1-s2.0-S0148296308000040-main

Available online at http://wendang.chazidian.com

JournalofBusinessResearch62(2009)61–

67

Informationdirection,websitereputationandeWOMeffect:

Amoderatingroleofproducttype

CheolParka,1,ThaeMinLeeb,?

b

SchoolofBusinessAdministration,KoreaUniversity,Jochiwon-Eup,Yongi-Gun,Chungnam,339-700,SouthKorea

DepartmentofBusinessAdministration,ChungbukNationalUniversity,Gaeshin-Dong,Heungduk-Gu,Cheongju,Chungbuk,361-763,SouthKorea

Received1December2006;receivedinrevisedform1October2007;accepted1November2007

a

Abstract

Thisarticleexamineshowtheelectronicwordofmouth(eWOM)informationdirection(positivevs.negative)andawebsite'sreputation(establishedvs.unestablished)contributetotheeWOMeffect.Thearticledescribesastudyfocusingonthemoderatingroleoftheproducttype(searchvs.experience).TheresultsoftheexperimentshowthattheeWOMeffectisgreaterfornegativeeWOMthanforpositiveeWOM,greaterforestablishedwebsitesthanforunestablishedwebsites,andgreaterforexperiencegoodsthanforsearchgoods.TheresultssupportthemoderatingeffectsofproducttypeontheeWOMinformationdirection-websitereputation-eWOMeffectrelationship.TheimpactofnegativeeWOMontheeWOMeffectisgreaterforexperiencegoodsthanforsearchgoods.Similarly,theimpactofwebsitereputationontheeWOMeffectisgreaterforexperiencegoodsthanforsearchgoods.ThefindingsprovidemanagerialimplicationsforanInternetmarketingstrategy.©2008ElsevierInc.Allrightsreserved.

Keywords:Internetmarketing;Electronicwordofmouth(eWOM);Direction;Websitereputation;Searchgoods;Experiencegoods

1.Introduction

ThearrivalandexpansionoftheInternethasextendedconsumers'optionsforgatheringproductinformationbyincludingotherconsumers'comments,postedontheInternet,andhasprovidedconsumersopportunitiestooffertheirownconsumption-relatedadvicebyengaginginelectronicword-of-mouth(eWOM)(Hennig-Thurauetal.,2004).Gruen,Osmon-bekovandCzaplewski(2006)suggestthatthecustomer-to-customerknow-howexchange,aspecificformofeWOM,hasa

Correspondingauthor.LeeistobecontactedatDepartmentofBusinessAdministration,ChungbukNationalUniversity,Gaeshin-Dong,Heungduk-Gu,Cheongju,Chungbuk,361-763,SouthKorea.Tel.:+82432612334;fax:+82432762347.

E-mailaddresses:cpark@korea.ac.kr(C.Park),leetm@chungbuk.ac.kr(T.M.Lee).1

Park,SchoolofBusinessAdministration,KoreaUniversity,Jochiwon-Eup,Yongi-Gun,Chungnam,339-700,SouthKorea.Tel.:+82418601566;fax:+82418601507.

0148-2963/$-seefrontmatter©2008ElsevierInc.Allrightsreserved.doi:10.1016/j.jbusres.2007.11.017

directrelationshipwithloyaltyintentions,aswellasanindirectrelationshipmediatedthroughtheoverallvalueofthefirm'soffering.Previousresearchontraditional(offline)WOMsuggeststhatfactorssuchassourcecredibility(DholakiaandSternthal,1977)andWOMdirections(positiveornegative)(Arndt,1967)areantecedentsoftheWOMeffect.

IneWOMhowever,unlikethecaseofWOMfrominterpersonalsources,recommendationsaretypicallyfromunknownindividualsandinatext-basedformat.So,onlineconsumershavedifficultyinusingsourcesimilaritytodeterminethecredibilityofinformation(Chatterjee,2001).Therefore,inordertobuildeffectiveInternetmarketingstrategy,marketersshouldunderstandtheeWOMeffect,notonlyintermsofsourcecredibility,butalsointermsofthemessageconfiguration(negative/positiveeWOMpostedontheestablished/unestablishedwebsite).However,littleresearchaddresseshoweWOMmessageconfigurationsaffecttheeWOMeffect.Therefore,thevehicleeffectisconsideredbyincorporatingawebsite'sreputation(Shamdasanietal.,2001)asadeterminantofthecredibilityoftheeWOM-specificsource.

