298 95 2MB
English Pages 225 Year 2008
Review of
Marketing Research
Review of Marketing Research VOLUME 4
Naresh K. Malhotra Editor
M.E.Sharpe Armonk,NewYork London,England
Copyright©2008byM.E.Sharpe,Inc. Allrightsreserved.Nopartofthisbookmaybereproducedinanyform withoutwrittenpermissionfromthepublisher,M.E.Sharpe,Inc., 80BusinessParkDrive,Armonk,NewYork10504. LibraryofCongressISSN:1548-6435 ISBN:978-0-7656-2092-7 PrintedintheUnitedStatesofAmerica Thepaperusedinthispublicationmeetstheminimumrequirementsof AmericanNationalStandardforInformationSciences PermanenceofPaperforPrintedLibraryMaterials, ANSIZ39.48-1984.
~
MV(c) 10 9 8 7 6 5 4 3 2 1
REVIEWOFMARKETINGRESEARCH EDITOR:NARESHK.MALHOTRA,GEORGIAINSTITUTEOFTECHNOLOGY EditorialBoard RickP.Bagozzi,UniversityofMichigan RussBelk,UniversityofUtah RuthBolton,ArizonaStateUniversity GeorgeDay,UniversityofPennsylvania WayneDeSarbo,PennsylvaniaStateUniversity JanHeide,UniversityofWisconsin,Madison MorrisB.Holbrook,ColumbiaUniversity MichaelHouston,UniversityofMinnesota ShelbyHunt,TexasTechUniversity DawnIacobucci,VanderbiltUniversity JeffInman,UniversityofPittsburgh ArunK.Jain,UniversityatBuffalo,StateUniversityofNewYork BarbaraKahn,UniversityofMiami WagnerKamakura,DukeUniversity DonaldLehmann,ColumbiaUniversity RobertF.Lusch,UniversityofArizona DebbieMacInnis,UniversityofSouthernCalifornia KentB.Monroe,UniversityofIllinois,Urbana EitanMuller,NewYorkUniversity NelsonNdubisi,MonashUniversity,Malaysia A.Parasuraman,UniversityofMiami WilliamPerreault,UniversityofNorthCarolina RobertA.Peterson,UniversityofTexas NigelPiercy,UniversityofWarwick JagmohanS.Raju,UniversityofPennsylvania VithalaRao,CornellUniversity BrianRatchford,UniversityofTexasatDallas JagdishN.Sheth,EmoryUniversity ItamarSimonson,StanfordUniversity Jan-BenedictE.M.Steenkamp,UniversityofNorthCarolina,ChapelHill DavidStewart,UniversityofCalifornia,Riverside RajanVaradarajan,TexasA&MUniversity MichelWedel,UniversityofMaryland BartonWeitz,UniversityofFlorida v
vI RUTH N. BOLTON AND CRINA O. TARASI
ADHOCREVIEWERS AnochaAribarg,UniversityofMichigan UlfBöckenholt,McGillUniversity EricBradlow,UniversityofPennsylvania MichaelKamins,UniversityofSouthernCalifornia DeborahMacInnis,UniversityofSouthernCalifornia JosephPancras,UniversityofConnecticut C.W.Park,UniversityofSouthernCalifornia S.Sriram,UniversityofConnecticut RajkumarVenkatesan,UniversityofVirginia GalZauberman,UniversityofNorthCarolina,ChapelHill
vi
CONTENTS ReviewofMarketingResearch:TakingStock NareshK.Malhotra
ix
Contents,Volume1 Contents,Volume2 Contents,Volume3
xv xvii xix
1. FormalChoiceModelsofInformalChoices:WhatChoiceModeling ResearchCan(andCan’t)LearnfromBehavioralTheory JordanJ.LouviereandRobertJ.Meyer
3
2. HowMuchtoUse?AnAction-GoalApproachtoUnderstandingFactors InfluencingConsumptionQuantity ValerieS.FolkesandShashiMatta
33
3. IntegratingPurchaseTiming,Choice,andQuantityDecisionsModels: AReviewofModelSpecifications,Estimations,andApplications V.KumarandAnitaManLuo
63
4. BrandExtensionResearch:ACross-CulturalPerspective MichaelA.Merz,DanaL.Alden,WayneD.Hoyer,andKalpeshKaushikDesai
92
5. AReviewofEye-TrackingResearchinMarketing MichelWedelandRikPieters
123
6. RoleTheoryApproachesforEffectivenessofMarketing-OrientedBoundary Spanners:ComparativeReview,ConfiguralExtension,andPotentialContributions JagdipSinghandArgunSaatcioglu
148
7. DesigningPriceContractsforProcurementandMarketingofIndustrialEquipment GeorgeJohn
183
AbouttheEditorandContributors
201
vii
REVIEWOFMARKETINGRESEARCH TakingStock NARESHK.MALHOTRA
Overview ReviewofMarketingResearch,nowinitsfourthvolume,isarecentpublicationcoveringthe importantareasofmarketingresearchwithamorecomprehensivestate-of-the-artorientation. Thechaptersinthispublicationwillreviewtheliteratureinaparticulararea,offeracriticalcommentary,developaninnovativeframework,anddiscussfuturedevelopments,aswellaspresent specificempiricalstudies.Thefirstfourvolumesfeaturesomeofthetopresearchersandscholars inourdisciplinewhohavereviewedanarrayofimportanttopics.Theresponsetothefirstthree volumeshasbeentrulygratifyingandwelookforwardtotheimpactofthefourthvolumewith greatanticipation. PublicationMission Thepurposeofthisseriesistoprovidecurrent,comprehensive,state-of-the-artarticlesinreview ofmarketingresearch.Wide-rangingparadigmaticortheoretical,orsubstantiveagendasareappropriateforthispublication.Thisincludesawiderangeoftheoreticalperspectives,paradigms, data(qualitative,survey,experimental,ethnographic,secondary,etc.),andtopicsrelatedtothe studyandexplanationofmarketing-relatedphenomenon.Wehopetoreflectaneclecticmixture oftheory,data,andresearchmethodsthatisindicativeofapublicationdrivenbyimportanttheoreticalandsubstantiveproblems.Weseekstudiesthatmakeimportanttheoretical,substantive, empirical,methodological,measurement,andmodelingcontributions.Anytopicthatfitsunder thebroadareaof“marketingresearch”isrelevant.Inshort,ourmissionistopublishthebest reviewsinthediscipline. Thus,thispublicationwillbridgethegapleftbycurrentmarketingresearchpublications.Current marketingresearchpublicationssuchastheJournalofMarketingResearch(USA),International JournalofMarketingResearch(UK),andInternationalJournalofResearchinMarketing(Europe) publishacademicarticleswithamajorconstraintonthelength.Incontrast,ReviewofMarketing Researchwillpublishmuchlongerarticlesthatarenotonlytheoreticallyrigorousbutalsomore expository,withafocusonimplementingnewmarketingresearchconceptsandprocedures.This willalsoservetodistinguishthispublicationfromMarketingResearchmagazinepublishedby theAmericanMarketingAssociation(AMA). ix
x REVIEW OF MARKETING RESEARCH
ArticlesinReviewofMarketingResearchshouldaddressthefollowingissues: • • • • • • • • • •
Criticallyreviewtheexistingliterature Summarizewhatweknowaboutthesubject—keyfindings Presentthemaintheoriesandframeworks Reviewandgiveanexpositionofkeymethodologies Identifythegapsinliterature Presentempiricalstudies(forempiricalpapersonly) Discussemergingtrendsandissues Focusoninternationaldevelopments Suggestdirectionsforfuturetheorydevelopmentandtesting Recommendguidelinesforimplementingnewproceduresandconcepts
ArticlesintheFirstVolume The inaugural volume exemplified the broad scope of the Review of Marketing Research. It containedadiversesetofreviewarticlescoveringareassuchasemotions,beauty,information search,businessandmarketingstrategy,organizationalperformance,referencescales,andcorrespondence analysis. These articles were contributed by some of the leading scholars in the field,fiveofthembeingformereditorsofmajorjournals(JournalofMarketingandJournalof ConsumerResearch). JohnsonandStewartprovidedareviewoftraditionalapproachestotheanalysisofemotionin thecontextofconsumerbehavior.Theyreviewedappraisaltheoryanddiscussedexamplesofits applicationinthecontextsofadvertising,customersatisfaction,productdesign,andretailshopping. Holbrookexploredandreviewedtheconceptofbeautyasexperiencedbyordinaryconsumersin theireverydaylives.Histypologyconceptualizedeverydayusageoftheterm“beauty”asfalling intoeightcategoriesdistinguishedonthebasisofthreedichotomies:(i)extrinsically/intrinsically motivated;(ii)thing(s)-/person(s)-based;and(iii)concrete/abstract.XiaandMonroefirstreviewed theliteratureonconsumerinformationsearch,andthentheliteratureonbrowsing.Theyproposed anextendedconsumerinformationacquisitionframeworkandoutlinedrelevantsubstantiveand methodologicalissuesforfutureresearch.HuntandMorganreviewedtheprogressandprospects of the “resource-advantage” (R-A) theory. They examined in detail the theory’s foundational premises,showedhowR-Atheoryprovidesatheoreticalfoundationforbusinessandmarketingstrategy,anddiscussedthetheory’sfutureprospects.BharadwajandVaradarajanprovided aninterdisciplinaryreviewandperspectiveonthedeterminantsoforganizationalperformance. Theyexaminedtheclassicalindustrialorganizationschool,theefficiency/revisionistschool,the strategicgroupsschool,thebusinesspolicyschool,thePIMSparadigm,theAustrianschool,and theresource-basedviewofthefirm.Theyproposedanintegrativemodelofbusinessperformance thatmodeledfirm-specificintangibles,industrystructure,andcompetitivestrategyvariablesas themajordeterminantsofbusinessperformance.VargoandLuschfocusedattentiononconsumer referencescales,thepsychologicalscalesusedtomakeevaluationsofmarketing-relatedstimuli,in consumersatisfaction/dissatisfaction(CS/D)andservicequality(SQ)researchandproposedsocial judgment-involvement(SJI)theoryasapotentialtheoreticalframeworktoaugment,replace,and/or elaboratethedisconfirmationmodelandlatitudemodelsassociatedwithCS/DandSQresearch. Finally, Malhotra, Charles, and Uslay reviewed the literature focusing on the methodological perspectives,issues,andapplicationsrelatedtocorrespondenceanalysis.Theyconcludedwitha listofthecreativeapplicationsandthetechnique’slimitations.
REVIEW OF MARKETING RESEARCH xi
ArticlesintheSecondVolume Thesecondvolumecontinuedtheemphasisofthefirstbyfeaturingabroadrangeoftopicscontributedbysomeofthetopscholarsinthediscipline.Thediversearticlesinthesecondvolumecanall begroupedunderthebroadumbrellaofconsumeraction.Bagozzidevelopedadetailedframework forconsumeractionintermsofautomaticity,purposiveness,andself-regulation.MacInnis,Patrick,andParkprovidedareviewofaffectiveforecastingandmisforecasting.Ratchford,Lee,and TalukdarreviewedtheliteraturerelatedtouseoftheInternetasavehicleforinformationsearch. Theydevelopedandempiricallytestedageneralmodelofthechoiceofinformationsourceswith encouragingresults.Miller,Malhotra,andKingreviewedthecategorizationliteratureanddevelopedacategorization-basedmodeloftheproductevaluationformationprocess,whichassistsin thepredictionofsetmembership(i.e.,evoked,inert,orinept).LamandParasuramanproposed anintegratedframeworkthatincorporatedamorecomprehensivesetofvariousindividual-level determinantsoftechnologyadoptionandusage.Recently,marketinghascomeunderincreased pressuretojustifyitsbudgetsandactivities.Lehmanndevelopedametricsvaluechaintocapturethevariouslevelsofmeasurementemployedinthisrespect.Finally,Oakley,Iacobucci,and Duhachekprovidedanexpositionofhierarchicallinearmodeling(HLM). ArticlesintheThirdVolume BoltonandTarasidescribedhowcompaniescaneffectivelycultivatecustomerrelationshipsand developcustomerportfoliosthatincreaseshareholdervalue.Theyreviewedtheextensiveliterature oncustomerrelationshipmanagement(CRM),customerassetmanagement,andcustomerportfoliomanagement,andsummarizedkeyfindings.Theyexaminedfiveorganizationalprocesses necessaryforeffectiveCRM:makingstrategicchoicesthatfosterorganizationallearning;creating valueforcustomersandthefirm;managingsourcesofvalue;investingresourcesacrossfunctions, organizationalunits,andchannels;andgloballyoptimizingproductandcustomerportfolios. ChandrasekaranandTelliscriticallyreviewedresearchonthediffusionofnewproductsprimarilyinthemarketingliteratureandalsoineconomicsandgeography.Whileotherreviewson thistopicareavailable,theirreviewdiffersfromprioronesintwoimportantaspects.First,the priorreviewsfocusontheS-curveofcumulativesalesofanewproduct,mostlycoveringgrowth. ChandrasekaranandTellisfocusedonphenomenaotherthantheS-curve,suchastakeoffand slowdown.Second,whilethepreviousreviewsfocusmainlyontheBassmodel,Chandrasekaran andTellisalsoconsideredothermodelsofdiffusionanddriversofnewproductdiffusion. EckhardtandHoustonreviewed,compared,andcontrastedculturalandcross-culturalpsychologicalmethods.Theypresentedtheunderlyingconceptionsofculturethatunderpinbothstreams, and discussed various methods associated with each approach. They identified the consumer researchquestionsbestansweredusingeachapproachanddiscussedhoweachapproachinforms the other. Finally, they examined how consumer research can benefit from understanding the differencesinthetwoapproaches.Whileculturalandcross-culturalperspectivesadoptdistinct viewsaboutcultureandpsychologicalprocesses,itispossibletoviewthemascomplementary ratherthanincompatible.SeveralsuggestionsbyMalhotraandhiscolleaguescanbeusefulin thisrespect(Malhotra2001;Malhotra,Agarwal,andPeterson1996;MalhotraandCharles2002; MalhotraandMcCort2001;Malhotraetal.2005).Forexample,onecanstartwithaneticapproach andthenmakeemicmodificationstoadapttothelocalcultures.Alternatively,onecanstartwith anemicperspectiveandthenmakeeticadaptationstogetanunderstandingacrosscultures.This systematictheorybuildingandtestingprocessisillustratedbyKimandMalhotra(2005).
xii REVIEW OF MARKETING RESEARCH
GrewalandCompeausynthesizedresearchfromconsumerbehavior,psychology,andapplied economicstoaddresshowpriceasaninformationcueaffectsconsumers’responsesinthecontext ofotherinformationcues.Theydevelopedaconceptualframework,usingadaptation-leveltheory andtransactionutilitytheory,thatsynthesizedpriorresearchonprice,referenceprice,andother informationcuesandtheireffectsonconsumers’priceexpectations,evaluations,andbehavioral intentions.Theirconceptualmodelcontributestoourunderstandingofthewayimperfectinformation affects consumers’ decision processes, goes well beyond the original price–perceived qualityparadigm,andintegratesknowledgefromconsumerresearch,psychology,andapplied economics. Sayman and Raju provided a review of research on store brands.Their review focused on integratingresearchinkeyareasandidentifyingdirectionsforfutureresearch.Thereislimited theoreticalandempiricalresearchregardingoptimalcounterstrategiesofnationalbrandsagainst storebrands;studiestendtofocusononeaspect,andnationalbrandqualityistypicallyassumed tobeexogenous.Researchershave,byandlarge,focusedonme-too-typestorebrands.Future researchshouldconsiderpremiumstorebrandproductsaswell. Merunka and Peterson examined an intrapersonal aspect of language, namely, whether the structure of a language, per se, influences the thoughts of those who speak it.They reviewed empiricalresearchconductedoverthepasthalf-centuryontheeffectsoflanguagestructureon avarietyofmentalactivities.Theyfoundsupportfortheweakformofthelinguisticrelativity hypothesis,thenotionthatthestructureofalanguagedoesindeedinfluence(butnotdetermine) cognition.Theestimationofindependentandjointeffectsoflanguageisdifficultatbest.We needcomprehensivestudiesthatincorporatetheorderinwhichbilingualsacquiretheirrespectivelanguages,howtheyacquiretheirlanguages,andwhentheyacquiretheirlanguages.Future researchshouldalsocomparethepossibleinfluenceofasinglelanguageonmentalprocessing acrossdifferentcultures. Belkdiscussedtheimplicationsofgettingvisualforresearch,teaching,andcommunicating.He identifiedbasicopportunities,threats,andconsequencesofbecomingvisual.Severaltechniques forcollectingvisualdatawerediscussedintherealmofinterviewingaswellasobservation.We mightwellbeenteringaGoldenAgeofvisualandmultimediamarketingresearchandBelkhelps ustogetagoodhandleonit. ArticlesinThisVolume Consistentwiththefirstthreevolumes,thisfourthvolumealsofeaturesabroadarrayoftopics withcontributionsfromsomeofthetopscholarsinthefield.Thesearticlesfallunderthebroad umbrellaoftheconsumerandthefirm. LouviereandMeyerconsidertheliteratureonbehavioral,economic,andstatisticalapproaches tomodelingconsumerchoicebehavior.Theyfocusondescriptivemodelsofchoiceinevolving markets,whereconsumersarelikelytohavepoorlydevelopedpreferencesandbeinfluencedby beliefsaboutfuturemarketchanges.Theycallforabetterallianceamongbehavioral,economic, andstatisticalapproachestomodelingconsumerchoicebehavior.Economicandstatisticalmodelerscanconstructivelylearnfrombehavioralresearchers.Anunderstandingoftheactualprocess thatisdrivingpreferencescanprovidebetteraprioriinsightsintothemodelstructuresandbest descriptiveaccountofchoices.Theauthorspositthatprimitivepattern-matchingheuristics,which behavioralresearcherssuggestoftenunderliechoicesinnewmarkets,canmanifestthemselves incomplexfunctionalformsofalgebraicchoicemodels,andfailingtomodelthevarianceinthe observedcomponentsofutilitycanresultinmisleadingconclusionsabouttheactualamountof
REVIEW OF MARKETING RESEARCH xiii
heterogeneitythatexistsinamarket.Theyalsoillustratethebenefitsofareversedialogue,how economictheorycanleadbehavioralresearcherstomoreparsimoniousexplanationsforapparent anomaliesinchoicetaskswherepreferencesareuncertain. FolkesandMattaidentifyfactorsthatinfluencehowmuchanindividualconsumesonasingle usageoccasionbydrawingonresearchinconsumerbehavioraswellasallieddisciplines.They developanintegratedframeworktounderstandhow,andatwhatstage,variousfactorsaffectusage quantitybasedonGollwitzer’s(1996)“actiongoals”model.Initially,factorssuchasaproduct’s priceandsocialnormsinfluenceconsumption-relatedgoalsandtheirperceiveddesirabilityand feasibility.Inthenextphase,factorssuchasself-controlstrategiesandproductinstructionsinfluencetheimplementationofthegoal.Finally,theconsumer’smotivationtousefeedback,andthe typeoffeedbackaboutconsumption,hasaninfluenceonsubsequentgoalsetting.Theirframework canaidmarketersinformulatingproducts,designingpackaging,andcreatingmessages.Itcan alsohelppublicpolicymakersidentifyeffectivestrategiestopromotethewell-beingofconsumersandoftheenvironment. KumarandLuoalsoexamineconsumption,butfromamodelingperspective.Inordertoallocate scarcemarketingresourcesefficientlyandeffectively,itisimportantforafirmtoknowwhatto sell,whentosell,andtowhom.KumarandLuoreviewhowthepurchasetiming,brandchoice, andpurchasequantitydecisionshavebeenmodeledhistorically,aswellastheissueswithineach decisionthathavebeenaddressed.Avastmajorityofthesestudiesusescannerdataortransactiondata.Sincerecentresearchhasshownthatcommonmethodvariancemaynotbeaserious problem(Malhotra,Kim,andPatil2006),surveyscanalsobeausefulsourceofsuchdataand shouldbeincreasinglyemployed.Theyalsoexaminethedifferencesamongvariousapproaches anddescribethecommonmethodsthathavebeenusedtomodelatleasttwoofthethreedecisions. Finally,theydescribethemanagerialimplicationsofmodelingthesedecisionsandsuggestways toaddressthefuturechallenges. Despitetheinterestinglobalbranding,studiesinvolvingbrandextensionstrategiesinforeign marketsremainverylimited.Thefactthatsofewstudiesexistlimitsourunderstandingofeffective brandextensionstrategyinacross-culturalcontext.Merz,Alden,Hoyer,andDesaiproposeanew conceptualframeworkandseveralpropositionsregardingeffectiveglobalbrandextensionstrategy inacross-culturalcontext.Indoingso,theyfirstreviewmorecommonlyexaminedantecedent variablesof(national)brandextensionevaluation.Then,theyproposeadefinitionofcultureand subsequentlyreviewtheexistingcross-culturalbrandextensionresearch.Theyexaminewaysin whichculturemayaffectconsumers’brandextensionevaluationanddeveloppropositionsthat areinneedofempiricalvalidation.ThesepropositionsaredevelopedbydrawinguponHofstede’s culturaldimensionsandRoth’stheoryofsocioeconomicsandserveasacross-culturalbrandextensionconceptualframeworkforstimulatingresearchaboutbrandextensionsacrosscultures. Giventhegrowingimportanceofvisualmarketinginpractice,WedelandPietersrevieweyetrackingresearchinmarketingandevaluateitseffectiveness.Specifically,theyrevieweye-tracking applicationsinadvertising(print,TV,andbanner),healthandnutritionwarnings,branding,and choiceandshelfsearchbehaviors.Theyalsoprovideacasestudyoftheapplicationofeye-trackingto adpre-testing.Finally,theydiscussfindings,identifycurrentgapsinourknowledge,andprovide anoutlookonfutureresearch. SinghandSaatcioglureviewdifferentapproachesforexaminingroletheoryimplicationsfor boundaryspannerssuchassalespeople,frontline,andcustomercontactemployees.Theyfocus onuniversalisticandcontingencyapproachesanddeveloptheconfiguralapproachbyextending configurationaltheoryprinciplestoroletheory.Theireffortiswelcome,asneitherthecontingencynortheconfiguralapproachhasreceivedmuchattentioninthemarketingliterature.They
xiv REVIEW OF MARKETING RESEARCH
compareandcontrastdifferentapproachesandreviewliteraturethathasremainedlessaccessible tomarketingresearchers.Theydiscussunderlyingassumptionsandpressforcriticalassessment oftheirecologicalvalidity.Finally,theyidentifypromisingbutasyetunchartedapproaches. Johnconsiderspricecontractdesigntemplatesgoverningprocurementandmarketingofindustrialequipment.Hearguesthatpriceformatschoicesprecedetheselectionofapricelevel. Thesepriceformatsareanintegralaspectoftheinstitutionalarrangementdevisedtogovernan exchange.Johnreviewsinstitutions,thatis,rulesofinteractionthatgovernthebehaviorofactorsindealingwithotheractors,withafocusontheirpricingelements.Hedevelopsadesign protocol and illustrates it by applying it to (a) the choice of fixed versus cost-plus prices for procuringcomponentsfromasupplier,and(b)thechoiceofleasingversussellingpriceformats forindustrialequipment. Itishopedthatcollectivelythechaptersinthisvolumewillsubstantiallyaidoureffortstounderstand,model,andmakepredictionsaboutboththefirmandtheconsumerandprovidefertile areasforfutureresearch. References Gollwitzer,PeterM.1996.“TheVolitionalBenefitsofPlanning.”InThePsychologyofAction,ed.PeterM. GollwitzerandJohnA.Bargh.NewYork:GuilfordPress. Kim,Sung,andNareshK.Malhotra.2005.“ALongitudinalModelofContinuedISUse:AnIntegrative ViewofFourMechanismsUnderlyingPost-AdoptionPhenomena.”ManagementScience51(5)(May): 741–755. Malhotra,NareshK.2001.“Cross-CulturalMarketingResearchintheTwenty-FirstCentury.”International MarketingReview18(3):230–234. Malhotra,NareshK.,JamesAgarwal,andMarkPeterson.1996.“Cross-CulturalMarketingResearch:MethodologicalIssuesandGuidelines.”InternationalMarketingReview13(5):7–43. Malhotra,NareshK.,andBetsyCharles.2002.“OvercomingtheAttributePrespecificationBiasinInternational Marketing Research by Using Nonattribute Based CorrespondenceAnalysis.” International MarketingReview19(1):65–79. Malhotra,NareshK.,SungKim,andAshutoshPatil.2006.“CommonMethodVarianceinISResearch: A Comparison ofAlternativeApproaches and a Reanalysis of Past Research,” Management Science (December):1865–1883. Malhotra, Naresh K., and Daniel McCort. 2001. “A Cross-Cultural Comparison of Behavioral Intention Models:TheoreticalConsiderationandanEmpiricalInvestigation.”InternationalMarketingReview18 (3):235–269. Malhotra,NareshK.,FrancisUlgado,JamesAgarwal,G.Shainesh,andLanWu.2005.“Dimensionsof ServiceQualityinDevelopedandDevelopingEconomies:Multi-CountryCross-CulturalComparisons.” InternationalMarketingReview22(3):256–278.
