ContentslistsavailableatScienceDirect
Research Policy
j ou rn a l h om epa g e :w w w . e l s e v i e r . c o m / l o c a t e / r e s p o l
Market failure in the diffusion of consumer-developed innovations:
Patterns in Finland
Jeroen P.J. de Jong
a,b, Eric von Hippel
c,∗, Fred Gault
d,e, Jari Kuusisto
f, Christina Raasch
gaRSMErasmusUniversity,Rotterdam,TheNetherlands
bUtrechtSchoolofEconomics,Utrecht,TheNetherlands
cMITSloanSchoolofManagement,Cambridge,MA,USA
dUNU-MERIT,Maastricht,TheNetherlands
eTshwaneUniversityofTechnology,TUT-IERI,SouthAfrica
fUniversityofVaasa,Vaasa,Finland
gTUMunichSchoolofManagement,München,Germany
a r t i c l e i n f o
Articlehistory:
Received18April2014
Receivedinrevisedform19June2015 Accepted29June2015
Availableonline13August2015 Keywords:
Userinnovation Commercialdiffusion Peer-to-peerdiffusion Generalvalue Marketfailure
a b s t r a c t
Empiricalstudieshaveshownthatmillionsofindividualusersdevelopnewproductsandservicesto servetheirownneeds.Theeconomicimpactofthisphenomenonincreasesifandasadoptersinaddition totheinitialinnovatorsalsogainbenefitsfromthoseuser-developedinnovations.Ithasbeenargued thatthediffusionofuser-developedinnovationsisnegativelyaffectedbyanewtypeofmarketfailure:
valuethatothersmaygainfromauser-developedproductcanoftenbeanexternalitytoconsumer- developers.Asaresult,consumerinnovatorsmaynotinvestinsupportingdiffusiontotheextentthat wouldbesociallyoptimal.Inthispaper,weutilizeabroadsampleofconsumersinFinlandtoexplorethe extenttowhichinnovationsdevelopedbyindividualusersaredeemedofpotentialvaluetoothers,and theextenttowhichtheydiffuseasafunctionofperceivedgeneralvalue.Ourempiricalanalysissupports thehypothesisthatamarketfailureisaffectingthediffusionofuserinnovationsdevelopedbyconsumers fortheirownuse.Implicationsandpossibleremediesarediscussed.
©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).
1. Introductionandoverview
Empiricalresearchfindsthattensofmillionsofcitizensspend tensofbillionsofdollarsannuallydevelopingandmodifyingcon- sumerproductstobetterservetheirownneeds(vonHippeletal., 2011). Drivenby theever-increasing qualityof freely available design and communication tools, single and collaborative user innovationis expectedtobecome evenmore prominent inthe future(BaldwinandvonHippel,2011).
The social welfare benefits of single and collaborative user innovationbycitizenswillbeconsiderably enhancedif citizen- developedinnovationsofgeneralvaluediffusetootherswhocan benefitfromthem.However,frommicroeconomictheory,thereis reasontohypothesizethatfree,peer-to-peerdiffusionofuserinno- vationwillbeinefficientlylowfromasocialwelfareperspective.
AsvonHippeletal.(2014)haveargued,wheninnovationdiffusion involvesfreerevealingrather thanmarkettransactions,innova-
∗Correspondingauthor.Fax:+16172532660.
E-mailaddress:evhippel@mit.edu(E.vonHippel).
torswillfindthebenefitsthataccruetoadopterstobepartiallyor entirelyanexternalityfromtheirpointofview.Asaresult,user innovatorscanbeexpectedtoinvestlessthanmightbesocially desirabletoinformorassistotherstoadopt,evenwhentheirinno- vationswouldbehighlyvaluabletoothers–amarketfailure.In thespecificcircumstancesfocuseduponhere,wesaythatamar- ketfailureexistsifuserinnovatorsandadopters,takentogether, wouldhavehighernetbenefitsfromtheuserinnovationiftheuser innovatorinvestedmoreindiffusion.Thistypeofmarketfailureis novelintheinnovationliterature.
Inthispaper,weempiricallyexplorethemarketfailurehypoth- esisjustdescribedviaasampleof176innovationsdevelopedfor personalusebyindividualconsumersinFinland.Inoverview,we foundthat85%oftheconsumerdevelopersreportthatwhatthey haddevelopedhighlysatisfiedtheirownneeds.Moreover,draw- ingonmultiplequestions,weconcludedthat,inourrespondents’
view,61%oftheirinnovationsaredeemedusefultosomeormany others.Still,actualcommercialand/orpeer-to-peerdiffusiononly occurredfor19%oftheinnovations.
Wefurtherfindthateffortexertedtoachievepeer-to-peerdif- fusionisnotaffectedbytheinnovators’assessmentofthegeneral http://dx.doi.org/10.1016/j.respol.2015.06.015
0048-7333/©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).
valueoftheinnovation,andconsequently,innovationswithhigher perceivedvaluetootherusersarenotmorelikelytospreadtopeers thanarelow-valuedones.Incontrast,commercialdiffusioneffort exertedisrelatedtoperceivedgeneralvalue.Thesefindingsarein linewiththeexistenceofamarketfailureofthetypehypothesized byvonHippeletal.(2014).
Ourconcludingdiscussionincreasesourunderstandingofits innerworkingsandpointsoutavenuesforfutureresearchaswell aspracticalimplicationsforpolicyandbusiness.
2. Reviewoftheliterature
Inthissection,wereviewtheliteratureonthefrequencyand importanceofinnovationandinnovationdiffusionbyusers,the pathwaysbywhichuserinnovationsdiffuse,users’incentivesto diffusetheirinnovations, andthelikelihood ofdiffusion-related marketfailure.
2.1. Extentofproductdevelopmentandmodificationby consumers
Representativenationalsurveysofcitizensaboveage18inthe UK,US,andJapan,showthatmillionsofindividualsineachofthese nationsdevelopormodifyconsumerproductstobetterservetheir personalneeds(vonHippeletal.,2011).In theUK,thefraction ofuserinnovatorswasfoundtobe6.1%ofthepopulation,inthe USitwas5.2%,andinJapanitwas3.7%.Thescopeofconsumer innovationinallthreenationswasfoundtobeverybroad,ranging fromimprovementstovehicles,toproductsusedinpatienthome care,toimprovementsinsportingproducts.
