the choice between formal and informal intellectual property: a literature review · 2013-05-30 ·...

NBER WORKING PAPER SERIES

THE CHOICE BETWEEN FORMAL AND INFORMAL INTELLECTUAL PROPERTY:A LITERATURE REVIEW

Bronwyn H. HallChristian Helmers

Mark RogersVania Sena

Working Paper 17983http://www.nber.org/papers/w17983

NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue

Cambridge, MA 02138April 2012

This work contains statistical data from UK ONS which is Crown copyright and reproduced with thepermission of the controller of HMSO and Queen’s Printer for Scotland. The use of the ONS statisticaldata in this work does not imply the endorsement of the ONS in relation to the interpretation or analysisof the statistical data. This work uses research datasets which may not exactly reproduce NationalStatistics aggregates. We gratefully acknowledge funding from the UK Intellectual Property Office.We thank Beth Webster, Ivan Png, and John Walsh for valuable comments. The views expressed hereinare those of the authors and do not necessarily reflect the views of the National Bureau of EconomicResearch.

At least one co-author has disclosed a financial relationship of potential relevance for this research.Further information is available online at http://www.nber.org/papers/w17983.ack

NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies officialNBER publications.

© 2012 by Bronwyn H. Hall, Christian Helmers, Mark Rogers, and Vania Sena. All rights reserved.Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission providedthat full credit, including © notice, is given to the source.

The Choice between Formal and Informal Intellectual Property: A Literature ReviewBronwyn H. Hall, Christian Helmers, Mark Rogers, and Vania SenaNBER Working Paper No. 17983April 2012, Revised May 2013JEL No. K11,L29,O34

ABSTRACT

We survey the economic literature, both theoretical and empirical, on the choice of intellectual propertyprotection by firms. Our focus is on the tradeoffs between using patents and disclosing versus the useof secrecy, although we also look briefly at the use of other means of formal intellectual property protection.

Bronwyn H. HallDept. of Economics549 Evans HallUC BerkeleyBerkeley, CA 94720-3880and [email protected]

Christian HelmersUniversidad Carlos III de Madridc/. Madrid, 12628903 Getafe, [email protected]

Mark RogersHarris Manchester CollegeOxford University

Vania SenaSouthend Campus Elmer ApproachSouthend-on-Sea SS1 [email protected]

2

TableofContents1. Introduction......................................................................................................................................................3 2. SurveyevidenceonthechoiceofIPprotectionmethods.............................................................8

2.1. (National) innovation surveys ......................................................................................................... 9 2.2. Cross-country comparisons ........................................................................................................... 13

3.Theory 17 3.1. Patent and secrecy as substitutes ................................................................................................... 18 3.1.1. Disclosure......................................................................................................................................20

3.1.2. Competitionforinnovation....................................................................................................25

3.1.3. Leadtimeandcomplexity.......................................................................................................27

3.1.4. Cumulativeorsequentialinnovation.................................................................................29

3.2. Combined patent-secrecy strategy ................................................................................................ 31 3.3. No patent, no secrecy -- disclosure .............................................................................................. 32 3.4. The life-cycle of firms and industry .............................................................................................. 34

4. EmpiricalEvidence.....................................................................................................................................35 4.1. Registered IP and secrecy as substitutes ...................................................................................... 36 4.1.1. Survey(CIS)–basedanalysis................................................................................................37

4.1.2. Invention‐levelevidence.........................................................................................................39

4.1.3. Evidencefromlegalchanges.................................................................................................41

4.1.4. Evidencefromlitigation..........................................................................................................42

4.2. Combined patent-secrecy strategy ................................................................................................ 43 4.3. No patent, no secrecy – disclosure ............................................................................................... 44 4.4. The Choice of other forms of IP .................................................................................................. 45 4.4.1. Trademarks...................................................................................................................................45

4.4.2. Copyright........................................................................................................................................47

4.4.3. MultipleandoverlappingIPuse..........................................................................................48

5. Patentsvssecrecy:welfare.....................................................................................................................49 6. Conclusions....................................................................................................................................................51 Bibliography............................................................................................................................................................54

3

TheChoicebetweenFormalandInformalIntellectualPropertyProtection:aReview

BronwynHalla,ChristianHelmersb,MarkRogersc,VaniaSenad

RevisedMay2013

Trade secret lawprovides farweakerprotection inmany respects than thepatent law. [...]The possibility that an inventor who believes his invention meets the standards ofpatentabilitywillsitback,relyontradesecretlaw,andafteroneyearofuseforfeitanyrighttopatentprotection[…]isremoteindeed.

USSupremeCourt(KewaneeOilCo.v.BicronCorp.,416U.S.470,1974)

Judges and lawyers have sometimes thought that because trade secret law provides lessprotection to the inventor than patent law does, no rational person with a patentableinventionwouldfailtoseekapatent.[…]Thisreasoningisincorrect.

Friedmanetal.(1991:62‐63)

1. Introduction

US$500,000perpatent: this figurecirculatedwidely followingGoogle’sannouncementofitstakeoverofMotorola(Economist17August2011).Thisvalueisobtainedbydividingtheprice paid by Google for the acquisition ofMotorola (US$12.5 billion) by the number ofpatentsheldbyMotorola(24,000).Applyingasimilarlogic,theacquisitionofthepatentsassigned toCanadianNortelbya consortiumof firms in July2011yieldsanevenhigherprice tag of US$750,000 per patent.1 These figures may seem extreme. However, datacoveringabroaderrangeofpatenttransactionssuggeststhattheaveragepriceoftradedUSPTO patents in 2012 was around US$370,000.2 Comparing the trading value ofcommercial secrets is more difficult, since the trading of secrets is by definition much

aUniversityofCaliforniaatBerkeley;UniversityofMaastricht;correspondingauthor:bhhall@econ.berkeley.edubUniversidadCarlosIIIdeMadridcUniversityofOxford(deceased)dUniversityofEssex1TheconsortiumcomprisedMicrosoft,Apple,Ericsson,EMC,Sony,andResearchinMotion.2IPOfferingsLLCPatentValueQuotient2012.

4

hardertoobserveempirically,butsomefigurescanstillbeobtainedfromcourtrulings.Forexample, in a recent ruling by a federal court in Virginia in September 2011, KolonIndustriesInc.washeldliabletopayingDuPontCo.theamountofUS$919.9millionforthetheftof149tradesecretsrelatedtotheproductionofKevlar,aspecialfiber(Bloomberg15September2011).ThissuggestsanaveragevalueofUS$6.3millionpertradesecret.

Thesecomputationsarecertainlynaïve,butthefiguresneverthelessindicatethatfirmsusebothpatentsandtradesecretstoprotectvaluableinventionsandcontrarytoacommonlyencountered belief, patents may not necessarily protect a company’s most valuableinventions.Suchinventionsmaybekeptsecretdespitebeingpatentable.

Theeconomicsliteraturehasfocusedoverwhelminglyontheuseofpatentsbycompaniesas a means to appropriating returns to innovation.3 Recent events related to theenforcement of patents on digital data transmission technologies in courts around theglobehavereinforcedtheimpressionthatinnovation–inparticularhigh‐techinnovation‐‐isinextricablylinkedwithpatents.Theabove‐citedcourtcaseonthemisappropriationofKevlar trade secrets, however, shows that firms choose secrecy over patenting even forhigh‐techproducts that arepatentable. In fact, theavailable empirical evidence,which isreviewed inmoredetailbelow, strongly suggests thatonlya small fractionof innovativecompanies relies on patents to protect their inventions. Hall et al. (2013), for example,showthatevenamongUKcompaniesthatconductsomeformofR&Dandreporttohavehadaninnovation,onlyaround4percentapplyforapatent.Thisismuchlowerthanwhatwouldbeexpectedifcompaniesprotectedinnovationsthroughpatentsbydefault.Infact,the available survey‐based evidence indicates that companies report heavier reliance onalternativemechanisms,suchasleadtimeandsecrecy(e.g.,Levinetal.1987;Cohenetal.,2000; Arundel 2001). The present article sets out to explore what we know aboutcompanies’ choices of how to protect their inventions, which is directly linkedwith thequestionofhowcompaniesappropriatereturnstotheirinnovations.

It turns out the answers to these questions are far from obvious. Companies have thechoicebetweenarangeofmechanismstoprotecttheirinnovativeactivitiesandoutput.Onthe one hand, they can choose formal intellectual property (IP),which includes patents,trademarks, registereddesigns, and copyright.On theother, firms can choosea rangeof‘alternative’ or informal appropriation mechanisms, such as secrecy, confidentialityagreements, lead time, or complexity. This paper reviews the theoretical and empiricalliteratureconcerningthechoiceamongformalIPandinformalappropriationmechanismsand combinations of these mechanisms in providing incentives for invention andinnovation,aswellasinshapingafirm’sabilitytocommerciallyexploititsknowledge.

Formal IP is designed to provide ex ante incentives to innovate by providing a rewardsystem thatmakes it easier for innovators tomake ex post profits if their innovation issuccessful,byallowingthemtoexcludeimitatorsforafiniteperiod.ThefinancialrewardtoanIPholderderivesfromthelegalrighttoexcludeothersfromusingtheinnovationandaddresses the fundamental problem of appropriability that governs the production of

3FortheoriginsoftheliteratureseeSchmookler(1966)andComanorandScherer(1969).

5

knowledge (Arrow 1962). Appropriability is a concern for inventors since one of theoutputsofinventiveandinnovationactivityisoftenknowledge,anonexcludableintangibleasset. Hence it is difficult to keep others from using this knowledge at a fraction of theinitial cost of the invention development.4 Although there may be important additionalreasonsforsettingupanIPsystem,theappropriabilityproblemisusuallyconsideredtobethe basic economic justification for an IP system because such a system allows theinventor/innovator to appropriate most of the returns from the initial innovationinvestmentbyexcludingthirdpartiesfromusingtheinnovation.Nevertheless,inpracticeinventionandinnovationdooccureveniffirmscannotaccess,orchoosenottouse,theIPsystem.Aswillbereviewedbelow,firmsusearangeofmechanisms–suchassecrecyorfirstmover advantage–bywhich they appropriate rewards to invention and innovationand the available empirical evidence suggests that firms rely on these alternativemechanismsmuchmorethanonformalIP.5

From a social point of view, granting a temporary property right on an innovation, forexampleintheformofapatent, is justifiedonthebasisthattheinventoris, inexchange,requiredtoexplaintheinnovationinaspecific,standardizedtechnicalformat(thatcanberead and understood by qualified third parties). The economic justification for thisdisclosureistoallowotherfirmstoavoidduplicationofresearch,possiblyacquireusefulknowledgeand,whenthepatentexpires,quicklyimitatetheinnovation.Theseissuesarestressedbyendogenousgrowththeories,whichdemonstratetheimportanceofknowledgespilloversamong firmsandsectors for sustained long‐rungrowth (Romer,1990). In thisrespect,thedisclosureofknowledgerequiredbyapatentapplication‐atleasttheoretically‐ allows knowledge to reach other firms and individual inventors and may help avoidwasteful duplication of research efforts; secrecy, on the contrary, may hinder thecirculation of new ideas and therefore slow down knowledge spillovers and economicgrowth.6

Theavailabilityanduseofthedifferentappropriabilityregimesdifferacrosstechnologiesandsectors;someofthedifferencesareduetodifferencesinlegalsystemsandexogenouscharacteristics of the technologies employed. Endogenous industry demographics andmarketstructurealsoaccountforsomeoftheobservableheterogeneityacrossindustriesinfirms’choicesbetweenformalandinformalIP.However,theappropriabilityregimealsodependsonfirms’strategiccompetitivebehaviorasdiscussedindetailfurtherbelow.

The main forms of formal IP are patents, trademarks, designs and copyright.7 The firstthreeof theseare registered rights,while copyright is anunregistered right. Inaddition,tradesecrecycanalsoberegardedasapartofIP,althoughinmostcommonlawcountries,4Insomecases,thefractionmaybefairlylarge,inthatsuccessfulimitationiscostlyevenwhentheimitatorhasacquiredtherelevantknowledge(Mansfieldetal.,1981).5Inbasictermsonecanthinkofinventionbeingonlythefirststepinacomplexprocesswiththeendpointbeingasuccessfulinnovation.Formally,apatentdescribestheinvention,andnottheinnovationthatmaycomelater.Inthispaperwedonotgenerallyrefertothisdistinctionunlessitiscritical.6However,thereisadebateabouttowhatextentfirmsusepatentdocumentstoobtaininformation.SeeSection5moredetaileddiscussion.7OtherregisteredIPincludesplantbreeders’rightsandsemiconductortopography(mask)rights.OtherunregisteredIPincludesunregistereddesigns,unregisteredtrademarks,andcompanysymbols.

6

includingtheUKandtheU.S.,tradesecretlawformspartofcommonlawandthereforeitsprotectionisweakerthaninothercountries.8SincetheunderlyingmechanismsdifferforregisteredandunregisteredIP,wedistinguishinthisreviewbetweenregisteredIP,payingparticularattentiontopatentsastheyprotecttechnologies,andunregisteredIPintheformof copyright aswell asmost informalmechanisms. Informal IPmay take various forms;secrecy, confidentiality agreements, lead time, and complexity (of design) are subsumedundertheinformalIPheading.9SimilartounregisteredformalIP,informalIPremains,byconstruction, largely unobservable or only partially observable to third parties, whichcreatesaformidablechallengeforempiricalworkaswillbediscussedindetailbelow.

The fundamental question thatwe address in this review is the following:what are thereasons why a firm with a given innovation that can be protected by formal IP wouldchoosenot torelyonsuch IP toprotectan innovation? Insearchofexplanations for thistype of firm behavior, we review the theoretical literature and assess the empiricalevidencetodeterminewhichofthetheoreticalargumentsaresupportedbytheavailabledata. As we will discuss below, the existing evidence shows that there are enormousdifferences in the use of IP at the firm‐level – differences that are beyond expecteddifferences in the applicability of IP (especially patents) to firms’ innovations.10 Theevidenceavailablefromvariousfirm‐levelsurveys,whichisreviewedbelow,suggeststhatonaverage,firmsrelymoreoninformalthanformalIPtoprotecttheirinventions,andthatmostfirmsusenoIPprotectionatall.

