An agent-based modeling approach to predict the evolution of market share of electric vehicles

ContentslistsavailableatSciVerseScienceDirect

TechnologicalForecasting&SocialChange

Anagent-basedmodelingapproachtopredicttheevolutionofmarketshareofelectricvehicles:AcasestudyfromIceland

EhsanShafieia,⁎,HedinnThorkelssona,EyjólfurIngiÁsgeirssona,BrynhildurDavidsdottirb,MarcoRabertoc,HlynurStefanssonaabcSchoolofScienceandEngineering,ReykjavikUniversity,Reykjavik,Iceland

EnvironmentandNaturalResources,SchoolofEngineeringandNaturalSciences,UniversityofIceland,Reykjavik,IcelandBusinessandManagementEngineering,FacultyofEngineering,UniversityofGenoa,Genoa,Italy

articleinfoabstract

Inthispaper,anagent-basedmodelisdevelopedtostudythemarketshareevolutionofpassengervehiclesinIceland,acountryrichindomesticrenewableenergy.Themodeltakesintoaccountinternalcombustionengine(ICE)vehiclesthatarecurrentlydominantinthemarketandelectricvehicles(EVs)thatarelikelytoenterthemarketinthefuture.Thevehiclescompeteformarketpenetrationthroughavehiclechoicealgorithmthataccountsforsocialinfluencesandconsumers'attractivenessforvehicleattributes.Themainresultprovidedbythemodelingapproachisthemarketshareofdifferentvehiclesduringthetimeperiod2012–2030.Theeffectsoffuelprices,vehicletaxes,futurepriceofEVsandrechargingconcernsonthemarketshareareanalyzedwiththehelpofthemodel.TheresultsshowthatEVswouldseizethemarketcompletelyinthescenariocombinedofhighgasolineprice,decreasingEVpricewithouttaxandnoworryabouttherechargingofEVs.ThesuccessfulpenetrationofEVsinthescenarioswithlowgasolinepriceandcombinationofmediumgasolinepriceandconstantEVpriceneedspolicysupportsliketaxexemptionandanenvironmentwhereconsumersdonotsufferfromrangeanxiety.

©2012ElsevierInc.Allrightsreserved.

Articlehistory:

Received2November2011

Receivedinrevisedform10May2012Accepted22May2012

Availableonline15June2012Keywords:

Agent-basedmodelVehiclechoice

MarketpenetrationElectricvehiclesIceland

1.Introduction

Afterthealmosttotaldisappearanceofelectricvehicles(EVs)fromthemarketintheearly20thcentury,EVshavebeguntoresurfaceinthe1990s[1].Comparedtointernalcombustionengine(ICE)vehicles,EVtechnologieshavethepotentialtoreducetheconsumptionofpetroleumproductsandCO2emissions[2].Inrecentyears,policy-makershavebeeninterestedinadoptionpatternsofEVtechnologyandindeterminingpoliciestoinfluencethevehiclemarket.ItisinlightofEVs'potentialadvantagesthatquestionsriseashowthevehiclefleetcouldevolvefrombeingentirelycomposedofthetraditionalICEvehiclestoamixedfleetofICEbasedvehiclesandEVsinthelong-run.Furthermore,itisofinteresttoknowhowtheimplementationprocesscanbeaffectedbysocial,technologicalandeconomicincentives.Suchquestionscanbedifficulttoanswerandthereisaneedtoprovideanalyticaltoolsformarketsimulationtoevaluatetheeffectivenessofdecisionsunderdifferentscenarios.

Duringthepastdecade,themethodologyofagent-basedmodels(ABM)hasbeenincreasinglyemployedtosimulateindividuals'behaviorandvehiclepurchasedecisions(see[3,4]foranoverview).ABMshavebeenusedtoconsidertheinterdependenciesamongthekeyagentsofvehiclemarketsuchasmanufacturers,consumers,andgovernmentalagencies.Somestudiescombinemanufacturerperspectivewithpurchasingbehaviorofconsumersinordertosimulatethemarketpenetrationofalternativefuelvehicles[e.g.4–6].Otherstudiesfocusonmodelingconsumervehiclechoice[e.g.7,8].

⁎Correspondingauthor.Tel.:+35996200;fax:+35996201.E-mailaddresses:ehsan@ru.is,ehsh90@gmail.com(E.Shafiei).0040-1625/$–seefrontmatter©2012ElsevierInc.Allrightsreserved.doi:10.1016/j.techfore.2012.05.011

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–16531639

Mostmodelsintroducedintheliteratureforvehiclechoicebehaviorarebasedonutilitytheory,whichassumesthatconsumerspurchasethevehiclesinthechoicesetwiththehighestutility[9,10].Theutilityofadoptingandusingatechnologyconsistsofanindividualpartandasocialeffectpart[11].Thestudiesthathaveemployedthisapproachformodelinginnovationdiffusionassumethatconsumeragentsdecideaccordingtoasimpleweightedutilityofindividualpreferenceandsocialinfluence[11–14].Inthesemodels,agentsadopttheproductswhentheutilityishigherthanaminimumthreshold.Theotherapproachhasbeenproposedinwhichindividualdecisionruleisderivedbasedontheprobabilityofvehiclepurchase[15].Todeterminetheprobabilityofvehicleadoption,MultinomialLogitModels(MNL)havebeenwidelyused[7,9,10,16,17].IntheMNLframework,consumersweighthevehicleattributessuchasprice,fuelconsumption,size,acceleration,safety,etc.,whendeterminingwhichvehicletopurchase.Theresultsofthesemodelingapproachesprovidethenumberandtypeofvehiclesinthemarket.

Inthispaper,anagent-basedmodelisdevelopedtostudythemarketshareevolutionofpassengervehiclesinIceland,whereover99%oftheelectricityisproducedfromrenewablelow-carbonenergysources,suchasgeothermalandhydropower.Whileheavyindustryisprevalentinusageoftheseenergyresourcesandthegeneralpopulationenjoysrelativelylowelectricitypricescomparedtotherestoftheworld,Icelandremainsmoreorlessentirelydependentuponforeigngasolineimportstosustainitstransportationneeds.Energycompanies,carmanufacturersanddealers,governmentandconsumerscouldbeconsideredasthemostimportantpotentialagentsthatcaninfluencethetransportationsystem.Inthispaperwefocusonconsumerbehaviors,asthereisnocarmanufacturerinIceland,andtheeffectsofotheragentsarestudiedexogenously.Fordeterminingthepersonalpreferencesofconsumers,weuseanintegratedformofMNL,whichincludestheutilityofconsumersandtheeffectsofsocialexposures.ThedecisionruleisemployedinanalgorithmtodeterminetheshareofEVsduringaplanningperiod.ThepaperseekstoanswercertainimportantquestionsregardingtheprocessofEVimplementationinIceland:HowlargeisthehypotheticalportionofEVswithinthepassengercarfleetaftertwodecades?Howwillenergypricesandtechnologicalimprovementinfluenceconsumerslookingforanewcar?HowwillconsumersreactwhentaxespresentlyleviedonimportedICEisappliedtotheimportedpriceofEVs?

