how to generate pdf in asp net mvc : Best pdf to text converter for SDK Library API wpf asp.net windows sharepoint movieforecast0-part1746

UsingOnlineRatingsasaProxyofWord-of-Mouth
inMotionPictureRevenueForecasting
ChrysanthosDellarocas†NeveenAwadFarag†Xiaoquan(Michael)Zhang
R.H.SmithSchoolofBusiness,UniversityofMaryland,CollegePark,MD20742
WayneStateUniversity,Detroit,MI48202
SloanSchoolofManagement,MassachusettsInstituteofTechnology,Cambridge,MA02139
cdell@rhsmith.umd.edu†nawad@wayne.edu†zxq@mit.edu
Abstract
The emergence e of online product t review w forums has s enabled d firms s to o monitor consumer
opinions about their products inreal-timeby miningpublicly availableinformationfrom the
Internet. Thispaperstudies s thevalueof onlineproductratingsinrevenueforecastingofnew
experiencegoods. Ourobjectiveistounderstandwhat t metrics ofonlineratingsarethemost
informativeindicatorsofaproduct’sfuturesalesandhowtheexplanatorypowerofsuchmetrics
comparestothatofothervariablesthathavetraditionallybeenusedforsimilarpurposesinthe
past.Wefocusourattentionononlinemovieratingsandincorporateourfindingsintopractical
motionpicture revenue forecasting models that t use e very early (opening weekend) ) box x office
andmovieratingsdatatogenerateremarkably accurateforecastsofamovie’sfuturerevenue
trajectory. Amongthe e metrics of online ratingsweconsidered,we foundthevalenceofuser
ratings tobethe mostsignificantexplanatoryvariable. Thegender r diversity of online raters
was alsosignificant,supportingthe theorythat word-of-mouththatis morewidelydispersed
amongdifferent socialgroups ismore effective. Interestingly,our r analysis founduser ratings
tobe more influentialin n predicting future e revenues than n average e professionalcritic reviews.
Overall,ourstudyhasestablishedthatonlineratingsareausefulsourceofinformationabout
amovie’slong-termprospects,enablingexhibitorsanddistributorstoobtainrevenueforecasts
ofagivenaccuracysoonerthanwitholdertechniques.
1
Best pdf to text converter for - SDK Library API:C# PDF Convert to Text SDK: Convert PDF to txt files in C#.net, ASP.NET MVC, WinForms, WPF application
C# PDF to Text (TXT) Converting Library to Convert PDF to Text
www.rasteredge.com
Best pdf to text converter for - SDK Library API:VB.NET PDF Convert to Text SDK: Convert PDF to txt files in vb.net, ASP.NET MVC, WinForms, WPF application
VB.NET Guide and Sample Codes to Convert PDF to Text in .NET Project
www.rasteredge.com
1 Introduction
Recentadvancesininformationtechnologyhaveenabledthecreationofadiversemosaicoftechnology-
mediatedword-of-mouthcommunitieswhereindividualsexchangeexperiencesandopinionsonava-
rietyoftopicsrangingfromproductsandservices,topoliticsandworldevents
1
.Onlinecommunities
allowopinionsofasingleindividualtoinstantlyreachthousands,orevenmillions,ofotherpeople.
This escalationinaudienceisalteringthedynamicsofmanyindustrieswhereword-of-mouthhas
traditionallyplayedanimportantrole.Forexample,theentertainmentindustryhasfoundthatthe
rapidspreadofword-of-mouthisshrinkingthelifecyclesofitsproducts(movies)andcausingitto
rethinkitsreleaseandmarketingstrategies
2
.
Rapid measurement t is the first prerequisite ofthefast reactions that are neededinthis new
playingfield. Fortunately, , inaddition to acceleratingits s diffusion, , theInternet t has made word-
of-mouthinstantly measurable: : traces s ofword-of-mouthcanbefoundinmanypubliclyavailable
Internetforums,suchasproductreviewsites,discussiongroups,chatrooms,andweblogs. This
publicdataprovidesorganizationswiththenewfoundabilitytomeasureword-of-mouthasithappens
bymonitoringinformationavailableontheInternet.