Table1usesdatacollated fromtheUKCommunity InnovationSurveys(CIS) to illustratethispoint.11Thetableshowsthe%share(usingsamplingweightstoproducepopulationestimates) of companies indicating medium or high importance to the company of allformalandinformalIPprotectionmechanisms,aswellaspatentingandtheuseofsecrecyindividually.Itshowsthatonly10‐11percentoffirmsrateformalIPorpatentprotectionasofmediumorhighimportance,whereasabout20percentratesomeformofinformalIPorsecrecyasimportant.

Of course, one reason for not using IP protection is that there is nothing that needsprotecting.InthesecondcolumnofTable1welookattheimportanceratingofIPforthe30 per cent of firms that have innovated during the past three years. These firms areindeedmorelikelytoconsiderIPprotectionimportant,with26‐28percentratingformalIPorpatentsasimportantandalmosthalfratinginformalIPorsecrecyasimportant.ButthatstillmeansthatabouthalftheinnovatingfirmsdonotthinkIPisofmuchimportance.

8Sinceenforcementthroughcommonlawisdifficultinpractice,tradesecretsareoftenenforcedthroughspecificcontracts,suchasconfidentialityornon‐disclosureagreements.Althoughthesedocumentsarenotalegalrequirementfortheenforcementoftradesecretsincourt,Almelingetal.(2010a,b)presentevidencefortheU.S.thatthesecretownerismorelikelytoprevailagainstemployeesorbusinesspartnersifsuchanagreementexists.9The“informal”labeldoesnotimplytheabsenceoflegalcontractsandobligations.10SeeforexampleRogersetal.(2007).11ThesedatacomefromtheUKCIS3,4,and5,coveringtheyears1998‐2006.

7

Table1

ImportanceofdifferentIPmechanismstoUKfirms1998‐2006(%)

Allfirms Innovators

FormalIP 10.6 27.7Patents 10.3 25.8InformalIP 18.9 48.5Secrecy 21.0 45.3

Note:FormalIPcontainspatents,trademarks,registereddesigns,andcopyright;InformalIPcontainssecrecy,leadtime,complexity,confidentiality.Data source: UK ONS CIS 3, 4, and 5; Table contains population‐weighted shares based on 38,760observations.

Table 2, drawn from the U.S. National Science Foundation’s new Business R&D andInnovationSurvey(BRDIS)survey,showssimilarresultsforU.S.firms.Lookingatallfirmsinallindustries,onlyasmallfractionfindanyformofIPimportanttothem,andtherankofimportance in terms of the share of firms is trademark, trade secret, copyright, designpatent, and utility patent. When only R&D‐doing firms are considered, the shares ofsomewhat important and very important increase substantially, as onemight expect. Inthiscase,utilitypatentsarestillnotasimportantastheothertypesofIPrights,withtheexceptionofdesignpatents.

Table2ImportanceofdifferentIPmechanismstoU.S.firmsin2008(%)

Allfirms R&D‐doing

firmsUtilitypatent 5 41Designpatent 6 33Trademark 15 60Copyright 12 50Tradesecret 14 67

Population‐weightedshareoffirmsthatratetheIPmechanismassomewhatorveryimportanttotheirfirm.

Source:National ScienceFoundation,NationalCenter for ScienceandEngineering Statistics,BusinessR&DandInnovationSurvey2008.Rowsmaynotsumto100duetorounding.

Hence,understandingwhy firmsmaypreferalternativemethods toprotect their IP is attheheartofunderstandingthefunctioningoftheIPsystem.IftheobjectiveoftheIPsystem

8

is to provide incentives to innovate, an improved understanding ofwhy firms choose torelyonformalIPtoprotectinnovationsinsomecircumstancesbutnotinothershasdirectimplicationsforthedesignofmechanismsthatsetoptimalincentivesforfirmstoinnovate.

Ourreviewcomplementsanumberofexisting literaturereviews thathave lookedat thechoicebetweendifferentmechanismstoprotectinventions.Antonetal.(2006)lookatthechoicebetweenpatents and secrecywithin the contextofweakpatents (i.e.patents thatstand a high chance of revocation in post‐grant administrational or court proceedings).Their discussion focuses on the trade‐off between strategically disclosing information tosoftencompetitionandtheeffectivenessofpatentsinprotectingthedisclosedinformation.Encaoua et al. (2006) also survey the literature on companies’ decision to patent. Theirinterestliesinreviewingtheavailableevidenceonwhetherthepatentsystemencouragesinnovation.

Therearesomerelatedtopicsthatoursurveydoesnotcover.First,wehavenotconsideredtheextensiveliteratureontheprivateandsocialvalueofintellectualpropertyrightsofallkinds, or theuseof IPRsas indicatorsof value. For these topics, seeGriliches (1990)onpatents as indicators, Lanjouw et al. (1998) on patent renewals, and Greenhalgh andRogers(2007a)onthevaluationofIP.Second,wehaveignoredthegrowingliteratureonuniversitypatenting and its implications for theopendiffusionof knowledge,which is aseparateandimportanttopicthatdeservesitsownsurvey.Moweryetal.(2004)providesausefuloverviewoftheU.S.Bayh‐DoleActanditsimpact.

Thestructureofthispaperisthefollowing.Section2discussestheempiricalevidenceonthe topic that is available from firm‐level surveys. Section 3 reviews the theoreticalliteratureonthechoicebetweenpatentsandsecrecyandSection4summarizesthemainresults from empirical analysis, while Section 5 briefly reviews implications for welfareanalysis.SomeconclusionsaredrawninSection6.

2. SurveyevidenceonthechoiceofIPprotectionmethods

BecausethereareaseveralwaysinwhichafirmcanprotectitsIPandsecurereturnstoitsinnovative activity that do not require formal registration, gaining even a partialunderstanding of the IP choice requires asking firms about the methods they use. Thismethodhasbeenemployedinanincreasingnumberofcountriessincethe1980sandwenow have survey evidence frommany countries that sheds light on the choice betweenformalandinformalIPprotectiontools.ThissectionreviewstheavailablesurveyevidenceonthechoicescompaniesmakebetweenformalIPandalternativeprotectionmechanisms.While the theoretical literature–which is reviewed in Section3 ‐ largely focuseson thepatent‐secrecytrade‐off,thesurveyscoveramuchwiderrangeofavailableappropriationmechanisms. In fact, an important finding from the survey‐based evidence is thatcompaniesconsidersomeoftheseotherappropriationmechanisms,notablyleadtime,tobemoreimportantthanpatentsandsecrecy.Table3summarizesthemainfindingsfromthesurveys.

9

2.1. (National)innovationsurveys

TheseminalstudiesinthisareaarethosebyLevinetal.(1987)–socalledYaleIsurvey‐andCohenetal.(2000)–theCarnegieMellonsurvey.Neitheroftheseworksattemptedtodirectly test the empirical implications from economic theory but both surveys wereconcernedwiththeextenttowhichfirmsindifferentindustrieschoselegalandnon‐legalmethods to secure returns from their inventions. The descriptive findings are broadlyconsistent across the two studies. On average, patents are not the most importantmechanism of IP appropriation while secrecy and lead time are, regardless of whetherproduct or process innovations are concerned. However, this is not entirely true forproduct innovationsand for industries thatarespecialized in theproductionof ‘discrete’productslikepharmaceuticalsandotherchemicalswherepatentsarestillthefavoritetooltosecurethereturnstointellectualproperty.Thesurveydatarevealseveralexplanationswhypatentsareconsideredmoreeffectiveforproductthanprocessinnovations.Processesmaynotbepatentable,ifpatentable,theyaremorelikelytodisclosetoomuchinformationtocompetitorsandtheyareseentobeeasytoinventaround.

AtroughlythesametimeastheYaleIsurvey(1981‐83),Mansfield(1986)surveyedabout100U.S.manufacturingfirms,askingthemtowhatextentpatentprotectionwasessentialfor the commercial introduction of their inventions. Mansfield asked specifically for theshare of innovations thatwould not have been developed or commercially introduced ifpatentprotectionhadnotbeenavailable.These counterfactualquestionson theeffectofpatents at the invention‐level are unique to Mansfield’s survey. He found that in twoindustries,pharmaceuticalsandchemicals,patentprotectionwasessentialfor30percentor more of the inventions. In another three industries (petroleum, machinery, andfabricated metals), patent protection was essential for about 10‐20 per cent of theinventions.Theremainingsevenindustries(electricalequipment,officeequipment,motorvehicles,instruments,primarymetals,rubber,andtextiles)showednorelianceonpatents.He also found that in the five industries where patents were relatively important, 84percentofpatentableinventionswerepatented,whereasthesharefellto66percentintheindustrieswherepatentswerenotimportant.HisresultsseemconsistentwiththoseoftheYaleIsurvey.

OneofthefirststudiestofollowupontheYalestudyoutsideoftheU.S.wasthatbyHarabi(1995)forSwitzerlandin1988.UsingthesameformatastheYaleIsurvey,heconfirmedthatSwiss firmsalsorankedpatentsvery lowasameansofappropriatingthereturnstoinnovation,except in thechemicals (includingpharmaceuticals)sectorandsomepartsofthemachinerysector.AsintheYaleIstudy,theSwissfirmsexpressedconcernthatpatentsrevealed toomuch informationandthat itwas tooeasy for firmsto inventaroundthem.However,theywereviewedbysomefirmsasuseful forobtaininglicensingincome.Asinthe Yale I survey, lead time was considered to be the most effective appropriationmechanism. Secrecy was seen as the second most effective appropriation mechanism –althoughsecrecywasratedasmoreeffectiveforprocessthanproductinnovations.

The Community Innovation Survey (CIS), which began in 1992 with seven Europeancountries, included a range of questions on the mechanisms available to companies to

10

appropriate returns to innovating.12 The wording of a typical question on one of thesesurveysisthefollowing:13

For thepast threeyears,please indicate the importance toyourenterpriseofeachof the followingmethods toprotect innovations:patents,registrationofdesigns,trademarks,copyright,etc.

[Therespondentisaskedtospecifyoneofnone,low,medium,orhigh.]

Brouwer andKleinknecht (1999)use the first roundof theCIS,which covers theperiod1990‐1992, for the Netherlands to study manufacturing firms’ preferences over thedifferentappropriabilitymechanisms.TheCISdataconfirmtheresultsfromtheYaleIandSwiss surveys: around half of innovating companies consider patents as insignificant asprotectionagainst imitators.Lead time,keepingqualifiedpeople in the firm,andsecrecy(especiallyforprocessinnovations)arerankedconsiderablyhigherthanpatents.TheCISdata also reveal largedifferences in the importanceof patents across industries. Patentsare considered to be important in chemicals and pharmaceuticals, whereas over 90 percentoffirmsinbasicmetalsregardpatentsasunimportant.Arundel(2001)confirmstheseresults for the firstCIS surveyusingdata forall sevencountries (Germany,Luxembourg,the Netherlands, Belgium, Denmark, Ireland, and Norway). The results show that firmssystematically regard lead‐time and secrecy asmore importantways to protect their IPthanpatents.Over50percentoffirmsranklead‐timeasthemostimportantmechanismtoappropriatereturnstotheirinnovationandnearly17percentregardsecrecyasthemostimportantwaytoprotectaninnovation.Incontrast,onlyabout10percentregardpatentsasthemosteffectivewaytosecurereturnsandonlyabout3percentconsiderregistereddesigns as the most important way to exploit an innovation. The relative greaterimportance of secrecy applies to firms across different size categories, although smallerfirmsregardsecrecyasevenmoreimportantthanlargercompanies.

Whilethesestudiesare informativeaboutthe importanceandeffectivenessattributedbyfirms to thedifferentmechanisms, therearenodataon firms’ actualpatentingactivities.Hall et al. (2013) combine three rounds of CIS data for the UK (1998‐2006)with firms’actualpatentfilingstoshowthatonlyasmallshare(4percent)ofinnovativecompaniesintheUK patents. Again, companies rate lead time, confidentiality agreements and secrecyhigher than patents as mechanisms to protect their innovations. Hardly surprisingly,patenting companies regard patents as much more effective a mechanism to protectinnovationsthanfirmsthatdonotpatent.However,thereishardlyanydifferencebetweencompaniesthatpatentandthosethatdonotwithregardtohowimportantsecrecyorleadtimeisratedbycompanies.Theresultsalsoshowthatthedecisiontopatentbyinnovativecompanies is largelyexplainedbya fewfactors: thetypeofactivitycarriedoutbya firm

12TherewasanearlierinnovationsurveyinFranceduringthelate1980sthatwasaprecursoroftheCIS.

13ThiswordingisfromtheCIS3andCIS4intheUK.Obviouslyinothercountries,thequestionisaskedintheappropriatelanguageandmaythereforevaryslightlyfromsurveytosurvey.

11

(industry),size,andthe typeof innovationcompaniesproduce(product innovations thatarenewtothemarket).

IntheU.S.,thereisnoequivalenttotheEuropeanCIS.ButtheNationalScienceFoundationandtheU.S.CensusBureaulaunchedtheBusinessR&DandInnovationSurveyin2009.Thesurveycollectsdataonfirms’R&DspendingandinnovativeaswellaspatentingactivitiesandisthusalotbroaderthantheSurveyofIndustrialResearchandDevelopmentwhichitreplaced.The surveyhas anumberof advantagesover theCIS: it collectsdataonglobalactivities of companies operating in the U.S., it asks for R&D employee headcounts byoccupationcategory,anditseparatessalesandR&Ddatabybusinessactivity.Sofaronlyafew results have been released. The survey reveals that in 2008, 21 per cent of R&Dperforming companies applied for a patent (U.S. NSF, 2013). Shares vary widely acrossindustries:morethan60percentofcompaniespatentedinbasicchemicalswhereas lessthan10percentpatented in food. Inanycase, these figuresaresignificantlyhigher thanthefiguresfoundbyHalletal.(2013)fortheUK.14

There is also growing survey evidence from countries other than the U.S. or Europe.15Hanel(2008)forexamplereportsresults froma1999surveyofCanadianmanufacturingfirms. The results are very similar to those for U.S. and European firms. Companiesconsistentlyregardalternativemechanisms,especiallyconfidentialityagreements,asmoreimportantthanpatents.Interestinglyfirmsalsoratetrademarkshigherthanpatents.AsintheU.S.andEuropeansurveys,patentsareneverthelessimportantindiscretetechnologies,pharmaceuticals,chemicals,andplastics.Thesurveyalsosuggeststhatinnovatorsof“new‐to‐the‐world” innovations relymore onpatents than secrecy. This is expected, given thenoveltyrequirementforpatenting,but italsosuggestsapositivecorrelationbetweenthevalueofinnovationsandthepropensitytopatentthem.

