what affects millennials’ mobility? part ii: the ... - steps€¦ · to a sample of 2400...

93
What Affects Millennials’ Mobility? PART II: The Impact of Residential Location, Individual Preferences and Lifestyles on Young Adults’ Travel Behavior in California March 2017 A Research Report from the National Center for Sustainable Transportation Dr. Giovanni Circella, University of California, Davis Farzad Alemi, University of California, Davis Kate Tiedeman, University of California, Davis Rosaria M. Berliner, University of California, Davis Yongsung Lee, Georgia Institute of Technology Dr. Lew Fulton, University of California, Davis Prof. Patricia L. Mokhtarian, Georgia Institute of Technology Prof. Susan Handy, University of California, Davis

Upload: others

Post on 29-Jun-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

WhatAffectsMillennials’Mobility?PARTII:TheImpactofResidentialLocation,IndividualPreferencesandLifestylesonYoungAdults’TravelBehaviorinCalifornia

March2017

AResearchReportfromtheNationalCenterforSustainableTransportation

Dr.GiovanniCircella,UniversityofCalifornia,DavisFarzadAlemi,UniversityofCalifornia,DavisKateTiedeman,UniversityofCalifornia,DavisRosariaM.Berliner,UniversityofCalifornia,DavisYongsungLee,GeorgiaInstituteofTechnologyDr.LewFulton,UniversityofCalifornia,DavisProf.PatriciaL.Mokhtarian,GeorgiaInstituteofTechnologyProf.SusanHandy,UniversityofCalifornia,Davis

Page 2: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

AbouttheNationalCenterforSustainableTransportationTheNationalCenterforSustainableTransportationisaconsortiumofleadinguniversitiescommittedtoadvancinganenvironmentallysustainabletransportationsystemthroughcutting-edgeresearch,directpolicyengagement,andeducationofourfutureleaders.Consortiummembersinclude:UniversityofCalifornia,Davis;UniversityofCalifornia,Riverside;UniversityofSouthernCalifornia;CaliforniaStateUniversity,LongBeach;GeorgiaInstituteofTechnology;andUniversityofVermont.Moreinformationcanbefoundat:ncst.ucdavis.edu.DisclaimerThecontentsofthisreportreflecttheviewsoftheauthors,whoareresponsibleforthefactsandtheaccuracyoftheinformationpresentedherein.ThisdocumentisdisseminatedunderthesponsorshipoftheUnitedStatesDepartmentofTransportation’sUniversityTransportationCentersprogram,intheinterestofinformationexchange.TheU.S.GovernmentandtheStateofCaliforniaassumesnoliabilityforthecontentsorusethereof.NordoesthecontentnecessarilyreflecttheofficialviewsorpoliciesoftheU.S.GovernmentandtheStateofCalifornia.Thisreportdoesnotconstituteastandard,specification,orregulation.AcknowledgmentsThisstudywasfundedbyagrantfromtheNationalCenterforSustainableTransportation(NCST),supportedbyUSDOTandCaltransthroughtheUniversityTransportationCentersprogram.TheauthorswouldliketothanktheNCST,USDOT,andCaltransfortheirsupportofuniversity-basedresearchintransportation,andespeciallyforthefundingprovidedinsupportofthisproject.AdditionalfundingforthedatacollectionforthisprojectwasprovidedbytheUCDavisSustainableTransportationEnergyPathways(STEPS)program.Theauthorsaregratefulforthissupport.Allerrorsoromissionsaretheresponsibilityoftheauthors,andnotthefundingorganizations.TheauthorswouldliketosincerelythankDanielSperling,RamPendyala,KayAxhausen,JoanWalker,ElisabettaCherchi,CinziaCirillo,DavidBunch,GilTal,KariWatkins,ScottLeVine,DeborahSalon,LauraPodolsky,EricGudz,FreshtaPirzada,AliaksandrMalokin,GouriMishra,CalvinThigpen,AlvaroRodriguezValencia,SimonBerrebi,AliceGrossman,AliEtezady,AtiyyaShaw,SarahMooney,RubemMondaini,AnissBahreinian(CaliforniaEnergyCommission),KatieBenouar,MelissaThompson,SoheilaKhoii,MohammadAssadi,DillonMiner,NicoleLongoria,PatrickTynerandDavidChursenoff(Caltrans),JohnOrr,ElisabethSanfordandGuyRousseau(AtlantaRegionalCommission),DavidOry(MetropolitanTransportationCommission),MikeAlba(LinkedInCorp.),KenLaberteaux(ToyotaMotorCorp.),andNataliaTinjacaMora(CamaradeComercioBogota,Colombia)fortheirthoughtfulcommentsandcontributionsduringthesurveydesign,datacollectionanddataanalysis.

Page 3: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

WhatAffectsMillennials’Mobility?PARTII:TheImpactofResidentialLocation,IndividualPreferencesandLifestyleson

YoungAdults’TravelBehaviorinCalifornia

ANationalCenterforSustainableTransportationResearchReport

March2017

GiovanniCircella,InstituteofTransportationStudies,UniversityofCalifornia,Davis

FarzadAlemi,InstituteofTransportationStudies,UniversityofCalifornia,Davis

KateTiedeman,InstituteofTransportationStudies,UniversityofCalifornia,Davis

RosariaM.Berliner,InstituteofTransportationStudies,UniversityofCalifornia,Davis

YongsungLee,SchoolofCityandRegionalPlanning,GeorgiaInstituteofTechnology

LewFulton,InstituteofTransportationStudies,UniversityofCalifornia,Davis

PatriciaL.Mokhtarian,SchoolofCivilandEnvironmentalEngineering,GeorgiaInstituteofTechnology

SusanHandy,InstituteofTransportationStudies,UniversityofCalifornia,Davis

Page 4: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

[pageleftintentionallyblank]

Page 5: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

TABLEOFCONTENTSEXECUTIVESUMMARY....................................................................................................................iIntroduction...................................................................................................................................1TheMobilityofMillennials.............................................................................................................4TheCaliforniaMillennials’Dataset.................................................................................................8

DataCleaningandRecodes......................................................................................................12Geocoding................................................................................................................................14WeightingandRaking..............................................................................................................19IntegrationofAdditionalLandUseDatafromOtherSources.................................................22FactorAnalysis.........................................................................................................................24

AdoptionofTechnology,IndividualAttitudesandMobilityChoicesofMillennialsvs.GenXers29InvestigatingMillennials’AttitudestowardsTransportationandTechnology........................33TravelBehaviorandtheAccessibilityofthePlaceofResidence.............................................46AdoptionofMultimodalTravelBehavior.................................................................................49

VehicleMilesTraveled.................................................................................................................55DependentVariable:Self-ReportedWeeklyVMT....................................................................56ExplanatoryVariables...............................................................................................................56Results......................................................................................................................................58

CarOwnership,VehicleTypeChoiceandPropensitytoChangeVehicleOwnership..................63VehicleTypeChoiceModel......................................................................................................67PropensitytoModifyVehicleOwnership................................................................................73

ConclusionsandNextStepsoftheResearch...............................................................................76References....................................................................................................................................81ListofAcronymsUsedintheDocument......................................................................................85

Page 6: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

i

WhatAffectsMillennials’Mobility?PARTII:TheImpactofResidentialLocation,IndividualPreferencesandLifestylesonYoungAdults’TravelBehaviorinCaliforniaEXECUTIVESUMMARYYoungadults(“millennials”,ormembersof“GenerationY”)areincreasinglyreportedtohavedifferentlifestylesandtravelbehaviorfrompreviousgenerationsatthesamestageinlife.Theypostponethetimeatwhichtheyobtainadriver’slicense,oftenchoosenottoownacar,drivelessiftheyownone,andusealternativenon-motorizedmeansoftransportationmoreoften.Severalexplanationshavebeenproposedtoexplainthebehaviorsofmillennials,includingtheirpreferenceforurbanlocationsclosertothevibrantpartsofacity,changesinhouseholdcomposition,andthesubstitutionoftravelforworkandsocializingwithtelecommutingandsocialmedia.However,researchinthisareahasbeenlimitedbyalackofcomprehensivedataonthefactorsaffectingmillennials’residentiallocationandtravelchoices(e.g.informationaboutindividualattitudes,lifestylesandadoptionofsharedmobilityisnotavailableintheU.S.NationalHouseholdTravelSurveyandmostregionalhouseholdtravelsurveys).Improvingtheunderstandingofthefactorsandcircumstancesbehindmillennials’mobilityisoftheutmostimportanceforscientificresearchandplanningprocesses.Millennialsmakeupasubstantialportionofthepopulation,andtheirtravelandconsumerbehaviorwillhavelargeeffectsonthefuturedemandfortravelandgoods.Further,millennialsareoftenearlyadoptersofnewtrendsandtechnologies;therefore,improvingtheunderstandingofmillennials’choiceswillincreasetheabilitytounderstandandpredictfuturetrendsatlarge.ThisstudybuildsonalargeresearcheffortlaunchedbytheNationalCenterforSustainableTransportationtoinvestigatetheemergingtransportationtrendsandtheimpactsoftheadoptionofnewtransportationtechnologiesinCalifornia,particularlyamongtheyoungercohorts,i.e.millennialsandthemembersoftheprecedingGenerationX.Duringthepreviousstagesoftheresearch,wedesignedadetailedonlinesurveythatweadministeredinfall2015toasampleof2400residentsofCalifornia,includingmillennials(youngadults,18-34in2015)andGenXers(35-50year-oldadults).Weusedaquotasamplingapproachtorecruitrespondentsfromeachagegroup(youngmillennials,oldermillennials,youngGenXers,andolderGenXers)acrossallcombinationsofmajorgeographicregionofCaliforniaandneighborhoodtype(urban,suburban,andrural). TheresultistheCaliforniaMillennialsDataset,acomprehensivedatasetthatcontainsinformationontherespondents’personalattitudes;lifestyles;adoptionofonlinesocialmediaanduseofinformationandcommunicationtechnology(ICT)devicesandservices;residential

Page 7: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

ii

locationandlivingarrangements;commutingandothertravelpatterns;autoownership;awareness,adoptionandfrequencyofuseofvarioussharedmobilityservices;majorlifeeventsinthepastthreeyears;expectationsforfutureevents;propensitytopurchaseanduseaprivatevehiclevs.touseothermeansoftravel;politicalideas,andsociodemographictraits.Thisreportsummarizestheanalysesoftheresidentiallocation,travelbehaviorandvehicleownershipofmillennialsandGenXers.Inthisstageoftheresearch,weaugmentedtheCaliforniaMillennialsDatasetwithadditionalvariablesmeasuringlanduseandbuiltenvironmentcharacteristicsfromothersourcesincludingtheU.S.EnvironmentalProtectionAgency’sSmartLocationDataset,andthewalkscore,bikescoreandtransitscorefromthecommercialwebsitewalkscore.com.Weweightedthedatatocorrectthedistributionofcasesinthesample,andtoreducethenon-representativenessofthedata,basedontheregionofCaliforniawheretherespondentslive,theneighborhoodtype,theagegroup,gender,studentandemploymentstatus,householdincome,raceandethnicity,andpresenceofchildreninthehousehold.Weapplieddatareductiontechniquestosummarizetheinformationrelatedtotheindividualattitudesandpreferences.Todothis,weperformedaprincipalaxisfactoranalysisonthe66attitudinalvariablesthatwerecollectedinthesurvey.Atotalof17factorswereextracted.SeveralkeydifferencesareobservedinthedistributionofthefactorscoresacrossvariousgroupsofmillennialsandGenXers.Forexample,wefindlargedifferencesintheattitudinalprofilesofmillennialsandGenXersonattitudinaldimensionssuchasmaterialism,thepropensitytoadoptnewtechnologies,andthedegreetowhichindividualsfeeltheyarewell-establishedintheirlife.Forotherattitudinalfactors,e.g.thepro-environmentalpolicyattitudes,thedifferencesassociatedwiththelocationwhererespondentsliveareremarkablylargerthanthedifferencesobservedacrossagegroups:urbandwellersconsistentlyreportstrongerpro-environmentalpolicyattitudesthannon-urbanresidents.Wealsofindthaturbanmillennialsareheavyadoptersoftechnology,smartphoneappsinparticular,andonaverageusetheseservicesmoreoftenforvariouspurposes,includingaccessinginformationaboutthemeans(orcombinationofmeans)oftransportationtouseforatrip,findinginformationaboutpotentialtripdestinations(e.g.acafé,orarestaurant),ornavigatinginrealtimeduringatrip.Largedifferencesarealsoobservedintheadoptionofsharedmobilityacrossbothagegroupsandurbanvs.non-urbanpopulations;notsurprisingly,millennialstendtoadoptthesetechnologicalservicesmoreoftenthanGenXers,particularlyinurbanareas.Wefurtheranalyzedtherelationshipsbetweenaccessibilityandtheadoptionofmultiplemodesoftransportation(multimodality,and/orintermodality)amongthevarioussub-segmentsofthepopulation.Forthisanalysis,weclassifiedmillennialsintwogroupsofindependentanddependentmillennialsbasedontheirlivingarrangementsandhouseholdcomposition.Infact,theresidentiallocationwheredependentmillennialslivehaslikelybeentheresultoftheirparents’choices,andnotofthemillennialsthemselves.WecomparedthelevelofaccessibilityoftheplaceofresidenceandtheadoptionofmultimodaltravelofthesetwogroupsofmillennialswiththoseofGenXers.Accessibilityandmultimodalityareusually

Page 8: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

iii

positivelycorrelated:residentsofmoreaccessibleneighborhoodsaremoreoftenmultimodaltravelers.However,millennials,andespeciallydependentmillennials,arefoundtomakethemostoftheirbuiltenvironmentpotential,eitherduetoindividualchoices,orthepresence(orlack)oftravelconstraints.Theyarelesslikelytobemono-driversandmorelikelytobemultimodalcommuters,eveniftheyoftenliveinneighborhoodsthatarelesssupportiveofsuchbehaviors.Ontheotherendofthespectrum,GenXersbyfarrelythemostoncars.Independentmillennialsmoreoftenchoosetoliveinaccessiblelocationsandtendtoadoptnon-motorizedandmultimodaltraveloptionsmoreoften.Weestimatedalog-linearmodelofthenumberofweeklyvehiclemilestraveled(VMT),usingbothapooledmodelfortheentiresampleandasegmentedmodelthatteststheeffectsofindividual,householdandlandusecharacteristicsontheVMTofmillennialsandGenXersseparately.Interestingly,themodelformillennialsexplainsthelowestamountofvarianceinthedata.Thisfindingsignalsthehigherheterogeneityandvariationamongthemembersofthisgroup,andtheincreaseddifficultyinexplainingtheirbehaviorsthroughtheestimationofeconometricandquantitativemodels.TraditionalbuiltenvironmentvariablessuchaspopulationdensityandlandusemixexplainalowerportionofVMTformillennialscomparedtoGenXers.Individualattitudesandstageinlife(currentlivingarrangementsandthepresenceofchildreninthehousehold)havelargereffectsonVMTformillennialsthanforGenXers.Wealsoinvestigatedtherelationshipsbehindcarownershipandthetypeofvehicleownedbyahousehold.Notsurprisingly,independentmillennialsthatliveinurbanareas,onaverage,ownfewercarsperdriverthanothergroups.Thisfindingcorroboratesthereducedneedsforacarindenser(andmoreaccessible)centralareas,wherealargerportionofindependentmillennialslive.However,suchaneffectmightbeshort-lived:manyoldermillennialswholiveinurbanareasreportthattheyplantopurchaseanewvehicleinthenearfuture.Thus,theirzero-orlow-vehicleownershipisprobablytheresultoftheirtransientstageofliferatherthanthelong-termeffectofpreferencestowardsvehicleownership.Duringfuturestagesoftheresearch,wedoplantostudyhowcarownershipvariesacrossdifferentgroupsofthepopulationthroughtheestimationofamodelthatinvestigateshowsociodemographiccharacteristics,individualpreferences,andlandusefeaturesaffectcarownership.Toinvestigatethepreferencetowardsthepurchaseofvariousvehicletypesamongdifferentgroupsofusers,wealsoestimatedamultinomiallogitmodel(MNL)ofvehicletypechoice,usingsociodemographictraits,builtenvironmentcharacteristics,andpersonalattitudesandpreferencesasexplanatoryvariables,fortheindividualsthatboughtorleasedavehiclethatismodelyear2010ornewer.Futurestagesoftheresearchwillfocusontheanalysisofadditionalcomponentsofmillennials’choices,includingcurrentresidentiallocation,futureaspirationstomodifyvehicleownershipandtravelchoices,theadoptionofsharedmobilityservices,andtherelationshipsbetweentheadoptionofsharedmobility,household’svehicleownership,andothercomponentsoftravelbehavior(e.g.thefrequencyofuseofothertransportationmodes).

Page 9: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

1

IntroductionYoungadults(oftenreferredtoas“millennials”,ormembersof“GenerationY”)areincreasinglyreportedtohavedifferentlifestylesandtravelbehaviorfrompreviousgenerationsatthesamestageinlife.Theypostponethetimetheyobtainadriver’slicense,oftenchoosetoliveinmorecentralurbanlocationsandchoosenottoownacar,drivelesseveniftheyownone,andusealternativenon-motorizedmeansoftransportationmoreoften.Severalpossibleexplanationshavebeenproposedtoexplaintheobservedbehaviorsofmillennials,includingtheirpreferenceformoreurbanlocations,changesinhouseholdcomposition,andsubstitutionoftravelforworkandsocializingwithtelecommutingandsocialmedia.ThebehaviorofmillennialshasanimportantroleinexplainingthechangesincartravelobservedinrecentyearsintheUnitedStatesandotherdevelopedcountries,wherethetotalvehiclemilestraveled(VMT)have,atleasttemporarily,“peaked”beforereboundingsharply,atleastintheUnitedStates,tonewrecordhighsinthefirsthalfof2016(FHWA,2016;Circellaetal.,2016a).Severalstudieshavestartedtoinvestigatethefactorsaffectingtheresidentiallocationandmobilitychoicesofmillennials.However,thedebateinthisfieldisstilldominatedbyspeculationsaboutthepotentialfactorsaffectingmillennials’behavior.Previousstudieshavebeenlimitedbythelackofinformationonspecificvariables(e.g.personalattitudesandpreferences,forstudiesbasedonNationalHouseholdTravelSurveydata),ortheuseofconveniencesamples(e.g.studiesonuniversitystudents).Certainly,theconnectedtech-savvymillennialsareapopularfigureinthemediaheadlines,andtheyareacommonpresenceinSanFrancisco,LosAngeles,oranyothermajorcityinthecountry.Notallmillennialsfitthisstereotype,though,andtherearelargemassesofyoungadultsthatstillbehaveinawaythatismoresimilartooldercohorts:theyarelikelytogetmarriedatayoungerage,oftenliveinsingle-familyhomes,drivealonefortheircommute,andraisetheirchildreninapredominantlysuburbanenvironment.Understandingthedifferentpatternsinlifestylesandbehaviorsamongthevarioussegmentsoftheheterogeneouspopulationofmillennials,andquantifyingtheirimpactontraveldemandandtheuseofvariousmeansoftransportation,isofextremeimportancetoresearchers,plannersandpolicy-makers.ThisstudybuildsonalargeresearcheffortlaunchedbytheNationalCenterforSustainableTransportationtoinvestigatetheemergingtransportationtrendsandtheimpactsoftheadoptionofnewtransportationtechnologiesinCalifornia,inparticularamongtheyoungercohorts,i.e.millennials.Duringthepreviousstagesoftheresearch,alargedatasetwascollectedwithacomprehensiveonlinesurveythatwasadministeredinfall2015toasampleof2400residentsofCalifornia,includingbothmillennials(youngadults,18-34in2015)andmembersoftheprecedingGenerationX(middle-ageadults,35-50).Weusedaquotasamplingprocesstoensurethatenoughrespondentsfromallagegroups(youngmillennials,oldermillennials,youngGenXers,andolderGenXers)weresampledfromeachcombinationofgeographicregionofCaliforniaandneighborhoodtype(urban,suburban,andrural),and

Page 10: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

2

controlledfordemographictargetsofthesampleforfivedimensions:gender,age,householdincome,raceandethnicity,andpresenceofchildreninthehousehold.TheresultistheCaliforniaMillennialsDataset,anunprecedenteddatasetwhichcontainsdetailedinformationontherespondents’personalattitudes,preferencesandenvironmentalconcerns;lifestyles;adoptionofonlinesocialmediaanduseofinformationandcommunicationtechnology(ICT)devicesandservices;residentiallocationandlivingarrangements;commutingandothertravel-relatedpatterns;autoownership;awareness,adoptionandfrequencyofuseofthemostcommonsharedmobilityservices(includingcar-sharing,bike-sharing,dynamicridesharingandon-demandrideservicessuchasUberorLyft);majorlifeeventshappenedinthepastthreeyears;expectationsforfutureeventsandpropensitytopurchaseanduseaprivatevehiclevs.touseothermeansoftravel;politicalideasandsociodemographictraits.Duringthisstageoftheresearch,webuiltontheCaliforniaMillennialsDataset,integratedthedatasetwithadditionaldataavailablefromothersources,andinvestigatedseveraltopicsrelatedtomillennials’mobilitychoicesandthechangingtrendsintraveldemandinCalifornia.Specifically,aspartofthestudy,wegeocodedtheresidentiallocationandtheprimarywork/studylocationreportedbyeachrespondentinthesample.Usingalsotheinformationfromthegeocodedresidentialandworklocationsoftherespondents,wedevelopedasetofqualitychecks,andfurthercleanedandrecodedtheinformationavailableinthedataset.Wematchedtherespondents’geocodedresidentiallocationwiththeinformationonthedominantneighborhoodtypeavailablefromanotherresearchprojectdevelopedatUCDavis.Further,wedevelopedasetofweights,usingbothcellweightsandtheiterativeproportionalfitting(IPF)rakingprocess,tocorrectforthenon-representativenessofthesampleintermsofdistributionbyregionofCalifornia,predominantneighborhoodtype,agegroup,gender,householdincome,studentandemploymentstatus,raceandethnicity,andpresenceofchildreninthehousehold.Basedonthegeocodedresidentiallocationoftherespondents,weintegratedthedatasetwithadditionalvariablesobtainedfromexternalsources.Theadditionalvariablesprovidedinformationonthecharacteristicsofthebuiltenvironmentintheplaceofresidenceandtravelaccessibilitybymode,frommultiplesourcesincludingtheU.S.EnvironmentalProtectionAgency(EPA)SmartLocationDataset,andthecommercialwebsiteWalkscore.com(whichalsocomputesabikescoreandtransitscore,inadditiontothebetter-knownwalkscore).Weappliedfactoranalysisasadatareductiontechniquetoinvestigatetherelationshipsrelatingthe66attitudinalvariablesavailableinthedatasetandtoextract17factorsthatmeasureattitudinalconstructsonseveraldimensionsofinterest.Wedevelopedanumberofanalysesusingtheinformationinthedataset,focusinginparticularontheimpactsoflandusecharacteristicsandthedifferentbehaviorsobserved,forexample,among“urban”millennialsvs.theothergroupsofyoungadultswholiveinsuburbanorruralareas,andthecorrespondinggroupsofGenXers.Thisreportsummarizesthefindingsfromthisstageoftheresearch.Intheremainderofthisreport,wefirstdiscussrecentstudiesthathaveinvestigatedseveralaspectsofmillennials’mobilityandcarownershipchoices.Wethenpresenttheinformation

Page 11: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

3

containedintheCaliforniaMillennialsDataset,summarizethedatacleaningandrecodingtasksthatwereperformedaspartofthisstageoftheresearch,describetheprocessthatwasusedtogeocodetheresidentialandworklocationsoftherespondents,theweightingprocessappliedtothedataset,andtheadditionaldatathatwereimportedfromexternalsourcesandthatwerematchedbasedonthegeocodedresidentiallocationoftherespondents,andpresenthowweappliedfactoranalysisonthe66attitudinalstatementsinthedatasettoextract17mainattitudinalfactors.ThefollowingsectionsinvestigatedifferencesamongmillennialsandthemembersoftheGenerationXthatliveinurban,suburbanandruralareas,startingfromtheuseofsocialmediaandsmartphoneappstocoordinatetravelalternativesandtoaccessinformationonthemeansoftransportationavailableforatrip,informationaboutpotentialtripdestinationsandreal-timetravelinformation,amongothers,andthenmovingtodiscussthedifferentattitudinalpatternsreportedbytheresidentsofvariousneighborhoodtypes,bygeneration.Wepresentseveralmeasuresofaccessibilityandinvestigatetheadoptionofmultimodaltravelamongdifferentgroupssegmentedbygenerationandneighborhoodtype.Thefollowingchapterpresentsasetofeconometricmodelsoftheindividuals’vehiclemilestraveled(VMT),whichwereestimatedasbothapooledmodel(fortheentiresample)andsegmentedmodelsformillennialsandGenXers.Themodelsallowidentifyingtheimpactsofindividualandhouseholdcharacteristics,stageinlife,landusecharacteristics,adoptionoftechnologyandpersonalattitudesontheamountofcartravelofmillennialsandGenXers.Wethenturnourattentiontocarownershipandvehicletypechoice,throughthecomparisonofthedifferentcarownershiplevelsfoundamongmembersofdifferentgenerationalgroupsthatliveinthevariousneighborhoodtypes.Weestimateadiscretechoicemodelofthevehicletypechoice,whichshedslightontheimpactofseveralgroupsofexplanatoryvariablesonthedecisiontobuyorleaseaspecifictypeofvehicles,anddiscussthedifferenttrendsinthepropensitytochangethelevelofvehicleownershipinthehousehold(e.g.propensitytobuyanewvehicle)observedamongthemembersofdifferentgenerationalgroupsthatliveinurbanvs.non-urbanlocations.Thefinalconclusionssummarizethefindingsfromthisstageoftheproject,andidentifydirectionsforfutureresearch.Theactivitiesdevelopedsofarinthisresearchprojectandthelargeamountofinformationthathasbeencollectedwillallowanumberofadditionalanalysesofpotentialinterestfortheresearchcommunity,plannersandpolicy-makers;thesewillbedevelopedduringthenextstagesofthismulti-yearresearchprogram.ThisPartIIReportbuildsonthePartIReporttitled"WhatAffectsMillennials’Mobility?PARTI:InvestigatingtheEnvironmentalConcerns,Lifestyles,Mobility-RelatedAttitudesandAdoptionofTechnologyofYoungAdultsinCalifornia”,whichprovideddetailedinformationonthemotivationsforthisstudy,previousstudiesfromtheliteratureonwhichthisresearchbuilds,thedatacollectioneffort,thecontentoftheonlinesurveythatwasusedinthestudy,thesamplingmethodologyandpreliminaryanalysisoftheCaliforniaMillennialsDataset.AdditionalinformationonthesetopicscanbefoundinthePartIprojectreport(seeCircellaetal.,2016b).

