pdf reader to byte array c# : How to add image to pdf in preview SDK control service wpf azure asp.net dnn online-reviews0-part1772

Do online reviews matter? — An empirical investigation of panel data
Wenjing Duan
a,
,Bin Gu
b,1
,Andrew B. Whinston
b,2
a
Funger515,SchoolofBusiness,TheGeorgeWashingtonUniversity,Washington,DC20052,UnitedStates
CBA5.202,McCombsSchoolofBusiness,TheUniversityofTexasatAustin,Austin,Texas78712,UnitedStates
a r t i c l e
i n f o
a b s t r a c t
Articlehistory:
Received13May2007
Receivedinrevisedform26March2008
Accepted2April2008
Availableonline10April2008
This study examines the persuasive effect and awareness effect of online user reviews on
movies'dailyboxofficeperformance.Incontrasttoearlierstudiesthattakeonlineuserreviews
asanexogenousfactor,weconsiderreviews bothinfluencingandinfluencedbymoviesales.
Theconsiderationof theendogenousnatureofonline userreviews significantlychanges the
analysis. Ourresultshows thattheratingofonlineuserreviewshasnosignificantimpacton
movies'boxoffice revenuesafteraccountingfortheendogeneity, indicatingthatonlineuser
reviewshavelittlepersuasiveeffecton consumer purchase decisions. Nevertheless, we find
thatboxofficesalesaresignificantlyinfluencedbythevolumeofonlineposting,suggestingthe
importance of awareness effect. The finding of awareness effect for online user reviews is
surprisingas onlinereviews underthe analysis arepostedtothesame websiteandarenot
expectedtoincreaseproductawareness. Weattributetheeffecttoonlineuserreviewsasan
indicatorof theintensity ofunderlyingword-of-mouththatplaysadominantroleindriving
boxofficerevenues.
©2008ElsevierB.V.Allrightsreserved.
Keywords:
Onlineuserreviews
Onlinefeedbacksystems
Word-of-mouth
Productsales
Motionpicture
Simultaneousequations
1.Introduction
On September 12, 2004, an anonymous consumer dis-
closed in his online journal that the ubiquitous, U-shaped
Kryptonitelock couldbeeasilyopenedwithaballpoint pen
[26].Withindays, thenewspenetratedvirtuallyeveryblog
(short for “web logs,” where individuals publish their
personal diaries) and Internet chat room. The onlineword-
of-mouth frenzy forced Kryptonite to announce a free
exchange program on September 22 for any affected lock.
The Kryptonite incident demonstrates the sheer power of
onlineword-of-mouthtoday.Withthehelpof the Internet,
information is no longer only controlled by news media or
large businesses. Everyone can share their thoughts with
millions of Internet users and influence others' decisions
throughonlineword-of-mouth.
Word-of-mouthhas beenrecognizedas one ofthe most
influential resources of information transmission since the
beginningofsociety,especiallyforexperiencegoods[21,22].
However, conventional interpersonal word-of-mouth com-
munication is only effective within limited social contact
boundaries,and the influencediminishesquicklyover time
and distance [17]. The advances of information technology
haveprofoundlychangedthewayinformationistransmitted,
andhavetranscendedthetraditionallimitationsofword-of-
mouth.Consumerscannoweasilyandfreelyaccessinforma-
tion and exchange opinions on companies, products, and
servicesonanunprecedentedscaleinrealtime.
Online customer review systems are one of the most
powerful channels to generate online word-of-mouth [11].
With the popularity of online word-of-mouth activity, an
increasingnumberofbusinesseshavestartedtoofferonline
word-of-mouth services.Amazon.com is well-knownforits
extensive customer review systems. Major television net-
workssuchasABC,CBS,andNBCsponsorUsenetnewsgroups
to elicit viewers to talk about their programs and shows.
Similarly,almosteverystudioandfilmdistributorhasutilized
the Web as a critical marketingvenue bycreating websites
and discussion forums for their movies [18]. The Web has
becomea mediumto reachaudiences directlyandgenerate
buzzeswithtremendousefficiencyandflexibility,regardless
DecisionSupportSystems45(2008)1007–1016
⁎ Correspondingauthor.Tel.:+12029943217;fax:+2029945830.
E-mailaddresses:wduan@gwu.edu(W.Duan),
Bin.Gu@mccombs.utexas.edu(B.Gu),abw@uts.cc.utexas.edu(A.B.Whinston).
1
Tel.:+15124711582;fax:+15124710587.
Tel.:+15124717962;fax:+15124710587.
0167-9236/$–seefrontmatter©2008ElsevierB.V.Allrightsreserved.
doi:10.1016/j.dss.2008.04.001
ContentslistsavailableatScienceDirect
Decision Support Systems
journal homepage: www.elsevier.com/locate/dss
How to add image to pdf in preview - insert images into PDF in C#.net, ASP.NET, MVC, Ajax, WinForms, WPF
Sample C# code to add image, picture, logo or digital photo into PDF document page using PDF page editor control
how to add an image to a pdf; how to add image to pdf file
How to add image to pdf in preview - VB.NET PDF insert image library: insert images into PDF in vb.net, ASP.NET, MVC, Ajax, WinForms, WPF
Guide VB.NET Programmers How to Add Images in PDF Document
add signature image to pdf; acrobat insert image into pdf
of geographic boundaries. The most successful example of
leveragingonlineword-of-mouthasthemajormarketingtool
isthe“megahit”TheBlairWitchProject(1999).Themoviewas
initiallyseenasateenagefrightflickwitha“tiny”production
budget of $60,000. Thanks to the large-scale discussions
generatedontheWeb,iteventuallybecameahugeboxoffice
success($248 millionworldwide).
InspiteofthewidespreadbeliefthattheInternetmayact
as a huge “megaphone” in promoting product sales, few
literaturehasprovidedevidencethatonlineword-of-mouth,
suchasproductreviewsandrecommendations,playsanyrole
in influencing consumers' choices and purchase decisions.
Therehavebeenanumberofrecentstudiesinvestigatingthe
impact of online word-of-mouth on product sales [5–
7,12,19,21,32]. However, theresultsaremixed. Someof the
research supports the view that online user review has a
significant impact on sales [7], while other research chal-
lengessuchaview[6,21,32].
The challenges and confusion mainly come from three
aspects. First,studiesdifferintheirview oftheinfluenceof
onlineuserreviews.Somefocusonuserreviews'persuasive
effect that influence a consumer's assessment of product
quality [5–7,19,32], while others focus on user reviews'
awareness effect that increase product awareness among
consumersthroughdispersion[12,21].Second,manystudies
treat word-of-mouth as exogenous [6,7,12,19]. Word-of-
mouth, however, isnot only the driving forces ofconsumer
purchasebutalsotheoutcomeofproductsales.Thecausality
between product sales and word-of-mouth works in both
directions.Ignoringthedualinfluencerandindicatorrolesof
word-of-mouth is one ofthemaincauses oftheconfusion.
Third, many researchers conduct their analyses in a cross-
sectionalcontext[5–7,32].Across-sectionalsetting,however,
cannot control for the intrinsic product heterogeneity. In
particular,itcannotexplainwhetherthedifferenceinproduct
salesisduetotheunobserveddifferencesinproductquality
ortheeffectofword-of-mouth.
Givenpreviouslimitationsandchallenges,weassessboth
thepersuasiveeffect andtheawarenesseffectofonlineuser
reviewsinthisstudyusingasimultaneousequationsystemto
fully capture the dual nature of online user reviews. In
addition,weexaminetherelationshipbetweenonlineword-
of-mouthandproductsalesinapaneldatasettingtocontrol
forindividualheterogeneities.Weutilizeonlineuserreviews
for motion pictures as our research context because rapid
spreadofword-of-mouthhas beenhistoricallyconsidereda
critical factor for financial success by the entertainment
industry[10,25,34,37].A recentreportbyForresterResearch
foundthatapproximately 50%ofyoungInternet surfersrely
on word-of-mouth recommendations to purchase CDs,
movies,videosorDVDs[21].Weconstruct apanel data set
includingdailyonlineuserreviewsanddailymovieboxoffice
sales.Oursimultaneousequationsystemtakesfulladvantage
of the panel data structure and specifies causality in both
directions.Usingthesimultaneousequationsystem,weseek
toclarifytheconfusioninpriorstudiesbyprovidingmeasures
ofthetrueeffectofonlineuserreviews.
Ourfindingschallengeconventional thinkingbyshowing
thatuserratingsdonotaffectmoviesalesaftercontrollingfor
endogeneity of user reviews and product heterogeneity,
suggesting little persuasive effect for online user reviews.
