Beyond general behavioral theories: Structural discrepancy in young motorcyclist's risky driving behavior and its policy implications

(1)

ContentslistsavailableatScienceDirect

Accident

Analysis

and

Prevention

jo u r n al hom e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / a a p

Beyond

general

behavioral

theories:

Structural

discrepancy

in

young

motorcyclist’s

risky

driving

behavior

and

its

policy

implications

Yi-Shih

Chung

a,∗

_,

_Jinn-Tsai

_Wong

b

a_Kainan_University,_Logistics_and_Shipping_Management,_No.1_Kainan_Road,_Luzhu_Shiang,_Taoyuan_33857,_Taiwan

b_National_Chiao_Tung_University,_Institute_of_Trafﬁc_and_{Transportation,}_4F,₁₁₈_Chung_Hsiao_W._Rd.,_Sec._1,_Taipei_10044,_Taiwan

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received31July2010

Receivedinrevisedform6March2011 Accepted17April2011

Keywords: Motorcyclist Riskybehavior Structuraldiscrepancy Structuralequationmodeling Demographicfactor

a

b

s

t

r

a

c

t

Whilemanystudiesexaminethemeanscoredifferencesofpsychologicaldeterminantsbetween

het-erogeneousdrivergroups,thisstudyrevealsastructuraldiscrepancyinacausalbehavioralframework.

Usingyoungmotorcyclists(ages18–28)assubjects,thisstudyinvestigatesthevariousrolesofkey

inﬂu-entialfactorsindeterminingriskydrivingbehavior.Multi-groupanalysisofstructuralequationmodeling

showsthatageandgenderaretwofactorsthatcaneffectivelydistinguishheterogeneousdrivergroups

exhibitingdifferentdecision-makingmechanismsinshapingtheirriskydrivingbehaviors.When

encoun-teringundesirabletrafﬁcconditions,roadragecanimmediatelyincreasemalemotorcyclists’intentions

toengageinriskydrivingbehaviors;ontheotherhand,youngfemalemotorcyclistsfurthercalculate

theirperceivedrisktodeterminewhethertoengageinriskydrivingbehaviors.Thisresultshowsthat

thereisasigniﬁcantlinkbetweenriskperceptionandtrafﬁcconditionawarenessforexperienceddrivers

(ages25–28),butnotforyoungerdrivers(ages18–24).Thisﬁndingshowsthatwhilewell-developed

theoriessuchasplannedbehaviorandriskhomeostasisprovidegeneralframeworkstoexplainrisky

driv-ingbehavior,heterogeneousdrivergroupsmayexhibitstructuraldiscrepanciesthatreﬂecttheirvarious

decision-makingmechanisms.Thissuggeststhat,inadditiontomeandifferences,understanding

struc-turaldiscrepanciesamongheterogeneousgroupscouldhelpresearchersidentifyeffectiveintervention

strategies.

1. Introduction

Intheireffortstoreducetrafﬁccollisions,researchersandtrafﬁc

engineershaveattemptedtoidentifythecausalityofrisky

behav-iorand trafﬁcaccidentsand discoverreliablefactorspredicting

driverbehaviors.However,driverbehaviorsaretypically

hetero-geneousandsometimesunpredictable.Eventhoughdriversshare

similarfeaturesandexperiencesimilartrafﬁcconditionsand

envi-ronments,theirbehaviorsmayvary.Thoughdifferentfactorsmay

inﬂuencedifferentriskydrivingbehaviors(Fernandesetal.,2010),

similarriskydrivingbehaviorscouldresultfromdifferentcauses.

Forexample,somedriversmayspeedjustforfunorexcitement,

whileothersspeedbecausetheyareunawareofpotentialdangers

inthedrivingenvironment(Forward,2010;Wongetal.,2010a).

Thesedifferencesmakeitdifﬁculttodeviseasingleintervention

strategyforalldrivers.

To accountfor theseheterogeneous risky drivingbehaviors,

previous researchers have focused on the correlation between

∗ Correspondingauthor.Tel.:+88633412500(6083);fax:+88633412361. E-mailaddresses:yishih.chung@gmail.com,zest@mail.knu.edu.tw(Y.-S.Chung).

various factors and risky driving behavior, testing the mean

differencesoffactorsbetweendifferentdrivergroups,and

iden-tifyingthecharacteristicsofvarioustypesofriskydriversusing

factororclusteranalysis(Jonah,1997).Morerecently,researchers

havebeguntoexaminethe“causalstructures”ofvariousdriving

behaviors and investigate the relationships between affecting

factors based on developed behavioral theories (Nelson et al.,

2009;Vanceetal.,2006;Wong etal.,2010b).For example,the

theoryof plannedbehavior proposesthatattitudes (the degree

towhichperformanceofthebehaviorispositivelyornegatively

valued),subjectivenorms(theperceivedsocialpressuretoengage

or not engagein a behavior), and perceivedbehavioral control

(people’sperceptionsoftheirabilitytoperformagivenbehavior)

are important predictors of behavior through the intention to

performthisbehavior(Ajzen,1991).Thistheoryhasbeenwidely

usedtoexplainriskydrivingbehaviorssuchasspeeding(Elliott

and Thomson, 2010), drunk driving (Chan et al., 2010a), and

dialinganddriving(Walshetal.,2008).Thesestudiesfocusonthe

associationandpredictabilityofriskydrivingbehaviorfactors,and

attempttoevaluatethehypotheticalcausalrelationshipbetween

factorsandriskydrivingbehavior basedonbehavioral theories.

Asa result,theyprovideuswitha betterunderstandingof the

formationofvariousriskydrivingbehaviors.

(2)

Whilesomestudieshavedemonstratedstructuralrelationships

for riskydriving behavior, it remains unclear how these

struc-turalrelationshipsappearamongheterogeneousdrivergroups,and

especiallymotorcyclists.Somedriversshowdistinctdriving

behav-iorsfromtheircounterparts:youngornovicedrivers,maledrivers,

studentdrivers,ordriverswithviolationoraccidenthistory(Bina

etal.,2006;ChangandYeh,2007;Dejoy,1992;Forward,2010; HamedandEasa,1998;Jonah,1997;Linetal.,2003;McKnightand McKnight,2003;Simons-Morton et al.,2005; Taubman-Ben-Ari

etal.,2004).Previousstudiesshowthatthesedriversexhibit

dif-ferentpsychologicalconditionswhiledriving.Forexample,Hamed

andEasa(1998)showedthatmaledriversdemonstrateahigher

levelofperceivedriskthanfemaledrivers.Clearly,various

psycho-logicalconditionsarepartiallyresponsibleforthedistinctdriving

behaviorsbetweenheterogeneousgroups.However,merely

exam-ining the differences in psychological conditions provides only

limitedinsightsintotheformationofheterogeneousdriving

behav-iors. This study uses structural equation modeling to examine

thediscrepancyofcausalbehavioralstructuresbetween

heteroge-neousdrivergroupsinacomprehensivemannerthatcomplements

previousstudies.

Given the growing popularity of motorcycles and the high

accident rate of young motorcyclists, this study selects young

motorcyclistsasthesubjects.Motorcyclesoffertheadvantagesof

lowinitialcostand,forsomemodels,goodfuelefﬁciency.High

fuelpricesin recentyears haveledtoanincreasingnumber of

registeredmotorcycles in somecountries. In theUnited States,

therearemorethan6.2millionregisteredmotorcycles.Morethan

5000motorcyclistswerekilledin2009,accountingfor12%ofall

highwayfatalities(NHTSA,2009).Thesituationisevenworsein

developingcountries,wherepoweredtwo-wheelersareaprimary

modeoftransportationinurbanareas.Forexample,motorcycles

accountfortwothirdsofallregisteredvehiclesinTaiwan,and45%

oftrafﬁcaccidentsinvolvemotorcyclists(MTC,2007a).Compared

tootherdrivers,youngmotorcyclistsaremorelikelytoengagein

riskydrivingbehaviorand becomeinvolved insevereaccidents

(Tsengetal.,2001;Haqueetal.,2009;Zamani-Alavijehetal.,2010).

Thismightresultfromvariousreasons,suchasa relativelylow

helmetuserate(AckaahandAfukaar,2010),enjoymentof

motor-cycling(Zamani-Alavijehetal.,2010),differentpersonalitytraits

(Chen,2009;Wongetal.,2010b),limitedawarenessofpotential

dangersontheroad(Haqueetal.,2009;Wongetal.,2010b),and

poordriving skillsand littleexperience(Changand Yeh,2007).

Thevarietyoffactorsdiscussedintheliteraturesuggestsaneed

toanalyzetheformationofmotorcyclist’sheterogeneousdriving

behaviorina holisticmanner.Therefore,this studyinvestigates

boththemeanscoredifferencesofpsychologicaldeterminantsand

thestructuraldiscrepancybetweenheterogeneousdrivergroupsin

acausalbehavioralstructure.

Theremainingsectionsofthispaperareorganizedasfollows:

Section2brieﬂyreviewsfactorsthatdeﬁnetheformationof

hetero-geneousdrivingbehaviors.Section3introducesthemethodology

usedinthisstudy.Section4presentsanalysisresults,andSection

5providesdiscussionand policyimplications.Finally, Section6

providesconcludingremarks.

