1

FourPathstoCompatibility

JosephFarrell*UniversityofCalifornia,Berkeley

and

TimothySimcoeBostonUniversitySchoolofManagementandNBER

AchapterfortheOxfordHandbookoftheDigitalEconomy,editedbyMartinPeitzandJoelWaldfogel.

THISDRAFT:March3,2011

ABSTRACT:Wedescribefourwaystoachieveproductcompatibility:decentralizedadoption,negotiationinaconsensusStandardSettingOrganization(SSO),followingaleader,andusingconvertersormulti‐homing.Eachmeanshascostsandbenefitsintermsofthelikelihoodofcoordination,thetimeandresourcesinvolved,andtheimplicationsforexpostcompetitionandinnovation.Wediscusswhatdetermineswhichtechnologiesfollowwhichpathtocompatibility,andconsiderhybridmechanismsthatcombinetwoormorepaths.

*ProfessorFarrelliscurrentlyonleaveservingasDirectoroftheBureauofEconomicsattheFederalTradeCommission,butthisarticleisauthoredinhisacademiccapacityanddoesnotpurporttorepresenttheviewsoftheCommissionoranyCommissioner.

2

1.IntroductionCompatibility standards are design rules that promote product inter‐operability,such as the thread‐size for mechanical nuts and bolts or the communicationprotocolssharedbyall Internetdevices.Products thatadheretostandardsshouldwork together well, which produces a range of benefits: users may shareinformation,or“mixandmatch”components;thecostofmarketentrydeclines;andthere is a division of labor, enabling specialization in component production andinnovation. This chapter describes four paths to compatibility – standards wars,negotiations,dictatorsandconverters–andexploreshowandwhentheyareused,asalternativesorincombination.While product inter‐operability may pose engineering challenges, we focus oneconomic incentive issues that arise when its costs and benefits are not evenlydistributed. For example, firms that control a technology platform may resistcompatibilitywithothersystems,orstandardsthatcouldreduceswitchingcostsfortheir installedbase.Specializedcomponentproducersmay fightagainststandardsthatthreatento“commoditize”theirproducts.Evenwhensupportforcompatibilityis widespread, rival firms may advocate competing designs that confer privatebenefitsbecauseofintellectualpropertyrights,lead‐timeadvantagesorproprietarycomplements.Giventhismixofcommonandconflictinginterests,wefocusonfournaturalwaystocoordinate design decisions (in the sense of achieving compatibility). The first isdecentralizedchoice,whichcanyieldcoordinatedoutcomeswhennetworkeffectsare strong, even if the resulting process is messy. Negotiations are a secondcoordinationmechanism.Inparticular,firmsoftenparticipateinvoluntaryStandardSettingOrganizations(SSOs),whichseekabroadconsensusonaspectsofproductdesignbeforeendorsingaparticulartechnology.Athirdroutetocompatibilityistofollow the lead of an influential decision‐maker, such as a large customer orplatformleader.Finally,participantsmayabandoneffortstocoordinateonasinglestandard and instead patch together partial compatibility through converters ormulti‐homing.1These four paths to compatibility have different costs and benefits,which can bemeasured in time and resources, the likelihood of successful coordination forcompatibility,andtheexpostimpactoncompetitionandinnovation.Whetherthesecomplexwelfare trade‐offs arewell internalized depends on how (and bywhom)the path to compatibility is chosen. A full treatment of the compatibility problemwouldspecify the selectionprocessandquantify the relativeperformanceofeachpath. In practice, while theory clarifies the potential trade‐offs, we have limitedempirical evidence on the comparative costs and benefits of each path, or theprecisenatureoftheselectionprocess.1Auserissaidto“multi‐home”whenitadoptsseveralincompatiblesystemsandcanthusworkwithothersonanyofthosesystems.

3

Sometimes the choice of a particular path to compatibility is a more‐or‐lessconsciousdecision.Forexample,firmscandecidewhethertojointhedeliberationsof an SSO or follow the lead of a dominant player. A dominant player can decidewhethertocommittoastandardandexpect(perhapshope)tobefollowed,ordefertoaconsensusnegotiation.Astheseexamplessuggest,itcanbeacomplexquestionwho,ifanyone,“chooses”themechanism,ifany,usedtocoordinate. Somemarketforces push toward efficiency, but it is not guaranteed. For example, a platformleaderhasageneralincentivetodictateefficientinterfacestandardsortoallowanefficient evolution process, but that incentive may coexist with, and perhaps beoverwhelmed by, incentives to stifle ex post competition. Likewise, competitionamongSSOsmayormaynotleadthemtowardbetterpolicies,andstandardswarsmayormaynottiptowardsthesuperiorplatform.Sometimes firms will start down one path to compatibility and then veer ontoanother.For instance,adecentralizedstandardswarmayberesolvedbyresort toanSSOorthroughtheinterventionofadominantfirm.SlownegotiationswithinanSSO can be accelerated by evidence that the market is tipping, and platformsponsorsmaypromotecomplementaryinnovationbyusinganSSOtoopenpartsoftheirplatform.Whiletheorysuggeststhatcertain“hybridpaths”canworkwell,weknow rather little about how different coordination mechanisms complement orinterferewithoneanother.This chapter begins by explaining something familiar to many readers: how thechoice of inter‐operability standards resembles a coordination game in whichplayershaveamixofcommonandconflicting incentives. Inparticular,weexplainhow compatibility produces broadly shared benefits, and discuss several reasonswhy firms may receive private benefits from coordinating on a preferredtechnology.Section3describescostsandbenefitsofourfourpathstocompatibility.Section 4 examines the selection process and the role of “hybrid” paths tocompatibility.Section5concludes.2.CostsandBenefitsofCompatibilityWhen all influential players favor compatibility, creating or upgrading standardsinvolves a coordination problem. When there is but one technology, or whenparticipantssharecommongoalsandnotionsofquality,thesolutionisprimarilyamatterofcommunicationthatcanbesolvedbyholdingameeting,orappointingafocal adopter whom all agree to follow. But if there are several technologies tochoose from and participants disagree about their relative merit, it turns a purecoordination game into a battle of the sexes, where players may try to “win” byarguingfor,orcommittingto,theirpreferredoutcome.

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐INSERTFIGURE1ABOUTHERE

4

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Figure1 illustrates thebasicdilemmainasymmetric two‐playergame.As longasC>B, the benefits of compatibility outweigh the payoffs from uncoordinatedadoptionofeachplayer’spreferredtechnology,andthegamehastwopure‐strategyNashEquilibria:bothadoptA,andbothadoptB.Eachplayergainssomeadditionalprivate benefit (equal to D) in the equilibrium that selects their preferredtechnology. When these private benefits are small (D ≈ 0), many coordinationmechanismswouldworkwell.ButasDgrowslarge,playerswillpushhardfortheirpreferredequilibrium.Whether theseefforts topromoteaparticularoutcomearesocially productive depends on a variety of factors about the players’ availableactionsandthedetailsoftheequilibriumselectionprocess.Below,weassumethatD>0, and compare the costs and benefits of four broadmethods for choosing anequilibrium. But first, this section explains why the payoffs in Figure 1 can be asensible way to model the choice of compatibility standards, particularly in theInformationandCommunicationstechnology(ICT)sector.Thebenefitsofcompatibility(C‐BinFigure1)comeintwoflavors:horizontalandvertical.Horizontalcompatibilityistheabilitytosharecomplementsacrossmultipleplatforms, and we call a platform horizontally open if its installed base ofcomplementscanbeeasilyaccessedfromrivalsystems.ManypartsoftheInternetare in this sense horizontally open. For example, web pages can be displayed oncompetingwebbrowsers,andrivalinstantmessengerprogramsallowuserstochat.Video game consoles and proprietary operating systems, such as MicrosoftWindows, by contrast, are horizontally closed: absent further action (such as“porting”),anapplicationwrittenforoneisnotusableonothers.These distinctions can be nuanced. For example, what if a platform’s set ofcomplements is readily available to users of rival platforms, but at an additionalcharge,aswithmanybanks’ATMnetworks?Similarly,Microsoftmayhavechoices(suchasthedegreeofsupportofferedtocross‐platformtoolslikeJava)thataffect,but do not fully determine, the speed and extent to which complements forWindowsbecomeavailableonotherplatforms.Benefitsofhorizontalcompatibilityincludetheabilitytocommunicatewithalargerinstalledbase(directnetworkeffects)andpositivefeedbackbetweenthesizeofaninstalled base and the supply of complementary goods (indirect network effects).Katz and Shapiro (1985) analyzed oligopoly with firm‐specific demand‐sideincreasing returns.Amore recent literatureonmany‐sidedplatforms (e.g.RochetandTirole2003;ParkerandVanAlstyne2005;Weyl2010)extendstheanalysisofindirectnetworkeffectsbyallowingexternalitiesandaccessprices tovaryacrossdifferentusergroups.22SeeDavidandGreenstein(1990)orShapiroandVarian(1998)forareviewoftheearlyliteratureonnetworkeffects,andRysman(2009)forareviewofthenascentliteratureontwo‐sidedmarkets.