IntheUK,vonHippeletal.(2012)estimatedthatconsumer- developersonaveragespent7.1 daysand £1098out-of-pocket costsperyear.Atthemacro-levelandwhenevaluatingperson-days ataverageUKworkforcesalaries,totalannualspendingbycon- sumersoninnovationwasestimatedto£3.2billion.Incomparison, estimatedannualR&Dexpendituresbycompaniesonconsumer productswere£2.2billion.Similarfindingshavebeenreportedfor theUSandJapan(vonHippeletal.,2011).Thesefindingsshowthat boththescaleandscopeofuserinnovationissubstantial.
2.2. Diffusionpathways
Consumersasuserinnovatorsaremotivatedtocreateinnova- tionstoservetheirownneeds–notthoseofothers,andconsumer needshave beenshown tobe heterogeneous (Frankeand von Hippel,2003).Atthesametime,whatoneconsumerrequiresmay fitwhatanotherwantsbetterthatcommercially-availableprod- ucts,andsosomeuserinnovationsmayprovetobeofgeneralvalue.
Whenuserinnovationsarevaluabletoothers,diffusionenhances socialwelfare(Gambardellaetal.,forthcoming).Userinnovations areespeciallylikelytobeofgeneralvaluewhentheyhavebeen developed by‘lead users’,who arecharacterizedby needsthat foreshadowgeneraldemand.Producerswhopurposefullyseekout innovationsdevelopedbyleadusersasabasisforcommercialprod- uctshavefoundthistobeaprofitablepractice(Lilienetal.,2002).
Thediffusionpathwaysuserinnovationsmightfollowareas showninFig.1(Baldwinetal.,2006;deJongandvonHippel,2013).
AtthetopofFig.1,weseethatuserswhoinnovatemaychooseto
Fig.1. Pathwaysviawhichuser-developedinnovationsdiffuse.
revealinformationregardingtheirinnovationswithoutchargeto otherusers(peers)interestedinadoptingthem.Thisdiffusionmay bepurposeful,orsimplybetheresultofspilloversofunprotected information,aswhenanovelproductisusedbyauserinnovator inapublicsetting(Strandburg,2008).
Diffusioncanalsobeaccomplishedlessdirectly,withproducers obtaininginformationfromuserinnovatorssothattheycanadopt theinnovation(andfurtherdevelopitifneeded)andthenofferitto abroadaudienceforgeneralsale.Ascanbeseenattheleftsideof Fig.1,theinformationmaybefreelyrevealedtotheproducerson thesametermsasitisrevealedtoadoptingusers:freelyrevealed informationhasnorestrictionsuponwhomayaccessit.Or,some userinnovatorsmaychoosetonotfreelyrevealtheirinnovation- relatedinformationbutinsteadreceivesomekindofcompensation (e.g.,pay,royalties,andfavors)(deJongandvonHippel,2009).Or alternatively,theymaystarttheirownfirmforthatsamepurpose (ShahandTripsas,2007).Inanyofthesecommercialpathways,the innovationendsupbeingofferedforgeneralsale,andinthatway diffused.
2.3. Prevalenceofinnovationdiffusionbyindividualusers
Ithasbeenempiricallydocumentedthatuserinnovatorsmay freelyrevealwhattheyhavedeveloped,forotherstoexamine,imi- tate,ormodifywithoutanycompensationtotheinnovator.The practicesvisibleinopensourcesoftwaredevelopmentwereimpor- tantinbringingthisphenomenontogeneralawareness.Inthese projectsitwasclearpolicythatprojectcontributorswouldrou- tinelyandsystematicallyfreelyrevealcodetheyhaddevelopedat privateexpense(Raymond, 1999).However,freerevealingdoes notimplythatotherswilladoptwhathasbeenfreelyrevealed.In thecaseofinnovationsbyindividualusers,surveyevidenceshows thatdiffusionexistsinonlyafractionoftheidentifiedcases.As canbeseeninTable1,thediffusionrate,viacommercialand/or peer-to-peerchannels,variesfrom5.0%to17.1%.Thisisthecase eventhough,ascanalsobeseeninTable1,onlyasmallpercentage ofindividualconsumershavelegallyprotectedtheirinnovation- relatedknowledgeasintellectualproperty.
Notethat,ontheirown,thefiguresfordiffusionshowninTable1 arenotevidenceforunder-diffusion.Although,thismatterwasnot studiedpriortotheempiricalstudywewillreportonhere,many orevenmostoftheinnovationsinearlierstudiesmayhavebeen ofinterestonlytotheinnovatinguser.Insuchcases,non-diffusion isnotevidenceofashortfallininvestmentindiffusionbytheuser
Table1
Protectionofanddiffusionofuserinnovationsdevelopedbyconsumers.
Source Country Datayear Sample ProtectionwithIPRs Diffusion
vonHippeletal.(2012) UnitedKingdom 2009 104innovationsbyconsumers≥18years 1.9% 17.1%
OgawaandPongtanalert(2011) USA 2010 114innovationsbyconsumers≥18years 8.8% 6.1%
OgawaandPongtanalert(2011) Japan 2011 83innovationsbyconsumers≥18years 0.0% 5.0%
innovator:itsimplyisareflectionoftheexpectedlackofadopter interest.Obviously,diffusionisusefulonlytotheextentthatuser innovationshavevaluetoothers.
2.4. Potentialmarketfailure
Ingeneral,amarketfailureexistswhenanotherpossibleout- comecan make atleast oneeconomic actor betteroff without makingsomeoneelseworseoff(KrugmanandWells,2006).Market failuresareaformofinefficiencythatmaycallforremedy.
Thetypeof possiblemarketfailurethatmight beassociated withthediffusionofindividual,freely-revealedinnovationsisthat individualusers, withnomechanisminplacetosharetheben- efits othersmight reapfromadopting theirinnovations, would viewpotentialadopters’benefitsasanexternalityandsounder- investininnovationdiffusionfromtheperspectiveofsocialwelfare.
Tounderstandthelevelofdiffusioninvestmentthatitwouldbe sociallyusefulforuserinnovatorstoexpend,considertheinnovat- inguserandthepoolofpotentialadopterswhocouldbenefitfrom hisorherinnovationasanintegratedsystemforwhichabenev- olentdictatorseekstomaximizetheaggregatesurplus.Assume thatinvestmentsinthediffusionofinnovation-relatedinforma- tionbyinnovatorswilllowerthecostsforallpotentialadopters.
