Citation
Predicting social networking sites continuance intention

Material Information

Title:
Predicting social networking sites continuance intention Should I stay or should I go?
Creator:
Sibona, Christopher ( author )
Language:
English
Physical Description:
1 electronic file (317 pages). : ;

Subjects

Subjects / Keywords:
Online social networks ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Colorado Denver.
Bibliography:
Includes bibliographic references.
System Details:
System requirements: Adobe Reader.
General Note:
Department of Computer Science and Engineering
Statement of Responsibility:
by Christopher Sibona.

Record Information

Source Institution:
University of Colorado Denver
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
904622891 ( OCLC )
ocn904622891

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

PREDICTINGSOCIALNETWORKINGSITESCONTINUANCE INTENTION:SHOULDISTAYORSHOULDIGO? by CHRISTOPHERSIBONA B.S.,VirginiaPolytechnicInstituteandStateUniversity,1993 M.B.A.,UniversityofColoradoDenver,2007 Athesissubmittedtothe FacultyoftheGraduateSchoolofthe UniversityofColoradoinpartialfulllment oftherequirementsforthedegreeof DoctorofPhilosophy ComputerScienceandInformationSystems 2014

PAGE 2

2014 CHRISTOPHERSIBONA ALLRIGHTSRESERVED

PAGE 3

ThisthesisfortheDoctorofPhilosophydegreeby ChristopherSibona hasbeenapprovedforthe ComputerScienceandInformationSystemsProgram by DawnGregg,Chair JudyScott,Adviser IlkyeunRa ZhipingWalter 9/26/2014 ii

PAGE 4

Sibona,ChristopherPh.D.,ComputerScienceandInformationSystems PredictingSocialNetworkingSiteContinuanceIntention:ShouldIStayor ShouldIGo? ThesisdirectedbyAssociateProfessorJudyScott ABSTRACT Thisresearchdevelopsandtestsmodelstopredictcontinuanceintentiononsocialnetworkingsites.Themodelsaddsnewfactorswhicharerelevanttosocial networkingsitescontinuanceintention.Thesocialnetworkingsitecontinuance modeladdsvefactors:personalinnovativeness,habit,attitudetowardalternatives,interpersonalinuence,andconsumerswitchingcoststoenhancethe predictivepowerofinformationsystemscontinuance.Interpersonalinuence, alternativeperceptionsandproceduralandrelationalcostsaretheorizedto haveadirecteectoncontinuanceintention.Personalinnovativenessand habitaretheorizedtohaveadirectandmoderatingeectsoncontinuance intention.Theresultshavealargepositiveeectoftheexplanatorypowerin explainingmoreofthevarianceofcontinuanceintentiononasocialnetworking site.TheinformationsystemsIScontinuancemodelexplainsapproximately 66.8%ofthevarianceandthesocialnetworkingsitecontinuancemodelwith theveaddedfactorsexplains76.7%ofthevarianceandisconsideredtohave alargeeectintheexplainedvariance.Allofthefactorshavestatistical signicance;thefactorswiththelargestpathcoecientsare,inorder,satisfaction&perceivedusefulness =0.3686,consumerswitchingcosts = 0.2496,alternativeperceptions =-0.2069,habit =0.1642,personal innovativeness =-0.0589andinterpersonalinuence =-0.0451.Habit andpersonalinnovativeness,asmoderators,werenotstatisticallysignicant anddidnotsubstantiallyaidintheinterpretationofthefactors.Theresearch helpsexplainstherelevantfactorsforwhyusersofsocialnetworkingsiteswill iii

PAGE 5

continuetouseorabandonasite. Theformandcontentofthisabstractareapproved.Irecommendits publication. Approved:JudyScott iv

PAGE 6

DEDICATION Formywife,Christine:thankyouforyourloveandsupportthroughthis longandarduousjourney.Adventureisouttherenowlet'sgohaveanewone! Formyparents:thankyouforteachingmetheimportanceofeducation andencouragingmetofollowthispath.Formyfatherwhoencouragedmeto gotocollegeandpursueacademicopportunities.Formymotherwholearned introductorycomputerprogrammingwhenIwasinmiddleschoolandshowed meherprogramswhichwereprintedongiantspoolsofpaper.AtrstIwas justlookingforpapertodrawonandlaterIwantedtounderstandhowthe instructionsonthosepagesmadecomputerswork.Thankyouforallofthe sacricesyouhavemadeandallthatyouhavegivenme. v

PAGE 7

ACKNOWLEDGMENTS Ourchiefwantissomeonewhowillinspireustobewhatweknowwecould be. RalphWaldoEmerson, ConductofLife. Therearemanypeoplewhoencouragedmetocontinuemyeducation. Inhighschool,PeterG.Osbornewasaninspiringcomputerscienceteacher whotaughtmyrstintroductoryclassesofcomputerscience.Twoprofessors duringmyundergraduatestudiesatVirginiaTech,Prof.SallieM.Henry andProf.DennisKafura,werehelpfulinincreasingmyunderstandingof object-orientedsystemsandsoftwareengineeringprinciplesatamuchdeeper level.Ingraduateschool,Prof.SusanKeaveneyandDr.WendyGuildboth recommendedtometopursueadoctoraldegree;withtheirencouragementI setforthonthatpath.Dr.Guildwrotethemostpersonal,helpful,funnyand interestingrecommendationletterformygraduateschoolstudies. Iwanttothankmyfellowclassmateswhowerehelpfulindiscussionsabout theory,relevance,rigorandndingwhatisimportant.Specialthanksto classmates:JonBrickey,GaryBorkan,MohammadAlsharo,JaeHoonChoi, MichaelErskine,JehadImlawiandHosseinKalantar.Bestoflucktoyouall. Thankstomyadviser,Prof.JudyScott,foryoursupportandthoughtfulnessinreviewinganddirectingthisresearchintohelpfulchannels.Thanks tomycommitteemembersDeanDawnGregg,Prof.IlkyeunRaandProf. ZhipingWalterfortheirfeedbackandinsight.ThankstoProf.Madhavan Parthasarathywhotoldmenottowasteanytimeandjuststartdoingresearch atthatverymoment.ThankstoProf.PeterG.Bryantwhoencouragedme topersevere.Bryant'sadviceremindedmeofWinstonS.Churchill'swords, Itisnotenoughthatwedoourbest;sometimeswemustdowhatisrequired. Thankyouallforyourinspiration,encouragementandsupport. vi

PAGE 8

CONTENTS Chapter 1Introduction........................1 1.1ImportanceofTopic.......................3 1.2ResearchProblemandScope...................4 1.3ResearchQuestions........................5 1.4ResearchContribution......................6 1.5OutlineofDissertation......................7 2AcceptanceandContinuanceTheories.............8 2.1TechnologyAcceptanceandInformationSystemsContinuance9 2.2ConsumerSwitchingBehavior..................13 2.2.1ExploratoryResearchinCustomerSwitchingBehavior14 2.2.2ServiceProviderSwitchingModel............15 2.2.3CustomerSwitchingBehaviorforOnlineServiceProviders16 2.2.4ConsumerSwitchingCosts................17 2.2.5VarietySeekingBehavior.................20 2.2.6ConsumerBehaviorResearchApplicationsinInformationSystems........................22 2.3ISContinuanceAppliedtoSocialNetworkingSites......22 2.3.1Maslow'sHierarchyofNeeds...............22 2.3.2NetworkExternalities...................23 2.3.3SocialNetworkingSiteStickiness............23 2.3.4ISContinuance,SocialExchangeTheory,SocialCapital Theory&FlowTheory..................24 2.3.5SatisfactionScoresofSocialNetworkingSites.....25 vii

PAGE 9

2.3.6NamingtheDependentVariableSocialNetworkingSite ContinuanceIntention..................26 2.4SummaryofTheories.......................26 3SocialNetworkingSites...................30 3.1SocialNetworkingSitesToday..................30 3.2SocialNetworkingSitesinDecline................32 3.2.1CommunitiesofPracticeandOnlineCommunitiesof Practice..........................33 3.2.2TakingBreaksonSocialNetworkingSites.......35 3.2.3EngagementLevelConcernsofFacebook.......36 4PredictingEnd-UserSatisfactionthroughPerceivedEase-of-Use andPerceivedUsefulness-Study1..............37 4.1LiteratureReview.........................38 4.2StudyDesign...........................41 4.2.1Survey...........................44 4.2.2DataCollection......................45 4.2.3DataAnalysis.......................46 4.3Results...............................47 4.4Discussion&Conclusion.....................52 5FactorsinSocialNetworkingSiteContinuanceIntention.....57 5.1ProductSubstitutionsandAlternativePerceptions......57 5.1.1AlternativeAttractivenessinConsumerServiceSwitching59 5.1.2Alternativeattractivenessappliedininformationsystemsresearch.......................62 5.2PersonalInnovativeness......................63 5.2.1PersonalInnovativenessHypotheses...........65 5.3InterpersonalInuences.....................65 viii

PAGE 10

5.3.1InterpersonalInuencesHypotheses...........67 5.4Habit................................67 5.4.1HabitHypotheses.....................69 5.5SwitchingCosts..........................69 5.5.1SwitchingCostsHypotheses...............71 6ResearchDesignandMethod................74 6.1InstrumentDesign........................74 6.2SurveyQuestions.........................75 6.2.1PerceivedUsefulness...................75 6.2.2Conrmation.......................76 6.2.3Satisfaction........................77 6.2.4Habit...........................78 6.2.5Personalinnovativeness..................79 6.2.6InterpersonalInuence..................80 6.2.7ProceduralandRelationalCosts.............81 6.2.8AlternativePerceptions..................87 6.2.9SocialNetworkingSiteContinuanceIntention.....91 6.2.10DemographicsAndScreening..............92 6.3SamplingAndParticipants....................93 6.4Analyticalmethods........................93 6.4.1GoodnessofFitMeasuresandPLS-SEM........94 6.4.2Higher-orderconstructs..................95 6.4.3Moderatingeects....................97 6.5Design...............................98 7Results..........................100 7.1CommonMethodVariance....................100 7.2Demographics...........................103 ix

PAGE 11

7.3ContinuanceDependentVariable................103 7.4OutlierAnalysis..........................106 7.5BaseModel-ISContinuanceModel...............112 7.5.1ModelDescription....................112 7.5.2MeasurementModel...................112 7.5.3StructuralModel.....................114 7.6Completemodel-NonModerated................118 7.6.1ModelDescription....................118 7.6.2MeasurementModel...................126 7.6.3StructuralModel.....................129 7.7Completemodel-ModeratingFactorAnalysis.........136 7.7.1ModelDescription....................136 7.7.2MeasurementModel...................138 7.7.3StructuralModel.....................141 7.7.4ModeratingEectAnalysis-Habit...........146 7.7.5ModeratingEectAnalysis-PersonalInnovativeness.148 7.7.6ModeratingEectAnalysis-TwoFactors-Habitand PersonalInnovativeness.................148 7.8CompleteNon-Moderatedmodel-BackwardStepwiseRenement151 7.8.1ModelDescription....................151 7.8.2Renements........................152 7.8.3MeasurementModel...................152 7.8.4StructuralModel.....................156 7.9Backwardstepwiserenementwitheectsizes.........160 7.10ThreeDatasetcomparisons....................167 7.11Summary.............................182 x

PAGE 12

8Discussion.........................187 8.1ISContinuanceTheory......................190 8.2Consumerswitchingcosts....................191 8.3AlternativePerceptions......................195 8.4Habit................................196 8.5PersonalInnovativeness......................198 8.6InterpersonalInuence......................200 8.7DemographicVariables......................201 8.8ImplicationsforResearchers...................203 8.9ImplicationsforPractice.....................206 9Limitations........................210 10Conclusion........................213 References ...........................215 Appendix ...........................229 ACommonMethodVariance.................229 BOuterLoadingAnalysis...................234 CSNSContinuanceandHabit.................238 C.1ModelDescription........................238 C.2MeasurementModel.......................238 C.3StructuralModel.........................240 C.4ModeratingEectAnalysis....................242 DSNSContinuanceandPersonalInnovativeness.........244 D.1ModelDescription........................244 D.2MeasurementModel.......................244 D.3StructuralModel.........................245 D.4ModeratingEectAnalysis....................249 xi

PAGE 13

ESNSContinuanceandInterpersonalInuence.........251 E.1ModelDescription........................251 E.2MeasurementModel.......................251 E.3StructuralModel.........................252 FSNSContinuanceandAlternativePerceptions.........257 F.1FourAlternativeSocialNetworkingSiteAccountImpact...257 F.2ModelDescription........................258 F.3MeasurementModel.......................260 F.4StructuralModel.........................260 GPredictingSNSContinuancethroughCosts..........265 G.1ModelDescription........................265 G.2MeasurementModel.......................270 G.3StructuralModel.........................272 HCompetingModelsCostvs.Satisfaction............276 H.1ModelDescription........................276 H.2MeasurementModel.......................276 H.3StructuralModel.........................279 ISpecicAlternativeProductEects..............285 I.1NumberofsocialnetworkingsiteAnalysis...........285 I.2AlternativesbySpecicProduct.................288 I.2.1Surveyrespondentswhousedallfouralternativeproducts288 I.2.2ModelDescription....................288 I.2.3MeasurementModel...................288 I.2.4StructuralModel.....................289 I.2.5Surveyrespondentswhousedatleastoneoffouralternativeproducts......................290 I.2.6ModelDescription....................290 xii

PAGE 14

I.2.7MeasurementModel...................292 I.2.8StructuralModel.....................293 xiii

PAGE 15

LISTOFTABLES Table 1TypologyofConsumerPerceptionsofSwitchingCosts....19 2InformationSystemsTheoreticalFoundation..........28 3ConsumerSwitchingTheoreticalFoundation..........29 4PewInternet-ReasonforFacebookBreaks...........33 5CompositeReliability.......................49 6Correlationoftheconstructs...................50 7Supportforhypothesesandtheeectsofthecontrolvariables51 8SocialNetworkingSiteContinuanceModelHypotheses....72 9Gender...............................103 10Age.................................104 11Education.............................105 12ContinuanceDescriptives.....................106 13Single-ItemContinuance.....................108 14BaseModel-ISContinuanceMeasurementModel.......114 15BaseModel-ISContinuanceMeasurementModel-DiscriminantValidity...........................114 16BaseModel-ISContinuanceStructuralModel.........115 17PathCoecients.........................116 18SNSCostMeasurementModel..................127 19CompleteModelwithModerators-DiscriminantValidity...128 20Indices...............................134 22SNSCostMeasurementModel..................139 23BaseModel-SNSContinuanceMeasurementModel-DiscriminantValidity...........................140 xiv

PAGE 16

24Indices...............................146 26Renement............................153 27SNSCostMeasurementModel..................154 28StaticallySignicantFactors-MeasurementModel-DiscriminantValidity...........................155 29Indices...............................156 31StepwiseRenementwitheectsize...............161 31StepwiseRenementwitheectsize...............162 31StepwiseRenementwitheectsize...............163 31StepwiseRenementwitheectsize...............164 31StepwiseRenementwitheectsize...............165 32CoecientofDeterminationComparisonAcrossThreeDatasets170 33ConstructComparisonacrossdatasets-1............171 34ConstructComparisonacrossdatasets-2............172 35ConstructComparisonacrossdatasets-3............173 36Comparisonacrossdatasets-4.................174 37PathComparisonsacrossthreedatasets-BootstrapSample.176 38PathComparisonsacrossthreedatasetsCohen'sd-Bootstrap Sample...............................179 39Cohen'sDCategorizationofPathCoecientDierences...180 40SocialNetworkingSiteContinuanceModelHypothesesEvaluated183 41SummaryTable..........................184 42GoodnessofFitcomparison...................186 43CommonMethodBiasAnalysis.................229 43CommonMethodBiasAnalysis.................230 43CommonMethodBiasAnalysis.................231 43CommonMethodBiasAnalysis.................232 xv

PAGE 17

43CommonMethodBiasAnalysis.................233 44OuterLoadingAnalysis.....................234 44OuterLoadingAnalysis.....................235 44OuterLoadingAnalysis.....................236 44OuterLoadingAnalysis.....................237 45BaseModel+HabitMeasurementModel............239 46BaseModel+HabitMeasurementModel-DiscriminantValidity239 47BaseModel+HabitStructuralModel.............240 49BaseModel+PersonalInnovativenessMeasurementModel..245 50BaseModel+PersonalInnovativeness-DiscriminantValidity246 51BaseModel+PersonalInnovativenessStructuralModel...247 53BaseModel+InterpersonalInuenceMeasurementModel..252 54BaseModel+InterpersonalInuenceMeasurementModelDiscriminantValidity.......................252 55SNSContinuanceStructuralModelandInterpersonalInuence253 57FourAlternativeSNSAccountImpacts.............257 58ISContinuanceMeasurementModelwithAlternativePerceptions260 59ISContinuanceMeasurementModelwithAlternativePerceptions-DiscriminantValidity...................261 60ISContinuanceStructuralModelandAlternativePerceptions262 62SNSCostMeasurementModel..................270 63SNSCostMeasurementModel-DiscriminantValidity....271 64Indices...............................272 66SNSCostMeasurementModel..................277 67SNSCostMeasurementModel-DiscriminantValidity....278 68Indices...............................279 70NumberofSocialNetworkingSitesUsed............286 xvi

PAGE 18

71SitesUsedMeasurementModel.................286 72SitesUsedStructuralModel...................286 73PathCoecients.........................287 74AlternativesbySpecicProductMeasurementModel.....289 75AlternativesbySpecicProductMeasurementModel-DiscriminantValidity...........................289 76AlternativesbySpecicProductStructuralModel.......290 77PathCoecients.........................291 78AlternativesbySpecicProductMeasurementModel.....293 79AlternativesbySpecicProductDiscriminantValidity....293 80AlternativesbySpecicProductStructuralModel.......294 81PathCoecients.........................295 xvii

PAGE 19

LISTOFFIGURES Figure 1TechnologyAcceptanceModel-Davis..........10 2InformationSystemContinuance-Bhattacherjee....13 3ResearchModel&ControlVariables..............44 4ModelPathCoecients&ControlVariables..........50 5SocialNetworkingSiteContinuanceModel...........73 6ContinuanceHistogram......................107 7InitialResidualAnalysis-Continuance.............110 8ResidualAnalysisAfterOutliersRemoved-Continuance...111 9BaseModel-ISContinuance..................112 10BaseModelHistograms......................113 11BaseModel-ISContinuance..................115 12BaseModelTotalEectsPathCoecientsonISContinuance117 13CompleteNon-ModeratedModel................118 14CompleteNon-ModeratedSimpliedModel..........119 15 SecondOrderConstruct: SatisfactionandPerceivedUsefulness Histogram.............................120 16BrandRelationshipHistogram..................121 17PersonalRelationshipHistogram................121 18ProceduralEconomicCostHistogram..............121 19ProceduralEvaluationCostHistogram.............122 20ProceduralLearningCostHistogram..............122 21ProceduralSetupCostHistogram................122 22 SecondOrderConstruct: RelationshipCostHistogram....123 23 SecondOrderConstruct: ProceduralCostHistogram.....123 xviii

PAGE 20

24 ThirdOrderConstruct: CostHistogram............123 25HabitHistogram.........................124 26PersonalInnovativenessHistogram...............124 27InterpersonalInuenceHistogram................124 28AlternativeAttractivenessHistogram..............125 29AttitudeToSwitchHistogram..................125 30 SecondOrderConstruct: AlternativePerceptionsHistogram.125 31CompleteNon-ModeratedModelPathCoecientsforSNSContinuance..............................134 32CompleteNon-ModeratedSimpliedModel..........135 33CompleteNon-ModeratedModel................135 34CompleteModeratedModel...................136 35CompleteModeratedSimpliedModel.............137 36CompleteModeratedSimpliedModel.............147 37CompleteModeratedModel...................147 38ModeratedModelPathCoecientsonSNSContinuance...150 39BaseModelAndCostsPathCoecientsonSNSContinuance159 40BackwardStepwiseRenement-Eectsize...........166 41ConstructMeanComparisonAcrossThreeDatasets......175 42PathComparisonAcrossThreeDatasets............177 43PathComparisonAcrossThreeDatasets-StatSigPathsonly178 44PathComparison-Cohen'sDEectSizeThreeDatasets...181 45SocialNetworkingSiteContinuanceModel...........182 46 R 2 and f 2 ContinuanceByModel................185 47HabitHistogram.........................238 48BaseModelAndHabitPathCoecientsonSNSContinuance243 49PersonalInnovativenessHistogram...............244 xix

PAGE 21

50BaseModelAndPersonalInnovativenessPathCoecientson SNSContinuance.........................250 51InterpersonalInuenceHistogram................251 52BaseModelAndHabitPathCoecientsonSNSContinuance256 53FourAlternativesSNSAccountImpact.............258 54AlternativeAttractivenessHistogram..............258 55AttitudeToSwitchHistogram..................259 56 SecondOrderConstruct: AlternativePerceptionsHistogram.259 57BaseModelAndHabitPathCoecientsonSNSContinuance264 58BrandRelationshipHistogram..................265 59PersonalRelationshipHistogram................266 60ProceduralEconomicCostHistogram..............266 61ProceduralEvaluationCostHistogram.............267 62ProceduralLearningCostHistogram..............267 63ProceduralSetupCostHistogram................268 64 SecondOrderConstruct: RelationshipCostHistogram....268 65 SecondOrderConstruct: ProceduralCostHistogram.....269 66 ThirdOrderConstruct: CostHistogram............269 67CostModelPathCoecientsonSNSContinuance.......275 68BaseModelAndCostsPathCoecientsonSNSContinuance284 69SpecicProductPathCoecientsonSNSContinuance....292 70SpecicProductatleast1alternateproductPathCoecients onSNSContinuance.......................296 xx

PAGE 22

1Introduction SocialnetworksitesSNSarecomposedofaseriesofdynamicprocesses;users jointhesite,connectwithusers,sharecontent,interactwithusers,disconnect fromusers,takebreaksfromthesiteandabandonthesite.Userswhoinitially adoptthesitemaycontinuetousethesocialnetworkingsite,takebreaksfrom thesiteorstopusingthesitealltogetherRaineetal.,2013;thefactorsthat contributetousers'continuanceintentionsonsocialnetworksitesarenotwell known.Thisresearchprovidesthetheoreticalbackground,motivationand methodstoexaminesocialnetworkingsitecontinuanceintentionsofusersby adaptingBhattacherjee'sinformationsystemsIScontinuancetheory tothesocialnetworkingsiteenvironment.Thegoaloftheresearchistoexaminetheroleoffactors alternativeperceptions e.g.dousersshowanintention todiscontinueuseofasitewhenothersocialnetworksitesarebeingconsidered, interpersonalinuence personalinnovativeness habit ,and consumer switchingcosts onusers' socialnetworkingsitecontinuanceintention ResearchonSNScoverareasthatfocusontheindividualuserbehavior motivationsofuse,tieformation friending,following,connecting ,etc ,tie dissolution unfriending,unfollowing,disconnecting ,etc.,individual-to-site relationships socialcapital,contextcollapse,sitesatisfaction, and network abandonment ,andlargerenvironmentalconcernslikeprivacypolicies.ResearchinSNSmayextendbeyondtheindividualtocollectiveintelligence; researchersmakepredictionsforelections,boxocerevenue,communicable diseasedetectionandtransmission,etc.wherecollectiveindividualactions areaggregatedandtheindividualbehaviorislessimportant.Otherresearch interestsincludeidentitymanagementHewittandForte,2006,trustDwyer etal.,2007,self-presentationStutzman,2006,surveillanceandprivacycon1

PAGE 23

cernsGrossandAcquisti,2005,socialcapitalEllisonetal.,2007.Muchof theacademicresearchonFacebookhasfocusedonidentitypresentationand privacyconcernsEllisonetal.,2007. TechnologyacceptancehasbeenstudiedbyinformationsystemsISresearchersforavarietyofclassesofapplicationssuchaswork-relatedapplicationshealthcare,scheduling,etc.,generalapplicationsemail,telecommunication,etc.,ande-commerceapplicationsKingandHe,2006.Technology usageresearchhasmovedbeyondtheinitialadoptionstageofthetechnologyacceptancemodelTAMofDavisDavis,1989;Davisetal.,1989to predictuser'scontinuanceintentionsBhattacherjee,2001;Bhattacherjeeand Premkumar,2004;Thongetal.,2006,i.e.whetherainformationsystemsuser continuestouseordiscontinuesuseofaninformationsystempost-acceptance. ForinformationsystemstobehelpfultheymustbeadoptedDavis,1989but thelonger-termconsequencesofaninformationsystemmaybemoredirectly relatedtoitscontinueduseratherthanitsinitialacceptanceBhattacherjee, 2001;Thongetal.,2006. Researchintosocialnetworkingsitestintothelargerdomainofinformationsystemsresearch.Researchregardinginformationsystemsusagehas beenasignicanttopicofinformationsystemsresearchLiaoetal.,2009;BhattacherjeeandPremkumar,2004;Bhattacherjee,2001;KingandHe,2006;Lee etal.,2003;Hovorkaetal.,2013.Researchofthe usedomain acceptance phaseoftechnologyaccountsforthelargestportionofgeneratedhypotheses intwocoreinformationsystemsjournals:MISQuarterlyMISQandInformationSystemsResearchISRHovorkaetal.,2013.Hovorkaetal. notedthatthedomain satisfactionwithtechnology relatedtoinformation systemscontinuancesharesanareaofinterestasitisreferencedbythetwo majordomainsofISresearch:ISdevelopmenttheoryandISuse.Researchin 2

PAGE 24

informationsystemcontinuancemayhelpidentifythepathwaysofmulti-level phenomenathatemergebyintegratingindividualuseandbroaderorganizationalsuccessHovorkaetal.,2013. ThetechnologyacceptancemodelTAMproposedbyDavisDavis,1989; Davisetal.,1989explainsmotivationsofinformationsystemacceptanceand predictsuser'sbehavioralintentionstoacceptanewinformationsystemBhattacherjee,2001.TAMiscenteredaroundinitialacceptanceofaninformation systemwhichisanimportantphaseinthelifecycleofinformationsystem usage.TAMisbasedontheFishbeinandAzjen'sTheoryofReasoned ActionTRA;TRAisanintentionmodelthathelpspredictandexplainbehaviorsacrossmultipledomainsincludinginformationsystemdomainsDavis etal.,1989. Themotivationofthisresearchistodevelopandtestamodelofsocial networkingsitecontinuancewithfactorsthatarehelpfulinunderstanding thephenomenon.Theresearchofacceptanceandcontinuancehasfocused ontheindividual'sattitudetowardthetechnologywithoutexamininghow competingproductsorservicesmayinuenceacceptanceorcontinuanceand howthecompetingproductsorservicesmayinuencecontinuanceintention. Psychosocialfactorslikehabit,personalinnovativeness,interpersonalinuenceandconsumerswitchingcostsmayalsoexertaninuenceoncontinuance intentionthatarenotcurrentlycapturedintheIScontinuancetheory. 1.1ImportanceofTopic Informationsystemscontinuanceonsocialnetworkingsitesisanimportant topicbecausewhetheruserscontinuetouseagiveninformationsystemhasa signicantimpactonwhetherasitecancontinueasagoingconcern.Bhattacherjeearguesthatthelong-termviabilityofinformationsystems 3

PAGE 25

dependsmoreonitscontinueduseorpost-adoptionbehaviorratherthan initialadoption.Thecontextofthisresearchissocialnetworkingsiteswhere AmericansspendthelargestshareoftheirtimeonlineNielsen,2011.Social networkingsitesallowresearcherstoexaminesystemswhicharehighlyvoluntaryinnaturebecausethereislittleovertdirectexternalpressuretojoin, use,ordiscontinueuseofsiteslikeFacebook,Twitter,LinkedIn,GooglePlus, etc.Researchingsocialnetworkingsitesallowsresearcherstoextendmodels forsuccessbeyondtraditionalareaslikee-commerceintoareasthathavebeen under-researchedSchaupp,2011. Socialnetworkingsiteusershavelargelyabandonedsitesthatwereonce popularlikeSix-degrees,FriendsterandMySpaceforothersocialnetworking sitesandindicatesthatsiteusersmakedecisionstodiscontinueuseofonce successfulsites.ThereareoveronebillionworldwideusersofthesocialnetworkingsiteFacebook 1 ;intheUnitedStatesthereareover150millionFacebookusers,37millionTwitterusers,and26millionGoogleplususersNielsen, 2012.Continuanceintentionandsocialnetworkingsiteshavebroadimplicationsforinformationsystemsandmayinformareasbeyondsocialnetworking sites,inparticularinareaswherethereisahighdegreeofvoluntariness,where habitualuseofaninformationsystemisexpected,whereinterpersonalinuenceisafactor,andwhereproceduralandrelationalcostsmayimpactIS continuancedecisions. 1.2ResearchProblemandScope Thegoalofthisresearchistodevelopandtestamodelofinformationsystems continuancethatmoredirectlyappliestosocialnetworkingsites;i.e.what factorsarerelevanttosocialnetworkingsiteuserswhentheychoosetocon1 http://newsroom.fb.com/Key-Facts 4

PAGE 26

tinueordiscontinueuseofasite.Theresearchwilluseinformationsystems continuancetheoryfromBhattacherjeeasatheoreticalfoundationand includeadditionalfactorssuchasalternativeperceptions,personalinnovativeness,interpersonalinuences,habitandconsumerswitchingcosts. Theresearchwillfocusontheindividuallevelofanalysiswhereusersmake decisionswhethertocontinueordiscontinueuseofanindividualsite.TheresearchmayhaveimplicationsbeyondtheindividuallevelasHovorkaetal. suggeststhatsatisfactionandcontinuancemayactasabridgebetween theindividualleveladoptionbehaviorsandorganizationallevelsuccessand mayhelpbuildmulti-leveltheorythatlinkstheindividuallevelofanalysisto theorganizationallevelofanalysis.Satisfactionhasbeenusedasanindependentfactorininformationsystemsandconsumerbehaviorresearchtopredict switchingorcontinuancebehavior;however,satisfactionhasbeenshownin consumerbehaviorresearchtoexplainapproximately25%ofthevarianceof customerswitchingordiscontinuancebehaviorBurnhametal.,2003;SzymanskiandHenard,2001.TheresearchaddsfactorsaddedtoIScontinuance theorythatmayhelpexplainmoreofthevarianceincontinuancedecisions. 1.3ResearchQuestions InformationsystemscontinuanceofBhattacherjeeextendsexpectationconrmationtheorytopredictcontinuanceintentionthroughthreefactors: perceivedusefulness,conrmationandsatisfaction.Satisfaction,asameasureofcontinuance,limitstheamountofexplainedvarianceofcontinuance intentionbyitselfastheremaybeadditionaldirectandmoderatingfactorsin continuancedecisions. Whataretheimportantmotivatingfactorsthatpredictanindividual's intentiontocontinuetouseasocialnetworkingsite? 5

PAGE 27

Howdopsychosocialfactorslike personalinnovativeness,habit, and interpersonalinuence predictanindividual'sintentiontocontinue touseasocialnetworkingsite? Howdo consumerswitchingcosts predictanindividual'sintention tocontinuetouseasocialnetworkingsite? Howdo alternativeperceptions predictanindividual'sintentionto continuetouseasocialnetworkingsite? 1.4ResearchContribution Therearetwomainexpectedcontributionsoftheresearchasproposed.The rstisaliterarycontributionofexaminingtheinformationsystemsandconsumerbehaviorresearchtodeterminerelevantacademicresearchrelatedto continuousintention. Thesecondcontributionisregardingcontinuanceintentiononsocialnetworkingsites.Theresearchmodelspecicallyexpandsbeyondsatisfaction asameasureofcontinuanceintentiontoincludeadditionalrelevantfactors. Factorssuchas habit havebeenstudiedinothercontextswithininformationsystemse.g.Limayemetal.andmayhaveanimpactonsocial networkingsites,inparticularitmayhelpexplainwhyasitelikeFacebook withrelativelylowlevelsofsatisfaction 2 continuetoshowhighusagepatterns. TherelevantcostsofconsumerswitchingbehaviorbasedfromBurnhametal. mayhaveastrongroletoplayinswitchingbehavior.Burnhametal. foundthattheconsumercostshadalargerroleinswitchingintention thansatisfaction.Satisfaction,intheconsumerbehaviorliterature,explains approximately25%ofthevarianceincontinuanceintentionBurnhametal., 2003;SzymanskiandHenard,2001.Theresearchexaminestheimpactsof 2 http://www.theacsi.org/index.php?option=com_content&view=article&id=147& catid=14&Itemid=212&i=Internet+Social+Media 6

PAGE 28

alternativeservicesubstitutionsoncontinuanceintentiontoincluderelevant socialnetworkingsitecompetitorsasafactorinanindividual'sdecisionto continuetouseasocialnetworkingsite.Themodelprovidesamorecomplete theoreticalfoundationforexaminingcontinuanceintention. 1.5OutlineofDissertation Thedissertationisorganizedasfollows.Theintroductioncoversthegeneral overviewoftheproblem,importance,researchquestions,potentialoutcomes -Section1.Aliteraturereviewofinformationsystemstechnologyacceptance andinformationsystemscontinuanceintention,consumerswitchingbehavior, andinformationsystemscontinuanceasappliedtosocialnetworkingsitesis developedinSection2.Relevantfactorstoinformationsystemscontinuance forsocialnetworkingsitesandtheirimportancetotheeldaredescribedin Section3.Aninitialstudypredictingend-usersatisfactionbasedonperceived ease-of-useandperceivedusefulnessisdescribedinSection4.Amodelof socialnetworkingsitecontinuanceisdevelopedwiththerelevanthypotheses inSection5.Theinstrumentdesign,datacollectionmethod,andanalytical techniquesaredescribedinSection6.1.Thestatisticalanalysisandresults forthemodelsarefoundinSection7.Thediscussionexplainsthendingsof resultsinSection8.TheconclusionSection10providesnalremarksabout thefactorsandgoalsofthisresearch.TheAppendixshowsadditionaldetails suchascommonmethodvarianceandindividualfactormodelsingreaterdetail theimpactofasingleintroducedfactore.g.habitonSNSContinuance. 7

PAGE 29

2AcceptanceandContinuanceTheories Technologyadoptionhasbeenwidelystudiedformorethantwodecadesafter thedevelopmentofthetechnologyacceptancemodelTAMbyDavisDavis etal.,1989;Davis,1989;themodelmaybethemostwidelystudiedtheoreticalmodelininformationsystemsVenkateshetal.,2007;Leeetal.,2003. TAMpredictsauser'sbehavioralintention-to-useagiveninformationsystem; themodelhasbeenshowntohavestrongpredictivevaliditytopredictactual systemuseVenkateshetal.,2007.Informationsystemsresearchregardingadoptionswaslargelyfocusedonmeasureslikesatisfactionandattitude priortoTAM'stheoreticaldevelopmentVenkateshetal.,2007.BhattacherjeedevelopedinformationsystemsIScontinuancetheorybasedon expectation-conrmationtheoryofOliverfromconsumerbehaviorresearchtoexaminepost-adoptionbehaviorbehaviorbeyondinitialadoption ofatechnology.IScontinuancetheorypredictscontinuancebehaviorusing threefactors:perceivedusefulness,conrmationandsatisfaction. TheinformationsystemscontinuancemodelofBhattacherjeeis basedonexpectation-conrmationtheoryfromtheconsumerbehaviorarea. Consumerbehaviorresearchinthemarketingeldhasexaminedreasonsfor consumerdiscontinuanceandswitchingfromoneprovidertoadierentprovider ofproductsandservices.Keaveneydevelopedaneightcategorytypologyforconsumerswitchingbehaviorforserviceproviders.BansalandTaylordevelopedapsychometricmodelbasedonKeaveneyand thetheoryofplannedbehaviorFishbeinandAzjen,1975calledthe service providerswitchingmodel SPSM.KeaveneyandParthasarathyexaminedcustomerswitchingbehaviorinthecontextofonlineserviceproviders andusedattitudes,behaviorsanddemographicstopredictswitchingbehavior. 8

PAGE 30

Burnhametal.examinedcustomerswitchingbehaviorthroughthree dierentcostsandsatisfactiontopredictswitchingbehavior.Givon developedastochasticmodelofcustomerswitchingbehaviorthroughthelens ofvarietyseekingbehavior.Manystudiesofconsumerswitchingbehavioruse satisfactionasafactorinthemodel;however,researchersalsoaddadditional factorstoincreasetheexplainedvarianceinthemodeltogainpredictivepower Keaveney,1995;KeaveneyandParthasarathy,2001;Burnhametal.,2003. Consumerbehaviorresearchmayprovideafoundationtoexplainhowinformationsystemsuserscontinue-to-useorswitchtodierentserviceprovidersseeSection2.2. Researchersoftenusetechnologyacceptancemodelandinformationsystemscontinuancetheoryasafoundationtoexploreadditionalfactorsinrelated studiesonadoptionandpost-adoptionbehavior.Liaoetal.compareda revisedtechnologyacceptancemodelTAM,informationsystemscontinuance andexpectation-conrmationtheorytocomparetheexplainedvarianceeach theorypredictsforauser'scontinuanceintention.ResearchershaveusedBhattacherjeeinformationsystemscontinuancetheorytodeterminecontinuanceintentioninmanydomainsandspecicallyinthesocialnetworkdomain. ResearchershaveusedIScontinuancewithMaslow'shierarchyofneedsCao etal.,2013,networkexternalitiesLinandLu,2011,commitment-trustXu etal.,2012,socialexchangetheory,socialcapitaltheoryandowtheoryHu andKettinger,2008topredictcontinuanceintentionforsocialnetworking sites-seeSection2.3. 2.1TechnologyAcceptanceandInformationSystemsContinuance Davisetal.adaptedthetheoryofreasonedactionTRAtoexplainIS acceptancethroughtwoconstructs:perceivedusefulnessPUandperceived 9

PAGE 31

Figure1:TechnologyAcceptanceModel-Davis ease-of-usePEOUtoexplainsystemusers'intention-to-useandadoption behavior-seeFigure 1 .Davis,p.320dened perceivedusefulness as,"thedegreetowhichapersonbelievesthatusingaparticularsystem wouldenhancehisorherjobperformance,"anddened perceivedease-ofuse as,"thedegreetowhichapersonbelievesthatusingaparticularsystem wouldbefreeofeort."TheinitialresearchbyDavis,p.320was aboutadoptionwithinorganizationalcontexts;however;TAMhasbeenused topredictacceptanceinmanyISdomainsoutsidetheorganizationalcontext. Venkateshetal.notedthatTAMhasbeenwidelyacceptedamong ISresearchersandwidelyreplicatedwithlittlematerialtheoreticaladvancementtwodecadesafteritspublication.Theresearchersnotedthattheremay beseveralreasonsforthetheory'sadoptionbyISresearchersincludingitsparsimoniousnature,therobustnessofthescalesandthegeneralizabilityofthe model.TAMwasshowntohavebetterpredictivepowerthanthemoregeneral theoryofreasonedactionuponwhichitisbasedVenkateshetal.,2007indicatingitismoreusefulthanamoregeneralizedpsychologically-basedmodel. TAMhasbeenreplicatedinmanysettingsovertimewithmanypopulationsandindierentcontextsandhasprovenrobustVenkateshetal.,2007. Leeetal.describesachronologyofTAMresearchwherethemodel wasrstintroduced,replicated,validated,extended,andelaborated.Sev10

PAGE 32

eralvariableshavebeenintroducedtoTAMinsubsequentstudiesthrough extensions,Leeetal.notesthatvariables voluntariness,relativeadvantage,compatibility,complexity,observability,trialability,image,selfecacy, enduser,supportobjectiveusability,personalinnovativeness,computerplayfulness,socialpresence,subjectivenorms/socialinuence,visibility,jobrelevance,computerattitude,accessibility,resultdemonstrability,management support,computeranxiety,perceivedenjoyment,systemoutputorinformationquality,facilitatingconditions and priorexperience haveallbeenstudied withinthecontextoftechnologyadoption.OneconcernregardingTAMisthat itsemphasisontheinitialphaseoftechnologyacceptancemaylimititsabilitytopredictandinformusageatpost-adoptionstagesBhattacherjee,2001; Venkateshetal.,2007;Leeetal.,2003.Venkateshetal.2007notedthat anemergingareabeyondtheinitialphaseofacceptanceistheapplicationof expectation-conrmationtheoryfrommarketingresearchbyOliverto informationsystemcontinuancebehaviorbyBhattacherjee. Bhattacherjeemovedbeyondrst-timeuseofaninformationsystemtoexaminethelongertermconsequencesofcontinueduseorcontinuance vs.acceptance.IScontinuanceisimportanttosoftwaredevelopmentrms becausecustomers,marketshareandrevenuedependonbothinitialadoption and continueduse.OneadvantageofIScontinuancetheoryoverinitialacceptanceisthatIScontinuancespecicallyaddressesthephenomenaof discontinuancethosewhodecidetostopuseafterinitialacceptanceBhattacherjee,2001.AnadditionaladvantageofIScontinuancetheoryisthatit providesamodelofthepsychologicalmotivationsofcontinueduseafterinitial acceptance. BhattacherjeebasedthetheoryofISContinuanceonOliver's expectation-conrmationtheoryfromconsumerbehaviorresearch.Themodel 11

PAGE 33

includesthreeconstructs,perceivedusefulness,conrmation,andsatisfaction topredictIScontinuanceintention-seeFigure2.IScontinuanceisprimarily determinedbytheconsumerssatisfactionofprioruse,i.e.consumerswhohave apositiveexperiencewithsoftwaretendtohavehighercontinuanceintentions thanothers.Satisfactionisdeterminedbytwofactors,the perceivedusefulness Oliver0callsthisfactor expectation and conrmation .Higherlevelsof conrmationindicatethattheusersinitialexpectationoftheexperiencehas beenexceededbyactualuseandlowerlevelsindicatethattheactualusedid notmeettheexpectation.Higherlevelsofconrmationareexpectedtohave apositiveimpactonsatisfactionofthesystem.Perceivedusefulnesslargely capturesexpectationperceptionsofpost-consumptionusageandisrelatedto theTAMfactorofthesamename.Higherlevelsofperceivedusefulnessinthe systemareexpectedtoleadtohigherlevelsofsatisfaction.PerceivedusefulnessisalsoexpectedtohaveadirecteectonIScontinuanceintention;i.e. higherlevelsofperceivedusefulnessareexpectedtoincreaseIScontinuance intention.Thelastrelationshipinthemodelisthathigherlevelsofconrmationareexpectedtohavehigherlevelsofperceivedusefulness-thisissimilar totheTAMrelationshipthatperceivedeaseofuseispositivelyrelatedto perceivedusefulness. Expectation-conrmationtheoryOliver,1980isdesignedtouseboth pre-consumptionandpost-consumptionexpectationstopredictrepurchaseintentions;however,Bhattacherjee'sresearchexaminedonlythepostconsumptionrelationshiptopredictIScontinuance.TheresultsofBhattacherjee'ssuggestedthatpost-consumptionbehaviorismoreimportantthan pre-consumptionbehaviorandBhattacherjee'ssubsequentresearchin2004 foundthatasusersgainedmoreexperiencewithasoftwareproducttheirsatisfactionlevelsregressedtowardthemeanandhadalesspredictivepoweron 12

PAGE 34

Figure2:InformationSystemContinuance-Bhattacherjee overallsatisfaction. Liaoetal.comparedthecompetingmodelsofTAM,IScontinuance theory,expectation-conrmationtheorytopredictcontinuanceintention.The modelsallsupportedthehypothesesthattheygenerated.Thendingssuggestthatexpectation-conrmationtheoryhasmorepredictivepowerthanIS continuancewhichhasbetterpredictivepowerthanTAM.However,asBhattacherjeeexplains,IScontinuancetheoryisusedtopredictandexplain howthefactorsinuencecontinuancedecisionsandmayhavestrongertheoreticalfoundationtopredictcontinuanceintention. 2.2ConsumerSwitchingBehavior Consumerresearchinmarketinghasexaminedreasonsforcustomersdiscontinuinguseofaproductorserviceorswitchingfromoneserviceprovider toadierentserviceproviderKeaveney,1995;KeaveneyandParthasarathy, 2001;BansalandTaylor,1999;Burnhametal.,2003;Sanchez-Garciaetal., 2012.KeaveneyandParthasarathynotethatserviceswitchinginthe faceofincreasedcompetitionisoftencalled churn intheconsumerbehavior area CustomerswitchingisaconcernbecausecustomeracquisitionandretentioncostscanbehighKeaveneyandParthasarathy,2001;Parthasarathy andBhattacherjee,1998.Rapidlychangingmarketsmayexperiencemore customerchurnwherecustomerretentionisaconcernofserviceproviders. 13

PAGE 35

Loweringtheratecustomerswitchingmayincreaserevenuesandlowercosts ofserviceprovidersKeaveneyandParthasarathy,2001.Companieswhohave higherretentionratesorlowerchurnratesmayincreaseconsumerwordof mouthandincreasemarketshareKeaveneyandParthasarathy,2001.Customersatisfactionandservicequalityarefoundtohaveastrongrelationship withswitchingintentionsinavarietyofstudiesKeaveney,1995;Keaveneyand Parthasarathy,2001;BansalandTaylor,1999;Burnhametal.,2003;SanchezGarciaetal.,2012.Studiesalsonotethatresearchshouldconsiderother causesofswitchingbeyondsatisfactionKeaveneyandParthasarathy,2001; ParthasarathyandBhattacherjee,1998;Burnhametal.,2003.Burnham etal.notesthatwhiletheevidenceisstrongthatsatisfactioninuencesrepurchasesbehaviorittypicallyexplainsapproximatelyonequarterof thevarianceofbehavioralintentionsSzymanskiandHenard,2001;Burnham etal.,2003. 2.2.1ExploratoryResearchinCustomerSwitchingBehavior Keaveneyresearchedcustomerswitchingbehaviorinserviceindustries byexaminingmorethan500servicecustomerswhocitedmorethan800criticalbehaviorsthatleadcustomerstoswitchserviceproviders.Themotivation behindtheresearchwastodeterminewhycustomersswitchserviceproviders, whateventsandcombinationofeventsleadacustomertochangeproviders, howdoserviceencountersandservicequalityinuencethedecisiontoswitch. Theresearchfound8categoriesofserviceswitchingbehaviorsand other : pricing,inconvenience,coreservicesfailures,failedserviceencounters,responsetofailedservices,competition,ethicalproblemsand, involuntaryswitching.Fewerthan5%ofincidentswereclassiedas other indicatingthattheeightmajorcategoriescoverthemajorityofswitchingbe14

