ContentslistsavailableatScienceDirect
Structural
Change
and
Economic
Dynamics
jo u r n al h om ep a g e :w w w . e l s e v i e r . c o m / l o c a t e/ s c e d
Investigating
patterns
of
carbon
convergence
in
an
uneven
economy:
The
case
of
Turkey
Sevil
Acar
a,b,∗,
A.
Erinc¸
Yeldan
caDepartmentofEconomics,Altınbas¸University,Istanbul,Turkey
bBo˘gazic¸iUniversity-CenterforClimateChangeandPolicyStudies,Turkey
cDepartmentofEconomics,IhsanDogramaciBilkentUniversity,Ankara,Turkey
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:
Received11August2017
Receivedinrevisedform9February2018
Accepted29April2018
Availableonline7May2018
Keywords: Carbonconvergence Climatepolicy Emissionintensity Sectoralemissions Turkey
a
b
s
t
r
a
c
t
Turkeyisknowntosufferfromseverevolatilityinitsgrowthpatterns,aswellasfromtheuneven sec-toralgrowthandemployment.Volatileratesofemissionsacrosssectorsarefurthermanifestationsof thisunevenstructure.Thepurposeofthisstudyistwo-fold:first,wecheckfordynamicpatternsof con-vergenceofcarbondioxide(CO2)emissionsacrosssectors;andsecond,usingevidencefrompaneldata
econometrics,wesearchforthedeterminantsoftheseprocessesutilizingmacroeconomicexplanatory variables.Wefindthat,basedonvariousalternatecriteria,CO2emissionsdisplayconditionalconvergence
mainlydrivenbythebusinesscycle.Furthermore,acrosssectors,hightechnologyactivitiesdisplay con-vergenceovertime;andyet,mediumtechnologysectorsthatconstitutethebulkoftheaggregatevalue addeddisplayeitherpoorlyconvergentordivergenttrends.Theseresultsrevealthatmuchofthe emis-sionsconvergenceisdrivenbythebusinesscycleratherthantheworkingsofdiscretionarymitigation policy.
©2018ElsevierB.V.Allrightsreserved.
1. Introduction
Environmentalconvergenceresearchhasbeeninspiredbythe conventionaleconomicconvergenceliterature,andhasused simi-larmethodologiestoinspectcross-countrydynamicsofemissions convergenceforvarioussamplesanddifferenttimespans.Theidea stemsfromthefactthattheevolution ofincome and pollution cannotbeseparatedfromeachother.Income generation gener-allyrequiresenergyuseandenergyisusuallygeneratedviathe extractionofnaturalresourcesandtheuseoffossilfuelsthatemit pollutantsandgasesincludingcarbon.Needlesstosay,theseare alldependentontheeconomicactivitiesofhouseholds,firmsand governments;andthechoicesoftheseactorsundercertainpolicy andtechnologyconstraintsplayaroleinchangingtheamountof emissionsthatarereleasedtotheatmosphere.
Inacontemporarysetting,CO2convergenceisespecially impor-tantinthisrespectduetotheongoingdiscussionsofinternational agreementsonemissionscutsandtheimplementationofpolicy toolssuchascarbontaxesorcarbontradingschemes.Aswidely
∗ Correspondingauthorat:DepartmentofEconomics,Altınbas¸University,
Istan-bul,Turkey.
E-mailaddresses:sevil.acar@altinbas.edu.tr,sevil.acar@boun.edu.tr(S.Acar),
yeldane@bilkent.edu.tr(A.E.Yeldan).
known,theParisAgreementaimstolimitglobalwarmingtobelow 2◦C, andfurthermotivate theinternationaleffortstolimititto 1.5◦C.Consequentlyitbearscollectiveresponsibilityforall coun-tries.Themostprominentfeatureofthepost-ParisAgreementera isthatallpartiestotheAgreementpledgesomeemission reduc-tiontargetsandadoptmeasurestocurbemissionsinaccordance withtheprincipleof“commonbutdifferentiatedresponsibilities andrespectivecapabilities”asstatedontheUnitedNations Frame-workConventiononClimateChange.Withinthisframework,all countriesneedtocontributetotheglobaleffortstofightclimate changebyvariousmeasuresandpoliciesinordertotransformtheir economicsectorsintolow-carbonforms.Hence,bothdeveloped anddevelopingcountrieshaveemissionreductionresponsibilities thatentaildecliningsectoralemissionsatdifferentiatedrates.
Put differently, a low-carbon economic development path acrosstheglobe requiresthat greenhousegas(GHG)emissions needtoconvergetolowerlevelsglobally,regionally,and secto-rially.Ifthereisweakevidenceofcarbonconvergenceinpercapita emissionsatthegloballevel,“globalagreementsthatimpose con-vergencemaygeneratelargere-distributionalimpacts,significant transfersofwealth,andthusaburdenoflargeadjustmentcosts” (Acaretal.,2018:7).Therefore,policymakersneedtooverreach singleallocationrulesandextendthepolicysphereacrossraising knowledgeabouttheimpactsofcombinedscenarios,and develop-ingnewapproachestoclimatechangeadaptationandmitigation. https://doi.org/10.1016/j.strueco.2018.04.006
Whatis more,“evenin thepossiblepresenceofcarbondioxide convergence, globalclimate policymust also consider whether countriesthentendtoconvergetohigherorlowerpercapita lev-els”(Acaretal.,2018:7).Giventhebroadconsensusthatthemain instrumentofclimatechangemitigationshouldfundamentallybe accompaniedwithincreasedefficiencyofenergyutilizationaswell astechnologicalandinstitutionalchangethatallowfordecoupling ofeconomicactivityfromenvironmentalpollution,thequestionof sectoralpatternsofcarbonemissionsbecomeofdueimportance.
Turkey is grappling withthe challenges of ensuring a cost-competitive energy supply for its growing population and the industrialsectors,ensuringenergysecurity,andreducingitsGHG emissions.Akeyhypothesisofthisstudyisthattheprojectedlackof decouplingbetweengrowthandGHGemissionsismostlydrivenby theunevenpatternsofgrowthandindustrializationacrossTurkey. Yeldanet al.(2013)suggestthat oneofthemaincausesofthe productivityslowdownofthe Turkisheconomy over the2010s isthedivergingpatternsofregionaldevelopmentandthe widen-inggapacrosshighversuslowincomeregions,aswellasmodern versustraditionalsectoralproduction(andconsumption)patterns. Besides,high-pollutionindustrieshavebeenrelocatingfromthe moredevelopedcitiesofthecountrytotheless-developedcentral andeasternregions.
On the other hand, across sectors Turkey is suffering from patterns of heterogeneity that are characteristic of many late-industrializing nations. Transfer of the rural labor surplus to modernsectorsoccursatanunevenpace,labormarketsremain fragmentedandheterogenoussimplybecausecapitalinmostcases remainsheterogenous.Underconditionsofdifferentiatedratesof growthsectorially,emissionstend tofollowdivergentpatterns, strainingpolicyeffectiveness.Webelievethatsuchrelocationof industriesthatbeardifferentenvironmentalpressurescouldalso playaroleininterregionalandintersectoraleconomicconvergence and/ordivergence.