Page 12: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

4

TheMobilityofMillennialsMillennials(i.e.theyoungadultsborninthe1980sand1990s,whobecameadultsinthe21stcentury)areoftenreportedtobehavedifferentlyfrompreviousgenerationsatthesamestageinlife.Severalstudieshavediscussedthechangingtrendsinmillennials’lifestylesandmobilitydecisions.Millennialsarefoundtopostponethetimetheyobtainadriver’slicense,oftenchoosetoliveinmorecentralurbanlocationsandchoosenottoownacar,drivelesseveniftheyownone,andusealternativenon-motorizedmeansoftransportationmoreoften(Blumenbergetal.2012;Kuhnimhofetal.2012;Blumenbergetal.2015;McDonald2015;Circellaetal.2016b).Severalpossibleexplanationshavebeenproposedtoexplaintheobservedbehaviorsofmillennials,includingtheirpreferenceformoreurbanlocationsclosertothevibrantpartsofacity,changesinhouseholdcomposition,andthesubstitutionoftravelforworkandsocializingwithtelecommutingandsocialmedia.Inthisstudy,wefollowthedefinitionof“millennials”thatisconsistentwiththerecentstudiespublishedbythePewResearchCenter,whichidentifymillennialsastheindividualsbornbetween1981and1997(i.e.theywere18to34-year-old,asof2015).Thissegmentofthepopulationmayhavedifferentbehaviorsandlifestylesfromoldergenerations,evenwhilecontrollingforstageoflife,causingthemtotraveldifferently.Severalstudieshavestartedtoinvestigatethechangingtrendsinmillennials’mobility,andthefactorsthatarelikelytoaffecttheirchoices.Foranextensivereviewoftheliteraturethathasfocusedonmillennials’behavior,pleaserefertothePartIreportfromthisproject(Circellaetal.,2016b).Itisdifficulttoseparatethegenerationalcomponentofmillennials’behaviorsfromotherfactorsaffectingtheirmobilitychoices,includingthechangingeconomicconditionsandfluctuationsinfuelprices,trafficcongestioninlargemetropolitanareas,changesintheurbanformofAmericancities,householdcompositionandpersonallifestyles,theeventualsubstitutionofphysicaltripswithinformationandcommunicationtechnologies(ICT),astrongertendencytowardsmultimodality,andtheincreasedavailabilityofalternativetraveloptionsincludingnewsharedmobilityservicessuchascar-sharingandon-demandrideservices(e.g.thoseprovidedbytransportationnetworkscompanies,orTNCs,suchasUberorLyft,intheAmericanmarket)(Wachs,2013,Polzinetal.,2014;BuehlerandHamre,2014).Recentsociodemographicshiftsandmodificationsinhabitsandlifestylesincludemodificationsinhouseholdcomposition,livingarrangements,changesinpersonalattitudes,reductionin(andpostponementof)childbearing,andtheincreaseddiversityinthepopulation(Zmudetal.,2014).Theincreaseddiversityofthepopulation,inparticular,maycontributetodecreasingtheaverageVMTpercapitaofyoungergenerations:BlumenbergandSmart(2014)foundthat(similarlytootherstudies)immigrantsaremorelikelytocarpoolthanthosebornintheUnitedStates,eveniflargedifferencesexistdependingontheoriginoftheindividualsandtheplacewheretheywereraised.BlumenbergandSmart(2014)analyzed2000censusdataand2001travelsurveydata,andfoundthatthepercentageofforeign-borninacensustractispositivelycorrelatedwithcarpoolingrates.Shin(2016)examinedethnicenclavesinthe2012-2013CaliforniaHouseholdTravelSurvey,andfoundsimilarresults.Specifically,theauthorfoundthat

Page 13: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

5

immigrantsresidinginethnicenclaveshavehigherratesofhousehold-externalcarpoolingfornon-worktrippurposesthanimmigrantsresidingoutsideethnicenclaves.Thestudypostulatesthatethnicenclavesmayofferstrongersocialnetworks,whichmayaffectmodechoice(Shin2016).Millennials’behaviordiffersfromthatoftheiroldercounterpartsduetoacomplexcombinationoflifecycle,periodandcohorteffects,includinglifestyle-relateddemographicchanges,suchasshiftsinemploymentrates,delaysinmarriageandchildbearing(PewResearchCenter2014),andshiftsinattitudesanduseofvirtualmobility,whicharebelievedtobemorespecificoftheircohort(assuggestedbyMcDonald,2015).IntheiranalysisofNationalHouseholdTravelSurvey(NHTS)data,Polzinetal.(2014)showedthatmillennialsexhibitdifferenttravelbehaviorthanthepreviousgenerationsatthesameage–specifically,20-34yearoldsin2001drovemoremilesperyearthan20-34yearoldin2009-andidentifiedseveralfactorssuchasresidentiallocation,race,employmentandeconomicstatus,livingarrangements,licensurestatus,amongothers,thatareexpectedtoinfluencemillennials’mobility.McDonald(2015)alsoanalyzedNHTSdataandhighlightedthatallAmericanstraveledlessfrom1995to2009,butmillennialtraveldecreasedthemost.Thestudyindicatedthatdemographicshiftstypicalofthe18to34agegroupcouldexplain10-25%ofdifferencesobservedintravelpatterns.Theauthorconcludedthatanadditionalportion(35-50%)couldbeexplainedbyothervariablessuchaschangingattitudesorvirtualmobility,evenifshecouldonlyinferthisasNHTSdatadonotcontaininformationonthesevariables.Theremainingpercentageisattributedtothegeneraldeclineintravelacrossallgenerations(McDonald2015).Moderntechnologicalinnovationsfurthercontributetoreshapingtransportation.TheadoptionofICT,e.g.onlineshopping,telecommuting,etc.,isattributedanimportantroleinreshapingindividuals’relationshipswiththeuseoftravelmodesandorganizationofactivities(cf.Mokhtarian,2009;CircellaandMokhtarian,2017;Circellaetal.,2016a).Sharedmobilityserviceshavefurtherreshapedtransportationthroughtheintroductionofoptionsthatgiveusersincreasedmobilityandaccessibilitywithoutincurringthecostsofowningavehicle.Sharedmobilityservicesrangefromcar-sharingservices,includingfleet-basedservicessuchasZipcarorCar2Goandpeer-to-peerservicessuchasTuro,toridesharingservices,includingdynamiccarpoolingsuchasCarmaandon-demandrideservices(alsoknownasridesourcing)suchasUberandLyft,andbike-sharingservices.Sharedmobilityservicesmodifyanumberofkeyfactorsrelatedtotraveldecisions,includingtravelcost,convenienceandsecurity(Tayloretal.2015).Theadoptionoftheseservicescanaffectthelevelofautoownershipofahousehold,andcontributetoshiftingindividuals’preferenceawayfromcarownershipwithpotentialsizableimpactsondailyschedules,lifestyles,andevenresidentiallocation.Notsurprisingly,earlyadoptersofsharedmobilityservicesarepredominantlywell-educatedyoungindividualswholiveinurbanareas(Rayleetal.2014;Tayloretal.,2015;Bucketal.;2013).Theseservicesareparticularlypopularamongmillennials,whoareheavyusersofICTdevicesandaremoreopentothesharingeconomy(Polzinetal.,2014;Zipcar2013;Bucketal.,2013;Rayleetal.,2014).

Page 14: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

6

Thereiscontinuedinterestininvestigatingmillennials’travelpatterns(andthereasonsbehindtheobserveddifferenceswiththeiroldercounterparts),alsoinconsiderationofthelargesizeofthissegmentofthepopulation,andthelikelylargeeffectsthattheirchoiceswillhaveonfutureconsumerexpenditures,demandforhousing,andtraveldemand.Inarecentanalysisof1990,2001,and2009NHTSdata,Blumenbergetal.(2016)foundthattherewasasignificantdropindriving(PersonalKilometersTraveled-PKT)inthe2000s.Theyexaminednumerousfactorsincludingdrivers’licensure,employment,webuse,andtransitionstoadulthood,includinganumberofvariablestodescribestageoflife,suchaslivingwithparents,etc.TheauthorsfoundnostatisticalrelationshipamongthemajorityofthesevariablesandPKT.However,andnotsurprisingly,employmentwasconsistentlyandpositivelyassociatedwithPKT.TheyconcludedthatdecliningemploymentduringtheGreatRecessioncontributedsignificantlytothedeclineinyouthtravelbetween2001and2009(Blumenbergetal.2016).Duringthattime,unemploymentmorethandoubled.Theauthorsfoundthattheeffectofemploymentwas32%greateramongolder(ages27–61)thanyounger(ages20–26)adults.TheyinterpretedtheseresultstosuggestthateconomicfactorswereattherootofthedeclineinpersonaltravelintheU.S.duringthe2000s.Garikapatietal.(2016)analyzedolderandyoungermillennials,andfoundthatoldermillennialsarebecomingincreasinglyliketheirGenXcounterpartsatasimilarage.However,itisunclearifmillennialswilladapttothesametravelpatternsofthepriorgenerationsoriflingeringdifferenceswillremainintheirtravelandtimeusepatterns.Theissuehasimportantplanningimplications.Forexample,realestatesalesdatasignalanincreaseinthenumberofmillennialsmovingtomoresuburbandevelopments,evenifwitha“delayeffect”associatedwiththelatertimeinwhichmembersofthesegenerationsestablishnewhouseholds.Ifsuchatrendexpandsinfutureyears,withanincreaseinsuburbanliving,itislikelytobringimportantconsequencesintermsnotonlyofthedemandforhousing,butalsooffuturetraveldemand,andtheuseofvarioustransportationmodes.Ontheotherhand,thereportedpreferencesofmillennialsforurbanlifestyleshasbeenpromptinghopesforafurtherincreaseinthepopularityofcentralurbanneighborhoods,whichhavealreadygonethroughaprocessofprogressiverenewalandregenerationduringrecentyears(Wachs,2013).Millennials,withtheirlowerper-capitaVMTandautoownershiparecreditedbymanyasimportantactorsthatcanhelpplanningagenciesandregulatorsreachthemilestonesofreductioninVMTandGHGemissionsfromtransportationoftenincludedaspartofplanningprocessesalsoastheresultofenvironmentalregulations(asinthecaseoftheSustainableCommunityStrategiesmandatedinCaliforniabytheSenateBill375andrelatedregulations).Thisgoalisalsomirroredinthechangeshappeningintherealestatetrends,andchangingregulationsinmanyjurisdictions,forexamplethroughtherevisionofparkingrequirementsfornewdevelopmentsandchangesinzoningregulations.Further,millennialsaremorelikelytoliveinmulti-generationalhouseholdsthanpreviousgenerationsatthesameage,withadditionalimplicationsintermsoftheiraccesstoprivatevehiclesownedbyahousehold,andcoordinationoftravelpatternswithotherhouseholdmembers.FryandPassel(2014)foundthatby2012,24%ofyoungadultslivedinmulti-generationalhouseholds,upfrom19%in2007,and11%in1980.Thisshareishigheramongmen(26%ofmale25-34yearoldsliveinmulti-generationalhouseholds,comparedto21%of

Page 15: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

7

women).Theauthorsconcludethatthismaybeamanifestationofthedelayedentrytoadulthood(alongwithlatermarriageandchildbearing)(Fry&Passel2014),whichareallfactorsassociatedwithpotentialimpactsonindividualtravelbehavior(i.e.duetothedelayedlifecycleeffects).InastudyofAustraliandriver’slicensingtrends,DelboscandCurrie(2014)concludedthatfull-timeemploymentandthepresenceofchildreninthehouseholdwerestrongpredictorsoflicensingstatus,withhigherlicensingratesamongyoungadultswhoworkfull-time(inparticulariftheyhavechildren),comparedtopart-timeworkersandstudents.Theypositthatchangesinlivingarrangementsandstateoflifemaycausereducedorpostponedlicensureofyoungadults(Delbosc&Currie2014).Thesameseemstobetrueforcarownership:inanexaminationofmillennialcarownership,KleinandSmart(2017)usedeightwavesofdatafromthePanelStudyofIncomeDynamics.Theyfound,consistentwithpreviousliterature,thatyoungadultsownfewercarsthanpreviousgenerationsatthesamelifestage.Inparticular,theauthorsfoundthateconomicallyindependentyoungadults(i.e.thosethathavealreadyestablishedtheirownhousehold)ownmorecarsthanexpectedfortheirincomeandpersonalwealth,thereforepositingthateconomicfactorsarethemainoneslimitingyouthcarownership.Asyoungadultsbecomeeconomicallyindependentfromtheirparents,theircarownershipratestendtoincrease.Thisconclusionsseemstoimplythatrecentlyobserved“peakcar”trendmayreverseinfutureyears,themoretheeconomyrecoversandmoremillennials“leavethenest”(Klein&Smart,2017).Youngergenerationsmayprefermultimodalmobility,aswell.Vijetal.(2015)usedcross-sectionaltraveldiarydatafromindividualsintheSanFranciscoBayAreain2000and2012todevelopalatentclassmodeloftravelmodechoicebehavior.Theirfindingsindicateshiftsintheregiontowardsgreatermultimodality.Duringtheobservedperiod,motorizedvehiclemodesharesdecreasedfrom85%in2000to81%,whiletheproportionofthepopulationthatonlyconsidersprivatevehiclewhendecidinghowtotraveldeclinedfrom42%to23%.Theauthorsofthestudyconcludedthatchangesineconomicandsocialfactorsandlevelofserviceofdifferenttravelmodeshadamarginaleffect,butdidnotaccountfortheentiredeclineinvehiclemodesharesobservedfrom2000to2012.Further,theyfoundthatthemodalshiftsexistacrosstheentirepopulation,andwerenotlimitedtoanyonegeneration(Vijetal.2015).Manyofthetopicsmentionedaboveareinvestigatedaspartofthisstudy.Understandingthefactorsaffectingmillennials’choices,andtheirpotentiallong-termimpactsontraveldemand,isextremelyimportanttoplanningprocessesandpolicy-making.Still,previousstudieshavebeenlimitedbyeither(1)thelackofinformationonspecificvariables,suchaspersonalattitudesortheadoptionofnewtechnologiesandemergingmobilityservices,forstudiesbasedonNHTSorotherhouseholdtravelsurveysatthestatewideormetropolitanplanningorganization(MPO)level;or(2)theuseofnon-randomsamples,suchasconveniencesamplesdrawnfromspecificsegmentsofthepopulation,e.g.universitystudents.Thisstudyhasbeendesignedwiththeaimofovercomingsomeoftheselimitations.

Page 16: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

8

TheCaliforniaMillennials’DatasetThisstudybuildsonalargeresearcheffortundertakentoinvestigatetherelationshipsamongmillennials’residentiallocation,individualattitudes,lifestyles,travelbehaviorandvehicleownership,theadoptionofsharedmobilityservices,andtheaspirationtopurchaseanduseavehiclevs.useothermeansoftransportationinCalifornia,whichwasdesignedtoovercomesomeofthelimitationsfrompreviousstudies.Duringthepreviousstageofthisproject,whichwasalsoprimarilyfundedbytheNationalCenterforSustainableCaliforniaandCaltrans,arichdatasetwascollectedinfall2015withacomprehensiveonlinesurveythatwasadministeredtoasampleof2400Californiaresidents,includingmillennials(i.e.youngadults,18-34,in2015)andmembersoftheprecedingGenerationX(i.e.middle-ageadults,35-50).WeusedaquotasamplingapproachtorecruitrespondentsfromeachofthesixmajorregionsofCaliforniaandthreedominantneighborhoodtypes(urban,suburbanandrural),whilecontrollingforsociodemographictargetsincludinghouseholdincome,gender,raceandethnicity,andpresenceofchildreninthehousehold.TheresultistheCaliforniaMillennialsDataset,anunprecedenteddatasetwhichcontainsinformationontherespondents’personalattitudesandpreferences,lifestyles,adoptionofonlinesocialmediaandinformationandcommunicationtechnology(ICT),residentiallocation,livingarrangements,commutingandothertravel-relatedpatterns,autoownership,awareness,adoptionandfrequencyofuseofthemostcommonsharedmobilityservices(includingcar-sharing,bike-sharing,dynamicridesharingandon-demandrideservicessuchasUberorLyft),propensitytopurchaseanduseaprivatevehiclesvs.useothermeansoftravel,majorlifeeventsthathavehappenedinthepastthreeyearsandthatmighthaveinfluencedthecurrentlifestyles,residentiallocationandtravelbehavior,environmentalconcerns,politicalideasandsociodemographictraits.Theanalysisoftherichamountofdatacontainedinthisdatasetallowsustoaddressanumberofresearchquestionsthathavereceivedattentioninrecentyearsinthescientificandplanningcommunity.TheremainderofthissectionprovidessummaryinformationontheCaliforniaMillennialsDataset,andondatahandling,cleaningandtransformationthatwerecarriedouttoexpandandintegratethedatasetwithadditionalinformationavailablefromotherdatasources,inordertodeveloptheanalysisofinterestforthisresearch.Formoredetailedinformationonthesurveycontent,datacollectioneffortandsamplingstrategybehindthecreationoftheCaliforniaMillennialsDataset,pleaserefertothePartIprojectreport(Circellaetal.,2016b).Thedatacollectionprocesswasspecificallydesignedtoinvestigatetherelationshipsassociatedwiththebehavioralprocessesandmobility-relateddecisionsofyoungadults(millennials),andtoinvestigatetheimpactthatseveralgroupsofvariables,includingchangesinlifestyles,sociodemographictrendsandtheadoptionofemergingmobilityservices,haveonthetraveldecisionthisdynamicsegmentofthepopulation.Inaddition,thepresenceofacontrolgroupcomposedofmembersoftheolderGenerationXisusefultoallowcomparisonsacrossgenerationsinthestudy,usingthesamemethodologiesfordatacollectionandselectionofrespondentsfortheentiresample.

Page 17: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

9

ThesurveyusedtocollecttheoriginalinformationincludedintheCaliforniaMillennialsDatasetincludes11sections,whichcollectedinformationonvariablesrelevantfortheanalysisofmillennials’mobilityandotheremergingtransportationtrends:

a. Individualattitudesandpreferences,measuredthroughtheagreementwithagroupof66statementsonafive-levelLikertscale,for20dimensionsincludingsocialhabits,lifestyles,adoptionoftechnology,environmentalconcerns,exercise/physicalactivity,individualism,materialism,timeorganization,etc.;

b. Useofonlinesocialmedia(Facebook,Twitter,amongothers),andadoptionofICTdevicesandservices,e.g.frequencyofuseofsmartphoneappstobooktransportationservices,purchasetickets,checktrafficconditions,ordecidewhatmodeoftransportationtouse;ownershipandregularuseofvariousICTdevices;adoptionandfrequencyofuseofe-shopping;

c. Residentiallocationandlivingarrangements,includingtheself-reportedcharacteristicsoftheneighborhoodwheretherespondentslive,detailedaddress(orclosesttwo-streetintersectionnearthehomeaddress),informationabouttenancy,yearstherespondenthaslivedatthataddressed,andinformationabouttheotherpeoplewholivewiththerespondents(e.g.partner,parents,children/grandchildren,siblingsorotherrelatives,eventualroommates/flatmates,etc.);

d. Employmentandwork/studyactivities,includingdetailedinformationaboutoccupation,typeofjob(s),fieldofoccupation,studentstatus,workschedule,numberofhoursworkedintheaverageweekforthemainoccupationandforanyvolunteeringactivities;

e. Transportationmodeperceptions,includingperceptionsofdriving,publictransportationandactivemodes(walking,biking).Theseperceptionsincludecomfort,reliability,safety,cost,privacy,andabilitytomultitaskwhileusingthesemodesoftransportation,amongothers;

f. Currenttravelchoices,includingdetailedinformationonthetypicalusageofvariousmeansoftransportation(privatevehicle,carpool,shuttle,publictransportation,bike,etc.)forbothcommutesandleisuretrips.Thissectionalsocollectedinformationontheself-reportedcommutedistanceandaveragetimespentcommuting,thelocationofmaincommutedestination(workorschool),theactivitiesconductedwhiletraveling,andtherespondent’slongdistancetravelpatterns(measuredintermsofthenumberoflongdistancetripsmadebydifferenttravelmodesforeitherbusinessorleisurepurposes,duringtheprevious12months).

g. Awareness,adoptionandfrequencyofuseofthemostcommonsharedmobilityservices(includingcar-sharing,bike-sharing,dynamicridesharingandon-demandrideservicessuchasUberorLyft);thesectioncollectedinformationaboutthesharedmobilityservicesthatareavailablewheretherespondentlives(e.g.peer-to-peercar-sharingsuchasTuro,fleet-basedcar-sharingsuchasZipcar,on-demandrideservicessuchasUberorLyft,etc.)andhowoftentherespondentusestheseservices.WealsocollectedinformationonwhytherespondentusedUber/Lyft,howthisimpactedtheiralternative

Page 18: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

10

modechoice,e.g.thedecisiononwhethertousepublictransportation,orchosenottodrive,andwhateventuallylimitsorpreventstheuseofon-demandrideservices.

h. Driver’slicensingstatusandvehicleownership,includinginformationonwhetherarespondenthasadrivers’license,thetypeoflicensetheyhave,andthelegalagetoobtainalicenseintheplacewheretherespondentgrewup.Thissectionalsoincludesquestionsonthepercentoftimeacar(and/ormotorcycle)isavailabletotheindividual,thenumberofvehiclesownerbytheindividual’shousehold,anddetailedinformation(year,makeandmodel)ofthevehiclethatisusedmostoften.Thissectionincludeddetailedquestionsonthefactorsbehindtherespondents’decisiontopurchasethevehicle(usedornew).Finally,thissectioncollectedinformationonthenumberofmilesarespondenttravelsperweekbycarandbybike,thetypeofparkingavailableattheplaceofresidence(ifany),andiftherespondenthasapublictransportationpass.

i. Previoustravelbehaviorandresidentiallocation(andinformationonthemajorlifeeventsfromthepastthreeyears):thissectioncollectedinformationaboutthelifeeventsfromthepastthreeyears(e.g.movingtoanewcityorstate,buyingahome,beginningstudy,movinginwithapartner,havingchildren,etc.).Thissectionalsocollectedinformationonwhyaparticipantmayhavemovedandtheimpactofseveralfactorsonthischoice(e.g.birthofachild,qualityoftheschooldistrict,housingprice,parkingavailability,easeofwalkingandbikingetc.).Thissectionalsocollectedinformationonhowmuchparticipantstravelbyeachmodenowcomparedtothreeyearsago.

j. Expectationsforfutureevents(andpropensitytopurchaseanduseaprivatevehiclevs.touseothermeansoftravel),includingiftheparticipantsexpects/planstomove,and/orforeseechangesinthehouseholdcompositionintheirjobsorschooltheyattend.Thisincludesdataonhowparticipantsexpecttotravelinthreeyearsfromnow,comparedtohowtheycurrentlytravel,bymode.Finally,thesectioncollectedinformationontheinterestinpurchasinganewvehicle(andthetypeofvehicletheywouldconsiderpurchasingorleasing)and/orinjoiningorleavingacar-sharingprogram.

k. Sociodemographictraits,includinggender,age,USstateorforeigncountrywheretheindividualwasraised,politicalviews,householdsizeandcomposition,individualandhouseholdincome,educationlevel,parents’education,andnumberofdriversinthehousehold.

Duringthesurveydesign,weengagedseveralstakeholdersandworkedwithcolleaguesatotherresearchinstitutions,Californiastateandlocalagencies,andotherpartnerorganizations,toobtainfeedbackonthesurveycontentandimprovethesurveytool.Weextensivelypretestedthesurvey,andtriedtobalancethetrade-offbetweenthecomplexityofthecontentofthesurvey(andtheamountofinformationthatiscollected)andthetimerequiredtocompletethesurvey.WeadministeredthesurveytoasampleofmillennialsandmembersofGenerationXinCalifornia.Weusedaweb-basedopinionpaneltoinvitemembersofthesesegmentsto

Page 19: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

11

completethesurvey,andusedaquotasamplingapproachtoensurethatenoughresponseswereincludedfromeachgeographicregionofCaliforniaandneighborhoodtypewheretherespondentlives(classifiedinpredominantlyurban,suburbanandruralareas).SociodemographictargetswereusedtomakesurethatthesamplemirroredthecharacteristicsoftheCaliforniapopulationonfivekeysociodemographicdimensions:sex,age,income,raceandethnicity,andpresenceofchildreninthehousehold.Forthepurposesofthisstudy,wedividedCaliforniainsixmajorregions:

• MTC–MetropolitanPlanningOrganization(SanFranciscoBayArea);• SACOG–SacramentoAreaCouncilofGovernments(Sacramentoregion);• SCAG–SouthernCaliforniaCouncilofGovernments(LosAngeles/SouthernCalifornia);• SANDAG-SanDiegoAssociationofGovernments(SanDiego);• CentralValley(eightcountiesinthecentralSanJoaquinValley);and• NorthernCaliforniaandOthers(restofstatenotincludedinthepreviousregions).

Atotalof5,466invitationsweresentout,and3,018completecaseswerecollected.Thehighresponserateof46.3%isnotsurprisingconsideringthedatacollectionmethodusedforthisproject,andthehigherpropensityofopinionpanelmemberstorespondtosurveyinvitations.Afterexcludingseverelyincomplete,inconsistentorunreliablecases,afinaldatasetthatincludedapproximately2,400validcaseswasusedtocomputeinitialdescriptivestatisticsandotheranalysesreportedinthePartIreport(Circellaetal.,2016b).Whilethesamplingmethodusedtorecruittheparticipantsforthisstudy(basedontheuseofanonlineopinionpanel)andtheuseofanonlinesurveymightrepresentapotentialsourceofbiasfortheresearch,andcautionshouldbeusedingeneralizingtheresultsfromthestudytotheentirepopulationofCalifornia,theuseofthesamemethodologyfortherecruitmentofbothmembersofthemillennialgenerationandoftheprecedinggenerationXensuresinternalconsistencyinthecollectionofthedataandcreationofthedataset.Inotherterms,ifanysamplingandresponsebiasesaffectthestudy,itisreasonabletoexpectthatthesimilarbiasesaffectboththemillennialsandGenerationXsubsamples.Forthisreasons,evenifeventualbiasesarepresentinthedatacollectionandsamplingapproachusedfortheresearch,thecomparisonsbetweentheobservedbehaviors,andrelationships,betweenmillennialsandGenXerspresentedinthisreportremainvalid.ThedatacollectioneffortwasdesignedasthefirststepofalongitudinalstudyoftheemergingtransportationtrendsinCalifornia,designedwitharotatingpanelstructure,withadditionalwavesofdatacollectionplannedinfutureyears.Theresearchteamiscurrentlyworkingwiththefundingagency,inordertodefinetheplanforthefuturecomponentsofthelongitudinal(panel)study,alsothroughtheintegrationoftheinformationcollectedwiththissurveywithadditionaltraveldiariesandtraveldatacollectedwithGPS-basedsmartphoneapps.Further,infuturestagesoftheresearch,weplantoexpandthedatacollectionalsothroughotherchannels,alsothroughthecreationofapaperversionofthesurvey,inordertoexpandthetargetpopulationforthestudy,andreachspecificsegmentsofthepopulation,e.g.elderlyor

Page 20: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

12

peoplethatarenotfamiliarwiththeuseoftechnologyorwhodonothaveeasyaccesstotheinternetandwouldnotlikelycompleteanonlinesurvey.Also,weareconsideringcreatingaversionofthesurveyinSpanish,inordertobetterreachtheCaliforniapopulationofLatinosandincreasetheresponserateamongtheHispanicminority.DataCleaningandRecodesInordertoenforcestrictqualitycontrolinthecollectionofrespondents,wedevisedseveralmeasurestoidentifyandremoveproblematicorinconsistentcasesfromthedataset.Amongthestrategiesthatweredevelopedforpurposesofqualityassurance,weusedacommonqualityassurancepracticeintheformoftwotothree“trap”questions(dependingontheversionofthesurveythatwasadministeredtotherespondent)thatwereincludedinvarioussectionsofthesurvey.FurtherdetailsaboutthetrapquestionsthatwereusedandthestrategiesthatwereusedtoidentifyinconsistenciesinthedatasetcanbefoundinthePartIprojectreport(Circellaetal.2016b).Inadditiontotheuseoftrapquestions,wecheckedtheconsistencyofresponsesthroughoutthesurveythroughtheapplicationofseveralcriteria.Theconsistencychecksthatwereusedalsoincludedverifyingthespeedwithwhichrespondentsansweredthesurvey.Forexample,weremovedindividualswhofailedatrapquestionandalsocompletedthesurveyinaveryshorttime(below20minutes)asasignoflackofattentionduringthecompletionofthesurvey.Theaverageresponsetimeforthissurveywasapproximately35minutes.Therefore,itwouldhavebeenextremelydifficulttocompletethesurveyinlessthan20minutes.Additionalcriteriathatwereusedduringtheprocessofdatacleaningandrecodingarediscussedinthesub-sectionsbelow.Thesecriteriaincludedcheckinginternalconsistencyofacase,analyzingsurveyresponseoutliers,andinconsistenciesbetweentheinformationreportedbytherespondentinthemainbodyofthesurveyandinthescreenerfromtheopinionpanel.1InternalconsistencyAspartoftheinternalconsistencychecks,weidentifiedandcarefullyreviewedcasesthatwereconsideredsuspiciousaccordingtooneormoreofthefollowingcriteria:

• Flatliners:Individualswho“flatlined”oneormoresectionsthathadconflictingstatements(e.g.respondentswhoansweredyestobothstatements:“Iexpecttomoveinthenextthreeyears”and“Iexpecttostayinmycurrenthouseinthenextthreeyears.”)

