NBERWORKINGPAPERSERIESTHEMACROECONOMICIMPACTOFEUROPE’SCARBONTAXESGilbertE.MetcalfJamesH.StockWorkingPaper27488http://www.nber.org/papers/w27488NATIONALBUREAUOFECONOMICRESEARCH1050MassachusettsAvenueCambridge,MA02138July2020WehavereceivedvaluablecommentsfromMeredithFowlieandGeoffroyDolphin.WewishtothankCelineRamsteinandOzgurBozçagaforhelpwiththeWorldBankcarbontaxdata,andXiaoxinZhangforexcellentresearchassistance.TheviewsexpressedhereinarethoseoftheauthorsanddonotnecessarilyreflecttheviewsoftheNationalBureauofEconomicResearch.NBERworkingpapersarecirculatedfordiscussionandcommentpurposes.Theyhavenotbeenpeer-reviewedorbeensubjecttothereviewbytheNBERBoardofDirectorsthataccompaniesofficialNBERpublications.©2020byGilbertE.MetcalfandJamesH.Stock.Allrightsreserved.Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicitpermissionprovidedthatfullcredit,including©notice,isgiventothesource.TheMacroeconomicImpactofEurope’sCarbonTaxesGilbertE.MetcalfandJamesH.StockNBERWorkingPaperNo.27488July2020JELNo.E62,H23,Q43,Q54ABSTRACTPolicymakersoftenexpressconcernabouttheimpactofcarbontaxesonemploymentandGDP.FocusingonEuropeancountriesthathaveimplementedcarbontaxesoverthepast30years,weestimatethemacroeconomicimpactsofthesetaxesonGDPandemploymentgrowthratesforvariousspecificationsandsamples.OurpointestimatessuggestazerotomodestpositiveimpactonGDPandtotalemploymentgrowthrates.Moreimportantly,wefindnorobustevidenceofanegativeeffectofthetaxonemploymentorGDPgrowth.Weexamineevidenceonwhetherthepositiveeffectsmightstemfromcountriesthatusedthecarbontaxrevenuestoreduceothertaxes;whiletheevidenceisconsistentwiththisview,itisinconclusive.Wealsoconsidertheimpactofthetaxesonemissionreductionsandfindacumulativereductionontheorderof4to6percentfora$40/tonCO2taxcovering30%ofemissions.Wearguethatreductionswouldlikelybegreaterforabroad-basedU.S.carbontaxsinceEuropeancarbontaxesdonotincludeinthetaxbasethosesectorswiththelowestmarginalcostsofcarbonpollutionabatement.GilbertE.MetcalfDepartmentofEconomicsTuftsUniversityMedford,MA02155andNBERgilbert.metcalf@tufts.eduJamesH.StockDepartmentofEconomicsHarvardUniversityLittauerCenterM26Cambridge,MA02138andNBERJames_Stock@harvard.edu1Economistswidelyagreethatputtingapriceoncarbonemissionsisthemostcost‐effectivewaytoreducegreenhousegasemissions.Thetwomoststraightforwardwaystoapplyapriceareacarbontaxandacapandtradesystem.Acarbontaxcanbeleviedonfossilfuelsandothersourcesofgreenhousegasemissionsbasedontheiremissions;acapandtradesystemlimitsemissionstosomeoverallamount(thecap)andallowspolluterstotradetherightstothosescarceemissionrights.InthecurrentCongresstherearenumerousbillstoestablishnationalcarbontaxsystemsandafewcapandtradebills.Thefiledbillsreflectagrowingconsensusthatactionisneededatthenationalleveltocurbourcarbonpollutionandthatacarbontaxisthemoststraightforwardwaytodothat.Thebillsalsoreflectabroadconsensusamongeconomists,astypifiedbythemorethan3,500economistswhosignedtheClimateLeadershipCouncil’sstatementincallingforacarbontaxas“themostcost‐effectivelevertoreducecarbonemissionsatthescaleandspeedthatisnecessary.”1Amajorstumblingblocktopricingcarbonpollutionisconcernabouttheeconomicimpactofthepolicy.TheTrumpAdministration’sretreatfromaclimatepolicyisemblematic.IninitiatingaprocesstowithdrawtheUnitedStatesfromtheglobalParisAgreement,forexample,thePresidentclaimedthatthecosttotheeconomywouldbe“closeto$3trillioninlostGDPand6.5millionindustrialjobs…”(Trump,2017).Howshouldweassesstheeconomiccostsofacarbontax?Untilrecently,mostanalyseswerebasedonmodelingfromlargescalecomputablegeneralequilibriummodels.Butwenowhaveenoughexperiencewithcarbontaxsystemsaroundtheworldtocarryoutstatistical1ThestatementwaspublishedinTheWallStreetJournalonJan.17,2019andisavailableathttps://clcouncil.org/economists‐statement/.Bothoftheauthorsofthispaperaresignatoriesofthatstatement.2analysesofexistingsystems.Thefirstcarbontaxwasimplementedin1990sothereisnowuptothreedecadesofdatatodrawon.Inthispaperwecarryoutananalysisofthe31countriesinEuropethatarepartoftheEUwideemissionstradingsystem(EU‐ETS).Whileallofthesecountriespriceaportionoftheiremissionsthroughthiscapandtradesystem,fifteenofthesecountriesalsoimposeacarbontax,mostlyonemissionsnotcoveredbytheEU‐ETS.BylimitingouranalysistocountriesthatarepartoftheEU‐ETS,wecanidentifytheincrementalimpactofcarbontaxesonemissions,output,andemploymentbyleveragingthevariationincarbontaxsystemswithinthisgroupofcountries.ThispaperbuildsonapreviousanalysisinMetcalfandStock(2020).Webuildonthatanalysisinanumberofwaysincluding,importantly,measuringtheimpactofthecarbontaxesonemissionsaswellaseconomicgrowth.Wefindthefollowing.Forawiderangeofspecifications,wefindnoevidenceofadverseeffectsonGDPgrowthortotalemployment.Wealsotestandgenerallycannotrejectthehypothesisthatthecarbontaxhasnolongruneffectongrowthrates.Thisfindingisconsistentwithmacroeconomictheorythatsuggestsgrowthratesaredrivenbyfundamentals,suchastechnologicalprogress,whichareunaffectedbychangesinrelativeprices.Itisalsoconsistentwithmostgeneralequilibriummodelingofclimatepolicy.Finally,wefindcumulativeemissionreductionsontheorderof4to6percentforataxof$40pertonofCO2covering30%ofemissions.Wearguethatthisislikelytobealowerboundonreductionsforabroad‐basedcarbontaxintheU.S.sinceEuropeancarbontaxesdonotincludeinthetaxbasethosesectorswiththelowestmarginalcostsofcarbonpollutionabatement.Europeancarbon3taxesgenerallyexcludetheelectricitysectorandcarbonintensiveindustriessincethoseemissionsarecoveredundertheEUEmissionTradingSystem.Thenextsectionprovidesbackgroundandaliteraturereviewthatplacesourpaperincontext.SectionIIIsurveysEuropeancarbontaxes.SectionIVdetailsourdataandtheeconometricapproachwetakeassesstheimpactofEuropeancarbontaxes.SectionVpresentsresultsfromtheanalysis.Thenextsectionpresentssomerobustnessresults.WeprovidesomeconcludingremarksinsectionVII.II.PreviousLiteratureMostanalysesoftheeconomicimpactofcarbontaxesrelyonlarge‐scalecomputablegeneralequilibriummodels.OnerepresentativemodelistheE3modeldescribedinGoulderandHafstead(2017).Theyestimatethata$40pertoncarbontaxfortheUnitedStatesstartingin2020andrisingat5percentrealannuallywouldreduceGDPbyjustoveronepercentin2035relativetoano‐taxcounterfactual.Whiledifferentmodelsgivedifferentresults,mostfindverymodestreductions(ifatall)inGDPfromimplementingacarbontax.2Goulderetal.