Inordertoanswersuchquestions,thepaperisorganizedasfollows.ModelstructureandthemethodologyarepresentedinSection2.Then,themainassumptionsanddatasourcesforapplicationoftheproposedmodelinIcelandcasestudyaredescribedinSection3.ScenarioassumptionsarepresentedinSection4.Section5isdevotedtotheresultsanddiscussion.Finally,conclusionsandsomeprospectsforfutureresearchareprovidedinSection6.2.Modeldescription:theoreticalbackgroundandmethodology

Weuseagent-basedmethodologytodevelopamodelforstudyingtheconsumerbehaviorsandmarketshareevolutionofpassengervehicles.ThemodeltakesintoaccountICEvehiclesthatarecurrentlydominantinthemarketandEVsthatarelikelytoenterthelightdutymarketinthefuture.Setofvehiclescompeteformarketpenetrationthroughavehiclechoicealgorithmthataccountsforsocialinfluencesandconsumers'attractivenessforvehicleattributessuchasprice,fuelconsumption,length,accelerationandluggagecapacity.Consumersareheterogeneouswithrespecttosocio-demographicattributes,theirpreferencesanddecisionsonvehiclepurchase.Theypurchasethevehiclesbasedonboththeirownpreferencesandimitatingthebehaviorofotherconsumers.Thecoreoftheagent-basedmodelhereisconsumerchoiceprobability,whichisdescribedinthefollowingsection.2.1.Modelingconsumerchoiceprobability

ConsumerbehaviorisheredescribedonthebasisofMultinomialLogit(MNL)model.StandardMNLsarewidelyusedtodetermineindividualpreferenceprobabilityandadecisionruleforselectingthevehiclefromthechoiceset[18].Inthisframework,differentvehiclesaredistinguishedbyvariousattributesandtheconsumersmakechoicesamongthemsoastomaximizetheirutility.Anyvehicleisthereforeidentifiablebyasetofattributes,whichinturneitherrepelorattractprospectivepurchaserstothevehicleindifferentways.

ThemainlimitationofStandardMNLmodelisthatdecisionsolelydependsonattributesofvehiclesandpersonalpreferencesofconsumers.However,thedecisiontoadoptnewtechnologiesdependsontheprocessofword-ofmouth,socialinfluencesandimitation.Inotherwords,adoptionofnewvehiclescanbestimulatedbythebehaviorofotheragents,especiallybyrecommendationoffriendsandneighbors.Theagentsaremorelikelytopurchaseaparticularvehicle,ifaconsiderablenumberofotheragentshavealreadyadoptedit.Thismeansthatateachtimestep,eachagentlooksatotheragentsandthenitdecidestoevaluateavehicleforpurchasing.

Following[19],weuseanintegratedformofMNL,whichincludestheutilityofconsumersandtheeffectsofsocialexposures.Inthismodel,the“willingnesstoconsider”avehicleisafunctionofsocialinfluencesandmarketconditionsandtheattractivenessofeachvehicleintheconsiderationsetisafunctionofvehicles'attributes.ThemathematicalformulationoftheintegratedformofMNLmodelispresentedinEq.(1):

󰀂A󰀃P

Wk;j;t:expβi;a;tÂXa;j;t

a¼1

Pi;j;t¼ð1Þ󰀂A󰀃VPPWk;j;t:expβi;a;tÂXa;j;t

j¼1

a¼1

where:Pi,j,tβi,a,tProbabilityofpurchasingvehiclejbyagentiattimetPreferencesofconsumeriforattributeaattimet

10E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Xa,j,tWk,j,tVAValueofattributeaforvehiclejattimet

WillingnesstoconsidervehiclejbydriversofvehicletypekattimetNumberofvehicles

Numberofattributesforthevehicles

TheconsumersmustbesufficientlyfamiliarwithEVstoincludethemintheirconsiderationset[19].Tocapturetheformationofconsumers'considerationset,[19]introducestheconceptofconsumer'swillingnesstoconsideravehicle.Fortheaggregatepopulation,averagewillingnesstoconsidervariesovertheinterval[0,1].InEq.(1)itisassumedthateveryconsumerconsidersICE(i.e.WICE,ICE=WEV,ICE=1).Moreover,theagentswhohavealreadyadoptedEVsincludeEVsintheirconsiderationset(WEV,EV=1).However,WICE,EVincreasesinresponsetosocialexposure,andalsodecaysovertime.Accordingto[19]totalexposuretoEVsincreasesthroughthreechannels:

1-Marketingandadvertisements.

2-DirectsocialexposureandwordofmouthcreatedbycontactswiththeagentswhohavealreadyadoptedEVs.3-IndirectwordofmoutharisingfromcontactswithICEdrivers.

WillingnesstoconsidermaydecaybecauseofdiminishingsocialexposuretoEVs.InsufficientorinfrequentsocialexposureforEVsmaycausetheconsumerstoforgetaboutEVs.Itleadstolossofconsumers'familiaritywithEVsand,thus,lossofwillingnesstoconsiderthemovertime.Forthedetailedformulationofwillingness-to-considerandtheassociatedparameterssee[19].

Theconsumer'spreferencesinEq.(1)arethemeasurementofhowattractedoraverseanindividualistowardcertainattributes.Asanexample,anagentwithlowincomecanbeassumedtobeaversetohighpricedcars.Thereforehisorherpreferencestowardthepriceofacarwouldberelativelylow,whereasanotheragentwithhigherincomemightbelessconcernedabouthighcarpricesandhispreferencestowardcarpricesthereforehigher.

AvailabilityofchargingfacilitiesandtherelativelylowrangeofEVsarethemajorconcernsforconsumersconsideringEVs.Eq.(1)assumesnodifferencesbetweenthetwotechnologiesofEVsandICE,inwhichtheconvenienceofICErefuelinginpopulargasstationsisjustasconvenientasrechargingEVswithlimitedrange.Therefore,weemployamultiplicativeformofrefuelingeffectproposedby[5]1tomodifytheamountofprobabilitiesinEq.(1).Accordingto[5],rechargingeffectishereconstructedasafunctionofsocialacceptabilityofchargingfacilitiesandindividualdrivingpatterns:

󰀄

RFEi;j;t¼

where:RFEi,j,tDPi,tStα>0

RefuelingeffectforconsumeriusingvehiclejattimetDrivingpatternsofconsumeriattimet

AvailabilityandsocialacceptabilityofchargingfacilitiesattimetScalingparameterusedtocalibratethemodel

1

1−DPi;tÂe−α:St

j¼ICEj¼EV

ð2Þ

TheprobabilitiesestimatedinEq.(1)arescaledbyrefuelingeffectfactorstoestimatethemodifiedvaluesasfollows:

󰀁Pi;j;t¼

Pi;j;tÂRFEi;j;t∑Pi;j;tÂRFEi;j;t

j

ð3Þ

2.2.Vehiclechoicealgorithm

Differentvehiclescompeteformarketpenetrationthroughavehiclechoicealgorithmthataccountsforsocialinfluencesand

consumers'attractivenessforvehicleattributessuchasprice,fuelconsumption,length,accelerationandluggagecapacity.Therearetwotypesofprobabilitieswithinthemodel.Oneistheprobabilityofwhetherornotanagentbuysacar;theotheristheprobabilityofwhichcaranagentactuallyendsupbuying.Toestimatetheseprobabilities,thealgorithmstartsfromthecalculationofconsumerchoiceprobabilityusingEq.(3).Then,theprobabilitiesareaggregatedintoacumulativedistributionfunctionofvehiclepurchases.Thefunctionisthesumofprobabilitiesforbuyingallofthevehiclesuptoandincludingaparticularvehiclej:

Qi;j;t¼

jXh¼1

󰀁Pi;h;t

ð4Þ

InSchwoon(2006)[5]anormalizedutility,asaweightedaverageofthedirectutilityassociatedwiththecharacteristicsofthecarandthemarketshare,is

scaledbyavariablecalledrefuelingeffect.

