美國(guó)電動(dòng)汽車(chē)需求和充電站供應(yīng)建模-2023-09-新能源

WorkingPaperSeries

CongressionalBudgetOffice

Washington,D.C.

ModelingtheDemandforElectricVehicles

andtheSupplyofChargingStations

intheUnitedStates

DavidAustin

CongressionalBudgetOffice

david.austin@

WorkingPaper2023-06

September2023

ToenhancethetransparencyoftheworkoftheCongressionalBudgetOfficeandtoencourageexternalreviewofthatwork,CBO’sworkingpaperseriesincludespapersthatprovidetechnicaldescriptionsofofficialCBOanalysesaswellaspapersthatrepresentindependentresearchbyCBOanalysts.Papersinthisseriesareavailableat

/xUzd7.

TheinformationinthispaperispreliminaryandisbeingcirculatedtostimulatediscussionandcriticalcommentasdevelopmentalworkforanalysisfortheCongress.

IamgratefultoNicholasChase,TerryM.Dinan(formerlyofCBO),MichaelFalkenheim,RonGecan,

EvanHerrnstadt,JosephKile,AaronKrupkin,RobertReese,andChadShirleyforhelpfulcommentsandsuggestions.IalsothankShanjunLiofCornellUniversityandJohnMaplesoftheEnergyInformation

Administration.Althoughthoseexpertsprovidedconsiderableassistance,theyarenotresponsibleforthecontentsofthispaper.RebeccaLanningeditedthepaper,andCaseyLabrackcreatedthegraphics.

ii

Abstract

Thispaperpresentsasimulationmodelofthemarketsforlight-dutyelectricvehicles(EVs)andtheassociatedpubliccharginginfrastructure,aswellasthenetworkinteractionsbetweenthem.Itillustratesthemodel’sattributesbysimulatingtheeffectsoffederalsubsidiesforpublicelectricvehiclechargersandofanextensionoftaxcreditsforelectricvehicles.Iprojectthatbytheearly2030sthechargersubsidies,whichweresignedintolawin2021aspartoftheInfrastructure

InvestmentandJobsAct,willhaveincreasedthesizeofthechargernetworkenoughtomeetthedemandforchargingthroughthemiddleofthatdecade.Thatincludestheadditionaldemandthattheexpansionitselfwillinduce:Iprojectthatthrough2030,salesofEVswillrisemorethan

20percentmorerapidlywiththeexpandedchargernetworkthantheywouldhaveotherwise.

IncludingtheadditionaleffectoftheEVtaxcreditsthatweresignedintolawaspartofthe2022reconciliationact,aswellaspastgrowthinEVsales,IprojectthatEVswillconstitutebetween27percentand60percentofnewlight-dutyvehiclesalesby2032,comparedwithabout

6.5percentin2022.AfterthesubsidyfundingfromtheInfrastructureInvestmentandJobsActhasbeenspentandtheavailableEVtaxcreditsclaimed,EVchargernetworksandtheEVfleetwillremainsomewhatlargerthantheywouldhavebeenintheabsenceofthosepolicies.

Keywords:Electricvehicles,charginginfrastructure,networkeffects,taxcredits,subsidies