CONTENTS,VOLUME1xV
CONTENTS,VOLUME1 ReviewofMarketingResearch NareshK.Malhotra
ix
1. AReappraisaloftheRoleofEmotioninConsumerBehavior:Traditionaland ContemporaryApproaches AllisonR.JohnsonandDavidW.Stewart
3
2. TheEyeoftheBeholder:BeautyasaConceptinEverydayDiscourseand theCollectivePhotographicEssay MorrisB.Holbrook
35
3. ConsumerInformationAcquisition:AReviewandanExtension LanXiaandKentB.Monroe
101
4. TheResource-AdvantageTheoryofCompetition:AReview ShelbyD.HuntandRobertM.Morgan
153
5. TowardanIntegratedModelofBusinessPerformance SundarG.BharadwajandRajanVaradarajan
207
6. Consumers’EvaluativeReferenceScalesandSocialJudgmentTheory: AReviewandExploratoryStudy StephenL.VargoandRobertF.Lusch
245
7. CorrespondenceAnalysis:MethodologicalPerspectives,Issues,andApplications NareshK.Malhotra,BetsyRushCharles,andCanUslay
285
AbouttheEditorandContributors Index
317 319
xv
CONTENTS,VOLUME2xVII
CONTENTS,VOLUME2 ReviewofMarketingResearch:SomeReflections NareshK.Malhotra
ix
1. ConsumerAction:Automaticity,Purposiveness,andSelf-Regulation RichardP.Bagozzi
3
2. LookingThroughtheCrystalBall:AffectiveForecastingand MisforecastinginConsumerBehavior DeborahJ.MacInnis,VanessaM.Patrick,andC.WhanPark
43
3. ConsumerUseoftheInternetinSearchforAutomobiles:Literature Review,aConceptualFramework,andanEmpiricalInvestigation BrianT.Ratchford,Myung-SooLee,andDebabrataTalukdar
81
4. Categorization:AReviewandanEmpiricalInvestigationofthe EvaluationFormationProcess GinaL.Miller,NareshK.Malhotra,andTraceyM.King
109
5. Individual-levelDeterminantsofConsumers’AdoptionandUsageof TechnologicalInnovations:APropositionalInventory ShunYinLamandA.Parasuraman
151
6. TheMetricsImperative:MakingMarketingMatter DonaldR.Lehmann
177
7. Multilevel,HierarchicalLinearModelsandMarketing:ThisIs NotYourAdviser’sOLSModel JamesL.Oakley,DawnIacobucci,andAdamDuhachek
203
AbouttheEditorandContributors Index
229 231
xvii
CONTENTSxIX
CONTENTS,VOLUME3 ReviewofMarketingResearch:ALookAhead NareshK.Malhotra Contents,Volume1 Contents,Volume2
ix xv xvii
1. ManagingCustomerRelationships RuthN.BoltonandCrinaO.Tarasi
3
2. ACriticalReviewofMarketingResearchonDiffusionofNewProducts DeepaChandrasekaranandGerardJ.Tellis
39
3. OntheDistinctionBetweenCulturalandCross-CulturalPsychological ApproachesandItsSignificanceforConsumerPsychology GianaM.EckhardtandMichaelJ.Houston
81
4. ConsumerResponsestoPriceandItsContextualInformationCues: ASynthesisofPastResearch,aConceptualFramework,andAvenuesfor FurtherResearch DhruvGrewalandLarryD.Compeau
109
5. StoreBrands:FromBacktotheFuture SerdarSaymanandJagmohanS.Raju
132
6. Language,Thought,andConsumerResearch DwightR.MerunkaandRobertA.Peterson
152
7. YouOughttoBeinPictures:EnvisioningMarketingResearch RussellW.Belk
193
AbouttheEditorandContributors
207
xix
Review of
Marketing Research
FORMAL CHOICE MODELS OF INFORMAL CHOICES 3
CHAPTER1
FORMALCHOICEMODELSOF INFORMALCHOICES WhatChoiceModelingResearchCan (andCan’t)LearnfromBehavioralTheory JORDANJ.LOUVIEREANDROBERTJ.MEYER
Abstract Inthispaperweillustratethebenefitsofforgingabetterallianceamongbehavioral,economic, andstatisticalapproachestomodelingconsumerchoicebehavior.Wefocusontheproblemsthat arisewhenbuildingdescriptivemodelsofchoiceinevolvingmarkets,whereconsumersarelikely tohavepoorlydevelopedpreferencesandbeinfluencedbybeliefsaboutfuturemarketchanges. Weillustratehowunderstandingtheactualprocessthatisdrivingpreferencescanprovideanalystswithbothbetteraprioriinsightsintothemodelstructuresthatarelikelytoprovidethebest descriptiveaccountofchoicesinsuchsettings,aswellashowstablethesestructuresarelikelyto beovertime.Weshow,forexample,thatanalogicalreasoningheuristics—acommonstrategyfor makingdecisionsunderpreferenceuncertainty—canproducechoicepatternsthatresemblethe outputofcomplexnonlinear,nonadditive,multi-attributeutilityrules.Likewise,becausenovice consumersarelikelytodisplaystrongindividualdifferencesinthevarianceofunobservedcomponentsofutility,methodsthatfailtorecognizesuchdifferenceswilltendtooverstatetheactual extentoftasteheterogeneitythatexistsinapopulation.Wealsoillustratethebenefitsofareverse dialogue,examininghoweconomictheorycanleadbehavioralresearcherstomoreparsimonious explanationsforapparentanomaliesinchoicetaskswherepreferencesareuncertain.Weshow, forexample,thatsomeadhocmodelsthathavebeenusedtostatisticallydescribethecompromise effectinchoicecanbededucedfromfirstprinciplesofrationalriskydecisionmaking. Introduction Choicemodelingresearchinmarketinghasevolvedthroughtheinterplayofthreedifferentapproachestothestudyofhumandecisionmaking.Oneapproachistheeconomicperspective,which seesconsumersasmakingchoicesinamannerthatisconsistentwithrandomutilitymaximization.Consumersareviewedashavingwell-developedpreferencefunctionsdefinedoverproduct attributes,andtheychoosethoseoptionswhoseattributesofferthemostattractivetradeoffseither atthetimeofchoiceorintheshortorlongrun(e.g.,McFadden1981).Asecondapproachis exemplifiedbybehavioralresearchersandpsychologists,whoarguethatactualchoiceprocesses 3
4 JORDAN J. LOUVIERE AND ROBERT J. MEYER
maybefarremovedfromtherationalmechanismsassumedbyeconomists.Thatis,totheextent thatpreferencesexistatall,theyarediscontinuousandimprecise,withchoicesbeingtheoutcome ofheuristicrulesthatareuniquelyconstructedinresponsetotheexternalappearanceofoptions inchoicesets(e.g.,Payne,Bettman,andJohnson1993).Athirdandfinalview,astatisticalapproachtomodelingchoices,hasgrownrapidlysincethelate1980s.Adherentsofthisapproach claimideologicalneutralityinthedebateoverpreferencesandprocesses.Thatis,choicesare simplyviewedasdata;anymodelofchoiceisfairgameaslongasitpassestestsofdescriptive andpredictivevalidityinagivencontext(e.g.,Abe1995;terHofstede,KimandWedel2002; Rossi,AllenbyandMcCulloch2005;KamakuraandWedel2004).Notsurprisingly,thestatistical paradigmislessconcernedwithwhetheranygivenmodelcanbededucedfromfirstprinciplesof utilitymaximizationorcognitivetheory. Althoughthethreeapproachesdifferphilosophicallyandtosomeextentmethodologically, intuitionsuggeststhatthattheymightusefullyconvergeovertimeasourunderstandingofchoice behaviorevolvesandprogresses.Yettheacademicrealityappearstobeverydifferent.Forexample, behavioralresearchershavetendedtofocusonlaboratorydemonstrationsoffailuresoftheassumptionsofstandardeconomicmodels(e.g.,contextinvariance)andhavegivenlimitedattentionto developingalternativemodelingparadigmsthatmightaccountforthesefailures(forexceptions, see,e.g.,Kivetz,Netzer,andSrinivasan2004;TverskyandSimonson1993).Adherentsofthe economicsviewofchoicemodeling,fortheirpart,havefrequentlybeendismissiveofbehavioral findings,arguingthatlabsettingsexaggeratethesizeoferrorsthatwouldbeobservedinreal marketsorthattheycanbecapturedthroughcomplexgeneralizationsofstandardmodels(e.g., Machina1982).Finally,statisticalmodelershavedonelittletoresolvetheoreticalgapsbetween thebehavioralandeconomiccamps.WhilethereismuchtobelearnedandgainedfromincorporatingstatisticaladvancesfromdiscretemultivariateandBayesianstatisticsinchoicemodeling (e.g.,RossiandAllenby2003),thereisalsomuchtobelostbyadoptingapurelystatisticalview ofwhatisinherentlyahumanbehavioralprocess. Thepurposeofthispaperistotakeasmallsteptowardfusingthesedifferentperspectivesin theanalysisofchoicedata.Wetakealimitedfirststepbyexploringonedimensionofthisfusion, namelywhatempiricaleconomicandstatisticalmodelerscanconstructivelylearnfrombehavioral researchers(andvisaversa)whenbuildingmodelsofconsumerchoiceinevolvingmarkets—a settingwhereconsumerpreferencesarelikelytobehighlyunstable.Forexample,weillustrate howprimitivepattern-matchingheuristics,whichbehavioralresearcherssuggestoftenunderlie choicesinnewmarkets,canmanifestthemselvesincomplexfunctionalformsofalgebraicchoice models,andhowfailingtomodelthevarianceintheobservedcomponentsofutilitycanlead analyststoreachmisleadingconclusionsabouttheactualamountofheterogeneitythatexistsina market.Inourfinaldiscussionwealsoillustratethebenefitsofareverseflowoflearning,showing howabetterunderstandingoftherationalbasesofchoiceunderuncertaintycansometimeslead behavioralresearcherstosimplerexplanationsforlaboratorychoiceanomalies,focusingonthe particularcaseofthecompromise-effect. ChoiceandMarketEvolution Letusbeginwithathoughtexperimentthatillustratesthetypesofmodelingchallengesthatwetry toaddressinthispaper.Considerasimplemarketwithone(monopoly)providerofagood,such asamonopolyproviderofcableorbroadbandservicesoranotherpublicutility.Inthismarket, consumersmustdecidewhetherornottochoosethegood.Atsomepointthegoodinquestionis launchedintothemarket,andweassumethatpriortolaunch,informationisavailableaboutthe
FORMAL CHOICE MODELS OF INFORMAL CHOICES 5
good,itsfeaturesandprice(s),andthelikelylaunchdate.Thus,priortolaunchsomeconsumers inthemarketareawarethattheservicewillbeprovided,andhavereasonablycompleteinformationaboutitsfeaturesandlikelyprices.Another,probablymuchlarger,proportionofconsumersis“vaguelyaware”thattheservicewillbeprovided,andhasincompleteinformationabout featuresandpossibleprices.Finally,athirdproportionisunawareofthegoodorthatitwillbe launched. Thismarketisthuscharacterizedbythreestylizedgroupsofconsumers,whocanbeviewed asbeingonacontinuumofawarenessandinformednessaboutthegood,ortheycanbeviewed asthreediscretesegments.Initially,themostawareandinformedarelikelytobeasmallminority;thevaguelyawareandinformed,whileprobablyalargerproportion,alsoarelikelyto beaminority;mostofthemarketismorelikelytobeunawareanduninformed.Thenthegood islaunched,andthingsbegintochange.Totheextentthatthegoodisofinteresttoconsumers andtheyarecapableofbuyingandconsumingit,whichallowsthemtoreceivetheassociated benefitsorproblemsolutionsthegoodprovides,weexpecttheproportionofconsumerswhoare awareandinformedtogrowovertime.Likewise,consumerswhoareunawareanduninformed willgraduallymoveintothevaguelyawareandinformedgroup,andinthiswaythemarketwill evolvefromthe“bottomup.” Amarketerwhowishestomodelthedecisionofwhetherconsumerschoosethegoodinthis marketwouldtypicallybeginwiththetoolsofrandomutilitytheory.EachconsumerninthepopulationwouldbeassumedtoassociatewiththenewserviceiautilityU nit = β n ' X it + ε nit ,where Xit,isavectorofthemeasuredattributesoftheservice(e.g.,price),βnisanassociatedparameter vectordescribingtheconsumer’stastesfortheseattributes,andεnitisanunobservedcomponent ofutility.Theunobservedcomponentεnitwouldtypicallybeassumedtofollowanindependently andidenticallydistributedextremevaluedistribution(overconsumers,choicealternatives,and servicecharacteristics).Ifthisassumptionissatisfied,theindividualchoiceprobabilitiescanbe representedbythewell-knownmultinomiallogitmodel
Pnit =
∑e
eβn ' Xit β n ' X kt
+ θ nt
(1)
k
whereθntistheconsumer’sutilityforunmeasuredoutsidegoods.Toextendexpression(1)tothe studyofpopulationormarketchoices,analyststypicallymakeassumptionsabouthowtastesβn varyoverthepopulation.Forexample,ifβncanbeassumedtohaveastationaryparametricdistribution,thenpopulationchoicecanbemodeledbyarandom-coefficientsormixedlogitmodel thatassumes U nit = β ' Xit + ηnit + ε nit,where ηnit = (β int − β)' Xit isarandomvariablethatcaptures unobservedindividualdeparturesfromacommonstrictutilityβ’X(e.g.,HensherandGreene 2003). Itshouldbeclearthatwhiletheaboveapproachmightprovideagoodstatisticaldescription oftheassociationthatexistsbetweenchoicesandserviceattributesataparticularpointintime (oroveraseriesofpointsintime)foraparticularsampleofpeopleduringthecourseofmarket evolution,itcapturesfewofthebehavioralfeaturesofservice-choicedynamicsmentionedabove. Forexample,themodel(asformulated)doesnotcharacterizehowparameterheterogeneitymight beassociatedwithfactorsthatunderliedifferentiallevelsofawarenessandinformationpossessed byconsumers,theprovider’sdecisionsaboutcommunicationsandaccess,andbeliefsheldby consumersaboutthemarket’sfuture(e.g.,thepossibilityofnewentrantsorexpectationsabout howthetechnologywillevolve).Whileanalystsmayacknowledgethattheseassociationsarelikely
6 JORDAN J. LOUVIERE AND ROBERT J. MEYER
tochangeasamarketevolves,exactlyhowthechangeswilloccurorwhattheirtrajectorieswill typicallybeliesoutsidethepurviewoftheanalysis.Thus,itisfairtosaythattheoverwhelming majorityofthesetypesofmodelsarepurelydescriptivewithlittlerealexplanatorycapability. Inthesectionsbelowwetrytoillustratemorepreciselyhowrealbehavioralprocessesunderlyingchoicesinmarketscanmanifestthemselvesinthedatathatareusedtoestimatereduced-form statistical models, and how overlooking behavioral processes may lead analysts to erroneous conclusionsaboutboththenatureofpreferencesinmarketsandhowmarketswillevolveover time.Morespecifically,weexplorewhatbehavioraltheorywouldpredictabouttheempiricalappearanceandstabilityoftypicallyestimatedrandomutilitymodelswhentheyareusedtodescribe thechoicebehaviorofconsumerswho: 1. Havehighlevelsofuncertaintyabouttheirpreferencesforgoodsinamarket(bothattributevaluationsandweights) 2. Useheuristicshort-cutsthatdonotutilizealltheproduct-attributeinformationavailable tothematthetimeofchoice 3. Havestrategicforesight—thatis,considerhowthecurrentchoicewillaffecttheutility gainedfromfuturechoices ModelingChoicebyNaïveConsumers Oneaspectofrandomutilitytheorythatelicitsfewquarrelsistheassumptionthatconsumersare guidedbyadesiretochoosetheoptionthatwillgivethemthemostutilityorpleasure.Butfor manyconsumers,particularlythoseinnewlyevolvingmarkets,theabilitytoachievethismaximizationgoalisinhibitedbythesimplefactthatpreferencesareuncertain.Forexample,ifanovice consumerwereforcedtodecidewhetheritwasworth$10amonthtoadoptabroadbandservice thatwouldincreasedownloadspeedsby100kilobytes,theaxiomsofutilitytheorywouldnothelp hermuchtomakethisdecision.Inordertomakeautility-maximizingdecision,shewouldneed toknowwhatakilobyteis,theamountofadditionalpleasurethatshecouldexpectfroma100kb increase,andhowtoexchangetheextrapleasurefordollars—knowledgefewnoviceconsumers arelikelytohave.Thechoicesweobserveinnewmarkets,therefore,reflectanambiguousmixof enduringpreferencesandtheheuristicsconsumersusetoovercomethelackofpreferences. Whataretheheuristicsconsumersusetoovercomealackofattributepreferenceknowledge? Theconsensusviewisthatnaïvechoicesareoftenmadeusinganalogicalreasoning.Thatis,when anewproductisencountered,consumersjudgeitbyrecallingproductsthattheyconsumedinthe pastwithsimilarattributes(e.g.,Bradlow,Hu,andHo2004;Gregan-PaxtonandRoedderJohn 1997;Norman1988).Forexample,thissortofpattern-matchingprocessisthoughttoexplainhow peoplewithlittleskillinmathematicscanlearntoplaycomplexequilibriaingames.Insteadof encodingandsolvingoptimalstrategies,mostgamesallowplayerstodiscoverequilibriasimply bybeingwillingtorepeatthemovesthatyieldedthehighestpayoffsinthepast(e.g.,Camererand Ho1999;FudenbergandLevine1998).Similarly,theexpertiseofwineconnoisseursprobably lieslessintheirskillsatusingalgebraicrulestopredictqualityandmoreintheirpossessionofa richmemorybankofpastreferentexamplesthatformthebasisofevaluations(knownassmell andtastememory). Thepervasiveuseofpattern-matchingheuristicsinnovelproductjudgmentswasillustrated inworkbyMeyer(1987),whoexaminedtheprocessbywhichconsumerslearntomakemulti- attributejudgmentsinanovelproductcategory.Inhisexperiments,participantswereshownaseries ofproductprofilesdescribedbyseveralunfamiliarattributesandlevels(copperalloysgenerated
FORMAL CHOICE MODELS OF INFORMAL CHOICES 7
bydifferentproductionmethods),andtheywereaskedtopredicteachproduct’slikelyquality (strength).Oncethesubjectsmadeaprediction,theyreceivedfeedbackaboutthe“true”quality ofanoption.Consistentwithbehaviorspreviouslyobservedintaskslikethis(e.g.,Mellers1981), after several rounds of feedback participants became quite good at making forecasts.That is, theyacted“asif ”theyhadlearnedthemulti-attributerulethatdeterminedtruequality,andwere usingittomakeforecasts.Yet,thesurprisingempiricaloutcomeassociatedwiththisexperiment wasthatinthecourseofmakingjudgments,asubsetofrespondentsaskedtoprovideconcurrent verbal protocols gave no indication that they actually made judgments using a multi-attribute rule.Instead,theyappearedtomaketheirjudgmentsusingapattern-matchingprocesswhereby theymadeforecastsabouthowsimilaranygivennewprofilewastoapreviouslyseenprofilewith knownqualities.Theirjudgmentsimprovedovertimenotbecausetheyweredevelopingbetter knowledgeofarule,butbecausethedatabaseofreferentexamplesimproved,andthisdatabase allowedthemtoeffectivelymimictheoutcomesofsucharule. Shouldchoicemodelersbeworriedbythisresult?Notnecessarily;inthesamewaythatagame theoristwouldbeindifferenttowhetherpeopleplayequilibriabecausetheyactuallycalculatethe optimalstrategyorstumbletheirwaytherebytrialanderror(FudenbergandLevine1998),one presumesthatrandomutilitytheoristswouldbehappytoviewmulti-attributeutilitytheoryasan “asif ”modelofthewayinwhichpeopleevaluateoptions.Ifpattern-matchingheuristicsproduce judgmentdatathatarewell-approximatedbystablelinear-additivemodels,thenoneclearlycan build a paradigm around this; mathematical convenience in this case would trump behavioral realism.Butistheresuchanisomorphism,andhowwidespreadmightitbe? It is easy to show that a mathematical equivalence exists between pattern-matching and linear-additiverules,butonlyunderalimitingconditionofproduct-classexperience:whenthe underlying(ortrue)rewardstructureislinear-additiveinattributesanddecisionmakershave had direct experience with all of the product profiles under study in a given multi-attribute space.Incontrast,themorelimitedaperson’sjudgmentalexperienceinanygivencontext,the lesslinear-additivitywilldescribetheirjudgments,eveniftheunderlyingrewardfunctionis linear-additive. Todemonstrate,considerthefollowingpattern-matchingmodelofexpectedvaluations: (2) EVi = δ iz (Vz ) + (1 − δ iz ) p whereEViistheanticipatedutilityofsomemulti-attributeprofilei,δizisazeroto1boundedmeasure ofthesubjectivesimilaritybetweenprofileiandtheexperiencedprofilezthatismostsimilarto i,Vzistheexperiencedutilityofprofilez,andpisajudgmentalprior.1Suppose,alsoforthesake ofsimplicity,thatexperiencedutilityisgivenbythelinear-additivemulti-attributemodel Xk (3) Vk = β’ whereXkisavectorofmeasuresofthevaluesattributesofprofilekthatislinearinV,andβisan associatedweightvector. Itshouldbeobviousthatexpression(2)isequivalentto(3)whenδizequals1forallcomparisons i,z—thatis,wheneveryprofilehasbeenexperiencedbytheconsumer.Inthiscaseittrulydoesn’t matterwhetherconsumersliterallycalculateutilitiestoformprojectivejudgments(expression2) ormerelyactasiftheydo(expression1).Butwhataboutamorerealisticcasewhereconsumers havelimitedexperiencesinamarket—thatis,foranygivenpossiblecompletefactorialarrayof profiles,whathappensifonlyasmallsubsethavevaluations?Insuchcases(2)and(3)willnot
8 JORDAN J. LOUVIERE AND ROBERT J. MEYER
beequivalent;instead,consumerswillrevealan“asif ”multi-attributejudgmentrulethatdeparts fromtheasymptotic(orfull-information)rule(3)inpredictableways.Specifically,ifweregress projectedpreferences(EVi)againsttheattributesofeachoption,wedonotrecoveratrueorasymptoticutilityfunction(2),butinsteadamulti-attributefunctionwhoseformisdistortedbythe similarityofeachitotheactuallyexperiencedprofilesz,theirtruevaluations,andthejudgmental prior—thatis,EVi = δ iz( β’ X z ) + (1 − δ iz ) p . Toillustratethepropertiesofsuchanapproximation,considerthecaseofasimplejudgment contextinwhichconsumersevaluatetheattractivenessofeachofanumberoftwo-attributeoptionsonasubjectivescale.Eachoptionisdescribedbyapairoflevelsthatrepresentacombinationoftwosix-levelattributes.Respondentsjudgeallthirty-sixcombinationsrepresentedbythe factorialarray.Asabove,forsimplicityweassumethatthetrueutilitythatwouldbeobserved byaconsumerconsumingeachoftheprofilesisgivenbytheadditiveruleVi=xil+xi2.Upon enteringthetask,however,aconsumerhasonlylimitedexperienceinthecategory,andhas directlyexperiencedonlyasmallsubsetofthethirty-sixprofiles.Weconsidertheimplications ofapproximatingapattern-matchingjudgmentrule(expression1)withalinearmodelintwo illustrativecases:(1)theconsumer’spreviousexperiencescorrespondtothetwoextremesof theattributespace(the1,1and6,6profiles,respectively),and(2)theconsumerhasexperienced threemid-valuedoptions:the2,2,3,3,and4,4options.Togenerateanumericexampleweused anormalizedsimilaritymetric
δ iz =
[1 − ( ∑ z | xiz −− x kz | / ∑ z | xmax(z ) −− x |] MAX min( z ) k
λ
whereλ=4,andassumedaprior(p)equaltothe1–12responsescalemean(6.5). InFigures1.1athrough1.1cweplottheresultingtwo-factorinteractiongraphsforeachof
Figure1.1a TrueX1*X2Interaction
FORMAL CHOICE MODELS OF INFORMAL CHOICES 9 Figure1.1b X1*X2Interactionwhenjudgmentsaremadebyapattern-matchingrulewith referentsat1,1and6,6
Figure1.1c X1*X2Interactionwhenjudgmentsaremadebyapattern-matchingrulewith referentsat2,2,3,3,and4,4
10 JORDAN J. LOUVIERE AND ROBERT J. MEYER
thesetwocases(1.1band1.1c)tobecontrastedwiththenormativeinteraction(1.1a).Figures1.1b and1.1cprovidegoodandbadnewsabouttheabilityoflinearmodelstomimicpattern-matching judgmentprocesses.Thegoodnewsisthattheyshowthataslongasaconsumer’spreviousexperiencesarewellchosen(inthiscaseattheextremesofutilitycontinua),andunderanappropriate (here, neutral) prior, simple linear models can do a good job of describing contemporaneous preferencesandyieldparameterestimatesthatareasymptoticallystable.2Specifically,inFigure 1.1bweseethatjudgmentsgeneratedbyapattern-matchingprocessthatisinformedbytheutility extremeswillforecastthetruevaluationsofunfamiliarintermediateprofileswell,andshouldbe welldescribedbyalinear-additivemodel. Thebadnewsisthatifexperiences(and/orpriors)arenotwellchosen,thevalueoflinear-models willbegreatlyreducedinstabilityanddescriptivevalidity.AsshowninFigure1.1c,whenreferentexperienceslieintheinterioroftheattributespace(whichmaybemoretypicalinpractice), therevealedpreferencesurfacebecomesnonlinear,displayinganinteractionatthemoredistant (relativetoexperience)tail.Thisimpliesthatnotonlywouldasimplelinearmodeldoapoorjob ofcapturingcontemporaneouspreferences(oneneedsadatadesignthatcanestimatenonlinear effectsandinteractions),butitalsowouldpoorlyforecastthepreferencestructureobservedat futuretimeswhentheconsumer’sscopeofexperienceinthecategoryexpands. Thislatterresultholdstwoimplicationsforappliedchoiceanalysts.Thefirstisthatlinearadditivemodelswilloftenbeillsuitedfordescribingtheassociationthatexistsbetweenproduct attributesandproductchoicesfornoviceconsumers.Differentialknowledgeofutilityoverany multi-attributespacewillproducenonlinearitiesand/orinteractionsthatsuchmodelswillfailto capture.Butcaremustbetakeninrecallingthattheseinteractionswouldnotbemanifestations ofenduringconditionalpreferences(forexample,asustainedincreasedsensitivitytoservice variationsgivenhigherpaidprices),butratherthetransienteffectsoflimitedknowledgeof preferencesovertheattributespace.Assuch,theywouldbeexpectedtodisplaylittletemporal stability, perhaps vanishing completely as consumers become fully knowledgeable about a productcategory. TheEffectofUnspecifiedVariabilityintheUnobservedComponentsofUtility Itmightbeargued,ofcourse,thatbecausenoviceconsumersarelikelytobeheterogeneousin thekindsofproductexperiencestheyhavehad,individual-leveldeparturesfromlinear-additivity duetouseofnaïvepattern-matchingrulescouldwellwashoutwhenmarketsareviewedinthe aggregate.Thatis,onemightbeabletoproceedwithtraditionallinear-additivechoicemodels undertheassumptionthattransientindividualdifferencesinfunctionalformwouldbecaptured bythevarianceoftheunobservedcomponentofutilityinatraditionallinear-additivemodel.If theeffectsofsubsequentlearningsimilarlycancelthemselvesoutacrossthepopulation,thenthe coefficientsofthelinear-additivemodel—thoughperhapswrongattheindividuallevel—would providegoodaggregatelong-termforecastsofpreferences. Istheproblemsolvedthateasily?Theanswer,unfortunately,is“no,”fortworeasons.First, becausechoicemodelestimatesareperfectlyconfoundedwiththevarianceoftheunobserved componentofutility(see,e.g.,SwaitandLouviere1993),changesinconsumerexperience thatalterthestructureoftheunobservedcomponentovertimewouldalsoinducetemporal changes in these estimates.Without a theory of what is driving the error terms, the exact natureofthesetemporalchangeswouldbeimpossibletopredict.Forexample,considera choiceanalysisthatrevealsnoviceconsumerstobestatisticallyinsensitivetovariationsin servicequality.Theconfoundofvalue(coefficient)andvarianceimpliesthatthemeaning
FORMAL CHOICE MODELS OF INFORMAL CHOICES 11 Table1.1
ConsumerPreferenceandVariabilityTypes VariabilityXPreference LowPrecision HighPrecision
Lowpricesensitivity; highqualitysensitivity 1.Vi=1.5Qi–0.5Pi
scale=0.6
3.Vi=1.5Qi–0.5Pi
scale=1.9
Highpricesensitivity; lowqualitysensitivity 2.Vi=0.5Qi–1.5Pi
scale=0.6
4.Vi=0.5Qi–1.5Pi
scale=1.9
Table1.2
EstimationRealizationfromTable1.1
VariabilityXPreference
Lowpricesensitivity; highqualitysensitivity
Highpricesensitivity; lowqualitysensitivity
LowVariability HighVariability
1.Vi=0.9Qi –0.3Pi 3.Vi=2.85Qi –0.9Pi
2.Vi=0.3Qi –0.9Pi 4.Vi=0.9Qi –2.85Pi
ofthisresultisfundamentallyambiguous:onecouldneverknowforsurewhetherconsumersaretrulyindifferenttoserviceorwhetherthetrueeffectofserviceisbeingtemporarily maskedbytheaggregateeffectofconsumersusingaheterogeneousmixofpattern-matching heuristics. If the latter is the case, parameters estimated now would be of little value for long-termplanningpurposes. Asecond,moresubtle,problemisthatifapopulationisheterogeneousinitscategoryknowledge,thevarianceoftheunobservedutilitycomponentshouldalsonotbeconstantacrossasample atanygivenpointintime.Hence,inferencesaboutpreferenceheterogeneityderivedfrommodel parameterswillbeconfoundedwithknowledgeheterogeneity,orvariationsinthestandarddeviationoftheunobservedcomponentofutilityacrossconsumers. Forexample,consideracaseinwhichapopulationischaracterizedbyamixtureofexperienced consumerswhoreliablychooseproductsbasedonagivensetofattributes,andlessexperiencedconsumerswhosechoicesarelessreliablylinkedtoattributes(e.g.,theymakejudgmentsbyreferringto oneortwoproductswithwhichtheyhavehaddirectpreviousexperience).Inbothgroups,theconsumersdifferintheirtruesensitivitytoprice(thatis,thesensitivitytopricethatwouldbeobservedifone controlsforallunobservedinfluencesonchoice).So,imaginethattherearefourtypesofconsumers asshowninTable1.1:(1)lowchoicevariabilitycombinedwithlowsensitivitytopriceand highsensitivitytoquality;(2)lowchoicevariabilitycombinedwithhighsensitivitytopriceand lowsensitivitytoquality;(3)highchoicevariabilitycombinedwithlowsensitivitytopriceandhigh sensitivitytoquality;and(4)highchoicevariabilitycombinedwithhighsensitivitytopriceand lowsensitivitytoquality.Inthistable,the“scale”correspondstotheinversevariance(orprecision)of theunobservedcomponentofutility. Now,considerparametersestimatedfromthesefourconsumertypes,asshowninTable1.2. Itshouldbenotedthat“scale”multipliesthesystematicutilitycomponent,sotheparametersin Table1.1are“true”parameters.Notethatifweknowthetrueparameters,wecanconcludethat consumertypes1and3areidenticalandtypes2and4areidentical,exceptforchoicevariability.
