Energy consumption and economic growth_ Evidence from the Commonwealth of Independent States

ContentslistsavailableatScienceDirect

EnergyEconomics

journalhomepage:www.elsevier.com/locate/eneco

Energyconsumptionandeconomicgrowth:EvidencefromtheCommonwealthofIndependentStates

NicholasApergisa,JamesE.Payneb,⁎

abDepartmentofBankingandFinancialManagement,UniversityofPiraeusKaraoliandDimitriou80Piraeus,ATTIKI18534,GreeceDepartmentofEconomics,IllinoisStateUniversity,Normal,IL61790-4200,UnitedStates

articleinfoabstract

ThisstudyexaminestherelationshipbetweenenergyconsumptionandeconomicgrowthforelevencountriesoftheCommonwealthofIndependentStatesovertheperiod1991–2005withinamultivariatepaneldataframework.BasedonPedroni's(1999,2004)heterogeneouspanelcointegrationtestandcorrespondingerrorcorrectionmodel,cointegrationispresentbetweenrealGDP,energyconsumption,realgrossfixedcapitalformation,andlaborforcewiththerespectivecoefficientspositiveandstatisticallysignificant.Theresultsoftheerrorcorrectionmodelrevealthepresenceofunidirectionalcausalityfromenergyconsumptiontoeconomicgrowthintheshort-runwhilebidirectionalcausalitybetweenenergyconsumptionandeconomicgrowthinthelong-run.Thus,theresultslendsupportforthefeedbackhypothesisassociatedwiththerelationshipbetweenenergyconsumptionandeconomicgrowth.

©2009ElsevierB.V.Allrightsreserved.

Articlehistory:

Received20November2008

Receivedinrevisedform26January2009Accepted26January2009

Availableonline3February2009Keywords:

EnergyconsumptionGrowth

PanelunitrootandcointegrationtestsGranger-causality

1.Introduction

Formedin1991,theCommonwealthofIndependentStates(CIS)iscomprisedoftwelvecountriesoftheformerSovietUnion.1ThoughmanyofthecountrieswithintheCISmaybeconsideredtransitioneconomies,thesecountriesplayanimportantroleinworldenergymarketsbothasproducersofoilandnaturalgasandastransitcentersforthedistributionofthesenaturalresources.Inlightofthisregion'simportancewithinworldenergymarkets,itissurprisingtherehavebeennopublishedempiricalstudieswhichexploretherelationshipbetweenenergyconsumptionandeconomicgrowthforthisgroupofcountries.2Thetaskofthisstudyistofillthisvoidintheempiricalliterature.Suchaninvestigationwillnotonlyprovideinsightswithrespecttotheroleofenergyconsumptionineconomicdevelopmentforthesecountries,butalsoserveasabasisfordiscussionofenergyandenvironmentalpolicies.

Table1providesanoverviewofthecompositionofenergyproductionandusage,theenvironmentalimpactofenergyconsump-

⁎Correspondingauthor.Tel.:+13094388588;fax:+13094385228.

E-mailaddresses:napergis@unipi.gr(N.Apergis),jepayne@ilstu.edu(J.E.Payne).1Georgiajoinedlaterin1993.TheBalticcountriesofEstonia,Latvia,andLithuaniadecidednottoparticipateintheCIS,butinsteadpursueaffiliationwiththeEuropeanUnion.

2SeePayne(forthcoming)forasurveyoftheinternationalliteratureonthecausalrelationshipbetweenenergyconsumptionandeconomicgrowth.ReynoldsandKolodzieji(2008)examinetheformerSovietUnionandfocusesexclusivelyonthebivariateGranger-causalityrelationshipbetweenoil,coal,andnaturalgasproductionandGDP.ReynoldsandKolodzieji(2008)findunidirectionalcausalityfromoilproductiontoGDPandunidirectionalcausalityfromGDPtocoalproductionandnaturalgasproduction,respectively.

0140-9883/$–seefrontmatter©2009ElsevierB.V.Allrightsreserved.doi:10.1016/j.eneco.2009.01.011

tion,andlevelofeconomicdevelopmentwithintheCISregion.3RussiadominatestheCISwithrespecttooilproductionandservesasamajoroilproducerintheworld.WhileRussia,Kazakhstan,andAzerbaijanarenetexportsofoil,theremainingCIScountriesarenetimporters.Inregardstonaturalgasproduction,Russiahastheworld'slargestnaturalgasreserveswith1,680trillioncubicfeet,nearlytwicethereservesofIranwhichhasthesecondlargestnaturalgasreserves.4Moreover,Russiaistheworld'slargestnaturalgasproduceraswellastheworld'slargestexporterofnaturalgas.ThoughTurkmenistanandUzbekistanfollowRussiaintermsofnaturalgasproduction,bothofthesecountrieshavestruggledtobringtheirsizeableoilandnaturalgasreservestoworldmarketsdueinlargeparttoaninsufficientpipelineinfrastructurefortheexportofnaturalgastoend-usemarkets.Inregardstonaturalgasconsumption,overhalfofUkraine'senergyconsumptionisfromnaturalgas.5TherestoftheCIScountriesrelyuponnaturalgasimportsfromRussiatomeettheirnaturalgasconsumptionneeds.Relativetooilandnaturalgasproduction,coalproductionandconsumptionintheCISisnotassignificant.CoalproductionintheCISisconcentratedinRussia(whohastheworld'ssecondlargestrecoverablecoalreservesbehindtheU.S.),Kazakhstan,andtheUkraine.

TheenergysourcesunderlyingelectricityproductionvariesquiteabitacrossCIScountriesaswell.Thepercentageofelectricityproductionfromoilrangesfrom27.68%inAzerbaijanto0.00%in

Turkmenistanwasexcludedgiventheunavailabilityofdataonthevariablestobeincludedinthepaneldatasetforestimationofthecausalrelationshipbetweenenergyconsumptionandeconomicgrowth.4CountryEnergyBriefs,EnergyInformationAgency,www.eia.doe.gov.5CountryEnergyBriefs,EnergyInformationAgency,www.eia.doe.gov.

32N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

Table1

Overviewofenergyproductionandconsumption(CommonwealthofIndependentStates,2005).