1E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–165311

Notethattheorderinwhichtheaggregationoccursisimportant.Withtheagent'stotalpopulation,alooprunsthroughallagentsanddetermineswhethertheypurchaseavehicleornot.Whetherornotanagentpurchasesavehicleduringaniterationisdeterminedbytwocriteria.Thefirstoneiscomparingthelifeofcurrent(in-use)vehiclewithastochasticlifetimeforthisvehicledeterminedbyanormaldistributionfunction.Ifthecurrentageofthevehicleisgreaterthanitslifetime,thenconsumerdecidestopurchaseanewvehicle.Otherwise,arandomvariablewithauniformdistributionisgeneratedandthenitiscomparedtothesocialgroup'slikelihoodofpurchasingavehicle.Whentheageofvehicleisbelowitslifetime,forsimplicity,itisassumedthattheconsumers'tendencytobuynewvehicleisnotinfluencedbythevehicleageanditisonlydeterminedbysocialgroupcharacteristicsintermsofhowoftendifferentconsumerspurchaseanewvehicle.

Shouldanagentdecidetopurchaseaspecifictypeofvehicle,aMonte-Carlosimulationisprocessed.ArandomuniformvariableisgeneratedagainanditiscomparedtothecumulativeprobabilityfunctionQi,j,tinEq.(4).TheinputofEq.(4)thatyieldstheclosestbuthigheroutcomethanthevalueoftheuniformvariableequalsthetypeofvehiclethatispurchased.

Whenpurchasingavehicle,anagent'spurchasehistoryismarkedwiththevehicletype.Foreachiteration,anagent'shistoryofcarpurchasesdevelops.Whenaniterationendsdataisextractedfromallagentsonvehicleownership,formingtheoverallvehiclefleetcomposition.Frominformationonvehiclefleetcomposition,thenumberorproportionofEVswithinthefleetcanbededucted.Fig.1showstheentireprocessofoneloop.Eachlooprepresentsoneyearsointhatsensepurchasesareassumedtohappenyearly.

SincethevehiclefleetisinitiallysupposedtobecomposedentirelyofICEbasedvehicles,aspecialiterationofthealgorithminFig.1isrunbeforevehiclepurchasingwouldnormallyoccur.Theiterationisthesameasusualexceptfortwodeviations:everyagentisassumedtopurchaseavehicleandEVsareexcludedfromthechoiceset.TheendresultisarandomlydistributedICEbasedvehiclefleetintermsofvehicletypeinthebaseyear.Attheendofthebaseyearallagentshavevehicles.TheloopisiteratedandagentsnowhavethepossibilitytobuyEVsaftertheinstillationprocess.Shouldanagentchoosetobuyanewvehiclehisoldvehiclewillvanishfromthesystemand,thus,atalltimestepsthenumberofvehiclesinthefleetstaysthesame.Thereasonforsuchsimplificationisthatdataonusedvehicleswouldbefarmorecomplicated,aswellasthepreferencesofconsumersduetothenaturethatpricesforsecondhandvehiclesarenotasstandardasnewvehicles.

Theproposedalgorithmtakesintoaccountthetimedependenceontheevolutionofthevehiclefleet.WhileothersimilarstudiesthatsimulatecarpurchaseswiththehelpofMNLhaveonlyiteratedpurchasesonce,thismodeliteratespurchasesmultipletimesovertime.Therefore,thehistoryofpreviousvehiclesisloggedandthus,anydecisionfromthenextpurchasecouldtakeintoaccountpreviousownership.Moreover,theproposedmodelconsiderslifetimeofvehiclesaswellasconsumers'tendencytoreplacetheirvehiclesbeforeendoftheirlifetime.

Fig.1.Algorithmforanagent'svehiclepurchasingprocess.

12E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Table1

Vehiclecategoriesandsomeidentifyingfeatures.CategoryBudgetFamilyStationSUV

Sportcar

VehiclebodystyleHatchbacks

Sedans,liftbacksStationwagonsSUVs

Convertibles,cabriolets

VehicledimensionsSmallMediumLongMediumMedium

VehicleweightLowMediumMediumHighMedium

InitialcostLow

Low–mediumMedium

Medium–highHigh

3.Descriptionofthecasestudy

Theproposedapproachforagent-basedmodelingofvehicleadoptionisappliedforsimulatingIcelandiclight-dutyvehiclefleet.Thebaseyearis2011andthetimehorizonofthestudyforanalysisoftheresultsis19years,beginningin2012andcontinuinguntil2030.Definitionofagents,mainassumptionsandthedatausedinthemodelaredescribedinthefollowing.3.1.Vehiclechoiceset

Themodelassumesthatanagenthasachoicefromalimitedsetofvehiclesreferredtoasthechoiceset.Thesimulatedvehiclefleetisassumedtobedividableintovariouscategories:budget,family,station,SUVandsports.Eachcarcategoryhasoneormorerepresentativesforacarpurchasertochoosefrom;oneormoreICEbasedvehiclesandpossiblyanEV.Insomecases,householdsmaybeinterestedinaparticularcategorythathasonlyoneoptionofanICEbasedvehicle.ThishastodowiththelackofEVsavailabletorepresentallvehiclecategories.Eachcategoryhasvariousidentifyingfeatures,whichareshowninTable1.

ThecategoriesinTable1servemainlytonotonlyassurethatsuitablevehiclecandidatesareconsideredinthemodel,butalsoallowforin-depthanalysisbasedonaparticularcategory.AllcategoriespresentedinTable1haveatleastoneICEbasedrepresentative.Aprospectivevehiclebuyingagentwithinthemodelhasachoiceof8cars,5ofwhichareICEbasedcarsand3thatareEVs.Thecars,alongwiththemanufacturer,enginetypeandcategoriesarepresentedinTable2.

FromTable2itcanbeseenthattherearenoEVcounterpartspresentinthemodelforstationwagonsandSUVs,astheauthorsdidnotfindsuitabledataforthem.3.2.Vehicles'attributes

Accordingtoasocialstudyonenvironmentalissuesperformedin2002byCOWI(aconsultingcompanyinDenmark)[20],themostimportantvehicleattributesforanyhouseholdarethepurchaseprice,fuelconsumption,length,luggagecapacity,acceleration,lowermediumanduppermedium.Thelasttwofeatureswarrantfurthermentioning.Thecoefficientvaluesrelatedtoloweranduppermediumsignifyanagent'sattraction(oraversion)tovehiclesthatarelessormoreexpensivethanwhattheagentfeelsshouldbeafairprice.Asanexample,assumetwoagentswithdifferingpreferencesrelatedtoavehicle'supper-mediumfeature.Oneagentwouldthereforebemoreaversetoavehiclepricedhigherthanthenormthantheother(i.e.he/shewouldfeelcheated),whereasanotheragent'sattractiontotheuppermediumfeatureofavehiclecanbeattributedtoasenseoffullnessandsatisfactionachievedwhenbuyinganoverpricedvehicle.ThevehiclesofTable2alongwiththeirrespectivevehicleattributesarepresentedinTable3.

UnitsoffuelconsumptioninTable3varydependingontheenginetypeofavehicle.FuelconsumptionofICEbasedvehiclesismeasuredinlitersper100kmdriven,whereasfuelconsumptionofEVsismeasuredinkWhper100kmdriven.Allpricesarebasedonmanufacturersuggestedretailprices(MSRP)inUSD,convertedtoEURbasedonanexchangerateof1.34EUR/USD.TheseattributesareusedtocreateXa,j,tinEq.(1),wherefollowingtheCOWIstudy[20],thelogarithmistakenofallattributesexceptforupperandlowermediumsandalsofuelconsumptionwhichitsmonetaryvalueisdirectlyusedinthemodel.