PAGE 36

haviors.Switchingbehaviorwascategorizedas simple onefactoror complex multi-factor Simpleswitchingaccountedforforty-vepercentofswitching, two-factorincidentsaccountedforthirty-sixpercentofswitching,threefactor incidentsaccountedforfteenpercentofincidentsandfour-or-more-factor incidentsaccountforfourpercent.Keaveneynotesthattheresearchexpands thereasonsforwhycustomersswitchserviceprovidersbeyondservicequality andsatisfactiontoincludethefactorsfoundintheresearch. Sixoftheeightreasonsforservicefailurearecontrollablebytherm wheretwoarebeyondthebusiness'scontrolKeaveney,1995.Companiescan managethepricing,inconvenience,servicefailure,serviceencounters,response tofailureandethicalissues,whereinvoluntaryswitchingandcompetitionare beyondtherm'scontrol.Customersmaychangeprovidersevenwhentheyare satisedwiththeexistingservices,e.g.anaccountantmayprovidesatisfactory servicebutadierentserviceprovidermayprovidemoredirectaccessandis whatthecustomerdesires.Anotherexampleinthestudywasofacustomer whowassatisedwithamechanicbutswitchedbecauseadierentmechanic alsoperformedsatisfactoryworkonaspouse'sautomobile.Aconfounding issueisthatdissatisedcustomersmayalsostaywithserviceprovidersdespite issueswiththeserviceKeaveney,1995. 2.2.2ServiceProviderSwitchingModel BansalandTayloradvancedthedescriptiveresearchofKeaveney anddevelopedapsychometricmodelofcustomerswitchingandnamedtheresultingmodeltheserviceproviderswitchingmodelSPSM.Theresearchers groupedKeaveney'sresultsintotwocomponents-serviceperformancecore servicefailure,serviceencounterfailure,responsetoservicefailure,andethicalproblemsandcostofswitchingprice,inconvenienceandcompetition. 15

PAGE 37

Keaveney'sresearchisnotan intentional modelasitcollecteddataonserviceswitchingthatoccurred.BansalandTaylorincludedFishbein andAzjentheoryofplannedbehaviortoincludeattitudeandbehaviorfactorstodevelopabehavioralintentionmodel.Themodelincludesfourdirect antecedentstopredictswitchingintention:satisfaction,perceivedswitching costs,attitudetowardswitchingandsubjectivenormstopredictswitching intentions. TheBansalandTaylormodelwasvalidatedandaccountsfor76% ofthevarianceinswitchingbehaviorintentions.Thelogisticregressionmodel couldcorrectlyclassify74%ofswitchingbehaviorsandaccountfor30%of thevariance.TheserviceproviderswitchingmodelSPSMusesattitude towardswitchingbehavioralintentionsofthetheoryofplannedbehavior, satisfactionserviceperformance,switchingcostsandself-ecacytopredict themajorityofthevarianceincustomerswitchingintentionsandcustomer switchingbehaviors. 2.2.3CustomerSwitchingBehaviorforOnlineServiceProviders KeaveneyandParthasarathyexaminedcustomerswitchingbehavior foronlineserviceproviderse.g.Americaonline,Compuserve,etc..The researchersnotethatsomecustomerselecttodiscontinueservicecompletely whileothercustomerschoosedierentserviceproviders.Theresearcheduseda combinationofattitudes,behaviorsanddemographicsaspredictorsofswitchingbehavior.Theresearchersusedexternalsourceofinformationmassmedia,marketinginformation,etc.,interpersonalsourceofinformationwordof mouth,experientialsourceofinformationproductengagement,serviceusagefrequencyofuse,intensityofuse,usage-overallandrisktakingbehavior, satisfactionandinvolvementtomeasureattitudesandbehaviors.Incomeand 16

PAGE 38

educationwereusedasdemographicpredictors. TheKeaveneyandParthasarathyresultsshowthatcustomerswho continuedtousetheirexistingserviceproviderreliedmoreonexternalsources ofinformatione.g.massmediawhenmakingtheirsubscriptiondecisions thanthosewhoswitched.Customerswhostayedwiththeirexistingprovider reliedlessoninterpersonalsourcesofinformatione.g.wordofmouthwhen makingthesubscriptiondecisionthanthosewhoswitched.Customerswho reliedmoreontheirownexperiencewithaservicecontinuedtousetheservicemorethanthosewhoswitched.Customerswhostayedwiththeirexisting providerusedtheserviceathigherfrequencyofuseandhadhigheroverall usage,butdidnothaveanygreaterintensityofusethanthosewhodiscontinueduse.Customerswhostayedwiththeirexistingproviderhadhigherlevels ofsatisfactionthanthosewholefttheservice.Onlineservicecontinuershad greaterinvolvementwiththeservicethanthosewhodiscontinueduse.One unexpectedndingwasthatcustomerswhocontinuedwithanexistingservice hadnowlowerpropensityforrisktakingthanthosewhoswitchedtheirservice provider. OnegoaloftheresearchbyKeaveneyandParthasarathywasto provideaframeworkforpredictingwhichcustomersmayleaveanexisting serviceproviderandgotoanother.Ifcustomerswhomayleaveaservice canbeidentiedthenstrategiesmaybeenactedtoretaincustomers.The researcherssuggestthatcompaniesincreasesatisfactionaswellasinvolvement, andserviceusageespeciallyfrequencytoincreasecustomerretentionand reducecustomerchurn. 2.2.4ConsumerSwitchingCosts Burnhametal.examinedconsumerswitchingcoststhroughatypology 17

PAGE 39

ofthreedierentcosts.Theexaminedproceduralswitchingcostslossoftime andeort,nancialswitchingcostsandrelationalswitchingcostspsychologicaloremotionaldiscomfortduetothelossofidentityandthebreaking ofbonds.Theresearchshowsthatwhilecompanieswillwanttomanage customersatisfactiontheyshouldalsomanagetheperceptionsthatcustomers haveofswitchingcosts.Themodeldenesthreegroupsprocedural,nancialandrelationshipofeightswitchingcosts:economicriskcosts,evaluation costs,learningcosts,setupcosts,benetlosscosts,monetarylosscosts,personalrelationshiplosscosts,brandrelationshipcoststoexamineconsumer switchingcosts-seeTable1. TheresearchbyBurnhametal.supportedthatsatisfactionand thethreeswitchingcostsprocedural,nancialandrelationalwereallsignicantpredictorsofserviceproviderswitchingintention.Consumerswithhigher perceivedproceduralswitchingcostswerelesslikelytochangeproviders.Servicesthatareperceivedtobemorecomplexanddiculttoevaluatewere lesslikelytobesubstitutedforcompetingproducts.Consumerswhothought therewouldbeanegativenancialimpactweremorelikelytostaywithwith aserviceprovider.Theresultssuggestthatloyaltyprograms,bundlingof servicesandcomplexserviceoeringscanreduceserviceproviderswitching. Consumerswhoperceivedhighrelationalcostswiththeserviceproviderwere lesslikelychangeserviceproviders;consumerscanstronglyidentifywitha serviceproviderandwillbelesslikelytochooseadierentserviceprovider. Relationalcostshadthehigheststandardizedestimate.44followedbyproceduralcosts.15,thennancial.13aspredictorsforcustomerswitching behavior.Satisfactionwasalsosupportedasasignicantpredictorofswitchingintentionswherehighersatisfactionwasassociatedwithlowerswitchingof serviceproviders. 18

PAGE 40

Table1:TypologyofConsumerPerceptionsofSwitchingCosts CostComponents Procedural SwitchingCosts EconomicRiskCosts -thecostsofaccepting uncertaintywiththepotentialforanegativeoutcome whenadoptinganewprovideraboutwhichthe consumerhasinsucientinformation.p.111 EvaluationCosts -thetimeandeortcosts associatedwiththesearchandanalysisneededto makeaswitchingdecision.p.111 SetupCosts -thetimeandeortcostsassociated withtheprocessofinitiatingarelationshipwitha newproviderorsettingupanewproductforinitial use.p.111 LearningCosts -thetimeandeortcostsof acquiringnewskillsorknow-howinordertousea newproductorserviceeectively.p.111 Financial SwitchingCosts BenetLossCosts -thecostsassociatedwith contractuallinkagesthatcreateeconomicbenetsfor stayingwithanincumbentrm.p.111 MonetaryLossCosts -theonetimenancial outlaysthatareincurredinswitchingprovidersother thanthoseusedtopurchasethenewproductitself. p.111 Relational SwitchCosts PersonalRelationshipLossCosts -theaective lossesassociatedwithbreakingthebondsof identicationthathavebeenformedwiththepeople withwhomthecustomerinteracts.p.111-112 BrandRelationshipLossCosts -theaective lossesassociatedwithbreakingthebondsof identicationthathavebeenformedwiththebrand orcompanywithwhichacustomerhasassociated. p.112 AdaptedfromBurnhametal. 19

PAGE 41

Adirectcomparisonofthethreeswitchingcostsvs.thesatisfactionmeasureshowedthattheswitchingcostsexplainedmoreofthevariance% thansatisfaction%inintentiontoswitchserviceprovidersBurnham etal.,2003.Theresultsshowthatswitchingcostsmayhavemorepredictivepowerforcustomerswitchingbehaviorthatsatisfaction.Burnhametal. stronglysuggestthatswitchingcostsshouldbeincorporatedintocustomerswitchingbehaviorresearchinadditiontosatisfactionmeasures.The researchersalsonotecostsmayapplytodierentsettingsinvaryingamounts, forexample,insocialnetworkingsiteslikeFacebookandTwitterthenancial costsmaynotapplyastheserviceisfree-seeSection5fortheapplicationof consumerbehaviorresearchtosocialnetworksitecontinuanceintention. 2.2.5VarietySeekingBehavior Givonexaminedcustomerswitchingbehaviorthroughthelensof varietyseekingbehavior -operationalizedas,asameasureofindividualtendency tovaryconsumption.Thistendencyismeasuredonacontinuumthatextends fromextremetendencytovaryconsumptiontoanextremetendencytoavoid varietyGivon,1984,p.2.McAlisterandPessemiernotethatsome researchersndthatvarietyseekingbehavioris inexplicable orsoinherently complextoberenderedoperationallyinexplicable,whileotherresearchershave attemptedamorerigorousexplanationofvarietyseekingbehavior.McAlister andPessemiernotedthatresearcherswhotriedtoexplainwhyconsumersexhibitvarietyseekingbehaviorclassifythebehaviorsintotwoclasses derived and direct. Derivedbehaviorrefersto,behaviorresultingfromexternalorinternalforcesthathavenothingtodowithapreferenceforchange inandofitself,McAlisterandPessemier,1982,p.313.Derivedbehavior istheresultofsomeothermotivationsuchasachangeintastes,changein 20

PAGE 42

constraints,satisfyingmultipleneeds,etc.Directbehavioriswherechange isamotivationinandofitselfchangeisinherentlysatisfyingandmaybe inuencedbypeersandbrokenintotwogroups:intrapersonalmotivesand interpersonalmotives.Intrapersonalmotivesaremotivessuchas desirefor theunfamiliar,alternationamongthefamiliar andinterpersonalmotivesare motivessuchas aliation and distinction Givonresearchedvarietyseekingbehaviortodeterminehowwidespreadthephenomenonisandatwhatintensity.Repeatbuyingorconsumer sightingbehaviorcanbestochasticwithinagivenproductclass.Consumer switchingisnotasimplemanifestationofsatisfactionordissatisfactionutility ordisutility.Consumersmayreceiveautilityfromcontinuingwithaknown productorserviceormayreceiveutilityfromswitchingtoacompetingproductorservice.Theresultsfoundthat50%ofhouseholdswereindierentto varietyseekingwithfoodproducts,likewisetheremaining50%wereactively lookingortryingtoavoidchange.Theresearchdidnotexaminesatisfaction withcurrentproductstomodelconsumerswitchingbehaviorbutdeveloped aprobabilisticmodelbasedonpastpurchasesandpsychologicalmeasuresto determinetheprobabilityofbrandswitching.Givonnotesthatitmay beeasiertointroducenewproductstoconsumerswhoexhibithigherlevelsof varietyseekingbehavior-butitmaybehardertokeepthemfromswitchingto alternateproductsaswell.Givonalsofounddemographicdierencesin varietyseekingbehaviorwhereyoungerconsumerstendedtobevarietyseekers comparedtotheiroldercounterparts. 21

PAGE 43

2.2.6ConsumerBehaviorResearchApplicationsinInformation Systems Consumerbehaviorresearchmayhelpserveasafoundationforinformation systemscontinuancebehavior.Bhattacherjeetheoryisanapplication ofexpectation-conrmationtheoryofOliver1980.Researchersintheconsumerbehaviorareamakeanappealtoothersintheeldtoincludefactorsin additiontosatisfactiontoexplainwhichuserscontinueordiscontinueuseor switchtoadierentserviceprovider.Informationsystemscontinuancetheory will,likewise,oftenserveasafoundationforIScontinuancebehavior,but additionalfactorsmaybeaddedtofurtherexplainhowindividualsdecideto continuetouseaninformationsystem. 2.3ISContinuanceAppliedtoSocialNetworkingSites Researchershaveappliedinformationsystemscontinuancetheorytopredict continuanceintentiononsocialnetworkingsites.Kimexaminedfactors perceivedusefulness and perceivedenjoyment asdirecteectsonboth satisfactionandcontinuanceintention.Theresearchexaminedthefactorsof interpersonalinuence and mediainuence asdirecteectson continuance intention .Allfactorshadasignicantroletoplayinthemodelexceptmedia inuencewhichwasnotsignicant. 2.3.1Maslow'sHierarchyofNeeds Caoetal.examinedcontinuanceintentionthroughthetheoreticallens ofMaslow'shierarchyofneeds.Themodelusedtwofactors,fulllmentof socialneedsandfulllmentofself-actualizationneeds,todeterminethedirect eectonbothsatisfactionandcontinuanceintention.Theresearchfoundthat fulllmentofself-actualizationneedshaddirecteectsonbothsatisfaction 22

PAGE 44

andcontinuanceintention.Fulllmentofsocialneedshadadirecteecton satisfactionandnostatisticallysignicanteectoncontinuanceintention. 2.3.2NetworkExternalities LinandLuexaminedtheroleofnetworkexternalitiestopredictIScontinuance.Theresearchersexaminedthenumberofmembers,numberofpeers andperceivedcomplementarityasantecedentstoperceivedbenetsusefulnessandenjoymentwhichsubsequentlypredictedIScontinuanceintention. Theresearchersfoundthatnumberofmemberspredictedusefulness,number ofpeerspredictedusefulness,enjoymentandcontinuanceintentionandperceivedcomplementaritypredictedusefulnessandenjoyment.Themodelfound signicantdierencesofcontinuanceintentionbetweenhowmenandwomen, wherethemodelwasabletoaccountformuchmoreofthevarianceinthe dependentvariablethanformen. 2.3.3SocialNetworkingSiteStickiness Xuetal.developedatheoreticalmodelforsocialnetworkingsite stickiness todeterminewhyuserscontinuetouseparticularsitesbasedona commitment-trustmodel.Xuetal.notethatsatisfactiondoesnot appeartobeameasurethatstronglypredictscontinuancebehavioronsocial networksites;theresearchersnotethatFacebookhasarelativelylowsatisfactionscoreof64outof100gureanddespitethissatisfactionscorethe sitecontinuestogrow.Xuetal.extendthecommitment-trustmodel ofwebsitestickinesswhere commitment and trust arefactorsthatpredict stickinessintention.Commitment ispredictedby investmentsize alternative quality and gratication Trust ispredictedby communicationquality opportunisticbehavior and gratication .Both commitment and trust have grati23

PAGE 45

cation asaantecedentwhichindicatesthatboththecommitmentandtrust factorsaremediatedfactorsbetweengraticationandstickinessintention. 2.3.4ISContinuance,SocialExchangeTheory,SocialCapitalTheory&FlowTheory HuandKettingerdevelopedamodelforinformationsystemscontinuanceofsocialnetworkingservicesbasedonIScontinuance,socialexchange theory,socialcapitaltheoryandowtheory.TheHuandKettinger8 modelusesISContinuancetheoryofBhattacherjeeasitstheoretical foundation.Themodelsuggeststhatusersatisfactionisthemajordriverof IScontinuancethroughperceivedusefulnessandconrmationofexpectations. HuandKettingeraugmenttheISContinuancetheorywithsocialexchangetheory,socialcapitaltheoryandowtheory.Socialexchangetheory assertsthatindividualbehaviorisguidedtomaximizegainsandmarginal utility-individualsattempttobalancecostsandbenetsofpotentialsocial exchanges.Thepotentialrewardsofasocialexchangearenotcontractualin nature-trustactsasamechanismthatencouragesmemberstocontributein asocialexchange.Socialcapitalgenerallyreferstotheskillsandknowledge thatareaccessibletoanindividualthroughtheirrelationshipswithothers Coleman,1988.Colemannotesthatanimportantformofsocialcapitalistheabilitytoacquireinformationthroughrelationships;information itselfmaybevaluedhighlyandisgenerallycostlytoacquire.Accesstoa largeandweakly-tiedtiednetworkmayprovidemorebenetsthanasmaller strongly-tiednetworkGranovetter,1973.Flowtheoryistheexperienceof completeabsorptioninthepresentmomentNakamuraandCsikszentmihalyi, 2009.Flowtheoryisrelatedtotheintrinsicroleofowexperienceintermsof concentrationandperceivedenjoymentthatpeopleexperiencewhentheyact 24

PAGE 46

withtotalinvolvement.Whenin ow anindividualisoperatingatfullcapacityNakamuraandCsikszentmihalyi,2009.Individualswhoexperience ow wanttoreturntothoseactivitiesbecauseofthepositiveexperientialrewards NakamuraandCsikszentmihalyi,2009.Themodelthengroupsfactorsinto threemajorcomponents-satisfaction,perceivedbenetsandperceivedcosts topredictusagecontinuance.Usagehistory,whichincludesfrequencyofprior usageandcomprehensiveofpriorusage,isusedtomoderatetheantecedentsof usagesatisfaction.Continuanceintentionispredictedbysixfactors:social inuence,owexperience,3usagesatisfaction,perceivedvalue, perceivedinformationrisksandperceivedeort.HuandKettinger providedaroadmapoffutureworkwheretheywouldcollectdatafromcollege studentsandanalyzedatawithpartialleastsquaresPLS. 2.3.5SatisfactionScoresofSocialNetworkingSites TheAmericanCustomerSatisfactionIndexACSImeasuressatisfactionratingsonascaleof0-100 3 forsevensitesandFacebookhasthelowestrating onthesocialnetworkingsitesthatweremeasuredatanindividuallevel;the ratingsare1Googleplus,Wikipedia,YouTube, Pinterest,Twitter,LinkedIn,Facebook. 4 The ACSIresultsalsoshowthatFacebooksatisfactionisdeclining;thechange fromthepreviousyearssatisfactionis-7.6%and-4.7%sincetherstyear ACSIcollectedsatisfactionratingsforsites.Despitetherelativelylow satisfactionscoresFacebookcontinuestoachievehighlevelsofcustomerattentionwherethereare1.11billionmonthlyactiveusersMarch2013. 5 3 http://www.theacsi.org/acsi-results/acsi-results 4 http://www.theacsi.org/index.php?option=com_content&view=article&id=147& catid=14&Itemid=212&i=Internet+Social+Media 5 http://newsroom.fb.com/Key-Facts 25

PAGE 47

2.3.6NamingtheDependentVariableSocialNetworkingSiteContinuanceIntention Thisresearchusestheterm socialnetworkingsitecontinuanceintention to makeclearthecontextofthestudy.Informationsystemscontinuancetheory ofBhattacherjeemeasures continuanceintention butusestheterm informationsystemscontinuanceintention andexpectation-conrmationtheory ofOlivermeasures continuanceintention andusestheterm intention. Informationsystemsresearcherstendtousethecontextofthestudytolabelthedependentvariable:Kimused SNSContinuanceIntention ,Xu etal.used stickinessintention ,HuandKettingerused usage continuance ,andLinandLuused ContinuedIntentiontoUse Researchersinmarketinghaveabitlessvarietyintheirnamingconventionsofcontinuanceintention.Caoetal.used continuanceintention Keaveneyuses switchingbehavior ,Bansaletal.use switchingintention/switchingbehavior ,KeaveneyandParthasarathyuse switching behavior ,Burnhametal.use intentiontostaywithincumbentprovider andGivonuses brandswitching .Marketingresearchersappeartohave anemphasisonswitchingtoadierentprovider.Mostoftheseresearchers arefocusedonserviceswitchingasopposedtoproductswitching,butsome studybothproductsandservices.Inthecontextofsocialnetworkingsite continuance,thefocusisonserviceswitching. 2.4SummaryofTheories Table2summarizesthetheoriesapplicabletoinformationsystemscontinuance andTable3summarizesthetheoriesfromtheconsumerswitchingliterature. Theresearchinconsumerbehaviorandinformationsystemscontinuanceintentionmayserveasatheoreticalfoundationforadditionalstudiesintocontin26

PAGE 48

uanceintentiononsocialnetworkingsites.Informationsystemscontinuance theoryislargelyguidedbysatisfaction;theconsumerbehaviorliteraturealso usessatisfactionasapredictorforswitchingbehavior.Theconsumerbehavior literaturesuggestthatanumberoffactorscanbeaddedtomoreaccurately andcompletelypredictcontinuanceintention. 27

PAGE 49

Table2:InformationSystemsTheoreticalFoundation StudyTheory/AreaIndependentVariablesDependent Variable Davis ; Davisetal. theoryofreasoned action,technology acceptancemodel TAM perceivedease-of-use, perceivedusefulness, attitudetowardusing behavioral intention-touse Bhattacherjee expectationconrmation Theory, information systems continuance conrmation, perceivedusefulness, satisfaction information systems continuance intention Caoetal. informationsystem continuance+ Maslow'shierarchy ofneeds fulllmentofsocial needs,fulllmentof self-actualization, satisfaction information systems continuance intention LinandLu motivationtheory +network externalities perceivedbenets usefulnessand enjoyment,network externalitiesnumber ofmembers,numberof peers,perceived complementarity information systems continuance intention Xuetal. commitment-trust theory+usesand graticationtheory commitment investmentsize, alternativequality, gratication,trust communication quality,opportunistic behavior, gratication, gratication stickiness intention Huand Kettinger informationsystem continuance,social exchangetheory, socialcapital theory,owtheory perceivedbenets, perceivedcosts,usage satisfaction usage continuance 28

PAGE 50

Table3:ConsumerSwitchingTheoreticalFoundation StudyTheoryName/areaIndependentVariablesDependent Variable Keaveney customerswitching behavior pricing,inconvenience, coreservicesfailures, failedservice encounters,responseto failedservices, competition,ethical problems,involuntary switching switching behavior Bansaland Taylor serviceprovider switchingmodel SPSM satisfaction,perceived switchingcosts, attitudetoward switching,subjective norms switching intention Keaveney and Parthasarathy switchingbehavior foronlineservice providers externalinuence, interpersonalinuence, usagefrequency, intensity,overall,risk takingbehavior, satisfaction, involvement switching behavior Burnham etal. typologyof switchingcosts proceduralswitching costs,nancial switchingcosts, relationalswitching costs,satisfaction switching intention Givon varietyseeking behavior brandpreference, varietyseeking probabilitymodel brand switching 29

PAGE 51

3SocialNetworkingSites 3.1SocialNetworkingSitesToday SocialnetworkssitesarewhereAmericansspendthelargestshareoftheir timeonline;Americansspendapproximately25%oftheirtimeonlineonSNS andblogsNielsen,2011.Socialnetworksitesspanavarietyofcontexts fromgeneralpurposesocialsiteslikeFacebook,Google+,andTwittermicroblogging,tocontentfocusedsiteslikeMySpacemusic,LinkedInprofessional,YouTubevideo,Flickrphotosharing,etc.Socialmediasocial networkingsitesandblogsreachthemajorityofU.S.Internetusers;approximately80%ofinternetusersaccessthesesitesandapproximately25% ofAmerican'stimeonlineisspentonthesesitesNielsen,2011.Americans spendmoretimeofFacebookthananyotherU.S.websiteNielsen,2011; Facebookhas1.11globalactivemonthlyusersMarch2013andisbelieved tobethelargestsocialnetworkingsite. 6 Socialnetworksresearchandsocialnetworkanalysisareamongtheearly formsofsociologicalresearchScott,1988.Socialnetworksareoftenmodeled asagraphwheretherelationshipsbetweenactorse.g.individuals,organizations,andtheirrelationshipsareshownBrandesetal.,2012.Therelationshipsbetweentheactorsareoftensocialrelationshipsbutmayalsoshowother relationships.Socialnetworkresearchoftenexaminespropertiesofthenetworklikecentrality,densityandconnectednessScott,1988.Socialnetwork analysishasbeenusedtoinvestigatesocialmobility,classstructure,perceptionsofclass,welfaresupport,academiccitations,etc.Scott,1988.Social networkresearchoftenwillfocusoncommunitystudiesego-centricstudies wherethefocusisonindividualactorsorinter-organizationalrelationships 6 https://newsroom.fb.com/Key-Facts 30

PAGE 52

wheretherelationshipsbetweenorganizationsarethefocusScott,1988. Theindividualandcollectiveactionsonasocialnetworkingsiteareexpected tosharesimilaritieswithsocialnetworksinface-to-facesettings;however, socialnetworkingsitesarelikelytohaveaspeciccontextforwhichtheserelationshipconnectionsareformedandunformed.Relationshipformationand dissolution,forexamplemaybeexperiencedindierentwaysonlinefromfaceto-facesettings.Inface-to-facesettingconnectionsbetweenindividualsmay bemorediculttoascertainastheremaybelesscertaintyaboutwhether aconnectionexistswhereasinonlinesettingsdyadicconnectionsareoften explicitandvisible. JoinsonexaminedmotivationsforusingFacebookandfoundthat theuserssaidthattheywantedtokeepintouchwithothersastheirmain motivationswhereotherusessuchassocialsurveillance,reconnectingwith others,etc.werelesscommon.boydexamined friending behavioron twoearlygeneralpurposesocialnetworksFriendsterandMySpace;boyd'sresearchshowsthat friends canbeactualfriends,oracquaintances,colleagues, peoplethathavenevermetinface-to-facesettings,etc.Ellisonetal.examinedthebenetsofFacebookuseandsocialcapital.Socialcapitalgenerally referstotheskillsandknowledgethatareaccessibletoanindividualthrough theirrelationshipswithothersColeman,1988.Colemannotesthatan importantformofsocialcapitalistheabilitytoacquireinformationthrough relationships;informationitselfmaybevaluedhighlyandisgenerallycostly toacquire.Accesstoalargeandweakly-tiedtiednetworkmayprovidemore benetsasmallerstrongly-tiednetworkGranovetter,1973. Thereareahostofreasonstostudysocialnetworkingsites;thesesiteshave changedthewaypeoplecommunicate,interact,collaborate,shareinformation, searchforjobs,providesocialsupport,buildsocialcapital,havepractical 31

PAGE 53

politicalimplicationsforsocialmovements 7 Nardietal.,2011.Facebook, MicrosoftandGooglesupporteconomicgrowthandinnovation;understanding howtheInternet-basedtoolsareusedmayhelpthethesetoolsreachtheir potentialNardietal.,2011.Wilsonetal.providesthreereasonswhy Facebookresearchisrelevanttosocialscientists-theonlineinteractionsleave concreteobservabledatathatcanbeanalyzed,thesitespopularity,andthe potentialbenetsanddangersofitsusee.g.socialcapitalasabenetand privacyasadanger. Theconsequencesofinformationtechnologyoftenemergeunpredictably fromcomplexsocialinteractionsbetweentechnologyandthesocialsystemsin whichtheyareembeddedMarkusandRobey,1988-onlinesocialmediais increasinglybeingusedinbusinesscontexts.Itisoftendiculttopredicthow technologyaectssocialstructuresandhowsocialstructuresshapetechnologicalinnovationOrlikowskiandRobey,1991.Informationtechnologyusers canappropriatetechnologyinwaysbeyondtheiroriginalintentOrlikowski andRobey,1991. 3.2SocialNetworkingSitesinDecline Socialnetworkingsitesmayfacegrowthanddeclinepatternstypicalofother onlineservicessuchasonlinecommunitiesofpractice.IriberriandLeroy identiedavestagemodelforcommunitiesofpracticethatcovered inceptiontodeaththatmaybeappliedtosocialnetworkingsitelifecycle models.Socialnetworkingsitesandonlinecommunitiesofpracticeareboth sociallybasedwebsiteswhereinformationissharedbetweenmembers.Pew InternetProjectaskedFacebookusersforthereasonswhytheydiscontinued useofthesiteformulti-weekperiodsRaineetal.,2013.Theresearchfound 7 http://somelab.net/some-lab-themes/ 32

PAGE 54

Table4:PewInternet-ReasonforFacebookBreaks StatedReasonforbreak% Wastoobusy/Didn'thavetimeforit21 Justwasn'tinterested/Justdidn'tlikeit10 Wasteoftime/Contentwasnotrelevant10 Toomuchdrama/gossip/negativity/conict9 Wasspendingtoomuchtimeusingthesite8 Onlyanintermittentorinfrequentuser8 Wentonvacation/trip/deployment8 Justgottired/boredwithit7 Norealreason/Justbecause6 Concernsaboutprivacy/security/ads/spam4 Didnothavecomputer/internetaccess2 Preferotherwaystocommunicate/Facebooknotreallife2 Healthorageissues2 Tookabreakforreligiousreasons1 Didn'tlikepostingallthetime/Didn'twanttoshare1 Total99 Source: PewResearchCenter'sInternet&AmericanLifeProjectOmnibus Survey,conductedDecember13to16,2012onlandlineandcellphones. N=316forFacebookuserswhohavetakenabreakfromusingthesitein thepastRaineetal.,2013. thatthetopreasonswerebeingtoobusy/didnotnottimeforit%, notinterested,didnotlike%,wasteoftime/contentnotrelevant %-seeTable4. 3.2.1CommunitiesofPracticeandOnlineCommunitiesofPractice CommunitiesofpracticeCoPandonlinecommunitiesofpracticeOCoPare sociallybasedorganizationswherememberscreate,holdandtransferknowledgeinacommunityRoberts,2006.LaveandWenger,p.98dened communityofpracticeassetofrelationsamongpersons,activitiesandthe 33

PAGE 55

world,overtimeandinrelationwithothertangentialandoverlappingcommunitiesofpractice.CoPareatypeofsocialnetworksbutaredenedwithinthe speciccontextofknowledgecreationandtransferamongitsmembers.These communitiesarenotsimplestableentitiesbutdynamicenterprisesthatevolve overtimeasnewmembersjoinandexistingmembersdepartRoberts,2006. CoPsarenotnotalwaysstronglydenedwitheasilydenablemembershipas thecommunityboundariesmayevidentRoberts,2006. Onlinecommunitiesofpracticetendtobemoreopenastheymaymore easilycrossorganizationalboundariesenabledbytechnologylikeweb2.0technologies.Thecommunitiestendtohavealifecycleofvestages: inception, creation,growth,maturity and death IriberriandLeroy,2009.Eachstage hasdierentchallengesandsuccessfactorsthatallowthecommunitytogrow orcauseitsdemiseIriberriandLeroy,2009.Socialnetworkingsitesand communitiesofpracticearelikelytohavemanycommonalitieshowevermany socialnetworksitesaremorefocusedonthecreationandmaintenanceofsocialrelationshipscomparedtotheknowledgefocusedcommunitiesofpractice IriberriandLeroy,2009.Generalpurposesocialnetworksitesmayhave moreconictsfromcontextcollapsewheretheusersarepostingaboutavarietyoftopicsvs.communitiesofpracticewhosepostingbehaviormaybe morefocusedonaparticulartopic.boydnotesthatonlinecommunitiesofpracticesharesimilaritieswithsocialnetworkingsitesbutnotesthat thesocialnetworkingsitesareaboutthemembersrstandinterestssecondthatisthecommunityisdeneegocentrically.Socialnetworkingsiteuserscan setthecontextofthediscussionsandsocialboundariesmorepersonallythan communitiesofpracticewherethesocialnormsaredenedatthecommunity levelboyd,2006;LaveandWenger,1991. 34

PAGE 56

3.2.2TakingBreaksonSocialNetworkingSites Facebookmaybeasocialnetworkingsitethatisinmaturityordeathphase decline.AccordingtothePewInternetProjectSurvey, ComingandGoing onFacebook ,Facebookusersaretakingmulti-weekbreaksfromthesiteRaine etal.,2013.ThesamePewInternetProjectsurveyfoundthat27%ofusers saidtheyplannedtospendlesstimeonthesite,3%saidtheywantedtospend moretimeonthesiteand69%saidtheyplannedtospendthesameamountof timeonthesite.ResearchshowsthatteenshavedeclininginterestinFacebook; teensstatedthatincreaseduseofadults,over-sharingandstressfuldramaare allreasonsfordecreasedenthusiasmMaddenetal.,2013.Teensstillshow interestinsocialnetworkingsitesastheiruptakeofthesocialnetworking siteTwitterhasshownsignicantgrowthMaddenetal.,2013.Twitteris usedby24%ofonlineteensin2013comparedto16%in2011.Thereare popularmediareportsthatdiscusswhyusersareleavingsiteslikeFacebook andturningtoothersocialnetworkingsites.Forexample,7ReasonsIDumped Facebook 8 statedthat:1Facebooksuckstimefrommylife,2Mostofmy Facebookfriendsaren'tactuallyfriends,3Thereareotherbetteroptions forphotosharing,4Facebookbringsouttheworstinpeople,5Ilearnmore onTwitter,6ThepresenceofadsonFacebookisgettingridiculous,7Less ismore.Thedescriptiveresearchbythepopularmediaaboutcontinuance onthesitedoesnotfocusonsatisfaction,themainantecedenttoinformation systemcontinuancebutonotherfactors. 8 http://www.forbes.com/sites/timmaurer/2013/06/20/7-reasons-i-dumpedfacebook/ 35

PAGE 57

3.2.3EngagementLevelConcernsofFacebook Facebooknotesthatitsactiveusergrowthratewilldeclineovertimeashigher marketpenetrationratesareachieved. 9 Facebook's10-Klingnotesthat,A numberofothersocialnetworkingcompaniesthatachievedearlypopularity havesinceseentheiractiveuserbasesorlevelsofengagementdecline,insome casesprecipitously.Thereisnoguaranteethatwewillnotexperienceasimilarerosionofouractiveuserbaseorengagementlevels.Ouruserengagement patternshavechangedovertimeandcanbediculttomeasure,particularly asusersengageincreasinglyviamobiledevicesandasweintroducenewand dierentservices.Thelingalsonotesthatuserretention,growth,andengagementmaybenegativelyaectedasusersincreasinglyengagewithother productsoractivities.Thereportsuggestthatsocialnetworkingsiteservices fromothercompaniesmaybeactingasproductsubstitutesandnotproduct complementsandmayerodeFacebook'sengagementlevels.Xuetal. notethatthesocialnetworkingsiteMySpacehadmorethan100millionusers before2008butfellto54.4millionbyNovember2010.MySpacehadamarket valuationof580millionwhenNewsCorpboughtthesiteinJuly2005to35 millionwhenSpecicMediapurchasedthesiteinJune2011. 10 Xuetal. notethatsocialnetworkingsitesrelyonusergeneratedcontentandsuggest thatthelongtermviabilityofasocialnetworkingsitedependsoncontinued userparticipation. 9 http://investor.fb.com/secfiling.cfm?filingID=1326801-14-23&CIK=1326801 10 http://allthingsd.com/20110629/exclusive-myspace-to-be-sold-tospecific-media-at-35-million/ 36

PAGE 58

4PredictingEnd-UserSatisfactionthroughPerceivedEase-of-Use andPerceivedUsefulness-Study1 Study1-Summary Thisstudypredictsend-usersatisfactiononFacebookthroughtwoconstructs:perceivedease-of-useandperceivedusefulness.Usefulnesswasoperationalizedtoreecttheperceptionthatthesiteishelpfulforfriendship maintenanceandsocialsurveillance.TheresultsshowthatFacebookusers perceivethesitetobebotheasy-to-useanduseful.Perceivedease-of-useand perceivedusefulnesswerestatisticallysignicantpredictorsforsatisfaction; theperceivedusefulnessofthesitehasagreaterimpactonsatisfactionthan perceivedease-of-use.Severalcontrolvariableswereincludedtoadjusttheresults.Satisfactionisstatisticallysignicantlyhigherforuserswhohavemore friendsandinteractwithmorepeoplecomparedtothosewhohadfewerfriends andfewerinteractions.Satisfactionisnotstatisticallydierentformalesand femaleswhenaccountingforusers'perceptionofusefulnessandease-of-use. ThisstudyappliestheISsuccessmodeldevelopedformoreutilitariansystemstoahedonicsocialnetworksite. Thisstudyexaminesend-usersatisfactionthroughtwoestablishedconstructs:perceivedease-of-useandperceivedusefulnessDavis,1989.Perceivedusefulnessisoperationalizedinthisstudyashowusefulthesiteisfor maintainingfriendshipsandconductingsocialsurveillance.Socialnetwork siteslikeFacebookallowuserstoaccumulatesocialcapital;however,thesite appearstobenetweak-tierelationshipsmorethanstrong-tierelationships Ellisonetal.,2007;Vitaketal.,2010.Relationshipstrengthmayvaryfrom weak-tostrong-ties,althoughthereissomeconsensusthatthemajorityof tiesonFacebookareweakVitaketal.,2010;LewisandWest,2009. AlthoughFacebookisapopularsiteintermsoftheproportionoftimespent onlinethatfactalonedoesnotindicatethattheusersarehighlysatisedwith thesite,orbelieveitiseitherusefuloreasy-to-use.Facebookusersmayhave 37

PAGE 59

avarietyofmotivationstousethesitebeyondtheseelements.SNSmight feelcompelledtousethesitethroughsocialpressureorbecausetheynd itnecessarytomonitortheirnetwork.Joinsonexaminedmotivations forusingFacebookandfoundthattheuserssaidthattheywantedtokeep intouchwithothersastheirmainmotivationswhereotherusessuchas socialsurveillance,reconnectingwithothers,etc.werelesscommon.Studies onend-usersatisfactionhavelargelyfocusedonutilitarianwebsitesleaving hedonicsystemsunder-researchedOngandDay,2010;Schaupp,2011.The purposeofthisresearchistodeterminehowtheperceivedease-of-useand perceivedusefulnessofFacebookcontributetoend-usersatisfaction.There isstrongevidencethatFacebook,attheU.S.nationallevel,isquitepopular; however,thisanalysisisattheuser-levelandbringsinsightintotheuser-level attributesforsatisfactionandspecicallyexaminesthehedonicwebsiteof Facebook.Thestudywasconductedbyanalyzingsurveyresponsesof1,552 FacebookuserswhowererecruitedthroughTwitter. 4.1LiteratureReview boydandEllisonadenedsocialnetworksitesbasedonthreesystem capabilities.Thesystems:allowindividualstoconstructapublicorsemipublicprolewithinaboundedsystem,articulatealistofotherusers withwhomtheyshareaconnection,andviewandtraversetheirlistof connectionsandthosemadebyotherswithinthesystemboydandEllison, 2007a,p.210.Afterusersjoinasitetheyareaskedtoidentifyothersin thenetworkwithwhomtheyhaveanexistingrelationship.Thelinksthatare generatedbetweenindividualsbecomevisibletoothersinthelocalsubspace. Therearegeneralsocialnetworksitese.g.Facebookandothersthatare contentfocused. 38

PAGE 60

Davisetal.1989adaptedTRAtoexplainISacceptancethroughtwo constructs:perceivedusefulnessPUandperceivedease-of-usePEOUto explainsystemusers'intention-to-useandadoptionbehavior.Davis, p.320dened perceivedusefulness as"thedegreetowhichapersonbelieves thatusingaparticularsystemwouldenhancehisorherjobperformance,"and dened perceivedease-of-use as"thedegreetowhichapersonbelievesthat usingaparticularsystemwouldbefreeofeort."Conceptually,anapplicationthatisperceivedtobeeasiertouseismorelikelytobeadopted,ceteris parabus.TheinitialresearchbyDavis,p.320wasaboutadoption withinorganizationalcontextswhereemployeesaregenerallyreinforcedfor goodperformancebyraises,promotions,bonuses,andotherrewards,however;TAMhasbeenusedtopredictacceptanceinmanyISdomainsoutside theorganizationalcontext.Thetwoconstructs,PEOUandPU,canpredicta user'sattitudetowardthesystemandtheirbehavioral intention-to-use Davis etal.,1989.Davisetal.statedthatTAMwasdevelopedtocreate amodeltoexplainthedeterminantsofcomputeracceptancethatwascapableofexplaininguserbehavioracrossabroadrangeofend-usercomputing technologiesanduserpopulations. ResearchrelatedtoTAMhasemphasizeddierentaspectsovertimeastheoryhasbecomemoreacceptedandwidespread.Leeetal.discussfour phasesintheprogressionofresearchrelatedtoTAM;thephasesare:model introduction,modelvalidation,modelextensionandmodelelaboration.Duringthemodelintroductionphase,researchersreplicatedTAMinavarietyof settingsandoverlongertimeperiods.Researcherscomparedandcontrasted thefeaturesofTAMwithTRAtodeterminewhichtheorybetterexplains adoption.Duringthevalidationphase,researchersexaminedtheinstrument inanumberofdierentsettingstodeterminewhethertheconstructshadac39

PAGE 61

ceptablediscriminantandconvergentvalidity.Themodelvalidationphase examinedTAM'sboundaryconditionsandinvestigatedthemoderatingeects suchas,culture,gender,task,usertimeandIStypeLeeetal.,2003,p. 757.Duringthemodelelaborationphase,TAMwasmodiedtoaddressconcernssuchasmulti-levelanalysisandaddressconcernsregardingthecontext voluntaryandmandatorysettings. DeLoneandMcLean;2003D&MexaminedISsuccessthroughsix variables:systemquality,informationquality,use,usersatisfaction,individual impactandorganizationalimpacttodeterminewhatmakesagiveninformationsystemsuccessful.TheD&Mmodelidentiedtherelationshipsbetween thevariablesandcautionedthatadditionalresearchwasneededPetterand McLean,2009.Attheindividuallevel,end-usersatisfactionisamajorcomponentofISsuccess;thevariableisoneofthemostwidelyusedsinglemeasures ofISsuccessDeLoneandMcLean,1992.DeLoneandMcLeancaution usersoftheirmodeltoselectthedependentvariableofsuccessthatisappropriatetotheobjectivesandthecontextoftheempiricalinvestigation.The D&M2003modelincludesafeedbackloopbetween use and usersatisfaction because use necessarilyprecedes usersatisfaction andgreater satisfaction will increase use .Asusersincreaseuseandsatisfactiontheuserwillaccumulate netbenetsthe2003successmodelredenedindividualandorganizational impactsas netbenets andcanhelpdeterminethesuccessofaninformation system.PetterandMcLeanmeta-analysisofISsuccessstudiesfound thattherewasstatisticallysignicantrelationshipbetweenintention-to-use basedonTAMandend-usersatisfaction. JoinsonexaminedFacebookusers'motivationsandusesofSNSs. Theresearchidentiedthemesfromopen-endedsurveyquestionsandfound that keepingintouch [withfriendsandacquaintances]receivedthelargest 40

PAGE 62

shareofresponses.3%,and socialsurveillance secondlargestcategory accountedforanadditional17.3%.Othercategoriesincludedreconnecting withlostcontacts,communicatingwithothers,viewingphotographs,etc.A minorityofrespondentswereinterestedinmakingnewfriends.5%.Joinsonfoundthatwomenusedthesitemoreoftenthanmenandwasthelargest positivepredictorforfrequencyofvisitstoFacebook.Theresearchalsofound thatthosewhousedthesiteforsocialinvestigationdenedinthestudyas virtualpeoplewatching,stalkingothers,etc.usedFacebookmoreoften.The largestnegativepredictorsforfrequencyofusewerethosewhouseFacebook toupdatetheirpersonalstatusorviewstatusupdatesfromtheirnetworkfollowedbythosewhousethesitetolookatphotographs.KwonandWen researchedhowsixfactorssocialidentity,altruism,telepresence,PEOU,PU, perceivedencouragementcouldpredictsystemuseonSNS.Theirresearch foundthatPEOUandPUhadthestrongeststatisticallysignicantrelationshiptopredictactualsystemusage. 4.2StudyDesign Basedonpreviousresearch,thisinvestigationwasdesignedtopredictendusersatisfactiononthesocialnetworksiteFacebookwithtwoconstructs: perceivedease-of-useandperceivedusefulness.Perceivedusefulnesshasbeen operationalizedinthiscasetomeanfriendshipmaintenanceandsocialsurveillance,whicharetwoofthedominantmotivationsofFacebookusersJoinson, 2008.ThisresearchusestheestablishedtheoryfromDavis'sTechnologyAcceptanceModelDavis,1989fortwooftheconstructs, perceivedease-of-use and perceivedusefulness ,topredict end-usersatisfaction. TAMwasdesigned topredictcomputerusers'behavioralintention-to-useasystemDavis,1989. InthisresearchthesurveypopulationalreadyadoptedFacebook;therefore, 41