5

Vertical compatibility is the ability of those other than the platform sponsor tosupply complements for the system. We call a platform vertically open ifindependent firms can supply complementswithout obtaining a platform leader’spermission.3 For example, the Hush‐a‐Phone case (238 F.2d 266, 1956), and theFCC’slaterCarterfonedecision(13F.C.C.2d420)openedtheU.S.telephonenetworkto independently supplied attachments such as faxes, modems and answeringmachines. Many computing platforms, including Microsoft Windows, use verticalopenness to attract independent software developers. Like horizontal openness,vertical openness can be a matter of degree rather than a sharp distinction. Forinstance, aplatform leadermayoffer technically liberal accesspoliciesbut chargeaccessfees.Vertical compatibility produces several types of benefits. There are benefits fromincreased variety when vertical compatibility allows users to “mix and match”components (Matutes andRegibeau 1988). Vertical openness also can reduce thecostofentry,strengtheningcompetitionincomplementarymarkets.Finally,verticalcompatibility leads to a “modular” system architecture and division of innovativelabor. Isolating a module that is likely to experience a sustained trajectory ofimprovements allows other components to take advantage of performance gainswhileprotectingthemfromthecostofredesign.Andwhenthelocusofdemandorthevalueofcomplementaryinnovationsishighlyuncertain,modularityandverticalopenness facilitate simultaneous design experiments (Bresnahan and Greenstein1999;BaldwinandClark2000).The benefits of horizontal or vertical compatibility are often broadly shared butneed not be symmetric: there can also be private benefits of having a preferredtechnologybecometheindustrystandard.Suchprivatebenefits,labeledDinFigure1,oftenleadtoconflictandcoordinationdifficultiesinthesearchforcompatibility.Oneimportantsourceofconflictisthepresenceofaninstalledbase.Upgradinganinstalledbasecanbecostly,andfirmstypicallyfavorstandardsthatpreservetheirinvestmentsinexistingdesigns.Moreover,platformleaderswithalargeinstalledbasewillfavordesignsthatpreserveorincreaseswitchingcosts,whileprospectiveentrantspushtoreducethem.Forexample,initsU.S.antitrustcase,MicrosoftwasconvictedofusingillegaltacticstopreventWindowsusers,developersandOEMsfrommigratingtotheindependentstandardsembodiedintheNetscapebrowserandJavaprogramminglanguage.3Distinguishingbetweenhorizontalandverticalcompatibilitymayhelpilluminatetheoftenmurkyconceptofopenstandards.Cargill(1994)suggeststhattheterm`open’hasbecome“anicontoconvenientlyrepresentallthatisgoodaboutcomputing,”sowhenconflictsemerge,allsidesclaimsupportofopenstandards.End‐userstypicallydefine“open”inhorizontalterms,sincetheyseekacommitmenttofuturecompetitionattheplatformlevel.Platformleaderstypicallyemphasizeverticalcompatibility,whichgrantsaccessto(butnotcontrolover)proprietarytechnology.Meanwhile,standardsmavenscallatechnologyopenifithasbeenendorsedbyanaccreditedSSO,andopen‐sourceadvocatesfocusonfreeaccesstotheunderlyingcode.

6

Designleadsareanothersourceofconflict.ShortICTproductlifecyclesleavefirmsa limitedwindowofopportunity to capitalizeon thedemandunleashedbyanewstandard,andfirst‐moveradvantagescanbeimportant.Thus,firmsmaytrytoblockordelayanewstandardifrivalshaveasignificantleadatimplementation.DeLacyetal(2006)describesucheffortsinthecontextofWi‐Fistandardsdevelopment.In some cases, there is conflict over the location of module boundaries, or whatengineers call the “protocol stack.” Since compatibility often promotes entry andcompetition, firms typically prefer to standardize components that complementtheirproprietarytechnology,butleaveroomfordifferentiationinareaswheretheyhave a technical edge. For example, Henderson (2003) describes how thenetworkingtechnologystart‐upEmberallegedlyjoinedseveralSSOstopreventnewstandardsfromimpingingonitscoretechnology.Conflictscanalsoemergewhenfirmsownintellectualpropertyrightsinaproposedstandard,whichtheyhopeto licenseto implementers,oruseas leverageinfuturenegotiations. Lerner, Strojwas and Tirole (2003) show that nearly all “modern”patentpoolsarelinkedtocompatibilitystandards,andSimcoe(2007)documentsarapid increase in intellectual property disclosures in the formal standard settingprocess. While data on licensing are scant, Simcoe, Graham and Feldman (2009)showthatpatentsdisclosedintheformalstandardsprocesshaveanunusuallyhighlitigationrate.Finally, conflicting interests can amplify technological uncertainty. In particular,when technical performance is hard tomeasure, firms and userswill growmoreskeptical of statements fromself‐interestedparticipants about thequalityof theirfavoreddesign.3.FourPathstoCompatibilityGiven thismix of conflict, common interest and incomplete information, choosingcompatibility standards can be a messy process. This section considers theperformance of four paths to compatibility – standardswars, SSOs, dictators andconverters–intermsoftheprobabilityofachievingcompatibility,theexpectedtimeandresourcecosts,andtheimplicationsforexpostcompetitionandinnovation.Wefind that economic theory helps articulate some of the complex trade‐offs amongthesepaths,butthereislittlesystematicevidence.3.1StandardsWarsStandardswarscanbesponsoredorunsponsored;forbrevitywefocushereonthesponsoredvariety,inwhichproponentsofalternativetechnologiesseektopreemptoneanotherinthemarketplace,eachhopingthatdecentralizedadoptionwillleadtotheir own solution becoming a de facto standard through positive feedback andincreasing returns. Standards wars have broken out over video formats, modem

7

protocols,Internetbrowsers,andtransmissionstandardsforelectricityandcellularradio.Thesewarscanbeintensewhenhorizontallyincompatibleplatformscompeteforamarketwithstrongnetworkeffects,whichtheyexpecttotiptowardsasinglewinnerwhowill likely acquiremarket power. Muchhas beenwritten about thetactics and outcomes in such wars, and we do not attempt to be comprehensivehere,onlytoremindthereaderofsomeofthedynamics.4Standards wars often involve a race to acquire early adopters and efforts tomanipulateuserexpectations,asdescribedinBesenandFarrell(1994)orShapiroand Varian (1998). Pre‐emption is one strategy for building an early lead in anadoption race. Another strategy is to aggressively court early adopters withmarketing, promotions and pricing. Firms may also work to influence users’expectationsregardingthelikelywinner,sincethesebeliefsmaybeself‐fulfilling.5Firmsthatfallbehindinaraceforearlyadoptersorexpectationsmayusebackwardcompatibility or bundling to catch up. Backward compatibility jump‐starts thesupply of complements for a new platform. For instance, many video gameplatformscanplaygameswrittenforolderconsolessoldbythesamefirm.Bundlingpromotes the adoption of new standards by linking them to existing technologyupgrades. For example, Sony bundled a Blu‐ray disc player with the Playstationgameconsole topromote thatvideo formatoverHD‐DVD,andBresnahanandYin(2007) argue that Microsoft took advantage of the Windows upgrade cycle toovertakeNetscapeinthebrowserwars.Given the rangeof tacticsused in a standardswar,doesdecentralized technologyadoptionprovideanattractiveroutetocoordination?Onesocialcost isthat itwilloften lead to the emergence of a horizontally closed platform. While one mightquestion the direction of causality (perhaps intractable conflicts over horizontalinter‐operabilityleadtostandardswars),alternativepathstocoordinationmayleadtogreaterexpostcompetitionandreducetheriskofstrandedinvestments.Theeconomiclogicofstandardswarsseemsconsistentwithconcernsthatmarketsmay “tip” prematurely and/or towards an inferior solution. While many cite theQWERTYkeyboardlayoutasanexample(e.g.David1990),LiebowitzandMargolis(1990) dispute the empirical evidence, and suggest that markets will typicallycoordinate on the best available technology, as long as the benefits of changingplatforms outweigh any switching costs. It is difficult to find natural experimentsthatmightresolvethisdebate,e.g.byrandomlyassigninganearly lead insettingswhere there are clear differences in platform quality. But even if standardswars

4ShapiroandVarian(1998)providemanyotherexamples,andWest(2009)containsalengthylistofstandardswars.FarrellandKlemperer(2007)reviewtheeconomictheory.5FarrellandSaloner(1986)modelsequentialtechnologyadoptionwithnetworkeffectsandshowhowoutcomesmaydependonusers’initialbeliefs.ThenovelStartUp(Kaplan,1986Ch.9)providesanentertainingaccountofthebattleforexpectations,andthestrategicuseofbackwardscompatibility,inpen‐basedcomputeroperatingsystems.