• Locationalconsistency:Forexample,individualswhoprovidedthesameaddressforworkandhome,thoughtheyindicatedthattheydidnottelecommute,orindividualswhoperceivedneighborhoodtypeasextremelydifferentfromtheobjectivemeasuresthatweredeterminedusinggeocodedvaluesforthehomeaddress.

1Theopinionpanelusedashortscreener,whichcontainedonlyninequestions,torecruitandselectparticipantsforthestudy.

Page 21: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

13

• Travelpattern:Weassessedmodeavailabilityforcommuteandleisuretripsaccordingtothereportedlocation,tripdistanceandtimeofthecommutingtrips,andthroughthecomparisonofthegeolocatedworkandhomeaddresses.Wealsoevaluatedthecasesthatreportedfrequentuseofmultiplemodes,andinconsistencyinthereportedmulti-taskingactivitiesduringthemostrecentcommutetrip.

• Useofemergingtransportation:Respondentswhoreportedthattheyusedservicesthatarenotavailableintheareaswheretheylive(thesurveyexplicitlyaskedrespondentswhethertheyusedtheserviceintheirhometownorwhiletravelingawayfromhome),orrespondentswhoreportedthattheyusedmultipleserviceswithveryhigh(andunrealistic)frequencyovershortperiodsoftime(e.g.respondentthatusedZimride,Turo,ZipcarandUberveryfrequently,especiallyiflocatedinlocationswheretheseservicesarenotlargelyavailable).

• Householdcomposition:Severalquestionsinthesurveyaskedinformationrelatedtothehouseholdcompositionandlivingarrangement,allowingtheresearcherstoestablishwhetherthereportednumberofchildrenandnumberofadultsinthehousehold,andtheirageranges,areconsistentwiththeinformationreportedabouttheotherindividualsthatliveinthehousehold(intheprevioussectionCofthesurvey)

Casesthatfailedoneormorecriterialistedabovewere,inmostcases,removedfromthedataset,unlesssomevalidreasonsfortheinternalconsistencywereidentified.ResponseoutlierWereviewedcasesthatposeproblemsrelatedtooneormoreofthefollowingcriteria:

• Dailyactivitypatterns:individualswhoreportactivitiesthatareimplausibleorimpossible(e.g.watchingTVfor24hoursinoneday).

• Longdistancetrips:Individualswhoreportedextremelyhighnumberoflongdistancetripsforeitherbusinessorleisuretrips(over100miles).

• MoneyspentonUber/Lyft:IndividualswhoreportspendingveryhighmonthlyamountofmoneyonUbercomparedtotheself-reportedfrequencyofthisservice.

• Numberofcars:Respondentswhoreportveryhighorverylownumberofcarscomparedtotheirhouseholdsizeandstructureandthereportedcommutepattern(e.g.individualsthatreportthattheytraveldrivingaloneinacaronadailybasis,butthenreportthattheyliveinazero-vehiclehousehold).

• Vehiclemilestraveled:IndividualswhoreportedillogicalaverageweeklyVMTforcommutesandtravelpatterns(e.g.individualsthatlikelyreportedannualVMT,bymistake,insteadoftheweeklyVMT,orthatreportedzeroVMT,butthenreportedthattheydrivealonetowork/schoolintheircommutepattern).

Theinformationassociatedwiththecasesidentifiedthroughoneofcriteriaabovewaseitherremovedfromthedataset,orrecodedaccordingly(e.g.somevariablevalueswererecodedto“missing”),dependingontheseverityoftheproblemsthatwereidentified.

Page 22: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

14

InconsistencybetweentheSurveyandScreenerQuestionsWealsoidentifiedinconsistenciesbetweentheinformationreportedinthesurveyandtheinformationthatwasreportedwhenansweringthequestionsthatwereproposedinthescreenerusedbytheonlinesurveycompanytopre-screenrespondentsduringtherecruitmentofparticipantsforthestudy.WedesignedthescreenertoensurethatasamplethatisasrepresentativeaspossibleofthepopulationinthestateofCaliforniacouldbeassembledforthisstudy.Thescreenercollectedinformationonthefollowingvariables:gender,agegroup,Hispanicorigin,race,householdincome,Zipcodeoftheplaceofresidence,neighborhoodtype,presenceofchildreninthehousehold,andnumberofchildreninthehousehold.Inparticular,wecheckedtheconsistencyforthefollowingvariables:

• Gender:wecomparedthescreenerdatawiththesurveydata.• Agegroup:Therewereseveralcasesforwhichtheagewasnotconsistentwiththe

reportedgroups:inthiscasewecheckedthescreeneragegroupswiththesurveyresponse.

• Neighborhoodtype:Wecomparedtheperceivedandgeocodedmeasuresofneighborhoodtype(suburban,urban,rural)andindividualreviewedcasesthathaddifferencesinthereportedneighborhoodtype,toidentifythereasonsforthedifferentinformation.

• PresenceofChildren:WeassessedthepresenceofchildreninthehomegiventheresponsesinsectionCandsectionKofthesurvey,andcomparedthemtotheinformationprovidedinthescreener.

Inmostcases,theinconsistenciesaboveledtorecodingthescreenerdata,giventhatthesurveyinformationwasconsideredmoreaccurate,e.g.thescreenercansometimesbefilledbyothermembersofthehousehold.However,caseswithmoresevereinconsistencieswereremovedfromthesample.Werecodedsomeresponsesonacasebycasebasis,reviewingallanswersprovidedbyarespondent.Insomesituations,werecodedavariableto“missing”value,whentheinformationaboutthatvariablecouldnotbeassessedwithcertainty.Inthecaseofthescreenerinconsistencieswerecodedeitherthesurveyorthescreenerdependingonthecase.Alistofrecodeswaspreparedandimplementedinthefinaldataset.Afterassessingthecaseswhichpresentedsomeinconsistenciesorotherreasonsfornotbeingconsideredreliable,weretained2155casesinthedatasetusedfortheanalysesinthisreport,fromthemorethan3000casesthatwereoriginallycollected(andapproximately2400casesthatwereusedfortheinitialanalysesinthePartIreport).GeocodingTomaketheCaliforniaMillennialsDatasetrichwithvariousinformationfromexternaldatasources,wefirstgeocodedtheresidential,school,andworkplaceaddressesofindividualrespondentsbyemployingoneofthereliablegeocodingmethods,theGoogleMapsapplication

Page 23: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

15

programminginterface(API).Othergeocodingmethodswerealsoconsidered,includingtheESRIDesktopArcGISgeocodingtoolboxandtheESRIArcGISonlinegeocodingtool.Thesetoolsweretestedandusedininitialcomponentsofthegeocodingprocess.However,theywerenotusedinthefinalgeocodingprocess,becauseofsomelimitationsthatmadethemnotwellsuitedforthisproject.Inparticular,theDesktopArcGIStoolboxneedsastreetnetworkinaspecificformasaninputforgeocoding,andmostusersusetheUSCensustopologicallyintegratedgeographicalencodingandreferencing(TIGER)AddressRange-Featureshapefileastheinput.AlthoughtheUSCensushaveregularlyupdatedthisshapefile,itisfarfrombeingperfect.Forexample,thefirstandlaststreetnumbersofstreetsegmentsinthisfileareoftennotrecentlyupdated.Moreover,becauseArcGISisnotasearchenginesuchasGoogleandBing,ifaddressesaremisspelled,itsgeocodingoutcomesarenotasgoodasthosefromonlinesearchenginesthatoftensuccessfullyfindfulladdressesalsoincaseofpartialonesbasedonprevioussearchesandselectionsfromotherusers.Thispropertyalsocomeswithsomedisadvantages,though,astheGoogleMapsAPImightsometimesreturnwrongaddressesastheresultofthepredictionsoftheirsearchengine.Still,inthisproject,itwasfoundtobepreferabletousetheGoogleMapsAPI,withsomeadditionalqualitychecksthatwereperformedbytheresearchteamasapost-process,toverifythattheaddressgeocodedbyGooglereasonablymatchedtheoriginaladdressprovidedbytheuser.AsfortheArcGISonline,althoughESRIclaimsthatitsgeocodingoutcomesaremoreaccuratethanthoseobtainedbyemployingtheUSCensusshapefiles,ESRIdidnotexplicitlyrevealthecharacteristicsoftheirgeodatabase.Afterintensiveexperimentations,wefoundthattheoutcomeoftheArcGISonlinewasnotdiscernablybetterthanthatoftheDesktopArcGIStoolbox.Somerespondentsreportedinaccurate,partial,anderroneousaddresses,butmanyoftheproblematicaddressesappearedtobeformattedcorrectly,sotheresearchteamwasabletocleanandgeocodetheseaddressesthroughamultipleiterationgeocodingprocess.Fourtypesofaddresseswereidentifiedinthedataset,basedonthetypeofinformationprovidedbytherespondents:

1. Fulladdresseswithstreetnumbers;2. Intersectionsoftwocloseststreets;3. One-streetaddresses;and4. Onlythenameofcitiesand/orZIPcode2.

Eachtypeofaddresspresentsuniquechallengesthataffectthegeographicaccuracyandprecisionofgeocodes.Althoughmisspellsandtheomissionofsomeinformationinthestreetnamesareusuallyaneasyfix,someofthereportedfulladdressesdidnotexist(i.e.,thestreetnameisreal,butthereportedstreetnumberisnotfoundonthatstreet).Moreover,wefoundanontrivialnumberofcaseswithtwonearbystreetswhichactuallydonotcrosseachother:notallpeopleareabletocorrectlyremembertwointersectingstreetsnearbytheirresidential

2ThesurveyrequiredeachrespondenttoreportavalidZIPcode.Thus,respondentsthatdidnotfeelcomfortableaboutprovidingadditionalinformationabouttheiraddress,ataminimumprovidedinformationthatallowedtheresearchteamtoidentifythecityandZIPcodeinwhichtheylive.

Page 24: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

16

location,somerespondentsreportedtwostreetsthatareactuallyparallel(andsometimesevenfarfromeachother).Inaddition,specificruleshadtobedefinedtotreatcasesinwhichthesurveyparticipantsreportedonlyonestreetinsteadoftheirresidentialaddress.TheresearchteamhadtodevelopasetofrulestoassignthemostlikelyCensustracttotheserespondents’residential,study,andworkaddresses.Lastly,caseswithonlyinformationabouttheZIPcodehadthelowestqualityofinformation:ZIPcodeareasareoftenlargeenoughtocovervarioustypesofneighborhoods(e.g.theycanincludebothsuburbanandurbanneighborhoods).Asanonlinesearchenginethatisspecializedtoreturnreliableoutcomesevenwithincompleteandpartiallyincorrectkeywords,GoogleMapsAPIworksononeofthemostupdatedgeodatabasesandproducesarichsetofinformationonthequalityofgeocodes,whichuserscanusetoexaminegeocodingoutcomes.BecausethegeodatabaseofGoogleMapsAPIisincorporatedwiththesatelliteimagesofGoogleMaps,GoogleMapsAPIproducesaresultfromadirectsearch,insteadofgeographicreferencingbasedonthefirstandlaststreetnumbersofstreetsegments(whichishowtheDesktopArcGIStoolboxandtheonlineArcGISwork).Moreover,foreachquery,GoogleMapsAPIreturnsaddressesthatitfindsfromitsgeodatabaseandtypesofgeocodingthatituses:thus,GoogleMapsAPIpresentstwowaysofexaminingthequalityofageocode.First,userscancompareinputandoutputaddressesanddeterminehowsimilartheoutputaddressfromGoogleistotheinputaddress(alsoincaseofincompleteandpartiallyincorrectaddresses).Inaddition,twocategoricalvariableshelpusersdeterminehowreliableindividualgeocodingoutcomesare.Table1summarizesthenumberofcasesinthedataset,bythetypeofaddressthatwasreported(andgeocoded):1,858caseshadhighlyreliableaddresses(withfulladdressortwo-streetintersections),233weremoderatelyreliable(one-streetaddresses),and64werelessreliablecases(withonlycitynamesand/orZIPcodes).Table1.TypeofAddressesGeocodedintheDataset

Qualityofgeocodingofresidences NumberofcasesFulladdressesorintersectionsofclosesttwostreets 1,858(86.2%)One-streetaddresses 233(10.8%)CitynamesandZIPcodes 64(3.0%)Total 2,155(100%)

Page 25: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

17

Figure1.DistributionofmillennialsandGenXersinthedataset,basedontheirgeocoded

residentialaddressTheoutcomesofthegeocodingofresidentialaddresseshelpedtheresearchteamdeterminethetypeofneighborhoodwheretherespondentsliveinCalifornia.ThisprojectusestheneighborhoodtypedevelopedinanotherprojectfromresearchersatUCDavis,whichanalyzedandclusteredthe8,036censustractsinCaliforniabasedonthepredominantneighborhoodcharacteristics(Salon,2015).Theprojectclassifiedeachcensustractasbelongingtooneoffivecategories:CentralCity,Urban,Suburb,Rural-In-Urban,andRural.Becausegeocodeswithone-streetaddressesandwithcitynamesandZIPcodesdonotpresenttheexactlocationsofresidences,theresearchteamvisuallyinspectedthesecasestoseewhetherornottheirneighboringCensustractsalsohavesimilarneighborhoodcharacteristics.Ifboththeidentifiedcensustractandtheneighboringcensustractsshowthesametypeofland-usepatterns,eveninthecaseoflowqualityofthegeocodedlocation(i.e.one-streetaddressesorcitynamesand

Page 26: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

18

ZIPcodes),theresearchteamwasabletoassigntheneighborhoodtypewithagoodmarginofreliability.Incontrast,ifone’sownneighborhoodtypediffersfromthatofitsneighboringCensustracts,weusedtheperceivedneighborhoodtypesthattheindividualsreportedinthesurveytodeterminewhichtypesofneighborhoodstherespondentsarelikelytolivein.Figure2summarizesthedistributionofcasesinthedatasetbyneighborhoodtype.

Figure2.Distributionofcasesinthedataset,bygeocodedresidentialaddressand

neighborhoodtype

Page 27: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

19

WeightingandRakingInordertocorrectfornon-representativenessofthesample,andreplicatethedistributionofthepopulationofMillennialsandGenerationXlivinginCalifornia,weusedacombinationofcellweightinganditerativeproportionalfitting(IPF)(Kalton&Flores-Cervantes2003).Weusedcellweightstoweighoursampleonthreedimensions–agegroup(18-24,25-34,35-44,45-50),neighborhoodtype(Rural,Suburban,Urban),andregion(CentralValley,NorthernCaliforniaandOthers,SACOG,SANDAG,SCAG,SFMTC).Thisweightingprocesscompensatesfortheeffectsofthequotasamplingprocessusedinthedatacollectionandtheintentionaloversamplingofsomeregions.Weintentionallyunderrepresentedtheresidentsofmajormetropolitanareas,mainlyLosAngelesandtoalowerextentSanFrancisco,inthedatacollection,andoversampledindividualswholiveinotherareas(ruralcountiesandlesspopulatedregions),inordertocollectenoughrespondentsforeachregion,andbuildrobustanalysesforallsubsamples.Atthetimethestudywaslaunched,weenvisionedasampleofatleast700casesselectedamongthepopulationofCaliforniamillennialsforthisresearch.Thesizeofthesamplesizewaslaterincreasedthroughtherecruitmentofadditionalparticipantsinthestudy,andalsoacontrolgroupcomposedofmembersofGenerationX,whichwasnotincludedintheoriginalscopeoftheresearch,wasadded,furtherenrichingthediversityofrespondentsinthesample.Whileanyremainingsamplingbiascanlimitthevalidityofthegeneralizationoftheresultsfromthissampletothepopulationofinterest,themethodusedinthisstudyremainsveryvalidforcomparisonsamongthetwosubsamplesofmillennialsandmembersofGenX,whowererecruitedwiththesamemethodology.Thesamplingmethodthatcontrolledforthedistributionofeachsubsampleonseveralsociodemographictraitsandtheapplicationofweightsallowustobuildrobustanalysesofthesedata.Todevelopourbaselinepopulationthatwasusedtodevelopthetargetforthecellweights,weusedtheAmericanCommunitySurvey20141-yearestimatedatapairedwithresidentialneighborhoodclassificationdatafromSalon(2015).While,theresidentialneighborhoodtypesforCaliforniacensustractswerederivedfromSalon(2015),weaggregatedthefiveneighborhoodtypesdeterminedinthatstudytothreemajorneighborhoodtypes,whereRural-in-UrbanandRuralareaswereclassifiedas“Rural”andCenterCityandUrbanareaswereclassifiedas“Urban”.Suburbanareasweretreatedas“Suburban”consistentwiththefiveneighborhoodtypeclassification.WeusedtheACSdatatobuildacrosstabulationbasedonagegroupbyregionandneighborhoodtype.ThefinalsetofcellweightsweregeneratedbycomparingthecrosstabulationofsurveyrespondentsandthepopulationofCaliforniaresidentsages18to50.Inadditiontocell-weightingonthethreedimensionsdescribedabove,weusedmultipleroundsofiterativeproportionalfitting(IPF)rakingtomirrorthedistributionoftheCaliforniapopulationonseveraladditionaldemographictargets.Thisallowedustocorrectthedistributionsinthesamplebyassigningspecificweightstooursamplebasedonsixdimensions–race,ethnicity,presenceofchildreninthehousehold,householdincome,student/employmentstatus,andsex,whichwereusedastargetsintheIPFprocess.Weused1-yearestimatesofthe

Page 28: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

20

PublicUseMicrodata(PUMS)from2015tocreatethetargetsfortheCaliforniapopulationfrom18-50(U.S.CensusBureau2014).AtotalofthreeiterationsoftheIPFmethodwasappliedinthisprocess.ForthefirstroundofapplicationofIPF,weusedthecellweightsasthestartingweights,andweightedonhouseholdincome,student/employmentstatusandsex.Theannualhouseholdincomewasclassifiedinthreebroadcategories:Low(<$35,000),Medium($35,000-$100,000)andHigh(>$100,000).Student/Employmentstatuswasclassifiedthroughafour-levelvariable,wheretheparticipantmaybeunemployed,workonly,beastudentonly,orbebothastudentandworker.ThesecondroundofIPFusedtheweightsgeneratedbymultiplyingthecellweightsandthefirstroundofIPFandweightedtheseonRaceandEthnicity.Duetoissuesrelatedtooursamplesize,weconsolidatedtheracecategoriesinthedatasetasthreemainracegroups–White,Asian/PacificIslander,andOther.ForEthnicity,weusedthetwocategoriesofHispanicandNon-Hispanic.ThethirdroundofIPFusedtheresultsofthepreviousiterationsandweightedonGenerationandPresenceofChildren.GenerationwasdefinedasGenerationY/Millennials(individualswhowere18to34in2015),andGenerationX(individualswhowere35to50in2015).Thepresenceofchildreninthehouseholdwasmeasuredwithabinaryvariable(children,nochildren).Table2summarizesthedescriptivestatisticsforboththeunweightedandweighteddataset.Thenumberofweightedcasesineachgroupmaynotsumexactlyto2155duetoroundingeffects.

Page 29: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

21

Table2.DemographicStatisticsintheCaliforniaMillennialsDataset Weighted Unweighted

Numberofcases

Percentageoftotal

Numberofcases

Percentageoftotal

Total 2155 100% 2155 100%Gender Male 1043 48.4% 876 40.6%Female 1090 50.6% 1257 58.4%Transgender 9 0.4% 8 0.4%DeclinetoAnswer 13 0.6% 14 0.6%PresenceofChildrenintheHousehold HouseholdwithoutChildren 1018 47.3% 1089 50.5%HouseholdwithChildren 1137 52.7% 1066 49.5%HHincome Prefernottoanswer 142 6.6% 158 7.3%Lessthan$20,000 167 7.7% 207 9.6%$20,001to$40,000 357 16.6% 392 18.2%$40,001to$60,000 311 14.4% 374 17.4%$60,001to$80,000 294 13.6% 356 16.5%$80,001to$100,000 194 9.0% 236 11.0%$100,001to$120,000 225 10.4% 157 7.3%$120,001to$140,000 120 5.5% 81 3.8%$140,001to$160,000 133 6.2% 75 3.5%Morethan$160,000 213 9.9% 119 5.5%Age YoungerMillennials(18-24) 473 21.9% 385 17.9%OlderMillennials(25-34) 714 33.1% 830 38.5%YoungerGenerationX(35-44) 608 28.2% 613 28.4%OlderGenerationX(45-50) 361 16.7% 327 15.2%Ethnicity Hispanic 907 42.1% 501 23.2%Non-Hispanic 1248 57.9% 1654 76.8%Race Black/AfricanAmerican 88 4.1% 98 4.5%AmericanIndian/NativeAmerican 49 2.3% 40 1.9%Asian/PacificIslander 326 15.1% 332 15.4%White/Caucasian 1269 58.9% 1399 64.9%Other/multi-racial 422 19.6% 286 13.3%Education

Prefernottoanswer 8 0.4% 8 0.4%Somegrade/highschool 44 2.0% 42 1.9%Highschool/GED 242 11.2% 278 12.9%Somecollege/technicalschool 595 27.6% 642 29.8%Associate’sdegree 232 10.8% 242 11.2%Bachelor’sdegree 710 32.9% 686 31.8%Graduatedegree(e.g.MS,PhD,MBA,etc.) 227 10.5% 197 9.1%Professionaldegree(e.g.JD,MD,DDS,etc.) 98 4.5% 60 2.8%AverageHHsize 3.24

3.20

Average#ofVehiclesintheHH 1.88

1.80

Page 30: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

22

IntegrationofAdditionalLandUseDatafromOtherSourcesKnowingthelocationofwork/schoolandhomeaddressoftherespondentsenablesustointegrateourdatasetwithotherexistingdataincludingSmartLocationDatasetpreparedbytheU.S.EnvironmentalProtectionAgency(EPA),andotherlanduseaccessibilitymeasuresincludingthewalk,bikeandtransitscoresfromotherwell-establishedsources(e.g.Walkscore.com).TheSmartLocationDatabasesummarizesnumerousdemographic,employmentinformation,andprovidesvariousstatisticalanddeterministicbuiltenvironmentindicatorsestimatedatthecensusblockgroup(CBG)level(Ramsey&Bell2014)3.Thesedemographicandlanduseindicatorswerematchedtoindividuals’residentialandwork/schoollocationbasedonthegeocodedlocationoftheself-reportedaddress.ThebuiltenvironmentalattributesthataremeasuredintheSmartLocationDatasetcanbeclassifiedintofivemaincategories:

• Densityindices:TheSmartLocationDatasetprovidesdifferentmeasureofdensity,includingpopulation,housing,activityandtotalnumberofemploymentandemploymentbytypeforeachcensusblockgroup.

• Diversityindices:Differentmeasuresoflandusediversitywereestimatedforeachcensusblockgroup,includingjobtohouseholdbalances,entropyindicesfor5-tierand8-tieremploymentcategories,employmentandhouseholdentropybasedontripproductionandattractions,tripequilibriumindex,regionaldiversity,andhouseholdworkersperjob.

• Urbandesignindices:Theseindicesestimatedvariousurbandesignmeasuresincludingstreetnetworkdensityandintersectiondensitybyautomobile,pedestrianandmultimodalfacilities.Exampleofthesevariablesarenetworkorintersectiondensityintermsofauto-orientedlinkspersquaremileineachcensusblockgroup.

• Transitindices:UsingtheGoogletransitdata(particularlythelocationoftransitstopsandtheirregularschedule),theSmartLocationDatasetprovidesdifferentmeasuresoftransitavailability,proximity,frequencyanddensity.Thetransitvariablesarecomprisedofdistancefromthepopulation-weightedcentroidtothenearesttransitstop,theproportionofblockgroupwithinaquartermileorhalfmileofatransitstop,theaggregatedfrequencyoftransitserviceperhourduringtheeveningpeakperiod,andtheaggregatefrequencyoftransitservicepersquaremile.Thesetransitmeasuresareonlyestimatedfortheareasforwhichthecorrespondingtransitagenciesprovidedtherequiredinformation.

• Destinationaccessibilityindices:Theseindicatorsaredevelopedtomeasuretheaccessibilityfromcensusblockgrouptocensusblockgroup.Thesevariablesmeasurethenumberofjobsorworking-agepopulationwithina45minutescommutebycaror

3The2010CensusTigerLine/polygonswereusedindefiningblockgroupboundaries,whichwerelatermergedwiththeinformationobtainedfromtheotherdatasetsincludingthe2010Censusdata,theAmericanCommunitySurvey,theLongitudinalEmployer-HouseholdDynamics,InfoUSA,NAVTEQ,PAD-US,TODDatabase,andGoogleTransitFeedspecification(GTFS)database.

Page 31: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

23

transitfromacertainblockgroup.Inaddition,theEPASmartLocationDatasetincludesrelativemeasuresofaccessibilityforeachcensusblockgroupbasedonthecomparisonwiththeaccessibilityofthecensusblockgroupsthatarelocatedwithinthesamemetropolitanareas.

Furthermore,byusingthelatitudesandlongitudesofallhomesandworkplaces,wecanappendadditionalvariablesthatcapturethecharacteristicsofspecificlocationsandthatareavailablefromreliablepublicandprivatedatabases.Inparticular,Walkscore.comhasbeenknownforitscompositemeasureofwalkability,the“walkscore,”whichmanyscholarshavefoundausefulvariabletounderstandrelationshipsbetweenthebuiltenvironmentandnon-motorizedtravelpatterns.Whilenotperfect4,Walkscore.comprovidesthreemeasures—walkscore,bikescore,andtransitscore—thatcapturetheeasinessofusingvarioustravelmodesatspecificlocations.SinceWalkscore.comprovidesanAPIservice,theresearchteamwasabletoextractthethreescoremeasuresbasedonthelatitudesandlongitudesofthegeocodedresidentiallocationofeachrespondent.Thesemeasuresprovideagoodproxyofthesupply-sidecharacteristicsofvariousneighborhoodsacrossCalifornia.Withthegeographicgeocodesofhomes,schools,andworkplacesofallindividualsinthedataset,infuturestagesofthisprojectweplantofurtherenrichtheCaliforniaMillennialDatasetwithavarietyoftransitandland-usevariablesfromotherreliablesourcessuchasAllTransit.comandGoogle.TheAlltransit.cnt.orgwebsiteprovidesawidearrayofmatricesontheperformanceoflocalpublictransportationsystemsforindividualcensusblockgroups.Byemployingthegeneraltransportationfeedspecification(GTFS)datasetsthattransitagenciesmaintain,anddirectlycollectinginformationabouttransitservicesfromtheagencieswithoutGTFSdatasets,thewebsitereturnsarichsetofvariablesundersixcategories,suchasjobs,economy,health,equity,transitquality,andmobility.Inaddition,twoamongthevariousGoogleAPIservices,theGooglePlacesAPIandGoogleDirectionAPI,provideuniqueinformationthatweplantouseinfuturestagesoftheprojecttoanalyzethelocationchoiceandthemodechoiceofMillennialsandGenXers.TheGooglePlacesAPIprovidesthegeographiccoordinatesofadiversesetofbusinesses.AsusersandbusinessownerscanaskGoogletocorrectcriticalinformationsuchasopeningandclosingofbusinesses,GooglePlacesAPIprovidesthehighlyaccurategeographiclocationsofbusinessesbytype.TheGoogleDirectionAPIcalculatesthedistanceanddurationofatripfromanorigintoadestinationbyfourmodes–driving,transit,biking,andwalking–basedonrealisticcongestioninformationthatvariesbytimeofdaybyusingtheirarchivedtrafficdata.

4Walkscore.commeasuresitsscoresbasedontheaccessibilitytopublicplaces.However,thedefinitionofpublicplaceshasbeenquestioned,assomeplacesthatareclassifiedas“private”,butdoprovidefreeaccesstothepublicandthereforecouldqualifyforthedefinitionofpotentialdestinationsfortrips,arenotconsideredinthecomputationofthescores.