PAGE 63

thisresearchusesend-usersatisfactionasthedependentvariable. Intentionto-use and systemusage havebeenshowninpreviousstudiestobelinkedto end-usersatisfactionPetterandMcLean,2009;DeLoneandMcLean,2003. Baroudietal.foundbothuser informationsatisfaction and userinvolvement tohaveastatisticallysignicantcorrelationwith systemusage. SeveralstudieshavefoundPEOUandPUtohavestatisticallysignicantrelationshipswithrelatedmeasuresofusersatisfactionandusageSrinivasan, 1985;PetterandMcLean,2009;DeLoneandMcLean,2003;Adamsonand Shine,2003.AdamsonandShinepredicted end-usersatisfaction using perceivedease-of-use and perceivedusefulness forabanktreasuryapplication. AdamsonandShine'sresearchintosatisfactionwasinthecontextofa mandatoryenvironment,i.e.theusersofthesystemwererequiredtousethe systemtocompletetheirworkandhadnoviablealternatives.ISresearchers havefocusedtheirattentionmoreonutilitarianusesofinformationsystems e-commerce,task-orientedcomputing,oce-typeapplicationscomparedto morehedonicuseslikeSNS,bloggingsites,entertainment,etc.Ongand Day,2010;Schaupp,2011.ThisresearchfocusesonFacebookwherethe usersvoluntarilychoosetousetheonlinesocialnetworkforlargelyhedonic purposes. Thereareseveralcontrolvariablesusedtoadjusttheprimaryconstructs inthestudy;thecontrolvariablesare:age,gender,locationresideinU.S.or outsideU.S.,numberofinteractionswithFacebookusers,numberoffriends onthesiteandyearsofsocialnetworksiteuse.Ageandgenderhavebeen foundtohavemoderatingeectsonbehavioralintentiontouseandusage GefenandStraub,1997;Joinson,2008;Leeetal.,2003;KingandHe,2006; PetterandMcLean,2009.Severalstudieshavefoundculturaldierences basedonlocationinend-usersatisfactionLeeetal.,2003;KingandHe, 42

PAGE 64

2006;PetterandMcLean,2009;thisstudyuseslocationasaproxytoculture todeterminewhetherU.S.Facebookusershavedierentlevelsofend-user satisfactionthanthosewhoresideoutsidetheU.S.The numberofinteractions measuresthenumberoffriendswithwhomtheusertypicallyinteractsandmay berelatedtothebridgingsocialcapitalthatusersobtainsfromthesiteYoder andStutzman,2011.Joinsonfoundseveraldierencesinfrequency ofuseandtimespentonthenetworksitethatvariedbasedonthe number offriends onthesiteandmayhaveanimpactonsatisfaction.Thevariable yearsofsocialnetworksiteuse isusedasaproxyforSNSself-ecacy.Users whohaveusedSNSforlongerperiodsoftimewhichincludessitesotherthan Facebook,suchasMySpacemayberelatedtothatuser'sSNSself-ecacy andmayhaveaneectonsatisfaction.Thecontrolvariablesarenotthe primarypredictivevariablesinthisresearchbutareusedtocontrolforuser dierences. Basedontheliteraturereviewandresearchquestion,thisstudyproposes thefollowinghypothesestopredicttherelationshipbetweenperceivedeaseof-use,perceivedusefulnessandend-usersatisfaction: H1:Perceivedease-of-useofFacebookincreasestheperceivedusefulnessof Facebook H2:Perceivedease-of-useofFacebookincreasestheend-usersatisfaction ofFacebook H3:PerceivedusefulnessofFacebookincreasestheend-usersatisfactionof Facebook Figure3showstheresearchmodelusedforthisinvestigation. 43

PAGE 65

Figure3:ResearchModel&ControlVariables 4.2.1Survey Theresearchwasconductedusingasurveytodeterminethesurveyrespondent'sopinionsandbehaviorsabouttheirsatisfactionwithFacebook.The surveywasconductedsolelyontheInternetusingacommerciallyavailable surveytool.ThesurveyusedestablishedquestionsfrompreviousstudiesoperationalizedtopredictFacebookend-usersatisfaction.Demographicitems wereusedascontrolvariablesintheanalysis. Partsoneandtwoofthesurveyaskedquestionsunfriendingbehaviorsbothunfriendingthesurveyrespondentperformedandunfriendingthatwas donetothesurveyrespondent.Partthree,thefocusofthisstudy,asked questionsaboutsatisfaction,perceivedusefulnessandperceivedease-of-useof Facebook.Partfourasksdemographicquestions:age,gender,education,the numberofyearsofsocialnetworkingsiteuseandwhetherthepersonlives intheUnitedStatesofAmerica.Thecontrolvariableswerecategoricalin natureandcollectedthroughself-reportmeasures,e.g.theuserswereasked howoftentheywenttoFacebookeachdayonaveragebutwerenotaskedto tracktheirusage,provideaccountnamesorotheridentifyinginformation. 44

PAGE 66

4.2.2DataCollection SurveyrecruitmentwasconductedbysendingTwitteruserswhopostedabout Facebookatweetthataskedtheusertotakeasurveyaboutthesite.Twitter wasusedtorecruitsurveyparticipantsforseveralreasons:Twitterhasalarge userpopulationwherethemajorityofusershavepubliclyaccessiblemessages; Twitterusershadagoodtwithresearchsocialnetworksites;itisasimple processtocontactapersononTwitterthroughthe@replymechanism;and thetweetscanbescreenedforrecruitmentpurposes.Thereisnotarandom sampleinthisresearch;apurposivesamplingmethodwasusedtorecruit participants.Therecruitmenttweetwassentinasingletweetof140characters andprovidedenoughinformationtotheTwitterusertotakethesurvey.It wasnotconsideredundulyburdensomebytheresearcherortheInstitutional ReviewBoardIRBtosendasinglerequesttotheTwitterusertotakea survey. TherecruitmenttweetwasdesignedtofollowthemethodologyofDillman etal.asmuchaspossiblewithintheconstraintsofTwitter.Dillman etal.statethatemailsforsurveyrecruitmentshouldincludethefollowingsections:universitysponsorshiplogo,header,etc.,informativesubject heading,currentdate,appealforhelp,statementastowhythesurveyrespondentwasselected,usefulnessofsurvey,directionsonhowtoaccesssurvey, clickablelink,individualizedIDfortracking,statementofcondentiality andvoluntaryinput,contactinformationaboutsurveycreator,expressionof thanks,andindicatethesurveyrespondent'simportance. Themeasuresusedtooperationalizetheconstructswereadaptedfrompreviousresearch.Table5showsconstructsandtheiritems.Validatedmeasures frompriorresearchwereusedinthesurveywherepossiblealthoughminor changesinquestionswerenecessarytobemeaningfulforthecontextofthis 45

PAGE 67

survey.Allofthesurveyitems,exceptopen-endedresponsesanddemographic information,weremeasuredusinga7-pointLikertscalefrom StronglyDisagree to StronglyAgree SurveyswerecollectedbetweenApril17thandSeptember15,2010for151 totaldays.7,327recruitmenttweetsweresentduringthetimeperiod.Atotal of2,865surveyswerestartedand1,552werecompleted;54%ofthosewho startedthesurveycompletedthesurvey.Thesurveyswerestartedby39.6% ofthosewhoweresenttweetsandcompletedby21.3%.Thesurveyis15pages inlengthandtook,onaverage,approximately18minutestocomplete.Twitter respondentsweregatheredbyscreeningtweetsthathadthetermunfriend, defriend,orunfriending.Tweetsthatmetascreeningcriterionweresent repliesinvitingthepersontotakethesurveyaboutunfriending.Thetweet replysentwasretweetedbymanypeoplewhoreceivedtheinitialtweet. 4.2.3DataAnalysis TherawdatawascollectedfromacommerciallyavailablesurveytoolSurvey MonkeyandanalyzedwithSPSSversion18andstructuralequationmodelingSEMwascompletedwithAMOSversion18andPLS-Graphbuild1130. AMOSwasusedfortheinitialmeasurementmodeltoexaminethecategorical variablesatamoregranularlevelthanPLSallowsinthesubsequentstructural model.SEMhelpsexplaintherelationshipbetweenmultiplevariablesincludinglatentunobservablefactorsinasingle,systematicandcomprehensive mannerHairetal.,2006;GerbingandAnderson,1988;Gefenetal.,2000. Theanalysisusedameasurementmodeltoassessgoodnessoftfortheoverall modelandfactoranalysis,constructvalidityandreliabilityconvergentand discriminantforthelatentconstructsHairetal.,2006;Gefenetal.,2000; Wetzelsetal.,2009.Afterthemeasurementmodelwasvalidated,astructural 46

PAGE 68

modelwasgeneratedtodeterminethepathcoecientsbetweentheconstructs andtheassociatederrorcoecientswiththecontrolvariablesincluded.This researchusestwoindependentvariables perceivedease-of-use and perceived usefulness topredictonedependentvariable satisfaction .Twoitemsfrom thesurveyweredroppedbecausetheirloadingswerebelowa.5threshold, ortheysharedlittleproportionofvarianceincommonwithotheritemsin itsconstructandwerestronglycorrelatedwiththeitemsofotherconstructs. Controlvariablesfortheresearchinclude:age,gender,whethertheperson livesintheUnitedStates,numberoffriends,numberoffriendswithwhom thepersoninteracts,andyearsofsocialnetworkingsiteuse. Statisticalanalysisalonecannotprovecausation,becauseitdoesnotestablishisolationortemporalorderingBollen,1989.Correlationanalysis, includingSEM,canbeusedtoshowthatthecorrelationsfoundinthedata areinaccordancewiththecausationpredictedbyanestablishedtheory-base Bollen,1989;Gefenetal.,2000.Structuralmodelsmayovertthemodel tothedatasoitimportanttohaveatheoreticalbasisfortestingthetheory intheanalysisGefenetal.,2000.Thisresearchappliesestablishedresearch modelsandtheoriesregardingend-usersatisfactionandisagoodcandidate toapplySEMinaconrmatorymanner. 4.3Results ThestructuralmodelwasgeneratedusingthepartialleastsquaresPLS method,sinceitfocusesonpredictionoftheconstructsratherthanexplanationoftherelationshipsbetweenitemsHairetal.,2006.Thelatentfactors wereanalyzedforconvergentanddiscriminantvaliditybyanalyzingthefactorloadings,averagevarianceextractedAVE,constructreliabilityCRand crossloadingthroughcorrelationofconstructs.Factorloadingsfortheitems 47

PAGE 69

forthelatentconstructsPEOU,PUandSATwereallgreaterthan.70andindicatethattheitemsconvergeonthelatentconstructatadequatelevelsHair etal.,2006-seeTable5.Compositereliabilityandvaliditywasexamined forappropriateness.Compositereliabilityforallfactorsrangefrom.88to.93 andexceed.70whichsuggestsadequatereliabilityHairetal.,2006.Table6 showsthecorrelationestimatesbetweentheconstructsandthesquarerootof theaveragevarianceextractedAVEforeachconstructonthediagonal.The squarerootsoftheAVEarefrom.83to.85.Thesevaluesarehigherthanthe correlationestimates,whichrangefrom.55to.72.Thecorrelationestimate betweenPUandSAT,andPEOUandSATarehigh.72and.69,respectively; however,thereislittlecross-loadingamongthemeasuredvariables.Overall, theseresultssupportthediscriminantvalidityofthemodel. Figure4showsthepathcoecientsforthestructuralmodel.Themodel isassessedbythepathcoecientsandthecoecientofdetermination R 2 values,sincePLSdoesnotproduceanoverallgoodnessoftindices.Allofthe hypothesesweresupported.PEOU'seectonPUissupported,0.50, p <.001. PEOU'seectonSATisalsosupported,0.41, p <.001.Finally,PU'seect onSATissupported,0.47, p <.001.The t valuesforthesignicantpaths rangefrom19.16to22.38. R 2 valuesmeasuretheproportionofthevariance ofthelatentendogenousvariablethatisexplainedbythelatentexogenous variablesHairetal.,2006.TheendogenousvariablesareSATandPU,and theexogenousvariablesarePEOUandPU.The R 2 valuesshowthatSAT hasthelargestshareofthevarianceexplainedthroughthefactors R 2 =.65, andPUthesecond R 2 =.34,whicharebetweensubstantial R 2 =.67and moderate R 2 =.33Chin,1998. TheeectofageonPEOUissignicant,-0.15, p <.001.Theeectof genderonPUissignicant,0.06, p <.01.Theeectofwhethertheperson 48

PAGE 70

Table5:CompositeReliability Construct Abbr ItemsMeanStd. Dev Composite Reliability No.of Items Perceived ease-of-use PEOU Easytolearn, easytouse, easytond whatIneed 5.391.14.883 Perceived Usefulness PU Usefulin maintaining friends,helpful inmaintaining friends,helpsme knowwhatmy friendsare doing,maintain friendcontact information 5.561.02.904 Satisfaction SAT Enjoy,satised, featuresIlike, highquality, betterthan otherSNS, recommendto others 5.081.23.936 MeanandStandardDeviationbasedonLikertScores1-7where1isstrongly disagreeand7isstronglyagree 49

PAGE 71

Table6:Correlationoftheconstructs PEOUPUSat PEOU .85 PU.55 .83 SAT.69.72 .83 DiagonalvaluesarethesquarerootsoftheAVE Figure4:ModelPathCoecients&ControlVariables livesintheU.S.onPUissignicant,-0.05, p <.05.Theeectsofthenumber offriendsonPEOU,0.14, p <.001;PU,0.09, p <.01;andSAT,0.06, p < .01aresignicant.Theeectofthenumberoffriendswithwhomtheperson interactsonPEOU,0.18, p <.001;andPU,0.13, p <.001aresignicant, Finally,theeectofthenumberofyearsofsocialnetworkingsiteuseonPEOU, -0.06, p <.05;andSAT,-0.05, p <.01aresignicant.Table7summarizes thehypothesessupportedandtheeectofCVs. 50

PAGE 72

Table7:Supportforhypothesesandtheeectsofthecontrolvariables HypothesisEstimate t -statstic H1:PEOU PU0.5019.16*** H2:PEOU SAT0.4119.21*** H3:PU SAT0.4722.38*** Age PEOU-0.155.69*** Age PU0.041.74 Age SAT0.031.75 Gender PEOU0.020.89 Gender PU0.062.91** Gender SAT0.000.19 U.S.vs.NonU.S. PEOU-0.010.36 U.S.vs.NonU.S. PU-0.052.18* U.S.vs.NonU.S. SAT-0.021.23 No.Friends PEOU0.144.16*** No.Friends PU0.093.00** No.Friends SAT0.063.19** No.Interactions PEOU0.186.37*** No.Interactions PU0.135.11*** No.Interactions SAT0.041.67 YrsSNSUse PEOU-0.062.08* YrsSNSUse PU0.010.59 YrsSNSUse SAT-0.052.84** p <.05;** p <.01;*** p <.001 51

PAGE 73

4.4Discussion&Conclusion Socialnetworkingsitesmakeupthelargestshareoftimespentonlineinthe U.S.accordingtoNielsen.Facebookitselfisverypopular,itisconsideredthelargestsocialnetworkingsitebynearlyafactorofthreewithover 140millionuniquevisitorsinMay,2011comparedtothesecondlargestsocialnetworksite,Blogger,whichhadapproximately50millionuniquevisitors forthesamemonthNielsen,2011.Popularity,byitself,doesnotindicate thattheusershavehighorlowend-usersatisfactionorshowhowthetwo constructs,perceivedusefulnessandperceivedease-of-usecontributetosatisfaction.PopularityatNielsen'slevelofanalysisisatthemacro-level,this researchinvestigatestherelationshipofend-usersatisfactionattheindividual level. TheDeLoneandMcLeanupdatedISsuccessmodelshowthatthere isafeedbackloopbetweenuse,end-usersatisfactionandthenetbenetsthat areaccrued.Ameta-analysisoftheD&MmodelbyPetterandMcLean foundthattherelationshipbetweenactualsystemusethroughselfreport,actualuse,depthofuseandimportanceofuse,dependingonthestudy andusersatisfactiontobeweakbutstillstatisticallysignicantthroughthe analysisof26relatedstudies.Themeta-analysisfoundthatintention-touseandusersatisfactionhasastrongandstatisticallysignicantrelationship withusersatisfactionthroughtheanalysisof9relatedstudies.Petterand McLeanproposethatonereasonintention-to-useandusersatisfaction ismorestronglylinkedisthattheintention-to-useismorereliablethanthe use measuresintherelatedstudies.WhenSNSusers use aninformationsystem andattain end-usersatisfaction theusermayachieve netbenets whichmay feedbackintomore use and end-usersatisfaction ThenetbenetsattheindividualthatmaybeaccruedthroughFace52

PAGE 74

bookusemaybesocialcapital.Socialcapitalgenerallyreferstotheskills andknowledgethatareaccessibletoanindividualthroughtheirrelationships withothersColeman,1988.Colemannotesthatanimportantformof socialcapitalistheabilitytoacquireinformationthroughrelationships;informationitselfmaybevaluedhighlyandisgenerallycostlytoacquire.Access toalargeandweakly-tiedtiednetworkmayprovidemorebenetsasmaller strongly-tiednetworkGranovetter,1973.Ellisonetal.foundastrong positiverelationshipwithFacebookuseandbridgingsocialcapital.Facebook usewasalsofoundtoincreaselevelsofbondingandmaintenanceofsocialcapitalinthesamestudybutatlowerlevelsthanbridgingsocialcapital.Vitak etal.2010foundthatFacebookusehadlimitedeectsonbondingcapital; i.e.Facebookusedidnotstronglyimpactstrong-tierelationships.Facebook usersmaybelievethatincreasedsocialcapitalisanetbenetofsiteuse. Thepurposeofthisresearchistodeterminehowtheperceivedease-ofuseandperceivedusefulnessofFacebookcontributetoend-usersatisfaction. ThisresearchusesTAMwithage,gender,whetherthepersonlivesinthe U.S.,numberoffriends,numberoffriendswithwhomthepersoninteracts, andyearsofsocialnetworkingsiteuseascontrolvariables.Theresultsshow thatFacebookusers somewhatagree thattheyaresatisedwiththesite. Facebookusersaresatisedwiththesitewhentheynditusefulandeasyto use.Thersthypothesis, H1 ,positsthatperceivedease-of-useofFacebook increasestheperceivedusefulnessofFacebook,whichissupported.Thesecond hypothesis, H2 ,positsthatperceivedease-of-useofFacebookincreasestheendusersatisfactionofFacebook,whichissupportedwiththethirdhypothesis, H3 ,whichpositsthattheperceivedusefulnessofFacebookincreasesend-user satisfactionwithFacebook.TheseresultscoincidewithpreviousTAMresearch usedtopredictacceptanceinmanyISdomains,asstatedintheliterature 53

PAGE 75

review.TheresultsarehelpfulinthathedonicwebsiteslikeFacebookappear tosharesimilarrelationshipsthatutilitarianwebsiteshaveregardingend-user satisfaction. Facebookusers'satisfactionwasimpactedmorebyitsusefulnessthanits easeofuse,butbothfactorsareimportantinpredictingsatisfaction.UsefulnesshasbeenoperationalizedinthisresearchintermsoffriendshipmaintenanceandsocialsurveillancebasedonpreviousstudiesJoinson,2008.The usefulnessfoundinthisstudyhasagoodtwithFacebook'sstatedpurpose wheretheystatethatthesiteisasocialutilitythathelpspeoplecommunicatemoreecientlywiththeirfriends,familyandcoworkers. 11 Userswho foundFacebookeasy-to-usehadhighersatisfactionwiththesitethanthose whofounditdiculttouse. Theresultsalsoshowinterestingeectsofcontrolvariablesonperceived ease-of-use,perceivedusefulness,andsatisfaction.Olderusersperceivethesite tobemorediculttousethanyoungerusers.Youngerusersmaybemore familiarwiththeInternet,state-of-the-artITtechnology,etc.thanolderusers; however,agedoesnotaectperceivedusefulnessorsatisfactiononFacebook. FemaleusersperceiveFacebooktobemoreusefulthanmaleusers.Thisresult issimilartoGefenandStraub'sstudyofwomenhavinghigherperceived usefulnessofemailthanmen.However,unliketheirresultthatmenhave higherperceivedease-of-useforemailusage,genderdoesnoteectperceived ease-of-useonFacebook.GenderdoesnoteectsatisfactiononFacebook either.U.S.-basedFacebookusersndthesitetobemoreusefulthanthose wholiveoutsidetheU.S.;however,therearenodierencesinperceivedeaseof-useorsatisfactionbasedonlocation.Thenumberoffriendscontrolvariable isnotableinthatiteectsallthreeconstructsandistheonlycontrolvariable 11 http://www.facebook.com/press.php 54

PAGE 76

todoso.Whenusershavemorefriends,theyhavehigherperceivedease-ofuse,perceivedusefulness,andsatisfactionthanuserswhohavefewerfriends. Whenusershavemoreinteractions,theyhavehigherperceivedease-of-use andperceivedusefulnessthanuserswhohavefewerinteractions.Ingeneral, increasingthenumberofinteractionswithusersmayleadtheusertoincreased siteuse,whichmakesusersfeelmoreateasewiththesite.However,number offriendswithwhomthepersoninteractsdoesnoteectsatisfaction.Finally, whenusershavemoreyearsofSNSexperience,theyhavehigherperceived ease-of-useandsatisfactionthanuserswhohavelessexperience.However, yearsofSNSusedoesnoteectperceivedusefulness. LimitationstothisstudyincludetheuseofTwittertorecruitsurveyrespondents.ThesurveyrespondentsweretweetingaboutaparticularactunfriendingonSNSsandmayhavemoreintensityofuseonSNSsthenthe generalFacebookuser.Thesurveyrespondentswereuserswho,ataminimum,usedtwoSNS,FacebookandTwitter.Anadvantagetosurveyingthis populationisthattheage,andlocationbiaseswerelowercomparedtotraditionalmethodsofFacebookresearchstudentsurveysandad-hocmethods likepostingiersoncampusandonlineforumrecruitment.ThisstudyexaminedFacebookonamacro-levelanddidnotdistinguishbetweenavarietyof featuresavailableonthesite,e.g.chatfeatures,commercialinterests,games, entertainment,etc. FutureresearchwillneedtoexpandthedenitionofusefulnessofFacebookbeyondfriendshipmaintenanceandsocialsurveillancetoexamineother aspectsofFacebooksuchastheentertainmentvalueandcommercialvalueof thenetwork.Facebookhasbeenadoptedbycommercialinterestswhotryto buildbrandloyaltyanddisseminateinformationthroughthelikefunctionality.WhenFacebookuserslikeapagetheymakeaconnectiontothatpage 55

PAGE 77

andcontentwillappearintheuser'stimelineandmayappearintheuser's newsfeed.Facebook'sstrategyhascontinuedtoevolvewhereithasbecome aplatformforapplicationsbeyondstatusupdates.MarkZuckerberg,Facebook'sfounder,stated,Thepastveyearshavebeenaboutbeingconnecting peopleandthenextvetotenyearsareaboutwhatareallthethingsthatcan bebuiltnowthattheseconnectionsareinplaceRao,2011.Theease-ofuseconstructcanbeexpandedtocovermorespecicpartsofFacebook,e.g. chatfeatures,gamingfeatures,privacycontrols,togainmoreinsightintothe usabilityofspecicFacebookareas.Futurestudiescanaddressthespecic functionsofFacebookatamoregranularleveltodeterminehowsatisedusers arewithfeatureslikeconnectingtocommercialparties,entertainment,games, chat,etc.beyondconnectingtoSNSusers. Facebookusers'satisfactioncanbepredictedbytwoconstructs,perceived ease-of-useandperceivedusefulness.ThisstudysuccessfullyappliesthetraditionalISsuccessmodeldevelopedformoreutilitarianpurposestoahedonic socialnetworksite. 56

PAGE 78

5FactorsinSocialNetworkingSiteContinuanceIntention Theorybuildingandtheorytestingarebothcriticalfortheadvancementof knowledgeBhattacherjee,2012.Theoriesshouldmatchempiricalrealities andempiricaldatashouldbeabletocontributetothedevelopmentofmeaningfultheoriesBhattacherjee,2012.Theorybuildingtendstobemorevaluablewhentherearefewertheoriesintheeldtoexplainphenomenaandtheory testingtendstobemorevaluablewhentherearemultiplecompetingtheories ofthesamephenomenaBhattacherjee,2012.Thissectionwilldescribethe factorsthatareimportantinsocialnetworkingsitecontinuanceanddevelop testablehypothesesforevaluation. Theproposedmodelofsocialnetworkingsitecontinuanceintentionfor socialnetworkingaddsveconstructstotheBhattacherjeemodelof IScontinuance: personalinnovativeness,habit alternativeperceptions,interpersonalinuence and consumerswitchingcosts -seeFigure5.Thereare vehypothesesgeneratedbasedonthebackgroundpresentedearlierandthe theoreticalargumentsprovidedinthissection. 5.1ProductSubstitutionsandAlternativePerceptions ThediusionofinnovationstheoryofRogersdenedamodelforwhich innovationsareadopted;themodelhadfourmainelements:an innovation is communicated over time tomembersofa socialsystem .Theinitialdiusions ofinnovationsmodelassumedthatanyinnovationisindependentofallother innovationsandhasbeenextendedincludemorecomplexitytoaccountfor otheradoptionsMahajanetal.,1990. Informationsystemsarenotintroducedinstaticenvironments,thereare productcomplementsandsubstitutionsthatmayaddorremovevaluefrom 57

PAGE 79

anyinnovationMahajanetal.,1990.Productsintroducedtothemarketplacemayhaveapositiveornegativeinuenceonadoptiondecisions.Certain productsmayrequireaproducttobeinthemarketplacebeforesupporting productscanbeintroducedproductcomplements.Newproductsmayreplaceproductsthathavealreadybeeninthemarketproductreplacement. Modelsofadoptionshouldmoreaccuratelyconsideradoptionbeyondasingle adoptionbyasinglepotentialadopterMahajanetal.,1990.Consumers mayberst-timeadoptersorrepeatconsumersandmaydecidetoeitherrepeattheirpastconsumerdecisionsorreplacecurrentproductsorserviceswith dierentsystems.Mahajanetal.recommendsthatstudiesofdiusions ofinnovationsconsiderhowcompetitionandrivalrymayinuencethegrowth ofaproductcategoryandhowitmayaecttheentry/exitpatternsofcompetitors.Theresultofanewproductenteringthemarketmaybeanexpansionof themarket,diversionofdemandfromincumbentstocompetitorsoracombinationoftheoutcomesMahajanetal.,1993.Newproductintroductionsoften expandthemarketbecauseproductvarietyincreasesandpotentialcustomers maybecomeawareofnewlyintroducedproducts,althoughthesemarketexpansioneectsaremoreoftenseeninlessmatureenvironmentsMahajan etal.,1993. ParthasarathyandBhattacherjeeexaminedtheroleofreplacement substitutionanddisenchantmentlowsatisfactiononservicediscontinuance. Theresearchshowedthatapproximately60%ofsubscriberstoonlineservices whodiscontinuedtheservicedidsobecausethesubscriberwasdisenchanted withtheservice.Approximately35%ofsubscribersdiscontinuedservicebecausethesubscriberchoseareplacementservice;5%ofusersdiscontinuedfor unknownreasons.ParthasarathyandBhattacherjeefoundweaksupportfortheirhypothesisthatsubscriberswhoreplacedtheironlineservices 58

PAGE 80

areearlyadoptersandthosewhoweredisenchantedwiththeirservicearelate adopters.Theresearchfoundthatthosewhoreplacedtheserviceusedthe servicemoreextensivelyduringtheinitialperiodcomparedtothosewhowere disenchantedwiththeservice.Thissuggeststhatsubscriberswhousedthe servicelesswereunableordisinclinedtofullyusetheservice.Theinability tosuccessfullyexploittheserviceoptionsmayleadtodis-conrmationgap betweentheexpectationsandconrmationresultswhereasthosewhocould morefullyutilizetheserviceweremorelikelytochangeserviceproviders. Facebookcitedthatsomeofitsusersmaysubstituteservicesfromother providersforitsownservice.Facebook's10-Klingsaidthefollowing: 12 Webelievethatsomeofourusers,particularlyouryounger users,areawareofandactivelyengagingwithotherproductsand servicessimilarto,orasasubstitutefor,Facebook.Forexample, webelievethatsomeofourusershavereducedtheirengagement withFacebookinfavorofincreasedengagementwithotherproductsandservicessuchasInstagram.Intheeventthatourusers increasinglyengagewithotherproductsandservices,wemayexperienceadeclineinuserengagementandourbusinesscouldbe harmed. 5.1.1AlternativeAttractivenessinConsumerServiceSwitching Themarketingliteratureoftenuses attractivenessofalternatives asameasure toexamineserviceproviderswitching Jonesetal.,2000;Bansaletal.,2005; SharmaandPatterson,2000. SharmaandPattersonnotethatrelationshipbetweenserviceproviderandserviceconsumeriscomplicateddueto 12 http://www.sec.gov/Archives/edgar/data/1326801/000132680113000003/fb12312012x10k.htm 59

PAGE 81

highswitchingcostsofchangingprovidersandbecausetherecanbealackof anattractivealternativeprovider. Jonesetal.,p.262denes attractivenessofalternatives asthe,customerperceptionsregardingtheextenttowhichviablecompetingalternatives areavailableinthemarketplace.Conceptually,consumerswhoperceivethat therearefeweralternativestousingagivenservicearemorelikelytostay withtheservice,andconsumerwhoperceivethattherearemoreavailable alternativesaremorelikelytoswitchJonesetal.,2000.Jonesetal.tested twohypothesesintheirwork;theresearcherstestedboththedirecteectof alternativeattractivenessandtheinteractioneectwherealternativeattractivenessinteractswiththerelationshipbetweensatisfactionandrepurchase intentionor,whatininformationsystemsistypicallycalledcontinuanceintention.Thedirecteectofalternativeattractivenesswasnotsupported, howevertheinteractioneectwassupported. Bansaletal.examinedboth alternativeattractiveness and attitude towardswitchin gtopredict stickinessintention[continuanceintention] and switchingbehavior .Alternativeattractivenessisdescribedasapullfactor,or afactorthatmaypulltheserviceuserawayfromanexistingserviceprovider toanewserviceprovider.Bansaletal.,p.100claimsthatalternative attractivenessisthe,onlyexistingvariablefromtheserviceswitchingliteraturethatconformstothisconceptualization[push-pulleects]isalternative attractiveness.The attitudetowardswitching measuresthedegreetowhich aserviceconsumermaybefavorablydisposedtoswitchingserviceproviders. Havingafavorableattitudetowardswitchingmayindicatethattheconsumer ismorewillingtoswitch.Theresearchsupportedthatalternativeattractivenesswasastatisticallysignicantfactorinpredictingcontinuanceintention. Theindividualeectofattitudetowardswitchingisunclearintheresearchre60

PAGE 82

sultsbecausethefactorisnotevaluatedasanindividualfactorbutthroughthe largerconstruct mooring. Attitudetowardswitching,however,wasthelargest factorinpredictingthe mooring eectsreasonswhypeoplewouldcontinue withanincumbentserviceprovider,whichwasstatisticallysignicantfactor inpredictingswitchingintention. SharmaandPattersonnotesthatalackofalternativeserviceproviders benetsserviceproviderstoretaincustomersandthatwhenservicecustomers areunawareofattractivealternativesthentheymaystayintherelationship despiteperceivedlowsatisfactionlevels.SharmaandPattersonexaminedtherelationshipofalternativeattractivenessasamoderatorbetweentrust andrelationshipcommitmentcontinuanceintentionandservicesatisfaction andrelationshipcommitmentcontinuanceintention.Theresearchfound thatalternativeattractivenesswasanimpactfactorinpredictingrelationship commitment.Thendingswere: 1.Satisfactionhasastrongerimpactonrelationshipcommitmentunder conditionsofhighalternativeattractiveness. 2.Trusthasastrongerimpactoncommitmentunderlowalternativeattractiveness. 3.Underconditionsofhighalternativeattractiveness,theimpactofsatisfactiononrelationshipcommitmentisstrongerthanthatoftrust. 4.Underconditionsoflowalternativeattractiveness,theimpactoftrust onrelationshipcommitmentisstrongerthanthatofservicesatisfaction. 61

PAGE 83

5.1.2Alternativeattractivenessappliedininformationsystemsresearch Zhangetal.examinedtheroleofgenderinbloggerswitchingbehaviorwithtwofactorssatisfactionandalternativeattractivenessandmediating factorsofgenderandsunkcost.Zhangetal.notesthatalternative attractivenessissimilartorelativeadvantageintheinformationsystemsdiffusionsandadoption.Theresearchfoundthatuserswhoperceivedattractive alternativestotheircurrentbloggingplatformhadhigherintentionstoswitch theirservice. Hsiehetal.includedthreerst-orderfactors,enjoyment,relative usefulnessandrelativeease-of-usetomeasurethesecond-order pulleects to thealternativeattractivenessofcompetingserviceproviders.Pulleectshad thelargestimpactofthreefactors,pusheects,mooringeectsandpulleects thatpredictedswitchingintentionofbloggerstonewbloggingplatforms.This studyexaminedbloggersintentiontomovefromanexistingbloggingplatform toFacebook. BhattacherjeeandParkusedthepush-pull-mooringframeworkto examinewhyinformationsystemsusersmayintendtomigratetocloudservice providers.Theresearchersusedtheconceptofalternativeattractivenessto measurethedierencebetweentheexistingserviceproviderproductorservice withsubstituteservices.BhattacherjeeandParkusedtwofactorsin thepullmodel,relativeusefulnessandexpectedomnipresenceforthesecond orderfactor pull whichtheyusedtopredictintentiontomigrateswitchservice providersfromtheexistingprovidertoacloudservice.Theresearchersfound thatrelativeusefulnesshadstatisticallysignicantpositiveeectonintention toswitchandexpectedomnipresencehadamarginallysignicantpositive impactonintentiontoswitch. 62

PAGE 84

AlternativePerceptionsHypotheses AlternativePerceptionswillhave adirecteectonSNSContinuanceintention.Theriseofbothgeneralpurposeandspecialinterestsitesmayshowthatthesystemenvironmentisnot staticbutdynamicasproductsareintroduced.Newsocialnetworkingsites maycomplementtheexistingsystemsorbeareplacement.Keaveney notesthatsomefactorsarebeyondthecontrolofabusiness,likecompetition. Someresearchersviewvarietyseekingasinexplicablebutothersmodelvariety seekingbehavioras derived or direct Givon,1984.Theresearchonvariety seekingbehavioralsoindicatesthatpeoplemakechoicestovaryconsumption amongcompetingproducts.Userswithhighlevelsofpersonalinnovativeness maybemorepronetoreplacinganexistingsystemwithanewonewhereas userswithlowerlevelsofpersonalinnovativenessmaybemorelikelytobe servicediscontinuersbasedonParthasarathyandBhattacherjee.Socialnetworkingsiteusersmayexhibithighlevelsofsatisfactionandmaystill replacesystemswithnewsystemsasinnovationsareintroducedtothemarketplace. Hypothesis1:Socialnetworkingsiteuserswithhighapositiveattitudeto switchandwhoareattractedtocompetingsocialnetworkingsiteswillnegativelyaectcontinuanceintentiononthesocialnetworkingsite. 5.2PersonalInnovativeness Individualadoptersmayhavedierentpredispositionstoadoptnewtechnologies.Somepotentialuserswillreadilyadoptanewinformationsystemwhile otherswillrejectthemAgarwalandPrasad,1998.AgarwalandPrasad proposedusingtheconstruct personalinnovativeness toexplicitlydenehowusersmayadopttechnologybasedonpsychometriccharacteristics. 63

PAGE 85

Individualswhohavehigherlevelsofpersonalinnovativenessmayadoptinnovationsearlierthanothersandmayactaschangeagentsandopinionleaders tofurtherdiuseanewtechnology. Rogersdevelopedvecategoriesofadopters: innovators,early adopters,earlymajority,latemajority and laggardsMahajanetal.,1990. Traditionallyadoptercategoriesaredenedby timeofadoption howeverthis categorizationhasalsobeencriticizedfornegativemethodologicalconsequences AgarwalandPrasad,1998;Mahajanetal.,1990.Theuseof timeofadoption limitstheabilitytocomparestudiesacrossdierentproductsandhas reliabilityandvalidityissues.Personalinnovativenessisnotdirectlyvisible inthetechnologyacceptancemodelTAMofDavisalthoughAgarwal andPrasadarguethatthereisstrongtheoreticalandempiricalsupportofthecharacteristicsroleinnovationinadoption.AgarwalandPrasad ,p.206denedpersonalinnovativenessas,thewillingnessofanindividualtotryoutanynewinformationtechnology.Personalinnovativeness ishypothesizedtomoderatetheantecedentsandconsequencesofperceptions relatedtoadoption,i.e.theperceptionsaboutnewinformationsystemsand theintentiontouseanewinformationsystemaremoderatedbypersonalinnovativenessAgarwalandPrasad,1998.AgarwalandPrasadalso recommendthatpersonalinnovativenessbeusedasacontrolvariableinindividualstudiesandthatthefactormayaccountforasignicantportionofthe varianceininnovation-relateddependentvariables.Basedontheproperties ofthefactor,personalinnovativenessisusedasadirectfactorforpredicting socialnetworkingsitecontinuanceandasamoderatingfactor. ThatcherandPerrewappliedpersonalinnovativenesstounderstandtherelationshipbetweendynamicindividualattributese.g.computer anxietyandcomputerself-ecacyandstableindividualattributese.g.per64

PAGE 86

sonalinnovativeness,negativeaectivity,andtraitanxiety.Theresults showedthatpersonalinnovativenesshasapositiverelationshipwithselfecacyandnegativerelationshipwithcomputeranxiety. 5.2.1PersonalInnovativenessHypotheses Personalinnovativenessisexpectedtohaveadirectandmoderatingeecton continuanceintentionbasedonAgarwalandPrasadandThatcherand Perrew.PersonalinnovativenessisderivedfromRogers'sdiusionsof innovationtheoryandadaptedtotheinformationsystemdomain. Hypothesis2a:Personalinnovativenesswillnegativelyaectcontinuance intentionofsocialnetworkingsites. Hypothesis2b:Therelationshipbetweensatisfactionandsocialnetworking sitecontinuanceintentionwillbemoderatedbypersonalinnovativenesson socialnetworkingsites. 5.3InterpersonalInuences Rogersbelievedthatdiusionsresearchshouldfocusbeyondtheindividualtodyads,cliques,networksorsystemofindividuals.Socialsystemscan beexpectedtoexertasignicantinuenceonindividualadoptionandcontinuanceordiscontinuanceoftechnology.Theroleofopinionleadsmayhavea strongerinuenceonpotentialusersthenothermembersofasocialnetwork. Granovettersuggeststhatweak-tierelationshipsmayhaveastronger eectonanindividualreceivercomparedtostrong-tierelationships;itislikely thatadoptionsofinnovationsmaybestronglyinuencedbyinterpersonalrelationshipsandbythosewhoarenotparticularlyclosetoanotherpotential adopter.Thesocialnetworkmayinuencetheadoptionorcontinuanceinten65

PAGE 87

tionsofsocialnetworkingsiteusers.ThemodelofdiusiondevelopedbyBass assumesthatpotentialadoptersofaninnovationareinuencedbyoneoftwo means-themassmediaandwordofmouthMahajanetal.,1990. Innovators, theearlieradoptersofaninnovation,areinuencedbymassmediaexternal inuence,whereas imitators areinuencedbywordofmouthcommunicationsinterpersonalinuence.Themodelsinformresearchbyprovidinguser typologiesandhowtheymaybeinuencedastheyadoptinnovativeproducts andservices. ParthasarathyandBhattacherjeeexaminedtheroleofinterpersonal inuenceonservicecontinuation;theresearchshowedthatuserswhodiscontinuedservicesaremoreinuencedbyinterpersonalsourcesofinformation duringtheadoptionphasethanthosewhocontinuedservices.Thatis,users whostoppedusingasystemafteracceptanceweremoreinuencedbytheir relationshipsthanthosewhocontinuedtousethesystem.Theresearchalso foundthatdiscontinuersofonlineserversarelessinuencedbyexternalsources e.g.mediaduringtheinitialacceptancethanthosewhocontinuedtheusing theservice.Earlyadoptersexhibitdierentpatternsofdiscontinuancecomparedtolateadopters;earlyadopterstendtobeinuencedbyexternalsources suchasmassmediaandlateadopterstendtobeinuencedbyinterpersonal informationParthasarathyandBhattacherjee,1998. Hardgraveetal.investigatedtheroleofvefactorsonsoftware developersdecisionstofollowmandateddevelopmentmethods.Theresearch foundthatinterpersonalinuenceswereastrongfactorinpredictingthebehavioralintentionsofsoftwaredeveloperstoconformtotheprescribedmethod andstrongerthanorganizationalmandates.Thefourfactorsthatdetermined intentiontofollowthemethodare usefulness,socialpressure,compatibility, organizationalmandate and complexity ,inorderofhighestpredictivepower. 66

PAGE 88

Therstfourfactorswerestatisticallysignicantpredictorsofintentionto followthemethod,complexitywasnotfoundtobeasignicantfactor.The researchshowsthatinterpersonalinuenceandexternalinuencecanaect useradoptionandcontinuanceintentions. Kimexaminedtheroleofinterpersonalinuenceoninformation systemcontinuanceforsocialnetworkingsites.Theresearchfoundthatwhen membersofauser'ssocialnetworkbelievedthatusingasocialnetworkingsite wasagoodideauserswerepositivelyinuencedtocontinueusingthesite. 5.3.1InterpersonalInuencesHypotheses Interpersonalinuenceisexpectedtohaveadirecteectoninformationsystem continuance.Kimfoundastatisticallysignicantrelationshipbetween howmembersperceivedtheusefulnessofsocialnetworkingsitesandcontinuanceintention.Socialnetworkingsitesareinherentlysocialsomembersof one'ssocialnetworkcanbeexpectedtohaveaninuenceonwhetherauser continuestouseasite. Hypothesis3:Interpersonalinuencewillnegativelyaectcontinuanceintentionofsocialnetworkingsites. 5.4Habit User habits mayhaveasignicantroleininformationsystemscontinuance decisionsLimayemetal.,2007.Limayemetal.,p.705dened habitforinformationsystemusageas,theextenttowhichpeopletendto performbehaviorsuseISautomaticallybecauseoflearning.Theresearch foundthatantecedentstoinformationsystemcontinuanceincludecontinuance intentionmoderatedbyhabit.Limayemetal.modelusedthreefactors 67

PAGE 89

topredicthabit:satisfaction,comprehensivenessofusage,andfrequencyof pastbehavior.Habit'srolewastestedasamoderatorbetweeninformation systemcontinuanceintentionandinformationsystemactualusageandas adirecteecttoinformationsystemactualusage.Limayemetal. reviewspriorresearchwherehabitisusedasa:directfactoronbehavior, asanindirecteectandasamoderatorbetweenintentionsandbehavior inseveralcontextswithinandoutsideofinformationsystemsresearch. GefenandStraubsuggeststhathabithasastrongtheoreticalrole asadirectfactorinbehavioralintentions.GefenandStraubargues thatrepeatedexposuretoasystemwilldevelop habitofuse andincreaseintentiontouseasystem.Theuseofhabitwithotherfactorsforcontinuance intentionallowsthatuserscanmakebotharationalassessmentofcontinued useperceivedusefulnessandperceivedease-of-useandfactorswhichlack rationalassessmentlikehabitGefenandStraub,2003.Thisresearchmodel includeshabitasbothdirectandmoderatingfactorasapredictorofcontinuanceintention. PewInternetProjectfoundthatwhileteenshavelessenthusiasmforthe socialnetworkingsiteFacebooktheycontinuetousethesiteastheteenagers statethesiteisanimportantpartofoverallteenagesocializingMaddenetal., 2013.Teensstatedthattheydislikedincreasingadultpresenceonthesite, thattheir friends sharedexcessiveinformationandfoundthatthesitehada lotofdrama.Despitethedicultyofthesiteteensfeltaneedtocontinue usingthesitesotheydidnotmissoutonsocialinteractionsandinformation. Thestudyalsofoundthatteenswithmorethan600friendsvisitedthesite severaltimesaday.Themajorityofteenswhousesocialnetworkingsites saidtheyvisitthesitedaily.ThePewInternetProjectstudymaysuggest thathabithasaroleinsocialnetworkingsitecontinuancebehaviorasteen 68

