionment policies. Futhemoe, both Engle- Gange (Engle & Gange, 1987) and Johansen s (Johansen, 1991, 1995) methods ae applied to exploe the long-un elationship(s) in a multivaiate famewok. Finally, in ode to pefom a moe compehensive analysis on the diection of causality, Gange causality was examined though thee diffeent channels (shot-un, long-un, and oveall system) using vecto eo coection (VEC) models. As a esult of these advantages, this study povides a unique set of pespectives and insights and contibutes to the gowing body of liteatue on this subject. 50 Jounal of Applied Business and Economics vol. 12(3) 2011

4 The oganization of the pape is as follows: the next section pesents the eview of liteatue; section 3 incopoates desciptions on the methodology and data; section 4 povides analyses and findings of the study; and section 5 explains the conclusions and policy implications. REVIEW OF LITERATURE In the liteatue, we can find cointegation and Gange causality analyses on the elationship between enegy consumption and income/output, but with consideable vaiations in esults. A cointegation test can analyse the existence of a long-un elationship but fails to discove the diection of causality. Theefoe, testing fo both cointegation and Gange causality is impotant fom policy pespective. Fou foms of hypotheses have been identified: (a) unidiectional causality fom enegy to income/output; (b) unidiectional causality fom income/output to enegy; (c) bidiectional causality; and (d) no causality. The existence of a unidiectional causality fom enegy to income/output suggests that enegy consevative polices may advesely impact income/output. On the othe hand, existence of unidiectional causality fom income/output to enegy infes that enegy consevation policies would be undetaken without any advese impacts on income/output. Existence of bidiectional causality suggests that an economic system exhibits feedback, theefoe, shocks to enegy consumption would have diect negative impacts on income/output and feedback impact on its own. Finally, existence of no causality suggests that enegy (income/output) shocks ae neutal to income/output (enegy). Kaft and Kaft (1978) accomplished the pioneeing wok on the causal elationship between enegy and income. The Gange causality tests on a bivaiate model fo the United States (US) fo the peiod 1947 to 1974 suggest evidence of unidiectional causality unning fom GNP to enegy consumption. Unidiectional causality fom GDP to enegy is also found in Fatai, Oxley and Scimgeou (2004) fo New Zealand and Austalia; the same was found by Lise and Van Montfot (2007) and Edal, Edal, and Esengün (2008) fo Tukey; Chiou-Wei, Chen, and Zhu (2008) fo the Philippines and Singapoe; Akinlo, Zhao, & Hu, (2008) fo China. On the othe hand, unidiectional causality fom enegy consumption to GDP was found in Sten (2000) fo the US; Soytas and Sai (2003) fo Tukey, Fance, Gemany, and Japan; Fatai, Oxley and Scimgeou (2004) fo India and Indonesia; and Belloumi (2009) fo Tunisia. Bidiectional causality has also been epoted by Fatai, Oxley and Scimgeou (2004) fo Thailand and the Philippines; Ghali and El- Sakka (2004) fo Canada; Lee and Chang (2005) fo Taiwan; Chiou-Wei, Chen, and Zhu (2008) fo Malaysia and Indonesia; and Akinlo (2008) fo Gambia, Ghana, and Senegal. Some studies also found existence of no causality between the vaiables, suppoting the view of neutality hypothesis (see, fo instance, Jobet and Kaanfil (2007) fo Tukey, and Chiou-Wei, Chen, and Zhu (2008) fo the US, South Koea, and Thailand). Table 2 summaizes some of the ecent time seies studies on enegy consumption and income fo diffeent counties. As can be obseved fom Table 1, seveal diffeences exist with espect to the diection of causality as descibed above. Jobet and Kaanfil (2007) wote an excellent eview of the existing liteatue and suggest that the conflicting esults ae due to diffeences in methodology and sample peiod. Kaanfil (2009) illustated the diffeences in empiical esults on causality between enegy and economic gowth fo India, Tukey, and the US, and suppots the view that estimation esults ae highly sensitive to the methodology used and the time peiod consideed. Edal, Edal, and Esengün (2008) identified the sample peiod as an impotant facto because of the tansfomation of an economy duing the development pocess. Fom the liteatue eview, it is obseved that the diection of causality between enegy and income vaies with espect to time, space, and methodology. Thee is meit in the application of altenative methodologies and the use of ecent available data in the face of fomulating new policies. Jounal of Applied Business and Economics vol. 12(3)