Armenia

PetroleumProductionConsumptionNaturalgasProductionConsumptionCoal

ProductionConsumptionElectricityProduction%Oil

%Naturalgas%Coal

%Hydroelectric%NuclearConsumptionEnergyintensityCarbondioxide

EmissionspercapitaRealGDPPercapita

0.0040.00

Azerbaijan440.98115.00

Belarus33.70156.00

Georgia1.9813.40

Kazakhstan1337.90234.00

Kyrgyzstan1.9612.00

Moldova0.0014.50

Russia9512.982757.00

Tajikistan0.2830.00

Ukraine100.39328.00

Uzbekistan124.91155.00

0.0060.00205.50366.606.10716.900.5052.10934.801075.301.1026.100.0086.5022,622.8016,153.101.4050.50685.103079.502108.001702.20

0.000.100.000.000.000.300.000.0095.4069.000.401.500.000.20320.60258.400.100.1066.5068.403.303.20

6.320.0028.940.0028.0743.004.204.9121.2227.6858.140.0014.180.0019.202.7730.968.4091.380.010.110.0029.503.147.270.9113.280.0085.810.007.404.9167.927.4110.6870.3411.570.0058.002.5116.420.009.523.6286.870.008.203.183.860.2398.140.0016.300.005.602.39951.162.2346.1617.4218.1515.71779.402.6217.090.002.350.0097.650.0014.702.80185.920.2718.6526.666.67.74142.201.8447.7113.6368.8.6812.840.0040.901.12

1.203.776.600.8613.331.121.9610.590.776.955.33

$4162$4575$81$3520$8699$1727$2191$11,858$1478$5583$2008

Notes:Petroleumproductionandconsumptionmeasuredinthousandsofbarrelsperday.Naturalgasproductionandconsumptionmeasuredinbillioncubicfeet.Coalproductionandconsumptionmeasuredinmillionshorttons.Electricityproductionandconsumptionmeasuredinbillionkilowatthours.Dataonpetroleum,naturalgas,coal,andelectricityconsumptionwereobtainedfromtheEnergyInformationAgency–InternationalEnergyDataandAnalysis(www.eia.doe.gov).Dataonelectricityproduction,energyintensitymeasuredinGDPperunitofenergyuse(constant2005PPPinternationaldollarsperkilogramofoilequivalent).Carbondioxideemissionsmeasuredinmetrictonspercapitain2004.RealGDPpercapitameasuredinconstant2005PPPinternationaldollars.Dataonelectricityproduction,energyintensity,carbondioxideemissions,andrealGDPpercapitawereobtainedfromWorldBankEconomicIndicatorsCD-ROM.

Armenia,Kyrgyzstan,andTajikistan;thepercentageofelectricityproductionfromnaturalgasrangesfrom98.14%inMoldovato2.35%inTajikistan;thepercentageofelectricityproductionfromcoalrangesfrom70.34%inKazakhstanto0.00%inArmenia,Azerbaijan,Georgia,Moldova,andTajikistan;thepercentageofelectricityproductionfromhydroelectricpowerrangesfrom97.65%inTajikistanto0.11%inBelarus;andthepercentageofelectricityproductionfromnuclearpowerrangesfrom47.74%intheUkraineto0.00%inAzerbaijan,Belarus,Georgia,Kazakhstan,Kyrgyzstan,Moldova,Tajikistan,andUzbekistan.

InadditiontothevariationinenergysourcesandconsumptionpatternsacrossCIScountries,thereisagreatdealofvariationintheefficiencyofenergyusageasmeasuredbyGDPperunitofenergyuse,rangingfrom1.12inUzbekistanto4.91inArmeniaandGeorgia.AsnotedbytheEnergyInformationAgency,thevariationintheefficiencyofenergyusageisnotsurprisinggiventhatmanyoftheCIScountrieshaveanagingandinefficientenergyinfrastructurethatareincriticalneedofcapitalinvestmentandmodernization.Ontheenvironmentalfront,CIScountriesfacetremendouschallengesinthereductionofgreenhousegasemissions.Asameasureoftheenvironmentalconsequencesofenergyproductionandconsumption,Table1reportscarbondioxideemissionsinmetrictonspercapita.Carbondioxideemissionspercapitarangefromalowof0.77MT/capitainTajikistantoahighof13.33MT/capitainKazakhstan.Itisinterestingtonotethatthecountrieswiththelowestcarbondioxideemissionspercapita(Tajikistan0.77,Georgia0.86,andKyrgyzstan1.12)havethehighestpercentageofelectricityproductionfromhydroelectricpower(Tajikistan97.65%,Kyrgyzstan86.87%,andGeorgia85.81%).Finally,thelevelofeconomicdevelopmentinCIScountriesexhibitsawidedisparityinpartduetothesuccesstowhichindividualcountrieshavemadethetransitiontowardsamarket-basedeconomy.AsreportedinTable1,realGDPpercapitarangesfrom$1,727inKyrgyzstanto$11,858inRussia.

Giventhebriefoverviewofthecompositionofenergyproductionandusage,theenvironmentalimpactofenergyconsumption,andthelevelofeconomicdevelopmentwithintheCISregion,Section2discussesthefourhypothesesassociatedwiththecausalrelationshipbetweenenergyconsumptionandeconomicgrowth.Section3discussesthedata,methodology,andempiricalresults.Section4providesconcludingremarks.

2.Surveyoftheenergyconsumption-growthhypothesesandtheCIS

Therelationshipbetweenenergyconsumptionandeconomicgrowthhasbeensynthesizedintofourtestablehypotheses.First,thegrowthhypothesisassertsthatenergyconsumptionservesavitalroleineconomicgrowthbothasadirectinputintheproductionprocessandindirectlyasacomplementtolaborandcapitalinputs.ThegrowthhypothesissuggeststhatanincreaseinenergyconsumptioncausesanincreaseinrealGDP(i.e.theeconomyisenergydependent).Underthegrowthhypothesis,energyconservationpolicieswhichreduceenergyconsumptionmayhaveanadverseimpactonrealGDP.However,Squalli(2007)suggeststhepossibilitythatanincreaseinenergyconsumptionmayhaveanegativeimpactonrealGDP.Suchapossibilitycouldresultfromexcessiveenergyconsumptioninrelativelyunproductivesectorsoftheeconomy,capacityconstraints,orinefficienciesinenergyproduction.AnotherpossibilityforthenegativeimpactofenergyconsumptiononrealGDPcouldbeattributedtothecaseofagrowingeconomywhichrequiresadecreasingamountofenergyconsumptionasproductionmovestowardlessenergyintensivesectorsoftheeconomy.