PAGE 90

behaviorappearstohaveastrongcomprehensivenessofusageandfrequency ofpastbehaviorwhicharebothantecedentstoLimayemetal.model forhowhabitmayimpactinformationsystemusage. 5.4.1HabitHypotheses HabitisexpectedtohaveadirectandmoderatingeectoncontinuanceintentionbasedonLimayemetal.,GefenandStraubandPew InternetProject'sresearchonteenageusageofsocialnetworkingsitesMaddenetal.,2013.Whensocialnetworkingsiteusershabituallyusethesystem theirsatisfactionhasasuppressanteectoncontinuanceintention,i.e.those whousethesitehabituallywillcontinuetousethesiteevenwhiletheirsatisfactionmaybelow. Hypothesis4a:Habitwillpositivelyaectcontinuanceintentionofsocial networkingsites. Hypothesis4b:Therelationshipbetweensatisfactionandsocialnetworking sitecontinuanceintentionwillbemoderatedbyhabit. 5.5SwitchingCosts TheresearchbyBurnhametal.supportedthatsatisfactionandthe threeswitchingcostsprocedural,nancialandrelationalwereallsignicant predictorsofserviceproviderswitchingintention.Theswitchingcostshada largerimpactoncontinuanceintentionthansatisfactionintheBurnhametal. study.Manyofthecostscanbeadaptedforuseinastudyforsocial networkingsitestoincreasetheamountofexplainedvarianceinthemodel. Proceduralswitchingcostsinclude:economicriskcosts,evaluationcosts, setupcosts,learningcosts.Burnhametal.,p.111describeseconomic 69

PAGE 91

riskcostsas,thecostsofacceptinguncertaintywiththepotentialforanegativeoutcomewhenadoptinganewprovideraboutwhichtheconsumerhas insucientinformation.Evaluationscostsarethecostsrelatedtodeterminingwhetheraswitchshouldbemade,includingcollectingandanalyzingdata regardingaswitch.Learningcostsarethenewskillsthatneedtobeacquired toeectivelyuseanewproduct.Setupcostsdescribethecostsofacquiringandsettingupaproductforinitialuse;e.g.onasocialnetworkauser mayneedtosetupproleinformationandestablishanetworkforthesiteto functionlikeasocialnetwork. Financialswitchingcostsofswitchingtoanewsocialnetworkingsiteare lowtonon-existent,e.g.switchingfromFriendstertoMySpace,MySpaceto Facebook,orFacebooktoTwitterhavenonancialcosttotheuserasallsites arefree.ThetwonancialcostsinBurnhametal.arebenetlosscosts e.g.accruedpoints,discounts,etc.andmonetarylosscostse.g.initiation fees,deposits,etc..Whilesocialnetworkingsiteusersmayloseinformationif theydiscontinuesiteusee.g.couponsthatmaybeoeredthecostsarelow. Themonetarylosscostsarealsolowtonon-existentonmanygeneralpurpose socialnetworkingsitesasthesitesaremostlyoeredforfree. Therelationalswitchingcostsincludepersonalrelationshiplosscostsand brandrelationshiplosscostBurnhametal.,2003.Therelationalcostsin Burnhametal.werethestrongestpredictorofcontinuanceintentionby afactorofthree 13 andmayalsobeastrongpredictorforcontinuanceintention onsocialnetworkingsiteasthesitesaredene,inpart,bytherelationshipsin thenetwork.Personalrelationshiplosscostsaredenedas,theaectivelosses associatedwithbreakingthebondsofidenticationthathavebeenformed withthepeoplewithwhomthecustomerinteractsBurnhametal.,2003,p. 13 Standardizedcoecients are:relational:0.44,procedural0.15andnancial0.13. 70

PAGE 92

111-112.Theconsumerbehaviorliteraturenotesthatthesecostsaredueto thechangeswiththeexistingserviceprovideremployeesanddoesnotdirectly applytosocialnetworkingsitesassuch.Themeasurecanbemodiedinthis contexttonotetheaectivelossesofbreakingthesocialnetworkingsiteties ofthesocialnetworkingsitemembers.Brandrelationshipcostsarerelatedto thelossesthatoccurwhenthebondsofidenticationthatwereformedwith abrandarenowbroken.Consumerformbondsthroughtheirpurchasesand theservicescanprovideasenseofidentity. Theproceduralandrelationalswitchingcostsmayserveasastrongfoundationforstudyingcontinuanceintentiononsocialnetworkingsites.The nancialcostsdonotneedtobeincludedinthemodelastheydonotapply;however,theproceduralandrelationalcostscanbeadaptedtopredict continuanceintentiononsocialnetworkingsites. 5.5.1SwitchingCostsHypotheses Usersofaninformationsystemmayconsidertheswitchingcostsinvolvedin changingsocialnetworkingsites.Burnhametal.citedthreecostsassociatedwithswitchingcosts,procedural,nancialandrelational.Thereare fewnancialconsiderationswhenswitchingsocialnetworkingsites;however, proceduralandrelationalcostsmayaectcontinuancedecisions. Hypothesis5:Greaterproceduralandrelationalswitchingcostswillpositivelyaectcontinuanceintentionofsocialnetworkingsites. 71

PAGE 93

Table8:SocialNetworkingSiteContinuanceModelHypotheses NumHypothesis 1 Socialnetworkingsiteuserswithhighapositiveattitudetoswitch andwhoareattractedtocompetingsocialnetworkingsiteswill negativelyaectcontinuanceintentiononthesocialnetworking site. 2a Personalinnovativenesswillnegativelyaectcontinuanceintention ofsocialnetworkingsites. 2b Therelationshipbetweensatisfactionandsocialnetworkingsite continuanceintentionwillbemoderatedbypersonalinnovativeness onsocialnetworkingsites. 3 Interpersonalinuencewillnegativelyaectcontinuanceintention ofsocialnetworkingsites. 4a Habitwillpositivelyaectcontinuanceintentionofsocial networkingsites. 4b Therelationshipbetweensatisfactionandsocialnetworkingsite continuanceintentionwillbemoderatedbyhabit. 5 Greaterproceduralandrelationalswitchingcostswillpositively aectcontinuanceintentionofsocialnetworkingsites. 72

PAGE 94

Figure5:SocialNetworkingSiteContinuanceModel 73

PAGE 95

6ResearchDesignandMethod 6.1InstrumentDesign Theinvestigationintosocialnetworkingsitecontinuancewillbeconducted throughaninternet-basedsurveytool.Thesurveyquestionsareacombinationofestablishedquestionsfrompreviousstudiesandadaptedforthis investigation-seeSection6.2.Thesurveywillbeginwithacoverlettertointroducethesurveyusingthe TailoredDesignMethod byDillmanetal.. Thesurveywillscreenuserstodetermineifthepersonisover18andaFacebookuser.ThesurveywasreviewedandapprovedbytheColoradoMultiple InstitutionalReviewBoardattheUniversityofColoradoDenverpriortoits administration;theprotocolnumberis COMIRBProtocol13-3126 74

PAGE 96

6.2SurveyQuestions 6.2.1PerceivedUsefulness Davis,p.320dened perceivedusefulness inthetechnologyacceptance modelTAMas,"thedegreetowhichapersonbelievesthatusingaparticularsystemwouldenhancehisorherjobperformance."Thequestionsare adaptedtotFacebook's perceivedusefulness inmaintainingfriendshipsand socialsurveillance.JoinsonexaminedmotivationsforusingFacebook andfoundthattheuserssaidthattheywantedtokeepintouchwithothers astheirmainmotivationswhereotherusessuchassocialsurveillance,reconnectingwithothers,etc.werelesscommon.SibonaandChoiusedthe questionstoexaminesatisfactionofFacebookthroughtheconstructs perceived usefulness and perceivedease-of-use PerceivedUsefulnessDavis etal. AdaptedstudyquestionsSibonaandChoi 1UsingWriteOnewouldimprove myperformanceintheMBA Program Facebookishelpfulinmaintaining friends 2UsingWriteOneintheMBA programwouldincreasemy productivity Facebookhelpsmeknowwhatmy friendsaredoing 3UsingWriteOnewouldenhance myeectivenessintheMBA Program Facebookisusefulinmaintaining friends 4IwouldndWriteOneusefulin theMBAProgram Facebookhelpsmemaintainmy friend'scontactinformation 75

PAGE 97

6.2.2Conrmation Bhattacherjee,p.353dened conrmation asan,[assessment]of theperceivedperformancevis--vistheiroriginalexpectationanddetermine theextenttowhichtheirexpectationisconrmed.Theconstructionwas operationalizedinthestudyas,Users'perceptionofthecongruencebetween expectationofonlinebankingdivisionOBDuseanditsactualperformance, Bhattacherjee,p.359. ConrmationBhattacherjee Adaptedstudyquestions 1MyexperiencewithusingOBD wasbetterthanwhatIexpected MyexperienceusingFacebookwas betterthanwhatIexpected 2TheservicelevelprovidedbyOBD wasbetterthanwhatIexpected Theservicelevelprovidedby FacebookwasbetterthanwhatI expected 3Overall,mostofmyexpectations fromusingOBDwereconrmed Overall,mostofmyexpectations fromusingFacebookwere conrmed 4Myexperiencewithusingthis websitewasbetterthanwhatI expectedKimetal.,2009 ThebenetsprovidedbyFacebook arebetterthanwhatIexpected 76

PAGE 98

6.2.3Satisfaction Oliver,p.461dened satisfaction asanadditivecombinationofthe expectationlevelandtheresultingdisconrmationBhattacherjee,p. 359operationalized satisfaction asthe,users'aectwithfeelingsabout prioronlinebankingdivisionOBDuse,basedonSprengetal. overall satisfaction scale. SatisfactionBhattacherjee Adaptedstudyquestions 1Howdoyoufeelaboutyour overallexperienceofOBDuse: verydissatised-verysatised Howdoyoufeelaboutyour overallexperienceofFacebook use:verydissatised-verysatised 2Howdoyoufeelaboutyour overallexperienceofOBD use:verydispleased-verypleased Howdoyoufeelaboutyour overallexperienceofFacebook use:verydispleased-verypleased 3Howdoyoufeelaboutyour overallexperienceofOBDuse: veryfrustrated-verycontented Howdoyoufeelaboutyour overallexperienceofFacebook use:veryunhappy-veryhappy 4Howdoyoufeelaboutyouroverall experienceofOBDuse:absolutely terrible-absolutelydelighted N/A 77

PAGE 99

6.2.4Habit Limayemetal.7,p.705denedhabitas,theextenttowhichpeople tendtoperformbehaviorsuseIS[informationsystems]automaticallybecauseoflearning.Limayemetal.notethathabit,asdened,has littleoverlapwithintentionandmayprovideadditionalexplanatorypower beyondinformationsystemusage.Limayemetal.usedhabitasa moderatingconstructbetweeninformationsystemscontinuanceintentionand informationsystemsconrmationusage.Thestudydoesnotdenetheitems forcontinuanceusage-buttheauthorsstatethatthevariableisbasedontwo formativequestions.Thisstudyuses habit asamoderatingconstructbetween satisfaction and continuanceintention. HabitLimayemetal. Adaptedstudyquestions 1UsingtheWWWhasbecome automatictome UsingtheFacebookhasbecome automatictome 2UsingtheWWWisnaturaltomeUsingtheFacebookisnaturalto me 3Whenfacedwithaparticulartask, usingtheWWWisanobvious choiceforme Whenfacedwithaparticulartask, usingtheFacebookisanobvious choiceforme 78

PAGE 100

6.2.5Personalinnovativeness AgarwalandPrasad,p.206denepersonalinnovativenessinthedomainofinformationtechnologyPIITas,thewillingnessofanindividual totryoutanynewinformationtechnology.AgarwalandPrasad,p. 206statethatpersonalinnovativenessarepersonaltraitsandthattheyare notinuencedbyenvironmentalorinternalvariables,i.e.thetraitsarenot expressedbysituationally. PersonalInnovativenessAgarwalandPrasad Adaptedstudyquestions 1IfIheardaboutanewinformation technologyIwouldlookforways toexperimentwithit IfIheardaboutanewsocial networkingsiteIwouldlookfor waystoexperimentwithit 2Ingeneral,Iamhesitanttotry outnewinformationtechnologies Ingeneral,Iamhesitanttotry outnewsocialnetworkingsite 3Amongmypeers,Iamusuallythe rsttotryoutnewinformation technologies Amongmypeers,Iamusuallythe rsttotryoutnewsocial networkingsite 4Iliketoexperimentwithnew informationtechnologies Iliketoexperimentwithnew socialnetworkingsite 5IliketoexplorenewWebsites McKnightetal.,2002 Iliketoexplorenewsocial networkingsites 79

PAGE 101

6.2.6InterpersonalInuence ParthasarathyandBhattacherjee,p.369dened interpersonalinuenceas, therelativeinuenceofexternalandinterpersonalinformationsourcesonrespondents'initialadoptiondecisions.Theconstructis basedoninnovationdiusiontheoryRogersthatexaminesinuence inpurchasingdecisionsbasedonexternalsourcesandinterpersonalsources ParthasarathyandBhattacherjee. InterpersonalInuenceParthasarathyandBhattacherjee Adaptedstudyquestions 1Howmuchdideachofthe followingsourcesinuenceyouto subscribeto[this]service?opinion offriends,colleagues,relativesor others Howmuchdidtheopinionofyour friendsinuenceyoutouse Facebook 2Howmuchdidtheopinionofyour classmatesinuenceyoutouse Facebook 3Howmuchdidtheopinionofyour familymembersinuenceyouto useFacebook 4Howmuchdidtheopinionof othersinuenceyoutouse Facebook 80

PAGE 102

6.2.7ProceduralandRelationalCosts Burnhametal.examinedthreecosts:procedural,nancialandrelational,inthestudyofconsumerbehaviorresearch.Thedominantgeneral purposesocialnetworkingsitestodayhavenonancialcostssothisstudyfocusesontheproceduralandrelationalcosts.Burnhametal.,p.112 dened proceduralswitchingcosts as:theeconomicrisk,evaluation,learning, andsetupcosts,thistypeofswitchingcostprimarilyinvolvestheexpenditure oftimeandeort,and relationalswitchingcosts asthepersonalrelationship lossandbrandrelationshiplosscosts,thistypeofswitchingcostinvolvespsychologicaloremotionaldiscomfortduetothelossofidentityandthebreaking ofbonds.TheBurnhametal.studyviewstherelationalcostsasbetweenbuyerandsellerwherethisstudyadaptsthequestionstoonlinesocial networkingsitesastherelationalcostsofdissolvingtheconnectionsbetween users. 81

PAGE 103

Procedural-EconomicRisk Costs -Burnhametal. Adaptedstudyquestions 1Iworrythattheserviceoeredby otherserviceproviderswon'twork aswellasexpected Iworrythatothersocialnetworks won'tworkaswellasexpected 2IfItrytoswitchserviceproviders, Imightendupwithbadservice forawhile. IfItrytoswitchtoanothersocial networkingsite,Imightendup withnegativeimpactsforawhile. 3Iamlikelytoendupwithabad dealnanciallyifIswitchtoa newserviceprovider. notapplicable 4Switchingtoanewservice providerwillprobablyresultin someunexpectedhassle Switchingtoanewsocial networkingsitewillprobably resultinsomeunexpectedhassle 5Idon'tknowwhatI'llendup havingtodealwithwhile switchingtoanewserviceprovider Idon'tknowwhatI'llendup havingtodealwithwhile switchingtoanewsocial networkingsite 82

PAGE 104

Procedural-EvaluationCosts -Burnhametal. Adaptedstudyquestions 1Icannotaordthetimetogetthe informationtofullyevaluateother serviceproviders. Icannotaordthetimetogetthe informationtofullyevaluateother socialnetworkingsites. 2Howmuchtime/eortdoesittake togettheinformationyouneedto feelcomfortableevaluatingnew serviceproviders?:verylittle-alot Ifeelverycomfortableevaluating anewsocialnetworkingsitetosee ifitsuitsme. 3Comparingthebenetsofmy serviceproviderwiththebenets ofotherserviceproviderstakes toomuchtime/eort, evenwhenI havetheinformation. Comparingthebenetsof Facebookwiththebenetsof othersocialnetworkingsitestakes toomuchtime/eort, evenwhenI havetheinformation. 4Itistoughtocomparetheother serviceproviders Itistoughtocomparetheother socialnetworkingsites 83

PAGE 105

Procedural-LearningCosts Burnhametal. Adaptedstudyquestions 1Learningtousethefeatures oeredbyanewserviceprovider aswellasIusemyservicewould taketime. Learningtousethefeatures oeredbyanewsocialnetworking siteaswellasIuseFacebookwill taketime. 2Thereisnotmuchinvolvedin understandinganewservice providerwell. Thereisnotmuchinvolvedin understandinganewsocial networkingsitewell. 3Evenafterswitching,itwould takeeorttogetuptospeed withanewservice Evenafterswitching,itwould takeeorttogetuptospeed withanewsocialnetworkingsite 4Gettingusedtohowanother serviceproviderworkswouldbe easy. Gettingusedtohowanother socialnetworkingsiteworkswould beeasy. 84

PAGE 106

Procedural-SetupCosts Burnhametal. Adaptedstudyquestions 1Ittakestimetogothroughthe stepsofswitchingtoanewservice provider. Ittakestimetogothroughthe stepsofswitchingtoanewsocial networkingsite. 2Switchingserviceproviders involvesanunpleasantsales process. Switchingsocialnetworkingsite involvesanunpleasantsales process. 3Theprocessofstartingupwitha newserviceisquick/easy Theprocessofstartingupwitha newsocialnetworkingsiteis quick/easy 4Therearealotofformalities involvedinswitchingtoanew serviceprovider. Therearealotofformalities involvedinswitchingtoanew socialnetworkingsite. 85

PAGE 107

Relationship-Personal RelationshipLossCosts Burnhametal. Adaptedstudyquestions 1Iwouldmissworkingwiththe peopleatmyserviceproviderifI switchedproviders. Iwouldmissmyfriendson FacebookifIswitchedtoa dierentsocialnetworkingsite. 2Iammorecomfortableinteracting withthepeopleworkingformy serviceproviderthanIwouldbeif Iswitchedproviders. Iammorecomfortableinteracting withmyfriendsonFacebookthan IwouldbeifIswitchedtoa dierentsocialnetworkingsite. 3ThepeoplewhereIcurrentlyget myservicemattertome. MyfriendsonFacebookmatterto me. 4IliketalkingtothepeoplewhereI getmyservice. Iliketalkingtomyfriendson Facebook. Relationship-Brand RelationshipLossCostsBurnhametal. Adaptedstudyquestions 1Ilikethepublicimagemyservice providerhas IlikethepublicimageofFacebook 2Isupportmyserviceproviderasa rm. IsupportFacebookasarm. 3Idonotcareaboutthe brand/companynameofthe serviceproviderIuse. IdonotcareabouttheFacebook brand. 86

PAGE 108

6.2.8AlternativePerceptions Themajorityoftheresearchexaminesalternativeserviceprovidersinonelarge disambiguatedgroupwithoutspecicity,i.e.thequestionresearcherstendto askiswhetherserviceconsumerisattractedto anyother serviceprovider notaspecicserviceprovider.Thereareexceptions,Hsiehetal.was lessconcernedabouttheincumbentbloggingplatformthatthebloggerwas usingandmoreconcernedaboutwhetherFacebook,inparticular,wouldbe consideredanattractivealternative. Theresearchquestionintheinitialproposalintendedtoexaminespecic alternativestoFacebook,likeTwitter,Instagram,Tumblr,andPinterest.It mayberiskytoaskonlyaboutthespecicalternativesgiventhepastresearch; IsuggestthatIasktwosetsofquestionsinthepilot,onesetregarding other socialnetworkingsites,ingeneral,andaspecicsetaboutthealternatives likeTwitter,Tumblr,etc. Thesurveyquestionsbelowcoverboth alternativeattractiveness and generalattitudetowardswitching fromBansaletal.andinahigher-order measureare alternativeperceptions .Bansaletal.usesbothsetsof questionsasdirectfactorstopredictswitchingintention.Thequestionshave beenadaptedtotthespeciccontextofthestudy. 87

PAGE 109

AlternativeAttractivenessBansaletal. Adaptedstudyquestions 1Allinall,competitorswouldbe muchmorefairthanmyhair stylistis Allinall,othersocialnetworking siteswouldbemuchmorefair thanFacebookis 2Overall,competitors'policies wouldbenetmemuchmorethan myhairstylist'spolicies Overall,othersocialnetworking sites'policieswouldbenetme muchmorethanFacebookpolicies 3Iwouldbemuchmoresatised withtheserviceavailablefrom competitorsthantheservice providedbymyhairstylist Iwouldbemuchmoresatised withtheserviceavailablefrom othersocialnetworkingsitesthan theserviceprovidedbyFacebook 4Ingeneral,Iwouldbemuchmore satisedwithcompetitorsthanI amwithmyhairstylist Ingeneral,Iwouldbemuchmore satisedwithothersocial networkingsitesthanIamwith Facebook 5Overall,competitorswouldbe bettertodobusinesswiththan myhairstylist Overall,othersocialnetworking siteswouldbebettertousethan Facebook 88

PAGE 110

AlternativeAttractivenessBansaletal. Adaptedstudyquestionsspecic 1Allinall,competitorswouldbe muchmorefairthanmyhair stylistis Allinall,wouldbe muchmorefairthanFacebookis 2Overall,competitors'policies wouldbenetmemuchmorethan myhairstylist'spolicies Overall,policies wouldbenetmemuchmorethan Facebookpolicies 3Iwouldbemuchmoresatised withtheserviceavailablefrom competitorsthantheservice providedbymyhairstylist Iwouldbemuchmoresatised withtheserviceavailablefrom thantheservice providedbyFacebook 4Ingeneral,Iwouldbemuchmore satisedwithcompetitorsthanI amwithmyhairstylist Ingeneral,Iwouldbemuchmore satisedwiththanIamwith Facebook 5Overall,competitorswouldbe bettertodobusinesswiththan myhairstylist Overall,wouldbe bettertousethanFacebook 89

PAGE 111

AttitudeTowardSwitchingBansaletal.Forme,switchingfrommyhair stylisttoanewhairstylistwithin thenext2monthswouldbe: AdaptedstudyquestionsForme,switchingfromFacebook toanewsocialnetworkingsite withinthenext6monthswould be: 1Abadidea...AgoodideaAbadidea...Agoodidea 2Useless...UsefulUseless...Useful 3Harmful...BenecialHarmful...Benecial 4Foolish...WiseFoolish...Wise 5Unpleasant...PleasantUnpleasant...Pleasant 6Undesirable...DesirableUndesirable...Desirable 90

PAGE 112

6.2.9SocialNetworkingSiteContinuanceIntention Bhattacherjee,p.359denedcontinuanceintentionasa,users'intentiontocontinueusingonlinebankingdivisionOBD,basedonMathieson behavioralintentionscale.Bhattacherjeestatesthat,basedon expectancy-conrmationtheory,continuanceintentionisprimarilydetermined byusersatisfaction. I nformationSystems ContinuanceIntentionBhattacherjee Adaptedstudyquestions 1IintendtocontinueusingOBD ratherthandiscontinueitsuse Iintendtocontinueusing Facebookratherthandiscontinue itsuse 2Myintentionsaretocontinue usingOBDthanuseany alternativemeanstraditional banking Myintentionsaretocontinue usingFacebookthanusean alternativesocialnetworkingsite Twitter,Pinterest,etc. 3IfIcouldIwouldliketo discontinuemyuseofOBD IfIcould,Iwouldliketo discontinuemyuseofFacebook 91

PAGE 113

6.2.10DemographicsAndScreening Thefollowingdemographicquestionswillbeasked:age,gender,andeducation.Screeningquestionswillbeaskedtoensurethatthesurveyrespondentis overtheageof18andhasaFacebookprole.MaddenandSmithnoted signicantgenderdierencesinthewaymenandwomenmanagetheirproles; womenweremorerestrictiveinhowtheymanagedtheirprivacysettings.Age hasbeenshowntobecorrelatedwithunfriendingbehavioraswell;Madden andSmithnotedthatyoungerFacebookusersunfriendedmembersof theirsocialnetworksmoreoftenthanolderusersMaddenandSmith. Thecontrolvariablesarenottheprimarypredictivevariablesinthisresearch butareusedtocontrolforuserdierences. 92

PAGE 114

6.3SamplingAndParticipants Asurveywasadministeredtothreedistinctgroups:apublicannouncement willbemadetogathersurveytakers,studentsattheUniversityofColorado DenverandOracleemployeesfromthreeocesinColoradotocollectdataon socialnetworkingsitecontinuanceintention.Thepublicsurveyrespondents willbeexaminedindepthasthatgroupmaybethemostrepresentativeof thegeneralpopulation.Studentsaretheselectedpopulationastheywere thepopulationwhoinitiallycouldjoinFacebookandhavethelongesttenure withthesocialnetworkingsite. 14 Youthappeartohavewaningenthusiasm forthesiteMaddenetal.,2013andmaybeexperiencinghighvariationin theircontinuanceintention.ThetotalnumberofteensusingFacebookhas notdecreasedintheU.S.howevertheamountoftimespentonthesiteis decreasingMaddenetal.,2013.Oracleoceworkerswillbeusedtoinclude olderusersinthesample.Acomparisonacrossthethreedatasetsisexamined indepthinChapter7.10. 6.4Analyticalmethods TheanalysiswasconductedusingpartialleastsquarestructuralequationmodelingPLS-SEM.Ameasurementmodelandstructuralmodelwasdeveloped andtestedtoevaluatethehypothesesgeneratedinSection5.Thefactorsare evaluatedforreliability,averagevarianceextractedAVE,anddiscriminant validity.Factorsarecheckedforoutliers,normality,homoscedasticity,andare transformedasnecessarypriortoPLS-SEMevaluation.Thefactorsareexaminedtoensuretheymeettheassumptionsoftheanalyticalmethodprior tothedevelopmentofthestructuralmodel.Hairetal.recommends thatthesamplesizeshouldmeetoneoftwoheuristics-thesampleshould 14 https://newsroom.fb.com/Timeline 93

PAGE 115

betentimesthelargestnumberofformativeindicatorsusedtomeasureone constructortentimesthelargestnumberofstructuralpathsdirectedat aparticularlatentconstructinthestructuralmodel.Themeasurementand structuralmodelareevaluatedbasedongoodnessoftmeasures,r-square, FornellandLarckercriterion,etc.followingtheheuristicsinHairetal. ;Wetzelsetal.;FornellandLarcker.Goodnessoft measuresarecomputedasthegeometricmeanofaveragecommunalityand averageR 2 forendogenouslatentconstructsaccordingtoWetzelsetal.. Theloadingsandstatisticalsignicanceforeachofthefactorsareevaluated andshownasnecessary.Thehypothesesaretestedforstatisticalsignicance throughboot-strapping. 6.4.1GoodnessofFitMeasuresandPLS-SEM Hairetal.,p.143statesthat,thereisnoadequateglobalmeasureof goodnessofmodelt[inPLS-SEM].Goodness-of-tmeasuresarerestricted toreectiveoutermodelsandcannotbecomputedforallmodelsinthisresearch,i.e.modelsarereective-formativeinnaturedonothaveacomputable goodness-of-tmeasureHenseleretal.,2009.Wetzelsetal.useda globaltmeasurebasedonTenenhausetal.wheregoodness-of-twas denedasthegeometricmeanoftheaveragecommunalityandaverageR 2 for endogenousconstructs.Henseleretal.statesthatthisapproachhas notbeensystematicallyanalyzedinsimulationstudiesandsuggeststhatit shouldberestrictedtoreectiveoutermodelsonly.Insubsequentsimulation studyresearchbyHenselerandSarstedt,p.577,theauthorsstate: Theunderlyingideawouldbethatthemodelwithahighertis thebetterormorevalidmodel.However,usingsimulateddata, wehaveillustratedthattheGoFandtheGoFrelarenotsuitable 94

PAGE 116

formodelvalidation.Neitheroftheseindicesisabletoseparate validmodelsfrominvalidmodels.Infact,researcherswouldbe mislediftheychoseforthemodelyieldingthehighestGoF.Instead,researchersshouldcarefullyevaluatethepathcoecients andparticularlytheirsignicanceinordertodecideuponwhich pathstoleaveinthemodelandwhichtodiscard. Therefore,goodness-of-tcomparisonswouldnotoeravalidwaytochoose whichmodelsarevalidfromasetcompetingmodelsHenselerandSarstedt, 2013.Researcherscontinuetoinvestigatehowtodeveloparelevantgoodnessof-tmeasureforPLS-SEMmodelsHairetal.,2011;Henseleretal.,2009. 6.4.2Higher-orderconstructs Severalhigher-orderreective-formativeconstructsweregeneratedinthemodel development.Todevelopsecondandthird-orderconstructsseveralmethodicaldecisionsweremadebasedonBeckeretal..Beckeretal. guidewasdevelopedtoaddressthefeaturesthatPLS-SEMtoolsprovide.A reective-formativeconstructsaregeneratedwhentherearelower-orderedconstructsthatarereectivelymeasuredthatdonotshareacommoncauseanda higher-orderconstructmediatestheinuenceonsubsequentendogenousconstructsBeckeretal.,2012.AnexamplefromtheBurnhametal.2003cost modelarethethreecostsprocedural,nancialandrelationalthatformthe higher-orderconstruct costs inthemodel. Costs areformedbythethreereectiveconstructs:procedural,nancialandrelational.Therearethreegeneral approachesforPLS-SEMtomeasurelatenthigher-orderconstructs:1.repeatedindicators,2.sequentiallatentvariablescoreortwo-stageapproach and3.ahybridapproachBeckeretal.,2012.Beckeretal.recommendstherepeatedindicatorapproachforreective-formativemodels.Model 95

PAGE 117

developmenthasoptionstoweightdierentaspectsofthemodel-themodel canbeweightedthefactorsorthepaths.Forreective-formativeconstructs therecommendationistoweightthepathsBeckeretal.,2012.Themodels developedinSection7followtherecommendations-theyaredevelopedusingtherepeatedindicatorapproachandusethepath-weightedmode.Becker etal.statesthattherecommendedapproachforreective-formative modelsisless-biasedandthereforeproducesmorepreciseparameterestimates andmorereliablehigher-orderconstructscores. Researcherscanusecovariance-basedstructuralequationmodelsCB-SEM orpartialleastsquaredstructuralequationmodelsPLS-SEMHairetal., 2011.Hairetal.statesseveralrulesofthumbinhowtochooseCBSEMorPLS-SEMforaresearchproject.PLS-SEMisdesignedtomaximize theexplainedvarianceofthedependentlatentconstructsanddiersfrom CB-SEMwherethegoalistoreproducethetheoreticalcovariancematrix. PLS-SEMisbetteratpredictingthekeytargetconstructs;however,CB-SEM isbetterfortheorytesting,theoryconrmationorcomparisonsofalternative theories.PLS-SEMhandlesformativeconstructsbetterthanCB-SEM.PLSSEMcanworkwithmorecomplexmodelswheremanyconstructsandmany indicatorsarepresentbetterthanCB-SEM.PLS-SEMworkswellwithrelativelylowsamplesizes,andwhenlargersamplesizesareusedtheresultsof CB-SEMandPLS-SEMaresimilar. Themodelthatisemployedinthisresearchusesseveralformativeconstructsinparticularreective-formativeconstructs,hasamediumlevelof complexityandseveralmediatingconstructsthatfavorPLS-SEMoverCBSEM.CB-SEMwouldprovidestrongerabilitytocomparealternativetheories thatareusedinthisresearch;however,giventheothersetofconstraintsPLSSEMwasdeemedtoprovideabettertwiththeresearchobjectives.The 96

PAGE 118

softwarethatwillbeusedisSmartPLS3.0Ringleetal.,2005. 6.4.3Moderatingeects TotestformoderatingeectsrecommendationsfromChinetal.were followed.TheinteractioneectsweretestedthroughhierarchicalPLS-SEM modelswherecompared.StandardizedindicatorswereusedfortheinteractioneectsfollowingtherecommendationsofChinetal..Interaction eectscanbeleftunstandardized,standardizedormean-centered.Chinetal. statesthatatleastoneoftwomethodsshouldbeused,standardized ormean-centered;therecommendationforLikert-scaleitemswhoseitemsare theoreticallyparallelasisthecaseinthisresearchistostandardizetheitems. Cohen'seectsize f 2 willbecalculatedtocomparetheaddedbenetofthe interactiontermwhere0.02,0.15and0.35areconsideredtohaveasmall, moderateorlargeeect,respectivelyChinetal.,2003.Theeectsizeofa moderatingtermmaybesmallbutmeaningfulifthepathcoecientschange inameaningfulmannerChinetal.,2003. Hairetal.statesthatideallymoderatorsandtheotherconstructsin theanalysisareuncorrelatedtohelpdistinguishthemoderatorseectsinthis casetheotherconstructsare satisfactionandperceivedusefulness and continuanceintention .Increasingstrengthofthecorrelationsmakesitdicultto haveavalidinterpretationoftheresultsHairetal.,2006.WhismanandMcClellandstatethatcollinearityisnotaproblemaslongasanapproach tostandardizingorcenteringtheinteractiontermistaken;interactioneects areoftendierentbecauseitisestimatinga very dierenteect.Moderator modelsestimatesimpleeectsofonevariablewhentheothervariableisxed at0.WhismanandMcClellandstatethattheinteractiontermisoften highlycorrelatedwithitscomponentse.g.theinteractionterm satisfaction 97

PAGE 119

&perceivedusefulnessxhabit islikelytobecorrelatedwithboth habit and satisfaction&perceivedusefulness .Changingtheoriginofthescalesusing eithermethodstandardizingorcenteringwillnotaectthetestoftheinteractionandwillreducethecorrelationsbetweentheinteractiontermandits componentstozeroWhismanandMcClelland,2005. 6.5Design Theresearchusesindependentconstructsandcovariatestopredictonedependentvariable-SNSContinuance.Theindependentvariablesarealternative attractiveness,attitudetoswitch,brandrelationship,conrmation,habit,interpersonalinuence,perceivedusefulness,personalinnovativeness,personal relationshiploss,proceduraleconomiccost,proceduralevaluationcost,procedurallearningcost,proceduralsetupcostandsatisfaction.Threecovariates areusedtoadjustandmeasuredemographicvariables;theseareage,gender andeducation.Second-andthird-orderconstructsweregeneratedtogroup reective-formativeconstructsintomeaningfulgroups.AlternativeattractivenessandattitudetoswitchfromBansalandTaylorweregroupas alternativeperceptions.Relationshipcostsincludesbrandrelationshipand personalrelationshipsasdenedbyBurnhametal..Thefourproceduralcostsproceduraleconomiccost,proceduralevaluationcost,procedural learningcost,proceduralsetupcostwerecombinedintooneproceduralcost constructBurnhametal.,2003.ThemainfactorsofISContinuanceBhattacherjee,2001,Satisfactionandperceivedusefulnesswerecombinedintoa higher-orderconstructtoallowforsimplerinterpretationoftheresults.A third-orderfactorof cost wasdevelopedtoincludetherelationshipcostsand proceduralcostsintooneconstruct. Specicalternativesocialnetworkingsiteswereexaminedbasedonques98

PAGE 120

tionsaboutsocialnetworkingsitesTwitter,TUMBLR,InstagramandPinterest.Thequestionsweredevelopedbyaskingsurveyrespondentswhich socialnetworkingsitestheyusedandtoaskmorespecicquestionsabout theiropinionsofthosesitesbasedonBansalandTaylor.Thespecicalternativesocialnetworkingsitesareexaminedindepthtodetermine howattitudesregardingthosespecicsitescanpredictwhetherapersonwill continueordiscontinueuseofFacebook. 99

PAGE 121

7Results Theresultswillcompareaseriesofcompetingmodelsthatpredictsocialnetworkingsitecontinuancethroughindependentconstructs.Thebasemodelis theIScontinuancemodelofBhattacherjee.Individualfactorswillbe addedtothebasemodeltodeterminetheireectsseparatelytodeterminethe independentimpactofafactorandanymoderatingeectsonmodelsthathave moderators;theindividualfactormodelsareintheAppendix.Threemodels arepresentedindetail,theISContinuancemodelofBhattacherjeeSection7.5,thenon-moderatedmodelwithadditionalfactors-Section7.6, andthecompletemoderatedmodel-Section7.7.Abackwardsregressionwas developedtoincludeonlythestatisticallysignicantfactors.ModelstodeterminehowattitudesaboutspecicalternativeproductsTwitter,Tumblr, Instagram,andPinterestpredictcontinuanceonFacebookareintheAppendixF.Asummaryofthendingsshowsthecoecientofdetermination R 2 andeectsize f 2 foreachofthemodelstoshowtheeectsofthefactors inpredictingcontinuance. 7.1CommonMethodVariance Thesurveyinstrumentwasdevelopedusingasinglemethodresearchdesign andcommonmethodvarianceshouldbemeasuredtodeterminehowmuchof thevarianceinthesurveyisduetothesinglesurveymethodPodsakoetal., 2003.Totestwhethercommonmethodbiaswasinthesurveyinstrument aHartmannsingle-factortestpost-hocstatisticaltestsweredeveloped.62 variableswereenteredintoanexploratoryfactoranalysis,usingtherotated principalcomponentsfactoranalysiswiththeVariamaxrotationfunctionto determinethenumberoffactorsthatarenecessarytoaccountforthevariance 100

PAGE 122

inthevariables.Theresultsshowedthat13factorswerepresentandthemost covarianceexplainedbyonefactoris21.46%ofthevariancewhichisbelow the.50recommendationandisconsideredacceptable. Asecondmethodtodeterminecommonmethodvariancewasemployed usingthe commonmethodfactor Podsakoetal.,2003;Liangetal.,2007.A rst-orderlatentvariablewascreatedforeveryindicatorinthemodel.Each rst-orderlatentvariablewasthenconnectedtoanewsecond-orderlatent variablefromthetheoreticalmodelusingtherepeatedindicatorapproach,i.e. satisfactioninthemodelhasthreeindicatorsthereforethreelatentvariables weregeneratedsat1,sat2,sat3andonesatisfactionexogenousvariablewas generatedusingtherepeatedindicatorapproachandconnectedtosat1,sat2, sat3.Anew method latentvariablewasthencreatedalsobytherepeated indicatorapproachthe method latentvariablecontainsallvariablesinthe model.The method latentvariablewasthenconnectedtoallrst-orderlatent variablesinthemodel.Thepathcoecientsbetweeneverylatentvariableand itssecond-orderconstruct substantive factorloadingwasthencomputedas wellasthepathcoecientbetweenthemethodlatentvariableandevery latentvariable.ForanextendeddiscussionontheapproachseeLiangetal. Liangetal.,AppendixE-UsingPLStoAssessCommonMethod Bias. Theaverageofall substantive loadings,and substantive explainedvariance loadingsquaredand method factorloadingsand method explainedvariance loadingsquaredwerethencalculated-seeTable43.Theresultsshowthat theaveragesubstantivelyexplainedvarianceoftheindicatorsis.6419,while theaveragemethod-basedvarianceis0.0101.Theratioofthesubstantive variancetomethodvarianceisapproximately63:1.Itisunlikelythatthesurveymethodisbiasedforthisstudybasedontheratioofsubstantivevariance 101

PAGE 123

tomethod-variance. 102

PAGE 124

Table9:Gender GenderFrequencyPercentCumulativepercent Male51339.439.4 Female78560.399.7 NA4.3100.0 Total1302100.0100.0 7.2Demographics Demographicinformationregardingage,genderandeducationwascollectedin thesurvey.TheTables9,10and11showthedistributionofeachdemographic variable.Themajorityofthesurveytakerswerewomen.7%.Amajority .7%ofthesurveytakerswere39yearsoldoryoungerandthemajorityof surveytakerscompletedabachelorsdegreeorhadmoreeducation. 7.3ContinuanceDependentVariable ThecontinuancefactorwasmeasuredonaLikert-scalesevenitemscalewith threequestions.Continuancerangedfrom1to7andhadameanof4.6922with astd.deviationof1.52.Figure6showsthatthedistributionofcontinuance isnotnormal,valuesattheextremerangeofdonotplantocontinueandplan 103

PAGE 125

Table10:Age AgeFrequencyPercentCumulativepercent 18-19141.1%1.1% 20-2413710.5%11.6% 25-2921116.2%27.8% 30-3419214.7%42.5% 35-3913510.4%52.9% 40-4417013.1%66.0% 45-4916512.7%78.6% 50-541259.6%88.2% 55-59866.6%94.9% 60-64372.8%97.7% 65-69201.5%99.2% 70-7450.4%99.6% 75+10.1%99.7% NA40.3%100.0% Total1302100.0100.0 104

PAGE 126

Table11:Education EducationFrequencyPercentCumulative Percent Lessthan highschool 10.1%0.1% HighSchool 292.2%2.3% Somecollege 14611.2%13.5% Associates 886.8%20.3% Bachelor 52940.6%60.9% Masters 42932.9%93.9% Doctorate 765.8%99.7% NA 40.3%100.0% Total1302100.0% 100.0% 105

PAGE 127

Table12:ContinuanceDescriptives MeasureStatisticStd.Error Mean4.69220.0423 95%CondenceIntervalforMeanLowerBound4.6092 95%CondenceIntervalforMeanUpperBound4.7751 5%TrimmedMean4.7669 Median4.9754 Variance2.3270 Std.Deviation1.5254 Minimum1.0000 Maximum7.0000 Range6.0000 InterquartileRange1.9578 Skewness-0.72710.0678 Kurtosis-0.04760.1355 tocontinueareover-representedinthedistribution.Themajorityofusersare onthepositivesideofcontinuanceindicatingthattheuserintendstocontinue useofFacebook.Userswereaskedasinglequestionsastowhethertheywere planningtoleavethenetwork,continueonthenetworkorhavealreadyleftin thenearterm.Approximately77%saidtheyintendedtocontinueuseofthe site-seeTable13. 7.4OutlierAnalysis Outlieranalysiswasperformedtoreducetheeectsinuentialcasesthathave adisproportionateinuenceintheregressionanalysisHairetal.,2006.Outlieranalysiswasperformedcomparingthepredictedvalue ^ y andactualvalue y toremoveoutliersatthe95%condencelevel a =.05.Hairetal. indicatesthatthisisthemostwidelyusedoutlierremovaltechniquetoaddressinuentialobservationsinregressionanalysis.Therearemanyoptions 106

PAGE 128

Figure6:ContinuanceHistogram 107

PAGE 129

Table13:Single-ItemContinuance FrequencyPercentCumulative Percent WillsoonstopusingFacebook 20015.4%15.4% WillContinueusingFacebook 99876.7%92.0% I'vealreadyleft 1037.9%99.9% NA 10.1%100.0% Total1302100.0% 0.1% 108

PAGE 130

toreducetheeectsofinuentialobservationsinaregressionanalysisincluding:correctingdata,removingcasesthatareconsideredvalidbutexceptional, decidingwhethertoremoveobservationsthathavenolikelyexplanation,and re-assessingtheconceptualmodelwhenobservationswhoseindividualcharacteristicsareunexceptionalbutwhencombinedwithotherobservationsare exceptionalHairetal.,2006.Theoptionchoseninthisanalysisistoremove potentiallyvalidbutexceptionalcasestoreducetheoutsizedeectinuential caseshaveontheanalysis.OutlieranalysiswasperformedinSPSSthrough automaticlinearmodelingregressionanalysis.Residualswereremovedwhen thepredictedvalue ^ y andactualvalue y wereoutsidethe95%condence level.Theinitialnumberofobservationswas1370andafteroutlierremoval 1302remainedintheanalysisor95.0%ofthesurveyrespondents-seeFigures 7and8fordistributiondetailsoftheinitialandretaineddataset,respectively. 109

PAGE 131

Figure7:InitialResidualAnalysis-Continuance 110

PAGE 132

Figure8:ResidualAnalysisAfterOutliersRemoved-Continuance 111

PAGE 133

Figure9:BaseModel-ISContinuance 7.5BaseModel-ISContinuanceModel 7.5.1ModelDescription ThebasemodelincludesthemeasuresfromtheIScontinuancemodelofBhattacherjeeasabasisofcomparisonagainsttheothermodelsinthis research.Thebasemodelusestheindependentconstructsconrmation,perceivedusefulness,satisfactiontopredictsocialnetworkingsitecontinuance intention.Perceivedusefulnessandsatisfactionaretheorizedtohavedirect eectsandconrmationistheorizedtobefullymediatedbybyperceived usefulnessandsatisfaction. 7.5.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliability valuesareallgreaterthanorequalto.7providingevidenceoftheconstruct's internalconsistencyandreliabilityHairetal.,2006.Compositereliability valuesareallgreaterthantheAVEscoresindicatingconvergentvalidityHair etal.,2006.ThesquarerootsoftheAVEareallgreaterthanthelatentvari112

PAGE 134

aConrmationHistogram bPerceivedUsefulnessHistogram cSatisfactionHistogram Figure10:BaseModelHistograms 113

PAGE 135

Table14:BaseModel-ISContinuanceMeasurementModel ConstructAVECompositeReliability CONF 0.69100.8968 CONTINUANCE 0.66770.8831 PU 0.68040.8938 SAT 0.93090.9758 Table15:BaseModel-ISContinuanceMeasurementModel-Discriminant Validity FornellandLarckerCriterion Conrmation Perceived Usefulness Satisfaction Continuance Conrmation 0.8311 Perceived Usefulness 0.7108 0.8472 Satisfaction 0.5994 0.6299 0.8249 Continuance 0.7887 0.7890 0.5990 0.9648 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagnol elementsarethelatentvariablecorrelations. ablecorrelationsindicatingdiscriminantvalidityFornellandLarcker criterion-See:15.Theindicatorsinthereectivemeasurementmodelsreach satisfactoryindicatorreliabilitylevels.Themeasurementmodelassessment substantiatesthatalltheconstructmeasuresarereliableandvalid. 7.5.3StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactiontopredictsocialnetworkingsitecontinuanceintention.Thepredictorsconrmation,perceived usefulnessandsatisfactionexplainapproximately66.8%ofthevarianceR 2 incontinuanceintentionandisconsideredtohaveamoderatelevelofexplana114