8

typically“get itright” intermsofselectingforquality,coordinationproblemsmayaffectthetimingofstandardsadoption.6Optimists argue that standards wars are speedy, since participants have strongincentivestoraceforearlyadopters,andthatfierceexantecompetitionoffsetsanysocial cost ofex post incompatibility. But competition for early adopters does notalwaystaketheformofcreatingandtransferringsurplus,anditsbenefitsmustbeweighedagainst thecostsof stranded investments ina losingplatform.Moreover,theuncertaintycreatedbyastandardswarmaycauseforward‐lookingusers,whofearchoosingthelosingplatform,todelaycommitmentsuntilthebattleisresolved.For example,Dranove andGandal (2003) find that preannouncement of theDIVXformat temporarily slowed the adoption of DVD. Augeau, Rysman andGreenstein(2006) suggest that the standards war in 56K modems also delayed consumeradoption.When it is costly to fight a standardswar,participantsmay seekanescape route,suchassometypeoftruce.7Forexample,the56Kmodemstandardswarendedinadoption of a compromise protocol incorporating elements of both technologies.ThebattlebetweenCDMAandTDMAcellularphonetechnologyendedinaduopolystalemate,witheachstandardcapturingasignificantshareoftheglobalmarket.Ironically, these escape routes and stalemates illustrate a final strength ofdecentralizedadoptionasapathtocompatibility:itcanrevealthatnetworkeffectsare weak, or that technologies initially perceived as competing standards canultimately coexist by serving different applications. For example, Bluetooth (IEEE802.15)was conceived as a homenetworking standard, but ceded thatmarket toWi‐Fi(IEEE802.11)andisnowwidelyusedinshort‐rangelow‐powerdevices,suchas wireless headsets, keyboards and remote controls. Similarly, the plethora ofdigital image formats (JPEG, GIF, TIFF, PNG, BMP, etc.) reflect trade‐offs betweenimage‐qualityandcompression,aswellascompatibilitywithspecificdevices.Since“war” isapoormetaphor for theprocessofmatchingdifferentiated technology tonichemarkets,Updegrove(2007)hasproposedthealternative labelof “standardsswarms” for settingswhere network effects areweak relative to the demand forvariety.3.2StandardSettingOrganizationsOne alternative to standards wars is for interested parties to try and coordinatethroughnegotiation.Thisprocessisoftencalledformalordejurestandardsetting,andtypicallyoccurswithinconsensusStandardSettingOrganizations(SSOs).

6CabralandKretschmer(2007)evensuggestthatwhentippingtowardsaninferiortechnologywouldbeverycostly,optimalgovernmentpolicymaybetoprolongastandardswarsoparticipantscangathermoreinformation.7Interestingly,manyofthewell‐knownstandardswarsthatdoresultina“fighttothedeath”involvemediaformats.

9

There are hundreds of SSOs, and many of these non‐profit institutions developstandards for safety and performance measurement, as well as productcompatibility.8Weuse abroaddefinitionof SSO that includes globally recognized“big I” standard setters, such as ITU and ISO; private consortia that manage aparticularplatform,suchastheIETFandW3C;andsmallerconsortiathatfocusonaparticulartechnology,suchastheUSBForumortheBlu‐rayDiscAssociation.9Thisdefinition could even be stretched to include collaborative product‐developmentgroups, such as open‐source software communities. While the largest SSOs havehundreds of sub‐committees and maintain thousands of specifications, smallconsortiacanresemblejointventures,whereinaselectgroupoffirmsdevelopandcross‐licenseasingleprotocolunderaso‐calledpromoter‐adopteragreement.Standards practitioners typically distinguish between consortia and “accredited”Standards Developing Organizations (SDOs). SDOs sometimes receive preferentialtreatment in trade, government purchasing, and perhaps antitrust in return foradhering tobestpracticesestablishedbyanational standardsagency, suchas theAmericanNationalStandardsInstitute(ANSI).10Table1hintsatthesizeandscopeofformalstandard‐settingintheUnitedStatesbycountingentriesinthe2006ANSIcatalog of American National Standards and listing the twenty largest ANSI‐accreditedSDOs.11

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐INSERTTABLE1ABOUTHERE

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐SSOs use a consensus process to reach decisions. Though definitions vary,consensustypicallyimpliessupportfromasubstantialmajorityofparticipants.For8Alistofroughly550SSOsisavailableatwww.consortiuminfo.org.Cargill(2002)andtheIntellectualPropertyOwnersAssociation(2009)suggestclassificationschemes.9TheseacronymsstandforInternationalTelecommunicationsUnion(ITU),InternationalOrganizationforStandards(ISO),InternetEngineeringTaskForce(IETF),andWorldWideWebConsortium(W3C).ForICTstandards,ISOandtheInternationalElectrotechnicalCommission(IEC)collaboratethroughagroupcalledtheJointTechnicalCommittee(JTC1).MurphyandYates(2009)describetherelationshipbetweenthesenationalstandardsbodiesandtheglobalstandardssystemadministeredbyISO.10Annex1oftheWorldTradeOrganization’sTechnicalBarrierstoTradeAgreementmakesexplicitreferencetotheISO/IECGuidelinesforSDOprocedure.IntheUnitedStates,OMBCircularA‐119givespreferencetoSDOsinfederalgovernmentpurchasing,andtheStandardsDevelopmentActof2004(HR1086)grantscertainantitrustexemptions.SpeechesbyMajoras(2005)andVarney(2010),andalsotheEuropeanCommission(2010)antitrustguidelinesonhorizontalco‐operationprovidesomeassurancetoSSOsregardingopendiscussionofroyaltyrates.11Compatibilitystandardsmakeuproughly43percentofthetotalstockofAmericanNationalStandards,withmuchoftheotherhalfrelatedtoperformancemeasurementandsafety.ThustheICTsector’sshareofstandardsproductionexceedsitsshareofGDP,andevenpatenting.Oneexplanationisthatinformationtechnologyis,bydesign,uniquelymodular,sotheratioofstandardstoproductsishigh.Forexample,Biddleetal(2010)estimatedthatatypicallaptopimplementsbetween250and500differentcompatibilitystandards.

10

example,mostaccreditedSDOsrequireasuper‐majorityvoteandaformalresponsetoany“goodfaith”objectionsbeforeapprovinganewstandard.SinceSSOstypicallylackenforcementpower,thisscreeningprocessmayserveasasignalofmembers’intentionstoadoptastandard,oranefforttoswaythemarket’sbeliefs.RysmanandSimcoe (2008) provide some empirical evidence that SSOs’ non‐bindingendorsementscanpromote technologydiffusionbystudyingcitationrates forU.S.patents disclosed in the standard‐setting process, and showing that an SSOendorsementleadstoameasurableincreaseinforwardcitations.Beyond using a loosely defined consensus process and relying on persuasion andnetwork effects to enforce their standards, SSOs’ internal rules and organizationvarywidely.Someareopentoanyinterestedparticipant,whileotherschargehighfeesandlimitmembershiptoaselectgroupoffirms.SomeSSOshaveacompletelytransparent process, while others reveal little information. Some SSOs requirememberstograntaroyalty‐freelicensetoanyintellectualpropertycontainedinastandard,whileothersarecloselyalignedwithroyalty‐bearingpatentpools.Therehasbeenlittleempiricalresearchonthe internalorganizationofSSOs,butLemley(2002)andChiao,LernerandTirole(2007)examinevariationinSSOs’intellectualpropertyrightspolicies.Given SSOs’ heterogeneity, what can we say about the costs and benefits of theconsensus process as a path to coordination? Since SSOs encourage explicitcomparisons and often have an engineering culture that emphasizes the role oftechnical quality, there is some reason to expect higher‐quality standards thanwould emerge from a standards war or an uninformed choice among competingtechnologies.ThispredictionappearsinthestochasticbargainingmodelofSimcoe(2010), aswell as thewar‐of‐attritionmodelofFarrell andSimcoe (2009),whereSSOsprovideaquality‐screeningmechanism.But technical evaluation and screening for quality can impose lengthy delays,especially when the consensus process gives participants the power to blockproposed solutions. A survey by theNationalResearchCouncil (1990) found thatstandards practitioners viewed delays as a major problem, and Cargill (2001)suggeststhattheopportunitycostsofdelayedstandardizationexplainabroadshiftfrom accredited SDOs towards less formal consortia. Farrell and Saloner (1988)developa formalmodel to compareoutcomes ina standardswar (grab‐the‐dollargame)toanSSO(warofattrition).Theirtheorypredictsabasictrade‐off:theformalconsensusprocessleadstocoordinationmoreoften,whilethestandardswarselectsawinnermorequickly.SSOs have sought ways to limit deadlocks and lengthy delays. Some grant aparticular party the power to break deadlocks, though such unilateral decisionscouldbeviewedasadistinctroutetocompatibility(seebelow).Anotherapproachis to start early in the life of a technology, before firms commit to alternativedesigns.Illustratingtheimpactofcommitmentondelays,Simcoe(2010)showshowdelays at the IETF increased as the Internetmatured into a commercial platform.