Theanalysisofthecurrentstudydepartsfromindividual sec-torsinTurkey.Tothisend,wecheckforevidenceonconvergence ofCO2emissionsacrosssectorsutilizingvariousmethodsincluding paneldataeconometricsoversectoraldata.Theseexamplesare,of course,notgeneralizations,butshowhowcertainsectoral expe-rienceswouldfitthebigpictureofglobalcarbonconvergence(or divergence).Incontinuationofevidenceonconvergence,ifany, wesearch for theleadingindicators oftheseprocesses byway ofdifferentiatingtheproductionsectorsaccordingtotheirlevel oftechnologyandenergyutilization.Wedistinguishbetween pri-mary(low)technology,mediumtechnology,andhightechnology sectors.Besides,themainresultsaregivenaninterpretationfrom theperspectiveofthesectorswhichexperiencedthehighest trans-formationpressuresowingtohighoilpricesovertheinvestigated period(especiallyinthe2000s),e.g.coke,refinedpetroleumand nuclearfuel;chemicals andchemical products;and rubberand plastics.Hence,thisstudycontributestotheliteratureintwomain aspects.First,tothebestofourknowledge,thisisthefirststudy toundertakeananalysisofsectoralcarbonconvergenceinTurkey. Second,itemphasizestheneedtodistinguishbetweendifferent economicsectorswithrespecttotheirtechnologicalsophistication. Consequently,suchcategorizationhelpstotrackdiffering conver-genceand/ordivergencepatternsinCO2emissionsduetothefact thatsectorsmighthavevaryinglevelsoftechnologicalopennessor easeofaccesstoenergysavingtechnologiesthatmayhelpthem adoptexternaleconomiesofscaleandreducepollutionintensities. Thepaperisorganizedasfollows:Inthefollowingsection,we summarizetheGHGstatisticsofTurkeythroughthelens of cli-matechangebycastingcomparativefiguresfromtherestofthe world.We providea briefsurveyonthetheoreticalbackground andpertinentliteratureinsectionthree.Next,weintroduceour methodologyand data sources insection four.We study
alter-Fig.1.CO2emissions(metrictonspercapita).
Source:WorldDevelopmentIndicators(2018).
nativeconfigurationsofsectoralconvergenceinsectionfive.We summarizetheresultsandconcludeinsectionsix.
2. AglanceatTurkey’sgaseousemissionsthroughthelens
ofclimatechangeandeconomicgrowth
Turkey’s economy is known to display wide swings in its patternsofgrowthbothinaggregateandalsoinitssectoral compo-sition.Thestop-and-gopatternsofoutputgrowtharemanifested not onlyin terms of mini-businesscyclesof economic activity, butalsointermsofgaseousemissionsacrosssectors.Asof2015, Turkey’s totalemissions ofgaseous pollutants(in termsof car-bondioxideequivalent(CO2 eq.)is estimatedtobe475million tons (mtons). About three quarters of this is reported to arise fromenergy-relatedactivities,while61 mtonsareattributedto industrialprocesses.AccordingtodatafromtheTurkishStatistical Institute(2015),withtotalemissionsof6tonsofCO2eq.percapita in2015,Turkeydisplaysalowerfigureinemissionsinbothtotal andpercapitaaccountsthanmanycountries.TheWorldBank’s WorldDevelopmentIndicators(2018)revealthatCO2 emissions reached9.5,6.4,5metrictonspercapitafortheOECD,European Union(EU)andtheworldrespectivelyin2014,whereasTurkey’s percapitaCO2emissionswereat4.5metrictonspercapitainthe sameyear(seeFig.1).
However,Turkeyisalsoknowntodisplayoneofthehighest ratesofgrowthinCO2eq.emissionsamongtheemergingmarket economies.Turkey’saggregateCO2eq.emissionsincreasedfrom 214milliontonsin1990,to475milliontonsin2015,corresponding toacumulativeincreaseof122%duringtheperiod.Severalstudies projectthatthecountry’sGHGemissionswillkeepacceleratingin thenearfuture,climbing,forinstance,upto675milliontonsby 2030(AcarandYeldan,2016).ThissuggeststhatTurkeywillbeon adivergenttrendagainstmanyoftheemergingmarketdeveloping economiesaswellastheworldaveragesoverthenextdecades.
TheseassessmentsaresuccinctlynarratedinFig.2aandbbelow, wherewedisplaytherateofchangeinaggregateCO2eq.emissions againstchangesinrealGDPoverthepost-1990era.Theclose asso-ciationbetweentherealrateofchangeinCO2emissionsandthe realbusinesscyclesoverthisperiodisclearlyvisibleforTurkey, suggestingthatthemuchdesireddecouplingofgaseousemissions fromrealeconomicactivityhasnotyettakenplace.This obser-vationfurtherrevealsthelow elasticityofgaseous emissionsin responsetorealgrowth,andthatthereturnstoabatementpolicies hadratherbeendismal.ThisfactcontrastswiththeAsianemerging economies,whereasubstantialdecouplingofgaseousemissions fromrealGDPgrowthisobservable(seeFig.2aandb).
Besides,althoughCO2emissionsfromelectricityandheat pro-duction(asapercentageoftotalfuelcombustion)makethebulkof
Fig.2. aandb.CO2andGDPgrowthratesinTurkeyversusAsiancountries.
theemissionsinTurkeyaswellasintheOECD,EUandtheworld, thedistributionofemissionsacrossothersectorsshows variabil-ity.Forinstance,theshareofemissionsfromresidentialbuildings andcommercialandpublicservicesisfoundtobemuchhigherin TurkeythantheOECD,EUandtheworldaveragesin2012. Sim-ilarly,theshareofCO2emissionsfrommanufacturingindustries andconstructioniswellabovetheOECDandEUaverageswhereas itremainsbelowtheworldaveragein2012.Morestrikingly,while Turkey’semissionsfrommanufacturingdeclinedsharplyfollowing theglobaleconomicturmoil,thoseemissionsdidnotexperience seculardeclinesintheOECDortheEUonaverage.Emissionsfrom transportasashareoftotalfuelcombustion,ontheotherhand,are lowerthanthecorrespondingaverages(seeFig.3).Thedeclining trendoftransport-relatedemissionsinTurkeyfrom1960to2012
contrastswithincliningtrendsofsuchemissionsintheOECDand EU(WorldDevelopmentIndicators,2016).
Akeyhypothesisinthispaperisthattheprojectedlackof decou-plingbetweengrowthandemissionsmitigationismostlydriven bytheunevenpatternsofgrowthandindustrializationacross sec-torsinTurkey.Yeldanetal.(2013)suggestthatoneofthemain causesoftheproductivityslowdownoftheTurkisheconomyover the2010sisduetothedivergingpatternsofregionalandsectoral developmentandthewideninggapacrosshighversuslowincome regions,aswellasmodernversustraditionalsectoralproduction (andconsumption)patterns.Wearguethatthelackofmitigationat theaggregatenationallevelfindsitsmanifestationinthiswidening gapacrossregionalandsectoralstratificationofincome,production capacities,aswellastheconsequentcarbonandgaseousemissions.
Fig.3.SectoralCO2emissionsintheworld,EU,OECD,andTurkeyin2012(%oftotalfuelcombustion).
Source:WorldDevelopmentIndicators(2016).
3. Backgroundtheoryandliterature
Itis widelyknownthatconvergencein percapitaincome is rootedintheSolowmodel(Solow,1956), whichstipulatesthat countries(orregions)atlowerpercapitaincome levelstendto experiencehigher growth rates thanthe richerones. Thisidea isduetothetraditional(mainstream)assumptionofdiminishing marginalreturnstocapital.Theimplicationsofthishypothesishave beentestedfrequentlyintheempiricsofgrowthliterature.
Convergenceintermsofenvironmentalquality,asarecent con-cept,hasbeenstudiedinseveralrecentarticlesandreports.As pointed outby Brännlundet al. (2015),animportant aspectof this literature hasbeento analyzetheconditions under which aneconomycanachieve economicgrowthcombinedwith non-deterioratingenvironmentalquality.Onegeneralcondition,which emanatesfromtheoptimalityconditionsinadynamic neoclassi-calgrowthmodel,ispollution(ˇ)convergence,implyingthat,in thelong-run,pollutionshouldbeboundedaswellasapproacha steady-stateleveleveninthepresenceofpositivegrowthinper capitaGDP.