2. Factorsdeﬁningtheformationofheterogeneousdriving

behaviors

2.1. Personalitytraits

Wongetal.(2010a)isoneofthefewstudiesexamining

struc-tural discrepancy between heterogeneous groups. The authors

dividedyoungmotorcyclistsintofourgroupsbasedontheir

per-sonalitytraitsandexaminedstructuraldiscrepanciesinconducting

riskydrivingbehavior.Theirresultsconﬁrmtheexistenceof

struc-tural discrepancies. For example,when encountering undesired

trafﬁc conditions, young aggressive motorcyclists immediately

increasedtheirintentiontoconductriskydrivingbehavior;

never-theless,theactionsofthoseintheriskygroupdependedfurther

ontheirconﬁdence and perceivedfun orexcitement. Although

personalitytraitscaneffectivelydemonstratestructural

discrep-ancyinriskydrivingbehavior,theyaredifﬁculttouseinpractice

becausetheyarelatentconstructsthatrequirereliableandvalid

measurements.

2.2. Demographicfactorsanddrivingexperience

Previousresearchshowsthatdemographicfactorsanddriving

experience have signiﬁcant effects on distinguishing

heteroge-neousdrivergroups.Ageandgenderarethetwomostcommonly

useddemographicfactors.Previousstudiesconsistentlyconnect

young male drivers to risky driving behaviors, and have

thor-oughly discussed theunderlying factorsthat distinguish young

maledrivers.Inexperienceinyoungdriversmightcausethemto

engageinriskydrivingbehavior,suchasafailuretoemployroutine

safeoperatingpracticesorfailuretorecognizepotentialdangers

inthedrivingenvironment(ChangandYeh,2007;McKnightand

McKnight,2003).Youngdriversmightalsoengageinrisky

driv-ingbehaviorssuchasspeedinganddrunkdrivingbecausetheir

immediatetemptationoverridestheirknowledgeofthepossible

consequences(Taubman-Ben-Arietal.,2004).

Similartoyoungdrivers,maledriversareover-representedin

riskydrivingbehaviorsandtrafﬁcaccidents(Simons-Mortonetal.,

2005).Comparedtofemaledrivers,maledriversaremoresensation

seekingandperceivecertainriskydrivingbehaviorsaslessserious

andlesslikelytoresultinaccidents.Consequently,previousstudies

consistentlyreportmoreriskydrivingbehaviorsinmales(Dejoy,

1992;Jonah,1997).

Occupation is another important demographic factor

distin-guishing risky drivingbehaviors, especially for studentdrivers.

Student drivers, who are young and over-represented in risky

drivingbehaviorandtrafﬁcaccidents,demonstrateriskydriving

behaviorpatternsdifferentfromnon-studentdriversbecauseof

theirdifferentlifestyle.Thisoftenresultsinvariousdriving

expo-sures,risklevels,alcoholconsumption,etc.(Binaetal.,2006;Lin

etal.,2003).

Drivingexperienceisanotherfrequentlydiscussedfactorinthe

literature,andshowsasigniﬁcantrelationshipwithriskydriving

behaviorandtrafﬁcaccidents.Less-experienceddriversgenerally

exhibitpoordrivingskills,leadingtodangerousdrivingsituations

andmoretrafﬁcaccidents(ChangandYeh,2007;Forward,2010).

Comparedtomoreexperienceddrivers,less-experienceddrivers

mayalsofail toanticipate hiddenhazardsand tend tocommit

drivingerrorsmorefrequently,duetoinappropriateattention

allo-cation(Chanetal.,2010b).

Notethatwhileageanddrivingexperiencearetypicallyhighly

correlated, they are two differentconcepts. For example, even

youngpeoplecanbeexperienceddriversiftheydrive

motorcy-clesfrequently.However,givensimilardrivingexperience,adult

drivers maydrivein a moresensible and reasonable waythan

young drivers because they are more physically and mentally

maturethantheyoungdrivers.

2.3. Violationandaccidenthistory

Trafﬁclawviolationsaretypicalaberrantdrivingbehaviorsthat

endangerdriversthemselvesand otherroadusers. While some

drivergroups, suchasyoungormaledrivers,exhibitmore

vio-lation behaviorsthan their counterparts, previousstudies have

(3)

predicttrafﬁclawviolationbehaviorsthandemographicfactors.

Yagil(1998)suggestedthatduetotheirweakerinstrumentaland

normativemotivestoobeythelawand lessperceivednegative

risk on disobeying the law, young drivers have stronger

con-nectionstotrafﬁc violationsthanold drivers.BeggandLangley

(2004) showed that personality traits are connected to trafﬁc

violationsmore directlythandemographic factorssuchas

gen-der,especially for repeatedviolationbehaviors. By focusing on

young drivers, Bingham et al. (2006) found that while

demo-graphic factors may predict different violation behaviors with

varioussigniﬁcance levels, psychologicaladjustment(e.g.

toler-anceofdeviance,peervs.parent-orientedness)consistentlyplays

a signiﬁcantrole indetermining trafﬁcviolationbehaviors,and

particularlyforyoungerdrivers(e.g.,thoseintheir20s).Blincoe

et al. (2006) discovered that unlike drivers who never exceed

speedlimits,driverswhoslowonlyatcameralocationsandthose

whoexceedlimitsregardlessofcamerasdidnotperceive

speed-ing as a serious trafﬁc law violation. Based on the theory of

plannedbehavior, Forward (2010)demonstrated that attitudes,

subjectivenorms,controlbeliefs,andperceivedbehavioral

con-trolareeffectiveindicatorsoftheintentiontospeedonamajor

road.Apparently,trafﬁclawviolationbehaviorsaresigniﬁcantly

relatedtodriver’spsychologicalconditions,whichmaychangeas

driversmature.

Asforaccidenthistory,previousstudiesshowthatdriverswho

havebeeninvolvedinaccidentstypicallyreporthigherlevelsof

risk-takingbehavior(Linetal.,2004).UsingGPSspeeddata,Jun

etal.(2011)foundthatdriverswhohadaccidentexperiencestend

todriveathigherspeedsthandriversnotinvolvedinaccidents.

Wells-Parkeretal.(2002)alsofoundthataccidentexperienceis

highlycorrelatedwithroadrage.Similartoviolationexperience,

accidentexperiencecanbeusedtodeﬁneheterogeneousdriver

groups,exhibitingdifferentdrivingbehaviors,various

psychologi-calconditions,andlevelsofmaturity.

3. Methodology

Threeelementsarerequiredtoexaminethestructural

discrep-ancybetweenheterogeneousdrivergroups:acausalbehavioral

structure,factorsdeﬁningheterogeneousdrivergroups,anddata.

Thefollowingsubsectionspresenttheseelementsandtheanalysis

procedure.

3.1. Adoptedcausalbehavioralstructure

Manyresearchershavedevelopedtheirownstructural

frame-works to explain risky driving behaviors or accidents (Sumer,

2003;Ulleberg,2001).Numerousfactorscaninﬂuencerisky

driv-ingbehaviors:demographiccharacteristics,environmentalfactors,

roadandvehicleconditions,enforcementintensiﬁcation,and

per-sonality(Sumer,2003).However,thewayinwhichthesefactors

relate to each other and connect with risky driving behavior

dependsontheresearcher’sscopeofstudyandpurpose.

Althoughmanystudiesshowadirectconnectionbetweenthe

aforementionedfactorsandriskydrivingbehaviors,otherstudies

demonstratethatthesefactorscouldaffectriskydriving

behav-iorthroughintermediatefactorssuchasdriver’sattitudeorrisk

perception(UllebergandRundmo,2003;Wongetal.,2010b).For

example,inadditiontodirecteffects,MachinandSankey(2008)

foundthatpersonalityhasindirecteffectsonspeedingviarisk

per-ception.Nelsonetal.(2009)examinedtherelationshipsbetween

driver’sperceived risk,reportedemotionality, perceived

impor-tance,andcellphoneusage(includinginitiatingandanswering)

whiledriving.Nordfjaernetal.(2010)foundsigniﬁcant

relation-shipsbetweendriverattitudesand driverbehavior inruraland

urbanareaswhilecontrollingforage,gender,educational

achieve-ment,andpersonality.

Inoneofthefewstudiestargetingmotorcyclists,Chen(2009)

foundthat an altruistic personalityhasa direct effect onrisky

drivingbehavior,whileapersonalitythatincludesanxiety,anger,

sensation-seeking, and lack of norms hasan indirect effect on

behavior through attitude towards risky driving. Wong et al.

(2010b)alsodevelopedaframeworkspeciﬁctoyoung

motorcy-clists.Theirframeworkconsiderstwoprimarybehaviortheories

–thetheoryof plannedbehaviorand therisk homeostasis

the-ory–tomediatebetweenpersonalityandriskydrivingbehavior.

Thisframeworkconsistsofthecomprehensivefactorsdiscussedin

previousstudies,suchasdriver’sattitudeandriskperception.

Theframeworkadoptedinthisstudyisthesameone

devel-opedbyWongetal.(2010b),asFig.1illustrates.Thisframework

includesthreelevels:(1)anexplanatorylevel,whichconsistsof

threepersonalitytraitsthatexplaintheinternalcharacteristicsof

individualdifferencesanddemonstratesconsistentpatternsand

tendenciesinindividualreactionstotheexternalenvironment;(2)

alatentintermediatelevel,whichcontainsﬁveconstructsthatact

associalcognitivefactorsmediatingbetweenpersonalitytraitsand

riskydrivingbehaviors;(3)adependentlevel,whichconsistsoftwo

constructsthatrepresentmotorcyclist’sriskydrivingbehaviors.