Unfortunately,unliketraditionalmedia,onlineword-of-mouthcurrentlylacks anacceptedset
of metrics. Therefore, , even though firms can collect t large e amounts of information n from online
communities,itisnotyetclearhowtheyshouldanalyzeit. Onlyahandfulofstudieshavelooked
attheinformationvalueofonlineword-of-mouth;eachhasstudiedadifferenttypeofcommunity
and(perhapsasa consequenceofthis) has foundadifferentmetric tobemost relevant. Godes
and Mayzlin (2004) ) studied d unstructured Usenet t conversations s about TV shows. They y related
variousmetricsoftheseconversationstoadynamicmodelofsalesandfoundthatthe"dispersion"
ofconversationsacrosscommunitieshadexplanatorypower,whereasthevolumeofconversations
didnot. Liu(2004)foundthatthevolumeofmessages s postedonInternetmessageboardsabout
upcoming andnewly releasedmovies was abetter predictor of their box office success than the
1
Examplesofsuchcommunitiesincludeonlineproductreviewforums,Internetdiscussiongroups,instantmessag-
ingchatrooms,mailinglistsandweblogs.SchindlerandBickart(2003)provideacomprehensiveoverview.
2
Movies are seeing much h more rapid d change e in n revenues between n the opening g weekend and second weekend,
suggestingthatpublicopinionisspreadingfaster(Lippman2003). RickSands,thechiefoperatingofficeratMiramax,
summarizedthistrendbystatingthat“Intheolddays. . . youcouldbuyyourgrossfortheweekendandovercome
badwordofmouth,becauseittooktimetofilteroutintothegeneralaudience. Thosedaysareover. . Today,there
isnofoolingthepublic” (Muñoz,2003).
2
SDK Library API:Online Convert PDF to Text file. Best free online PDF txt
Online PDF to Text Converter. Download Free Trial. Convert a PDF to Text. Just upload your file by clicking on the blue button or drag
www.rasteredge.com
SDK Library API:C# PDF Convert to Images SDK: Convert PDF to png, gif images in C#
Graphics, and REImage in C#.NET Project. Best PDF converter SDK for Visual Studio .NET for converting PDF to image in C#.NET application.
www.rasteredge.com
valence(percentageofmessagesthatexpresspositiveopinions)ofthesemessages.
This paper focuses on another r important t type of f online word-of-mouth: numerical l product
ratings postedby consumers online. Inthe e lastfew years,a number ofpopular websites(such
as Amazon,Epinions,Yahoo,etc.) haveattemptedtointroducestructureintotheconversations
posted thereinbyallowingusers s tosubmit numericalratings about the topicbeing discussedin
additionto(orinsteadof)amoredetailedtextreview. Theintroductionofnumericalratingshas
significantlyloweredthecostofsubmittingevaluationsonline. Thishasledtoarapidincreaseof
thenumberofconsumerswhobecameactivecontributors.
Our objective isto establish h evidencefor r the usefulness of online product ratings inrevenue
forecastingofnewexperiencegoods.Furthermore,weareinterestedinunderstandingwhatmetrics
of onlineratings are the mostinformativeindicators ofa product’s future sales andhowtheex-
planatorypowerofsuchmetricscomparestothatofothervariables(suchasmarketingexpenditures
andexpert reviews)thathavetraditionallybeenusedforsimilarpurposesinthepast. . Wefocus
our attention n ononline movie ratingsandincorporate our r findings into practicalmotion n picture
revenueforecastingmodelsthatuseveryearly(openingweekend)boxofficeandmovieratingsdata
togenerateremarkablyaccurateforecastsofamovie’sfuturerevenuetrajectory
3
.
Anumberoffactorsmakethemotionpictureindustryanidealtestbedforthistypeofstudy.
First,itisanindustrywhereword-of-mouthplaysanimportantrole. Second,thereiswidespread
availabilityofmovieratingsontheInternet;themostpopularsites(Yahoo!Movies,IMDB,Rotten-
Tomatoes.com)receivehundredsofratingswithinhoursofanewmovie’srelease.Third,production,
marketinganddailyboxofficedataareeasilyavailableformostmovies,makingiteasytocorre-
latethedynamic evolutionofamovie’s performance to thatof onlineratings. Fourth,asizable
academicliteratureexistsonmotionpicturerevenueforecasting(Section2providesanoverview).