Assumealsothatadditionalinvestmentsbytheinnovatorwould loweradopters’costsatadecliningrate.Aggregatesurplusisthen maximizedatthepointwhereanadditionaldollarofinvestmentby theinnovatorindiffusionproducesanincreaseofexactlyadollarof benefitfortheentirepoolofpotentialadopters.Inotherwords,that socialoptimumobtainsifthemarginalcostofdiffusionequalsthe marginalbenefitderivedfromit.Ofcourse,individualinnovators mayobtainnon-priceddiffusionbenefitssuchashelpwithinnova- tiondevelopment,orreputationalbenefits,oraltruism(Frankeand Shah,2003;LakhaniandWolf,2005).Ifthesearesufficientlyhigh, theycouldoffsetthelackofdirectfinancialrewardthata mar- ketprovides.Whethertheydosoisamatterrequiringempirical investigationsofthetypeconductedinthispaper.
Thisnovelformofmarketfailurewasfirstdescribed byvon Hippeletal.(2014),andanexistenceproofwasprovidedwithin themedicalfield.Here,itwasfoundthatclinicians’effortstodif- fusevaluablediscoveriesmadeduringtheirclinicalpracticewas verylow,andunrelatedtothegeneralvalueofthosediscoveries.
Inthispaper,webroadentheexplorationofthispotentiallyvery fundamentalformofmarketfailuretoamuchmoregeneralcate- goryofuserinnovators–abroadsampleofcitizensinFinland.We alsoexploremoredeeplythedetailsofcauseandconsequence.
3. Researchmethods 3.1. Sampleidentification
OurresearchwassupportedbyagrantfromFinland’snational innovationagency,Tekes.Toobtainaninitialandrepresentative sample,wefirstcontacted10,000individualcitizensaged18–65 yearsby telephonetoinvitethem totakeanelectronicsurvey.
Thesecitizenswere randomlydrawnfromFinland’s Population RegisterCentre.In ordertoincreaseoursubsampleofvalidated userinnovations(seelater),wealsosoughttobuildaconvenience sampleconsistingofpeoplewhoseattributes,basedonprevious studies,madeitmorelikelythattheywereinnovators:highlyedu- cated,technicalworkers,andmales(vonHippeletal.,2011).Tothis end,weinvitedmembersofseverallaborunionswhosemembers hadthesecharacteristicstoparticipateinoursurvey(Theunions whosemembershipswecontactedwere:TheUnionofProfessional EngineersinFinland,theFinnishInventorsNationalFederation,
AcademicEngineersandArchitectsinFinland,theTradeUnionof Education,AKAVA,SEFE,andTheFinnishMetalworkers’Union).
AllinvitationswereissuedbetweenAugust,2012andJanuary, 2013.Eachinviteewassentahyperlinksothatshe/hecouldfind thesurveyandparticipate.Toavoidduplication,respondentswere askedtofillinthesurveysattheirhomeaddress.Eachrespondent wasallowedtocompletethesurveyonlyonce;toensurethis,our softwarerecordedeachrespondent’sIPaddress.
Completed questionnaireswere obtainedfrom 2048Finnish citizens.Therepresentativeandlikelyuserinnovatorsubsamples contained993and1055respondents,respectively.Wecombined theseintoanoverallsample,asexploratoryt-testsand2-tests showedthatnoneofthekeyvariablesinourstudy(reportedlater) significantlydifferedbetweenthetwosamples.Withrespectto demographiccharacteristics,58%oftherespondentsintheoverall sampleweremale.Fortyonepercenthadabachelors,mastersor Ph.D.degree,and46%workedinatechnicaljoborbusiness(e.g., engineering,medicine,naturalsciences,design,andIT).Respon- dents’averageagewas44.9yearsatthetimeofthesurvey,and86%
wereemployed(includingself-employedandbusinessowners).
3.2. Samplescreening
Toidentifyrespondentswhohadengagedinuserinnovation,we appliedarefinedversionofaprocedureinitiallydevelopedinthe UnitedKingdom(vonHippeletal.,2012).Thescreeningprocedure firstcastsabroadnettoincreasethechanceofcapturinganyprod- uctdevelopmentsormodificationsofrespondents.Then,careful screeningisappliedtoeliminateanyfalsepositivescaptured.
At the start of our questionnaire we stated: “The following questionsrelatetoanycreative activitiesinyour leisuretime.You mayhavecreatedanyproductsorapplicationsforpersonaluse,to helpotherpeople,tolearnorjustforfun”.Respondents’recallwas assistedbyofferingalistofninespecificcues:hadtheycreated any(1)computersoftware;(2)householdfixturesorfurnishing;
(3)vehicle-related;(4)toolsorequipment;(5)sports,hobbyor entertainment;(6)childoreducation-related;(7)health,careor medical;(8)foodorclothing;or(9)anyotheritems.Weaskedif respondentshadcreatedanyoftheseitemsforpersonaluseinthe pastthreeyears.
Outofthe2048respondents,initially624reportedatleastone creationwithrespecttothenineaforementionedcues.Wenext appliedtwoscreeningquestionsasonestepinourvalidationpro- cess.Weasked(1)whethertherespondentknewofanequivalent productavailableonthemarketthathe/shecouldhavebought;
and(2)whethertheinnovationhadbeendevelopedaspartofthe respondent’sjob.Apositiveanswertoeitherquestioneliminated theclaimedinnovationfromfurtherconsideration.Wewerenot interestedinreplications ofexistingproducts, butrather aimed forcreations/modificationsthatwereatleastnewtotheconsumer him/herself.Wealsowantedtoincludeonlyinnovationsthatindi- vidualshaddevelopedasconsumersratherthanasbusinesssector employees.Applicationofthesescreensreducedour624affirma- tiverepliesto251potentialinnovators.
Next, the survey script asked respondents to describe their developmentanditsintendedfunction.Theseopen-endeddescrip- tions were examined and discussed by two members of the researchteam. Casesregarded as false positives due tolack of novel,user-developedcontentsuchas:“Iinstalleda[manufacturer- developed] software upgrade on my personal computer” were removedatthisstage.Whennodescriptionwasprovided,wetook therespondent’sclaimsascorrect(e.g.,“Iamtooafraidtorevealit.
[But]theproblemIhadalmostonadailybasisisnowsolved!”).