Sofarwefocusedinthediscussiononfirms’assessmentoftheimportanceofthedifferentappropriation mechanisms for protecting innovations. Some of the surveys also offerinformationonmotivesforfirms’patentingdecisions.Cohenetal.(2000)findthatfirmsinthe U.S. use patenting for strategic reasons rather than for protecting their intellectualproperty.Respondentsreportedthattheyusedpatentingtoblockcompetitors,toimprovegoodwill reputation and to improve bargaining power in the market. A similar type ofanalysis conducted on a sample of patenting firms confirms these findings for Germany(Blindetal.,2006).AlthoughmostGermanfirmsindicatethattheyusepatentsprimarilytoprotecttheirinnovationsfromimitationandhelpsecurefreedomtooperate,firmsalsousepatents for offensive blocking of competitors (strategic patenting) and to improve the

14Hall etal. (2013)useabroaderdefinitionofR&D,whichmeans thesamplecontainsabroader rangeofcompanies. The companies in the U.S. sample may also take into consideration patent filings by foreignaffiliates.MoreoverthedifferencemayalsobepartlyduetoundercountingofpatentsbyHalletal.(2013)ontheonehand(becausepatentshavetobematchedbynametocompanies)andover‐reportingbycompaniesintheU.S.survey.15InnovationsurveysfollowingtheCISmodelareconductedinawiderangeofcountries, includingmiddleincomeeconomies,forexampleBrazil’sPINTECsurvey.

12

company’s image.16 Jung andWalsh (2010) also focus on patenting companies to learnabouttheirmotivesforpatenting.Theirsurvey, theGeorgiaTechInventorSurvey,asksasample of U.S.‐based inventors listed on triadic (filed with the USPTO, EPO, and JPO)patentsfortheirmotivestopatent.TheresultsconfirmthefindingsbyCohenetal.(2000)and Blind et al. (2006): a substantial share of inventors lists offensive and defensiveblockingasanimportantmotivationforpatenting.Offensiveblocking,whichincludestheuse of patent fences, is found to be more common in discrete technologies, such aspharmaceuticals. JungandWalshalsoask inventors forreasonswhypatented inventionsarenotusedbycompanies,whereuse isdefinedinthreeways:1)companiesexploit thepatented invention themselves,2) they license itout, or3)use it ina spin‐out company.Themostcommonlycitedreasonsfornon‐usearethatnocommercialusehasbeenfoundyet,thatthepatentisusedpassivelytoblockcompetitorsandpreventinventingaround,orthattheinventionwassubjecttomarketortechnologicalobsolescence.

Some of the surveys discussed above also explore potential differences in the use ofappropriationmechanismsbetweensmallandlargefirms.ItiseasytoseefromsummaryTable3thatthechoiceofsamplingframeaffectsthefrequencywithwhichthefirmsinthesampleuseanykindofIPprotectionmethod.Forexample,comparetheshareoffirmsthatrate patents or secrecy of high importance in the Arundel (2001) or Brouwer andKleinknecht(1999)sample,whichcoversawidesizerangetotheshareintheArundeletal.(1995)report,whichisrestrictedtolargeR&D‐doingfirms.Incontrast,leadtime(beingfirsttomarket)isratedrelativelyhighlyinallsamples.

The 2008 Berkeley Patent Survey conducted by the Berkeley Center for Law andTechnology specifically targeted small, new companies in the U.S. (Graham etal., 2010).Grahametal.(2010)andSichelmanandGraham(2010)summarizetheevidencefromtheinformationobtainedonaround1,300high‐tech(bio‐techandsoftware)start‐upsfoundedintheU.S.since1998.Theynoteimportantdifferencesinpatentingbehaviorandthewayinwhichpatents areused across industries.While for some industries, such asbio‐tech,patenting is a vital part of corporate strategy, firms in other sectors, notably software,essentiallyavoidthepatentsystemaltogether.Theyalsopointoutthatstrategicmotivestopatentasdescribedaboveareimportantforstart‐ups,contrastingthecommonlyheldviewthat strategicpatenting isonlypracticedby largeenterprises: indeedstart‐upsvalue thereputationeffectthatpatentownershipmaybringabout.Despitetheimportanceofpatentsreported by bio‐tech andmedical device companies, companies still report that patentsprovidehardlyanyincentivesforR&D.Thesurveyalsoasksfirmsdirectlywhytheychoosenottopatentanditturnsoutthatthemostsignificantbarriertopatenting(acrossthetwoindustries) is financial. However, bio‐tech firms rate concerns about the disclosure ofinformation contained in a patent publication as a greater obstacle than costswhile theoppositeistrueforsoftware.Tosomeextentthisdifferencereflectsthedifferinguseofthesystembyfirmsinthetwoareas:bio‐techsworryaboutdisclosurebecausetheydopatent(in spiteof the cost),whereas software firmsworrymoreabout costbecause theydon’t.

16TheGermancarmanufacturerAudiprovidesanexampleforhowacompany’simagecanbeinfluencedbyits patent filings.When launching a new carmodel in 2006, the company launched amulti‐million dollaradvertisingcampaignstating`Todate,NASAhasfiled6,509patents.IndevelopingtheA6,Audifiled9,621.’

13

Software companies also indicate that patents are of little use for them becauseenforcementwouldbeprohibitivelycostlyandinventingaroundisrelativelyeasy,makingpatentsineffective.Instead,softwarecompaniesconsiderleadtimeasthemostimportantwaytoappropriatereturnstoinnovating.

WhereasGrahametal.(2010)focusonstart‐upcompanies,HyytinenandPajarinen(2002,2003)andLeiponenandByma(2009)collectdataonawidersetof small companies. IntheirsurveyofsmallFinnishcompanies,thereisadistinctionbetweenfirms’ratingoftheeffectiveness of the different appropriability mechanisms and their actual use bycompanies.25percentof thesmall companies in thesample report thatpatentsare themostimportantappropriabilitymechanism,only15percentreportsecrecytobethemostimportant mechanism. This contrasts with the information on the actual use of thesemechanisms.62percentofcompaniesindicatethattheyrelyonsecrecyandonly16percentreportthattheypatent.ThisdisagreeswithMansfield’sfindingsfortheU.S.wherehefoundthatcompaniesthatindicatedthatpatentsplayednoimportantrolewerestillfoundtopatenttheoverwhelmingshareoftheirpatentableinventions.ThedifferenceisprobablyexplainedbythefactthatMansfieldcontactedmainlylargeestablishedfirms.

Oneimportantfactorthatneedstobeconsideredinevaluatingtheresultsofthesesurveysisthepositionheldbytherespondent,especiallyinlargefirms,wherethereisconsiderabledifferentiation in areas of responsibility. Many of the surveys of R&D‐doing firms aredirectedtotheR&Dmanager,asinthecaseofLevinetal.(1987);theCISquestionnairesdonotspecifywhoshouldfill themout. Incontrasttothemajorityofthesurveys,CockburnandHenderson(2003)surveysenior IP in‐housecounselsacrossarangeof industries intheU.S.ThedatacollectedfromIPcounselssuggestthattheavailabilityofpatentshasanimportantpositiveeffectonR&Dspending.IPcounselsdonotthinkthatpatentsrevealtoomuch valuable information to competitors. Moreover, patents are seen as importantinstruments to settle IP disputes and avoid litigation. Half of all the respondents to thissurveyidentifypatentsasthecompany’smostvaluableIPasset,andtradesecretsareonlyseenby12per cent of respondents as a company’smost valuable IP asset. Thuswe cansafelyconcludethatthelegaldepartmentsoffirmsgenerallyratepatentsmorehighlythaneitherfirmCEOsorR&Dmanagers.

Thefactthatsurveyevidenceresultsdependonthechoiceofrespondenttellsusthatweshouldbecautiousinassigningtoomuchnumericalcredibilitytothesharesthatresult.Inparticularitmaybesomewhatmisleadingtocompareacrosssurveysunlessthequestionsandthetypesofrespondentsareidentical.Nevertheless,withoutthesesurveyswewouldperhapsnothavelearnedtheextenttowhichfirmsrelyoninformalmethodslikeleadtime,secrecy, product complexity, and sales and service activity to secure returns to theirinnovativeactivities,asopposedtoformalmethodslikepatentsandtrademarks.

2.2. Cross‐countrycomparisons

Mostempiricalstudieshavebeenconductedonsurveydataforasinglecountry;whilethisallowsresearcherstoidentifythefirms’andindustrycharacteristicsthatcanexplainfirms’preference for secrecy (or for patents), single country studies do not allow the

14

identification of the characteristics of the patent legislation that can influence thispreference. For instance, in countrieswhere the procedure to apply for a patent is verycumbersome,some firmsmay find thewholeprocesssoexpensivetheyprefer toopt forsecrecy to protect their IP (see for example Moser (2012) for a historical comparisonbetween Britain and the U.S.). Some researchers have used cross‐country data (orinternationaldatasurveys)tounderstandhowthenationalpatentlegislationcanaffectthechoice between patents and secrecy. We focus here on the PACE survey developed byArundeletal.in1995andonthesurveysofJapaneseandU.S.firms(Pitkethly,2001;Cohenetal.2002).

The PACE survey was directed to the European Union’s 840 largest manufacturing andindustrialfirmslocatedinGermany,theUK,Italy,Belgium,theNetherlands,Belgium,Spain,Denmark, and France. The findings of the PACE report confirm important industryvariationsregardingtheeffectivenessof IPprotectiontools forEuropeanfirms.As inthecaseofLevinetal.(1987)andCohenetal.(2000)fortheU.S.,patentsplayanoutstandingroleinthepharmaceuticalandchemicalindustryforbothproductandprocessinventions.Secrecy is important in protecting process inventions in most industries. Arundel et al.(1995) suggest that differences in IP legislation and enforcement can explain whyEuropean firmstendtouseadifferentmixof IP tools thanU.S. firms.ArundelandKabla(1998)combinethePACEsurveydatawithdataonFrenchfirmscollectedbytheFrenchOffice of Industrial Studies and Statistics (SESSI). The data are used to investigatedifferences in patent propensities across industries. As expected, patenting propensitiesarehigherforproductinnovators(36percent)thanprocessinnovators(25percent).Thesectors with the highest patenting propensities are pharmaceuticals and precisioninstruments;thelowestpatentingpropensitiesarefoundintextilesregardlessofproductorprocessinnovators.ThesurveysalsorevealthatfirmsaremorelikelytorelyonpatentprotectioniftheyexporttotheU.S.orJapan.

Cohenetal.(2002)compareresultsfromasurveyamong593Japaneseand826U.S.firmsregardingtheimportanceofpatentsasappropriationmechanisms.Itemergesthatslightlymorerespondents in Japanratedpatentsasaneffectivemeans toprotecting innovationsthanintheU.S.(38percentand36percentrespectivelyforproductand25percentand24percentrespectivelyforprocessinnovations).ThemorestrikingresultisthatsecrecyisregardedasamuchlesseffectivewaytoprotectinnovationsinJapanthanintheU.S.(26per cent and 51 per cent respectively for product and 29 per cent and 53 per centrespectively for process innovations). Hence, while in the U.S. other appropriationmechanisms, aboveall secrecyand lead time, are regardedas themosteffectivewaysofprotecting innovations, in Japan, patents are equally important as any of the othermechanisms.Theauthorsexplain thesedifferences in the importanceofpatents in Japanand theU.S. by institutionaldifferences in the countries’ patent systems. For instance, atthe timeof the survey,patent lawsallowed innovators to apply for apatentearly in theinnovationprocessinJapanduetoafirst‐to‐fileruleofpriority(asopposedtotheU.S.first‐to‐invent rule of priority). Also, Japanese patentswere subject to “pre‐grant opposition”while no analogous oppositionprocess existed in theU.S. Thesedifferences implied thatJapanesefirmsratedpatentsasastrongertooltoprotecttheirIP.Pitkethly’s(2001)surveyofpatenting JapaneseandUKcompaniesalsosuggests that Japanesecompaniesconsider

15

patents as more important a source of technical information than UK companies.17 AsimilarresultisseenbyCohenetal.(2002),whofindthatJapanesefirmsaremuchmorelikely to seepatentsas informationsourcescontributing to thecompletionof innovationprojectsandtoviewpatentsasanimportantsourceofinformation.

Thesedifferencesobtainedfromcross‐countrycomparisonsareinteresting.However,thecomparisonofrelativelysimilarIPsystems(especiallywithintheEuropeanUnion)limitsthe degree to which such differences can explain observed differences in the choicebetweenformalandinformalappropriationmechanisms.

17Pitkethlysurveyedin1994anon‐randomsampleof211Japaneseand259UKcompaniesthathaveappliedforpatents.ThesurveyfocusesonIPmanagementpractices.

16

Table3:Summaryofmainsurveyresults

Survey Levinetal.(1987)

BrouwerandKleinknecht(1999)

Arundel(2001)

Cohenetal.(2000)

Blindetal.(2006)

Arundeletal.(1995);ArundelandKabla(1998)

Cohenetal.(2002)

Periodcovered 1981‐83 1990‐1992 1990‐1992 1994 2002 1990‐1992 1994

Country U.S. NLDE,LU,NL,BE,DK,IE,NO

U.S. DEUK,DE,IT,NL,BE,ES,DK,FR

U.S.,JP

Coverage

650linesofbusiness,R&D‐doingmfgpubliclytradedfirms

1000‐2000mfgfirms

2849R&Ddoingmfgfirms

1165largeR&D‐doingmfgfirms

522firmswith≥3EPOpatentapplications

414PACE+190FrenchlargeR&D‐doingmfgfirms

593largeR&D‐doingmfgfirms

Highim

portance

PatentsProd:4.3*Proc:3.5*

Prod:25%Proc:18%

Prod:11%Proc:7%

Prod:35%Proc:23%

79%Prod:67%Proc:46%

Prod:JP38%;US36%Proc:JP25%,US24%

Secrecy Prod:3.6*Proc:4.3*

Prod:33%Proc:41%

Prod:17%Proc:20%

Prod:51%Proc:51%

58% Prod:54%Proc:65%


Leadtime

Prod:5.4*Proc:5.1*

Prod:57%Proc:56%

Prod:54%Proc:47%

Prod:53%Proc:38%

88%Prod:67%Proc:46%


Sector

PatentsHigh:pharmaLow:pulp,paper

High:pharma/chemicals/petroleumLow:basicmetals

n.a.