Severalofthesestudieshaveattemptedtomodeltheimpactofword-of-mouthonmovierevenues;
most,however,havereliedonmoretraditionalexplanatoryvariables,suchasamovie’sstarpower,
marketingexpenditures,distributionstrategy,or professionalcriticreviews. Thesestudies,thus,
serveasausefulbenchmarkforassessingtheaddedvalueofonlineratings.
3
Throughout the e paper, , our perspective e is s that online ratings s consititute a a valuable real-time e “window” into
consumer attitudes that can be exploited by firms toforecast future revenues earlier than with more traditional
methods. Our r studydoes not t attempt t toconsider the important question of whether online ratings influence (as
opposedtopredict)futurerevenues.
3
SDK Library API:Purchase RasterEdge Product License Online
Buy Now. Raster XImage.Raster for .NET. Best .NET imaging SDK Buy Now. OCR XImage.OCR for .NET. Scan text from raster images, like jpeg, tiff, scanned pdf.
www.rasteredge.com
SDK Library API:C# PDF Text Highlight Library: add, delete, update PDF text
PDF Page in C#.NET Class. Best PDF document reader SDK control that can highlight PDF text in Visual C# .NET framework application.
www.rasteredge.com
Wedevelopedandtestedafamilyofforecastingmodels,basedonanovelextensionoftheBass
modelofproductdiffusion(Bass1969)thattakesintoconsiderationtheuniqueproperties ofthe
motionpictureindustry. Amongthemetricsofonlineratingsweconsidered,wefoundthevalence
(averagenumericalvalue)ofuserratingstobethemostsignificantexplanatoryvariable.Thegender
diversityofonlineraterswasalsosignificant,supportingthetheorythatword-of-mouththatismore
widelydispersedamongdifferent socialgroupsis moreeffective(Godes andMayzlin2004). The
dailyvolumeofonlineratingswashighlycorrelatedwiththecorrespondingboxofficerevenues. It
is,therefore,bestviewedasaproxyofsalesvolume. Ourresultssupportthehypothesisthatthe
impactofword-of-mouthonfuturesalesisproportionaltothevolumeofpastadopters;wedidnot
findanyspecialsignificanceofthevolumeofonlineratingsbeyondthat.
Interestingly,ouranalysisfounduserratingstobemoreinfluentialinpredictingfuturerevenues
thanaverage professionalcritic ratings. Giventhe e amount of attention that t critic c ratings have
beenreceivinguntilnow,thisresult has considerablepracticalconsequences. Atthesame e time,
thecorrelationbetweenuserandexpert ratings wasrelativelylow;higher predictivepowercould
beachievedbycombiningthem. Thisfindingprovidessupportforthecredibilityofuserratings,
butalsosuggeststhattheyshouldbestbeviewedasacomplement,ratherthanasasubstitute,of
expertreviews.
Using only openingweekendbox officeandonline ratings s data, our r best modelwas able to
forecastthetotalrevenueofmovies inarandomly chosenhold-out subset ofour samplewithan
averagerelativeabsoluteerrorof14.1%. AswediscussinSection5,suchlevelsofaccuracywould
haverequiredtheuseoftwoweeksofboxofficedatausingoldertechniques.