Second,theconservationhypothesissuggeststhatenergycon-servationpoliciesdesignedtoreduceenergyconsumptionandwastemaynothaveanadverseimpactonrealGDP.TheconservationhypothesisisconfirmedifanincreaseinrealGDPcausesanincreasein

N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

3

Table2

Testsofdynamicheterogeneity(CommonwealthofIndependentStates,1991–2005).Specification

ADF(3)

AR(4)

White'stestPanelA:testsofdynamicheterogeneityacrossgroups(Russiaincluded)Y–E–K–L32.12a46.79a98.02aPanelB:testsofdynamicheterogeneityacrossgroups(Russiaexcluded)Y–E–K–L26.74a35.28a79.41aNotes:ADF(3)reportstheparameterequalitytest(F-test)acrossallrelationshipsinthepanel.AR(4)displpaystheF-testofparameterqualityconductedinafourth-orderautoregressivemodeloftherelationships.White'stestofequalityofvariancesacrosstheinvestigatedrelationshipsinthepanel.The1%significanceleveldenotedby“a”.

energyconsumption.Alternatively,Squalli(2007)arguesthatagrowingeconomythatisconstrainedbypoliticalinfluences,infra-structure,orthemismanagementofresourcesmaygenerateineffi-cienciesalongwiththereductionintheconsumptionforgoodsandservicesincludingenergy.Third,theneutralityhypothesisconsidersenergyconsumptiontobeasmallcomponentofoveralloutputandthusmayhavelittleornoimpactonrealGDP.Asinthecaseoftheconservationhypothesis,energyconservationpolicieswouldnothaveanadverseimpactonrealGDP.TheneutralityhypothesisissupportedbytheabsenceofacausalrelationshipbetweenenergyconsumptionandrealGDP.Fourth,underthefeedbackhypothesis,energyconsumptionandrealGDPareinterrelatedandmayverywellserveascomplementstoeachother.ThepresenceofbidirectionalcausalitybetweenenergyconsumptionandrealGDPsupportsthefeedbackhypothesiswherebyanenergypolicyorientedtowardimprovementsinenergyconsumptionefficiencymaynothaveanadverseimpactonrealGDP.

Whiletherehavebeenanumberofempiricalstudiesondevelopingeconomiesyieldingmixedresultsastotheaforemen-tionedhypotheses,theempiricalstudiesontheenergyconsumption–economicgrowthnexuswithrespecttoCIScountrieshasyieldedonlyonepublishedstudyfocusingontheformerSovietUnion.6ReynoldsandKolodzieji(2008)examinetherelationshipbetweenoil,coal,andnaturalgasproductionandGDPintheformerSovietUnion.UsingGranger-causalitytestswithinabivariateframework,ReynoldsandKolodzieji(2008)findunidirectionalcausalityfromoilproductiontoGDPandunidirectionalcausalityfromGDPtocoalproductionandnaturalgasproduction,respectively.AspointedoutbyLütkepohl(1982),acommonproblemofsuchbivariateanalysisisthepossibilityofomittedvariablebias.First,giventheomittedvariablebiasassociatedwithpreviousstudies,thisstudyexaminestherelationshipbetweenenergyconsumptionandeconomicgrowthwithinamulti-variateframeworkbyincludingmeasuresofcapitalandlabor.7Second,tocircumventthereductioninthepowerandsizepropertiesofconventionalunitrootandcointegrationtestswhenusingalimiteddataset,thepanelunitrootandcointegrationmethodologyofPedroni(1999,2004)willbeutilized.Specifically,thePedroni(1999,2004)approachwillprovideadditionalpowerincombiningthecross-sectionandtimeseriesdatawhileallowingforheterogeneityacrosscountries.8Third,thesignandmagnitudeoftherespectivecoeffi-cientsfromtheerrorcorrectionmodelwillbediscussedinrelationto

6Nachaneetal.(1988),Ebohon(1996),MasihandMasih(1996,1998),MurrayandNan(1996),Jumbe(2004),MorimotoandHope(2004),Lee(2005),Wolde-Rufael(2005,2006),Chontanawatetal.(2006,2008),MahadevanAsafu-Adjaye(2007),MozumderandMarathe(2007),NarayanandSingh(2007),Akinlo(2008),andApergisandPayne(2009)forasampleofstudiespertainingtodevelopingeconomies.7Stern(1993,2000),GhaliandEl-Sakka(2004),Huangetal.(2008),andApergisandPayne(2009),amongothers,employamultivariateframeworkaswellwiththeinclusionofbothlaborandcapitalmeasures.8StudiesbyLee(2005),Al-Iriani(2006),Chenetal.(2007),MahadevanandAsafu-Adjaye(2007),Mehrara(2007),LeeandChang(2008),Leeetal.(2008),NarayanandSmyth(2008),andApergisandPayne(2009)employpanelcointegrationandcorrespondingerrorcorrectionmodelstoinferthecausalrelationshipbetweenenergyconsumptionmeasuresandeconomicgrowth.

thevarioushypothesesontheenergyconsumption–growthnexus.

Fourth,giventheappearanceofRussia'sdominancewithintheCommonwealthofIndependentStates,theempiricalanalysiswillutilizetwopaneldatasets,onewhichincludesRussiaandtheothernot,inordertoinferthesensitivityoftheresultsfortheregion.3.Data,methodology,andresults

Annualdatafrom1991to2005wereobtainedfromtheWorldBankDevelopmentIndicatorsforArmenia,Azerbaijan,Belarus,Georgia,Kazakhstan,Kyrgyzstan,Moldova,Russia,Tajikistan,Ukraine,andUzbekistan.ThemultivariatepaneldataframeworkincludesthenaturallogarithmofrealGDP(Y)andrealgrossfixedcapitalformation(K)bothinconstant2000U.S.dollars,laborforce(L),andenergyusage(E)inkilowattperoilequivalent.93.1.Testsofdynamicheterogeneity

Amajorconcernofthepaneldatasetsistheheterogeneityofthecountries.Thedynamicheterogeneity(i.e.variationoftheinterceptovercountriesandtime)oftherelevantvariablesisexaminedusingthemethodologysetforthbyHoltz-Eakinetal.(1985)andHoltz-Eakin(1986).TestsfordynamicheterogeneitybeginswiththeestimationoftheADF(3)equationforeachrelationshiptotestthenullhypothesisthattheregressionparametersareequalacrosstheseequationsviaanF-test.AsshowninPanelsAandBofTable2,theteststatistics(F=32.12andF=26.74)forparameterequalityrejectthenullhypothesisatthe1%significancelevel.Next,asimilartestofthenullhypothesisofparameterequalityisundertakenbyestimatinga4th-orderautoregressivemodelforeachoftherelationshipsunderinvestigationusingstandardChow-typeFtests.TheresultsofthistestinPanelsAandB(F=46.79andF=35.28)rejectthenullhypothesiswhichindicatesheterogeneityinthecross-sectionalparameters.Additionally,White'stestforgroup-wiseheteroskedasticityisusedtotestthenullhypothesisofhomogeneityerrorvarianceacrosscountries.TheWhite'stestiscalculatedbyregressingthesquaredresidualoftheADF(3)regressionsontheoriginalregressorsandtheirsquares.Thechi-squareteststatistics,(NT)×R2,forWhite'stestare98.02and79.41,respectively,whichrejectthenullhypothesisofhomogeneityerrorvarianceacrosscountries.Overall,theresultsforeachpaneldatasetindicatethattherelationshipsexhibithetero-geneityinboththedynamicsanderrorvariancesacrosscountries.3.2.Panelunitroottests