Thisresultisconsistentwithearlierfindingswithregardtothe
impact ofmovie critics.Eliashberg andShugan[16] showed
that movie critics' ratings are predictors of movie perfor-
mance,buttheydonotinfluencemovieperformance.Wefind
that, in the onlineuserreview setting,userratings share a
similarcharacteristic.Theyreflectmoviequality,buttheydo
notinfluencemoviesales.Thisresultindicatesthatconsumers
arefullycapableofinferringthetruequalityofamoviefrom
onlinereviewswithoutbeinginfluencedbytheratingsofthe
reviews per se.Our analyses also show that the numberof
postings is significantly correlated with movie sales after
taking into account of the causality issue, indicating the
presence of significant awareness effect. The finding is
surprising as online user reviews are posted to the same
website and, as a result, not expected to increase product
awareness.Weattributethe awareness effectto onlineuser
reviewsasanindicatoroftheintensityofunderlyingword-of-
mouth which plays a dominant role in driving box office
revenues.Moreover,wefindthatthenumberofuserreviews
onlineissignificantlydrivenbymoviesales,confirmingthat
userreviewisnotonlyaninfluencerof,butalsoanindicatorof
sales.Inaddition,ourresultsshowthatthenumberofpostings
is positively autocorrelated, demonstrating the self-driving
essenceofonlineword-of-mouth.Finally,fromthedataofthe
first two weeks, we obtainedsignificantly different results.
Suchadifferencecapturestherapidly-changingnatureofthe
effectofword-of-mouthontheInternet.
Ourpaperenrichestheempiricalresearchontheimpactof
online word-of-mouth. From the methodology perspective,
we demonstrate theimportance of controlling forthe dual
role of online word-of-mouth as an influencer and an
indicatorofproductsales,andtheimportanceofcontrolling
fortheunobservedbutinherentproductheterogeneityinthe
analysis of online word-of-mouth. From the managerial
perspective, weidentifyboththepersuasiveandawareness
effect of online user review. We show that consumers are
rational ininferringmoviequalityfromonline userreviews
withoutbeingundulyinfluencedbytherating,thuspresent-
ing a challenge to businesses that try to influence sales
through“planting”positiveproduct reviews.Ourfindingsof
awareness effect,also suggest that the underlying word-of-
mouth process could have a significant impact on sales,
suggesting that businesses should embrace and facilitate
word-of-mouthactivities.
The rest of the paper is organized as follows. The next
section provides the literature review followed by the
discussion of research objectives and hypotheses. We then
describeoursource of dataand the empirical model. Main
findingsarepresentedanddiscussednext,andthepaperends
withadiscussionoflimitationsandfutureresearch.
2.Literaturereview
Researchontheimpactofinterpersonalcommunicationis
common in the economics literature. The early studies of
Learning from Others provide evidence that word-of-mouth
communicationmayaffectothers'decisionsindifferentsocial
contexts [33]. Smallwood and Conlisk [42] showed that a
product may capture the entire market regardless of its
qualitythroughsometypeoflearningprocess.Banerjee[2,3]
presented two models indicating that people place such a
1008
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
C# WinForms Viewer: Load, View, Convert, Annotate and Edit PDF
Convert PDF to Tiff image (.tif, .tiff). • Convert PDF to HTML (.htm, .html). PDF Annotation. • Add sticky notes to PDF document in preview.
how to add a picture to a pdf file; how to add image to pdf acrobat
C# WPF Viewer: Load, View, Convert, Annotate and Edit PDF
PDF to Tiff image (.tif, .tiff). • Convert PDF to HTML (.htm, .html). PDF Annotation. • Add sticky notes to PDF document. • Highlight PDF text in preview.
adding images to pdf files; adding images to pdf forms
significant weightonotherpeople'sopinionsthattheymay
even ignore their own private information. Kirman [27]
demonstrated asimilarresult that Learningfrom Otherscan
cause a significant differentiation in market share between
two products withthe same quality.Ellisonand Fudenberg
[17]studiedasimplemodelofword-of-mouthcommunica-
tion and found that social learning is often most efficient
whencommunicationbetweenagentsislimited.
A number of previous empirical studies have been
conductedto examinethe impact ofinterpersonal word-of-
mouth,butresultsaremixed.KatzandLazarsfeld[24]found
that word-of-mouth plays the most important role in
influencing the purchase of household goods and food.
Coleman et al. [9]usedword-of-mouth toexplainadoption
oftetracyclineamongphysicians.FosterandRosenzweig[20]
attributedadoptionofhigh-yieldvarietiesofseedsbyfarmers
toword-of-moutheffect.However,VandenBulteandLilien
[45]castdoubtontheroleofword-of-mouthasasalesdriver.
Theyre-examinedtheanalysisbyColemanetal.[9]andfound
thatmarketingefforts,notword-of-mouth,playsadominant
roleinphysicians'adoptiondecision.
The utilization of the Internet as a venue for publicizing
feedback and recommendations on products and businesses
hasgainedgrowingpopularity.However,littleisknownaboutif
onlineword-of-mouthhasanyinfluenceonconsumerpurchase
decisions.Dellarocas[11]providesacomprehensivereviewof
thecurrentprogressandchallengeofstudyingonlinefeedback
systems.Chatterjee[5]usedsurveystoexaminetheimpactof
negativeonlineuserreviews.Theresultsindicatethattheuseof
onlineword-of-mouthinformationdependsonaconsumer's
intentto purchase online.Consumers who aremorefamiliar
withaspecificretailerare lesslikelyaffectedby thenegative
reviews.Chenetal.[7]studiedtheunderlyingpatternsofonline
consumerpostingbehavior throughonline reviewsforauto-
mobiles.They foundthat automobile characteristics such as
qualityandpricehaveasignificantimpactonusers'inclination
topost.Chenetal.[6]empiricallyinvestigatedtheimpactsof
bothonlineuserreviewsandrecommendationinformationon
book sales in Amazon.com from the consumer search cost
perspective.Theyfoundthatrecommendationsarepositively
associatedwithsales,whileconsumerratingsarenotfoundto
berelatedtosales.Theyalsofoundthatrecommendationsare
more important for less popular books. Li and Hitt [28]
investigated the self-selection effect and information role of
onlineproductreviews.Byanalyzing thedataofonline book
reviews,theyfoundthataverageratingdeclinesovertimeand
earlyconsumerreviewsdemonstrate positivebiasdueto the
self-selectioneffect.
Onlineuser reviewscan influenceproduct salesthrough
either awareness effects or persuasive effects. Awareness
effects indicate that reviews convey the existence of the
product and thereby put it in the choice set of consumers.
Persuasive effects, in contrast, are to shape consumers'
attitudesandevaluationtowardstheproductandultimately
influence their purchase decision. These two effects have
been studiedintensively inpriorliteraturein marketing on
the effect of advertising. It is found that advertising has a
significantpositiveeffectonbrandawareness,butnoeffecton
perceivedquality[8]. A recent work by Godes andMayzlin
[21] focused on measuring g the e influence e of f dispersion of
word-of-mouth, a concept closely relatedto the awareness
effect.Theyexaminedword-of-mouthcommunicationforTV
showswithinandacross different Usenet newsgroups.They
found that dispersion of word-of-mouth is significantly
correlatedwithaTVshow'sperformanceearlyon,whilevo-
lumeexhibitssignificanceonlyinlaterperiods.Ascrosscross-
newsgroup dispersion creates more awareness than within
newsgroup dispersion, the results indicate that awareness
effect of online word-of-mouth has a significant influence.
Theirempiricalanalysestookintoaccountthedualnatureof
word-of-mouthcommunicationasbothaninfluencerandan
outcome. However, the system of seemingly unrelated
regressions(SUR)doesnot handletheendogeneityofword-
of-mouth when it acts as aninfluencerof the sales. In this
paper,ourfocusistoexamineboththepersuasiveeffectand
awareness effect ofonline userreviews which is critical to
understandtheinfluenceofonlineuserfeedbacksystems.We
usea simultaneousequationsystem that fullycharacterizes
theinterdependentrelationshipbetweenonlineuserreviews
and movie revenues. Moreover, we use movie review data
that are essentially different from Usenet newsgroup con-
versations.Inadditiontomeasuringvolume,wemeasureuser
ratingsthatareoftenconsideredadrivingforceofconsumers'
productchoice,whichisnotavailableforUsenetnewsgroup
dataandthushasnotconsideredbyGodesandMayzlin[21].