62C.Park,T.M.Lee/JournalofBusinessResearch62(2009)61–67

Inaddition,thisstudysuggeststheeWOMinformationdirectionasamessagecharacteristicvariable.

Oneinteresting,butunansweredquestion,isdoestheimpactoftheeWOMinformationdirectionandwebsitereputationontheeWOMeffectvarywiththeproducttype?Inordertoaddressthisvoid,thisstudyfocusesonthemoderatingeffectofproducttype(searchvs.experience)ininvestigatingtheimpactofeWOMinformationdirectionandwebsitereputationontheeWOMeffect.

Therefore,theobjectivesofthisstudyincludethefollowingpoints.First,toexaminetheimpactoftheeWOMinformationdirection(positivevs.negativeeWOM)ontheeWOMeffect;second,toinvestigatetheinfluenceofwebsitereputation(establishedvs.unestablishedwebsite)ontheeWOMeffect;andfinally,toexaminehowtheinteractionsofeWOMinformationdirectionandwebsitereputationwithproducttype(searchgoodsvs.experiencegoods)influencetheeWOMeffect.

2.Conceptualbackgroundandresearchhypotheses2.1.eWOMdirection

Asignificantfindinginimpressionformationliteratureisthenegativityeffect,namelythatpeopleplacemoreweightonnegativeinformationratherthanpositiveinformationinthesubjects'formingoverallevaluations(SkowronskiandCarl-ston,1989).Evidencesuggeststhatnegativeinformationismoreattentiongrabbingingeneralandreceivesgreaterscrutinythanpositiveinformation(HomerandYoon,1992).PreviousresearchsuggeststhattheWOMinfluenceappearstobeasymmetricalinthatanegativeWOMhasastrongerinfluenceoncustomers'brandevaluations(Arndt,1967;Mizerski,1982;Richins,1983;Wright,1974)andonthepurchaseintentionsofpotentialbuyers(BrownandReingen,1987;Weinbergeretal.,1981)thanapositiveWOM.Theseargumentsofferarationaleforthefollowinghypothesis.H1:TheeWOMeffectisgreaterfornegativeeWOMinformationthanforpositiveeWOMinformation.

2.2.Websitereputation

Sourcevariables,whetherintheformofamessagespokespersonoramediavehicleitselfessentiallymoderatetheimpactofapersuasivemessagebyinfluencingtheaudience'sconfidencebasedontheendorsementbythesource(Shamdasanietal.,2001).ThisvehiclesourceeffectislikelytobeappliedtotheeWOMcontext.ConsumerconfidencebroughtaboutbyacrediblewebsitemightinfluencetheeWOMpostedonthewebsite.

AccordingtoShamdasanietal.(2001),duetolowbarriersofentryandlackofgatekeepersontheWeb,well-establishedandreputablewebsiteshavebeenmorereadilyacceptedbyconsumersthanhaveunknownsites.Thepreviousdiscussionsuggeststhefollowinghypothesis.H2:TheeWOMeffectisgreaterforwebsiteswithestablishedreputationsthanforthosewithoutestablishedreputations.

2.3.Categorizationofgoods:searchvs.experience

AccordingtoNelson(1974),searchgoodsaredefinedasthosecharacterizedbyproductattributeswherecompleteinformationaboutthegoodscanbeacquiredpriortopurchase;experiencegoodsarecharacterizedbyattributesthatcannotbeknownuntilthepurchaseandafteruseoftheproductorforwhichaninformationsearchismorecostlyand/ordifficultthandirectproductexperience(Klein,1998).MauteandForrester(1991)examinetheapplicabilityofthissearch/experienceframeworktotheinformationsearchprocess.MauteandForrester(1991)suggestsearchandexperiencequalitiesasmoderatorsofthelinkbetweensearchantecedentsandexternalsearchefforts.ByextendingthisframeworktothestudyoftheinfluenceofeWOMinformationcontentontheeWOMeffectacrossproductcategories,thisstudyincludesMauteandForrester'sconceptualizationofproductinformationattributesasmoderatorsoftheeWOMinformation–eWOMeffectrelationship.