Inspiredbypreviouseconomicconvergenceresearch, environ-mentalconvergenceliteraturedevotesitselftoinvestigatewhether convergence acrossenvironmental indicators or theamount of pollutants(particularly,emissions) existsacrossvariousregions and time periods(e.g.List, 1999;Strazicich and List,2003; Lee andList,2004;Nguyen-Van,2005;Aldy,2006,2007;Bulteetal., 2007;Ezcurra,2007;WesterlundandBasher,2008;Camareroetal., 2008;PanopoulouandPantelidis,2009;BrockandTaylor,2010; andCamareroetal.,2013).Thebulkofthisresearchhasfocused oncarbonconvergence;byutilizingeithercross-countrydataor paneldatacomprisingofcountries(seePetterssonetal.(2014)for acomprehensivereview).Attheexpenseofover-generalization, themainfindingofthisresearchisthatconvergenceinpercapita carbondioxideemissionsisrealizedtosomeextentbetweenthe developed (OECD)countries,while evidencing relatively persis-tentgapsordivergenceatthegloballevel.Inaddition,studieson regionalconvergencehavealsoinvestigatedpatternsofpollutants acrossregions.Forinstance,List(1999)testsforconvergenceof SO2andNOxfor10USregionsduringtheperiod1929–1994and findslimitedevidenceofconvergence.Similarly,LeeandList(2004) conductunitroottestsforNOx inUS statesfrom1900to1994 demonstratingthatNOxemissionsarenot convergingsincethe seriesarenon-stationaryandcontainaunitroot.Aldy(2007)and
Bulteetal.(2007)arealsoamongthosewhoconcentrateonUS regionalemissions.
Researchonsectoralconvergence,however,remainsrelatively limited.Somestudiesfocus on“environmentalperformance”in search for environmentalconvergence. For instance,Brännlund etal.(2015)investigatetheconvergenceofCO2performanceacross the14Swedishmanufacturingsectorsfrom1990to2008.Theyfirst calculateanenvironmentalperformanceindexderivedfrom pro-ductionofboththegoodandbadoutputs.Thentheyestimatethe growthofthisindex(i.e.therateofchangeintheratiooftheinverse emissionintensity)basedontheinitialvalueoftheindexandother factorssuchassectoralcapitalintensity,fossilfueluse,fossilfuel price,valueaddedandEUETSparticipation.Theydetectconditional ˇ-convergenceinCO2performancetogetherwiththecontribution ofhigherfossilfuelpricestoimprovedCO2 performancein the Swedishindustrialsectorswhereastheyfindnosignificanteffect ofEUETSparticipation.Similarly,MorleoandGilli(2016)analyze theenvironmentalperformancesof14 manufacturingsectorsin theEU(excludingCrotia),using theratiobetweenvalueadded andcarbondioxideemissions(i.e.environmentalproductivity)as theperformanceindicatorfortheperiod1995–2009.Usingdata fromtheWorldInput-OutputDatabase(WIOD),theauthorsfind thatthereisevidenceforˇ-convergenceconditionalontherole ofvariablesliketradeopennessandpolicyforsectorial environ-mentalperformance.Thisimpliesthatsectorswithlowerinitial levelsofenvironmentalproductivitydemonstratehighergrowth ratesofenvironmentalproductivitythanthosethathavealready experiencedgoodenvironmentalperformances.Besides,theyfind thattradeopennessfostersenvironmentalproductivity,whereas technology,proxiedbythecompoundgrowthrateofthe knowl-edgestock(measuredthroughpatentapplications),appearstobe insignificant.Ontheotherhand,theydonotdetectanyevidence of-convergenceforthesamesample.
Recently, more studies have started to concentrate on sec-toralGHGemissionsinsearchforconvergence.WangandZhang (2014)studypercapitaCO2emissionsin28provincesandsix sec-torsinChina.Theyevidenceconvergenceinallthesectorsfrom 1996to 2010,and report differentfactorsthat lead to conver-gence.Forinstance,GDPpercapitaandpopulationdensityarethe determinantsofconvergenceintheindustrysectoraswellasin thetransportation,storage,postal,andtelecommunications ser-vicessector.ApartfromGDPpercapitaandpopulationdensity, tradeopennessalsoinfluencesconvergenceinthewholesale,retail,
trade,andcateringservices.Finally,convergenceofemissionsdue toresidentialconsumptionismainlyshapedbypopulationdensity. Anotherstudythatquestsforsectoralemissionsconvergence isMoutinhoetal.(2014),whichanalyzesCO2intensityofthe Por-tugueseindustry.Theauthorsfindsigmaconvergenceforallsectors aswellasprovideevidenceforthesignificantrolesoffossilfueluse andenergyconsumption indeterminingsectoralCO2 emissions andemissionsintensity.Withafocusonasinglesector,Moutinho (2015) divides thetourism sector into various subsectorswith respecttotheirenergyuseandinvestigatesthecarbonconvergence (anddivergence)patternsinthesesectors.Hedetectssigma con-vergencegenerallyinaccommodationandfoodservices,transport andwholesaleandretailtradesub-sectorswhenthedispersionof theiremissionintensitiesbetween1996and2003isconsidered; howeverhefindsdivergenceinthecorrespondingsectorsbetween 2003and2006.
IntheirsearchofwhetherthemeasurestakenwithintheEUto meetitscommitmentsforclimatechangemitigationhavehadthe desirableimpacts,Morales-Lageetal.(2017)questionif conver-genceoccurredovertheperiod1960–2012insectoralpercapita CO2emissionsintheEU.ByutilizingtimeseriestechniquestoCO2 dataforthe28membercountries,theauthorstestforstochastic andclubconvergence.Asaresult,theydetectmeagerevidenceof clubconvergenceduetothefactthatwhileseveralEUcountries keepincreasingtheiremissions,theothersdobetteratlimiting emissionsconsiderably.Yet,theauthorsobservemajordifferences amongthesectorsandsub-sectorsconsidered.For instance,the transportsector(asasub-sectorofenergy)isfoundtwodisplaytwo convergenceclubswithnineandsevenmembercountries respec-tively,whereasthereareeightnon-convergingcountriesinterms ofemissionsfromtransport.Similarly,twoconvergenceclubsofsix membercountrieseacharedetectedforthemanufacturingsector (again,asasub-sectorofenergy),whiletwodivergingclubswith twoandninecountriesrespectivelyareidentified.Finally, diver-genceforfifteen countriesisdetectedinemissionsfrompower generationandheating,which,ingeneral,isasignificantsource ofemissionsthroughouttheEU.
ApergisandPayne (2017)conducta similaranalysisfor fifty USstates(includingtheDistrictofColumbia)utilizingthe Phillips-Sulclub convergenceapproachfor theperiod1980–2013.Their findingsevidencethattherearemultipleconvergenceclubs“inthe aggregate,bysector(residential,commercial,industrial,transport, andelectricpower),and fortwoofthethreefossilfuelsources (naturalgasandcoal)withfullpanelclubconvergenceinthecaseof petroleum”(ApergisandPayne,2017:365).Theauthorsinterpret theseresultsasawaytorecognizetheneedtodesigndifferential environmentalpoliciesthatwouldidentifythedifferencesinthe convergencepathsofvarioussectors.
Oliveira and Bourscheidt (2017) investigate per capita GHG emissionsconvergence for a multi-sectorial panelof countries. Theymake useof random and fixedeffects panel datamodels aswellasArellanoandBond’s(1991)GMMestimator.Theyfind strongandrobustevidenceof“percapitaconvergenceinCH4 emis-sionsintheagriculture,food,andservicessectors”,whereas“the evidenceofconvergenceinCO2 emissionswasmoderateinthe followingsectors:agriculture,food,non-durablegoods manufac-turing,andservices”(OliveiraandBourscheidt,2017:402).
Tothebestofourknowledge,thecurrentstudyisthefirstof itstypeintheanalysisofsectoralcarbonconvergenceinTurkey. Despitenotsearchingforconvergence,Kumbaro˘glu(2011) con-ductsasectordecompositionanalysisofTurkey’sCO2emissions duringtheperiod1990–2007,andhighlightsthescaleeffectasthe majorsourceofemissiongrowthintheelectricity, manufactur-ing,and transportsectors.Heattributesemissiongrowthinthe householdandagriculturesectorstoenergyintensity.Comparing theleadinggrowthenginesoftheTurkisheconomyfortheperiods
1995–2002and 2003-2009,As¸ıcı(2015)alsoillustratesthatthe latterperiodischaracterizedbygrowthinmoreenergyintensive sectorsand“thecompositionoftheeconomicactivityis concen-tratedinmoreCO2andNOxintensivesectors”(As¸ıcı,2015:1738).