(2019)alsoconsideraU.S.carbontaxstartingat$40pertonandrisingat2percentannually.TheyfindtheGDPcostsoverthe2016–2050perioddiscountedat3percentequaltolessthanone‐thirdofonepercentofGDP.2TrumpcitedaNERA(2017)studycommissionedbyanindustrygrouptoanalyzehowmeetingan80percentreductionby2050wouldaffectvariousindustrysectors.Amongotherissues,theheadlinenumbercitedbyTrump(7percentreductioninGDP)isfromaNERAscenarioinwhichsectorspecificregulationsareimposedwithverydifferentmarginalabatementcostsacrosssectors.Ifmarginalabatementcostsareallowedtoequalizeacrosssectorsinthatstudy,thecostsarereducedbyovertwo‐thirds.4Turningtotheempiricalliterature,Metcalf(2019)findsnoadverseGDPimpactoftheBritishColumbiacarbontaxbasedonadifference‐in‐differenceanalysisofapanelofCanadianprovincesoverthetimeperiod1990–2016.UsingapanelofEuropeancountriesoverthetimeperiod1985–2017,hefinds,ifanything,amodestpositiveimpactonGDP.ThatimposingacarbontaxmighthavepositiveimpactsonGDPisnotimplausibleonceoneconsidersthegovernments’useofcarbontaxrevenue.Intheearly1990s,forexample,carbontaxeswereimposedinanumberofScandinaviancountriesasarevenuesourcetofinancereductionsinmarginaltaxratesfortheirincometaxes(seeBrannlundandGren,1999,forbackgroundonthesereforms).Variationintheuseofrevenuesfromnewlyenactedcarbontaxescoulddifferentiallyimpacteconomicgrowthandissomethingweexploreinthispaper.Bernardetal.(2018)useavectorautoregression(VAR)toestimatetheimpactoftheBCcarbontaxonprovincialGDP,controllingforthepre‐taxpriceofgasoline(ordiesel)andUSeconomicvariables;theyfindnoimpactofthetaxonGDP.Inearlierworkwithamorelimitedversionofthedatasetusedinthispaper,we(MetcalfandStock(2020))uselocalprojectionstoestimatetheimpactofcarbontaxesinEuropeancountriesonGDPandfoundnoadverseimpactsofthetaxoneconomicgrowthoremployment.TheseresultsareconsistentwithYamazaki(2017)analysisoftheemploymenteffectsoftheBritishColumbiacarbontax.Yamazakifoundmodestpositiveimpactsonemploymentintheprovince.Whileaggregateimpactsweresmall,hefoundsignificantjobshiftingfromcarbonintensivetonon‐carbonintensivesectors.Focusingonemissions,LinandLi(2011)estimatedifference‐in‐differenceregressionscomparingindividualcountrieswithcarbontaxes(Finland,theNetherlands,Norway,Denmark,5andSweden)withasetofcontrolcountriesandfindmixedresults.In4ofthe5countries,thegrowthrateofemissionsfallsbybetween0.5and1.7.OnlytheestimateforFinlandisstatisticallysignificantatthe10percentlevel,withthecoefficientsuggestingadropinthegrowthrateofemissionsof1.7percent.Martinetal.(2014)assesstheUnitedKingdom’sClimateChangeLevy’s(CCL)impactonenergyandemissionsindicatorsforvariousmanufacturingsectors.AsdiscussedinMetcalf(2019),theCCLisnotatruecarbontaxgivenitsdifferentialtaxationoffossilfuels.WhileCO2emissionsfallby8.4percent,butimpreciselyestimated,theirresultsarealsoconsistentwiththeCCLleadingtofuelsubstitutionawayfromelectricityandtowardcoal.Thisfollowsfromthelowertaxrateoncoalthannaturalgas.ArecentpaperbyAndersson(2019)focusesontheimpactofSweden’scarbontaxontransportationemissions.HefocusesontransportationasthisisthesectormostimpactedbytheSwedishcarbontax.Hefindsanemissionsreductionontheorderof11percent.WhilethismightappearmodestgiventhefactthatSwedenhasthehighestcarbontaxintheworld,mostanalystsarguethatthetransportationsectoristhemostdifficultsectortodecarbonizegiventheefficiencyoftheinternalcombustionengine.TurningtoBritishColumbia,RiversandSchaufele(2015)findthattheprovince’scarbontax,whichcoversgasoline,diesel,andnaturalgas,significantlyreducesgasolineconsumption.Theyestimatethatthecarbontaxhasastrongerimpactongasolinedemand–byafactoroffour–thanacomparableincreaseinthepriceofgasoline,asurprisingfindingthattheauthorsattributetothehighsalienceofthecarbontax.Metcalf(2019)estimateddifference‐in‐differenceregressionsusingCanadianprovincedataandfindthattheBCtaxreducedemissions6ontheorderof5to8percentsinceit’simpositionin2008.Prettis(2019)estimatesa5percentreductionintransportationemissionsfromtheBCcarbontax,withpotentiallylargerlong‐runemissions,butdoesnotdetectaneconomy‐wideemissionsreductionattributabletothetax.Asnotedattheoutset,thispaperbuildsonMetcalfandStock(2020).InadditiontoconsideringadditionaleconometricmodelspecificationsforemploymentandGDP,wealsoassessthecarbontaxesimpactsoncountryemissions.Wealsotestwhethermacroeconomicoutcomesareaffectedbytheuseofcarbontaxrevenue.Specifically,weconsiderwhethergreentaxreforms–reformswherecarbontaxrevenuesareusedtolowerexistingdistortionarytaxrates–hasadifferentimpactonmacrooutcomesthanwhentherevenueissimplyaddedtogeneralrevenue.3III.CarbonTaxesinEuropeCarbontaxeswerefirstenactedinEuropewithFinlandleadingthewayin1990.FollowinganearlywaveofcarbontaxenactmentsprimarilyintheNordiccountries,morecountriesenactedcarbontaxesandcurrentlysixteenEuropeancountrieshavecarbontaxesinplace.Wefocusontheso‐calledEU+countriesthatarealsopartoftheEU‐ETSandsoexcludeUkrainefromouranalysis.WefocusonEU+countriestocontrolconsistentlyfortheimpactoftheETSongrowth.TheETSwentintoeffectwithapilotphase(PhaseI)in2005.InPhaseI,powerstationsandcertainenergyintensivesectorsweresubjecttothecap.4PhaseII(2008–3Wecan’truleoutthepossibilitythataddingcarbontaxrevenuestogeneralrevenueallowsacountrytoavoidafuturetaxincreaseasopposedtoanincreaseinspending.Inthatcase,wewouldnotexpectadifferentoutcomethanwhentherevenueisexplicitlyearmarkedforreductionsindistortionarytaxrates.4Thesectorsarepowerstationsandothercombustionplantsofatleast20MW,oilrefineries,cokeovens,ironandsteelplants,cementclinker,glass,lime,bricks,ceramics,pulp,andpaperandboard.Aluminum,petrochemicals,ammonia,nitric,adipic,andglyoxylicacidproduction.andCO2capture,transport,andstoragewereaddedinPhaseIII.72012)addeddomesticaviation(in2012),andPhaseIII(2013–2020)addedvariousadditionalsectors.5Countrieswithcarbontaxesareasfollows(listedinchronologicalorderofenactment):6Finland(1990):ThefirstwaveofcarbontaxesinEuropebeganwithFinland’senactmentofataxin1990.AswasalsothecaseinsubsequentNordiccarbontaximplementations,Finland’scarbontaxwasenactedduringatimeofincometaxreformtolowerhighmarginalincometaxrates(CarlandFedor,2016).Between1990and1994,Finlandtaxedallfossilfuelsincludinggasoline,diesel,fueloil,coal,naturalgaswithsomefuels(e.g.naturalgas)taxedatalowerratepertoncarbondioxide.7Between1994and1997,Finlandswitchedtoacombinedcarbonandenergytaxbeforeswitchingbacktoacarbontaxin1997.Thetaxbasesince1997waspredominantlymotorandheatingfuels.Taxrevenuesarenotspecificallyearmarkedforparticularusesbutcoincideswithreductionsinincometaxratesbothinitiallyandovertime(OECD,2019).Itraisedalittleunder$1.5billionin2018,accordingtotheWorldBankCarbonPricingDashboard.Poland(1990):AccordingtotheWorldBank’sStatesandTrendsofCarbonPricing,Polandhasthelowestcarbontaxrateamongthosecountrieswithacarbontax.Italsocoversasmallshare5Twenty‐fiveofthethirty‐onecountriesinoursamplehavebeensubjecttotheETSfromitsinception.RomaniaandBulgariajoinedin2007whileNorway,Iceland,andLiechtensteinjoinedtheETSstartingwithPhaseIIin2008.CroatiajoinedtheETSasofPhaseIIIin2013.SeeEuropeanCommission(2015)forahistoryandmembershipoftheETS.6ThissectiondrawsonmaterialfromSumneretal.