High:medicalequipment,pharmaLow:printing/publishing

High:rubber&plastic,biotechLow:construction/mining

High:pharmaLow:prod:utilities;proc:electricalequip.

n.a.

Secrecy n.a. n.a. n.a.

High:misc.chemicalsLow:printing/publishing

n.a.

High:FoodLow:prod:fab.metals;proc:utilities

n.a.

%patentingcompanies

n.a. 2.7% n.a.Prod:49%Proc:31%

100% 86%Prod:JP62%;US54%Proc:JP42%,US32%

Notes:CH:Switzerland,DE:Germany,LU:Luxemburg,NL:Netherlands,BE:Belgium,DK:Denmark, IE: Ireland,ES:Spain, IT: Italy,NO:Norway,CA:Canada,FR:France, JP: Japan;mfg:manufacturing;prod:product innovation;proc:process innovation; *mean scores (range:1=notat all effective,7=veryeffective).

17

3. Theory

In this sectionwe review theoretical arguments that suggestwhyandhow firms choosebetweenformalandinformalIPprotection.WeconcentrateourdiscussiononpatentsandsecrecyasthesearethemaintypesofformalandinformalIPconsideredinthetheoreticalliterature.Itisalsotruethat,inprinciple,thechoicebetweenpatentsandsecrecyinvolvesanexplicitandfairlystarktrade‐offbetweendisclosureandnondisclosureofaninventiveidea.Incontrast,considerthetraditionaluseof intellectualpropertyinformofcopyrightfor protection of software by protecting its expression in bits and bytes. This use iscommonly combined with the use of trade secrecy for the code that exists in a formcomprehensiblebyhumans,sothatinprinciplenotrade‐offbetweencopyrightandsecrecyexists.Thatis,asoftwareprogramcanbebothpublishedinmachineformundercopyrightand protected (at least barring considerable effort in reverse engineering) by secrecy.18Thesameconsiderationsapplytosuchinformalinstrumentsasleadtimeandcomplexity–inprinciplethesecaneasilybecombinedwithanyoftheformalmechanismssothereisnoreal need to choose. However, in principle, full secrecy rules out the use of the patentinstrumentforprotectingIP,whichiswhythereisafocusonthischoiceinthetheoreticalliterature.

Table 4 shows schematically the four options available to a firm facedwith the decisionwhethertopatentormaintainaninventionsecret.Ofcourse,thisassumesthatfirmshaveachoice. If for example the invention does not represent patentable subject matter, theoption of obtaining patent protection does not arise. Importantly, the table refers toinventions as the unit of observation, not products. Products often embody a number ofinventions, which can be protected by different mechanisms (see Section 4 for morediscussion).Table4alsoshowsthatapartfromeitherpatentingorsecrecyinthelower‐leftandupper‐rightquadrantrespectively,firmsmayinfactcombinesecrecyandpatentingtoprotect a given invention (upper‐left quadrant) or discard both options (lower‐rightquadrant).Inwhatfollows,wediscusseachoftheseoptionsinturn,althoughwefocusonthechoiceontheoff‐diagonalsbecausethatisthemostwidelymodeled.

Table4:Patentingvs.secrecy

Patent Don’tpatent

Secrecy Patent‐secrecycombination Secrecyonly

Non‐secrecy Patentonly Disclosure–publishing

Source:Graham(2004)

18Thelegalsituationwithrespecttoreverseengineeringofsoftwarecodeissomewhatcomplexandevolving.SeeSamuelsonandScotchmer(2002)forafullerdiscussion.

18

3.1. Patentandsecrecyassubstitutes

Muchof thetheoretical literatureregardsa firm’schoicebetweenpatentsandsecrecyasmutuallyexclusive(Friedmanetal.1991).Thechoiceisexplainedbytheinherenttrade‐offbetween the benefits from using patents and its costs relative to relying on secrecy.Benefits and costs are not only a function of the invention that qualifies for patentprotection,butalsoofdefensiveoroffensivestrategicconsiderationstakingintoaccountafirm’s competitors’ behavior.This ishardly surprising in lightof thenatureof IP, i.e., itsvalue lies in affecting third parties’ behavior rather than directly affecting a firm’s owninputsintoproduction.19

There are differences in the legal protection conferred by patents and secrecy. First,patents are granted only on patent eligible subject matter. The definition differs acrossjurisdictions, for example the EPO does not consider computer enabled inventions(software)perseaspatenteligiblesubjectmatterwhereastheUSPTOdoes.Hence,secrecyis applicable to amuch broader range of inventions than patents, as there is no formalsubjectrestriction.20Patentabilityrequiresaninventiontosatisfythenoveltyandinventivestep thresholds, whereas there is no such requirement for secrecy (even when non‐disclosureagreementsareused).21Thisimpliesthatsecrecycanprotectworkinprogress,whereasonlyinventionsthathavereachedacertainstageofdevelopmentcanbepatentedas theyneedtomeet thesestatutoryrequirements forpatentability.Second, ifgranted,apatentoffers20yearsofstatutoryprotection.Secrecy,incontrast,canpotentiallyprotectthe invention indefinitely.However,an invention that isbeingkeptsecretmay legallybepracticedbya firm thathas independentlydiscoveredor reverseengineered it,whereasthesameisnottrueofapatentedinvention.22

Sometechnologiesareeasiertoprotectwithapatent,andoncegranted,apatentiseasiertoenforce,becausethetechnologyisinherentlyeasiertodescribeanddelimit.Thisfactormay affect thewillingness of firms touse thepatent system.The leading exampleof theimportance of this is the chemicals sector, including pharmaceuticals, where patents

19A firm’sownfreedomtooperate isensuredbyblockingthirdparties fromclaimingpropertyrightsonaspecificinvention.20 Secrecy usually protects commercial information, financial data (e.g. pricing and costing data), businessmethods, business strategies, business plans, customer/supplier lists, technical designs, drawings, blueprints,maps,negativeinformation(i.e.thingsthathavebeentriedbutdidnotwork),prototypes,formulae,recipes–forwhichpatentprotectionwouldnormallynotbeavailable.21 According to Article 39 TRIPS, any “undisclosed information” is protected by secrecy if secrecy ismaintained through “reasonable efforts,” although the effective protection of trade secrecy variessubstantially by jurisdiction. Some jurisdictions may impose additional requirements such as that theinformationbenon‐trivial.Contrarytopatents,tradesecrecyprotectiondoesnotrequireaninventivestep.Itsufficesthattheundisclosedinformationhaseconomicvalue,i.e.athirdpartywouldbenefiteconomicallyifitgainedaccesstotheinformation.Dependingonthejurisdiction,tradesecrecymayalsobeprotectedbyunfaircompetitionlaw,tortlaw,criminallaw,aswellasbreachofconfidenceandcontractualobligations.22Thesituationisslightlymorecomplicatedunderafirst‐to‐inventprioritysystemsuchasprevailedintheU.S.until2012.Inthiscase,thefirmholdinganinventionsecretcouldinprinciplefileforapatentifanotherfirm filed for a patent on the same invention if it could show that it invented first. An “interference”proceedingatthepatentofficemightbetheoutcomeandthereisnoguaranteethatthefirstfirmwouldwin,giventhedifficultyofestablishingpriority.

19

protectaspecificcompoundthatcanbedescribedinaprecisechemicalformula.Thisleadsto fewer (although not zero) disputes over the exact breadth of the patent and easiernotice, in the property rights sense (Bessen and Meurer, 2008). The contrast with theinformation technology sector, where the precise breadth of patent is often unclear, isstriking, and is one of the reasons that patents in that sector are generally not usedprimarilyforappropriatingthereturnstoinnovation.

Applying for a patent requires direct and indirect financial expenditures. If a patent isgranted it protects an invention only in the jurisdiction in which it was granted. Thepublished patent is visible, however, also outside of jurisdictionswhere the patent is inforce. Thus the invention could be legally imitated and used in jurisdictions where thepatent isnot in force.Moreover,tokeepthepatent inforce,maintenancefeeshavetobepaid toeachpatentofficewhichhasvalidated thepatent.However,secrecy isalsocostlybecauseoftenitisvitalthatconfidentialityagreementsareusedandtheknowledgeoftheinventionisguarded.Keeping innovationssecretusuallyrequiresconsiderableeffortandactive knowledge management in the form of an internal secrecy policy, which may becostlytoimplementandmaintain.Forexample,firmsmayrelyonthesplittingofR&Dintodifferent componentsacross researchersandresearch labssuch that individualpiecesofR&Ddonotallowacompleteunderstandingandfunctioningofagiventechnology.23

Both patents and secrecy are expensive to enforce. In the case of patents, enforcementrequiresactivemonitoringofpotentialinfringementandthefinancialresourcestoengageinlitigation.24Enforcementofsecrecywillalsobecostlyandmaybedifficulttoachieveincourt.25 The legal protection of secrecy is alsonarrower as onlymisappropriation of theinventioncanbeclaimedincourt.Thevalueofapatentorsecrecydependsonthefinancialcapability of the owner to undertake legal action in case of infringement or breach ofconfidentiality isdetectedand the likelihoodof successof thecourt case.WesummarizetheseconsiderationsinTable5.

23Zhao(2006)providesempiricalevidencethatmultinationalfirmstendtosplitknowledgemoreifpartoftheresearchisexecutedincountrieswithweakIPrightsprotection.Aratherlow‐techproductalsoillustratesthis point: Thomas’s EnglishMuffin. A recent U.S. court case suggests that Thomas splits the recipe of itsEnglishMuffins,whichisatradesecret,intoseparatecomponents,suchasthebasicrecipe,themoisturelevelofthemixture,andthebakingprocess.Reportedlyonlysevenkeyemployeesknowallstepsrequiredtomakethe muffins while all other employees only have knowledge about their specific assigned task in themanufacturing process. This case also serves to illustrate the threat to secrecy that emerges from themovement of personnel, as the court case was triggered by concerns that one of the seven “informed”employeesmightrevealhisknowledgeafterhavingacceptedajobwiththecompetitor.SeeNewYorkTimes,August6,2010.24ForexampleincaseoftheUK,totalcostsfortheplaintiffanddefendantofacourtactioninapatentcasearetypicallybetween£1and£6million(HelmersandMcDonagh,2012).25Dataoncourtcasesallegingthemisappropriationoftradesecretsshowthatintheoverwhelmingshareofcases(former)employeesandbusinesspartnersareaccusedofmisappropriation(Almelingetal.,2010a,b).This may indicate that proving misappropriation of trade secrets by third parties in court, includingmisappropriationbycompetitors,maybeverydifficult.Moreover,theownerofatradesecretmayhavelittleincentivetopursueacaseincourtifthiscouldleadtomoreconfidentialinformationbeingrevealedduringcourt proceedings. As Neil Wilkoff put it in a recent IPKat blogpost, “Whatever the amount of any courtjudgmentforbreach,thetradesecretsmokeisoutofthebottleandnotevenAladdinwillbeabletoretrieveit.”

20

Table5:Factorsaffectingthepatent‐secrecychoice

Patents Secrecy

Disclosure(codifiableknowledge) Yes No

Disclosure(tacitknowledge) No No

Easeofdelimitinginvention Yes Notclear

Reverseengineeringallowedingeneral

No Yes

Subjectmatter Statutory Broader

Timing Afterinvention Work‐in‐progress

Processvs.product Both Easierforprocess

Length 20years Longer(potentially)

Costtoobtain Higher Nonzero

Enforcementcost Expensive Expensive

Second,thereisuncertaintyinvolvedintheuseofbothpatentsandsecrecy.Withregardtopatents, there are three instances where outcomes are uncertain: whether a patent isgranted,whetherthepatentisinvalidatedpostgrant(eitherbythepatentofficeoracivilcourt),andwhether infringementcanbeproven, that iswhether itcanbeshownthataninfringingactionfallswithinthevalidclaimsofapatent.Incaseofsecrecy,theuncertaintyconcernsmostlywhethertheinventioncanbemaintainedsecret,althoughthequestionofwhetherforexamplebreachofconfidentialitycanbeenforcedmaybealsorelevant.

Thetheoreticalliteraturediscussedbelowtakesthesefactorsasgivenandinsteadfocusesonspecificelementsofthepatent‐secrecytrade‐off.Inthefollowingsubsectionswediscusstheoreticalmodelsofthepatent‐secrecychoice,classifyingthemintomodelsthatfocusondisclosure,competitionforinnovation,leadtimeandcomplexity,andtheconsequencesofsequentialorcumulativeinnovation.

3.1.1. Disclosure

In principle, once an inventor decides to disclose an invention, he forfeits the option toprotecttheinventionthroughsecrecy.Moreover,theinformationdisclosedinapatentmaybeonlyimperfectlyprotectedbythepatentsimplybecauseinfringementhastobedetectedandenforcedthroughlitigation,whichmayhaveanuncertainoutcome.Thissuggeststhatdisclosurerequiredbyapatentmayhaveanimportanteffectonaninventor’sdecisiontopatent ormaintain an invention secret.Horstmanet al. (1985) show that an innovator’spropensity to patent falls if patents reveal information to competitors. The model byHorstmanetal.,however,doesnotallowfirmstostrategicallychoosetheamountdisclosedthrough a patent. The seminal model by Anton and Yao (2004) [AY] in contrast looks

21

directlyattheroleofdisclosureinafirm’sdecisiontopatent.Inthemodeldisclosurehastwoeffects,ontheonehandithelpscompetitorsinnovateandontheotheritcanbeusedstrategically to transmit information on the innovator’s competitive edge to a laggardcompetitor.We sketch the AYmodel below, using a general downward‐sloping demandfunctionratherthanthelineardemandfunctionusedbyAY.