Movie industry experts appear to agree that word-of-
mouthisacritical factorunderlyingamovie'sstayingpower
whichleadstoits ultimatefinancial success.However,prior
research on the relationship between word-of-mouth and
marketperformanceofmotionpicturesissurprisinglylimited.
NeelameghamandChintagunta[35]empiricallyassessedthe
relationshipbetweenword-of-mouth and weekly revenues,
butfailedtoobtainanysignificantresults.Theyattributedthe
failure to the inadequacy of the measurement of word-of-
mouth,whichmayalsoexplainthelackofsignificantresultsof
theword-of-moutheffect inthe previousliterature. Elberse
andEliashberg[15]usedrevenuesperscreenintheprevious
weekasaproxyofword-of-mouthintheiranalysisofdemand
andsupply ofmotionpictures.They foundsuchameasure-
ment ofword-of-mouthto be a key predictor of boxoffice
revenues. Dellarocas et al. [12] employed a modified Bass
Diffusionmodeltostudytheeffectsofonlineuserreviewsin
forecasting movie revenues. Their results showed that the
early online user review information can help generate
accurate future forecasts of movie revenues. Reinstein and
Snyder [40] apply a difference-in-difference approach to
uncover the impact of movie critics on sales. They have
identifiedmarginalpositiveinfluenceofmoviecriticsonthe
demand.Extendingearliermodels[4,16],Liu[32]examined
the relationship between online user feedback and movie
salesbasedon weekly dataregressions.Theresultssuggest
that word-of-mouth valence is not correlated with movie
sales, but online messagevolume is significantly correlated
withtheweeklymoviesales.Incontrasttothecross-sectional
andsingle-equationOLSsettingusedinLiu[32],weproposea
simultaneous equation panel data analysis in this study to
capturethedualnatureofword-of-mouthanditsinteraction
withsales.Inaddition,weuse daily dataas opposed to the
weeklydatatocapturetheunprecedentedspeedofinforma-
tiontransmissionontheInternet.
On the other hand, a range of studies have provided
evidenceforapositiverelationshipbetweencriticalreviews
1009
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
How to C#: Preview Document Content Using XDoc.Word
C# DLLs for Word File Preview. Add references: Get Preview From File. You may get document preview image from an existing Word file in C#.net.
attach image to pdf form; acrobat add image to pdf
How to C#: Preview Document Content Using XDoc.PowerPoint
Add necessary XDoc.PowerPoint DLL libraries into your created C# application as You may get document preview image from an existing PowerPoint file in C#.net.
add an image to a pdf; adding image to pdf in preview
and theatrical success [29,30,31,38,39,41,43,46]. Eliashberg
andShugan[16] triedto distinguishcritics'roleas“influen-
cers”,whoseopinions influencetheiraudienceandthusthe
boxoffice,fromtheirrolesas“predictors”,asmerelyaleading
indicator of their respective audience with no significant
influence on actual box office revenues. The authors found
that critical reviews correlate withlateandcumulativebox
officerevenuesbutdonothaveasignificantcorrelationwith
early box office performance. This finding implies that a
critical review is more likely to be a “predictor” than an
“influencer”.ArecentstudybySorensenandRasmussen[44]
evaluated the impact of New York Times book reviews on
sales. Their results suggest that “any publicity is good
publicity:”evennegativereviewsleadtoincreasesinsales.
3.Hypotheses
Inthisstudy,weaimto investigatetheimpactofonline
word-of-mouth on product sales. As a context for our in-
quiry, we choose online user reviews for motion pictures.
There are several reasons for choosing such a research
context.First,giventhatpricemayplayanimportantrolein
consumers' purchasing decisions and product satisfaction,
choosingthemovieindustryto studyonlineword-of-mouth
has its unique advantage. Movie ticket prices are typically
determinedinthelocal markets.Therefore,wecanruleout
thepossibilitythatpriceismediatingconsumers'purchasing
decisions. Second, word-of-mouth has traditionally been
considereda critical factorin influencing box office perfor-
mance,but thereis no consistent documented support. We
wouldaddresssuch ashortcoming by exploring theimpact
of online word-of-mouth on movie sales. Third, compared
with other products such as books and music CDs, which
usuallyhaveonly salesrank data,motionpictureboxoffice
sales data is publicly available, thus significantly reducing
measurement error. Finally, both online user reviews and
moviesalesarehigh-frequencydatathatcanbecollectedon
a daily basis. This provides sufficient observations for
empirical analysis.
Online user reviews have two effects on consumer
purchase decisions.First, most review sites allow a user to
provide bothanoverall rating (often denoted bya letter or
stargrade)andadetailedreview.Theratingandreviewcould
influenceotherconsumers'perceptionofproductquality.The
effect is equivalent to the persuasive effect studied in the
advertisingliterature.Duanetal.[14]reportthatabout22%of
usersofCNETsortproductsbyuserratings.Inaddition,prior
research also suggests that review ratings have a positive
impact on movie sales [40,41]. To measure the persuasive
effectofonlineuserreviews,weconsiderratingaspartofthe
measurementofword-of-mouth.Besidesinfluencingauser's
perception of product quality, online user reviews also
increase product awareness among consumers. The aware-
nesseffectismostsignificantwhenuserreviewsisdispersed
to online communities that are previously unaware of the
product.Foronlineuserreviewspostedonretailwebsite,we
expectnodirectawarenesseffectsinceconsumerswhovisit
theproduct pageareawareoftheproductinthefirstplace.
However,weexpect volume ofonlineuserreviewsto bean
indicatorof the intensity of the underlying word-of-mouth
effect. Previous theoretical and empirical researchprovides
support for the positive relationship between volume of
word-of-mouthandproductsales[21,32,33].Wethusderive
thefollowinghypotheses:
H1. Number of user postings has a positiveimpact on box
officerevenues.
H2. Userreviewratingshaveapositiveimpactonboxoffice
revenues.
We measure user review ratings from two different
perspectives,i.e.cumulative ratinganddailyrating.Cumulative
ratingisthearithmeticaverageofalltheprevioususerreview
ratings, while daily rating is the arithmetic average of user
review ratings posted in a single day. Cumulative rating
represents the summary score posted by the user review
website. Dailyratingreflects themost recentword-of-mouth
informationdisseminatedbyuserswhohavejustwatchedthe
movie.Consideringthatsomeoftheusersmayonlybrowsethe
overallratingwhileotherstendtoreadthemostrecentposts
morecarefully,weseparateH2intotwoparts.
H2a. Cumulativeuserreviewratinghasapositiveimpacton
boxofficerevenues.
H2b. Dailyuserreviewratingshaveapositiveimpactonbox
officerevenues.
DeVanyandWalls[13]exploredthedemandandsupply
dynamics andthe path ofthe distributionoffilmrevenues.
Their results indicate that weekly revenues are autocorre-
lated: more recent revenue increase is more likely to
experience additional revenue growth. Recent research by
Elberse and Eliashberg [15] verify that previous week per
screenrevenuesarepositivelycorrelatedwithcurrent week
sales. Such a positive autocorrelationof movie salesresults
fromthenatureoftheconsumerdemandofmotionpictures
[13].Whilethepreviousstudiesconstrainautocorrelationto
weeklydata,weextendittodailydatainthisstudy.
H3. Daily box office revenues are autocorrelated: a movie
whichexperiencedincreasingrevenuesinthepreviousdayis
more likely to experience additional growth than a movie
whichexperiencedgrowthinthedistantpast.
Word-of-mouthnotonlyleadstofuturesales,itisalsoan
outcomeofprevioussales.Forexample,Chenetal.[7]found
thatthenumberofonlinepostingsispositivelyrelatedtopast
automobilesalescontrollingforpriceandquality.Godesand
Mayzlin[21]illustratedthatthenumberofUsenetpostingsis
positivelycorrelatedwith aTV show's performance. Hence,
wehypothesize:
H4. Boxofficerevenueshaveapositiveimpactonthevolume
ofword-of-mouth.
Previousresearchofword-of-mouthindicatesthatvolume
ofword-of-mouthcommunicationpeaksinashortperiodof
time[21]. Such a buzz effect indicates that word-of-mouth
often leads to more word-of-mouth, suggesting a positive
autocorrelation.Thus,wehavethefollowinghypothesis:
H5. Dailynumberofuserreviewsisautocorrelated:arecent
increaseinthenumberofpostingsforamovieismorelikely
toelicitmoreuserreviewsinthefollowingday.