12 JORDAN J. LOUVIERE AND ROBERT J. MEYER
Yet,withthescaleconfound,wewouldconcludethatnotwoofthefourconsumertypesarealike, althoughwemightincorrectlyconcludethat1and4sharesensitivitytoquality,while2and3share sensitivitytoprice.Asimilarresultwouldobtainifinsteadofdiscreteconsumertypes/classesone hadcontinuousdistributionswiththetrueparametersasmeans. AreVarianceEffectsonModelParametersReal? Anobviousobjectiontotheabovediscussionandstylizedexampleisthattheyonlyestablish thepossibilitythatanalysesofpreferenceheterogeneitybasedonstandardmethodsmaybemisleading.Inturn,thisbegsthequestionsastothedegreetowhichindividualsindeeddifferinthe varianceoftheunobservedutilitycomponent,andthedegreetowhichpreferenceparametersare confoundedbythisvariability. Anillustrationofthemagnitudeandnatureofvarianceeffectswasrecentlyprovidedby LouviereandEagle(2006).Theyreporttheresultsof66choiceexperimentswherebychoice modelswereestimatedforsingleindividuals,whichallowsonetoestimatethesizeofthevarianceoftheunobservedcomponentofutilityinchoice.Toprovideaflavoroftheseanalyses, inAppendix1.1wereporttheresultsof21individual-levelmodelestimatesfrom2ofthe66 experiments,reflectingchoicesamonghypotheticalpizzaproductsandcross-countryflights. Thesetwocontextsarereportedsimplyforconvenienceandbecausetheyhavefewerparameters thanothercontexts.Thetwoexperimentsusedacommonunderlyingoptimallyefficientdesign toestimatethemaineffectsofa23x43factorialbasedonthreeoptionsperset.Participants weremembersofanAustralianopt-inonlinepanel;completionratesforbothconditionswere over80percent. Tables1A.1and1A.2intheappendixdisplaytheindividual-levelmultinomiallogitmodel estimatesforthesubjectswhoparticipatedintheexperiments,andTables1A.3and1A.4give summarystatisticsfortheexperiments.Theindividual-levelmodelestimatesallowonetocalculate residualsfrommodelpredictions,inturnallowingonetoregressdesignmatrixcodesonresiduals. Thisauxiliaryregressionallowsonetodetermineif(a)significantunobservedvariabilityremains afterMNLestimation,and(b)theremainingunobservedvariabilityissystematic(i.e.,residuals systematicallyrelatedtodesignelements).Bothconditionsproducedsimilarresults,with18ofthe 21individualsinthepizzaconditionand17intheflightsconditionexhibitingregressionresults thataresignificantata90percentlevel.Thus,thevastmajorityofindividualsinbothconditions havesignificantremainingunobservedvariabilitythatissystematicallyrelatedtodesignattribute levels.Thus,itisunlikelythattheindividualssatisfiedconstanterrorvarianceassumptions. InFigures1.2aand1.2bwegraphtheindividuals’meansquaredmodelresidualsagainsttheir airfareandpriceutilityestimates,forflightsandpizzas,respectively.Bothgraphsareconsistent withrandomutilitytheory,whichpredictsthataserrorvariancesincrease(measuredbymean squareresiduals),modelparameterestimatesshouldconvergetozero.Bothgraphsdisplaythis result,allowingoneto“see”thatthemagnitudesoftheairfareandpriceestimatesareafunction ofthevariabilityineachindividual’schoices.Thus,inthecaseofairfaresandpizzaprices,alarge proportionofparameterdifferencesbetweenindividualscanbeexplainedsimplybydifferencesin individuals’choicevariability.Thus,modelestimatesoftheseeffectsaresignificantlyconfounded withindividualdifferencesinvariability. Morespecificdetailsofthebreakdownoftheseeffectsareasfollows: 1. Summary of variance explained by choice variability (MSR) between individuals (flights)
FORMAL CHOICE MODELS OF INFORMAL CHOICES 13 Figure1.2a MSRvs.FareEstimates:Flights
Figure1.2b MSRvs.MNLPriceEstimates:Pizzas
a.Traveltimes—6.8% b.Airfares—22.1% c.Airlinebrands—21.1% d.Frequentflyerprogram—0.4% e.Numberofstops—24.5% f.Freedrinks—9.1% g.Alternative-SpecificConstants(ASCs)—20.5% h.Averageacrossallindividualestimates—16.2% 2. Summaryofvarianceexplainedbychoicevariability(MSR)betweenindividuals(pizzas) a.Pizzachain—1.2% b.Pizzaprices—21.2% c.Numberoftoppings—23.6%
14 JORDAN J. LOUVIERE AND ROBERT J. MEYER
d.Freebread—0.0% e.Freedrinks—8.6% f.Freedessert—12.9% g.ASCs—19.3% h.Averageacrossallindividualestimates—13.6% Theimportanceoftheseresultsisthattheydemonstratethedangerofinterpretingempirical varianceinchoicemodelparametersasuniquelyreflectiveofeitherpreferencesorpreference heterogeneity.Specifically,iftrueparametersareconfoundedwitherrorvariances,choicemodels canforecastfuturechoicebehaviorwellonlyifthevariancesaretemporallystationaryand/or theydonotco-varywithotherfactorsthatwereconstantinthedatasourceusedforestimation, butarenotconstantattimesorplacesorinsegmentstobepredicted.Asnotedabove,thereis goodreasontosuspectthattheyoftenwillnotbeconstant,particularlyinthecaseofdeveloping markets,butalsoinmanycaseswherenoattemptsaremadetounderstandotherpossiblesources ofvariabilityinchoices(atypicalcaseinchoicemodeling). DealingwithComplexChoicesinMatureMarkets:Short-CutHeuristicsand TheirRepresentation Theabovediscussionfocusedonproblemsmodelingchoicesinearlystagesofmarketevolution whenthesetofmarketalternativesissmall,buttheuncertaintyinhowtoevaluatethesealternatives ishigh.Asmarketsapproachmaturity,however,theoppositeproblemoccurs;whiletheremaybe littleuncertaintyinforecastingtheutilitythatconsumerswillderivefromindividualofferings, theirchoicesmaywellbemoredifficultiftherearealargenumberofdifferentiatedofferings.In thissectionwetakeupwhatweknowabouthowinformedconsumersmakechoicesfromcomplex sets,andtheimplicationsofthisforstandardchoiceanalyses. Apervasivefindingofworkthathasstudiedprocessesassociatedwithcomplexchoicesisthat decisionsareoftenguidedbynoncompensatoryscreeningrulesthatacteithertoproduceunique choiceoutcomesortosequentiallyreducechoicesetstocognitivelymanageablesizes(e.g.,Payne, Bettman,andJohnson1993).Forexample,individualsmayeliminatealternativesifthey(a)fail onacriticalproductattribute(aconjunctiveeliminationrule),(b)failtoofferatleastonedistinctivebenefitacrossattributes(adisjunctiveeliminationrule),or(c)areunattractivebyvirtueof theirrank-orderpositioninaset(arankeliminationrule;e.g.,Einhorn1970).Intheory,theuse ofsuchdecisionrulescouldbeproblematicforrandomutilitymodelsbecausetheyimplythat indirectutilityfunctionsarenotstrictlylinearandadditive.Instead,optionsareevaluatedusing noncompensatoryrulesthatmakeonlylimiteduseofattributeinformation. Asweshowedearlier,undersomeconditionslinearmodelscanprovideareasonablyclose first approximation to decisions made by pattern-matching rules. Fortunately, they also can provideareasonablyclosefirstapproximationtononcompensatoryscreeningorelimination rules—aslongasoneiswillingtoconsidernonadditiveformsthatadmitinteractionsamong cues(see,e.g.,Einhorn1970).Forexample,consideraconsumerwhomakesaseriesofbinary judgmentsaboutwhethereachofseveralproductsdescribedbytwoattributes,X1,andX2,is acceptableornot.Theconsumermakesherjudgmentsusinganoncompensatoryconjunctive screeningrule,asfollows:
Acceptableif[X1>α]and[X2>δ].
(4)
FORMAL CHOICE MODELS OF INFORMAL CHOICES 15
Itiseasytoshowthatacontinuousalgebraicanalogto(4)existsaslongasthereisimprecision intheattributethresholdsαandδthatdrivejudgmentsforindividualswithinagivendataset.If αandδactasrandomvariables,theprobabilitythataproductprofile(Xi1,Xi2)willbejudgedas acceptableisgivenbytheproductofthemarginalprobabilitiesthattherealized(andunobserved) valuesofαandδarelessthanXi1andXi2atthemomentofchoice;thatis,
(5)
Pr(i|Xi1,Xi2)=Pr(Xi1>α)Pr(Xi2>δ).
Thecontinuous-functionalanalogof(5)thatfollowsdependsontheassumedfunctionalform ofthedistributionoferrorsassociatedwiththeacceptancethresholdsαandδ.Forexample,if thecumulativedensitiesoftheerrorsassociatedwithαandδcanbeapproximatedbyalinear probabilitymodeloftheformPr(Xi1>α)=αo+α1Xi1andPr(Xi1>δ)=bb+b1Xi2thendatageneratedbyaconjunctiveprocessforgeneratingacceptabilityjudgmentswouldcorrespondtothe bilinearutilityfunction
Pr(i|Xi1,Xi2) = ( a0 + a1 X i1 )(b0 + b1 Xi 2 ) .
(6)
= k0 + k1 Xi1 + k2 Xi 2 + k3 Xi1 Xi 2
(e.g.,KeeneyandRaiffa1976).Expression(6)containsanimportantimplication,namelythatif theerrorsareassociatedwiththresholdsandthedistributionoftheseerrorsisapproximatelylinear, conjunctivescreeningprocessesaremathematicallyequivalenttoalinearprobabilitymodelthat recognizeslinear-by-linearinteractionsbetweenattributes.Byextension,thefindingoffan-like (linear-by-linear)interactionamongapairofattributesinamulti-attributejudgmentexperiment (e.g.,afull-factorialconjointdesign)haslongbeenseenassuggestingthelikelyuseofnoisy screeningrulesinjudgmentbydecisionmakers(e.g.,Louviere1988). Notethatwecanextendthisideatoanyarbitrarysetofconjunctiveordisjunctivescreeningrules anderrordistributions.Specifically,forasetofindependentacceptabilityjudgmentsgeneratedby ageneralfamilyofstochasticscreeningrulesoftheform(X>,low high>low
high>low high>low
high>low high>low
high>low
high>low
high>low high>low,highlow
Uncertaintyavoidance (highvs.low)
high>low*
Powerdistance (highvs.low)
indiv.collect.
indiv.collect.
modern>lessmodern modernlessmodern modern>lessmodern
modern>lessmodern
modern>lessmodern
Individualismvs.collectivism Socioeconomics (individualismvs. collectivism) (modernvs.lessmodern)
Culturaldimensionsandsocioeconomics
Note:PB=parentbrand;BE=brandextension. *High>lowindicatesthattheperceivedsimilaritybetweenthePBandtheBEconstitutesamoreimportantdeterminantofBEevaluationincultureshigh inpowerdistancethanincultureslowinpowerdistance.
Socialrisk Financialrisk
Brandextension’sproductcategory characteristics:
PBcharacteristics: PerceivedqualityofthePB Numberofproductcategories affiliatedwithabrand
RelationshipbetweenthePB andtheBE: Perceivedsimilaritybetween thePBandtheBE Brandconceptconsistency betweenthePBandtheBE
Determinantcategorywith exemplardeterminantsof BEevaluation
TheCross-CulturalBrandExtensionResearchAgenda
Table4.1
BRAND EXTENSION RESEARCH 117
118
M.A. MERZ, D.L. ALDEN, W.D. HOYER, AND K.K DESAI
index=104),Greece(uncertaintyavoidanceindex=112),ortheUnitedStates(individualism =91). Second,managersneedtocarefullyexaminethetargetmarketintermsofthepreferredbrand concept(e.g.,symbolicversusfunctional)andbrandconceptinconsistencytolerance(e.g.,symbolicparentbrandandextensionversussymbolicparentbrandbutfunctionalextension).Roth (1995)hasfoundthatabrand’sconceptaffectsitsmarketshareindifferentculturalmarkets.The samecanbehypothesizedtobethecaseforbrandextensions.Moreover,managersneedtopay particularattentioniftheydecidetopositiontheirnewbrandextensionwithabrandconceptthat is different from the parent brand’s concept. Some cultures—just as is the case for perceived similaritybetweentheparentanditsextension—aremorelikelytobetolerantintermsofbrand conceptinconsistencies(e.g.,Austria,Singapore,andEcuador)thanothers(e.g.,Malaysia,Greece, andtheUnitedStates,asperthescoresnotedabove). However,itisunclearhowthesedifferentdimensionsinteractwitheachother.Forexample,a culturemayberelativelylowinpowerdistance,yetrelativelyhighinindividualism,uncertainty avoidance,andsocioeconomics.(e.g.,Germany).Similarly,aculturemayberelativelylowin powerdistanceanduncertaintyavoidance,yetrelativelyhighinindividualismandsocioeconomics(e.g.,theUnitedStates).Insuchinstances,itisnotclearwhetherafar,orincongruent,brand extensionstrategywillbeevaluatedmorefavorablyeventhoughsomeoftheculturaldimensions wouldpredictso.Inasimilarvein,itisnotclearwhatweighttheseculturaldimensionswould takeon,iftheypredictedopposingoutcomes(e.g.,theindividualismanduncertaintyavoidance dimensionspredictthatcongruentbrandextensionswillbesuccessful,whilethepowerdistance andsocioeconomicsdimensionspredictthatincongruentbrandextensionswillbesuccessful). Itispossiblethatallofthesedimensionswilltakeonequalweightintermsoftheirimmediate effectonbrandextensionevaluation,orthatsomeculturaldimensionswillhaverelativelymore weightthanothers.Asaresult,futurecross-culturalbrandextensionresearchshouldalsoexamine theinterplaybetweenthedifferentculturaldimensionsandtheirindividualaswellasaggregated effectsonbrandextensionevaluation.Moreover,itseemslikelythatthedifferentdeterminants havevariable“importanceweights”andthattheseweightsarenotthesamefromculturetoculture andindustrytoindustry.Futureresearch,therefore,shouldalsoaddressthisissue. ImplicationsofParentBrandCharacteristics Brandextensionresearchhasfurthermoredemonstratedthatparentbrandcharacteristicsaffect brandextensionevaluation.Specifically,ithasbeenfoundthattheperceivedquality,orreputation,oftheparentbrandaswellasthenumberofproductsaffiliatedwithabrandpositivelyaffect brandextensionevaluationandeventuallysuccess(DacinandSmith1994;VölcknerandSattler 2006;SmithandPark1992).Fromacross-culturalperspective,however,wehaveuncoveredthe possibilitythatthetwodeterminantsofperceivedqualityandnumberofproductsaffiliatedwith abrandmightplayasignificantlymoreimportantroleinsomenationalmarketsthaninothers. Accordingly,wehavehypothesizedthattheperceivedqualityoftheparentbrandwillbeamore importantdeterminantofbrandextensionevaluationinmarketswhoseculturesarehigh(versus low)onpowerdistance,uncertaintyavoidance,collectivism,andsocioeconomics.Similarly,we hypothesizedthatthenumberofproductsaffiliatedwiththeparentbrandwillconstituteamore importantdeterminantofbrandextensionevaluationincultureshighonuncertaintyorientation andcollectivism,andlowonsocioeconomics.Itisunclearwhateffectpowerdistancehasonthe relationshipbetweenthenumberofproductsaffiliatedwiththeparentbrandandbrandextension evaluation.Overall,ourfindingssuggestthefollowingforinternationalmarketingmanagers.