TheAYmodelisathree‐stagesignalinggamewithtworisk‐neutralfirms–theinnovator,i,and thecompetitor, j. In the first stage (“protectionanddisclosurestage”), the innovatorinvestsinR&D;theoutcomeisaprocessinnovationallowingtheinnovatortoproduceatmarginalcost,c.TheoutcomeoftheinvestmentinR&Disuncertain:incaseoffailure,thefirmusestheexistingtechnologyandproducesatamarginalcost c >c.TheR&Doutcomeisobservedbytheinnovatorbutnotbythecompetitor.26Theinnovatorhastodecidehowtoprotecthisinnovation.Hehastwooptions:hecaneitheruseapatent(P)orcankeeptheinnovationsecret(S).RegardlessofwhethertheinnovatorchoosesPorS, informationontheprocessinnovationleakswhichallowstheimitatortoreplicatetheprocessinnovationand produce atmarginal cost s (where s ≥ c). That is, AY assume that disclosure underpatentingisthesameasundersecrecy,withtheonlydifferencebeingthattheimitatorisliabletodamagesifheisfoundtoinfringethepatent.Thisisafairlyinnocuousassumptionwhoserelaxationwouldcomplicatethemodelwithoutaddingmuchinsight.Inthesecondstage(“infringement‐riskingimitationstage”),havingobservedthechoiceoftheinnovatorofthepreviousstage,thecompetitor(whodoesnotinvestinR&D)hastochoosewhetherto imitate and risk infringement (if the innovator has chosen P) or use the existingtechnology.Inthethirdstage(“competitionstage”),thetwofirmscompeteasCournot.

CompetitionStage.Therearethreedifferenttypesofcompetitionregimes,dependingonthechoicesoftheinnovatorandofthecompetitorinthepreviousstagesofthegame.Iftheinnovatorhasdecidedtorelyonsecrecy,thecompetitorcanimitatetheinnovationatnocost and produce at marginal cost s. In this case, the market reduces to pure Cournotcompetition. If the innovator has obtained a patent the competitor can imitate and riskinfringementor stickwith theexisting technology. Ifheuses theexisting technology,hismarginalcostis c .FirmscompeteunderCournot,althoughtheimitatornowfacesalargermarginalcost c s .Iftheimitatordecidestoimitatethepatentedinvention,hismarginalcostissbutheisexposedtotheriskofbeingcaughtinfringingthepatent.Inthemodel,theimitator is found to infringe the patent with probability γ. If infringement is found, theimitatorhastopaytheinnovatordamages.Inthemodel,thesearemodeledasroyaltiesonthe revenues: τpqj where τ is the royalty rate. Hence γτ = g defines the expectedinfringementdamagesrate.AYinterpretthisasanindicatorofthestrengthofthepropertyrightsavailabletotheinnovator.Ifg=1,thentheinnovatorhasfullpropertyrights;ifg=0hehasnopropertyrightsintheinvention.

Two of the three competition regimes are straight Cournot, with costs (c,s) or (c,c ).Therefore we focus on the third case, where there is a patent and the imitator risks

26EncaouaandLefouili(2005)obtainresultssimilartoAYinafullinformationset‐up.

22

infringement.Inthiscase,theprofitsfortheimitatorjareequaltothedifferencebetweentotalrevenuesandtotalproductioncostsnetofexpectedinfringementdamages:

( )j j jp s q gpq (1.1)

The profits for the innovator i can be expressed as the difference between the totalrevenuesandtotalcostsplustheexpecteddamagesfromtheinfringementofthepatent:

( )i i jp c q gpq (1.2)

We assume a monotonic demand function p = F(Q) = F(qi+qj) with F’<0. From this wecompute best response functions for the innovator and the imitator, and solve for theequilibrium output choices as a function of the degree of disclosure and the strength ofpropertyrights.

* **

* * * ** *

1( )

'( ) 1

1 1( ) (1 ) ( )

'( ) '( ) (1 )

j

i j

sq F Q

F Q g

gsq F Q c gq g F Q c

F Q F Q g

(1.3)

With some tedious manipulation, we can use equation (1.3) to show that the weakerpropertyrightsprotectionis(lowerg),themoretheimitatorproduces,andconverselyasgincreases.27Onecanalsoshowthatassincreasesrelativetoc(lessdisclosure),theratioofthe innovator output to imitator output grows. In passing, note that j will not produceunlessthemarketpricecovershismarginalcostandtheroyaltieshehastopay[p*(1‐g)>s].

Theexpectedprofitsinequilibriumare:

2*2 *

*2 * *

1(1 ) 2

'( *) 1

12

'( *) 1

j

i

sg p sp

F Q g

gsp gcp c c cp

F Q g

(1.4)

AssumingalineardemandfunctionasinAY,onecanuseequations(1.4)toshowthetrade‐off theinnovatorfaceswhenchoosingtheoptimaldegreeofdisclosure.Ontheonehand,disclosing toomuch information (lower s) reduces profits as the imitator cannowmoreeasilyreplicatetheprocessinnovation.Ontheother,theprofitsoftheinnovatorincrease(and those of the imitator fall) if the imitator infers that the innovator has lower costs(smallers).

27Therearesomeauxiliaryconditions:F’’hastobesufficientlysmallandpositive(orzero,asinthecaseoflineardemand),andithastobeprofitablefortheimitatortoproduce.

23

Infringement‐risking imitationstage.Thecompetitorhastochoosewhetherto imitate(riskinginfringement)orusetheexistingtechnology.Hecanobservewhethertheprocessinnovationhasbeenpatentedandthedegreeofdisclosure.Hecannotobservethemarginalcost of the innovator and therefore is not fully aware of the extent of the competitiveadvantageconferredbytheinnovation.Iftheinnovatorhasnotpatented,theimitatorcanuse the disclosed knowledge and produce at cost s without infringement risk. If theinnovator has patented, the imitator has two alternatives: he can imitate and riskinfringementorusetheexistingtechnology.

Supposetheinnovatorchoosestopatentanddisclosess.Ifthecompetitordoesnotimitate,hefacesacostdisadvantageofc versusc,andprofitsequalto 2( ) / '( )N Np c F Q ,whereNdenotes the price and quantity with no imitation. If he decides to imitate, the costdisadvantage issversusc,butthere istheriskofhavingtopaydamages.Usingequation(1.4),theexpectedpayoffsforthecompetitorunderthedifferentscenariosare:

2 2( ) (1 )( / (1 ))

'( ) '( )N I

N I

p c g p s g

F Q F Q

(1.5)

Theequality in (1.5) identifies anequalpayoff line (EPL) in the (i,j)marginal cost spacealongwhichtheexpectedpayoffsarethesameandthereforethecompetitorisindifferentbetween imitating and non‐imitating. Above this line (where s is closer to c , that is,disclosure isweak), the competitor chooses not to imitatewhile below it, he chooses toimitate.TheintersectionbetweentheEPLandthelinewherec=s(i.e.the45°line)givesthe cut‐off level of the inferred costs above which the competitor has no incentive toimitate.Inthecaseoflineardemandwithp=α‐βQ,AYshowthatthispointisgivenbythefollowingexpression:

* 2 (1 ) ( ) 1

( , , )1 (1 ) ( )

c g h gc g c

g h g

(1.6)

where h g 3 / 3 g 1 g .

So the competitorwill imitatewhen, given the expected damages implied byg, the costdisadvantageofusingthenewtechnologyovertheinferredcostoftheinnovatorissmallerthanthecostdisadvantageofusingtheexistingtechnologycomparedtotheinferredcostoftheinnovator.

Protection and disclosure stage. The innovator decides whether to use a patent orsecrecy and howmuch to disclose. The equilibrium involves three distinct regions: 1) aregionwhereaninnovatorwithhighmarginalcostsandasmallinventionpreferstopatenttheinventionandfullydisclosethetechnicalknowledge;2)aregionwheretheinnovatorwithamediumsizeinventionpreferstopatentandpartiallydisclosetheinvention;and3)aregionwheretheinnovatorwithalargeinventionchoosessecrecyandpartiallydisclosestheknowledge.

24

If the innovation is relativelysmall (c>c*), thecost reductionofsoverc is insufficient tojustify the risk of infringing a patent and therefore the competitor has no incentive toimitate.Inthiscase,theinnovatorhasastrongincentivetopatentandtodisclosefullytosignallowcostwhichhasapositiveimpactonhisexpectedprofits.AYshowthattheresultholds evenunder aweakproperty rights regime. In addition, as 1g (strongpropertyrights),thecutoffc*tendstowardzero,andtheinnovatorwillalwayspatentanddisclosefully.

Forc<c*, the competitor finds imitationattractive and therefore the innovatorno longerhasanincentivetodisclosefully.Instead,theinnovatorfacesatrade‐offbetweensignalinglow costs (and obtaining larger damage payments) and transferring knowledge to acompetitor.Inotherwords,itisworthsignalinglowcoststhroughdisclosureandtoobtaindamagesifthepatentisinfringed.Sotheinnovatorwillnowdisclosepartially.

For large innovations, however, the trade‐off to signaling via partial disclosure andpatenting becomes less attractive. The innovator has two sources of profits underimitation:revenuesfromdamagesifthepatentisinfringedandprofits.Astheinnovator’smarginal cost of production falls (i.e. the innovation gets larger), the innovator faces thesametrade‐offasinthemediuminnovationregion.Ontheonehand,theinnovatormaybeinducedtodisclosemoreknowledgetosignallowcost.Ontheother, lowermarginalcost(foragivenamountofdisclosure)impliesthatitcanproducemorethanitscompetitor.Thecompetitorfacestwooppositeforcesaswell:ontheonehand,lowermarginalcostoftheinnovatorimplieslowerprofitsforthecompetitor.Ontheother,iftheinnovatordisclosesmore,thecompetitor’smarginalcostsfall.AYshowthatinequilibrium,alargeinnovationleads the innovator to rely more strongly on the cost advantage and less on expecteddamages. For g sufficiently small and c sufficiently large, the cost‐advantage effectdominates.Inequilibrium,givinguppropertyrightssignalsalargeinnovationandpermitslessdisclosureofvaluableenablingknowledge.

To support this counterintuitive result empirically, AY point to the example of the FordMotor company in 1913, after introducing the moving assembly line process. Fordencouraged wide disclosure of this innovation, which was not patent protected, butaccordingtoHounshell(1984), thedisclosurewas insufficienttoallowfull imitation.Thebasic rationalewas that Fordwanted to signal to competitors that it had extremely lowproduction costs in order to discourage them.Although an interesting example, it is notclear that patenting was even an option in this case, as the innovation was essentiallyorganizational.ButitdoescapturethemotivationthattheAYmodelrelieson.Aswewillsee later, the general conclusions of the AY model are overturned if the innovator isthreatenedbyarivalinnovatorwhomightpatentfirst.Inthisexample,the(partial)lackofpatentability and the first‐to‐invent system that prevailed then would probably havediscouragedpatentingbyarivalanyway.

Zaby (2010)arrivesat the sameconclusionasAY, that is, important innovationsarenotpatentedwhereasthelessimportantonesare.Zabyanalyzesthepatentingdecisionofaninnovator who is aware that patenting an invention involves disclosure of knowledgewhichmayenablea competitor to imitate its invention.Thekeydifferences fromAYare

25

that:a)thereisnoasymmetricinformationonthesizeoftheinnovation;b)theinnovatorcannotchoosetheamountoftechnicalknowledgetodisclosewithapatentbecauseZabyassumesthatpatentingisalwaysassociatedwithfulldisclosurewhilesecrecyisnot;andc)competition is Bertrandwith different qualities rather than Cournot. In her asymmetricduopoly model, one firm is a successful inventor while the second firm is not but mayeventually develop the capability of making a closely related invention. The strategicinteractionsbetweenthe two firmsaremodeledasa three‐stagegamewhere in the firststage,theinnovatorhastodecidewhethertopatentornot.Inthesecondstage,bothfirmschoosetheirqualities,withfirm2restrictedtoaregionthatisnotcoveredbythepatent,and in the last stage, they compete in prices. While a patent may protect the firm’sinvention,thefirmmayruntwoadditionalrisks:first,apatentrequiresthefulldisclosureoftheprotectedinvention;second,thecompetitormaystillenterthemarketwithanon‐infringingproductasthepatentcannotcoverallpossibleproductqualities.

HegerandZaby(2013)useasimplifiedversionoftheZaby(2010)modeltolookattheroleofdisclosureindeterminingacompany’schoicebetweenpatentingandsecrecy.ThemodelisbasedonSalop’scirclemodel(Salop,1979),wherefirmschoosealocationinaverticallydifferentiated product circle. There is a single innovator that can choose to patent tobroadentheproductspaceheoccupies.Allotherfirmsdecidewhethertoenterthemarkettakingintoaccounttheinnovator’schoice.Allfirmsfacesomefixedmarketentrycoststhatdecline with more disclosure. Hence, in this model, the innovator trades off theexclusionaryeffectapatenthasoncompetitionwithitsdisclosureeffectwhichpromotesentry and hence competition. Since this trade‐off varies across companies, the modelimpliesthatpatentingpropensitiesdifferacrossfirmsbecausethedisclosureandexclusioneffectsofapatentdiffer.

3.1.2. Competitionforinnovation

Thediscussiononthechoicebetweensecrecyandpatentinguptonowhasfocusedontheinteraction between a single innovator and one or several imitators. This excludes thepossibility of simultaneousdiscovery. Lemley (2012)provides ample anecdotal evidencefor the widespread occurrence of such simultaneous discovery, i.e. a situation in whichindependent researchers produce (almost) simultaneous inventions. The examples showthatwell‐known,disruptiveinventions,suchasthesteamengine,thetelegraph,thesewingmachine,thetelephone,andthelightbulb,werenottheoutputofindividualgenius,despitewidely held beliefs. Instead, they were the outcome of cumulative research andexperimentationwheredifferentinventorsfollowedverysimilarleadsatthesametime.InallthehistoricexamplesprovidedbyLemley,thesuccessfulinventorobtainedapatentonthe invention, which in most cases produced considerable income for the inventor. Incontrast, theother inventorsworking simultaneouslyon the sameor a similar inventioncameawayempty‐handed.Lemley(2012:755) thereforeargues that “[t]hepatent isn’tacarrotsomuchasastickwithwhichtothreatentheslow.”Thatis,patentsofferincentivestoinnovatenotsomuchbecausetheyofferarewardintheformofapatent,butbecausepatentspunishthoseinventorsthatinvestinresearchbutfailtoobtainthepatent.Hence,ina settingwithsimultaneous inventions, thedecisionwhether topatentormaintainan

26

invention secret is dominated by the concern over a competitor obtaining a patent first,whichoverturnstheAYresultabove.