Paststudiesprovideevidenceattestingtothenotionthatthecharacteristicsoftheproduct(i.e.,thesubjectofWOMcommunications)mayaffecthowconsumersprocesstheWOMmessage(SundaramandWebster,1999).

Ifconsumershavedifficultyinjudgingproductqualityorifjudgmentalcriteriaareambiguous,thevalueofavailableinformationforthepurposesofanalysisincreases(Bone,1995).ThissuggeststhattheinfluenceofWOMmaybegreaterinsomesituationsthaninothers;specifically,WOMeffectsshouldbegreaterwhentheconsumerfacesanambiguoussituation(Bone,1995).

Heightenedperceivedriskalongthesearch-experienceproductcontinuum,whichresultsfromalackofknowledgeandinformation,increasestheimportanceofexperience(Hsiehetal.,2005).Consumers,especiallywhentheyintendtopurchaseexperiencegoods,canreducetheiruncertaintybyreferringtoeWOMinformation,suchasonlineconsumerreviews,thattypicallycontainexpertadviceandcommentsfromexperiencedusers.Forthisreason,theeWOMeffectisexpectedtobegreaterforexperiencegoodsratherthanforsearchgoods,leadingtothefollowinghypothesis.H3:TheeWOMeffectisgreaterforexperiencegoodsthanforsearchgoods.

2.4.Interactioneffect:eWOMdirectionandproducttypeHigherlevelsofbrandfamiliarity,generatedthroughdirectorindirectbrand-relatedexperiences,areassociatedwithawelldevelopedknowledgestructureaboutthebrandanditsattributes(AlbaandHutchinson,1987).Giventhatconsumersarenotlikelytochangetheirattitudestowardfamiliarbrands(HoyerandMacInnis,1997),exposureoffamiliarbrandtoWOMcommunications,eitherpositiveornegative,isnotlikelytoproducesignificantchangesinconsumers'preexist-ingbrandevaluations(SundaramandWebster,1999).Lessfamiliarbrands,thatareusuallyassociatedwithconsumers'underdevelopedknowledgestructures,aremoresusceptibletochangeintheirbrandevaluationsfollowingexposureto

C.Park,T.M.Lee/JournalofBusinessResearch62(2009)61–6763

anynewbrand-relatedcommunications.Consumers,lessfamiliarwithabrand,aremoreamenabletoprocessingnewbrand-relatedinformationandchangingtheirbrandevalua-tionsbasedonthedirectionoftheinformation(SundaramandWebster,1999).GiventhatfindingsearchattributesontheInternetiseasierthanfindingexperienceattributes,con-sumersarelikelytopossessamuchmoredetailed,richcognitivestructureforsearchgoodsthanforexperiencegoods.Inaddition,HochandHa(1986)suggestthatadvertising'sframing(positive/negative)effectisstrongerwhentheproductisambiguous,thatis,whenqualityishardforaconsumertodetermine.Therefore,theevaluationofexperiencegoods,comparedwithsearchgoods,maysustaingreaterdamageintheeWOMeffectduetonegativeinformationbecausethenegativeeWOMinformationwillmagnifyconsumers'prevailinguncertaintyandfearrelatedtotheirpoorcognitiveknowledgestructure.Theseargumentsgiverisetothefollowinghypothesis.H4:Theimpactdiffer-enceofnegativeandpositiveeWOMinformationontheeWOMeffectisgreaterforexperiencegoodsthanforsearchgoods.2.5.Interactioneffect:websitereputationandproducttypeThemessagesourceisacrucialdeterminantofpersuasionespeciallywhenconsumersevaluateexperienceattributesforaproduct(JainandPosavac,2001).JainandPosavac(2001)arguethatwhenconsumershavenopriorknowledgeofabrand'squalities,theuseofhigh-credibilitysourcescommu-nicatesexperienceattributeinformationmorepersuasively.ThisstudyincludesthewebsitereputationwhichisexpectedtoserveasextrinsiccuesofeWOMinformationqualityespeciallyinthecaseofpurchasingexperiencegoods.