4. Methodology,dataandsources
Thenotionofconvergencecanbeinvestigatedthroughthree concepts:sigma()convergence,stochasticconvergence,andbeta (ˇ)convergence.
Tobeginwith, sigma()convergencetakesintoaccountthe dynamicsandtheintra-distributionalbehaviorofaselected emis-sionseries.BarroandSala-i-Martin(1992)describe-convergence asthedecreaseinthecross-sectionvarianceofemissionsovertime. Uptothisaim,cross-sectional varianceorstandarddeviationis simplyplottedtodetectconvergence.Otherstudieshaveexamined thebehaviorofrelativeemissions(REit),whererelativeemissions aremeasuredasthelogofonecountry’sorsector’semissions(y) attimetdividedbytheyearlysampleaverage ¯yt,asnotatedby CarlinoandMills(1993)asfollows:
REit=ln(yit/yt) (1)
Second, stochastic convergence focuses on the time series characteristicsoftheemissionseries.Usingtimeseriesanalysis, stochasticconvergencecanbeexploredtodetectwhethershocks toemissionsforcountryorsectorirelativetoanothercountryor sectorj(ortheaverageofthesample)aretemporary(seePettersson etal.(2014)forfurtherdetails).Ifthetimeseriesofinterestdoesnot containaunitrootandisproventobetrendstationary,theseries isfoundtobestochasticallyconverging.Manystudiesincluding StrazicichandList(2003);LanneandLinski(2004);McKitrickand Strazicich(2005);Romero-Ávila(2008); Westerlundand Basher (2008); Lee and Chang (2009); Nourry (2009), and Yavuz and Yilanci(2013)makeuseofvariousunitrootteststotrace stochas-ticconvergenceofemissionsindifferentsamplesofcountries.This methodcanalsobeimplementedforpaneldatabyusingpanelunit roottechniques,whichwillbeemployedinthenextsectionofthe currentstudy.
Third,beta(ˇ)convergenceoccurs“whentheemissionsofa poorercountry,withlowerinitialslevelsofemissionspercapita, tendtogrowfasterthantheonesfromarichcountryandthereisa catching-upeffectwiththemorepollutingcountries”(Pettersson etal., 2014:149). ˇ-convergencecanbe investigatedboth in a cross-sectionandpaneldatasetting.Thecross-sectionalapproach implies thatconvergence is examinedbyregressing thelogged periodgrowthrateofemissionsln(yit/yi0)(forthewholesample) ontheinitialloggedemissionlevels,lnyi0,andanerrortermεifor country,region,orsectoriasinbelow(Petterssonetal.,2014:150): ln(yit/yi0)=˛+ˇln(yi0)+εi (2) where εi is theerrortermfor countryorregioni. Accordingly, ˇ<0impliesconvergence.Similarly,panelˇ-convergencecanbe analyzedasinthefollowingequation(Petterssonetal.,2014:151): ln(yit/yi,t−)=˛+ˇln(yi,t−)+ıi+t+εit (3) whereln(yit/yi,t−)isthegrowthrateofemissionsbetweent− and t, and ı demonstrates sector-specific effects, and repre-sentsperiod-specificeffects.Thismodelspecificationhelpstotest whetheremissiongrowthratesconvergeacrosscross-sectionunits bytime;i.e.whethertheyareeagertoslowdowninthelong-run astheyapproachtheirownlong-rungrowthpath.
Intheirmeta-analysisofthecarbonconvergenceliterature,Acar et al.(2018) detectthat the choiceof differentcarbon conver-genceapproachesmayendupwithdifferentresults.Forinstance, thechoice of theˇ-convergence concepttends toincrease the
Table1
Variablesusedintheanalysis.
Abbr. Definitionofthevariable Unit Datasource CO2 CO2emissions Gg(kt) WIOD VA SectoralValue-added TLs(million) WIOD KSTOCK Capitalstock TLs(million) WIOD EN Emissionrelevantenergyuse TJ WIOD
likelihoodofconvergence,whereasthechoiceofthestochastic con-vergenceconceptoftentendstoprovecarbondivergence.Besides, resultsmayvaryalotevenamongthestudiesthatfocuson stochas-ticconvergence,becauseutilizingdifferentunitroottestsinsearch forstationarityinemissionsseriesusuallyimpliesthatdifferent assumptionsprevail.AnoteworthyexamplefromPetterssonetal. (2014)isthat“accountingforstructuralbreaksandcross-sectional dependenciestendstofavourthestochasticconvergence hypoth-esisinstudiesbasedonpaneldata”(Acaretal.,2018).
Inourinvestigationofˇ-convergence,weutilizetwoseparate emissionsindicators,onebeingthegrowthrateofsectoral emis-sions(CO2)andtheotherbeingthegrowthofsectoralemission intensitydefinedastheratioofCO2 emissionstosectoralvalue added (CO2/VA).Themotivation behindstudying theformer is highlylinkedtotheabsoluteemissionreductiontargetsinregards totheglobalclimatechangeconcerns.Scientistshavewarnedthat lifeonEarthwillfaceunforeseeableadverseconsequencesandbe seriouslythreatenediftheincreaseinsurfacetemperatureexceeds 2◦C;thus,thescientificcommunitysetanultimatetargettolimit theriseinglobaltemperaturebythisamount(2◦C).Bringingthis aimtothesectorallevel, ideally,thetotal allowableCO2 emis-sionsshouldbeloweredsignificantly.Hence,sectoral“absolute” emissionsneedtoconvergetolowerlevelsinlinewiththeglobal climatechangeconcerns.Whiletheuseoftotalsectoralemissions issimpleandintuitive,ithassomeshortcomings.Forinstance,it doesnotrecognizeemissionreductionactionsimplementedbefore 1995,whichisthestartingyearforoursampleperiod,anditdoes noteasilyaccommodatechangesinasector’scircumstances.That’s whywealsomakeuseofrelativeemissions.Thereasonofthechoice ofsectoralvalueaddedasthedenominatorinthelatteristhatwe investigateaheterogeneoussampleofsectorsinouranalysis,and hencevalueaddedisusedasaproxyforeachsector’scontribution toGDP.
Alongside, we focus on the coefficient (ˇ) of the previous emissionsandemissionintensitiesrespectivelyinsearchfor con-vergence,wherethenullhypothesisofdivergenceisH0:ˇ=0forall i;andthealternativehypothesisofconvergenceisHa:ˇ<0forall i.AnegativesignforˇimpliesunconditionalconvergenceinCO2 emissions.Addingcontrolvariablessuchassectoralvalueadded (VA)measuredin fixed prices, realcapital stock (KSTOCK) and energyuse(EN)toEq.(3)entailstestingconditionalconvergence. Ourmodelsareestimatedviapanelfixed-effectsanddynamic panel(Arellano-Bond)specifications.Panelconvergencehas fre-quentlybeenaddressedbyeitherfixedorrandomeffects inthe literature.However,itisplausibletoincludesomedynamiceffects intothestandardpanelmodelsincegrowthofemissions accom-modatesdynamic effects withrespecttothepreviousemission growthrates.Ineconometrictheory,thesedynamiceffectscanbe integratedintothemodelviatheinclusionofalaggeddependent variableamongtheregressors.Whiledoingso,thelagged depen-dentvariablemightbecorrelatedwiththeerrortermespecially insmallsamples,whichcomesoutasaproblem.Aninstrumental variablespecificationispreferredtotacklethisproblemand,more specifically,theGeneralisedMethodofMoments(GMM)modelcan beemployedusingthelaggedvaluesofthevariablesintheoriginal modelasinstruments.Amongseveralapproachestodynamicpanel datamodels,Arellano-Bondspecificationis themostcommonly
Table2
Valueaddedsharesofsectorsaccordingtotechnologyutilization.
Value-addedshares
1995 2013
Primary/LowTechnologySectors 0.18 0.11 MediumTechnologySectors 0.74 0.81 HighTechnologySectors 0.08 0.08 Source:WIODdatabasedontheOECDclassificationoftechnologyadoption.
usedone.Itaccountsforindividualorfixedeffectsbydifferencing thedata.Besides,itisthemorefavourableapproachandresults inconsistentestimateswhenthenumberofcross-sections,N,is higherthanthenumberoftimeperiods,T(Baltagi,2005:136).