Kultti et al. (2006) makes the fairly obvious point that when there is a possibility ofsimultaneousinvention,evenwhensecrecyoffersstrongerprotection,firmswillprefertopatent to prevent their competitors frompatenting the same invention. They consider aduopolywhere firmscansimultaneouslydevelop thesame inventionandhenceapatentraceemerges.Theinnovatorwillbewillingtopatenttogainsomecompetitiveadvantageifthereisahighprobabilitythatthecompetitordevelopsthesameinvention.Whenthereisastrong likelihoodofsimultaneous invention,patentingtakesonadefensiverole:nowthechoice is not between patenting and secrecy, but between patenting and allowing acompetitor to patent. Since the inventor that obtained the patent always earns higherprofits,thereisastrongincentivetopatent.28Thistypeofresulthassimilaritieswiththeolder patent race models where, in a ‘winner take all race’, firms compete in research(Wright, 1983; Gilbert and Newberry, 1982). Kultti et al.’s main result that patenting ispreferred remains true even if patentingper se offers slightly less chances of protectionthansecrecy.However,thisresultonlyholdsiftheprotectionofferedbythepatentsystemis above a certain threshold: if the protection from patenting falls (say due to weakenforcement),thenatsomepointsecrecywillbepreferred.

Themainargumentofthepaperisgiveninasimplemodel.TwofirmsinvestinR&Dwhichresultseitherinaninnovation(withprobabilityq)ornot(withprobability1‐q).Supposethattheinnovationcanbeprotectedonlybysecrecy.Secrecyiseffectivewithprobability.IfonlyonefirmsucceedsinitsR&D,itearnsmonopolyprofits.Ifbothfirmssucceed,eachfirmearnsduopolyprofits / 2D Mp p< .Thefirmscanchoosetheirsuccessprobabilityby

paying 2/)( 2Rqqc . If firm1chooses 1q andfirm2chooses 2q theexpectedprofits forfirm1areequalto:

21 2 1 2 1(1 ) / 2M Dq q q q Rq (1.7)

Ifwe introduce the possibility that a firm can patent, now firm 1 has to choose how toprotectitsinnovationbeforeitlearnswhetherthecompetitorhassuccessfullyinnovated.Ithastwooptions:itcaneitherpatentorusesecrecy.Ifbothfirmsaresuccessfulandfileforapatent,eachfirmobtainsthepatentwithprobabilityonehalf.Thestrengthofthepatent

p isdefinedastheprobabilitythatapatentholdercanexcludethecompetitorfromusing

the invention. Using equation (1.7), it is easy to show that patenting is the dominantstrategy,evenifthepatentoffersweakerprotectionthansecrecy.Theintuitionisthatiftheprobability that the second inventor can innovate is large, the innovator may prefer topatenttomakesurethattheotherinventorcannotpatent.

28This incentive isalsoaffectedbythe legalsituationsurrounding“prioruserrights”.Theseare therightsgiven to the original innovator if he relied on secrecy but a subsequent imitator obtained a patent on theinnovation.Generallysuchrightsarelimited,althoughsomejurisdictionsallowthemincertaincases(notabletheU.S.,forbusinessmethods).

27

Kultti et al. (2007) extends themodel to an infinite horizon discrete timemodelwith acontinuumofinventorsandideas.Theresultsaresimilar–ifpatentstrengthisstrong,allfirmspatentandifpatentstrengthisweak,allfirmschoosesecrecy.However,thereisanintermediate range of patent strength where some firms choose to patent and otherschoosesecrecy.Theythengoontocharacterizetheoptimalpatentstrengthfromawelfareperspective. While intuitive, the model ignores the possibility that firms might torpedoeach other through strategic disclosure in form of defensive publications which wouldjeopardize thepossibility topatent the resulting invention. If such strategicdisclosure ispossible,thiswouldgiverisetostrategicinteractionintheformofprivatearrangementsinwhichfavorableexclusivebilaterallicensingagreementsarenegotiatedwhichcouldundotheincentivesfortherace(seediscussioninSection3.3below).

Mosel(2011)revisitstheresultsobtainedbyAYdiscussedabovebyextendingtheKulttietal. framework to allow for heterogeneity in the “quality” of innovations. The maindifferencesfromAYarethat(a)bothcompaniesconductR&D,(b)patentingiscostly,and(c)thereisfulldisclosureifafirmdecidestopatent.29ThismodelproducesresultsthatarethereverseofthoseinAY:smallinnovationsarenotpatented(becauseofthefilingcosts)while large innovations always are (because the benefit of patenting first relative to thecompetitoroutweighs thepatentingcosts).Thisresultseemstoaccordmuchbetterwithreality (seeSections2and4),althoughempiricallyamore importantdeterminantof thechoiceofsecrecyoverpatentingmaybetheabilitytokeeptheinventionsecret.Themodelallows such variability via a single parameter, but the parameter is held constant in thestaticanalysis.

3.1.3. Leadtimeandcomplexity

ManyofthefirmsurveysthataskabouttheuseofmethodstoappropriatethereturnstoR&D and innovation find that firms rate lead time (the first mover advantage) andcomplexityatleastasifnotmoreimportantthanpatentingforprotectingtheirinnovations.The discussion so far attributed to disclosure the decisive role in a company’s decisionwhethertoprotectaninventionthroughpatentingorsecrecy,althoughtheworkbyZaby(2010)hintedatthepossibilitythatatechnologicalleadmightplayinthedecision,andanumberofmodelsshowthatwhenthisleadislarge,theinventormayprefernottopatentandusesecrecy.Patentswillbepreferredbytheinventoronlywhenitstechnologicalleadismoderate.

Schneider (2008) studies a model similar to Zaby’s model, but where the form ofcompetition is left unspecified, in order to allow for a full range of duopoly profits. Heassumes that there is one lead innovator and one potential follower innovator. If thefollower innovates, the two products compete. The follower only tries to innovate if heexpectstomakeprofits,whichdependsonthecostofR&Dandnatureofcompetition.Thisgivesrisetothepossibilitythattheleadinnovatorchoosestokeepaninventionsecretin29Themodelalsoallows for thepossibility thatpatentapplicationsget rejected,which isassumed tobeafunctionofthe“quality”ofaninnovationandapatentrenewaldecision.Butthesefeaturesarenotcrucialforthemainresultsofthemodel.

28

order to prevent disclosure,whichwill raise the cost of the follower’s R&D. Schneider’smodel suggests that this is only rational “when the speed of discovery [of the leadinnovator] of the subsequent invention is high, relative to the competitor's” (Schneider,2008:1349).Inotherwords,theleadinnovatorhastheabilitytogenerateaseriesofnewproductsandsecrecypreventsacompetitorenteringtherace.ThisconclusionisessentiallythesameasthatinthepreviouslycitedZaby(2010).Bothreachtheconclusionthatfirmswithalargetechnologicalleadovertheircompetitorsmayprefersecrecytopatenting.

Thereisasmallliteraturethatexplorestheroleofstrategicinteractionamongimitatorsininfluencing the innovator’s incentives to patent. In this literature, an innovator canstrategically delay entry by imitators, i.e. create lead time, by providing incentives forimitators to free‐ride on each other’s imitation efforts. This creates a delay betweeninnovationandimitationevenintheabsenceofpatents.Thedecisiontopatent,therefore,depends on whether profits from secrecy exploiting the strategically created lead timeexceed those obtained from patenting. In themodel by Henry and Ruiz‐Aliseda (2012),innovatorsachieveleadtimebyinvestingintechnologythatprotectsaninvention,suchasproduct complexity. In Henry and Ponce (2011), innovators use licensing to potentialimitators strategically to create lead time. In Anton and Yao (1994), an innovator alsoexploits(thethreatof)licensingtoappropriatereturnstohisinvention.Butthemechanismis different because this model assumes imperfect information and hence the need todiscloseinformationtoattractlicensees.Wediscusseachofthesemodelsinturn.

Henry and Ruiz‐Aliseda (2012) allow an innovator to invest in making it harder forpotential competitors to imitate an innovation via reverse engineering. If imitation iscostly,imitatorsalsohavetodecidewhethertoprotecttheimitationfromcopyingbecauseiftheimitatorfailstopatent,allotherimitatorscanimitateatzerocosts.HenryandRuiz‐Aliseda (2012) show that if imitation costs are sufficiently large, imitators will delayreverse‐engineeringinthehopethatanotherimitatorpaysthecostsfirstwhichwouldthenallow them to imitate without incurring the costs associated with overcoming copy‐protection. Hence, innovators may choose secrecy over patents for technologies wheresecrecy can be accompanied by product complexity so that reverse‐engineering can bemade very costly. Henry and Ponce (2011) use a similar model of competition amongimitatorstoshowthatinnovatorscanstrategicallycreateleadtimeovertheircompetitorswithout using a patent. They analyze a set‐up in which an inventor can sell specificknowledge on an invention to potential imitators and these potential imitators have thechoicebetweencostlyimitationandacquisitionoftheknowledge.Inequilibrium,inventorschoosetoselltheirtechnologyinawaythatallowsacquiringfirmstore‐selltheknowledgetootherfirms.Asaresult,oncethefirstimitatorhasacquiredtheknowledgeandenteredthemarket, he will compete with the innovator in themarket for knowledge and driveprices for the knowledge to zero. This is nevertheless optimal for the inventor becausepotentialimitatorsdonotimmediatelyenterthemarket,butwaitinthehopethatanotherfirm enters first and drives down the price of the required knowledge. This produces asituationinwhichtheinventorenjoysatemporarymonopolypositionwithoutrecoursetoapatent.

29

TheHenry‐Poncemodel implies that themore tradableknowledge is,holding thepatentterm life constant, the more likely firms are to rely on secrecy rather than patents(assumingpatentingismoreexpensivethanmaintaininganinventionsecret).Theintuitionbehindthisisthatthecertainlengthoftheprotectiongrantedbyapatenthastooutweightherelativecostsassociatedwithpatentsbecauseinventorscanalsoreapmonopolyprofitsfromthedelayedentryofimitators.

ThemechanismsusedbyaninventortoappropriatereturnsfrominnovatinginHenryandPonce(2011)arefundamentallydifferentfromtheseminalAntonandYao(1994)model.In Anton and Yao (1994), a financially constrained inventor sells an invention to twocompanies, which invent themselves with some probability. They assume that theinvention cannot be protected by a patent and that the companies cannot observe thequalityoftheinventionexante.AntonandYaoshowthatinequilibriumtheinventorfirstfullyrevealstheinventiontothebuyerbeforesigningacontract.Althoughthebuyercoulduse the disclosed inventionwithout compensating the inventor, the buyerwill offer theinventorexpostacontractwithastrictlypositivepayoff.Thisisoptimalfromthebuyer’sperspective to keep the inventor from also selling the invention to its competitor. Themodel, therefore, shows how an inventor can obtain rents without recourse to patentprotection in a setting characterized by incomplete information due to the threat ofcompetitionamongthebuyers.Whilethesemodelsofferplausiblewaysforcompaniestoappropriate returns to innovationwithout patents, it appears, however, that in practicefirmsfindpatentsconvenientwhenconstructingknowledgecontracts(seealsoSections2and4).

3.1.4. Cumulativeorsequentialinnovation

Theabovemodelsofsecrecyandpatentingassumethatinnovationsarediscreteanduseaoneinvention(oroneinnovation)‐onepatentmodel.Theinnovationusuallytakestheformofaprocess innovationthat lowersproductioncosts.However, inreality innovationsareoften complex, involving many inventions covered by patents and also cumulative (orsequential), i.e., inventors build on the innovations of others. At a very basic level, inindustrieswhere innovation is cumulative, secrecy can lead to duplicationof efforts. Forexample,Erkal(2005)obtainsthisresultinamodelwithtwosequentialinnovations(eachoneinvolvingaracebetweentwoinnovators).Thekeyassumptionofthemodelisthatifthefirstinnovatorreliesonsecrecy,thenthesubsequentinnovatorsdonothaveasmuchknowledge,whichnotsurprisinglydoesleadtoduplicationofefforts.Thusincaseswheresequentialinnovationisimportantandthereareseveralinnovators,shearguesthatitmaybeworthwhile increasing thebreadthofpatentprotection (since thiswill encourage theuseofpatentsandassociateddisclosure).

ScotchmerandGreen(1990)werethefirsttolookatthequestionofwhethertheinventorofa “weak”or “partial” invention thatprovides information forsubsequentresearchwillchoosetopatentitorkeepitsecret,andwhetherthenoveltyrequirementshouldbeweaktoencouragedisclosureofsuchaninvention.Theyconcludethatsucharequirementwillnothelpmuchbecausethefirmcanalwayschoosenottopatent,evenifitispossible.Theyalsoshowthatthefirstinventionismorelikelytobepatentedunderafirst‐to‐filesystem

30

(whichisnowthenormaroundtheworld)thanunderafirst‐to‐inventsystem.However,neither of these systems is first best, somethingwhich they have also explored in otherwork(GreenandScotchmer1995).

Ponce (2011) lays out a model that is similar to Scotchmer and Green’s but adds theassumption thatdisclosure (patenting) createsprior artwhich increases thedifficulty ofpatentingsubsequentinventions.InPonce’smodel,thefirstinnovatortradesofftheeffectdisclosurehasonloweringtheprobabilitythattherivalinnovatorpatentswiththeeffectthat the disclosed information has in helping the rival to successfully innovate. Whendisclosurehaslittleeffectontheprobabilityofobtainingsubsequentinventions,loweringthenoveltystandardforpatentingincreasesdisclosure,whereaswhendisclosurestronglyaffectssubsequentinnovation,loweringthestandardleadstomoreuseofsecrecy.Derivingwelfare effects for the choice of patentability standard is complex, but generally followsfromtheseobservations.

Bhattarchaya and Guriev (2006) look at a different situation: instead of sequentialinvention, they consider sequential R&D, where R&D is split between companies thatperformR(esearch)andD(evelopment)separately.Inthemodel,thereisasingleinventorandtwocompaniesthatdeveloptheinventionintoamarketableinnovation.Theexistenceof two developers is a key difference from the Anton and Yao (1994) model discussedabove, where the equilibrium result has a single firm developing the invention. InBhattarchaya andGuriev’smodel, the inventor comes upwith an invention that he thenlicensestoadeveloper.Developersproduceinnovationsusingeitherthelicensedinventionorwhateverinformationabouttheinventiontheycanobtainthroughapatentpublicationor disclosure during licensing negotiations. The information on the invention aswell asowndevelopmenteffortproducewithsomelikelihoodamarketableinnovation.Ifonlyonedeveloper succeeds in obtaining an innovation, he acts as a monopolist in the productmarket,whereas ifbothdevelopers succeed, theycompeteunderBertrand.The inventorhasthechoicetopatentortomaintainhisinventionsecret.Patentingdisclosesinformationto both developers. If the invention is patented, in equilibrium an exclusive licensingcontract with one developer is signed. If the invention is protected through secrecy, inprinciplenoinformationabouttheinventionleaks.Informationisonlytransmittedbytheinventor to negotiate a licensing agreement. Moreover, if secrecy is used to protect theinvention, Bhattarchaya and Guriev (2006) assume that the inventor could license theinventiontoonedeveloperandthensubsequentlylicenseitopportunisticallytotheseconddeveloper too. Even if the second developer did not acquire a license, he could use theinformation obtained during the licensing negotiations to innovate. To avoid thisopportunistic behavior, developers optimally offer the inventor a contract that gives theinventorashareofsales.