Beforeproceedingtotestsofcointegrationtodeterminethelong-runrelationshipamongthevariablesinquestion,panelunitroottestsforeachofthevariablesneedtobeexamined.Whilethereareanumberofpanelunitroottestsintheliterature,thepresenceofparameterheterogeneityattributedtothedifferenteconomiccondi-tionsandstagesofeconomicdevelopmentineachcountrysuggeststheuseoftheImetal.(IPS,2003)panelunitroottest.10Specifically,theImetal.(2003)panelunitroottestallowsforheterogeneousautoregressivecoefficients.11Imetal.(2003)panelunitroottest

9TheuseofrealgrossfixedcapitalformationasaproxyforcapitalfollowstheworkbySoytasandSari(2006,2007,forthcoming)andSoytasetal.(2007).Specifically,inaccordancewiththeperpetualinventorymethodassumingaconstantdepreciationrateindicatesthatchangesininvestmentcloselyfollowchangesinthecapitalstock.10TheBreitung(2000)andLevinetal.(2002)panelunitroottestsassumeahomogeneousautoregressiveunitrootunderthealternativehypothesiswhereasImetal.(2003)allowsforaheterogeneousautoregressiveunitrootunderthealternativehypothesis.MaddalaandWu(1999)andChoi(2001)usethenon-parametricFisherstatisticincomparableunitroottests.ThepanelunitroottestsofHadri(2000)andCarrion-i-Silvestreetal.(2005)examinethenullhypothesisofstationarity.11TheBreitung(2000),Hadri(2000),Choi(2001),Levinetal.(2002),andCarrion-i-Silvestreetal.(2005)panelunitroottestsalsoindicatedthattherespectivevariablescontainaunitroot.Resultsareavailableuponrequestfromtheauthors.

4N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

averagestheaugmentedDickey–Fuller(ADF)unitroottestsallowingfordifferentordersofserialcorrelation,\"Pwhile

it=pi

j=1’ij\"it−j+uit,renderingthefollowingexpressionyXp

it=ρiyit−1+

i

j

=1’ij\"it−j

+δiXit+uitð1Þ

wherei=1,...,Nforeachcountryinthepanel;t=1,...,Treferstothetimeperiod;Xitrepresentstheexogenousvariablesinthemodelincludingfixedeffectsorindividualtimetrend;ρiaretheautore-gressivecoefficients;pirepresentsthenumberoflagsintheADFregression;andεitarethestationaryerrorterms.Ifρib1,yitisconsideredweaklytrendstationarywhereasifρi=1,thenyitcontainsaunitroot.Thenullhypothesisisthateachseriesinthepanelcontainsaunitroot.Thealternativehypothesisisthatatleastoneoftheindividualseriesinthepanelisstationary.Imetal.(2003)specifyat-barstatisticastheaverageoftheindividualADFstatisticswhichisnormallydistributedunderthenullhypothesiswiththecriticalvaluesprovidedbyImetal.(2003).Table3reportstheresultsoftheIPSpanelunitroottestsforbothpaneldatasets.Thepanelunitroottestsrevealthateachvariableisintegratedoforderone.3.3.Panelcointegrationtests

Inlightoftheheterogeneityinboththedynamicsanderrorvariancesineachpanel,thePedroni(1999,2004)heterogeneouspanelcointegrationtest,whichallowsforcross-sectioninterdepen-dencewithdifferentindividualeffects,isspecifiedasfollows:Yit=αit+δit+β1iEit+β2iKit+β3iLit+\"it

ð2Þ

wherei=1,...,Nforeachcountryinthepanelandt=1,...,Treferstothetimeperiod.Theparametersαitandδiallowforthepossibilityofcountry-specificfixedeffectsanddeterministictrends,respectively.εitaretheestimatedresidualsrepresentingdeviationsfromthelong-runrelationship.Sinceallvariablesareexpressedinnaturallogarithms,theβ′sparametersofthemodelcanbeinterpretedaselasticityestimates.

Totestthenullhypothesisofnocointegration,ρi=1,thefollowingunitroottestisconductedontheresidualsasfollows:\"it=ρi\"it−1+wit

ð3Þ

Pedroni(1999,2004)proposestwosetsoftestsforcointegration.Thepaneltestsarebasedonthewithindimensionapproach(i.e.panelcointegrationstatistics)whichincludesfourstatistics:panelv-statistic,panelρ-statistic,panelPP-statistic,andpanelADF-statistic.Thesestatisticsessentiallypooltheautoregressivecoefficientsacrossdifferentcountriesfortheunitroottestsontheestimatedresiduals.Thesestatisticstakeintoaccountcommontimefactorsandhetero-geneityacrosscountries.Thegrouptestsarebasedonthebetweendimensionapproach(i.e.groupmeanpanelcointegrationstatistics)whichincludesthreestatistics:groupρ-statistic,groupPP-statistic,andgroupADF-statistic.Thesestatisticsarebasedonaveragesofthe

Table3

IPSpanelunitroottests(CommonwealthofIndependentStates,1991–2005).VariablesRussiaincludedRussiaexcludedY−1.27(2)−1.06(2)ΔY−5.11(1)a−4.52(1)aE−1.37(3)−1.17(3)ΔE−6.44(2)a−4.86(2)aK−1.46(3)−1.29(3)ΔK−5.36(2)a−4.63(1)aL−1.49(3)−1.31(3)ΔL

−5.90(1)a−5.16(2)aNotes:The1%significanceleveldenotedby“a”(Imetal.,2003).Lineartrendtermincluded.Numbersinparenthesesaretheaugmentedlagsincludedintheunitroottest.

Table4

Panelcointegrationtests(CommonwealthofIndependentStates,1991–2005).WithindimensionBetweendimensionTeststatistics:

Teststatistics:

PanelA:panelcointegrationtest(Russiaincluded)Panelv-statistic35.65291aGroupρ-statistic−33.51088aPanelρ-statistic−30.77302aGroupPP-statistic

−29.49234aPanelPP-statistic−34.73611aGroupADF-statistic−4.24632aPanelADF-statistic−4.206aPanelB:panelcointegrationtest(Russiaexcluded)Panelv-statistic26.33719aGroupρ-statistic−23.48226aPanelρ-statistic−21.35539aGroupPP-statistic−27.06619aPanelPP-statistic−25.88329aGroupADF-statistic−1.17682

PanelADF-statistic−1.37922

Notes:Criticalvaluesatthe1%significanceleveldenotedby“a”:panelv(24.56),panelρ(−17.60),panelPP(−25.59),panelADF(−2.97),groupρ(−21.12),groupPP(−25.59),andgroupADF(−3.18).