JELClassification:H23,H54,L98

iii

Contents

Abstract ii

Contents iii

1.Introduction 1

2.DemandforElectricVehicles 3

2.1FactorsThatInfluenceEVMarketShare 4

2.2AttributeDrift 6

2.3ModelCalibrationAdjustmenttoAttributeDriftTerm 7

3.SupplyofEVChargers 10

4.ModelBaseCase:FederalChargerSubsidiesandEVTaxCredits 12

4.1FederalChargerSubsidies:DescriptionofModelingApproach 13

4.2FederalEVTaxCredits:DescriptionofModelingApproach 14

4.3PolicyEffectsonEVChargerNetworks 17

4.4EffectsonMarketShareofNewEVs 20

4.5ModelUncertainty 23

4.6ComparisonofCBO’sandOtherProjections 24

AppendixA:ParameterValues 26

AppendixB:SensitivityAnalyses 28

SensitivityAnalysis1:ChangethePriceElasticityofNew-EVMarketShareby25Percent.28

SensitivityAnalysis2:ChangetheSensitivityofEVDemandWithRespecttoEVChargers

by25Percent 29

SensitivityAnalysis3:AlternativeTimeTrendsfortheAttributeDriftTerm 31

SensitivityAnalysis4:ChangetheRateofDecreaseinEVProductionCostsby50Percent.32

SensitivityAnalysis5:CompareICCT’sLow,Moderate,andHighCasesforEVTax-Credit

Eligibility 34

SensitivityAnalysis6:ChangetheAmountofEVLeasingThatOccursasaWayofClaiming

EVTaxCredits 36

AppendixC:AProposedRuleonTailpipeEmissions 38

References 41

1

1.Introduction

Thispaperprovidesanoverviewofananalysisthatjointlymodelsthedemandforlight-duty,

plug-inelectricvehicles(EVs)andthesupplyofpublicEVchargers.AstheCongressconsidersoradoptsarangeofpoliciesthatwouldsubsidizesalesofplug-inelectricvehicles,eitherdirectlyorindirectly,thismodelgivestheCongressionalBudgetOfficeanadditionaltoolforestimatingtheeffectsofthosepoliciesonthefederalbudget,carbondioxideemissions,andthedemandforelectricityandgasoline.Themodelprojectssalesofnewelectricvehicles—includingplug-in

hybrid-electricvehicles—asashareofsalesofallnewpassengervehicles.Thatshareismodeledasafunctionofpredictedvehiclecosts(production,operation,andmaintenance);thesizeoftheEVchargernetwork(includingslowerlevel2,orL2,chargerswithrespecttothenumberof

registeredEVsandfasterL3,or“DCfast,”chargerswithrespecttothesizeofthenational

highwaysystem);andconsumers’preferencesforEVsorsimilarcarswithinternalcombustionengines.ThemodelalsoprojectsthestockofEVchargersasafunctionofthepredictedcostofsupplyingachargerandtheprojectedsizeoftheelectricvehiclefleet.

ThemodelbuildsonrecentworkbyColeetal.(2023),whomodelthedemandforEVsandthesupplyofEVchargersasjointlydetermined.1ThedemandforEVsispresentedasdependingonthesizeoftheEVchargernetwork:Thelargerthatnetwork,themoreutilityanEVprovides,

becauseitcanbedriventomoreplacesandrechargedmoreeasily.Similarly,inColeetal.

(2023)thesupplyofchargersdependsonsalesofEVs:ThelargertheEVfleet,themoreuseeachEVchargerreceivesandthemorerevenueitgeneratesperunitoftime.

Themodeldescribedheretakesasimilarapproachbutwithtwoimportantdifferences.First,it

incorporatesintoitsbasecasetheanticipatedeffectsofupto$7.5billioninfederalEVchargersubsidiesprovidedbytheInfrastructureInvestmentandJobsAct(IIJA;PublicLaw117-58)in

2021,alongwiththeanticipatedeffectsoftheEVtaxcreditsprovidedbythe2022reconciliationact(P.L.117-169).2

Beforetheenactmentofthe2022reconciliationact,CBOandthestaffoftheJointCommitteeonTaxation(JCT)preparedanestimateofitsbudgetaryeffects(CBO2022).Asrequiredbythe

CongressionalBudgetActof1974,CBOestimatedtheeffectsofthespendingprovisions,andJCTestimatedtheeffectsofthetaxprovisions.Bystatute,thecostestimateforthe2022

reconciliationactpublishedbyCBOdirectlyincorporatedJCT’sestimatesofthebudgetary

effectsoftheenergy-relatedtaxprovisionsofthatbill,includingthoserelatedtoelectricpower,

1Coleetal.(2023),citingZhouandLi(2018)andSpringel(2021).

2ThebasecasepresentedinthispaperisnotaninputtoCBO’sbudgetbaseline.Itisanoutputofastand-aloneCBOmodelbasedonrecentmarkettrendsandfindingsfromtheliterature.

2

electricvehicles,carboncaptureandsequestration,andcleanenergymanufacturing.Themodeldescribedinthisworkingpaperwasnotusedinpreparingthatcostestimate.

