原油现货和期货价格关系协整线性和非线性因果关系外文翻译.docx
《原油现货和期货价格关系协整线性和非线性因果关系外文翻译.docx》由会员分享,可在线阅读,更多相关《原油现货和期货价格关系协整线性和非线性因果关系外文翻译.docx(13页珍藏版)》请在冰点文库上搜索。
原油现货和期货价格关系协整线性和非线性因果关系外文翻译
外文题目:
TheRelationshipbetweenCrudeOilSpotandFuturesPrices:
Cointegration,LinearandNonlinearCausality
出处:
EnergyEconomics
作者:
SteliosD.Bekiros,CeesG.HDiks
原文:
TheRelationshipbetweenCrudeOilSpotandFuturesPrices:
Cointegration,LinearandNonlinearCausality!
Abstract
Thepresentstudyinvestigatesthelinearandnonlinearcausallinkagesbetweendailyspotandfuturespricesformaturitiesofone,two,threeandfourmonthsofWestTexasIntermediate(WTI)crudeoil.ThedatacovertwoperiodsOctober1991-October1999andNovember1999-October2007,withthelatterbeingsignificantlymoreturbulent.ApartfromtheconventionallinearGrangertestweapplyanewnonparametrictestfornonlinearcausalitybyDiksandPanchenkoaftercontrollingforcointegration.Inadditiontothetraditionalpairwiseanalysis,wetestforcausalitywhilecorrectingfortheeffectsoftheothervariables.Tocheckifanyoftheobservedcausalityisstrictlynonlinearinnature,wealsoexaminethenonlinearcausalrelationshipsofVECMfilteredresiduals.Finally,weinvestigatethehypothesisofnonlinearnon-causalityaftercontrollingforconditionalheteroskedasticityinthedatausingaGARCH-BEKKmodel.WhilstthelinearcausalrelationshipsdisappearafterVECMcointegrationfiltering,nonlinearcausallinkagesinsomecasespersistevenafterGARCHfilteringinbothperiods.Thisindicatesthatspotandfuturesreturnsmayexhibitasymmetriesandstatisticallysignificanthigher-ordermoments.Moreover,theresultsimplythatifnonlineareffectsareaccountedfor,neithermarketleadsorlagstheotherconsistently,videlicetthepatternofleadsandlagschangesovertime.
Keywords:
Nonparametricnonlinearcausality;OilFuturesMarket;Cointegration;
Theroleoffuturesmarketsinprovidinganefficientpricediscoverymechanismhasbeenanareaofextensiveempiricalresearch.SeveralstudieshavedealtwiththeLead-lagrelationshipsbetweenspotandfuturespricesofcommoditieswiththeobjectiveofinvestigatingtheissueofmarketefficiency.GarbadeandSilber(1983)firstpresentedamodeltoexaminethepricediscoveryroleoffuturespricesandtheeffectofarbitrageonpricechangesinspotandfuturesmarketsofcommodities.TheGarbade-SilbermodelwasappliedtothefeedercattlemarketbyOellermannetal.(1989)andtothelivehogcommoditymarketbySchroederandGoodwin(1991),whileasimilarstudybySilvapulleandMoosa(1999)examinedtheoilmarket.BoppandSitzer(1987)testedthehypothesisthatfuturespricesaregoodpredictorsofspotpricesintheheatingoilmarket,whileSerletisandBanack(1990)andChenandLin(2004)testedformarketefficiencyusingcointegrationanalysis.CrowderandHamed(1993)andSadorsky(2000)alsousedcointegrationtotestthesimpleefficiencyhypothesisandthearbitrageconditionforcrudeoilfutures.Finally,SchwarzandSzakmary(1994)examinedthepricediscoveryprocessinthemarketsofcrudeandheatingoil.
Intheory,sincebothfuturesandspotprices“refect”thesameaggregatevalueoftheunderlyingassetandconsideringthatinstantaneousarbitrageispossible,futuresshouldneitherleadnorlagthespotprice.However,theempiricalevidenceisdiverse,althoughthemajorityofstudiesindicatethatfuturesinfluencespotpricesbutnotviceversa.Theusualrationalizationofthisresultisthatthefuturespricesrespondtonewinformationmorequicklythanspotprices,duetolowertransactioncostsandflexibilityofshortselling.Withreferencetotheoilmarket,ifnewinformationindicatesthatoilpricesarelikelytorise,perhapsbecauseofanOPECdecisiontorestrictproduction,oranimminentharshwinter,aspeculatorhasthechoiceofeitherbuyingcrudeoilfuturesorspot.Whilstspotpurchasesrequiremoreinitialoutlayandmaytakelongertoimplement,futurestransactionscanbeimplementedimmediatelybyspeculatorswithoutaninterestinthephysicalcommodityperseandwithlittleup-frontcash.Moreover,hedgerswhoareinterestedforthephysicalcommodityandhavestorageconstraintswillbuyfuturescontracts.Therefore,bothhedgersandspeculatorswillreacttothenewinformationbypreferringfuturesratherthanspottransactions.Spotpriceswillreactwithalagbecausespottransactionscannotbeexecutedsoquickly(SilvapulleandMoosa,1999).Furthermore,thepricediscoverymechanism,asillustratedbyGarbadeandSilber(1983),supportsthehypothesisthatfuturespricesleadspotprices.Theirstudyofsevencommoditymarketsindicatedthat,althoughfuturesmarketsleadspotmarkets,thelatterdonotjustechotheformer.Futurestradingcanalsofacilitatetheallocationofproductionandconsumptionovertime,particularlybyprovidingamarketschemeininventoryholdings(Houthakker,1992).Inthiscase,iffuturespricesforlatedeliveriesareabovethoseforearlyones,delayofconsumptionbecomesattractiveandchangesinfuturespricesresultinsubsequentchangesinspotprices.AccordingtoNewberry(1992)futuresmarketsprovideopportunitiesformarketmanipulationbythebetterinformedorlargerattheexpenseofothermarketparticipants.Forexample,itisprofitablefortheOPECtointerveneinthefuturesmarkettoinfluencetheproductiondecisionsofitscompetitorsinthespotmarket.Finally,supportforthehypothesisthatcausalityrunsfromfuturestospotpricescanalsobefoundinthemodelofdeterminationoffuturespricesproposedbyMoosaandAl-Loughani(1995).Intheirmodelthefuturespriceisdeterminedbyarbitrageurswhosedemanddependsonthedifferencebetweenthearbitrageandactualfuturespriceandbyspeculatorswhosedemandforfuturescontractsdependsonthedifferencebetweentheexpectedspotandtheactualfuturesprice.Thereferencepointinbothcasesisthefuturespriceandnotthespotprice(SilvapulleandMoosa,1999).