IftheconsumerisexposedtoeWOMinformationfromahighlycrediblesource(highlyreputablewebsite)endorsingexperiencegoods,theeWOMinformationshouldbemorepersuasivethananyeWOMinformationfromalesscrediblesource.Thispredictionshouldholdbecauseitisamaineffectprediction:thatthosehighlycrediblesources(establishedreputationwebsites)shouldbemorepersuasivethansourceswithlittleornocredibility(unestablishedwebsites).Thisbelievabilityisevenmoreimportantintheambiguouscircumstanceofexperiencegoodswhereconsumersdonotpossesspriorknowledgeorexperience.H5:Theimpactdifferenceofestablishedandunestablishedwebsite'sreputationontheeWOMeffectisgreaterforexperiencegoodsthanforsearchgoods.3.Method3.1.Researchdesign

ThisstudyutilizedanexperimentthatmanipulatedtheeWOMinformationdirection,thewebsitereputation,andtheproducttype,thusresultingina2(positivevs.negative)×2(establishedvs.unestablished)×2(searchvs.experience)between-subjectsfactorialdesign.Subjectswereassignedrandomlytotheexpe-riment'streatments.

3.2.Developmentofstimuli

Anin-depthfocusgroup,involvingeightparticipants,wasconductedtobrainstormforsearchandexperiencegoods.Fromthelistof20productsgenerated,6productswerethenchosentobeincludedinthepretest.

BasedonNelson(1974),producttypewasmeasuredona5-pointscale(1=productqualitycanbeevaluatedpriortopurchase,5=productqualitycannotbeevaluateduntilpurchaseanduseoftheproduct)forasampleof20malesand20females.TheresultsindicatedthattheproducttypemeanforalanguageschoolprogramwasthehighestandtheTOIEC(TestofEnglishforInternationalCommunication)bookwasthelowest.Themeansdifferenceofthesetwoproductswassignificant(Mbook=1.5,Mlanguageschool=4.4,t(39)=?25.76,p=.000),whileevaluationsmadebymaleandfemalerespondentsforthesetwoproductsshowednosignificantdifference.AccordingtotheFGI(FocusGroupInterview)andpretestresults,theTOIECbookforsearchgoodsandthelanguageschoolprogramforexperiencegoodswereselectedforthemainexperiment.DevelopmentofwebsitesandeWOM.Inordertoensureconsistencyacrosstreatments,alltreatmentwebpageswerecreatedfromoneactualwebpage,keepinganidenticalstyleandlayoutthroughout,differingonlyintheInternetshoppingmallname,producttype,andeWOMinformationdirection,thatconstitutedthemanipulation.

AfterreviewingsamplesofeWOMpostedonthewebsite,apanelofdoctoralcandidatesandmarketingfacultyfamiliarwitheWOMliteraturegeneratedalistofeightkeyitemsforeachproduct.

http://wendang.chazidian.com(openedasanInternetshoppingmallfirstinKoreainJune,1996)http://wendang.chazidian.com(astart-upInternetshoppingmallinKorea)wasselectedasanunestablishedwebsite.

Fictitiousbrandnameswerecreated(“Point”fortheTOEICbookand“Nice”forthelanguageschoolprogram)toavoidpossibleerrorsources.Fig.1displayssampleexperimental

stimulus.

Fig.1.AsampleexperimentaleWOM(positive/search/established).