TheseconddifferencefromtheapproachtakenbyColeetal.(2023)isthatIcalibratethemodelsothatitsbase-yearsupplyofchargersmatchesthemostrecentfull-yearnumberreportedintheAlternativeFuelsDataCenter’sstationlocatordatabaseandsothat,oncetheIIJAsubsidiesareexhausted,itprojectsaratioofL2chargerstoEVsapproximatelyequaltothecurrentratioof

about3chargingportsper100EVs.3Thatservesasmyestimateoftheoptimal(profit-

maximizing)ratioforchargersuppliers,evenasthenumberofregisteredEVscontinuesto

increase.Coleetal.(2023)calibratetheirmodeltoafutureratiothatisaboutthreetimeshigher.

Inthispaper,IdiscussthedemandforEVsandparticularlymyapproachtowardattributedrift,orchangesincertainattributesofEVownershipthataffectconsumers’preferencesforEVsorinternalcombustionenginevehicles(ICEVs).(Thoseattributes,whicharenototherwise

specifiedinmymodel,includetheavailabilityandperformanceofEVchargers,social

influences,andmanyattributesofthevehiclesthemselves.)TheattributedrifttermallowsthemodeltomoreaccuratelyaccountforrecenttrendsinEVsales.Thosesalesexceedwhatthe

modelwouldpredictsolelyonthebasisoftheothertermsinthedemandequation:therelativeownershipcostsofEVsandICEVs,thenumberofpubliclyaccessiblechargersperEVandperhighwaymile,andpastsalesofEVs.

Next,IdiscussthesupplyofEVchargers.ThemodelhasasupplyequationgivingthenumberofchargingstationsbyyearasafunctionofthesizeoftheEVfleet,thecostofsupplyingachargerthatyearcomparedwiththeanticipatedlowercostthenextyear,andthenumberofcharging

stationsthatexistedthepreviousyear.Ifinanyyearitisoptimal,inthemodel,forsupplierstoaddnonewEVchargers,thesizeofthechargernetworkwilldeclinethatyearbythenumberofchargersthatwillfail,basedonratesoffailurethatincreasewithchargerage.ThediscussionofchargersupplyalsodescribeshowImodelthefederalIIJAsubsidiesforEVchargers.

IusethemodeltoprojectannualsalesofnewEVsandthesupplyofnewchargingstations

through2050.Thedemandandsupplyequationsinteract:ThesizeofthechargernetworkaffectsEVdemand,andthesizeoftheEVfleetaffectsthesupplyofchargers.Detailsaboutthevaluesusedforthemodel’sparametersandsixsensitivityanalysesofinfluentialparametersare

providedin

AppendixA

and

AppendixB,

respectively.

ThemodeldescribedinthispaperprovidesCBOwithatoolforestimatingtheeffectsof

developmentsintheautomobileindustry,andinfederalpolicytowardEVsales,onthefederal

3ThenumberofEVchargersperchargingstationvaries,currentlyaveragingabout2chargersperstationaccordingtothestationlocatordatabase.EVchargersmayhavemultipleportsforchargingmorethanonevehicleatatime,

justasfuelpumpswithmultiplenozzlescanrefuelmorethanonevehicleatatimeatagasstation.

3

budgetandtheeconomy.Thispaperprovidestransparencyintothatmodel,whichwas

developedwhiletheTreasuryDepartmentwasdevelopingguidanceoneligibilityrequirementsforelectricvehiclestoqualifyfortaxcredits.ThepercentageofEVsthatwillultimatelyqualifyforthosecreditsisnotknownwithprecision.Theprojectionsprovidedinthispaperreflectthatandothersourcesofuncertainty(see

AppendixA)

.

2.DemandforElectricVehicles

Inthemodel,vehiclebuyerschoosebetweenelectricandinternalcombustionversionsoftheirdesiredvehicle.Ianalyzecarsandlighttrucksseparatelybutdonotdistinguishbetween

differentvehiclemakesandmodels.Inthevehiclesalesdatausedinthissimulationmodel,Iclassifyplug-inhybridvehiclesasEVsandnon-plug-inhybridvehiclesasICEVs.