ThesectoralvariablesusedinthemodelsaredescribedinTable1 below:
AllourdataareadaptedfromtheWorldInput-OutputDatabase (WIOD)1,andarefurthersupplementedbytheTurkstatdataonCO
2 emissionsincludingemissionsfromenergy,industrialprocesses andproductuse,agriculture,andwaste.Thenamesandthe clas-sificationofthesectorsthatareunderconsiderationareprovided inAppendixA2.Thesummaryofdescriptivestatisticsforthe vari-ablesofinterestisprovidedinAppendixA3.Wefurtherclassify oursectorsintermsoftheirtechnologylevels,asprimary(low), mediumandhightechnology-drivenactivities.Thiscategorization isbasedontheOECDclassificationoftechnologyadoption.WIOD datarevealsthat,thebulkofthemanufacturingsectorsdisplay mediumtechnologycharacteristicsandtheshareofmedium tech-nologysectorsaccountfor81%oftotalvalueaddedin2013(see Table2).
Fig.4furtherdisplaysthedistributionofsectoralCO2emissions inTurkeyoverthesampleperiod.Inabsoluteemissions, Electric-ity,GasandWaterSupply(no.17)andTransport(no.21)standout astheprominentsectors,whereasLeatherandFootwear(no.5)is theleastemittingsector.Thetimedispersionoftheemissionsis thewidestforHotelsandRestaurants(no.20)aswellasWoodand ProductsofWoodandCork(no.6)asillustratedintheboxplot. Theboxesareboundedbythefirstandthirdquartilesofthedata, enclosingthemiddle50%ofthesample.Thedotsillustratethe out-liers;thelinesacrosseachboxshowthemedians;andthe“+”signs indicatethe“mean”observationsforeachsector.Itisrevealedthat thesectorsunderconsiderationbehavequitedifferentlyintheir meanandmedianemissionsduringthe1995–2013period.When wecomparesectoralemissionswithrespecttosectoralvalueadded amounts,Coke,Refined Petroleumand NuclearFuel(no.8)and Electricity,GasandWaterSupply(no.17)arenoticeablythe sec-torswhichareperformingbadly.Othereconomy2(no.22)releases thelowestamountofCO2pervalueaddedamongothersectors.
5. Empiricalresultsonpatternsofconvergenceofsectoral
gaseousemissions
5.1. -convergence
Inordertoperformadistributionalanalysisofemissionsinthe Turkishsectors,weplotthenaturallogarithmoftheratioofCO2 ineachsectordividedbyaverageCO2 emissionsinallsectorsin thatyear,i.e. logrelativeemissions.Tothatend, Fig.5
demon-1SeeTimmeretal.(2015)andthewebsitehttp://www.wiod.org/newsite/home.
htmforthedetailsoftheWIOD.
2Othereconomyiscomprisedofthefollowingsectors:Postand
Telecommu-nications;FinancialIntermediation;RealEstateActivities;RentingofMachinery
andEquipmentandOtherBusinessActivities;PublicAdministrationandDefence;
CompulsorySocialSecurity;Education,HealthandSocialWork;OtherCommunity,
Fig.4.SectoralCO2emissionsdistributionfor1995–2013. Note:Thenamesofthesectorsfrom1to22correspondingtothex-axisareprovidedinAppendixA2.
Fig.5. EvolutionoflogrelativeCO2emissionsineachsector,1995–2013.
stratessignsofconvergencetosomeextent,especiallyaccelerating followingtherecentglobalcrisis.Ithastobenoted,inthis junc-ture,thatthe2008/09crisishadaprofoundimpactonthenatureof thisconvergence.Fig.6isadirectillustratorofthisphenomenon, whereaveragelogrelativeCO2emissionsforallsectorsincrease initially,makeapeakin2003,declinesubstantiallyafterwards,and hitthebottomin2008.Therehasbeenarecoveryinmeansectoral emissionsfollowingtheglobalturmoil.
Finally,assuggestedbyBarroandSala-i-Martin(1992),Table3 displays-convergenceformulatedby“standarddeviation”,which servesasameasureofcross-sectionalvariationofemissionsover time.Apparently,thestandarddeviationofemissionsdecreased by7%from1995to2013,documentingsigmaconvergenceinthe sectors.
5.2. Stochasticconvergence
Inordertotestforstochasticconvergence,wefirsttestfor cross-sectionaldependenceforthethreerelevantvariablesderivedfrom thesample:naturallogofCO2emissions(LNCO2),CO2emissions asashareofsectoralvalueadded(CO2/VA),andlogrelativeCO2 emissions(LNRELCO2).Theresultsaredisplayedin A4. Accord-ingly,werunfirstgenerationpanelunitroottestsforCO2/VAas wecannotrejectcross-sectionindependence,whereaswerun sec-ondgenerationpanelunitroottestsforLNCO2andLNRELCO2as thecross-sectionsforthesevariablesexhibitcross-section depen-dence.
Amongseveralfirstgenerationpanelunitroottests,Imetal. (2003)andBreitung(2000)tests,whicharethetwowidelyused
Fig.6.MeanlogrelativeCO2emissionsforallsectorsbyyear.
Table3
Standarddeviationofcross-sectoralCO2emissionsfrom1995to2013.
1995 2013 %Changebetween1995and2013 Standarddeviationofcross-sectoral(log)CO2emissionsfrom1995to2013 1,66 1,55 −7%
panelunitroottests,areemployedhere.Themethodologyisas follows.ConsideringanAR(1)processforpaneldata,yitismodeled as:
yit=iyit-1+Xit␦i+εit (4)
wheretandistandfortimeandcross-sectionunits,respectively. IndividualunitroottestssuchasIPS, Fisher-ADF,andFisher-PP allowdifferingiacrosscross-sections,whereascommonunitroot testssuchasLLC,BreitungandHadriassumeacommonunitroot process,therebytakingidenticali=acrosscross-sections,i.e.for alli.IPStestprovidesindividualtestsforeachseries.Thenulland alternativehypothesesoftheIPStestareasfollows:Ho:Allpanels containunitroots.Ha:Somepanelsarestationary.Inotherwords,IPS assumesthatatleastoneoftheseriesisstationaryunderthe alter-nativehypothesis.Ontheotherhandthecorrespondinghypotheses fortheBreitungunit-roottestarestatedasfollows:Ho:Panels con-tainunitroots.Ha:Panelsarestationary.Breitungillustratesthatthe IPStestssuffersfromasignificantlossofpowerwhen individual-specifictrends are includedtothe test and hisalternative test statistic“doesnotemployabiasadjustment”(Baltagi,2005:243). Assuch,theBreitungtestimpliesstrongerresultsthantheIPS.In bothtests,therejectionofaunitrootandthepresenceof station-arityimplyconvergence,whereasthenon-rejectionofaunitroot impliesdivergence.
For LNCO2 and LNRELCO2, we employ Pesaran’s CADF test (2007), which is a second generationpanel unit root test. The testallowstheindividualautoregressiverootstodifferacrossthe cross-sectional unitsandis normallydistributed underthenull hypothesisofnon-stationarity.
According to Table 4, the IPS and Breitung test results for CO2/VAsuggestthatunitrootscannotberejectedinthemajority ofthespecifications,implyingnon-stationarityandhence, stochas-ticdivergence.Howeverbothtestsimplyconvergencewhenthe seriesarede-trended.Pesaran’sCADFtestalsoshowsthatsectoral emissionsandrelativeemissionsinlogarithmsdonotconvergeas
thenull hypothesisofunitrootscannotberejected.Tosumup, theseresultsprovidestrongsupportforadivergingpatternin sec-toralemissionlevelsandpoorevidenceforconvergenceinemission intensity.