Overall,ourstudyprovidespositiveevidencethatonlineratingsareausefulsourceofinformation
about amovie’s long-termprospects. Froma a managerialperspective,the added valueof online
ratings is s that they y allow w forecasts of a a given accuracy y to be e obtained sooner r than with older
techniques. Theabilitytogenerateveryearlyforecastshasthepotentialtoalterthewaythatthe
movieindustryisusingsuchtools. Currently,post-releaseforecastsareprimarilyofvaluetomovie
exhibitorswhousethemtobettermanagetheyieldfromtheirexhibitioncapacity. Webelievethat
thereal-time availability ofreliableestimatesofword-of-mouthcanhave importantimplications
formotionpicturemarketingaswell.Suchinformationmayallowmoviedistributorstofine-tunea
movie’scampaign,ortodevelopentirelynewmarketingstrategiesthatcanattempttorespondto
4
SDK Library API:C# WPF PDF Viewer SDK to convert and export PDF document to other
Best PDF Viewer control as well as a powerful .NET An advanced .NET WPF PDF converter library for converting Export PDF text content to TXT file with original
www.rasteredge.com
SDK Library API:C# PDF Print Library: Print PDF documents in C#.net, ASP.NET
XDoc.HTML5 Viewer. XDoc.Windows Viewer. XDoc.Converter. C#.NET rotate PDF pages, C#.NET search text in PDF A best PDF printer control for Visual Studio .NET and
www.rasteredge.com
anaudience’sinitialreceptionofanewmovie.
Therestofthepaperisorganizedasfollows.Section2discussesrelatedwork.Section3describes
ourdataset. Section4introducesourforecastingmodels. . Section5presentstheresultsoffitting
ourmodelstoourdatasetandcomparestheirforecastingaccuracytothatofoldermodels. Finally,
Section6summarizesourfindings,discussesthemanagerialimplicationsofthiswork,pointstoits
limitations,andsuggestspotentialavenuesforfutureresearch.
2 RelatedWork
Ourworkrelatestotwoimportantstreamsofpastresearch: forecastingmodelsofmotionpicture
revenuesandmethodologiesformeasuringword-of-mouth.
Forecastingmodelsofmotionpicturerevenues
Predictingthesuccessofamotionpicturehaslargelybeenviewedintheindustryasa“wildguess”
(Litmanand Ahn1998). Despite e suchdifficulty, severalresearchers have proposed models that
attempttoforecastmotionpicturerevenues. Suchmodelscanbeclassifiedalongtwodimensions.
Oneclassificationcanbebasedonthetypeofforecastingmodelemployed:
1. Econometricmodelsidentifyfactorsthatpredictorinfluencemotionpictureboxofficesuccess.
Alargevarietyoffactors havebeenexamined. Some e studieshavelookedat moviecharac-
teristics,suchasstarpower(DeVanyandWalls1999;Ravid1999),moviegenreandMPAA
ratings(AustinandGordon1987),andacademyawards(DoddsandHolbrook1988). Oth-
ershaveexaminedamovie’smediaadvertising(FaberandO’Guinn1984),timingofrelease
(Krider andWeinberg1996),distributionstrategy (JonesandMason1990;Jones andRitz
1991)andcompetition(Ainslie,DrezeandZufryden2003). Severalresearchershavestudied
theroleofprofessionalcriticreviews(EliashbergandShugan1997;ReinsteinandSnyder2005;
Basuroy,ChatterjeeandRavid2003).Finally,afewintegrativestudiesexaminedtherelative
contributionofacombinationoffactors(Litman1983;NeelameghamandChintagunta1999;
ElberseandEliashberg2003;Boatwrightetal.2005).
2. Behavioralmodels s focusonfactorsinvolvedinindividualdecisionmakingtowardsselecting
5
SDK Library API:VB.NET PDF Convert to Tiff SDK: Convert PDF to tiff images in vb.
Free VB.NET Guide to Render and Convert PDF Document to TIFF in Visual Basic Class. Best VB.NET adobe PDF to Tiff converter SDK for Visual Studio .NET.
www.rasteredge.com
SDK Library API:C# PDF Text Add Library: add, delete, edit PDF text in C#.net, ASP
XDoc.HTML5 Viewer. XDoc.Windows Viewer. XDoc.Converter. View & Process. A best PDF annotation SDK control for Visual Studio .NET can help to add text to PDF
www.rasteredge.com
a movie to watch(Eliashberg andSawney, 1994; ; Sawney y andEliashberg, , 1996; ; Zufryden,
1996; De e Silva 1998; Eliashberg et t al. 2000). Such h models s usually y employ y a hierarchical
framework that t develops forecasting models s by y relating g behavioral l traits of f consumers to
aggregateeconometricfactors.