Thiscontracthasthedrawbackthatit lowerstheoptimaleffortexertedbythedevelopersince he loses a share of his profits to the inventor. Since an inventor can move fromsecrecy to patenting but not vice versa, secrecy has to offer at least the same payoff aspatenting to be optimal. The trade‐off that emerges is that secrecy reduces informationleakagebutatthesametimetheroyaltyratereducesthedevelopers’incentivetoinvestindevelopmentwhich lowersexpectedrevenues.BhattarchayaandGurievshowthatthis is

31

the case for sufficiently large disclosure and a valuable invention. The optimal share ofrevenues paid to the inventor decreases in disclosure (lower return from opportunisticbehavior) and the valueof the invention (higher expected revenue). Since a lower shareincreases the incentives for the developer to exert effort and to produce a successfulinnovation,thepayofffortheinventorincreases.Thisresult–secrecyispreferredformorevaluableandlesscodifiableinventions‐‐isreminiscentoftheAntonandYao(2004)modeldiscussedearlier.However,themechanismsarefundamentallydifferent.

3.2. Combinedpatent‐secrecystrategy

Whilewehavedescribedthechoicebetweenpatentingandsecrecyasamutuallyexclusivedecision, the upper‐left quadrant of Table 4 suggests the possibility that formal andinformalmethodsmayinfactbecombinedtoprotectaninvention.

Arora (1997), for example, documents the early days of the organic chemical industrywhich provide an example of a situation in which firms resorted to both secrecy andpatenting to protect innovations. Arora argues that certain chemical innovations werecomposed of tacit elements, notably the specific combination of different compounds,whichwereprotectedbysecrecy,andcodifiedknowledge,i.e., individualcompoundsthatwere protected by patents. Arora argues, more generally, that knowledge based on“inductiveandempiricistprocedures”ishardtoprotectthroughpatentsbecausethistypeof knowledge is hard to codify and the corresponding claims would have to be narrowwhich would disclose a great deal of information. Hence, according to Arora, for suchinventions firms prefer to patent the codified aspects and to keep the remainder secret.Even ineasily codifiable technologies, suchas chemical compounds, thebestpracticesofproducingacompoundmaybekeptsecret,30whichsuggeststhatpatentsandsecrecycanact as complements. This also explains why companies in the pharmaceutical industrycommonlyreportrelyingheavilyonbothpatentsandsecrecy(Cohenetal.2000,Tables1and2).

Graham(2004)studiesthecasewherefirmscombinepatentingandsecrecybystagingtherevelationofinformation,keepingthecodifiedpartofaninventionsecretwhilepreservingthe option to obtain patent protection in the future. He observes that in the U.S. patentsystempriorto1999,patentapplicationsremainedsecretuntilthepatentissued.Byusinga continuation, continuation‐in‐part, or a division, assignees were thus able to keep apending patent application secret for an extended period of timewhilemaintaining theearly priority. In thisway, firmswere able to effectively combine the benefits of patentprotectionwith trade secrecy and to avoid the trade‐off between patent protection anddisclosure.Prior to1995,when the termof thepatentwas17years fromthegrantdaterather than20years fromtheapplicationdate, therewas littlecost in termsof lengthofpatent term to this strategy. Graham suggests that the combination of secrecy andpatentingthroughcontinuationwasparticularlyinterestingtofirmsthathadafirst‐mover30 However, in the US before the 2012 America Invents Act, patenteeswere required to disclose the bestmodeforpracticingthepatentedinventionaccordingtothePatentActof1952(formorediscussionseeLoveandSeaman,2012).

32

advantage innewtechnological fields inwhich the incumbentswere threatenedbyentrythatcoulddisplacetheincumbent’stechnology.However,iflead‐timeisimportant,Grahamarguesthatfirmswerelesslikelytousesecrecyandcontinuationduetothefactthatlead‐time and secrecy are substitutes. Hegde et al. (2009) show that firmsmay still combinesecrecywithpatenting to some extent because continuations still offer the possibility toalterindividualclaimstherebyeffectivelyextendingsecrecywithregardtospecificclaims.

Althoughtheavailableanecdotalevidencesuggeststhatfirmsindeedcombinesecrecywithpatenting,therearenotheoreticalmodelsthatexplorethispossibilityandtheincentivestodoso inmoredetail.Especially thepossibility torelyonpatentsandsecrecyatdifferentstagesoftheresearchanddevelopmentprocessofaninnovationmightofferaninterestingavenueforfurtherresearch.

3.3. Nopatent,nosecrecy‐‐disclosure

The final cell of Table 4 contains the case where firms choose to simply disclose aninvention, forexample in the formofadefensivepublication,withouthavingrecourse topatentprotection.

In contrast to secrecy, defensive publications still guarantee a firm’s freedom‐to‐operateandde facto secrecymay evenbemaintained as the information that is revealed canberestricted or substantially disguised. Defensive publications may be used in particularstrategically by firms to influence the state of prior art relevant to competitors’ patentapplications(Lichtmanetal.,2005;BakerandMezzetti,2005;Bar,2006).Hence,disclosingpreviously unknown information to the public can raise the inventive step threshold,jeopardizing competitors’ patent applications. Ponce (2007) considers this type ofvoluntarydisclosureinanenvironmentwithtwoinnovatorswhohavetochoosewhethertopatentornotinasequentialfashion.Thefirstinnovatorwhopatentsmayfacetheriskofbeing imitated. However, the use of secrecy is associated with two additional threats:duplication (when the second innovator develops a similar invention) and exclusion(occurringwhenanimitatorobtainsapatentforasimilarinnovation).Inthisenvironment,thefirstinnovatormayprefersecrecyiftheprotectionitcangetfromsecrecyislargerthanthe protection offered by a patent net of the patenting costs. Still, a firmwhich opts forsecrecy will want to disclose knowledge as it may stop the second innovator fromdeveloping the same innovationandpatenting it. If so, in equilibrium the first innovatordisclosesasufficientlylargeamountofinformationsuchthatthesecondinnovatorhasnoincentive to patent. In this type of equilibrium where patent protection is weak andproductmarketcompetition isnot too intense,both innovatorsopt forsecrecybut theremaybestilldisclosureoftechnicalknowledge.

Raising thepriorartbarmaynotbe theonly functionof strategicdisclosure.Gill (2008)offersamodelinwhichaninnovatorthathasaleadoveritscompetitorusesdisclosuretopersuade the competitor to leave the patent race. Gill’s model has two stages in whichinnovatorsinvest inaR&Dproject.Theleaderobtainsanintermediateresearchoutcomefirst andhas todecidewhether to invest further to turn it intoamarketable innovation.Theinnovatorcanchoosetodisclosetheintermediateresearchresults(butnottheamount

33

ofdisclosure).Thishastwoeffects inthemodel:ontheonehanddisclosureoffersusefulknowledge to the competitor, on the other it informs the follower about the leader’scommitment to the research project. When the latter effect outweighs the former,disclosuremakesthefollowerquitthepatentraceafterthefirststage.InGill’smodel,theleader only discloses if he continues investing in the R&D project because disclosure iscostly. Disclosure then occurs only when development costs in the second stage aresufficiently large. The intuition for this is simple. If development costs are high, theinnovatorhasan incentivetoabandontheprojecthimselfunlesschancestowintheracearesufficientlyhigh.Disclosurehelpsincreasingthechancesbyprovidingacrediblesignaltothefollowerofcommitmenttostaying intherace,therebypushingthefolloweroutofthe race. Perhaps more interestingly, for intermediate development costs, the modelsuggeststhatleadtimeitself issufficienttopushthefolloweroutoftheracebecausethefollowerstillbelievestheleaderinvestsinthesecondstagedevelopment.Hence,theleaderdoesnotneedtodisclosetodriveoutthefolloweralthoughtheoptiontodiscloseiscrucialforthiseffecttowork.

Anton and Yao (2008) look at the situation in which sellers waive the signing ofconfidentiality agreements when they negotiate the sale of innovation protected bysecrecy.31 This counterintuitive behavior, voluntarily foregoing to ability to enforcesecrecy, can be optimal to attract buyers who have incomplete information on theinnovation on offer. Anton and Yao (2008) show that if buyers have incompleteinformationontheinnovationonsale,waivingtradesecrecyisoptimalforsellerstoattractpotential buyers as it eliminates the threat of expost litigation for buyers if they fail toacquire the innovation. While the model offers an explanation for why innovators maychoose to disclose an invention that could be protected by trade secrecy and anondisclosureagreement, the innovatorwould strictlybenefit fromusinga registered IPright, suchas apatent, if hewereable todo so.Hence, themodel applies to a setting inwhichaninnovatorchoosesbetweenformalIPprotectionintheformoftradesecrecyandnoIPprotection.

Theopensourcesoftwaresectorprovidesanexampleofanentiresectorthatisbasedonthe absence of patenting and secrecy. However this does notmean that IP protection isforegone completely in this sector. Frequently various forms of copyright protectionaccompanied by a General Public License (GPL) are used, to ensure that the particularsoftwareinquestionremainsopen.

GambardellaandHall (2006)studythegeneralphenomenonofknowledgeproduction inan open (no patent, no secrecy) environment. They show that an equilibriumwith freesharing of knowledge and no IP protection can exist but that such an equilibrium is31 Confidentiality agreements are often considered necessary to claim misappropriation of proprietaryinformation in contract negotiations. For example, in a recentNewYork court case, Edelman v. StarwoodCapital Group, LLC (NY Slip Op. 09309), investor Edelman failed in his claim for misappropriation ofconfidential information against investor Starwood LLC despite having expressly marked the documentsprovidedtoStarwoodduringtheirfailedbusinessnegotiationsasconfidential.Thecourtpointedoutthattoobtaintradesecrecyprotectioninthiscontext,EdelmanwouldhavehadtosignaconfidentialityagreementwithStarwood.

34

unstableagainstadefectorwhochoosestoprivatizehisknowledge(thiscaseiseffectivelyan example of Olson’s (1965) Logic of CollectiveAction). They show that such defectionbecomesmoreprobable as thenumber of knowledgeproducers growsor the value of aparticular participant’s addition to knowledge rises. Therefore without some kind ofcoordination, production of the public knowledge good (science or research software ordatabase) can be sub‐optimal. The authors show that if "lead" researchers (or “lead”programmers) are able to establish a norm of contribution to the public good, a betteroutcomecanbeachieved,andthataGPLisapossiblesuchmechanism.Theygiveasmallexampleof thebreakdownofano‐IPequilibriumbasedon theobservedprivatizationofeconometricsoftwareproduction.

3.4. Thelife‐cycleoffirmsandindustry

Theuseofpatentsfortheappropriationofreturnstoinvestmentininnovationappearstoevolve as firms and industries mature. History is replete with examples of newtechnologiesthatdevelopedwithoutpatentsandwithconsiderablesharingofknowledge,suchastextilelooms(ForayandHilaire‐Perez,2001),theCleveland(UK)ironindustryofover the period 1850‐1875 (Allen, 1983), and Cornish pumping equipment (Nuvolari,2002).Tosomeextent,amodernexampleisthesoftwareindustryintheU.S.,whichreliedheavilyonsecrecyandcopyrightuntilaseriesoflegaldecisionsinthemid‐1990srenderedsoftware patentable.32 This example illustrates that after the initial growth of a newtechnologyandindustry,aconstituencyofsuccessfulfirmsdevelopsthatwishestoprotecttheir technology via patents. Such a constituency will sue to expand patentable subjectmatteriftheirtechnologyisnotalreadyincluded,andinanycasewilltransitionthesectorfromanopenknowledge‐sharingonetoonewhereIP ismoreprotected,asdescribedbyGambardellaandHall(2006).

Boldrin and Levine (2012) make a similar argument, emphasizing that firms switch tousing (that is, enforcing) patentswhen the industrymatures inwhich they operate anddemandstagnates.Thisargumentfollowsthe“industryshakeout”literaturepioneeredbyJovanovichandMacDonald(1994)andKlepper(1996).IntheJovanovichandMacDonald(1994)model,firmsinanindustrycompetetomarketanexogenousinvention.Firmsenterinto themarket to turn the invention (with some exogenously fixed) probability into aninnovation. The invention is refined at a later stage, offering firms that had innovatedbeforetheopportunitytoinnovateagain.Newfirmscankeepenteringduringthatprocessandattempttoinnovate.Sincetheinnovationlowersmarginalproductioncosts,firmsthatinnovate increase output. As a result prices fall. As output expands andprices fall, firmsthatfailedtoinnovateexit.Themodelproducesashakeoutofcompanies,i.e.,asuddenexitof a large number of firms, if price falls fast. This can occur either because innovatingcreates a large drop in marginal costs or because of a high likelihood of innovation.Klepper’s (1996)model allows firms to generate both product and process innovations.Process innovation reduces average costs. Because of increasing returns to process

32 Of course, the ultimate causal source for these eventswas not the courts, but the firms in the industry(includingthelargest,IBM)whosuedthePatentOfficewhentheirpatentapplicationswererejected.

35

innovation,largerfirmshavemoreincentivestoinvestinprocessinnovation.Thisimpliesthat earlier incumbents have a cost advantage over entrants. Entrants may still find itprofitabletoenterthemarketifprofitsfromtheirproductinnovationaresufficientlylarge.However,ascompanies’outputgrows,prices fall tosucha lowlevelthatentryceases. Inthemodel, firms carry out different types of product innovation, whichmeans as entrystops, fewer product innovations are available in themarket. Because of this aswell asincreasing returns to process innovation, process innovation expands at the expense ofproduct innovation. Themarket is left with fewer and larger companies that produce asmallerrangeandnumberofnewproducts.