Vehicles'lifetimeisanotherattributethatdirectlyaffectsthevehiclepurchasedecisioninthemodel.Basedon[23]weuseanormaldistributionfunctionN(15.9,4.2)withthemeanof15.9yearsandastandarddeviationof4.2years.

Table2

Vehiclesusedinthemodel.ManufacturerMitsubishiNissanTesla

VolkswagenSubaruToyotaMitsubishiPorsche

Namei-MievLeaf

RoadsterGolfOutbackYaris

Outlander911Carrera

EngineTypeEVEVEVICEICEICEICEICE

CategoryBudgetFamilySportsFamilyStationBudgetSUVSports

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Table3

Vehicleattributesusedinthemodel[21,22].Vehiclenamei-MievLeaf

RoadsterGolfOutbackYaris

Outlander911Carrera

FueltypeElectricityElectricityElectricityGasolineGasolineGasolineGasolineGasoline

Purchaseprice(USD)29,99033,720109,00019,75523,19513,71521,99577,800

Fuelconsumption

(literorkWhper100km)1921.217.78.710.667.39.411.2

Length(mm)33944539394201478043004665

Uppermedium00100001

Lowermedium00000100

Luggagecapacity(mm3)3683301414259923681025142

13

Acceleration(seconds)9.19.13.98.69.78.77.44.7

3.3.Syntheticpopulation

Ideallyeveryagentwouldhaveitsownuniquepreferencetowardvehicleattributes.WithasyntheticpopulationcomposedofeveryhouseholdinIceland,however,rationalizingpreferencesassignedtoeachagentfromendlesspossibilitiesisoutofthescopeofthisarticle.Instead,anassumptionismadethateachagent'spreferencesarecorrelatedcloselyenoughwithaparticularsocialdemographicgroup(hereaftersocialgroup)towarrantthathisorherpreferencesbeconsideredthesameasthoseofsaidgroup.Anagentisclassifiedasamemberofasocialgroupbasedonsex,respectiveincomelevel(low,mediumorhigh)orfamilystatus(single,inarelationship,withchildrenorlivingwithparents).Eachsocialgrouphasitsdistinctivepreferencetowardvehicles'attributes.Forexample,agroupwithlowincomeismuchmoresensitivetowardboththeinitialpurchaseandrefuelingcostsofavehicle,whereasanothergroupwithhigherincomeisnotoverlyconcerned.

Householdsareclassifiedintogroupsbasedon8differentsocialstandingsand3incomelevelsaspertheCOWIreport[20].Atotalof24socialgroupsandtheirdescriptionsarepresentedinTable4.

IncomelevelsinIcelandfortheyear2008areobtainedfromtheCreditInforeport[24]andarelistedinTable5.

AccordingtotheCreditInforeportthereareatotalof175,677householdspresentinIceland[24].ThenumbersofhouseholdsbelongingtoparticularsocialgroupswithcertainincomelevelsareasseeninTable6.

Table4

Socialdemographicgroupsusedinthemodel.

Female

LowincomeMediumincomeHighincome

Singlefemale

FemalewithparentsSinglefemale

FemalewithparentsSinglefemale

Femalewithparents

Male

Singlemale

MalewithparentsSinglemale

MalewithparentsSinglemale

Malewithparents

FemaleandmaleFemaleFemaleFemaleFemaleFemaleFemale

withsignificantwithsignificantwithsignificantwithsignificantwithsignificantwithsignificant

other

otherandchildrenother

otherandchildrenother

otherandchildren

MaleandfemaleMaleMaleMaleMaleMaleMale

withwithwithwithwithwith

significantsignificantsignificantsignificantsignificantsignificant

other

otherandchildrenother

otherandchildrenother

otherandchildren

Table5

IncomelevelsinIceland2008.SocialstatusSingle

Inrelationship

Lowincome

[1000ISK/month]0–2490–9

Mediumincome[1000ISK/month]250–9550–799

Highincome

[1000ISK/month]550+800+

Table6

DistributionofIcelandichouseholdsintolimitedsocialgroups,2008[24].Socialgroup

SinglefemaleSinglemale

Femalelivingw.parentsMalelivingw.parents

Couplew/ochildren(femalebuyer)Couplew.children(femalebuyer)Couplew/ochildren(malebuyer)Couplew.children(malebuyer)Grandtotal

Lowincome41,41139,0

Mediumincome10,62315,284

HighIncome9993049

Total53,03357,873

11,484730513,47217,4786021901130,97733,794

175,677

14E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Table7

DistributionofIcelandichouseholdsintoallsocialgroups,2008.Socialgroup

SinglefemaleSinglemale

Femalelivingw.parentsMalelivingw.parents

Couplew/ochildren(femalebuyer)Couplew.children(femalebuyer)Couplew/ochildren(malebuyer)Couplew.children(malebuyer)Grandtotal

Lowincome20,70519,77020,70619,770574236525742365356.7%

Mediumincome531172531272673687396736873932.4%

Highincome49915245001525301045053011450610.9%

Total26,51328,93826,51828,93715,48816,815,4168175677

Weight15.1%16.5%15.1%16.5%8.8%9.6%8.8%9.6%100%

WithnoinformationavailablewithintheCreditInforeportontheratioofsinglesthatlivewiththeirparents,itisassumedthat50%ofallsinglesarelivingwiththeirparents.Thereisneitheranydataavailableonwhichpersonwithinaparticularhouseholddecidesuponvehiclepurchases,requiringanotherassumptionthatvehiclepurchasingdecisionsinIcelandichouseholdsaresplitevenlybetweenthesexes.AmendingTable6withtheaforementionedassumptionresultsinTable7,containingthesplitofhouseholdsintoall24socialgroupsoriginallyintroducedinTable4.3.4.Vehiclepurchaseprobability

Apartfromvehiclelifetime,howoftenanagentpurchasesavehicleisdeterminedbyitssocialgroupandincome.AccordingtoCOWIreport,4%ofDanishhouseholdspurchasedavehicleintheyear1997[20].ConsideringthefactthatcurrentIcelandicpurchasinghabitsmaydiffer,anassumptionismadethatallvehiclesimportedtoIcelandinacertainyearrepresentvehiclepurchasinghouseholds.ImportedvehiclesinIcelandintheyears2007,2008and2009are18,876,10,703and2550respectively[25].Thenumberofimportedvehiclesvariesabnormallyduetoasharpturnfromeconomicboom(2007)torecession(2009).Therefore,anaverageimportvalueofthethreeyearsisusedtorepresenttheaveragenumberofimportedpassengervehicles.Assumingthattheaveragenumberofimportedpassengervehicles(10,710vehicles)representsthenumberofvehiclepurchasinghouseholdsinIceland,thepercentageofvehiclepurchasesperhouseholdinIcelandis6%.Therefore,Icelandersareassumedtopurchase50%morevehiclesperhouseholdthanDanesand,thus,theprobabilitythatanagentbelongingtoaparticularsocialgroupbuysavehicle,aspresentedintheCOWIreport[20],isscaledupbyafactorof1.5.TheresultingprobabilitiesofvehiclepurchasedependingonsocialgrouparepresentedinTable8.3.5.Consumers'preferences

AnagentintendingtobuyavehiclehastomakeachoicebetweenvehiclesofthechoicesetpresentedinTable3.Howandwhyanagentchoosesavehicleisbaseduponaparticularvehicle'sattributesandtheagent'sweightedpreferencestowardsaidattributes.EachsocialgroupisdefinedbycertainweightsofpreferencecalledconsumerpreferencesinEq.(1).DataonsocialgrouppreferencesintermsofvehicleattributesforIcelandichouseholdsdoesnotexist,butthesubjectwasanalyzed,processedandpublishedinDenmarkbyCOWI[20].DuetoculturaltiesbetweenIcelandandDenmarkastwoScandinaviancountries,thereisenoughsimilaritybetweenthemtojustifytheuseofDanishsocialstudydataunderIcelandiccircumstances.Therefore,wesimplyassumethatIcelandersbehaveasDanes.Thevaluesforeachofthe24socialgroupscanbeseeninTable9forlowincomegroups,Table10formediumandTable11forhighincomesocialgroups.