PAGE 136

Table16:BaseModel-ISContinuanceStructuralModel GoodnessofFit .6490 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 PerceivedUsefulness 0.35920.2435 Satisfaction 0.64690.6021 Continuance 0.6680.4751 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. Figure11:BaseModel-ISContinuance tion 15 ;themodelalsoexhibitspredictiverelevanceQ 2 whereitsvalueis.4751 scoresabovezeroindicatepredictiverelevanceinPLSpathmodels.Satisfactionhadthestrongestpredictiveabilityforsatisfactionwithastandardized pathcoecient of.631 t =31.1585followedbyperceivedusefulnesswith astandardizedpathcoecient of.1971 t =10.6813.Thecoecientsare bothpositiveandindicatethathigherlevelssatisfactionandhigherlevelsof perceivedusefulnessareassociatedwithhigherlevelsofcontinuanceintention. 15 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 115

PAGE 137

Table17:PathCoecients RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients Satisfaction Continuance0.631031.1585 PerceivedUsefulness Continuance0.246310.6813 OtherPathCoecients PerceivedUsefulness Satisfaction0.19718.7811 Conrmation Satisfaction0.670534.0352 Conrmation PerceivedUsefulness0.599432.7735 TotalEects Sample Mean M T Statistics ContinuancePathCoecients CONF CONTINUANCE0.645449.0332 SAT CONTINUANCE0.631531.1585 PU CONTINUANCE0.370815.4962 age CONTINUANCE0.07194.6425 gender CONTINUANCE0.04112.6427 ed CONTINUANCE0.01440.8395 OtherPathCoecients CONF SAT0.788874.4869 PU SAT0.19848.7811 CONF PU0.599532.7735 116

PAGE 138

Figure12:BaseModelTotalEectsPathCoecientsonISContinuance 117

PAGE 139

Figure13:CompleteNon-ModeratedModel 7.6Completemodel-NonModerated 7.6.1ModelDescription ThemodelincludesthemeasuresfromtheIScontinuancemodelBhattacherjee,2001andfactors:consumerswitchingcosts,habit,personalinnovativeness,interpersonalinuence,andalternativeperception.Thetwopredictors forSNSContinuance,satisfactionandperceivedusefulnesshavebeencombinedintoasinglereective-formativeconstructthatallowsforamoredirect comparisonofthecompetingmodelsandallowsthemodeltobecompared atahigherlevelHairetal.,2006.TheBurnhametal.costmodel wasinitiallydevelopedasaformative-reectivemodelanddoesnotneeda transformation.Thecostmodelusestheindependentconstructseconomic risk,evaluationcosts,learningcostsandsetupcostsproceduralcostsand personalrelationshiplossandbrandrelationshiplossrelationalcoststopredictcontinuanceintention.TheIScontinuancemodelusestheindependent 118

PAGE 140

Onlyhigher-orderfactorsareshown Figure14:CompleteNon-ModeratedSimpliedModel constructsconrmation,perceivedusefulness,satisfactiontopredictinformationsystemscontinuanceintention.Outerloadinganalysisforeachitemand therelatedconstructcompositereliabilityandCronbach'salphacanbefound inAppendixB.FormoderatingeectsofthecompletemodelseeSection7.7. Thetheoreticalmodelhasindependentfactorsforsatisfactionandperceivedusefulnessaspredictorsforcontinuanceintentionanddoesnotcombine thesefactorsintoasingleconstruct.Theanalysishere,asstatedabovecombinesthetwofactors,andthemeasurementmodelandstructuralmodelwas conductedonthismodel.Anadditionalmodelwasgeneratedwithtoshowthe eectsofsatisfactionandperceivedusefulnessoncontinuanceintentionwith themoderatedfactorsasshowninFigure13theresultsofwhichareshownin Figure33.Thetablesandguresinthissectionusetheanalysisofthecombinedfactorsforperceivedusefulnessandsatisfactionexceptasnotedinthe 119

PAGE 141

CombinedsecondorderconstructofSatisfactionandPerceivedUsefulness Figure15: SecondOrderConstruct: SatisfactionandPerceivedUsefulness Histogram twoFigures13and33.Asdesigned,withrepeatedindicatorsforhigher-order factorsandpath-weighting,thedierencesbetweenthetwomodelswithPLSSEMissmall.ThecoecientofdeterminationR 2 forthesimpliedmodelis 0.7667andforthecompletenon-moderatedmodelis0.7693,orthechange r inthecoecientofdeterminationofthetwomodelsis0.0026.Theeectsize f 2 betweenthetwomodelsis0.0113andisconsideredlessthansmall.02is consideredsmallbasedon Chinetal. .Theoretically,theindependent factormodelisstronger,butthereistensionbetweenthestrongertheoretical modelandthemoreparsimoniousmodelthatallowsfactorstobeevaluated atasimilarlevelHairetal.,2006.Thesimpliedmodelallowsfactorstobe analyzedatamoreappropriatelevelwithveadditionalfactorsbeyondthe factorsinIScontinuancemodelofBhattacherjeethatareconsideredin thenon-moderatedmodel.Thehigher-ordermodelallowsthosefactorstobe analyzedatalevelthatismoreappropriateforthisresearch. 120

PAGE 142

Figure16:BrandRelationshipHistogram Figure17:PersonalRelationshipHistogram Figure18:ProceduralEconomicCostHistogram 121

PAGE 143

Figure19:ProceduralEvaluationCostHistogram Figure20:ProceduralLearningCostHistogram Figure21:ProceduralSetupCostHistogram 122

PAGE 144

Figure22: SecondOrderConstruct: RelationshipCostHistogram CombinedsecondorderconstructofBrandRelationshipandPersonal RelationshipCosts CombinedsecondorderconstructofProceduralEconomicCost,Procedural EvaluationCost,ProceduralLearningCostandProceduralSetupCost Figure23: SecondOrderConstruct: ProceduralCostHistogram CombinedThirdorderconstructofRelationshipCostandProceduralCost Figure24: ThirdOrderConstruct: CostHistogram 123

PAGE 145

Figure25:HabitHistogram Figure26:PersonalInnovativenessHistogram Figure27:InterpersonalInuenceHistogram 124

PAGE 146

Figure28:AlternativeAttractivenessHistogram Figure29:AttitudeToSwitchHistogram Combinedsecond-orderconstructofAlternativeAttractivenessandAttitude toSwitch Figure30: SecondOrderConstruct: AlternativePerceptionsHistogram 125

PAGE 147

7.6.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliability valuesareallgreaterthanorequalto.7providingevidenceoftheconstruct's internalconsistencyreliabilityHairetal.,2006.CompositereliabilityvaluesareallgreaterthantheAVEscoresindicatingconvergentvalidityHair etal.,2006.ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelationsindicatingdiscriminantvalidityFornellandLarcker criterion-See:19.Theindicatorsinthereectivemeasurementmodelsreach satisfactoryindicatorreliabilitylevels.Themeasurementmodelassessment substantiatesthatalltheconstructmeasuresarereliableandvalid. 126

PAGE 148

Table18:SNSCostMeasurementModel ConstructAVECompositeReliability PROC_SETUP 0.59380.8530 PersRelationLoss 0.64160.8990 AlterAttract 0.64750.9012 AttitudeToSwitch 0.73800.9440 BrandRelationship 0.68090.8638 CONF 0.69100.8969 CONTINUANCE 0.71660.8831 HABIT 0.76710.9079 InterpersInu 0.60870.8599 PERSINNOV 0.70380.9219 PROCCOST 0.56460.8382 PROCEVAL 0.60130.8181 PROCLEARN 0.60410.8590 PU 0.68050.8939 SAT 0.93090.9758 127

PAGE 149

Table19:CompleteModelwithModerators-DiscriminantValidity FornellandLarckerCriterion PROC SETUP Proc RelationLoss Alter Attract Attitude To Switch Brand Relationship CONFCONTINUANCE HABITInterpers Inu PERS INNOV PROC COST PROC EVAL PROC LEARN PUSAT PROCSETUP 0.7706 ProcRelationLoss 0.3388 0.8010 AlterAttract -0.2262 -0.4442 0.8047 AttitudeTo Switch -0.4500 -0.4143 0.5752 0.8591 Brand Relationship 0.1169 0.4936 -0.4867 -0.3264 0.8252 CONF 0.2251 0.6732 -0.4942 -0.4412 0.6519 0.8313 CONTINUANCE 0.2888 0.7318 -0.6122 -0.5544 0.6098 0.7116 0.8465 HABIT 0.2623 0.7443 -0.4132 -0.3846 0.4975 0.6372 0.6952 0.8758 InterpersInu 0.1981 0.1940 -0.0219 -0.1205 0.1650 0.2108 0.1058 0.1440 0.7802 PERSINNOV -0.4141 -0.0284 0.2811 0.4140 0.0501 -0.0396 -0.1633 -0.0082 -0.0555 0.8389 PROCCOST 0.7094 0.4668 -0.2153 -0.4307 0.2221 0.3395 0.3733 0.3685 0.2338 -0.3249 0.7514 PROCEVAL 0.5409 0.1478 -0.1995 -0.3975 0.0776 0.1424 0.2034 0.0921 0.1297 -0.6325 0.4839 0.7754 PROCLEARN 0.6849 0.3127 -0.2452 -0.3530 0.1776 0.2003 0.2723 0.2056 0.1820 -0.3679 0.6428 0.5005 0.7772 PU 0.2808 0.7555 -0.3716 -0.3510 0.4734 0.5991 0.6306 0.6550 0.2569 -0.0284 0.3850 0.1103 0.2428 0.8249 SAT 0.1785 0.6674 -0.5658 -0.4567 0.6370 0.7879 0.7861 0.6219 0.1096 -0.0370 0.2519 0.1115 0.2015 0.5982 0.9648 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagonalelementsarethelatentvariablecorrelations. 128

PAGE 150

7.6.3StructuralModel ThestructuralmodelwasassessedtodeterminehowtheindependentconstructsfromISContinuancetheoryandconsumerswitchingcostspredictinformationsystemscontinuanceintention..Thepredictorsexplainedapproximately76.7%ofthevarianceR 2 incontinuanceintentionandisconsidered tohaveasubstantiallevelofexplanation 16 ;themodelalsoexhibitspredictive relevanceQ 2 whereitsvalueis.5403scoresabovezeroindicatepredictive relevanceinPLSpathmodels.Satisfactionandperceivedusefulness,ina combinedmeasure,predictcontinuanceintentionwithastandardizedpath coecient of.368 t =13.5952.Allofthecostspredictcontinuance intentionwithastandardizedpathcoecient of0.250 t =7.9023.Alternativeperceptionhadthethirdlargestpathcoecient =-0.207, t =10.3132 followedbyhabit =.164, t =6.7738,personalinnovativeness =-.059, t =4.0897andinterpersonalinuence =-.045, t =3.2342.Coecients thatarepositiveindicatethatareassociatedwithhigherlevelsofcontinuance intention. RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SatAndPU CONTINUANCE0.368613.5952 COSTS CONTINUANCE0.24967.9023 AlternativePerception CONTINUANCE-0.206910.3132 16 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 129

PAGE 151

RelationPath Coecient Tstatistic HABIT CONTINUANCE0.16426.7738 PERSINNOV CONTINUANCE-0.05894.0897 InterpersInu CONTINUANCE-0.04513.2536 OtherPathCoecients Relationship_costs COSTS0.979186.4142 PersRelationLoss Relationship_costs0.735335.1657 CONF SAT0.670633.1904 SAT SatAndPU0.785132.4689 CONF PU0.599331.6692 AlterAttract AlternativePerception0.662117.6155 BrandRelationship Relationship_costs0.405315.9571 PROC_COST ProceduralCosts0.937313.3217 AttitudeToSwitch AlternativePerception0.455210.9372 PU SatAndPU0.307410.4468 PU SAT0.19718.883 ProceduralCosts COSTS0.0381.7011 PROC_LEARN ProceduralCosts0.14681.6372 PROC_EVAL ProceduralCosts-0.11591.6114 PROC_SETUP ProceduralCosts0.01750.1869 130

PAGE 152

RelationPath Coecient Tstatistic TotalEects ContinuancePathCoecents SatAndPU CONTINUANCE0.36513.5952 CONF CONTINUANCE0.293213.2753 SAT CONTINUANCE0.286312.0838 PU CONTINUANCE0.168810.8282 AlternativePerception CONTINUANCE-0.207210.3132 AlterAttract CONTINUANCE-0.1379.3630 COSTS CONTINUANCE0.25687.9023 Relationship_costs CONTINUANCE0.25087.8161 PersRelationLoss CONTINUANCE0.18467.6372 AttitudeToSwitch CONTINUANCE-0.09437.1346 BrandRelationship CONTINUANCE0.10137.0277 HABIT CONTINUANCE0.16426.7738 age CONTINUANCE0.06134.3408 PERSINNOV CONTINUANCE-0.0594.0897 InterpersInu CONTINUANCE-0.04343.2342 PROC_COST CONTINUANCE0.00961.6601 ProceduralCosts CONTINUANCE0.01041.6396 131

PAGE 153

RelationPath Coecient Tstatistic education CONTINUANCE0.02101.4686 gender CONTINUANCE0.01671.2233 PROC_EVAL CONTINUANCE-0.00111.1223 PROC_LEARN CONTINUANCE0.00151.0321 PROC_SETUP CONTINUANCE0.00020.1520 OtherPathCoecients Relationship_costs COSTS0.976786.4142 CONF SatAndPU0.803275.3986 CONF SAT0.788869.2248 PersRelationLoss COSTS0.718735.7847 PersRelationLoss Relationship_costs0.735935.1657 SAT SatAndPU0.784232.4689 CONF PU0.599831.6692 AlterAttract AlternativePerception0.661617.6155 BrandRelationship Relationship_costs0.403915.9571 PU SatAndPU0.462715.8195 BrandRelationship COSTS0.394615.1307 PROC_COST ProceduralCosts0.933113.3217 AttitudeToSwitch AlternativePerception0.454410.9372 132

PAGE 154

RelationPath Coecient Tstatistic PU SAT0.19778.8830 PROC_COST COSTS0.03771.7277 ProceduralCosts COSTS0.04071.7011 PROC_LEARN ProceduralCosts0.14491.6372 PROC_EVAL ProceduralCosts-0.11331.6114 PROC_EVAL COSTS-0.00421.1616 PROC_LEARN COSTS0.00581.0693 PROC_SETUP ProceduralCosts0.01570.1869 PROC_SETUP COSTS0.00090.1551 133

PAGE 155

Table20:Indices GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 AlternativePerception 0.99240.5352 CONTINUANCE 0.76670.5403 COSTS 0.99420.2178 PU 0.35890.2433 ProceduralCosts 0.99330.3252 Relationship_costs 0.99880.5006 SAT 0.64570.6010 SatAndPU 0.99990.6083 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. Figure31:CompleteNon-ModeratedModelPathCoecientsforSNSContinuance 134

PAGE 156

Figure32:CompleteNon-ModeratedSimpliedModel Figure33:CompleteNon-ModeratedModel 135

PAGE 157

Figure34:CompleteModeratedModel 7.7Completemodel-ModeratingFactorAnalysis 7.7.1ModelDescription ThemodelincludesthemeasuresfromtheIScontinuancemodelBhattacherjee,2001andfactors:consumerswitchingcosts,habit,personalinnovativeness,interpersonalinuence,andalternativeperceptionandtwomoderating factors,satisfactionandperceivedusefulnessxhabitandsatisfactionandperceivedusefulnessxpersonalinnovativeness.ThetwopredictorsforIScontinuance,satisfactionandperceivedusefulnesshavebeencombinedintoasingle reective-formativeconstructthatallowsforamoredirectcomparisonofthe competingmodels;thiscombinedfactorwasusedforthemoderatingfactoranalysis.TheBurnhametal.consumerswitchingcostmodelwas initiallydevelopedasaformative-reectivemodelanddoesnotneedatransformation.Thecostmodelusestheindependentconstructseconomicrisk, 136

PAGE 158

Figure35:CompleteModeratedSimpliedModel evaluationcosts,learningcostsandsetupcostsproceduralcostsandpersonalrelationshiplossandbrandrelationshiplossrelationalcoststopredict socialnetworkingsitecontinuanceintention.TheISContinuancemodeluses theindependentconstructsconrmation,perceivedusefulness,satisfactionto predictsocialnetworkingsitecontinuanceintention. Thetheoreticalmodelhasindependentfactorsforsatisfactionandperceivedusefulnessaspredictorsforcontinuanceintentionanddoesnotcombine thesefactorsintoasingleconstruct.Theanalysishere,asstatedabovecombinesthetwofactors,andthemeasurementmodelandstructuralmodelwas conductedonthismodel.Anadditionalmodelwasgeneratedwithtoshowthe eectsofsatisfactionandperceivedusefulnessoncontinuanceintentionwith themoderatedfactorsasshowninFigure34theresultsofwhichareshown inFigure37.Thetablesandguresinthissectionusetheanalysisofthe 137

PAGE 159

combinedfactorsforperceivedusefulnessandsatisfactionexceptasnotedin thetwoFigures34and37.ThedierencesbetweenthecoecientsofdeterminationR 2 forthesimpliedandcompletedmodelaresmall,theR 2 forthe simpliedmodelis.7687andthetheoreticalmodelis.7707orthechange r inthecoecientofdeterminationofthetwomodelsis0.0020.Theeectsize f 2 betweenthetwomodelsis0.0087andisconsideredlessthansmall.02 isconsideredsmallbasedon Chinetal. .The coecientsforthe simpliedandcompletemodelarenotsignicantforthemoderatorsineither model. 7.7.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliability valuesareallgreaterthanorequalto.70providingevidenceoftheconstruct's internalconsistencyreliabilityHairetal.,2006.CompositereliabilityvaluesareallgreaterthantheAVEscoresindicatingconvergentvalidityHair etal.,2006.ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelationsindicatingdiscriminantvalidityFornellandLarcker criterion-See:23.Theindicatorsinthereectivemeasurementmodelsreach satisfactoryindicatorreliabilitylevels.Themeasurementmodelassessment substantiatesthatalltheconstructmeasuresarereliableandvalid.Theresultingtablesshowthecompletemodelincludingthemoderatingeectsalthough themoderatingeectsarelessthansmall. 138

PAGE 160

Table22:SNSCostMeasurementModel ConstructAVECompositeReliability PROC_SETUP 0.59380.8530 PersRelationLoss 0.64160.8990 AlterAttract 0.64750.9012 AttitudeToSwitch 0.73800.9440 BrandRelationship 0.68090.8638 CONF 0.69100.8969 CONTINUANCE 0.71650.8831 HABIT 0.76710.9079 InterpersInu 0.60890.8600 PERSINNOV 0.70380.9219 PROC_COST 0.56460.8382 PROC_EVAL 0.60130.8181 PROC_LEARN 0.60410.8590 PU 0.68050.8939 SAT 0.93090.9758 SatAndPU*HABIT 0.41990.9366 SatAndPU*PERSINNOV 0.42650.9537 139

PAGE 161

Table23:BaseModel-SNSContinuanceMeasurementModel-DiscriminantValidity FornellandLarckerCriterion PROC SETUP Proc RelationLoss Alter Attract Attitude To Switch Brand Relationship CONFCONTINUANCE HABITInterpers Inu PERS INNOV PROC COST PROC EVAL PROC LEARN PUSATSat And PU* Habit Sat And PU* Pers Innov PROCSETUP 0.77 PersRelationLoss 0.34 0.80 AlterAttract -0.23 -0.44 0.80 AttitudeTo Switch -0.45 -0.41 0.58 0.86 Brand Relationship 0.12 0.49 -0.49 -0.33 0.83 CONF 0.23 0.67 -0.49 -0.44 0.65 0.83 CONTINUANCE 0.29 0.73 -0.61 -0.55 0.61 0.71 0.85 HABIT 0.26 0.74 -0.41 -0.38 0.50 0.64 0.70 0.88 InterpersInu 0.20 0.19 -0.02 -0.12 0.17 0.21 0.11 0.14 0.78 PERSINNOV -0.41 -0.03 0.28 0.41 0.05 -0.04 -0.16 -0.01 -0.06 0.84 PROCCOST 0.71 0.47 -0.22 -0.43 0.22 0.34 0.37 0.37 0.23 -0.32 0.75 PROCEVAL 0.54 0.15 -0.20 -0.40 0.08 0.14 0.20 0.09 0.13 -0.63 0.48 0.78 PROCLEARN 0.68 0.31 -0.25 -0.35 0.18 0.20 0.27 0.21 0.18 -0.37 0.64 0.50 0.78 PU 0.28 0.76 -0.37 -0.35 0.47 0.60 0.63 0.66 0.26 -0.03 0.39 0.11 0.24 0.82 SAT 0.18 0.67 -0.57 -0.46 0.64 0.79 0.79 0.62 0.11 -0.04 0.25 0.11 0.20 0.60 0.96 SATAndPU* HABIT -0.11 -0.43 0.09 0.10 -0.21 -0.28 -0.35 -0.43 -0.14 -0.01 -0.20 -0.04 -0.11 -0.44 -0.28 0.65 SATAndPU* PERSINNOV 0.05 -0.11 -0.09 -0.07 -0.04 -0.06 -0.07 -0.09 0.05 -0.05 -0.01 0.05 0.03 -0.03 -0.03 0.36 0.65 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagonalelementsarethelatentvariablecorrelations. 140

PAGE 162

7.7.3StructuralModel ThestructuralmodelwasassessedtodeterminehowtheindependentconstructsfromISContinuancetheoryandconsumerswitchingcostspredictinformationsystemscontinuanceintention..Thepredictorsexplainedapproximately76.9%ofthevarianceR 2 incontinuanceintentionandisconsidered tohaveasubstantiallevelofexplanation 17 ;themodelalsoexhibitspredictive relevanceQ 2 whereitsvalueis0.5401scoresabovezeroindicatepredictiverelevanceinPLSpathmodels.Satisfactionandperceivedusefulness,in acombinedmeasure,predictcontinuanceintentionwithastandardizedpath coecient of0.3672 t =14.0414.Costshadthesecondlargestpath coecient =0.2381, t =7.8851,followedbyalternativeperception =-0.2191, t =10.5662,habit =.1564, t =6.1575,personalinnovativeness =0.0618, t =3.6746,interpersonalinuence =-0.0439, t = 3.0296,moderatingfactorsatisfactionandperceivedusefulnessxpersonalinnovativeness =-0.0334, t =1.5965,andmoderatingfactorsatisfaction andperceivedusefulnessxhabit =-0.0223, t =0.9708. RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SatAndPU CONTINUANCE0.367214.0414 COSTS CONTINUANCE0.23817.8851 AlternativePerception CONTINUANCE-0.219110.5662 17 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 141

PAGE 163

RelationPath Coecient Tstatistic HABIT CONTINUANCE0.15646.1575 age CONTINUANCE0.06184.6029 PERSINNOV CONTINUANCE-0.05703.6746 InterpersInu CONTINUANCE-0.04393.0296 SatAndPU*PERSINNOV CONTINUANCE-0.03341.5965 SatAndPU*HABIT CONTINUANCE-0.02230.9708 education CONTINUANCE0.02061.4701 gender CONTINUANCE0.01801.2129 OtherPathCoecients Relationship_costs COSTS0.980286.0357 PROC_COST ProceduralCosts0.947714.0233 SAT SatAndPU0.784634.1109 PersRelationLoss Relationship_costs0.733734.2792 CONF SAT0.670133.1767 AlterAttract AlternativePerception0.660818.8908 CONF PU0.599131.3832 AttitudeToSwitch AlternativePerception0.456711.5569 BrandRelationship Relationship_costs0.40715.7015 PU SatAndPU0.308211.0637 142

PAGE 164

RelationPath Coecient Tstatistic PU SAT0.19688.5347 PROC_LEARN ProceduralCosts0.15031.6705 PROC_EVAL ProceduralCosts-0.12391.7226 ProceduralCosts COSTS0.03581.5859 PROC_SETUP ProceduralCosts0.00450.0496 TotalEects ContinuancePathCoecents SatAndPU CONTINUANCE0.367214.0414 CONF CONTINUANCE0.294813.6612 SAT CONTINUANCE0.288112.511 COSTS CONTINUANCE0.23817.8851 Relationship_costs CONTINUANCE0.23347.8567 AlternativePerception CONTINUANCE-0.219110.5662 PersRelationLoss CONTINUANCE0.17137.585 PU CONTINUANCE0.169911.1883 HABIT CONTINUANCE0.15646.1575 AlterAttract CONTINUANCE-0.14489.9052 AttitudeToSwitch CONTINUANCE-0.10017.1652 BrandRelationship CONTINUANCE0.09507.2462 143

PAGE 165

RelationPath Coecient Tstatistic age CONTINUANCE0.06184.6029 PERSINNOV CONTINUANCE-0.05703.6746 InterpersInu CONTINUANCE-0.04393.0296 SatAndPU*PERSINNOV CONTINUANCE-0.03341.5965 SatAndPU*HABIT CONTINUANCE-0.02230.9708 education CONTINUANCE0.02061.4701 gender CONTINUANCE0.0181.2129 ProceduralCosts CONTINUANCE0.00851.5315 PROC_COST CONTINUANCE0.00811.5399 PROC_LEARN CONTINUANCE0.00131.0715 PROC_EVAL CONTINUANCE-0.00111.2122 PROC_SETUP CONTINUANCE0.00000.0405 OtherPathCoecients Relationship_costs COSTS0.980286.0357 PROC_COST ProceduralCosts0.947714.0233 CONF SatAndPU0.802981.0546 CONF SAT0.787974.9908 SAT SatAndPU0.784634.1109 PersRelationLoss Relationship_costs0.733734.2792 144

PAGE 166

RelationPath Coecient Tstatistic PersRelationLoss COSTS0.719235.4088 AlterAttract AlternativePerception0.660818.8908 CONF PU0.599131.3832 PU SatAndPU0.462616.2700 AttitudeToSwitch AlternativePerception0.456711.5569 BrandRelationship Relationship_costs0.407015.7015 BrandRelationship COSTS0.398914.8049 PU SAT0.19688.5347 PROC_LEARN ProceduralCosts0.15031.6705 PROC_EVAL ProceduralCosts-0.12391.7226 ProceduralCosts COSTS0.03581.5859 PROC_COST COSTS0.03391.6038 PROC_LEARN COSTS0.00541.0816 PROC_SETUP ProceduralCosts0.00450.0496 PROC_EVAL COSTS-0.00441.2504 PROC_SETUP COSTS0.00020.0414 145

PAGE 167

Table24:Indices GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 AlternativePerception 0.99240.5354 CONTINUANCE 0.76870.5401 COSTS 0.99420.2179 PU 0.35890.2433 ProceduralCosts 0.99330.3252 Relationship_costs 0.99880.5007 SAT 0.64570.6010 SatAndPU 0.99990.6084 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. 7.7.4ModeratingEectAnalysis-Habit TheresultsofthemoderatingtermarelessthansmallaccordingtoChinetal. .TheR 2 forthenon-moderatedmodelis0.7667andforthemoderated modelitis0.7677.Thepathcoecient for satisfactionandperceived usefulness inthemoderatedmodelis0.364andnon-moderatedmodelare 0.367.Thepathcoecientforhabitinthemoderatedmodelis0.1553and non-moderatedmodelare0.1642.Thepathcoecientfor satisfactionand perceivedusefulnessxhabit inthemoderatedmodelis-0.0348anditssign isnegativeindicatesincreasesinhabitdecreasestheeectsatisfactionand perceivedusefulnessoncontinuanceintentionasexpectedfromthetheoretical model.Themoderatingt-statisticsfor satisfactionandperceivedusefulnessx habit indicatesthatthemoderatorisnotstatisticallysignicant t =1.8136 andtheeectsize f 2 inthechangeofthecoecientofdeterminationR 2 is 0.0043whichdoesnotmeetthethresholdfortobeconsideredashavingan 146

PAGE 168

Figure36:CompleteModeratedSimpliedModel Figure37:CompleteModeratedModel 147

PAGE 169

eectthresholdforsmalleects f 2 is.02basedon Chinetal. 7.7.5ModeratingEectAnalysis-PersonalInnovativeness TheresultsofthemoderatingtermarelessthansmallaccordingtoChinetal. .TheR 2 forthenon-moderatedmodelis.7667andforthemoderated modelitis.7683.Thepathcoecient for satisfactionandperceivedusefulness inthemoderatedmodelis.3702andnon-moderatedmodelare.367. Thepathcoecientforpersonalinnovativenessinthemoderatedmodelis -0.0581andnon-moderatedmodelare.0591.Thepathcoecientfor satisfactionandperceivedusefulnessxpersonalinnovativeness inthemoderated modelis-0.0401.Themoderatingt-statisticsfor satisfactionandperceived usefulnessxpersonalinnovativeness indicatesthatthemoderatorisstatisticallysignicant t =1.9931andtheeectsize f 2 inthechangeofthecoecient ofdeterminationR 2 is0.0069whichdoesnotmeetthethresholdfortobe consideredashavinganeectthresholdforsmalleects f 2 is.02basedon Chinetal. 7.7.6ModeratingEectAnalysis-TwoFactors-HabitandPersonalInnovativeness TheresultsofthemoderatingtermarelessthansmallaccordingtoChin etal..TheR 2 forthenon-moderatedmodelis0.7667andforthe moderatedmodelitis0.7687.Thepathcoecient for satisfactionand perceivedusefulness inthemoderatedmodelis0.3672andnon-moderated modelare.3426.Thepathcoecientforhabitinthemoderatedmodelis 0.1564andnon-moderatedmodelare0.1642.Thepathcoecientforpersonal innovativenessinthemoderatedmodelis-0.0570andnon-moderatedmodel are-0.0589. 148

PAGE 170

Thepathcoecientfor satisfactionandperceivedusefulnessxhabit inthe moderatedmodelis-0.0223 t =0.9708.Thepathcoecientfor satisfaction andperceivedusefulnessxpersonalinnovativeness inthemoderatedmodelis0.0334 t =1.5965.Themoderatingt-statisticsforbothmoderatorsindicates thatneithermoderatorsisstatisticallysignicantandtheeectsize f 2 is0.0086 whichdoesnotmeetthethresholdfortobeconsideredashavinganeect thresholdforsmalleects f 2 is.02basedon Chinetal. Formorene-grainanalysisofthemoderatingfactorsseeAppendixCand D. 149

PAGE 171

Figure38:ModeratedModelPathCoecientsonSNSContinuance 150

PAGE 172

7.8CompleteNon-Moderatedmodel-BackwardStepwiseRenement 7.8.1ModelDescription Thismodelincludesthebackwardstepwiserenementofthecompletenonmoderatedmodelwheretheconstructswiththeleaststatisticalsignicance areremovedinastepwisemannerofleastimpacttotheoverallcoecient ofdetermination.ThemodelincludesthemeasuresfromISContinuance modelBhattacherjee,2001,theconsumerswitchingcostmodelBurnham etal.,2003andhabit,personalinnovativeness,interpersonalinuence,and alternativeperceptions.ThetwopredictorsforIScontinuance,satisfaction andperceivedusefulnesshavebeencombinedintoasinglereective-formative constructthatallowsforamoredirectcomparisonofthecompetingmodels. TheBurnhametal.costmodelwasinitiallydevelopedasaformativereectivemodelanddoesnotneedatransformation.Theconsumerswitching costmodelusestheindependentconstructseconomicrisk,evaluationcosts, learningcostsandsetupcostsproceduralcostsandpersonalrelationshiploss andbrandrelationshiplossrelationalcoststopredictinformationsystems continuanceintention.TheIScontinuancemodelusestheindependentconstructsconrmation,perceivedusefulness,satisfactiontopredictinformation systemscontinuanceintention.Twomoderatorshavebeendropped,habit xsatisfactionandpersonalinnovativenessxsatisfactionaspreviousanalyses haveshowntherelationshipwithcontinuanceintentiondidnothaveamoderatingeect. 151

PAGE 173

7.8.2Renements Renementsweremadebyexaminingthetotaleectstablesforexogenous constructsandthosewiththeloweststatisticalrelevancewereremovedina stepwisemanner.Thecovariatesofage,genderandeducationareincluded untilthelaststeptoadjustresultsforcovariatesasthesevariablesareunder secondaryconsideration.Therenementsremovedtheproceduralcostsinthe Burnhametal.costmodelasallofthecostswerebelowthethreshold. Twocovariateswereremoved,genderandeducation,asthevalueswerebelow thethreshold;agewasretainedasastatisticallysignicantcovariate.See Table26fordetails. 7.8.3MeasurementModel Afterallconstructsandcovariatesthathadt-statisticvalueslessthan2.58 wereremoved,themeasurementmodelwasassessedforreliabilityconstruct indicatorreliabilityandinternalconsistencyandvalidityconvergentand discriminant.AllaveragevarianceextractedAVEvaluesareabovethe.50 threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliabilityvaluesareallgreaterthanorequalto.7providingevidenceof theconstruct'sinternalconsistencyreliabilityHairetal.,2006.Composite reliabilityvaluesareallgreaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006.ThesquarerootsoftheAVEareallgreaterthanthe latentvariablecorrelationsindicatingdiscriminantvalidityFornellandLarckercriterion-See:28.Theindicatorsinthereectivemeasurement modelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodel assessmentsubstantiatesthatalltheconstructmeasuresarereliableandvalid. Theresultingtablesshowthecompletemodelincludingthemoderatingeects althoughthemoderatingeectsarelessthansmall. 152

PAGE 174

Table26:Renement StepConstructRemovedTotal EectPath Coecient T-statisticR 2 Initial Model .767 1PROC_SETUP CONTINUANCE 0.00020.1451.767 2PROC_LEARN CONTINUANCE 0.00141.1316.767 3PROC_EVAL CONTINUANCE -0.00070.8428.767 4PROC_COST CONTINUANCE 0.01681.1299.766 AllConstructswithT-statisticslessthan1.96and2.58removed StepCovariatesRemovedTotal EectPath Coecient T-statisticR 2 Initial Model .766 1Gender0.01671.1986.766 2Education0.02111.3944.765 3Age0.06404.3821.762 AllCovariatesremoved R 2 forcontinuanceafterremoval 153

PAGE 175

Table27:SNSCostMeasurementModel ConstructAVECompositeReliability PersRelationLoss0.64240.8993 AlterAttract0.64760.9013 AttitudeToSwitch0.73790.9440 BrandRelationship0.68130.8640 CONF0.69070.8968 CONTINUANCE0.71790.8838 HABIT0.76570.9072 InterpersInu0.60900.8600 PERSINNOV0.70430.9221 PU0.68040.8938 SAT0.93080.9758 154

PAGE 176

Table28:StaticallySignicantFactors-MeasurementModel-DiscriminantValidity FornellandLarckerCriterion Proc Relation Loss Alter Attract Attitude to Switch Brand Relationship CONFCONTINUANCE HABITInterpers Inu PERS INNOV PUSAT Proc Relation Loss 0.801 Alter Attract -0.449 0.805 Attitude ToSwitch -0.416 0.575 0.859 BrandRelationship 0.493 -0.488 -0.327 0.825 CONF 0.673 -0.497 -0.443 0.651 0.831 CONTINUANCE 0.732 -0.614 -0.556 0.609 0.711 0.847 HABIT 0.744 -0.418 -0.387 0.497 0.635 0.695 0.875 Interpers Inu 0.196 -0.022 -0.121 0.166 0.214 0.108 0.147 0.780 PERS INNOV -0.031 0.280 0.414 0.049 -0.042 -0.165 -0.012 -0.055 0.839 PU 0.757 -0.372 -0.351 0.473 0.599 0.631 0.656 0.258 -0.029 0.825 SAT 0.670 -0.567 -0.459 0.637 0.789 0.787 0.623 0.112 -0.038 0.599 0.965 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagonalelementsarethelatentvariablecorrelations. 155

PAGE 177

Table29:Indices GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 AlternativePerception0.99250.5357 CONTINUANCE0.76540.5459 COSTS0.99880.5011 PU0.35910.2435 SAT0.64690.6021 SatAndPU0.99990.6097 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. 7.8.4StructuralModel ThestructuralmodelwasassessedtodeterminehowtheindependentconstructsfromISContinuancetheoryandconsumerswitchingcostspredictinformationsystemscontinuanceintention..Thepredictorsexplainedapproximately76.5%ofthevarianceR 2 incontinuanceintentionandisconsidered tohaveasubstantiallevelofexplanation 18 ;themodelalsoexhibitspredictive relevanceQ 2 whereitsvalueis.5459scoresabovezeroindicatepredictive relevanceinPLSpathmodels.Satisfactionandperceivedusefulness,ina combinedmeasure,predictcontinuanceintentionwithastandardizedpath coecient of.373.Allofthecostspredictcontinuanceintentionwitha standardizedpathcoecientof.239.Alternativeperceptionshadthethird largestpathcoecient-.208,followedbyhabit.171,personalinnovativeness-.072andinterpersonalinuence-.041.Coecientsthatarepositive indicatethatareassociatedwithhigherlevelsofcontinuanceintention. 18 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 156

PAGE 178

RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SatAndPU CONTINUANCE0.373713.6107 COSTS CONTINUANCE0.23937.953 AlternativePerception CONTINUANCE-0.207510.6461 HABIT CONTINUANCE0.17127.2769 PERSINNOV CONTINUANCE-0.07214.7918 InterpersInu CONTINUANCE-0.04272.9938 OtherPathCoecients CONF SAT0.670633.0243 PersRelationLoss COSTS0.725732.2281 CONF PU0.599331.1625 SAT SatAndPU0.78531.1428 AlterAttract AlternativePerception0.66217.3973 BrandRelationship COSTS0.416115.3979 AttitudeToSwitch AlternativePerception0.455410.7596 PU SatAndPU0.30759.8987 PU SAT0.19718.7098 TotalEects 157

PAGE 179

RelationPath Coecient Tstatistic ContinuancePathCoecents SatAndPU CONTINUANCE0.372213.6107 CONF CONTINUANCE0.29913.2167 SAT CONTINUANCE0.292111.7331 PU CONTINUANCE0.171711.0174 AlternativePerception CONTINUANCE-0.209110.6461 AlterAttract CONTINUANCE-0.13889.5387 COSTS CONTINUANCE0.24167.953 PersRelationLoss CONTINUANCE0.17517.5119 BrandRelationship CONTINUANCE0.10087.4619 HABIT CONTINUANCE0.16877.2769 AttitudeToSwitch CONTINUANCE-0.09457.1759 PERSINNOV CONTINUANCE-0.07164.7918 age CONTINUANCE0.06414.6439 InterpersInu CONTINUANCE-0.0412.9938 OtherPathCoecients CONF SatAndPU0.803474.8767 CONF SAT0.788870.1917 PersRelationLoss COSTS0.724332.2281 158

PAGE 180

RelationPath Coecient Tstatistic CONF PU0.599731.1625 SAT SatAndPU0.784431.1428 AlterAttract AlternativePerception0.664317.3973 BrandRelationship COSTS0.417515.3979 PU SatAndPU0.461515.372 AttitudeToSwitch AlternativePerception0.451510.7596 PU SAT0.19628.7098 Figure39:BaseModelAndCostsPathCoecientsonSNSContinuance 159

PAGE 181

7.9Backwardstepwiserenementwitheectsizes Abackwardstepwiserenementmodelwasdevelopedwheretheinitialmodel containedalltheoreticalconstructsofthenon-moderatedmodelandwereremovedfromthemodelinastepwisemannerbasedonthetotaleectpath coecientsmallestpathcoecientrst.TheinitialmodelhadfourteenconstructsandthreecovariatesandaninitialR 2 of.767.Thenalmodelhadone construct,satisfaction,andhadanR 2 of.623.Satisfaction,alone,canexplain 62.3%ofthevarianceincontinuanceintention.Table31showsthestepsin theanalysisstartingwiththeremovaloftheconstructwiththeleasttotal eectoncontinuance,proceduralsetupcost,throughthelastremainingconstruct,satisfaction.Theeectsize f 2 iscalculatedtoshowthedierencethat theconstructhadineectascomparedtothepreviousstepandtheinitial modelbasedonthechangeinthecoecientofdetermination R 160

PAGE 182

Table31:StepwiseRenementwitheectsize StepConstruct Removed Total Eect Path Coecient T-statistic R after removal f 2 previousstep f 2 description -previous step f 2 initial f 2 description -initial Initial Model 0.7670 1 PROC_SETUP CONTINUANCE 0.00020.14510.76700.0000lessthan small 0.0000lessthan small 2 PROC_LEARN CONTINUANCE 0.00141.13160.76700.0000lessthan small 0.0000lessthan small 3 PROC_EVAL CONTINUANCE -0.00070.84280.76700.0000lessthan small 0.0000lessthan small 161

PAGE 183

Table31:StepwiseRenementwitheectsize StepConstruct Removed Total Eect Path Coecient T-statistic R after removal f 2 previousstep f 2 description -previous step f 2 initial f 2 description -initial 4 PROC_COST CONTINUANCE 0.01681.12990.76600.0043lessthan small 0.0043lessthan small 5 gender CONTINUANCE 0.01661.24810.76600.0000lessthan small 0.0043lessthan small 6 education CONTINUANCE 0.02111.48030.76500.0043lessthan small 0.0085lessthan small 7 InterpersInu CONTINUANCE -0.04333.11880.76400.0043lessthan small 0.0127lessthan small 162

PAGE 184

Table31:StepwiseRenementwitheectsize StepConstruct Removed Total Eect Path Coecient T-statistic R after removal f 2 previousstep f 2 description -previous step f 2 initial f 2 description -initial 8 PERSINNOV CONTINUANCE -0.06824.49370.76000.0169lessthan small 0.0292smallto medium 9 age CONTINUANCE 0.06704.87480.75600.0167lessthan small 0.0451smallto medium 10 BrandRelationship CONTINUANCE 0.09817.41680.75100.0205smallto medium 0.0643smallto medium 163

PAGE 185

Table31:StepwiseRenementwitheectsize StepConstruct Removed Total Eect Path Coecient T-statistic R after removal f 2 previousstep f 2 description -previous step f 2 initial f 2 description -initial 11 AttitudeToSwitch CONTINUANCE -0.12268.44820.73800.0522smallto medium 0.1107smallto medium 12 HABIT CONTINUANCE 0.17076.27070.72600.0458smallto medium 0.1496smallto medium 13 PU CONTINUANCE 0.224012.41210.7270-0.0036lessthan small 0.1465smallto medium 14 AlterAttract CONTINUANCE -0.215210.32650.69700.1099smallto medium 0.2310mediumto large 164

PAGE 186

Table31:StepwiseRenementwitheectsize StepConstruct Removed Total Eect Path Coecient T-statistic R after removal f 2 previousstep f 2 description -previous step f 2 initial f 2 description -initial 15 PersRelationLoss CONTINUANCE 0.372015.92910.62300.2442mediumto large 0.3820large 16 CONF CONTINUANCE 0.622144.65530.62300.0000lessthan small 0.3820large Final Model SAT CONTINUANCE 0.789470.54150.62300.0000lessthan small 0.3820large 165

PAGE 187

Keyparenthesisindicateeectsizefrominitialmodel:
PAGE 188

7.10ThreeDatasetcomparisons Threedistinctsetsofdatawerecollectedinthecourseofthisstudy;apublic dataset,astudentdatasetandanOracledataset.Theresultsanalyzedused onlythepublicdatasetthosewhocametothesurveythroughapublicnotice.Thestudentdatasetwasgatheredthroughstudentcontactsandcontains 217cases.TheOracledatasetwasgatheredthroughapublicnoticetothree dierentOraclelocationsinthestateofColoradoandcontains46cases.The datawasanalyzedtoseeifthereweredierencesbetweenthethreedatasets. Theconstructsarerelativelysimilaracrossthethreedatasetswhilethereare largerdierencesindemographicsbetweenthestudentdatasetandthetwo othersetswheretheageofthestudentsismuchyoungerandtheeducation levelisalsolower. Thedatasetssharemanysimilaritiesindierentanalyses,butthereare dierencesaswell.TheR 2 ishighestfortheOracledatasetat.8838,followed bythepublicdataset.7667thenthestudentdataset.6896-seeTable32 forthecompletenon-moderatedmodel. Therearetwentyconstructsanddemographicvariablesintheanalysis; somestatisticallysignicantdierencescanbefoundamongtheconstructs, buttherearemanysimilarities.Inat-testbetweenthepublicdatasetand studentdatasettherewerethreeconstructswithstatisticallysignicantdierencesandallthreedemographicswerestatisticallysignicantdierent-and theremaining17constructshadnostatisticallysignicantdierences.InattestbetweenthepublicdatasetandOracledatasetthereweresevenconstructs withstatisticallysignicantdierencesandonedemographicagethatwere statisticallysignicantdierent-theremaining14constructsandtwodemographicseducationandgenderhadnostatisticallysignicantdierences.In at-testbetweenthestudentdatasetandOracledatasettherewereninecon167

