Market failure in the diffusion of consumer-developed innovations : patterns in Finland

(1)

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

^g

aRSMErasmusUniversity,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.

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

(2)

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£^3.2^billion.Încomparison, estimatedannualR&Dexpendituresbycompaniesonconsumer productswere£^2.2^billion.^Similar^findings^have^been^reported^for 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%

(3)

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-testsand␹²-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-endeddescriptions 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

(4)

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.

(5)

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.

Finland^a(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

(6)

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.

(7)

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**

(8)

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:

Wald␹²(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.