BRAND EXTENSION RESEARCH 119
First,managersshouldassesstheparentbrand’sperceivedquality,orreputation,amongtargetmarketconsumers,aswellasthetargetmarketconsumers’perceptionsaboutthenumberof productsaffiliatedwiththeparentbrand.Managersshouldthenensurethattheproductcharacteristics,whicharemoreimportantinsomeculturesthaninothers,aresatisfactorilyperceived. Forexample,ahigh-qualityparentbrandinthetargetmarketismoreimportantincultureshighin powerdistance(e.g.,Malaysia),uncertaintyavoidance(e.g.,Greece),andcollectivism(e.g.,China orEcuador).Asaresult,internationalmarketingmanagersshouldassurethattheparentbrand’s perceivedquality,orreputation,issufficientinagivennationalmarketforthebrandextensionto befavorablyevaluatedandhencesuccessful.Similarreasoningcanbeputforwardforthenumber ofproductsaffiliatedwiththeparentbrand. Second, the previous reasoning implies that it might be helpful and strongly advisable for managers—depending on the culture—to support the introduction of brand extensions with a communicationstrategythataimsathighlightingthegoodqualityoftheparentbrandand/orthe severalproductsthatarealreadyaffiliatedwiththeparentbrand.Thisseemstobeofparticular importanceforcultureshighinuncertaintyavoidanceandcollectivism.Whilesomenationalbrand extensionresearchhasexaminedseveralcommunicationstrategies(e.g.,Barone,Miniard,and Romeo2000;Bridges,Keller,andSood2000;Lane2000),nobrandextensionstudyexiststhat examinesthisissueacrosscultures. Again,however,itisunclearhowthedifferentculturaldimensionsinteractwitheachotherand affecttherelationshipbetweentheparentbrandcharacteristicsandbrandextensionsevaluation.In addition,itisunclearhowthetwodeterminantsofperceivedquality,orreputation,andnumberof productsaffiliatedwiththeparentbrandinteractwitheachotherandincombinationaffectbrand extensionevaluation(againstthebackgroundoftheculturaldimensions).Moreresearchisneeded tounderstandtheseinteractionprocesses. ImplicationsofBrandExtension’sProductCategoryCharacteristics Priorresearchonbrandextensionshasalsofoundthatthebrandextension’sproductcategory characteristicsaffectbrandextensionevaluation.Inparticular,ithasbeenfoundthatconsumers’ perceivedsocialandfinancialriskofthebrandextensioncategoryaffectconsumers’evaluation ofbrandextensions(DelVecchioandSmith2005;VölcknerandSattler2006).However,wehave arguedthatthisrelationshipislikelytobemoderatedbyculture.Asaresult,wehavehypothesized thattheperceivedsocialriskoftheextensioncategoryconstitutesamoreimportantdeterminant ofbrandextensionevaluationincultureshighinpowerdistance,collectivism,andsocioeconomicsthanincultureslowonthesedimensions.Similarly,consumers’perceptionoffinancialrisk oftheextensioncategoryislikelytoconstituteamoreimportantdeterminantofbrandextension evaluationincultureshighinpowerdistance,uncertaintyavoidance,individualism,andlowin socioeconomicsthaninculturesthataretheoppositeonthesedimensions.Thisfindinghasthe followingimplicationforinternationalmarketingmanagers. Managersneedtomakesurethattheintroductionofthebrandextensiongoesalongwithan assessmentofthetargetmarketconsumers’perceivedrisk(socialandfinancial)oftheproduct category,intowhichthebrandisintendedtobeextended.Insomecultures(e.g.,highpower distanceanduncertaintyavoidance),riskperceptionplaysanimportantdeterminantofbrandextensionevaluationandneedstobereducedincaseitisperceivedtoohigh.Inothercultures(e.g., lowpowerdistanceanduncertaintyavoidance),riskperceptionplaysalessimportantroleinthe evaluationofbrandextensions,andhencesuccess,andmightnotneedtobeaddressedtomake consumersacceptandbuythenewbrandextension.Asaresult,itistheinternationalmarketing
120
M.A. MERZ, D.L. ALDEN, W.D. HOYER, AND K.K DESAI
managers’tasktoassessthelevelofperceivedriskofthebrandextensionproductcategoryin thetargetmarketandtodecidewhethermarketingmedianeedtobeusedtoreducetheperceived riskoftheextensioncategory.Again,itisunclearhowthedifferentculturaldimensionsinteract andincombinationaffectbrandextensionevaluation.Therefore,furtherresearchshouldaddress thisissue,too. Overall,internationalmanagersneedatoolthatenablesthemtodecidewhichofthedeterminants ofbrandextensionevaluationaremostimportantinaparticularcultureaccordingtoindividual dimensionsaswellasacombinationofthesedimensions.Amainobjectiveofcross-culturalbrand extensionresearchshouldbetoundertaketheempiricalanalysesrequiredtomorefullydevelop suchaconceptualtoolforinternationalmarketingmanagers. References Aaker,DavidA.andErikJoachimsthaker.1999.“TheLureofGlobalBranding.”HarvardBusinessReview 77(November/December):137–144. Aaker,DavidA.,andKevinLaneKeller.1990.“ConsumerEvaluationsofBrandExtensions.”Journalof Marketing54(1):27–41. ———.1993.“InterpretingCross-CulturalReplicationsofBrandExtensionResearch.”InternationalJournal ofResearchinMarketing10(1):55–59. Alden,DanaL.,Jan-BenedictE.M.Steenkamp,andRajeevBatra.2006.“ConsumerAttitudestowardMarketingGlobalization:Antecedent,ConsequentandStructuralFactors.”InternationalJournalofResearch inMarketing26(3):227–239. Barone,MichaelJ.,JeanB.Miniard,andPaulW.Romeo.2000.“TheInfluenceofPositiveMoodonBrand ExtensionEvaluations.”JournalofConsumerResearch26(March):386–400. Batra,Rajeev,VenkatramRamaswamy,DanaL.Alden,Jan-BenedictE.M.Steenkamp,andS.Ramachander. 2000.“EffectsofBrandLocalandNonlocalOriginonConsumerAttitudesinDevelopingCountries.” JournalofConsumerPsychology9(2):83–95. Bearden,WilliamO.,andMichaelJ.Etzel.1982.“ReferenceGroupInfluenceonProductandBrandPurchase Decisions.”JournalofConsumerResearch9(September):183–194. Belk, RussellW. 1988. “ThirdWorld Consumer Culture.” In Marketing’s Contribution to Development ResearchinMarketing,ed.E.KumcuandA.F.Firat,103–127.Greenwich,CT:JAIPress. Belk,RussellW.,KennethD.Bahn,andRobertN.Mayer.1982.“DevelopmentalRecognitionofConsumptionSymbolism.”JournalofConsumerResearch9(1):4–17. Bottomley, PaulA., and John R. Doyle. 1996. “The Formation ofAttitudesTowards Brand Extensions: TestingandGeneralisingAakerandKeller’sModel.”InternationalJournalofResearchinMarketing, 13(4),365–77. Bottomley,PaulA.,andStephenJ.S.Holden.2001.“DoWeReallyKnowHowConsumersEvaluateBrand Extensions?EmpiricalGeneralizationsBasedonSecondaryAnalysisofEightStudies.”JournalofMarketingResearch38(4):494–500. Boush,DavidM.,andBarbaraLoken.1991.“AProcess-TracingStudyofBrandExtensionEvaluation.” JournalofMarketingResearch28(1):16–28. Boush,DavidM.,S.Shipp,B.Loken,E.Gencturk,S.Crockett,E.Kennedy,B.Minshall,D.Misurell,L. Rochford, and J. Stroble. 1987. “Affect Generalization to Similar and Dissimilar Brand Extensions.” Psychology&Marketing4(3):225–237. Bridges,Sheri,KevinLaneKeller,andSanjaySood.2000.“CommunicationStrategiesforBrandExtensions: EnhancingPerceivedFitbyEstablishingExplanatoryLinks.”JournalofAdvertising29(4):1–11. Broniarczyk,SusanM.,andJosephW.Alba.1994.“TheImportanceoftheBrandinBrandExtension.” JournalofMarketingResearch31(2):214–228. BusinessWeek.2006.“TheBestGlobalBrands.”August7,54–56. Chiu,Liang-Hwang.1972.“ACross-CulturalComparisonofCognitiveStylesinChineseandAmerican Children.”InternationalJournalofPsychology7(4):235–242. Czellar,Sandor.2003.“ConsumerAttitudeTowardBrandExtensions:AnIntegrativeModelandResearch Propositions.”InternationalJournalofResearchinMarketing20(1):97–115.
BRAND EXTENSION RESEARCH 121 Dacin,PeterA.,andDanielC.Smith.1994.“TheEffectofBrandPortfolioCharacteristicsonConsumer EvaluationsofBrandExtensions.”JournalofMarketingResearch31(2):229–242. Dahl,Stephan.2004.“InterculturalResearch:TheCurrentStateofKnowledge.”Workingpaper,Middlesex University,BusinessSchoolLondon. DelVecchio,Devon,andDanielC.Smith.2005.“Brand-ExtensionPricePremiums:TheEffectsofPerceivedFitandExtensionProductCategoryRisk.”JournaloftheAcademyofMarketingScience33(2): 184–196. Erdem,Tülin, Joffre Swait, andAnaValenzuela. 2006. “Brands as Signals:A Cross-CountryValidation Study.”JournalofMarketing70(January):34–49. Escalas,JenniferEdson,andJamesR.Bettman.2003.“YouAreWhatTheyEat:TheInfluenceofReference GroupsonConsumers’ConnectionstoBrands.”JournalofConsumerPsychology13(3):339–348. Grewal,Dhruv,JerryGottlieb,andHowardMarmorstein.1994.“TheModeratingEffectsofMessageFramingandSourceCredibilityonthePrice-PerceivedRiskRelationship.”JournalofConsumerResearch 21(1):145–53. Han,JinK.,andBerndH.Schmitt.1997.“Product-CategoryDynamicsandCorporateIdentityinBrandExtensions: AComparisonofHongKongandU.S.Consumers.”JournalofInternationalMarketing5(1):77–92. Hannerz,Ulf.1992.CulturalComplexity:StudiesintheSocialOrganizationofMeaning.NewYork:ColumbiaUniversityPress. Harrell,GilbertD.1986.ConsumerBehavior.SanDiego,CA:HarcourtBrace. Hem,LeifE.,LesliedeChernatony,andNinaIversen.2003.“FactorsInfluencingSuccessfulBrandExtensions.”JournalofMarketingManagement19:781–806. Hermans,HubertJ.M.,andHarryJ.G.Kempen.1998.“MovingCultures.”AmericanPsychologist53(10): 1111–1120. Hofstede,Geert.1991.CulturesandOrganizations:SoftwareoftheMind.NewYork:McGraw-Hill. ———. 2001. Culture’s Consequences—ComparingValues, Behaviors, Institutions, and Organizations acrossNations.ThousandOaks,CA:Sage. Holt,DouglasB.,JohnA.Quelch,andEarlL.Taylor.2004.“HowGlobalBrandsCompete.”HarvardBusinessReview82(9):68–75. Interbrand.2006.“BestGlobalBrands2006—ARankingbyBrandValue.”Report.Availableatwww.ourfishbowl.com/images/surveys/BGB06Report_072706.pdf(accessedNovember2006). Jain,ShailendraPratap,KalpeshKaushikDesai,andHuifangMao.2007.“TheInfluenceofChronicand SituationalSelf-ConstrualonCategorization.”JournalofConsumerResearch34(1):66–67. Ji,Li-Jun,KaipingPeng,andRichardE.Nisbett.2000.“Culture,ControlandPerceptionofRelationships intheEnvironment.”JournalofPersonalityandSocialPsychology78(May):943–955. John,DeborahRoedder,BarbaraLoken,andChristopherJoiner.1998.“TheNegativeImpactofExtensions: CanFlagshipProductsBeDiluted?”JournalofMarketing62(January):19–32. Keyfitz,Nathan.1982.“DevelopmentandtheEliminationofPoverty.”EconomicDevelopmentandCultural Change30(3):649–670. Lane,VickiR.2000.“TheImpactofAdRepetitionandAdContentonConsumerPerceptionsofIncongruent Extensions.”JournalofMarketing64(April):80–91. Loken,Barbara,andDeborahRoedderJohn.1993.“DilutingBrandBeliefs:WhenDoBrandExtensions HaveaNegativeImpact?”JournalofMarketing57(July):71–84. Mandel,Naomi.2003.“ShiftingSelvesandDecisionMaking:TheEffectsofSelf-ConstrualPrimingon ConsumerRisk-Taking.”JournalofConsumerResearch30(June):30–40. Mandler,George.1982.“TheStructureofValue:AccountingforTaste.”AffectandCognition:The17th AnnualCarnegieSymposium,eds.MargaretS.ClarkandSusanT.Fiske,3–36.Hillsdale,NJ:Lawrence Erlbaum. Merz,MichaelA.,andDanaL.Alden.2007.“UncertaintyOrientationandtheEvaluationofBrandExtensions.”Workingpaper,UniversityofHawaii. Merz,MichaelA.,YiHe,andDanaL.Alden.Forthcoming.“ACategorizationApproachtoAnalyzingthe GlobalConsumerCultureDebate.”InternationalMarketingReview. Meyers-Levy,Joan,andAliceM.Tybout.1989.“SchemaCongruityasaBasisforProductEvaluation.” JournalofConsumerResearch16(1):39–54. Milberg,SandraJ.,C.WhanPark,andMichaelS.McCarthy.1997.“ManagingNegativeFeedbackEffects AssociatedwithBrandExtensions:TheImpactofAlternativeBrandingStrategies.”JournalofConsumer Psychology6(2):119–140.
122
M.A. MERZ, D.L. ALDEN, W.D. HOYER, AND K.K DESAI
Miller,J.G.1997.“TheoreticalIssuesinCulturalPsychology.”InHandbookofCross-CulturalPsychology, ed.Y.H.PoortingaJ.W.Berry,andJ.Pandey.Vol.1.Boston:Allyn&Bacon. Monga,AlokparnaBasu,andDeborahRoedderJohn.2004.“ConsumerResponsetoBrandExtensions:Does CultureMatter?”AdvancesinConsumerResearch,31,216–222. ———.2007.“CulturalDifferencesinBrandExtensionEvaluation:TheInfluenceofAnalyticversusHolistic Thinking.”JournalofConsumerResearch33(March):529–536. Ng,SharonandMichaelHouston.2004.“ExemplarsandBeliefs?ImplicationsofRepresentationalDifferencesonBrandEvaluationsAcrossCultures.”AdvancesinConsumerResearch31:223. Nisbett,RichardE.,DavidH.Krantz,andChristopherJepson.1983.“TheUseofStatisticalHeuristicsin EverydayInductiveReasoning.”PsychologicalReview90(4):339–363. Nisbett, Richard E., Kaiping Peng, Incheol Choi, andAra Norenzayan. 2001. “Culture and Systems of Thought:HolisticVersusAnalyticCognition.”PsychologicalReview108(April):291–310. Ozanne,JulieL.,MerrieBrucks,andDhruvGrewal.1992.“AStudyofInformationSearchBehaviorduring theCategorizationofNewProducts.”JournalofConsumerResearch18(March):452–463. Park,C.Whan,BernardJ.Jaworski,andDeborahJ.Maclnnis.1986.“StrategicBrandConcept-ImageManagement.”JournalofMarketing50(4):135–145. Park,C.Whan, Sandra Milberg, and Robert Lawson.1991.“EvaluationofBrandExtensions:TheRole ofProductFeatureSimilarityandBrandConceptConsistency.”JournalofConsumerResearch18(2): 185–193. Roth,MartinS.1995.“TheEffectsofCultureandSocioeconomicsonthePerformanceofGlobalBrand ImageStrategies.”JournalofMarketingResearch32(2):163–175. Schwartz,ShalomH.1992.“UniversalsintheContentandStructureofValues:TheoreticalAdvancesand EmpiricalTestsin20Countries.”InAdvancesinExperimentalSocialPsychology,ed.M.Zanna,Vol. 35.SanDiego,CA:AcademicPress. ———.1994.“BeyondIndividualism/Collectivism:NewCulturalDimensionsofValues.”InIndividualism andCollectivism:Theory,MethodandApplications,ed.U.Kim,H.C.Triandis,C.Kagitcibasi,S-C.Choi, andG.Yoon,85–119.NewburyPark,CA:Sage. Smith,DanielC.,andC.WhanPark.1992.“TheEffectsofBrandExtensionsonMarketShareandAdvertisingEfficiency.”JournalofMarketingResearch29(3):296–313. Spencer-Oatey,H.2000.CulturallySpeaking:ManagingRapportinTalkAcrossCultures.London:Continuum. Stayman,DouglasM.,DanaL.Alden,andKarenH.Smith.1992.“SomeEffectsofSchematicProcessing on Consumer Expectations and Disconfirmation Judgments.” Journal of Consumer Research 19 (2): 240–255. Steenkamp,Jan-BenedictE.M.,RajeevBatra,andDanaL.Alden.2003.“HowPerceivedBrandGlobalness CreatesBrandValue.”JournalofInternationalBusinessStudies34(1):53–65. Sujan,Mita,andJamesR.Bettman.1989.“TheEffectsofBrandPositioningStrategiesonConsumers’Brand andCategoryPerceptions:SomeInsightsfromSchemaResearch.”JournalofMarketingResearch26 (November):454–467. Sunde,Lorraine,andRoderickJ.Brodie.1993.“ConsumerEvaluationsofBrandExtensions:FurtherEmpiricalResults.”InternationalJournalofResearchinMarketing10(1):47–53. Völckner,Franziska,andHenrikSattler.2006.“DriversofBrandExtensionSuccess.”JournalofMarketing 70(2):18–34. Wernerfelt,Birger.1988.“UmbrellaBrandingasaSignalofNewProductQuality:AnExampleofSignaling byPostingaBond.”RandJournalofEconomics19(Autumn):458–466. Yoon,Yeosun,andZeynepGurhan-Canli.2004.“Cross-CulturalDifferencesinBrandExtensionEvaluations: TheEffectofHolisticandAnalyticalProcessing.”AdvancesinConsumerResearch31:224.
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 123
CHAPTER5
AREVIEWOFEYE-TRACKING RESEARCHINMARKETING MICHELWEDELANDRIKPIETERS
Abstract Motivatedfromthegrowingimportanceofvisualmarketinginpractice,werevieweye-tracking researchtoevaluateitseffectiveness.Weprovideacasestudyoftheapplicationofeye-trackingto adpretesting.Werevieweye-trackingapplicationsinadvertising(print,TV,andbanner),health andnutritionwarnings,branding,andchoiceandshelfsearchbehaviors.Wethendiscussfindings, identifycurrentgapsinourknowledge,andprovideanoutlookonfutureresearch. Introduction Consumersareexposeddailytohundredsofadvertisementsontelevision,innewspapers,magazines,theyellowpages,andonInternetsites,aswellastoahostofothervisualmarketingstimuli. Companiestrackcloselywhatconsumerssee,inordertohelprendertheirvisualmarketingefforts moreeffective.Asaconsequence,therehasbeenarapidgrowthincommercialapplicationsof eye-tracking technology in the United States and Europe. Firms such as Kraft Foods, PepsiCo, Pfizer,P&G,andUnileverareleadingusersofthemethodology.Providersofeye-trackingdatanow conducthundredsofstudieseachyear.Suchcommercialresearchcompaniesinclude,forexample, PerceptionResearchServicesintheUnitedStatesandVerifyInternationalinEurope.Thegrowthof eye-trackingisinlargepartdrivenbytechnologicalinnovationsinthedevelopmentofeye-tracking devicesandsharpdeclinesinthecostsofthesedevices.Untilrecently,thecommercialuseofeyetrackingwascumbersome,timeconsuming,andexpensive.Thissituationhaschangedinrecent yearsduetonewgenerationsofinfraredeye-trackers,whichenableeyemovementrecordingof consumersundernaturalexposureconditions,withlargeamountsofstimuli,andathighprecision andlowcost.Therefore,ourgoalistoprovideareviewoftheapplicationsofeye-trackingresearch inmarketingtodate,andtheinsightsthatithasyielded.Wefirstprovidesomebackgroundoneyetrackingandvisualattention,thenprovideacasestudyontheuseofeyemovements,andproceed toreviewapplicationsofeye-trackinginanumberofareasofmarketing. TheoreticalBackground Inordertocloselyprocessaspecificobjectorlocationinavisualmarketingstimulus,consumers havetomovetheireyes.Thisisrequiredbecauseacuityacrosstheretinaofthehumaneyerapidly fallsoffwithincreasedeccentricityfromthefovea,whichisthecentralandmostsensitivepartofthe retinadirectlyoppositeofthelens.Thismakesitinterestingtostudyeyemovementsasindicators 123
124
MICHEL WEDEL AND RIK PIETERS
ofinformationacquisitionbehavior(Russo1978).Whatwebelievetobesmoothmovementsof oureyesinfactconsistoftwoverydifferentcomponents:fixationsandsaccades(Buswell1935). Saccadesarerapid,ballisticjumpsoftheeyes,typicallylastingaround20–40milliseconds,that servetoprojectspecificlocationsofthesceneontothefovea.Thesaccadeisthefastestmovement inthehumanbody,andhumansmakearoundonehundredandseventythousandofthemaday. Fixationsaremomentsduringwhichtheeyeisrelativelystill,typicallylastingaround200–500 milliseconds(Rayner1998).Duringafixationacontiguousareaofthesceneisprojectedonto thefoveafordetailedvisualprocessing.Atanygivenpointintime,onlyabout8percentofthe visualfieldisprojectedonthefoveaandavailablefordetailedprocessing. Yettheamountofinformationthatistransmittedthroughtheopticnerveexceedswhatthebraincan process,sothebrainhasevolvedattentionalmechanismsthatselectasubsetofrelevantinformation forenhancedprocessing.Whenattentionselectsaparticularlocationorobjectinascene,processing ofitisenhanced,andprocessingofnonselectedlocationsandobjectsissimultaneouslysuppressed. Inprinciple,attentionmayoperateonspatiallocations,visualfeatures,orobjectsinthescene.Both traitsandstatesoftheconsumer(top-downfactors)andcharacteristicsofthevisualmarketingstimulus (bottom-upfactors)caninfluenceattention.Muchisknownabouttheinfluenceofstimulus-based (bottom-up)factorsonattention,andthatknowledgecanbegainfullyappliedinthedesignofvisual marketingstimuli.Anobject“popsout”andisfoundinstantaneouslyonthefirsteyefixation,based onpre-attentiveprocesses,whenitstandsoutinthescenebecauseofasingleperceptualfeature,such asthewaythenewHeinzgreenketchupstandsoutamongthehomogeneouslyredcompetingbrands ontheshelf.Thesebasicperceptualfeaturesincludecolor,edges,luminance,shapes,andsizesof objectsinascene.Thebottom-upcontrolofattentionbythosefeaturesislargelyinvoluntarily.Much lessisknownabouttheinfluenceoftop-down,voluntarymechanismsinguidingattention.Yet,both bottom-upandtop-downprocessesinfluenceattention.Top-downfactorssuchasconsumers’search goals(“findthenewHeinzbottle,”or“findthecheapestcakemix”)andmemoryforbasicfeatures (“whatisthecolorofthenewHeinzbottle?”)influencevisualprocessingbyselectivelyenhancing visualfeaturesthatarediagnosticandbyselectivelysuppressingfeaturesthatarenondiagnostic. Suchvoluntary,top-downenhancementandsuppressionofperceptualfeaturesiseffortful,usually slower.Thus,attentionprioritizesobjectsincommercialscenesasaresultoftheintegratedeffect ofbottom-upandtop-downfactors.Attentionisreflectedineye-movements.Informationextracted duringfixationseventuallycontributestodownstreammarketingeffectsofinterestsuchaslearning (memory),preferenceformation,choice,andsales. Thepatternoffixationsandsaccadesacrossastimulussuchasanadvertisementiscalleda scanpath(NotonandStark1971).Eye-trackersrecordthepatternsoffixationsandsaccadesthat consumersmakeacrossavisualstimulus.Themostcommontypesforcommercialpurposesuse infraredcornealreflectionmethodology,whichmeasuresthedistanceandangleofthereflection ofinfraredlightfromthecenterofthepupiltodeterminethepointoffixationoftheperson,aftercalibration(YoungandSheena1975).Suchinfraredeye-trackinghasatemporalandspatial resolutionofsufficientaccuracyforcommercialandacademicapplicationsinmarketing.Modern table-mounteddevices(e.g.,TobiiSystems1)integrateminiaturehead-andeye-trackingcameras intoanLCDmonitor,whichenablesindividualstofreelymovetheirheadsduringeye-tracking. Thecomparativelylowcostsofnewgenerationsofeye-trackingsystems,theirshortcalibration times,naturalexposureconditions,andunobtrusivemeasurementhavegreatlycontributedtothe recentgrowthofeye-movementapplicationsinmarketingpracticeaswellastheirusefortheory developmentandtestinginacademicresearch. Figure5.1showsaneye-trackingdevice(top),andtheeyemovementsofasampleofconsumers acrossamagazinead(bottom).Eye-trackingprovidesmoment-to-momentmeasuresofthepoint-
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 125 Figure5.1 TobiiEye-Tracker(Top),andScanpathsof100ConsumersSuperimposedon aCarAd(Bottom)
126
MICHEL WEDEL AND RIK PIETERS
of-regardofconsumersexposedtovisualmarketingstimuli,includingtelevisioncommercials, printads,catalogs,yellowpagesadvertising,outdooradvertising,point-of-purchasematerial, andWebpages.Suchspatio-temporalmeasuresofvisualattentionareintimatelyrelatedtothe attentionprocessesofprimeinterest,andcannotbeobtainedotherwise.Becauseitisincreasingly evidentthatattentionismorecentraltotheprocessingandeffectivenessofvisualstimulithan previously believed (see subsequent discussion), and because the current technology enables measurementsatunprecedentedscaleandprecision,weseeaprominentroleforeye-trackingin theorydevelopmentandvisualmarketingpractice.Toillustratetheuseofeye-trackinginvisual marketingpractice,weprovideanapplicationinanadvertisingpretestcontexttoimproveprint adlayout,below. Surprisingly,inviewofthedominanceofvisualstimuliinmarketing,theapplicationofeyetrackingmethodsinpractice,and,asillustrated,theimportanceofvisualprocessinginconsumer behavior,aperusalofthemarketingliteraturerevealsalackofcontinuedfocusoneye-tracking.In thepast,thestudyofvisualattentionintheacademicmarketingliteraturemayhavebeenhampered bytheacceptedwisdomthat(1)attentionisamereprecondition,agatethroughwhichinformation entersonitswaytohigher-ordercognitiveprocessesofmoreinterest,(2)gainingandretaining attentioniseasy,forexample,throughcontrastwithcompetitorsandrepetition,and(3)measuring attentionwitheye-trackingisdifficultbecauseofthecostsofearliereye-trackinggenerations,and unnecessarybecausemeasuringitthroughmemorymeasuresiseasyandvalid. Academicresearchhasshownthatthestudyofvisualattentionisimportantinitsownrightand thattheacceptedwisdomisflawed.First,psychologicalresearchrevealsthatvisualattentionis notonlyagate,assuggestedbyhierarchicalprocessingmodelssuchasAIDA(Attitude-InterestDesire-Action;Starch1923;Strong1920),butalsoandmoreimportant,akeycoordinatingmechanismthatservestomaintaininformationprocessingandothergoalsovertime(LaBerge1995). Attentionmightevenbeclosertoactualbehaviorthanintuitioninformsus,andeyemovements couldbemorethanthetipoftheiceberg(Russo1978),asrecentworkbyRizzolattiandhiscolleaguesaboutthepotentiallyclosecorrespondenceofeyemovementsandhigher-ordercognitive processesreveals(Rizzolatti,Riggio,andSheliga1994). Second,gainingandretainingattentionisdifficult,becauseofthehighlevelsofadvertising clutter—thatis,thenumberofadsthatconsumersareexposedtoatagivenlocationortimein newspapers,magazines,orWebpages(BurkeandSrull1988;Keller1991;Kent1993)andthe numberofSKUs(StockKeepingUnits)onshelvesofbrick-and-mortarretailersorWebpagesof onlineretailers(Mulvihill2002).Breakingthroughtheclutterandgainingconsumers’attention isdifficultineachofthosesituations.AsJones(2004,p.146)putsit,“ifconsumers’attention isnotcaughtinthefirstplace,allfindings[onmessageappreciationandacceptance]in‘forced exposure’researcharevalueless.” Thirdandfinally,althoughitislikelytruethatstandardmeasuresofattentionbasedonposthocrecallarebiased,measuringvisualattentioniseasy.Biel(1993,p.29)states,“Iamtroubled bythefactthat(a)attentionisclearlyanelementthatdeterminestheeffectivenessofadvertising, but(b)recall’sfailingsmakeitunacceptableasacandidatemeasureoftheattentiondimension. Thatwouldleadmetopushhardforresearchanddevelopmenttoprovideausefulmeasureof attention.Myviewisthatabehaviouralmeasureofattentionmaybeafruitfulareainwhichto search.”Eye-trackingprovidessuchmeasures.Eyemovementsaretightlycoupledwithvisual attention,makingthememinentindicatorsofthevisualattentionprocess,whichisnoweasyto assesswithmoderneye-trackingequipment.Eyemovementsarebehavioralmeasuresoftheunobservablevisualattentionprocessofprimeinterest.“Theeyesdon’tlie.Ifyouwanttoknowwhat peoplearepayingattentionto,followwhattheyarelookingat”(DavenportandBeck2001,p.19).