PAGE 189

structswithstatisticallysignicantdierencesandtwodemographicsage& educationthatwerestatisticallysignicantdierent-theremaining12constructsandgenderhadnostatisticallysignicantdierences-SeeTables33, 34,35,and36andFigure41.Theconstructsarenotablysimilarwithsome exceptionsasnoted. DatasetpathcomparisonsusingMANOVAinSPSSbycollectingabootstrapsampleof500iterationsacrossthethreedatasetsareshowninTable37 and38andFigures42,43signicantpathsonlyand44Cohen'sD.The MANOVAanalysiswasfoundtobesignicantusingfourmeasures,Pilia's Trace,Wilk'sLambda,Hotelling'sTrace,andRoy'sLargestRoot.Thepaths coecients ofthepublicandstudentdatasetaresimilarinnatureandthe pathcoecientsbetweentheOracledatasetandthetwootherdatasetsdier innotableways.TheOracledatasetismuchsmallerwhichmayleadtosome over-ttingofthePLS-SEMmodel,howeverthemodelsalsodierindemographicswheretheOracledatasetcontainsanolderpopulation.Ageaccounts forasmallamountofthevarianceinthemodel = : 0623 soitisnotthat theageoftheOraclesampleisolder.Theconsumerswitchingcostfactorsare muchlargerfortheOracledatasetwhere,inparticular,therelationshipcosts andbrandrelationshipcostsweremuchhigherthantheothertwodatasets. SatisfactionandperceivedusefulnesshadlittleeectintheOraclesampleas wellbecausethemodelmajorfactorswerefromtheconsumerswitchingcost model.Therewere11statisticallysignicantdierencesbetweenthepublicandstudentdatasetand19statisticallysignicantdierencesbetweenthe publicandOracledatasetand20statisticallysignicantdierencesbetween thestudentandOracledataset.Thereareatotalof22constructsmeasured inthecompletemodel. Tables37and38showthepathcoecientsfortheoriginalsampleand 168

PAGE 190

theirstandarderroraswellasabootstrapsampleof500iterationsforeach datasetandacomparisonofstatisticalsignicancetoadierentdataset.Hair etal.notesthatlargesamplesizesof400orlargerreducesampling errortosuchasmalllevelthatsmalldierencesareconsideredstatistically signicant.EectssizeswerecalculatedinTable38andshowninFigure44. LargeeectsmaybeconsideredwhereCohen'sDisgreaterthan5.Forthis research,Cohen'sDofsmall,mediumandlargeforthepathcoecientsmaybe consideredatthe5,10,and20level.Therewhere15pathcoecientsthatwere substantiallysimilarbetweenthepublicandstudentdataset,2smalleects betweenthepublicandstudentdataset,4medium,and1large.Therewhere 6thatweresubstantiallysimilarbetweenthepublicandOracledataset,2 smalltomediumeects,1mediumtolargeeect,and13large.Therewhere4 eectsthatweresubstantiallysimilarbetweenthestudentandOracledataset, 1smalltomediumeect,4mediumtolargeeects,and13large.Theseresults suggestthatthepathcoecientsbetweenthepublicandstudentdatasetare largelysimilar,whereasthepublic-Oracleandstudent-Oracledatasethavea greaternumberandlargerdierences. 169

PAGE 191

Table32:CoecientofDeterminationComparisonAcrossThreeDatasets Public R 2 Student R 2 Oracle R 2 CONTINUANCE0.76670.68960.8838 PublicDatasetN=1302,StudentDatasetN=217,OracleDatasetN=46 170

PAGE 192

Table33:ConstructComparisonacrossdatasets-1 ValuePROC_SETUPProcRelation Loss AlterAttractAlternative Perception AttitudeTo Switch Brand Relationship AveragePublicScore4.21774.87133.80593.72243.64283.4080 Std.DevPublic1.14311.27901.01650.94701.02891.3414 AverageStudentScore4.18034.58793.72003.72853.76323.9829 std.DevStudent0.98101.09670.97860.93591.03621.1098 AverageOracleScore4.47335.15963.61183.72273.54393.5276 std.devOracle1.02261.25000.91040.79341.13331.2483 t-testpublictostudent0.64900.0021***0.24660.92940.11100.0000*** t-testpublictoOracle0.13510.13300.20170.99830.52330.5515 t-teststudenttoOracle0.06890.0019***0.49130.96860.20080.0141** 171

PAGE 193

Table34:ConstructComparisonacrossdatasets-2 ValueCONFCONTINUANCECOSTSHABITInterpersInuPERS INNOV AveragePublicScore4.30234.69224.36934.83923.38563.5773 Std.DevPublic1.25081.52541.09711.49251.44791.4575 AverageStudentScore4.43814.68984.34074.75753.90003.6583 std.DevStudent0.91691.21620.91791.36201.33601.3083 AverageOracleScore4.28264.86704.74094.25853.29962.4441 std.devOracle1.35911.55971.16391.49991.44251.5157 t-testpublictostudent0.12550.98300.71650.45050.0000***0.4418 t-testpublictoOracle0.91660.44540.0244**0.0096***0.69210.0000*** t-teststudenttoOracle0.34210.39540.0112**0.0275**0.0068***0.0000*** 172

PAGE 194

Table35:ConstructComparisonacrossdatasets-3 ValuePROC_COSTPROC_EVALPROC_LEARNPUProcedural Costs Relationship_costs AveragePublicScore4.22354.17934.14725.10114.21044.3789 Std.DevPublic1.14351.24191.17321.24401.15291.1495 AverageStudentScore4.16944.30674.15694.95664.20214.3718 std.DevStudent1.05131.07871.07641.15981.24260.9578 AverageOracleScore4.57904.65844.55265.08884.15904.7169 std.devOracle1.05561.24671.00661.25261.35321.0794 t-testpublictostudent0.51420.15450.90930.11010.92200.9312 t-testpublictoOracle0.0380**0.0102**0.0208**0.94780.76760.0497** t-teststudenttoOracle0.0172**0.05190.0228**0.48920.83370.0309** 173

PAGE 195

Table36:Comparisonacrossdatasets-4 ValueSATSatAndPUageeducationgender AveragePublicScore4.48724.69305.37445.08471.6048 Std.DevPublic1.49601.29192.41631.13390.4883 AverageStudentScore4.63324.73972.26733.46541.4747 std.DevStudent1.12540.99451.13960.85520.5005 AverageOracleScore4.71784.85967.23915.04351.6304 std.devOracle1.58961.32672.17241.21030.4880 t-testpublictostudent0.16960.61180.0000***0.0000***0.0003*** t-testpublictoOracle0.30540.39060.0000***0.80880.7262 t-teststudenttoOracle0.66910.48610.0000***0.0000***0.0552 174

PAGE 196

Figure41:ConstructMeanComparisonAcrossThreeDatasets 175

PAGE 197

Table37:PathComparisonsacrossthreedatasets-BootstrapSample PathPublic Path Student Path Oracle Path STERR p publicstudent p publicoracle p studentoracle SatAndPU CONT0.36580.34660.02240.00670.0131**0.0000***0.0000*** CONF CONT0.29370.23930.02050.00580.0000***0.0000***0.0000*** SAT CONT0.28740.27640.01910.00490.07550.0000***0.0000*** COST CONT0.25760.25760.90900.00761.00000.0000***0.0000*** Relationship_costs CONT0.25190.24760.84140.00851.00000.0000***0.0000*** AlternativePerception CONT-0.2076-0.1649-0.06470.00470.0000***0.0000***0.0000*** PersRelationLoss CONT0.18470.17930.65900.00921.00000.0000***0.0000*** PU CONT0.16870.16530.00590.00350.99330.0000***0.0000*** HABIT CONT0.15960.22070.01350.00480.0000***0.0000***0.0000*** AlterAttract CONT-0.1369-0.1206-0.04060.00350.0000***0.0000***0.0000*** BrandRelationship CONT0.10250.10130.27290.00441.00000.0000***0.0000*** AttitudeToSwitch CONT-0.0948-0.0598-0.02640.00240.0000***0.0000***0.0000*** age CONT0.06170.0560-0.00300.00240.0480**0.0000***0.0000*** PERSINNOV CONT-0.0582-0.06100.00300.00331.00000.0000***0.0000*** InterpersInu CONT-0.0441-0.0036-0.05750.00280.0000***0.0000***0.0000*** education CONT0.0212-0.02860.01810.00280.0000***0.77050.0000*** gender CONT0.00990.02010.05010.00250.0000***0.0000***0.0000*** ProceduralCosts CONT0.00990.02010.05010.00500.13120.0000***0.0000*** PROC_COST CONT0.00930.01510.00560.00330.24490.79970.0131** PROC_LEARN CONT0.00150.01140.00950.00230.0001***0.0018***1.0000 PROC_EVAL CONT-0.0011-0.0047-0.00060.00431.00001.00001.0000 PROC_SETUP CONT0.0000-0.00280.01050.00280.99230.0007***0.0000*** TotalStatisticallySigDierences111920 PublicDatasetN=1302,StudentDatasetN=217,OracleDatasetN=46 CONT:ContinuanceIntention 176

PAGE 198

PublicDatasetN=1302,StudentDatasetN=217,OracleDatasetN=46 Figure42:PathComparisonAcrossThreeDatasets 177

PAGE 199

PublicDatasetN=1302,StudentDatasetN=217,OracleDatasetN=46 Figure43:PathComparisonAcrossThreeDatasets-StatSigPathsonly 178

PAGE 200

Table38:PathComparisonsacrossthreedatasetsCohen'sd-BootstrapSample PathPublic Path Student Path Oracle Path Cohen'sd publicstudent Cohen'sd publicoracle Cohen'sd studentoracle SatAndPU CONT0.36580.34660.02242.8548 50.9002 48.0454 CONF CONT0.29370.23930.0205 9.4272 47.4070 37.9798 SAT CONT0.28740.27640.01912.2413 54.8358 52.5945 COST CONT0.25760.25760.9090-0.5244 -85.7817 -85.2573 Relationship_costs CONT0.25190.24760.84140.5040 -69.0345 -69.5385 AlternativePerception CONT-0.2076-0.1649-0.0647 -9.1375 -30.5789 -21.4414 PersRelationLoss CONT0.18470.17930.65900.5863 -51.5961 -52.1824 PU CONT0.16870.16530.00590.9722 46.9885 46.0163 HABIT CONT0.15960.22070.0135 -12.7453 30.5029 43.2482 AlterAttract CONT-0.1369-0.1206-0.0406-4.5920 -27.1536 -22.5616 BrandRelationship CONT0.10250.10130.27290.2633 -39.0932 -39.3565 AttitudeToSwitch CONT-0.0948-0.0598-0.0264 -14.5377 -28.4227 -13.8850 age CONT0.06170.0560-0.00302.4119 27.2705 24.8586 PERSINNOV CONT-0.0582-0.06100.00300.8804 -18.7312 -19.6116 InterpersInu CONT-0.0441-0.0036-0.0575 -14.25024.7248 18.9750 education CONT0.0212-0.02860.0181 17.97491.1344 -16.8406 gender CONT0.00990.02010.0501 24.1529 -5.4382 -29.5911 ProceduralCosts CONT0.00990.02010.0501-2.0182 -7.9617 -5.9435 PROC_COST CONT0.00930.01510.0056-1.74251.11142.8539 PROC_LEARN CONT0.00150.01140.0095-4.2702-3.44170.8286 PROC_EVAL CONT-0.0011-0.0047-0.00060.8331-0.1274-0.9605 PROC_SETUP CONT0.0000-0.00280.01050.9729-3.7013-4.6742 TotalStatisticallySigDierences Shadeofgreydenotessmalltomedium,mediumtolarge,andlargeeectssize CONT:ContinuanceIntention 179

PAGE 201

Table39:Cohen'sDCategorizationofPathCoecientDierences EectCohen'sDValuePublic-StudentNPublic-OracleNStudent-OracleN LessthanSmall <5 1564 Smalltomedium >=5 221 Mediumtolarge >=10 414 Large >=20 11313 Total222222 180

PAGE 202

PublicDatasetN=1302,StudentDatasetN=217,OracleDatasetN=46 Figure44:PathComparison-Cohen'sDEectSizeThreeDatasets 181

PAGE 203

7.11Summary Theresearchexaminedmanymodelstodeterminewhichtheindividualeects oftheconstructsandanoverallmodelevaluationwithmultipleeects.An evaluationofthehypothesesexaminedisshownisTable40.Asummarytable oftheresultsisprovided-seeTable41. Figure45:SocialNetworkingSiteContinuanceModel 182

PAGE 204

Table40:SocialNetworkingSiteContinuanceModelHypothesesEvaluated NumResultHypothesis 1 YSocialnetworkingsiteuserswithhighapositiveattitudeto switchandwhoareattractedtocompetingsocial networkingsiteswillnegativelyaectcontinuanceintention onthesocialnetworkingsite. 2a YPersonalinnovativenesswillnegativelyaectcontinuance intentionofsocialnetworkingsites. 2b NTherelationshipbetweensatisfactionandsocialnetworking sitecontinuanceintentionwillbenegativelymoderatedby personalinnovativenessonsocialnetworkingsites. 3 YInterpersonalinuencewillnegativelyaectcontinuance intentionofsocialnetworkingsites. 4a YHabitwillpositivelyaectcontinuanceintentionofsocial networkingsites. 4b NTherelationshipbetweensatisfactionandsocialnetworking sitecontinuanceintentionwillbenegativelymoderatedby habit. 5 YGreaterproceduralandrelationalswitchingcostswill positivelyaectcontinuanceintentionofsocialnetworking sites. 183

PAGE 205

Table41:SummaryTable ModelName R 2 R 2 Interpretation f 2 f 2 Interpretation specicproductoneormorecompetitors0.3713 weaktomoderate-0.4714 largeneg specicproductallfourcompetitors0.6072 moderatetosubstantial-0.1540 mediumtolargeneg costonly0.6433 moderatetosubstantial-0.0684 smalltomediumneg basemodelISContinuance 0.6677 moderatetosubstantial 0.0000 lessthansmall basemodel+interpersonalinuence0.6685 moderatetosubstantial0.0024 lessthansmall basemodel+personalinnovativeness0.6839 moderatetosubstantial0.0512 smalltomedium basemodel+habit0.7116 moderatetosubstantial0.1522 mediumtolarge basemodel+cost0.7165 moderatetosubstantial0.1721 mediumtolarge basemodel+alternativeperceptions0.7205 moderatetosubstantial0.1889 mediumtolarge completewithstatsig.predictors0.7654 substantial0.4165 large completemodelwithallpredictors0.7667 substantial0.4243 large completemodelwithtwomoderators0.7687 substantial0.4367 large BasemodelisIScontinuance basemodel+indicatesoneaddedfactor completemodelindicatesthatmultiplefactorsareaddedandnoproductspecicfactorsareadded specicproductmodelisforattitudestowardsspecicproductslikeTwitter,TUMBLR,PinterestandInstagram R 2 and f 2 interpretationbasedonHairetal. 184

PAGE 206

R 2 Interpretations:Weak=0.25,Moderate=0.50,Substantial=0.75 f 2 Interpretations:Small=0.02,Medium=0.15,Large=0.35 R 2 and f 2 interpretationbasedonHairetal. Figure46: R 2 and f 2 ContinuanceByModel 185

PAGE 207

Table42:GoodnessofFitcomparison ModelNameGoodnessofFitGoFInterpretation specicproductoneormorecompetitors0.5427large specicproductallfourcompetitors0.6908large costonlyN/A basemodelISContinuance0.6490large basemodel+interpersonalinuence0.6360large basemodel+personalinnovativeness0.6394large basemodel+habit0.6590large basemodel+costN/A basemodel+alternativeperceptionsN/A completemodelwithstatsig.predictorsN/A completenon-moderatedmodelN/A completemoderatedmodelN/A GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 186

PAGE 208

8Discussion Thegoalofthisresearchwastoextendinformationsystemscontinuanceof Bhattacherjeewhichpredictscontinuanceintentionthroughthreefactors:perceivedusefulness,conrmationandsatisfactiontoexplainmoreof varianceofcontinuanceintention.ResearcherwhouseonlythefactorsinIS continuancelimittheamountofexplainedvarianceofcontinuanceintention whenadditionaldirectandmoderatingfactorsincontinuancedecisionshave importantimpacts.Thisresearchexaminedimportantmotivatingfactorsto predictanindividual'sintentiontocontinuetouseasocialnetworkingsite, including: Howdopsychosocialfactorslike personalinnovativeness,habit, and interpersonalinuence predictanindividual'sintentiontocontinuetouse asocialnetworkingsite? Howdo consumerswitchingcosts predictanindividual'sintentionto continuetouseasocialnetworkingsite? Howdo alternativeperceptions predictanindividual'sintentiontocontinuetouseasocialnetworkingsite? ThisresearchusedfactorsfromBhattacherjee'sBhattacherjee,2001IScontinuancetheoryandveadditionalfactorstopredictcontinuanceintention onthesocialnetworkingsiteFacebook.Theveadditionalfactorsinclude: Burnhametal.'sBurnhametal.,2003consumerswitchingcosts,alternativeattractiveness,habit,personalinnovativeness,andinterpersonalinuence. Thenon-moderatedmodelexplainsapproximately76.7%ofthevarianceof continuanceintention;asubstantialamountofthevarianceaccordingtoHair etal..Themoderatedmodelexplained76.9%.Thehabitandper187

PAGE 209

sonalinnovativeness,asmoderatorswerenotstatisticallysignicantandwere droppedfromthemodel.Theresultsexaminedinthediscussionwillfocuson thenon-moderatedmodelexceptwherenoted. Fivehypotheseswereconrmed,satisfaction&perceivedusefulness,asa combinedsecond-orderconstruct,wasthemajorstatisticallysignicantfactor inthemodel =0.3686,followedbycosts =0.2496,alternativeperceptions =-0.2069,habit =0.1642,personalinnovativeness =-0.0589 andinterpersonalinuence =-0.0451.Twohypotheseswererejected;habit andpersonalinnovativeness,asmoderatorswerenotstatisticallysignicant anddidnotsubstantiallyaidintheinterpretationofthefactors. Thefactorspredictingcontinuanceintentionhadtheoreticalsupportfrom pastresearchstudies.Theresultssupportthatsatisfactionandperceived usefulnessareimportantfactorsforSNScontinuance.Surveyrespondents whoweresatisedwithFacebookandwhoperceivedthesiteasusefulhad higherlevelsofcontinuanceintentionandsupportthemajorpredictorsinIS continuanceBhattacherjee,2001.Consumerswitchingcostswasthesecond mostimportantfactorandsupportsthatsurveyrespondentwhoperceived thatleavingFacebookwouldbehavehighcostsweremorelikelytostayand supportsBurnhametal..Burnhametal.denedthreecosts, procedural,nancialandrelationalcostsandthisresearchusedtwocosts, proceduralandrelationalanddidnotincludenancialasthesitesstudieddo nothaveanynancialcosttouse.Therelationalcostshadastrongeecton whetherasurveyrespondentwouldcontinuancetostayonFacebookwhere theproceduralcostswerenotanimportantfactor.Surveyrespondentswho perceivedthatalternativesocialnetworkingsiteswereviablecandidateshave lowerlevelsofcontinuanceintentiononFacebookandsupportBansaletal. .Habithasastrongdirecteectoncontinuanceintentionandindicates 188

PAGE 210

thatsurveyrespondentswhouseFacebookoutofhabitaremorelikelyto continuetousethesiteandsupportsLimayemetal..Habitdidnot haveastatisticallysignicantmoderatingeectregardingitsinteractionwith satisfactiontopredictcontinuanceintentionandwasnotsupported.Social networkingsiteuserswhohavehigherlevelsofpersonalinnovativenessonsocialnetworkingsiteshadhigherlevelsofdiscontinuanceintentionandsupports AgarwalandPrasad.Surveyrespondentswhoweremoreinuencedto jointhenetworkbytheirpersonalcontactshadhigherlevelsofdiscontinuanceandsupportsParthasarathyandBhattacherjee.Fourfactorshave bothstatisticalandpracticalsignicance:satisfaction&perceivedusefulness, consumerswitchingcosts,alternativeperceptionsandhabit.Twofactorshave statisticalsignicancebutnotpracticalsignicance:personalinnovativeness andinterpersonalinuence. Theresearchresultsarewithinthecontextsocialnetworkingsitecontinuanceintention.Theseresultsshowhowthismodelperformsforsocial networkingsiteusers,and,specically,Facebook.Thefactorscomefroma bothmarketingresearchandinformationsystemsresearchandsomeofthe factorshavebeenappliedspecicallytothesocialnetworkingsitecontext. Thefactorsinthisstudyandtheresultsareexpectedtobegeneralizableto othersocialnetworkingsitessuchasTwitter,Tumblr,InstagramandPinterest.Thesefactorsmaybelesshelpfulinstudyingcontinuanceintentionin othercontextsininformationsystemsresearche.g.wordprocessing,general businessapplications.Socialnetworkingsitesarelargelyhedonicandhave ahighdegreeofvoluntarinesssothefactorsinthisresearchwereappliedbecausetheymayhavestrongpredictivepowerincontinuanceintentionwithin thiscontext. 189

PAGE 211

8.1ISContinuanceTheory Theresultssupportthatsatisfaction&perceivedusefulnessareimportant predictorsofcontinuancebehaviorsupportingBhattacherjee'sIScontinuancetheoryandOliver'sexpectation-conrmationtheoryfromconsumerbehaviorresearch.Satisfactionwasastrongerpredictorthanperceived usefulnessforcontinuanceintention.Bhattacherjeesuggeststhatsatisfaction ismorerealisticandstablemeasureovertimecomparedtoperceivedusefulness.Perceivedusefulnessisaperceptualmeasurethatmayincludeboth pre-usageandpost-usagebeliefsandisexpectedtodecreaseovertimeasfamiliaritywiththeservicegrowsBhattacherjee,2001;Oliver,1980. Facebookhasbeenpopularsinceatleast2009whenithad360million users 19 andwouldallowmanyifnotmostsurveyrespondentsampletimeto determinewhetherthesiteisusefulandreachastablelevel.Bhattacherjeesuggeststhatperceivedusefulnessismoreimportantattheacceptancestageandsatisfactionismoreimportantinthecontinuancestage; i.e.perceivedusefulnessisabetterpredictorfortechnologyacceptancethan informationsystemscontinuance.Theresultsfromthisstudyconrmthat satisfactionisamoreimportantfactorinongoingcontinuanceintentionthan perceivedusefulness. TheconrmationfactorfromBhattacherjee'sIScontinuanceand Oliver'sexpectation-conrmationtheorycanbeexaminedthroughits eectsonperceivedusefulnessandsatisfactionanditstotaleectsoncontinuanceintention.Conrmationhadastrongeectonbothperceivedusefulness andsatisfactionandalargereectonsatisfactionthanperceivedusefulness. Theresultsindicatethatmeetingorexceedingsocialnetworkingsiteusers' expectationthroughtheiruseisanimportantfactorincontinuanceintention. 19 http://newsroom.fb.com/company-info/ 190

PAGE 212

Conrmationisthe strongest rst-orderfactoroncontinuanceinthemodel exceedingthatofsatisfactionandcostswhentotaleectsaremeasured.Conrmationoccurswhenusersassesswhethertheirperceivedperformanceexpectationismetthroughuse.Theresultssupporttherolethatconrmation isanimportantfactorinSNScontinuance. Itisoftenusefultoexceeduserexpectationsconrmationwhereusers mayhavelowexpectationsandproductperformanceexceedsthoseexpectationsleadingtohighersatisfactionandcontinuanceBhattacherjee,2001; Oliver,1980.Whenusers'expectationsarenotmetitcanleadtodisconrmation,dissatisfactionandlowercontinuanceintention.Consumer'sexpectationsshouldexceedathresholdlevelofexpectationaboutaserviceinthe acceptancestageotherwisetheymaynotchoosetouseitatall;i.e.ifan expectationofaserviceissolowastodeemaproductnotworthytotrythan itmaynotbeusedatall.Oliverfoundthatconrmationandsatisfactionhavesimilarstrengthsinpredictingconsumerrepurchaseintentions continuance;theseresultsareinlinewiththatnding.Thetotaleectsfor bothconrmationandsatisfactionarestatisticallysignicantconrmation =0.2961;satisfaction =0.2894. 8.2Consumerswitchingcosts Theresultsshowthatconsumerswitchingcostshadastrongroleinpredicting continuanceintentiononFacebook =0.2496.Theswitchingcostsindicate thatFacebookusersmakedecisionsonwhethertostayorleaveFacebookbased onthecostsinvolvedwithleavinghighercostsindicatethattheuseris more likelytocontinuewiththeservice.Thecosts,however,didnotexceedthat ofsatisfactionandperceivedusefulnessasahigher-orderconstruct.Switching costsisaformativemeasurewhichtheoreticallyincludesproceduralcosts, 191

PAGE 213

nancialcostsandrelationalcostsBurnhametal.,2003;thisstudyuses proceduralcostsandrelationalcostsonlyastherearenonancialcoststo useFacebook.Therelationalcostscomposedofpersonalrelationshipcosts andbrandrelationshiphadastrongeectoncontinuanceintentionandthe proceduralcostssetup,economic,evaluationandlearninghadlittletono eectaseitheradirecteectoncostsandasatotaleectoncontinuance intention. Facebookuserswhoperceivedthatswitchingtoanewsitewouldhavea positivecosttotheirrelationshipswhichincludesbothbrandandpersonal hadhighercontinuanceintentiontostayonFacebook.Inthetotaleectsrstorderconstructresults,personalrelationshipcostisthefthhighestfactor, aheadofperceivedusefulnessbutbehindsatisfactionsatisfaction =0.2894, personalrelationshipcost =0.1797,andperceivedusefulness =0.1704. ThepersonalrelationshipitemswereadaptedfromBurnhametal.and measurewhetherpersonalrelationshipswouldsuerifthesurveyrespondent switchedoraddedadierentsocialnetworkingsite.Theitemsmeasurehow comfortablethesurveyrespondentiswithinteractingwiththeirfriendson Facebookcomparedtoothersites,whethertheirfriendsonFacebookmatterto them,iftheyenjoytalkingtotheirfriendsontheplatform,andiftheywould misstheirfriendsiftheyswitchedtoadierentsite.Thesurveyquestions themselvesareagnosticastowhetherswitchingwouldmeantheywouldstop usingFacebook,orsimplyaddanadditionalsocialnetworkingsitetotheirset ofsocialnetworkingsite.Regardlessoftheambiguity,surveyrespondentswho saidthattherearehighcoststotheirpersonalrelationshipsiftheyswitchedto asocialnetworkingsitehadhigherintentionstostayonFacebook.Theresults supportthatpersonalrelationshipcostisanimportantfactorincontinuance intentiononFacebook. 192

PAGE 214

ThebrandrelationshipcostsareanimportantfactorinhowsurveyrespondentsvaluedtherelationalcostsofleavingFacebookandoverallcontinuance intention.Inthetotaleectsrst-orderconstructresults,brandrelationship costistheseventhhighestfactor,behindofalternativeattractivenessalternativeattractiveness =-0.1370,brandrelationship =0.0990.Thebrand constructitemsmeasurewhetherthesurveyrespondentslikethebrand,supportthecompanyandbelievethecompanyhasapositivepublicimage.Survey respondentswhosawthebrandinapositivemannerhadhighercontinuance intentionsuggestingthatFacebookshouldattempttogarnerapositiveimage withitsusercommunity. Theproceduralcostsofeconomic,learning,evaluationandsetuphadno predictivevalueforcontinuanceintention.Itcouldbethatthemajorityof thesurveyrespondentshadatleastoneothersocialnetworkingsite.5% usedatleastoneothersocialnetworkingsiteand19.4%usedthreeormore sotheproceduralcostsmaynotbeburdensometothissample.Thesurvey recruitmentviaapublicannouncementmayhaverecruitedpeoplewhomore familiarwiththeproceduralcostsofaddingasocialnetworkingsiteasthe majorityhaveaddedothersitestotheirsetofsocialnetworkingsitesanddid notfeelthatthecostsweredemanding.Thesurveyrespondentswerenotasked toactuallyuseorshownadierentsocialnetworkingsitesoitmayhavebeen diculttoevaluatethecostsasanintellectualexercise.Bhattacherjeeand Parkusedlearningcostsandsetupcostsinpastresearchandfound theproceduralcoststoberelevant;however,theydemonstratedaspecic applicationtotheirrespondentspriortoconductingasurvey.Thesurvey respondentsintheBhattacherjeeandParkresearchdidnotusethe applicationsotheirpre-acceptanceopinionsmayhavebeenstrongerwithout use. 193

PAGE 215

BothBhattacherjeeandOliverindicatethatpre-acceptance beliefscansignicantlydecreaseovertimewithactualusageofaservice,e.g. perceivedusefulnessasapre-acceptancebelieftendstoregresstothemean throughactualusage.Itcouldbethattheconsumerswitchingcostsarean importantpre-acceptancemeasurebutnotimportantinpost-consumptionusage. Theimpactofconsumerswitchingcostswerelowerthanexpected;there areseveralpossibleexplanationsfortheresultsfoundinthisresearch.Burnhametal.3statedthatapproximately25%ofcontinuanceintentionin servicesarearesultthelevelofconsumersatisfaction;inthisresearchthe IScontinuancemodelbyitselfexplainsamuchlargerproportionofcontinuanceintentionR 2 =.668or66.8%ofthevariance.Thatsatisfactionand perceivedusefulnessexplainsomuchofthevarianceofcontinuanceintention maybewhyconsumerswitchingintentionhaslowerexplanatorypowerthan satisfaction.Anotherpossibleexplanationisthatsocialnetworkingsitesare hedonicinnaturesosatisfactionmaybeaprimarydriverincontinuancevs. consumerswitchingcosts.Consumerswitchingcostsarecomprisedofthree costs,procedural,nancialandrelational,butformostsocialnetworkingsites andFacebookinparticular,therearenonancialcosts.Removingoneofthe threefactorsmayalsohaveanegativeimpactontheexplanatorypowerof themodel.Theproceduralcostshadpracticallynoexplanatorypowerincontinuanceintention,thismaybethatwhensurveyrespondentsconsideredthe setupcostsforsocialnetworkingsitestheywerejustnotasrelevantasother factors.Therangeofproceduralcostsdidrangefromlowtohighandhad areasonabledistribution,butsimplydonotappeartomatterincontinuance intentionasotherfactorsaremoreimportant.Consumerswitchingcosts,by themselves,explainamoderatetosubstantialamountofcontinuanceinten194

PAGE 216

tionR 2 =0.6433or64.3%ofthevariance.Consumerswitchingcostsdo appeartobeastrongpredictorincontinuanceintention,justnotasmuchas satisfactionandperceivedusefulnessareintermsoftheirexplanatorypower. 8.3AlternativePerceptions Alternativeperceptionswerethethirdlargestfactor =-0.2069inSNSContinuancewheresurveyrespondentwhobelievedalternativesocialnetworking siteswereviablecandidateshadhighlowerlevelsofcontinuanceintentionon Facebook.Alternativeperceptionsarecomposedoftwoconstructs,attitude towardswitchingandalternativeattractiveness.Alternativeattractiveness hasalargereectoncontinuanceintention =-0.1370thanattitudetowardswitching =-0.0942andindicatesthatsurveyrespondentsthosewho wereattractedtoalternativesweremorelikelytodiscontinueuseofFacebook comparedtothosewhohaveamorefavorableattitudetowardswitching. SurveyrespondentswereaskedquestionsregardingwhetheradierentsocialnetworkingsitewouldbeanattractivealternativetoFacebookwithout specifyinganalternativesite.Thealternativeattractivenessquestionswere askedintermsofwouldadierentsocialnetworkingsitebemorefair,have policiesthatbenetyoumore,bemoresatisfying,etc.thanFacebookand notdidnotoeraspecicsitetocompareagainstFacebooke.g.Twitteror Pinterest.Thosewhofeltthatattractedtoothersiteshadlowerlevelsof continuanceintentiononFacebook.The attitudetowardswitching measures thedegreetowhichaserviceconsumermaybefavorablydisposedtoswitchingserviceproviders;questionswereaskedintermsofwhetheritwouldbea goodidea,useful,benecial,etc.toswitchtoanewsocialnetworkingsitein thenextsixmonths.Surveyrespondentswhohadfavorableattitudestoward switchingweremorelikelytodiscontinueuseofFacebook. 195

PAGE 217

ThisresearchexaminedhowattitudesaboutspecicalternativesTwitter, Instagram,PinterestandTumblrpredictcontinuanceintentiononFacebook. Theresearchfoundthatholdinganaccountonanalternativesitehadlow explanatorypowerofcontinuanceintentionR 2 =.051,orexplainedapproximately5.1%ofcontinuanceintention-AppendixF.Instagrammaybe consideredacomplementtoFacebookashavinganaccount increased continuanceintentiononFacebook,whereasTwittermaybeconsideredasubstitute toFacebookashavinganaccount decreased continuanceintentiononFacebook.Theattitudesformedbyusingspecicalternativeproductsexplained asmalltomoderateamountofcontinuanceintentiononFacebookforsocial networkingsiteusersthatusedatleastonealternativeproductR 2 =.371,or explainedapproximately37.1%ofcontinuanceintention.Generalalternative perceptionsaboutalternativeserviceswerestrongpredictorsforcontinuance intentionwheretheyarethethirdlargestfactorofcontinuanceintentioninthis study.Theseresultssuggestthatsocialnetworkingsitesmaydevelopageneral alternativeperceptionaboutasiteoutsidethespecicsocialnetworkingsite servicesthatarehelpfulinpredictingcontinuanceintention. 8.4Habit Habitwasthefourthlargestfactor =0.1642inSNSContinuancewhere surveyrespondentswhohavehigherhabitualuseofFacebookaremorelikely tocontinueusingthesite.Limayemetal.,p.705denedhabitfor informationsystemusageas,theextenttowhichpeopletendtoperform behaviorsuseISautomaticallybecauseoflearning.Surveyrespondentswho usethesiteaspartofaroutinemaycontinuetousethesiteaspartoftheir dailycourse.Habitmayhelpexplainwhyteenswhoshowlessenthusiasm forthesocialnetworkingsiteFacebookcontinuetousethesiteasitisan 196

PAGE 218

importantpartofoverallteenagesocializingMaddenetal.,2013.Teens statedmanyreasonsfordislikingtheFacebookincludinganincreasedadult presence,excessivesharingofinformationandtheamountofdramabut continuetokeeptheiraccounts.Facebookmaybeaplatformwherepeopleare lessexcitedaboutusingitorhavelowerenjoymentthroughusebutcontinueto useitastheplatformmaybeimportantmeansofcommunication.Facebook usersmaybeincorporatingthesiteintotheirlivesbutmaybelessexcited aboutusingthesite. TheroleofhabitmayincreaseovertimeasanimportantfactorinSNS continuanceassatisfactionmaydecreaseasithistoricallydoesandregresses tothemeanBhattacherjeeandPremkumar,2004.Facebookwasstarted in2004andopentoallusersover13yearsoldin2006somanyofthesites' usershavehadconsiderableexperiencewiththesite. 20 Asusersbecomemore familiarwithaproducttheycontinuallysetandresettheirexpectationsand conrmtheirexpectationsovertimeBhattacherjeeandPremkumar,2004. Peopletendtohavecomplexviewsaboutnewcommunicationtechnologies astheyareadopted;someareskepticalofthetechnologywhileothersembraceitBaym,2010.Newcommunicationtechnologyisoftenviewedas providinggreatbenets;soonaftertheradiowasinventeditwasthoughthat everyhomewouldbeanextensionofCarnegieHallorHarvardUniversity andwhentelevisionwasinventeditwasexpectedtobringworldpeaceand socialharmonyMosco,2004.Somebelievethatthenewtechnologydevalues face-to-facecommunicationandexpressconcernthatcommunicationwillgrow increasinglyshallow.Otherswillembracethetechnologyandtoutitsbenets tocreatestrongerrelationshipsandnewopportunitiestointeractwithnew people.Asthetechnologyisadoptedbygreaternumbers,peopleseethepros 20 http://newsroom.fb.com/company-info/ 197

PAGE 219

andconsofnewmediaandacceptitforwhatitprovidesBaym.Habit maybeafactorinwhy,assatisfactionandexpectationsbecomeincreasingly morerealisticpeoplecontinuetousethesite.Facebookusemaybecomepart ofpeople'sdailyroutineseveniftheyarenottheydonotndthesiteparticularlyenjoyableitmaybeachannelinwhichtheycommunicatewithmembers oftheirsocialnetwork. Habitdidnotmakeastatisticallysignicantdierenceintermsofmoderatingtherelationshipbetweensatisfactionandcontinuanceintentionashypothesized.IntheindividualfactormodelAppendixCwhereIScontinuance andhabitaretheonlyfactorspresentitwasastatisticallysignicantrelationshipwherethosewithlowersatisfactionlevelswouldcontinuetouseFacebook iftheyhadhigherlevelsofhabit.Themoderatingt-statisticforhabitxsatisfaction t =4.3531indicatesthatitisstatisticallysignicantalthoughthe eectsize f 2 is.0139whichdoesnotmeetthethresholdfortobeconsidered ashavinganeectthresholdforsmalleects f 2 is.02basedon Chinetal. 8.5PersonalInnovativeness Personalinnovativenesswasthefthlargestfactor =-0.0589inSNSContinuancewheresurveyrespondentswhohadhigherlevelsofpersonalinnovativenessweremorelikelytodiscontinueuseofFacebook.AgarwalandPrasad ,p.206denedpersonalinnovativenessas,thewillingnessofanindividualtotryoutanynewinformationtechnology.Personalinnovativeness explicitlydeneshowusersmayadopttechnologybasedonpsychometriccharacteristics.Individualswhohavehigherlevelsofpersonalinnovativenessmay adoptinnovationsearlierthanothersandmayactaschangeagentsandopinion leaderstofurtherdiuseanewtechnology.Personalinnovativenessissimi198

PAGE 220

lartoRogers'svecategoriesofadopters: innovators,earlyadopters, earlymajority,latemajority and laggards howevertheseadoptercategories aredenedby timeofadoption andnotpsychometricattributesAgarwal andPrasad,1998;Mahajanetal.,1990.ThosewhoadoptFacebookearlier mayhavehigherpersonalinnovativenessonsocialnetworkingsitesandmay alsobemovingtonewerormoreinnovativesites. Personalinnovativenessdidnotmakeastatisticallysignicantdierence intermsofmoderatingtherelationshipbetweensatisfactionandcontinuance intentionashypothesized.IntheindividualfactormodelAppendixDwhere IScontinuanceandpersonalinnovativenessaretheonlyfactorspresentitwas notastatisticallysignicantrelationshipwherethosewithlowersatisfaction levelswouldcontinuetouseFacebookiftheyhadhigherlevelsofpersonalinnovativeness.Themoderatingt-statisticforpersonalinnovativenessindicates thatitisnotstatisticallysignicant t =1.4588;theeectsize f 2 is0.0028 whichdoesnotmeetthethresholdfortobeconsideredashavinganeect thresholdforsmalleects f 2 is.0200basedon Chinetal. AgarwalandPrasadrecommendedtoresearcherstousepersonalinnovativenessasacontrolvariableinfuturestudieswhenexaminingindividual usageintentionandnotspecicallycontinuance.Studiesregardingtheroleof personalinnovativenesshavebeeninconsistentwherepersonalinnovativeness hasbeenshowntoavaluableantecedenttosomefactorssuchascognitiveabsorption,perceivedeaseofuseandperceivedusefulnessbutoftennothavinga strongrelationshiptousageintentionLuetal.,2005.ThatcherandPerrew suggestthatpersonalinnovativenessshouldbeusedasanantecedentto both beliefs e.g.ease-of-useand attitudeformation e.g.acceptanceandcontinuance.Arguably,personalinnovativenessmayhaveastrongroletoplay on acceptance moresothan continuance asithasstrongertheoreticalsup199

PAGE 221

portfromRogers diusionofinnovations theory.AhujaandThatcher refertopersonalinnovativenessasatraitand,assuchshouldnotvary acrossdierentenvironments,againsuggestingthatpersonalinnovativeness maybemoreappropriateasacontrolvariable.boydandEllisonbdiscussseveralcaseswhereearlyadoptershaveleftsocialnetworkingsiteslike SixDegreesandFriendsterfornewsitessuchasMySpaceafterencountering dicultywithearlierinnovationsandsuggeststhatbeinganearlyadopters hasanimpactonpost-acceptancecontinuancebehavior.Inthecontextof thisresearch,itmaybemoreappropriatetoviewpersonalinnovativenessasa controlvariable,andoneinwhichthereisastatisticallysignicantrelationship butdoesnotrisetothelevelofpracticalsignicance. 8.6InterpersonalInuence Interpersonalinuenceisthenalandsixthfactor =-0.0451inSNSContinuancewheresurveyrespondentswhohadhigherlevelsofinterpersonalinuenceweremorelikelytodiscontinueuseofFacebook.Examiningtherole ofinterpersonalinuenceallowstheresearchtoextendbeyondtheindividual tosystemsofindividualsRogers,1976.Surveyrespondentswhocamewere inuencedbytheironlineandoinesocialnetworkfriends,familymembers, classmatesandothershadhigherlevelsofdiscontinuancecomparedtothose whowerelessinuencedbytheirpersonalrelationshipstouseFacebook.The researchsupportsParthasarathyandBhattacherjeeregardingtherole ofinterpersonalinuenceonservicecontinuation.ParthasarathyandBhattacherjeefoundthatuserswhostoppedusingasystemafteracceptance weremoreinuencedbytheirinterpersonalrelationshipsthanthosewhocontinuedtousethesystem.ParthasarathyandBhattacherjeealsofound thatdiscontinuersofonlineserversarelessinuencedbyexternalsourcese.g. 200

PAGE 222

mediaduringtheinitialacceptancethanthosewhocontinuedtheusingthe service.Kimexaminedtheroleofinterpersonalinuenceoninformationsystemcontinuanceforsocialnetworkingsites.Kimfoundthat whenmembersofauser'ssocialnetworkbelievedthatusingasocialnetworkingsitewasagoodideauserswerepositivelyinuencedtocontinueusingthe site. Theseresultssuggestthatuserswhojoinedbecauseoftheirpersonalrelationshipshavehigherlevelsofdiscontinuanceandaremorelikelytostopusing theFacebook.Theresultsconrmthattheroleinterpersonalrelationships atthetimeofadoptioninuencehowpeoplemakediscontinuancedecisions. ParthasarathyandBhattacherjeetheorizethatindividualswhoadopt basedontheirinterpersonalrelationstendtobecomemoredisenchantedwith servicescomparedtothosewhoadoptbasedonmasscommunication.The sameeectsmaybeseenonFacebookcontinuance,thosewhowereinuenced bytheirinterpersonalrelationshipsmaybedisenchantedwiththesiteandhave higherdiscontinuanceintention.Facebookisasocialcommunicationtooland memberswhojointhesiteareaskedtomakeconnectionswithpeoplethey know.SomeFacebookusersmayjoinbecausetheyareaskedbytheirexisting friendstojoininterpersonalrelationships,theymayusethesiteanddecide itisnotwhattheyexpectedorthattheserviceisnotusefulanddecideto discontinue.Thefactorallowsagreaterunderstandingofhowfactorsoutside thespecicproductandserviceimpacttheconsumers'intentiontocontinue use. 8.7DemographicVariables Threedemographicvariablesweretestedascovariatesforcontinuanceintentioninthisresearch;age,genderandeducationwereidentiedaspotential 201

PAGE 223

inuencesinpredictingcontinuanceintention.Agewastheonlystatisticallysignicantrelationshipinthenon-moderatedmodelwitha =0 : 0613 t =4.3408whereolderFacebookusersshowhigherlevelsofcontinuancethan youngerusers.ThisndingsupportsMaddenndingswhereyounger usersweremorelikelytoindicatetheyweregoingtospendlesstimeonFacebookthatolderusers.Maddenfoundthat38%ofthosebetween18 and29intendedtospendlesstimeonFacebookcomparedtothoseinthe30 to49yearagegroup%and50+%.Genderwasnotstatisticallysignicant = 0.0167, t =1.2233anddoessupportthatgenderisastatistically signicantfactor.MaddenfoundgenderdierencesinhowgenderisassociatedwithFacebookuseamonginternetusers.Maddenfoundthat 62%ofmaleinternetusersuseFacebookat72%offemaleinternetusersused Facebook.Thisresearchshowsthatthesedierencesmaybebetterexplained byotherfactorsregardingcontinuanceintentionbasedonotherfactorssatisfaction,consumerswitchingcosts,etc.thangenderalone.Educationwas notfoundtobeastatisticallysignicantfactor = 0.021, t =1.4686;those withhighereducationdidnothaveastatisticallysignicantrelationshipwith continuanceintention.Madden3foundsomedierencesineducational attainmentandtheirusageofFacebookwhere60%ofinternetuserswhohad lessthanahighschooleducationtohighschoolgraduateusedFacebook,and 73%ofofinternetuserswhohadsomecollegeexperienceusedFacebook,and 68%ofinternetuserswhowerecollegegraduatesandbeyondusedFacebook. Thisresearchshowsthatthesedierencesmaybebetterexplainedbyother factorsregardingcontinuanceintentionthateducationalattainment.Overall, thedemographicvariableswereusedtoadjusttheresultsofthemainfactors underconsiderationandnoneofthedemographicsvariablesrisetothelevel ofpracticalsignicance. 202

PAGE 224

8.8ImplicationsforResearchers Therewereveadditionalfactorsexaminedintodeterminetherelationship withIScontinuancetheoryandcontinuanceintention.Allveofthefactors werestatisticallysignicantpredictors;however,thepracticalsignicanceof thecoecientsshouldbeexaminedaswellasthestatisticalsignicance.Statisticalsignicanceinthisresearchmaybemoreeasilyfoundbecausethereis alargesamplesize N =1302.Threefactorsappeartohavebothstatistical andpracticalsignicanceinadditiontosatisfaction&perceivedusefulness: consumerswitchingcosts,alternativeperceptionsandhabit.Twofactorshave statisticalsignicancebutlackpracticalsignicance:personalinnovativeness andinterpersonalinuence. Theresultsndmanyfactorsthatarehelpfulinunderstandcontinuance intention.Satisfaction&perceivedusefulnessandconsumerswitchingcosts, alternativeperceptionsandhabithaveastrongrelationshipwithcontinuance intentionwhereaspersonalinnovativenessandinterpersonalinuencemaybe small.Theeectsofsatisfaction&perceivedusefulnesscouldbemisleadingintheparsimoniousmodelofIScontinuance.TheISContinuancemodel explained66.8%ofthevarianceofcontinuanceintentionwhereasaddingve additionalfactorsraisedtheamountofexplainedvarianceto76.7%.Theeect ofthefactorsincreasedtheexplainedvarianceofthemodelfrom moderateto substantial to substantial ;theeectsofthenewfactorsarealsoconsideredto belarge.TheIScontinuancemodelhadhighloadingfactorsforsatisfaction totaleects =0.6315andperceivedusefulnesstotaleects =0.3708. TheSNScontinuancemodelnon-moderatedreducedtheinuenceofsatisfaction&perceivedasacombinedsecond-orderfactorto =0.3686,when otherfactorsareabletoexerttheirinuence.Satisfaction'sroleincontinuanceintentionremainsrelativelyhighintermsofitsinuenceoncontinuance 203