11

Butearlystandardizationalsohasdownsides;inparticular,privateactorshavelittleincentive to contribute technology if they see no commercial opportunity, soanticipatorystandardsrelyheavilyonparticipationfrompublicsector institutionssuchasacademiaorgovernmentlabs.Athirdwaytoresolvedeadlocksistoagreeonpartialorincompletestandards.Suchstandards often include “vendor specific options” to facilitate productdifferentiation.AndSSOparticipants sometimesagree to a “framework” thatdoesnot achieve full compatibility, but standardizes those parts of an interfacewherecompromise can be reached (thus lowering the ex post cost of achievingcompatibilitythroughconverters).12Finally,SSOsmayworkfasterifcompetinginterestsareplacedinseparateforums,such as independent working groups within a large SSO or even independentconsortia.LernerandTirole(2006)modelforumshoppingwhenthereisfreeentryinto certification, and show that technology sponsors will choose the friendliestpossibleSSO,subjecttotheconstraintthatcertificationswaysuserbeliefsenoughtoinduceadoption.This“competingforums”approachworkswell ifthereisdemandfor variety and converters are cheap. But if network effects are strong, forumshoppingmayproduceescalatingcommitmentsinadvanceofastandardswar.Forexample, the Blu‐ray and HD‐DVD camps each established an independentimplementers’forumtopromotetheirownvideoformat.Beyondprovidingaforumfornegotiationandcertificationactivities,SSOsareoftena locus of collaborative research and development. Thus, onemight ask whetherselectingthispathtocoordinationhassignificantimplicationsforinnovation?SomeformsofinnovationwithinSSOsraiseapublicgoodsproblem:incentivesareweakifallfirmshavefreeaccesstoimprovements,especiallyinhighlycompetitiveindustries.Weiss and Toyofuku (1996) gather evidence of free riding in 10BaseTstandards development. Cabral and Salant (2008) study a model wherestandardization leads to free riding in R&D, and find that firms may favorincompatibilityifithelpsthemsustainahighrateofinnovation.Eisenmann(2008)suggests that SSOsoften strugglewith “architectural” innovations that spanmanycomponent technologies, since it is difficult to coordinate the decisions ofspecialized firms with narrow interests in the outcomes of a particular workinggrouportechnicalcommittee.However,suchproblemsneednotpreventall innovationwithinSSOs.Firmsoftencontributeproprietarytechnologytoopenplatforms,indicatingthatthebenefitsofstandardizingapreferredtechnologyoutweighthetemptationtofree‐rideinthosecases. Where SSOs encourage horizontal openness, that should encourage

12Thedecisiontoadoptaframeworkorincorporatevendor‐specificoptionsintoastandardmaybeobservable,andhenceamenabletoempiricalresearch,sincethework‐processandpublishedoutputofmanyimportantSSOs(e.g.3GPPandtheIETF)arepubliclyaccessible.

12

innovation in complementary markets by expanding the addressable market orinstalled base. And while standards can reduce the scope for horizontaldifferentiation in the market for a focal component, increased competition maystimulate the search for extensions and other “vertical” quality enhancements, asemphasized in Bresnahan’s (2002) analysis of divided technical leadership in thepersonalcomputerindustryandthequality‐laddermodelofAcemogluetal(2010).Finally, in horizontally closed platforms, SSOs may encourage innovation byenablingcommitmentstoverticalopenness.13Inparticular,whenaplatformleadercontrolssomebottleneckresource,smallentrantsmayfearhigheraccesspricesorotherpolicychangesthatwouldcaptureashareoftheirinnovationrents.Platformleaders might solve this hold‐up problem by using SSOs to commit to ex postcompetition (see generally Farrell and Gallini (1988)). For instance, XeroxresearchersusedanSSOtogiveawaytheEthernetprotocol,andMicrosofttookthesamestrategywithActiveX(SirbuandHughes1986;VarianandShapiro1998,254).OneimportantwaySSOsaddresspotentialhold‐upproblemsisbyrequiringfirmsto disclose essential patents, and to license them on reasonable and non‐discriminatory (RAND) terms. Thesepolicies seek to prevent patent holders fromdemanding royalties that reflect coordination problems and the sunk costs ofimplementation, as opposed to the way that well‐informed ex ante negotiationwouldreflectbenefitsoftheirtechnologyoverthenextbestsolution.14Althoughthe“reasonable”royaltyrequirementcanbehard toenforce,andSSOscannotprotectimplementersfromnon‐participatingfirms,theseintellectualpropertypoliciesarenevertheless an important method for platform leaders to commit to verticalopenness.Insummary,StandardSettingOrganizationsareaheterogeneoussetofinstitutionslinked by their use of the consensus process. This process emphasizes technicalperformance, and may select for high‐quality standards, but can also producelengthy delays when participants disagree. SSOs also provide a forum forcollaborative innovation, and a way for platform leaders to commit to verticalopennessinordertopromotemarketentryandcomplementaryinnovation.3.3ImposingaStandardAthirdpathtocoordinationisforsomeonewithsufficientclouttosimplyimposeastandard. This dominant player might be a platform leader, a large customer orcomplementor,oragovernmentagency. Apotentialadvantageofcoordinationbyfiat isspeed. Inparticular,dictatorscanavoidorresolvedeadlocksthatemergeinboth standardswars and SSOs. System‐wide architectural transitionsmay also beeasierwhenadefactoplatformleaderinternalizesthebenefitsofa“bigpush”andis

13FurmanandStern(2006)andMurrayandStern(2007)studytheimpactof“vertical”openaccesspoliciesoninnovationoutsideofnetworkindustries.14Farrelletal(2007)describeanextensivelegalandeconomicliteratureonSSOIPRpolicies.

13

thereforewillingtobearmuchofthecost.However,sincedictatorsarenotalwaysbenevolent or capable of spotting the best technology, ex post competition,innovationincentivesandtechnicalqualitywilloftendependonwhoisincharge.Platformleadersoftendictatestandardsforverticalinter‐operability.Forexample,AT&ThistoricallysettherulesforconnectingtotheUStelephonenetwork,andIBMhaslongdefinedtheinterfacesusedinthemarketforplug‐compatiblemainframes.More recently, Apple has maintained tight control over new applications for itsiPhone/iPadplatform.Inprinciple,platformleadersshouldhaveanincentivetousetheircontroloverkeyinterfaces so as toorganize the supplyof complements efficiently.However, fearsthatincumbentmonopolistswillblockentryorhold‐upcomplementaryinnovatorsoften lead to calls for policy‐makers to intervene in support of vertically openinterfaces.15FarrellandWeiser(2003)summarizeargumentsforandagainstmandatoryverticalopenness,and introducethetermICE(InternalizingComplementaryExternalities)to summarize the laissez faire position that a platform leader has incentives toorganizethesupplyofcomplementsefficiently.WhenICEholds,aplatformleader’schoice of vertical openness is efficient; open interfaces promote entry andcompetition in complementary markets, while closed interfaces might betterencouragecoordinationandsystemicinnovation.However,FarrellandWeisernotemultipleexceptionstotheICEprinciple,makingitdifficulttodiscerntheefficiencyofaplatformleader’sverticalpoliciesinpractice.Forexample,aplatformsponsormay inefficiently limit access if it faces regulated prices in its primarymarket; ifcontrol over complements is a key tool for price discrimination; if it has a largeinstalled base; or if a supply of independent complements would strengthen acompetingplatform.16In addition to platform leaders, large customers or complementers can act as defacto standard setters. For instance, Wal‐Mart played an important role in thestandardizationforRadioFrequencyIdentification(RFID)chipsbycommittingtoaparticular specification. Similarly, movie studios played a significant role inresolvingthestandardswarbetweenBlu‐rayandHD‐DVD.Andinsomecases,the

15Callsformandatoryverticalopennessoftenproducefiercedebates.Forexample,priortothe1956consentdecree,IBMresistedpublishingthetechnicalspecificationsthatwouldallowvendorstooffer“plugcompatible”mainframesandperipherals.Morerecentarethedebatesover“netneutrality”andISPs’freedomtochargepricesbasedonthequality‐of‐serviceprovidedtodifferentwebsitesorInternetapplications.16Weyl(2010)offersanalternativeprice‐theoreticanalysisofamonopolythatcontrolsaccesstobothsidesofatwo‐sidedplatform.Inhismodel,themonopolytariffsexhibittwotypesofdeviationfromperfectlycompetitivepricing:a“classicalmarketpowerdistortion”(whichresemblesaLernermarkuprule)anda“Spence(1975)distortion”wherebythemonopolistinternalizesthenetworkbenefitstothemarginal,asopposedtotheaverage,platformadopter.