5.3. ˇ-convergence
Asdescribedinsectionfour,ˇ-convergenceisanalyzedviapanel dataregressiontechniqueshere.Table5demonstratestheresults oftheanalyseswhichareundertakenforthewholesample.Models 1FEand1ABrepresentFixedEffectsandDynamicGMM (Arellano-Bond)modelsrespectivelywiththegrowthrateoflogsectoralCO2 emissionsasthedependentvariable;whereas2FEand2AB repre-sentthecorrespondingmodelswithgrowthrateofCO2/VAasthe dependentvariable.Accordingly,theindependentvariablesarein naturallogarithmsinmodels1FEand1AB,whiletheyare trans-formedintosharesinvalueaddedofeachsectorinmodels2FEand 2AB.
Theresultsimplyconditionalˇ-convergenceinallcases,with theexceptionof1FE,withslightdifferencesregardingtheeffects oftheexplanatoryvariables.It appearsthatsectoralenergyuse increasestheemissionsgrowthratesignificantlywhereas indus-trialvalueaddeddecreasesemissionsgrowthratecontributingto convergenceinmodel1AB.Thismightstemfromtheexistenceof economiesofscaleastheindustryproduceshighervalueadded. Thatistosay,whenthesectorshavealoweroutputlevel,they wouldproducesomeamountofthe“bads”,i.e.emissions.As sec-torsgrow,theydonotnecessarilyincrease theiramountofCO2 proportionallytotheiroutputgrowthsincetheywouldrequire rel-ativelylessenergyorotherinputsperoutputastheproduction scaleincreases.Besides,sectoralcapitalstockhasaslightly signifi-cantpositiveimpactonthegrowthofemissionsinonlyoneofthe models(1AB).
Finally,Table6displaystheresultsoftheconvergenceanalysis inthreesectorsclassifiedaccordingtotheirtechnologylevels:
pri-Table4
Panelunitroottests.
LNCO2 CO2/VA LNRELCO2
IPSteststatistic Withadriftandtrend −0.7533(0.2256) Withoutadrift,withtrend −2.3591(0.0092)
Withadrift,withouttrend −1.7694(0.0384)
Withoutadriftandtrend 2.8945(0.9981) Breitungteststatistic Withadriftandtrend 0.4861(0.6866) Withoutadrift,withtrend 0.2554(0.6008) Withadrift,withouttrend −1.8267(0.0339)
Withoutadriftandtrend 1.2558(0.8954)
Pesaran’sCADFteststatistic Constant −0.426(0.335) −1.247(0.106) Constantandtrend −0.0667(0.252) −0.942(0.173) p-valueinparentheses.
Table5
Fixedeffectsanddynamicpaneldataestimation(Arellano-Bond)resultsforthewholesample.
(1FE) (1AB) (2FE) (2AB)
GrowthofCO2 GrowthofCO2 GrowthofCO2/VA GrowthofCO2/VA
CO21 −0.190*** −0.222*** −0.012*** −0.006** (−2.73) (−3.98) (−4.35) (−2.41) VA 1 −0.049 −0.053* (−1.08) (−1.94) KSTOCK1 0.009 0.053* 0.00001 −0.0002 (0.19) (1.91) (0.03) (−0.84) EN1 −0.112 0.197*** 0.0009*** 0.0006** (−1.23) (3.20) (3.14) (2.40) Constant 3.059*** −0.469** 0.126*** 0.053** (4.89) (−2.03) (2.77) (2.11) Observations 396 396 396 396 F 21.26 12.90 P>F 0.000 0.000 r2o 0.017 0.006 chi2 20.95 7.19 P>chi2 0.000 0.066 Sargan 315.24 316.37 sarganp 0.847 0.795 ar1 −1.97 −11.06 ar1p 0.049 0.000 ar2 0.68 0.55 ar2p 0.495 0.585
tstatisticsinparentheses.Denotations(F:F-Value,r2o:OverallR-Square,chi2:Chi-Square,p:P-Value).
* p<0.1.
** p<0.05.
***p<0.01.
Table6
Fixedeffectsestimationresultsforsectorsclassifiedwithrespecttotechnology.
LOWTEC Growthof CO2 MEDTEC Growthof CO2 HITEC Growthof CO2 LOWTEC Growthof CO2/VA MEDTEC Growthof CO2/VA HITEC Growthof CO2/VA CO21 −0.426* −0.124 −0.694*** −0.050 −0.009** −0.021** (−1.71) (−1.54) (−2.94) (−1.66) (−2.52) (−2.46) VA1 −0.363 −0.014 −0.054 (−0.90) (−0.18) (−0.87) KSTOCK1 0.287 −0.025 −0.131 0.001 −0.001 0.006** (1.30) (−0.38) (−0.92) (0.31) (−1.12) (2.55) EN1 0.218 −0.183 0.376 0.003 0.000 0.001 (0.82) (−1.65) (1.33) (1.03) (0.56) (1.55) Constant 0.499 3.416*** 3.319** 0.189 0.289*** −0.055 (0.33) (4.20) (2.62) (0.54) (3.41) (−0.48) Observations 54 270 72 54 270 72 F 1.813 13.199 8.074 2.236 11.565 4.584 p>F 0.117 0.000 0.000 0.066 0.000 0.001 r2o 0.000 0.013 0.075 0.098 0.005 0.015
tstatisticsinparentheses.Denotations(F:F-Value,chi2:Chi-Square,p:P-Value,r2o:OverallR-Square).
* p<0.1.
** p<0.05.
mary(low)technology(LOWTEC),mediumtechnology(MEDTEC) andhightechnology(HITEC).Asthenumberofobservationsdoes not satisfy model assumptions, we are not able to conduct a dynamicanalysisforthespecifiedsectors.Henceweproceedwith panelfixedeffects.
Theresultsimplythatthesampleofmedium-techsectorsdoes notsupportˇ-convergenceinCO2emissionlevels,whereas low-andhigh-techsectorsexperienceabsoluteconvergence(although low-techsectors donot havea highlysignificantcoefficientfor theirpastemissions,implyingweakerconvergence).Thelackof support for convergence in the absolute level of emissions of mediumtechnologysectors,whichconsistofthebulkofthe Turk-ishmanufacturingindustries,isclearlythemaindrivingfactorin therelativelylow degreeof convergenceat theaggregatelevel (observedviathecorrespondingbetacoefficientsabove).
Whenwedealwithemissionintensities(CO2/VA)instead,we findthatthemedium-techandhigh-techsectorsprovideevidence forconvergencewhilelow-techsectorsdonot.Itcanbearguedthat theconvergenceasobservedwithinthehightechnologysectorscan beattributedtotheirdynamicandopencharacter.Opennessand relativeeaseinaccesstoadvancedtechnologywouldhavehelped thesesectorstointernalizetheexternaleconomies ofscaleand therebyreducetheirpollutionintensities.Besides,owingtohigh oilpricesover theinvestigated period(especiallyinthe2000s), high-techsectors,whichareatthesametimeoil-intensive sec-torssuchascoke,refinedpetroleumandnuclearfuel;chemicals andchemicalproducts;andrubberandplastics,experiencedhigh transformationpressures.Ontheotherhand,capitalstocksinthe high-techsectorsplayapositiverolesoastoacceleratethegrowth rateofCO2/VA.
6. Conclusion
Inthispaperwesearchedfortheexistenceandnatureof conver-genceofcarbondioxideemissionsfortheTurkisheconomyunder conditionsofunevengrowth.Weappliedaseriesof economet-ricteststodeducepatternsofconvergence,bothattheaggregate –economy-widelevel,aswellasacrosssectors.
Thesimplestmetricweutilizedwasthemeasureofstandard deviationsfromthemean,i.e.,the“-convergence”.Thismeasure wasfoundtoindicateconvergenceintheaggregate.Acloser inves-tigationrevealsthatthemaindrivingfactorbehindthisresulthad beenthebusinesscycle.Inparticular,therepercussionsofthe2009 globalcrisisareobservedtohaveaprofoundimpacton accelerat-ingtheconvergenceoftheCO2emissionsbywayofeveningout thefluctuationsoftheaggregateeconomicactivity.