individualautoregressivecoefficientsassociatedwiththeunitroottestsoftheresidualsforeachcountryinthepanel.Allseventestsaredistributedasymptoticallyasstandardnormal.Oftheseventests,thepanelv-statisticisaone-sidedtestwherelargepositivevaluesrejectthenullhypothesisofnocointegrationwhereaslargenegativevaluesfortheremainingteststatisticsrejectthenullhypothesisofnocointegration.12PanelsAandBofTable4reportboththewithinandbetweendimensionpanelcointegrationteststatisticsforeachpaneldataset.Allsevenpanelcointegrationtestsrejectthenullhypothesisofnocointegrationatthe1%significancelevelforthepanelwhichincludesRussiawhileinthecaseofthepanelwhichexcludesRussiafiveofthesevenpanelcointegrationtestsrejectthenullhypothesis.Thus,theevidencesuggeststhatinbothpaneldatasetsthereisalong-runequilibriumrelationshipbetweenrealGDP,energyusage,capital,andlabor.

FollowingPedroni(2000),thefullymodifiedOLS(FMOLS)techniqueforheterogeneouscointegratedpanelsisestimated.13PanelsAandBofTable5displaytheFMOLSresults.Allthecoefficientsarepositiveandstatisticallysignificantatthe1%significancelevelandgiventhevariablesareexpressedinnaturallogarithms,thecoeffi-cientscanbeinterpretedaselasticityestimates.InPanelAwhichincludesRussia,theresultsindicatethata1%increaseinenergyusageincreasesrealGDPby0.42%;a1%increaseincapitalincreasesrealGDPby0.12%;anda1%increaseinthelaborforceincreasesrealGDPby0.67%.Ontheotherhand,theresultsinPanelBwhichexcludeRussiarevealmuchsmallerelasticityestimates.Inthiscase,a1%increaseinenergyusageincreasesrealGDPby0.24%;a1%increaseincapitalincreasesrealGDPby0.06%;anda1%increaseinthelaborforceincreasesrealGDPby0.33%.AcomparisonoftheelasticityestimatesinPanelsAandBimplythattheinclusionofRussiahasanimpactontheresponsivenessofrealGDPtothevariousfactorsofproduction(energyusage,capital,andlabor).WithrespecttotheotherpanelstudiesreportingFMOLSestimates,theelasticityofenergyusagefortheCISregioniswithintherangeofotherstudies:Lee(2005)foreighteendevelopingcountriesreportstheelasticityofenergyusageat0.50%;0.32%byLeeandChang(2008)forsixteenAsiancountries;0.25%byLeeetal.(2008)fortwenty-twoOECDcountries;0.12%byNarayanandSmyth(2008)forG7countries;and0.28%byApergisandPayne(2009)forsixCentralAmericancountries.

Giventhevariablesarecointegrated,apanelvectorerrorcorrectionmodel(Pesaranetal.,1999)isestimatedtoperformGranger-causalitytests.Thetwo-stepprocedureofEngleandGranger

12SeePedroni(1999)fordetailsontheheterogeneouspanelandgroupmeanpanelcointegrationstatistics.13NotethattheestimatesfromeitherFMOLSorDOLSareasymptoticallyequivalentformorethan60observations(Banerjee,1999).Thepaneldatasetofthisstudycontains165observations.

N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

5

Table5

PanelFMOLSlong-runestimates(CommonwealthofIndependentStates,1991–2005).PanelA:FMOLSestimates(Russiaincluded)Y=0.761+0.42E+0.12K+0.67L(4.62)a(4.73)a(5.)a(5.17)aAdj.R2=0.68LM=1.24

RESET=2.19

HE=1.15

[0.41][0.29][0.36]

PanelB:FMOLSestimates(Russiaexcluded)Y=0.344+0.24E+0.06K+0.33L(3.77)a(3.14)a(1.76)b(3.12)aAdj.R2=0.46LM=1.69

RESET=2.75

HE=1.09

[0.49]

[0.36][0.32]

Notes:t-statisticsarereportedinparenthesesandprobabilityvaluesinbrackets.LMistheLaGrangemultipliertestforserialcorrelation.RESETisthemisspecificationtest.HEisWhite'sheteroskedasticitytest.Significanceatthe1and5-percentlevelsdenotedby“a”and“b”,respectively.

(1987)isperformedbyfirstestimatingthelong-runmodelspecifiedinEq.(2)inordertoobtaintheestimatedresiduals.Next,definingthelaggedresidualsfromEq.(2)astheerrorcorrectionterm,thefollowingdynamicerrorcorrectionmodelisestimated:

ΔY+Pqk=1γPqPq

it=ω1j11ikΔYit−k+k=1γ12ikΔEit−k+k=1γ+Pq

13ikΔKit−kk=1γ14ikΔLit−k+λ1i\"it−1+u1it

ð4aÞ

ΔE=ωPqPqPq

it2j+k=1γ21ikΔYit−k+k=1γ22ikΔEit−k+k=1γ23ikΔKit−k+Pq

k=1γ24ik

ΔLit−k+λ2i\"it−1+u2itð4bÞ

ΔKPqPqPq

it=ω3j+k=1γ31ikΔYit−k+k=1γ32ikΔEit−k+k=1γ33ikΔKit−k

+Pq

k=1γ34ik

ΔLit−k+λ3i\"it−1+u3itð4cÞ

ΔLPqk=1γPqPq

it=ω4j+41ikΔYit−k+k=γ42ikΔEit−k+γΔKit−k

+Pq

1k=143ikk=1γ44ikΔLit−k+λ4i\"it−1+u4it

ð4dÞ

whereΔisthefirst-differenceoperator;qisthelaglengthsetattwobasedonlikelihoodratiotests;anduistheseriallyuncorrelatederrorterm.14IntherealGDPEq.,(4a),short-runcausalityfromenergyusage,capital,andthelabortorealGDParetested,respectively,basedonH0:γ12ik=0∀ik,H0:γ13ik=0∀ik,andH0:γ14-ik=0∀ik.IntheenergyusageEq.,(4b),short-runcausalityfromrealGDP,capital,andthelaborforcetoenergyusagearetested,respectively,basedonH0:γ21ik=0∀ik,H0:γ23ik=0∀ik,andH0:γ24-ik=0∀ik.InthecapitalEq.,(4c),short-runcausalityfromrealGDP,energyusage,andthelaborforcetocapitalaretested,respectively,basedonH0:γ31ik=0∀ik,H0:γ32ik=0∀ik,andH0:γ34ik=0∀ik.Finally,inthelaborforceEq.,(4d),short-runcausalityfromrealGDP,energyusage,andcapitaltothelaborforcearetested,respectively,basedonH0:γ41ik=0∀ik,H0:γ42ik=0∀ik,andH0:γ43ik=0∀ik.Thenullhypothesisofnolong-runcausalityineachEq.(4a)–(4d),istestedbyexaminingthesignificanceofthet-statisticforthecoefficientontherespectiveerrorcorrectiontermrepresentedbyλ.