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 127
Visualattentiontoadvertisinghasbeenshowntohavesystematicdownstreameffectsonbrand memory,indicatingitspredictivevalidity(Pieters,Warlop,andWedel2002;WedelandPieters 2000).Further,validityissupportedbyexperimentsthatdemonstratearelationbetweenattention tobrandsonshelves,measuredwitheye-tracking,andin-storedecisionmaking(Chandon2002; PietersandWarlop1999;RussoandLeclerc1994).Finally,apreliminaryindicationthatattention isassociatedwithsaleswasprovidedbyTreistmanandGregg(1979),whoobservedthatoneof twoadsthatpeoplelookedatlongerinaneye-trackingexperimentalsoattainedhighersales.With suchreassuringfindingsonpredictivevalidity,thecurrenteaseofeye-movementrecordings,and theemergenceofstrongtheoriesofvisualattention,thedoorisopenforfurtherresearchonand theestablishmentoftheacademicfieldofvisualmarketing,buildingonandextendingwhathas becomeknowninrecentyearsthrougheye-trackingresearch. ACaseStudy Thiscasestudyinvolvesperformancemetricsandestablishedbenchmarkstoimproveanunderperformingprintad.Itwasconductedin2000byVerifyInternationalforafull-pageprintadof Robyn,aEuropeanUnileverliquidlaundrydetergentbrand.DatacollectionwasdoneatVerify’s eye-trackingfacilities.Thedesignwasan“after-onlywithcontrolgroup”design;115(initialad) and109(improvedad)regularconsumerswererecruitedconsecutivelytoparticipateinthetest,and wereexposedtotheinitialadandtheimprovedad,respectively.Theads,astheywouldappearina regularmagazinewithaneditorialcounterpage,werepresentedonNEC21-inchLCDmonitorsin full-colorbitmapswitha1,280x1,024pixelresolution.Participantsinthetestcontinuedtoanext pagebytouchingthelowerrightcornerofthe(touch-sensitive)screen,aswhenpaging.Infrared cornealreflectionmethodologywasusedforeye-tracking.Theequipmentleavesparticipantsfree tomovetheirheads,camerastrackingboththepositionoftheeyeandhead. ThetoppanelofFigure5.2showstheinitialad,withthreekeyregionsofinterest(ROI),the key ad design objects (brand, text, pictorial), and eye fixations of 115 consumers (50 percent malesand50percentfemales)obtainedthrougheye-tracking,superimposed.Table5.1shows theaveragefixationsontheeditorialcounterpage,theadpage,anditsthreeobjects,andthree keymetricsderivedfromthead.Alsoshownarebenchmarks,thescoresofthebest30percent ofadsfromthedatabaseof6,000recentlytestedads.ThethreemetricsusedaretheCrea-Score (percentageofsampleattendingtheadinthefirsttenseconds,ameasureofattentionretention), BrandContact(percentageofsamplefixatingthebrand),andStimulusPerformance(percentage ofsamplefixatingbrand,text,andpictorial).Thescoresarerespectively40,48and30percent (withstandarderrorsaround4–5percent).Itisclearthattheinitialadfallsshortofthebenchmark.Theseresultsprovidedinputtoredesigntheadintermsofthesizesandlocationsofthe keydesignelements(headline,pictorial,brand,andtext),withoutaffectingitscreativecontent. Althoughformalprocedurescanbeusedtodesigntheadbasedonsuchinformation,hereheuristic procedureswereused. Thenewadwasretested.TheimprovedadisshowninthelowerpanelofFigure5.2,againwith ROIandfixationpointsoverlaid,andTable5.1showsthemetricsforit.Theretestrevealsthatthe improvedad,evenwithrelativelysimpleandinexpensivechangesinthevisuallayoutinterms ofthedesignelements,andvirtuallynochangeinitscreativecontent,performsmuchbetter.The Crea-Score,BrandContact,andStimulusPerformancemetricsimproveto48,61,and45percent, respectively,thelattertwochangesbeingstatisticallysignificant.Buteventheseimprovedmetrics fallshortofthebenchmarkofthetop30percentofads.Althoughtheretestresultswerealready encouraging,roomforfurtheradimprovementremained,andformaloptimizationmethodsto
128
MICHEL WEDEL AND RIK PIETERS
Figure5.2 RobynTestCaseResults:InitialAd(Top)andImprovedAd(Bottom),With SameEditorialCounterpage;RegionsofInterest(RoundedBoxes),and FixationPointsSuperimposed
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 129 Table5.1
Verify/UnileverRobynPretestCase InitialAd (N=115)
ImprovedAd (N=109)
AverageFixationFrequencies Editorialpage Adpage Brand Text Pictorial
7.74 10.63 1.42 5.00 6.83
6.47 13.22 2.40 6.61 7.19
PerformanceMetrics Crea-Score BrandContact StimulusPerformance
40 48 30
48 61 45
Improvement (%)
Benchmarks
20 27 50
46 83 58
maximizeattentionbyoptimizingthesurfacesizesandlocationsofthekeydesignobjectsinads couldbeutilized.Thekeynotionisthattheeye-tracking-basedattentionmetricsusedinthiscase studyareadequatemetricstomeasureadeffectiveness.Theprecedingdiscussiononthetheoryof visualattentionandeyemovementshasarguedthistobethecase,andwewillreturntothisissue inthediscussionsection.Notethattheeye-trackingonlytakesafewsecondsofparticipants’time toconduct,doesnotrequireverbalquestioningthatmightdisruptongoingadprocessing,andcan bedoneatlowcost,whichcontributestoitsincreasingpopularityinvisualmarketingpractice. Eye-TrackingResearchforVisualMarketing In1924,inthefirsteye-movementanalysisofprintadsthatweknowof,Nixonobservedeyemovementsofconsumerswhowerepagingthroughamagazinewithprintads,whilehidinghimselfin aboxbehindacurtain.Sometimelater,Karslake(1940)usedthePurdueEyeCameratocollect eye-movementdataonadvertisementsappearingintheSaturdayEveningPost.In1950,Fittsand hiscolleagues(Fitts,Jones,andMilton1950)examinedtheeyemovementsofpilotslandingan airplane,whichisthefirstusabilitystudythatprovidedfindingsnowcentralinWebandinterface research.Afteraperiodofrelativesilenceinmarketing,newimpetusfortheuseofeye-tracking camefromRusso’spioneeringarticleof1978,“Eye-FixationsCanSavetheWorld,”inwhichhe arguedforstudyingeye-movementstoevaluatemarketingeffectiveness,focusingonconsumer decisionprocesses.Russocomparedfivecognitiveprocesstracingmethods,includinginformation displayboards,input-outputanalysis,andverbalprotocols,onsevencriteria.Themethodswere ratedonsevenperformanceattributes,includingquality,validity,obtrusiveness,easeofuse,and costoftheequipment.Heconcludedthateye-fixationmethodologyscoreshighonmanyofthese criteriaandoffersadvantagesnotofferedbyothermethods.Asprocesstracingdata,eyemovements offerdetailandvalidity.Verbalprotocolswereconcludedtobecomplementarywitheyefixations, noothermethodbeingmoredifferentfromeyefixations,andRussosuggestedsimultaneoususe ofthesetwomethodsinmarketingresearch. SinceRusso’sarticle,applicationsofeye-trackinghaveappearedinsubstantiveareasofvisual marketinginvolvingin-storechoicedecisionsandshelfsearch,printadvertising,TVcommercials, e-commerce,labelingandeducationalmessages,andbranding.Theareaofvisualsearchalone haswitnessedasteeprisefromalittleovertenpublishedpapersperyearintheearlyninetiesto
130
MICHEL WEDEL AND RIK PIETERS
overtwohundredadecadelater.Rayner(1998)offeredanextensivereviewofeye-trackingin readingandrelatedareas,andDuchowski(2003)recentlysurveyedeye-trackingapplicationsin variousareas,includingengineeringandpsychology,andprovidesanumberofmarketingapplicationsaswell. Inthispaper,werevieweye-movementresearchinmarketing.Wefocusonbroadcategories ofvisualmarketingstimuliandtasks:choiceandsearchbehavior,printadvertising,publicpolicy information,televisioncommercials,Webusabilityandadvertising,andreadingtasksinsurvey designandbranding.Thefundamentaldifferencesinthenatureofthestimuliinthoseareasallow foruniqueinsightsinthevisualattentionprocessesinquestion.Wespecificallyfocusoninsights ontheeffectsofbottom-upandtop-downfactorsonthevisualattentionprocess.Weexaminethe insightsonbottom-upspace-based,feature-based,andobject-basedattentionthatthesestudies haveafforded.Wealsosummarizethetop-downroleofspecificgoalsandtasksthatconsumers engageinwhenexposedtovisualmarketingstimuli,forexampletoselectoneobjectoutofmultiplecompetingonesonadisplay,duringsearchandchoice,ortocomprehend,evaluate,and/or memorizevisualmarketingstimuliasawhole,whenexposedtoprintandtelevisionadvertisements. Weexaminevisualmarketinginsightsgainedthrougheye-trackingofstationarystimulisuchas printadsandyellowpageads,anddynamicstimulisuchastelevisioncommercials,andweexplore Webusability,whereconsumerswithmultiplegoalsareexposedtostaticanddynamicscenes. ChoiceandSearchBehavior Inachoiceorsearchtask,theconsumer’sgoalistoselectoneoutofasetofmultipleobjects.In achoicetask,preferenceuncertaintyneedstobereduced,thatis,theparticipantneedstodecide whichobjecttochooseamongtheavailablealternatives.Inatargetsearchtask,spatialuncertainty iskey,thatis,itisunclearwheretheobjectofinterestisamongitsdistracters.Focusingonthe choiceprocessitself,vanRaaij(1977)useddirectobservationsofeyemovementsapplyinga one-waymirrorandrecordingcamera.Hisstudyrevealedapervasiveusebyconsumersofpaired comparisonsbetweenalternatives. RussoandLeclerc(1994)investigatedthechoiceprocessfornondurables,buildingonearlier workbyRussoandRosen(1975).Inalaboratorysimulationofsupermarketshelves,likevan Raaij,RussoandLeclercuseddirectobservationofeyemovementsfromvideorecordingsthrough aone-waymirror.Theyheroicallyobservedthecompletescanpathofeyemovementsacrossthe alternativesintheset,identifyingthreedifferentstagesinthechoiceprocess:orientation,evaluation,andverification,respectively.Orientationconsistedofanoverviewoftheproductdisplay.In theevaluationstage,whichwasthelongest,directcomparisonsbetweentwoorthreealternative productsweremade.Theverificationstageinvolvedfurtherexaminationofthealreadychosen brand.Modelsofplannedanalysisofchoicealternativesweredisconfirmedinfavorofanadaptive andconstructiveprocess.Inparticular,thelaststagethattheyobservedisofinterestandnovel, andfurtherresearchmaystatisticallytesttheirthree-statesequentialmodelofchoice. Focusingontop-downfactors,PietersandWarlop(1999)studiedtheimpactoftimepressure andtaskmotivationonvisualattentionduringbrandchoice.Allbrandswerenew,toruleout (top-down)memoryeffectsonchoice.Analysisofeyemovementsrevealedthatvisualattention adapts rapidly to differences in time pressure and task motivation, two important contextual factors,underscoringthefindingsofRussoandLeclerc(1994).Underhightimepressure,consumersacceleratedinformationacquisition,asrevealedbyadecreaseintheaveragedurationof eyefixations.Moreover,participantsfilteredinformationbyskippingtextualinformationonthe packaging.Underhightimepressure,consumersalsoshiftedtoaprocessing-by-attributestrategy,
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 131
indicatedbyincreasingnumbersofbetween-brandsaccades.Highlymotivatedconsumersdeceleratedinformationacquisition,indicatedbylongeraveragefixationdurations,andprocessing-byattributediminished,indicatedbyreducedlevelsofinter-brandsaccades.Whenmotivationwas high,consumersskippedfewerofthebrandnamesandmoreofthepictorialobjects.Thisreveals thecontentofconsumers’brandschemas,whichspecifyforinstancethatrelevantinformation duringbrandchoiceisconsideredtobetextualratherthanpictorial.Theresultsofthisstudy demonstratethepotentialofusingeye-movementanalysistoinferhighercognitiveprocesses, theimportanceoftaskandindividualfactors,withimplicationsforthepretestingofthedesign ofpackagesandshelves.Theyalsoshowedthatbrandpreferencecanbepredictedfrompatterns ofeyemovements. Inastudyontheimpactofpoint-of-purchase(POP)marketingonbrandchoice,Chandon (2002)developadecision-pathmodelofvisualattentionandconsiderationfornewandexisting brands.Itenablesestimatingaproduct’svisual(bottom-upfromexposuretothedisplay)and memory-basedequity(top-down,i.e.,theaddedvaluederivedfromtheseprocesses),andtheir sensitivitytoperson,brand,andPOPmarketingfactors.Animportantfindingisthatlookingat abrandincreaseditsconsiderationprobabilityby30to120percent.Chandonandassociates concludedthattheimpactofvisualequityislargestforbrandswithlowermemory-basedequity, whichisinlinewiththeinterplayofbottom-upandtop-downmechanismsinvisualattention, andsuggeststhatthecurrentpracticeofallocatingshelfspaceaccordingtomarketshareneed notbeoptimal. Infourstudies,Janiszewski(1998)investigatedexploratorysearchbehaviorandstudiedrelationshipsbetweenthesizeofobjectsinproductdisplaysandtheamountofattentiondevotedtothem, buildingonandextendingAnstis’s(1974)theory.Explorationissimilartotheorientationstageof RussoandLeclerc(1994).Janiszewski’sresearchrevealedtheinfluenceofbottom-upmechanisms incompetitionforattention,anddemonstratedtheinfluenceofdisplaylayout,aswellasofobject sizeandeccentricityfromthefixationpointonattention.Partofhisstudiesusedretailercatalog displaysasstimuli.Heconcludedthatthecompetitionforattentioncreatedbyitemssurrounding afocalitem(measuredasthesummedattentionaldemandofthesedistracteritems)influencesthe amountoftimeapersonspendslookingattheitemandthelikelihoodthatthepersonwillrecall informationabouttheitem.Thisresearchhasimplicationsforvisualmerchandising,advertising, andcatalogpagelayout,andshows,forexample,thatthecurrentpracticeofincreasingattention toaspecificiteminacatalogdisplaybyincreasingitssizeand/orremovingotheritemsmaybe suboptimal,andthatcarefullyrearrangingthelayoutoftheitemscouldmaximizeattentionto thedisplayasawhole. Lohse(1997)collectedeye-movementdatawhilethirty-twoconsumerschoseamongfortyeightbusinessesinatelephonedirectory,andrecordedthreechoicesamongtheselistingsunder ahypotheticalgoal,suchas“yourcarneedsrepair.”Thisworkisimportantbecauseitsynthesizesresearchonchoice(RussoandLeclerc1994),onbrandsearch(Janiszewski1998),andon attentiontoadvertising(reviewedinthenextsection).Afractionalfactorialdesignwasusedto systematicallyvarycolor(red/black),presenceofgraphics(yes/no),fonttype(bold/plain),and ad size (small/large, display/in column), among other things, as important bottom-up factors. Analysisofthefixationdatarevealedthatconsumersscanaccordingtoserialposition,butnot exhaustively,andasaresult,someadsareneverseen.Adsizehadastrongeffect:consumers noticedalmostallofthequarter-pagedisplayadsbutonlyone-quarteroftheplainlistings.This, ofcourse,isinlinewiththefindingsbyJaniszewski(1998)onsize.Participantsfixatedcolor beforenoncolorads,noticedmorecoloradsthannoncolorads,andviewedcolorads21percent longerthanequivalentadswithoutcolor.Also,theyviewed42percentmoreboldlistingsthan
132
MICHEL WEDEL AND RIK PIETERS
plainlistings.Theseresultssupporttheimportanceofbasicperceptualfeaturessuchascolorand size,andofcontrastsinthesefeaturesinthedesignofadsandchoicesets.Consumersspent54 percentmoretimeviewingadsforbusinessesthattheyendedupchoosing,whichdemonstrates theimportanceofattentionforsubsequentchoicebehavior;consistentwiththefindingsofRusso andLeclerc(1994)andPietersandWarlop(1999).Inafollow-upstudy,LohseandWu(2001), examinedeye-movementdatafromChineseconsumersengagedinchoosingbusinessesfromthe yellowpagesofaChinesetelephonedirectoryonacomputerscreen,andfoundconvergentevidence fortheirpreviousfindings,acrossadifferentcultureanddifferentadstimuli,whichunderlinesthe generalizabilityofthefindings.Itremainstheonlystudytodateaddressingculturaldifferences inattentionandeyemovements. PrintAdvertising Thestudyofeyemovementswithprintadsparallelsstudiesonsceneperceptioninpsychology, whichusuallyinvolvesanintegrationofvariousobjectsinthesceneintoameaningfulwhole (HendersonandHollingworth1998;Yarbus1967).Thisdiffersfromchoiceandsearchtaskswhere thegoalistoselectoneoutofasetofmultipleobjectsinadisplay. Kroeber-Riel(1979)citesthepioneeringresearchinthedoctoraldissertationofWitt(1977)in whichsixtyparticipantswereshownadvertisementsonaprojectionscreenforafixeddurationof sixseconds.Offourdifferentads,twoversionswerecreatedbymanipulationofthepictorial,to invokemildversusintensearousal.Thus,Wittexaminestheinfluenceofemotionsastop-down factors,andinterestinglyusesmanipulationofthepictorialtoinducethesetop-downeffects.The averagenumbersoffixationsonthepictorialofthetwotypesofadswere3.9and5.5respectively, indicatingmoreinformationintakeunderhigherarousal.Fixationsonthetextobjectoftheads werenotaffected,however,indicatingthattheactivationwasnottransferredtootherobjectsof thead,whichWittexplainedasalackofconceptualintegrationofthesetwoobjects.Witt(1977) alsodemonstratedarelationshipbetweenfixationfrequencyonanobjectandsubsequentrecall ofit,althoughthecorrelationsweremoderate(0.3–0.4). Inanotherearlystudyinmarketing,TreistmanandGregg(1979),bothworkingatPerception Research Services (PRS) at the time, investigated the diagnosticity of fixation data, ability to discriminatebetweendifferentprintadvertisements,andabilitytopredictwhichofseveralads hadthehighestsalespotential.Theyrecordedvisualselectionofthreeaddesignelements:pack shot,productcopy,andpriceinformation;theproportionoftimespentoneachofthesedesign elements; and the sequence in which they were visited (among other things) for a number of cosmeticsads.Thisstudythuslooksattheeffectsofadobjects(elements)asbottom-upfactors influencing attention. The researchers observed relationships of eye-movement patterns with involvement,familiarityasatop-downinfluencingfactor,andwithpurchaseintentasadesirable down-streameffect.Further,combiningeye-trackingdatawithpurchaseinterestdatacorrectly pinpointedthebetteradvertisement.Althoughmanyofthefindingsweredescriptive,thisstudy laysanimportantfoundationforfurtherresearchinvisualattentiontoprintads,inestablishing theroleoftop-downfactorsanddown-streammeasuresofeffectiveness. Leven(1991)conductedastudyontheinfluenceofspace-andobject-basedaspectsofthe adlayoutonattention.Tothataim,149consumerswereexposedto10printadvertisementsfor detergentandcosmeticsproductsunderafreeviewingtask,whiletheireyemovementswererecordedwithinfraredeye-tracking.Consumerswereseatedabouttwometersfromabacklitslide projectorscreen(1-by-1meter)onwhichtheadswereshown,andtheycouldproceedtothenext adbypushingabutton.Averagegazedurationundertheseconditionswas7.2seconds,whichis
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 133
comparabletoresearchunderhighinvolvementconditions(Rayneretal.2001),buthigherthan typicalgazedurationsundermorenatural,lowinvolvement,conditions(PietersandWedel2004). Leven(1991)foundthatthecenterofadswasfixatedmuchmorefrequently,withlongergaze durationsthanthesides,andthattheupper-rightcornerwasfixatedleast,andwiththeshortest gazedurations,whichsuggeststhatthislocationwasonlyexploredratherthanattendedtoin depth.Inaddition,hisresearchindicatedapreferentialscanpathsequencestartinginthemiddle, goingtothetop,andafterothersteps,endinginthelower-rightcorner,whereoftenthebrand and/orsloganarelocated.Additionalanalysesrevealedthatthenumberofdifferentadobjects (sixweredistinguished:headline,brandname,product/packshot,bodytext,peopleandobjects) fixatedrapidlyincreasedtofourafterabouttwo-and-halfseconds,andthantaperedoff,reaching sixobjectsafteronly8seconds.Thissuggeststhatparticipantsquicklyscannedtheadsbefore attendingtothemindetail,whichwereturntolater. Inastudyonattentiontoprintads,Rosbergen,Pieters,andWedel(1997)proposeamethodologytoaccountforheterogeneityinvisualattentioneffectsacrossconsumers.Forthatpurpose theyusedamixtureregressionmodel,whichaccountsforunobservedsegmentsofconsumers basedontheireyemovementsacrossfouraddesignobjects—pictorial,brand,headlineandbody text—similartothoseusedbyTreistmanandGregg(1979)andLohse(1997).Thus,theprimary focushereisonthebottom-upeffectsofsizeasabasicperceptualfeature.Inanexperiment, theyrecordedconsumers’eye-movementsduringnaturalexposuretoaconsumermagazine,in which experimentally designed ads were inserted. Three consumer segments were identified thatexhibiteddistinctpatternsofvisualattention,respectivelycalledthescanning(attentionto headlineandpictorial),initial(attentiontoheadline,pictorial,andbrand),andsustained(attentiontoallfourobjects)attentionsegments.Totaladviewingtimeincreasedfromthefirsttothe thirdsegment,andthesegmentswereshowntodifferinproductinvolvement,brandattitude,and recall,thussupportingtheinfluenceoftop-downfactorsandthepredictivevalidityofeye-tracking fordownstreammeasures.Byrevealinghoweachofthethreesegments’attentionpatternswas differentiallysensitivetocharacteristicsoftheads,suchasthesizeoftheirheadline,thisstudy showedthecombinedeffectsof(bottom-up)salienceand(top-down)informativenessonattentiontoadvertising.Thisalsoillustratestheimportanceofaccountingforconsumerheterogeneity instudiesofvisualattention,andthecontributionofstatisticalmodelsinmakinginferenceson covertvisualattentionfromoverteye-movementdata. Visualattentionduringrepeatedexposurestoprintadvertisementswasthetopicofeye-tracking researchonwearoutbyPieters,Rosbergen,andWedel(1999).Theysetouttotestscanpaththeory, proposedbyNotonandStark(1971),whichpostulatesthatsequencesoffixationsoccurringupon firstexposuretostationarystimulireoccurduringsubsequentexposures.Theyproposedastatistical modelcomprisedofsubmodelsforattentiondurationandforbetween-andwithin-objectsaccades, withthelatterdescribedbya(heterogeneous)Markovmodelthatcapturesfirst-orderdependencies offixationsintime.Theydistinguishthesamefourad-designobjectsasinRosbergen,Pieters,and Wedel(1997)andotherstudiescitedabove.Theirresultsshowedthatattentiondurationacross advertisingrepetitionsdecreasessignificantly—byasmuchas50percentonaverage.Herethe repetitionswerethreecloselyspacedexposures,mimickingtheincreasinglypopularpracticeof advertisingschedulingforhighimpact.Inspiteofthat,andinsupportofscanpaththeory,the switchingprobabilitiesbetweenadobjectsremainedfairlyconstantacrossrepetitions.Theseresults supportsimilarfindingsbyHarris(1993)onthestabilityofscanpaths.Whereasscanpaththeory specifiesthatthestabilityofscanpathsisduetostorageofsuchpathsin(visuo-spatial)memory andthusatop-downeffect,itnowseemsmorelikelythatthestabilityofscanpathsisinfactdriven bottom-upbythelayoutofthead,althoughtherehavebeennodirecttestsofthisspeculation
134
MICHEL WEDEL AND RIK PIETERS
yet.Thisresearchatteststotheimportanceofvisualattentioninunderstandingmechanismsof advertisingeffectivenessandadwearout. Followinguponthepreviouslydescribedstudy,Liechty,Pieters,andWedel(2003),develop ahiddenMarkovmodeltocapturethetemporaldimensionofvisualattentiontoprintads.Their modelisbasedonevidence,citedearlier,thatwhilevisuallyexploringprintads,consumers’attentionswitchesbetweentwounobservedstates.Theideaisthatduringsceneperception,people trytoreducetwotypesofuncertainty,namely,whattheidentityofthevariousobjectsinthescene isandhowthesecontributetotheoverallmeaningofthescene,whichrequiresdetailed,focalattention,andwherethevariousinformativeobjectsandlocationsinthesceneare.Liechty,Pieters, andWedel(2003)inferthesetwovisualattentionstates,theirrelativeprevalenceandsequencein time,fromthecompletesequenceofeye-trackingdata.Forthatpurpose,theydevelopanextensionofexistinghiddenMarkovmodels(formulatedandestimatedintheBayesianframework). Themodeldescribestheobservedtime-seriesofsaccadesonaspatial(48-cell)gridoverlaidon theads,forwhichafirst-orderMarkovmodelisused,similartoPieters,Rosbergen,andWedel (1999).Buttheiraccountofthevisualattentionprocessisprimarilyspatial,asitisinLeven(1991). Ahiddentwo-stateMarkovprocess,indicativeofattentionswitchingbetweenglobalandlocal states,isassumedtodrivethoseobservedeyemovements.Individualsareassumedtomakeshorter saccadesinthelocalthanintheglobalstate.Notethatthismodelformalizesandallowsexplicit testsof,amongothers,RussoandLeclerc’s(1994)andLeven’s(1991)findingsofunobserved attentionstatessequentiallydrivingeyemovements.Themodelwasestimatedoneye-movement datacollectedinastudyofsixty-nineconsumersexposedtoseventeenprintadvertisementsintheir naturalcontext.Participantsnearlyalwaysstartedinthelocalattentionstate(withaprobabilityof over90percent),buttheprobabilityofterminatingexposuretotheadvertisementsismuchlarger whenintheglobalattentionstate(23percent).Duringexposuretoanadvertisement,participants jumpbetweenthelocalandglobalstates2.6timesonaverageandtendtospendlongerinthelocal (1.13seconds)thanintheglobal(0.22seconds)state.Byswitchingbackandforthbetweenlocal andglobalattention,theproblemofinterpretingthecomplexvisualsceneisbrokendownintoa sequenceofsimplerlocalizedinterpretationsofthemostsalientandinformativeregions.Although thelinkbetweenthetwostatesofthemodelandthe“where”and“what”pathwaysinthevisual brainisnotuncontested(ReichleandNelson2003;alsoseeWedel,Pieters,andLiechty2003), takentogether,theseresultsshowthatthelocalattentionstateisdominant,withahighprobability ofstartinginthatstate,wherethereareoccasionalshortjumpstotheglobalattentionstatewith ahighprobabilityofendingattentiontothestimulusinthatstate.Theauthorsdidrevealthatthe two-statemodelprovidesamuchbetterdescriptionofthedatathanaone-statemodel,butdidnot testforathree-statemodel.Thisresearchdemonstrateshowonecanuseeye-movementmeasures asindicatorsofunderlyingtop-downfactorstotesttheoriesofattention,whichFeng(2003)refers toasadesirableapproachto“reverseinference.”Suchareverseinferenceapproachisimportant becauseitallowsustoinfercognitiveprocessesfromobservedeye-movements,andenablestests ofattentiontheoriesproposedinpsychologyandcognitiveneuroscience.Thesearefacilitatedby thelargesamplesizesofstimuliandparticipants,andthetemporalandspatialresolutionofcurrentdatacollectedthrougheye-tracking. Amongthepioneersoneye-trackingresearchinreading,Raynerandcoauthors(2001)used printadvertisementsinastudyonattentiontotextandpictorialinformationundervariouslevels ofinvolvement.Littleresearchinpsychologyhadaddressedthatissuepreviously.Intheirexperiment,halfoftheparticipantswereinstructedtopayspecialattentiontocarads,andtheotherhalf weretoldtopayspecialattentiontoskin-careads,whereasbothgroupswereexposedtoboththe carandskin-careads.Thus,Rayneretal.chosetostudytop-downeffectsbymanipulatingtask