64C.Park,T.M.Lee/JournalofBusinessResearch62(2009)61–67

3.3.Pretesting

Pretestswereconductedwithasampleof20malesand20females.Toassessthepositivity/negativityoftheframingmanipulation,thesubjectsratedtheextenttowhichtheyfeltthattheeWOMinformationstressedthenegative/positiveaspectsofthediscussedobjectona5-pointscale(1=negativeinformationto5=positiveinformation).Thewebsitereputationwasmeasuredwith10itemsona5-pointsemanticdifferentialscale(α=0.92),adaptedfromShamdasanietal.(2001),measuringthewebsite'sreputationintermsofitsexpertiseandtrustworthiness.TheordereffectwascontrolledbyrandomizingthearrangementofeWOMinformation.ThepretestresultssuggestthattheeWOMdirection(Mnegative=1.5,Mpositive=4.7,t(39)=?29.24,p=.000)andwebsitereputation(meanvalueof10items)(Munestablished=1.6,Mestablished=4.0,t(39)=28.92,p=.000)measuresyieldsignificantdifferencesintheexpecteddirection.TheseresultsindicatethatthemanipulationsoftheeWOMinformationdirectionandwebsitereputationaresuccessful.3.4.Measures

TheeWOMeffectwasmeasuredby3itemsona5-pointratingsofagreement(1=stronglydisagreeto5=stronglyagree)withthefollowingthreestatementsadaptedfromJeonandPark(2003):(1)IwillrefertothiseWOMinformationinapurchasedecision.(2)Overall,IthinkthiseWOMinformationiscredible.(3)ThiseWOMinformationwillcruciallyaffectmypurchasedecision.Thereliabilitycoefficientforthescalewas0.73.TheeWOMinformationdirection,websitereputa-tion,andproducttypeweremeasuredforthesameitemsforthepretests.

Covariates.Pastresearchsuggestthatbrandfamiliarityaffectconsumerpurchasedecisions(LaneandJacobson,1995).ConsumersarelikelytobehaveinasimilarpatternwhenprocessingeWOMinformation.Thus,eWOMfamiliarityarelikelytoaffecteWOM-basedpurchasedecisions.TocontrolfortheeWOMfamiliarityeffect,eWOMfamiliarityisincludedinthemodelasacovariate.eWOMfamiliaritywasmeasuredwith3itemsona5-pointsemanticdifferentialscale(α=0.82),usingstatementsadaptedfromKentandAllen(1994):RegardingeWOMinformation,Iam(1)unfamiliar/familiar,(2)inexperienced/experienced,and(3)notknowl-edgeable/knowledgeable.3.5.Experimentalprocedure

Undergraduatestudents,takingbusinessclassesatamajoruniversityinSeoul,hadnoticeofanonlineexperiment.Classassistantrequestedwebmastertopostanotice/linkconcerningtheonlineexperimentontheseveralclasses'websites.Inordertomotivatesubjectstorespond,thesubjectsweregivenfreemealvouchersupontheirsurveysubmission.Theonlinesurveywasprogrammedsuchthatrespondentswouldbeassignedrandomlytooneoftheeightconditionsintheexperiment,with

Table1

ANCOVAresultsSource

dfMSF-valueCovariate

eWOMfamiliarity

16.2422.98***eWOMinformationdirection(A)122.5482.99***Websitereputation(B)129.25107.70***Producttype(C)149.45182.06***A×B10.060.20A×C11.314.81*B×C12.137.86**A×B×C13.7513.79***

Error4310.27

Total

439

*pb.05,**pb.01,***pb.001.

thecellhavingtheleastresponsesgivenpriority,henceensuringequalcellsizes.Ashortdescriptionofthewebsiteswasgiven;then,subjectswereshowntheeWOMinformationdirection,websitereputation,andproducttypecombinationsasdesig-natedbythetreatmentcell.Followingthisexposure,eWOMeffectsweremeasuredbasedontheeWOMstimulithesubjectshadjustseen.Covariatesandmanipulationchecksforexperimentalstimuliwerethenmeasured(withthesameitemsforpretests),followedbyquestionsaboutgeneraldemographics.