Thethreepersonalitytraitsusedinthisstudyaresensation

seek-ing,amiability,andimpatience.Sensationseekingisdeﬁnedasa

personalitytraitinvolvinganindividualdesireforexcitementor

stimuli.Amiabilityreferstoafriendly,sociable,andcongenial

per-sonalitytrait.Impatienceisthepersonalitytraitofbeingannoyed

easilyduetoundesiredconditions,suchasdelays.

Theﬁvesocialcognitivefactorsusedinthisstudyinclude

rid-ingconﬁdence,affectiveriskperception,utilityperception,trafﬁc

conditionunawareness,andattitudetowardsunsaferiding.

Rid-ing conﬁdencerefers totheperceived behavioral control, asin

thetheoryofplannedbehavior(Wongetal.,2010b).Affectiverisk

perceptionincludestheconcernofriskydrivingbehaviors,which

reﬂectstheriskthatdriversassigntosuchbehaviorbasedontheir

experienceinsteadofactualriderrisks.Utilityperception

repre-sentsriskybehaviorbeliefs,andismeasuredbyacceptingcertain

riskyridingbehaviorstosavetimeorsimplyforfun.Trafﬁc

condi-tionunawarenessreferstotheindividual’ssituationalawareness

inagivenridingenvironment,andreﬂectsthedriver’sprevailing

mannersorsafetyculture.Finally,attitudestowardsunsaferiding

indicatethecontinuoustendencyofpeopletolikeordislikesuch

behavior.

Fastridingandridingviolationaretwocommonriskyriding

behaviorsinTaiwan,andwerethereforechosenasthetwotypes

ofriskydrivingbehaviorinthedependentlevel.

AppendixA(TableA1)providesquestionnaireitemsforeach

construct.Readersinterestedinthedevelopmentofthisframework

andadetaileddiscussionoftheseconstructscanrefertoWonget

al.(2010b).

3.2. Selectedgroupingfactors

Whileboth psychologicalfactorsand manifestvariables1 _are

effectiveindicatorsofheterogeneousdrivingbehaviors,thisstudy

choosesmanifestvariablesasthegroupingvariables.Unlike

psy-chologicalfactors,whicharelatentconstructsthatrequirereliable

andvalidmeasurements,manifestvariablesarerelativelyeasyto

use.

Thisstudyusesﬁvemanifestvariablestodeﬁneheterogeneous

drivergroups:gender,age,occupation,violationexperience,and

1_Manifest_variables_typically_refer_to_measurements_that_researchers_can_directly observeorobtain,suchasgenderorage(Hatcher,1994).

(4)

Fig.1.Riskyridingbehaviormodel.

accidentexperience.Theseﬁvefactorswerechosenfortheir

impor-tanceintheliterature.Samplesizeisanotherconsiderationwhen

choosinggrouping variables. Thatis,each drivergroup deﬁned

bythese variables should containat least 150samples.This is

thesmallestsamplesizeneededtoprovidereliableSEMresults,

giventhattheframeworkwasvalidatedusingalargersamplesize

(Fabrigaretal.,2010;Iacobucci,2010).

Amongthesefactors,gender,occupation,violationexperience,

andaccidentexperiencearecategoricalvariablesandthusdivided

thesamplesbasedontheiroriginaldeﬁnitioninthesurvey,i.e.

“malevs.female,”“non-studentvs.student,”“violationexperience

vs.noviolationexperience,”and“accidentexperiencevs.no

acci-dentexperience.”Violationexperienceincludesallkindsoftrafﬁc

lawviolations,whileaccidentexperiencerepresentstheaccidents

occurringinthepasttwoyears.Thisstudytreatstheonly

contin-uousvariable,age,asacategoricalvariable.Twenty-fouryearsof

agewaschosenasthecut-offpointforthesampleforthefollowing

reason.Accordingtoanationwidereport(MTC,2007b),Taiwanese

motorcyclistsaged24orbelowhaveasigniﬁcantlyhigheraccident

ratethanthoseageabove24.Thisisalsoatransitionpointformost

youngmalesinTaiwan,astheyaregraduatingfromschools,

ﬁnish-ingobligatorymilitaryservice,andsteppingintosociety.Thesame

istrueforyoungfemales,whoareﬁnishingtheirgraduatedegrees

andstartingtheircareersatthisage.Therefore,thisstudydivides

thesamplesintotwogroups–18–24and25–28.

Therearetworeasonswhythisstudyusesage,ratherthan

driv-ingexperience,todeﬁneheterogeneousdrivergroups.First,ageisa

morecomprehensivemeasurethandrivingexperience.The

matu-rityofmotorcycledriversdoesnotmerelydependondrivingskills

andknowledge;itisalsodeterminedbytheirphysicalandmental

ability.Therefore,groupingdriversbyagecanexplainthestructural

discrepancyinacausalbehavioralframeworkmoreclearlythan

drivingexperience.Second,whilestructuraldiscrepancycouldalso

appear between groups withdifferent driving experiences, the

resultmaybeaffectedbytheadopteddeﬁnitionofdriving

experi-ences.

3.3. Data

Thisstudy uses 91 items torepresent theconstructs in the

adoptedframework.Thequestionnaireusedinthisstudyincludes

these91itemsandbackgroundinformation, includinggrouping

variables.Collegestudentsandtransportationprofessionalswere

invitedtoparticipateinapilottest.Theveriﬁedquestionnairewas

administeredtoparticipantssatisfying threecriteria:(1) 18–28

yearsold,(2)holdavalidridinglicense,and(3)have

motorcycle-ridingexperienceduringthepastmonth.Motorcyclistsaged18–28

havethehighestaccident ratein Taiwan,andarethereforethe

subjectsofthisstudy.GiventheInternet’shighpenetrationrate

inTaiwan(morethan70%ofTaiwanesepeoplehaveaccesstothe

Internet)andtohelpreachyoungriders,thequestionnaire was

postedontheInternet.Subjectscompletingthequestionnaire

qual-iﬁedforaprizedrawingtoencourageparticipation,andatotalof

683validsampleswerecollected.Thecompositereliabilityofmost

constructs satisﬁedthe conventional threshold of 0.7 (Hatcher,

1994;Wongetal.,2010a).Wongetal.(2010a)usedthissurvey

datatoinvestigateheterogeneousdrivingbehaviorindriverswith

distinctpersonalitytraits.Thisstudyadoptsthesamedatatotake

advantageofthedatavalidityandthecomparabilityof

heteroge-neousdrivergroupsdeﬁnedbypersonalitytraitswiththosedeﬁned

bymanifestvariables.Readersinterestedinquestionnaire

develop-mentanddetailedcharacteristicsofcollectedsamplescanreferto

Wongetal.(2010a,2010b).

3.4. Analysisprocedure

Theanalysisinthisstudyincludesthree steps:First,

partici-pantsweregroupedseparatelybasedonselectedgroupingfactors.

Second,theresultinggroupswereexaminedfortheirmeanscore

differencesinlatentconstructssuchaspersonality,attitude,and

riskydrivingbehavior.Third,thestructuraldiscrepancybetween

groupswastestedandinvestigatedusingmulti-groupanalysisin

(5)

4. Results

4.1. Clustercharacteristics

Table1presentsthemeanscoresandstatisticaltestsofthe

con-structmeansbetweenclustersdeﬁnedbygender,age,occupation,

violationexperience,andaccidentexperience.Theseresultsreveal

signiﬁcantdifferencesinconstructscoresbetweenclustersdeﬁned

bygenderandviolationvariables.Genderclustersrevealedthat

malemotorcyclistsaremoresensationseekingandlessimpatient2

thanfemalemotorcyclists.Malesalsohaveahigherlevelofdriving

conﬁdence,perceivelessriskandmoreutilityfromriskydriving

behavior,andarelessawareoftrafﬁcconditions.Finally,malesare

morelikelytoexhibitriskydrivingbehavior,includingfastdriving

anddrivingviolations.Theaforementionedcontrastsbetweenmale

andfemaledriversarealsoapparentbetweendriverswith

traf-ﬁclawviolationsandthosewithouttrafﬁclawviolations.Drivers

withviolationexperiencearemoresensationseekingandamiable

thandriverswithoutsuchexperience,haveahigherlevelof

driv-ingconﬁdence,perceivelessriskandmoreutilityfromriskydriving

behavior,arelessawareoftrafﬁcconditions,andaremorelikelyto

exhibitriskydrivingbehaviors.

Clustersdeﬁnedbyoccupationandaccidentexperienceshow

limiteddifferencesinconstructs.Comparedtonon-student

motor-cyclists, student motorcyclists are more sensation seeking and

impatient, and have stronger attitudes towards unsafedriving.

Driverswithaccidentexperiencearelessamiable,perceivemore

utilityfromriskydriving,andaremorelikelytodemonstrate

fast-ridingbehaviorthandriverswithoutaccidentexperience.

Ageistheonlyvariablethatdoesnotshowanysigniﬁcant

differ-encebetweenclusters.Inotherwords,thereisinsufﬁcientevidence

toshowdifferencesbetween18–24-year-oldand25–28-year-old

motorcyclistsintheirpersonalitytraits,drivingconﬁdence,

atti-tude,trafﬁcawareness,riskperception,utilityperception,andrisky

drivingbehaviors.