Another classificationcanbebasedonthetiming oftheforecast. Mostoftheproposedmodels
aredesignedtoproduceforecastsbeforeamovie’sinitialrelease(Litman1983;Zufryden1996;De
Silva1998; Eliashbergetal. 2000)whileothers s focusonforecasting later-week revenues after r a
movie’s early boxoffice revenues become known(Shawhney andEliashberg 1996; ; Neelamegham
andChintagunta1999).Thelattercategorytendstogeneratemoreaccurateforecastingresultsdue
tothefactthatthesemodelshaveaccesstomoreexplanatoryvariables,includingearlyboxoffice
receipts,criticreviews,andword-of-moutheffects.
Ourstudy proposes afamily ofdiffusionmodels whosegoalis toforecastlater-weekrevenues
very soon n (i.e. within2-3days) ) aftera movie’sinitialrelease. Thenovelty y ofourapproachlies
intheexaminationofvarious metricsofonlineratings asaproxyofword-of-mouth. . Tothebest
of our r knowledge we e are the first t to examine the e use e of these metrics in the context of movie
revenueforecasting
4
. Ourcontributionliesinestablishingwhichmetrics s ofonlineratingsarethe
bestpredictorsofmotionpictureperformanceandincomparingthepredictivepowerofthesenew
metrics tothat ofmoretraditionalexplanatoryvariables usedinpastresearchsuchasamovie’s
marketingexpenditures,distributionstrategy,andprofessionalcriticreviews.
Methodologiesfor measuringword-of-mouth
Traditionalattemptstomeasureword-of-moutharebasedontwoprincipaltechniques:inferenceand
survey. Forexample,Bass(1969)andthosewhohaveextendedhismodeltypicallyuseaggregated
salesdatatoinferthemodel’scoefficientofinternalinfluence,which,inturnisassumed torelate
toword-of-mouth. Asanotherexample,Reingenetal. (1984)conductasurveyofthemembersof
asororityinwhichtheycomparebrandpreferencecongruitybetweenwomenthatlivedinthesame
4
Concurrently and d independently y Liu u (2004) studied the e impact t of unstructured online discussions on movie
revenues. Our r study, in contrast, focuses on numerical online ratings. Whereas s Liu u found that the e volume of
discussion was significant but its valence(positiveor negative) marginally so, our study finds thevalence of user
ratingstobethemostsignificantvariable. Furthermore,ourmodifiedBassmodelhelpsdisentanglethedifferentway
inwhichthevolumeandvalenceofratingsbothrelatetotheevolutionofamovie’sboxofficerevenues.
6
SDK Library API:C# PDF Text Box Edit Library: add, delete, update PDF text box in
with .NET PDF Library. A best PDF annotator for Visual Studio .NET supports to add text box to PDF file in Visual C#.NET project.
www.rasteredge.com
houseandthosethatdidnot. Theyfindthatthosethatlivedtogetherhadmorecongruentbrand
preferencesthanthosethatdidnot. Thestudytheninfersthatthosethatlivedtogetherhadmore
opportunitiesforinteractionandthus,thatword-of-mouthcommunicationwasmoreprevalent.
Surveysremainthemostpopularmethodtostudywordofmouth,largelybecauseindividuals
canbeaskeddirectlyabouttheircommunicationhabits;theerrorthenliesintheself-reportingof
behavior.Severalwell-knownstudies,suchasBowmanandNarayandas(2001),BrownandReingen
(1987),ReingenandKernan(1986)andRichins(1983),basetheiranalysesonproprietarysurveys
designedtotestaspecifichypothesisrelatedtoword-of-mouth.
TheadventoftheInternetintroducedathirdtechniqueformeasuringwordofmouth:directly
throughonlinediscussion groups s and d online review forums. Researchers s caneasily gather large
amounts of datafrom m suchforums. Nevertheless, , soundmethodological principles s for analyzing
suchdataarestillintheprocessofbeingestablished.Previousresearchhaslookedatunstructured
onlinediscussionforumsandhasusedvolumeanddispersionwhenexaminingonlinewordofmouth.