Page 32: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

24

FactorAnalysisInthissection,wediscussthevariabledimensionreductionmethodthatwasappliedontheattitudinalstatementsfromsectionsAandJofthesurvey.Theattitudinalvariablesweremeasuredaskingtherespondentsfortheiragreementwith66statementsusinga5-levelLikerttypescale(fromstronglydisagreetostronglyagree).The66attitudinalstatementsweredesignedtomeasuretheindividual’sattitudesrelatedto28pre-determinedunobservableconstructs,includingattitudestowardbiking,carownership,changesvs.routine,environmentalconcern,landuse,masculinity,roleofgovernment,multitasking,etc.Theseattitudinalconstructscanexplainvariabilityindecisionsaboutcarownership,travelmodechoice,residentiallocationandmanyotherdecisionsthatmadebydifferentsegmentofpopulation.Asdiscussedearlier,outof2155respondents191individualshavefailedinansweringcorrectlytooneofthetrapquestionsincludedinthesurvey.ThreetrapquestionswereembeddedinthesectionsAandGofthereport,tocontrolforthequalityoftheresponses.Informationrelatedtotheindividualswhofailedtwoormoretrapquestionswasautomaticallyremovedfromthedataset.Theremaining191casesthatfailedonlyonetrapquestionareexpectedtocontainlowerqualityinformation,whichcouldskewtheresultofthefactoranalysisandsignificantlychangethefactorextractionandloadingprocess.Hence,weonlyperformedthefinalfactoranalysisontheindividualswithhigherqualityoftheresponses,i.e.therespondentswhodidnotfailanytrapquestion(N=1964cases).5Thefirstandmostchallengingstepinfactoranalysisistodeterminethenumberoffactorstobeextracted.Thedefaultinmoststatisticalsoftwarepackagesistoretainallfactorswitheigenvaluesgreaterthan1.0orgreaterthanavalueclosetoone,e.g.0.7(asdiscussedbyJolliffe,1972).Ontheotherhand,VelicerandJackson(1990)showedthatusingthiscriterionmayleadtotoomanyextractedfactors.UsingaMonteCarlosimulation,theauthorsfoundthat36%ofthesamplesretainedtoomanyfactorsusingthiscriterion.Hence,alternativeapproaches(basedonmultiplecriteria)havebeenrecommendedtoidentifythenumberoffactors,includingscreetestplot,Velicer’sMAPcriteria,parallelanalysis,andmostimportantlytheinterpretabilityoftheextractedfactors.Basedonmultiplecriteriaincludingtheevaluationoftheeigenvalues,screeplot,strengthoftherelationship,andinterpretability,arangeforthenumberoffactorwasfirstidentified.Thenfactorsolutionswiththosenumbersweretestedtoseewhichsolutionproducesthebestoutcomeconceptuallyandnumerically.Asexpected,somevariableswerefoundtohavesmallloadingsonanyfactors(smallerthan0.29).Intheotherwords,somestatementsdidnotloadonanyfactorsinanymeaningfulway.Thesestandalonestatementseitherbelongstosinglestatementconstruct(e.g.“Ilikeridingabike”isagoodattitudinalvariablethatcanbeusedinisolationtopredictbicyclingbehavior)orperceiveddifferentlybyrespondents(e.g.statement5Wecomparedtheresultsfromafactoranalysisthatwasperformedonthefulldataset,whichincludedalsotheselowerqualitycases.Thecomparisonconfirmedthehigheramountofnoiseinthesolutionthatwasestimatedusingthefulldataset.

Page 33: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

25

usedforcapturingtheeffectsofpeerpressureareoftendifficulttobeusedinbehavioralresearchduetothereluctantattitudeofmostrespondentstoreportpeerpressure,andsocialdesirabilitybias).Additionally,somestatementswithweakfactorloadingswereincludedinfactorsmeasuringacompletelydifferentattitudinalconstruct.Forexample,attitudestowardmasculinity(ormachismo),whichweremeasuredbystatementsincluding“Itismoreimportantformenthanforwomentohaveahigh-payingcareer”and“Atwork,it’sperfectlyfineforwomentohaveauthorityovermen”,loadedwellinthefactorthatmeasuredthepro-environmentalpolicyattitudesofindividuals.This,whileisasignofanotherlatentattributeofindividuals(e.g.whichmeasuressomeconservativism,ortraditionalthinking),makestheinterpretabilityofthefactormorecomplicated,intermsoftheirrelationshipwithenvironmentalchoices,andtravelbehavior.Forthisreason,thosetwostatementswereremovedfromthefactoranalysis.Table3showsthe14-standalonestatementsthatareexcludedfromthefactoranalysis.Onecanusethesestandalonestatementsasanordinalorasastandardizedvariablefordescriptivestatisticsandasexplanatoryvariablesformodelingpurposes,evenifthestatementsarenotincludedinthefactoranalysis.Table3.StandaloneStatements

AttitudinalStatementsIwouldpaymoneytoreducemytraveltime.ItismoreimportantformenthanforwomentohaveAhigh-payingcareer.Atwork,itisperfectlyfineforwomentohaveauthorityovermen.IavoiddoingthingsthatIknowmyfriendswouldnotapprove.Backgroundmusic/radio/TVistoodistractingforme.Ilikestickingtoaroutine.ItrytomakegooduseofthetimeIspendcommuting.Ilikeridingabike.Ifeelpositivelyaboutthelevelofinvestmentoccurringinmylocalroadsandlocaltransit.TheairqualityintheregionwhereIliveconcernsme.Havingchildrenmeansyouhavetohaveacar.Individualsshouldgenerallyputtheneedsofthegroupaheadoftheirown.Itisprettyhardformyfriendstogetmetochangemymind.IamuncomfortablebeingaroundpeopleIdonotknow.Aftercarefulanalysisoftheresultsandexcludingthestandalonestatements,weperformedthefactoranalysisonthe52remainingstatements.Basedonmultiplecriteria,atotalnumberof17factorswereidentified.Thefollowingsubsectionssummarizethecriteriathatwereusedtodeterminetheoptimalnumberoffactors.

Page 34: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

26

Eigenvaluegreaterthanone(orvalueclosetoone)Table4showstheinitialeigenvaluesfordifferentnumberoffactors.Asindicatedinthistable,16factorshaveeigenvaluesgreaterthan1.00and10factorshaveeigenvaluesbetween0.99and0.7.Hence,theoptimalnumberoffactorscouldbeintherangebetween16and26.Table4.Eigenvalues

Factor InitialEigenvalues Factor Initial

Eigenvalues Factor InitialEigenvalues

1 4.88 21 0.81 41 0.482 3.66 22 0.80 42 0.463 2.84 23 0.77 43 0.444 2.69 24 0.75 44 0.445 2.21 25 0.74 45 0.436 1.81 26 0.72 46 0.427 1.64 27 0.68 47 0.408 1.57 28 0.68 48 0.409 1.51 29 0.65 49 0.3710 1.28 30 0.64 50 0.3511 1.26 31 0.63 51 0.3412 1.20 32 0.62 52 0.2113 1.12 33 0.59 14 1.10 34 0.59 15 1.03 35 0.56 16 1.01 36 0.55 17 0.91 37 0.55 18 0.90 38 0.54 19 0.84 39 0.53 20 0.84 40 0.50

Screetest(i.e.elbowrule)Thesecondcriteriaforchoosingthenumberoffactorswasthescreetest.Accordingtothiscriterionthepercentofvarianceexplainedbytheindividualfactorswould“leveloff”asthesolutionreachesthemostappropriatenumberoffactors.Beyondthisnumberoffactors,additionalfactorswouldaccountforrandomerrors.Thisruleshouldbeappliedtoafinalun-rotatedsolution.Usingall52statementsusedinthefactoranalysis,weplottedthechangesinvarianceexplainedbydifferentnumbersoffactor.Theresultindicatesthatthedesirablenumberoffactorscanbebetween10and17(wherethepercentofvarianceexplainedbyindividualfactorsstartedtoleveloff).StrengthoftherelationshipInthiscriterionwecheckedwhethertherotatedfactorloadingsaregreaterthan|0.3|.Toidentifynon-trivialfactorsthatcouldbeobtained,researchersusedifferentcut-offs.Some

Page 35: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

27

researchersusemorerelaxedcriteriasuchasacut-offof|0.2|,whichseemsverylow,andsomeothersuseverystringentcriteriasuchasacut-offof|0.7|.Inourstudy,weusedacut-offvalueof|0.3|.Interpretability“Variablesthatloadnear1areclearlyimportantintheinterpretationofthefactor,andvariablesthatloadnear0areclearlyunimportant.Simplestructurethussimplifiesthetaskofinterpretingthefactors”(BryantandYarnold,1995,page132-133).Thus,forsimplicitywecontrolledthatallloadedstatementsconceptuallyconveyasimilarcontent(construct).Asdiscussedearlier,forexample,wehadtoexcludethemasculinity/machismostatements,whichloadedonthepro-environmentalpolicyfactor(withnegativedirection):thesetwogroupsofstatementsseemtocaptureratherdifferentconstructs.Table5.ModeratelyCorrelatedFactors

Factororvariable FactororVariable CorrelationPro-environmentalpolicies Mustowncar -0.356

Pro-environmentalpolicies Responsivetoenvironmentaleffectandpriceoftravel 0.301

Usingtheabovecriteriaensurestherobustnessandvalidity(convergentvalidityanddiscriminantvalidity)ofthefactorsolution.Furthermore,duetoexistenceofcorrelationamongfactors(seeTable5forthemosthighlycorrelatedfactors,withcorrelationshigherthan|0.3|),wechoseanobliquerotation:obliquerotationmayshowsomelevelsofcorrelationamongfactors,whichisnotidealinstatisticalanalysis,butitcancaptureindividualfactorsthatarebettersupportedbythedata,becauseitallowstohavefactorsthatarenotorthogonaltooneanother.Thefactoranalysisextractionmethodthatwasusedforthefinalsolutionwasthemaximumlikelihoodmethod.Thismethodproducesparameterestimatesthataremostlikelytohaveproducedtheobservedcorrelationmatrixifthesampleisfromamultivariatenormaldistribution(asreportedintheIBM’sSPSSManual).Maximumlikelihoodallowsthecomputationofawiderangeofgoodnessoffitmeasuresandsignificancetests.Thegoodnessoffittestofthefinalfactorsolutionwasstatisticallysignificant,withavalueofchi-squareof1336.94,andanumberofdegreesoffreedomequalto578.TheresultsoffinalfactorsolutionarepresentedinTable6.

Page 36: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

28

Table6.FinalResultsoftheFactorAnalysisFactorsandLoadedstatements FactorLoadingPro-storeshopping Iprefertoshopinastoreratherthanonline. 0.998Ienjoyshoppingonline. -0.413Pro-environmentalpolicies Weshouldraisethepriceofgasolinetoreducethenegativeimpactsontheenvironment. 0.937Weshouldraisethepriceofgasolinetoprovidefundingforbetterpublictransportation. 0.841Thegovernmentshouldputrestrictionsoncartravelinordertoreducecongestion. 0.331VarietySeeking Iliketryingthingsthatarenewanddifferent. 0.592Ihaveastronginterestintravelingtoothercountries. 0.405Pro-exercise Theimportanceofexerciseisoverrated. -0.822Gettingregularexerciseisveryimportanttome. 0.587Pleasantcommute Mycommuteisstressful. -0.802Mycommuteisgenerallypleasant. 0.689Trafficcongestionisamajorproblemformepersonally. -0.544ThetimeIspendcommutingisgenerallywastedtime. -0.501Gettingstuckintrafficdoesnotbothermethatmuch. 0.305Pro-suburban Iprefertoliveinaspacioushome,evenifitisfartherfrompublictransportationandmanyplacesIgoto. 0.764

IprefertoliveclosetotransitevenifitmeansIwillhaveasmallerhomeandliveinamorecrowdedarea. -0.69

Iliketheideaoflivingsomewherewithlargeyardsandlotsofspacebetweenhomes. 0.428Iliketheideaofhavingdifferenttypesofbusinesses(suchasstores,offices,restaurants,banks,andlibrary)mixedinwiththehomesinmyneighborhood. -0.357

Responsivetoenvironmentaleffectandpriceoftravel TheenvironmentalimpactsofthevariousmeansoftransportationaffectthechoicesImake. 0.739

Iamcommittedtousingalesspollutingmeansoftransportationasmuchaspossible. 0.598ThepriceoffuelaffectsthechoicesImakeaboutmydailytravel. 0.532Toimproveairquality,Iamwillingtopayalittlemoretouseahybridorotherclean-fuelvehicle. 0.384

EstablishedinLife I’malreadywell-establishedinmyfieldofwork. 0.704I’mstilltryingtofigureoutmycareer(e.g.whatIwanttodo,whereI’llendup). -0.636Iamgenerallysatisfiedwithmylife. 0.387Longtermsuburbanite Ipicturemyselflivinglong-terminasuburbansetting. 0.819Ahouseinthesuburbsisthebestplaceforkidstogrowup. 0.568Ipicturemyselflivinglong-terminanurbansetting. -0.310Mustowncar Idefinitelywanttoownacar. 0.697Iamfinewithnotowningacar,aslongasIcanuseorrentoneanytimeIneedit. -0.500Carasatool

Page 37: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

29

Thefunctionalityofacarismoreimportanttomethanitsbrand. 0.579Tome,acarisjustawaytogetfromplacetoplace. 0.480Climatechangeconcerned Greenhousegasesfromhumanactivitiesarecreatingmajorproblems. 0.796Anyclimatechangethatmaybeoccurringispartofanaturalcycle. -0.656ItispointlessformetotrytoohardtobemoreenvironmentallyfriendlybecauseIamjustoneperson. -0.307

Technologyembracing HavingWi-Fiand/or3G/4GconnectivityeverywhereIgoisessentialtome. 0.609Gettingaroundiseasierthaneverwithmysmartphone. 0.492Learninghowtousenewtechnologiesisoftenfrustrating. -0.359Technologycreatesatleastasmanyproblemsasitdoessolutions. -0.310Monochronic(Pro-monotasking) It’sbesttofinishoneprojectbeforestartinganother. 0.518Iliketojuggletwoormoreactivitiesatthesametime. -0.346Time/modeconstrained Myschedulemakesithardorimpossibleformetousepublictransportation. 0.580IamtoobusytodomanythingsI’dliketodo. 0.443Mostofthetime,Ihavenoreasonablealternativetodriving. 0.388Pro-social Socialmedia(e.g.Facebook)makesmylifemoreinteresting. 0.505Peoplearegenerallytrustworthy. 0.442Ienjoythesocialaspectsofshoppinginstores. 0.323Materialism Iwould/doenjoyhavingalotofluxurythings. 0.441IprefertominimizethematerialgoodsIpossess. -0.412Forme,alotofthefunofhavingsomethingniceisshowingItoff. 0.387Iliketobeamongthefirstpeopletohavethelatesttechnology. 0.380Tome,owningacarisasymbolofsuccess. 0.316TheBartlettmethodwasusedforgeneratingthefinalstandardizedfactorscores.Theresultingscoresfromthismethodareexpectedtobeunbiasedand,therefore,moreaccuratereflectionsofthecases’locationonthelatentcontinuuminthepopulation.

AdoptionofTechnology,IndividualAttitudesandMobilityChoicesofMillennialsvs.GenXersTheanalysisoftheCaliforniaMillennialsDatasetallowsustoinvestigateseveraltrendsassociatedwiththepersonaltravel-relatedattitudesofmillennialsandtheirmeasuresoftravelbehavior,andcomparethemwiththeattitudinalandbehavioralpatternsobservedamongmembersoftheolderGenerationX.Inthispartofthereportwesummarizetheobservedtrendsin(1)theuseofmoderntechnologies,socialmediaandsmartphoneapplicationsfortravelschedulingpurposes,(2)thedistributionofattitudinalpatterns,asmeasuredbythefactorscoresthatwerecomputedforallrespondentsincludedinthedataset,and(3)measuresoftravelbehaviorandadoptionofsharedmobilityservices,averageaccessibilityintheplaceofresidenceandadoptionofmultimodaltravelamongvarioussegmentsofthepopulation.In

Page 38: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

30

particular,wefocusondifferencesobservedamongvariousgroupsofmillennialsvs.olderadults,basedonthelocationwhereindividualslive.Figure3showstheuseofsocialmediasuchasFacebooktocoordinatetravelfornon-workactivitiesbyagegroup(millennialsvs.GenerationX)andneighborhoodtype(urban,suburbanandrural)wheretheindividuallives.Notsurprisingly,millennialsaremoreinclinedtofrequentlyusesocialmediatocoordinatefortheirnon-workrelatedtravel,withurbanmillennialsbeinginparticulartheheaviestadoptersoftheseservicestocoordinatetheiractivities.

Figure3.Theuseofsocialmediatocoordinatetravelbyagegroupandneighborhoodtype

Millennialsalsoreportedthattheyusesmartphoneinconnectionwiththeirdailytravelmoreoftencomparedtotheiroldercounterparts.Thefollowingsetoffiguressummarizestheuseofsmartphonetochecktrafficconditions(Figure4),checkwhenabusortrainarrives(Figure5),decidewhatmodeorcombinationofmodestouse(Figure6),learnhowtogetto/explorenewplaces(Figure7),andnavigateinrealtime(Figure8).Inparticular,andconsistentwithexpectations,urbanpopulationsarefoundtousetheirsmartphonemoreoftenforalltheseactivitiesbothamongMillennialsandGenXers.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never Sometimes Often Everytime or most often

Page 39: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

31

Figure4.Useofsmartphonetochecktrafficandtoplanthetravelrouteordeparturetimeby

agegroupandneighborhoodtype

Figure5.Useofsmartphonetocheckwhenabusortrainwillbearrivingbyagegroupand

neighborhoodtypeThedifferencesacrossneighborhoodtypesareparticularlylargefortheuseofsmartphonetechnologytocheckwhatmodesoftransportation,orcombinationsofmodes,touse,whichislikelytobeaneffectoftheavailabilityofmultipletraveloptionsindenserurbanareas.Inlatersectionsofthereport,wewillreturntodiscussingthemeasuresoftravelaccessibility,bymode,forthemembersofthevariousgenerations.Weplantofurtherinvestigate,infuturestepsoftheresearch,howtheuseofthesetechnologies,andthevariouslevelsofaccessibility

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never At least once a year At least once a Month At least once a week Daily

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never At least once a year At least once a Month At least once a week Daily

Page 40: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

32

intheareaswhereindividualslive,affecttheirtravelpatterns,anissueofsignificantimportancetoplanningprocesses.

Figure6.Useofsmartphonetodecidemeansoftransportationtousebyagegroupand

neighborhoodtype

Figure7.Useofsmartphonetolearnhowtogettoanewplacebyagegroupand

neighborhoodtype

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never At least once a year At least once a Month At least once a week Daily

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never At least once a year At least once a Month At least once a week Daily

Page 41: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

33

Figure8.Useofsmartphonetonavigateinrealtimebyagegroupandneighborhoodtype

InvestigatingMillennials’AttitudestowardsTransportationandTechnologyThissectiondescribesthedifferingattitudinalprofilesobservedamongmillennialsandmembersoftheGenerationXbytheneighborhoodtypetheyliveinusingthecomputedfactorscores.Personalattitudesandpreferencesarelikelytobeimportantfactorsaffectingindividualchoicesrelatedtohousing,travelandactivityscheduling.Still,todate,informationaboutindividualattitudes,preferences,andlifestylesisrarelycollectedintransportationsurveys.Inthissection,weexplorehowaverageattitudesdifferamongvarioussegmentsofthepopulationofmillennialsandGenXerswholiveindifferentneighborhoodtypes,withrespecttoseveralconstructsthatwereexploredintheattitudinalsectionofthesurvey,andthroughthefactoranalysispresentedinthepreviouschapter.Thenextsetoffigurespresentstheaveragefactorscores(and95%confidenceintervals)forvariousgroupsofindividuals,classifiedbyagegroupandneighborhoodtype(urban/non-urban)inwhichtherespondentslive.Itisimportanttoremindthereadersthat,asallfiguresinthissectionreportinformationforthestandardizedfactorscores(e.g.withzeromean,andvarianceequalto1),any(eventual)differencesacrossgroupsshouldbeevaluatedaccordingly.Forexample,ifagrouphasamoderatelypositiveaveragefactorscoreforthepro-environmentalpolicyfactorscores,thatmeansthattheindividualsthatbelongtothatgroup,onaverage,tendtohavestrongerpro-environmentalpolicyattitudes,comparedtotheaveragefortheentiresample(whosemeanforthisvariableiszero).

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

Seldom or never At least once a year At least once a Month At least once a week Daily

Page 42: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

34

Accordingly,thefigurespresentedinthissectionshouldnotbeinterpretedintermsofwhatindividualshaveacertainattitudinalcharacteristics(e.g.whatgroupsare“pro-environmentalpolicy”)but,rather,inrelativetermsasacomparisonacrossgroups(e.g.thefigureshelpanswerthequestion“aretheindividualsthatbelongtotheyoungergenerationsmorelikelytohavehigher“pro-environmentalpolicy”attitudesthanthosethatbelongtotheoldergeneration?Andwhatabouturbanvs.suburbanresidents?”).Similarly,inthosecasesinwhichallindividualsinthesampleeventuallyshareasimilarattitudetowardsatopic(e.g.positive“pro-environmentalpolicy”attitudes),thecomparisonacrossgroupsoftheaveragevaluesforthestandardizedfactorscoreshelpsdistinguishwhatgroupsofindividualstendtohaveevenstrongerattitudes(agreeevenmorethanothers)withsuchattitudinalconstruct.

Figure9.Average“pro-environmentalpolicy”factorscorebyagegroupandneighborhood

type(95%confidenceintervalsarereportedinthefigureforeachgroup)Figure9presentsthedifferencesintheattitudestowardpro-environmentalgovernmentpolicy,asmeasuredbytheaveragefactorscorethatwasextractedinthefactoranalysisforindividualsfrombothgenerationsthatliveinurbanvs.non-urbanareas:individualswithahigheraveragefactorscoretendtohavehigherdegreeofagreementwiththefollowingstatements:“We

Page 43: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

35

shouldraisethepriceofgasolinetoreducethenegativeimpactsontheenvironment.”,“Weshouldraisethepriceofgasolinetoprovidefundingforbetterpublictransportation.”and“Thegovernmentshouldputrestrictionsoncartravelinordertoreducecongestion.”Urbanrespondentsofallagesappeartobehighersupportiveofpro-environmentalpolicies,whilenon-urbanresidents’agreementwiththesestatementsappearstodeclineastheageofrespondentsincreases.Urbanresidents,acrossallagegroups,alsopresentmoreheterogeneityforthisattitudinaldimension,asshownbythelargerconfidenceintervalsaroundthemean.Next,Figure10showstheaveragevaluesforthevarietyseekingattitudinalfactorscore,byagegroupandneighborhoodtype.Thisfactorcapturesindividuallevelsofagreementwithstatementsconsistingof“Iliketryingthingsthatarenewanddifferent”and“Ihaveastronginterestintravelingtoothercountries”.Urbanrespondentshavehigherscoresacrossagegroups,particularlyintheagerangesof25to34and35to44.

Figure10.Average“varietyseeking”factorscoresbyagegroupandneighborhoodtype(95%

confidenceintervalsarereportedinthefigureforeachgroup)

Page 44: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

36

Again,muchlargervarianceisobservedamongurbandwellers,probablyasthecombinedeffectoftheheterogeneityassociatedwiththesegroupsofindividuals,aswellasthesmallersamplesizesthatareavailablefortheurbansubsamples.6Individualsinthehighestagegroup(45-50)arethosethathavethelowestvaluesforthisfactorscore.

Figure11.Average“responsivetoenvironmentaleffectsandpriceoftravel”factorscorebyagegroupandneighborhoodtype(95%confidenceintervalsarereportedinthefigurefor

eachgroup)Figure11reportstheresponsivenessoftravelerstopriceandenvironmentaleffectoftransportation.Thosethathaveahighervalueforthisfactorscoretendtoagreewiththefollowingstatements:“TheenvironmentalimpactsofthevariousmeansoftransportationaffectthechoicesImake”,“Iamcommittedtousingalesspollutingmeansoftransportationas

6Urbanresidentsincludevariousgroupsofindividualswithdifferentlifestyles,includinggroupsofindividualswhoareinatransientstageoftheirlife,youngerindividualswhoarestilldevelopingtheirtrainingandeducation,individualsthatlivewithotherroommatesandhousemates,temporaryresidents,professionalsandotherhighly-educatedworkers,youngcoupleswithnochildren,membersofminorities,etc.Theproportionoftemporaryresidents(andtenantswhorenttheirhousingunits)isusuallyhigherinurbanareas,andtheaverageturnoverofresidentsinahousingunitisfaster.Inaddition,awidevarietyofurbanneighborhoodsexist,eachwithdifferentcharacteristicsandvariouslevelsofaccessibilitybyvarioustransportationmodes.

Page 45: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

37

muchaspossible”,“ThepriceoffuelaffectsthechoicesImakeaboutmydailytravel”and“Toimproveairquality,Iamwillingtopayalittlemoretouseahybridorotherclean-fuelvehicle”.Thisfactorcapturesrespondents’willingnesstochangetheirtravelmodebasedonboththeenvironmentalimpactsoftransportationandgasprice.AsindicatedinFigure11,urbanrespondentsofallagegroupshavehigheraveragefactorscoresthannon-urbanrespondents.Interestingly,non-urbanrespondents’tendencytoagreewiththesestatementsappearstodeclinebyagegroup,withtheindividualsbetween35and50yearold(GenXers)agreeingtheleastwiththesestatements.However,amongurbanrespondents,theaveragefactorscoreappearsrelativelyconstantbyagegroup.Thismaysuggestthaturbanrespondentsofallagesviewtheenvironmentpositivelyandconsidertheenvironmentalimpactsoftransportation-relateddecisionaswellaspriceoffuelwhenamakingtransportationchoices.Thismaybealsoaffectedbytheavailabilityofmoreoptions(i.e.transitservices,bikelanesandshorterdistancesthatcanbecoveredwithvariousmodes).Further,thisattitudinalfactorscoremightsignalthebehaviorofindividualsthatmayeventuallyself-selecttoliveinanurbanneighborhoodtypeduetotheseunderlyingpreferences(e.g.theymovedtoanareathatbettermatchestheirpreferences).Figure12showsthedifferencesintheaverageclimatechangeconcernfactorscorebyagegroupandneighborhoodtype.Thosethathavehighervaluesforthisfactorscoretendedtoagreewiththestatement“Greenhousegasesfromhumanactivitiesarecreatingmajorproblems”,andtendedtodisagreewiththefollowingstatements:“Anyclimatechangethatmaybeoccurringispartofanaturalcycle”,and“ItispointlessformetotrytoohardtobemoreenvironmentallyfriendlybecauseIamjustoneperson.”Thepatternofresponsesissimilartothefactormeasuringtheagreementwiththegovernmentintervention,whereurbanrespondentshavealmostuniformlyhigherscoresforthisfactor,whilefornon-urbanrespondentsexpresslowerconcernforclimatechange,onaverage.Differencesbetweenurbanandnon-urbanrespondentstendtoincreasewithage.

Page 46: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

38

Figure12.Average“climatechangeconcerned”factorscorebyagegroupandthe

neighborhoodtype(95%confidenceintervalsarereportedinthefigureforeachgroup)Figure13reportstheaveragevaluesfortheestablishedinlifefactorscorebyagegroupandareaswheretherespondentslive.Thisfactorcapturesrespondents’opinionabouttheirlifestagethroughtheirlevelofagreementwiththestatements“I’malreadywell-establishedinmyfieldofwork”,“Iamgenerallysatisfiedwithmylife”,and“I’mstilltryingtofigureoutmycareer(e.g.whatIwanttodo,whereI’llendup).”Itisnotsurprisingtoseethatasindividualsbecomemoreestablishedintheirlife,theirlevelofsatisfactionincreases(althoughthisseemstocounteractthestereotypeoftheoptimisticmillennialgeneration,whothinkpositiveeveniftheyareinatransientstageoftheirlife,asoftenreportedbythemedia).