Thediscussion aboveshows various patternsin mean score

differences of constructs deﬁnedby the selectedﬁve variables.

Theclustersdeﬁnedbygenderand violationexperienceexhibit

themostsigniﬁcantdifferencesthroughoutmostconstructsinthe

behavioralstructure,whilethosedeﬁnedbyagearenotvery

dif-ferent,andthosedeﬁnedbyoccupationandaccidentexperience

arepartiallydifferent.

4.2. Multi-groupequivalencetests

Thissubsectionshowstheresultsofmulti-groupstatisticaltests.

Thesetestsincludetwosteps.Theﬁrststepistoconductan

equiv-alencetestofthemeasurement modeltoexaminewhetherthe

questionnaireitemsforeachconstructandthevarianceand

covari-ancerelationshipsbetweenconstructsareconsistentlyvalidand

reliableacrossgroups.Thesecondstepistoimplementthe

equiva-lencetestofthestructuremodeltoanalyzewhethertheproposed

causalrelationship is consistently appropriate acrossgroups. If

themeasurementmodelequivalenceissatisﬁedandtheproposed

causalrelationshipsareinappropriatefor groups,path analyses

mustbeperformedseparatelyforeachgrouptodeterminethebest

causalrelationshipsforeachcluster.

Table 2 summarizes the equivalence tests of measurement

andstructure consistencyacrossclusters.Resultsshowthatthe

consistency of measurement models between clusters deﬁned

byallﬁvevariables issatisfactoryatthe0.05signiﬁcance level;

2_Impatience_represents_how_respondents_feel_involved_in_undesired_trafﬁc con-ditions,suchasblockedviewsortrafﬁcjams. TableA1 showstheassociated

questionnaireitems. Table

1 Mean score of constructs and statistical tests between clusters defined by demographic factors and driving experience. Construct Overall mean Gender Age Occupation Violation Accident experience Male Female t 18–24 25–28 t Student Non-student t Yes No t Yes No t Sensation seeking 0.93 1.00 0.84 5.95 ** 0.94 0.90 1.62 0.95 0.89 2.00 * 0.99 0.87 4.80 ** 0.94 0.92 0.55 Amiability 2.49 2.53 2.46 1.13 2.46 2.56 − 1.51 2.49 2.51 − 0.34 2.58 2.41 2.84 ** 2.39 2.54 − 2.13 * Impatience 1.80 1.74 1.88 − 2.72 ** 1.83 1.76 1.25 1.84 1.73 1.98 * 1.8 1.81 − 0.15 1.86 1.78 1.30 Riding confidence 2.04 2.17 1.89 7.25 ** 2.01 2.08 − 1.68 2.04 2.04 − 0.03 2.15 1.94 5.17 ** 1.98 2.06 − 1.85 Affective risk perception 2.30 2.21 2.40 − 4.18 ** 2.32 2.26 1.14 2.30 2.29 0.29 2.26 2.34 − 1.72 2.33 2.28 1.01 Utility perception 1.96 2.08 1.83 4.91 ** 1.99 1.92 1.27 1.99 1.90 1.60 2.08 1.86 4.29 ** 2.07 1.92 2.55 ** Attitude towards unsafe riding 1.15 1.30 0.98 7.48 ** 1.15 1.15 − 0.11 1.18 1.07 2.37 * 1.24 1.07 3.84 ** 1.15 1.15 0.06 Unawareness of traffic conditions 0.87 0.93 0.81 3.21 ** 0.89 0.85 0.92 0.88 0.87 0.25 0.93 0.82 2.64 ** 0.92 0.85 1.71 Risky riding behavior 1.14 1.26 1.00 6.97 ** 1.14 1.12 0.44 1.15 1.11 1.09 1.23 1.05 4.94 ** 1.17 1.12 1.00 Fast riding 1.47 1.66 1.26 7.2 ** 1.46 1.49 − 0.59 1.47 1.47 0.04 1.71 1.25 8.31 ** 1.58 1.43 2.35 * Riding violation 0.78 0.87 0.68 4.82 ** 0.79 0.78 0.11 0.79 0.77 0.62 0.86 0.71 3.82 ** 0.81 0.77 0.88 *p < 0.05. ** p < 0.01.

(6)

Table2

Statisticaltestsofmulti-groupequivalence.

Variable Test df Chi-square p-value

Gender Equivalenceofmeasurementmodel 17 23.047 0.15

Equivalenceofstructuremodel 40 62.587 0.01*

Age Equivalenceofmeasurementmodel 17 25.910 0.08

Equivalenceofstructuremodel 40 56.706 0.04*

Occupation Equivalenceofmeasurementmodel 17 13.299 0.72

Equivalenceofstructuremodel 40 42.169 0.38

Accidentexperience Equivalenceofmeasurementmodel 17 15.782 0.54

Violationexperience Equivalenceofmeasurementmodel 17 12.230 0.79

* _p_<_0.05.

in otherwords,the constructindicators (i.e.,the questionnaire

items)used forthewholesampleare alsosatisfactoryforeach

group.Ontheotherhand,thisstudyprovidessufﬁcientevidence

for the age and gender variables to reject structure

equiva-lence between the associated clusters. This suggests that the

pathstructureformalemotorcyclistsdiffersfromthatforfemale

motorcyclists,whilethepathstructurefor18–24-year-old

motor-cyclistsdiffers from that for 25–28-year-old motorcyclists. The

resultsofnon-signiﬁcantdifferencesfortheremainingthree

vari-ablessuggestthatthepathstructureforthewholesample(i.e.,

thestructure shown in Fig.1) isalso appropriate for

motorcy-clistswithdifferentoccupations,violationexperience,andaccident

experience.

Basedontheresultsofsigniﬁcantstructuraldifferencesforage

andgendervariables,thisstudyre-calibratesthepathstructuresfor

theassociatedclusterstoreﬂectappropriatecausalrelationships

betweenconstructs.

4.3. Discrepancyofcausalstructuresbetweenclusters

Fig.2showsthebeststructuresforgroupsdeﬁnedbygenderand

agevariables.Inthisfigure,thepathsforsignificantcoefficientsand

therelationshipsbetweenconstructsarelabeledwithplus/minus

signstoindicatepositive/negativerelationships.AppendixA

sum-marizes the estimation results (Table A2). The goodness of ﬁt

indicatorsforallclustersmostlysatisfyornearlysatisfythe

con-ventionalrequirements, including2_/df_<_2, _GFI_{(goodness-of-ﬁt}

index)greaterthanorequalto0.9,andRMSEA(rootmeansquare

errorofapproximation)lessthanorequalto0.05(Hatcher,1994).

Therefore,theseestimationresultsareappropriatefordiscussing

thestructuraldiscrepancybetweenmaleandfemalemotorcyclists

andbetween18–24-year-oldand25–28-year-oldmotorcyclists.

4.3.1. Maleandfemalemotorcyclists

Fig. 2(a) and (b) show the best structures for male and

femalemotorcyclists.Thesegroupsexhibitseveralkeydifferences.

Compared to female motorcyclists, male motorcyclists exhibit

additionallinksfromamiability(F2)toriskperception(F5),

impa-tience (F3)to utility perception (F6), riding conﬁdence (F4) to

attitudetowardsunsaferiding(F7),andfromutilityperception(F6)

toriskyridingbehavior(F9).Ontheotherhand,female

motorcy-clistsexhibitadditionallinksfromsensationseeking(F1)torisk

perception(F5),riskperception(F5)tounawarenessoftrafﬁc

con-ditions(F8),andfromutilityperception(F6)toattitudetowards

unsaferiding(F7).Thesedifferentlinksrevealstructural

discrepan-ciesanddistinctwaysinwhichpersonalitytraitsaffectriskydriving

behaviorsinyoungmaleandfemalemotorcyclists.Toclarifythis

point,Table3listsallthepathsfromthreepersonalitytraits

(con-structsF1,F2,and F3)toriskydrivingbehaviors(construct F9),

andcalculatesthecorrespondingpatheffectsandthepercentage

oftheircontributiontotheformationofriskydrivingbehavior.

Halfoftheeffectofthesensationseekingpersonalitytraiton

riskydrivingbehavioroccursthroughmaleandfemaledriver

atti-tudes(i.e.,pathF1→F7→F9)asTable3shows.Theotherhalfofits

effectoccursthroughdifferentpathsformaleandfemale

motor-cyclists.Perceivedutilitygreatlyinﬂuencessensation-seekingmale

motorcyclists(i.e.,pathF1→F6→F9),accountingfor38.9%oftotal

effect.Inotherwords,sensation-seekingmalemotorcyclistsare

morelikely toconduct riskydriving behaviordue totheir

per-ceivedutilityfromriskydriving,suchasexcitementorfun. On

theotherhand,thewayinwhichsensation-seekingfemale

motor-cyclists conduct risky driving behavior is rather sophisticated.

Themostobviousexamplesarethetwopathsthroughrisk

per-ception,F1→F5→F7→F9andF1→F5→F8→F9,whichaccount

for morethan 35% of thetotal effect.These two pathssuggest

that,unlikestraightforwardresponseoftheirmalecounterpart’s,

sensation-seekingfemalemotorcycliststakeastepbackand

care-fullycalculatetheirperceiveddrivingriskbeforeconductingrisky

drivingbehaviors.