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 135
instructions.Viewerstendedtospendmoretimelookingatthetext(about70percent)thanthe picturepartofthead,andspentmoretimelookingatthetypeofadtowhichtheywereinstructed topayattention.Fixationdurationsandsaccadelengthswerebothlongeronthepictorialthanthe text,butmorefixationsweremadeonthetext.Fixationsonthetextseemedtofollowpredictable patternsfoundinotherstudiesonreading.Here,participantstendedtoreadthelargeprint(inthe headlines)first,thenthesmallerprint(inthebodytext),andthenattendedtothepicture.Although someoftheresultsofthestudymaybecausedbydemandeffects,theparticipantsbeingspecificallyinstructedtopayattentiontotheads(Yarbus1967),avaluablefindingisthattheattention toadvertisementsisstronglyinfluencedbythetaskinstructions(lookingatcarversusskin-care ads).Oneoftheconclusionsofthestudywasthatperceptionofadsandothercomplexscenesis largelybasedontextualinformation,andthelackoftransferfrompictorialtotextthatwasfound isconsistentwithfindingsfromotherstudies(Witt1977).Amoreextensivestudyontopdown effectsonattentiontoprintadswasreportedonbyPietersandWedel(2007).Inthestudy,five processinggoalsdirectedattheadasawholeorthebrandelement,andinducingonmemoryor learningwereexperimentallyvaried.Theresultsdemonstrateamarkedeffectofthesegoalson consumereye-movements,inlinewithwhatYarbus(1967)alreadyreported.Forexample,text receivesmostattentionunderabrandlearninggoal,whileanadlearninggoaluniformlypromotes attentiontomostelements.Theheadlineappearedtobeinvarianttothesetop-downeffects,howeverandreceivedequalattentionunderallconditions. Wedel and Pieters (2000) exposed 88 consumers to 65 print ads appearing in their natural contextin2magazines,inanattempttoformalizetheattention-to-memoryrelationshipobserved inseveralpreviousstudies(Lohse1997;RosbergenPieters,andWedel1997;Witt1977).The frequencyofeyefixationsonthebrand,pictorial,andtextobjectsofeachadwasrecordedfor eachconsumer,andconsumerswerelateraskedtoidentifythebrandsfrompixilatedimagesof thebrands,inaperceptualmemorytask.Thisindirectmemorytaskisimportantlydifferentfrom thelesssensitiverecalltasksusedinpreviousresearch.Accuracyandlatencyofmemorywere assessed.Acontributionofthisstudywasthattheauthorsprovideanaccountoftheprocessing thattakesplacetostoreinformationinlong-termmemory,whichhadbeenlackinginprevious research.TheydevelopedahierarchicalBayesianmodelforthatpurpose,groundedinattentionand memorytheories.Themodeldescribestheprocessbywhicheyefixationsonprintadvertisements leadtomemoryfortheadvertisedbrands.Ateachfixationaninformationchunkisextractedfrom thead,theamountofinformationcontainedinitvaryingrandomlyacrossadsandconsumers, andisaccumulatedinlong-termmemoryacrossmultiplefixationstotheadobjects.Thetotal amountofaccumulatedinformationinfluencesboththeaccuracyandlatencyofbrandmemory, andaccuratememoryisassumedtooccurwhentheaccumulatedinformationexceedsathreshold. Thestudylooksatsizeofadelementsasbasicfeaturesthataffectattentionbottom-up.Wedel andPietersconfirmedagainthatacrossthetwomagazines,brandsurfacesizehadaparticularly prominenteffectonfixations.Onaverage,thebrandwasmuchsmallerthanthepictorial(about 10times)andthetext(about3to5times),butthebrandreceivedbyfarthemostfixationsper unitofitssurface,followedbythetextobject—supportingtheRayneretal.(2001)findings—and thepictorial.However,whereasfixationstothepictorialandthebrandsystematicallypromoted accuratebrandmemory,textfixationsdidnot.Thisfindingaddsinsighttopreviousresultsonthe attention-memoryrelationship,butcastsdoubtsontheassumedcentralroleoftextinadvertising. Theauthorsalsofoundarecencyand(smaller)primacyeffectonmemory,whilecontrollingfor attention—indicatingtheroleofmemoryretrievalcues. Pieters,Warlop,andWedel(2002)extendedthestudyofWedelandPietersbyfirstinvestigating theinfluenceofadoriginalityandfamiliarityonconsumers’eyefixationsonthebrand,text,andpicto-
136
MICHEL WEDEL AND RIK PIETERS
rialobjectsofadvertisements,andhowthesefactors,alongwiththeinformationextractedduringeye fixations,promotememoryfortheadvertisedbrand.Adoriginalityisusedcommonlyinadvertising tobreakthroughcompetitiveclutter.However,itcouldhavedetrimentaleffectswhenconsumerspay moreattentiontotheadattheexpenseoftheadvertisedbranditself.Moreover,thepositiveeffectsof originalitycanquicklyfadewhentheadbecomesfamiliarandthenoveltywearsoff.InPieters,Warlop, andWedel’s(2002)study,infraredeye-trackingwasappliedtocollecteye-fixationdatafrom119consumerswhopagedthroughtwogeneral-audiencemagazinescontaining58full-pageadvertisements. Insupportoftheresearchers’hypothesis,originaladvertisementsdrewmoreratherthanlessattention totheadvertisedbrand.Moreimportant,however,advertisementsthatwerebothoriginalandfamiliar attractedthelargestamountofattentiontotheadvertisedbrand,whichimprovedsubsequentbrand memoryaswell.Inaddition,originalityandfamiliaritywerefoundtoalsopromotebrandmemory directly.Theseresultsrevealthedoublebenefitsofadoriginalityandhoweye-trackingresearchcan detectthisinpre-andpost-testing,andthussupportaddevelopmentandmediaplanning Whereasmanyofthepreviouslydescribedstudiesdocumentedtheeffectsofsurfacesizesofad objectsonvisualattention,theseresultshadlimitedgeneralizabilityduetoarelativelysmalland selectsetofadvertisements.Therefore,PietersandWedel(2004)conductedananalysisof1,363 printadvertisementstestedwithinfraredeye-trackingmethodologyamongover3,600consumers. Datawereaggregatedacrossconsumers.Theyexaminedattentioncaptureandtransferforthree adobjects—brand,pictorial,andtext.Attentioncapturewasoperationalizedasthepercentageof participantsfixatingaselectedadobjectatleastonce.Attentiontransferwasoperationalizedas theeffectofgazedurationforoneoftheadobjectsongazefortheotherobjects,comparableto Janiszewski’s(1998)operationalizationofattentioncompetitionondisplays.Eachofthethree adobjectshadauniqueeffectonattentiontoadvertisements,whichareatoddswithcommonly heldideasinmarketingpractice.Thepictorialissuperiorincapturingattention,independentofits size,whichgoesagainstrecommendationsinadpracticetomakethepictorialsizelargeoreven aslargeaspossible.Thetextobjectbestcapturesattentionindirectproportiontoitssurfacesize, thatis,attentiontothetextincreasesmostbyincreasingitssize,amongtheseobjects.Thebrand objectmosteffectivelytransfersattentiontotheotherobjects.Thatis,longergazestothebrand objectcarryovertothoseonotherobjects,butnotnecessarilytheotherwayaround.Theauthors foundlittleornotransferfromthepictorialtobrandandtextobjects,whichgeneralizedprevious observationsby,amongothersWitt(1977)andRayneretal.(2001),butareatoddswithcommon thinkingabouttheinfluenceofthepictorialinattentionguidance.Onlyincrementsinthetext object’ssurfacesizeproducedanetgaininattentiontotheadasawhole.Brandfamiliaritywas showntoreduceattentiontothebrandobjectbutsimultaneouslyincreasedattentiontothetext object,ratherthanhavingauniformeffectacrossallobjects.Thestudyhasobviousimplications fortheroleofthedesignobjectsinprintaddesigninpractice.Finally,Pieters,WedelandZhang (2007)investigatedtheoptimaldesignoffeatureadvertisements.Theseadsareparticularlyproneto clutter,becausemultiplefeatureadsoftenappeartogetheronasinglepage.Theauthorsproposed twomeasuresofclutter,andestimatedamodelthatallowedthemtooptimallydesignafeaturead displaypage.Usingdatafromcloseto1,500newspaperfeatureads,theyshowedthatthedesign canbesignificantlyimprovedtoattractmoreattentiontotheentireaddisplay,bydecreasingthe sizeofthepictorial,andincreasingthesizesofpromotionandpriceelementsinparticular. PublicPolicyandSocialMarketingInformation Whereasprintadsarescenescontainingmultimodalinformationforcommercialpurposes,common healthwarningsandnutritionlabelsarebasiciconicandtextualcomponentsofadsandpackages.
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 137
Thereisahistoryinsocialmarketingandpublicpolicyresearchofeye-trackingmandatoryhealth warningsandnutritionlabels.Inaseriesofstudies,agroupofcollaborators(Fischeretal.1989;Fox etal.1998;Fletcheretal.1995;Krugmanetal.1994),extensivelydocumentedadolescentsviewing tobaccoandbeerads.Thesestudiesareinterestedinattentiontotheselabelsaswholeobjects,on whichbottom-upandtop-downfactorsmayjointlyoperate.Inonestudy,(Fischeretal.1989)with sixty-oneadolescentsexposedtofivedifferenttobaccoadvertisements,theaverageviewingtimeof thewarningamountedtoonly8percentofthetotaladvertisementviewingtime.Almosthalfofthe participantsdidnotfixateonthewarningintheadatall,whileabout36percentoftheparticipants appearedtohavereadsomeofthewarning.Followingadviewing,participantswereaskedtoidentify thewarningstheywereexposedtoinalistthatincludedothersimulatedwarnings.Participantsdid onlyslightlybetterthanrandomguessinginthisrecognitiontest.Theauthorsconcludedthatthe federallymandatedwarningsmustbeviewedasalargelyineffectivepublichealthmessage. InastudybyKrugmanandassociates(1994)among326adolescents,newwarningsweredevelopedandtested.Twomandatedwarningsandtwonewlydevelopedwarningswereembedded inmagazineadsfortwocigarettebrands.Participantsviewedeachadaslongasdesiredwhileeye movementswererecorded.Participantssubsequentlycompletedamaskedrecalltask.Themetrics usedwerethenumberofparticipantswhonoticedthewarning,theirtimetofirstfixationonthe warning,andthetimespentfixatingonthewarning.Resultsindicatedthatwithinthecompetitive readingenvironmentcreatedbyacigaretteadvertisement,newwarningsattractgreaterreadership, withquickerattentiontothenewthantofamiliar,mandatedwarnings.Thedifferenceinattention capturewasabout10percent,thenewwarningscapturingmoreoftheparticipants’attention;the proposedwarningsalsocapturedattentionfaster,timestothefirstfixationonthewarningbeing between0.8and2.4seconds.Attentionengagementwasalsolongerforthenewwarnings,by abouthalfasecond,or1–2fixations.Onaverage,participantsspent2–3secondsviewingthe warnings.Theauthorsfoundtheattentionmeasurestobesignificantlyrelatedtoasubsequent maskedrecallmeasure.Next,Fletcheretal.(1995)providedstrongerevidenceoftheassociation betweendwell-timeandcontentrecall,byincludingprintadsfortwonontobaccoproductsthat alsocontainedwarnings,alongwiththeabovecigaretteadsandtheirwarnings.Thesefindings areinlinewithfindingsonattentioneffectsonmemoryforprintads,inparticularthosebyWedel andPieters(2000).Implicationsofthesestudieswerethattheeffectivenessofwarningscanbe improvedthroughsimplicityandnovelty.Simplicitycanbeachievedbyreducingthenumberof wordsinthetext(24fortheexistingwarnings),andnoveltybyamorerapidchangeofwarning contentandformatsorbyoriginality,topreventwearout(seePieters,Warlop,andWedel2002). Thesestudiesthusdemonstratetheeffectofbothbasicfeaturesandsemanticcontent. Foxetal.(1998)followeduponthesestudiesandexposed148fourteen-toeighteen-year-olds tofivefrequentlyrunprintadstargetedtothissegment,whilerecordingtheireyemovements. Twooftheadswereforcigarettes,andonewasforabeerbrand.Interestingly,thereweresignificantdifferencesinthegazedurationsforthetwocigaretteadsthatcarriedovertothewarnings: warningsinadsthatareviewedlongeralsoreceivealongergaze(thegazesonthewarningswere relativelylong,2and2.5seconds,respectively).Thisagainpointstotheimportanceofinvestigating thewarninginthecontextofthead,andtotheeffectsofthecontextonattentioningeneral(see Janiszewski1998).Thealcoholwarningreceivedlessattention(1.6seconds)thanthecigarette warnings,andlessthanthelogointheads. Two studies of product labeling used visual search tasks while tracking eye movements, investigating the effect of perceptual features on search performance. Laughery et al. (1993) studiedwarningsonalcoholicbeveragecontainerswiththeaimofprovidingrecommendations forformatspecifications.Inoneofthreeexperiments,theauthorstrackedeyemovementsfor
138
MICHEL WEDEL AND RIK PIETERS
twenty-fourparticipantsexposedtothirty-eightlabels,selectedfromafactorialdesignwithtext color,pictorialcolor,presenceofborder,andaniconasfactors.Thus,thisstudylookedatthe effectsofbasicperceptualfeatures,bottom-up.Participantswereaskedtoidentifythewarning, andresponselatencytimeswerecollectedinasearchtask.Thestudysuggests,althoughoverall effectsweresmall,thatperceptualfeaturesusedinconjunctionmayhelpimprovethespeedof visuallydetectingthewarning(by50percenttoabout.5seconds)aswellasshortenthelatency ofrespondingtoit. Goldberg,Probart,andZak(1999)evaluatedfoodlabels’abilitytopromoteafastandaccurate visualsearchfornutritioninformation.Theylookedatexperienceasatop-downmoderatingfactor.Participants(fivepracticedlabelreadersandfivenonreaders)viewed180nutritionlabelson acomputerandwereaskedtofindanswerstospecificquestions(e.g.,aboutservingsize)under differentfixedexposuredurations(1,3,and5seconds)inavisualsearchtask.Labelmanipulations includedseveralalternativelinearrangements,locationofthetargetitem,andlabelsize.Dependent measuresincludedsearchtimeandfixationfrequency,aswellastheaccuracyanddurationofthe identifyingfixation.Practicedlabelreadersacquiredthetargetmorequicklyandaccuratelythan didless-practicedreaders,revealingastrongtop-downeffect.Surprisingly,labelsizedidnothavea significanteffect.Thinnerascomparedtothickeranchoringlineshadalargereffectonvisualsearch time.Targetsnearthecenterofthelabelrequiredone-thirdmoretimeandwerehardertofindthan targetsatthetoporbottomofthelabel,whichmaybecausedbymoredenseinformationnearthe center.Thisstudy,countertotheLaugheryetal.(1993)findings,showsthatlow-levelperceptual featuresaswellasspatialcharacteristicsmayinfluencevisualsearchperformance. TelevisionCommercials Eye-movementdataacrossdynamicscenessuchastelevisioncommercialsare,whatmightbe called,“doublydynamic,”sincetheeyesmoveoverimagesthatthemselvesmovetoo.Collectingandanalyzingsuchattentiondataischallenging,whichmaybeoneofthereasonsforthe comparativepaucityofresearchinthisarea.However,afewstudiesdidtakeupthegauntlet. Usingtelevisedsoftdrinkadsasstimuli,JaniszewskiandWarlop(1993)foundevidenceinthree experimentsthatconditioningprocedures,aimedtoestablishassociativelearningaboutthetarget brand,enhanceattentiontothebrand.Theaccountofvisualattentionisthusprimarilyobject (brand)based,andthelearningeffectsinvestigatedtop-downeffects.Thisisthusanimportant demonstrationofatop-downinfluenceonattentioninthecontextofdynamicstimuli.Janiszewski andWarlopconcludedthatitseemspossibletousethesequenceofscenesinaTVcommercial toencourageattentiontothebrandthroughconditioning,andtosimultaneouslymakesemantic informationavailableforcomparingchoicealternatives,asinputintochoicebehavior—aspeculationthatfutureresearchmighttest. Inastudyoftelevisedadvertisingimagesduringsoccergames,d’YdewalleandTamsin(1993), followinguponanearlierstudybyd’Ydewalleetal.(1988)usedeye-trackingtoinvestigatethe incidentalprocessingof(static)advertisingbillboardsplacedaroundthesoccerfield,andtheir subsequentrecallandrecognition.Bothstudiesfoundthatparticipantsdonotlookattheadvertising panelssubstantially,andbecausetheydonotlookatthem,theyareunabletorecallandtorecognize them.Theydidfindspecificeffectsofscenesthatcontainedclose-upsofthepanels,pointingtothe potentialofscene-basedanalyses,butdidnotinvestigatetheeffectofbasicfeaturesortop-down factors.Yetthesearetheonlyacademicstudiesthathaveinvestigatedtheeffectivenessofbillboard productplacementsinsportgames,andshouldserveasanimportantbaseforfutureresearchin thisarea.Inasecondstudyondynamicstimuli,d’Ydewalle,Desmet,andVanRensbergen(1998),
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 139
investigatedtheeffectsofsuccessiveshotsinmovies,andtheextenttowhichshottransitionsare perceptuallydisruptive.TheirfindingsareofobviousrelevancetoTVcommercialsaswell.This isoneofthefewstudiestodatethatinvestigatesthebottom-upeffectofmovementsonattention. Theresearchersdistinguishthreetypesoftransitionsbetweenshots.First-ordertransitionsrefer eithertosmalldisplacementsofthecamerapositionortosmallchangesoftheimagesize,possiblydisturbingtheperceptionofapparentmovementandleadingtotheimpressionofjumping. Second-ordertransitionsfollowfromareversalofthecameraposition,leadingtoachangeofthe left–rightpositionofthemainactorsandacompletechangeofthebackground.Withthird-order transitions,thelinearsequenceofactionsinthenarrativestoryisnotobeyed,andthesearemeant toreinforcethenarrativecontinuityofthestory.Theexperimentofd’Ydewalle,Desmet,andVan Rensbergen(1998)involvedeye-trackingon76participantswatching4experimentallydesigned versionsofa7-minutemoviesegment.Asadependentmeasure,d’Ydewalleandcoauthorsused thestandarddeviationsofthex-ylocationsoftheindividualfixationsin200-millisecond-long segments,asameasureofattentionalhomogeneityorfocus,causedbybottom-upeffects.The resultsshowedthattherewasanincreasedspatialvarianceofeyemovements200–400millisecondsafterbothsecond-andthird-ordershottransitions.Suchanincreasewasnotobtainedafter afirst-ordertransition,suggestingthattheincreasedspreadofthedistributionofeyemovements aftersecond-andthird-ordertransitionsisduetopostperceptual,cognitiveeffectsandthatthese transitionsdisruptperceptionandcognition. AokiandItoh(2000)proposedamethodologyforanalyzingviewers’attentiontotelevision commercials.Theirresults,basedonasmallsampleoffourparticipants,suggestthatproduct preferenceisnotamajorfactorforattentiontoTVadvertisements,butthatitmaybeinfluenced morebybottom-updesignfactorssuchasscene-changefrequencyandtheuseofcelebrities.In asecondstudy,AokiandItoh(2001)investigateinfluencesofauditoryinformationonhuman visualattentionduringviewingoftelevisioncommercials.Theycharacterizepatternsofgazetowardsceneswithandwithoutsoundsandfoundthatparticipants’gazepatternswereaffectedby auditoryinformation.Clearlytheseinitialstudiesprovidesomevaluableinsights,butthedetailed studyofattentiontoTVcommercialsremainsawideopenarea. WebUsabilityandAdvertising E-marketingthroughWebstores,electronicauctions,brandsites,bannerandpop-upadsonWeb sites,searchenginemarketing,andsoforthislargeandgrowing,andhasbecomeanintegral partoffirmstrategyandtactics.Giventhegrowthofexpendituresone-marketing,surprisingly littleacademicmarketingresearchtodatehasaddressedthisfield.Thereisanemergingstream ofresearchinappliedpsychologyandengineeringonhuman–computerinteractionandWebusability(e.g.,Cowen2001;JacobandKarn2003),butitstheoriesandfindingshavenotyetbeen fullyinfusedinvisualmarketing,andthereareonlyafewstudiesonWebadvertising.Because Webdesigncombinesstaticanddynamicvisualfeatures,andrequiresconsumerstoengagein search,choice,andsceneperceptiontasksoftensimultaneously,itcouldbecomeafertileground fortestingandapplyingintegrativetheoriesofthehumanperception–actioncycleanditsimplicationsformarketing. ThePoynterInstitute2hasconductedprobablythebest-knownWebusabilitystudiesusing eye-tracking.IntheirEyetrackIII,forinstance,eyemovementswererecordedoftheforty-six peoplewhowereexaminingmocknewsWebsitesandrealmultimediacontent.Oneofthefindingswasaspatialeffectoneyemovements:theeyesmostoftenfixatedfirstintheupperleft ofthepage.Onlyafterperusingthetopportionofthepageforsometimedidtheeyesexplore
140
MICHEL WEDEL AND RIK PIETERS
furtherdownthepage.Dominantheadlineswerefoundtomostoftendrawtheeyefirstupon enteringthepage,whichisconsistentwiththeworkofRayneretal.(2001)inthecontextof printads.However,(even)headlineshadlessthanasecondofasitevisitor’sattention,which underlinesthespeedofconsumerprocessingandtheimportanceofimmediatecommunication bythelargesttext. DrezeandHussherr(2003),inastudyonthedesignofaFrenchWebportalamongfiftyparticipants,examinedwhybanneradsseemtobeineffective,andwhatadvertiserscandotoimprove theireffectiveness.Click-throughratesarestillthedominantmeasureofInternetadvertising effectiveness,butaredisappointinglylowinmanycases,anditisthusimportanttounderstand theirdeterminants.DrezeandHussherrusedeye-trackingtoinvestigateWebsurfers’attentionto onlineadvertising,combinedwithasurveyofInternetusers’recall,recognition,andawarenessof banneradvertising.Theyclaimthattheirstudysuggeststhatthereasonwhyclick-throughrates arelowisthatsurfersavoidlookingatbanneradsduringtheironlineactivities.Outoftheeight banneradsthatparticipantswereexposedto,theylookedatfouronaverage.Theresearchers findstronglocationandsizeeffectsofregionsofinterestontheportal’sWebpages,reconfirmingfindingsonbottom-upeffectsofspaceandsizefromotherapplicationareasofeye-tracking research.But,theyalsofindanegativeeffectofregionsofinterestthatcontainthebannerad. Thismayimplythatpartofasurfer’sprocessingofbannerswillbedoneperipherallyand/or pre-attentively.Ifperipheralprocessingdoestakeplace,theclickthroughratemaynotbethe optimal,oronly,preferredmeasuretoassessbanneradperformance.Alternatively,thesefindings couldmeanthatrelativetopagecontent,banneradsattractlessattention,forexample,because theyaresmallerorbecausetheirlocations,whicharestoredinconsumers’spatialmemory,can easilybeignoredwithoutactiveavoidancetakingplace.Interestingly,“experts”spentlesstime lookingattheWebpagesandinspectedfewerregionsthannovices,whereastheirscanpaths seemedmoreregular,confirmingthetop-downeffectofexpertisepreviouslydocumentedby Goldberg,Probart,andZak(1999)forfoodlabels.Thus,expertsappeartobemoreefficient andsystematicintheirvisualexploration(seealsotheresultsofNotonandStark[1971]on explorationofpaintingsbyexpertsandnovices).Menandwomanbehavedsimilarly,butolder peoplespentmoretimeinspectingWebpagesthanyoungones,presumablybecausetheyhave lessautomatizedvisualroutinesforWebhandlingandthusneedmoresustainedattention.Banneradsdidhaveanimpactontraditionalmemory-basedmeasures,advertisingrecall,brand recognition,andbrandawareness.Althoughbrandawarenessisanimportantadeffectiveness measure,giventheproblemswithsuchmeasuresreportedpreviously,theconclusionthatthese arepreferredmeasuresofadeffectivenesswouldneedfurtherstudy.Advertisersshouldrely moreonvisualattentionmeasuresratherthanclickthrough(orinadditiontoit),whichisone keyconclusiontobedrawnfromthisstudy. Inatestofscanpaththeory(NotonandStark1971)forattentiontorepeatedInternetimages, JosephsonandHolmes(2002)exposedeightparticipantsrepeatedlytoaportalpage,anadvertisingpage,andanewsstorypageontheWebforadurationoftensecondseach.Theseauthors usedstring-editingtechniquestoassessthecompletesequenceofeyefixationsacrosspredefined areasofinterestontheWebpages,andhowthesesequenceschangedacrossexposures.TheresultsindicatebothsubstantialdifferencesbetweenparticipantsintheirscanpathforthesameWeb page,whichmaybeindicativeoftop-downeffects,aswellassubstantialdifferencesbetween Webpages,whichmaybeindicativeofbottom-upeffectsduetotheWebpagelayout.Thefindingscallforfurtherresearchthatallowsforstatisticaltestsofdifferencesbetween(largersamples of)stimuliandparticipantstoestablishthecontributionoftop-downandbottom-upeffectson attentiontoWebpages.