TheestimatedcoefficientsinTables9–11showthepreferenceweightstowardvariousvehicleattributes.ByreplacingtheseparametersintheconsumerutilityfunctioninEq.(1),theprobabilitiesofvehiclepurchasebyeachconsumercanbecalculated.Eachsocialgroupisdefinedbycertainweightofpreferences.Ascanbeseenfromthepreferencevaluesinthesetables,theconsumerswithlowerincomearemoreconcernedaboutbothvehiclepricesandfuelcostastheabsolutevaluesoftheir

Table8

Scaledsocialgroupprobabilitiesofvehiclepurchase.Socialgroups

ProbabilityofvehiclepurchaseLowincome

SinglefemaleSinglemale

Femalelivingw.parentsMalelivingwithparents

Couplewithoutchildren(femalebuyer)Couplewithoutchildren(malebuyer)Couplewithchildren(femalebuyer)Couplewithchildren(malebuyer)

0.90%1.50%1.50%4.95%2.55%13.50%3.15%8.40%

Mediumincome2.85%4.50%9.30%8.40%6.90%19.65%6.60%16.80%

Highincome11.40%24.30%18.75%33.75%13.35%28.50%13.80%27.90%

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Table9

Preferencevaluesofsocialgroupswithlowincome[20].Socialgroups

SinglefemaleSinglemale

FemalewithparentsMalewithparents

Femalewithsignificantother

FemalewithsignificantotherandchildrenMalewithsignificantother

Malewithsignificantotherandchildren

Log(price)−3.681−3.345−3.123−2.918−4.601−4.312−4.257−3.918

€/100km−0.500−0.218−0.502−0.253−0.313−0.411−0.394−0.346

Log

(length)0.6692.832

−0.253−0.204−0.666−0.287−0.656Lowermedium

Uppermedium−0.119

−4.494

5.2819.84.1851.177

−0.247−0.428−0.253−0.273

Log

(luggagecapacity)

15

−Log

(acceleration)2.0172.4432.5094.4432.1852.3422.4262.220

3.0303.065

Table10

Preferencevaluesofsocialgroupswithmediumincome[20].Socialgroups

SinglefemaleSinglemale

FemalewithparentsMalewithparents

Femalewithsignificantother

FemalewithsignificantotherandchildrenMalewithsignificantother

Malewithsignificantotherandchildren

Log(price)−4.156−3.148−3.014−2.855−3.200−3.887−3.249−3.4

€/100km−0.397−0.286−0.381−0.333−0.447−0.383−0.442−0.413

Log(length)1.5074.2591.7825.4086.6819.3981.1691.417

Lowermedium−0.319

Uppermedium−0.318−0.159−0.229−0.1−0.412−0.280−0.385

Log

(luggagecapacity)−1.3712.0172.047−1.0983.0381.4723.538

−Log

(acceleration)2.0333.0911.7963.4652.5312.1863.3042.596

−0.181−0.322−0.144−0.1

Table11

Preferencevaluesofsocialgroupswithhighincome[20].Socialgroups

SinglefemaleSinglemale

FemalewithparentsMalewithparents

Femalewithsignificantother

FemalewithsignificantotherandchildrenMalewithsignificantother

Malewithsignificantotherandchildren

Log(price)−2.250−1.047−4.796−2.149−1.012−1.810−1.221−1.362

€/100km0.000−0.293−0.234−0.277−0.443−0.288−0.334−0.269

Log(length)

Lowermedium

Uppermedium

Log

(luggagecapacity)2.8502.443−5.158

−Log

(acceleration)4.7852.4386.1243.1712.7722.5883.5072.414

7.5674.8611.3420.3388.7298.149

−0.680−0.473−0.270−0.704−0.728−0.832

−0.457−0.430−0.904

preferencestowardthesaidattributesarehigher.Theemptycellsindicatenosignificantpreferencestowardthecorrespondingattributes.

3.6.Rechargingandrangeeffects

AvailabilityofrechargingstationsforEVscanreducetheconsumerworriesassociatedwithrangeanxiety.ForEVs,itisassumedthatthepervasivenessandsocialacceptabilityofchargingfacilitiesarelinearlycorrelatedtotheproportionofconsumerswhohaveadoptedEVs.Therefore,marketshareofEVsineachperiodisusedasaproxyforStinEq.(2).InEq.(2),thescalingparameterαissetto3asthecentralcasevaluesuggestedby[5].

Basedontheamountofkilometersavehicleisdrivenperyear,[5]estimatesanaveragevalueof0.49fordrivingpatterns(DPi,t)inGermany.ByadjustingthisvaluewithaverageannualkilometersofavehicleinIceland,thevalueof0.78isapproximatedtoreflecttheaggregationlevelofIceland.

4.Scenarioassumptions

Gasolineprice,EVpurchasepriceandimporttaxonvehiclesarethemostimportantexogenousfactorsthatdrivethemodelbehavior.Therefore,tochecktheextenttowhichthemarketshareofEVsischanged,variousscenarioscanbedefinedbasedonaforementionedaspects.

16E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Fig.2.ScenariosforthepriceofEVs.

4.1.Gasolineprice

GasolinepriceisforecastedbasedonthethreescenariosforthefuturepriceofcrudeoildevelopedbytheU.S.EnergyInformationAdministration[26].Todoso,theaveragegrowthrateofcrudeoilpriceineachscenario(inconstantpriceof2009)isappliedtotheimportedpriceofgasolinetoIceland.Astheresult,wehave3scenariosforgasolinepriceasfollows:1-Lowgasolineprice(annualgrowthrate:−1%)2-Mediumgasolineprice(annualgrowthrate:3%)3-Highgasolineprice(annualgrowthrate:6%)

Theconsumerpriceofgasolineisassumedtobe205ISK/linthebaseyear.SinceIcelandisacountryrichingeothermalandhydroelectricrenewableenergyresources,itisreasonablyassumedthatelectricitypriceremainsconstantduringthestudyperiodas9.85ISK/kWh[27].4.2.EVprice

TwoscenariosaredefinedforthepriceofEVsasillustratedinFig.2.Forthefirstcase,EVpricesareassumedtostaythesamethroughoutthetimeperiodconsidered.Therefore,EVsarepresentedasbeingalwaysmoreexpensivethanICEbasedvehicles.Inthesecondcase,EVswilllowerrapidlyinpriceastheresultofmassproductionandtechnologicallearning.Thedecreasingratesinthisscenarioarebasedonthedatain[28].4.3.TaxonimportedEVs

Twocasesareimplementedregardingtaxonimportedvehicles:

1-Equaltaxation:equalityintaxationforbothICEbasedvehiclesandEVsbyaddingexcisedutyof45%toimportedvehicles.2-AddedincentivesforEVs:addinganincentiveforEVsbyremovingexcisedutyfromimportedEVs.5.Resultsanddiscussions

ThissectionpresentstheevolutionofmarketshareofEVsasthemostimportantresultsofusingthemodel.Combinationofthreecasesforgasolineprice(seeSection4.1)andtwocasesforEVprice(seeSection4.2)makessixdifferentscenarios.Fig.3showshowthesescenarioswouldaffectthepenetrationrateandmarketshareofEVsintotalfleet.Presentationoftheresultsfortheadditional10-yearperiodfrom2031to2040inFig.3isjustforillustratingthedynamicsandpatternofmarketsharedevelopment,assumingnofurtherchangesinthevalueofscenarioparametersthroughoutthistimeperiod.