1010
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
VB.NET PDF File Compress Library: Compress reduce PDF size in vb.
enables compressing and decompressing in preview in ASP.NET to reduce or minimize original PDF document size Reduce image resources: Since images are usually or
how to add jpg to pdf file; how to add a jpeg to a pdf
How to C#: Preview Document Content Using XDoc.excel
Add necessary references: RasterEdge.Imaging.Basic.dll. Get Preview From File. You may get document preview image from an existing Excel file in C#.net.
add picture to pdf form; how to add photo to pdf in preview
Online user reviews are considered public goods since
providingreviewsontheInternetcostsreviewers'effort,but
benefits all the users.Public goods theory suggests that an
individualcontributeslesswhentherearesubstantialsources
of contribution [12,23]. Li and Hitt [28] evaluate the self-
selectioneffectinonlinebookreviews.Theyfoundthatusers
who are most likely to contribute post their reviews early,
leavingthosewhoarelesslikelytocontribute.Boththepublic
goods theory and the self-selection effect lead to the
followinghypothesis:
H6. Daily number of user reviews is negatively correlated
withthecumulativenumberofreviews.
4.Researchmethodology
4.1.Description ofthe data
Thedataforthisstudywas collectedfromthreesources,
Yahoo!Movies(YM:http://www.movies.yahoo.com),Variety.
com (Variety:http://www.variety.com), and BoxOfficeMojo.
com (Mojo: http://www.boxofficemojo.com). We matched
thelistofmovies,basedontheVariety'syear2003–2004box
officerank intheUSmarket,withthatonYM andMojofor
userreviewsanddailyboxofficeinformation.Bythetimewe
collectedthedata,moviesstillplayinginthetheaterwerenot
includedinthesample.Oursampleincludesthemoviesthat
have a complete history of user reviews from their release
dates and havethe completecorrespondingdailybox office
revenuedataaswell.
3
Thefinal datasetincludes71movies
releasedbetweenJuly2003 andMay2004.
For each movie, we collected the following information
fromYM:eachuserreview'syahooID,postdate,overallgrade,
gradeforstory,acting,direction,andvisual,andlengthofthe
full review. We also collected the Average User Grade and
AverageCriticGradepostedonYMbythedatewecollectedthe
data.
4
The letter grade of each individual user review was
convertedintoanumericalvaluebyassigning13toA+,12to
A… and 3 to D. This set of data was aggregated, for each
movie,byaddingupgradesandtakingthearithmeticaverage
foreachday.
5
Similarly,wecalculatedthecumulativeaverage
gradeforeachmovie.Wethusconstructedourmeasurement
ofcumulativeratinganddailyrating.Wealsosummedupthe
dailyandcumulativenumberofpostsforeachmovie.
Havingmatchedeachmovieby titleandreleasedate,we
collected the following information from Mojo: daily gross
revenues, daily rank, number of theaters engaged, average
revenue per theater, anddaily gross-to-date revenues. Sum-
mary data were also collected for each movie including
estimatedmarketing costs, production budget,MPAA rating,
producer,domesticgrossrevenues,andoverseagrossrevenues.
Usersstartpostingreviewsusuallyrightontheopeningdayof
themovie onYM,andreviewskeepemerginglong after the
movie's theater lifetime.
6
The specific post date information
providedbyYMforeachreviewcanbematchedwithdailybox
officerevenues.Table1presentsthesummarystatisticsforour
sample.Table2providesthedescriptionandmeasurementof
thekeyvariablesusedintheempiricalanalysis.
YM posts anassessment ofAverage User Grade which is
calculatedbasedonalltheuserratings.However,onlythose
withdetailedreviews will be postedonthewebsite,which
meanswewerenotableto collectalluserratings.
7
Totestif
suchasampleisarepresentativeofallthepostingsintermsof
thereviewgrade,wekepttrackof12new-releasemoviesfor
two weeks from their opening date. We collected all the
reviewspostedandtheupdatedAverageUserGradeprovided
by YM at 2–6 min intervals. At each interval, the average
reviewgradecalculatedbasedonthepostswithfullreviews
wascomparedwiththeAverageUserGradepostedonYM.We
observed almost perfect correlation for all the movies,
suggesting that the portion of user reviews shown on YM
sitesisagoodproxyofall theuserratings(this partofdata
andanalysiswillnotbereportedhere).
Similar to the pattern of box office life cycle of motion
pictures,weobservedthat,for most movies, thenumberof
user reviews skyrockets immediately aftertheopening and
drops significantly afterwards. Most movies are shown in
theaters for eight to ten weeks. Typically, the box office
receiptspeakatthetimeofinitialfilmrelease,followedbyan
exponential decay over time. Since word-of-mouth effect
decreasesovertimeveryquickly,itisessentialtocapturethe
dynamics in the early periods. We therefore constructed a
balanced panel data set of the 71 movies for the first two
weeks in this study. The descriptive statistics for some key
variablesofthefirsttwoweeksispresentedinTables3and4.
The tables show that the average numberof postings drops
significantlyfromweek1toweek2(themeanvalueofDAILY-
POSTchangesfrom88.54to36.95andthemaximumnumber
decreasesfrom633to231).Suchadifferenceimpliesthatmost
buzz is created in the early period and the intensity keeps
changingovertime,thusmakingtheimportanceofusingdaily
dataevenmoreevident.Tables5and6presentthepooleddata
correlation matrix of the key variables for week 1 and 2.
DAILYREVENUE
it
and DAILYPOST
it
in general have a strong
positive correlation(0.65 forthe first week and0.56 forthe
3
All the movies in our sample were nation-wide releases from their
openingdays.
Theaveragecriticgradewascalculated basedon13–15criticsreviews
invitedbyYMandpostedontheYMwebsite.
WecontactedYahoo!Moviestoverifysuchanumericaltransformation
oftheoriginallettergrade.Wewerenotified that theaverageusergrade
postedontheYahoo!Movieswebsiteiscalculatedinthesameway.
Table1
Summarystatistics
Variable
N
Mean
Std.Dev.
Min.
Max.
Budget(M)
64
46.06
32.17
4.00
150.00
Est.marketingcosts(M)
57
24.00
7.13
10.00
50.00
USgross(M)
71
66.16
51.21
10.39
377.03
Totaluserposts
71
1350.24
882.80
342.00
4562.00
Avg.UserGrade
71
8.89
1.02
6.00
11.00
Avg.CriticGrade
71
7.46
1.75
3.00
11.00
6
Reviewsthatwerepostedlatere.g.,after2–3monthsofmovie'srelease
datewereprobablybasedon experienceotherthaninthetheater,suchas
fromTV, DVD,orothervenues.Althoughwecollected allthereviews,we
onlyusedthosethatwerepostedduringthemovie'stheatrerunningtime.
Alotofusersonlyprovidealettergradeinsteadofafullreview,which
willnotbeshownonYM,butwillbeaggregatedintotheAverageUserGrade
onYMwebsite.
1011
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
How to C#: Set Image Thumbnail in C#.NET
VB.NET How-to, VB.NET PDF, VB.NET Word following steps below, you can create an image viewer WinForm Open or create a new WinForms application, add necessary dll
add an image to a pdf in preview; how to add an image to a pdf in reader
C# PDF remove image library: remove, delete images from PDF in C#.
Generally speaking, using well-designed APIs, C# developers can do following things. Remove Image from PDF Page Using C#. Add necessary references:
add image pdf acrobat; add image in pdf using java
secondweek).However, such acorrelation does not indicate
anydirectionofcausalityoranytimingsequenceinpriori.We
havetofullycharacterizetheirinterdependentrelationshipsin
the empirical analysis to uncover the true impact of user
reviews.
Table7providespairwisecorrelationsforsomecharacter-
isticvariablesinoursample.Similartotheobservationsofthe
first two weeks' data, the total numberof user posts hasa
relatively high positive correlation with US gross revenues
(0.68). This indicates the intrinsic connection between the
number of posts and box office performance, but does not
designateany causal relationship. In addition, we observed
thatAverageUserGradeandAverageCriticGradedonothavea
very high correlation (0.56), suggesting that online user
reviews may carry different information from that of the
professional criticalreviews.
4.2.Empiricalmodelspecification
As we are interested in the interdependence between
movies' box office revenues and online word-of-mouth
information, we developed the following two-equation
system:one equationwith daily revenuesasthe dependent
variable(therevenueequation)andonewithdailynumberof
postsas the dependent variable(the postequation).Sucha
systemcapturestheinteractionbetweenthetwo dependent
variablesovertime,andtheequationofdailynumberofposts
alsocharacterizesthedynamicsofthevolumeofonlineword-
of-mouth.