Finally,weonlyincludedinnovationsintooursamplethatpro- ducedsomeleveloffunctionalnovelty.Thisincludeddevelopments thatwereacustomizedversionofexistingproductsthatwerenot
Table2
Objectsofvalidateduserinnovations,andexamples(n=176).
Object Freq. Examples
Tools&equipment 20% Atoolthathelpstochange tyreswithlessbackpain.There arenosimilarproductsonthe market.Thisoneisforpersonal use.
Householdfixtures&furnishing 20% Afoldawaybathtub.Iam havingasmallbathroomand wantedtoavoidbigand expensiverenovationwork.
Sports,hobby&entertainment 17% Newdeviceforbeekeeping, helpsliftingthecompartments ofthebeehive.Thisisusually heavyliftingwhichneedstobe donebytwopersons,butnot anymore.
Foodandclothing 12% AhamburgermoldthatIcould notfindintheshops.Iwanted extralargehamburgers,but thetoolswerenotavailable.
Transport&vehicle 11% Ihavemademyownstuntbike footrests.Theyaremuch stronger,lighterandsaferthan availablecommercialproducts Help,care&medical 7% Toolstohelpmybrotherwho
isdisabledandwhocanonly useonearm.Hecannowpeel, diceandsliceandworkwith anythingfrombreadtofruit withonehand.
Computersoftware 6% Softwarethatisabletotake screenshotssimultaneously fromseveralcameras.Iliketo seewhathappensinmystreet.
Children&education 4% Aseatbeltcontrolthatguides thebelttocomedownoverthe collarbone/shoulderandnot forexampleoverthethroat.
Mychilddoesnotget frustratedanymorewhenthe seatbeltisinhisface.Heno longerwearsitonlypartly,e.g.
onlyonthehip,soitissafer.
Other 3% Acylinderwovenofacid-proof
steelnettoneutralizewell water.Thecylinderisfilled withdolomitelimegrainsand lowereddownintothewell.
ThepH-valueofthewellwater risesandthereforethelifeof thepipingandplumbing fixtureswillincrease.Existing productsdidnotmatchwith mysituation.
Total 100%
availableonthemarket,and that providedimportant valuefor thedeveloper.Itexcludedpurelyaestheticimprovements,asin:“I paintedapicturemoreappropriatetomysettingthananyavailable”.
(Thisisarestrictionthatfutureresearchersmaywanttorevisit:
ineffect,allpurelyartisticinnovationisexcluded).Afterthecom- pletionofourscreeningprocess,wehadasampleof176validated innovations.Table2providestheirfrequenciesandsomeexamples.
3.3. Variables
Asweexplainedearlieron,ourmajorgoalistoexploretherela- tionshipsbetweenthevalueassociatedwithproductinnovations developedbycitizensforothers,theextenttowhichrespondents hadmadeanefforttodiffusetheirinnovationscommerciallyand/or peertopeer,andhowoftencommercialand/orpeer-to-peerdif-
fusion had beenobserved.Table3 providesan overviewof the variablesintheanalyseswewillpresentlater.
First,inthesurveyweaskedrespondentsif,asfarastheycould assess,diffusionoftheirinnovationhadoccurred.Commercialdif- fusionwasmeasuredwithadichotomousindicator–‘yes’ifthe innovationwascommercialized eitherbya ventureoranexist- ingcommercialproducer–otherwise‘no’.Similarly,peer-to-peer diffusionwasmeasuredwithadichotomousindicator–‘yes’ifthe innovationwasadoptedbyotherindividualusers,–otherwise‘no’.
Next,weaskedwhethertheinnovatorhadmadeanefforttodif- fusehis/herinnovation.Diffusioneffortviacommercialpathways wasindicatedifhe/shehadshowntheinnovationtoabusinessor entrepreneur.Diffusionefforttopeerswasindicatedifhe/shehad revealedtheinnovationtootherindividuals.
Withrespecttovaluecreatedbytheinnovations,wefirstasked respondentswhetherornottheirinnovationhadservedtheirown needs.Next,weaskedfourquestions(showninTable3)todeter- minetowhatextenteachinnovatorthoughthisorherinnovation couldalsoservetheneedsofothers.Aprincipalcomponentanal- ysisthenwasapplied(detailsavailableonrequest).Thisshowed thatthefourquestionscouldbecondensedintoasingledimension, andwesavedtheresultingfactor-scoreasanindicatorofperceived generalvalue.
Asan independentcheckonthevalidity of innovators’self- assessmentofgeneralvalue,weprovidedthreeindependentcoders withtheopen-endeddescriptionsofallinnovationsinourvali- datedsample.Basedonthesedescriptions,eachoftheseindividuals codedtheinnovationsashavingvalueto:no,few,ormanyother consumers.Cohen’skappawascalculated tobe0.45, indicating fairormoderateagreementamongthecoders(LandisandKoch, 1977;Cicchetti,1994).Wenextcomputedtheaveragescoreofthe threecoderstoobtainanindependentmeasureofgeneralvalue.
Thismeasurewaspositivelyandsignificantlyrelatedtoself-rated generalvalueprovidedbytheinnovators(r=.37,p<.001).
Ourindependentcodersthusprovidesupportforthevalidity oftheself-ratedgeneralvaluemeasureprovidedbyoursample ofinnovators.However,ofthetwoavailablemeasures,weelected tousetheinnovators’self-ratinginfurtheranalyses.Wereasoned that theself-rating, although sufferingfrom potentialbiases to bediscussedinSection5.2,islikelytobethemostaccurateone.
Considerthat,formostinnovations,theopen-endeddescriptions relieduponbytheindependentcoderswerenotverydetailed.By comparison,theinnovatorsthemselveshavemuchricherinforma- tiononthenatureoftheirinnovations.Inaddition,theinnovators’
assessmentofgeneralvaluewasmadeaftertheirinnovationswere completed,puttouse,andperhapsalsoobservedbyothers.Self- ratingaccuracyshouldthereforebeadditionallyenhancedbythis post-innovationinformation–whichwasavailabletotheinnova- torsbutnottotheindependentcoders.