(2011),CarlandFedor(2016),MartenandDender(2019),andWorldBankGroup(2019b).7https://sustainabledevelopment.un.org/index.php?page=view&type=99&nr=183&menu=1449,accessedMarch29,2020.8ofemissions(lessthanfivepercent).Revenuesareontheorderof$1millionannuallyandareearmarkedforenvironmentalspending(OECD,2019).Norway(1991):NorwaywasalsooneoftheearlyNordiccarbontaxadopters,withthetaxadoptedaspartofbroaderincometaxreform(seealsoBrannlundandGren(1999)whodocumentthegreentaxreformsinDenmark,Norway,andSweden).LikeotherNordiccountries,thetaxesgointogeneralrevenueandarenotearmarkedthoughitisunderstoodthatthey,inpart,makeupforlostincometaxrevenuefromloweringincometaxrates.Thetaxcoversoil,gasoline,andnaturalgasinthetransportationandindustrialsectors(primarily)withmodestcontributionsfromagricultureaswellastheresidentialandcommercialsectors(OECD,2018).8Sweden(1991):Sweden’scarbontaxappliesprimarilytotransportandheatingfuels(HammarandAkerfeldt,2011).Initiallyenactedwithlowerratesfortradeintensivesectorsoftheeconomy,thegapbetweenthestandardandreducedrateshasgraduallybeenreduceduntilitwaseliminatedin2018.AsAndersson(2019)notes,roughly90percentofthetaxrevenuecomesfromthetransportsector.AswithotherNordiccountries,Sweden’scarbontaxwasenactedaspartofabroadertaxreformthatsawreductionsinincometaxratesaswellasanexpansionofthevalueaddedtaxtocovergasolineanddiesel.Denmark(1992):Denmarkisanothercountrythatimplementedacarbontaxaspartofabroadertaxreformmovementintheearly1990s.LikeotherNordiccountries,thetaxesgointogeneralrevenueandarenotearmarkedthoughitisunderstoodthatthey,inpart,makeupfor8EmissionsfromoffshoreoilandgasproductionaresubjectbothtotheEU‐ETSandthecarbontax(Svenningsenetal.,2019,p.80).9lostincometaxrevenuefromloweringincometaxrates.Thetaxcoversoil,gasoline,andnaturalgasinthetransportationandindustrialsectors(primarily)withmodestcontributionsfromagricultureaswellastheresidentialandcommercialsectors(OECD,2018).Slovenia(1996):SloveniawasthefirstformerSovietblockcountrytoimplementacarbontax.Revenuesareearmarked,inpart,forgreenspending(atleastin2005).Thetaxfallsprimarilyonemissionsfromthebuildingsandtransportsector.Estonia(2000):Estoniahasamodestcarbontaxthatappliesprimarilytotransportfuels(OECD,2018).Revenuesarenotearmarkedbutgointogeneralrevenue.Latvia(2004):Latvia’scarbontaxappliestoemissionsfromindustryandthepowersectornotsubjecttotheEUETS(WorldBankCarbonPricingDashboard).Carbontaxrevenueisappliedtogeneralrevenue.Switzerland(2008):WhileSwitzerlandisnotpartoftheEU‐ETS,ithasitsowncapandtradesystem(since2008)andhasformallylinkeditssystemwiththeEU‐ETSasof2019.ItscarbontaxappliestoenergyrelatedCO2emissionsnotcoveredbythecountry’sETS.Firmsintradeexposedsectorswithlargecarbontaxburdenscanoptoutofthesystemiftheycommittoasetreductioninemissionsbygivendates.Switzerlandearmarksone‐thirdofthecarbontaxrevenueforsubsidizeenergyreductioninthebuildingsector,eitherthroughenergyefficiencyinvestmentsordistributedgeneration–geothermalprimarily–investments(CarlandFedor,2016,andOECD,2019).99LiechtensteinisrequiredasaresultofabilateraltreatywithSwitzerlandtoenactthesameenvironmentallawsasinSwitzerland.Asaresult,italsoenactedacarbontaxin2008.Allrevenuesareaddedtogeneralrevenue.SimilarexemptionsexistasinSwitzerland.WehavenotincludedLiechtensteinintheempiricalanalysis.10Ireland(2010):Irelandenactedacarbontaxinthewakeofthe2008fiscalcrisisthatconfrontedthecountrywiththeriskofcollapseoftheIrishbankingsystemandcollapseofpropertyvalues(Converyetal.,2013).Revenuesenterthegeneralbudgetthoughthereissomeminorearmarkingforenvironmentalprojects.10ThetaxappliestoallfossilfuelemissionsandcanoverlapwithcoverageoftheETS.Firms,however,areeligibleforarebateofthecarbontaxforemissionssubjecttotheETS.Iceland(2010):LikeIreland,Icelandenactedacarbontaxfollowingthefinancialcrisisof2008.Initially,itpeggedthetaxratetotheEU‐ETSallowancepricebutsoonraisedtheratetoraiseadditionalrevenueforthenationalbudget(Svenningsen,etal.,2019).Thecarbontaxappliestofossilfuelconsumptioninthecountry;giventheextensiveuseofgeothermalandhydropowerforpowergeneration(nearlyall)andhouseholdheating(over90percent),thecarbontaxisprimarilyataxontransportfuels.UnitedKingdom(2013):TheUKCarbonPriceFloor(CPF)appliestopowergeneratorssubjecttotheEU‐ETSandisdesignedtosetafloorpriceonpower‐relatedCO2emissions.Revenuesarepartofgeneralrevenueandarenotearmarked.Confusingly,theUKalsoappliesaClimateChangeLevywhichisnotatruecarbontaxasitappliesdifferentratespertonofcarbondioxidetodifferentfuelswith,forexample,coaltaxedathalftherateofnaturalgas(Martin,etal.,2014).10IntheDecember2009FinancialStatementoftheIrishMinisterforFinance,theMinisterstatedthattherevenuewouldbeused“toboostenergyefficiency,tosupportruraltransportandtoalleviatefuelpoverty.TheCarbonTaxwillalsoallowustomaintainorreducepayrolltaxes.”SeetheStatementathttp://budget.gov.ie/Budgets/2010/FinancialStatement.aspx#item9,accessedonMarch29,2020.Intheabsenceofexplicitearmarking,weviewthisstatementasconsistentwithusingtherevenuetocontributetogeneralrevenueandavoidapayrolltaxincrease.11Spain(2014):Spain’scarbontaxisataxonfluorinatedgases.Thetaxisappliedwhenequipmentisrechargedandcanberebatedtotheextentthegasesarerecoveredandrecycled.Revenueentersthegeneralrevenuewithnoearmarking.France(2014):France’scarbontaxcoversnearlyallfossilfuelsnotsubjecttotheEU‐ETS.Certainindustrialprocessemissionsareexemptfromthetax.Initially,revenuefromthetaxwasearmarkedtofinancegreenspending(CarlandFedor,206).Overtimetheshareofearmarkedrevenuehasdeclinedwiththeshareallocatedtogeneralrevenuerising.Portugal(2015):Portugal’scarbontaxgenerallyappliestoemissionsfromsectorsnotcoveredbytheEU‐ETS.AuniquefeatureofPortugal’staxisthattherateisindexedtotheaverageallowancepriceoftheEU‐ETSfromthepreviousyear(WorldBankCarbonPricingDashboardandPereiraetal.,2016).Thecarbontaxwasenactedaspartofabroadersetoftaxreformsandrevenuefromthetaxwasusedtolowerexistingincometaxes.Table1summarizesinformationaboutcarbontaxesacrosstheseEU+countries.Figure1showsthetimetrendofcarbontaxratesintheEU+countriessincetheirenactment.Thereisconsiderablevariationinratesaswellastimeofenactmentforthetaxes.