Thepathslinkingamiabilitytoriskydrivingbehaviorandthe

associatedtotaleffectsareextremelydifferentbetweenmaleand

femalemotorcyclists.Amiable malemotorcyclists conductrisky

drivingbehaviorsbasedonabalancebetweentwopaths:the

pos-itiveeffectsofridingconﬁdence(i.e.pathF2→F4→F7→F9)and

thenegativeeffectsofriskperception(i.e.pathF2→F5→F7→F9).

Amiablemalemotorcyclistsarelesslikelytoconductriskydriving

behaviorsbecausetheirperceivedriskduetoriskydriving

out-weighstheirconﬁdence.Comparedtolessamiablemaledrivers,

amiablefemalemotorcyclists,throughtheirridingconﬁdence(F4),

utilityperception(F6)andattitudetowardsunsaferiding(F7),are

morelikely toconductrisky drivingbehaviorbecausethispath

consistsofonlypositivelinks.

Asformotorcyclistswithanimpatientpersonality,malesand

females alike dependona balancebetweenpositive and

nega-tivepatheffectstodeterminetheirriskydrivingbehaviors.Most

of the positive path effects for male and female motorcyclists

comefromthesamepath,F3→F7→F9.Thisshows that

impa-tientmotorcyclists,whethermaleorfemale,conductriskydriving

behaviorsimplybecausetheirattitudestowardsunsaferidingare

triggeredbytheirimpatiencewithtrafﬁc conditions.Regarding

pathswithnegativeeffects,bothmaleandfemalemotorcyclists

have the path F3→F5→F7→F9. Thisindicates that perceived

risktriggersasaferattitude,whichreducesamotorcyclist’s

inten-tiontoengageinriskydrivingbehavior.Youngimpatientfemale

motorcyclistsshowanadditionalpathwithasubstantialnegative

impact:pathF3→F5→F8→F9.Thispathincludestheconstruct

ofunawareness of trafﬁcconditions (F8),suggestingthat when

(7)

Fig.2.Calibratedstructuresandrelationshipsbetweenconstructs.

Table3

Effectsa_of_personality_traits_on_risky_riding_behavior.

Personalitytrait Impactpath Gender Age

Male Female 18–24 25–28 Sensationseeking F1→F4→F7→F9 0.030(10.1)b _0.024_(5.8) F1→F4→F6→F7→F9 0.002(1.1) 0.008(3.3) F1→F4→F8→F9 0.027(11.2) F1→F5→F7→F9 0.030(16.7) 0.080(19.1) F1→F5→F8→F9 0.036(20.0) 0.044(10.5) F1→F6→F9 0.116(38.9) 0.105(25.1) F1→F6→F7→F9 0.032(17.8) 0.052(21.5) F1→F7→F9 0.152(51.0) 0.080(44.4) 0.155(64.0) 0.165(39.5) Totaleffectc _0.298_(100.0) _0.180_(100.0) _0.242_(100.0) _0.418_(100.0) Amiability F2→F4→F6→F7→F9 0.002(100.0) 0.006(120.0) F2→F4→F7→F9 0.024(−342.9) 0.014(100.0) F2→F4→F8→F9 0.020(400.0) F2→F5→F6→F7→F9 −0.001(−20.0) F2→F5→F7→F9 −0.031(442.9) −0.020(−400.0) Totaleffect −0.007(100.0) 0.002(100.0) 0.005(100.0) 0.014(100.0) Impatience F3→F4→F6→F7→F9 −0.007(10.1) 0.010(10.0) F3→F5→F6→F7→F9 −0.004(−4.0) F3→F5→F7→F9 −0.067(−46.5) −0.057(82.6) −0.071(−71.0) −0.098(251.3) F3→F5→F8→F9 −0.069(100.0) −0.054(138.5) F3→F6→F9 0.030(20.8) 0.031(−79.5) F3→F7→F9 0.181(125.7) 0.064(−92.7) 0.165(165.0) 0.082(−210.3) Totaleffect 0.144(100.0) −0.069(100.0) 0.100(100.0) −0.039(100.0)

a_Effects_are_the_products_of_{coefﬁcients}_along_the_path_between_the_two_speciﬁed_constructs_that_involve_intervening_constructs. b_Numbers_in_the_parentheses_represent_percentages.

(8)

attitude,aremoreawareoftrafﬁcconditions,andarelesslikely

toconductrisky drivingbehaviors.Due tothis pathdifference,

thetotaleffectofimpatienceisnegativeforfemalemotorcyclists,

but positive for male motorcyclists. Whenyoung motorcyclists

areimpatient,malesaremorelikely todemonstraterisky

driv-ingbehavior,butfemalesdonot.Thisdifferencemaydependon

whetherperceivedrisk affects trafﬁcconditionawareness. This

phenomenonagainrevealsthatfemaleshavearelativelycautious

drivingstyle.

Anoticeablelinkforbothmaleandfemalemotorcyclistsisthe

positivelinkbetweenridingconﬁdence(F4)andattitudetowards

unsaferiding(F7).Thisistheonly linkthatmale motorcyclists

havebutfemalemotorcyclistsdonot,andthatdistinguisheshow

ridingconﬁdenceaffectsattitudesdifferentlyinmaleandfemale

motorcyclists. While male motorcyclists with a higher level of

driving conﬁdence directly relate to stronger attitude towards

unsaferiding,femalemotorcyclistswithhigherconﬁdence

con-sidertheperceivedutilityfromriskydriving,andthendetermine

theirattitudeondriving.Inotherwords,malemotorcyclistshave

stronger attitude towards unsafe riding even without

perceiv-ingany beneﬁtsas long astheyhave a higher level of driving

conﬁdence.

4.3.2. 18–24-year-oldand25–28-year-oldmotorcyclists

Fig.2(c) and (d) shows thebest structures for

18–24-year-oldand25–28-year-oldmotorcyclists.Thestructurediscrepancy

betweenthese two groups is more obviousthan that between

maleandfemalegroups.Thisresultseemstocontradicttheone

showninTable1,wherethemeanscoredifferencesofconstructs

betweenagegroupsaremostlynon-signiﬁcant.Thisresultreveals

thatmerelyinvestigatingthemeanscoresofconstructswithout

exploringthestructuralrelationshipsoftheconstructscouldlead

toincorrectconclusions.

The18–24-year-old grouphasﬁvelinks not sharedwithits

counterpart:amiability (F2)torisk perception(F5), riding

con-ﬁdence(F4)toutilityperception(F6),riding conﬁdence(F4)to

unawarenessoftrafﬁcconditions(F8),riskyperception(F5)to

util-ityperception(F6),andutilityperception(F6)toattitudetowards

unsaferiding(F7).Ontheotherhand,the25–28-year-oldgroup

hasfourdifferentlinks:sensationseeking(F1)toriskyperception

(F5),impatience(F3)toutilityperception(F6),ridingconﬁdence

(F4)toattitudetowardsunsaferiding(F7),riskyperception(F5)to

unawarenessoftrafﬁcconditions(F8),andutilityperception(F6)

toriskyridingbehavior(F9).Table3showshowtheselinks

pro-ducedifferentpathsconnectingpersonalitytraitsandriskydriving

behavior.

As for the personality trait of sensation seeking, the path

F1→F7→F9playsthemostsigniﬁcantrolein determiningthe

total effectfor both groups.For the 18–24-year-oldgroup, this

path accounts for 64% of the total effect. In other words, the

primaryreasonwhy18–24-year-oldmotorcyclistsconductrisky

driving behavior is their unsafe driving attitude, which is due

to their sensation-seeking personality trait. Utility perception

(F6) also plays a critical role for both groups, accounting for

morethan20%oftotal effects.Thecorrespondingpathsinclude

pathF1→F6→F7→F9 for the18–24-year-old group and path

F1→F6→F9 for the other group. In spite of these

similari-ties,riskperception(F5)(showninpathsF1→F5→F7→F9and

F1→F5→F8→F9) is unique to the more experienced group,

motorcyclistsagedbetween25and28.Thesetwopathsare

respon-sibleforapproximately30%ofthetotaleffect.Withmoresensation

seeking,25–28-year-oldmotorcyclistsperceivelessrisk,reducing

theirawarenessoftrafﬁcconditionsandexhibitingastronger

atti-tudetowardsunsaferiding.Asaresult,theyconductmorerisky

drivingbehavior.Inotherwords,sensationseekerscantransform

experienceintoincorrectriskperceptions,whichendangersroad

safety.Obviously,thisresultisnotwhatwewouldliketosee.Due

totheinexperienceofnovicemotorcyclists(i.e.,18–24-year-olds),

attitudedeterminesmostoftheirriskydrivingbehavior.

Thepathsstartingfromamiabilityareextremelydifferentfor

thetwoagegroups.Themoreexperiencedgrouponlycontainsone

path,F2→F4→F7→F9,indicatingthelikelihoodof

25–28-year-oldamiablemotorcycliststoconductriskydrivingbehaviorswhen

theyaremoreconﬁdentabouttheirowndrivingskills.Theyounger

groupexhibitsfourpathsstartingfromamiability;twoofthepaths

producepositiveeffectsandtheothertwoproducenegativeeffects.

Whethertheyconductriskydrivingbehaviordependsprimarilyon

thebalancebetweenthepositiveeffectresultingfrom

unaware-nessoftrafﬁcconditionsduetoridingconﬁdence(F4→F8)and

thenegativeeffectresultingfromasaferattitudeduetoperceived

risk(F5→F7).