Fig.3showsthecasesthatthereisnoexcisetaxaccountedforEVs.ICEbasedvehicleshaveanexcisetaxbetween30%and45%dependingonenginesize.Asshowninthisfigure,theproportionofEVsinthescenariowithhighgasolinepriceanddecreasingEVpricerisesquicklyonaccountoflowerrefuelingcostandcompetitivepurchasepriceofEVs.ThisscenarioisthebestcaseforEVs,whichthepercentageofEVswithinthetotalvehiclefleetatyear2030is93%.Asshouldbeexpected,thescenariooflowgasolinepricealongwithconstantEVpriceleadstothelowestpenetrationfromthebeginning,endingupwiththelowestmarketshareofEVs(i.e.48%in2030).

AccordingtoFig.3,anS-shapedgrowthpatternisobservedforthemarketshareofEVsinallscenarios.DominationofICEinthebeginningofthestudyperiodleadstoalowlevelofwillingnesstoconsiderEVsandunfavorablerechargingeffect.Asaresult,despiteEVs'competitivenessincostandenergyconsumption,theirmarketsharehasaslowgrowthintheearlystages.Lowerenergypriceand,tosomeextent,competitivepurchasepriceofEVsduetotaxexemptiongraduallyenhancethepurchase

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–165317

Fig.3.EvolutionofmarketshareofEVsintotalfleet.

probabilityofEVsand,thus,provideafoundationforovercomingtheinitialslowgrowthstagethatisthenfollowedbyarapidgrowthstage.Theprocessofgrowthcontinuesinallscenariosastherearepositivefeedbackloopsbetweenmarketshareandwillingnesstoconsiderandbetweenmarketshareandrechargingeffect.ThesereinforcingmechanismsleadtorapidgrowthofEVs'marketpenetration.Itshouldbenotedthatthispatternrepresentsanartifactofthemodeland,hence,itcouldberealisticifnootherbalancingeffects(negativefeedbackloops)suchasconsumers'concernsaboutlimitedrangeofEVsandmarketpenetrationofotheralternativefuelvehiclesexist.Therefore,theresultsonthevaluesofmarketsharesshouldbeviewedwithsomecautionduetothelimitedscopeofthemodel.

ThecomparativeanalysisofthescenariosprovidesinsightsonthepatternofEVadoption.Theresultsshowthatinthelowestpenetrationscenario(Lowgasolineprice+ConstantEVprice),thetimetoachievethehalfofthemarketshareisdelayedtill2030.Inotherwords,incomparisonwiththebestcase(Highgasolineprice+DecreasingEVprice),thetimetoseizethemarketshareof50%ispostponedbyonedecade.

Assumingthatthenumberofvehiclesisconstantduringthestudyperiod,thepotentialforreductionofgasolineconsumptioninIcelandisdepictedinFig.4.Anannualdrivingdistanceof11,965km(astheaveragevalueduringtheperiod2006–2010[29])isassumedforthecalculationofgasolineconsumption.Thehighestpotentialinthebestcasescenarioshows93%reductioningasolineconsumptionduringthestudyperiod.

Thepercentageshareofeachvehiclein2030isshowninFig.5.EvolutionofthevehiclefleetwithtimeseemstoindicatetheICEbasedYaris,GolfandOutbackbeingswitchedfortheelectrici-MievandLeaf,whilethemoreexpensiveRoadster,911CarreraandOutlanderdonotgetswappedoutasrapidly.Boththe911CarreraandtheOutlanderhavelessthan1%shareofthetotalvehiclefleetbytheendoftheyear2030.

Figs.3–5indicatethatthenegativeimpactoflowgasolinepricesonEVmarketsharemaybebalancedbythepositiveimpactofEVpricereduction.Asaresult,themarketshareinthescenariowithlow(medium)gasolinepriceanddecreasingEVpricewouldberemarkablysimilartotheonewithmedium(high)gasolinepriceandconstantEVprice.

Toshowtheimpactofpolicysupportbytaxlevies,weperformedadditionalsensitivityanalysiswithrespecttotaxuponimportedEVs.Forsuchpurpose,thebestandtheworstcasescenariospresentedinFig.3(astheextremecasesinourstudy)arecombinedwiththescenarioonEVtaxation.Withanexcisetaxof45%onimportedEVs,theoutcomesindicateasignificantlysmallerportionofEVsduringthestudyperiod,asillustratedinFig.6.Inthelowextremecase,itcanbeseenthatbytaxleviesonEVs,theywouldnotsuccessfullypenetratethemarketandonly6%ofvehicleswithintheoverallvehiclefleetwouldbeEVsin2030.Inthehighextremecase,however,EVtaxationleadstoreductionofmarketshareby14%in2030.Thesensitivityanalysis,showedthatbymovingfromthebesttotheworstcasescenario,thenegativeeffectofEVtaxationonmarketpenetrationisincreased,indicatingthatthemarketforEVsdoesnottakeoffwithoutpolicysupportinthescenarioswithlowgasolinepriceand,tosomeextent,inthescenariowithmediumgasolinepricealongwithconstantEVprice.

Fig.7illustratestheimpactofeliminationofconsumers'concernsaboutrechargingandEVrangeonthepatternofmarketpenetration.Eliminationoftherefuelingeffectfromthemodelleadstoasituationinwhichtheconsumershavenoconcerns

18E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Fig.4.PotentialforreductionofgasolineconsumptioninIceland.

Fig.5.Compositionofvehiclefleetin2030(EVsareshowningreencolor).

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–165319

Fig.6.ImpactofexcisetaxonevolutionofmarketshareofEVs.

aboutrechargingavailabilityandlowrangeofEVs.TheresultsshowthatovercomingtheworryofconsumersaboutEVrechargingfacilitiesresultsinhigherpenetrationraterightfromthebeginning,endingupwithhighermarketshareofEVs.EVswouldeventuallyseizethemarketcompletelyinthescenariocombinedofhighgasolineprice,decreasingEVprice(withouttax)andnoworryabouttherechargingofEVs.

Toevaluatethesensitivityoftheresultstochangesinthevaluesofconsumerpreferences(presentedinTables9–12)weperformedadditionalsensitivityanalysis.Theparametersofinterestarepreferencevaluestowardvehiclepriceandfuelcostasthetwoimportantattributesinthemodel.Fig.8showstheeffectsofchangesinvehiclepricepreferencesonthemarketshareofEVs.Forthehighandlowpreferencestowardvehicleprice,weexaminethesensitivityofthebestandtheworstcasescenarios,

Fig.7.ComparisonoftheresultswiththecaseofnoconcernsaboutchargingofEVs.