14

pivotal “customer” isactuallyausergroup,aswhenCableLabs–a consortiumofbroadcasters–developedtheDOCSISprotocolforcablemodems.Theinterestsoflargecomplementorsanddirectcustomersareoftenatleastlooselyalignedwiththoseofend‐users,totheextentthattheirowncompetitivepositionsarenot threatened.Thusconsumersmaywellbenefit fromchoicesmadebythosepowerfulplayers.However,evenwell‐informedpowerfulplayersmayfinditusefultogather informationwithinanSSObeforemakingadecision.Farrell andSimcoe(2009) model this hybrid process, and find that it often outperforms bothuninformed immediate random choice and an SSO‐based screening process thatlacksadominantthird‐party.Government is a third potential dictator of standards. In some cases, thegovernment exerts influence as a large customer. For example, the US Office ofManagement and Budget Circular A‐119 encourages government agencies to usevoluntary consensus standards. And in response to requests from the EuropeanUnion, Microsoft submitted its Open Office XML file formats to ISO. Morecontroversially,governmentsmayuseregulatoryauthoritytopromoteastandard.For example, the U.S. Federal Communication Commission coordinated a switchfrom analog (NTSC) to digital (ATSC) television broadcasting. Sometimes supportfor standards is even legislated, as in the 2009 stimulus package,which containsincentivesforphysicianstoadoptstandardizedelectronicmedicalrecords(butdoesnottakeapositiononanyspecifictechnology).In general, government involvement canbe relativelyuncontroversialwhen thereare large gains from coordination and little scope for innovation or uncertaintyabout the relative merits of different solutions. For instance, it is useful to havestandards for daylight‐saving time and driving on the right side of the road.Government involvement may also be appropriate where private control of aninterface would lead to extreme market power primarily because of severecoordination problems as opposed to differences in quality. But governmentintervention in highly technical standard‐setting processes can pose problemsincluding lack of expertise, regulatory capture, and lock‐in on the government‐supportedstandard.3.4ConvertersandMulti­homingConverters,adapters, translatorsandmulti‐homingareways toreduce thedegreeorcostofincompatibility.Forexample,computersuseawidevarietyoffileformatsto store audio and video, but most software can read several types of files.Econometriciansusetranslationprograms,suchasStatTransfer,tosharedatawithusersofdifferentstatisticalsoftware.EventheInternet’scorenetworkingprotocolsarguably function as a cross‐platform converter: as long as all machines andnetworks run TCP/IP, it is possible to connect many different platforms andapplicationsoverawidevarietyofphysicalnetworkconfigurations.

15

One benefit of using converters to achieve compatibility is that no single partyincurs the full costs of switching. Rather, everyone can choose their preferredsystem,butcanalsotapintoanotherplatform’ssupplyofcomplements,albeitatacostandperhapswith somedegradation.Since translatorsneednotwork inbothdirections,developmentcostsaretypicallyincurredbythepartywhobenefits,orbya thirdpartywhoexpects toprofitbychargingthosewhobenefit.Andconverterscanavert the longdeadlocksthatmayoccur inastandardswaroranopen‐endednegotiation,sincethereisnoneedtoagreeinadvanceonacommonstandard:eachplatform simply publishes its own interface specifications and lets the other sidebuildaconverter(assumingunanimoussupportfortheconverter‐basedsolution).Sometimesusersorcomplementersmayjoinseveralplatforms;suchmulti‐homingcan resemble a converter solution. For example, most retailers accept severalpayment card systems, so consumers canpick one and for themost part not riskbeing unable to transact. Corts and Lederman (2009) show that video game‐developersincreasinglymulti‐home,andarguethatmulti‐platformcontentexplainsdecliningconcentration in theconsolemarketover time.And insteadof seekingacommon standard for all computer cables, most machines provide a variety ofsocketstoaccommodatedifferentconnectorssuchasUSB,SCSI,HDMIandEthernet.Multi‐homing preserves platform variety andmay align the costs and benefits ofhorizontalcompatibility.However,dedicatedconvertersorcoordinationonasingleplatformbecomemoreefficientasthecostsofplatformadoptionincrease.Of course, multi‐homing or converters cannot eliminate conflicting interests, andcanopennewpossibilities for strategicbehavior. For example, firmsmay seekanadvantagebyprovidingconverterstoaccessarival’scomplementswhileattemptingtoisolatetheirownnetwork.Ataritriedthisstrategybydevelopingaconvertertoallow its users to play games written for the rival Nintendo platform. However,Nintendowasable toblockAtari’seffortsbyasserting intellectualpropertybasedonanencryptionchipembeddedineachnewgame(ShapiroandVarian,1998).Firms may use one‐way converters to create market power on either side of avertical interface. MacKie‐Mason and Netz (2007) suggest that Intel pursued thisstrategyby includinga “hostcontroller” in theUSB2.0specification,andallowingperipheral devices to speak with the host‐controller, but delaying the release ofinformation about the link between the host‐controller and Intel’s chipsets andmotherboards.Converters can also favor a particular platform by degrading, rather than fullyblocking, inter‐operability. Many computer users will be familiar with thefrustrations of document portability, even though most word processors andspreadsheetscontainconvertersthatread,andsometimeswrite,inthefileformatsusedbyrivalsoftware.Finally,convertersmayworkpoorlyfortechnicalreasons.Thismaybeparticularlysalient for vertical interfaces, since allowing designs to proliferate undercuts the

16

benefits associated with modularity and specialization across components. Forexample,mostoperatingsystemsdonotprovideafullyspecifiedinterfaceforthird‐partyhardware (e.g.printersorkeyboards),and the “devicedriver” software thatacts as a translator is widely believed to be the most common cause of systemfailures(Ganapathietal.,2006).In summary, converters are attractive because they preserve flexibility forimplementers.However,inastandardswar,firmsmayworktoblockconverters,asAtari did. Firms may also gain competitive advantage by using converters tomanipulate a vertical interface. And even when there is little conflict, dedicatedcompatibilitystandardsmaydominateconvertersforheavilyusedinterfaces,whereperformanceandscalabilityareimportant.4.ChoosingapathWhat determines which path to compatibility is followed, or attempted, in aparticular case? Whenwill that choice be efficient?While data on the origins ofcompatibilitystandardsarescant,thissectionofferssomeremarksontheselectionprocess.17Choosingapath to compatibility can itselfbea coordinationproblem, creatinganelement of circularity in analysis of this choice. We try to sidestep this logicaldilemmabygroupingplatformsintotwocategories:thosewithadominantplatformleader,andsharedplatformsthatdefault toeithercollectivegovernance(SSOs)orsplintering and standards wars. Eisenmann (2008) suggests that this distinctionbetween shared and proprietary platforms emerges early in the technology lifecycle, basedon firms’ strategicdecisionsabouthorizontal openness. Inparticular,he predicts that platform leaders will predominate in “winner‐take‐all” marketswherenetworkeffectsarestrong(relativetothedemandforvariety),multi‐homingiscostly,andthefixedcostsofcreatinganewplatformaresubstantial.This life‐cycleperspectiveofplatformgovernance isconsistentwiththe intriguing(though unsystematic) observation that many technologies settle on a particularpath to compatibility, even a specific agency, and adhere to it over time. Forexample, the ITU has managed international inter‐operability oftelecommunications networks since 1865, and JEDEC has been the dominant SSOfor creating open standards for semiconductor inter‐operability (particularly inmemory chips) since 1968. Likewise, for products such as operating systems andvideogameconsoles,proprietaryplatformleadershiphasbeenthedominantmodeofcoordinationacrossseveralgenerationsoftechnology.

17OneexceptiontothepaucityofdataisapaperbyBiddleetal(2010)thatidentifies250compatibilitystandardsusedinatypicallaptop.TheauthorsworkedwithInteltoestimatethat20percentofthesestandardscomefromindividualcompanies,44percentfromconsortiaand36percentfromaccreditedSDOs.

17

Nevertheless,thereareseveralwell‐knowncasesofdominantfirmslosingdefactocontrol over a platform. The most famous example is IBM and the personalcomputer architecture. Other examples include the demise of micro‐computingincumbents like Digital Equipment; Google replacing AltaVista as the dominantsearchengine;Microsoft’swell‐documentedstrugglestoadapttotheInternet;andthe ongoing displacement of the Symbian cellular phone operating system byalternatives from Research in Motion (Blackberry), Apple (iPhone) and Google(Android).Bresnahan(2001)suggeststhatakeyconditionforsuch“epochal”shiftsinplatformleadershipisdisruptivetechnicalchangeatadjacentlayersofthelargersystem, since it is hard to displace a platform leader through direct horizontalcompetition when network effects are strong. While this observation certainlyaccordswiththefactsofwell‐knowncases,itisnotveryamenabletoformaltestinggiventheinfrequentnatureofsuchmajorshifts.4.1SelectionandefficiencyWhen there is a clear platform leader, the ICE principle suggests that leaderwillhave an incentive to choose an efficient coordination process regarding verticalcompatibility. For example, the platform leader might delegate standard‐settingactivities toanSSOwhen fearsofhold‐up impedecomplementary innovation,butimposeastandardwhenSSOnegotiationsdeadlock.Unfortunately,theICEprincipleissubjecttomanycaveats,bringingbackquestionsaboutwhetheraplatform leader choosesaparticularpath for itsefficiencyor forother reasons such as its impact on ex post competition. For instance, Gawer andHenderson (2007) use Intel’s decision to disseminate USB as an example of ICE,whileMacKie‐MasonandNetz(2007)arguethatIntelmanipulatedUSB2.0togainacompetitiveadvantage.Whereonestudyemphasizestheinitialdecisiontogiveupcontrolovera technology, theotheremphasizes theuseofone‐wayconverters toexclude competitors and gain lead‐time advantages in complementary markets.These competingUSBnarrativeshighlight thedifficultyofdeterminingaplatformleader’smotives.Without a platform leader, it is less clear why the private costs and benefits ofchoosing an efficient path to compatibility would be aligned. If firms are ex antesymmetric,andcommittoapathbeforelearningthemeritsofcompetingsolutions,they would have an ex ante incentive to choose the efficient mechanism. Butstandard setting is typically voluntary, and firms do not commit to abide byconsensus decisions. Thus, when asymmetries are present, or emerge over time,firms may deviate to a path that favors their individual interests. While thesedeviations fromcollectivegovernancemay leadto the“forking”ofstandards, theydonotnecessarilyblocktheSSOpath,andinsomecasestheremainingparticipantscanstillachieveconverter‐basedcompatibility.Some observers suggest that this chaotic situation can deliver the virtues of bothdecentralized adoption and collective choice. For example, Greenstein (2009)