4.Analysisandresults

Atotalof440responseswerecollectedwith55responsesineachoftheeighttreatmentconditions.Respondentswereaged19–28with55.5%menand44.5%women.Allexperimentalmanipulationswereagainassessedandfoundtobeeffectiveinthemainstudy.4.1.Hypothesistesting

Thehypothesesweretestedusinga2×2×2analysisofcovariancewitheWOMfamiliarityascovariate.ANCOVAresultsareavailableinTable1.Table2presentsthemeansandstandarddeviationsoftheeightcellsinvolvedintheexperi-mentaldesignforthisstudy.Fig.2illustratesthesummaryofhypothesestesting.AsshowninFig.2,allthehypothesesaresupported.Figs.3and4displaytheinteractioneffectsforeWOMinformationdirection-producttypeandwebsitereputa-tion-producttype.

TheANCOVAresults(seeTable1)includestatisticallysignificant3-wayinteractioneffect(F(1,431)=13.79,pb.001).Thestructureoftheobserved3-wayinteractionappearsinFig.5.

ComparisonsshowthattheimpactontheeWOMeffectresultingfromnegativeeWOMwithestablishedwebsitesissignificantlygreaterthanpositiveeWOMforsearchgoods(3.6versus3.0;pb.001)andexperiencegoods(4.4versus4.0;pb.001).However,thisdifferenceisinsignificantinthecaseoftheeWOMforsearchgoodspostedontheunestablishedwebsites(3.0versus2.9;pN.20).Incontrast,forexperience

C.Park,T.M.Lee/JournalofBusinessResearch62(2009)61–67

Table2

ThecellmeansandstandarddeviationsfortheeWOMeffecteWOMinformationdirectionPositiveNegativeTotal

EstablishedreputationSearch3.0(0.53)3.6(0.51)3.3(0.59)3.7(0.68)

Experience4.0(0.44)4.4(0.41)4.2(0.47)

UnestablishedreputationSearch2.9(0.44)3.0(0.68)3.0(0.58)3.2(0.70)

Experience3.1(0.48)3.9(0.70)3.5(0.70)

3.3(0.63)3.7(0.76)

65

Note:Thestandarddeviationsareinparentheses.

Fig.3.TheeffectsoftheeWOMinformationdirectionontheeWOMeffectasafunctionofproducttype.

goods,negativeeWOMwithunestablishedwebsitesresultinsignificantlygreaterimpactontheeWOMeffectthanpositiveeWOM(3.9versus3.1;pb.001).

5.Generaldiscussion,implications,andlimitations5.1.Generaldiscussion

First,theeWOMeffectisgreaterfornegativeeWOMthanforpositiveeWOM.Inaddition,whereasmostpreviousresearchhasfocusedonthenegativityeffectandthereasonsforthateffect,thisresearchshowsthattheproducttypeassociatedwiththeeWOMdiscussionsmoderatesthiseffect.Specifically,thenegativityeffectappearstobemoresignificantwhentheeWOMisforexperiencegoodsratherthanforsearchgoods.

GiventhatfindingsearchattributesontheInternetiseasierthanfindingexperienceattributes,consumersarelikelytopossessamuchmoredetailed,richcognitivestructureofsearch

goodsthanofexperiencegoods.Empiricalresultssuggestthatexperiencegoods,comparedwithsearchgoods,sustaingreaterdamageintheeWOMeffectduetonegativeeWOMinformation.ThereasonisthatthenegativeeWOMinformationmagnifyconsumers'prevailinguncertaintyandfearinitiatedbytheirpoorcognitiveknowledgestructureofexperiencegoods.ThisresearchprovidesaninsightintowhyandhowtheeWOMnegativityeffectisdifferentacrossproducttypes.

Second,thewebsitereputationhasasignificantimpactontheeWOMeffect.Specifically,theeWOMeffectisgreaterforwebsiteswithestablishedreputationsthanforunestablishedwebsites.Thisresultempiricallyvalidatesthevehiclesourceeffect(websitereputation)(Shamdasanietal.,2001)intheeWOMcontext.Thisprovidesaninsightintoregardingawebsite'sreputationasdimensionsforsourcecredibilityintheeWOMcontext.

Anotherinterestingfindingofthisstudyisthattheimpactofawebsite'sreputationontheeWOMeffectisgreater

for

Fig.2.Asummaryofhypotheses

testing.