1650E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

Table12

AverageandstandarddeviationofEVs'marketshare(resultsof20simulationswithdifferentrandomseed).Year

Highoilprice+decreasingEVpriceAverage

20112012201320142015201620172018201920202021202220232024202520262027202820292030

0.000.741.983.796.369.14.20.3927.3134.82.4749.9457.12.0070.5176.5882.0086.6690.4293.32

StdDev0.000.030.040.050.070.100.140.170.180.190.170.160.130.130.130.120.100.090.060.05

Lowoilprice+constantEVpriceAverage0.000.601.412.353.374.495.737.148.7210.12.6215.0717.9421.2725.1329.4434.0738.8043.4347.82

StdDev0.000.020.020.040.040.050.050.070.090.080.100.130.150.200.240.260.290.290.270.26

presentedinFig.3,tothepreferencevaluesthatare20%higherandlowerthanthepreferencevaluesinTables9–12.ThesensitivityofthemarketshareevolutionofEVsto±20%changesinpreferencevaluesforfuelcostisillustratedinFig.9.Theresultsindicatethatthemarketsharesaremoresensitivetochangesinthepreferencesoffuelcostthantochangesinthoseofvehicleprice.Moreover,thevariationaroundtheworstcasescenariowouldbehigherthanthatofthebestone.AhigherpreferencevaluestowardvehiclepricewouldreducethemarketshareofEVs.However,byincreasingthevalueofpreferencestowardfuelcost,EVswouldpenetratefaster.

Finally,duetothestochasticnatureoftheproposedmodel,weperformed20simulations,eachwithadifferentrandomseed.Thefindingsshownegligiblevariationsacrossdifferentrunsofeachscenario.Table12showstheaverageandstandarddeviation

Fig.8.Sensitivityanalysiswithrespecttoconsumers'preferencestowardvehicleprice.

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–16531651

Fig.9.Sensitivityanalysiswithrespecttoconsumers'preferencestowardfuelcost.

ofthemarketsharevaluesforthebestandtheworstcasescenariospresentedinFig.3.Theresultsindicatethatthemodelisstatisticallyrobustintermsofrepeatability.

6.Conclusion

Inthispaper,asimulationofthefuturedevelopmentoftheIcelandicpassengervehiclefleetwasgeneratedbymeansoftheagent-basedcomputationalapproach.Avehiclechoicealgorithmwasproposedthataccountsforsocialinfluencesandconsumers'attractivenessforvariousvehicleattributes.Theproposedalgorithmtakesintoaccountthetimedependenceontheevolutionofthevehiclefleet.Decisionpurchasesaredeterminedbyvehiclelifetimeandsocialgroupprobabilitiesforvehiclereplacement.Themodelhaspotentialtoanalyzetheeffectsoffuelprices,vehicles'attributes,consumerpreferencesandsocialinfluences.

DifferentscenarioswerethendefinedandthemodelwasiteratedtoobservethechangesinthemarketshareevolutionofEVs.TheresultsweregiveninratioofEVscomparedtoICEvehiclesfortheperiod2012–2030.TheresultsshowedthatEVscanbesuccessfullyadoptedinmoderateandfavorablescenarios.Ofthescenariostestedinthismodel,thecombinationofhighgasolineprice,decreasingEVprice,droppingtaxonEVsandeliminationofconsumerconcernsaboutrecharginghasthegreatesteffectregardingtheratioofEVsversusICEbasedvehicles.InthisscenarioEVscanseizethemarkettotallyin2030.SuccessfuladoptionofEVsinadversescenarios,liketheoneswithlowgasolinepriceandtheonemadeupofmediumgasolinepriceandconstantEVprice,needssupportpoliciessuchasreductioninimporttaxesandinfrastructuredevelopmentforrechargingtoovercometheconsumersworries.

Anumberofassumptionsandsimplificationshavebeenmadeinthemodel.First,themodeltakesintoaccountonlythepurchaseofnewvehiclesand,thus,whenanagentbuysanewvehicle,hispreviousvehiclegetsscrapped.RelaxingthisassumptionmaynegativelyaffectthetimingandmarketpenetrationrateofEVs.

AsignificantpartofconsumerbehaviorinthemodeloriginatesfromDanishdata.TheoriginofthedatamighthaveaneffectontheresultsinregardstopurchasedecisionssincetheretailpriceofDanishcarsisratherhigherthaninIceland,coupledwithmorereadilyavailablepublictransportationsuchassubwaysystemsandbikelanesinDenmark.ThecreditabilityofthecasestudywouldincreasesignificantlywithIcelandicdataonconsumerpreferencestocomplementit.

Consumersbehavedifferentlytoconsiderthevehiclesintheirconsiderationchoiceset.Moreover,theyresponddifferentlytochangingrechargingconditions,astheyareheterogeneousintheirdrivingpatterns.Inthisstudy,however,averagevaluesforwillingnesstoconsideranddrivingpatternhavebeenconsideredfortheaggregatepopulation.Providingarealempiricalstudyfordeterminingbehavioraldifferencesofindividualswouldenhancetheprecisionofresults.

TheproblemofrangeanxietyattachedtoEVshasbeenpartiallyanalyzedinthepaper.AvailabilityofchargingstationsanddrivingpatternofconsumeragentsarenotsufficienttoaddressrangeanxietyproblemandthereisaneedtoconsidertheeffectsofrechargingtimeandlimitedrangeofEVsonconsumers'preferences.

1652E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–1653

TheresultsonmarketshareofEVsseemtobeoptimisticastheEVwasconsideredastheonlyrepresentativeofalternativefuelvehiclesinthemodel.Includinghybridelectricvehicles(HEV)andplug-inhybridelectricvehicles(PHEV)astheotheroptionswiththepotentialtodisplaceaportionofICEmarketsharemayhaveabalancingeffectonadoptionofEVs.Moreover,theboundariesofthemodelcanbeextendedtoendogenizetheeffectsofincreasingelectricitydemandduetoelectrificationoftransportationsystemonthepriceofelectricity.Bydoingso,thenegativeimpactofhigherelectricitypriceonadoptionofEVscanbetakenintoaccount.

Whatmightadvancethemodelfurtheristoallowandcreatespatialenvironmentforagentstointeractwitheachother.Agentscouldinfluenceotheragentswithintheirneighborhood,forexampleifanagentbuysanEVandreallylikesittheagentmightinfluencethepreferencesofotheragentsregardingthesaidvehicle.ByaddingthespatialfeaturetothemodelthebenefitsofABMwouldbefullyexplored.References

[1]K.G.Høyer,Thehistoryofalternativefuelsintransportation:thecaseofelectricandhybridcars,Util.Pol.16(2008)63–71.

[2]M.Duvall,E.Knipping,M.Alexander,L.Tonachel,C.Clark,EnvironmentalAssessmentofPlug-InHybridElectricVehicles,Vol.1,NationwideGreenhouseGas

Emissions,EPRI,PaloAlto,CA,2007,Report1015325.

[3]K.Kieckhäfer,G.Walther,J.Axmann,T.Spengler,Integratingagent-basedsimulationandsystemdynamicstosupportproductstrategydecisionsinthe

automotiveindustry,in:M.D.Rossetti,R.R.Hill,B.Johansson,A.Dunkin,R.G.Ingalls(Eds.),Proceedingsofthe2009WinterSimulationConference,2009.[4]T.Zhang,S.Gensler,R.Garcia,Astudyofthediffusionofalternativefuelvehicles:anagent-basedmodelingapproach,J.Prod.Innovat.Manag.28(2011)152–168.[5]M.Schwoon,Simulatingtheadoptionoffuelcellvehicles,J.Evol.Econ.16(2006)435–472.

[6]O.Vliet,B.Vries,A.Faaij,W.Turkenburg,W.Jager,Multi-agentsimulationofadoptionofalternativefuels,Transport.Res.:D-Tr.E.15(2010)326–342.[7]G.Mueller,P.Haan,Howmuchdoincentivesaffectcarpurchase?Agent-basedmicrosimulationofconsumerchoiceofnewcars,PartI:modelstructure,

simulationofboundedrationality,andmodelvalidation,EnergyPolicy37(2009)1072–1082.