Page 47: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

39

Figure13.Average“establishedinlife”factorscorebyagegroupandneighborhoodtype

(95%confidenceintervalsarereportedinthefigureforeachgroup)Bothlifesatisfactionandstabilityincreasesbyage.BothyoungerandoldermillennialstendtohaveloweraveragescoresthanmembersofGenerationX.Thisisunsurprisinggiventhatthemillennialsareoftenunderemployedandinmanycasesstillnotindependent(livingwiththeirparents),butlargedifferencesareobservedbetweenyoungandoldmillennials,withtheurbanmillennialshavingthelowestaveragescoresforthisfactor.Alsoforthisfactor,muchlargervarianceisobservedamongurbandwellers,evenifthey,onaverage,havehigherscoresthantheirnon-urbancounterparts.Figure14reportstheaveragefactorscoreandconfidenceintervalforthelong-termsuburbanitelifestylefactorscore.Thosethathavehigherscoresforthisfactortendtoagreewiththefollowingstatement“Ipicturemyselflivinglong-terminanurbansetting”andtheytendtodisagreewith“Ipicturemyselflivinglong-terminasuburbansetting”and“Ahouseinthesuburbsisthebestplaceforkidstogrowup.”

Page 48: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

40

Figure14.Average“long-termsuburbanite”factorscorebyageandneighborhoodtype(95%

confidenceintervalsarereportedinthefigureforeachgroup)Ingeneral,inclinationtowardsuburbanitelifestyleislowerforindividuallivinginurbanneighborhoodcomparedtotheircohortlivinginsuburbanorruralareas.Notsurprisingly,GenXerswholiveinsuburbanareashavethehighestaveragescoresforthisfactor.Veryinterestingly,andsomewhatunexpectedly,though,thetrendamongmillennialsshowthatmanymillennialsstillseethemselvesliving“longterm”inasuburbanarea.Thisfindinghasextremelyimportantplanningimplications:ifconfirmedbyfuturedecisionsaboutresidentiallocation,thetrendwouldconfirmthatthehigherpreferenceforcentralurbanareasamongmillennialsmightbeonlyatransitionassociatedwiththeirstageinlife.Similarly,thehopeofmanypolicy-makersthatmillennialsmightcontinuetoembraceurbanlifestylesandcontinuetosupporttheregenerationofthecentralareasofcitiesalsoastheyagemightnotbefullysupported,withimportantimplicationsonthefuturedemandforhousingandtravel.Figure15presentstheaveragefactorscoreandconfidenceintervalforthe“mustownacar”factor.Thosethathavehighscoresforthisfactortendedtoagreewiththefollowingstatement:“Idefinitelywanttoownacar”anddisagreewiththestatement:“Iamfinewithnotowningacar,aslongasIcanuseorrentoneanytimeIneedit”.

Page 49: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

41

Figure15.Average“mustowncar”factorscorebyageandneighborhoodtype(95%

confidenceintervalsarereportedinthefigureforeachgroup)Exceptyoungermillennials,theurbanrespondentsofallagegroupstendtodisagreewiththisfactor,indicatingthattheyarelessinclinedtoownacar.Fornon-urbanrespondents,carownershipattitudesappeartobestrongerwithage.Ingeneral,membersoftheGenerationXhaveahigherpreferencetowardsowningacarthanmillennials.Theurbanpopulationinthecentralagegroups(25-34and35-44)havethelowestscoresforthisfactor,thussuggestingthatthesegroupsdonotrecognizelargeimportancetoowningacar,aslongastheycanaccesssufficientmobilityservicesthroughotherchannels.However,theratherhighscoresforthisfactoramongyoungmillennials(agroupthatisfoundtohavelowercarownershiplevels)seemstoconfirmthatformanyindividualsinthisgroup,carownershipisstillseenhashavingavalue,evenifthecurrentlowercarownershiplevelsmightbeassociatedwithtemporaryconditions,suchaslowerincome,studentstatusandlackofemployment.

Page 50: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

42

Figure16reportstheaveragefactorscoreandconfidenceintervalforthecarsasatoolfactorscore.Thisfactorcapturestherespondents’levelofagreementwiththestatements“Thefunctionalityofacarismoreimportanttomethanitsbrand”and“Tome,acarisjustawaytogetfromplacetoplace”.Alsointhiscase,theloweraveragefactorscoreforyoungmillennialswholiveinurbanareasseemstosuggestthattheirlowerlevelsofcarownershipareonlyatemporarystatus.

Figure16.Average“carasatool”factorscorebyageandneighborhoodtype(95%confidence

intervalsarereportedinthefigureforeachgroup)Figure17presentstheaveragefactorscoreandconfidenceintervalcapturingrespondents’inabilitytouseothertravelalternativesduetotheirtimeandtravelmodeconstraintsimposedbyeithertheirbusyscheduleorunavailabilityofdifferentoptionsfortraveling.Thisfactorisbasedonthethreeattitudinalstatements:“Myschedulemakesithardorimpossibleformetousepublictransportation,”“IamtoobusytodomanythingsI’dliketodo,”and“Mostofthetime,Ihavenoreasonablealternativetodriving”.

Page 51: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

43

Amongurbanresidents,oldermillennialstendtohavehigheraveragescoresforthisfactor,whileamongnon-urbanresidents,theoldermembersofGenerationXhavethehighestaveragescores.Thismaybeduetophysicalconstraintsorotherliferesponsibilities,suchashavingchildren.

Figure17.Averagetimeandmodeconstrainfactorscorebyageandneighborhoodtype(95%

confidenceintervalsarereportedinthefigureforeachgroup)Figure18reportstheaveragefactorscoreandconfidenceintervalfortherespondents’feelingsregardingtheadoptionoftechnology.Thisfactorcapturesthetechnologicalembracementconstructthroughthestatements“Learninghowtousenewtechnologiesisoftenfrustrating”(withnegativesign),“Technologycreatesatleastasmanyproblemsasitdoessolutions”(withnegativesign),“HavingWi-Fiand/or3G/4GconnectivityeverywhereIgoisessentialtome”and“Gettingaroundiseasierthaneverwithmysmartphone.”Respondentsshowedaclearpatternwithdistinctivefeaturesbetweenurbanandnon-urbandwellers.Forurbanresidents,thefactorscoreispositiveorclosetozero–indicatingeitherpositiveorneutralfeelingsabouttheroleoftechnologyacrossallagegroups.Fornon-urban

Page 52: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

44

residentstechnologyexcitementdecreaseswithage.Youngmillennials(18-24)havehigherenthusiasmabouttechnology,whileoldmillennials(25-34)haveslightlylowerpropensitytowardstechnology,andthemembersofGenerationXreportthelowestembracementoforrelianceontechnology.

Figure18.Average“technologyembracing”factorscorebyageandneighborhoodtype(95%

confidenceintervalsarereportedinthefigureforeachgroup)Thelastfactorscorethatisdescribedinthisreportismaterialism.Figure19showsthedifferencesintheaveragescoreforthisfactorbyagegroupandneighborhoodtype.Thosethathavehighervaluesforthisfactortendtoagreewiththefollowingstatements:“Iwould/doenjoyhavingalotofluxurythings”,“Forme,alotofthefunofhavingsomethingniceisshowingItoff”,“Iliketobeamongthefirstpeopletohavethelatesttechnology”,“Tome,owningacarisasymbolofsuccess”.Also,thosewithhighfactorscorestendtodisagreewiththestatement:“IprefertominimizethematerialgoodsIpossess”.Theaveragescoresforthisindextendtodecreasewiththeincreasingageoftherespondents.Youngmillennialsinbothurbanandnon-urbanareashavethehighestaveragescores,perhaps

Page 53: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

45

duetotheirinterestinhavingthelatestgadgets,ortheirstageoflife–wherefewhavechildrenormortgagesthatpreventthemfromacquiringorwantingtoacquirematerialgoods.OlderGenerationXmembershavethelowestaveragescoresforthematerialismfactor.Infuturestagesoftheresearch,itwillbeveryinterestingtoexplorehowthemembersofthefollowingGenerationZ(under18yearolds,asoftoday),willbehaveinfutureyears,comparedtothesegenerationsthatwearestudying.Inaddition,non-urbanrespondents(apartfromtheyoungmillennials)tendtohaveloweraveragevaluesforthisfactor,andthushavelowermaterialisticattitudes,thantheirurbancounterparts.

Figure19.Average“materialism”factorscorebyageandneighborhoodtype(95%confidence

intervalsarereportedinthefigureforeachgroup)

Page 54: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

46

TravelBehaviorandtheAccessibilityofthePlaceofResidenceInthePartIreportforthisresearchstudy(Circellaetal.,2016b),wediscussedanumberofobserveddifferencesinthetravelbehaviorofmillennialsvs.theoldercounterpartsbelongingtotheprecedingGenerationX.Amongtheobserveddifferences,theanalysisofthecollecteddatahighlightedthatmillennialstendtodriveless,andthisdifferencesholdsevenaftercontrollingfortheneighborhoodtypewheretherespondentslive.Further,alargerproportionofmillennialsreportnottohaveavaliddriver’slicenseatthetimetheycompletedthesurvey.Millennialsalsoarelesslikelytodriveduringtheircommute,andmoreoftenuseactivemodesoftransportation,includingwalkingandbiking,aswellasridingpublictransit.Amongtheindividualsthatphysicallycommutetoworkatleastoneperweek,millennialstendtomorefrequentlyengageintravelmultitasking(i.e.carryoutanactivitywhiletraveling)duringtheircommute,comparedtoGenXersinallregionsofCalifornia.Thehigheradoptionofmultitasking,whichcorrelateswiththelargeradoptionofICTdevicesamongmillennials,mightbeassociatedwithadifferentevaluationoftheutilityoftravelalternatives,andthereforeexplainatleastinparttheobserveddifferencesinmodechoice.Oneofthereasonsthatmaybebehindtheobserveddifferencesintravelpatternsbetweenmembersofthedifferentgenerationrelatestothecharacteristicsofthebuiltenvironmentoftheresidentiallocationandthework/schoollocationwhereindividualstravel.Forexample,thefollowingTable7andTable8respectivelyreporttheaveragefrequencyofuse(byday)ofon-demandrideservicessuchasUberLyftandofcar-sharingservicessuchasZipcarorTuro.Table7.AverageFrequencyofUseofUber/LyftbyGenerationandNeighborhoodType

Millennials(N=1157) GenerationX(N=998)

NeighborhoodType Rural 0.004 0.003Suburban 0.010 0.007Urban 0.056 0.039Note:Numbersinthetablemeasuretheaveragenumberofper-capitatripsperdaybyneighborhoodtype(ordinalfrequencycategoriesweretransformedintodiscretenumbersoftripstocomputethedatainthistable)Whilecarsharingservicesarecertainlymorerarelyusedthanon-demandrideservicessuchasUberorLyft,Tables7-8reportsomesimilartrends,withresidentsofdenser,morecentrallocationsusingtheseservicesmoreoftenthansuburbanorruralresidents.Inalltheseareas,millennialstendtousesharedmobilityservicesmoreoftenthanGenXers.Thus,consideringalsothedifferentdistributionsoftheurbanvs.non-urbanpopulationsofmillennialsandolderpeers,acompositeeffectmightexplaintheadoptionofthesetrends:notonlymillennialsaremorelikelytoadopttheseservicesthanolderpeers,holdingthecharacteristicsoftheneighborhoodconstant,butmillennialsarealsomorelikelytoliveinurbanareas.Thejointdecisionsoftheresidentiallocationwhereanindividualdecidestolive,andthetypeoftravel

Page 55: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

47

behaviortheyhaveisanimportanttopictoexploreinordertoinvestigatethereasonsbehind,andtheimpactsofmillennials’decision.Table8.AverageFrequencyofUseofZipcar/TurobyGenerationandNeighborhoodType

Millennials(N=1157) GenerationX(N=998)

NeighborhoodType Rural 0.00211 0.00010Suburban 0.00202 0.00070Urban 0.00984 0.00098Note:Numbersinthetablemeasuretheaveragenumberofper-capitatripsperdaybyneighborhoodtype(ordinalfrequencycategoriesweretransformedintodiscretenumbersoftripstocomputethedatainthistable)Tounderstandtheimpactofbuiltenvironmentalcharacteristics,weintegratedourdatasetwithotherinformationusingthegeocodedself-reportedresidentiallocationaddress.Figures20-22,presenttheaveragevaluesofsomeresidentiallocationaccessibilitymeasuresforthedifferentagegroupandbydifferentmodes.Thefiguresrespectivelypresenttheaveragewalkscore,bikescore,andtransitscoreofmillennialsandgenerationXbyneighborhoodtype.

Figure20.Walkscorebyagegroupandneighborhoodtype

Page 56: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

48

Figure21.Bikescorebyagegroupandneighborhoodtype

TheaveragescoresobservedacrosstheresidentiallocationofGenXersandmillennialsareverysimilarwithinaneighborhoodtype.Forexample,theaveragewalkscore(Figure20)foranurbanmillennialwas80.8,comparedto80.4foramemberofGenerationXinanurbanarea(thoughmoremillennialstendtoliveinsuchneighborhoods,thanGenXers).However,largedifferencesinthewalkscoresarefoundacrossneighborhoodtypes:forexample,millennialswholiveinsuburbanareashaveanaveragewalkscoreof51.7,comparedtoaveragewalkscoreof30.7inruralareas.Thedifferencesinthebikescores(Figure21)wereslightlylesspronounced,duethemorehomogenouscharacteristicsofbikeaccessibility(e.g.manysuburbanneighborhoodsandruralareasareratherbike-friendly),forexamplewithmillennials’bikescoresrangingfrom70(urban)to56.7(suburban)to51.2(rural).Transitscoresshowedasimilarpattern,thoughwithamoresignificantdropintheaveragescoresinruralareas.Urbanmillennialshadanaveragetransitscoreof60,whilesuburbanmillennialshadanaveragescoreof35,andruralmillennialshadanaveragescoreof22.

Page 57: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

49

Figure22.Transitscorebyagegroupandneighborhoodtype

TherewerenosignificantdifferencesintheaverageaccessibilitymeasuredbythesescoresbetweenmillennialsandGenerationX,withtheonlyexceptionofthewalkscoresforruralmillennialswhichwere2percentagepointshigherthanthoseforruralGenXers,suggestingthatmillennialsmayliveinslightlymorewalkableruralareas.However,forthemostpart,millennialsandGenXershaveaverageaccessibilityscoreswithinapoint.AdoptionofMultimodalTravelBehaviorAspreviouslydescribedinthisreport,inordertoenrichtheCaliforniaMillennialsDatasetwithlandusedataavailablefromothersources,wedevelopseveralmeasuresofaccessibilityusingtwomainsourcesofdatathatwereimportedbasedonthegeocodedresidentiallocationoftherespondents:theSmartLocationDatabase(SLD)developbytheUSEnvironmentalProtectionAgencyandWalkscore.com.SLDdataprovidelandusemeasuresondensity,diversity,design,accesstotransit,anddestinationaccessibilityattheCensus2010blockgrouplevel(Ramsey&Bell,2014).Wecomplementedthesedatawiththewalkscores,bikescores,andtransitscoresavailablefromthecommercialwebsitewalkscore.com,whichreflectmoremicro-levelbuiltenvironmentcharacteristicsavailableatafinerlevelofspatialdetailthanthecensusblock

Page 58: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

50

groupandarebasedonrecentlyupdateddatasources.7FortherespondentsthatprovidedavalidstreetaddresswecomputedaccessibilitymeasuresbasedonthecensusblockgroupfortheSLDmeasures,andonthelatitudeandlongitudeoftheresidenceforthescoresobtainedfromWalkscore.com Inthisanalysis,wefurtherclassifiedmillennialsintwogroups:theindependentmillennialswhodonotlivewiththeirparents,andthedependentmillennialswholivewiththeirparents.Weassumethatindependentmillennialshavemoreflexibilityinchoosingtheirresidentiallocation,butdependentmillennialsareaffectedbytheirparentsintheirresidentialchoiceandmodechoiceforvarioustrips.FortherespondentsthatprovidedavalidstreetaddresswecomputedaccessibilitymeasuresbasedonthecensusblockgroupfortheSLDmeasures,andonthelatitudeandlongitudeoftheresidenceforthescoresfromWalkscore.com.Further,foreachrespondentinthedataset,wecomputedseveralmultimodalityindicesusinginformationonthemode(s)thattheindividualusedfortheirlastcommutetour.8Weclassifyrespondentsbasedontheirmono-vs.multi-modalitystatusasmono-car(i.e.individualswhodrovealoneorcarpooledfortheirentirecommutetour),mono-transit(i.e.individualswhoonlyusedpublictransportationservicessuchasbus,commuterrail,andlightrailfortheentiretyoftheircommutetour),mono-walk(i.e.individualswhoonlywalkedtoworkorschool),mono-bike(i.e.individualswhoonlybikedtoschoolorwork),andmono-other(i.e.individualsthatexclusivelyusedothermodesoftransportation,e.g.on-demandrideservices,ferry,etc.fortheircommute).Wealsodefinedtwointer-modalindicesforindividualswhousedmorethanonemodesduringtheircommutetour:intermodal-car(anindexthatidentifiesindividualswhousedacarastheirmaincommutemodeinconjunctionwithothersecondarymodes)andinter-modalgreen(thatidentifiesindividualswhousedanynon-carmodeastheirprimarymodeoftransportation,combinedwithothersecondarymodes). Wecomputedtheseindicesforallrespondentsthatcommutetoworkorschoolatleastonceperweek,andhaveavalidgeocodedaddress.Thesampleavailableforthisanalysisconsistsof483independentmillennials,320dependentmillennials,and584GenXers.Figure23reportsthesummarystatisticsforthetwolargestmetropolitanareasofCalifornia,SanFranciscoandLosAngeles,comparingtheaverageforfouroftheeightmultimodalityindicesthatwerecreatedandtheaverageaccessibilitymeasuresforthethreegroupsthathavebeenidentified.

7Therearesomelimitationsintheuseofthewalkscorewhencomparingdifferentneighborhoods:forexample,manycommunitieswherethehomeownersmaintaintheparks,communitycentersandotheramenitiesgetlowscoresfromWalkscore.combecausethefacilitiesarenotconsidered“public”,eventhoughanyonewholivesanywherenearhasaccessandthecommunitiesarenotgated.Despitetheselimitations,thescoreprovidesausefulmeasureofaneighborhood’swalkability,withastandardizedscorethatcanbeeasilycomparedacrosslocations.8Additionalmeasuresofmultimodalitywerecomputedfornon-commuting/leisuretrips,butarenotfurtherdiscussesinthisreport,forbrevity.

Page 59: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

51

Figure23.Accessibilitylevelandadoptionofmultimodality,bygenerationalgroup,in(a)the

SanFranciscoBayArea(MTC);and(b)GreaterLosAngelesregion(SCAG)Inbothregions,9independentmillennialshavethehighestvaluesforallaccessibilitymeasures.Importantdifferencesareobservedamongdependentandindependentmillennials.Dependentmillennialstendtoliveinareasthathavethelowestlevelsofaccessibilitybynon-carmodes,

9Thetrendsinbothregionsaresimilar,withtheonlyexceptionthatlevelsofaccessibilitybynon-automodesarehigherinSanFrancisco/MTC,whilethepercentageofmono-carcommuters,inparticularamongGenXers,ishigherinLosAngeles/SCAG.

Page 60: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

52

probablyduetotheresidentiallocationchosenbyothermembersofthehouseholds(e.g.youngadultswhostillwiththeirparents).Independentmillennials,ontheotherhand,aremoreoftenfoundtoliveinlocationswithhigheraccessibility.Suchlocationsaremoreconducivetotheadoptionofgreenerandnon-autocommutemodes(and/ormayreinforcethepropensityofyoungadultstousesuchmodes),asmoreoftendonebytheindividualinthisgroup.Attheotherendofthespectrum,GenXersrelyheavilyontheuseofcarsfortheircommute.Interestingly,inbothregionsGenXersarefoundtoenjoybettertravelaccessibilitythandependentmillennials.Thisseemstosignalthatatleastsomedependentmillennialstendtodrivelessandhaveamoremultimodaltravelbehaviordespitelivinginneighborhoodsthatarelessconducivetomultimodalityandtotheuseofnon-automodes.Severalexplanationscouldbebehindthisfinding,includingtheimpactoflowerincomeandweakereconomicconditions(whichconstitutepotentialconstraintstomillennials’useofprivatevehicles),reducedfamilyobligations(e.g.millennialswholivewiththeirparentsarelesslikelytohavetheirownchildrentoescorttoschoolorextracurricularactivities,thereforetheyhavefewerconstraintsofthistype,andmorespaceforindividualchoices),and/orhigherpropensitytowardssuchbehaviors.Mostlikely,acombinationofthesefactorsisbehindthesepatterns. Table9,below,summarizestheaccessibilitymeasuresandmultimodalityscoresthatwerecomputedforthevariousregionsofCalifornia.Next,Figure24summarizestheadoptionofmultimodalbehaviorbyregionofCaliforniaandsub-segmentofthepopulation.Insummary,accessibilityandmultimodalityarepositivelycorrelated:residentsofneighborhoodswithbetteraccessibilityaremoreoftenfoundtobemultimodalcommuters.However,millennials,andespeciallydependentmillennials,arefoundtomakethemostoftheirbuiltenvironmentpotential,eitherduetoindividualchoicesorthepresence(orlack)oftravelconstraints. Theyarelesslikelytobemono-driversandmorelikelytobemultimodalcommuters,eveniftheyliveinneighborhoodsthatarelesssupportiveofsuchbehaviors.Thissuggeststhattheconnectionbetweenthebuiltenvironmentandtravelpatternsmaydifferbygeneration:infuturestepsoftheresearchweplantofurtherinvestigate(andmodel)therelationshipsbetweenaccessibilityandmultimodalbehavioramongthemembersofthedifferentgenerations,whilecontrollingforotherfactorsaffectingresidentialandtravelchoices.FurtherinformationcanbefoundinCircellaetal.(2017).

Page 61: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

53

Table9.AverageaccessibilitymeasuresanduseofcommutemodesbyregionofCaliforniaandgenerationalgroup

Samplesize(N)

Housingunits/acre

People/acre

Jobs/acre Walkscore Bikescore Transitscore Always

car

Othermode(transit,walking,biking)

Morethanonemode

CentralValley IndependentMillennials 73 2.5 7.0 1.4 37.0 53.7 27.8 74.4% 7.3% 18.3%

DependentMillennials 35 3.0 8.5 2.2 41.0 55.9 30.9 60.0% 22.9% 17.1%GenXers 82 3.2 8.8 2.1 42.7 56.8 29.8 83.6% 9.6% 6.8%

MTC IndependentMillennials 179 11.1 24.9 14.9 66.6 77.1 56.7 56.6% 17.1% 26.4%

DependentMillennials 67 5.9 16.0 3.7 53.4 61.7 44.1 56.7% 14.9% 28.4%GenXers 129 8.7 19.6 7.9 60.3 70.7 51.8 72.6% 10.6% 16.8%

NorthernCaliforniaandRestofStateIndependentMillennials 53 3.5 8.9 2.6 47.6 82.6 32.2 60.4% 18.8% 20.8%DependentMillennials 26 2.4 6.8 1.5 30.9 52.3 18.3 80.8% 0.0% 19.2%GenXers 48 2.3 5.6 2.3 36.2 86.2 17.0 81.1% 11.3% 7.5%

SACOG IndependentMillennials 90 4.1 9.2 3.7 48.8 79.3 32.2 76.8% 13.7% 9.5%

DependentMillennials 32 3.4 8.8 1.7 41.3 66.0 28.9 68.8% 6.3% 25.0%GenXers 95 3.3 8.3 5.7 42.0 73.8 33.4 82.2% 11.1% 6.7%

SANDAG IndependentMillennials 114 6.4 14.0 5.2 51.0 50.3 38.4 73.8% 8.4% 17.8%

DependentMillennials 43 4.5 11.6 2.3 44.2 41.2 33.1 62.8% 7.0% 30.2%GenXers 107 7.0 15.0 7.5 57.7 54.1 41.3 80.7% 5.3% 14.0%

SCAG IndependentMillennials 156 7.5 17.8 11.9 62.3 62.2 51.3 68.0% 9.5% 22.5%

DependentMillennials 61 4.4 13.8 2.5 48.0 55.2 34.6 68.9% 6.6% 24.6%GenXers 169 6.6 16.0 5.1 57.3 62.7 42.8 84.0% 6.4% 9.6%

Total IndependentMillennials 665 6.6 15.2 8.1 54.8 63.6 44.0 68.3% 11.9% 19.8%

DependentMillennials 264 4.3 12.0 2.5 45.3 54.5 34.8 64.8% 10.2% 25.0%GenXers 630 6.1 14.1 5.8 52.8 62.9 42.0 79.8% 8.7% 11.4%

Page 62: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

54

Figure24.Adoptionofmultimodalbehavior,byregionofCaliforniaandsub-segmentofthe

population

Page 63: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

55

VehicleMilesTraveled

Aspointedoutintheliterature,millennialsmaytraveldifferently,andfordifferentreasons,thanpreviousgenerations.Inthissection,weinvestigatethereasonsaffectingmillennials’vehiclemilestraveled(VMT)whilecomparingthemwiththecorrespondingpatternsobservedamongthemembersoftheprecedingGenerationX.Asobservedinmanypreviousstudies,millennialstendtopostponemarriage,householdcreation,andchildbearing,andtheyhavefewertotalchildrenthanpreviousgenerations(PewResearchCenter2014;McDonald2015).AllthesepatternsmightcontributetolowerVMT.HouseholdswithoutchildrentendtohavelowerVMTthanthosewithchildren(Santosetal.2011;LeVine&Jones2012).However,oldermillennialsmayalreadyexhibitpatternssimilartooldergenerations,indicatingthatmillennialsmay“drivelater,ratherthandriveless”(Garikapatietal.2016).Surveysofmillennialsreportthatthemembersofthiscohortseemtohavestrongerpreferencefordenseurbanareas(PewResearchCenter2014;Polzinetal.2014;BRS2013;Zmudetal.2014),andaremorecommittedtoenvironmentalcauses(Hanksetal.2008;Strauss&Howe2000),whichmayalsocontributetoreducingVMT(Ewing&Cervero2010).ThebuiltenvironmentisastrongdeterminantofVMT:numerousstudieshaveconnectedpopulationdensity,employmentdensity,andregionaldiversity(amongotherdimensionsofthebuiltenvironment)withvehiclemilestraveled.Vehiclemilestraveledseemstobemoststronglyrelatedtomeasuresofaccessibilitytodestinations.Moregenerally,residentsofmoretraditionaldenseurbanneighborhoodstendtodrivelessthanthosethatliveinlessdensesuburbanneighborhoods(Santosetal.2011;Ewing&Cervero2001;Caoetal.2009b;Cervero&Duncan2003;Cervero&Duncan2006).However,itisunclearhowthiseffectsmayaffectfuturetraveldemand,asmillennialsageandmovetoadifferentstageoflife.Inotherwords,theoftenreportedmillennials’preferenceforurbanareasandreduceduseofpersonalvehiclesmightbeatemporarytrend,associatedwiththeirstageinlife,andmaynotbealastingtrend(Myers2016).Anotherwell-studiedcorrelateofVMTisvirtualmobilityoradoptionofinformationcommunicationtechnology(ICT),whichisanotherfactoroftenindicatedaspotentiallyaffectingtheamountofindividuals’travelandmodechoice(Mokhtarian2009;Salomon&Mokhtarian2008;Contrino&McGuckin2006,CircellaandMokhtarian,2017).Millennialsaremorelikelytoadoptvirtualmobilityoptions,suchasonlineshopping,telecommuting,ride-sharing,andotherreal-timetransportationservices(Blumenbergetal.2012;Zipcar2013).McDonald(2015)suggestedthatthemillennialuseofvirtualmobilitymightexplainasignificantportioninthedropindrivingobservedamongthemembersofthisgeneration.Moregenerally,theubiquityofthesmartphoneadoptionalongwithincreasesinmobilityoptionshavecreatedaclassof“real-timeriders”(ITSAmerica2015)thatspansallcohorts.Still,millennialsarethefirstgenerationofso-called‘digitalnatives’(Prensky2001),havinggrownupwiththeinternet,andarelikelytobetheusersthatmostbenefitfromtheavailabilityofmoderntechnologiesandemergingtransportationtechnologies(includingthemodernsharedmobilityservices).This

Page 64: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

56

mayapplytofollowinggenerations,too,butisonlybecomingapparentwithmillennials(Lyons2014).Itisthusimportanttostudythecohorteffectsandexploretheimpactoftraditionalexplanatoryvariablessuchasthebuiltenvironmentandsocioeconomicfactorsandtheirlikelyeffectsonthetravelbehaviorofthemembersofdifferentgenerations.Inthissection,westudytheself-reportedVMTofmillennialsandGenerationX,andinvestigatetheimpactofmultipleexplanatoryvariables,includingsociodemographics,landusecharacteristicsandindividuals’attitudes,ontheself-reportedVMT.DependentVariable:Self-ReportedWeeklyVMT

Thefollowingsectionspresenttheresultsofamodelthatwasestimatedusingtheself-reportedweeklyVMTasthedependentvariable.TheweeklyVMTreportedbyindividualsrangedfrom0to1000,withameanof115milesperweek,andamedianof75.Informationforthisvariable,whichislikelytobeaffectedbysomelimitationstypicalofanyself-reportedmeasuresoftravelbehavior(e.g.eventualunder-reportingoftripsandVMT)wascollectedinasimilarwayforallrespondentsinthedataset.Weusedalog-transformationoftheVMTvariable,inordertoreducethedeviationfromthenormalityofthevariable.DuetothepresenceofcaseswithVMTequaltozero,andinordertoavoidtakingthelogarithmofzero,thefinaldependentvariablethatwasusedinthemodelwasln(VMT+1),asoftendoneinsimilarmodelsintheliterature.ExplanatoryVariables

Sociodemographic

Weusedseveralsociodemographicvariablesinourmodel.Thevariablesusedincludedbothindividualandhouseholdcharacteristics.Further,inordertoallownon-linearrelationshipsofVMTwithage,wealsoincludedaquadratictermfortheageoftherespondent(i.e.“squaredage”wasalsoincluded)inthemodel.Itisexpectedthatvehicletravelincreasesasadolescentsbecomeadults,peaksduringadultage(30-60years)whenemploymentandchildrearingresponsibilitiesaregreatest,andthendeclinesasindividualsretireandage(LeVine&Jones2012).Wealsotestedtheinclusionofageinsegmentedmodelstomodeltheeffectofageineachgeneration:millennialsandGenerationX.Weincludedoccupationoremployment,codedasstudentonly,workeronly,studentandworker,andunemployed.HouseholdincomewasalsoincludedasadeterminantofVMT.PreviousstudieshavefoundthatincomehasapositiveeffectonVMT(Brownstone&Golob2009;Rentziouetal.2012;Greeneetal.1995).Inthisstudy,weusedthreeannualhouseholdincomebrackets(respectively,lowerthan$35,000,$35,000-$100,000,andhigherthan$100,000)toallowincometohaveanonlinearrelationshipwithVMT.