18

argues that a proliferation of SSOs combined with widespread independenttechnical experimentation is a sign of “healthy standards competition” on thecommercial Internet. This optimistic view emphasizes the virtues of “standardsswarms.”Whennetworkeffectsareweak(astheabsenceofaplatformleadermightsometimes suggest), and substantial market or technological uncertainty exists,decentralized choice can identify promising standards for a particular niche,withSSOs emerging to facilitate coordination as needed. Unfortunately, there is noguarantee thatmixingdecentralizedadoptionwithSSOscaptures thebenefitsandavoidsthecostsofeitherpath in isolation. Inparticular,eitherpathmay leadtoastalemate,andwhendecentralizedadoptionistheoutsideoptionthereisalwaysadangerofstrandedinvestmentsorselectingthewrongsystem.A second optimistic argument holds that newways to govern shared technologyplatforms will arise in response to market pressures and technologicalopportunities.Forexample,Cargill(2001)andMurphyandYates(2009)claimthataccreditedSDOs lostmarketshare tosmall consortiaduring the1980sand1990sbecausetheSDOs’ponderousdecision‐makingprocedureswereill‐matchedtorapidICT product lifecycles (see also Besen and Farrell 1991). Smaller and less formalorganizations might work faster by relaxing the definition of consensus, takingadvantageofnewtechnologiesforcollaboration,andallowingcompetingfactionstoworkinisolationfromoneanother.Berners‐Lee(1999)citesdelaysattheIETFasaprimarymotiveforcreatingtheWorldWideWebConsortium(W3C),andFigure2showsthatconsortiaareindeedontherise.18

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐INSERTFIGURE2ABOUTHERE

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐Whilethisevolutionaryhypothesisisintriguing,itisnotobviousthatorganizationalexperimentation and competition will evolve an efficient path to compatibility.Simcoe (2010) shows that consortia still experience coordination delays whenparticipants have conflicting interests over commercially significant technology.And the proliferation of SSOs also increases the potential for forum shopping, asemphasized by Lerner and Tirole. We view competition between SSOs as apromising topic for further research.19Atpresent, it remainsunclearwhether thecurrentproliferationoforganizationalmodelsforSSOsistheoutcomeof,orpartof,an evolutionary process, or simply confusion regarding how best to organize acomplexmulti‐lateralnegotiation.18Observingtheslowdownatmanyconsortia,Cargillsuggeststhattheytoowillbesupplanted,perhapsbytheopen‐sourcesoftwaredevelopmentmodel,orotherbottom‐upeffortstoestablishdefactostandards.19Aparallelliteratureonvoluntarycertificationprograms,reviewedinDranoveandJin(2010),mayofferinsightsoncompetitionbetweenSSOsthatcanalsobeappliedtocompatibilitystandards.Forinstance,theyciteseveralrecentstudiesthatexaminetheproliferationofcompeting“eco‐labeling”initiatives(e.g.EnergyStarversusLEEDforconstructionorSustainableForestInitiativeversusForestStewardshipCouncilforlumber).

19

4.2HybridpathsWhile markets, committees, converters, and dictators offer distinct paths tocompatibility,theycansometimesbecombined.Forexample,standardswarsmayberesolvedthroughnegotiationsatanSSO,ortheinterventionofadominantfirm;andslowSSOnegotiationsmaybeacceleratedbyanagreementtouseconverters,orbyevidencethatthemarketistippingtowardsaparticularsolution.Farrell and Saloner (1988) model a hybrid coordination process that combinesmarkets and committees. In theirmodel, the hybrid path combines the virtues ofstandardswarsandSSOswithoutrealizingallofthecosts.Inparticular,themarketworks faster than an SSO, while the committee reduces the chance of inefficientsplintering. The IETF’s informal motto of “rough consensus and running code”reflectsasimilar logic.Byemphasizing“runningcode,” the IETFsignals that firmsshouldnotwait forevery issue tobe resolvedwithina committee, and that somelevelofexperimentation isdesirable.However, it remains important toachieveatleast“roughconsensus”beforeimplementation.SynergiesofhybridstylecanalsooccurbetweenSSOs.Independentfirmsandsmallconsortia often work to pre‐establish a standard, before submitting it to anaccreditedSDOsforcertification.Forexample,SunMicrosystemsusedISO’sPubliclyAccessibleSpecification(PAS)processtocertifytheJavaprogramminglanguageandODF document format (Cargill 1997). Similarly, Microsoft used ISO’s fast‐trackprocedures to standardize its Open Office XML document formats. As describedabove,platformleadersmayvalueSDOcertificationifitprovidesacrediblesignalofvertical openness that attracts complementary innovators. However, critics claimthat fast‐track procedures can undermine the “due process” and “balance ofinterest” requirements which distinguish SDOs from consortia, leading users orcomplementerstoadoptproprietarytechnologyoutofafalsesenseofsecurity.Asecondhybridpathtocompatibilityoccurswhenparticipantsinastandardswaruseconverterstofashionanescaperoute.Forexample,the56Kmodemstandardswar was resolved by adopting a formal standard that contained elements ofcompeting systems. Converters can also reduce the scope of conflicting interestswithinanSSO,especiallywhenparticipantsadopta“framework”thatfallsshortoffullcompatibility.An alternative escape route (and third hybrid path) relies on a dictator to breakdeadlockswithinanSSO.FarrellandSimcoe(2009)analyzeamodelofconsensusstandardsettingasawarofattrition,inwhichapoorlyinformedbutneutralthirdparty can break deadlocks by imposing a standard. They find that this hybridprocesswilloften(butnotalways)outperformanuninterruptedscreeningprocess,or an immediate uninformed choice. In practice, there are many examples of adominantplayerinterveningtoaccelerateaconsensusprocess,suchasthecaseofDOCSIS(cablemodems)orelectronichealthrecords,bothmentionedabove.

20

Thus,informally,therearesomereasonstohopethatastandardssystemwithmanypathstocompatibilitywillperformwell.Platformleadersoftenhaveanincentivetochoosetheefficientpath,andagreatervarietyof“pure”pathsmeansmoreoptionstochoosefrom.SSOsmayevolveinresponsetomarketpressuresandtechnologicalopportunities.Andboththeoryandpracticalobservationsuggestthatmanypathstocompatibilitycanbecombinedincomplementaryways.Atthispointinourunderstanding,however,anyoptimismshouldbeverycautious.Various exceptions to the ICE principle show that platform leaders may weighefficiency againstex post competitionwhen choosing a path to compatibility. It isnotclearwhencompetitionamongSSOswill lead tomoreefficient institutions,asopposedtoincreasedforumshoppingandtechnologysplintering.Andwhilehybridpaths can work well, they highlight the complex welfare trade‐offs among theprobabilityofcoordination,thecostsofnegotiation,andtheimplicationsforexpostcompetitionandinnovation.5.ConclusionsCompatibility standards can emerge through market competition, negotiatedconsensus, converters or the actions of a dominant firm. These four paths tocompatibility have different costs and benefits which depend on whether aparticularstandardpromotesverticalorhorizontal inter‐operability; thepresenceof an installed base or proprietary complements; firms’ sunk investments inalternativedesigns;andthedistributionofintellectualpropertyrights.Whenchoosingapathtocompatibility,therearetrade‐offsbetweentheprobabilityof coordination, expected costs in time and resources, and the implications forexpost competitionand innovation.There isanargument thataplatform leaderwillinternalize these costs and benefits and choose the socially efficient path tocompatibility. But that argument has many exceptions. Others argue thatdecentralizedexperimentationwithdifferenttechnologies,looselycoordinatedbyacombinationofmarketsandSSOs,willtypicallyproducegoodoutcomes.However,itis hard to predict how far competition among SSOs leads them toward optimalpolicies,orhowreliablystandardswarsselectthesuperiorplatform.Amid these complex questions, there is certainly scope for beneficial governmentinvolvement, whether as a large end‐user, a regulator, or a third‐party withtechnical expertise. But direct government intervention in highly technicalstandard‐setting processes can pose problems including lack of expertise,regulatorycapture,andlock‐inongovernment‐supportedstandards.Viewing the economic literature on compatibility standards in terms of our fourbroadpaths also suggests several researchopportunities. First, there is very littledata on the relative market share of these alternative paths. Thus, it is unclearwhethereconomistshavefocusedonthemostimportantormostcommonmodesof