Asfortheeffectofimpatientpersonalitytrait,bothagegroups

dependonthebalancebetween positiveand negativepathsto

determinethelikelihoodofconductingriskydrivingbehavior.The

mostsigniﬁcantpositiveandnegativepathsarethesameforboth

groups:pathF3→F5→F7→F9andpathF3→F7→F9. The

for-merproducesanegativeeffect,whilethelatterproducesapositive

effect.Thesetwopathsimplythatyoung,impatientmotorcyclists

conductriskydrivingbehaviorpartiallyduetoanunsafeattitude.

However,theriskperception(F5),whichmediatesbetween

impa-tience(F3)andattitudetowardsunsaferiding(F7),hasanegative

effectthatpartiallyoffsetsthepositiveeffect.Thepositiveeffect

oftheyounger-agedgroupoutweighsthenegativeeffectandthe

consequenttotaleffectispositive,indicatingahigherlikelihood

ofriskydrivingbehaviorforanimpatient,novicemotorcyclist.On

theotherhand,themoreexperiencedgroupshowsamore

pow-erfulnegativeeffectfrompathF3→F5→F7→F9.Thisgroupalso

hasoneadditionalnegativepath,F3→F5→F8→F9,and

conse-quentlyproduces a negativetotal effect.Impatient experienced

motorcyclistsarelesslikelytoconductriskydrivingbehaviordue

toasaferattitudeandgreatertrafﬁcconditionawarenessresulting

frommoreperceiveddrivingrisk.

5. Discussionandthepolicyimplications

Thisstudyinvestigatestherolesofmanifestvariables,including

age,gender,occupation,violationexperience,andaccident

experi-ence,ontheheterogeneityofyoungmotorcyclistsindetermining

riskydriving behaviorby examiningtheconstructmeanscores

andtheirstructuraldiscrepancies.Usingstatisticaltestsand

multi-groupanalysisofstructuralequationmodeling,thisstudyshows

that structuraldiscrepancy existsbetweensome drivergroups,

whichisnotexplicitlyimpliedbymeanscoredifferences.Mean

scoredifferencesdonotnecessarilyindicatestructuraldiscrepancy,

asdemonstratedbydrivergroupswithdifferentviolation

experi-ences.Inaddition,thesimilaritybetweenmeanscoresdoesnot

suggeststructurallikeness,asillustratedbydifferentagegroups

ofyoungmotorcyclists.Results showthatwhilethetheoriesof

plannedbehaviorandriskhomeostasismightexplainthegeneral

causalstructureofriskydrivingbehavior,thesigniﬁcanceofcausal

linksbetweenconstructsmayvaryamongheterogeneousdriver

groups.Thus, interventionstrategies thatfocusonreducingthe

strengthofconstructmeanscoresmayhaveaslighteffecton

pre-ventingriskydrivingbehavioriftheconstructdoesnotconnect

withriskydrivingbehaviororifthetotaleffectsofthe

correspond-ingpathsarerelativelysmall.

5.1. Meanscoredifferencesandstructuraldiscrepancies

Themeanscoredifferencesobtainedinthisstudyagreewiththe

(9)

andviolationexperience(ChangandYeh,2007).Male

motorcy-clistsand motorcyclists withviolationexperienceare generally

moresensationseeking,haveahigherlevelofdrivingconﬁdence,

perceivelessriskandhigherutility(e.g.,funorexcitement),have

astrongerinclinationtowardunsafedriving,aremoreunaware

oftrafﬁcconditions,andshowmoreriskydrivingbehaviors.The

limited differences between driver groups deﬁned by accident

experiencemaybeduetotherelativelyinfrequentandrandom

occurrenceofaccidents.Theenhancedperceivedriskdueto

acci-dentexperiencemayalsofadeawayandbecomenon-signiﬁcant

astimegoesby(Linetal.,2004).Thelimiteddifferencesbetween

drivergroupsdeﬁnedbyoccupationsuggeststhatstudentdrivers

aregenerallymoresensationseekingandimpatient,butarenot

sig-niﬁcantlydifferentfromyoungnon-studentdriversinexhibiting

riskydrivingbehavior.

Youngmotorcyclistsofdifferentagegroupsdidnotshowany

signiﬁcantdifferenceinthemeanscoresofpsychological

determi-nantsandriskydrivingbehavior.Thisresultseemstocontradict

previousstudies,inwhichnovicedriverssigniﬁcantlyreducetheir

riskydriving behaviorstwo orthree yearsafter obtainingtheir

driver’slicense(Langleyetal.,1996;Simpson,2003).Taiwan’slack

ofasophisticatedlicensingprogrammaybereasonforthe

differ-entﬁndingsinthisstudy.While18istheminimumagetoobtain

amotorcyclist’slicense,Taiwanesemotorcyclistsrequireacertain

periodoftimetobecomeexperienceddrivers.Thelicensing

pro-cedureformopedsand lightmotorcyclesinTaiwanrequires no

priorexperienceorcompulsorytraining.Instead,novice

motorcy-clistslearnpracticaldrivingskillsbythemselvesafterobtaininga

driver’slicense(Changand Yeh,2007;Chen,2009;Wongetal.,

2010b).

Thesigniﬁcantstructuraldiscrepancybetweendriversof

dif-ferent age ranges suggests that even though it is difﬁcult to

alterthepsychologicalstatusofyoungmotorcyclists,their

driv-ingbehaviorscan beadapted throughlearning andexperience.

Thisstudyshowsthatriskperceptionisacriticalconstruct

deter-miningthestructural discrepancy between18–24-year-old and

25–28-year-olddrivers.Thelattergroupisparticularlycautious

becauseoftheirawarenessoftrafﬁcconditionsandperceivedroad

risk. On the contrary,the younger group doesnot exhibit this

causalpath.Thismayreﬂecttheover-simpliﬁedtestsfor

acquir-ing a motorcyclist’s licensein Taiwan: a written test only for

mopedsandlightmotorcycleswithanenginecapacityoflessthan

50cm3_,_and _a _written_and _track_test3 _for _those_with_an _engine

capacitylessthan250cm3_._Moreover,_a_{motorcyclist’s}_license_can

be immediatelyobtained after passing the exams withoutany

sophisticatedlicensingproceduressuchaslearner’spermit,

pro-bationarylicensing,provisionallicensing,orgraduatedlicensing

(Simpson,2003).Theresultsabovesuggestthatsomemeasures,

suchasgraduatedlicensingschemes,maybenecessarytoaddress

thisproblem.

In addition to age groups, this study revealsstructural

dis-crepanciesbetweenmaleandfemalemotorcyclists.Comparedto

femalemotorcyclists,malemotorcyclists,exhibitasimplercausal

behavioral structurein termsoffewer linkswithnegativepath

coefﬁcientsandfewerpathsconnectingpersonalitytorisky

driv-ingbehavior. Thisdiscrepancy partiallyexplains whymale and

femalemotorcyclistsbehavedifferentlyevenwhentheyhave

sim-ilarpersonalitytraitsorencountersimilartrafﬁcconditions.For

example,seekingexcitementorfunisasimple reasonformale

motorcycliststoconductriskydrivingbehavioriftheyare

rela-tivelysensationseeking.Ontheotherhand,theperceivedriskand

awarenessoftrafﬁcconditionsmightinhibitfemalemotorcyclists

3_This_is_a_road_test_at_an_indoor_site_where_all_driving_conditions_are_{pre-speciﬁed} andnoothervehiclesarepresent.

fromconductingriskydrivingbehavioreveniftheyarerelatively

sensationseeking.

Thoughtherearestructuraldiscrepanciesinsomedrivergroups,

criticalconstructsconsistentlyplayrolesindifferentdrivergroups.

Amongthesocialpsychologicaldeterminantsadoptedinthisstudy,

theutilityperceptionanddriverattitudeconstructsincludemore

incidentand emanatinglinks andareassociated withrelatively

strongerpatheffects.Thisresultechoestheﬁndingsofmany

stud-ies(Ajzen,1991;Chen,2009;Forward,2010;Iversen,2004;Kim andYamashita,2007),andreinforcestheimportanceof

educat-ingyoungmotorcyclistsregardingsafedrivingandtheseverityof

trafﬁcaccidentscausedbyriskydriving.

5.2. Amorecomprehensiveapproachtodevisingintervention

strategy

Previousstudiesusethemeanscoredifferencetodevise

inter-vention strategies to reduce risky driving behavior and trafﬁc

accidents. However, this study shows that the effectiveness of

reducing thestrength ofconstructmeanscoresdependsonthe

importanceoftheconstructsinacausalbehavioralstructure.Thus,

theﬁndingsofthisstudyprovideatleastthreetypesofintervention

strategies:reducingthestrengthofundesiredconstructs,adding

desiredlinks,andremovingorreducingthestrengthofundesired

links.

For example,a driver’sattitudeis apparentlythemost

criti-cal constructconnecting personalityand risky drivingbehavior

becauseit is associated withthemost pathsand thestrongest

total effects. Thus, a unit change in the driver’s attitude has

a greater effect on risky driving behavior than other

con-structs. Nonetheless, it can be very difﬁcult to changedriver’s

attitude.