OthervariablesinTable3wereincludedintheregressionanal- yseswepresentlatertoinvestigatethecorrelatesofdiffusionvia commercialpathwaysorpeer-to-peer.Weincludedrespondents’
educationalattainment(dummyforthosewithatleastabache- lordegree).Wealsoincludedmotivesforinnovating,whichwere measuredbydistributing100pointsoverfivepre-definedmotives (takenfromHienerthetal.,2014).Threedummyvariableswere includedindicatingcollaborationpartnersiftheywereinvolved, reasoningthatsuchinnovationsmorelikelydiffuse.Wealsoadded respondents’willingnesstofreelysharetheirinnovation-related knowledgewithatleastsomeothers,and,likewise,adummyindi- cating iftherespondent was willingtoshare for somekindof compensation.Finally,weincludedtwocountvariablesregarding theinnovator’sperceptionofthecostanddifficultyforothersto adopttheirinnovation.Adoptionmayrequirethatadoptersinvest fromalittletoalotoftimeormoneytounderstand,replicate,and applyaninnovation.
Table3 Variables.
Variable Description Values
Commercialdiffusion Innovationwascommercializedinaventureoradoptedbyaproducerforgeneral sale
0=no;1=yes Peerdiffusion Innovationwasadoptedbyotherusersfornon-commercialpurposes 0=no;1=yes Diffusioneffort:commercial Innovatorshowedtheinnovationtoabusinessorentrepreneur 0=no;1=yes Diffusioneffort:topeers Innovatorrevealedtheinnovationtootherindividuals 0=no;1=yes
Personalvalue Responsetotheitem,‘thisinnovationworkedforme,itsolvedmypersonalneed.’ 1(barely/notatall)–4(perfectly) Generalvalue Factor-scoreoffourindicatorslistedbelow(standardizedalpha=.75;meanr=.42,
IRCs>=.50;varianceexplained57%)
Range:−1.48–2.25 ..Thisinnovationwouldbeofvaluetootherpeople(1=tonone,2=tofew,3=to
many,4=to(nearly)all)
..Ithinkthisinnovationcanbecomeavaluablecommercialproduct(1=not,2=to asmallmarket,3=toareasonablemarket,4=toasubstantialmarket)
..Myinnovationwouldenableotherpeopletodosomethingtheycouldnotdo before(0=no,1=yes)
..Myinnovationwouldhelpotherpeopletosavemoney(0=no,1=yes)
Education Dummyvariableforthosewithatleastabachelordegree 0=no;1=yes
Motives: Innovator’smotivestodeveloptheinnovation,withimportanceindicatedby distributing100pointsacross...
Personalneed ..Ipersonallyneededit 0–100points
Sales ..Iwantedtosellit/makemoney 0–100points
Learning ..Iwantedtolearn/developmyskills 0–100points
Helping ..Iwashelpingotherpeople 0–100points
Enjoyment ..Ididitforthefunofdoingit 0–100points
Collaboration: Innovationwasdevelopedincollaborationwithothers...
Relatives/friends ..Relatives/friends 0=no;1=yes
Business ..Businesses/producers 0=no;1=yes
Club/community ..Membersofacommunityorclub 0=no;1=yes
Willingnesstofreelyreveal Innovatoriswillingtofreelysharehis/herinnovation-relatedknowledge 0=no;1=yes Willingnesstotrade Innovatoriswillingtorevealhis/herinnovation-relatedforacompensation(e.g.,
money,royalties,favors,anddiscounts)
0=no;1=yes Commercialadoptionbarriers Countvariableofthreetypesofadoptioncostsforcommercialadopters(required
learningeffort,time/moneyinvestment,oranyother)
0–3barriers Peeradoptionbarriers Countvariableofthreetypesofadoptioncostsforpeers/otherusers(required
learningeffort,time/moneyinvestment,oranyother)
0–3barriers
4. Findings
We start with overall study findings (Section4.1).We next explore thevalue that innovatingconsumers derive fromtheir innovations,andfromtheprocessofdevelopingthem,aswellas perceivedvaluetoothers(Section4.2).Wethenexplorelevelsof diffusionacrossdifferentlevelsofperceivedgeneralvalueofthe innovations(Section4.3).Next,ourmainanalysisisconcernedwith thefactorsassociatedwithcommercialandpeer-to-peerdiffusion (Section4.4).
4.1. Frequencyandnatureofuserinnovationbyconsumersin Finland
Ourfirstanalysesfocusedonoursubsampleof993respondents whohadbeendrawnatrandomandrecruitedonthephone.We estimatethat inFinland,5.4%oftheconsumerpopulationaged 18–65yearshasengagedinuserinnovationinthepastthreeyears (5.9%intheunweightedsample—seeTable4notes).
Withrespecttodiffusion-relatedmatters,weseeinTable4that onlyasmallfractionofconsumerinnovatorsprotecttheirinnova- tionsfromcopyingviaintellectualpropertyrights.Wealsoseethat only19%oftheconsumer-developedinnovationsdiffuse.Ascanbe seeninTable1,thesefindingsaresimilartofindingsobtainedin theUK,theUS,andJapan.Recallthat,ontheirown,thefiguresfor diffusionshowninTable1arenotevidenceforunder-diffusion.
Consumerinnovatorsthatseektoservetheirownneedsmayonly sometimesdevelop innovationsthatare ofpotentialinterestto othersaswell.Diffusionisbeneficialonlytotheextentthatuser innovationshavevaluetoothers.
Table4
ExtentofconsumerinnovationanddiffusioninFinland.
Finlanda(n=993) Percentageofconsumerswhodeveloped
ormodifiedaconsumerproductinthe previousthreeyears...
...Inthegeneralpopulationaged18and over(forFinland,aged18–65)
5.4%
...Amongsthighlyeducated(atleast bachelordegree)
7.7%
...Amongstthoseinatechnicaljobor business
8.8%
...Amongstmales 6.3%
Estimatednumberofconsumerinnovators aged18andover(forFinland,aged18–65)
0.17million Diffusion:percentageofconsumer
innovations...
...ProtectedwithIPRs 4.7%
...Actualdiffusiontopeersand/or commercially
19.0%
aViaacomparisonwithpopulationstatisticsfor2012obtainedfromStatistics Finland,wefoundthatmales,youngercitizensaged18–24,andthosewithonly primaryeducationwereunder-represented.Thissamplingbiaswascorrectedforby computingweightsforallrespondentsacrossallcombinationsofgender,education andageclasses(detailsavailableonrequest).
4.2. Personalandgeneralvalueofinnovations
Withrespect to personal utility,most respondentsreported beinghighlysatisfiedwiththeusevalueoftheirinnovation.Inour combinedsampleof176validatedinnovations,85%ofthedevel- opersreportedthattheirinnovationverywellorperfectlysolved theirownpersonalneeds.Morespecifically,theirresponsedistri- butiontothestatement.‘Theinnovationworkedforme,itsolved
Table5
Perceivedgeneralvalueofuserinnovationsbyconsumers.