Aconsumer’schoiceofanEVoranICEVdependsonexpectedownershipcosts,thedensityof

theEVchargernetwork,andshiftsinconsumers’preferencesforEVs(asmodeledbyattributedrift).Iestimateavehicle’sexpectedownershipcostasitspurchaseprice—accountingfortheEVtaxcreditscontainedinthe2022reconciliationact—plusitsexpectedoperatingand

maintenancecostsoveritsfirsteightyears,discountedtothedateofpurchase.4ImeasurethedensityofthechargernetworkintermsofthenumberofrapidchargersperhighwaymileandthenumberofslowerchargersperEV.5

Attributedriftcanbethoughtofasreflectingtheeffectsonconsumers’preferencesforEVsorICEVsoffactorsnototherwiseincludedinthemodel.Examplesofsuchfactorsinclude

improvementsinEVsorthechargernetwork,breadthandavailabilityofEVmodelsrelativetoICEVs,andsocialinfluencessuchasproportionofEVownershipamongacquaintancesand

otherdrivers.Withoutattributedrift,themodelwouldprojectfutureEVsalessolelyonthebasisoftrendsinvehiclecostsandnetworkdensity.Icalibratethedrifttermsothatthemodel’s

4InvaluingexpectedfuturesavingsonmaintenanceandoperatingcostsforanEVversusanICEV,Iuseadiscountrateofabout10percent.Thatratecombinesapreferenceforreceivingvaluetodayversusinthefuturewithan

observedtendencyforconsumerstoundervaluefuturesavingsfromenergy-efficienttechnologies.Iuse3percentas

consumers’rateoftimepreference,andImodelconsumersasundervaluingfuturesavingsinenergyand

maintenancecostsbyanaverageof25percent.(Theactualreductioncomesfromaprobabilitydensitythataverages25percent.)SeeAllcottandWozny(2014);seealsoHelfandandWolverton(2011).Finally,Imodelconsumersasvaluingfuturesavingsovereightyearsratherthanovertheexpectedlifeofthevehicle.Thecombinationofthose

factorsamountstodiscountingfuturesavingsatanannualrateofabout10percent.InsensitivitytestingIfindthatcountingjustfiveyearsofexpectedsavingsdoesnotsubstantiallychangethemodel’sprojections.

5AlthoughmyEVprojectionsdonotaccountforpaneltrucksorotherfreight-deliveryvehicles,thosevehicleswilltypicallyrechargeinprivatefleetfacilitiesovernightoratdedicatedtruckstopsandwillthusnottendtocompete

withpassengervehiclesforaccesstothepubliccharginginfrastructure.L3rapidchargerscanrechargemostcarstoabout80percentofcapacityinabout20minutesandaremostsuitableforplacementalonghighways.SlowerL2

chargerstakefourorfivehourstoprovideacomparablechargeandaremoresuitableforplacementinparkingfacilities.Forchargingtimes,seeAlternativeFuelsDataCenter,“DevelopingInfrastructuretoChargeElectricVehicles.”

4

projectedEVsalesthrough2030—ignoringtheexpectedinfluenceofthefederalIIJAchargersubsidiesandEVtaxcredits—areconsistentwiththeobservedEVsalestrendoverthepast

severalyears.(ThenextsectionpresentsEVsalesprojectionsthatarebasedondifferent

assumptionsaboutattributedrift.)Finally,Iadjusttheinterceptandattributedrifttermssothat,giventheotherparametervaluesinthemodel,itsbase-yearEVsalesmatchobservedtotalsfrommostrecentfullyear,currently2021.

2.1FactorsThatInfluenceEVMarketShare

ImodelthedemandforEVsasarisingfromanunderlyingconsumerutilityfunction:

ui,j,t=aj+ln?pj,t??βp+ln?L2tΤTotEvt?1??βL2+ln?L3tΤHWYMiles??βL3+ψt+Ei,j,t,

whereiindexesindividualconsumers,jisvehicletype(carorlighttruck),andtistimeinyears.Theβtermsareparametersassociatedwith,respectively,thesensitivityofthedemandforEVs

tochangesinthefollowing:ownershipcostpj,tofanEVoftypejinyeartrelativetothatfora

comparableICEV;thenumberofslowerL2chargersperregisteredEVinyeart;andthenumberoffasterL3chargersperhighwaymile.6Finally,ψtisattributedrift,andEi,j,tisanidiosyncratictasteshockdistributedastypeI(Gumbel)extremevalue(Coleetal.2023).Ifurtherdescribetheparametersbelow.