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 141
ReadingTasksinSurveyDesignandBrandExtensionResearch Eye-trackinginWebdesignandWebadvertisingservestoimprovehuman–computerinteraction byadaptinginterfacestonaturalhumanperceptionanddecisionprocesses.Eye-trackingcanalso servesuchergonomicfunctionsinotherareas,suchassurveyandquestionnairedesigninmarketingresearchandinbrandextensions,forwhichreadingtaskshavebeenused.Inparticular,for large-scalepaper-and-pencilmarketingandWebsurveyresearch,extensivepretestingofthemain questionnairescanbeconductedtodetectpotentialproblemswithquestioncomprehensibility, responseformats,andquestionnairerouting(Grovesetal.2004).RedlineandLankford(2001), forinstance,examinedhowparticipantsattendtoandmakeuseofbranchinginstructions(e.g., toskipanumberofquestions),andthusthenavigationalpathsinquestionnaires.Notfollowing theproperpathscouldincreaseitemnon-responseorcauseinaccurateresponses,andcouldalso increaseirritationandconfusion,allofwhichiscostly.Theyofferedthreeversionsofaquestionnaireaboutlife-styleissues,usingdifferentbranchinginstructions,totwenty-fiveparticipantsin athree-groupdesign.Eyemovementswererecordedonthequestion,answercategories,check box,andbranchinginstructions,andreturnsweepsbetweenquestioncomponentswereexamined. Theanalysesrevealedthaterrors-of-commission,wherebranchinginstructionsarenotfollowed asevidencedfromeyemovements,variedsignificantlyacrossthethreeconditions.Oneconclusionwasthatmanybranchinginstructionswerenotreadatall,andthattobeeffective,branching instructionsneededtoimmediatelyprecedeorfollowthespecificresponsecategorytowhichthey related.Thisisinlinewithearlierworkinreadingresearchshowingthatmostprocessingtakes placeduringorimmediatelyaftereyefixationsontherelevantmessagecomponents(“processing immediacy”),andthatthusthetimelagbetweentheinstructioninthemessageandtherequired actionoftheconsumershouldbeminimized. Anotheremergingareawhereeye-movementanalysisprovidesnewinsightsisbrandextension research.Asacaseinpoint,Stewart,Pickering,andSturt(2004)proposeaningeniouseye-movementbasedtechniquetoassesstheeffectivenessofbrandextensions.Brandextensionsarecommon,as muchas95percentofallnewproductintroductionsinvolvethem,yetfailuresarealsocommonand harmful,cannibalizingexistingbrandsanddamagingbrandequity.Whereasarangeofestablished explicitmethodsexistsforgaugingthesuccessofanextension,itisnotsufficientlycleartowhat extentthesemethodscantapthefastcognitiveprocessingduringbrandextensionevaluationinthe marketplace.Theauthorsproposeanewmethodbasedonthenotionthatambiguitiescreatedby implausibleextensionscauseprocessingdifficultiesthatshouldsurfaceineye-movementpatterns. Eighteenparticipantswerepresentedwiththreetypesofsentences(e.g.,Iwantedtotakeapicture onPolaroid,butitcosttoomuch;IwantedtorecordasongonPolaroid,butitcosttoomuch;I wantedtoserveatrifleonPolaroid,butitcosttoomuch)thatcouldonlybefullyunderstoodifthey madeaninferencethatwasbasedonextendingthemeaningassociatedwiththebrandname.The threeconditionswerestandardbrandusage,theplausiblebrandextension,andtheimplausiblebrand extension.Stewartandcoauthorspredictthatplausiblebrandextensionswillcauselittledifficulty duringinitialreading,butthatimplausibleextensionswillcauseimmediatedisruption.Eyemovementsofeighteenparticipantswhowerereadingthesesentencesweretracked.Thesentenceswere classifiedintobrandregion,spilloverregion(immediatelyfollowingthebrand),andend-of-sentence region(theremainderofthesentence).Themeasuresusedwere,amongothers,regression-pathtime (thegazefromthetimetheregionisfirstenteredtothetimeitislastexited),probabilityoffirst-pass regression(proportionoftimesinwhichtheregionisexitedtotheleftafterafirst-passfixationin theregion),andtotalgazetimeontheregion. Analysesofthesemeasuresconfirmedthepredictionsandyieldedtwomainresults.First,therewas
142
MICHEL WEDEL AND RIK PIETERS
evidenceofdifficultyinprocessingtheimplausibleconditionincomparisonwiththenormalusage condition,asreflectedin200millisecondslongerviewingtimesonaverage,whichissubstantial. Second,theresultsshowedthatevenplausibleextensionsledtominorreadingdifficulties,inparticular laterduringtheprocessingofthesentence.Theauthorsinterpretthistoresultfromtheconstruction ofanewinterpretationassociatedwiththebrand.Implausibleextensions,incontrast,causeimmediatedisruption,resultingfromthedifficultyofprovidingacoherentinterpretation.Thestudyshows thateye-tracking,duetoitshightemporalresolution,revealstheeasewithwhichparticipantscan interpretdifferentbrandextensions.Theanalysis,inwhichmeasurementsandmetricsarederived fromsubstantivetheoryontheunderlyingcognitiveprocess,providesacompellingdemonstration oftheextenttowhichconsumersfindparticularbrandextensionsplausible. Conclusion Thisreviewdemonstratesthatsincethefirsteye-trackingstudiesinmarketingbyNixon(1924),a sizablebodyofliteraturehasaccumulatedusingeye-trackingtechnologytoassessvisualmarketingeffectiveness.Withregardtobottom-upfactors,thatbodyofliteraturehasdemonstratedthe effectsofspatiallocationofvisualmarketingstimuli,theeffectofbasicfeaturessuchascolor, lines/edges,movement,andinparticular,size,ofspecificobjects,andofcontext.Top-downeffectsthathavebeendemonstratedincludethoseofmemory,involvement,attitudes,processing states,emotions,goals,andexpertise.And,theseattentionaleffectshavebeendemonstratedfor exploration,search,andchoicetasks.Butthebodyofevidenceonthesefactorsisfarfromcomprehensiveorconclusive. The research does reveal the power of eye-tracking to the development and growth of the visualmarketingdisciplineasaunifiedmethodologytotrackvisualmarketingeffectivenessin manyareasofmarketing.First,eye-trackingprovidesinsightsintocommunicationprocessingand effectivenessthatcannotbeobtainedbytraditionalmeasuresbecauseofthespeedandlackof consciousaccesstotherapidattentionalprocessestakingplaceduringcommunicationexposure. For instance, the opportunities to unintrusively monitor consumers’ visual attention to visual marketingstimulifrommomenttomomentintimerevealedtheinfluenceofinformationintake duringeyefixationsonbrandconsiderationandpreferenceduringstimulus-basedchoice,and theswitchingbetweenattentionstatesduringadprocessing.Second,eye-trackingcomplements traditionalmeasuresbyprovidingmoredetailaboutthespatio-temporaldynamicsofattentionand ittherebyenablesdevelopingandtestingintegratedmodelsofvisualcommunicationprocessing andeffectiveness.Forexample,combiningeye-trackingmeasureswithtraditionalverbalmeasuresofbrandmemoryallowedstructuralmodelsoftheattentionstorageandretrievalprocess, andtestsoftheinfluenceofattentionenhancementandinhibitioninWebadvertising.Third,eyetrackingresearchchallengesreceivedknowledgeaboutthedeterminantsofvisualcommunication performancethathavebeenobtainedbytraditionalmeasuresofcommunicationprocessingand effectiveness.Forexample,suchresearchhasshownthatcurrentideasabouttheinfluenceofthe contentandsizeofadelementsonattentioncaptureandtransfer,andabouttheorderofattending toadobjects,needupdating. Yet,severalareasremainunder-researched,andweidentifyfiveimportantoneshere.Firstand foremost,moreworkisneededthatapplieseye-trackingmeasurestomakereverseinferencesabout fundamentalcommunicationprocesses.Todate,eye-trackingresearchhasoftenbeendescriptive, forinstance,byrelatingperceptualaspectsofthecommunicationstimulus,suchasthesize,color, orpositionofthebrand,text,orpictorial,directlytomeasuresofvisualattention.Whilethishas ledtomanynewinsights,muchmorecanbegainedfromusingeye-trackingmeasuresasindicators
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 143
ofthelatentprocessesofinterest,takingastructural-theory-drivenapproachtomodelingthem. Suchareverseinferenceapproach(Feng2003),inwhichkeyindicatorsofunobservablecognitive processesarederivedfromtheobservedeyemovements,willnotonlylikelyleadtomoreinsights intovisualmarketingeffectiveness,butwillalsoenablerealistictestsofattentiontheoriesfrom cognitiveneuroscienceandperceptionpsychologythatsofarhavehadtorelyonabstractstimuli andimpoverishedexposureconditions.Webelievethatintegratingeye-trackingdatainareverse inferenceapproachholdsmuchpromiseforthetheoryandpracticeofvisualmarketing. Second,moreresearchabouttheinfluenceofstimulusinformativenessonvisualattentionis calledfor.Todate,eye-trackingresearchinmarketinghastypicallyexaminedtheinfluenceof salience,asastimulus-characteristic,residinginlocalcontrastsinperceptualfeaturessuchassize, luminance,color.Asaconsequence,muchlessisknownabouttherolethattheinformativenessof objectstotheconsumers’goalsplayinattentioncaptureandretention.Carefullydesignedexperimentsandstatisticalmodelsneedtobecombinedtodisentangleeffectsofattentiontofeatures, space, and objects in commercial scenes from observed eye movements. Evidently, top-down factorssuchasconsumertraitsandstateshaveanimpactonattentionpatterns,andsomestates mayevenbeprimedbottom-up(endogenously)bycertainobjectsinthevisualmarketingstimuli themselves,suchaswhenattentiontoasnackbrandinaretailer’sbillboardprimesaspecific weight-lossgoalinconsumers,orwhenthetextinprintadsprimesamoregeneralbrand-learninggoal.ThepioneeringworkofAlfredYarbus(1967)ontheinfluenceofgoalsontheattention patternsisalandmarkthatcanbebuildupon.Thefinalattentionpatternsofconsumersacross visualmarketingstimuliaretheproductofthesalienceandtheinformativenessofobjectstothe consumer,andweknowverylittleaboutthelatter.Moreresearchontheinterplaybetweensalience andinformativenessinattentiontovisualmarketingstimuli,andontherolethatconsumergoals playinthisprocess,isclearlyneeded. Third,moreworkiswelcomeonotherstaticvisualmarketingstimulibesidesprintads,and ondynamicstimulusmodes,suchastelevisionandWebsitesandadvertising.Whereasresearch todatehastaughtusmuchaboutvisualattentiontoprintads,lessisknownaboutattentionto otherstaticvisualmarketingstimuli,suchasproductandbrandpackages,shelfdesignandlayout,point-of-purchasematerial,outdoorcommunication.Also,nexttonothingisknownabout attentiontotelevisioncommercials,productplacement,in-scriptsponsoring,andsimilardynamic visualmarketingstimuli.Issuesoftheoptimallengthandpacingoftelevisioncommercials,or optimizingbrandexposuresinthem,arereadytobestudied.Becauseofthe“doublydynamic” characterofsuchdata,thisresearchwillrequiremajortheoreticalanddataanalyticefforts,but therewardsarelikelytobehighaswell,inviewofthedominantroleofsuchdynamicstimuli inmarketingpractice.Inasimilarvein,moreresearchonWebsitesandadvertisingisawaited, whichmayevenbemorechallenging,becausestaticanddynamicmodesaremixedinthem,but thepossibleinsightstobegleanedarelikelytobeequivalentlyrewardingaswell.Suchwidening thescopeofthestimulusmodesunderresearchscrutinyseemsparticularlyfruitful. Fourth,thetimeisripetodevelopandteststandardizedattentionmetricstogaugetheeffects ofvisualmarketing.Alargesetofeye-trackingmeasureshasbeenusedinpreviousmarketing research,suchasgazeduration,fixationduration,fixationfrequency,firstfixation,andthelike. Originally such disaggregate, micro measures were often developed in the context of reading research,andtheyhavebeenshowntobeusefulaswellforvisualmarketing.Yet,toadvance theapplicationofeye-trackingresearchinmarketingpractice,itseemsimportanttoarriveata smallersetofmetrics—aggregatedoverconsumers,time,and/orspace—thattapkeyoutcomes oftheattentionprocessofinterestandthatareattunedtotheneedsofvisualmarketingpractice. Similartomarketingmetricssuchasconversionratiosorpriceelasticities,attentionmetricsmight
144
MICHEL WEDEL AND RIK PIETERS
becomestandardsintheindustrythatmarketerscancomparetobenchmarksinordertomanage andassesstheirvisualmarketingefforts.Webelievethatthedevelopmentofstandardizedattentionmetricswillacceleratethefurtherdevelopmentofvisualmarketing. Finally,althoughthereisevidencethatevensmallattentionaleffectsoncommercialvisual stimuli,suchasprintads,catalogues,shelves,labels,andpackagingdesigns,affectmemory,attitudes,preferences,andintentions(Fletcheretal.1995;Janiszewski1998;Krugmanetal.1994; PietersandWarlop1999;Pieters,Warlop,andWedel2002;Rosbergen,Pieters,andWedel1997; TreistmanandGregg1979;WedelandPieters2000),moreresearchisneededthatinvestigatesthe relationshipofattentiontodownstreammarketingeffects.TreistmanandGregg(1979)observed thatoneoftwoadsthatpeoplelookedatlongerinaneye-trackingexperimentalsoreceivedmore sales.Inviewoftherapidincreaseofeye-trackingapplicationsinpracticeandtheincreasing clutterinadvertisingpractice,investigatingtherolethatvisualattentionplaysinstimulatingsales wouldfillanimportantgapincurrentknowledge. Inconclusion,eyemovementsundernormalconditionsarestronglyanddirectlyconnected tohigh-ordercognitiveprocesses,andattentionplaysamuchmorecentralroleincommunicationprocessingandeffectivenessthanpreviouslybelieved.Researchtodatehasconfirmedthe effectsofahostofbottom-upandtop-downfactorsontheattentionpatternstoarangeofcommercialstimuli.Thecurrentopportunitiestocollecteye-trackingdataonlargesamplesofstimuli andconsumersatcomparativelylowcostsareunprecedented.Eyemovementsmaycontribute significantlytovisualmarketingreachingitsfullpotential,byprovidinganexceptionalviewon consumers’moment-to-momentprocessingofvisualmarketingstimuli,withpredictivevalidity fordownstreameffects.Andthereisnowasubstantialbodyofresearchforfuturevisualmarketingresearchtobuildon. Acknowledgments TheauthorsthankAnochaAribargandUlfBöckenholtforusefulcomments. Notes 1.www.Tobii.com. 2.http://poynterextra.org/eyetrack2004/main.htm,lastaccessedApril3,2006.
References Anstis,S.M.1974.“AChartDemonstratingVariationsinAcuitywithRetinalPosition.”VisionResearch 14:589–592. Aoki,Hirotaka,andKenjiItoh.2000.“AnalysisofCognitiveAttitudestoCommercialFilmsonBasisof Eye-TrackingData.”ProceedingsofHumanFactorsandErgonomicSociety1:38–41. ———.2001.“AnalysisofCognitiveProcessesDuringViewingTelevisionCommercialsBasedonSemantic StructureofScenesandEye-MovementData.”JournalofJapanIndustrialManagementAssociation52 (2):101–116. Biel,AlexanderL.1993.“AdResearchintheUS:HurdleorHelp?ABriefReviewoftheStateoftheArt.” AdMap329(May):27–29. Burke,R.R.,andT.K.Srull.1988.“CompetitiveInferenceandConsumerMemoryforAdvertising.”Journal ofConsumerResearch15(1):55–68. Buswell,G.T.1935.HowPeopleLookatPictures:AStudyofthePsychologyofPerceptioninArt.Chicago: UniversityofChicagoPress. Chandon,Pierre.2002.“DoWeKnowWhatWeLookAt?AnEye-TrackingStudyofVisualAttentionand MemoryforBrandsatthePointofPurchase.”WorkingPaper,INSEAD,Fontainebleau.