PAGE 225

intentiontotaleects =0.2894howeverperceivedusefulnessdropsto =0.1704totaleectsbelowpersonalrelationshiploss,alternativeperceptionsandconsumerswitchingcosts.Thedirecteectsofperceivedusefulness oncontinuanceintentionshowninthemoderatedmodelaremuchlower =0.0432,lackpracticalsignicanceandisnotstatisticallysignicant t = 1.8277asadirecteect. Bhattacherjeeincludesperceivedusefulnessbecausethefactoris theonlybeliefthatisdemonstratedtoconsistentlyinuenceuserintention acrosstemporalstagesofISuse,Bhattacherjee,p.355.Perceived usefulnessisaninuentialfactorinTAM-basedstudiesBhattacherjee,2001, butitseectsincontinuanceintentionmaybelessinuentialwhenotherfactorsareconsideredsuchastheonesinthisstudy.ISContinuancetheoryis basedonOliver'sexpectation-conrmationtheorywhereperceivedusefulnessisnotpresent.Thejusticationforreplacing expectation with perceived usefulness isthatperceivedusefulnessmaybemorespecictothecontextof informationsystemsandexpectationsmayhaveamorebroadcontext.Oliver's expectation-conrmationtheorydoesnotincludeadirectrelationship between expectation and continuanceintention becausethemodelproposes thatcontinuanceintentionisprimarilypredictedbysatisfaction.Bhattacherjee,includesadirecttheoreticalrelationshipbetweenperceivedusefulnessandcontinuanceintentionbecauseofitsinclusioninTAMDavis. Bhattacherjeestates: Althoughtheusefulness-intentionassociationwasoriginallyderivedinanacceptancecontext,itislikelytoholdtrueincontinuancecontexts,becausehumantendenciesforsubconsciously pursuinginstrumentalbehaviorsorstrivingforrewardsareindependentofthetimingorstageofsuchbehaviors. 204

PAGE 226

Whileaddingadirectlinkbetweenperceivedusefulnessandcontinuance intentionmaybejustiedonpracticalgroundsinthatcontinuanceresembles acceptance,itstheoreticalinclusionissuspectandthefactormayalsobeless inuentialthanexpectedwhenadditionalfactorsareaddedtothemodel. Lookingbeyondsatisfactionandperceivedusefulnesswashelpfulwhere threefactorshadbothstatisticalandpracticalimplications-consumerswitchingcosts,alternativeperceptionsandhabit.Consumerswitchingcostsshowed thatboththerelationshipwiththebrandandpersonalrelationshipswiththose inthenetworkhadlargeimpactsoncontinuanceintention.Alternativeperceptionsshowthatcontinuanceintentionisalsoimpactedbyfactorsbeyond theserviceinthestudyrelativeadvantageofothersites.Researchersshould extendbeyondtheparticularserviceunderinvestigationtodetermineifother sitesareimpactingcontinuancedecisions.Lastly,habithasaroleforcontinuanceintentionwhereusersmaycontinueusingasitebecauseitispart ofaroutine.Thesefactorshadlargeimpactoncontinuanceintentionand maymoreappropriatelydenehowsatisfactionandperceivedusefulnesspredictcontinuanceintentioninthattheneteectofsatisfactionandperceived usefulnessissignicantlylowerwiththesefactorsinplace. Theresearchusedmultipleanalysestodeterminehowspecicalternatives socialnetworkingsiteservicesandgeneralalternativeperceptionspredictcontinuanceintention.Thereweresmalldierencesinwhetherhavinganaccount withanalternativesitewasassociatedwithhigherorlowercontinuanceintentiononFacebook.ThespecicattitudesformedfromalternativesTwitter, Instagram,PinterestandTumblrwerehelpfulinpredictingcontinuanceintention,butthegeneralalternativeperceptionsweremuchstrongerregarding thepredictivepowerofcontinuanceintention.Theresultssuggestthatgeneral alternativeperceptionsaremorehelpfulindeterminingcontinuanceintention 205

PAGE 227

comparedtoexamininganindividualalternativeperceptione.g.isTwittera viablealternativetoFacebook. 8.9ImplicationsforPractice Thereareseveralimplicationsforpracticeasaresultofthisresearch.Satisfaction&perceivedusefulnesscontinuetobeanimportantfactorincontinuanceintention.Companieswhosegoalsincludecustomerretentionshould attempttoincreaseusersatisfactionandperceivedusefulness.Usefulnessmay beanimportantpredictorforoverallsatisfactionbutwhenotherfactorsare includedinamodelthenitsdirecteectsaremodest.Conrmationofexpectationsthroughproductuseisthemostimportantrst-orderfactorinthe model.ConrmationsaresetandresetthroughuseovertimeBhattacherjee andPremkumar,2004soitmaybeimportanttoaddnewfeaturestohedonic socialnetworkingsitesthatuserswilluseandenjoy. Thereweretwomajorfactorsregardingconsumerswitchingcoststhateffectcontinuanceintention.Facebookshouldfocusonthepersonalbondsthat existonthesiteassurveyrespondentswhobelievetheirpersonalrelationshipswouldbedamagedbyswitchingtootherservices.Facebookfocuseson existingpersonalrelationshipsandlatenttiessothatitsmemberscanconnect toothers.Totheextentthattheserelationshipsareimportanttoitsusers theymaycontinuetousethesiteathigherratesthanothers.Brandrelationshipswerealsoanimportantcomponenttoconsumerswitchingcostswhere surveyrespondentswhovaluedtheFacebookbrandhadstrongercontinuance intention.Facebookcouldtrytobuilditsreputationwithinitsservicesothat peoplehavefavorableattitudestowardthebrand.OneareainwhichFacebookappearstohaveissuesregardingitsbrandiswithitsprivacypolicies 21 21 FacebookSEC10-QquarterlyReport,April25,2014 http://investor.fb.com/ secfiling.cfm?filingID=1326801-14-23&CIK=1326801 206

PAGE 228

boyd,2008;boydandHargittai,2010.boydsuggeststhatchanges toFacebook'sprivacypoliciesinthepastleaditsuserstofeelthattheyhad increasedexposureandthatthecompanyallowedaninvasionofprivacyand thesepolicychangesmayleaduserstotrustFacebookless.boydandHargittai,2010,WebstatethatnewscoverageregardingFacebook'sprivacypolicies relayaconsistentmessage,donottrustFacebook.Advocatessuggestedthat thatMay31st2010bedeclaredQuitFacebookDayhoweverfewusersactuallydidquitboydandHargittai,2010andthesitecontinuestogainusers. 22 Facebook's10-QApril25,2014states,Wewillalsocontinuetoexperience, media,legislative,orregulatoryscrutinyofourdecisionsregardinguserprivacyorotherissues,whichmayadverselyaectourreputationandbrand. Competingservicemaywishtocontinuetohighlighttheperceiveddiculty insettingFacebook'sprivacysettingsintheintendedmanner. Facebookwillcontinuetofacealternativeproductsinthemarketplace. Someproductsmaybecomplementarye.g.Instagramwhileothersmaybe moreappropriatelyviewedassubstitutesTwitter.Instagramusersshowan increaseofcontinuanceintentiononFacebook;itmaybethatstrongintegrationbetweenthetwoproductsishelpfultobothproducts.Theproductsmay becomplementaryinthatthestatusupdatesinFacebookaretextbasedand Instagramaddsavisualcomponent.UserswhohaveTwitteraccounts,onthe otherhand,havelowerintentionstocontinueusingFacebook.Itappearsthat TwittermaybebehavingmorelikeasubstituteproductwhereuseofTwitter leadstolowerintentionstocontinueusingFacebook.Theusersmayusethe sitesdierently;cross-postingoftextispossiblebetweenthetwositescan beconguredbutmaynotbeparticularlyhelpfultotheusers.BothTwitterupdatesandFacebookupdatesarelargelytextbased,withTwitteronly 22 http://newsroom.fb.com/company-info/ 207

PAGE 229

allowingfortweetsof140characterscomparedtoFacebook'smoregenerous textallocation.Inthiscase,havingtextfromonesitecross-postedtoanother sitemaynotbeconsideredparticularlyuseful.PinterestandTumblraccount holdershadnostatisticallysignicantdierenceincontinuanceintentionon Facebook.Themeanvaluesforaccountholdersindicatethatthemajorityof thesurveyrespondentsplannedtocontinueusingFacebook;buttheresults showdierencesinthestrengthofthatrelationship.Facebookmaycontinue itsmergersandacquisitionpoliciestocontinuetoaidFacebook.Facebook purchasedWhatsAppforupto$19billiondollarsin2014toincreaseitsmobilecommunicationpresence. 23 Facebookhasalsodevelopedadditionalmobile productsthatexistoutsideitsFacebookdomaine.g.Paper 24 .Facebookmay needtodevelopandacquireproductsthatkeeptheiruserbaseloyaltothe brand. Thecompanycanpromoteitsroleasacommunicationmediumtohelp buildahabitofuseandincreasecontinuanceintention.Facebookusersalreadyappeartohaveahighlevelofhabitofusehoweverovertimehabitmay increasinglybecomeimportantassatisfactionmaydecreaseovertimeasusers havemorerealisticexpectations.TeensmayhaveadecreasedinterestinFacebookbuttheycontinuetousethesiteandretaintheiraccountsMadden, 2013;Maddenetal.,2013.SowhiledecreasedusemaybeaconcernforFacebookthelargerconcernmaybeifitsuserbasediscontinuesandremovetheir accountsaltogether.Garciaetal.suggeststhattheusercommunity canhaveimportantimpactsonasitesabilitytoremainanongoingconcern. Iftheuserbasequitssitesmaygointorapiddecline.Promotingahabituation ofusecanhelpFacebookremainaviablecompany. 23 http://www.businessweek.com/articles/2014-02-19/facebook-acquireswhatsapp-for-19-billion 24 http://newsroom.fb.com/news/2014/01/introducing-paper-stories-fromfacebook-2/ 208

PAGE 230

Therearetwofactorswhichexhibitstatisticalsignicancebutnotpractical signicance:personalinnovativenessandinterpersonalinuence.Thereisnot signicantjustication,basedonthisresearch,tostartmajorinitiativesto keepthosewhoshowmorepersonalinnovativeness.Personalinnovativenessis associatedwithhighervarietyseekingbehavior,peoplewithhigherpersonal innovativenesstendtobeeasiertoattracttoaservice and moredicult toretain.Thestrengthoftherelationshipbetweenpersonalinnovativenessis notstrongandthereforeotherinitiativeswouldtakehigherprecedence.Those whocametoFacebookduetotheirpersonalrelationshipleaveathigherlevels thanthosewhosaidtheycameforotherreasons.Facebookmaywantto havealargermassmediapresencetoattractandretainusersduetothese ndings;however,anyeectislikelytobesmall.Facebookhasaadoption ratethatisquitehighamongteenagers%in2012Madden,2013and 67%ofAmericanadultsareFacebookusersRaineetal.,2013.Raineetal. foundthatmanyFacebookusersplantospendlesstimeonthesitein thecomingyearwhere38%of18-29yearolds,26%of30-49,and17%of50 +saidtheyplannedtospendlesstimeonFacebook.Theresultsfromthis researchindicatethatabout8%ofFacebookusershaveleftand15%planto stopusingFacebooksoon.Facebookdoeshavecauseforconcernregarding lowerusageratesbutthesiteremainsstickyformanyusers. 209

PAGE 231

9Limitations Participantsinthepresentstudywerenotrecruitedrandomly.Respondents wererecruitedviaapublicannouncementandwillnotberepresentativeof allFacebookusers.Therecruitmentmethodmayalsohaveledtotheoverrepresentationofthosewhomoreinterestinthetopicingeneral.Thereare threedatasetsinthisresearch;apublicdatasetrecruitedbyapublicannouncement,astudentdatasetandanOracledataset.Thedatasetsarelargelysimilar toeachother;however,therearedierencesaswell.Theconstructmeasurementsseemtoberelativelysimilar,butthereweresomedierencesinthe pathcoecients.TheOracledatasethadamuchlargerpathcoecients fortheconsumerswitchingcostmodelthanthepublicandstudentdataset indicatingthattheseuserscontinuanceintentionisbasedmoreonthecosts associatedwithmovingawayfromasocialnetworkingsitethansatisfaction withthesite.Thisdierencemaybeexploredinthefuturewithpost-hoc nitemixtureanalysisFIMIXMcLachlanandPeel. TheresearchinvestigatesvemajorfactorsinadditiontoIScontinuance theorytopredictcontinuanceintentiononFacebookbuttherearemaybeadditionalfactorsthatarerelevanttoswitchingintention.Thisresearchaskedsurveyrespondentsaboutfourdierentproceduralcostseconomiccosts,learning costs,setupcostsandevaluationcostswhichmaybediculttoanswerwithoutevaluatingaspeciccompetitor.BhattacherjeeandParkexamined switchingintentionsafterdemonstratingonespecicapplicationtocapture subjects'perceptionsofnewservicesandhadsomeresultsthatdierfrom thesendingslearningcostsandsetupcostsweresignicantfactors.Users mayswitchsocialnetworkingsitesdespitebeinggenerallysatisedwiththe siteandthecostsifotherconnectionsintheirnetworkareabandoningthesite 210

PAGE 232

Garciaetal.,2013.Popularityofasitemaybeephemeral,oncepopularsites likeFriendster,MySpace,etc.areeithernolongerexistorindeclineGarcia etal.,2013.Factorssuchasthefrequencyofserviceusageandoverallusage werenotstudiedinthisresearchbuthavebeenshowninpastresearchto haveaneectoncontinuanceintentionKeaveneyandParthasarathy,2001. Theroleofprivacyhasnotbeenstudiedinthisresearchbutmayalsohave aneectoncontinuanceintention.Theroleofprivacymaybecapturedby auser'sperceptionofthebrandfromBurnhametal..Futurestudies maywanttoaddfactorsnotstudiedinthisresearchtobetterunderstandthe predictivepowerofthosefactors. Thisresearchexaminedanindividual'sintentiontocontinuetouseasite, butdoesnotexaminetheissueattheorganizationallevelorusemulti-level analysis.Garciaetal.suggeststhattheroleoftheindividualhas organizationalimpactsandcanhelpdeterminewhetherasiteisgrowing,in decline,orsteady. Thesurveyquestionsthemselvesareagnosticastowhether switching means thatauserwouldstopusingFacebook,orsimplyaddanadditionalsocialnetworkingsitetotheirsetofsocialnetworkingsite.Thesurveyquestioncould moreclearlydenewhether switching toanewsitemeansremovingtheiraccountsfromFacebookoraddingaccountstoothersocialnetworkingsitesand retainingtheirFacebookaccount.Thatis,ausermaysaytheywouldcontinuetouseFacebookif switching toPinterestaccountmeanttheyaddeda PinterestaccountandretainedtheirFacebookaccount;butanotherinterpretationcouldmeanthat switching toaPinterestaccountmeanstheywilladd PinterestandremovetheirFacebookaccount. Thepracticalsignicanceofthecoecientsshouldbeexaminedaswell asthestatisticalsignicance.Statisticalsignicanceinthisresearchmaybe 211

PAGE 233

moreeasilyfoundbecausethereisalargesamplesize N =1302.Practical signicanceisrelatedtothemagnitudeofthepathcoecients.Age,for example,inthisresearchisstatisticallysignicant t =3.826butthepath coecientissosmall =0.055tonotoermuchpracticalsignicance. Hairetal.statesthatfactorloadings .5aregenerallyacceptableand .3to .4areminimallyacceptable.Manyofthefactorloadingshaveboth statisticalandpracticalsignicance,butsomestatisticallysignicantfactors donotmeetthepracticalthreshold. 212

PAGE 234

10Conclusion Understandingthedynamicsofthesocialnetworkingsitesprovidesbenetsto researchersandthecommunity.Theinformationsystemcontinuancemodel hasbeenenhancedtoincludevefactorswhichhelpexplainusercontinuance decisions.Consumerswitchingcostshaveastrongdirecteectoncontinuance intentionasfoundinBurnhametal..Consumerswitchingcostsare comprisedofproceduralandrelationalcostsofwhichrelationalcostswereimportantfactorandproceduralcostswerenot.Relationalcostsarethelosses thatoccurfromleavingabrandandleavingtherelationshipsonsocialnetworkingsitetogotoanother.Proceduralcostsincludethetimeittakesto evaluateanewservice,setuptheserviceandlearnhowtousetheservice. Socialnetworkingsitesarebeingintroducedinadynamicenvironmentwhere newgeneralpurposesocialnetworkingsitesareintroducede.g.Friendster, MySpace,Facebook,Twitter,GooglePlusand alternativeperceptions have astrongaectoncontinuanceintention.Facebookuserswhondalternativeservicestobeaviablealternativeshowhigherdiscontinuanceintention onFacebook.Theroleof habit showsthatuserswhouseasocialnetworking sitewithhighratesofhabitcontinuetousethesiteathigherlevels.The moderatingeectofhabitonsatisfactionforcontinuanceintentionwasnot supported.Personalinnovativenessmaybeconsideredacontrolvariablein thisresearch;therewasasmalldirecteectwherethosewithhigherpersonal innovativenesshadhigherlevelsofdiscontinuance.Theroleof personalinnovativeness didnothavemoderatingeectonSNSContinuancewhereusers withhighlevelsofpersonalinnovativenessdidnotmoderatetherelationship betweensatisfactionandcontinuanceintention. Interpersonalinuence had asmalldirecteectoncontinuanceintentionwherethosewhojoinedthesite 213

PAGE 235

duetotheirpersonalrelationshipsweremorelikelytodiscontinueuseofthe site. Thegoalofthemodelsinthisresearchwastondfactorsthatexplained moreofthevarianceofcontinuanceintentiononsocialnetworkingsitesby addingveadditionalfactorstoIScontinuanceBhattacherjee,2001.The non-moderatedmodelexplainsapproximately76.7%ofthevarianceofcontinuanceintention,asubstantialamountaccordingtoHairetal.andis consideredalargeincreaseovertheIScontinuancemodellevelofexplained varianceof.8%.Theresearchexaminedveconstructs,personalinnovativeness,habit,alternativeperceptions,interpersonalinuence,andproceduralandrelationalswitchingcoststopredictcontinuanceintentions.Five ofsevenhypothesesweresupportedbythisresearch-allmajordirecteects weresupportedandtwomoderatingeectswerenotsupported. 214

PAGE 236

REFERENCES Adamson,I.andShine,J..Extendingthenewtechnologyacceptance modeltomeasuretheenduserinformationsystemssatisfactioninamandatoryenvironment:Abank'streasury. Technologyanalysis&strategicmanagement ,15:441. Agarwal,R.andPrasad,J..Aconceptualandoperationaldenition ofpersonalinnovativenessinthedomainofinformationtechnology. InformationSystemsResearch ,9:pp.204. Ahuja,M.K.andThatcher,J.B..Movingbeyondintentionsand towardthetheoryoftrying:Eectsofworkenvironmentandgenderon post-adoptioninformationtechnologyuse. MISQuarterly ,29:pp.427 459. Bansal,H.S.andTaylor,S.F..Theserviceproviderswitchingmodel spsm:Amodelofconsumerswitchingbehaviorintheservicesindustry. JournalofServiceResearch ,2:200. Bansal,H.S.,Taylor,S.F.,andJames,Y.S..Migratingtonewservice providers:Towardaunifyingframeworkofconsumers'switchingbehaviors. JournaloftheAcademyofMarketingScience ,33:96. Baroudi,J.J.,Olson,M.H.,andIves,B..Anempiricalstudyofthe impactofuserinvolvementonsystemusageandinformationsatisfaction. Commun.ACM ,29:232. Baym,N.K.0. PersonalConnectionsintheDigitalAge .DigitalMedia andSociety.Polity. 215

PAGE 237

Becker,J.-M.,Klein,K.,andWetzels,M..Hierarchicallatentvariable modelsinpls-sem:Guidelinesforusingreective-formativetypemodels. LongRangePlanning ,45:359. Bhattacherjee,A..Understandinginformationsystemscontinuance: Anexpectation-conrmationmodel. MISQuarterly ,25:351. Bhattacherjee,A.. SocialScienceResearch:Principles,Methods,and Practices .AnolBhattacherjee,2ndedition. Bhattacherjee,A.andPark,S.C..Whyend-usersmovetothecloud: amigration-theoreticanalysis. EuropeanJournalofInformationSystems advanceonlinepublication:1. Bhattacherjee,A.andPremkumar,G..Understandingchangesinbelief andattitudetowardinformationtechnologyusage:Atheoreticalmodeland longitudinaltest. MISQuarterly ,28:pp.229. Bollen,K.A.. StructuralEquationswithLatentVariables .JohnWiley &Sons. boyd,d..Facebook'sprivacytrainwreck:Exposure,invasion,andsocial convergence. Convergence:TheInternationalJournalofResearchintoNew MediaTechnologies ,14:13. boyd,d.andHargittai,E..Facebookprivacysettings:Whocares? FirstMonday ,15. boyd,D.M..Friends,friendstersandtop8:Writingcommunityinto beingonsocialnetworksites. FirstMonday ,11:Online. boyd,D.M.andEllison,N.B.a.Socialnetworksites:Denition, 216

PAGE 238

historyandscholarship. JournalofComputer-MediatedCommunication 13:1. boyd,d.m.andEllison,N.B.b.Socialnetworksites:Denition, history,andscholarship. JournalofComputer-MediatedCommunication 13:210. Brandes,U.,Freeman,L.C.,andWagner,D..Socialnetworks. InTamassia,R.,editor, HandbookofGraphDrawingandVisualization DiscreteMathematicsandItsApplications ,chapter26.Chapmanand Hall/CRC. Burnham,T.A.,Frels,J.K.,andMahajan,V..Consumerswitching costs:Atypology,antecedents,andconsequences. JournaloftheAcademy ofMarketingScience ,31:109. Cao,H.,Jiang,J.,Oh,L.-B.,Li,H.,Liao,X.,andChen,Z..Amaslow's hierarchyofneedsanalysisofsocialnetworkingservicescontinuance. Journal ofServiceManagement ,24:pp.170. Chin,W.W.. ModernMethodsforBusinessResearch ,chapterThe partialleastsquaresapproachtostructuralequationmodelling,pages295 336.LawrenceErlbaumAssociates,Jahway,NJ. Chin,W.W.,Marcolin,B.L.,andNewsted,P.R..Apartialleast squareslatentvariablemodelingapproachformeasuringinteractioneects: Resultsfromamontecarlosimulationstudyandanelectronic-mailemotion/adoptionstudy. InformationSystemsResearch ,14:189. Coleman,J.S..Socialcapitalinthecreationofhumancapital. AmericanJournalofSociology ,94:s95s120. 217

PAGE 239

Davis,F.D..Perceivedusefulness,perceivedeaseofuse,anduser acceptanceofinformationtechnology. MISQuarterly ,13:319. Davis,F.D.,Bagozzi,R.P.,andWarshaw,P.R..Useracceptanceof computertechnology:Acomparisonoftwotheoreticalmodels. Management Science ,35:pp.982. DeLone,W.H.andMcLean,E.R..Informationsystemssuccess:The questforthedependentvariable. InformationSystemsResearch ,3:60. DeLone,W.H.andMcLean,E.R..Thedeloneandmcleanmodelof informationsystemsuccess:Aten-yearupdate. JounrnalofManagement InformationSystems ,19:9. Dillman,D.A.,Smyth,J.D.,andChristian,L.M.. Internet,Mail, andMixed-ModeSurveys:TheTailoredDesignMethod .Wiley,3rdedition. Dwyer,C.,Hiltz,S.R.,andPasserini,K..Trustandprivacyconcern withinsocialnetworkingsites:Acomparisonoffacebookandmyspace.In ProceedingsoftheThirteenthAmericasConferenceonInformationSystems Ellison,N.B.,Steineld,C.,andLampe,C..Thebenetsoffacebook "friends:"socialcapitalandcollegestudents'useofonlinesocialnetwork sites. JournalofComputer-MediatedCommunication ,12:1143. Fishbein,M.andAzjen,I.. Belief,Attitude,Intention,andBehavior: AnIntroductiontoTheoryandResearch. Addison-Wesley. Fornell,C.andLarcker,D.F..Evaluatingstructuralequationmodels withunobservablevariablesandmeasurementerror. JournalofMarketing ResearchJMR ,18:3950. 218

PAGE 240

Garcia,D.,Mavrodiev,P.,andSchweitzer,F..Socialresilienceinonline communities:Theautopsyoffriendster.In ProceedingsoftheFirstACM ConferenceonOnlineSocialNetworks ,COSN'13,pages39,NewYork, NY,USA.ACM. Gefen,D.,Straub,D.,andBoudreau,M.-C..Structuralequationmodelingandregression:Guidelinesforresearchpractice. Communicationsof theAssociationforInformationSystems ,4:1. Gefen,D.andStraub,D.W..Genderdierencesintheperception anduseofe-mail:Anextensiontothetechnologyacceptancemodel. MIS Quarterly ,21:389. Gefen,David,E.K.andStraub,D.W..Trustandtaminonline shopping:Anintegratedmodel. MISQuarterly ,27:51. Gerbing,D.W.andAnderson,J.C..Anupdatedparadigmforscale developmentincorporatingunidimensionalityanditsassessment. Journalof MarketingResearch ,25:186. Givon,M..Varietyseekingthroughbrandswitching. MarketingScience ,3:1. Granovetter,M.S..Thestrengthofweakties. AmericanJournalof Sociology ,78:1360. Gross,R.andAcquisti,A..Informationrevelationandprivacyinonline socialnetworks.In Proceedingsofthe2005ACMworkshoponPrivacyin theelectronicsociety ,volume1,pages71. Hair,J.F.,Ringle,C.M.,andSarstedt,M..Pls-sem:Indeedasilver bullet. JournalofMarketingTheoryandPractice ,19:139. 219

PAGE 241

Hair,J.F.,Sarstedt,M.,Ringle,C.M.,andMena,J.A..Anassessment oftheuseofpartialleastsquaresstructuralequationmodelinginmarketing research. JournaloftheAcademyofMarketingScience ,40:414. Hair,J.J.F.,Black,W.C.,Babin,B.J.,Anderson,R.E.,andTatham,R.L. MultivariateDataAnalysis .PearsonEducation,Inc.,6thedition. Hardgrave,B.C.,Davis,F.D.,andRiemenschneider,C.K..Investigatingdeterminantsofsoftwaredevelopers'intentionstofollowmethodologies. JournalofManagementInformationSystems ,20:pp.123. Henseler,J.,Ringle,C.M.,andSinkovics,R.R.2009.Theuseofpartialleast squarespathmodelingininternationalmarketing. AdvancesinInternational Marketing ,20:27719. Henseler,J.andSarstedt,M..Goodness-of-tindicesforpartialleast squarespathmodeling. ComputationalStatistics ,28:565. Hewitt,A.andForte,A..Crossingboundaries:Identitymanagementandstudent/facultyrelationshipsonthefacebook. Posterpresentedat CSCW,Ban,Alberta,Canada Hovorka,D.S.,Larsen,K.R.,Birt,J.,andFinnie,G..Ameta-theoretic approachtotheoryintegrationininformationsystems.In Proceedingsofthe 201346thHawaiiInternationalConferenceonSystemSciences ,pages4656 4665. Hsieh,J.-K.,Hsieh,Y.-C.,Chiu,H.-C.,andFeng,Y.-C..Post-adoption switchingbehaviorforonlineservicesubstitutes:Aperspectiveofthepushpull-mooringframework. ComputersinHumanBehavior ,28:1912. 220

PAGE 242

Hu,T.andKettinger,W.J..Whypeoplecontinuetousesocialnetworkingservices:Developingacomprehensivemodel.In InternationalConferenceonInformationSystemsICIS .ICIS2008Proceedings. Iriberri,A.andLeroy,G..Alife-cycleperspectiveononlinecommunity success. ACMComputingSurveys ,41:11:1. Joinson,A.N..Lookingat,lookinguporkeepingupwithpeople?: motivesanduseoffacebook.In twenty-sixthannualSIGCHIconferenceon Humanfactorsincomputingsystems Jones,M.A.,Mothersbaugh,D.L.,andBeatty,S.E..Switchingbarriersandrepurchaseintentionsinservices. JournalofRetailing ,76:259. Keaveney,S.M..Customerswitchingbehaviorinserviceindustries: Anexploratorystudy. JournalofMarketing ,59:pp.71. Keaveney,S.M.andParthasarathy,M..Customerswitchingbehavior inonlineservices:Anexploratorystudyoftheroleofselectedattitudinal, behavioral,anddemographicfactors. JournaloftheAcademyofMarketing Science ,29:274. Kim,B..Understandingantecedentsofcontinuanceintentioninsocialnetworkingservices. Cyberpsychology,Behavior,andSocialNetworking 14:199. Kim,D.J.,Ferrin,D.L.,andRao,H.R..Trustandsatisfaction, twosteppingstonesforsuccessfule-commercerelationships:Alongitudinal exploration. InformationSystemsResearch ,20:237. King,W.R.andHe,J..Ameta-analysisofthetechnologyacceptance model. Information&Management ,43:740. 221

PAGE 243

Kwon,O.andWen,Y..Anempiricalstudyofthefactorsaecting socialnetworkserviceuse. ComputersinHumanBehavior ,26:254. Lave,J.andWenger,E.. SituatedLearning:LegitimatePeripheral Participation .CambridgeUniversityPress. Lee,Y.,Kozar,K.A.,andLarsen,K.R..Thetechnologyacceptance model:Past,present,andfuture. CommunicationsoftheAssociationfor InformationSystems ,12:752. Lewis,J.andWest,A..'friending':London-basedundergraduates' experienceoffacebook. newmediaandsociety ,11:1209. Liang,H.,Saraf,N.,Hu,Q.,andXue,Y..Assimilationofenterprise systems:Theeectofinstitutionalpressuresandthemediatingroleoftop management. MISQuarterly ,31:59. Liao,C.,Palvia,P.,andChen,J.-L..Informationtechnologyadoption behaviorlifecycle:Towardatechnologycontinuancetheorytct. InternationalJournalofInformationManagement ,29:309320. Limayem,M.,Hirt,S.G.,andCheung,C.M.K..Howhabitlimitsthe predictivepowerofintention:Thecaseofinformationsystemscontinuance. MISQuarterly ,31:pp.705. Lin,K.-Y.andLu,H.-P..Whypeopleusesocialnetworkingsites: Anempiricalstudyintegratingnetworkexternalitiesandmotivationtheory. ComputersinHumanBehavior ,27:11521161.Group Awarenessin{CSCL}Environments. Lu,J.,Yao,J.E.,andYu,C.-S..Personalinnovativeness,socialinuencesandadoptionofwirelessinternetservicesviamobiletechnology. The 222

PAGE 244

JournalofStrategicInformationSystems ,14:245268.TheFutureis UNWIRED:OrganizationalandStrategicPerspectives. Madden,M..Teenshaven'tabandonedfacebookyet.Commentary, PewResearchInternetProject,1615LStreet,NW,Suite700,Washington, DC20036. Madden,M.,Lenhart,A.,Cortesi,S.,Gasser,U.,Duggan,M.,Smith,A.,and Beaton,M..Teens,socialmediaandprivacy.Technicalreport,Pew ResearchCenter'sInternetandAmericanLifeProject. Madden,M.andSmith,A..Privacymanagementonsocialmediasites. Technicalreport,PewResearchCenter'sInternetandAmericanLifeProject. Mahajan,V.,Muller,E.,andBass,F.M..Newproductdiusionmodels inmarketing:Areviewanddirectionsforresearch. JournalofMarketing 54:pp.1. Mahajan,V.,Sharma,S.,andBuzzell,R.D..Assessingtheimpact ofcompetitiveentryonmarketexpansionandincumbentsales. Journalof Marketing ,57:pp.39. Markus,M.andRobey,D..Informationtechnologyandorganizationalchange:causalstructureintheoryandresearch. ManagementScience 34:583. Mathieson,K..Predictinguserintentions:Comparingthetechnology acceptancemodelwiththetheoryofplannedbehavior. InformationSystems Research ,2:pp.173. McAlister,L.andPessemier,E..Varietyseekingbehavior:Aninterdisciplinaryreview. JournalofConsumerResearch ,9:311. 223

PAGE 245

McKnight,D.H.,Choudhury,V.,andKacmar,C..Developingandvalidatingtrustmeasuresfore-commerce:Anintegrativetypology. Information SystemsResearch ,13:pp.334. McLachlan,G.andPeel,D.2004. Finitemixturemodels .JohnWiley& Sons. Mosco,V.. Seekingconvergenceinpolicyandpractice ,chapterFrom HeretoBanality:MythsAboutNewMediaandCommunicationPolicy, pages25.CanadianCentreforPolicyAlternatives. Nakamura,J.andCsikszentmihalyi,M.. OxfordHandbookofPositivePsychology ,chapterFlowtheoryandresearch,pages195.Oxford UniversityPress. Nardi,B.,Ellison,N.,andLampe,C..Understandingscience:context forsenatorcoburn's'underthemicroscope'report. Interactions ,18:32 35. Nielsen.Stateofthemedia:Thesocialmediareport-q32011.Technicalreport,Nielsen. Nielsen.Thesocialmediareport2012:Socialmediacomesofage. Technicalreport,Nielsen. Oliver,R.L..Acognitivemodeloftheantecedentsandconsequences ofsatisfactiondecisions. JournalofMarketingResearch ,17:pp.460. Ong,C.-S.andDay,M.-Y..Anintegratedevaluationmodelofuser satisfactionwithsocialmediaservices.In IEEEInternationalConference onInformationReuseandIntegration ,pages195.IEEE. 224

PAGE 246

Orlikowski,W.J.andRobey,D..Informationtechnologyandthe structuringoforganizations. InformationSystemsResearch ,2:143. Parthasarathy,M.andBhattacherjee,A..Understandingpost-adoption behaviorinthecontextofonlineservices. InformationSystemsResearch 9:pp.362. Petter,S.andMcLean,E.R..Ameta-analyticassessmentofthedelone andmcleanissuccessmodel:Anexamiiniationofissuccessattheindividual level. Information&Management ,46:159. Podsako,P.M.,MacKenzie,S.B.,Lee,J.-Y.,andPodsako,N.P.. Commonmethodbiasesinbehavioralresearch:Acriticalreviewoftheliteratureandrecommendedremedies. JournalofAppliedPsychology ,88:879 903. Raine,L.,Smith,A.,andDuggan,M..Comingandgoingonfacebook.Technicalreport,PewInternet&AmericanLifeProject,PewResearchCenter'sInternet&AmericanLifeProject,1615LSt.,NW-Suite700, Washington,D.C.20036. Rao,L..Facebook'szuckerberg:Ifiwerestartingacompanynow,i wouldhavestayedinboston. TechCrunch ,pageonline. Ringle,C.M.,Wende,S.,andWill,A..Smartpls2.0m3.Technical report,SmartPLS,Hamburg,Germany.2.0beta. Roberts,J..Limitstocommunitiesofpractice. JournalofManagement Studies ,43:623. Rogers,E.M.. DiusionofInnovations .FreePress. 225

PAGE 247

Rogers,E.M..Newproductadoptionanddiusion. JournalofConsumerResearch ,2:pp.290. Sanchez-Garcia,I.,Pieters,R.,Zeelenberg,M.,andBigna,E..When satisedconsumersdonotreturn:Varietyseeking'seectonshort-and long-termintentions. Psychology&Marketing ,29:15. Schaupp,L.C.2011.Websitesuccess:Antecedentsofwebsitesatisfaction andre-use. JournalofInternetCommerce ,9:42. Scott,J..Socialnetworkanalysis. Sociology ,22:109. Sharma,N.andPatterson,P.G..Switchingcosts,alternativeattractivenessandexperienceasmoderatorsofrelationshipcommitmentin professional,consumerservices. InternationalJournalofServiceIndustry Management ,11:470. Sibona,C.andChoi,J.H..Factorsaectingend-usersatisfactionon facebook.In SixthInternationalAAAIConferenceonWeblogsandSocial MediaICWSM ,volume6th,pages1. Spreng,R.A.,MacKenzie,S.B.,andOlshavsky,R.W..Areexaminationofthedeterminantsofconsumersatisfaction. JournalofMarketing 60:pp.15. Srinivasan,A..Alternativemeasuresofsystemeectiveness:Associationsandimplications. MISQuarterly ,9:pp.243. Stutzman,F..Anevaluationofidentity-sharingbehaviorinsocial networkcommunities. JournaloftheInternationalDigitalMediaandArts Association ,3:10. 226

PAGE 248

Szymanski,D.M.andHenard,D.H..Customersatisfaction:ametaanalysisoftheempiricalevidence. Journaloftheacademyofmarketing science ,29:16. Tenenhaus,M.,Vinzi,V.E.,Chatelin,Y.-M.,andLauro,C..{PLS} pathmodeling. ComputationalStatistics&DataAnalysis ,48:159205. PartialLeastSquares. Thatcher,J.B.andPerrew,P.L..Anempiricalexaminationofindividualtraitsasantecedentstocomputeranxietyandcomputerself-ecacy. MISQuarterly ,26:pp.381. Thong,J.Y.,Hong,S.-J.,andTam,K.Y.2006.Theeectsofpost-adoption beliefsontheexpectation-conrmationmodelforinformationtechnology continuance. InternationalJournalofHuman-ComputerStudies ,64:799 810. Venkatesh,V.,Davis,F.,andMorris,M.G..DeadOrAlive?The Development,TrajectoryAndFutureOfTechnologyAdoptionResearch. JournaloftheAssociationforInformationSystems ,8:267. Vitak,J.,Ellison,N.B.,andSteineld,C..Thetiesthatbond:Reexaminingtherelationshipbetweenfacebookuseandbondingsocialcapital. In Proceedingsofthe201144thHawaiiInternationalConferenceonSystem Sciences ,volume44,pages1. Wetzels,M.,Odekerken-Schrder,G.,andvanOppen,C..Usingpls pathmodelingforassessinghierarchicalconstructmodels:Guidelinesand empiricalillustration. MISQuarterly ,33:177195. Whisman,M.andMcClelland,G..Designing,testing,andinterpreting interactionsandmoderatoreectsinfamilyresearch. Journaloffamilypsy227

PAGE 249

chology:JFP:journaloftheDivisionofFamilyPsychologyoftheAmerican PsychologicalAssociationDivision43 ,19:111. Wilson,R.E.,Gosling,S.D.,andGraham,L.T..Areviewoffacebookresearchinthesocialsciences. PerspectivesonPsychologicalScience 7:203. Xu,C.,Ryan,S.,Magro,M.,andWen,C..Whydopeoplestick withaspecicsocialnetworkingsite?anintegratedrelationshipanduses graticationperspective.In AMCIS2012 ,pages1. Yoder,C.andStutzman,F..Identifyingsocialcapitalinthefacebook interface.In annualconferenceonHumanfactorsincomputingsystems ACM. Zhang,K.Z.,Lee,M.K.,Cheung,C.M.,andChen,H..Understanding theroleofgenderinbloggersswitchingbehavior. DecisionSupportSystems 47:540. 228

PAGE 250

ACommonMethodVariance Commonmethodvarianceresultsareshownusingacommonmethodfactor asdescribedbyPodsakoetal.andLiangetal..Theanalysis wasperformedonthecompletenon-moderatedmodel. Table43:CommonMethodBiasAnalysis ConstructIndicatorSubstantive Factor Loading S Path Coecient S Method Factor Loading M Path Coecient M Alternative Attractiveness AltAtt10.72350.52350.16400.0269 AltAtt20.73690.54300.15720.0247 AltAtt30.86530.74870.04780.0023 AltAtt40.81650.6667-0.07230.0052 AltAtt50.76500.5852-0.13700.0188 Attitude Toward Switching AttSwitch10.89440.80000.00990.0001 AttSwitch20.93000.86490.08370.0070 AttSwitch30.76380.5834-0.01120.0001 AttSwitch40.84880.7205-0.00670.0000 229

PAGE 251

Table43:CommonMethodBiasAnalysis ConstructIndicatorSubstantive Factor Loading S Path Coecient S Method Factor Loading M Path Coecient M AttSwitch50.84390.7122-0.01080.0001 AttSwitch60.89680.8043-0.03120.0010 BrandRelationshipBrand10.76910.59150.09940.0099 Brand20.84970.7220-0.07590.0058 Brand30.76320.5825-0.03770.0014 ContinuanceCont10.90280.81500.02570.0007 Cont20.74430.55400.11030.0122 Cont30.87870.7721-0.16660.0278 HabitHabit10.91330.8341-0.01960.0004 Habit20.93470.8737-0.02240.0005 Habit30.65030.42290.05760.0033 Interpersonal Inuence Interpers10.89660.80390.00240.0000 Interpers20.78380.6143-0.10130.0103 Interpers30.41220.16990.03640.0013 230

PAGE 252

Table43:CommonMethodBiasAnalysis ConstructIndicatorSubstantive Factor Loading S Path Coecient S Method Factor Loading M Path Coecient M Interpers40.85490.7309-0.01180.0001 Personal Innovativeness PI10.84510.71420.02480.0006 PI20.71210.5071-0.07440.0055 PI30.81220.65970.11930.0142 PI40.91320.83390.03130.0010 PI50.87420.7642-0.09380.0088 Perceived usefulness PU10.90820.8248-0.08630.0074 PU20.68500.46920.12110.0147 PU30.90330.8160-0.03940.0016 PU40.68160.4646-0.02590.0007 ProcCostProcCost10.62430.38980.09340.0087 ProcCost20.76420.5840-0.03100.0010 ProcCost30.79940.63900.02120.0004 231

PAGE 253

Table43:CommonMethodBiasAnalysis ConstructIndicatorSubstantive Factor Loading S Path Coecient S Method Factor Loading M Path Coecient M ProcCost40.79870.6379-0.08750.0077 ProcEvalProcEval10.84610.71590.06190.0038 ProcEval20.87550.7665-0.10240.0105 ProcEval30.64880.42090.02740.0008 ProcEval40.20700.04280.20700.0428 ProcLearnProcLearn10.77490.60050.01160.0001 ProcLearn20.58250.3393-0.17140.0294 ProcLearn30.75160.56490.11300.0128 ProcLearn40.81070.6572-0.06840.0047 Procedural Relationship ProcRel10.88520.7836-0.07810.0061 ProcRel20.43500.18920.43500.1892 ProcRel30.91440.8361-0.19370.0375 ProcRel40.81840.6698-0.03950.0016 ProcRel50.88200.7779-0.18440.0340 232

PAGE 254

Table43:CommonMethodBiasAnalysis ConstructIndicatorSubstantive Factor Loading S Path Coecient S Method Factor Loading M Path Coecient M P:rocedural Setup ProcSetup10.71890.51680.01390.0002 ProcSetup20.76460.5846-0.00030.0000 ProcSetup30.67380.45400.02270.0005 ProcSetup40.73760.5441-0.03770.0014 SatisfactionSat10.96030.92220.00970.0001 Sat20.98010.9606-0.02700.0007 Sat30.95030.90310.00740.0001 ConrmationConf10.86790.75330.02530.0006 Conf20.83090.6904-0.00810.0001 Conf30.61670.38030.03830.0015 Conf40.86050.7405-0.05100.0026 Average0.78940.64190.00120.0101 233

PAGE 255

BOuterLoadingAnalysis Outerloadingsanalysisofthecompletenon-mediatedmodel. Table44:OuterLoadingAnalysis ConstructIndicatorLoadingC.R.Cronbach's Alpha AlternativeAttractivenessAltAtt10.74540.90120.8627 AltAtt20.7031 AltAtt30.8594 AltAtt40.8595 AltAtt50.8428 AttitudeTowardSwitchingAttSwitch10.9090.94400.9286 AttSwitch20.8336 AttSwitch30.7989 AttSwitch40.8689 AttSwitch50.8522 AttSwitch60.8872 BrandRelationshipBrand10.88120.86380.7634 Brand20.875 Brand30.7076 ContinuanceCont10.91180.88310.8004 Cont20.8413 Cont30.7813 234

PAGE 256

Table44:OuterLoadingAnalysis ConstructIndicatorLoadingC.R.Cronbach's Alpha HabitHabit10.90880.90790.8469 Habit20.9082 Habit30.8067 InterpersonalInuenceInterpers10.85270.85990.7954 Interpers20.6336 Interpers30.7569 Interpers40.8564 PersonalInnovativenessPI10.85450.92190.8987 PI20.7111 PI30.8989 PI40.8913 PI50.8252 PerceivedusefulnessPU10.89740.89390.8397 PU20.7870 PU30.9057 PU40.6907 ProcCostProcCost10.7150.83820.7421 ProcCost20.7469 ProcCost30.7975 235