Page 65: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

57

Builtenvironment

EwingandCervero(2010)summarizesthefindingsfromtheliteratureregardingtheeffectsofthebuiltenvironmentcharacteristicsonVMTandtravelbehavior.Inthisstudy,weusethegeocodedinformationontheresidentiallocationreportedbytherespondentstomatcheachcasewithadditionalinformationaboutthelocallandusecharacteristics,includingtheneighborhoodtypeasdeterminedinapreviousstudydevelopedbyresearchersatUCDavis(Salon2015).

Withinacensusblock,characteristicsmayvaryenoughthatresidentneighborhoodperceptionsandexperiencemayvary(Handy2002;Handyetal.2005;Bagleyetal.2002).Inaddition,notonlytheobjectivecharacteristicsofthebuiltenvironmentbutalsotheperceivedneighborhoodcharacteristicsarefoundtobegoodpredictorsoftravelbehavior(Handyetal.2006).Inthisstudy,weusedgrosspopulationdensity(people/acre),grossemploymentdensity(jobs/acre),jobdiversity,andtotalroadnetworkdensityasobjectivemeasuresofthelocalbuiltenvironmentcharacteristics.Bothpopulationandemploymentdensitywerepreviouslyfoundtobesignificant(Cervero&Murakami2010)inpredictingVMT.Inaddition,weusedregionaldiversity,basedonpopulationandtotalemployment,deviationoftheratioofjobs/popinacensusblockgroupfromtheregionalaverages,andtripproductionsandtripattractionsequilibriumindex.Thesevariablesaregoodmeasuresofthecharacteristicsofthelanduse.Wherethesemeasuresaremorebalanced,thelocalmixoflandusesisthoughttoreducetraveltimeanddistance(Cervero&Duncan2006).TechnologyAdoptionandUseofSocialMedia

Theadoptionofinformationcommunicationtechnology(ICT)hasbeenoftenreportedasapotentialfactoraffectingtravelbehavior,which,dependingonthelocalcontextandindividuals’characteristics,mayleadtosubstitutionof,generationof,modificationoforneutralitywiththeamountoftravel(SalomonandMokhtarian2008,CircellaandMokhtarian,2017).Inthisstudy,wecontrolledforseveralmeasuresofICTadoptionanduse.Inthefinalmodel,weuseavariablethatmeasuresthefrequencywithwhichtherespondentstelecommutefortheirworktoaccountforthepotentialimpactsoftelecommutingonweeklyVMT.NewmobilityservicessuchasUberandLyftmayfunctionsimilarly,generatingnewtripsandincreasingVMTorreplacingdrivingmiles(Tayloretal.2015;Hallock&Inglis2015;Shaheenetal.2015).Wecreatedvariablestoassesstherespondent’sfrequencyofusingnewsharedmobilityservices,including:on-demandrideservices(e.g.UberandLyft),carsharing(includingfleet-basedandpeer-to-peerservicessuchasZipcar,Car2GoandTuro),bikesharing,ridesharing(includingpeer-to-peercarpoolinganddynamicridesharingsuchasZimrideandcarpoolingthatarrangedviaFacebookorCraigslist).Inthestudy,wetransformedthefrequenciesofuseoftheseservices,whichwerereportedasordinalvariablesinthesurvey,intomonthlyfrequenciesbyassumingthat‘‘5ormoretimesaweek”canbeconsidered5timesaweek(5/7),“3-4timesaweek”canbeconsideredthreeandahalftimesaweek(3.5/7),‘‘1–2

Page 66: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

58

timesaweek’’canbeconsidered1.5timesaweek(1.5/7),‘‘1–3timesamonth’’becomes2timesamonth(2/30),“lessthanonceamonth”becomes3timesperyear,and“Iuseditinthepast”(butnotanymore)isapproximatedtozero.LifestylePreferencesandIndividuals’Attitudes

Asdescribedinanearliersectionofthisreport,weappliedfactoranalysisasadatareductiontechniquetoanalyzetheattitudinalvariablesinthesurveyandcompute17factorsscores.InthefinalVMTmodel,weincludedthefollowingfactorscores:

a. Establishedinlife:Individualswhoscorehighlyonthisfactorstronglyagreedwithstatementsincluding“I’malreadywell-establishedinmyfieldofwork”andtheytendedtodisagreewiththestatement“I’mstilltryingtofigureoutmycareer(e.g.whatIwanttodo,whereI’llendup).”

b. Preferenceforsuburbanneighborhoods(pro-suburban):Individualswhoscorehighlyonthisfactortendedtoagreewiththestatementsthatemphasizedthepreferenceforlivinginspacioushomesthatwerefurtherawayfrompublictransitandlivinginalocationwithlargeyardsandlotsofspacebetweenhomes.

c. Responsivenesstotheenvironmentalimpactsandpriceoftravel:Individualswhoscorehighlyonthisfactortendedtoagreewiththestatements“TheenvironmentalimpactsofthevariousmeansoftransportationaffectthechoicesImake”,“Iamcommittedtousingalesspollutingmeansoftransportationasmuchaspossible”,“ThepriceoffuelaffectsthechoicesImakeaboutmydailytravel”and“Toimproveairquality,Iamwillingtopayalittlemoretouseahybridorotherclean-fuelvehicle”.Thisfactorcapturesrespondents’willingnesstochangetravelplansbasedonbothgaspricesandenvironmentalconcerns.

d. “Mustownacar”:Individualswhoscorehighlyonthisfactoragreedwith“Idefinitelywanttoownacar”anddisagreedwiththestatement“Iamfinewithnotowningacar,aslongasIcanuseorrentoneanytimeIneedit”.

e. Time/modeconstrain:Thisfactorcapturestheattitudeofrespondentswhomorelikelydrivebynecessityasopposedtobychoice.Thisistheamalgamationoffourattitudinalstatements.Thosethatloadedpositivelyontothisfactortendedtoagreewiththefollowingstatements“Myschedulemakesithardorimpossibleformetousepublictransportation,”“IamtoobusytodomanythingsI’dliketodo,”and“Mostofthetime,Ihavenoreasonablealternativetodriving”.Respondentswhoscoredhighonthisfactormayhavenoreasonablealternativetodriving.Thisfactorislikelystronglycorrelatedwiththebuiltenvironmentcharacteristicsandneighborhoodtypewheretherespondentslive.

Results

Weestimatedweightedlog-linearmodelsusingthelog-transformationoftheself-reportedmeasureofweeklyVMTasthedependentvariable.Table10summarizestheestimatedcoefficientsforapooledmodel,whichincludesbothmillennialsandthemembersof

Page 67: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

59

GenerationX(N=1801),andtheseparatemodelsformillennials(N=976)andGenerationX(N=825).Allmodelshaveverysatisfactorygoodnessoffit,withR-Squaredbetween0.448and0.517(adjustedR-Squaredbetween0.439and0.509).Therefore,themodelsareabletoexplainapproximately50%ofthevarianceofthedependentvariable,avaluethatisremarkableconsideringthemanysourcesofpotentialnoisethataffectindividuals’VMTandthatcannotbeusuallycapturedineconometricmodels.Interestingly,themillennials’modelhasthelowestgoodnessoffit(R-squaredof0.448,comparedto0.517fortheGenerationX’smodel).Thisconfirmsthelargerheterogeneityinmillennials’mobilitychoices,andtheincreaseddifficultyofpredictingtheirbehaviors.Inotherwords,whileitiseasiertopredictGenXers’weeklyVMT,usingtherichsetofvariablesavailableforthisresearch.Moresourcesofnoise(e.g.impactofunobservedvariables,and/ordifferencesinindividualtastes,habits,etc.)characterizethemillennials’group.Still,ourmodeldoesaremarkablejobinexplainingthevariationinmillennials’VMT,duetotheabundanceofvariables,suchaslandusecharacteristics,individualattitudes,andadoptionoftechnology,whicharenotavailableinotherdatasets.Thefollowingsub-sectionsdiscusstheimpactsofthevariousgroupsofvariablesthatwerecontrolledforintheVMTmodels.Table10.ResultsofthePooledandSegmentedModeloflog(VMT+1)

PooledModel Millennials GenerationX

B p B p B p

(Intercept) 0.487 0.302 -5.253 <.001

1.162 <.001

Occupation StudentOnly 0.437 0.004

0.818 <.001

-0.441 0.155

StudentandWorker 0.73 <.001

0.839 <.001

0.608 0.001

WorksOnly 0.687 <.001

0.742 <.001

0.711 <.001

Sex(Male) 0.271 <.001

0.205 0.014

0.305 <.001

Age 0.038 0.163 0.453 <.001

Age2 -0.001 0.138 -0.008 <.001

HouseholdIncome >$100k 0.291 <.001

0.462 <.001

0.342 0.004

$35k-100k 0.155 0.031

0.194 0.042

0.21 0.062

LiveswithParents -0.235 0.003

-0.176 0.083 -0.383 0.004

LiveswithChildren 0.206 0.001

0.314 0.001

CarAvailability(%) 0.027 <.001

0.026 <.001

0.029 <.001

TelecommutingFrequency -0.506 0.001

-0.693 <.001

PopulationDensity -0.007 <.001

-0.013 <.001

Diversity -0.557 <.001

-0.325 0.072 -0.9 <.001

FSpro-suburban 0.054 0.064 0.066 0.085

FSresponsive_env_price -0.055 0.047

FSestablished_in_life 0.107 0.001

0.091 0.057 0.066 0.122

Page 68: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

60

FSmust_own_car 0.106 <.001

0.119 0.003

0.073 0.082

FStime_mode_constrain 0.165 <.001

0.153 <.001

Uber/LyftFrequency -1.407 0.05

Observations 1801 976

825 R2/adj.R2 .480/.474

.448/.439 .517/.509

Socio-demographics

Inourpooledmodel,aswellasinthesegmentedmodels,variablessuchashouseholdincome,gender,presenceofownchildreninthehousehold,andoccupation/employmentstatuswereallfoundtohaveastatisticallysignificanteffectonanindividual’sVMT.Interestingly,theeffectsofage(whichiscontrolledforthroughtheuseofboththeAgevariable,andthequadratictermAge2,tocontrolfornon-lineareffectsofageontheamountofcartravel)arefoundtobesignificantinthepooledmodelandinthemillennialmodelonly.ThefindingsconfirmtheassumptionthattherelationshipbetweenVMTandagedoesnotfollowalinearrelationship.Inparticular,theestimatedmodelcoefficientspredictthat(aftercontrollingfortheeffectsofothervariables,suchasHHincome,presenceofchildren,etc.)theeffectsofageonVMTappeartopeakatage35.Similarly,whenseparatingthetwosegmentsofthepopulationinthedataset,boththeAgeandAge2termsarenotsignificantintheGenerationXmodel,suggestingthattheremainingdifferencesinVMTattributabletoageamongthisgrouparenegligible(i.e.forindividualsintheagegroup35-50,individualVMThasalready“peaked”,andtheremainingchangesinVMTareexplainedthroughtheimpactofothervariables).Acrossallmodels,malerespondentshadhigherVMTsthanfemalerespondents,thoughtheeffectwasmuchsmallerinthemillennialmodel.Inthepooledmodel,mendrove30%moremilesperweekthanwomen,allelseequal,whileinthemillennialmodel,mendrove24%moremilesthanwomen.AmongthemembersofGenerationX,mendrove33%morethanwomen.Thismayindicatethatgenderdifferencesaresmallerwithinthemillennialgeneration,aswomenhavesaturatedtheworkforce(andtherearesmallergenderdifferencesinlifestyles,income,etc.)andmenshareinmorehouseholdobligations.Forthemillennialmodel,individualsthatwerebothemployedandstudentshadhigherVMTsthanindividualsthatworkedonly,orwerestudents–goingtobothworkandschool,assumingthattheyareindifferentlocations,resultsinahigherVMT.IntheGenerationXmodel,thosewiththehighestVMTsonlyworkedandwerenotstudents.Householdcompositionisfoundtobeaveryimportantfactoraffectingtheamountofindividuals’cartravel.Inparticular,individualsthatlivewiththeirparentstendtodrivefewermilesperweekthanthosethathavealreadyestablishedtheirownhousehold.Veryinterestingly,thepresenceofchildreninthehouseholdisfoundtobeaveryimportantpredictorofVMTformillennials:youngadultsthathavetheirownchildrentendtodrivemore(startingfromalowerbaselinevaluefortheirgeneration,comparedtotheolderGenXers)toaveryremarkableextent.Theeffectofhavingtheirownchildrenlivinginthehouseholdisalso

Page 69: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

61

foundtohaveasignificanteffectinthepooledmodel,butisnotfoundtobeasignificantpredictorofVMTforthemembersoftheGenerationX.

Caravailability(measuredasthepercentoftimeacarisavailabletotheindividual)wasalwaysfoundtobepositivelycorrelatedwithvehiclemilestraveled.Inthepooledmodel,foreachadditionalpercentincrementofcaravailabilitythereisa3%increaseinVMT.Thisvariablewasusedinplaceofthetypicalcarsperhouseholdorcarsperlicenseddriverasamorepreciseestimateofvehicleavailability.BuiltEnvironment

Theestimatedcoefficientsindicatethat,asexpected,populationdensityisnegativelycorrelatedwithvehiclemilestraveledinboththepooledmodelandGenerationXmodel.Thisisconsistentwithearlierfindingsintheliterature(Ewing&Cervero2001).Inthepooledmodeltheeffectofdensityisasmall,butsignificant:anincreaseinaunitofpopulationdensity,reportedinpopulationperacrepercensusblock,resultsinadecreaseinVMTof0.07%.However,thisvariablewasnotfoundtobesignificantinthemillennialsmodel.Regionaldiversity,measuredasthecensusblockgroupdeviationfromjobstopopulationratiofromtheregion’s,wasnegativelycorrelatedwithVMTacrossallmodels.Forexample,inthepooledmodel,aunitincreaseinregionaldiversityresultedinaVMTdecreaseof43%.ThisvariablehasevenlargereffectsamongGenerationX.Overall,theimpactoflandusecharacteristicsappearstobelargeramongtheoldergroup.Millennials’VMTseemtobeaffectedtoalargerdegreebyothergroupsofvariablesthatwerecontrolledforinthemodel.TechnologyAdoption

Wecontrolledfortheadoptionoftechnologythroughseveralvariablesinthemodelestimation.Inthefinalmodel,weincludeavariablethataccountsfortheeffectofthefrequencyoftelecommuting(fortheindividualsthateitherworkorworkandgotoschool),whichwasfound(notsurprisingly)tohaveastatisticallysignificant,andnegative,effectonVMTinboththepooledandtheGenerationXmodels.Veryinterestingly,thefrequencyoftelecommutingwasnotfoundtobesignificantintheVMTmodelformillennials.Whethermillennialsadopttelecommutingornot,thisdoesnotseemtohaveasignificanteffectonVMT,perhapsbecauseofthepotentialsubstitutionofcommutetripswithcartripsdoneforotherreasons.

Wealsocontrolledfortheimpactoftheadoptionofnewsharedmobilityservices.Inparticular,inthefinalmodel,weincludedavariablethataccountedforthefrequencyofuseofon-demandrideservicessuchasthoseprovidedbyUberorLyft.Thefrequencyofuseoftheseserviceswasfoundtohaveasignificant(ata0.10levelofsignificance)andnegativeeffectonmillennials’VMT.Thissuggeststhatmillennialswhouseondemandrideservicestendtodriveless.Thedirectionofcausalityofthisfindingisunclearthough:theadoptionofon-demandrideservicesmightleadsomemillennialstodriveless(asaconsequenceoftheadoptionoftheseservices,and/orthesubstitutionoftripsthatwouldhavebeenotherwisemadebydrivingtheircar),orthereversemightbealsotrue:millennialsthathaveloweraccesstoaprivatevehicle

Page 70: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

62

(e.g.theyliveinzero-orlow-vehicle-owninghouseholds)mightadopttheseservicesmoreoften,asawaytocompensatefortheirlowerautoaccessibility.Thistopicwillbefurtherinvestigatedinfutureextensionsoftheresearch,throughtheapplicationoflatentclassanalysisandtheestimationoflatentclassmodelstoanalyzedifferentbehaviorsamongdifferentgroupsofusers,andtheestimationofbivariatemodelsthatcanjointlyestimateanindividual’samountofcartravel(e.g.VMT,orthefrequencyofuseofpublictransportation)andthefrequencyofuseofmodernsharedmobilityservices(includingon-demandrideservices,suchasUberandLyft).

PersonalAttitudesandPreferences

Weusedseveralfactorscoresthatwerecomputedinthefactoranalysisofattitudinalvariables,tocontrolfortheimpactofindividualattitudesandpreferencesontheindividual’samountofcartravel.Inparticular,thefactorscoresmeasuringtheindividuals’perceivedlackofalternativestodriving,theirdegreeofresponsivenesstotheenvironmentaleffectsandpriceoftraveloptions,thedegreetheyfeeltheyarewellestablishedinlife,thepreferencetoownacar(vs.accessingonewhenneeded),andthepreferenceforsuburbanneighborhoodswerefoundtohavesignificanteffectsandwereincludedinthefinalpooledmodel(andinseveralcasesalsointhesegmentedmodels).

AllattitudinalfactorscoreswerefoundtohaveanimportanteffectinexplainingindividualVMT.IndividualsthatreportedthattheydonothavereasonablealternativetodrivingwerefoundtoreporthigherVMT.Further,VMTwasfoundalsotoincreasewiththedegreebywhicharespondentfeelsestablishedinlife(aone-unitincreaseinthisfactorresultedinan11%increaseinVMT).Theseindividualslikelyhavemoreresponsibilities,havehighersocioeconomicstatusandarebetterestablishedintheircareers,resultinginhigherVMT.Formillennials,inparticular,thosethathaveaunithigherscoreforthisvariabledrive9.5%more,whilethisvariableisnotsignificantintheGenerationXmodel.

Attitudeswereimportanttocontrolforasaproxyforresidentialself-selection,whichisoftenaconfoundingfactorintherelationshipwithVMTandstructuralvariables.Interestingly,thefactorscoremeasuringthedegreebywhicharespondentisresponsivetotheenvironmenteffectsandpriceoftraveloptionswasfoundtobesignificantonlyinthepooledmodel(withtheexpectedsign).Theweaksignificanceofthisvariablemayindicatethatconsideringtheenvironmenteffectsoftraveloptionswhenchoosingonwhethertodrivemightnothavesizableeffectsonone’sVMT,orthatthiseffectisalreadycapturedbyanothervariable,suchasresidentialselectionorcaravailability,orchoosingtoownacaringeneral.

ThosewholikecarsanddefinitelywanttoownonearemorelikelytohavehigherVMTsthanthosewhodonotloadpositivelyontothatfactor–theimpactofthisvariableislargerformillennials(andisnotfoundtobesignificantintheGenerationXmodel).Similarly,thefactorscoreforthepro-suburbanattitudewaspositivelycorrelatedwithVMTinthepooledandintheGenerationXmodel.Anincreaseofoneunitinthisfactorscore(beingmore“prosuburbs”)isassociatedwithanincreaseof5.8%inVMT.ThefullanalysiscanbefoundinTiedemanetal.(2017).

Page 71: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

63

CarOwnership,VehicleTypeChoiceandPropensitytoChangeVehicle

Ownership

Morethan17.4millionvehiclesweresoldintheUnitedStatesin2015,breakingthepreviousrecordof17.3millionvehiclessoldin2000(HarwellandMufson2016).Therecentincreaseincarsaleshaspromptedspeculationsonwhetherthecarmarkethasdefinitelyreboundedafterthetemporarydecreaseincarsalesduringtheyearsofeconomicrecession,10thoughacertain“delay”effectmightalsobebehindtherecordvolumesofcarsalesin2015:vehiclessalesduringtheyearmighthavebeengrownalsobecausemanyconsumerspostponedthetimeofreplacementoftheirvehiclesduringtheeconomiccrisis.

Figure25.Averageratioofavailablecars(includingminivans,SUVs,pickuptrucks)per

householdmemberwithadriver’slicense

10Thediscussionabouttheapparent“peak”incarsalesobservedduringthepastfewyearshasalsobeenconnectedwiththeobservedpeakincartravelthatwasobservedduringtheearly2000s.ForamorecompletediscussionofthefactorsassociatedwiththeobservedchangesinpassengertraveltrendsintheU.S.seeCircellaetal.(2016b).

Page 72: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

64

Severalfactorsaffectvehicleownershiprate,suchasindividualandhouseholdcharacteristics,availabilityandaccessibilityofdifferentmodesoftransportation,thequalityoftravelofferedbycarvs.theotheralternatives,characteristicsofbuiltenvironment,individual’slifestylesandpersonalattitudesandpreferencestowardstheuseofcarsand/orothermodes.Inthissection,wefirstlookattwodifferentmeasuresofcarownershipusingthedataavailableforthisproject:(1)averageratioofavailablecars(includingminivans,SUVs,pickuptrucks)perhouseholdmemberwithdrivinglicense,and(2)averageratioofavailablevehiclestohouseholdmembers.Wethendevelopamodelofvehicletypechoiceandanalyzethefactorsthataffectthedecisiononwhatvehicletoown.Figure25presentsthedistributionoftheratioofcarsperhouseholdmemberswithadriver’slicensebyagegroupandneighborhoodtype.Asdiscussedearlier,millennialswholivewithparentsareexpectedtobehavedifferentlycomparedtomillennialswhodonotlivewiththeirparentsandtheyhavealreadyestablishedtheirindependenthousehold.Asshowninthegraph,exceptdependentmillennialswholiveinsuburbanareas,theaveragenumberofvehiclesperhouseholddriverdecreasesfromruralneighborhoodtosuburbanandurbanareas.Thiscouldbeduetohigheravailabilityandaccessibilityofdifferenttravelalternativesandhighercostofcarownershipinurbanareas.Further,andmoreinteresting,theratioofvehiclesperdriversissensiblylowerforonecategory:theindependentmillennialswholiveinurbanareashavethelowestratioofcarperhouseholddrivers.Thisgroupofindividualsistheonlyonethathasanaverageratioofcarsperhouseholddriversthatislowerthan0.8.Incontrast,allgroupsofGenXershavemuchhigherratiosofvehiclesperhouseholddrivers(whichforruralGenXersisapproximatelyequalto1).