DAILYREVENUE
it
¼h
t
þa
1
DAILYPOST
it
þa
2
CUMURATING
i;t1
þa
3
DAILYREVENUE
i;t1
þa
4
WEEKEND
it
þA
i
þe
it
ð1Þ
DAILYPOST
it
¼g
t
þb
1
DAILYREVENUE
it
þb
2
DAILYPOST
i;t1
þb
3
CUMUPOST
i;t1
þb
4
WEEKEND
it
þq
i
þr
it
:
ð2Þ
Leti =1,…Nindex themovies.Fortherevenueequation,
DAILYREVENUE
it
denotesthedailygrossrevenuesofmovieiat
day t, and its one-day lagged variable is defined as
DAILYREVENUE
i,t− 1
.Sincetheadaptationofsupply(allocation
ofnumberoftheaters andscreens)to demandusually takes
placeinthelaterperiodofamovie'slifecycle,thereareunique
advantages of investigating early box office data without
worrying about the adjustment of the supply from movie
distributors.Inaddition,numberofmoviesshowingintheaters
usuallydoesnotchangeinagivenweek,thusthecompetition
andsubstitutioneffectsofvariousmoviesshowingonthesame
daycanbecontrolledthroughourdailypaneldatasetting.H1
suggests that the numberof postings ispositively correlated
withboxofficerevenue.Thus,weexpectα
1
N0.H3acknowl-
edges thatdailyboxofficerevenuesarepositivelyautocorre-
latedandthusweexpectα
3
N0.WedefineDAILYPOST
it
asthe
total numberofuserreviews postedformoviei at dayt and
DAILYPOST
i,t−1
asthetotalnumberofuserreviewspostedfor
movieiatdayt−1.
CUMURATING
i,t−1
representsthecumulativeaverageuser
review grade ofmoviei up to day t −1.Since YM provides
Table2
Variables,descriptions,andmeasures
Variable
Descriptionandmeasure
DAILYREVENUE
it
Dailyrevenueformovieiindayt
(inthousands,USdollars)
DAILYREVENUE
i,t−1
Dailyrevenueformovieiindayt−1
(inthousands,USdollars)
CUMUPOST
it
Cumulativenumberofreviewsposted
formovieiuntildayt
DAILYPOST
it
Numberofuserreviewspostedfor
movieiindayt
DAILYPOST
i,t−1
Numberofuserreviewsposted
formovieiindayt−1
CUMURATING
i,t−1
CumulativeAverageUserGrade
formovieiuntildayt−1
DAILYRATING
i,t−1
DailyAverageUserGradeformovie
iuntildayt−1
WEEKEND
it
Adummyvariableindicatingifdaytisa
weekend(codedas1ifdayisFriday,
Saturday,andSunday,0otherwise)
Table3
Week1descriptivestatistics
Variable
N
Mean
Std.Dev.
Min.
Max.
DAILYREVENUE(M)
497
3.69
4.00
0.15
34.45
CUMURATING
497
9.69
1.32
5.85
12.20
DAILYRATING
497
9.58
1.50
3.33
12.86
CUMUPOST
497
435.98
392.73
3.00
1,958.00
DAILYPOST
497
88.54
94.46
3.00
633.00
Table4
Week2descriptivestatistics
Variable
N
Mean
Std.Dev.
Min.
Max.
DAILYREVENUE(M)
497
2.15
2.34
0.089
19.15
CUMURATING
497
9.59
1.35
6.37
11.78
DAILYRATING
497
9.42
1.73
1.33
12.71
CUMUPOST
497
794.41
564.22
77.00
2,575.00
DAILYPOST
497
36.95
31.53
2.00
231.00
Table5
Week1correlationmatrix
Variable
DAILYREVENUE
CUMURATING
i,t−1
DAILYRATING
i,t−1
CUMUPOST
i,t−1
DAILYPOST
DAILYREVENUE
1.00
0.19
0.18
0.05
0.65
CUMURATING
i,t−1
0.19
1.00
0.69
0.19
0.17
DAILYRATING
i,t−1
0.18
0.69
1.00
0.13
0.18
CUMUPOST
i,t−1
0.05
0.19
0.13
1.00
0.17
DAILYPOST
0.65
0.17
0.18
0.17
1.00
1012
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
AverageUserGradeonthetopofeachmovie'spage,itisthe
most noticeable information on the website. H2a indicates
thatthecumulativeuserreview ratinghasapositiveimpact
onmovierevenues.We thenexpect α
2
N0.To testH2b, we
usedDAILYRATING
i,t−1
tosubstituteCUMURATING
i,t−1
inthe
revenue equation, which is formulated in Eq. (3) and
estimatedwithEq.(2).
DAILYREVENUE
it
¼d
t
þg
1
DAILYPOST
it
þg
2
DAILYRATING
i;t1
þg
3
DAILYREVENUE
i;t1
þg
4
WEEKEND
it
þu
i
þf
it
ð3Þ
Followingthediscussionabove,weanticipateγ
1
N0,γ
2
N
0,andγ
3
N0.
For the post equation, the addition of variable
DAILYREVENUE
it
indicates that the number of postings is
influencedby thenumberof peoplewho havewatchedthe
movie.Therefore,H4predicts that β
1
N0.DAILYPOST
i,t− 1
is
theone-daylaggedvariableofdaily numberofpostingsand
β
2
N 0 is suggested to be positive by H5. CUMUPOST
i,t − 1
denotesthecumulativenumberofuserreviewsposteduntil
dayt −1. H6 suggeststhatusershaveless incentive topost
reviewsgivenasufficientnumberofexistingreviews.There-
fore,weexpectedthatβ
3
b0.
Adummyvariable,WEEKEND
it
,isincludedinallequations
toidentifythepotentialdifferenceofconsumers'movie-going
behaviorbetweenthe weekend andweekdays. θ
t
t
,andδ
t
representinterceptsthatdenotetheaggregatetimeeffectfor
eachmovie.Foreachequation,wealsoincorporatethefixed
effects,μ
i
i
,and φ
i
, to capturetheidiosyncratic character-
istics associated with each movie, such as its budget,
marketing costs, genre, distributor, as well as its intrinsic
quality. The fixed effects capture all non-time-varying
unobservedheterogeneityofeachmovie,thus wewereable
to control for unobserved differences across movies. In
addition, fixed-effects estimation allows the error term to
arbitrarilycorrelatewithotherexplanatoryvariables,making
theestimationmoreflexibleandrobust.
5.Resultsanddiscussions
5.1.Estimationresultsfor the first week
Athree-stageleast-square(3SLS)procedurewasemployed
to simultaneously estimate the system of two equations
(either(1.)and(2.)or(3.)and(2.)).OLSresultsarepresented
for comparison. OLS estimation is inconsistent because the
regressorsofalltheequationsincludeendogenousandlagged
variables.Wearealsoconcernedabouttheconsistencyof3SLS
estimation procedure since we include lagged endogenous
variablesintheequation,andafixed-effectsmodelmaysuffer
fromfinitesamplebias[36].Inaddition,thelaggedvariables
contributetotheidentificationofthesystemofequations.We
thenestimate a model suggested by Arellano andBond[1]
usingaGMM-basedmethodandfindqualitativelyequivalent
results to 3SLS. Estimation results for the first week are
presented in Tables8 and 9Table8 shows the results for
estimatingEqs.(1)and(2)(cumulativeuserreviewrating),and
Table9presentstheresultsforanalyzingEqs.(3)and(2)(daily
userreviewrating).
In Table 8, for the revenue equation (3SLS estimation),
DAILYPOST
it
are significant predictors for DAILYREVENUE
it
,
supportingH1.ThepositiverelationshipbetweenDAILYPOST
it
Table6
Week2correlationmatrix
Variable
DAILYREVENUE
CUMURATING
i,t−1
DAILYRATING
i,t−1
CUMUPOST
i,t−1
DAILYPOST
DAILYREVENUE
1.00
0.20
0.13
0.28
0.56
CUMURATING
i,t−1
0.20
1.00
0.61
0.26
0.21
DAILYRATING
i,t−1
0.13
0.61
1.00
0.19
0.17
CUMUPOST
i,t−1
0.28
0.26
0.19
1.00
0.57
DAILYPOST
0.56
0.21
0.17
0.57
1.00
Table7
Correlationmatrixofmoviesummaryvariables
Budget Est.
marketing
costs
US
gross
Totaluser
review
Avg.User
Grade
Avg.Critic
Grade
Budget
1.00
0.68
0.44
0.37
0.19
0.15
Est.
marketing
costs
0.68
1.00
0.69
0.57
0.17
−0.008
USgross
0.44
0.69
1.00
0.68
0.41
0.36
Totaluser
posts
0.37
0.57
0.68
1.00
0.43
0.16
Avg.User
Grade
0.19
0.17
0.41
0.43
1.00
0.56
Avg.Critic
Grade
0.15
−0.008
0.36
0.16
0.56
1.00
Table8
First(opening)weekestimation:OLSand3SLS(cumulativerating)
Variable
OLS(fixed-effects
estimation)
3SLS(simultaneous
fixed-effectsestimation)
Coefficient(Std.Err.)