(Notethatthescaleofthetopgraphdiffersfromthatofthenexttwo.)Figure2showsGDPpercapitagrowthratesbeforeandaftereachcountry’senactmentofthecarbontax.Thedotsindicatemeanvaluesandbars90percentconfidenceintervals.Thereisnoclearpatterninchangesingrowthratesfollowingenactmentofthecarbontax.11Wethereforeturnnexttoaneconometricanalysis.11Theeventstudygraphsarebasedonregressionswithoutcontrolsandsimplyillustratetheimportanceofundertakingamoresystematicanalysis.GraphsforemploymentandemissionsareincludedintheAppendix.12IV.DataandEconometricApproachOurdataonrealGDPandcarbontaxratescomefromtheWorldBankGroup(2019a).12EmploymentdataarefromtheEUEurostatdatabase.DataontheshareofgreenhousegasemissionscoveredbythetaxcomefromWorldBankGroup(2019a),andenergypriceandenergyexcisetaxdataarefromtheInternationalEnergyAgency(2019).DataoncountrycarbondioxideemissionsfromfuelconsumptionarefromEurostatandcovertheyears1990through2018.Wefocusoncarbondioxideemissionsfromfuelcombustioninroadtransport,thecommercialandinstitutionalsector,andthehouseholdsector.Thesearesectorsmosttypicallyincludedincountrylevelcarbontaxes.IdentifyingthedynamiccausaleffectofacarbontaxonGDPgrowthiscomplicatedbythepossibilityofsimultaneity:pooreconomicoutcomescouldleadthetaxauthoritiestoreducetherateortopostponeaplannedincrease.13AsdiscussedinMetcalfandStock(2020),itisusefultothinkofchangestoacarbontaxashavingtwocomponents,onerespondingtohistoricaleconomicgrowth,theotherbeingunpredictedbypastgrowth.Changesinthelattercategorycouldincludetaxchangesbasedonhistoricallylegislatedschedules,changesinambitionbasedontheenvironmentalpreferencesofthepartyinpower,orresponsestointernationalclimatepolicypressure.Ouridentifyingassumptionisthatthislattercategoryof12RealcarbontaxratesarenominaltaxratesdividedbytheGDPdeflator(homecountrycurrency),convertedtoUSdollarsat2018exchangerates.WeusednationalstatisticalagencydataforGDPandprices,insteadofWorldBankdata,forIrelandandNorway.ForIreland,weusedadjustedGrossNationalIncome,whicheliminatesdistortionsfromintellectualpropertyinflowsduetoIreland’sstatusasataxhaven(Worstall,2016),andtheCPI.Norwaymaintainsdualaccounts,onshoreandoffshore,thelatterincludingoilrevenues;weuseonshoreGDPanditsdeflatortoavoidspuriouslyconfoundingcarbontaxeffectswithNorway’soffshoreoilproduction.13BritishColumbia,forexample,hasannouncedadelayinthe2020scheduledincreaseinitscarbontaxduetotheCOVID‐10pandemicshocktotheeconomy.Seeinformationathttps://www2.gov.bc.ca/gov/content/environment/climate‐change/planning‐and‐action/carbon‐tax,accessedonJune5,2020.13changes–thosenotpredictedbyhistoricalown‐countryGDPgrowthandcurrentandpastinternationaleconomicshocks–areexogenous.ThisassumptionallowsustoestimatethedynamiceffectonGDPgrowthoftheunexpectedcomponentofacarbontaxusingtheJordà(2005)localprojection(LP)method,adaptedtopaneldata.Specifically,weuseOLStoestimateasequenceofpaneldataregressions,(1)100Δ𝑙𝑛𝐺𝐷𝑃𝛼𝛩𝜏𝛽𝐿𝜏𝛿𝐿𝛥𝑙𝑛𝐺𝐷𝑃𝛾𝑢.where𝜏istherealcarbontaxrateforcountryiatdatetand𝛩istheeffectofanunexpectedchangeinthecarbontaxrateattimetonannualGDPgrowthhperiodshence.Inallregressions,thetaxrateisinteractedwithits2019shareofitsemissioncoverage.Thisspecificationassumesthatanydamage(orbenefit)ofthetaxtoaneconomywouldbe,inthefirstinstance,proportionaltothecoveredshareoftheeconomy.Allregressionsincludebothcountryandyearfixedeffects.Includingtheformeraddressesthepossibilitythatcountrieswithhighermeangrowthratesmightbetheonesmorelikelytoadoptandincreaseacarbontax,inwhichcasethetaxcouldspuriouslyappearbeneficial.Inprinciple,underourexogeneityassumptionitshouldnotbenecessarytoincludeyeareffects,butwedosofortworeasons.First,becausethecountriesareallEuropean,theysharecommonpoliticalpressures,whichcouldinducecommonchangesincarbonprices,andhavecommoneconomicinfluences.Thesecommoninfluences,evenifexogenous,couldappearasconfounders,soweidentifytheeffectofthetaxincreasefromcountry‐levelsurprisesincarbonpricesaftercontrollingforcommonmovements(yeareffects).Second,evenifyeareffectsarenotneededforidentification,becauseofcommonmacroeconomic14movements(suchastheglobalrecessionof2009),includingyeareffectscouldreducestandarderrors.Standarderrorsareheteroskedasticity‐robust(Plagborg‐MollerandWolf(2019)).WealsoestimatepanelstructuralVAR(SVAR)regressionswiththetaxrateandGDPgrowthasdependentvariables,fourannuallagsofeachasregressors,andcountryandyearfixedeffects.ThisisapanelversionofthestandardtimeseriesstructuralVAR.TheidentificationconditionsarethesameasintheLPregression.Inpopulationtheestimandisthesame.AlthoughtheSVARandLPmethodshavethesameidentifyingconditionandthesameestimand,infinitesamplestheycandifferandtheywillhavedifferentstandarderrors.Thus,usingtheSVARestimatorprovidesarobustnesscheckontheLPestimator.SVARstandarderrorsarecomputedbyparametricbootstrap.14InMetcalfandStock(2020),wealsoestimateddistributedlagregressions.Theseregressionsrequirethestrongeridentificationconditionthatthecarbontaxisstrictlyexogenous,thatis,thereisnofeedbackfromGDPgrowthtothetaxrate.WetestthisconditionbycomputingatestoffeedbackfromGDPgrowthtothetaxrate,thatis,apanelGrangerCausalitytestofthecoefficientsonGDPgrowthinaregressionofthecarbontaxrateonitslagsandlaggedGDPgrowth.ToensurestationaryregressorssothatstandardFcriticalvaluescanbeused,wecomputethistestusingthegrowthrateofthevariables.Asdiscussedinthenextsection,thetesttendstorejectlackoffeedbacktothetaxrate(atleastatthe10percentlevel),indicatingthatthedistributedlagidentifyingconditionsarenotsupportedbythedata.Accordingly,wedonotpresentdistributedlagresultshere.14SeeStockandWatson(2018)andPlagborg‐MollerandWolf(2019)fordetailsonmethodologyandrelationbetweenVARsandLPs.15Weconsiderthecounterfactualofaone‐timepermanentincreaseinthecarbontaxby$40,forataxthatcovers30%ofthecountry’semissions,acoverageratethatisclosetothesamplemean.WecomputethisdynamicresponsefromtheLPandSVARimpulseresponsesusingthemethodinSims(1986),whichentailscomputingthesequenceofshocksnecessarytoyieldthespecifiedcounterfactualcarbontaxincrease.Inpractice,thecarbontaxseriesishighlypersistentsothedynamiceffectsweestimatehereareclosetothe(scaled)structuralimpulseresponsefunctionsatthehorizonsweconsider.AkeyissueinthedynamicmodelisthelongruneffectofthecarbontaxonthegrowthrateofGDP,thatis,whetheracarbontaxpermanentlychangesnotjustthelevelofGDPbutalsotheslopeoftheGDPgrowthpath.Thestandardtheoryunderlyingcomputabledynamicequilibriummodelsofacarbontaxmodelsthelong‐rungrowthrateasdeterminedbyfundamentals,andthatthosefundamentalsarenotaffectedbytherelativepricechangeinducedbycarbontax.Ifso,thetaxmightaffectGDPgrowthintheshortrunbutwouldreverttothelongrungrowthrateintime.Ineffect,thetaxwouldshiftGDPtoanewlevelafterwhichitwouldmoveinparallelwithitspathhadthecarbontaxnotbeenimposed,seeforexampletheGoulderandHafstead(2017)E3modelorNordhaus’sDICEmodel.15This“parallelpath”hypothesisimposesatestablerestrictionontheLPandSVARspecifications,specificallythatthelong‐runeffectofashocktothecarbontaxonGDPgrowthiszero.WeestimatedtheLPandSVARspecificationsbothwiththiszerolong‐rungrowtheffectrestrictionimposed(“restricted”case)andnotimposed(“unrestricted”).InboththeSVARand15ModelshavebeendevelopedthatallowthelongrungrowthrateofGDPtobeaffectedbyclimatedamages.See,forexample,Moyeretal.(2014).Butthisismoretheexceptionthantherule.16LPspecifications,intheunrestrictedcaseτentersinlevels,intherestrictedcase,τentersinfirstdifferences.FortheSVAR,therestrictionisthatthelong‐runstructuralimpulseresponsefromthetaxtoGDPgrowthinthelevelsspecificationiszero,whichwetestdirectly.BecausetheLPapproachcomputesimpulseresponsesouttoamaximumfinitehorizonhinequation(1),itdoesnotestimatethelong‐termeffectatarbitrarilydistanthorizons.Consequently,fortheLPtestofthelong‐runrestriction,weapproximatethelong‐termeffectbytheeffectatthe8‐yearhorizon.ThediscussioninthissectionhasfocusedontheeffectofacarbontaxonGDPgrowth.Weusethesamemethodstoanalyzetheeffectonthegrowthrateofemploymentandemissions.V.ResultsWebeginwithresultsforGDP,thenturntoemploymentandemissions.A.GDPFigure3showstheIRFfortheLPmodelforrealGDP,estimatedusingall31countriesoverthefull1985‐2018sample,wherethecarbontaxrateisinteractedwiththeshareofemissionscoveredbythetax.Figure3ashowsresultsfromtheunrestrictedmodel,thatis,themodelthatallowsforanonzerolong‐termeffectofthetaxonGDPgrowth.Thepredictedeffectispositiveineachyearthroughyear6exceptforyear4.Innoyear,however,istheeffectsignificantatthe5%level(inmostyearsitiswithinonestandarderrorofzero).Theresultsfortherestrictedmodel(figure3b),inwhichazerolong‐termeffectofthetaxrateonGDPgrowthisimposed,aresimilartothosefortheunrestrictedmodel.Again,thepointestimateisgenerallynomorethanonestandarddeviationawayfromzero.17Figure4showstheIRFsfortheSVARmodel.TheunrestrictedIRF(figure4a)isalwayspositiveandnearconstantatabout0.4percentagepoints.TherestrictedIRF(figure4b)ismuchclosertozeroand,infact,theperiod6estimateisessentiallyzero(thoughnotimposedexplicitlybythemodel).Ingeneral,theSVARresultsareverysimilartotheLPresults.Inallcases,thestandarderrorbandsintheSVARrestrictedmodelsapproachzeroinlateryears.Fortherestrictedmodels,thisisaconsequenceofthejointstationarityofGDPgrowthandthechangeinthecarbontax.Fortheunrestrictedmodels,thisconvergenceoftheIRFtozeroneednotbethecase,buttherestrictionissupportedbythedatasotheunrestrictedandrestrictedestimatesaresimilar.TheSVARandLPmodelsareconsistentestimatorsofthesameobjectsinpopulation,butevensoitisstrikinghowsimilartheempiricalresultsareusingthetwomethods.Thisistruegenericallyforthesedata,acrossdependentvariablesandregressors.WethereforehenceforthonlyreportLPresultsinthetext,relegatingselectedSVARestimatesforkeyspecificationsintheAppendix.Table2summarizestheresultsfortestsoftherestrictionthatthelong‐runeffectonthegrowthrateofGDPofthetaxiszero.NeithertheLPnorSVARtestsrejectthishypothesis:FortheLPmodel,theteststatisticequals0.33(p‐value=0.75).FortheSVARmodel,theteststatisticis1.34(p‐value=0.18).Wefindsimilartestresultsforothercutsofthedata(discussedbelow)andreportresultsinTable2.Table3reportstestsofthehypothesisthatthecarbontaxrateisstrictlyexogenous(Grangercausalitytests).Theteststatisticequals2.21forourfullsampleandhasap‐valueof0.066,rejectingstrictexogeneityatthe10percentlevel.WerejectstrictexogeneityfortheGDPregressionsatthe10percentlevelorlowerforothercutsofthedata(seeTable3).18TheresultsillustratedinFigures3and4donotsuggestparticularlylargepositiveimpactsofacarbontaxonGDP.Butneitherdotheysupportaclaimoflargeadverseimpacts.Itispossible,however,thateffectsaccumulateovertime.Figure5showscumulativeIRSsfortheLPmodel.TheunrestrictedmodelcumulativeIRF(toppanel)showsapositiveimpactongrowthbyyear6ofroughly2percentagepointsbutstandarderrorbandsarelargewiththe95percentconfidenceintervalrangingfrom‐2to+6percentagepoints.Whenrestrictingthelong‐rungrowthratetobezero,theimpactisnownegligibleinallsixyears.Inthisandsubsequentsub‐samples,wefindnoevidencetosupporttheviewthatEuropeancarbontaxeshavehadasignificantimpactonGDP,eitherpositiveornegative.B.TotalEmploymentFigure6showsIRFsfortotalemployment.Inboththeunrestrictedandrestrictedcases,employmentinitiallyrisesandthensubsequentlyfalls.Thecumulativeimpact(Figure7)ispositiveoverasixyearperiodwithapointestimateof1.15percentagepointsintheunrestrictedLPmodeland0.35percentagepointsintherestrictedmodel.Inneithermodeldowerejectazerocumulativeimpact.Results(intheAppendix)formanufacturingemploymentaresimilartototalemploymentbuttheestimatesarelessprecise.AswithGDP,wecanrejectsignificantnegativeemploymentimpactsfromthecarbontax.C.EmissionsImpulseresponsefunctionsmeasureannualchangesinthevariablesofinterestfollowingapolicychangesuchasthe$40pertoncarbontaxmodeledthroughout.WefocusontheeffectofthecarbontaxonthelevelofCO2emissionsinroadtransportandthecommercial,institutional,andhouseholdsectors.Thislevelseffectisestimatedbythecumulativestructural19impulseresponsefunction,becauseemissionsenteringrowthrates.UnlikeGDPandemployment,thereisnoaprioriexpectationofa“parallelpath”hypothesis,thatis,thatinthelongrunthegrowthrateofemissionswouldbeunaffectedbythecarbontax.Infact,abasicpremiseofclimatepolicyisthatataxcouldhelpbendthecurveonemissionsgrowthrates(throughinnovationandgreentechnologicalprogress).Butwewouldbesurprisedifwefoundevidenceofachangeinthelongrunemissionsgrowthrategiventhelengthofoursampleandthemagnitudeofmostcountrytaxrates.Whilewecanrejectthehypothesisofzerolong‐runchangesintheemissionsgrowthrateforthefullsampleintheLPmodel,wecannotrejectitintheSVARmodelnorcanwerejectthezerolong‐runchangeinothersamplesforeithermodel(Table3).ResultsforthefullsampleareshowninFigure8.Emissionsfallby6.5percentagepointsbytheendofyear6intheunrestrictedmodel(toppanel)thoughstandarderrorbandsarewideandwecannotrejectnochangeinemissions.Intherestrictedmodel(bottompanel),emissionsfallbyasmuchas5.6percentagepointsinthefirst4yearsbutstabilizeata3.8percentreductionbyyear6.Theestimatesareimprecise,witha95%confidenceintervalincludingbothnoemissionsreductionsandlargeremissionsreductions.Wealsonotethatwhilestrictexogeneityisrejectedinsomeorallofthesamples(ata10percentlevelorbetter)forGDPandemploymentmeasures,itisgenerallynotrejectedforemissions.Thisisperhapsnotsurprisingsincewewouldexpectthatadversemacroshockswouldbereadilyobservabletodecision‐makersandmightleadtopolicychangesinclimatepolicy.Whileatransitoryincreaseinemissionsmight,throughthepoliticalprocess,spur20greaterambition,theempiricalevidencedoesnotsuggestthatthischannelisasignificantdeterminantoftaxratechanges.VI.RobustnessThefindingofanoverallslightpositiveeffectoneconomicactivityisintriguing,andraisesthequestionastowhetherthispositiveeffectcouldarisefromtheuseofthecarbontaxrevenuetoimprovetheoverallefficiencyofthetaxsystem,givingrisetoadoubledividend.Anotherpossibilityisthatcountrieswithalongexperiencewiththecarbontaxhaveadifferentresponsethancountrieswithlessexperience.Thereisinfactconsiderablevariationintaxrates,useofrevenues,orthelengthoftimethecarbontaxhasbeenineffect.WeexploreinthissectionwhetheranyofthesefactorsmatterforGDPoremploymentgrowthaswellasemissions.A.RevenueRecyclingWebeginbyaskingwhethergrowthimpactsarelargerforthosecountriesthatstatedanintentiontorecyclethecarbontaxrevenuethroughcutsinincomeorpayrolltaxrates.TheDoubleDividendHypothesissuggeststhisshouldbeefficiencyenhancingand,presumably,improvegrowthprospects(e.g.Goulder,1995).Thereislimiteddataonhowcountriesactuallyusecarbontaxrevenues.Manyearlymovingcountries(Denmark,Sweden,Norway,Finland)enactedcarbontaxesaspartofaGreenTaxReformdesignedtoreducemarginalincometaxrates.Switzerlandexplicitlyearmarkedtwo‐thirdsofcarbontaxrevenuefortaxcuts.PortugalalsoearmarkedrevenuefortaxcutsaspartofaGreenTaxReform.Wetreatthisgroupofsixcountriesasagroupthatpartiallyorfullyusedcarbontaxrevenuetolowerexistingincometaxrates.Whileourdesignationisnecessarilyimpreciseandrecognizingthattaxrevenuesare21fungible,weinvestigatewhethergrowthimpactsarelargerforthisgroupofcountriesrelativetothefullsampleofcarbontaxenactingcountries.Notetoothatweconsidercountriestoberevenuerecyclingbasedonstatedintentionsratherthanactualoutcomes.Regardlessofwhatcountriessaytheyaregoingtodo,wecannotobservethecounterfactualoutcomehadthecarbontaxesnotbeenimplementedanditispossiblethatsomeofthesecountriesusedcarbontaxrevenuefornon‐revenuerecyclingpurposes(andviceversaforthosecountriesthathavenotstatedanintentiontorecyclecarbontaxrevenuesthroughlowertaxrates).Figure9showstheresultsforGDPgrowth(toppanel)andtotalemployment(bottompanel)fromtheLPregressionsfocusingonthesesixcountriesrelativetocountrieswithnocarbontax.GDPgrowthisinitiallyabitlargerinthissubsample(0.5percentgrowthinGDPrateinyear2versus0.2percentinfullcarbontaxsampleasperfigure3b)butthecoefficientsareimpreciselyestimated,andwecannotrejectthatthegrowthratesarethesame(andbothequaltozero).Theemploymentimpactsareinitiallylarger(bottompanel)thaninthefullsample(comparetoFigure6b).Thegrowthrateinyear2is0.9percentagepointshigherwhereasinthefullsampleforthecomparableregression,itis0.5percentagepointshigher.Thecumulativeimpactbyyear6is0.8percentagepointshigheremploymentgrowthversus0.4percentagepointsinthefullsample.AswithGDPgrowth,wecan’trejectthattheyarethesame(andequaltozero).Withonlysixcountriesinthetreatmentgroup(andshortspansofthecarbontaxforSwitzerlandandPortugal),itisdifficulttomakedefinitivestatementsaboutrevenuerecyclingwithourdata.Anotherwaytogetatthisquestionistolookatthosecountriesthatarenotdeemedrevenuerecyclingcountries.Thisisalargergroupandperhapswecanobservemeaningful22differenceshere.Figure10showsGDPgrowthimpacts(toppanel)andtotalemploymentimpacts.TheestimatesarequitenoisybutsuggestaninitialdeclineinGDPandemploymentfollowedbyareboundinyears3–5.Theeffectfadesawaybyyear6.ThecumulativeimpactforGDPispositivebutbarelyexceedsitsstandarderror.Thecumulativeimpactbyyearsixfortotalemploymentisnegativebutlessthanitsstandarderror.Thepointestimatesprovidesomemodestsupportforgrowthenhancingbenefitsofrecyclingcarbontaxrevenuesthroughtaxcuts,butstandarderrorsarelargeandourmeasureofrevenuerecyclingreflectsonlystatedinitialintentions,notactualuseoftherevenues,sooneshouldbecautiousbeforedrawingconclusionsabouttheefficiencybenefitsofrevenuerecyclingthroughtaxcutsbasedontheseregressions.Emissionsinitiallyfallfasterinthenon‐revenuerecyclingcountriesrelativetotherevenuerecyclingcountries,butemissionsreductionsbegintoconverge:emissionsare4.9percentlowerinthenon‐revenuerecyclingcountriesbyyear6whiletheyare3.1percentlowerintherevenuerecyclingcountries(Figure11).Asinthefullsample,theestimatesareimprecise,andwecannotrejectnochangeincumulativeemissions.16B.LargeCarbonTaxesCountriesalsodifferinthemagnitudeoftheircarbontaxrates.Wewouldexpectlargerimpactsincountrieswithhighertaxrates,holdingallelseequal.Herewefocusonthosecountrieswhoseshare‐weightedcarbontaxratesareatleast$10pertoninatleastoneyear16Resultslooksomewhatdifferentintheregressionswhereweallowthelongrungrowthrateofemissionstochange.Cumulativeemissionsfallinbothsampleswithalargercumulativedeclineinthenon‐revenuerecyclingcountries.Emissionsfallby3.7percentintherevenuerecyclingcountriesbyyear6andby18.6percentinthenon‐revenuerecyclingcountries.Forthislattersetofcountries,the95percentconfidenceintervalinyear6runsfrom‐4.7to‐32.6percent.23(thuscorrespondingto$30/toncoveringone‐thirdofemissions).ThosecountriesareDenmark,Finland,France,Ireland,Norway,Sweden,andSwitzerland.Figure12showstheimpactofthetaxinthesecountriesonGDPgrowth.Inboththeunrestrictedmodel(toppanel)andrestrictedmodel(bottompanel)theGDPgrowthimpactisinitiallypositive,dipsnegativeinyears3and4,andthenreturnstopositive.Thecumulativeimpactbyyear6is1.75percentagepointsintheunrestrictedmodeland‐0.16percentagepointsintherestrictedmodel.Asimilarpatternholdsforemployment(notshown).Surprisingly,emissionsfallbyonly2percentintheunrestrictedandrestrictedmodelsbyyear6(Figure13);thisisasmalleramountthaninthefullsamplewhereemissionsfellby6.5percentintheunrestrictedmodeland3.8percentintherestrictedmodel.Aswiththefullsample,coefficientestimatesareimpreciselyestimated.C.ScandinaviaWealsoconsideredwhetherourresultsarebeingdrivenbytheearlyScandinavianadopters(Denmark,Finland,NorwayandSweden).TheGDPandtotalemploymentimpulseresponsefunctionsfortheScandinaviancountriesarequitenoisyandhoveraroundzero(withlargestandarderrors).DroppingthesecountriesfromtheEUdatasetalsoincreasethestandarderrorsconsiderably.WhiletheScandinaviancountriesarenotdrivingresults,theyhelpreducestandarderrorsconsiderably.Wereportthoseresultsintheappendix.D.AlternativeTaxRateMeasuresOurtaxrateseriesisbasedondatacollectedbytheWorldBankandusesthehighestcarbontaxrate(pertonCO2)whentherearemultiplerates.Inmostcases,thishighestrateistherateongasolineanddieselforroaduse.Wemultiplythatratebytheshareofemissions24coveredbythecarbontaxin2018.RecentlyDolphinetal.(2019)haveconstructedaseriesoftaxratesoncarbondioxideemissionsbuiltupfromfuelleveltaxrates.Workingwithsectorandfuelspecificdata,theycomputetheshareofvariousfossilfuelsineachsectorcoveredbyacarbontaxandconstructanemissionsweightedcarbonprice(ECP)astheweightedaverageofsector‐fuelspecificcarbontaxrates,weightedbytheiremissionssharein2013.Were‐estimatedourregressionsusingtheirdata,andselectedresultsareshownintheAppendix.17Despitedifferencesinthebaseyearforfixingemissionsshares,exchangerates,anddifferentmethodologiesforconstructingtheshareweightedtaxrates,theresultsusingtheDophinetal(2019)emissions‐weightedcarbonpriceareverysimilartothosereportedusingourpricederivedfromWorldBankdata.Weconcludefromthisthatourresultsarerobusttohowcarbontaxesaremeasured–whetherbasedoncountryreportedcarbontaxratesorbuiltupfromsectorspecificexcisetaxratesfortaxingcarbon.VII.ConclusionPlacingapriceoncarbonpollutioniswidelyviewedasthemostcost‐effectiveapproachtoreducingemissions.Resistancetothisapproachissignificantinpartduetoconcernsabouttheeconomicimpactonjobsandgrowth.UsingvariationintheuseofcarbontaxesinEuropeancountriesthatareallpartoftheEUEmissionTradingSystem(ETS),wefindnoevidencetosupportclaimsthatthetaxwouldadverselyimpactemploymentorGDPgrowth.Wefindmodestevidenceforemissionsreductionsarisingfromthetax.Itisworthnoting,however,thatsincethesectorsforwhichemissionreductionsareleastcostlyarealready17Wethanktheauthorsforsharingthecountryaggregatetaxrateswithus.25coveredbytheETSandsonotsubjecttothecarbontax,emissionreductionsincountriesthatapplyabroad‐basedcarbontaxshouldviewthesereductionsaslowerboundsonpotentialemissionreductions.26ReferencesAndersson,JuliusJ.2019."CarbonTaxesandCO2Emissions:SwedenasaCaseStudy."AmericanEconomicJournal:EconomicPolicy,11(4),1-30.Bernard,Jean-Thomas;MaralKichianandMisbahulIslam.2018."EffectsofB.C.’SCarbonTaxonGDP,"USAEE,ResearchPaperSeriesNo.18-329.Brannlund,RunarandIng-MarieGreneds.1999.GreenTaxes:EconomicTheoryandEmpiricalEvidencefromScandinavia.Cheltenham,UK:EdwardElgar.Carl,JeremyandDavidFedor.2016."TrackingGlobalCarbonRevenues:ASurveyofCarbonTaxesVersusCap-and-TradeintheRealWorld."EnergyPolicy,96,50-77.Convery,Frank;LouiseDunneandDeirdreJoyce.2013."Ireland'sCarbonTaxandtheFiscalCrisis,"OECD,OECDEnvironment:Paris,WorkingPaperNo.59.Dolphin,Geoffroy;MichaelG.PolittandDavidM.Newbery.2019."ThePoliticalEconomyofCarbonPricing:APanelAnalysis."OxfordEconomicPapers,72(2),472-500.EuropeanCommission.2015."EUETSHandbook,"EuropeanCommission:Brussels,EuropeanCommission.2019."EurostatDatabase,"https://ec.europa.eu/eurostat/data/database,AccessedonDecember20,2019.Goulder,LawrenceH.1995."EnvironmentalTaxationandthe'DoubleDividend':AReader'sGuide."InternationalTaxandPublicFinance,2,157-183.Goulder,LawrenceH.andMarcHafstead.2017.ConfrontingtheClimateChallenge.NewYork:ColumbiaUniversityPress.Goulder,LawrenceH.;MarcA.C.Hafstead;GyuRimKimandXianlingLong.2019."ImpactsofaCarbonTaxacrossUSHouseholdIncomeGroups:WhatAretheEquity-EfficiencyTrade-Offs?"JournalofPublicEconomics,175,44-64.Hammar,HenrikandSusanneAkerfeldt.2011."CO2TaxationinSweden:20YearsofExperienceandLookingAhead."Jordà,Òscar.2005."EstimationandInferenceofImpulseResponsesbyLocalProjections."AmericanEconomicReview,95(1),161-182.Lin,BoqiangandXuehuiLi.2011."TheEffectofCarbonPriceonPerCapitaCO2Emissions."EnergyPolicy,39,5137-5146.Marten,MelanieandKurtvanDender.2019."TheUseofRevenuesfromCarbonPricing,"Paris,OECDTaxationWorkingPapersNo.43.27Martin,Ralf;LaurieB.dePreuxandUlrichJ.Wagner.2014."TheImpactofaCarbonTaxonManufacturing:EvidencefromMicrodata."JournalofPublicEconomics,117,1-14.Metcalf,GilbertE.2019."OntheEconomicsofaCarbonTaxfortheUnitedStates."BrookingsPapersonEconomicActivity,(Spring),405-458.Metcalf,GilbertE.andJamesH.Stock.2020."MeasuringtheMacroeconomicImpactsofCarbonTaxes."AmericanEconomicReview:PapersandProceedings,110(May),101-106.Moyer,ElisabethJ.;MarkD.Wooley;NathanJ.Matteson;MichaelJ.GlotterandDavidA.Weisbach.2014."ClimateImpactsonEconomicGrowthasDriversofUncertaintyintheSocialCostofCarbon."JournalofLegalStudies,43(2),401-425.NERA.2017."ImpactsofGreenhouseGasRegulationsontheIndustrialSector,"NERA,Washington,DC.Pereira,AlfredoM.;RuiM.PereiraandPedroG.Rodrigues.2016."ANewCarbonTaxinPortugal:AMissedOpportunitytoAchievetheTripleDividend?"EnergyPolicy,93,110-118.Plagborg-Moller,MikkelandChristianK.Wolf.2019."LocalProjectionsandVARsEstimatetheSameImpulseResponses,"Princeton:DepartmentofEconomics,PrincetonUniversity,Prettis,Felix.2019."DoesaCarbonTaxReduceCO2Emissions?EvidencefromBritishColumbia,"DepartmentofEconomics,UniversityofVictoria,Victoria,BC.Rivers,NicholasandBrandonSchaufele.2015."SalienceofCarbonTaxesintheGasolineMarket."JournalofEnvironmentalEconomicsandManagement,74,23-36.Stock,JamesH.andMarkW.Watson.2018."IdentificationandEstimationofDynamicCausalEffectsinMacroeconomics."EconomicJournal,128,917-948.Sumner,Jenny;LoriBirdandHillaryDobos.2011."CarbonTaxes:AReviewofExperienceandPolicyDesignConsiderations."ClimatePolicy,11,922-943