21

organizing the search for compatibility, or merely the routes they find mostinteresting. Our impression is that standards wars and platform leaders havereceivedmoreacademicattentionthanhaveSSOsandconverters. Possibly this isbecausetheformerpathsarerepletewithopportunitiesforinterestinglystrategic,yetfamiliarlymarket‐based,competitivestrategies,whilethelatteroptionsleadtolesstractableormoreforeignquestionsofsocialchoiceandbargaining.Asecondtopicforresearchistheselectionofapathtocompatibility,particularlyintheearlystagesofatechnology lifecycle.Manystudiesassumeeitherastandardswaror aplatform leader (whomightdelegate the choiceof standards forverticalcompatibility to an SSO). But we know little about how the rules for collectivegovernanceofasharedplatformemergeorevolveovertime.Andthereisnotmuchresearch on forum shopping by technology sponsors, or the nature and effects ofcompetition among SSOs. Developing a better understanding of how a particularpath is chosen represents a crucial first step towards quantifying the cost‐benefittradeoffsacrosspaths(unlesstheassignmentisrandom),andadjudicatingdebatesovertheefficiencyoftheselectionprocess.Finally, there is an opportunity to examine interactions among the four paths tocompatibility.Despitesomefirststepstowardsmodeling“hybrid”paths,thereisnogeneraltheoryandverylittleempiricalevidenceonwhochoosesthemechanism(s)andhow,oronwhether the fourpaths tend to complementor interferewithoneanother.

22

REFERENCESAcemoglu, D., G. Gancia, and F. Zilibotti (2010). “Competing Engines: of Growth: Innovation and

Standardization.”NBERWorkingPaper15958.Arthur, W. Brian. 1989. “Competing Technologies, Increasing Returns, and Lock‐In by Historical

Events.”EconomicJournal,97:642‐65.Augereau, A., S. Greenstein and M. Rysman (2006). “Coordination versus Differentiation in a

StandardsWar:56Kmodems.”RandJournalofEconomics,34(4):889‐911.Baldwin, C.Y. and Clark, K.B. (2000). Design Rules. Vol. 1: The Power of Modularity. Cambridge,

Massachusetts:MITPress.Bekkers,R.,G.Duysters,etal.(2002)."IntellectualPropertyRights,StrategicTechnologyAgreements

andMarketStructure:ThecaseofGSM."ResearchPolicy,31:1141‐1161.Berners‐Lee,T.andM.Fischetti(1999).WeavingtheWeb:Theoriginaldesignandultimatedestiny

oftheWorldWideWebbyitsinventor.SanFrancisco:HarperSanFrancisco.Besen,S.M.andJ.Farrell(1991).“TheRoleoftheITUinStandardization:Pre‐Eminence,Impotence,

orRubberStamp?”TelecommunicationsPolicy,15(4):311‐321.Besen, S. M. and J. Farrell (1994). Choosing how to compete ‐ strategies and tactics in

standardization.JournalofEconomicPerspectives,8(2):117–131.Besen,S.M.andG.Saloner(1989).TheEconomicsofTelecommunicationsStandards.Changingthe

Rules: Technological Change, International Competition, and Regulation inTelecommunications.R.CrandallandK.Flamm.Washington,Brookings:177‐220

Biddle,B.,A.WhiteandS.Woods(2010).“HowManyStandardsinaLaptop?(AndOtherEmpirical

Questions)”AvailableatSSRN:http://ssrn.com/abstract=1619440.Boudreau, K. (2010). “Open Platform Strategies and Innovation: GrantingAccess versusDevolving

Control.“ManagementScience,56(10):1849‐1872.Bresnahan, T. (2002). “The Economics of the Microsoft Case.” Stanford Law and Economics Olin

WorkingPaperNo.232.(AvailableatSSRN:http://ssrn.com/abstract=304701.)Bresnahan, T. and S. Greenstein (1999). “Technological competition and the structure of the

computerindustry.”JournalofIndustrialEconomics,47(1):1–40.Bresnahan,T.andP‐LYin(2007).“StandardSettinginMarkets:TheBrowserWar,”inStandardsand

PublicPolicy,CambridgeUniversityPress2007,ShaneGreensteinandVictorStango,editors.Cabral,L. andT.Kretschmer (2007) “Standardsbattlesandpublicpolicy.” inStandardsandPublic

Policy,CambridgeUniversityPress2007,ShaneGreensteinandVictorStango,editors.Cabral, L. andD. Salant (2008). “Evolving technologies and standards regulation.”Working Paper.

(AvailableatSSRN:http://ssrn.com/abstract=1120862).Cargill,C.(2002).“IntellectualpropertyrightsandStandardsSettingOrganizations:AnOverviewof

FailedEvolution.”(Availableatwww.ftc.gov/opp/intellect/020418cargill.pdf)

23

Cargill, C. (2001). Evolutionary pressures in standardization: Considerations on ANSI’s nationalstandardsstrategy.TestimonyBeforetheU.S.HouseofRepresentativesScienceCommittee.(http://www.sun.com/software/standards/Testimony‐Final‐reformatted.pdf).

Cargill, C. F. (1989). InformationTechnology Standardization :Theory, Process, andOrganizations.

Bedford,MA,DigitalPress.Cargill,C.F. (1997).OpenSystemsAtandardization :ABusinessApproach.UpperSaddleRiver,NJ,

PrenticeHallPTR.Chiao,B.,J.Lerner,andTirole,J.(2007).“TheRulesofStandardSettingOrganizations:AnEmpirical

Analysis.”RANDJournalofEconomics.38(4):905‐930.Corts,K.andM.Lederman(2009).“Softwareexclusivityandthescopeofindirectnetworkeffectsin

theU.S. home video gamemarket” International Journal of Industrial Organization, 27(2):121‐136.

David,P.A.andS.Greenstein(1990)."TheEconomicsofCompatibilityStandards:AnIntroductionto

RecentResearch."EconomicsofInnovationandNewTechnology,1(1):3‐42.David,P.(1988).“ClioandtheEconomicsofQWERTY.”AmericanEconomicReview,75:332‐337.DeLacey, B., K. Herman, D. Kiron and J. Lerner (2006) “Strategic Behavior in Standard‐Setting

Organizations.”HarvardNOMWorkingPaperNo.903214.Dranove,D.andN.Gandal(2003).“TheDVDvs.DIVXStandardWar:EmpiricalEvidenceofNetwork

EffectsandPreannouncementEffects,”TheJournalofEconomicsandManagementStrategy,12(3):363‐386.

Dranove,D.andG.Jin(2010).“QualityDisclosureandCertification:TheoryandPractice.”Journalof

EconomicLiterature,48(4):935–63.Eisenmann, T. and L. Barley (2006). Atheros Communications.HarvardBusiness School Case 806‐

093.Eisenmann,T.(2008).“ManagingProprietaryandSharedPlatforms.”CaliforniaManagementReview,

50(4).Eisenmann,ThomasR.,GeoffreyParkerandMarshallVanAlstyne."OpeningPlatforms:When,How

and Why?" Chap. 6 in Platforms, Markets and Innovation, edited by Annabelle Gawer.Cheltenham,UKandNorthampton,Mass.:EdwardElgarPublishing,2009,paperbacked.

European Commission (2010). “Guidelines on the applicability of Article 101 of the Treaty on the

Functioning of the European Union to horizontal co‐operation agreements.” Available athttp://ec.europa.eu/competition/consultations/2010_horizontals/guidelines_en.pdf

Farrell,J.andN.Gallini(1988).“Second‐SourcingasaCommitment:MonopolyIncentivestoAttract

Competition.”TheQuarterlyJournalofEconomics,103(4):673‐94.Farrell,J.andP.Weiser(2003).“Modularity,VerticalIntegration,andOpenAccessPolicies:Towards

aConvergenceofAntitrustandRegulationintheInternetAge.”HarvardJournalofLawandTechnology,(17)1:85‐135.

Farrell, J., J. Hayes, C. Shapiro, and T. Sullivan (2007). “Standard setting, patents and hold‐up.”

24

AntitrustLawJournal,74:603–670.Farrell,J.andP.Klemperer(2007).“Coordinationandlock‐in:Competitionwithswitchingcostsand

network effects.” In M. Armstrong and R. H. Porter (Eds.), Handbook of IndustrialOrganization,Volume3,pp.1967–2056.Elsevier.

Farrell,J.andT.Simcoe(2009).“ChoosingtheRulesforConsensusStandardization.”WorkingPaper.