Coefficient(Std.Err.)
Equation1:RevenueequationwithDAILYREVENUEasdependentvariable
DAILYREVENUE
i,t−1
0.28(0.04)***
0.21(0.05)***
CUMURATING
i,t−1
0.56(0.27)**
0.18(0.18)
DAILYPOST
it
0.01(0.002)***
0.02(0.002)***
WEEKEND
it
3.08(0.32)***
2.96(0.29)***
N=426,R
2
=0.87
N=426,R
2
=0.89
Equation2:PostequationwithDAILYPOSTasdependentvariable
DAILYREVENUE
it
5.35(1.42)***
19.19(3.21)***
CUMUPOST
i,t−1
−0.26(0.03)***
−0.27(0.04)***
DAILYPOST
i,t−1
0.24(0.04)***
0.11(0.03)**
WEEKEND
it
11.63(9.98)
−40.85(14.31)***
N=426,R
2
=0.84
N=426,R
2
=0.82
***pb0.01,**pb0.05,*pb0.10.
Note:Timedummies(foreachday)andmoviedummies(fixedeffectforeach
movie)usedinestimatingthemodelarenotreported.
1013
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
andDAILYREVENUE
it
impliesthathighervolume ofword-of-
mouthgeneratedonthewebiscorrelatedwithhigheroffline
box office revenues. The result indicates that number of
onlineuserreviewscouldbeagoodindicatoroftheintensity
ofunderlyingword-of-moutheffectandincreaseawareness
among potential moviegoers. However, CUMURATING
i,t − 1
doesnot have asignificant impact on DAILYREVENUE
it
after
wecontrolfortheendogeneityofuserreviews,rejectingH2a.
The result contrasts sharply with the results from the OLS
regression, indicating the importance of controlling for the
interdependence between product sales and online user
reviews. We also find that the previous day's box office
revenueispredictiveoftoday'sboxofficerevenue,supporting
H3.ThesignificanceofthecoefficientofvariableWEEKEND
it
alsoverifiesourassumptionthattheatersenjoysignificantly
higherrevenuesonweekends.
Forthepostequation(3SLSestimation),thecoefficientof
DAILYREVENUE
it
is positive and significant, indicating that
volume of word-of-mouth is also strongly affectedby sales.
This result supports H4 and verifies that word-of-mouth
informationisnotonlyaninfluencerto,butalsoanindicatorof
revenues. The positive and significant coefficient of
DAILYPOST
i,t − 1
supports H5, signifying the self-driving
progression ofonline word-of-mouth in the openingweek.
Suchafindingimpliesthatearlybuzzgeneratedforaproduct
on the web is a significant driver forlater word-of-mouth
interests,especiallyfornew-releasedmovies.CUMUPOST
i,t−1
,
asexpectedinH6,isnegativelycorrelatedwiththedependent
variable(DAILYPOST
it
).Usershavelessincentivetospendtime
to post reviews if previousreviews alreadyprovide enough
information.Analternativeinterpretationisduetotheself-
selection effect. Users who are most likely to post will
contribute their reviews immediately after they watch the
movie,whilelaterusersmayjusttendtobrowsethereviews
withmuchlessincentivetopost.WealsofindWEEKEND
it
isa
significant negativepredictorimplying that, onaverage,the
number of reviews posted onweekdays is more than that
postedonweekends.
Table9presentsresultsofestimatingEqs.(3)and(2).The
significance ofthecoefficients remains the samecompared
withthoseinTable8.ThefactthatneitherDAILYRATING
i,t−1
norCUMURATING
i,t−1
hasasignificantrelationshipwithbox
officerevenuesindicates that onlineuserreviewshavelittle
persuasive effect and may not play an essential role in
influencing consumers' movie-going behavior. People often
believethatbadreviewgradeswoulddrivedownsales and
good reviews would increase sales. However, our results
indicate that online review ratings do not significantly
influence box office revenues after controlling for the
inherent movie heterogeneity. To put it differently, movies
boxofficesalesarenotinfluencedbytime-seriesvariationin
user ratings, which suggests that consumers do not blindly
follow the ratings posted by other users. Instead, they are
more likely to read the review and make an independent
judgmentabout thetruequalityofthemovie.However,we
findthat the numberofreviews plays animportant role in
influencing sales. There are two plausible explanations for
thisfinding.First,increasesinthenumberofreviewsprovide
moreinformationaboutthemovie,thusattractingmoreusers
tothetheatre.Thisinformationeffect,however,shalldiminish
quicklywiththenumberofreviewsposted.GiventhatYMhas
morethan1000 onlinereviewsformost movies,webelieve
the average information effect shall be quite small. Second,
postingreviewsonlineultimatelyreflectsauser'sincentiveto
discussthe movie withotherusers.As such,thenumberof
online reviews reflects the awareness effect of underlying
word-of-mouthinterests.Theonlineuserreviewscollectedin
ourdatarepresentasnapshotoftheoverall word-of-mouth
spreadaround.Thestrongrelationshipbetweenthenumber
ofonlineuserreviewsandboxofficesalessuggeststhatmovie
salesaresignificantlydrivenbytheawarenesseffect.
There are some major changes in the significance of
variables if we compare 3SLS with OLS. In particular,
CUMURATING
i,t− 1
isasignificantpredictorinOLSestimation
(Table8).Thismightexplainwhysomeofthepreviousresearch
foundthatonlineratingisasignificantinfluencerforproduct
sales.SimpleOLSregressiondoesnotcorrectlycharacterizethe
impactofonlineuserratingsgiventhecorrelationbetweenthe
errortermandtheendogenousvariable.Inourspecificsetting,
theeffectofCUMURATING
i,t− 1
isoverestimatedinOLSgiven
the endogeneity of DAILYPOST
it
. We also noticed that the
coefficientofDAILYPOST
it
increasesfrom0.01inOLSto0.02in
3SLS,whichis anoteworthy difference.Thisimpliesthat not
considering the endogeneity of DAILYPOST
it
leads to under-
estimation of its impact on revenues. Other significant
differencesofcoefficientincludeDAILYREVENUE
it
(5.35inOLS
to19.19 in3SLS),DAILYPOST
i,t−1
(0.24 inOLSto0.11in3SLS),
andDAILYPOST
it
(11.63inOLSto−40.85in3SLS).InTable9,we
observeddifferencessimilartothatinTable8.Thedifferencesof
theresultsbetween3SLSandOLSsubstantiateourdiscussionof
theinconsistencyofOLSestimation.
5.2.Estimation resultsfor thesecondweek
Inorderto capturetherapidly-changingnatureofword-
of-mouth communication, particularly on the Internet, we
also estimated the two-equation system using the second
week'sdata.OLSresultsstillshowedamajordivergencefrom
3SLS for the data of the second week, which we will not
Table9
First(opening)weekestimation:OLSand3SLS(dailyrating)
Variable
OLS(fixed-effects
estimation)
3SLS(simultaneous
fixed-effects
estimation)
Coefficient
(Std.Err.)
Coefficient(Std.Err.)
Equation1:RevenueequationwithDAILYREVENUEasdependentvariable
DAILYREVENUE
i,t−1
0.28(0.04)***
0.20(0.05)***
DAILYRATING
i,t−1
0.56(0.27)**
0.07(0.07)
DAILYPOST
it
0.01(0.002)***
0.02(0.002)***
WEEKEND
it
3.08(0.32)***
2.96(0.29)***
N=426,R
2
=0.87
N=426,R
2
=0.89
Equation2:PostequationwithDAILYPOSTasdependentvariable
DAILYREVENUE
it
5.35(1.42)***
19.19(3.21)***
CUMUPOST
i,t−1
−0.26(0.03)***
−0.20(0.04)***
DAILYPOST
i,t−1
0.24(0.04)***
0.11(0.02)***
WEEKEND
it
11.63(9.98)
−40.85(14.31)***
N=426,R
2
=0.84
N=426,R
2
=0.82
***pb0.01,**pb0.05,*pb0.10.
Note:Timedummies(foreachday)andmoviedummies(fixedeffectforeach
movie)usedinestimatingthemodelarenotreported.