ToensurethatEVandICEVownershipcosts(pj,t)arecomparedonanequalbasis,Icalculate

thecostsforbothtypesofvehicleusingthenumberofmilesthatapotentialbuyerwouldexpecttodriveinanEV.Untilrecently,EVswereestimatedtobedrivenonlyabout60percentasmanymilesasICEVs,onaverage(Burligetal.2021).WithexpansionintheEVchargernetworkandimprovedbatterycapacities,thatratioappearstobeincreasingovertime.Toreflectthat,ImodeltheratioofmilesdrivenbyEVsversusICEVsascurrentlyaveraging60percentandgradually

increasingto100percentby2035.

ThatincreasecontributestogrowthintheprojecteddemandforEVs,becauseitmeansthat

expectedannualenergysavingsfromEVsversusICEVsarealsoincreasing.Somecurrent

evidencesuggeststhatmanyEVsarealreadybeingdrivenasmanymilesascomparableICEVs(Spilleretal.2023).Ifso,currentdemandforEVsmayalreadyreflectmuchofthoseenergy

savings,andthusthemodelmaybeslightlyoverstatingthatsourceofgrowthinprojected

demand.However,thecontributionofhigherfutureenergysavingstogrowthinthedemandforEVsisrelativelysmall.

6ItreatHwyMilesasconstant,althoughitmayincreasegraduallyovertime.Ialsoholdconstanttheenergy

efficiencyofEVs(althoughbatterycostscontinuallydecline,whichisequivalenttoincreasingenergyefficiency

fromtheperspectiveofownershipcosts)andICEVsaftermodelyear2026becausecorporateaveragefueleconomy(CAFE)standardsforlatermodelyearshavenotyetbeenspecified.Through2026,ItreatbothaverageICEVfueleconomyandmanufacturingcostsasrisingwithincreasinglystringentCAFEstandards(see

AppendixA)

.

5

LiketheeffectofanticipatedgrowthinEVmilesdriven,therateofdecreaseinEVbatterycostsisalsofavorabletoEVs.Thatratecouldslowifitbecamemoredifficulttominethescarce

materialsusedinbatteries.Conversely,innovationsinbatterytechnology,whichwouldbespurredbyconcernsaboutthescarcityofmaterials,couldsustainorincreasebatterycostreductions(see

AppendixA)

.

ExistingempiricalresearchonthesensitivityofEVdemandtothesizeofthechargernetworkdoesnotdistinguishbetweenL2andL3chargers.Thus,IassignthesamevaluetobothβL2andβL3.Evenso,withabout15.4registeredEVsperhighwaymileintheUnitedStatesattheendof2022—aratiothatwillincreaseovertime—eachadditionalL3chargeristhereforemodeledashavingln?TotEVΤHWYMiles?=ln?15.4?=2.7timesmoreinfluenceonEVdemandat

presentthaneachnewL2chargerhas.7

Thatconsumerutilityfunction—or,moreprecisely,thetypeIextreme-valuetermreflecting

variationinindividuals,vehicles,andtimeinconsumers’preferencesforpassengervehicles—

yieldsatractableandlogicallyappealingexpression(itsvaluesrangefrom0to1)forthemarketshareofEVsversusICEVs.Withthatutilityfunction,thesharesofEVsamongallnewlight-

dutyvehiclesaregivenbyapairoflogisticfunctions,oneeachforcarsandlighttrucks.Logisticfunctions’familiarS-shapedcurvesareusefulformodelingtechnologydiffusionbecausethey

asymptoteatsharesof0and1:

EVs?arej,t=

e?aj+xtβj+ψt?