5 TABLE 2 OVERVIEW OF SOME TIME SERIES STUDIES ON ENERGY USE AND INCOME Autho(s) County Peiod Methods Sten (2000) USA Fatai, Oxley and Austalia, 1960 Scimgeou New Zealand 99 (2004) and 4 Asian counties Ghali and El- Sakka (2004) Canada Multivaiate VECM Cointegation & Gange causality Multivaiate VECM Results Cointegation Gange causality EC to GDP fo Austalia No, fo New Zealand GDP to EC, fo Austalia and New Zealand Bidiectional Naayan and Smyth (2005) Jobet and Kaanfil (2007) Lise and Van Montfot (2007) Akinlo (2008) Chiou-Wei, Chen, and Zhu (2008) Edal, Edal, and Esengün (2008) Austalia Tukey Tukey Sub Sahaan 03 counties USA and nine newly industialized counties in Asia Tukey VECM Income to ELC Cointegation & No No causality Gange causality VECM GDP to enegy Cointegation & Gange causality Linea & nonlinea Gange causality Cointegation & Gange causality Mixed esults fo othe counties Bidiectional(3); GDP to EC (3); No causality (5) No causality fo USA; Mixed esults fo othe counties Bidiectional Naayan and Pasad (2008) Zhao, & Hu, (2008) 30 OECD counties including Austalia China Belloumi (2009) Tunisia Bootstap appoach Cointegation, VECM & Gange causality - ELC to GDP (8) GDP to ELC (6) GDP to EC in the shot un VECM Bidiectional in long-un; EC to GDP in shotun Notes: EC and ELC epesent final enegy consumption electicity consumption, espectively. Numbe in a paenthesis is the numbe of counties in the categoy. VECM stands fo Vecto Eo Coection Model. 52 Jounal of Applied Business and Economics vol. 12(3) 2011

6 METHODOLOGY Vaiables and Data One poblem egading the pai wise Gange causality test is that it can poduce misleading esults when both seies ae caused by a common thid vaiable. To solve this poblem, we applied a multivaiate appoach so that the tue elationship between enegy consumption and eal GDP could be investigated. Specifically, ou model consists of thee vaiables; total enegy consumption (eg), eal GDP (y), and capital accumulation (ks). As mentioned in the intoduction, goss fixed capital fomation is a eliable poxy fo capital accumulation. All vaiables ae in the fom of natual logaithm; theefoe, thei fist diffeences appoximate the gowth ates. The annual model is estimated with time seies data spanning fom 1965 to The annual time seies data wee collected fom the Austalian Bueau of Statistics (ABS) and Wold Development Indicatos CD ROM 2009 (Wold Bank, 2009). Data fo the GDP chain pice index and the goss fixed capital fomation wee collected fom the ABS, while data fo total enegy consumption was collected fom the Wold Bank (2009). Popeties of Data and Gange Causality A majo shotcoming of the ealy liteatue on causality test between income and enegy elationship is the failue to captue the stationay popety of data (Ghali & El-Sakka, 2004; D. Sten & Cleveland, 2004). This is impotant because Gange causality implies that the test cannot be pefomed in case of non-stationay seies. A seies is called non-stationay if its distibution shifts ove time. Accodingly, if a non-stationay seies needs to be diffeenced d times in ode to find a stationay fom; the seies is called integated of ode d, that is, I (d). While non-stationay seies have to be diffeenced to find a stationay seies, a majo poblem of diffeencing is the loss of infomation egading seies levels. Engle and Gange (1987) devised a way to un a model with non-stationay vaiables in level fom, populaly known as the eo coection model. Engle and Gange (1987) postulated that a linea combination of two o moe non-stationay seies might be stationay if they have the same ode of integation. Futhemoe, if such a linea combination exits, the seies ae cointegated implying long-un elationships. Following Gange (1988) and Engle and Gange (1987), VEC specifications fo the set vaiables in this study can be witten as follows: yt 1 1, iyt i 1, iegti 1, iksti v1 t1 1, t (1) eg t 2 y 2, i ti eg 2, i ti ks 2, i ti v 2 t1 3 3yt i 3, iegti 3, i 3ksti v3 t1 3 t ks, In equations (1) to (3), the paametes v 1, v 2, and v 3 epesent the adjustment coefficients, and