Second,wefocusedonthedynamicsofstochasticconvergence. ThisanalysiswascarriedbothonthelevelofCO2emissions,and alsoonCO2intensity,i.e.,CO2pervalueadded(CO2/VA).Wefound thatsectoralCO2emissionsperunitofvalueaddeddepict stochas-ticconvergence(whende-trended)corroboratingourfindingthat theCO2emissionsfollowthebusinesscycle.Attheaggregatelevel ofCO2emissions,however,patternsofconvergencearedissipated andgivewaytoadivergingtrend.Wethensearchedforevidence onˇ-convergencetestedinconditionalterms.Hereweregressedthe rateofgrowthofthelevelofCO2emissionsontheoneperiodlagged valueof thefollowing explanatoryvariables:CO2, valueadded, physicalcapitalstock,andenergyutilization.Inasecondvariantof thismodel,therateofchangeofCO2/VAintensitieswereregressed againsttheperunitvalueaddedratiosofthesamevariables,K/VA andEN/VA.Ourresultsimpliedconditionalconvergenceinmost ofthecasesspecified.Energyuseappearedtobethemost promi-nentindicatorthatdroveemissionsgrowthandemissionintensity growthinthewholesample.
Finally,wedistinguishedtheaggregateeconomyundera three-tiersectoralspecificationbasedontheirtechnologycharacteristics: low,medium,andhigh.Wefindthatwhilethehightechnology sec-torsdisplaystrongconvergence,themediumtechnologysectors –thebulkofTurkey’seconomyaccountingfor80%ofthe aggre-gatevalueadded,doesnotsupportˇ-convergenceinCO2emission levels.Ourresultsfurtherrevealedthatthephysicalcapitalstock failstogenerateastatisticallysignificantimpactonCO2emissions (exceptitspositiveroleonthehigh-techsectors’emission inten-sitygrowth).Thisisanunexpectedresultgiventheratherstrong capitalintensityoftheTurkishgrowthpath,especiallyover the 2000s.Weinterprettheseobservationsasaresultofthelackof anyviablede-couplingduetothepersistentstructuralrelianceon energyresourceswithheavycoalandotherfossilfuelintensities. Nevertheless,ourresultsregardingphysicalcapitalareconsistent withBrännlundetal.(2015),whostudycarbondioxide conver-genceacrossSwedishindustrialsectorsandfindthathighercapital intensityintheselectedindustriesgivesrisetosloweremissions convergence.Theyexplainthisfindingbythefactthatthe replace-mentofindustrialequipment,buildings,andinfrastructurewith lowcarbononesisusuallycostlyandtime-consuming,leadingto slowerornocarbonconvergenceforthecapital-intensivesectors. To sumup,ourfindings do not supportanyconclusive evi-denceonthepatternsofconvergence.Testsofunitrootsindicate thatCO2/VAshowsdivergence,withevidenceofconvergenceonly whentheseriesarede-trended.Yet,wefinddiverging patterns acrosssectorsonthebasisof“aggregate”emissions.If,ontheother hand,weintroduceanalysesof“conditional”beta-convergence,we findconvergence.Thediversityoftheseresultsleadustosuggest thatsectoralemissionsfollowingeneralthebusinesscyclerather thanindicatingde-couplingofsectoralemissionsfrom correspond-ingvalueaddedincreases.
Still,thereareimportantaspectstoconsiderthatmayfurther advancesuchanalyses.Thefirstimportantpointisthatcarbon con-vergenceisahistoricalprocess.Inotherwords,carbonconvergence evolvesovertime.Thisimplies,inturn,thattherelevant histori-calcontextoftheprocessofsectoralcarbonconvergenceislikely tochange.However,itisdifficulttodeterminewhattherelevant historicalcontextexactlyis.Worldenergypricesareonesuch fac-tor,butalsoglobalpowerpoliticsandenergysecurityissuesare certainlyprimecandidatesespeciallywhenemissionsemanating fromoilcombustionareconsidered.Thisbringsus toasecond importantpointwhencarbondioxideconvergenceisconsidered; namelythattheprimaryenergysourcesthatgiverisetocarbon emissions(oil,coalorothersources)arerelatedtopartlydifferent technologies,andhence,sectorsthatutilizetheafore-mentioned sourcesofenergyastheirinputsareexpectedtoexperience dif-ferent transformations.Forinstance,highoilpricesconstitutea powerful transformation pressure and an incentive for techno-logical changein theoil-consumingsectors. Aresponsetosuch transformationpressureswouldbeconsistentwithclearevidence forˇ-convergenceinhigh-techandmedium-techoil-consuming sectors.
Severalpolicyimplicationscouldbederivedfromtheseresults inrelationtothemeasuresappliedtocurbemissionsinhigh emit-tingsectors.First,asectoralfocusshouldbethemaincentreof emissionreductiontargetsiftheaimsof‘greening’aretaken seri-ously.Inouranalyses,mostlythelow andmediumtechsectors provetobeexperiencing non-convergingpatternsin emissions. Highemittersliketransport,electricity,gasandwatersupply sec-torsneedtobethemainsectorsifthecountryintendstoimplement effectivepoliciestomitigateCO2emissions.Second,asthe emis-siongrowthratesaremostlyattributabletotheenergyintensities inthesectors,itappearsnecessarytoreconsiderthepatternsof energyusetakingintoaccountthefactthatfossilfuelsarecurrently themostdominantenergysourcesforthesesectors.Third,as
tech-nologylevelmakesadifferenceintheconvergencecharacteristics, the countrycould try to transform or diversify its technologi-cal sophistication towards cleaner options. Repetto (1990: 38) suggeststhat “technologiesthatreduceenvironmentaldamages contributeto economic productivity,even though theyare not costlesstoinstallortooperate”.Demotivatingordisincentivizing technologicalchangesthattriggerthereleaseofharmfulemissions intotheenvironmentorthatfacilitatetheintensiveuseofnatural resourceinputsmightbeadesirablepolicyoption.
Lastbutnottheleast,thereisaclearneedforfurther environ-mentalpoliciesandregulationstocopewithfuturecarbondioxide emissionsacrossthesectors.AsectoralCO2convergenceanalysis ofthiskindmightprovideinsightsabouttheimpactsofrelevant energyandclimatepoliciesonindustrieswithdiffering characteris-ticswithrespecttotechnology,capitalcompositionandenergyuse. Weproposethattheefficacyofenvironmental(andenergy)policy ultimatelyrestswiththerateofconvergenceofsectoralemissions alongwithawarranteddownwardtrendinemissionsperunitof output.
Acknowledgements
Theauthorsgratefullyacknowledgetheresearchsupport pro-videdbyTUBITAK, undergrantno114K941.Apreviousversion ofthepaperwaspresentedattheMiddleEastEconomic Associ-ation(MEEA)meetingsinconjunctionwiththeASSAConference, SanFrancisco,inJanuary2016.WearegratefultoMineC¸ınarand theparticipantsoftheMEEAfortheirinvaluablecommentsand toYasinKütükforhisinvaluableassistancewiththeeconometric analysis.Needlestostate,noneofthembearsanyresponsibilityfor theresultsandviewsexpressedinthepaper.
AppendixA. Supplementarydata
Supplementarymaterialrelated tothis article canbefound, intheonlineversion,atdoi:https://doi.org/10.1016/j.strueco.2018. 04.006.
References
Acar,S.,Söderholm,P.,Brännlund,R.,2018.Convergenceofpercapitacarbon
dioxideemissions:implicationsandmeta-analysis.Clim.Policy18(4),
512–525https://doi.org/10.1080/14693062.2017.1314244.
Acar,S.,Yeldan,A.E.,2016.EnvironmentalimpactsofcoalsubsidiesinTurkey:a
generalequilibriumanalysis.EnergyPolicy90,1–15,http://dx.doi.org/10.
1016/j.enpol.2015.12.003.
Aldy,J.E.,2006.Percapitacarbondioxideemissions:convergenceordivergence?
Environ.Resour.Econ.33(4),533–555.
Aldy,J.E.,2007.Divergenceinstate-levelpercapitacarbondioxideemissions.Land
Econ.83(3),353–369.