1+e?aj+xtβj+ψt?,

whereαjisacalibrationparameterforsettingthemodel’sinitialEVsharefornewvehiclesoftypej∈{car,truck}tothecurrentlyobservedvalue;Xtismatrixshorthandforthefactorspj,t,(L2t/TotEVt-1),and(L3t/HwyMiles)thatunderliethedemandforEVsinthismodel;βjisvectorshorthandforthethreecorrespondingdemand-responseparametersβp,βL2,andβL3;andΨtistheattributedriftparameter,discussedingreaterdetailinthenextsection.

TheβjparametersreflecthowsalesofnewEVsrespondtochangesinvehicleownershipcosts(includingpurchaseprice)orinthechargernetworkandhavethefollowingsigns:βp<0

(increasesinownershipcostsreducethedemandforEVs),βL2>0,andβL3>0(increasesinL2chargersperEVorinL3chargersperhighwaymileincreasethedemandforEVs).Theβjparametersarerelatedtoelasticitiesofdemandwhendemandismeasuredintermsofmarket

7AsofDecember2022,therewereabout3.4millionelectricvehiclesregisteredintheUnitedStates,includingEVs

andplug-inhybridEVs(AlternativeFuelsDataCenter,“ElectricVehicleRegistrationsbyState”).TheNationalHighwaySystemintheUnitedStatescurrentlyincludesabout224,000milesofhighway,includingabout

67,000milesofinterstatehighwaysandotherfreewaysand157,000milesofotherprincipalarterialhighways(FederalHighwayAdministration2021).

6

share(ratherthanunitsales).Forexample,thepriceelasticityofmarketsharefornewEVs,inyeart,isgivenbyβp??1?EVs?aTet?—meaningthata1percentdecreaseintherelative

ownershipcostofanEVversusanICEVwouldincreasethemarketshareofnewEVsbyβp??1?EVs?aTet?percentinyeart.8

AslongasthemarketshareofnewEVsremainssmall,thepriceelasticityofmarketsharewillapproximatelyequalβp.TheelasticitywilldeclineastheshareofnewEVsincreases.

Ultimately,astheshareofnewEVsapproaches100percent,theelasticitywillapproachzero.

ThechargernetworkelasticitiesofmarketsharefornewEVshavethesameformasthepriceelasticity:Anincreaseof1percentinthenumberofL2chargersperEV,orinL3chargersperhighwaymile,wouldincreasethemarketshareofnewEVsby{βL2OTβL3}??1?EVs?aTet?.

AppendixA

liststheparametervaluesanddatausedinthemodel,alongwiththeirsources.

Vehiclesalesareaggregatetotalsforplug-invehicles(EVsandplug-inhybridEVs)andvehiclesthatdonotplugin(ICEVsthatrunongasolineordieselandhybridEVsthatdonotplugin).

Vehicleproductioncostsandfuelcostsaregivenforrepresentativevehicles:EVsratherthanplug-inhybridEVsbecauseEVsoutsellthembyasubstantialmargin,andgasoline-powered

ICEVs(ofaveragefuelefficiency)ratherthandieselornon-plug-inhybridEVs,forthesamereason(EIA2022b).EVproductioncostsreflectestimatesforcarbatterieswithacapacityof70kilowatt-hoursandbatteriesforlighttruckswithacapacityof120kilowatt-hours,both

providingarangeof240miles.Futuredecreasesinbatteryproductioncostsaretreatedas

reducingthepricesofnewEVs.9

AppendixB

presentstheresultsofsensitivityanalysesonthemostinfluentialparametersinthemodel,includingtheβj,ttermsrelatingtotheownershipcostandnetworksizeelasticitiesofdemandfornewEVs.

2.2AttributeDrift

ThetermattributedriftreferstochangesinunspecifiedattributesofEVownership(including

attributesofthevehiclesthemselves;theiravailabilityinthenew-vehiclemarketplacerelativetoICEVs;attributesofEVchargernetworks,includingavailabilityandperformanceofvehicle

chargers;andtheinfluenceofsocialfactors)thataffectconsumers’preferencesforEVsrelativetoICEVs.IntheEVdemandequation,attributedriftattimetis

8RelativeownershipcostsofEVsandICEVsdependonpurchasepricesandthepresentvalueofexpectedlifetimecostsoffuelandmaintenance,givenanassumednumberofmilestraveledperyearasavehicleages.Purchase

pricesareestimatedasmarkupsofprojectedproductioncosts,includingthecostofEVbatteries.Theprice-andcharger-supplyelasticitiesofdemandaredrawnfromprobabilitydensitiesthatreflecttherangeandqualityofestimatesfoundintheliterature(see

AppendixA)

.