t-1 ae the cointegation vectos deived fom the long-un elationships (in level). Using the VEC models, Gange causality tests between the vaiables can be investigated by thee diffeent channels: (i) A Wald coefficient test o joint F-test applied to the coefficients of each explanatoy vaiable. Fo example, to investigate whethe causality (shot-un) uns fom enegy consumption to GDP, we test the null hypotheses whethe 11 = 11 =. = 1p =0. (ii) (iii) Statistical significance of lagged eo coection tems to investigate the long-un causality. A joint F-test o Wald coefficient test applied jointly to the sum of the lagged dynamic tems and lagged eo coection tems to investigate the Gange endogeneity o system causality. 3, t (2) (3) Jounal of Applied Business and Economics vol. 12(3)

7 ANALYSIS AND FINDINGS Unit Root Test Given the impotance of the stationay popety of data fo identifying the causal elationship, thee diffeent tests ae applied to find obust esults. These tests ae Augmented Dicky-Fulle (ADF), Dicky- Fulle GLS (DF-GLS), and Phillips-Peon (PP). The DF-GLS test is implemented along with the conventional DF, as the fome pefoms bette in the case of small sample sizes and powe (Elliott, Thomas, & Stock, 1996). The null hypothesis fo all tests indicates that the seies has a unit oot (nonstationay) in the level foms. We applied the Akaike Infomation Citeion (AIC) (Akaike, 1969) to select the lag stuctue fo the ADF and DF-GLS, while the bandwidth fo the PP test was selected with the Newey-West Batlett Kenel. Six was chosen as the maximum lag length. Table 3 epots the unit oot test esults fo the vaiables, togethe with the optimal lag lengths epoted in the paentheses. The battey of unit oot tests pesented in the table almost unanimously indicates that all the vaiables ae non-stationay in level (with and without tend) and stationay in the fist diffeence fom. This indicates that the integation of the vaiables in the study is of ode one, e.g. I (1). TABLE 3 UNIT ROOT TEST RESULTS ADF DFGLS PP Vaiables Level Fist diffeence Level Fist diffeence Level Fist diffeence y (0) -5.95* (0) 1.09(2) -2.44^(1) * eg (1) -8.37* (0) 0.62(3) -2.03^(2) * ks (0) -6.01* (0) 2.14(0) -5.29*(0) * Note: each test uses an intecept and tend when they ae significant. *,^, and denote significance at the 1%, 5% and 10% citical levels, espectively. Cointegation Tests The stationay popeties of the vaiables allow us to exploit the infomation content of the data in level fom using cointegation theoy. Cointegation theoy suggests that a linea combination of two o moe non-stationay seies would be stationay (Engle & Gange, 1987). If this is the case, then a long un elationship among the vaiables can be established. This futhe leads to the conclusion that causality exits among the vaiables, even though the diection of causality is not pecisely detemined. Testing a linea combination of two o moe non-stationay seies involves unning a egession equation in level and then checking whethe the esidual is stationay. The estimation of the long un elationships (t-statistics in the paenthesis) ae: yt egt 0.39kst 1 t (2.35) (18.87) (13.82) egt yt 0.40kst 2 (-0.59) (18.87) (-6.88) kst egt 2.14yt 3 (-4.27) (-6.89) (13.82) t t (4) (5) (6) Equations (4) to (6), it epesent the estimated esiduals fom the equilibium egessions. Now, we test whethe the estimated esiduals fom the estimated equations ae stationay o not. Fo this pupose, we use the Dickey-Fulle (DF) unit oot tests (Endes, 2004). The actual it is not obsevable; theefoe, it 54 Jounal of Applied Business and Economics vol. 12(3) 2011

8 will be inappopiate to use the usual DF citical values to measue the significance levels of the estimated coefficients. We follow Mackinnon s (1991) Response Suface Estimation pocedue to find the citical values fo the cointegation test. The DF test esults indicate that the estimated esiduals fom all thee equations ae stationay at the 10% level. As the null fo the DF test is no cointegation, it is not possible to eject the hypothesis that the vaiables ae not cointegated (Engle & Gange, 1987). It can now be concluded that that cointegating elationship(s) of the ode (1,1) exist among the set of vaiables in the study. One poblem with the above Engle and Gange (1987) method of the cointegation test is that it ignoes the possibility of the existence of two o moe