The theorybehindmeasuringdispersion,or the spreadofcommunicationacross communities,is
thatwordofmouthspreadsquicklywithincommunities,butslowlyacrossthem(Granovetter1973).
GodesandMayzlin(2004)havefoundthatthedispersionofconversationsaboutweeklyTVshows
acrossInternetcommunitieshaspositivecorrelationwiththeevolutionofviewershipoftheseshows.
Thetheorybehindvolumeisthatthemoreconsumersdiscussaproduct,thehigherthechancethat
otherconsumerswillbecomeawareofit.Inarecentpaper,Duanet.al.(2005)explorethedynamic
relationshipbetweenonlineuserreviewsandmotionpicturebox officerevenues. . They y findthat,
whereasthevolumeofonlinepostingsshowssignificantcorrelationwithboxofficesales,thevalence
(averagenumericalrating)ofthosepostingsdoes nothaveasignificantimpact. . Inthisstudywe
extendpreviousattemptstomeasuretheimpactofonlineword-of-mouthbylookingatstructured
productratingforums andsuggestingmethodologies that endogenize theimpact ofvolume and,
thus,allowustobetterexploretheimpactofthevalenceofonlinefeedback.
7
Variable 
Min  Mean  Max 
Box office (aggregate; in millions) 
2.5  68.1  403.7 
Production Budget (in millions) 
 46.1 
140 
Marketing Budget (in millions) 
 24.3 
50 
Exhibition longevity (in weeks) 
14 
51 
Screens in opening week 
 2,393  3,615 
Volume of total user ratings 
67 
689  6,295 
Volume of first week user ratings 
312  3,802 
Volume of critic ratings 
13 
20 
Average aggregate user rating (range 1-5) 
1.9 
3.4 
4.4 
Average critic rating (range 1-5) 
1.4 
3.1 
4.6 
Total number of movies 
80 
Total number of user ratings 
55,156   
Total number of critic ratings 
1,040   
Total number of unique users 
34,893   
Table1:Keysummarystatisticsofourdataset.
3 DataSet
DataCollectionMethodology
DataforthisstudywerecollectedfromYahoo! Movies(http://movies.yahoo.com)andBoxOffice-
Mojo(http://www.boxofficemojo.com). FromYahoo! Movies,wecollectedthenamesofallmovies
releasedduring2002.Forthepurposeofouranalysis,weexcludedtitlesthatwere(a)notreleasedin
theUnitedStates,(b)notatheatricalrelease(e.g. DVDreleases),or(c)notreleasednation-wide.
Foreachoftheremainingtitleswecollecteddetailedratingsinformation,includingallprofessional
criticreviews (textandletterratings,whichweconvertedtoanumberbetween1and5) andall
userreviews(dateandtimeofreview,userid,reviewtext,integerratingsbetween1and5).
WeusedBoxofficemojotoobtainweeklyboxoffice,budgetandmarketingexpensesdata. This
information was s missing for r severalmovies from the e publicly accessible parts s of f that site. We
obtainedadataset of80movieswithcompleteproduction,weekly boxoffice,critic reviews and
daily user review data
5
. Our r finaldata a set consists of 1188 weekly box office data, 1040 critic
reviews(anaverageof13reviewspermovie),and55156userreviewsfrom34893individualusers
(anaverageof689reviewspermovieand1.5reviewsperuser). Table1providessomekeysummary
statistics.
5
The final movie sample was found d to havesimilar r overall profile with h the e full set of nationally-released d 2002
movies(intermsofgenre,budget,andmarketing),ensuringthatnobiaswasintroducedbyconsideringonlyamovie
subset.
8
Age 
2002 Yahoo! 
Movie Raters 
2001 US 
Moviegoers* 
<18 
13% 
15% 
18-29 
58% 
35% 
30-44 
23% 
28% 
45+ 
6% 
22% 
Gender 
Men 
74% 
49% 
Women 
26% 
51% 
* Source: Newspaper Association of America (NAA) 
Table2: EstimateddemographicprofileofYahoo! Moviesraters.