Duetostructuraldiscrepancies,achangeintheconstructscore

meanscanhavedifferenteffectsondrivergroups.Forexample,

strategies to reduce theoverconﬁdence of youngmotorcyclists

could have a greater effect on male motorcyclists than female

motorcyclists.Thisisbecauseridingconﬁdenceisassociatedwith

morepathsandstrongertotaleffectsformaledriversthanfemale

drivers.

The links between constructs are also critical elements to

consider when devising intervention strategies. One way is to

builddesiredlinks.Forexample,thenegativelinkbetweenrisk

perception and unawareness of trafﬁc conditions helps reduce

risky driving behavior for female motorcyclists, but this link

is absentfor male motorcyclists. Educatingyoungmale

motor-cyclists about the possible dangers and related risks hidden

in the driving environment could build this link. Anotherway

to devise an intervention strategy is to avoid undesired links.

For example, research shows a linkbetween utilityperception

and risky driving behavior for male motorcyclists, but not for

femalemotorcyclists. Thislinkincreasesthepossibility ofmale

motorcyclists conducting risky driving behavior because of the

excitement or fun of it. Educating young male motorcyclists

aboutthepossibleseverityof riskydriving behavioror

provid-ing them withaccident archivesmight reduce the strength of

thislink.

6. Concludingremarks

Previous studies on this topic use psychologicaland

demo-graphicfactorstodistinguishheterogeneousdrivergroups. This

studycomplementspreviousstudiesbydemonstratingthe

advan-tagesofusingdemographicfactorstodivideyoungmotorcyclists

intogroupswithsigniﬁcantstructuraldiscrepanciesandexplore

(10)

Wongetal.(2010a)usedthesamesurveydatatoinvestigate

howdriverswithdifferentpersonalitytraitsshapedriskydriving

behavior in distinct ways; their study shows signiﬁcant

struc-turaldiscrepancies among fourdrivergroups:risky,aggressive,

conservative,andnervousdrivers.Thestructurestheydiscovered

using personality traits are somewhat different from the

cur-rentstructures, discovered by demographic factors. This might

result from two reasons. First, because a driver has a mixture

ofpersonalitytraits,thestructuresdeﬁnedbydemographic

fac-tors couldbeaverages ofthe structuresdeﬁned bypersonality

traits.For example, althoughmale drivers are more aggressive

thanfemaledrivers,notallmaledriversareaggressive.

Accord-ing to Wong et al. (2010b), the average male motorcyclist is

33.33%aggressive,26.78%conservative,23.77%risky,and16.12%

nervous.Second,personalitytraitsareonlyoneofthefactors

char-acterizingdemographic groups.Theﬁndingsof thisstudyshow

that demographic factors,such as gender, are effective

indica-torstodistinguishheterogeneousdrivergroupsintermsofboth

meanscoredifferencesandstructuraldiscrepancy.However,the

otherfactorscausingstructuraldiscrepancyarewellworthfurther

studies.

Despite the high correlation between age and driving

experience4_in_the_survey_data_(correlation_coefﬁcient_0.79,

signif-icantlyat0.001),ageanddrivingexperiencefunctiondifferently

inlicensingprocedures(Simpson,2003).Therefore,it wouldbe

helpfultodistinguishbetweentheeffectsofageanddriving

experi-enceonstructuraldiscrepancytodevisesafetystrategies,especially

forgraduatedlicensing programs.Nevertheless, duetothe

lim-ited sample size, this study didnot control driving experience

whileexaminingthestructuraldiscrepanciesbetweendifferentage

groups.

Because this study focuses onyoung motorcyclists, the age

ofparticipantsrangedfrom18 to28 yearsonly. Thismay

par-tiallyresultinthenon-signiﬁcantdifferencesinpersonalitytraits

between the two age subgroups. Future studies could expand

this age range to better reﬂect the characteristics of young

driver’s heterogeneous behaviors and make the results more

persuasive.

Acknowledgements

Theauthorswould liketothanktheanonymousrefereesfor

theirhelpfulsuggestionsandcomments.Thisworkwaspartially

supportedbytheNationalScienceCouncilofTaiwan(NSC

97-2221-E-009-116-MY3&NSC98-2410-H-424-018).

AppendixA. AppendixA

4 _Deﬁned_as_the_number_of_years_the_participant_has_ridden_a_motorcycle.

TableA1

Questionnaireitemsforeachconstruct. Explanatoryconstructs:personalitytraits

Sensationseeking Ioftencraveexcitement.

Isometimesdothingsjustforkicksorthrills.

ItisOKtogetaroundlawsandrulesaslongasyoudonotbreak themdirectly.

Ifsomethingworks,itislessimportantwhetheritisrightorwrong. Amiability

FewpeoplethinkIamselﬁshandegotistical. Fewpeoplethinkofmeascalmandcalculating. Impatience

PedestriansblockmywaywhileIamridinginanalley. Iamstuckinatrafﬁcjam.

Iamridingbehindatruckandmyviewsareblocked. Someoneisweavinginandoutoftrafﬁc.

Latentintermediateconstructs Ridingconﬁdence

Icanhandleanyunexpectedsituationevenwhenridingon unfamiliarroads.

IfIrunintodangerwhileriding,Ihavetheskillstogetoutofitsafely. Affectiveriskperception

Rushrunningatthestartinstanceofthegreenlight. Ridebetweentwolanesoffastmovingtrafﬁc.

Ridesoclosetothefrontvehiclethatitwouldbedifﬁculttostopin anemergency.

Mergeontomajorroadsfromaminorroadwhenthereisoncoming trafﬁc.

RidesofastintoacornerthatIfeellikeIamlosingcontrol. Utilityperception

Ridingisnotonlyfortransportationbutalsoforfunorrecreation. Ridingamotorcyclemakesmefeelrelaxed.

Attitudetowardsunsaferiding

Itisacceptabletorideintheoppositelaneofatwo-laneroadfor convenience.

Inordertosavetime,ridingagainstthedirectiononaone-wayroad isacceptable.

Withgoodskills,speedingisOK.

IthinkitisOKtospeedifthetrafﬁcconditionallowsmetodoso. Unawarenessoftrafﬁcconditions

Donotusemirrortochecksurroundingvehicleswhileridingor turning.

Donotuseturnsignalswhenturning.

Donotusemirrortochecksurroundingvehicleswhileridingor turning.

Dependentconstructs:riskyridingbehavior Fastriding

Inordertoridefaster,Isqueezethroughanextremelynarrowspace betweenonevehicleandanother.

Comparedtothesurroundingtrafﬁcﬂow,Iridemuchfaster. Disregardthespeedlimitlateatnightorinearlymorning. Ridingviolation

Drinkandride. Runthroughredlights. Ridethewrongway.

(11)

TableA2

Standardizedpathcoefﬁcientsandgoodnessofﬁts.

Path Gender Age

Male Female 18–24 25–28 F1→F4 0.293** _0.207** _0.299** _0.351** F2→F4 0.239** _0.241** _0.227** _0.209# F1→F5 −0.205** _−0.312** F2→F5 0.146# _0.111# F3→F5 0.315** _0.395** _0.385** _0.381** F1→F6 0.492** _0.493** _0.477** _0.546** F3→F6 0.128* _0.159* F4→F6 0.183** _0.126# _0.23** F5→F6 −0.086# F1→F7 0.312** _0.216* _0.288** _0.322** F3→F7 0.371** _0.172* _0.307** _0.16# F4→F7 0.208** _0.133 F5→F7 −0.438** _−0.391** _−0.343** _−0.5** F6→F7 0.173* _0.203** F4→F8 0.111# F5→F8 −0.223** _−0.19* F6→F9 0.235* _0.192* F7→F9 0.487** _0.372** _0.538** _0.512** F8→F9 0.782** _0.788** _0.807** _0.74** F9→F10 0.575 0.652 0.591 0.622 F9→F11 0.64** _0.799** _0.723** _0.639**

Goodnessofﬁt Chi-square/df=1.745 Chi-square/df=1.728 Chi-square/df=1.775 Chi-square/df=1.742

GFI=0.905 GFI=0.893 GFI=0.917 GFI=0.867

AGFI=0.882 AGFI=0.868 AGFI=0.897 AGFI=0.835

RMSEA=0.045 RMSEA=0.048 RMSEA=0.042 RMSEA=0.055

*_p_<_0.05. **_p_<_0.01.

#_p_<_0.1

References

Ackaah,W.,Afukaar,F.K.,2010.Prevalenceofhelmetuseamongmotorcycleusers intamalemetropolis,ghana:anobservationalstudy.TrafﬁcInj.Prev.11(5), 522–525.

Ajzen,I.,1991.Thetheoryofplannedbehavior.Organ.Behav.Hum.Decis.Process. 50(2),179–211.

Begg,D.J.,Langley,J.D.,2004.Identifyingpredictorsofpersistentnon-alcoholor drug-relatedriskydrivingbehavioursamongacohortofyoungadults.Accid. Anal.Prev.36(6),1067–1071.

Bina,M.,Graziano,F.,Bonino,S.,2006.Riskydrivingandlifestylesinadolescence. Accid.Anal.Prev.38(3),472–481.

Bingham,C.R.,Shope,J.T.,Raghunathan,T.,2006.Patternsoftrafﬁcoffensesfrom adolescentlicensureintoearlyyoungadulthood.J.Adolesc.Health39(1),35–42. Blincoe,K.M.,Jones,A.P.,Sauerzapf,V.,Haynes,R.,2006.Speedingdrivers’attitudes andperceptionsofspeedcamerasinruralEngland.Accid.Anal.Prev.38(2), 371–378.