cointegation vectos. This is paticulaly impotant as the model involves moe than two vaiables. To exploe this possibility, we employ Johansen s (1988, 1995) pocedues to check whethe the above esults egading cointegation fom Engle and Gange (1987) ae consistent with the Johansen method. In the Johansen cointegation test, the pesence and numbe of cointegation vecto(s) is identified using two diffeent likelihood atio tests one is based on tace statistics and the othe on maximum eigenvalue. Howeve, befoe going to the cointegation test, it is necessay to un an unesticted VAR (vecto autoegession) to detemine the maximum lag length because the Johansen appoach is sensitive to the lag length. A maximum lag length of fou in the estimated VAR suggests that the esidual in the model is out of any seial coelation. We then apply a numbe of lag length citeion, such as a LR test statistic (LR), a final pediction eo (FPE), Akaike infomation citeion (AIC), Schwaz infomation citeion (SIC), and Hannan-Quinn infomation citeion (HQ) though the unesticted VAR to detemine the optimal leg length fo the Johansen cointegation test. All citeions unanimously suggest no lag fo the cointegation test. As thee is no pioy to include tend in the long-un equation, and given the visual plot of the data, we only allow a linea tend in data but intecept (no tend) in the long-un equation and the VAR. As shown in Table 4, both tace and maximum eigenvalue statistics indicate the pesence of one cointegating equation at the 5% level. TABLE 4 COINTEGRATION TEST RESULTS Tace Ho Eigenvalue Statistic 5% Citical Value Pob.** = 0 * Maximum Eigenvalue H o Eigenvalue Statistic 5% Citical Value Pob.** = 0 * Tace test indicates one cointegating equation at the 5% level Max-eigenvalue test indicates one cointegating equation at the 0.05 level * denotes ejection of the hypothesis at the 0.05 level **MacKinnon (1999) p-values Gange Causality Tests While the cointegation esults suggest a long-un elationship, they do not indicate the diection of causality. Theefoe, Gange causality test between y and eg is pefomed in equations (1) to (3), whee t-1 = y t eg t ks t-1 is the nomalized long-un elationship to y. Note that in the cointegation system, one vaiable does not Gange cause anothe if the lagged value of the fist diffeence fom of the fome does not ente in the eo coection model of the latte. Fo example, eg Jounal of Applied Business and Economics vol. 12(3)

9 does not Gange cause y, if the lagged value eg does not ente equation (1). Befoe testing the Gange causality, it is necessay to pefom a numbe of diagnostic tests to validate standad assumptions. Seial coelation LM tests asset no seial coelation in the esidual and CUSUM (Cumulative Sum of Recusive Residuals) and CUSUM squae tests validate the paamete stability. The optimal lags ae chosen based on minimum AIC. The CUSUM and CUSUM Squaes test esults epoted in Figues 2, 3 and 4 suggest that the stability of the paametes is not ejected fo all thee equations. Finally, the existence of Gange causality can be tested using a Wald coefficient test. Table 5 epots the Gange causality test esults. FIGURE 2 CUSUM AND CUSUM SQUARE TESTS FOR EQUATION 1 FIGURE 3 CUSUM AND CUSUM SQUARE TESTS FOR EQUATION 2 FIGURE 4 CUSUM AND CUSUM SQUARE TESTS FOR EQUATION 3 56 Jounal of Applied Business and Economics vol. 12(3) 2011

10 TABLE 5 TEST OF GRANGER CAUSALITY Equation Shot-un Long-un Eo coection EC DGP EC EG EC GCF F-statistics T-statistics Joint F-statistics b 3.17 b a b 4.75 b a b 0.28 c 4.20 b a b a 4.45 b 4.76 b - Note: supescipts a, b and c denote significance at 1%, 5% and 10% citical level espectively. Some impotant obsevations can be made fom the estimation esults. Fist, eo coection tems ae significant in all thee equations, suggesting convegence to long-un equilibium wheneve thee is a deviation fom the cointegating elationship. Second, thee is significant shot-un bidiectional causality between enegy consumption and output in Austalia. Regading the shot-un Gange causality between enegy consumption and capital fomation enegy consumption does not Gange cause capital fomation, but capital fomation Gange causes enegy consumption. The eo coection tems in equations 1 and 3 equations ae significant at the 1% level, while the eo coection tem in equation 2 is significant at the 10% level. Finally, the joint significance