Page 73: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

65

Figure26.Averageratioofavailablecars(includingminivans,SUVs,pickuptrucks)per

householdmember

Incontrasttotheratioofcarsperhouseholddrivers,theratioofcarsperhouseholdmembers(Figure26)variesinasmallrangeacrossalldifferentagegroupsandneighborhoodtypes.Theresultindicatesthatbothdependentandindependentmillennialswholiveinurbanneighborhoodshavelowercaravailabilitycomparedtotheirpeerswholiveinruralandsuburbanareas.Inthiscase,thelowerratioofcars/householdmembersisobservedamongdependentmillennials.Duringfuturestagesoftheresearch,wedoplantostudyhowcarownershipvariesacrossdifferentgroupsofthepopulation:theresearchteamiscurrentlyworkingattheestimationofcarownershipmodelsthatinvestigatehowvarioussociodemographiccharacteristics,individualpreferences,andlandusefeaturesaffecthouseholdcarownership.Further,animportantfocusoftheresearchhasbeenfocusedonwhataffectsthetypeofvehiclesthatindividuals(andthehouseholdsinwhichtheylive)prefertoown.Withcheapergasandastrongereconomy,consumersareflockingtonewandusedcarlotslookingfortheirnewcar.Recenttrendshavealsoshownaresurgenceinvehiclesalesforlargervehicles

Page 74: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

66

(includingSUVs,crossoversandpick-uptrucks).Inthispart,weexamineindividualswhoownatleastonevehicleinthehousehold,inordertobetterunderstandhowindividualattitudes,lifestyles,builtenvironmentcharacteristics,andsocio-demographictraitsaffectthetypeofvehicletheyown.Asalreadymentioned,therecenttrendsincarsalessomehowcontrasttheobservedtrendsinvehicleuseandsalesfromthepastfewyears,whichshowedanapparentpeakincarownershipanduseintheUnitedStatesaswellasotherdevelopedcountries(SchoettleandSivak2013;Kuhnimhof,Armoogum,etal.2012).Thistrendhasbeenevenstrongeramongyoungadults,ormillennials.Severalstudieshavereportedthatyoungadultstendtodelaydrivinglicensure,ownfewerornovehicles,anddrivelessevenwhentheyhaveaccesstoacarinthehousehold(McDonald2015;Polzin,Chu,andGodfrey2014;Blumenbergetal.2012).However,todate,therehavebeennostudiesthatspecificallyfocusedoninvestigatingthevehicleownershipandvehicletypechoiceamongyoungadults.Alargervarietyofvehicletypes,includingsedan,hatchback,two-seater,pick-uptruck,SUV,minivan,coupe,etc.arenowadaysavailableonthemarket.However,ratherlimitedknowledgeexistsonthemotivationsaffectingbuyersofthesevehicles,beyondtheimpactofpurchasepriceandvehiclecharacteristicssuchasnumberofseats,operatingcosts,etc.Ourstudyaimstocontributeclosingthisgapbyinvestigatingtheeffectsofindividualattitudesandpreferences,generationaldifferences,andindividualcharacteristicsonvehicletypechoice.Veryfewauthors,todate,haveinvestigatedtheimpactsofattitudesandpreferencesonvehicletypechoice(BaltasandSaridakis2013;ChooandMokhtarian2004).Furthermore,nostudyhassofarlookedatgenerationaldifferencesinvehicletypechoice.Sincethe1980s,researchershavebeenexaminingvehicletypechoice(BeggsandCardell1980;BerkovecandRust1985;ManskiandSherman1980;LaveandTrain1979).Inordertomodelvehicletypechoice,studiestypicallyuseeithermultinomiallogit(MNL)(ChooandMokhtarian2004;Kitamuraetal.2000;LaveandTrain1979;ManskiandSherman1980;FredManneringandWinston1985)ornestedlogitmodels(BerkovecandRust1985;F.Mannering,Winston,andStarkey2002).LaveandTrain(1979)usedMNLtoinvestigatethevehicletypepurchasedandfoundthatlargerhouseholdsthathavetwoormorevehiclesaremorelikelytochoosesmallercars(LaveandTrain1979).Moreover,theyfoundthatolderpeopleandhouseholdswithhighVMTtendtochooselargervehicles.Unsurprisingly,vehiclepricenegativelyaffectsthechoiceofeachtypeofvehicle(LaveandTrain1979).ManskiandSherman(1980)werethefirstresearcherstotrytomodelthenumberofvehiclesandvehicletypechoicesimultaneously(ManskiandSherman1980).InestimatinganMNLmodel,theauthorsfoundthatseatingandluggagespacepositivelyaffectthevehicletypechoice,andinparticularlargerone-vehiclehouseholdsandhouseholdswithlowincomearelesslikelytochoosevehicleswithhigheroperatingcosts(ManskiandSherman1980).Similarly,ManneringandWinston(1985)developedamultinomiallogitmodeltomodelthechoiceamong10vehicletypealternativesbasedonyear,make,andmodel(FredManneringandWinston1985).Theyfoundthatbrandloyaltyhasasignificanteffectonthechoiceofthehousehold’svehiclemake.SimilartothefindingsofLaveandTrain(1979),vehiclepurchasepriceandoperatingexpendituresnegatively

Page 75: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

67

affectthechoiceofavehicletype(FredManneringandWinston1985).Kitamuraetal.(2000)usedamultinomiallogitmodeltoinvestigatethechoiceofvehiclebodytype(e.g.4-doorsedan,2-doorcoupe,etc.)andfoundthatmalesaremorelikelytousepick-uptrucks,andyoungerindividualsweremorelikelytouseSUVs,pick-uptrucks,andsportscars(Kitamuraetal.2000).Unsurprisingly,largerhouseholdsaremorelikelytousevansorwagonsasthesetypesofvehicleshavelargerspaceandseatingcapacity(Kitamuraetal.2000).Eventhoughseveralresearchershaveexploredthefactorsaffectingahousehold’svehicletypechoice,theliteratureismorelimitedregardingtheimpactofindividualattitudes,preferences,andlifestylesonthischoice.Amongthestudiesthatinvestigatedtheimpactofattitudinalvariablesonvehiclechoice,ChooandMokhtarian(2004)foundthattravelattitudes,personalitytraits,andlifestyleshavesignificanteffectsonthevehicletypechoice(ChooandMokhtarian2004).Morespecifically,peoplewholiveinhighdensityareasaremorelikelytodrivemoreexpensivecars,suchasluxuryandluxurySUVs(ChooandMokhtarian2004),andadislikeoftravelispositivelyassociatedwithdrivingaluxuryvehicle(ChooandMokhtarian2004).BaltasandSaridakis(2013)developedamultinomiallogitmodeltomodelthechoiceof12mutuallyexclusivevehicletypealternatives(BaltasandSaridakis2013).Theywerethefirstresearcherstodemonstratethatthepurposeofcaruse,theconsumer’sinvolvementwithcars,andtheconsumer’sattachmenttocars,havesignificanteffectsoncartypechoice.Further,theirmodelshowedthatthepropensitytopurchaseasmallcarisstatisticallyrelatedtotheirrelianceonfriendsandfamilymembersforadvice.SimilartothefindingsofChooandMokhtarian(2004),BaltasandSaridakis(2013)foundthatthosewhopreferluxuryvehiclesaremorelikelytoliveinurbanareas.Despitetherecentinterestoftheliteratureininvestigatingthebehaviorofthemillennialgeneration,toourknowledge,nopreviousworkhasbeendoneinvestigatingthevehicletypechoiceofyoungadults.VehicleTypeChoiceModel

Forthisanalysis,weestimatedamultinomiallogitmodel(MNL)toexploretherelationshipamongvehicletypechoice(thedependentvariableinthemodel)andsocio-demographiccharacteristics,residentiallocationandlandusecharacteristics,andpersonalattitudesandpreferences.Weusedonlyasubsetofthedatainestimatingthismodel,foranumberofreasons.First,werestrictedtheanalysestotheindividualswhoindicatedthattherewasatleastonevehicleinthehouseholdandprovidedvalidyear,make,andmodelinformationfortheprimaryhouseholdvehicle.Second,inordertofocusontherespondentsthatcurrentlyownavehiclethatwaspurchasedbythemunderconditionsrathersimilartotheircurrentlivingconditions,andremovethepossiblebiasofrespondentswhoweregiftedcarsfromotherfamilymembersorpurchasedanoldcaroutofcontingencies(e.g.itwasoneofthefewavailableinalimitedpricerange)ratherthanchoosingitbasedonpersonalpreferencesandtastes,wenarrowedthesubsetofanalysistotheindividualswhoownedorleasedausedornewvehiclethatismodelyear2010ornewer.

Page 76: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

68

DependentVariable:VehicleType

Surveyrespondentswhoindicatedthattherewasatleastonevehicleinthehouseholdwereaskedaquestionabouttheyear,make,andmodelofthevehiclethattheyusemost.Toassigneachvehicletoavehicletype,weusedtheEnvironmentalProtectionAgency’s(EPA)FuelEconomyDatasetwhichprovidesvehicleclassificationdataforallconsumervehiclesfrom1984tothepresent(EPA2016).Wematchedeachcompleteyear,make,andmodel,withacorrespondingvehicleclassificationbasedontheinformationprovidedbytheEPAdataset.Byusingthevehicle’smodelyear,wewereabletotakeintoaccountmodelredesignsthatinsomecasesmovedvehiclesfromonevehicletypetoanother.Forexample,the1984HondaAccordisclassifiedintheEPAdatasetasasubcompactcarbutthe2016HondaAccordisclassifiedasamidsizecar11.TheEPAhasmorethan15differentvehicletypeswhenaccountingforthedifferentdrivetrainoptions.Aswedonothaveinformationaboutthetrimlevelordrivetrainofthevehiclemodelinoursurvey,weaggregatedsomevehicletypessuchasSportUtilityVehicle2WDandSportUtilityVehicle4WDinjustonecategoryregardlessofthespecifictrimlevelordrivetrainthateachvehiclehas.Forthisanalysisweusedsixdifferentvehicletypechoices:

1. Small/compact2. Midsize3. Large4. Luxury5. SUV6. LuxurySUV

Weexcluded“pick-up”trucksand“sportcars”fromtheanalysisduetothesmallnumberofpick-uptrucksandsportcarsownedbytherespondentsinoursample,andtheverydifferentcharacteristicsofthesevehicles,whichwouldhavesignificantlyincreasedtheheterogeneityofanyonevehicleclassification(ifthevehicleswereincludedinthatcategory).Asmallnumberofrespondentsthedatasetreportedthattheyownseveralvehiclesthatcanbeclassifiedas“crossovers”or“minivans”.ThesevehiclesweremergedintheSUVcategory,duetothesimilarsizeofthesevehicles,andthemanysimilaritiesandoverlapsamongthevehiclesthatbelongtothesecategories.Sociodemographics

Weincludedindividualandhouseholdsocio-demographicandsocio-economiccharacteristicsasexplanatoryvariables.Wecontrolledforagethroughtheuseofthe“age”variable.Tocontrolforthenon-linearityofageinthismodel,wealsoincludedan“age-squared”variable.Wealsocontrolledforhouseholdcompositionthroughseveralvariablesincludingthenumberofchildrenandadultsinthehousehold.Householdswithchildrenareexpectedtomorelikely

11Pleasenotethatonlymodelyear2010ornewerwereincludedintheanalysisofthispaper.Theexamplepresentedhereisonlyforexplanatorypurposesontheprocessthatwasusedintheresearch

Page 77: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

69

ownlargervehicles,vansandSUVs(thelasttwocategoriesaremergedunder“SUV”inthisanalysis)duetotheirincreasedseatingcapacityandcomfortforriders.Finally,ascustomaryinmodelsofthistype,wecontrolledfortheimpactofothersocio-demographicvariablessuchasgenderandhouseholdincome(expectinghouseholdincometobeanimportantdriverforthepurchaseoflargerandmoreexpensive/luxuryvehicles).ResidentialLocationandLandUseCharacteristics

Inadditiontocontrollingforthetraditionalsocio-demographicandsocio-economicvariables,wealsocontrolledforthecharacteristicsoftheresidentiallocationthroughtheuseofaninteractionterm,whichallowedtheimpactoftheannualhouseholdincometovaryforthehouseholdsthatliveinurbanneighborhoods(usingthenon-urbanHHsasthereferencecategory).Weexpectedthat,holdingallelseequal,thosewholiveinurbanneighborhoodswouldbemorelikelytoownsmall/compactvehiclesandlesslikelytoownSUVs.IndividualPreferencesandAttitudes

Aspreviouslydescribed,thesurveyincluded66separatestatementsthatwereincludedinthestudytomeasuretheindividual’sattitudesaboutanumberofdimensionsrelatedtotheenvironment,travel,adoptionoftechnology,multi-tasking,lifesatisfaction,landuse,theroleofgovernment,etc.fromwhichweextracted17attitudinalfactors.Weincludedthreefactorscoresasexplanatoryvariablesinthefinalvehicletypechoicemodel:

a. Utilitariancaruse(carasatool):Individualswhoscorehighonthisfactortendedtoagreewithstatementssuchas“Thefunctionalityofacarismoreimportanttomethanitsbrand”.

b. Establishedinlife:Individualswhoscorehighlyonthisfactorstronglyagreedwithstatementsincluding“I’malreadywell-establishedinmyfieldofwork”Theytendedtodisagreewiththestatement:“I’mstilltryingtofigureoutmycareer(e.g.whatIwanttodo,whereI’llendup).”

c. Individualswithmultipletransportationmodesavailableandnotimerestraints(Reversedtime/modeconstrained).Thiscapturesrespondentsthatfeelasthoughtheyhavemultipletransportationoptionsavailabletothemandarenotconstrainedbytime.Thosethatloadedpositivelyontothisfactortendedtodisagreeorstronglydisagreewiththefollowingstatements:“Myschedulemakesithardorimpossibleformetousepublictransportation,”(indicatingthattherescheduledoesNOTmakeithardforthemtousepublictransit)“IamtoobusytodomanythingsI’dliketodo,”(indicatingthattheyareNOTtoobusytodotheactivitiesthattheywouldliketodo)and“Mostofthetime,Ihavenoreasonablealternativetodriving”(indicatingthattheyDOhavereasonablealternativestodriving).Theserespondentsmayhavenoalternativetodriving.

Page 78: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

70

Results

Sinceourdependentvariable,primaryvehicletype,consistsofsixmutuallyexclusivecategories,wedevelopedamultinomiallogitmodelforvehicletypechoice.Asmentionedintheprevioussection,thesesixcategoriesare:Small/Compact,Midsizecar,Largecar,SUV,Luxury,andLuxurySUV.Thefinalmodelhasfivealternativespecificconstantsand22alternativespecificvariablesthatrepresentninedifferentvariables.Thetablebelowpresentstheestimatedcoefficients(withtherespectivep-valuesinparentheses).Therho-squaredvalueofthefinalmodelis0.252,whichisquitegoodforamodelofthistype.Incomparison,therho-squaredforthemarketsharemodelis0.116,whichindicatesthatthemodelwithonlytheconstantsexplainsabout12%oftheinformationinthedata,andthatourfullmodelisabletocontributesignificantlytoexplainingthechoiceofvehicletype,despitetheobviousdifficultiesassociatedwiththeheterogeneityinthechoicesofvehicletype,andimpactsofeventualunobservedvariablesthatmightaffectthechoiceofthevehicleoneowns.Inparticular,aspointedoutinpreviouspapersintheliterature,thechoiceofthevehicletobuyisusuallyachoicethatismadeatthehousehold,andnotindividual,level.Additionally,thechoiceofthevehicletobuyisaffectedbytheothervehicle(s)thatthehouseholdeventuallyowns(orplanstopurchaseinthenearfuture).Thus,thechoiceofthevariousvehiclesthatareownedbyahousehold(forhouseholdsthatownmorethanonevehicle)isajointchoice,andshouldbemodelassuch.Unfortunately,inthisdatasetweonlyhaveinformationonthenumberofvehiclesowned/leasedbyahousehold(and,therefore,weknowoftheeventualpresenceofothervehicles,inadditiontothe“primary”vehicle),butwedonothaveinformationaboutthetypeofvehiclesthatareowned,apartfromtheprimaryvehicle.Thissomehowlimitstheabilityofthemodeltopredictthechoicesofahouseholdthatmightdecide,forexample,toownaSUVandacompactcartofulfilltheirmobilityneeds.12Despitethislimitation,theestimatedmodelprovidessomeusefulinformationontherelationshipbetweenvariousgroupsofvariableandthetypeofprimaryvehiclethatanindividualowns.

12Inourdataset,theinformationaboutsuchaHHwouldbeincludedas“owninganSUVastheprimaryvehiclethatisusedmostoftenandanotherunknownvehicle”,oras“owningacompactcarastheprimaryvehiclethatisusedmostoftenandanotherunknownvehicle”.

Page 79: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

71

Table11.EstimatedCoefficientofVehicleTypeChoiceModel

DependentVariable:VehicleType

Small/Compact

Midsize LargeCars

SUV Luxury LuxurySUV

Age 0.059(0.013)

(base) 0.252(0.000)

0.222(0.000)

1.176(0.000)

Age2 -0.001

(0.045)(base) -0.003

(0.001)-0.002(0.000)

-0.014(0.000)

Female (base) 0.603(0.000)

Numberofchildrenunder18

yearsoldinthehousehold

-0.266(0.048)

(base) 0.488(0.000)

-0.521(0.014)

FSCarasatool 0.355(0.005)

(base)

FSTime/Modeconstraint

(reversed)

-0.35(0.007)

(base) -0.584(0.047)

FSEstablishedinlife

-0.242(0.067)

(base) 0.5(0.025)

Householdincome 0.109(0.084)

(base) 0.22(0.014)

0.479(0.001)

InteractionHHIncomewith

urbanneighborhoodtype

(base) 0.121(0.026)

-0.16(0.078)

Constant -0.911(0.000)

(base) -6.181(0.000)

-5.646(0.000)

-2.561(0.000)

-29.696(0.000)

Numberofobservations 529 Log-likelihoodat0 -947.84 Log-likelihoodatmarketshare -801.05 Log-likelihoodatconvergence -708.53 !"#% ('()*+,-(!"#% ) 0.252(0.200)!/0% ('()*+,-(!/0% ) 0.116(0.088)Note:p-valuesarereportedinparenthesesbelowtheestimatedcoefficientsThesocio-demographiccharacteristicsusedinthemodelprovideinterestinginsightintovehicletypechoice:weusedageandagesquaredtoallowforanon-linearrelationshipofthisvariablewiththechoiceofcertainvehicletypes.Asshowninthemodel,theprobabilitythatanindividualownsalargecar,SUV,orLuxurySUVincreaseswithage.Similarly,thosewhoareolderaremorelikelytoassociatedhigherutilitywithandownasmall/compactvehicle(probably,asaneffectoftheHHvehiclefleetcomposition,asdiscussedabove),eveniftoalesserdegree.ThosewithchildrenlivingathomearemorelikelytoownSUVs(and/orvans,whichwerealsoincludedinthiscategory)andlesslikelytoownsmall/compactandluxuryvehicles:parentsneedtheutilityofanSUVwhichisnotofferedbysmallervehicles.Parentsaremorelikelytoassociatevaluewiththeseatingspace,storagecapacity,andgeneralcomforttypically

Page 80: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

72

associatedwiththesevehicles.Alsolookingathouseholdincome,ourmodelshowsthathigherhouseholdincomehasapositiveimpactonthelikelihoodtoownasmall/compactvehicle,aluxuryvehicle,andaluxurySUV.Whiletheimpactofhouseholdincomeonluxurybrandvehicles(eithercarsorSUVs)isprettystraightforward,theimpactofhouseholdincomeonthelikelihoodtoownasmall/compactcarislikelyassociatedwiththejointchoiceofthemultiplevehiclesownedbymoreaffluenthouseholdsdescribedabove.Forinstance,inahighincome2car–2personhousehold,thesurveyrespondentmayhaveequalaccesstobothvehicles;however,onevehicleismainlyusedbythespouse,leavingtherespondentwiththeothervehiclewhoseinformationisreportedinthesurvey.WomenwerefoundtobemorelikelytoownluxurySUVs.ThisreaffirmsfindingsfromarecentEdmonds.comstudywhichfoundthatwomennowaccountfor41%ofnewluxuryvehiclepurchases(http://www.detroitnews.com/story/business/autos/2016/09/06/women-buying-luxury-vehicles/89936258/).Theinclusionoffactorsextractedfromtheattitudinalvariablesprovideimportantinsightsintofurtherunderstandingvehicletypechoicebehavior.Asdescribedinthemethodologysection,weincludedthreefactorsasexplanatoryvariablesinthemodel.Theinclusionofthe“Establishedinlife”factorwasanattempttocapturetheeffectofstageoflike(inparticular,arelevantvariabletocaptureyoungermillennials’behaviorsandlifestyles).Inthisinstance,thosewhohavehighervaluesforthisfactoraremorelikelytoownluxuryvehiclesandlesslikelytoownsmallorcompactvehicles.Thisresultisnotsurprising,consideringthatluxuryvehiclesareexpensiveandindividualswhoaremorecertainaboutlife(andperceivethattheyarelessinatransientandunstablestageoftheirlife)havearemorelikelytobeabletopurchasemoreexpensivevehicles.Thosewhorecognizehigher“utilitarian”valuetotheuseofacar(i.e.havehigher“carasatool”factorscores)aremorelikelytoownsmallorcompactvehicles.Small/compactvehicles,inmostcases,donotfillanichemarketandtheyaresimplyseenasawaytogetfromorigintodestinationwhileminimizingpurchaseandmaintenancecost;theyarenotascomfortableasluxuryvehiclesandtheydonotprovidethespaceofanSUV.LandUseCharacteristics

Theinteractiontermofhouseholdincomeandurbanneighborhoodtypewasincludedinthemodelasawaytoaccountforthedifferentbehaviorofurbanhouseholdsregardingthechoiceofthevehicletoown.13Inadditiontothebaseeffectofhouseholdincomeonthevehicletypechoicethatwasdiscussedearlier,wefindthat,notsurprisingly,individualswithhighhouseholdincomesthatliveinurbanneighborhoodsaremorelikelytoownluxuryvehiclesandlesslikelytoownluxurySUVs.Theseeffectsareintroducedinthemodelascorrectionstothebaseeffect

13Inadditiontotheimpactonthetypeofvehiclethatisowned,landusecharacteristicsareexpectedtoaffectthenumberofvehiclesthatareownedbyahousehold.Weplantoexplorethisrelationshipinfuturestepsoftheresearch,throughtheestimationofacarownershipmodelthataccountsfortheimpactofindividualandlandusecharacteristicsonthenumberofvehiclesownedbyahousehold.

Page 81: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

73

ofthehouseholdincomeonvehiclechoice,meaningthathigherincomehouseholdsthatliveinurbanareasarenotaslikelytoownaluxurySUVasthehigherincomehouseholdsthatliveinotherneighborhoodtypes(thoughtheyarestillmorelikelytochoosethesevehiclesthanlowerincomehouseholds),andtheyareevenmorelikelytoownaluxurycar(andnotanSUV)thanthehighincomehouseholdsthatliveinotherneighborhoods.MoredetailscanbefoundinBerlinerandCircella(2017).PropensitytoModifyVehicleOwnership

IntheCaliforniaMillennialDataset,wealsocollectedinformationabouttherespondents’self-reportedwillingnesstobuy/leaseavehicle(Figure27)andtheirpropensitytosell/getridoftheircurrentlyownvehiclewithinthenextthreeyears(Figure28).AsshowninFigure27,millennialsingeneral,andoldermillennialsinparticular,moreoftenreportthattheyaremoreinclinedtopurchase/leaseacarwithinthenextthreeyears,comparedwithotheragegroups.Thisisconsistentwithexpectations,becausemillennials,particularlymillennialswholivesinurbanneighborhood,havelowercaravailabilitycomparedtotheiroldercounterpart,whohasalreadyacquiredavehicleorhashigheraccessibilitytoacarotherwiseownedinthehousehold.Thistrendalsoconfirmsthatveryoftenmillennialsareinatransientlifestage,andtheirzero-orlow-carownershipmightbeonlyatemporaryfactor,subjecttochangeduringtheirnearfuture.Thefindingmayhave,inparticular,consequencesonthecarownershipstatusofurbanmillennials.Thisgroupofyoungadultsareoftenfoundtoliveindenseneighborhoodandnottoownacar.Highexpectationshavebeenposedonthisgroupineventuallycontinuingtotransformthefutureoftransportation,andeventuallyhelpinthetransitiontowardsmoresustainablemobility.However,thehighpropensitytopurchaseacarduringthenextthreeyearsoftherespondentsincludedinthisgrouprepresentapotentialthreattosomesustainabilitygoals,andsignalsthatprobablymostpartofthelowcarownershipstatusofthisgroupisnotlikelytolastastheseindividualsageandtransitioninthefollowingstagesoftheirlife.

Page 82: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

74

Figure27.Distributionofindividual’swillingnesstopurchase/leaseavehiclewithinthenext

threeyearsbyagegroupandneighborhoodtype

Figure28presentsindividual’spropensitytosell/replaceoftheircurrentlyownedvehiclewithinthenextthreeyears.Asindicatedinthegraph,carownershipdecreasesfromruraltosuburbanandurbanneighborhoodsamongbothmillennialsandGenXers.Interestingly,thepropensitytosellacarishigheramongmillennialswhoownacarandlivesinurbanneighborhoodcomparedtoboththeirolderpeerswholiveinthesameareas,andtomillennialswholiveinotherneighborhoodtypes.Incontrast,thewillingnesstoreplacetheircurrentvehiclesishigheramongthemembersofGenerationX,inparticularamongthosewholiveinruralneighborhood.Weplantofurtherinvestigatethetopicsthataresummarizedinthesefigures,throughthedevelopmentofmodelsofthepropensitytochangethelevelofvehicleownershipinthehousehold,andinvestigatethefactorsaffectingthesetrends.Thistopicandthetypeofvehiclethattherespondentswouldconsiderbuying,asalsoreportedinthesurveyareofpotentialinteresttoautomakersandplanningagencies.Theywilllikelyaffectfuturedemandforcarsalesanduse.Further,infuturestagesoftheresearch,weplantoinvestigatetherelationships

0% 20% 40% 60% 80% 100%

Urban

Non-Urban

Urban

Non-Urban

Urban

Non-Urban

Urban

Non-Urban45

to 5

035

to 4

425

to 3

418

to 2

4

Yes

No

Not sure

Page 83: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

75

betweentheadoptionofsharedmobilityservicesandthepropensityofrespondentstomodifytheirlevelofvehicleownership.14

Figure28.Distributionofindividuals’propensitytosell/getridoftheirvehiclewithinthenext

threeyearsbyagegroupandneighborhoodtype

14Thiswillprovideadditionalinformationonthelikelychangesincarownershipanduse,astheadoptionofsharedmobilityservicesbecomemorepopularinfutureyears.

0% 20% 40% 60% 80% 100%

Rural

Suburban

Urban

Rural

Suburban

Urban

Gen

XG

en Y

No, I don’t currently own a vehicle

No, I don’t plan to sell/get rid of any of the vehicles in my household

Yes, I plan to reduce the number of vehicles in my household

Yes, I plan to replace my current vehicle with another one

I am not sure

Page 84: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

76

ConclusionsandNextStepsoftheResearch

Millennialsincludeaverylargesegmentofthepopulation,whooftenareearlyadoptersofnewtrendsandtechnologiesthatlaterareadoptedbyothersegmentsofsociety.Thus,improvingtheunderstandingofthefactorsandcircumstancesbehindmillennials’mobilitychoicesisofoutmostimportanceforscientificresearchaswellasforplanningprocesses.Previousstudieshavehighlightedhowmillennialsoftenhavedifferenttastes,lifestyles,consumerandtravelbehaviorfromthoseofpreviousgenerationsatthesamestageinlife.Still,today’syoungadultsareina“transitional”stageoflife,inwhichtheyarebuildingthebasisfortheirfuturelife,familyandworkcareer.Thus,theircurrentchoicesareexpectedtobeasumoflifecycle,periodandgenerationaleffects:theircurrentbehaviorsarenotnecessarilygoingtolastasmillennialsbecomeolder,andtransitiontomorestablelifestages.Thisstudyinvestigatesmillennials’choices,throughtheanalysisofacomprehensivedatasetthatincludesinformationonmanyofthevariablesthathavebeenattributedaroleinaffectingnewtraveltrendsandadoptionofemergingtransportationservices.Thesevariablesweredifficulttocontrolinpreviousstudies,whichwereoftenlimitedbythelackofavailabilityofinformationonspecificvariables(suchasstudiesbasedontheanalysisofNHTSdata),ortheuseofnon-representativesamples(asinthecaseofconveniencesamples,e.g.collectedamonguniversitystudents).ThestudybuildsonanextensiveresearcheffortcarriedoutwiththecollectionoftheCaliforniaMillennialsDataset,anunprecedenteddatasetcollectedin2015,whichincludesinformationonindividualpreferences,lifestyles,adoptionoftechnology,carownershipandtravelbehaviorforapproximately2400residentsofCalifornia,includingbothmillennials(youngadults,18-34,in2015)andmembersofprecedingGenerationX(middle-ageadults,35-50).ThestudyallowstheinvestigationofseveralcomponentsoftheemergingtrendsintraveldemandandadoptionoftransportationtechnologyinCalifornia.Inthisstageofthestudy,wematchedtheinformationcontainedintheCaliforniaMillennialsDatasetwithadditionalvariablesofinterestincludinglanduseandbuiltenvironmentdataavailablefromothersources,basedonthegeocodedresidentiallocationoftherespondents.ThedataprovideawidevarietyoflanduseandaccessibilitymeasuresavailablethroughtheUSEPASmartLocationDatasetandthewalkscore,bikescoreandtransitscoreobtainedfromthecommercialwebsiteWalkscore.com.Usingthegeocodedinformationontheresidentiallocation,andtheinformationprovidedbytherespondentsinthesurvey,wecarefullycleanedandrecodedthedata,toimprovethequalityoftheresponsesandidentifyinternalandexternalinconsistenciesandpotentialoutliersthatmayleadtonoiseinthedata.Further,wedevelopedasetofweights,throughtheapplicationofbothcellweightsandtheiterativeproportionalfitting(IPF)rakingapproach,tocorrectthedistributionofcasesinthesample,andreducethenon-representativenessofthedatabasedontheregionofCaliforniawheretherespondentslive,neighborhoodtype,age,gender,studentandemploymentstatus,householdincome,raceandethnicityandpresenceofchildreninthehousehold.