Theaimofthepresentstudyistotestfortheexistenceoflinearandnonlineacausallead-lagrelationshipsbetweenspotandfuturespricesofWestTexasIntermediate(WTI)crudeoil,whichisusedasanindicatorofworldoilpricesandistheunderlyingcommodityofNewYorkMercantileExchange's(NYMEX)oilfuturescontracts.Weapplyathree-stepempiricalframeworkforexaminingdynamicrelationshipsbetweenspotandfuturesprices.First,weexplorenonlinearandlineardynamiclinkagesapplyingthenonparametricDiks-Panchenkocausalitytest,andaftercontrollingforcointegration,aparametriclinearGrangercausalitytest.Inthesecondstep,afterfilteringthereturnseriesusingtheproperlyspecifiedVARorVECMmodel,theseriesofresidualsareexaminedbythenonparametricDiks-Panchenkocausalitytest.InadditiontoapplyingtheusualbivariateVARorVECMmodeltoeachpairoftimeseries,wealsoconsiderresidualsofafullfive-variatemodeltoaccountforthepossibleeffectoftheothervariables.Thisstepensuresthatanyremainingcausalityisstrictlynonlinearinnature,astheVARorVECMmodelhasalreadypurgedtheresidualsoflineardependence.Finally,inthelaststep,weinvestigatethenullhypothesisofnonlinearnon-causalityaftercontrollingforconditionalheteroskedasticityinthedatausingaGARCH-BEKKmodel,againbothinabivariateandinafive-variaterepresentation.Ourapproachincorporatestheentirevariance-covariancestructureofthespotandfuturepricesinterrelationship.TheempiricalmethodologyemployedwiththemultivariateGARCH-BEKKmodelcannotonlyhelptounderstandtheshort-runmovements,butalsoexplicitlycapturethevolatilitypersistencemechanism.Improvedknowledgeofthedirectionandnatureofcausalityandinterdependencebetweenthespotandfuturesmarkets,andconsequentlythedegreeoftheirintegration,willexpandtheinformationsetavailabletopolicymakers,internationalportfoliomanagersandmultinationalcorporationsfordecision-making.
Theremainderofthepaperisorganizedasfollows.Section2brieflyreviewsthelinearGrangercausalityframeworkandprovidesadescriptionoftheDiks-PanchenkononparametrictestfornonlinearGrangercausality.Section3describesthedatausedandSection4presentstheresults.Section5concludeswithasummaryandsuggestionsforfutureresearch.
Figure1displaysthespotandfuturepriceandreturnstimeseries.Thefollowingnotationisused:
“WTISpot”isthespotpriceand“WTIF1”,“WTIF2”,“WTIF3”and“WTIF4”arethefuturespricesformaturitiesofone,two,threeandfourmonthsrespectively.DescriptivestatisticsforWTIspotandfutureslog-dailyreturnsarereportedinTable1.Specifically,thereturnsaredefinedas
wherePtistheclosingpriceondayt.ThedifferencesbetweenthetwoperiodsarequiteevidentinTable1whereasignificantincreaseinvariancecanbeobservedaswellasahigherdispersionofthereturnsdistributioninPeriodIIreflectedinthelowerkurtosis.Additionally,PeriodIIwitnessedmanyoccasionalnegativespikesasitcanbealsoinferredfromtheskewness.TheresultsfromtestingnonstationarityarepresentedinTable2.
Specifically,Table2reportstheAugmentedDickey-Fuller(ADF)testforthelogarithmiclevelsandlog-dailyreturns.ThelaglengthswhichareconsistentlyzeroinallcaseswereselectedusingtheSchwartzInformationCriterion(SIC).Allthevariablesappeartobenonstationaryinlog-levelsandstationaryinlog-returnsbasedonthereportedp-values.Table1alsoreportsthecorrelationmatrixatlag0(contemporaneouscorrelation)forbothperiods.Significantsamplecross-correlationsarenotedforspotandfuturesreturnsindicatingahighinterrelationshipbetweenthetwomarkets.However,sincelinearcorrel