Availableathttp://people.bu.edu/tsimcoeFarrell, J. andG. Saloner (1988). "Coordination throughCommitteesandMarkets."Rand Journalof

Economics,19(2):235‐252.Farrell, J.andC.Shapiro(1992)."StandardSettinginHigh‐DefinitionTelevision."BrookingsPapers

onEconomicActivity:1‐93.Farrell, J. (2007) “Should competition policy favor compatibility?” in Shane Greenstein and Victor

Stango,eds.,StandardsandPublicPolicy,Cambridge:CambridgeUniversityPress,2007,pp.372‐388.

Furman,J.andS.Stern(2006).“Climbingatoptheshouldersofgiants:Theimpactofinstitutionsoncumulativeresearch.”NBERWorkingPaperNo.12523.

Gawer A. and R. Henderson (2007). "Platform Owner Entry and Innovation in Complementary

Markets:EvidencefromIntel,"JournalofEconomics&ManagementStrategy,16(1):1‐34.Gallagher, S. and J.West (2009). “Reconceptualizing and expanding the positive feedback network

effects model: A case study,” Journal of Engineering and TechnologyManagement, 26(3):131‐147.

Ganapathi, A., V. Ganapathi and D. Patterson (2006). “Windows XP Kernel Crash Analysis.”

Proceedingsofthe20thLargeInstallationSystemAdministrationConference,pp.149‐159.Greenstein, S. (2009), “Glimmers and Signs of Innovative Health in the Commercial Internet,”

ColoradoHighTechnologyLawJournal.forthcoming.Henderson, R. and K. Clark (1990). “Architectural Innovation: The Reconfiguration of Existing

Product Technologies and the Failure of Established Firms.” Administrative ScienceQuarterly,35(1):9‐30.

Henderson,R.(2003).EmberCorp.:DevelopingtheNextUbiquitousNetworkingStandard.Harvard

BusinessSchoolCaseNumber9‐703‐448.Intellectual Property Owners Association (2009). Standards Primer: An Overview of Standards

Setting Bodies and Patent‐Related Issues that Arise in the Context of Standards SettingActivities. Section 16. (standardslaw.org/seminar/class‐2/excerpts‐from‐ipo‐standards‐primer/)

Katz,M.L.andC.Shapiro(1985).“Networkexternalities,competitionandcompatibility.”American

EconomicReview,75:424–440.Liebowitz,S.J.andS.Margolis(1990)“TheFableoftheKeys,”JournalofLawandEconomics,33(1):

1‐25.

25

Lemley,M.(2002)."IntellectualPropertyRightsandStandardSettingOrganizations."CaliforniaLawReview,90:1889‐1981.

Lerner, J. and J. Tirole (2006). “A model of forum shopping.” American Economic Review, 96(4):

1091–1113Lerner, J., J. Tirole and M. Strojwas (2003). "Cooperative Marketing Agreements Between

Competitors:EvidencefromPatentPools,"NBERWorkingPapers9680.Majoras,D. (2005). “Recognizing thePro‐competitivePotentialofRoyaltyDiscussions inStandard‐

Setting.”StanfordUniversityStandardizationandtheLawConference.September23,2005.http://www.ftc.gov/speeches/majoras/050923stanford.pdf.

Matutes, C. and Regibeau, P. (1988). ""Mix and Match": Product Compatibility without Network

Externalities,"RANDJournalofEconomics,19(2):221‐234.Mackie‐Mason, J. and J. Netz (2007) “Manipulating interface standards as an anticompetitive

strategy,” in Standards and Public Policy, Cambridge University Press 2007, ShaneGreensteinandVictorStango,editors.

Murray, F. and S. Stern (2007). “Do formal intellectual property rights hinder the free flow of

scientific knowledge? An empirical test of the anti‐commons hypothesis.” Journal ofEconomicBehaviorandOrganization,63(4):648–687.

MurphyC.andJ.Yates(2009).TheInternationalOrganizationforStandardisation(ISO).Routledge.Parker,G.andM.VanAlstyne(2005).“Two‐SidedNetworkEffects:ATheoryofInformationProduct

Design.”ManagementScience,51(10):1494‐1504.Rochet, J‐C. and J. Tirole. (2003) “Platform Competition in Two‐sided Markets” Journal of the

EuropeanEconomicAssociation,1(4):990‐1029.Russell,A.L.(2006).“RoughConsensusandRunningCodeandtheInternet‐OSIStandardsWar.”IEEE

AnnalsoftheHistoryofComputing,July‐September,48–61. Rysman,M.(2009)."TheEconomicsofTwo‐SidedMarkets."JournalofEconomicPerspectives,23(3):

125–43.Rysman, M. and T. Simcoe (2008). “Patents and the performance of voluntary standard setting

organizations.”ManagementScience,54(11):1920–1934.Shapiro,C.(2001).Navigatingthepatentthicket:Crosslicenses,patentpools,andstandardsetting.

InA. Jaffe, J.Lerner,andS.Stern(Eds.), InnovationPolicyandtheEconomy,Volume1,pp.119–150.MITPress.

Shapiro,C.andH.R.Varian(1998).InformationRules:AStrategicGuidetotheNetworkEconomy.

Boston,Mass.,HarvardBusinessSchoolPress.Simcoe, T. (2010). “Standard Setting Committees: Consensus Governance for Shared Technology

Platforms.”AmericanEconomicReview,forthcoming.(http://ssrn.com/abstract=899595).Simcoe, T., S. J. Graham and M. Feldman. (2009) “Competing on Standards? Entrepreneurship,

Intellectual Property and Platform Technologies.” Journal of Economics and ManagementStrategy,18(3):775‐816.

26

Simcoe,T (2007). “Explaining the Increase in Intellectual PropertyDisclosure.” in StandardsEdge:

TheGoldenMean.BolinGroup.Spence,M.(1975).“Monopoly,Quality,andRegulation.”BellJournalofEconomics,6(2):417–429.Varney,C.A.(2010)“PromotingInnovationThroughPatentandAntitrustLawandPolicy,”Remarks

prepared for joinUSPTOFTCWorkshopon the IntersectionofPatentPolicyandAntitrustPolicy.May26,2010.Availableathttp://www.justice.gov/atr/public/speeches/260101.htm

Weiss, M. and R. Toyofuku (1996). “Free‐ridership in the standards‐setting process: the case of

10BaseT.”StandardView4(4),205–212.Weiss, M. B. and M. Sirbu (1990). “Technological choice in voluntary standards committees: An

empiricalanalysis.”EconomicsofInnovationandNewTechnology,1(1):111–134.West,J.(2007)“TheEconomicRealitiesofOpenStandards:Black,WhiteandManyShadesofGray,”

in Shane Greenstein and Victor Stango, eds., Standards and Public Policy, Cambridge:CambridgeUniversityPress,2007,pp.87‐122.

WeylG.(2010)"APriceTheoryofMulti‐SidedPlatforms,"AmericanEconomicReview,100(4):1642‐1672.

27

TABLE1:MAJORANSIACCREDITEDSSOs

Acronym Standards ICT Full Name

INCITS 10,503 Y International Committee for Information Technology Standards

ASTM 8,339 N American Society for Testing and Materials IEEE 7,873 Y Institute of Electrical and Electronics Engineers UL 7,469 N Underwriters Laboratories ASME 7,026 N American Society of Mechanical Engineers ANSI/TIA 4,760 Y Telecommunications Industry Association ANSI/T1 3,876 Y ANSI Telecommunications Subcommittee

ANSI/ASHRAE 3,070 N American Society of Heating, Refrigerating and Air-Conditioning Engineers

AWS 2,517 N American Welding Society ANSI/NFPA 2,365 N National Fire Protection Association ANSI/EIA 2,011 Y Electronic Industries Association ANSI/SCTE 1,803 Y Society of Cable Telecommunications Engineers ANSI/AWWA 1,759 N American Water Works Association ANSI/AAMI 1,621 Y American Association of Medical Imaging ANSI/NSF 1,612 N National Sanitation Foundation ANSI/ANS 1,225 N American Nuclear Society ANSI/API 1,225 N American Petroleum Institute ANSI/X9 940 N Financial Industry Standards ANSI/IPC 891 Y Association Connecting Electronics Industries ANSI/ISA 872 Y International Society of Automation Total ICT 30,786 43% Notes: List of largest ANSI accredited Standards Developing Organizations based on a count ofdocumentslistedinthe2006ANSIcatalogofAmericanNationalStandards.The“Standards”columnshowstheactualdocumentcount.The“ICT”columnindicatestheauthors’ judgmentastowhetherthatorganization’sprimaryfocusiscreatingcompatibilitystandards.

28

FIGURE1:COMPATIBILITYCHOICEASACOORDINATIONGAME

FIGURE2:THEGROWTHOFCONSORTIA

Notes:Figureshowsthecumulativenumberofnewconsortiafoundedduringeachfive‐yearperiod,based on the authors’ analysis of the list of ICT consortiamaintained by AndrewUpdegrove, andpublishedatwww.consortiuminfo.org.