[8]M.J.Eppstein,D.K.Grover,J.S.Marshall,D.M.Rizzo,Anagent-basedmodeltostudymarketpenetrationofplug-inhybridelectricvehicles,EnergyPolicy39

(2011)37–3802.

[9]C.R.Bhata,S.Sen,N.Eluru,Theimpactofdemographics,builtenvironmentattributes,vehiclecharacteristics,andgasolinepricesonhouseholdvehicle

holdingsanduse,Transport.Res.:B-Meth.43(2009)1–18.

[10]A.Mohammadian,E.J.Miller,Empiricalinvestigationofhouseholdvehicletypechoicedecisions,Transp.Res.Rec.18(2003)99–106.

[11]M.A.Janssen,W.Jager,Simulatingmarketdynamics:theinteractionsofconsumerpsychologyandstructureofsocialnetworks,Artif.Life9(4)(2003)

343–356.

[12]S.A.Delre,W.Jager,T.H.A.Bijmolt,M.A.Janssen,Willitspreadornot?Theeffectsofsocialinfluencesandnetworktopologyoninnovationdiffusion,J.Prod.

Innovat.Manag.27(2010)267–282.

[13]S.A.Delre,W.Jager,T.H.A.Bijmolt,M.A.Janssen,Targetingandtimingpromotionalactivities:anagent-basedmodelforthetakeoffofnewproducts,J.Bus.

Res.60(2007)826–835.

[14]S.A.Delre,W.Jager,M.A.Janssen,Diffusiondynamicsinsmall-worldnetworkswithheterogeneousconsumers,Comput.Math.Organ.Theory13(2007)185–202.[15]A.Namatame,H.Morita,K.Matsuyama,Agent-basedmodelingforthestudyofdiffusiondynamics,Proceedingsofthe2009SpringSimulation

Multiconference,SanDiego,California,USA,March22–272009.

[16]A.Mohammadian,E.J.Miller,Dynamicmodelingofhouseholdautomobiletransactions,Transp.Res.Rec.1831(2003)98–105.[17]R.Garcia,ModelingVehicleChoiceBehaviorUsingAgent-BasedApproach,UniversityofMichigan,Michigan,2007.[18]K.E.Train,DiscreteChoiceModelswithSimulation,2nded.CambridgeUniversityPress,NewYork,2009.

[19]J.Struben,J.D.Sterman,Transitionchallengesforalternativefuelvehicleandtransportationsystems,Environ.Plann.B35(6)(2008)1070–1097.

[20]COWI,FiscalmeasurestoreduceCO2emissionsfromnewpassengercars—mainreport,EuropeanCommission'sDirectorate-GeneralforEnvironment,

COWIA/S,2002.

[21]Yahoo!Autos,http:www.autos.yahoo.com/new-cars.html(Accessed:September2,2011).[22]www.green.autoblog.com(Accessed:September2,2011).

[23]D.B.Mueller,J.Cao,E.Kongar,M.Altonji,P.H.Weiner,T.E.Graedel,Servicelifetimesofmineralenduses,u.s.geologicalsurvey,MineralsResourcesExternal

ResearchProgramAwardNumber:06HQGR0174,2007.

[24]CreditInfo,UpplýsingagjöfCreditinfoÍsland-skýrslaIIIvegnasérstakraaðstæðnaáíslenskumfjármálamarkaði,CreditInfo,Reykjavík,2008.

[25](Accessed:May28,2010).StatisticsIceland,Utanríkisverslun,http://hagstofa.is/?PageID=744&src=/temp/Dialog/varval.asp?ma=UTA03201%26ti=Innflutningur+

nokkurra+v%F6rutegunda+eftir+m%E1nu%F0um+1999%2D2010%26path=../Database/utanrikisverslun/Innflutningur/%26lang=3%26units=Einingar/Tonn/CIF%20millj%F3nir%20kr%F3na.

[26]U.S.EnergyInformationAdministration,AnnualEnergyOutlook,2011.

[27]OrkuveitaReykjavíkur.(Accessed:Jun1,2010).OrkuveitaReykjavíkur.http://www.or.is/Heimili/Verdskraogskilmalar/Rafmagn.

[28]C.Thiel,A.Perujo,A.Mercier,CostandCO2aspectsoffuturevehicleoptionsinEuropeundernewenergypolicyscenarios,EnergyPolicy38(2010)

7142–7151.

[29]TheRoadTrafficDirectorate,http://www.us.is/id/1000910(Accessed:October10,2011).

Dr.EhsanShafieiiscurrentlyapostdoctoralresearcheratReykjavikUniversityinIceland.HereceivedhisPhDinMechanicalEngineering(EnergySystems)fromSharifUniversityofTechnology,Tehran,Iranin2009.HehaspreviouslyworkedatIranMinistryofEnergyasenergy-systemmodelerfrom2005to2010.HisareasofresearchinterestsincludeEnergy-SystemsModeling,ClimateChangeandMitigation,OptimalControlTheoryandSystem-Dynamics.

HedinnThorkelssonhasanMScdegreeinFinancialEngineeringfromtheSchoolofScienceandEngineering,ReykjavikUniversityinIceland.Hehasworkedontheresearchduringhisgraduatestudies.

Dr.EyjólfurIngiÁsgeirssonisanassistantprofessorintheSchoolofScienceandEngineeringatReykjavíkUniversityinIceland.HereceivedhisPhDinOperationsResearchfromColumbiaUniversityin2007.HealsohasaCSdegreeinIndustrialEngineeringandaBSdegreeinComputerScience,bothfromUniversityofIceland.Hisresearchinterestsaredesignandanalysisofalgorithms,optimization,planning&scheduling,timetabling,combinatoricoptimization,problemsolving,AI,andapplicationsofoperationsresearch.

Dr.BrynhildurDavidsdottirisanassociateprofessorofEnvironmentandNaturalResourcesattheUniversityofIceland,andistheDirectoroftheGraduatePrograminEnvironmentandNaturalResourcesattheUniversityofIceland.Herresearchfocusesmostlyoncomplexsystemsmodelingwithafocusonalternativeenergyfutures,energyandclimatemitigationandadaptationpolicy,soilsustainabilityandecosystemservicesofsoils,sustainabilityindicatorsinvariouscontexts,andactionresearchonhowtofacilitatethecreationofsustainablecommunities.

E.Shafieietal./TechnologicalForecasting&SocialChange79(2012)1638–16531653

Dr.MarcoRabertoisassistantprofessorofBusinessandManagementEngineeringattheFacultyofEngineeringoftheUniversityofGenoa,Italy.HegottheLaureaDegreeinPhysicsin1999andtheDoctoraldegreeinElectronicsandComputerScienceEngineeringin2003attheUniversityofGenoa.Hismainresearchinterestsregardagent-basedmodelingandsimulationofeconomicsystemsandfinancialmarketswithparticularfocusontheissuesoffinancialinstabilityandtheinterplaybetweenfinancialandmonetaryvariablesandtherealeconomy.

Dr.HlynurStefanssonisanAssistantProfessorattheSchoolofScienceandEngineering,ReykjavikUniversityinIceland.HereceivedhisPhDinEngineeringfromImperialCollegeLondonandM.Sc.inInformaticsfromtheInstituteofMathematicalModelingattheTechnicalUniversityofDenmark.Hisresearchinterestsareinthefieldofdecisionscience,simulationandoptimizationincludingvariousapplicationswithaspecialfocusonsustainableenergyproductionandfutureenergystrategies.