Generalvalue Thisinnovation... ...Helpsother peopletosave money(yes)
...Enablespeopletodonew things(yes)
...Wouldbevaluabletoothers (manyornearlyall)
...Canbecomeavaluable commercialproduct(toa reasonable/substantialmarket)
ClusterI:valuabletomany(17%) 70% 67% 74% 93%
ClusterII:valuabletosome(44%) 66% 68% 42% 0%
ClusterIII:valuabletonone(39%) 0% 0% 0% 3%
Note:Percentagesincellsbasedon30,77and69validatedinnovationsforclusterI,IIandIII,respectively.
Table6
Diffusionofinnovationsanddiffusioneffortacrossclustersofgeneralvalue.
Perceivedgeneralvalue Diffusionofinnovations Diffusioneffort
Anytype Peer-to-peer Commercial Topeers Commercially
ClusterI:valuabletomany 19% 12% 15% 23% 19%
ClusterII:valuabletosome 25% 19% 9% 21% 6%
ClusterIII:valuabletonone 15% 15% 0% 12% 0%
Total 19% 16% 6% 18% 6%
mypersonalneed’was‘barelyornotatall’(3%),‘somewhat’(12%),
‘verywell’(43%)and‘perfectly’(42%).
We next asked innovating respondents about the extentto whichtheythoughtthatotherswouldfindtheirinnovationsvalu- able.RecallfromSection3thatfourindicators,showninTable5, wereusedtoassessgeneralvalue.Utilizingdatacollectedforthese indicators,weappliedclusteranalysistooursample.Classifica- tionwasobtainedusingthetwo-stepclusterproceduresuggested byMilliganandSokol(1980).First,hierarchicalclusteringbased onWard’smethodwithsquaredEuclidiandistanceswasapplied togroupthereportedinnovationsintohomogeneousclusters.To assesstherobustnessofvariousclusteroptions,wesavedarangeof initialsolutionswithtwotofiveclusters.Inasecondstepwepro- ceededwithk-meansclusteranalyses,usingtheinitialhierarchical solutionsasstartingvalues.CoefficientKappa(chance-corrected coefficientofagreement)betweeneachinitialanditeratedsolu- tion(cf.Singh,1990)thenindicatedthatathree-clustersolution wasoptimal(k=.94).
NotefromTable5that 17%oftheinnovationcasesareself- assessedbytheinnovators‘likelytobeusefultomany’(Cluster I)and44%are‘likelytobeusefultosome’(ClusterII).Thisisa veryinterestingfinding–firstofkind–withrespecttothesocial welfarepotentialofconsumerinnovation.Itisinprinciplepossible thatveryfewconsumer-developedinnovationswouldbeofinter- esttoothers,duetohighheterogeneityofuserneedthathasbeen foundamongconsumers(FrankeandReisinger,2003;Frankeand vonHippel,2003).Or,itispossiblethat,despitehighheterogeneity ofneed,user-developedinnovationswillinmanycasesbeagoodor betterfittotheneedsofothersthanavailablecommercialoptions.
Empiricalfindingssuchastheonesdescribedhereareneededto understandthismatter.
AnillustrativeexampleofaninnovationinCluster1,‘likelyto beusefultomany’,is“Iamsufferingfromachingfeet.Mydeviceis differentfromtheonesavailableonthemarkets.Itgivesmassageon 80–100pressurepoints,simultaneouslyorasawave,basedonsimple processor-guidedprogram.Thenoveltyvalueisbasedonmassagethat canbeadjustedeasilyaccordingtomyneeds.Itcanimitateacupunc- turewithoutpuncturingmyskin”.Anexampleofaninnovationin Cluster3,‘likelyvaluabletononebeyondtheinnovator’was“Iwork inthegarageandthereisoftenneedfortoolsthatarenotavailableon themarkets.Forexample,Idevelopedadrillandsleevecombination thathelpsingettingbroken8mmboltoutofa5cmdeephole.Hardly newsworthy,thesenoveltoolsassistmeinmyownactivities”.
4.3. Levelsofdiffusion
InTable6,weseeinitialevidencecompatiblewithamarket failurewithrespecttothepeer-to-peerdiffusionofinnovations developedbyindividualsinFinland.Overall,ascanbeseen,only aminorityofinnovationsdeemedtobeofvaluetoothersdidin factdiffuse,anddiffusioneffortwasseldomexertedbyinnovating individuals.
Inthecaseofpeer-to-peerdiffusion,wefindthatthereisno significantrelationshipbetweenthelikelihoodofdiffusionandthe generalvalueoftheinnovation(2=.8withdf=2,p=.646).Inaddi- tion,therewasnosignificantrelationshipbetweenthelikelihood thatinnovatorsweremakinganefforttodiffusetopeers(byreveal- ingtheirinnovationtootherpotentialusers)andthegeneralvalue oftheinnovation(2=2.5withdf=2,p=.285).Bothfindingsarein linewiththepresenceofamarketfailureinthecaseofpeer-to-peer diffusion.Ifthevalueofaninnovationtoothersisentirelyanexter- nalityforindividualinnovators,thereisnoreasonthatdiffusion effortshouldbecorrelatedwiththegeneralvalueoftheinnovation.
Inthecaseofdiffusiontocommercialfirms,thingsarediffer- ent.Here,wefindasignificantrelationshipbetweenthelikelihood ofdiffusionandtheperceivedgeneralvalueofaninnovationas assessedbytheindividualinnovator(2=8.0withdf=2,p=.018).
Thereisalsoarelationshipbetweenefforttodiffusecommercially andhigherperceivedgeneralvalue(2=12.2withdf=2,p=.002).
Thesefindingsareinlinewiththeexistenceofamarketfailure affectingpeer-to-peerdiffusioneffortand diffusionaccomplish- ment in the case of innovations developed by individual user innovators.Incontrast,thefailureappearstobemitigatedinthe case ofcommercialdiffusioneffortof innovationsdevelopedby individuals.