9Thetechnologicaladvancesthatthosecostdecreasesrepresentcouldinsteadbeusedtoincreasebatteryrangewhileholdingvehiclepricesfixed;eitherusewouldincreaseEVsales.

7

ψt=μ+ψt?1+ζt,

whereψ0=0,μisaconstanttimetrend,andζtisamean-zeroannualdeparturefromthattrend.

Greaterattributedrift—ahighertrendvalueμ—projectsamorerapidshifttoEVs.Inthe

model’ssimulations,trendμisrandomlydrawn,withameanvaluethatdependsonβp:ThemoreresponsiveconsumersaretochangesintherelativecostsofEVownership,themore

rapidlytheshifttoEVswouldoccurwithdecreasesinthosecosts.Bothμandζtaredrawnfromnormaldensityfunctions:

μ~N?a??βp?,b??βp??,ζt~N?0,C??μ??.

ThatfunctionalformforattributedriftfollowsColeetal.(2023),whociteArchsmithetal.(2021).

Tospecifydrift,themodelerspecifiesparametersa,b,andc,withabeingthekeyparameter

becauseitdirectlyaffectsthedemandtrend.Theparametersbandcspecifyvariancesfortrendμanddepartureζt,respectively.Ateachiterationofapolicysimulation,anewpriceelasticityisdrawnfromaspecifiedprobabilitydistribution.10Thatselectiondeterminesthevalueofβpthatwillbeusedforthatiteration.Withβpdetermined,trendμisthenselected:Onaverageitwill

haveavalueofa·?βp?,althoughwithavarianceofb·?βp?.Whenahigh(low)priceelasticityisdrawn,forthatiterationconsumerswillbemore(less)responsiveeveryyeartochangesinthe

factorsthataffectconsumers’preferencesforEVsversusICEVs.Withμdetermined,anewζtisdrawnforeachyearofeachiteration,determiningtherandom,annualdeparturefromlong-termattributedriftμ.Thatdeparturewillbezeroonaverage,withavarianceofc·?μ?.

2.3ModelCalibrationAdjustmenttoAttributeDriftTerm

Onaverage,thevalueofattributedriftψtwillchangebya·?βp?eachyear.Thus,theparameteradirectlyinfluencesprojectedsharesofnewEVs.Toprovideanempiricalbasisforthevalueofathatisusedinthemodel,IconsidertrendsinactualU.S.salesoflight-dutyEVsoverthepast

decade.

Plottingnew-EVmarketsharessince2011showsthatEVsaleshavebeenrisingatanincreasingrate(see

Figure1)

.Until2017,themarketsharefornewEVshadbeenincreasingbylessthan0.2percentagepointsperyear,onaverage.Sincethen,thenew-EVsharehasincreasedbyabout1percentagepointperyear,andithasrisenevenmorerapidlysince2020.Therelationship

betweenEVmarketshareandyeariswelldescribedbyaquadratictrendline.

10ThatdistributionreflectstherangeofestimatedEVpriceelasticitiesintherefereedstudiesCBOreliedonforthoseestimates(see

AppendixA)

.

8

Figure1.

MarketShareofNewPlug-inEVs,WithQuadraticTrendLine

Percent

25

20

15

5

0

y=0.0913x2-0.7107x+1.5144

Historical-Growthscenario

203

20112015202020250

Datasource:CongressionalBudgetOffice,usingdatafromArgonneNationalLaboratory,EnergySystemsandInfrastructureAnalysisDivision.See

/publication/58964#data.

EV=electricvehicle.

Fittingaquadraticcurvetonew-EVmarketsharessince2011yieldstheequation

EVshare=1.514–0.711*years+0.091*years2.

Thetrendl