of the sum of the lags of the explanatoy vaiable and the eo-coection tem ae used to test the oveall causality in the system. Testing the joint significance of the lagged dynamic tems of an explanatoy vaiable and eo-coection tem is consideed a stong endogeneity test as it is moe estictive than a single vaiable. Test esults suggest the existence of Gange endogeneity. As such, bidiectional Gange causality uns between enegy consumption and eal GDP in Austalia in both shot- and long-un. CONCLUSION AND POLICY IMPLICATION This study investigated the linkages between enegy consumption and output/income in Austalia using multivaiate cointegation and causality analyses duing the peiod of Empiically, while many of the ealie studies have investigated the elationship, the esults ae at best mixed. Moeove, thee ae few studies in the Austalian context and the majoity use a bivaiate famewok to exploe cointegation and Gange causality. None of them consideed capital accumulation as a contol vaiable. Based on the pediction that both enegy consumption and capital fomation play impotant oles in the economic upliftment pocess, and given the possibility that enegy policy may also affect pocuement of enegy efficient machiney, the study contols fo goss fixed capital fomation. The study also used the latest available data and employs two altenative methods to exploe the long-un elationships between enegy consumption and eal GDP. VEC models investigated the diection of causality though thee diffeent channels: shot-un, long-un, and oveall system. The cointegation analysis in the study shows that enegy consumption and eal GDP ae cointegated and a bidiectional causality exists between enegy consumption and eal GDP in Austalia in the longun. Results fom the VEC models suggest the existence of bidiectional causality between enegy consumption and eal GDP in Austalia in both the shot- and long-un. While thee is a stong evidence of long-un causality between enegy consumption and capital accumulation, enegy consumption does not Gange cause capital accumulation, but capital accumulation Gange causes enegy consumption in the shot-un. Tests fo the joint significance of the lagged dynamic tems of an explanatoy vaiable and eo-coection tems suggested the existence of Gange endogeneity in the system. Jounal of Applied Business and Economics vol. 12(3)

11 Being a high pe capita cabon dioxide emitte, as well as pevailing hot and dy weathe conditions and enduing vulneability to climate vaiation, Austalia has geat inteest in emissions contol at both national and global levels. Howeve, cabon emissions ae lagely attibuted to enegy consumption. Given the long-un elationship and the diection of causality between enegy consumption and eal GDP, as found in this study, dastic measues of cabon eductions by placing constaints on the enegy secto seem to pose significant economic costs fo the Austalian economy. In a county like Austalia, whee enegy consumption appeas to be an impotant deteminant of long-un gowth, diect measues to educe cabon fom enegy elated souces pose a significant theat to the county s economic gowth. Pesently, the fossil fuel enegy consumption accounts fo about 95% of the total enegy consumption in Austalia, and coal alone accounts fo about 80% of the total electicity poduction (Wold Bank, 2009). Convesely, clean enegy consumption is only 1% of total enegy consumption. It is, theefoe, necessay to focus on the divesification of enegy souces, paticulaly towads the clean enegy souces (natual gas, nuclea powe, sola, and wind) to attain the taget of cabon eduction while maintaining a sustainable output gowth. REFERENCES Akaike, H. (1969), Fitting autoegessive models fo egession. Annals of the Institute of Statistical Mathematics, 21, Akinlo, A. (2008). Enegy consumption and economic gowth: Evidence fom 11 Sub-Sahaa Afican counties. Enegy Economics, 30(5), Belloumi, M. (2009). Enegy consumption and GDP in Tunisia: Cointegation and causality analysis. Enegy Policy, 37(7), Chiou-Wei, S., Chen, C., & Zhu, Z. (2008). Economic gowth and enegy consumption evisited Evidence fom linea and nonlinea Gange causality. Enegy Economics, 30(6), Depatment of Climate Change (2008). Austalia's National Geenhouse Accounts, National Geenhouse Gas Inventoy 2006 Commonwealth of Austalia. Elliott, G., Thomas, J., & Stock, J. (1996). Efficient