PanelsAandBofTable6reporttheresultsoftheshort-runandlong-runGranger-causalitytestsforeachpaneldataset.InPanelAwhichincludesRussia,theshort-rundynamicsinEq.(4a)showthatenergyusage,realgrossfixedcapitalformation,andthelaborforcehaveapositiveandstatisticallysignificantimpactonrealGDP.An

14ThenumberoflagsintheVARsystemwasdeterminedvialikelihoodratiotests

(Sims,1980).Amaximumnumberofsixlagswereconsideredandtherestrictionthatalower-ordersystemrelativetoahigher-ordersystemistestedwhichresultedintheselectionoftwolagsintheVARsystem.

examinationofthesumofthelaggedcoefficientsontherespective

variablesindicatesthatenergyusage(0.281)hasagreaterimpactonrealGDPthaneitherrealgrossfixedcapitalformation(0.195)orthelaborforce(0.159)whichreiteratestheimportanceofenergyasafactorofproductionintheeconomicgrowthprocessinthecaseoftheCommonwealthofIndependentStates.IntermsofEq.(4b),itappearsthatrealGDPandthelaborforcedonothaveastatisticallysignificantimpactonenergyusageintheshort-run.Yet,realgrossfixedcapitalformationhasapositiveandstatisticallysignificantimpactonenergyusage,indicativeoftheircomplementarity.InregardstoEq.(4c),itisnotsurprisingthatrealGDPhasapositiveandstatisticallysignificantimpactonrealgrossfixedcapitalformation.Moreover,bothenergyusageandthelaborforceappeartoserveascomplementstorealgrossfixedcapitalformationgiventheirpositiveandstatisticallysignificantimpact.WithrespecttoEq.(4d),bothrealGDPandrealgrossfixedcapitalformationhaveapositiveandstatisticallysignificantimpactonthelaborforceintheshort-runwhiletheimpactofenergyusageisstatisticallyinsignificant.

Nowletuscomparetheshort-rundynamicswhenRussiaisexcludedfromtheanalysis.Thoughenergyusage,realgrossfixedcapitalformation,andthelaborforceinEq.(4a)ofPanelBhaveapositiveandstatisticallysignificantimpactonrealGDP,energyusagedoesnothaveaslargeanimpact.ThisfindingcanbeattributedtothefactthatwiththeexclusionofRussiatheremainingCIScountriespaleincomparisonwiththerespecttotheoil,naturalgas,andcoalproductionandarelargelynetimportersofenergy.InEq.(4b)onlyrealgrossfixedcapitalformationhasapositiveandstatisticallysignificantimpactonenergyusage,comparabletotheresultsinPanelA.SimilartothefindingsinPanelA,theresultsforEq.(4c)showthatrealGDP,energyusage,andthelaborforcehaveapositiveandstatisticallysignificantimpactonrealgrossfixedcapitalformation.Finally,theresultsforthelaborforceEq.(4d)areagaincomparabletothosereportedinPanelA.Thus,theresultsreportedinbothPanelsA

Table6

Panelcausalitytestresults(CommonwealthofIndependentStates,1991-2005).PanelA:panelerrorcorrectionmodel(Russiaincluded)DependentSourcesofcausation(independentvariables)variableShort-run

Long-runΔY

ΔE

ΔK

ΔL

ECT(4a)ΔY–

5.48(0.281)6.42(0.195)4.72(0.159)−0.013[0.01]a[0.00]a[0.00]a[0.00]a[0.02]b[0.00]a[0.00]a(4b)ΔE1.21(0.116)

–

4.05(0.305)1.19(0.038)−0.137[0.33][0.00]a[0.01]b[0.00]a[0.46][0.01]a[0.00]a(4c)ΔK6.29(0.9)10.46(0.462)–

11.(0.278)−0.163[0.00]a[0.00]a[0.00]a[0.00]a[0.00]a[0.00]a[0.00]a(4d)ΔL

7.12(0.462)0.02(0.048)7.05(0.392)–

−0.033[0.00]a[0.00]a[0.81][0.28]

[0.00]a[0.00]a[0.02]bPanelB:panelerrorcorrectionmodel(Russiaexcluded)DependentSourcesofcausation(independentvariables)variableShort-run

Long-runΔY

ΔE

ΔK

ΔL

ECT(4a)ΔY–

2.14(0.126)3.57(0.139)5.66(0.234)−0.012[0.05]b[0.04]b[0.03]b[0.00]b[0.01]a[0.00]a[0.02]b(4b)ΔE1.06(0.084)

–3.19(0.217)

0.83(0.031)−0.095[0.47][0.03]b[0.03]b[0.02]b[0.46][0.03]b[0.03]b(4c)ΔK3.46(0.369)5.18(0.377)–10.46(0.261)−0.128[0.01]a[0.01]a[0.00]a[0.00]a[0.00]a[0.00]a[0.00]a(4d)ΔL

9.33(0.527)0.01(0.037)9.21(0.469)–

−0.025[0.00]a[0.00]a[0.86][0.39][0.00]a[0.00]a[0.04]bNotes:PartialF-statisticsreportedwithrespecttoshort-runchangesintheindependentvariables.Thesumofthelaggedcoefficientsfortherespectiveshort-runchangesisdenotedinparentheses.ECTrepresentsthecoefficientoftheerrorcorrectionterm.ProbabilityvaluesareinbracketsandreportedunderneaththecorrespondingpartialF-statisticandsumofthelaggedcoefficients,respectively.Significanceatthe1and5-percentlevelsaredenotedby“a”and“b”,respectively.

6N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

andBindicateshort-rununidirectionalcausalityfromenergyconsumptiontoeconomicgrowth.WhetherRussiaisincludedorexcludedfromtheempiricalanalysis,theshort-rungrowthdynamicssuggestthattheCISregionmaybeconsideredenergydependent.

Whilethegrowthhypothesismaybesupportedintheshort-run,onemustalsoconsiderthelong-rundynamics.TheerrorcorrectiontermsreportedinPanelsAandBreflectthespeedofadjustmentintheresponsetoshockstowardslong-runequilibrium.TheerrorcorrectiontermsinbothPanelsAandBarestatisticallysignificantataleastthe5%level;however,thecoefficientestimatesreflectsomevariationinthespeedofadjustmentwhencomparingthetwomodels.AcomparisonoftheerrorcorrectiontermsintherealGDPequations(4a)revealscoefficientestimatesofsimilarmagnitudewitharelativelyslowadjustmenttowardequilibrium.InPanelA,thecoefficientestimatesoftheerrorcorrectiontermsassociatedwiththeequationsforenergyusages(4b),realgrossfixedcapitalformation(4c),andthelaborforce(4d)areslightlylargerthanthosereportedinPanelB.However,inbothPanelsAandB,thespeedofadjustmenttowardequilibriumisstillratherslow.RegardlessofwhetherRussiaisincludedorexcludedinthemodelestimation,thelong-rundynamicssuggestthereislong-runbidirectionalcausalitybetweenenergyconsumptionandeconomicgrowth.Thus,inthelong-runtheresultssupportthefeedbackhypothesis.