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 145 Cowen,Laura.2001.“AnEyeMovementAnalysisofWeb-PageUsability.”M.A.thesis,LancasterUniversity,UK. Davenport,ThomasH.,andJohnC.Beck.2001.TheAttentionEconomy:UnderstandingtheCurrencyof Business.Boston,MA:HarvardBusinessSchoolPress. Dreze,Xavier,andFrançois-XavierHussherr.2003.“InternetAdvertising:IsAnybodyWatching?”Journal ofInteractiveMarketing17(4):8–23. Duchowski,AndrewT.2002.“ABreadth-FirstSurveyofEye-TrackingApplications.”BehaviorResearch Methods,Instruments,andComputers34(4):455–470. ———.2003.Eye-TrackingMethodology:TheoryandPractice.NewYork:SpringerVerlag. d’Ydewalle,G.,G.Desmet,andJ.VanRensbergen.1998.“FilmPerception:TheProcessingofFilmCuts.” InEyeGuidanceinReadingandScenePerception,ed.G.Underwood,357–368.Amsterdam:Elsevier. d’Ydewalle,G.,andF.Tamsin.1993.“OntheVisualProcessingandMemoryofIncidentalInformation: AdvertisingPanelsinSoccerGames.”VisualSearch2:401–408. d’Ydewalle, G., P.VandenAbeele, J.Van Ronsgerger, and D. Concke. 1988. “Incidental Processing of AdvertisementsWhileWatchingSoccerGameBroadcasts.”InPracticalAspectsofMemory:Current ResearchandIssues,ed.M.M.Gruneberg,P.E.Morris,andR.N.Sykes,470–483.Vol.1.Memoryin EverydayLife.Chichester,NY:Wiley. Feng,Gary2003.“FromEyeMovementtoCognition:TowardaGeneralFrameworkofInference.CommentaryonLiechtyetal.2003.”Psychometrika68:551–556. Fischer,PaulM.JohnW.Richards,EarlJ.Berman,DeanM.Krugman.1989.“RecallandEye-Tracking StudyofAdolescentsViewingTobaccoAdvertisements.”JournaloftheAmericanMedicalAssociation 261(1):84–89. Fitts,P.M.,R.E.Jones,J.L.Milton.1950.“EyeMovementsofAircraftPilotsduringInstrument-Landing Approaches.”AeronauticalEngineeringReview9(2):24–29. Fletcher,JamesE.,DeanM.Krugman,RichardJ.Foit,PaulFischer,andTinaRojas.1995.“MaskedRecall andEye-TrackingofAdolescentsExposedtoCautionaryNoticesinMagazineAds.”MarketingandPublic PolicyConferenceProceedings,ed.P.S.EllenandP.J.Kaufman,128–135.Vol.5.Atlanta:GeorgiaState UniversityPress. Fox,RichardJ.,DeanM.Krugman,JamesE.Fletcher,andPaulM.Fischer.1998.“Adolescents’AttentiontoBeerandCigarettePrintAdsandAssociatedProductWarnings.”JournalofAdvertising27(3): 57–68. Goldberg,JosephH.,ClaudiaK.Probart,andR.E.Zak.1999.“VisualSearchofFoodNutritionLabels.” HumanFactors(September):425–437. Groves,RobertM.,FloydJ.Fowler,Jr.,MickP.Couper,JamesM.Lepkowski,EleanorSinger,andRoger Tourangeau.2004.SurveyMethodology,Hoboken,NJ:Wiley-Interscience. Harris,C.M.1993.“OntheReversibilityofMarkovScanninginFreeViewing.”VisualSearch2:123–135. Henderson,JohnM.,andAndrewHollingworth1998.“EyeMovementsDuringSceneViewing:AnOverview.”InEyeGuidanceinReadingandScenePerception,ed.GeoffreyUnderwood,269–293.Amsterdam: Elsevier. Jacob,R.J.K.,andK.S.Karn.2003.“Eye-TrackinginHuman-ComputerInteractionandUsabilityResearch: ReadytoDeliverthePromises.”InTheMind’sEye:CognitiveandAppliedAspectsofEyeMovement Research,ed.R.Radach,J.Hyona,andH.Deubel,573–605.Boston:North-Holland/Elsevier. Janiszewski,Chris.1998.“TheInfluenceofDisplayCharacteristicsonVisualExploratorySearchBehavior.” JournalofConsumerResearch25(December):290–301. Josephson,Sheree,andMichaelE.Holmes.2002.“VisualAttentiontoRepeatedInternetImages:Testing theScanpathTheoryontheWorldWideWeb.”ProceedingsofETRA2002,Eye-TrackingResearch& ApplicationsSymposium,43–49. Karslake,J.S.1940.“ThePurdueEye-Camera:APracticalApparatusforStudyingtheAttention-Valueof Advertisements.”JournalofAppliedPsychology24:417–440. Keller,KevinL.1991.“MemoryandEvaluationEffectsinCompetitiveAdvertisingEnvironments.”Journal ofConsumerResearch17(March):463–476. Kent, Robert. 1993. “CompetitiveVersus Noncompetitive Clutter in TelevisionAdvertising.” Journal of AdvertisingResearch33(2):40–46. Kroeber-Riel,W.1979.“ActivationResearch:PsychobiologicalApproachesinConsumerResearch.”Journal ofConsumerResearch5(March):240–250. Krugman,DeanM.,R.J.Fox,J.E.Fletcher,P.M.Fischer,andT.H.Rojas.1994.“DoAdolescentsAttendto
146
MICHEL WEDEL AND RIK PIETERS
WarningsinCigaretteAdvertising?AnEye-TrackingApproach.”JournalofAdvertisingResearch34: 39–52. LaBerge,David.1995.AttentionalProcessing:TheBrain’sArtofMindfulness.Cambridge:HarvardUniversityPress. Laughery,KennethR.StephenL.Young,KentP.Vaubel,andJohnW.Brelsford,Jr.1993.“TheNoticeabilityof WarningsonAlcoholicBeverageContainers.”JournalofPublicPolicyandMarketing12(1):38–56. Leven,Wilfried. 1991. Blickverhalten von Konsumenten: Grundlagen, Messung und Anwendung in der Werbeforschung.Heidelberg:Physica-Verlag. Liechty,John,RikPieters,andMichelWedel.2003.“GlobalandLocalCovertVisualAttention:Evidence fromaBayesianHiddenMarkovModel.”Psychometrika68(4):519–541. Lohse,GeraldL.1997.“ConsumerEyeMovementPatternsonYellowPageAdvertising.”JournalofAdvertising26:61–73. Lohse, Gerald L., and D.J.Wu. 2001. “Eye Movement Patterns on ChineseYellow PagesAdvertising.” ElectronicMarkets11(2):87–96. Mulvihill, Geoff. 2002. “Campbell Upgrades Soups to Battle Its Competitors.”Available at www.freep. com/features/food/ffill31_20021231.htm. Nixon,H.K.1924.“AttentionandInterestinAdvertising.”ArchivesofPsychology72:5–67. Noton,D.,andL.Stark.1971.“EyeMovementsandVisualPerception.”ScientificAmerican224:34–43. Pieters,Rik,EdwardRosbergen,andMichelWedel.1999.“VisualAttentiontoRepeatedPrintAdvertising: ATestofScanpathTheory.”JournalofMarketingResearch36(November):424–438. Pieters,Rik,andLukWarlop.1999.“VisualAttentionduringBrandChoice:TheImpactofTimePressure andTaskMotivation.”InternationalJournalofResearchinMarketing16:1–17. Pieters,Rik,LukWarlop,andMichelWedel.2002.“BreakingThroughtheClutter:BenefitsofAdvertisement OriginalityandFamiliarityonBrandAttentionandMemory.”ManagementScience48(6):765–781. Pieters,Rik,andMichelWedel.2004.“AttentionCaptureandTransferinAdvertising:Brand,Pictorialand Text-SizeEffects.”JournalofMarketing68(April):36–50. Pieters,Rik,andMichelWedel.2007.“GoalControlofAttentiontoAdvertising:TheYarbusImplication.” JournalofConsumerResearch34(August):224–233. Pieters,Rik,MichelWedel,andJieZhang.2007.“OptimalFeatureAdvertisingUnderCompetitiveClutter.” ManagementScience,forthcoming. Poffenberger,AlbertT.1925.PsychologyinAdvertising.Chicago:A.W.Shaw. Rayner, Keith. 1998. “Eye Movements in Reading and Information Processing: 20Years of Research.” PsychologicalBulletin124(3):372–422. Rayner,Keith,KarenM.Rotello,AndrewJ.Stewart,JessicaKeir,andSusanA.Duffy.2001.“Integrating Text and Pictorial Information: Eye MovementsWhen Looking at PrintAdvertisements.” Journal of ExperimentalPsychology:Applied7(1):219–226. Redline,CleoD.,andChristopherP.Lankford.2001.“Eye-MovementAnalysis:ANewToolforEvaluatingtheDesignofVisuallyAdministeredInstruments(PaperandWeb).”PaperpresentedattheAmerican AssociationforPublicOpinionResearchmeeting,Montreal,Canada,May. Reichle,E.D.andNelson,J.R.2003.“Localvs.globalattention:Aretwostatesnecessary?Commenton Liechtyetal.2003.”Psychometrika,68,543–549. Rimey,R.D.,andC.M.Brown.1991.“ControllingEyeMovementswithHiddenMarkovModels.”InternationalJournalofComputerVision7(1):47–65. Rizzolatti,Giacomo,L.Riggio,andB.M.Sheliga.1994.“SpaceandSelectiveAttention.”InAttentionand Performance,ed.C.UmiltàandM.Moscovitch,231–265.Cambridge,MA:MITPress. Rosbergen,Edward,RikPieters,andMichelWedel.1997.“VisualAttentiontoAdvertising:ASegment-Level Analysis.”JournalofConsumerResearch24(December):305–314. Russo,J.Edward.1978.“EyeFixationsCanSavetheWorld:ACriticalEvaluationandaComparisonbetween EyefixationsandOtherInformationProcessingMethodologies.”InAdvancesinConsumerResearch, ed.H.K.Hunt.AnnArbor,MI:AssociationforConsumerResearch,561–570. Russo,J.Edward,andFranceLeclerc.1994.“AnEye-FixationAnalysisofChoiceProcessesforConsumer Nondurables.”JournalofConsumerResearch21(September):274–290. Russo, J. Edwards, and Larry D. Rosen. 1975. “An Eye FixationAnalysis of Multialternative Choice.” MemoryandCognition3(3):267–276. Starch,Daniel.1923.PrinciplesofAdvertising.Chicago:A.W.Shaw.
A REVIEW OF EYE-TRACKING RESEARCH IN MARKETING 147 Stark,L.W.,andH.C.Ellis.1981.“ScanpathsRevisited:CognitiveModelsDirectActiveLooking.”InEye Movements:CognitionandVisualPerception,ed.D.F.Fisher,R.A.Monty,andJ.W.Senders.Hillsdale: LawrenceErlbaumAssociates. Stewart,A.J.,M.J.Pickering,andP.Sturt.2004.“UsingEyeMovementsDuringReadingasanImplicit MeasureoftheAcceptabilityofBrandExtensions.”AppliedCognitivePsychology18:697–709. Strong,EdwardK.1920.“TheoriesofSelling.”JournalofAppliedPsychology9(February):75–86. Treistman,Joan,andJohnP.Gregg.1979.“Visual,Verbal,andSalesResponsestoPrintAds.”Journalof AdvertisingResearch19(4):41–47. Ungerleider,L.G.,andM.Mishkin.1982.“TwoCorticalVisualSystems.”InTheAnalysisofVisualBehavior, ed.D.Ingle,R.J.W.Mansfeld,andM.S.Goodale,549–586.Cambridge,MA:MITPress. vanRaaij,FredW.1977.“ConsumerInformationProcessingforDifferentInformationStructuresandFormats.”InAdvancesinConsumerResearch,ed.W.D.Perreault,Jr.,176–184.Vol.4.Atlanta:Association forConsumerResearch. Wedel,Michel,andRikPieters.2000.“EyeFixationsonAdvertisementsandMemoryforBrands:AModel andFindings.”MarketingScience19(4):297–312. Wedel,M.,Pieters,R.,&Liechty,J.(2003).“Evidenceforcovertattentionswitchingfromeye-movements; ReplytoCommentariesonLiechtyetal.2003.”Psychometrika,68(4),557–562. Witt, Dieter. 1977. “EmotionalAdvertising: The Relationship Between Eye-Movement Patterns and Memory—EmpiricalStudywiththeEye-MovementMonitor.”Ph.D.,UniversityofSaarland.Described inKroeber-Riel1979. Yarbus,AlfredL.1967.EyeMovementsandVision.NewYork:PlenumPress. Young,L.R.,andD.Sheena.1975.“SurveyofEyeMovementRecordingMethods.”BehavioralResearch MethodsandInstrumentation7:397–429.
CHAPTER6
ROLETHEORYAPPROACHESFOR EFFECTIVENESSOFMARKETING-ORIENTED BOUNDARYSPANNERS ComparativeReview,ConfiguralExtension,and PotentialContributions JAGDIPSINGHANDARGUNSAATCIOGLU
Abstract Roletheoryhasprovedremarkablypromisinginexaminingeffectivenessofmarketing-oriented boundaryspanners.Thispaperreviewsdifferentapproachesforexaminingroletheoryimplicationsforboundaryspanners—namelyuniversalisticandcontingencyapproaches—anddevelops theconfiguralapproachbyextendingconfigurationaltheoryprinciplestoroletheory.Neitherthe contingencynortheconfiguralapproachhasreceivedmuchattentioninthemarketingliterature. Wecompareandcontrastdifferentapproaches,outliningbodiesofworkthathaveremainedless accessibletomarketingresearchers.Bytriangulatingacrossthealternativeapproaches,weexpose underlyingassumptionsandpressforcriticalassessmentoftheirecologicalvalidity.Weidentifyopportunitiesforpotentialcontributionsbyexploringpromisingbutasyetunchartedapproaches. Marketingorientedboundaryspannerssuchassalespeople,frontline,andcustomercontactemployeesfillcriticalrolesthatinfluenceorganizationaleffectivenessandsustainability.Consider thefollowing: • Boundaryspannersarestrategicallyimportantbecausetheyrepresentthe“face”oftheorganizationtocustomersandpublic,andarecriticalnodeswhereknowledgeaboutmarkets andconsumersisaccumulated. • Boundaryspanningworkisrarelyroutinizedandinvolvessignificantpeople-orientedwork. Boundaryspannersarerequiredtoconstantlyinteractwithcustomers,undertaketasksthat involveemotionallabor,andprovidediscretiontotailortheirbehaviorstoindividualcustomer needs,problems,anddemands. • Boundaryspanningworkissensitivetointernalandexternalorganizationalenvironments. Variationinconsumerdemands(e.g.,seasonaland/oreconomicvariationsindemandfor products/services) and in internal operations (e.g., new product/service introductions or interfacetechnologies)oftenaffectsboundaryspannersunpredictably. 148
EFFECTIVENESS OF MARKETING-ORIENTED BOUNDARY SPANNERS 149
• Boundaryspannersareorganizationallymonitoredandcontrolled(e.g.,viahumansupervision,electronic,audio,andvideodevices).Organizationsareincreasinglyconcernedabout theproductivityofboundaryspanners,whilekeepingthequalityofcustomerservicedelivered infocus. • Boundaryspanningworkishighlystressful.Suchworkislikenedtoa“three-corneredfight” withthecustomer(demandingattentionandservice)andtheorganization(demandingefficiencyandproductivity)atthetwoendsandtheboundaryspanner“caughtinthemiddle.” • Boundary spanning roles are profit centers.They are expected to cross-sell, up-sell, and more-sellwhileintheprocessofprovidinghigh-qualityservice/information.Thisdualaccountabilityinjectscompetingpressuresonboundaryspanners. Ofthevarioustheoriesappliedtostudyeffectivenessofmarketing-orientedboundaryspanners, roletheoryisarguablythemostpromisingsofar.Withitsrootsinsociologydatingbacktothe fifties(Merton1949;Rommetveit1954),earlygroundedresearchonworkorganizationscanbe tracedtothesixties.Themuch-citedworkofKahnetal.(1964)andBelasco’s(1966)researchwith salespeoplewereimportantstepsintranslatingsociologicalnotionsofroletheoryintomeaningfulandrelevantconstructsforthestudyofmarketing-orientedboundaryspanners.Forinstance, independentofKahnetal.(1964),Belasco(1966)madesomeimportantobservationsonthedifferentroledemandsexperiencedbysalespeople:(a)intellectualdemandsthatrequireintelligence, problem-solvingskills,andjobknowledgeabilities,(b)emotionaldemandsfromdealingwith issuessuchas“advocacyconflict”—internalconflictfrombeinganadvocateforthecustomer andthecompanyatthesametime,and(c)interactionaldemandsthatarisefromtheintensityand adaptabilityrequiredinthediverserangeofinteractions.Withouteffectivecopingmechanisms, Belascofearedthatsalespersoneffectivenesswouldbeseriouslyunderminedregardlessoftheir intelligence,jobknowledge,and/orskills.AlthoughKahnetal.andBelascoprovidedrichtheory forprobingboundaryspanningroles,scientificprogresslingeredtillRizzo,House,andLirtzman (1970)publishedvalidatedscalesofroleconflictandambiguity,andstirredupresearchinterestin thistopic.By1985,JacksonandSchulerreportedover200articlesonroleconflictandambiguity inorganizationalsettingsthatwerepublishedbetween1970and1983;ofthese,96wereoriginal empiricalstudiesthattheymeta-analyzed.Afewyearslater,BrownandPeterson(1993)were abletolocate59studiesthatfocusedspecificallyonsalespersonroleconflictandambiguityand itsinfluenceonperformanceandsatisfaction. Despitethisvolumeofresearch,academicandpractitionerperspectivesonboundaryrolestress aredefinedbyconvergenceandcontrasts.Bothperspectivesconvergeontheviewthatboundaryrolestressorsincurheavycostsfortheorganizationandindividualalikebecauseoflowered productivity,reducedmotivationandcommitment,andincreasedhealthcosts(Cavanaughetal. 2000;MaslachandLeiter1997).Insomeprofessions,especiallyinvolvingfrontlineandcustomer contactwork,stressorshavebeendescribedasreachingepidemicproportions(Marino1997). Contrastingperspectivesemergewhentheinfluenceofboundaryrolestressorsisconsidered.In contrasttotheconvergenceintheacademicliterature,practitionershavelongarguedaboutthe potentialofboundaryrolestressorstopromoteperformance,enhancemotivation,andsparkcreativity(MohrmanandCohen1995;Newton1995).Withregularity,thepopularpresshasfancied workplacemantrassuchas,“itisbettertoburnoutthantorustout,”presumablytoassureboundary spannersthattheyarenotaloneinfacingstressandthatstresscanbeturnedintoanopportunity todevelopandenhanceoneself.Paradoxically,whilethisnotionof“eustress”hasdeeprootsin theacademicliterature(Selye1976;YerkesandDodson1908),empiricalstudieshavegenerally producedweakandmixedevidence.Thus,whilemuchacademicresearchsuggestsredesigning
150 JAGDIP SINGH AND ARGUN SAATCIOGLU
andreconfiguringboundaryworkinawaythatreduces,ifnoteliminates,criticalrolestressors (TubreandCollins2000),practitionersviewsuchrecommendationswithlittlerelevancesince theyholdthatthenatureofcustomerinterface(e.g.,people-oriented,nonroutinizedwork)andits unpredictability(e.g.,variabilityininternalandexternalconditions)makerolestressaninherent aspectofboundaryroles. Tobridgetheseperspectives,apromisingapproachhasbeenproposedbyKarasek(1979)and hiscolleaguesthatviewsasingularfocusonrolestressorsasmyopicandmisguided.Instead,it arguesthatthestudyofrolestressorsmustsimultaneouslyconsiderthejobscope—thedegreeof autonomy,feedback,andparticipationaffordedtoboundaryroleemployees(Karasek1979;Xie andJohns1995).Notingthatgreaterjobscopemaymakeallthedifferencebetween“eustress”and “distress,”andbetween“healthy”and“unhealthy”work,thisapproachistheoreticallyappealing becauseofitsconceptualrichness,andmanageriallyattractiveasevidencedbythepopularityof empowermentprograms.Unfortunately,suchstressor–scopemodelshavereceivedlimitedempiricalattentioninthemarketingliterature.Assuch,thepotentialofthestressor–scopeframeworkto provideinsightsandbridgeperspectivesislargelyunrealized. Thepurposeofthisreviewistoilluminateandstrengthentheprecedingbridgetogerminatenew researchideasanddirectionsforunderstandingtheeffectivenessofmarketing-orientedboundary spanners.Specifically,weprovideareviewofthreedifferenttheoreticalperspectivesonboundary rolestressandeffectiveness,includinguniversalistic,contingency,andconfiguralperspectives. Theuniversalisticperspectivereflectsmuchcurrentresearchinmarketingandisgroundedinthe roleepisodemodelofKahnetal.(1964).Thecontingencyperspectiveisbasedoncontributions toKarasek’smodel.Becausemanyofthesecontributionshaveoccurredoutsidethemarketing literature,weprovideadetaileddiscussionofthetheoryunderlyingthisperspective,andreview theassociatedempiricalliterature.Finally,wedeveloptheconfiguralperspectiveasatheoretical contributionofthispaper.Thisperspectiveextendsideasfromconfigurationalapproachestoposit nonlinearandhigher-ordereffectsofroletheorythatcannotberepresentedbycontingencyapproaches.Whiletheseperspectiveshavecompetingelements,ourorientationiscomparativeand complementary.Wecomparetheseperspectivesfortheirtheoreticaldistinctivenesstoencourage futureresearchthatapproachesthestudyofboundaryrolestressandeffectivenessfrommultiple, notsingular,perspectivestouncoverconvergentandanomalousideas.Focusingonasingular theoreticalperspective,asreflectedinmuchmarketingliterature,limitsthevisionofunderstanding. Inaddition,attheirboundaries,thesetheoreticalperspectivesofferopportunitiesforinteresting andcreativeworkthathasremainedasyetuntappedandunexploited.Weprovideanoutlinefor futureresearchdirectionstothisend. TheoreticalPerspectivesonRoleStressorsandBoundarySpanningRoles UniversalisticPerspective Thisperspectivepositsthatrolestressorsinvariablyhavedysfunctionalconsequencesforboundary spanneroutcomesincludingperformance,satisfaction,andcommitment,regardlessofjobcontext, scope,and/ornatureoftheorganization. Theory RootedinKahnetal.’swork,thisperspectivepositsthatrolestressorshavealinearrelationship withboundaryspanneroutcomes.Thecommonlyexaminedrolestressorsincluderoleconflict,
EFFECTIVENESS OF MARKETING-ORIENTED BOUNDARY SPANNERS 151
roleambiguity,androleoverload(BehrmanandPerreault1984;Belasco1966;Kahnetal.1964). Roleconflictoccurswhenaboundaryspannerbelievesthattheexpectationsanddemandsoftwo ormoremembersofhisorherrolesetareincompatible(e.g.,bossandcustomer).Roleambiguityrelatestotheperceivedlackofinformationneededbyanemployeetoperformhisorherrole adequatelyandhisorheruncertaintyabouttheexpectationsofdifferentrolesetmembers.Role overloadoccurswhenthefrontlineemployeeperceivesthatthecumulativeroledemandsexceed hisorherabilitiesandmotivationtoperformatask.Theinfluenceofdifferentrolestressorson boundaryspannerperformanceandwell-beingissupportedconceptuallybytheroleepisodemodel ofKahnandcolleagues(1964),whichpositsthat(1)boundaryspannersinteractwithdifferent rolesenders(e.g.,customers,boss,co-workers)inmanyepisodestoobtaininformation,direction, taskdemands,andassistance;(2)rolesenderdemandsandexpectationstaketheformofperceived stressorswhenaboundaryspannerbelievesthatthereisconflict(e.g.,amongdemands),ambiguity (e.g.,aboutexpectations),oroverload(e.g.,ofdemandsandexpectations);(3)perceivedstressors areinfluencedbyaperson’spsychological,dispositional,andsociologicalcharacteristics;and(4) persistentstressorsarelikelytooverwhelmtheperson’sresourcesandtherebyhaveadysfunctional impactonhisorherbehavioralandpsychologicaloutcomes(e.g.,jobperformance,satisfaction). HobfollandFreedy(1993)haveconceptualizedtheinfluenceofrolestressorswithinaconservation ofresourcesframework.Boundaryspannersarethoughttoregulatetheirbehaviorstocopewith rolestressorsinawaythatconservestheirvaluedresources;however,regulationfailuresoccur whenstressorsoverwhelmanindividual’scopingresources,resultinginimpairedperformance andwell-being.Whileitallowsforthepossibilitythatdifferentboundaryspannersmayperceive disparatelevelsofrolestressorsinsimilarworksituations,thisperspectiveisuniversalisticinits predictionsaboutthelinearanddirecteffectofperceivedrolestressorsonoutcomes. EmpiricalFindingsandAssessment Thisperspectivehasproducedsignificantempiricalworksummarizedinseveralmeta-analyses andreviews(BrownandPeterson1993;FisherandGitelson1983;JacksonandSchuler1985; KingandKing1990).Thepreponderanceofevidencesuggeststhattheinfluenceofrolestressors isconsistent,compelling,andinvariablydysfunctional(BehrmanandPerreault1984;Brownand Peterson1993;FisherandGitelson1983;Rhoads,Singh,andGoodell1994;Singh1993;Spector,Dwyer,andJex1988).Forexample,BrownandPeterson(1993),intheirmeta-study,found correlationsof–.24,–.36,–.28,and.36betweenroleambiguityandjobperformance,satisfaction, commitment,andpropensitytoleave,respectively.Likewise,thecorrelationsbetweenroleconflict andjobperformance,satisfaction,commitment,andpropensitytoleavewere–.07,–.33,–.34, and.28respectively.Althoughroleoverloadisnotasfrequentlystudiedasotherrolestressors, ingeneralthecorrelationsbetweenroleoverloadandthedifferentjoboutcomesparallelthose obtainedforroleconflictandambiguityintermsofbothmagnitudeanddirection.Forinstance, Singh,Goolsby,andRhoads(1994)reportcorrelationsof–.14,–.39,.25,and.09betweenrole overloadandjobsatisfaction,organizationalcommitment,turnoverintentions,andjobperformance,respectively.Assuch,thereisbroadevidencesupportedbymeta-analyticalresultsthat rolestressorshavesignificantlinearanddysfunctionalrelationshipswithcriticaljoboutcomes (BrownandPeterson1993;FisherandGitelson1983;JacksonandSchuler1985). Whileresearchershavecalledforexploringmoderatingvariables,themajorityoftheempirical workhasdownplayedtheeffectofsituationalorcontextualvariability.Forinstance,inBrown andPeterson’smeta-analysis,theinfluenceofsupervisorybehaviorsandjob/taskvariablesaccountedforlessthan10percentofexplainedvariance.Consequently,BrownandPeterson1993,
152 JAGDIP SINGH AND ARGUN SAATCIOGLU Figure6.1 AGraphicalRepresentationofKarasek’sJobDemands(RoleStressors)– DecisionLatitudeModel Job Demands Low
High
Low-Strain
Job Latitude
Low
High
Passive
Active 3
4
1
2
High-Strain
(p.68)claimedthattheirrolestressmodelthatexcludesjobcontextvariablesindicates“considerablerobustnessandgeneralizability...acrossrelationshipsandstudycontexts.”Likewise, inanothermeta-analysis,Churchilletal.(1985,p.109)foundthat,onaverage,organizational contextvariablesexplain“only1%ofthevariationinperformance”andthatthisinfluencewas the“lowest...amongthesixcategoriesofpredictorsstudied.”ThesefindingsparallelJackson andSchuler’s(1985)meta-analysisinthatcontextualvariableshavesignificant,negativebut ratherweak(