.Svenningsen,LeaSkraep;LivLaerkeHansen;MichaelMunkSorenson;EmelievonBahr;HrafnhildurBragadottir;KennetChristianUggeldahl;HanneSoiland;OyvindLone;JorgenSchouandTroelsHansen.2019."TheUseofEconomicInstrumentsinNordicEnvironmentalPolicy2014-2017,"NordicCouncilofMinisters,Copenhagen.Trump,DonaldJ.2017."StatementbyPresidentTrumpontheParisClimateAccord,"https://www.whitehouse.gov/briefings-statements/statement-president-trump-paris-climate-accord/,AccessedonDecember10,2019.28WorldBankGroup.2019a."CarbonPricingDashboard,"https://carbonpricingdashboard.worldbank.org/map_data,AccessedonDec.1,2019.WorldBankGroup.2019b."StatesandTrendsofCarbonPricing2019,"WorldBankGroup,WashingtonDC.Worstall,Tim.2016."AbsolutelyFascinating-Apple'sEUTaxBillExplainsIreland's26%GDPRise,"https://www.forbes.com/sites/timworstall/2016/09/08/absolutely-fascinating-apples-eu-tax-bill-explains-irelands-26-gdp-rise/#15945ea91a70,AccessedonDec.10,2019.Yamazaki,Akio.2017."JobsandClimatePolicy:EvidencefromBritishColumbia'sRevenue-NeutralCarbonTax."JournalofEnvironmentalEconomicsandManagement,83,197-216.29Table1.EU+CarbonTaxesCountryYearofEnactmentRatein2018(USDpermetricton)IntendedRevenueRecycling?ShareofGreenhouseGasEmissionsin2019CoveredbyTaxCarbonTaxRevenuein2018(USDMillions)Denmark(DNK)199224.92Yes40%543.4Estonia(EST)20003.65No3%2.8Finland(FIN)199070.65Yes36%1,458.6France(FRA)201457.57No35%9,263.0Iceland(ISL)201025.88No29%44.0Ireland(IRL)201024.92No49%488.8Latvia(LVA)20049.01No15%9.1Norway(NOR)199149.30Yes62%1,659.8Poland(POL)19900.16No4%1.2Portugal(PRT)201511.54Yes29%154.9Slovenia(SVN)199629.74No24%83.1Spain(ESP)201430.87No3%123.6Sweden(SWE)1991128.91Yes40%2,572.3Switzerland(CHE)200880.70Yes33%1,177.7UK(GBR)201325.71No23%1,091.0Notes:Coverageistheshareofacountry’semissionscoveredbythecarbontax.Seetextforrevenuerecyclingdetails.Source:WorldBankGroup(2019a)30Table2.TestofLongRunEffectofCarbonTaxonGrowthRatesandEmissionsFullSampleGDPGDPperCapitaTotalEmploymentManufacturingEmploymentEmissionsLP0.330.75‐0.170.87‐0.630.53‐3.370.001‐2.090.04SVAR1.340.182.070.040.620.53‐0.510.61‐1.260.21RevenueRecyclingCountriesLP0.050.96‐0.440.66‐0.720.47‐3.590.000‐0.950.34SVAR1.390.162.430.020.170.87‐0.460.65‐0.400.69LargeCarbonTaxCountriesLP‐0.410.69‐0.360.720.140.89‐3.290.001‐0.530.60SVAR1.000.320.920.361.230.22‐0.160.87‐0.340.73ScandinavianCountriesLP‐0.440.66‐0.600.550.800.42‐2.200.030.190.85SVAR0.950.341.000.321.040.300.160.870.160.87Tablereportsresultsofthetestthatthereisnolong‐runchangeinthegrowthrate.Failuretorejectthenullsupportsthenolong‐runchangehypothesis.Thetablereportsthet‐statisticforthetestinthetoprowanditsp‐valueinthesecondrow.Seetextfordescriptionoftest.31Table3.TestofStrictExogeneityGDPGDPperCapitaTotalEmploymentManufacturingEmploymentEmissionsFullSample2.210.0662.530.0381.040.3840.820.510.650.63RevenueRecyclingCountries1.980.0942.020.0891.050.380.630.640.700.59LargeCarbonTaxCountries3.860.0044.090.0032.410.0471.970.0961.570.18ScandinavianCountries2.550.0372.540.0383.650.0061.600.173.810.004TablereportsresultsofastrictexogeneitytestthatthereisnofeedbackfromshockstoGDPtotaxrates.ThetablereportstheFstatisticwith(8,inf)degreesoffreedominthetoprowanditsp‐valueinthesecondrow.Seetextfordescriptionoftest.32Figure1.RealCarbonTaxRatesOverTime(SeeTable1forCountryAbbreviations)33Figure2.CarbonTaxEnactmentandGDPPerCapitaGrowthRate34Figure3a.GDPIRFforLPRegression–UnrestrictedFigure3b.GDPIRFforLPRegression–Restricted35Figure4a.GDPIRFforSVARRegression–UnrestrictedFigure4b.GDPIRFforSVARRegression–Restricted36Figure5a.GDPCIRFforLPRegression–UnrestrictedFigure5b.GDPCIRFforLPRegression–Restricted37Figure6a.TotalEmploymentIRFforLPRegression–UnrestrictedFigure6b.TotalEmploymentIRFforLPRegression–Restricted38Figure7a.TotalEmploymentCIRFforLPRegression–UnrestrictedFigure7b.TotalEmploymentCIRFforLPRegression–Restricted39Figure8a.EmissionsCIRFforLPRegression–UnrestrictedFigure8b.EmissionsCIRFforLPRegression–Restricted40Figure9a.GDPIRFforLPRegression–Restricted:RevenueRecyclingCarbonTaxCountriesOnlyFigure9b.TotalEmploymentIRFforLPRegression–Restricted:RevenueRecyclingCarbonTaxCountriesOnly41Figure10a.GEPIRFforLPRegression–Restricted:Non‐RevenueRecyclingCarbonTaxCountriesOnlyFigure10b.TotalEmploymentIRFforLPRegression–Restricted:Non‐RevenueRecyclingCarbonTaxCountriesOnly42Figure11a.EmissionsCIRFforLPRegression–Restricted:RevenueRecyclingCountriesFigure11b.EmissionsCIRFforLPRegression–Restricted:Non–RevenueRecyclingCountries43Figure12a.GDPIRFforLPRegression–Unrestricted:LargeCarbonTaxCountriesOnlyFigure12b.GDPIRFforLPRegression–Restricted:LargeCarbonTaxCountriesOnly44Figure13a.EmissionsCIRFforLPRegression–Unrestricted:LargeCarbonTaxCountriesFigure13b.TotalEmissionsCIRFforLPRegression–Restricted:LargeCarbonTaxCountries45Appendix–AdditionalTablesandFiguresI.EventStudyFiguresFigureA1.CarbonTaxEnactmentandTotalEmploymentGrowthRateFigureA2.CarbonTaxEnactmentandManufacturingEmploymentGrowthRate46FigureA3.CarbonTaxEnactmentandCO2Emissions47II.ImpulseResponseFunctions–ManufacturingEmploymentFigureA4.ManufacturingEmploymentIRFforLPRegression–UnrestrictedFigureA5.ManufacturingEmploymentIRFforLPRegression–Restricted48III.ImpulseResponseFunctionsFocusingonRoleofScandinavianCountriesFigureA6.GDPIRFforLPRegression–Restricted:ExcludingScandinavianCountriesFigureA7.TotalEmploymentIRFforLPRegression–Restricted:ExcludingScandinavianCountries49FigureA8.GDPIRFforLPRegression–Restricted:ScandinavianCountriesOnlyFigureA9.TotalEmploymentIRFforLPRegression–Restricted:ScandinavianCountriesOnly50IVResultsUsingEmissionsWeightedCarbonPricesfromDolphin,etal.(2019)BelowareimpulseresponsefunctionsfortheGDPandTotalEmploymentgrowthratesandacumulativeimpulseresponsefunctionfortotalemissionsfortheLPrestrictedmodelusingDolphinetal.’semissionsweightedcarbonprices.ResultsareverysimilarwhentheLPunrestrictedorSVARapproachesareused.TheseresultsareroughlycomparabletoourssincethemeanemissionsshareintheDolphinetal.datafortheirfixed‐weightcarbonpriceis34percent.18Thethreefiguresthatfollowshouldbecomparedtofigures3b,6b,and8brespectivelyinthepaper.FigureA10.GDPIRFforLPRegression–Restricted:ECPtaxdataused18PersonalcommunicationwithGeoffroyDolphinonJune3,2020.51FigureA10.TotalEmploymentIRFforLPRegression–Restricted:ECPtaxdatausedFigureA11.EmissionsCIRFforLPRegression–Restricted:ECPtaxdataused
VIP