4.4. Factorsdistinguishingdiffusingfromnon-diffusing innovations
Next,weexplorewhethertherelationshipsbetweenperceived generalvalue,diffusionanddiffusioneffortareconfirmedinamore elaborativemultivariateframework.Aswillbeseen,inthisframe- work we included as control variables educational attainment, innovationmotives,externalcollaborationindicators,willingness tofreelyrevealortradeinnovations,andperceivedadoptionbar- riers.Thisincreasedrichnessenabledustomorebroadlyassess factors distinguishing diffusing from non-diffusing innovations.
Table7providesdescriptivestatisticsandbivariatecorrelations.
Weestimatedfourprobitregressionmodels,asshowninTable8, which due to thecross-sectional natureof our data shouldbe interpreted in correlational and not causal terms. To ease the interpretation of significant effects we report marginal effect parameters.Varianceinflationfactorsoftheprobitmodelsdidnot exceed2.0,indicatingthatmulticollinearitywasnotpresent.
Thefirstregressionwassignificantandhelpfultoexplaindiffu- sionviacommercialpathways(2=28.8withdf=12,p=.004).
Weexcludedwillingnesstotradeasanindependentvariable:for allcasesthathaddiffusedcommerciallytheinnovatorwaswilling totransferhis/herknowledgeforacompensation.Thissuggests thatsuchwillingnessis a necessityforcommercialdiffusionto occur. The marginal effect estimates showed that after includ- ingthecontrolvariables, generalvalueisstill positivelyrelated withcommercialdiffusion.Inaddition,educationalattainmentwas marginallysignificant.
Withrespecttotheeffectofgeneralvalueoncommercialdif- fusion,ourestimatessuggestthatoneadditionalunitofgeneral value(i.e.,astandarddeviation,asgeneralvalueisastandardized factor-score)increasestheprobabilityofcommercialdiffusionby 5.8%.Giventheestimatedbaselineprobabilityof6.3%,athighlevels ofperceivedgeneralvaluetheprobabilityofcommercialdiffusion almostdoubles.
ThethirdmodelinTable8exploresthecorrelatesofinnova- tors’efforttodiffusecommercially.Again,willingnesstotradewas notincluded,asitperfectlycorrelatedwiththedependentvari- able.Overallmodelfit wasevenstronger comparedtothefirst model(2=35.6withdf=12,p=.000).Afterenteringthecon- trolvariablestherelationshipbetweenperceivedgeneralvalueand commercialdiffusioneffortwasstillsignificant.
Athighlevelsofgeneralvalue(onestandarddeviationaboveits meanscore)theprobabilityofexertingcommercialdiffusioneffort increasesby4.5%.Moreover,wefoundthatinnovatorsweremore likelytoexertcommercialdiffusioneffortwhentheirinnovation motivewassalesrelated,whentheinnovationwasdevelopedin collaborationwithcluborcommunitymembers,andathighlev- elsofpersonalusevalue.Overall,theseobservationsconfirmthat innovationswithhighgeneralvaluearemorelikelytospreadifa marketincentiveisoperating.
ThesecondmodelinTable8explainsthecorrelatesofaccom- plisheddiffusiontopeers,thatis,tootherusers.Afterincluding thecontrolvariableswefindthatperceivedgeneralvalueisnot relatedwithpeer-to-peerdiffusion,echoingTable6.Onlyinnova- tors’willingnesstofreelyrevealtheirinnovationseemstoincrease theprobabilitythatinnovationsspreadtootherusers.However, overallmodelfitisnotsignificant(2=15.0withdf=13,p>.10) indicatingthatourabilitytoexplainpeer-to-peerdiffusionislim- ited.
Modelfitwasacceptableinthefourthmodel,inwhichpeerdif- fusioneffortwasthedependentvariable(2=23.7withdf=13, p=.032).Here,weagainfindthatarelationshipbetweengeneral valueand peerdiffusioneffortis lacking,which is inlinewith themarketfailureweproposed.Rather,innovationsdevelopedin collaborationwithothers aremore likely tobeshowntoother individuals.
5. Discussion
Inthis study,we have extensivelyanalyzed therelationship betweentheperceivedgeneralvalueofuserinnovations,theextent towhichtheydiffuse,andtheextenttowhicheffortisexertedby userinnovatorstosupportdiffusion.Ourgoalhasbeentoinves- tigatethemeritofthe“under-diffusionofuserinnovationdueto
marketfailure”hypothesis(vonHippeletal.,2014),andtoexplore Table7 Descriptivestatistics(n=176). VariableMSD(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18) 1.Commercialdiffusion.06.24 2.Peerdiffusion.16.37.12 3.Diffusioneffort:commercial.06.24.15ˆ.17* 4.Diffusioneffort:topeers.18.38.08.30**.02 5.Personalvalue3.23.78-.01.04.09.09 6.Generalvalue.001.00.30**.02.29**.08.08 7.Education.46.50.17*−.06−.03−.01.03.08 8.Motive:personalneed51.130.5−.05−.04−.16*−.10.05−.22**.04 9.Motive:sales3.48.1.18*-.01.31**.00.00.37**.11−.28** 10.Motive:learning12.213.5.05.03.03.12.04.16*.17*−.47**.14ˆ 11.Motive:helping13.323.0−.01.03.09.00−.04.08−.15*−.61**−.08−.08 12.Motive:enjoyment19.916.7−.02.01−.01.08−.07−.03−.05−.47**.03.09−.16* 13.Collaboration:relatives/friends.22.42−.02−.03−.08.09−.07−.11.05−.04−.02−.04.04.06 14.Collaboration:business.05.21.18*.14ˆ.18*.19*.08.16*.08−.02.18*−.10.02.00.02 15.Collaboration:club/community.04.20.07.16*.20*.14ˆ−.06.07.11−.15ˆ−.05.18*.11.01−.04.09 16.Willingnesstofreelyreveal.84.37−.24**.10−.17*.02.04−.32**−.05.13−.46**−.13.07−.01−.07−.13−.07 17.Willingnesstotrade.91.28.08−.06.08.07.06.12.02.06−.04−.07.13−.21**−.09−.03.06.23** 18.Commercialadoptionbarriers.18.39.14ˆ.11.08.07−.08.15ˆ.00−.11.10.15ˆ.00.03−.14ˆ.11.21**−.06.09 19.Peeradoptionbarriers.26.47.08.07.08−.06−.01.17*.06−.19*.12.14ˆ.01.16*−.03.18*.21**−.15ˆ−.06.44**
Table8
Probitregressionmodelsofdiffusion,anddiffusioneffort(n=176).