DemographicsofOnlineRaters
Wewereabletocollectpartialraterdemographicdatabyminingtheuserprofilesthatareassociated
withtheraters’YahooIDs. About85%ofratersinourdatasetlistedtheirgenderand34%their
age. Fromthatinformation,weconstructedanestimateofthedemographicprofileoftheYahoo!
Moviesraterpopulation(Table2). Wefoundthatthedemographicbreakdownofonlineratersis
substantially skewedrelativetothatofUSmoviegoers. Mostnotably,a a disproportionatelyhigh
percentage of f online e ratings were provided by y young g males under 30. This s suggests s that some
rebalancingofonlineratingsmightberequiredtoimprovetheirvalueinforecastingrevenues.
RelationshipbetweenUserandCriticRatings
Sincemuchworkhasbeendoneonusingcriticreviewstopredictmovierevenue(Eliashbergand
Shugan1997;ReinsteinandSnyder2005;Basuroy,ChatterjeeandRavid2003),itisnaturaltoask
howwelluserratingscorrelatewithcritic ratings. Table3 3 depicts the correlationbetweencritic
anduserratings. Allscoresarerelativelylow. Interestingly,firstweekuserreviewsexhibithigher
correlation with critic reviews thandolater r week reviews. Also, , reviews posted by y male users
correlatebetterthanreviewspostedbyfemaleusers. Thelowcorrelationbetweenuserandcritic
ratingsemphasizestheimportanceofexamininguserreviewsasapredictivetool,astheinformation
providedby users s is s substantially different from m the informationprovidedby y professional l movie
critics.
9
Raters 
All  Male  Female 
First week 
0.63  0.61 
0.46 
Second week 0.58 0.57 
0.53 
Third week 
0.53  0.46 
0.45 
All weeks 
0.59  0.58 
0.49 
Table3: Correlationofcriticanduserratings.
PreponderanceofExtremeUserRatings
Figure1aplotsthehistogramofaverageuserratingsforallmoviesinourdataset.Thehistogram
ofaveragecriticratings(normalizedtolieinthesameintervalas userratings)isalsoplottedfor
comparison.Userratingsarelessevenlydistributedthancriticratings,withthemajorityofmovies
receivinganaverageuserratingbetween3.5and4.5. Evenmorerevealingisaplotoftherelative
incidenceofthevarioustypesofratings(Figure1b). Criticsseemtoberatingmoviesona(slightly
upwardlybiased)curve. Incontrast,themajority y ofuserratings lieatthetwoextremes of the
ratings scale, , witha strong emphasis onthe positive end: : almost t half of all l posted ratings are
equaltothehighestpossiblerating,18%ofratingsareequaltothelowestpossiblerating,andonly
about30%areintermediatevalues.Thepreponderanceofextremereviewsisconsistentwithsimilar
findingsrelatedtoonlineproductreviewsonAmazon.comandothersites(AdmatiandPfleiderer
2000). Itis s alsoconsistentwithpastresearchonword-of-mouththatfinds thatpeoplearemore
likelytoengage ininterpersonalcommunicationwhentheyhave very positiveandvery negative
experiences(Anderson1998)
6
.
DynamicsofRatingsVolume
Online reviews are (at t least in n principle) ) contributed by people who have e watched the movies
beingrated. Itis,thus,expectedthattheirdailyvolumewillexhibitastrongcorrelationwiththe
correspondingbox officerevenues andwill l decline over r time. Figure2confirmsthis s for “Spider-
Man”. Observethatthevolumeofdailyratingsfollowscloselythebox-officepeaksandvalleysthat
areassociated withweekendsandweekdays,especially y during the first two weeks. Most t movies
6
Itisimportant tonoteherethattheskeweddistribution ofonlineratingsisnotacauseofalarmanddoesnot
diminish their information n value. In n an n interesting g theoretical paper, Fudenberg and d Banerjee (2004) prove that
thepresence ofreporting bias (i.e. . higher r propensitytocommunicateextremerather thanaverageoutcomes)ina
populationdoesnotdiminishtheabilityofword-of-mouthtoenableperfectsociallearning.
10
Documents you may be interested
Documents you may be interested