ThoughenergyconsumptionplaysanimportantroleinthegrowthprospectsfortheCIScountriesintheshort-run,energyconsumptionandeconomicgrowthareinterrelatedinthelong-run.Indeed,thefeedbackhypothesisassertsthatenergyconsumptionandeconomicgrowthareinterrelatedandmayverywellserveascomplementstoeachother.Forinstance,inordertomaintainthevitalityoftheenergysectorintheshort-run,thesecountriesneedresourcesforcapitalinvestmenttoreplacetheagingandinefficientenergyinfrastructurecurrentlyinexistencewhichcanonlyoccurthrougheconomicgrowth.Moreover,giventhesupportforthefeedbackhypothesisinthelong-run,energypoliciesorientedtowardsimprovementsinenergyproductionandconsumptionefficiencymaynothaveanadverseimpactonrealGDPandmayactuallyservetoimproveupontheenvironmentaldamageassociatedwithexcessiveenergycon-sumptionandinefficientenergyproductionmethodsthatmaystillexistinmanyoftheCIScountries.4.Concludingremarks

Understandingtherelationshipbetweenenergyconsumptionandeconomicgrowthisvitalintheeffectivedesignandimplementationofenergyandenvironmentalpolicies.InthecaseoftheCommon-wealthofIndependentStates(CIS),thedatarevealsagreatdealofvariationacrosscountriesintermsofthelevelofeconomicdevelopmentaswellasvariationinthecompositionofenergyproductionandusage.TheCISisdominatedbyRussiaintermsofoilandnaturalgasproductionwithmanyofthesmallercountriesservingasnetimportersofRussia'soilandnaturalgasproductionaswellasservingastransitcentersforthedistributionofthesenaturalresourcestoglobalenergymarkets.ThisstudyemploysapaneldatasetforelevencountrieswithintheCommonwealthofIndependentStates(Armenia,Azerbaijan,Belarus,Georgia,Kazakhstan,Kyrgyz-stan,Moldova,Russia,Tajikistan,Ukraine,andUzbekistan)overtheperiod1991–2005toexaminethecausalrelationshipbetweenenergyconsumptionandeconomicgrowthtakingintoaccounttheroleofcapitalandlaborinthegrowthprocess.GiventhedominanceofRussiaamongtheCIScountries,theanalysisisundertakenusingtwopaneldatasets:onewithRussiaincludedandtheotherexcludingRussia.

Inbothpaneldatasetsheterogeneouspanelcointegrationtestsrevealthereisalong-runequilibriumrelationshipbetweenrealGDP,energyconsumption,realgrossfixedcapitalformation,andthelaborforce.Thelong-runelasticityestimateswithrespecttoeachfactorof

productionaregenerallylargerinmagnitudewiththeinclusionofRussia,reflectingthedominantroleRussiaplaysintheregion'seconomicdevelopment.Furthermore,theestimationofapanelvectorerrorcorrectionmodelssuggestthereisshort-rununidirectionalcausalityfromenergyconsumptiontoeconomicgrowthwhereasinthelong-runthereisbidirectionalcausalitybetweenenergycon-sumptionandeconomicgrowthindicativeofthefeedbackhypothesis.Thefeedbackhypothesisassertsthatenergypolicieswhichimprovetheefficiencyintheproductionandconsumptionofenergymaynothaveadetrimentalimpactoneconomicgrowth,butmayalsoenhanceenvironmentalqualityassuchpolicieswillcurbexcessiveenergyconsumptionandinefficientenergyproductionmethods.References

Akinlo,A.E.,2008.Energyconsumptionandeconomicgrowth:evidencefrom11sub-SaharaAfricancountries.EnergyEconomics30,2391–2400.

Al-Iriani,M.A.,2006.Energy-GDPrelationshiprevisited:anexamplefromGCC

countriesusingpanelcausality.EnergyPolicy34,3342–3350.

Apergis,N.,Payne,J.E.,2009.EnergyConsumptionandEconomicGrowthinCentral

America:EvidencefromaPanelCointegrationandErrorCorrectionModel.EnergyEconomics31,211–216.

Banerjee,A.,1999.Paneldataunitrootsandcointegration:anoverview.OxfordBulletin

ofEconomicsandStatisticsS1(61),607–629.

Breitung,J.,2000.Thelocalpowerofsomeunitroottestsforpaneldata.Advancesin

Econometrics15,161–177.

Carrion-i-Silvestre,J.L.,delBarrio-Castro,T.,Lopez-Bazo,E.,2005.Breakingthepanels:

anapplicationtoGDPpercapita’.EconometricsJournal8,159–175.

Chen,S.T.,Kuo,H.I.,Chen,C.C.,2007.TheRelationshipbetweenGDPandelectricity

consumptionin10Asiancountries.EnergyPolicy35,2611–2621.

Choi,I.,2001.Unitroottestsforpaneldata.JournalofInternationalMoneyandFinance

20,249–272.

Chontanawat,J.,Hunt,L.C.,Pierse,R.,2006.Causalitybetweenenergyconsumptionand

GDP:evidencefrom30OECDand78non-OECDcountries.SurreyEnergyEconomicsDiscussionPaperSeries,vol.113.

Chontanawat,J.,Hunt,L.C.,Pierse,R.,2008.Doesenergyconsumptioncauseeconomic

growth?Evidencefromasystematicstudyofover100countries.JournalofPolicyModeling30,209–220.

Ebohon,O.J.,1996.Energy,economicgrowthandcausalityindevelopingcountries:a

casestudyofTanzaniaandNigeria.EnergyPolicy24,447–453.

Engle,R.F.,Granger,C.W.J.,1987.Cointegrationanderrorcorrection:representation,

estimation,andtesting.Econometrica55,251–276.

Ghali,K.H.,El-Sakka,M.I.T.,2004.EnergyandoutputgrowthinCanada:amultivariate

cointegrationanalysis.EnergyEconomics26,225–238.

Hadri,K.,2000.Testingforstationarityinheterogeneouspaneldata.Econometric

Journal3,148–161.

Holtz-Eakin,D.,1986.Testingforindividualeffectsindynamicmodelsusingpaneldata.