Chan,D.C.N.,Wu,A.M.S.,Hung,E.P.W.,2010a.Invulnerabilityandtheintentionto drinkanddrive:anapplicationofthetheoryofplannedbehavior.Accid.Anal. Prev.42(6),1549–1555.

Chan,E.,Pradhan,A.K.,Pollatsek,A.,Knodler,M.A.,Fisher,D.L.,2010b.Are driv-ingsimulatorseffectivetoolsforevaluatingnovicedrivers’hazardanticipation, speedmanagement,andattentionmaintenanceskills?Transp.Res.Pt.F-Trafﬁc Psychol.Behav.13(5),343–353.

Chang,H.L.,Yeh,T.H.,2007.Motorcyclistaccidentinvolvementbyage,gender,and riskybehaviorsinTaipei,Taiwan.Transp.Res.PartF-TrafﬁcPsychol.Behav.10 (2),109–122.

Chen,C.F.,2009.Personality,safetyattitudesandriskydrivingbehaviors-Evidence fromyoungTaiwanesemotorcyclists.Accid.Anal.Prev.41(5),963–968. Dejoy,D.M.,1992.Anexaminationofgenderdifferencesintrafﬁcaccidentrisk

perception.Accid.Anal.Prev.24(3),237–246.

Elliott,M.A.,Thomson,J.A.,2010.Thesocialcognitivedeterminantsofoffending drivers’speedingbehaviour.Accid.Anal.Prev.42(6),1595–1605.

Fabrigar,L.R.,Porter,R.D.,Norris,M.E.,2010.Somethingsyoushouldknowabout structuralequationmodelingbutneverthoughttoask.J.Consum.Psychol.20 (2),221–225.

Fernandes,R.,Hatﬁeld,J.,SoamesJob,R.F.,2010.Asystematicinvestigationofthe differentialpredictorsforspeeding,drink-driving,drivingwhilefatigued,and notwearingaseatbelt,amongyoungdrivers.Transp.Res.PartF-TrafﬁcPsychol. Behav.13(3),179–196.

Forward,S.E.,2010.Intentiontospeedinaruralarea:reasonedbutnotreasonable. Transp.Res.PartF-TrafﬁcPsychol.Behav.13(4),223–232.

Hamed,M.M.,Easa,S.M.,1998.Orderedprobabilitymodelingofseatbeltusage.J. Transp.Eng.-ASCE124(3),271–276.

Haque,M.M.,Chin,H.,Huang,H.,2009.Modelingfaultamongmotorcyclistsinvolved incrashes.Accid.Anal.Prev.41(2),327–335.

Hatcher,L.,1994.AStep-by-stepApproachtoUsingtheSASSystemforFactor Anal-ysisandStructuralEquationModeling.SASInstituteInc,Cary.

Iacobucci,D.,2010.Structuralequationsmodeling:ﬁtindices,samplesize,and advancedtopics.J.Consum.Psychol.20(1),90–98.

Iversen,H.,2004.Risk-takingattitudesandriskydrivingbehaviour.Transp.Res.Part F-TrafﬁcPsychol.Behav.7(3),135–150.

Jonah,B.A.,1997.Sensationseekingandriskydriving:areviewandsynthesisofthe literature.Accid.Anal.Prev.29(5),651–665.

Jun,J.,Guensler,R.,Ogle,J.,2011.Differencesinobservedspeedpatternsbetween crash-involvedandcrash-not-involveddrivers:applicationofin-vehicle moni-toringtechnology.Transp.Res.PartC:EmergingTechnol.19(4),569–578. Kim,K., Yamashita,E.Y.,2007. Attitudesof commercialmotorvehicledrivers

towardssafetybelts.Accid.Anal.Prev.39(6),1097–1106.

Langley,J.D.,Wagenaar,A.C.,Begg,D.J.,1996.AnevaluationoftheNewZealand graduateddriverlicensingsystem.Accid.Anal.Prev.28(2),139–146. Lin,M.R.,Chang,S.H.,Pai,L.,Keyl,P.M.,2003.Alongitudinalstudyofriskfactorsfor

motorcyclecrashesamongjuniorcollegestudentsinTaiwan.Accid.Anal.Prev. 35(2),243–252.

Lin,M.R.,Huang,W.,Hwang,H.F.,Wu,H.D.I.,Yen,L.L.,2004.Theeffectofcrash experienceonchangesinrisktakingamongurbanandruralyoungpeople.Accid. Anal.Prev.36(2),213–222.

Machin,M.A.,Sankey,K.S.,2008.Relationshipsbetweenyoungdrivers’personality characteristics,riskperceptions,anddrivingbehaviour.Accid.Anal.Prev.40(2), 541–547.

McKnight,A.J.,McKnight,A.S.,2003.Youngnovicedrivers:carelessorclueless? Accid.Anal.Prev.35(6),921–925.

MTC,2007a.MonthlyStatisticsofTransportationandCommunications.Taipei, Taiwan: Ministryof Transportationand Communications, Executive Yuan, Taiwan.

MTC,2007b.SpecialReport:SurveyofMotorcycleUsageinTaiwan.Taipei,Taiwan: Ministry ofTransportation and Communications, Executive Yuan, Taiwan. Taiwan.

Nelson,E.,Atchley,P.,Little,T.D.,2009.Theeffectsofperceptionofriskand impor-tanceofansweringandinitiatingacellularphonecallwhiledriving.Accid.Anal. Prev.41(3),438–444.

NHTSA,2009.2007MotorcyclesTrafﬁcSafetyFactSheet:NationalHighwayTrafﬁc SafetyAdministration.

Nordfjaern,T.,Jorgensen,S.H.,Rundmo,T.,2010.Aninvestigationofdriverattitudes andbehaviourinruralandurbanareasinNorway.Saf.Sci.48(3),348–356. Simons-Morton,B.,Lerner,N.,Singer,J.,2005.Theobservedeffectsofteenage

pas-sengersonthe.riskydrivingbehaviorofteenagedrivers.Accid.Anal.Prev.37 (6),973–982.

Simpson,H.M.,2003.Theevolutionandeffectivenessofgraduatedlicensing.J.Saf. Res.34(1),25–34.

Sumer,N.,2003.Personalityandbehavioralpredictorsoftrafﬁcaccidents:testinga contextualmediatedmodel.Accid.Anal.Prev.35(6),949–964.

(12)

Taubman-Ben-Ari,O.,Mikulincer,M.,Gillath,O.,2004.Themultidimensional driv-ingstyleinventory–scaleconstructandvalidation.Accid.Anal.Prev.36(3), 323–332.

Tseng,P.-Y.,Huang,Y.-S.,Jiang,S.-Y.,2001.Analysisofaccidentrisksbydriverage.In: ProceedingsoftheInternationalConferenceofTrafﬁcSafetyandEnforcement, Taoyuan,Taiwan.

Ulleberg,P.,2001.Personalitysubtypesofyoungdrivers.Relationshiptorisk-taking preferences,accidentinvolvement,andresponsetoatrafﬁcsafetycampaign. Transp.Res.PartF-TrafﬁcPsychol.Behav.4(4),279–297.

Ulleberg,P.,Rundmo,T.,2003.Personality,attitudesandriskperceptionas predic-torsofriskydrivingbehaviouramongyoungdrivers.Saf.Sci.41(5),427–443. Vance,D.E.,Roenker,D.L.,Cissell,G.M.,Edwards,J.D.,Wadley,V.G.,Ball,K.K.,2006.

Predictorsofdrivingexposureandavoidanceinaﬁeldstudyofolderdrivers fromthestateofMaryland.Accid.Anal.Prev.38(4),823–831.

Walsh,S.P.,White,K.M.,Hyde,M.K.,Watson,B.,2008.Diallinganddriving:factors inﬂuencingintentionstouseamobilephonewhiledriving.Accid.Anal.Prev.40 (6),1893–1900.

Wells-Parker,E.,Ceminsky,J.,Hallberg,V.,Snow,R.W.,Dunaway,G.,Guiling,S., Williams,M.,Anderson,B.,2002.Anexploratory studyoftherelationship betweenroadrageandcrashexperienceinarepresentativesampleofUSdrivers. Accid.Anal.Prev.34(3),271–278.

Wong, J.T., Chung, Y.S., Hsiao, L.W., 2010a. Structural discrepancy of riding behavioramongheterogeneousyoungmotorcyclistsinTaiwan. Transporta-tionResearch Record.Journalof theTransportationResearch Board2194, 107–114.

Wong,J.T.,Chung,Y.S.,Huang,S.H.,2010b.Determinantsbehindyoung motorcy-clists’riskyridingbehavior.Accid.Anal.Prev.42(1),275–281.

Yagil,D.,1998.Instrumentalandnormativemotivesforcompliancewithtrafﬁclaws amongyoungandolderdrivers.Accid.Anal.Prev.30(4),417–424.

Zamani-Alavijeh,F.,Niknami,S.,Bazargan,M.,Mohamadi,E.,Montazeri,A., Ghofra-nipour,F.,Ahmadi,F.,Tavaﬁan,S.S.,ShahrzadBazargan,H.,2010.Risk-taking behaviorsamongmotorcyclistsinmiddleeastcountries:acaseofIslamic Repub-licofIran.TrafﬁcInj.Prev.11(1),25–34.