PAGE 257

Table44:OuterLoadingAnalysis ConstructIndicatorLoadingC.R.Cronbach's Alpha ProcCost40.7439 ProcEvalProcEval10.80740.81810.6703 ProcEval20.6889 ProcEval30.8231 ProcEval4N/A ProcLearnProcLearn10.80980.8590.7832 ProcLearn20.7201 ProcLearn30.8057 ProcLearn40.7699 PersonalRelationshipLossProcRel10.83060.89900.8585 ProcRel20.6845 ProcRel30.8043 ProcRel40.8338 ProcRel50.8412 P:roceduralSetupProcSetup10.81190.85300.7707 ProcSetup20.8042 ProcSetup30.6525 ProcSetup40.8021 SatisfactionSat10.96850.97580.9629 236

PAGE 258

Table44:OuterLoadingAnalysis ConstructIndicatorLoadingC.R.Cronbach's Alpha Sat20.9700 Sat30.9559 ConrmationConf10.91450.89690.8427 Conf20.8817 Conf30.5842 Conf40.8994 237

PAGE 259

CSNSContinuanceandHabit C.1ModelDescription ThebasemodelincludesthemeasuresfromtheISContinuanceModelof Bhattacherjeeandexaminesthedirectandmoderatingeectsof habit Themodelusestheindependentconstructsconrmation,perceivedusefulness, satisfactionandhabittopredictinformationsystemscontinuanceintention. HabitistheorizedtohaveamoderatingeectonContinuancethroughthe satisfactioncontinuancepath. Figure47:HabitHistogram C.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations 238

PAGE 260

Table45:BaseModel+HabitMeasurementModel ConstructAVECompositeReliability Conrmation 0.69070.8968 PerceivedUsefulness 0.68040.8938 Satisfaction 0.93080.9758 Continuance 0.71770.8837 Habit 0.76570.9072 Sat*Habit 0.7660.9671 Table46:BaseModel+HabitMeasurementModel-DiscriminantValidity FornellandLarckerCriterion CONFCONTINUANCE HABIT PUSATSAT* HABIT CONF 0.8311 CONTINUANCE 0.7118 0.8472 HABIT 0.6348 0.6970 0.8750 PU 0.5994 0.6324 0.6562 0.8249 SAT 0.7887 0.7879 0.6233 0.5990 0.9648 SAT* HABIT -0.2618 -0.3627 -0.4440 -0.3133 -0.2713 0.8752 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. indicatingdiscriminantvalidityFornellandLarckercriterion-See:46. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. 239

PAGE 261

Table47:BaseModel+HabitStructuralModel GoodnessofFitmoderatedmodel .6590 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 PerceivedUsefulness 0.35920.2435 Satisfaction 0.64690.6021 Continuance 0.71160.5057 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. C.3StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactionandhabittopredict informationsystemscontinuanceintention..Thepredictorsconrmation,perceivedusefulness,satisfaction,habitandsatisfactionmoderatedbyhabitexplainapproximately71.2%ofthevarianceR 2 incontinuanceintentionandis consideredtohaveamoderatelevelofexplanation 25 ;themodelalsoexhibits predictiverelevanceQ 2 whereitsvalueis0.5057scoresabovezeroindicate predictiverelevanceinPLSpathmodels.Satisfactionhadthestrongestpredictiveabilityforsatisfactionwithastandardizedpathcoecientof.5224 t =22.8887 followedbyHabit.2633, t =10.3241andperceivedusefulness withastandardizedpathcoecientof0.2268 t =8.8468.Themoderating eectofHabit*Satisfactionhadthesmallesteect-0.0683, t =4.3531.The coecientsthatarepositiveindicatethatareassociatedwithhigherlevelsof continuanceintention.Themoderatingeectisnegativemostlikelydueto multi-collinearity. 25 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 240

PAGE 262

RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SAT CONTINUANCE0.522422.8887 HABIT CONTINUANCE0.263110.3241 PU CONTINUANCE0.12384.931 SAT*HABIT CONTINUANCE-0.06834.3531 OtherPathCoecients CONF PU0.599430.9255 CONF SAT0.670533.4842 PU SAT0.19718.8051 TotalEects ContinuancePathCoecients CONF CONTINUANCE0.487025.3226 SAT CONTINUANCE0.522622.8887 HABIT CONTINUANCE0.263310.3241 PU CONTINUANCE0.22758.8468 age CONTINUANCE0.09636.2138 SAT*HABIT CONTINUANCE-0.06714.3531 gender CONTINUANCE0.01991.3412 241

PAGE 263

RelationPath Coecient Tstatistic Education CONTINUANCE0.01450.8938 OtherPathCoecients CONF PU0.600030.9255 CONF SAT0.789068.2235 PU SAT0.19758.8051 C.4ModeratingEectAnalysis TheresultsofthemoderatingtermarelessthansmallaccordingtoChinetal. .TheR 2 forthenon-moderatedmodelis.7076andforthemoderated modelitis.7116.Thepathcoecientfor satisfaction inthemoderatedand non-moderatedmodelare.5224and.5222,respectively.Thereisnochange inthepathcoecientfor perceivedusefulness to satisfaction .1971.The moderatingt-statistic t =4.3531indicatesthatitisstatisticallysignicant althoughtheeectsize f 2 is.0139whichdoesnotmeetthethresholdfortobe consideredashavinganeectthresholdforsmalleects f 2 is.02basedon Chinetal. 242

PAGE 264

Figure48:BaseModelAndHabitPathCoecientsonSNSContinuance 243

PAGE 265

DSNSContinuanceandPersonalInnovativeness D.1ModelDescription ThebasemodelincludesthemeasuresfromtheISContinuanceModelofBhattacherjeeandexaminesthedirectandmoderatingeectsof personal innovativeness .Themodelusestheindependentconstructsconrmation,perceivedusefulness,satisfactionandpersonalinnovativenesstopredictinformationsystemscontinuanceintention.Personalinnovativenessistheorizedto haveamoderatingeectoncontinuancethroughthesatisfactioncontinuance path. Figure49:PersonalInnovativenessHistogram D.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal 244

PAGE 266

Table49:BaseModel+PersonalInnovativenessMeasurementModel ConstructAVECompositeReliability Conrmation 0.69070.8968 PerceivedUsefulness 0.68040.8938 Satisfaction 0.93080.9758 Continuance 0.71790.8839 Personalinnovativeness 0.70400.9220 Sat*Personalinnovativeness 0.63080.9616 consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations indicatingdiscriminantvalidityFornellandLarckercriterion-See:50. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. D.3StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactionandpersonalinnovativenesstopredictinformationsystemscontinuanceintention..Thepredictorsconrmation,perceivedusefulness,satisfaction,personalinnovativeness andsatisfactionmoderatedbyhabitexplainapproximately68.4%ofthevarianceR 2 incontinuanceintentionandisconsideredtohaveamoderatelevel ofexplanation 26 ;themodelalsoexhibitspredictiverelevanceQ 2 whereits valueis.4878scoresabovezeroindicatepredictiverelevanceinPLSpath 26 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 245

PAGE 267

Table50:BaseModel+PersonalInnovativeness-DiscriminantValidity FornellandLarckerCriterion CONF CONTINUANCE PUPersonal Innovativeness SATSAT* Personal Innovativeness CONF 0.8311 CONTINUANCE 0.7108 0.8473 PU 0.5994 0.6302 0.8249 Personal Innovativeness -0.0427 -0.1618 -0.0291 0.8390 SAT 0.7887 0.7877 0.5990 -0.0382 0.9648 SAT* Personal Innovativeness -0.0284 -0.0548 -0.0368 -0.0294 -0.0306 0.7942 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. 246

PAGE 268

Table51:BaseModel+PersonalInnovativenessStructuralModel GoodnessofFitmoderatedmodel 0.6394 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 PerceivedUsefulness 0.35920.2435 Satisfaction 0.64690.6021 Continuance 0.68390.4878 R 2 coecientofdetermination Q 2 predictiverelevance.Scoresabovezeroindicatepredictiverelevance. models.Satisfactionhadthestrongestpredictiveabilityforsatisfactionwith astandardizedpathcoecientof.6242 t =28.9068followedbyperceived usefulnesswithastandardizedpathcoecientof.2465 t =9.962andpersonalinnovativeness-.1303, t =8.3021.Themoderatingeectofpersonal innovativeness*Satisfactionhadthesmallesteect-0.0305, t =1.4605.The coecientsthatarepositiveindicatethatareassociatedwithhigherlevelsof continuanceintention. 247

PAGE 269

RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SAT CONTINUANCE0.624228.9068 PU CONTINUANCE0.24659.962 PersonalInnovativeness CONTINUANCE-0.13038.3021 SAT*PersonalInnovativeness CONTINUANCE -0.03051.4605 OtherPathCoecients CONF PU0.599431.596 CONF SAT0.670532.1036 PU SAT0.19718.4888 TotalEects ContinuancePathCoecients CONF CONTINUANCE0.638946.4045 SAT CONTINUANCE0.623428.9068 PU CONTINUANCE0.369314.9739 PersonalInnovativeness CONTINUANCE-0.13128.3021 age CONTINUANCE0.07204.2785 gender CONTINUANCE0.03732.3325 248

PAGE 270

RelationPath Coecient Tstatistic SAT*PersonalInnovativeness CONTINUANCE -0.02921.4588 Ed CONTINUANCE0.00820.5400 OtherPathCoecients CONF PU0.599531.5960 CONF SAT0.788169.4189 PU SAT0.19798.4888 D.4ModeratingEectAnalysis TheresultsofthemoderatingtermarelessthansmallaccordingtoChinetal. .TheR 2 forthenon-moderatedmodelis0.6830andforthemoderated modelitis0.6839.ThepathcoecientforSatisfactioninthemoderatedand non-moderatedmodelare.6242and.6247,respectively.Thepathcoecient forperceivedusefulnessinthemoderatedandnon-moderatedmodelare0.2465 and0.2473,respectively..Thepathcoecientforpersonalinnovativeness changedfrom-.1303moderatedmodelto-.1293non-moderatedmodel. Themoderatingt-statisticindicatesthatitisnotstatisticallysignicant t = 1.4588;theeectsize f 2 is0.0028whichdoesnotmeetthethresholdfortobe consideredashavinganeectthresholdforsmalleects f 2 is.0200based on Chinetal. 249

PAGE 271

Figure50:BaseModelAndPersonalInnovativenessPathCoecientsonSNS Continuance 250

PAGE 272

ESNSContinuanceandInterpersonalInuence E.1ModelDescription ThebasemodelincludesthemeasuresfromtheISContinuanceModelof Bhattacherjeeandexaminesthedirecteectsof interpersonalinuence .Themodelusestheindependentconstructsconrmation,perceivedusefulness,satisfactionandinterpersonalinuencetopredictinformationsystems continuanceintention. Figure51:InterpersonalInuenceHistogram E.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations indicatingdiscriminantvalidityFornellandLarckercriterion-See:54. 251

PAGE 273

Table53:BaseModel+InterpersonalInuenceMeasurementModel ConstructAVECompositeReliability CONF 0.6910.8969 CONTINUANCE 0.71640.8831 InterpersInuence 0.60720.8591 PU 0.68040.8938 SAT 0.93090.9758 Table54:BaseModel+InterpersonalInuenceMeasurementModel-DiscriminantValidity FornellandLarckerCriterion CONF CONTINUANCE Interpers Inuence PUSAT CONF 0.8313 CONTINUANCE 0.7112 0.8464 Interpers Inuence 0.2111 0.1028 0.7792 PU 0.5992 0.6293 0.2565 0.8249 SAT 0.7880 0.7889 0.1101 0.5983 0.9648 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. E.3StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactionandhabittopredict 252

PAGE 274

Table55:SNSContinuanceStructuralModelandInterpersonalInuence GoodnessofFit 0.6360 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 CONTINUANCE 0.66850.4716 PU 0.35900.2433 SAT 0.64570.6010 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. informationsystemscontinuanceintention..Thepredictorsconrmation,perceivedusefulness,satisfaction,habitandsatisfactionmoderatedbyhabitexplainapproximately66.9%ofthevarianceR 2 incontinuanceintentionandis consideredtohaveamoderatelevelofexplanation 27 ;themodelalsoexhibits predictiverelevanceQ 2 whereitsvalueis0.4716scoresabovezeroindicate predictiverelevanceinPLSpathmodels.Satisfactionhadthestrongestpredictiveabilityforsatisfactionwithastandardizedpathcoecientof.6294 t =30.4700followedbyperceivedusefulness.2543, t =10.3447.Interpersonalinuenceddidnothaveastatisticallysignicanteectoncontinuance; thepathcoecientwas-.0289 t =1.7005.Thecoecientsthatarepositive indicatethatareassociatedwithhigherlevelsofcontinuanceintention. RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients 27 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 253

PAGE 275

RelationPath Coecient Tstatistic SAT CONTINUANCE0.629430.4700 PU CONTINUANCE0.254310.3447 InterpersInuence CONTINUANCE-0.02891.7005 OtherPathCoecients CONF SAT0.670032.7312 CONF PU0.599231.5449 PU SAT0.19688.8316 TotalEects ContinuancePathCoecients CONF CONTINUANCE0.648746.0435 SAT CONTINUANCE0.630030.4700 PU CONTINUANCE0.376915.0685 age CONTINUANCE0.07114.1547 gender CONTINUANCE0.04222.5614 InterpersInuence CONTINUANCE-0.0271.7005 education CONTINUANCE0.01330.7563 OtherPathCoecients CONF PU0.599831.5449 CONF SAT0.788268.8001 254

PAGE 276

RelationPath Coecient Tstatistic PU SAT0.19618.8316 255

PAGE 277

Figure52:BaseModelAndHabitPathCoecientsonSNSContinuance 256

PAGE 278

FSNSContinuanceandAlternativePerceptions F.1FourAlternativeSocialNetworkingSiteAccountImpact ANCOVAwasanalyzedtodeterminewhethersimplyhavinganaccountwith oneoffoursitesimpactscontinuanceintentionandmeasureanystatistically signicantdierencesbetweenaccountholdersofthespecicsitesandnonaccountholders.Theanalysisincludedcovariatesageandgendertoadjust fordierences.ThedependentvariableiscontinuanceintentiononFacebook. Twositeshadstatisticallysignicantimpacts;Twitteruserswere more likely to discontinue useofFacebookthanthosewhodidnotuseTwitter,andInstagramuserswere more likelyto continue touseFacebookthanthosewhodid notuseInstagram.WhetherauserhadaPinterestorTumblraccountdidnot makeastatisticallysignicantdierenceoncontinuanceintention.TheadjustedcoecientofdeterminationR 2 was.057andindicatesthatthemodel haslessthanaweaklevelofexplanatorypowerHairetal.,2011.See Table 57 and Figure53 fordetails. Table57:FourAlternativeSNSAccountImpacts NCI p UsesTwitter7084.483 0.043* DoesNotUseTwitter2104.866 UsesPinterest4604.594 0.430 DoNotUsePinterest4584.738 UsesInstagram4204.863 0.016* DoNotUseInstagram2944.431 UsesTumblr1594.798 0.136 DoNotUseTumblr7594.542 CIisthecontinuanceintentionon Facebook ,*: p <.05 257

PAGE 279

CIisthecontinuanceintentionon Facebook Figure53:FourAlternativesSNSAccountImpact Figure54:AlternativeAttractivenessHistogram F.2ModelDescription ThebasemodelincludesthemeasuresfromtheISContinuanceModelofBhattacherjeeandexaminesthedirecteectsof alternativeperceptions .The modelusestheindependentconstructsconrmation,perceivedusefulness,satisfactionandalternativeperceptionstopredictinformationsystemscontinuanceintention.Alternativeperceptionsisareective-formativeconstructthat iscomposedofalternativeattractivenessandattitudetowardswitching. 258

PAGE 280

Figure55:AttitudeToSwitchHistogram Combinedsecond-orderconstructofAlternativeAttractivenessandAttitude toSwitch Figure56: SecondOrderConstruct: AlternativePerceptionsHistogram 259

PAGE 281

Table58:ISContinuanceMeasurementModelwithAlternativePerceptions ConstructAVECompositeReliability AlterAttract 0.64760.9012 AttitudeToSwitch 0.73800.9440 CONF 0.69100.8969 CONTINUANCE 0.71660.8832 PU 0.68040.8938 SAT 0.93090.9758 F.3MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations indicatingdiscriminantvalidityFornellandLarckercriterion-See:59. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. F.4StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactionandhabittopredict 260

PAGE 282

Table59:ISContinuanceMeasurementModelwithAlternativePerceptionsDiscriminantValidity FornellandLarckerCriterion Alter Attract Attitude To Switch CONF CONTINUANCE PUSAT Alter Attract 0.8047 Attitude ToSwitch 0.5751 0.8591 CONF -0.4942 -0.4413 0.8313 CONTINUANCE -0.6133 -0.5523 0.7109 0.8465 PU -0.3717 -0.3510 0.5992 0.6289 0.8249 SAT -0.5658 -0.4567 0.7880 0.7874 0.5983 0.9648 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. informationsystemscontinuanceintention..Thepredictorsconrmation,perceivedusefulness,satisfaction,habitandsatisfactionmoderatedbyhabitexplainapproximately72.1%ofthevarianceR 2 incontinuanceintentionandis consideredtohaveamoderatelevelofexplanation 28 ;themodelalsoexhibits predictiverelevanceQ 2 whereitsvalueis.5097scoresabovezeroindicate predictiverelevanceinPLSpathmodels.Satisfactionhadthestrongestpredictiveabilityforsatisfactionwithastandardizedpathcoecientof.472 t = 20.9223followedbyalternativeperceptions-.294, t =14.1585andperceived usefulness.216,t=10.0472.Thecoecientsthatarepositiveindicatethat areassociatedwithhigherlevelsofcontinuanceintention.Surveyrespondents whohadhigherattitudestowardswitchingandwereattractedtoalternatives werelesslikelytocontinueuseofFacebook. 28 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 261

PAGE 283

Table60:ISContinuanceStructuralModelandAlternativePerceptions GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 CONTINUANCE 0.72050.5097 PU 0.35900.2433 SAT 0.64570.6010 AlternativePerceptions N/A0.5344 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SAT CONTINUANCE0.472020.9223 AlternativePerception CONTINUANCE-0.294014.1585 PU CONTINUANCE0.215810.0472 OtherPathCoecients CONF SAT0.670032.0814 CONF PU0.599231.5416 AlterAttract AlternativePerception0.666718.6091 AttitudeToSwitch AlternativePerception0.450211.2022 PU SAT0.19688.6327 TotalEects 262

PAGE 284

RelationPath Coecient Tstatistic ContinuancePathCoecients CONF CONTINUANCE0.501226.5944 SAT CONTINUANCE0.472020.9223 AlternativePerception CONTINUANCE-0.294014.1585 PU CONTINUANCE0.308613.8792 AlterAttract CONTINUANCE-0.196012.1855 AttitudeToSwitch CONTINUANCE-0.13238.2601 age CONTINUANCE0.06003.9292 gender CONTINUANCE0.04953.3283 education CONTINUANCE0.00910.5881 OtherPathCoecients CONF SAT0.78865.9009 CONF PU0.599231.5416 AlterAttract AlternativePerception0.666718.6091 AttitudeToSwitch AlternativePerception0.450211.2022 PU SAT0.19688.6327 263

PAGE 285

Figure57:BaseModelAndHabitPathCoecientsonSNSContinuance 264

PAGE 286

GPredictingSNSContinuancethroughCosts G.1ModelDescription ThemodelincludesthemeasuresfromtheBurnhametal.andexamines theeectsoftwocategoriesofcosts:proceduralandrelational.TheBurnham etal.includesthreecosts-procedural,nancialandrelational;however,forthemajorityofsocialnetworkingsites,andforFacebookinparticular, therearenonancialcosts.Themodelusestheindependentconstructseconomicrisk,evaluationcosts,learningcostsandsetupcostsproceduralcosts andpersonalrelationshiplossandbrandrelationshiplossrelationalcoststo predictinformationsystemscontinuanceintention.Thecostsreectthecost toleavethecurrentserviceprovider,i.e.thehigherthecostthemorelikely thecustomeristostaywiththecurrentprovider.Themodelisgenerated withsecond-andthird-orderconstructsthatcombinetheeectsoflower-level constructs;thehigherorderconstructsarereective-formativeinnature. Figure58:BrandRelationshipHistogram 265

PAGE 287

Figure59:PersonalRelationshipHistogram Figure60:ProceduralEconomicCostHistogram 266

PAGE 288

Figure61:ProceduralEvaluationCostHistogram Figure62:ProceduralLearningCostHistogram 267

PAGE 289

Figure63:ProceduralSetupCostHistogram CombinedsecondorderconstructofBrandRelationshipandPersonalRelationshipCosts Figure64: SecondOrderConstruct: RelationshipCostHistogram 268

PAGE 290

CombinedsecondorderconstructofProceduralEconomicCost,Procedural EvaluationCost,ProceduralLearningCostandProceduralSetupCost Figure65: SecondOrderConstruct: ProceduralCostHistogram CombinedThirdorderconstructofRelationshipCostandProceduralCost Figure66: ThirdOrderConstruct: CostHistogram 269

PAGE 291

Table62:SNSCostMeasurementModel ConstructAVECompositeReliability BRAND_RELATIONSHIP 0.68090.8638 CONTINUANCE 0.71650.8830 PROC_COST 0.56460.8382 PROC_EVAL 0.60130.8181 PROC_LEARN 0.60410.8590 PROC_RELATIONSHIP_LOSS 0.64160.8990 PROC_SETUP 0.59380.8530 G.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations indicatingdiscriminantvalidityFornellandLarckercriterion-See:63. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindicatorreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. 270

PAGE 292

Table63:SNSCostMeasurementModel-DiscriminantValidity FornellandLarckerCriterion BRAND RELATIONSHIP CONTINUANCE PROC COST PROC EVAL PROC LEARN RELATIONSHIP LOSS PROC SETUP BRAND RELATIONSHIP 0.8252 CONTINUANCE 0.6097 0.8465 PROC COST 0.2221 0.3734 0.7514 PROC EVAL 0.0776 0.2027 0.4839 0.7754 PROC LEARN 0.1776 0.2721 0.6428 0.5004 0.7772 RELATIONSHIP LOSS 0.4936 0.7325 0.4668 0.1478 0.3127 0.8010 PROC SETUP 0.1169 0.2886 0.7094 0.5408 0.6849 0.3388 0.7706 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagonalelementsarethelatentvariablecorrelations. 271

PAGE 293

Table64:Indices GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 CONTINUANCE 0.64330.4507 ProceduralCost 0.99330.3249 Cost 0.99330.2178 RELATIONSHIPCOSTS 0.99880.5006 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. G.3StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructsconrmation,perceivedusefulness,satisfactionandhabittopredict informationsystemscontinuanceintention..Thepredictorsconrmation,perceivedusefulness,satisfaction,habitandsatisfactionmoderatedbyhabitexplainapproximately64.3%ofthevarianceR 2 incontinuanceintentionand isconsideredtohaveamoderatelevelofexplanation 29 ;themodelalsoexhibitspredictiverelevanceQ 2 whereitsvalueis.4507scoresabovezero indicatepredictiverelevanceinPLSpathmodels.Costistheonlydirect predictorforcontinuanceintentionandhasstandardizedpathcoecientof .796 t =67.9214.Therearetwopredictorsforcost,theproceduralandrelationshipcostswhererelationshipcostshasahigherpathcoecientof.981 t =86.6208andproceduralcostshavethelowerpathcoecientof.035 t = 1.5556.Therelationshiparestatisticallysignicantandtheproceduralcosts arenotstatisticallysignicant.Thecoecientsthatarepositiveindicatethat areassociatedwithhigherlevelsofcontinuanceintention. 29 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 272

PAGE 294

RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients COST CONTINUANCE0.796467.9214 OtherPathCoecients RELATIONSHIP_COSTS COST0.980686.6208 PROC_RELATIONSHIP_LOSS RELATIONSHIP_COSTS 0.733934.5113 BRAND_RELATIONSHIP RELATIONSHIP_COSTS 0.406716.0826 PROC_COST PROCEDURAL_COST0.948313.2494 PROC_EVAL PROCEDURAL_COST-0.1251.7337 PROC_LEARN PROCEDURAL_COST0.15061.6895 PROCEDURAL_COST COST0.03511.5556 PROC_SETUP PROCEDURAL_COST0.0040.0412 TotalEects ContinuancePathCoecients COST CONTINUANCE0.796467.9214 RELATIONSHIP_COSTS CONTINUANCE0.78152.828 PROC_RELATIONSHIP_LOSS CONTINUANCE 0.573231.1804 273

PAGE 295

RelationPath Coecient Tstatistic BRAND_RELATIONSHIP CONTINUANCE0.317615.1406 Ed CONTINUANCE0.05653.2817 Age CONTINUANCE0.04242.6689 PROC_COST CONTINUANCE0.02651.562 PROCEDURAL_COST CONTINUANCE0.0281.5484 PROC_EVAL CONTINUANCE-0.00351.1836 PROC_LEARN CONTINUANCE0.00421.068 Gender CONTINUANCE0.01380.7793 PROC_SETUP CONTINUANCE0.00010.0333 OtherPathCoecients RELATIONSHIP_COSTS COST0.980686.6208 PROC_RELATIONSHIP_LOSS COST0.719735.6677 PROC_RELATIONSHIP_LOSS RELATIONSHIP_COSTS 0.733934.5113 BRAND_RELATIONSHIP RELATIONSHIP_COSTS 0.406716.0826 BRAND_RELATIONSHIP COST0.398815.172 PROC_COST PROCEDURAL_COST0.948313.2494 PROC_EVAL PROCEDURAL_COST-0.1251.7337 PROC_LEARN PROCEDURAL_COST0.15061.6895 274

PAGE 296

Figure67:CostModelPathCoecientsonSNSContinuance RelationPath Coecient Tstatistic PROC_COST COST0.03331.5698 PROCEDURAL_COST COST0.03511.5556 PROC_EVAL COST-0.00441.1887 PROC_LEARN COST0.00531.0707 PROC_SETUP PROCEDURAL_COST0.00400.0412 PROC_SETUP COST0.00010.0334 275

PAGE 297

HCompetingModelsCostvs.Satisfaction H.1ModelDescription ThemodelincludesthemeasuresfromtheConsumerSwitchCostmodel Burnhametal.,2003andISContinuancemodelBhattacherjee,2001.The twopredictorsforISContinuance,satisfactionandperceivedusefulnesshave beencombinedintoasinglereective-formativeconstructthatallowsfora moredirectcomparisonofthecompetingmodels.TheBurnhametal. costmodelwasinitiallydevelopedasaformative-reectivemodelanddoesnot needatransformation.Thecostmodelusestheindependentconstructseconomicrisk,evaluationcosts,learningcostsandsetupcostsproceduralcosts andpersonalrelationshiplossandbrandrelationshiplossrelationalcoststo predictinformationsystemscontinuanceintention.TheISContinuancemodel usestheindependentconstructsconrmation,perceivedusefulness,satisfactiontopredictinformationsystemscontinuanceintention. H.2MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providing supportforconvergentvalidityHairetal.,2006.Compositereliabilityvalues areallgreaterthanorequalto.7providingevidenceoftheconstruct'sinternal consistencyreliabilityHairetal.,2006.Compositereliabilityvaluesareall greaterthantheAVEscoresindicatingconvergentvalidityHairetal.,2006. ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelations indicatingdiscriminantvalidityFornellandLarckercriterion-See:67. Theindicatorsinthereectivemeasurementmodelsreachsatisfactoryindica276

PAGE 298

Table66:SNSCostMeasurementModel ConstructAVECompositeReliability PROC_SETUP 0.59380.8530 PersRelationLoss 0.64160.8990 BrandRelationship 0.68100.8638 CONF 0.69100.8969 CONTINUANCE 0.71650.8831 PROC_COST 0.56460.8382 PROC_EVAL 0.60130.8181 PROC_LEARN 0.60410.8590 PU 0.68050.8939 SAT 0.93090.9758 torreliabilitylevels.Themeasurementmodelassessmentsubstantiatesthat alltheconstructmeasuresarereliableandvalid.Theresultingtablesshow thecompletemodelincludingthemoderatingeectsalthoughthemoderating eectsarelessthansmall. 277

PAGE 299

Table67:SNSCostMeasurementModel-DiscriminantValidity FornellandLarckerCriterion PROC SETUP Proc Relation Loss Brand Relationship CONFCONTINUANCE PROC COST PROC EVAL PROC LEARN PUSAT PROC SETUP 0.7706 Proc Relation Loss 0.3388 0.8010 BrandRelationship 0.1169 0.4935 0.8252 CONF 0.2251 0.6732 0.6519 0.8313 CONTINUANCE 0.2852 0.7313 0.6111 0.7117 0.8465 PROC COST 0.7094 0.4667 0.2222 0.3395 0.3702 0.7514 PROC EVAL 0.5408 0.1478 0.0776 0.1425 0.1993 0.4839 0.7754 PROC LEARN 0.6849 0.3126 0.1776 0.2003 0.2705 0.6428 0.5004 0.7772 PU 0.2808 0.7556 0.4734 0.5991 0.6303 0.3851 0.1103 0.2427 0.8249 SAT 0.1785 0.6674 0.6370 0.7879 0.7878 0.2520 0.1116 0.2015 0.5982 0.9648 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;non-diagonalelementsarethelatentvariablecorrelations. 278

PAGE 300

Table68:Indices GoodnessofFit NotApplicableinreective-formativemodels EndogenousConstructsR 2 Q 2 CONTINUANCE 0.71650.5049 ProceduralCost 0.99330.3242 RelationshipCost 0.99880.5006 Satisfaction 0.64570.6010 SatAndPU 0.99990.6079 Costs 0.99440.2170 PerceivedUsefulness 0.35890.2433 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. H.3StructuralModel ThestructuralmodelwasassessedtodeterminehowtheindependentconstructsfromISContinuancetheoryandconsumerswitchingcostspredictinformationsystemscontinuanceintention..Thepredictorsexplainedapproximately71.7%ofthevarianceR 2 incontinuanceintentionandisconsidered tohaveamoderatelevelofexplanation 30 ;themodelalsoexhibitspredictive relevanceQ 2 whereitsvalueis.5049scoresabovezeroindicatepredictive relevanceinPLSpathmodels.Satisfactionandperceivedusefulness,ina combinedmeasure,predictcontinuanceintentionwithastandardizedpath coecientof.480 t =17.4745.Allofthecostspredictcontinuanceintention withastandardizedpathcoecientof.399 t =14.321.Theresultsindicate thatISContinuanceexplainsmoreofthevariancethantheconsumerswitch costmodel.Coecientsthatarepositiveindicatethatareassociatedwith higherlevelsofcontinuanceintention. 30 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 279

PAGE 301

RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SatAndPU CONTINUANCE0.480317.4745 COSTS CONTINUANCE0.398614.321 age CONTINUANCE0.05433.8155 education CONTINUANCE0.03582.2875 gender CONTINUANCE0.0130.8378 OtherPathCoecients ProcRelationLoss Relationshipcosts0.730235.0565 BrandRelationship Relationshipcosts0.411316.4335 CONF PU0.599131.1508 CONF SAT0.670132.756 PROC_COST ProceduralCosts0.949613.9198 PROC_EVAL ProceduralCosts-0.12831.7679 PROC_LEARN ProceduralCosts0.15461.7937 PU SAT0.19688.9423 PU SatAndPU0.304910.3242 ProceduralCosts COSTS0.03151.4065 Relationshipcosts COSTS0.982487.6927 280

PAGE 302

RelationPath Coecient Tstatistic SAT SatAndPU0.787332.3647 PROCSETUP ProceduralCosts0.00020.0018 TotalEects ContinuancePathCoecients SatAndPU CONTINUANCE0.480317.4745 CONF CONTINUANCE0.385616.7192 SAT CONTINUANCE0.378114.4612 Relationship_costs CONTINUANCE0.391614.3975 COSTS CONTINUANCE0.398614.321 PersRelationLoss CONTINUANCE0.285912.9775 PU CONTINUANCE0.220812.6321 BrandRelationship CONTINUANCE0.161111.265 age CONTINUANCE0.05433.8155 education CONTINUANCE0.03582.2875 PROC_COST CONTINUANCE0.01191.3768 ProceduralCosts CONTINUANCE0.01261.3556 PROC_EVAL CONTINUANCE-0.00161.0628 PROC_LEARN CONTINUANCE0.00190.9744 gender CONTINUANCE0.0130.8378 281

PAGE 303

RelationPath Coecient Tstatistic PROC_SETUP CONTINUANCE00.0015 OtherPathCoecients Relationship_costs COSTS0.982487.6927 CONF SatAndPU0.80373.349 CONF SAT0.787967.2319 PersRelationLoss COSTS0.717335.5389 PersRelationLoss Relationship_costs0.730235.0565 SAT SatAndPU0.787332.3647 CONF PU0.599131.1508 BrandRelationship Relationship_costs0.411316.4335 BrandRelationship COSTS0.404115.5954 PU SatAndPU0.459815.4002 PROC_COST ProceduralCosts0.949613.9198 PU SAT0.19688.9423 PROC_LEARN ProceduralCosts0.15461.7937 PROC_EVAL ProceduralCosts-0.12831.7679 PROC_COST COSTS0.02991.4302 ProceduralCosts COSTS0.03151.4065 PROC_EVAL COSTS-0.0041.0908 282

PAGE 304

RelationPath Coecient Tstatistic PROC_LEARN COSTS0.00491.002 PROC_SETUP ProceduralCosts0.00020.0018 PROC_SETUP COSTS00.0015 283

PAGE 305

Figure68:BaseModelAndCostsPathCoecientsonSNSContinuance 284

PAGE 306

ISpecicAlternativeProductEects Anexaminationofspecicservicealternativeswasconductedtodetermine whetherthenumberofsocialnetworkingsitesauserusesfromthesetof Instagram,Pinterest,TumblrandTwitter,andhowattitudesaboutthesame setofsitespredictscontinuanceintention. I.1NumberofsocialnetworkingsiteAnalysis Todeterminewhetherthenumberofsocialnetworkingsitesauserhascan predictcontinuanceintentionthesurveyrespondentwasaskediftheyused Instagram,Pinterest,TumblrandTwitter.Thetotalnumberofsitesused wasthencalculated.Themajorityofsurveyrespondentsusedatleastone othersocialnetworkingsitefromthecompetingset.5%.Thecoecient ofdeterminationis.0512andindicatesthatapproximately5.1%ofthevariance isexplainedbythemodelandisconsideredweak. AllaveragevarianceextractedAVEvaluesareabovethe.50threshold, providingsupportforconvergentvalidityHairetal.,2006.Compositereliabilityvaluesareallgreaterthanorequalto.7providingevidenceofthe construct'sinternalconsistencyreliabilityHairetal.,2006. FornellandLarckercriterionisnotpresentedasallthepredictorsin thismodelaresingleitemmeasures.Numberofsitesusedpredictscontinuance intentionwithastandardizedpathcoecientof-0.0706t-statistic=1.0929 andindicatesthatthereisnotastatisticallysignicantrelationshipwiththe numberofsitesusedandcontinuanceintention,i.e.thenumberofsitesa surveyrespondentusesdoesnotpredictwhetherthepersonwillcontinuetouse Facebookordiscontinueuse.Thecovariates,genderandage,hadstatistically signicanteectsinpredictingcontinuanceintention,andeducationhadno 285

PAGE 307

Table70:NumberofSocialNetworkingSitesUsed Number ofSites Used Count % 0 384 29.49% 1 395 30.34% 2 271 20.81% 3 198 15.21% 4 54 4.15% Total 1302 100% Table71:SitesUsedMeasurementModel ConstructAVECompositeReliability CONTINUANCE 0.71490.8824 statisticallysignicanteectinpredictingcontinuanceintention. Table72:SitesUsedStructuralModel GoodnessofFit 0.2081 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 Continuance .0512.0357 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. 286

PAGE 308

Table73:PathCoecients RelationPath Coecient Tstatistic TheoreticalModel ContinuancePathCoecients SNSUse Continuance-0.07061.0929 OtherPathCoecients TUMBLR SNSUse0.94073.4590 Pinterest SNSUse-0.43721.0852 Twitter SNSUse0.15950.4233 Instagram SNSUse-0.05790.1614 TotalEects Sample Mean M T Statistics ContinuancePathCoecents gender Continuance0.17254.0262 age Continuance0.10752.3474 TUMBLR Continuance-0.07871.4126 Pinterest Continuance0.03490.8343 SNSUse Continuance-0.05580.6505 Twitter Continuance-0.01270.2208 Instagram Continuance0.00690.1025 education Continuance0.00220.0705 OtherPathCoecients TUMBLR SNSUse0.41051.5834 Pinterest SNSUse-0.00380.8878 Twitter SNSUse0.22810.4847 Instagram SNSUse0.10310.1683 287

PAGE 309

I.2AlternativesbySpecicProduct I.2.1Surveyrespondentswhousedallfouralternativeproducts I.2.2ModelDescription ThemodelincludesalternativeattractivenessmeasuresforalternativeproductsInstagram,Pinterest,TumblrandTwittertopredictcontinuanceintention.TheindependentconstructsInstagramAlternative,PinterestAlternative,TumblrAlternativeandTwitterAlternativetopredictinformationsystemscontinuanceintention.Surveyrespondentswhoperceiveanalternative producttobeabetteralternativetoFacebookaretheorizedtohavedirect eectsonFacebookcontinuance.Theusersinthissampleuse all fourproducts-thesamplesizeis54users.Thesamplerepresentsapproximately6%of theuserswhohaveatleastonealternativesocialnetworkingsite,and4%of thetotalsurveyrespondents.Theadvantageofusingthissampleisthatno replacementvaluesformissingvaluesarenecessaryandtheabilitytopredict howusersofmultiplesiteswillcontinueordiscontinueuseofFacebook. I.2.3MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliability valuesareallgreaterthanorequalto.7providingevidenceoftheconstruct's internalconsistencyreliabilityHairetal.,2006.CompositereliabilityvaluesareallgreaterthantheAVEscoresindicatingconvergentvalidityHair etal.,2006.ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelationsindicatingdiscriminantvalidityFornellandLarcker 288

PAGE 310

Table74:AlternativesbySpecicProductMeasurementModel ConstructAVECompositeReliability Continuance 0.74370.8969 Instagram 0.7330.8917 Pinterest 0.84230.9412 TUMBLR 0.86470.9504 Twitter 0.74540.8973 Table75:AlternativesbySpecicProductMeasurementModel-Discriminant Validity FornellandLarckerCriterion Continuance Instagram Pinterest TUMBLRTwitter Continuance 0.8624 Instagram -0.5791 0.8562 Pinterest -0.5821 0.6850 0.9178 TUMBLR -0.7284 0.6153 0.6102 0.9299 Twitter -0.6553 0.6250 0.5596 0.7417 0.8634 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. criterion-See:75.Themeasurementmodelassessmentsubstantiatesthatall theconstructmeasuresarereliableandvalid. I.2.4StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructstopredictinformationsystemscontinuanceintention.Thepredictors InstagramAlternative,PinterestAlternative,TumblrAlternativeandTwitter Alternativeexplainapproximately60.7%ofthevarianceR 2 incontinuance intentionandisconsideredtohaveamoderatelevelofexplanation 31 ;the modelalsoexhibitspredictiverelevanceQ 2 whereitsvalueis.3864scores abovezeroindicatepredictiverelevanceinPLSpathmodels.TumblrAlter31 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 289

PAGE 311

Table76:AlternativesbySpecicProductStructuralModel GoodnessofFit .6908 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 Continuance 0.6072.4563 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. nativehadthestrongestpredictiveabilityforsatisfactionwithastandardized pathcoecientof-.440 t-statistic =2.740followedbyTwitterAlternative withastandardizedpathcoecientof-.217t-statistic=1.387,Pinterest -.113,t-statistic=.792andInstagram-.089,t-statistic=.4994.OnlyTumblrAlternativeisastatisticallysignicantpredictor.Thecoecientsareall negativeindicatethatincreasingperceptionthatanalternativeproductisa viablesubstituteindicateslowerlevelsofcontinuanceintention. I.2.5Surveyrespondentswhousedatleastoneoffouralternative products I.2.6ModelDescription ThemodelincludesalternativeattractivenessmeasuresforalternativeproductsInstagram,Pinterest,TumblrandTwittertopredictcontinuanceintention.TheindependentconstructsInstagramAlternative,PinterestAlternative,TumblrAlternativeandTwitterAlternativetopredictinformationsystemscontinuanceintention.Surveyrespondentswhoperceiveanalternative producttobeabetteralternativetoFacebookaretheorizedtohavedirecteffectsonFacebookcontinuance.Theusersinthissampleuse all fourproducts -thesamplesizeis918users.Thesamplerepresentsapproximately70.5%of 290

PAGE 312

Table77:PathCoecients RelationPath Coecient Tstatistic TheoreticalModel TUMBLR Continuance-0.43982.7401 Twitter Continuance-0.21671.3873 Pinterest Continuance-0.11330.792 Instagram Continuance-0.0890.4994 TotalEects Sample Mean M T Statistics TUMBLR Continuance-0.43982.7401 education Continuance0.15891.4531 Twitter Continuance-0.21671.3873 Pinterest Continuance-0.11330.7920 Instagram Continuance-0.0890.4994 age Continuance-0.01350.1327 gender Continuance-0.0150.1304 291

PAGE 313

Figure69:SpecicProductPathCoecientsonSNSContinuance thetotalsurveyrespondents.Theadvantageofusingthissampleisthatthe sampleismuchlargersosmallerbutstillstatisticallysignicanteectscanbe found.Missingvaluesarereplacedwiththemeanoftheconstructtopredict howusersofmultiplesiteswillcontinueordiscontinueuseofFacebook;using themeanwillnotbiastheresult,butstatisticalsignicancemaybelowered toaccountfortheuseofthemeanvalue. I.2.7MeasurementModel Themeasurementmodelisassessedforreliabilityconstructindicatorreliabilityandinternalconsistencyandvalidityconvergentanddiscriminant.All averagevarianceextractedAVEvaluesareabovethe.50threshold,providingsupportforconvergentvalidityHairetal.,2006.Compositereliability 292

PAGE 314

Table78:AlternativesbySpecicProductMeasurementModel ConstructAVECompositeReliability Continuance 0.70610.8779 Instagram 0.81510.9295 Pinterest 0.81440.9292 TUMBLR 0.80960.9272 Twitter 0.82030.9316 Table79:AlternativesbySpecicProductDiscriminantValidity FornellandLarckerCriterion Continuance Instagram Pinterest TUMBLRTwitter Continuance 0.8403 Instagram -0.3133 0.9028 Pinterest -0.3854 0.3310 0.9024 TUMBLR -0.2835 0.1841 0.1866 0.8998 Twitter -0.5389 0.3433 0.3651 0.2549 0.9057 Note:DiagonalelementsinboldarethesquarerootoftheAVEs;nondiagonalelementsarethelatentvariablecorrelations. valuesareallgreaterthanorequalto.7providingevidenceoftheconstruct's internalconsistencyreliabilityHairetal.,2006.CompositereliabilityvaluesareallgreaterthantheAVEscoresindicatingconvergentvalidityHair etal.,2006.ThesquarerootsoftheAVEareallgreaterthanthelatentvariablecorrelationsindicatingdiscriminantvalidityFornellandLarcker criterion-See:79.Themeasurementmodelassessmentsubstantiatesthatall theconstructmeasuresarereliableandvalid. I.2.8StructuralModel Thestructuralmodelwasassessedtodeterminehowtheindependentconstructstopredictsocialnetworkingsitecontinuanceintention.Thepredictors InstagramAlternative,PinterestAlternative,TumblrAlternativeandTwitter 293

PAGE 315

Table80:AlternativesbySpecicProductStructuralModel GoodnessofFit 0.5427 GoodnessofFitmeasures:GoF small =.1,GoF medium =.25GoF large =.36Wetzelsetal.,2009 EndogenousConstructsR 2 Q 2 Continuance 0.37130.2604 R2coecientofdetermination Q2predictiverelevance.Scoresabovezeroindicatepredictiverelevance. Alternativeexplainapproximately37.1%ofthevarianceR 2 incontinuance intentionandisconsideredtohaveamoderatelevelofexplanation 32 ;themodel alsoexhibitspredictiverelevanceQ 2 whereitsvalueis.2604scoresabove zeroindicatepredictiverelevanceinPLSpathmodels.TwitterAlternative hadthestrongestpredictiveabilityforsatisfactionwithastandardizedpath coecientof-0.3889 t-statistic =12.2996followedbyPinterestAlternative withastandardizedpathcoecientof-0.1830t-statistic=5.5089,TUMBLR-0.1256,t-statistic=3.7429andInstagram-0.0962,t-statistic=3.0192. Allalternativesareconsideredstatisticallysignicantpredictor.Thecoecientsareallnegativeindicatethatincreasingperceptionthatanalternative productisaviablesubstituteindicateslowerlevelsofcontinuanceintention. 32 Hairetal.-R 2 of0.75issubstantial,0.50ismoderate,and0.25isweak. 294

PAGE 316

Table81:PathCoecients RelationPath Coecient Tstatistic TheoreticalModel Twitter Continuance-0.388912.2996 Pinterest Continuance-0.18305.5089 TUMBLR Continuance-0.12563.7429 Instagram Continuance-0.09623.0192 TotalEects Sample Mean M T Statistics Twitter Continuance-0.388912.2996 Pinterest Continuance-0.1835.5089 gender Continuance0.12284.4128 TUMBLR Continuance-0.12563.7429 Instagram Continuance-0.09623.0192 age Continuance0.03771.395 education Continuance0.02540.9392 295

PAGE 317

Figure70:SpecicProductatleast1alternateproductPathCoecientson SNSContinuance 296