NBERTechnicalPaperSeries,No.57.

Holtz-Eakin,D.,Newey,W.,Rosen,H.,1985.Implementingcausalitytestswithpaneldata

withanexamplefromlocalpublicfinance.NBERTechnicalWorkingPaper,No.48.Huang,B.N.,Hwang,M.J.,Yang,C.W.,2008.Causalrelationshipbetweenenergy

consumptionandGDPgrowthrevisited:adynamicpaneldataapproach.EcologicalEconomics67,41–.

Im,K.S.,Pesaran,M.H.,Shin,Y.,2003.Testingforunitrootsinheterogeneouspanels.

JournalofEconometrics115,53–74.

Jumbe,C.B.L.,2004.Cointegrationandcausalitybetweenelectricityconsumptionand

GDP:empiricalevidencefromMalawi.EnergyEconomics26,61–68.

Lee,C.C.,2005.EnergyconsumptionandGDPindevelopingcountries:acointegrated

panelanalysis.EnergyEconomics27,415–427.

Lee,C.C.,Chang,C.P.,2008.EnergyconsumptionandeconomicgrowthinAsian

economies:amorecomprehensiveanalysisusingpaneldata.ResourceandEnergyEconomics30,50–65.

Lee,C.C.,Chang,C.P.,Chen,P.F.,2008.Energy-incomecausalityinOECDcountries

revisited:thekeyroleofcapitalstock.EnergyEconomics30,2359–2373.

Levin,A.,Lin,C.F.,Chu,C.,2002.Unitroottestsinpaneldata:asymptoticandfinite-sampleproperties.JournalofEconometrics108,1–24.

Lütkepohl,H.,1982.Non-causalityduetoomittedvariables.JournalofEconometrics19,

267–378.

Maddala,G.S.,Wu,S.,1999.Acomparativestudyofunitroottestswithpaneldataanda

newsimpletest.OxfordBulletinofEconomicsandStatistics61,631–652.

Mahadevan,R.,Asafu-Adjaye,J.,2007.Energyconsumption,economicgrowthand

prices:areassessmentusingpanelVECMfordevelopedanddevelopingcountries.EnergyPolicy35,2481–2490.

Masih,A.M.M.,Masih,R.,1996.Energyconsumption,realincomeandtemporal

causality:resultsfromamulti-countrystudybasedoncointegrationanderror-correctionmodellingtechniques.EnergyEconomics18,165–183.

Masih,A.M.M.,Masih,R.,1998.Amultivariatecointegratedmodelingapproachin

testingtemporalcausalitybetweenenergyconsumption,realincome,andpriceswithanapplicationtotwoAsianLDCs.AppliedEconomics30,1287–1298.

Mehrara,M.,2007.Energyconsumptionandeconomicgrowth:thecaseofoilexporting

countries.EnergyPolicy35,2939–2945.

N.Apergis,J.E.Payne/EnergyEconomics31(2009)1–7

7

Morimoto,R.,Hope,C.,2004.TheimpactofelectricitysupplyoneconomicgrowthinSri

Lanka.EnergyEconomics26,77–85.

Mozumder,P.,Marathe,A.,2007.Causalityrelationshipbetweenelectricityconsump-tionandGDPinBangladesh.EnergyPolicy35,395–402.

Murray,D.A.,Nan,G.D.,1996.Adefinitionofthegrossdomesticproduct–electrification

interrelationship.JournalofEnergyandDevelopment19,275–283.

Nachane,D.M.,Nadkarni,R.M.,Karnik,A.V.,1988.Cointegrationandcausalitytesting

oftheenergy–GDPrelationship:across-countrystudy.AppliedEconomics20,1511–1531.

Narayan,P.N.,Singh,B.,2007.TheelectricityconsumptionandGDPnexusfortheFiji

Islands.EnergyEconomics29,1141–1150.

Narayan,P.K.,Smyth,R.,2008.EnergyconsumptionandrealGDPinG7countries:new

evidencefrompanelcointegrationwithstructuralbreaks.EnergyEconomics30,2331–2341.

Payne,J.E.,forthcoming.SurveyoftheInternationalEvidenceontheCausalRelationship

betweenEnergyConsumptionandGrowth.JournalofEconomicStudies.

Pedroni,P.,1999.Criticalvaluesforcointegrationtestsinheterogeneouspanelswith

multipleregressors.OxfordBulletinofEconomicsandStatistics61,653–670.

Pedroni,P.,2000.FullymodifiedOLSforheterogeneouscointegratedpanels.Advanced

inEconometrics15,93–130.

Pedroni,P.,2004.Panelcointegration:asymptoticandfinitesamplepropertiesof

pooledtimeseriestestswithanapplicationtothePPPhypothesis:newresults.EconometricTheory20,597–627.

Pesaran,H.M.,Shin,Y.,Smith,R.P.,1999.Pooledmeangroupestimationofdynamic

heterogeneouspanels.JournaloftheAmericanStatisticalAssociation94,621–634.

Reynolds,D.B.,Kolodzieji,M.,2008.FormerSovietUnionoilproductionandGDP

decline:Grangercausalityandthemulti-cycleHubbertcurve.EnergyEconomics30,271–2.

Sims,C.,1980.Macroeconomicsandreality.Econometrica48,1–48.

Soytas,U.,Sari,R.,2006.CanChinacontributemoretothefightagainstglobalwarming.

JournalofPolicyModeling28,837–846.

Soytas,U.,Sari,R.,2007.Therelationshipbetweenenergyandproduction:evidence

fromTurkishmanufacturingindustry.EnergyEconomics29,1151–1165.

Soytas,U.,Sari,R.,forthcoming.EnergyConsumption,EconomicGrowth,andCarbon

Emissions:ChallengesFacedbyanEUCandidateMember.EcologicalEconomics.Soytas,U.,Sari,R.,Ewing,B.T.,2007.Energyconsumption,income,andcarbonemissions

intheUnitedStates.EcologicalEconomics62,482–4.

Squalli,J.,2007.Electricityconsumptionandeconomicgrowth:boundsandcausality

analysesofOPECcountries.EnergyEconomics29,1192–1205.

Stern,D.I.,1993.EnergyandeconomicgrowthintheUSA:amultivariateapproach.

EnergyEconomics15,137–150.

Stern,D.I.,2000.AmultivariatecointegrationanalysisoftheroleofenergyintheUS

macroeconomy.EnergyEconomics22,267–283.

Wolde-Rufael,Y.,2005.Energydemandandeconomicgrowth:theAfricanexperience.

JournalofPolicyModeling27,1–903.

Wolde-Rufael,Y.,2006.Electricityconsumptionandeconomicgrowth:atimeseries

experiencefor17Africancountries.EnergyPolicy34,1106–1114.