1014
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
discussindetail.Instead,ourdiscussionwillfocusonthe3SLS
estimation.TheresultsareshowninTables10and11.
Fortherevenueequation,thecoefficientsofthevariables
ofthesecondweekaresimilartothoseofthefirstweek,but
theimpactofthevolumeofword-of-mouthisstronger.From
the first to second week, the coefficient of DAILYPOST
it
changes from around 0.02 to 0.05 in both Eqs. (1) and (3).
Thischangecanbeattributedtothedifferencesofconsumer
preference in the early period of a movie's theoretical life
cycle. The very early consumers (in the opening week) are
those with particular interest in the movie (e.g., fan of a
particularsubject,star,director,andetc.).Suchaself-selected
portion of early consumers does not have much word-of-
mouthtorefertoandotherpeople'sopiniondoesnothavea
very strong impact on them either. However, the later
followers are almost entirely driven by theword-of-mouth
generated.LiandHitt[28]analyzedandverifiedtheexistence
oftheself-selectioneffectintheearlyperiodofproducts'life
cycles.Ourresultsalsoindicatethatsuchdynamicstookplace
in a very short time frame with the help of the Internet,
suggestingthatusingshortertimeperioddata(e.g.,dailydata
inthisresearch)ismoreappropriateforinvestigatingonline
word-of-mouth.
For the post equation, DAILYPOST
i,t − 1
is no longer
significant (in both Tables 10 0 and 11), which implies that
theself-drivingeffectofword-of-mouthhasdroppeddrasti-
cally in the second week. This is also consistent with the
prediction of H6 implying that earlier users are more
enthusiastic and easily driven by other consumers' posts.
Suchafindingalso demonstrates theveryvolatilenature of
onlineword-of-mouth.Itisalsoobservedthatthecoefficient
of DAILYREVENUE
it
has droppedsignificantly (from 19.19 in
the first week to 3.92 in the second week for the
CUMURATING
i,t− 1
equation,andfrom 19.19 to 5.97 forthe
DAILYRATING
i,t − 1
equation), though it still does remain
significant.Thisresultisconsistentwithourdiscussionofthe
public good nature of online user review. An increasingly
smallproportionofpeoplewhohavewatchedthemoviehave
the incentive to write reviews on the Internet given the
existingnumberofpostings.
6.Conclusions
Theobjectiveofthisresearchistoinvestigatetheimpactand
characteristics of online word-of-mouth. Our results yield
interestingandimportantinsightsforbothacademicresearchers
andpractitioners.
Wedevelopedasimultaneousequationsystemtocapture
the interdependent relationship between online word-of-
mouth and movie sales. Our model fully specifies the dual
causal relationship and reveals the true effect of word-of-
mouthonmoviesales.Incontrasttoearlieronlineword-of-
mouthstudies,wefoundthathigherratingsdo not leadto
higher sales, but the number of posts is significantly
associated with movie sales. These results suggest that
consumers are not influenced by the persuasive effect of
onlineword-of-mouth,althoughtheyareaffectedbyaware-
nesseffectgeneratedby theunderlyingprocessofword-of-
mouth. Businesses shall therefore focus more on the
mechanisms that facilitate dispersion of underlying word-
of-mouthexchangeratherthantrytoinfluenceonlineratings.
Ourempirical analysisconductedindifferent timeperiods
captured the fast-changing nature of online word-of-mouth
communication.We foundthat word-of-mouth has a greater
impactonmoviesalesinthelaterperiodbutatthesametimethe
buzzeffectofword-of-mouthstartstodiminish.Thesignificant
differencesbetweenthetimeperiodssuggesttheimportanceof
employingadynamicsysteminstudyingtheeffectofword-of-
mouth in the digital environment. As online word-of-mouth
startstoestablishanenlargingpresenceinpeople'sroutinelife,it
iscriticalforfirmsandorganizationstounderstandtheeffectsof
onlineword-of-mouthontheirmanagerialdecisions.
Ourresearchhasestablishedarelationshipbetweenonline
word-of-mouthinformationandofflinemovie sales.However,
we did not directly observe how word-of-mouth information
wouldaffectconsumers'choicesandpurchasingdecisions.One
importantandinterestingextensionofourresearchwillbeto
investigatetheconsumer'sdecisionundertheinfluenceofword-
of-mouthinformation,especiallyinthedigital environment.In
Table10
Secondweekestimation:OLSand3SLS(cumulativerating)
Variable
OLS(fixed-effects
estimation)
3SLS(simultaneous
fixed-effectsestimation)
Coefficient(Std.Err.)
Coefficient(Std.Err.)
Equation1:RevenueequationwithDAILYREVENUEasdependentvariable
DAILYREVENUE
i,t−1
0.31(0.04)***
0.22(0.05)***
CUMURATING
i,t−1
2.04(0.76)***
0.83(0.56)
DAILYPOST
it
0.01(0.003)***
0.05(0.01)***
WEEKEND
it
1.58(0.22)***
1.44(0.24)***
N=426,R
2
=0.86
N=426,R
2
=0.83
Equation2:PostequationwithDAILYPOSTasdependentvariable
DAILYREVENUE
it
−0.02(0.80)
3.92(2.16)*
CUMUPOST
i,t−1
−0.21(0.03)***
−0.16(0.03)***
DAILYPOST
i,t−1
0.05(0.05)
0.05(0.03)
WEEKEND
it
6.30(3.58)*
−2.32(5.48)
N=426,R
2
=0.79
N=426,R
2
=0.82
***pb0.01,**pb0.05,*pb0.10.
Note:Timedummies(foreachday)andmoviedummies(fixedeffectforeach
movie)usedinestimatingthemodelarenotreported.
Table11
Secondweekestimation:OLSand3SLS(dailyrating)
Variable
OLS(fixed-effects
estimation)
3SLS(simultaneous
fixed-effects
estimation)
Coefficient(Std.Err.)
Coefficient
(Std.Err.)
Equation1:RevenueequationwithDAILYREVENUEasdependentvariable
DAILYREVENUE
i,t−1
0.31(0.04)***
0.19(0.05)***
DAILYRATING
i,t−1
0.08(0.05)*
0.04(0.03)
DAILYPOST
it
0.009(0.003)***
0.05(0.01)***
WEEKEND
it
1.57(0.22)***
1.50(0.25)***
N=426,R
2
=0.86
N=426,R
2
=0.83
Equation2:PostequationwithDAILYPOSTasdependentvariable
DAILYREVENUE
it
−0.02(0.80)
5.97(2.32)**
CUMUPOST
i,t−1
−0.21(0.03)***
−0.13(0.03)***
DAILYPOST
i,t−1
0.05(0.05)
0.04(0.02)
WEEKEND
it
6.30(3.58)*
−6.79(5.84)
N=426,R
2
=0.79
N=426,R
2
=0.80
***pb0.01,**pb0.05,*pb0.10.
Note:Timedummies(foreachday)andmoviedummies(fixedeffectforeach
movie)usedinestimatingthemodelarenotreported.
1015
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
addition, not all word-of-mouth is equal. Consumers needto
distinguishthe“true” and“honest”opinionsfrom all kinds of
feedback and recommendations on the web. Under such
circumstances,howconsumerschoosetheirinformationsource
and the mechanisms that help consumers to find trusted
information sources will be of particular interest for future
research.Moreover,furtherstudytocharacterizeandidentifythe
impactoftheonlineword-of-mouthinformationfromdifferent
resources and formats wouldalso bebeneficial to our under-
standinganddesignofonlinefeedbackandinformationsystems.
Thepresentstudyhasseveralotherlimitations.Ouranalysis
is, by necessity,restrictedto online userswho chooseto post
reviews and post them on YM. Thus, our estimates are
conditionedonsuchauserpopulation.Whilesucharestriction
does not bias the panel estimation results, they should be
interpretedasapplyingtoaself-selectedsetofonlineusers.All
themoviesinoursamplearenation-widereleases.Itwouldbe
interesting infuture researchtocomparethewide andlimited
release movies. Furthermore, we have focused on only one
entertainmentproductinthisstudy.Whilewebelieveourresults
arerelatively generalizable,itcertainlywouldbeimportantto
replicateandextendsuchastudytootherindustries.
References
[1] M.Arellano,S.R.Bond,Sometestsofspecificationforpaneldata:Monte
Carlo evidence and an application to employment equations, The
ReviewofEconomicStudies58(2)(April1991)277–297.