tests fo an autoegessive unit oot. Econometica, 64:4, Endes, W. (2004). Applied econometic time seies (2nd ed.): John Wiley & Sons (ASIA) Pte Ltd. Engle, R., & Gange, C. (1987). Co-integation and eo coection: epesentation, estimation, and testing. Econometica: Jounal of the Econometic Society, Edal, G., Edal, H., & Esengün, K. (2008). The causality between enegy consumption and economic gowth in Tukey. Enegy Policy, 36(10), Fatai, K., Oxley, L., & Scimgeou, F. G. (2004). Modelling the causal elationship between enegy consumption and GDP in New Zealand, Austalia, India, Indonesia, The Philippines and Thailand. Mathematics and Computes in Simulation, 64(3-4), Ganaut, R. (2008). The Ganaut climate change eview: Cambidge Univesity Pess. Ghali, K., & El-Sakka, M. (2004). Enegy use and output gowth in Canada: a multivaiate cointegation analysis. Enegy Economics, 26(2), Jounal of Applied Business and Economics vol. 12(3) 2011

12 Gange, C. (1988). Causality, cointegation, and contol. Jounal of Economic Dynamics and Contol, 12(2/3), IPCC (2007a). Climate Change 2007: Synthesis Repot. Geneva, Switzeland: Contibution of Woking Goups I, II and III to the Fouth Assessment Repot of the Integovenmental Panel on Climate Change. IPCC (2007b). Climate Change 2007: The Physical Science Basis. Contibution of Woking Goup I to the Fouth Assessment Repot of the Integovenmental Panel on Climate Change. Jobet, T., & Kaanfil, F. (2007). Sectoal enegy consumption by souce and economic gowth in Tukey. Enegy Policy, 35(11), Johansen, S. (1988). Statistical analysis of cointegation vectos. Jounal of Economic Dynamics and Contol, 12(2-3), Johansen, S. (1991). Estimation and hypothesis testing of cointegation vectos in Gaussian vecto autoegessive models. Econometica: Jounal of the Econometic Society, Johansen, S. (1995). Likelihood-based infeence in cointegated vecto autoegessive models: Oxfod Univesity Pess, USA. Kaanfil, F. (2009). How many times again will we examine the enegy-income nexus using a limited ange of taditional econometic tools? Enegy Policy, 37(4), Kaft, J., & Kaft, A. (1978). On the elationship between enegy and GNP. Jounal of Enegy and Development, 3(2), Lee, C., & Chang, C. (2005). Stuctual beaks, enegy consumption, and economic gowth evisited: Evidence fom Taiwan. Enegy Economics, 27(6), Lise, W., & Van Montfot, K. (2007). Enegy consumption and GDP in Tukey: Is thee a co integation elationship? Enegy Economics, 29(6), Mackinnon, J. (1991). Citical Values fo Cointegation Tests. In R. Engle & C. Gange (Eds.), Long-un Economic Relationships McKibbin, W., & Wilcoxen, P. (2002). The ole of economics in climate change policy. Jounal of Economic Pespectives, Naayan, P. K., & Pasad, A. (2008). Electicity consumption-eal GDP causality nexus: Evidence fom a bootstapped causality test fo 30 OECD counties. Enegy Policy, 36(2), Naayan, P. K., & Smyth, R. (2005). The esidential demand fo electicity in Austalia: an application of the bounds testing appoach to cointegation. Enegy Policy, 33(4), Pickeing, K., & Owen, L. (1997). An intoduction to global envionmental issues (2nd ed.): Routledge. Soytas, U., & Sai, R. (2003). Enegy consumption and GDP: causality elationship in G-7 counties and emeging makets. Enegy Economics, 25(1), Jounal of Applied Business and Economics vol. 12(3)

13 Soytas, U., & Sai, R. (2009). Enegy consumption, economic gowth, and cabon emissions: Challenges faced by an EU candidate membe. Ecological Economics, 68(6), Soytas, U., Sai, R., & Ewing, B. T. (2007). Enegy consumption, income, and cabon emissions in the United States. Ecological Economics, 62(3-4), Sten, D. (2000). A multivaiate cointegation analysis of the ole of enegy in the US macoeconomy. Enegy Economics 22, Sten, D., & Cleveland, C. (2004). Enegy and Economic Gowth, Depatment of Economics, Rensselae Polytechnic Institute. Sten, N. (2008). The economics of climate change. Ameican Economic Review, 98(2), Thompson, P., & Taylo, T. (1995). The capital-enegy substitutability debate: A new look. The Review of Economics and Statistics, 77(3), Wills, I. (2006). Economics and the envionment: a signalling and incentives appoach (2nd ed.): Allen & Unwin. Wold Bank (2009). Wold Development Indicatos CD-ROM. China at both aggegated and disaggegated levels. Enegy Economics, 30(6), Jounal of Applied Business and Economics vol. 12(3) 2011