Apergis,N.,Payne,J.E.,2017.PercapitacarbondioxideemissionsacrossU.S.states
bysectorandfossilfuelsource:evidencefromclubconvergencetests.Energy Econ.63,365–372.
Arellano,M.,Bond,S.,1991.Sometestsofspecificationforpaneldata:MonteCarlo
evidenceandanapplicationtoemploymentequations.Rev.Econ.Stud.58, 277–297.
As¸ıcı,A.A.,2015.OnthesustainabilityoftheeconomicgrowthpathofTurkey:
1995–2009.Renew.Sustain.EnergyRev.52,1731–1741.
Baltagi,B.H.,2005.EconometricAnalysisofPanelData,3rdedition.John
Wiley&SonsLtd.,WestSussex.
Barro,R.J.,Sala-I-Martin,X.,1992.Convergence.J.Polit.Econ.100(2),223–251.
Brännlund,R.,Lundgren,T.,Söderholm,P.,2015.Convergenceofcarbondioxide
performanceacrossSwedishindustrialsectors:anenvironmentalindex approach.EnergyEcon.51,227–235.
Breitung,J.,2000.Thelocalpowerofsomeunitroottestsforpaneldata.In:Baltagi,
B.(Ed.),NonstationaryPanels,PanelCointegration,andDynamicPanels, AdvancesinEconometrics,vol.15.JAI,Amsterdam.
Brock,W.A.,Taylor,M.S.,2010.Thegreensolowmodel.J.Econ.Growth15,
127–153.
Bulte,E.,List,J.A.,Strazicich,M.C.,2007.Regulatoryfederalismandthedistribution
ofairpollutantemissions.J.Reg.Sci.47(1),155–178.
Camarero,M.,Picazo-Tadeo,A.J.,Tamarit,C.,2013.ArethedeterminantsofCO2
emissionsconvergingamongOECDcountries?Econ.Lett.118,159–162.
Camarero,M.,Picazo-Tadeo,A.J.,Tamarit,C.,2008.Istheenvironmental
performanceofindustrializedcountriesconverging?A‘SURE’approachto testingforconvergence.Ecol.Econ.66(4),653–661.
Carlino,G.A.,Mills,L.O.,1993.AreU.S.regionalincomesconverging?J.Monet.
Econ.32,335–346.
Ezcurra,R.,2007.Istherecross-countryconvergenceincarbondioxideemissions?
EnergyPolicy35(2),1363–1372.
Im,K.S.,Pesaran,M.H.,Shin,Y.,2003.Testingforunitrootsinheterogeneous
panels.J.Econometr.115,53–74.
Kumbaro˘glu,G.,2011.AsectoraldecompositionanalysisofTurkishCO2emissions
over1990-2007.Energy36,2419–2433.
Lanne,M.,Linski,M.,2004.Trendsandbreaksinper-capitacarbondioxide
emissions,1870-2028.EnergyJ.25(4),41–65.
Lee,C.C.,Chang,C.P.,2009.Stochasticconvergenceofpercapitacarbondioxide
emissionsandmultiplestructuralbreaksinOECDcountries.Econ.Model.26, 1375–1381.
Lee,J.,List,J.A.,2004.Examiningtrendsofcriteriaairpollutants:aretheeffectsof
governmentalinterventiontransitory?Environ.Resour.Econ.29(1),21–37.
List,J.A.,1999.HaveairpollutantemissionsconvergedamongU.S.Regions?
Evidencefromunitroottest.South.Econ.J.66(1),144–155.
McKitrick,R.,Strazicich,M.C.,2005.Stationarityofglobalpercapitacarbondioxide
emissions:implicationsforglobalwarmingscenarios.In:WorkingPapers 0503.UniversityofGuelph,DepartmentofEconomicsandFinance,Ontario, Canada.
Morales-Lage,R.,Bengochea-Morancho,A.,Camarero,M.,Martínez-Zarzoso,I.,
2017.StochasticandclubconvergenceofsectoralCO2emissionsinthe
EuropeanUnion.In:UniversitatJaumeIWorkingPapers,No:2017/1.
Morleo,G.,Gilli,M.,2016.EmissionidiCO2neiPaesieuropei:un’analisiempirica
sullaconvergenzatrasettorimanifatturieri.CO2EmissionsinEurope:An
EmpiricalAnalysisontheConvergenceamongManufacturingSectors]Rivista dieconomia,culturaericercasociale5,77–103.
Moutinho,V.,2015.Isthereconvergenceandcausalitybetweenthedriversof
energy-relatedcarbondioxideemissionsamongthePortuguesetourism industry?Int.J.EnergyEcon.Policy5(3),828–840.
Moutinho,V.,Robaina-Alves,M.,Mota,J.,2014.Carbondioxideemissionsintensity
ofPortugueseindustryandenergysectors:aconvergenceanalysisand econometricapproach.Renew.Sustain.EnergyRev.40,438–449.
Nguyen-Van,P.,2005.DistributiondynamicsofCO2emissions.Environ.Resour.
Econ.32(4),495–508.
Nourry,M.,2009.Re-examiningtheempiricalevidenceforstochasticconvergence
oftwoairpollutantswithapair-wiseapproach.Environ.Resour.Econ.44(4), 555–570.
Oliveira,G.,Bourscheidt,D.M.,2017.Multi-sectorialconvergenceingreenhouse
gasemissions.J.Environ.Manage.196,402–410.
Panopoulou,E.,Pantelidis,T.,2009.Clubconvergenceincarbondioxideemissions.
Environ.Resour.Econ.44(1),47–70.
Pesaran,2007.Asimplepanelunitroottestinthepresenceofcross-section
dependence.J.Appl.Econ.22(2),265–312.
Pettersson,F.,Maddison,D.,Acar,S.,Söderholm,P.,2014.Convergenceofcarbon
dioxideemissions:areviewoftheliterature.Int.Rev.Environ.Resour.Econ.7 (2),141–178.
Romero-Ávila,D.,2008.Convergenceincarbondioxideemissionsamong
industrialisedcountriesrevisited.EnergyEcon.30(5),2265–2282.
Repetto,R.,1990.Environmentalproductivityandwhyitissoimportant.
Challenge33,35–38.
Solow,R.,1956.Acontributiontothetheoryofeconomicgrowth.Q.J.Econ.70(1),
65–94.
Strazicich,M.C.,List,J.A.,2003.AreCO2emissionlevelsconvergingamong
industrialcountries?Environ.Resour.Econ.24(3),263–271.
Timmer,M.P.,Dietzenbacher,E.,Los,B.,Stehrer,R.,deVries,G.J.,2015.An
illustrateduserguidetotheworldinput–outputdatabase:thecaseofglobal automotiveproduction.Rev.Int.Econ.23,575–605.
TurkishStatisticalInstitute(TurkStat),Availableat2015.GreenhouseGas
EmissionsInventory,2013.http://www.turkstat.gov.tr/PreHaberBultenleri.
do?id=18744.
Wang,J.,Zhang,K.,2014.Convergenceofcarbondioxideemissionsindifferent
sectorsinChina.Energy65,605–611.
Westerlund,J.,Basher,S.A.,2008.Testingforconvergenceincarbondioxide
emissionsusingacenturyofpaneldata.Environ.Resour.Econ.40(1),109–120.
WorldDevelopmentIndicators,2018.Dataaccessedon2ndFebruary,2018.
Retrievedfromhttp://data.worldbank.org/indicator.
WorldDevelopmentIndicators,2016.Dataaccessedon28thFebruary,2016.
Retrievedfromhttp://data.worldbank.org/indicator.
WorldInput-OutputDatabase(WIOD),2015.WorldInput-OutputTables.Available
athttp://www.wiod.org/newsite/home.htm.
Yavuz,N.C.,Yilanci,V.,2013.Convergenceinpercapitacarbondioxideemissions
amongG7countries:aTARpanelunitrootapproach.Environ.Resour.Econ.54 (2),283–291.
Yeldan,A.E.,Tas¸c¸ı,K.,Voyvoda,E.,Özsan,E.,2013.Escapefromthemiddleincome