[2] A. Banerjee, A simple modelof herd behavior, Quarterly Journalof
Economics110(3)(1992)797–817.
[3] A.Banerjee,Theeconomicsofrumours,ReviewofEconomicStudies60
(1993)309–327.
[4] S.Basuroy,S.Chatterjee,S.A.Ravid,Howcriticalarecriticalreviews?The
boxofficeeffectsof filmcritics, starpower,and budgets, Journalof
Marketing67(October2003)103–117.
[5] P. Chatterjee, Online review: do consumers usethem?Advances in
ConsumerResearch28(2001)129–133.
[6] P.-Y.Chen,S.-Y.Wu,J.Yoon,Theimpactofonlinerecommendationsand
consumer feedback on sales, Proceedings of the International Con-
ferenceonInformationSystems,2004,pp.711–724,ICIS.
[7] Y. Chen, S. Fay, Q.Wang,Marketingimplicationsofonline consumer
productreviews,Workingpaper,departmentofmarketing,University
ofFlorida,2004.
[8] C.R.Clark,U.Doraszelski,M.Draganska,Informationorpersuasion?An
empiricalinvestigationoftheeffectofadvertisingonbrandawareness
andperceivedqualityusingpaneldata,StanfordUniversityGraduate
SchoolofBusinessResearchPaper,2007,No.1971.
[9] J.S.Coleman,E.Katz,H.Menzel,Medicalinnovation:adiffusionstudy,
Bobbs-Merrill,Indianapolis,IN,1966.
[10] D.Delen,R.Sharda,P.Kumar,Movieforecastguru:aweb-basedDSSfor
Hollywoodmanagers,DecisionSupportSystems43(2007)1151–1170.
[11] C.Dellarocas,Thedigitizationofwordofmouth:promiseandchallenges
ofonlinefeedbackmechanisms, ManagementScience49(10) (2003)
1407–1424.
[12] C.Dellarocas,N.F.Awad,X.Zhang,Exploringthevalueofonlineproduct
reviews in forecastingsales: thecase of motion pictures,Journalof
InteractiveMarketing21(4)(2007)23–45.
[13] A.DeVany,W.D.Walls,Bose–Einsteindynamicsandadaptivecontract-
inginthemotionpictureindustry,EconomicJournal,106(439)(1996)
1493–1514.
[14] W.Duan, B.Gu, A.B. Whinston, Informationalcascadesand software
adoptionontheInternet:anempiricalinvestigation,forthcoming,MIS
Quarterly(inpress).
[15] A.Elberse,J.Eliashberg,Demandandsupplydynamicsforsequentially
releasedproductsininternationalmarkets:thecaseofmotionpictures,
MarketingScience,22(3)(2003)329–354.
[16] J.Eliashberg,S.M.Shugan,Filmcritics:influencersorpredictors?Journal
ofMarketing61(April1997)68–78.
[17] G. Ellison, D. Fudenberg, Word-of-mouth communication and social
learning,QuarterlyJournalofEconomics110(1)(1995)93–125.
[18] H.Fattah,Hollywood,theinternet,&kids,AmericanDemographics23(5)
(2001)51–57.
[19] C.Forman,A.Ghose,B.Wiesenfeld,Examiningtherelationshipbetween
reviewsandsales:theroleofrevieweridentitydiscloserinelectronic
markets,”forthcoming,InformationSystemsResearch(inpress).
[20] A.D. Foster, M.R. Rosenzweig, Learning by doing and learning from
others: humancapitalandtechnicalchangeinagriculture,Journalof
PoliticalEconomy103(6)(1995)1176–1210.
[21] D. Godes, D. Mayzlin, Using online conversations to study word of
mouthcommunication,MarketingScience23(4)(2004)545–560Fall.
[22] M.Granovetter,Thestrengthofweakties,AmericanJournalofSociology
78(6)(1973)1360–1380.
[23] B.Gu, S.Jarvenpaa,Onlinediscussionboardsfortechnicalsupport:the
effectoftokenrecognitiononcustomercontributions,Proceedingsofthe
InternationalConferenceonInformationSystems,2003,pp.110–120,ICIS.
[24] E.Katz,P.F.Lazarsfeld,Personalinfluence,FreePress,Glencoe,IL,1955.
[25] M. Khouja, M. Hadzikadic, H.K. Rajagopalan, L. Tsay, Application of
complexadaptivesystemstopricingofreproducibleinformationgoods,
DecisionSupportSystems44(2007)725–739.
[26] D. Kirkpatrick, O. Ryan, Why there's no escaping the blog? U.S.
EditionFortune,January10th2005.
[27] A.Kirman,Ants,rationality,andrecruitment,TheQuarterlyJournalof
Economics108(1)(February1993)137–156.
[28] X. Li,L.M. Hitt, Self selection and information roleofonlineproduct
reviews,forthcomingatInformationSystemsResearch(inpress).
[29] B.R.Litman,Predictingsuccessoftheatricalmovies:anempiricalstudy,
JournalofPopularCulture16(1983)159–175.
[30] B.R.Litman,H.Ahn,Predictingfinancialsuccessesofmotionpictures,
NeedhamHeights,Allyn&Bacon,MA,1998.
[31] B.R.Litman,L.S.Kohl,Predictingfinancialsuccessofmotion pictures:
the80'sexperience,JournalofMediaEconomics2(1989)35–50.
[32] Y.Liu,Wordofmouthformovies:itsdynamicsandimpactonboxoffice
revenue,JournalofMarketing70(July2006)74–89.
[33] D.L.McFadden,K.E.Train,Consumer'sevaluationofnewproducts:learning
fromselfandothers,JournalofPoliticalEconomy104(4)(1996)683–703.
[34] L.Munoz,High-TechwordofmouthaimsmoviesinaFlash,LosAngeles
Times,2003.
[35] R. Neelamegham, P. Chintagunta, ABayesian model toforecast new
productperformanceindomesticandinternationalmarkets,Marketing
Science18(2)(1999)115–136.
[36] S.Nickell,Biasesindynamicmodelswithfixedeffects,Econometrica49
(1981)1417–1426.
[37] B. Orwall, E. Ramstad, Web's reach forces Hollywood to rethink
America-firstpolicy,WallStreetJournal.EasternEdition235(2000).
[38] J.Prag,J.Casavant,Anempiricalstudyofthedeterminantsofrevenues
andmarketingexpendituresinthemotionpictureindustry,Journalof
CulturalEconomics18(1994)217–235.
[39] S.A. Ravid, Information, blockbusters, and stars: a study of thefilm
industry,JournalofBusiness72(4)(1999)463–492.
[40] D.A. Reinstein, C.M. Snyder, The influence of expert reviews on
consumerdemandforexperiencegoods:acasestudyofmoviecritics,
TheJournalofIndustryEconomicsLIII(1)(March2005)27–51.
[41] M.S.Sawhney,J.Eliashberg,Aparsimoniousmodelforforecastinggrossbox-
officerevenuesofmotionpictures,MarketingScience15(2)(1996)113–131.
[42] D.E.Smallwood,J.Conlisk,Productqualityinmarketswhereconsumers
areimperfectlyinformed,QuarterlyJournalofEconomics93(February
1979)1–23.
[43] S. Sochay, Predicting theperformance of motion pictures,Journalof
MediaEconomics,7(4)(1994)1–20.
[44] A.T.Sorensen,S.J.Rasmussen,Isanypublicitygoodpublicity?Anoteonthe
impactofbookreview,NBERWorkingpaper,StanfordUniversity,April2004.
[45] C.VandenBulte,G.Lilien,Medicalinnovationrevisited:socialcontagion
versusmarketingeffort,AmericanJournalofSociology106(5)(2001)
1409–
1435.
[46] F.Zufryden,Newfilmwebsitepromotionandbox-officeperformance,
JournalofAdvertisingResearch40(1/2)(2000)55–64.
Wenjing Duan is an Assistant Professor in Information Systems and
TechnologyManagementatTheGeorgeWashingtonUniversity.
BinGuisanAssistantProfessorinInformationSystemsatTheUniversityof
TexasatAustin.
Andrew B.WhinstonistheHughRoyCullen CentennialChairinBusiness
Administration, Professorof Information Systems, Computer Science and
Economics,JonNewtonCentennialIC2Fellow,andDirectoroftheCenterfor
ResearchinE-CommerceatTheUniversityofTexasatAustin.
1016
W.Duanetal./DecisionSupportSystems45(2008)1007–1016
Documents you may be interested
Documents you may be interested