Diffusionobserved Diffusioneffort
Commercial Peer-to-peer Commercial Peer-to-peer
dy/dx dy/dx dy/dx dy/dx
Baselineestimate .063 .164 .060 .176
Effectparameters:
Education .055ˆ (.033) −.064 (.054) −.045 (.031) −.043 (.058)
Motive:sales −.001 (.003) .002 (.005) .005* (.002) −.001 (.004)
Motive:learning .000 (.001) .002 (.003) −.001 (.001) .005* (.002)
Motive:helping −.000 (.001) .000 (.001) .000 (.000) .000 (.001)
Motive:enjoyment −.000 (.001) .000 (.002) .000 (.001) .003 (.002)
Collaboration:relatives/friends −.005 (.041) .009 (.069) −.027 (.033) .141* (.069)
Collaboration:business .059 (.077) .248 (.178) .002 (.044) .442* (.175)
Collaboration:club/community −.024 (.041) .336 (.212) .448* (.187) .362 (.215)
Willingnesstofreelyreveal −.050 (.062) .157** (.047) .002 (.045) .080 (.082)
Willingnesstotrade −.175 (.138) .100 (.091)
Commercialadoptionbarriers .047 (.035) −.029 (.036)
Peeradoptionbarriers .027 (.063) −.121 (.072)
Personalvalue −.006 (.018) .012 (.035) .057** (.021) .041 (.034)
Generalvalue .058* (.023) .007 (.033) .045* (.018) .045 (.031)
Modelfit:
Wald2(df) 28.8(12) 15.0(13) 35.6(12) 23.7(13)
Significance(Waldp-value) .004 .304 .000 .032
PseudoR2 .299 .090 .415 .170
Note:Robuststandarderrorsinparentheses.Forinnovationmotives,personalneedisthereferencegroup.Two-tailedsignificance**p<.01,*p<.05,ˆp<.10.
factorsaffectingit.Ouranalyseswerebasedonabroadsampleof consumersinFinland.
Intheremainderofthissection,wediscusstheevidenceinfavor oftheunder-diffusionofconsumerinnovationhypothesis.Then, wediscussthelimitationsofourstudyandoffersuggestionsforfur- therresearch.Finally,weconsiderwhetheranythingcanbedoneto increasediffusionofgenerallyvaluableinnovationsthatconsumers develop.
5.1. Evaluatingthemarketfailurehypothesis
Themarket failurehypothesis proposedin von Hippelet al.
(2014)offersa reason whygenerally valuable userinnovations mightnotdiffuse.Wefindthattheirhypothesisisinlinewithour empiricalfindings.
Unlike producers, consumers develop innovations for them- selvesprimarily–andthereisnonecessaryreasonwhyconsumer creationswouldalsooffervaluetoothers.Nonetheless,wefound that61%ofuserinnovationswerebelievedbytheirdevelopersto bevaluabletoatleastsomeotherusers–andsotobeapoten- tialsourceofbroadersocialandeconomicvalue.Ofcourse,this broadervalueisonlyrealizedtotheextentthattheinnovations actuallydiffuse.Inlinewiththemarketfailurehypothesiswehave exploredinthispaper,wefoundthatfewinnovatorsexertaneffort todiffusetheirinnovationscommerciallyortootherusers.Only6%
reportedthattheyhadexertedefforttoshowtheirinnovationtoan entrepreneurorbusinessforgeneralsale,while18%hadrevealed theirinnovationtootherindividuals.Asforobserveddiffusion,16%
oftheconsumer-developedinnovationsdiddiffusepeertopeer, while6%diffusedviacommercialpathways.
Notably,innovators’reasonfornottryingtodiffusecannothave beenawishtoprotecttheiridea,e.g.,forreasonsofrivalrywith otherusers.Veryfewrespondentshadappliedforintellectualprop- ertyrights,and84%oftheinnovatorssaidtheywerewillingto freelyrevealtheirinnovation.Thesituationcanbeexplainedby thelackofaconnectionbetweenthebenefitsobtainedbyadopters andanyinvestmentsbyconsumerinnovatorstodiffusetheirinno- vations.Adopters’benefitsareanexternalityfromtheperspective ofconsumerinnovatorswhofreelyreveal.
Takingatfacevaluetheinnovator’ssubjectiveassessmentofthe valueoftheinnovationtoothers,wecansaythattheinnovator’s decisionnottoshareisindividuallyrational,butsociallyinefficient.
Otherswouldbenefit,intheinnovator’sownassessment,ifshe/he decidedtosharemore;andyetshe/hedoesnotdoso.Findingsfrom theprobitregressionspresentedaboveclearlyshowthatdiffusion efforttopeers,anddiffusionaccomplishedtopeers,arenotrelated toperceivedgeneralvalue.No matterwhatvaluetheinnovator believeshisinnovationtohaveforothers,diffusionisunaffected.
Thisdisconnectindicatesanexternalityastherootcause:valueto othersisnotinternalizedandthereforenotfactoredintodiffusion effortdecision-making.
Bycontrast,wefindthatinthepresenceofamarketincentive, innovatorsaremorelikelytoputeffortintodiffusion.Inthecase ofthecommercialdiffusionpathway,themarketfailurewehave documentedinthecaseofpeer-to-peerdiffusionseemedreduced inintensity.Thisisreasonable,giventhatamarketincentiveslink canthensometimesexisttorewardeveninnovatorswhofreely revealtheirinnovations.
5.2. Limitationsandsuggestionsforfurtherresearch
Thisstudyhaslimitationsthatshouldbedealtwithinfuture research.Wedescribethreebelowwhichallarisefromthefactthat ourresearchutilizedtheinnovators’ownperceptionofthegeneral valueoftheirinnovation,ratherthanrelyingupontheevaluations ofindependentraters.Wefeeltherearebenefitsfromthischoice, butitisalsotruethatrelianceoninnovators’self-evaluationsdoes introducesomepotentialproblems.
Onthebenefitsside,itisimportanttounderstandthattheinno- vators’view–however,imperfect–ispreciselytheoneneeded forouranalysisofpossiblemarketfailure.Themarketfailurewe explorerestsonwhetherinnovatingusersthinktheirinnovations areorarenotofgeneralvalue,andwhethertheirviewonthismat- teraffects theefforttheydevoteefforttodiffusion.Aswehave seen,thisconnectionisnotsignificantlypresentinthecaseofpeer- to-peerdiffusion,whichisconsistentwithadiagnosisofmarket failure.