Page 85: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

77

WedevelopedanumberofanalysestoinvestigatethecomplexrelationshipsbehindresidentiallocationandmobilitychoicesofCaliforniamillennialsandmembersofGenerationX.First,throughtheuseofdatareductiontechniques,weappliedafactoranalysisapproachtothe66variablesthatcollectedinformationontherespondents’attitudesandpreferencestowardsanumberofdimensions,includingtravelmodepreferences,adoptionoftechnology,environmentalconcerns,landusepreferences,etc.Weextractedasetof17factorsthatmeasuresthemainattitudinalconstructsonanumberoftopics,andcanbeusedintheanalysisofchoicesrelatedtotravelbehavior,residentiallocation,andcarownershipanduse.Weanalyzedtheattitudinalprofilesandindividualcharacteristicsformanysubgroupsofindividuals:notsurprisingly,millennialsthatliveinurban,suburbanorruralareasoftenmanifestratherdifferentattitudinalpatternsfromtheircounterpartsinolderagegroups.Wealsoanalyzedtheadoptionandfrequencyofuseofsmartphoneappsamongdifferentsociodemographicgroups:urbanmillennialsareheavyadoptersoftheseservices,andonaverageshowhigheradoptionofthesetechnologiesforvariouspurposes,includingaccessinginformationaboutthemeans(orcombinationofmeans)oftransportationtouseforatrip,findinginformationabouttripdestinationsornavigatinginreal-timeduringatrip.Largedifferencesarealsoobservedintheadoptionofsharedmobilityservicesamongurbanandnon-urbanpopulations:notsurprisingly,millennialstendtoadoptthesenewtechnologicaltransportationservicesmoreoftenthanthemembersofGenX,inparticularinurbanareas.Wefurtheranalyzedtherelationshipbetweenaccessibilityandadoptionofmultiplemodesoftransportation(multimodality,and/orintermodality)amongthemembersofvarioussub-segmentsofthepopulation.Forthisanalysis,wefurtherclassifiedmillennialsintwogroups,dependingontheirlivingarrangementsandhouseholdcomposition,identifyingtheindependentmillennials(whodonotliveanymorewiththeirparents,andhavealreadyestablishedtheirownhousehold),andthedependentmillennials(wholivewiththeirparents),asabetterwaytocontrolfortheresidentiallocationoftherespondents(astheresidentiallocationfordependentmillennialshaslikelybeenchosenbytheirparents,andnotbythemillennialsthemselves).WecomparedthelevelofaccessibilityoftheplaceofresidenceandtheadoptionofmultimodaltravelofthetwogroupsofmillennialswiththoseoftheoldermembersoftheGenerationX.Independentmillennialswerefound,onaverage,tohavethehighestvaluesforallaccessibilitymeasures.Further,importantdifferencesareobservedamongdependentandindependentmillennials:dependentmillennialstendtoliveinareasthathavethelowestlevelsofaccessibilitybynon-carmodes,probablyduetotheresidentiallocationchosenbyothermembersofthehouseholds(e.g.youngadultswholivewiththeirparents).Thissharplycontraststheresidentiallocationofindependentmillennialswhoaremoreoftenfoundtoliveinlocationswithhigheraccessibility.Centrallocationsaremoreconducivetotheadoptionofgreenerandnon-autocommutemodes(and/ormayreinforcethepropensityofyoungadultstousesuchmodesortoadoptmultimodaltravel).Attheotherendofthespectrum,GenXersrelyheavilyontheuseofcarsfortheircommute.Interestingly,atleastapartofdependent

Page 86: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

78

millennialsarefoundtodrivelessthantheirolderpeersinspiteoflivinginneighborhoodsthatarelessconducivetomultimodalityandtotheuseofnon-automodes.Thefindingssuggestthatahighercomponentoftheadoptionofmultimodalbehaviorsisassociatedwithmakingthesedecisionsbychoice,ratherthannecessity.Insummary,andnotsurprisingly,accessibilityandmultimodalityarepositivelycorrelated:residentsofmoreaccessibleneighborhoodsaremoreoftenfoundtobemultimodalcommuters.However,millennials,andespeciallydependentmillennials,arefoundtomakethemostoftheirbuiltenvironmentpotential,eitherduetoindividualchoices,orthepresence(orlack)oftravelconstraints.Theyarelesslikelytobemono-driversandmorelikelytobemultimodalcommuters,eveniftheyliveinneighborhoodsthatarelesssupportiveofsuchbehaviors.Thissuggeststhattheconnectionbetweenthebuiltenvironmentandtravelpatternsmaydifferbygeneration:infuturestepsoftheresearchweplantofurtherinvestigate(andmodel)therelationshipsbetweenaccessibilityandmultimodalbehavioramongthemembersofthedifferentgenerations,whilecontrollingforotherfactorsaffectingresidentialandtravelchoices.Inordertoinvestigatetheimpactsofvariousgroupsofvariablesonthemobilitychoices,andinparticularoncaruse,ofthemembersofthevariousgenerations,weestimatedalog-linearmodelofthenumberofweeklyvehiclemilestraveled(VMT).Weestimatedbothapooledmodelfortheentiresample,andasegmentedmodelthatallowedustocontrolfortheeffectsofindividual,householdandlandusecharacteristicsontheVMTofmillennialsandGenXers,separately.Allmodelshaveexcellentgoodnessoffit:however,andveryinterestingly,amongthethreemodelsthatarepresented,themodelformillennialsexplainsthelowestamountofvarianceinthedata.Thisfindingsignalsthehigherheterogeneityandtastevariationamongthemembersofthisgroup,andtheincreaseddifficultyinexplainingtheirbehaviorsthroughtheestimationofeconometricandquantitativemodels.Traditionalbuiltenvironmentvariablessuchaspopulationdensityanddiversityofhousing/jobsdonotexplainasmuchvariationinVMTformillennialsasforGenerationX.Attitudinalvariablesandvariablesmeasuringthestageoflifeoftherespondents(inparticular,thelivingarrangementsandthepresenceofchildreninthehousehold)explainmorevariationformillennialsthanGenerationX,confirmingthatmillennials’travelchoicesarebestexplainedbytheirattitudesandstageoflifethanbymoretraditionalvariablesusedinotherstudies.Weinvestigatetherelationshipofindividualsbelongingtothevariousagegroupswithcarownershipandthetypeofvehiclethatisownedinthehousehold.Notsurprisingly,independentmillennialsthatliveinurbanareasarefoundtoownfewercarsperdriverinthehousehold.Thisfinding,whichmatchesthereducedneedsforacarindenser(andmoreaccessible)centralareas,andthestereotypeofmillennialsthatmoreoftenprefertoownfewervehiclesandadoptothermodesoftransportationmoreoften,mightbeshort-livedthough.Manyoldermillennialswholiveinurbanareasactuallyreportthattheydoplantopurchaseanewvehicleinthenearfuture,thusconfirmingthattheirzero-orlow-vehicleownershipstatusisprobablytheresultoftheindividuals’transientstageoflife,ratherthanthelong-termeffect

Page 87: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

79

ofstrongpreferencestowardsvehicleownershipanduse.Duringfuturestagesoftheresearch,weplantostudyhowcarownershipvariesacrossdifferentgroupsofthepopulationthroughtheestimationofcarownershipmodelsthatinvestigatehowvarioussociodemographiccharacteristics,individualpreferences,andlandusefeaturesaffecthouseholdcarownership,andtheuseoflatentclassanalysis(andlatentclassmodeling)tofurtheridentifytheimpactoftasteheterogeneityamongdifferentgroupsofindividualswithregardwithvehicleownershipandtravelbehavior.Inordertoinvestigatethepreferencetowardsthepurchaseofvariousvehicletypesamongdifferentgroupsofusers,inthisstageoftheresearchweestimatedamultinomiallogitmodel(MNL)ofvehicletypechoice,usingsocio-demographiccharacteristics,residentiallocationandlandusecharacteristics,andpersonalattitudesandpreferencesasexplanatoryvariables.Wefocusedonindividualsthatboughtorleasedausedornewvehiclethatismodelyear2010ornewerforthisanalysis,inordertoavoidthenoiseassociatedwiththeeventualpresenceofvehiclesthatweregiftedtotheindividualbyotherfamilymembers,orvehiclesthatwerepurchasedoutofcontingencies(e.g.asinthecaseofoldervehicles,forwhichonlyfewavailableoptionsmightbeavailableinalimitedpricerange).Duringthenextstagesoftheresearch,weplantocapitalizeonthisambitiousresearchprogramfortheinvestigationofthemobilityofmillennialsinCalifornia.Inparticular,weplantofurtherinvestigatetheheterogeneityinthepopulationofmillennials(andolderadults)throughthedevelopmentofclusterorlatentclassanalysistoanalyzedifferentprofilesofpeople,andinvestigatetheproportionofmillennialsandGenXersthatliveinurbanareas,havedynamiclifestyles,areheavyusersofsocialmedia,ownzero(orfew)cars,usepublictransportation,andadoptnewtechnologies,andwhatdifferencesexistwiththeothersegmentsofthemillennialpopulation.Further,weplantoinvestigate(andmodel)therelationshipsbetweenaccessibilityandmultimodalbehavioramongthemembersofthedifferentgenerations,whilecontrollingforotherfactorsaffectingresidentialandtravelchoices,includinghouseholdsizeandcomposition,individualattitudesandlifestyles,andadoptionoftechnology.Wealsoplantoinvestigatetherelationshipsbehindtheadoptionofsharedmobilityservicesandothercomponentsoftravelbehavior,amongvarioussub-segmentsofthepopulation.Inparticular,weplantoevaluatetherelationshipsandlatentconstructsbehindtheadoptionofsharedmobilityservices,suchascarsharingoron-demandrideservicessuchasUberorLyft,andanalyzetheimpactofvariousfactorsaffectingtheuseoftheseservicesinvariousgeographicregionsandneighborhoodtypes,andamongdifferentsegmentsofthepopulation,throughtheestimationofmultivariatemodelsoftheadoptionandfrequencyofuseofeachtypeofsharedmobilityservices.Wewillinvestigatetheimpactofresidentiallocationandneighborhoodcharacteristicsonthesechoices,andestimatebivariatemodelstoexploretherelationshipsbetweentheadoptionofsharedmobilityservicesand:

a) Theuseofothertravelmodes,includingdrivingaloneandusingpublictransportation;b) Autoownership;and

Page 88: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

80

c) Theindividual’sreportedwillingnesstochangethelevelofautoownership,e.g.reducingthenumberofvehiclesinthehousehold,buyinganewvehicle,etc.

Further,thestudywillexploreheterogeneityintravelers’behavior,withrespecttotheadoptionofsharedmobilityservices,travelbehavior,individuallifestylesandtastes,asawaytoinvestigatedifferencesintheobservedrelationshipsamongvariousgroupsofindividuals.Thestudywillprovideimportantinsightsintotheimpactoftheadoptionofnewsharedmobilityservicesonothercomponentsoftraveldemand,VMTandautoownershipinvariousregionsofCaliforniaandlandusetypes,controllingforindividualcharacteristicsanddifferencesamongsegmentsofthepopulation.Finally,thedatacollectioneffortforthisstudywasdesignedasthefirststepofalongitudinalstudyoftheemergingtransportationtrendsinCalifornia,designedwitharotatingpanelstructure,withadditionalwavesofdatacollectionplannedinfutureyears.Infuturestagesofthisresearch,weplantoexpandthedatacollectionalsothroughotherchannels,eventuallyalsothroughthecreationofapaperversionofthesurvey,inordertoexpandthetargetpopulationforthestudy,andreachspecificsegmentsofthepopulation,e.g.elderlyorpeoplethatarenotfamiliarwiththeuseoftechnologyorwhodonothaveeasyaccesstotheinternetandwouldnotlikelycompleteanonlinesurvey.Also,weareconsideringcreatingaversionofthesurveyinSpanish,inordertobetterreachtheCaliforniapopulationofLatinosandincreasetheresponserateamongtheHispanicminority.Theanalysisoftheinformationcollectedthroughmultiplewavesofsurveywillprovidevaluableinformationonthelikelychangeshappeningintraveldemand,andwillprovideinsightsintotheimpactsoftheadoptionofanumberofnewtransportationservicesonfuturetransportationinthestate.

Page 89: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

81

References

Bagley,M.N.,Mokhtarian,P.L.&Kitamura,R.,2002.Amethodologyforthedisaggregate,multidimensionalmeasurementofresidentialneighbourhoodtype.UrbanStudies,39(4),pp.689–704.

Berliner,R.M.,andG.Circella.2017.“CalifornianMillennialsDriveSmallerCars:EstimatingVehicleTypeChoiceofMillennials.”Presentedatthe96thTransportationResearchBoardMeeting,Washington,D.C.,January2017.

Blumenberg,E.,A.Brown,K.Ralph,B.D.Taylor,andC.T.Voulgaris.2015.Typecastingneighborhoodsandtravelers:AnalyzingthegeographyoftravelbehavioramongteensandyoungadultsintheU.S.ProjectReporttotheFederalHighwayAdministration,September2015.

Blumenberg,E.,B.D.Taylor,M.Smart,K.Ralph,M.Wander,andS.Brumbagh..,2012.What’sYouthGottoDowithIt?ExploringtheTravelBehaviorofTeensandYoungAdults.ProjectReport,UniversityofCaliforniaLosAngeles,September2012,LosAngeles.

Blumenberg,E.,K.Ralph,M.Smart,andB.D.Taylor.,2016.Whoknowsaboutkidsthesedays?AnalyzingthedeterminantsofyouthandadultmobilityintheU.S.between1990and2009.TransportationResearchPartA:PolicyandPractice,93,pp.39–54.

Blumenberg,E.&Smart,M.,2014.Brothercanyousparearide?Carpoolinginimmigrantneighbourhoods.UrbanStudies,51(9),pp.1871–1890.

Brownstone,D.&Golob,T.F.,2009.Theimpactofresidentialdensityonvehicleusageandenergyconsumption.JournalofUrbanEconomics,65(1),pp.91–98.

BRS,2013.Americans’ViewsontheirCommunities,Housing,andTransportation,Washington,D.C.Availableat:http://uli.org/wp-content/uploads/ULI-Documents/America-in-2013-Final-Report.pdf(LastaccessedonMarch30,2017).

Cao,X.,Mokhtarian,P.L.&Handy,S.L.,2009a.Examiningtheimpactsofresidentialself-selectionontravelbehaviour:afocusonempiricalfindings.TransportReviews,29(3),pp.359–395.

Cao,X.,Mokhtarian,P.L.&Handy,S.L.,2009b.Therelationshipbetweenthebuiltenvironmentandnonworktravel:AcasestudyofNorthernCalifornia.TransportationResearchPartA:PolicyandPractice,43(5),pp.548–559.

Cervero,R.&Duncan,M.,2006.’WhichReducesVehicleTravelMore:Jobs-HousingBalanceorRetail-HousingMixing?JournaloftheAmericanPlanningAssociation,72(4),pp.475–490.

Cervero,R.&Duncan,M.,2003.Walking,Bicycling,andUrbanLandscapes:EvidenceFromtheSanFranciscoBayArea.AmericanJournalofPublicHealth,93(9),pp.1478–1483.

Cervero,R.&Murakami,J.,2010.Effectsofbuiltenvironmentsonvehiclemilestraveled:evidencefrom370USurbanizedareas.EnvironmentandplanningA,42(2),pp.400–418.

Circella,G.andP.L.Mokhtarian(2017)ImpactofInformationCommunicationTechnologyonTransportation,inHansonS.andG.Giuliano,eds.TheGeographyofUrbanTransportation,4thEdition,Chapter4,pp.86-109,NewYork:TheGuilfordPress.

Circella,G.,K.Tiedeman,S.Handy,F.Alemi,andP.L.Mokhtarian.2016a.WhatAffectsU.S.

PassengerTravel?CurrentTrendsandFuturePerspectives.WhitePaperfromtheNationalCenterforSustainableTransportation.UniversityofCalifornia,Davis,February2016;

Page 90: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

82

availableathttps://ncst.ucdavis.edu/wp-content/uploads/2014/08/06-15-2016-NCST_White_Paper_US_Passenger_Travel_Final_February_2016_Caltrans3.pdf(LastaccessedonMarch30,2017).

Circella,G.,L.Fulton,F.Alemi,R.M.Berliner,K.Tiedeman,P.L.Mokhtarian,andS.Handy.2016b.WhatAffectsMillennials’Mobility?PARTI:InvestigatingtheEnvironmental

Concerns,Lifestyles,Mobility-RelatedAttitudesandAdoptionofTechnologyofYoung

AdultsinCalifornia.ProjectReport,NationalCenterforSustainableTransportation.UniversityofCalifornia,Davis,May2016;availableathttp://ncst.ucdavis.edu/wp-content/uploads/2014/08/05-26-2016-NCST_Report_Millennials_Part_I_2016_May_26_FINAL1.pdf(LastaccessedonMarch30,2017).

Circella,G.,F.Alemi,R.M.Berliner,K.Tiedeman,Y.Lee,L.Fulton,S.Handy,andP.L.Mokhtarian.2017.“MultimodalBehaviorofMillennials:ExploringDifferencesinTravelChoicesBetweenYoungAdultsandGen-XersinCalifornia.”Presentedatthe96thTransportationResearchBoardAnnualMeeting,Washington,D.C.,January2017.

Contrino,H.&McGuckin,N.,2006.AnExplorationoftheInternet’sEffectonTravel,Citeseer.Delbosc,A.&Currie,G.,2014.Changingdemographicsandyoungadultdriverlicensedeclinein

Melbourne,Australia(1994--2009).Transportation,41(3),pp.529–542.Diana,M.&Mokhtarian,P.L.,2009.GroupingTravelersontheBasisoftheirDifferentCarand

TransitLevelsofUse.Transportation,36,pp.455-467.Ewing,R.&Cervero,R.,2010.TravelandtheBuiltEnvironment.JournaloftheAmerican

PlanningAssociation,76(3),pp.265–294.Ewing,R.&Cervero,R.,2001.Travelandthebuiltenvironment:asynthesis.Transportation

ResearchRecord:JournaloftheTransportationResearchBoard,(1780),pp.87–114.Frändberg,L.&Vilhelmson,B.,2011.Moreorlesstravel:personalmobilitytrendsinthe

Swedishpopulationfocusinggenderandcohort.JournalofTransportGeography,19(6),pp.1235–1244.

Fry,R.A.&Passel,J.S.,2014.Inpost-recessionera,youngadultsdrivecontinuingriseinmulti-generationalliving,PewResearchCenter,Social&DemographicTrendsProject;availableathttp://www.pewsocialtrends.org/2014/07/17/in-post-recession-era-young-adults-drive-continuing-rise-in-multi-generational-living/(LastaccessedonMarch30,2017).

Garikapati,V.M.,R.M.Pendyala,E.A.Morris,P.L.Mokhtarian,andN.C.McDonald,2016.ActivityPatterns,TimeUse,andTravelofMillennials:AGenerationinTransition?TransportReviews36(5),pp.558-584.

Goodwin,P.,2012.Peaktravel,peakcarandthefutureofmobility.Greene,D.L.,Chin,S.-M.&Gibson,R.,1995.Aggregatevehicletravelforecastingmodel,Oak

RidgeNationalLab.,TN(UnitedStates).Hallock,L.&Inglis,J.,2015.TheInnovativeTransportationIndex:TheCitiesWhereNew

TechnologiesandToolsCanReduceYourNeedtoOwnaCar.,Availableat:http://www.uspirg.org/sites/pirg/files/reports/Innovative_Transportation_Index_USPIRG.pdf(LastaccessedonMarch30,2017).

Handy,S.,Cao,X.&Mokhtarian,P.,2005.Correlationorcausalitybetweenthebuiltenvironmentandtravelbehavior?EvidencefromNorthernCalifornia.Transportation

Page 91: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

83

ResearchPartD:TransportandEnvironment,10(6),pp.427–444.Handy,S.,Cao,X.&Mokhtarian,P.L.,2006.Self-selectionintherelationshipbetweenthebuilt

environmentandwalking:EmpiricalevidencefromNorthernCalifornia.JournaloftheAmericanPlanningAssociation,72(1),pp.55–74.

Handy,S.L.,2002.TravelBehavior--LandUseInteractions:AnOverviewandAssessmentoftheResearch.InMahmassaniH.,ed.InPerpetualMotion:TravelBehaviorResearch

OpportunitiesandApplicationChallenges,Amsterdam:Pergamon.Hanks,K.W.Odom,D.Roedl,andE.Blevis.2008.Sustainablemillennials:attitudestowards

sustainabilityandthematerialeffectsofinteractivetechnologies.InProceedingsoftheSIGCHIConferenceonHumanFactorsinComputingSystems.ACM,pp.333–342.

ITSAmerica,2015.RiseoftheReal-TimeTraveler:Anexplorationoftrendsandinnovationin

urbanmobility.Kalton,G.&Flores-Cervantes,I.,2003.Weightingmethods.JournalofOfficialStatistics,19(2),

p.81.Klein,N.J.&Smart,M.J.,2017.Millennialsandcarownership:Lessmoney,fewercars.

TransportPolicy,53,pp.20–29.Kuhnimhof,T.,J.Armoogum,R.Buehler,J.Dargay,J.M.Denstadli,andT.Yamamoto.2012.

MenShapeaDownwardTrendinCarUseamongYoungAdults—EvidencefromSixIndustrializedCountries.TransportReviews,32(6),pp.761–779.

Lyons,G.,2014.Viewpoint:Transport’sdigitalagetransition.JournalofTransportandLandUse,8(2),pp.1-19.

McDonald,N.C.,2015.AreMillennialsReallythe“Go-Nowhere”Generation?JournaloftheAmericanPlanningAssociation,pp.1–14.

Metz,D.,2012.Demographicdeterminantsofdailytraveldemand.TransportPolicy,21,pp.20–25.

Metz,D.,2013.Peakcarandbeyond:thefourtheraoftravel.TransportReviews,33(3),pp.255–270.

Mokhtarian,P.,2009.Iftelecommunicationissuchagoodsubstitutefortravel,whydoescongestioncontinuetogetworse?TransportationLetters,1(1),pp.1–17.

Myers,D.,2016.PeakMillennials:ThreeReinforcingCyclesThatAmplifytheRiseandFallofUrbanConcentrationbyMillennials.HousingPolicyDebate,pp.1–20.

PewResearchCenter,2014.ComparingMillennialstoothergenerations.Availableat:http://www.pewsocialtrends.org/2015/03/19/comparing-millennials-to-other-generations/(LastaccessedonMarch30,2017).

Polzin,S.E.,Chu,X.&Godfrey,J.,2014.Theimpactofmillennials’travelbehavioronfuturepersonalvehicletravel.EnergyStrategyReviews,5,pp.59–65.

Prensky,M.,2001.DigitalNatives,DigitalImmigrantsPart1.OntheHorizon,9(5),pp.1–6.Puentes,R.,2013.HaveAmericansHitPeakTravel?ITFRoundTablesLong-runTrendsinCar

Use,152,p.91.Ramsey,K.&Bell,A.,2014.Smartlocationdatabase.Washington,DC.Rentziou,A.,Gkritza,K.&Souleyrette,R.R.,2012.VMT,energyconsumption,andGHG

emissionsforecastingforpassengertransportation.TransportationResearchPartA:PolicyandPractice,46(3),pp.487–500.

Page 92: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

84

Salomon,I.&Mokhtarian,P.L.,2008.Cantelecommunicationshelpsolvetransportationproblems?Adecadelater:Aretheprospectsanybetter.InD.A.Hensher&K.J.Button.,eds.HandbookofTransportModelling.Amsterdam,pp.519–540.

Salon,D.,2015.Heterogeneityintherelationshipbetweenthebuiltenvironmentanddriving:Focusonneighborhoodtypeandtravelpurpose.ResearchinTransportationEconomics,52,pp.34–45.

Santos,A.,N.McGuckin,H.Y.Nakamoto,D.Gray,andS.Liss.2011.Summaryoftraveltrends:

2009nationalhouseholdtravelsurvey.Shaheen,S.,N.Chan,A.Bansal,andA.Cohen.2015.SharedMobilityaSustainablityand

TechnologyWorkshop:Definition,IndustryDevelopmentandEarlyUnderstanding.UniversityofCaliforniaBerkeley.

Shin,E.J.,2016.UnravelingtheEffectsofResidenceinanEthnicEnclaveonImmigrants’TravelModeChoices.JournalofPlanningEducationandResearch,p.0739456X16663309.

Sivak,M.,2014a.HasmotorizationintheUSpeaked?Part4:Householdswithoutalight-dutyvehicle.TransportResearchInstitute,UniversityofMichigan,AnnArbor.

Sivak,M.,2014b.HasmotorizationintheUSpeaked?Part5:Updatethrough2012.Strauss,W.&Howe,N.,2000.Millennialsrising:Thenextgreatgeneration.NewYork:Vintage.Taylor,B.,R.Chin,M.Crotty,J.Dill,L.Hoel,M.Manville,S.Polzin,B.Schaller,S.Shaheen,D.

Sperling,M.Zafar,andS.Zielinski.2015.SpecialReport319:Betweenpublicandprivatemobility:Examiningtheriseoftechnology-enabledtransportationservices.

Tiedeman,K.,G.Circella,F.Alemi,andR.M.Berliner.2017.“WhatDrivesMillennials:ComparisonofVehicleMilesTraveledBetweenMillennialsandGenerationXinCalifornia.”Presentedatthe96thTransportationResearchBoardAnnualMeeting,Washington,DC.,January2017.

U.S.CensusBureau,2014.AmericanCommunitySurvey,2014AmericanCommunitySurvey1-YearEstimates.

Vij,A.,Gorripaty,S.&Walker,J.L.,2015.Fromtrendspottingtotrend’splaining:UnderstandingmodalpreferenceshiftsintheSanFranciscoBayArea.

LeVine,S.&Jones,P.,2012.OntheMove:MakingSenseofcarandtraintraveltrendsinBritain.,p.121.

Wachs,M.,2013.Turningcitiesinsideout:transportationandtheresurgenceofdowntownsinNorthAmerica.Transportation,40(6),pp.1159–1172.

Zipcar,2013.Millennials&Technology:ASurveyCommissionedbyZipcar(Powerpoint),Zipcar’smillennialstudy.

Zmud,J.P.,V.P.Barabba,M.Bradley,J.R.Kuzmyak,M.Zmud,andD.Orrell.2014.StrategicIssuesFacingTransportation,Volume6:TheEffectsofSocio-DemographicsonFuture

TravelDemand.

Page 93: What Affects Millennials’ Mobility? PART II: The ... - STEPS€¦ · to a sample of 2400 residents of California, including millennials (young adults, 18-34 in 2015) and Gen Xers

85

ListofAcronymsUsedintheDocument

ACOP AmericanConsumerOpinionPanelCaltrans CaliforniaDepartmentofTransportationCEC CaliforniaEnergyCommissionEPA (UnitedStates)EnvironmentalProtectionAgencyFHWA FederalHighwayAdministrationGenX GenerationX(Middle-agedadults,35-50y.o.in2015)GenY GenerationY(Youngadults,18-34y.o.in2015)GHG GreenhouseGasHH HouseholdICT InformationandCommunicationTechnologyIPF IterativeProportionalFittingIT InformationTechnologyIRB InstitutionalReviewBoardITS InstituteofTransportationStudiesLDT LightDutyTrucksLTE LongTermEvolution(a4Gmobilecommunicationsstandard)LU LandUseMNL MultinomialLogit(Model)MPO MetropolitanPlanningOrganizationsMTC MetropolitanPlanningOrganization(SanFranciscoBayArea)NCST NationalCenterforSustainableTransportationNHTS NationalHouseholdTravelSurveySACOG SacramentoAreaCouncilofGovernmentsSANDAG SanDiegoAssociationofGovernmentsSCAG SouthernCaliforniaCouncilofGovernmentsSTEPS SustainableTransportationEnergyPathwaysSUV SportUtilityVehicleTDM TransportationDemandManagementTNC TransportationNetworkCompanyTRB TransportationResearchBoardUC UniversityofCaliforniaUCDavis UniversityofCalifornia,DavisUCLA UniversityofCalifornia,LosAngelesUSDOT UnitedStatesDepartmentofTransportationVMT VehicleMilesTraveled