Boldrin and Levine (2012) argue that industries are characterized by the absence ofpatentsduringtheinnovationphases.Followingashakeout,theremainingfirmsbegintopreferpatentsoveralternativemechanisms to secure theiroligopoly rents.Theysuggestthat patterns observed in patent litigation in the U.S. confirm this hypothesis. Notably,companiesthatseedeclineintheirrevenuesharesofamarket,suchasTexasInstruments,begintorelyonpatentlitigationtoextractrentsfromentrants.

Althoughtrue,thefactthatdecliningfirmsturntoenforcingtheirpatentsisnotnecessarilyalwaysanegativething.Inprinciple,abankruptormoney‐losingfirmnormallyattemptstocaptureasmuchsalvagevalueasitcanfromtheassetsitowns.Thefactthatfirmscandothisiftheyfailhelpsthemtoraisemoneyfrominvestorsexante,sincetheinvestorsstandtoloselessintheeventoffailurethaniftheassetsweresimplyabandoned.Totheextentthat patentable technology has salvage value and serves as an asset to the firm, it isappropriatethattheexitingfirmtrytorealizethatvalueviaenforcementorsaletoanotherentitythatisabletorealizethevalue.33

4. EmpiricalEvidence

ThesurveysreviewedinSection2gatheredalargeamountofinformationoncompanies’perceptionson the importanceof thedifferent appropriationmechanismsand their self‐reported use. The theoreticalmodels discussed in the preceding section look at specificaspects of the trade‐off that companies face in choosing a mechanism to appropriatereturns to innovation. This section discusses the empirical literature that analyzes thedeterminants of firms’ choices of the different appropriation mechanisms and thecorrespondingoutcomes.Inveryfewcases,thisliteratureattemptstorelatetheirresultstothe theoretical literature discussed earlier, but for themostpart the empirical literaturehasevolvedseparately,partlyduetolackoftheappropriatedata,aswediscussbelow.

33Muchofthecritiqueofthiskindofactivityisbasedonthefactsthatthepatentsthusenforcedmaybeoflowqualityandmayrelyonakindofholdupstrategywhereitischeaperfortheaccusedinfringertosettleratherthantotestthepatent’svalidity(e.g.,LemleyandShapiro,2005).Thatis,inprincipleitisappropriatefor a non‐producing entity to realize value from a patent portfolio, but various features of current patentadministrationandlegalenforcementmakesuchstrategiesmoreofataxoninnovationthananincentiveforinnovativeentry.

36

4.1. RegisteredIPandsecrecyassubstitutes

Theempirical literaturehasfocusedoverwhelminglyonthedeterminantsofacompany’sdecision toopt forsecrecyoraregistered IPright,mostly in the formofpatents.A largepartofthis literaturehasmostlybeenconductedusingdatafromvariousCISwhichhavetheadvantagethattheyidentifyfirmsthathaveaproductorprocessinnovationandalsocollect information on the use of alternative appropriability methods. The drawback ofusingtheCISdata,however, isthatdataareavailableonlyatthefirmlevel.Thequestionandthewaythe theoretical literaturemodels thechoiceproblem,however,concerns theinvention. Since companies commonlypursuea varietyof activities, theCISdatamaybeinherently limited in answering the research question. Another issue concerns thepatentability of a firm’s inventions. The theoretical literature assumes that both patentsandsecrecyareavailabletoafirm.If,however,aninventionisnotpatentable(e.g.softwarewith the European Patent Office), the choice problem collapses. This creates the risk ofconfoundingtheshareofpatentedinventionsinthesetofpatentableinventionswiththeshare of patented inventions. If not all inventions are patentable, this procedureunderestimatesafirm’sdecisiontopatentandbiasestheinferenceiftherearesystematicdifferencesacrossfirmsintheshareofpatentableinventionsthattheycreate.Asdiscussedbelow,thisproblem,however,canbeattenuatedatleastpartlybyconditioningthesetofcompaniesonthetypeofinnovationthattheyreport(e.g.,productinnovationthatis‘newtothemarket’).

There isalsoa fundamentalendogeneityproblemin theanalysisof thesesurveydata. InhisdiscussionoftheoriginalLevinetal.(1987)YaleIstudy,Gilbert(1987:823)noteswithregardtoasetofquestionsoncompetitionandriskofimitationthat“thereisanempiricalproblemwithsurveysoftherelationshipbetweencompetitionandR&D.IfR&Dreallydoeshaveaneffectonentryandcompetition,thenthesampleisnecessarilybiased.”Halletal.(2013)offeranotherexampleoftheendogeneityproblem.Theyfindthatpatentingfirmsratetheeffectivenessofpatentsasamechanismtoappropriatereturnssignificantlyhigherthan firms thatdonotpatent.Do firmspatentmorebecause theyconsiderpatents tobeeffectivemeanstocapitalizeonaninvention?Ordotheyratepatentsaseffectivebecausetheyhavepatented (avoiding cognitivedissonance)?Regardlessof theseproblems, someempirical regularities emerge across the various studies that use CIS data on differentcountries(andthereforeslightlydifferentinstitutionalsettings).Thestudiesidentifysomecharacteristics of the industry and the innovative firms that affect the choice betweensecrecyandpatents.

The more insightful studies in this area, however, rely on invention‐level data incombination with largely exogenous variation in the availability of IP protection acrossjurisdictions.Thethirdtypeofanalysisdoesnotuseinvention‐leveldata,butreliesonlyonexogenous variation in the availability of IP protection across jurisdictions to infer theeffectofchangesinthestrengtheningofsecrecyprotectiononresearchinputsandoutputs.This approach avoids the aggregation problem of the survey‐based studies, but reliesinstead on strong assumptions regarding confounding factors that may influence theobservedeffectofalegalchangeontheoutcomevariable.Wealsobrieflydiscussafourthapproachthatovercomesthedifficultyinobservingafirm’suseofsecrecybyusingdataon

37

court cases. The case‐level data reveals information on the companies using (andmisappropriating)tradesecretsandoftenadditionalinformationonthecharacteristicsandvalue of the inventions protected by secrecy. The problem with this type of analysis,however, is theendogenousselectioninto litigation.Wediscussthesefourapproaches inturn.

4.1.1. Survey(CIS)–basedanalysis

AsreviewedinSection2,therearealargenumberofstudiesthatusesurveydatatolearnabout the determinants of a firm’s assessment of the effectiveness of the differentmechanisms and their use. This literature does not test directly any of the theoreticalmodels.Instead,itanalyzesdeterminantsoftheuseofdifferentappropriationmechanismsbycompanies.Despite theendogeneityandmeasurementproblemsdiscussedabove, theavailability of some, albeit aggregate, information on a firm’s innovative activities,including those that are not patented, as well as some self‐reported measure on firms’reliance of the different appropriationmechanisms has proven useful to establish somerobustcorrelations.

Innovation,company,andindustrycharacteristics

Productvs.processinnovations:studiesusingsurveybaseddataconsistentlyfindthattheuseofpatentsismoreassociatedwithproductinnovationsthanwithprocessinnovations.The regression models differ slightly across studies, but product innovations areconsistently found to be positively correlated with patenting (inter alia Brouwer andKleinknecht,1999;Arundeletal.,1995,Halletal.,2013).Theseresultscanbeinterpretedinseveralways.Mostobviously,itiseasiertousesecrecyforprocessinnovationthanforproduct innovations.Product innovationsaremore likely torepresentpatentablesubjectmatter, which means the uncertainty associated with the grant process may be lower.Patents on process innovations may also reveal more information than patents onproducts, which means firms prefer to patent product innovations and keep processinnovationssecret.Finally,product innovationsmaybeonaveragemorevaluable,hencethecostsassociatedwithpatentingweighlessheavily.

Size: one of the main findings of the Yale and Carnegie‐Mellon surveys is thatappropriability strategies vary across firms of different size. Larger companies regardpatents as more important an appropriation strategy than smaller companies. ThesefindingsareconfirmedinalargenumberofstudiesusingtheCIS(interaliaBrouwerandKleinknecht,1999;Arundeletal.,1995,Halletal.,2013).Theprincipleexplanationforthisfindingisdoubtlessthatlargefirmsgenerallyfindtheuseofthepatentsystemlowercostper patent than smaller firms for fixed cost reasons. Equally Small and Medium‐sizedEnterprises (SMEs)may suffer from financial constraints (seebelow) and thereforemaydecidethatapplyingforapatentisfinanciallytooonerous.However,firmsthatspecializeinknowledgeproductionandproofofinnovativeconceptaremorelikelytobeSMEsandfor these firms patents can be quite important sincemost of their assets are knowledgeassets. In addition, some startupsmay find that having patents improves their access tofinancing(HsuandZiedonis2007,interalia).

38

R&Dintensity:notsurprisingly,patentingisassociatedwiththeR&Dperformancewithinfirms. In general, it is roughly proportional in the cross section of manufacturing firms(Bound et al. 1984), but somewhat less than proportional within firm (Hausman et al.1984). Studies that look at the choice between patenting and secrecy for protection ofinnovationgenerallyfindthatR&D‐performinginnovatingfirmsaremorelikelytooptforpatents than other innovating firms (presumably because they are more likely to havepatentableinventions,e.g.,Halletal.2013).ThenumbersinTable2forU.S. firmsclearlyconfirmthisfact.

Incremental vs. large innovations: Anton and Yao (2004) suggest that firms whichproduce large innovations should relymore on secrecy than on patents to protect theirinventions.Whilecounterintuitive,thispredictionistestedinapaperbyPajak(2009)whouses theFrenchversionof theCIS4 tomodel the choicebetweenpatentingand secrecy(binary variables of self‐reported use) where the size of innovation (i.e. whether theinnovationisnewtothemarketoronlytotheindustry)appearsamongtheindependentvariables.Hefindstheratherequivocalresultthatinonethirdofinnovativeindustries(7outof21)alargerinnovationispositivelycorrelatedwithasmallerpatent‐to‐secrecyratio,inlinewiththepredictionsofAntonandYao(2004),althoughfortheothertwo‐thirds,alargerinnovationfavorspatentingoversecrecy.Moreover,forhissampleofsmallfirmsinintermediate goods sectors, Pajak finds that firms reporting innovationsnew to the firmare more likely to use patents, whereas the same firms seem to prefer secrecy forinventionsnew to themarket.Thisempirical findingshouldbe interpretedwithcaution,however,astheestimatesaremerelycorrelations,thesamplesizeissmall(72firms)andtheshareofinnovatingsmallfirmsislessthan10percent(thatis,only7firms)

FinancialConstraints:Applyingforapatentandmanagingapatentportfolioisexpensive.Afirmnotonlyhastomeetthedirectmonetaryexpensesassociatedwiththeapplicationprocess but it also has to monitor themarket for potential infringement and take legalactionifthepatentistoserveitsintendedpurposeofexclusion.Notsurprisinglyfirmsthatreport that they are financially constrained tend to use unregistered IP methods. Asdiscussed earlier, themost important reason cited by startups for not patenting is cost(Grahametal.2010).SeealsoCordesetal. (1999),whoreportonasurveyofsmallhightechnologyfirmsdonefortheU.S.SmallBusinessAdministrationwhichfoundthatcostofapplyingandenforcementwastheleadingreasonthesefirmsdidnotgenerallyusepatents.However,Halletal. (2013) foundonlyveryweakevidence that thepresenceof financialconstraintsreducedtheprobabilityofpatenting,oncefirmssizeandsectorwerecontrolledfor.

Knowledgemanagementpractice:JensenandWebster(2009)usesurveydataonasetof785Australianfirmstounderstandtheinteractionbetweenfirms’knowledgemanagementpractices and their choice of knowledge appropriationmechanisms. They find firms thatpursuea “closed learningstyle” torelymoreonpatentsandsecrecy.Whereas firms thatbase their technological learning on a more open model that involves exchange acrossfirmsaremorelikelytorelyonleadtime,brandnames,andcontroloverthedistributionprocess. This provides additional evidence that patents and secrecy can act ascomplementary forms of knowledge appropriation mechanisms within a “closed”

39

knowledge management model that relies on the acquisition of knowledge throughlicenses,thereadingof(patent)publications,andin‐houseR&D.

Incontrast,CoshandZhang(2011)findthatUKfirmsengagedinopeninnovationpracticesaremore likely to use IP protectionmethods of all types.Hagedoorn andRidder (2012)interview and survey European firms involved in open innovation, and report that theytend touse formalcontracts togovern their interactionswithpartners,and thatover90per cent viewed patents and trade secrets as the most important mechanisms forprotectiontheirintellectualpropertyintheserelationships,instrongcontrasttotheJensenandWebsterresult.

ImpactonPerformanceandKnowledgeSpillovers

AsmallliteraturebasedonsurveydataanalyzestheimpactthatthechoiceofIPinstrumenthas on the firms’ performance. This literature is not very developed and while issuesassociated to the identification strategyareunresolved, it is still interesting to reportonsomeresults thatcanofferguidance for futureempiricalanalysis.Hanel (2008)analyzesthe use of IP protection for the Canadianmanufacturing industry, paying attention to apossibleeffectonprofits.Asafirststep,hefocusesonthepropensityofinnovativefirmstoprotect their IP. Small firms use IP protection tools less often, whereas ‘world‐first’inventorsuseeverykindofIPprotectionmorefrequentlythanotherfirms.Inthesecondstagehe focuseson the impact that theuseof IPprotectionhason the firms’profits.HefindsthatfirmswhichprotecttheirIPincreasedormaintainedtheirprofit.

Hussinger (2006) uses 626 manufacturing firms from the Mannheim Innovation Panel(1998‐2000,CISIII)toanalyzetheimpactonthepercentageofsalesofnewproductsoftheuseofpatentsandsecrecy.Thereisastrongpositivecorrelationbetweenpatentsandsaleswithnewproducts,whereasthereisnorelationshipbetweensecrecyandinnovativesales.This finding is consistent with the hypothesis that patents are preferred to secrecy forprotecting valuable inventions in themarket phase, but is not supportive of the severalmodelsthatsuggestthatsmallerratherthanlargeinventionswillbepatented.Thispaperisnoteworthyforthefactthattheauthoruseslaggedpatentholdingsasaninstrumentforthe firm’s current evaluation of patent importance, controlling to some extent for theendogeneityofthechoiceofIP.Unfortunatelysheisunableto lookatprocessinnovationduetothelackofdataonthedegreeorimportanceofthiskindofinnovation(asopposedtosimplyitspresence).

4.1.2. Invention‐levelevidence

Themainlimitationsofthesurvey‐basedliteratureisontheonehandtheendogeneityoffirms’ self‐reportedeffectivenessof the various IPmechanisms and their use andon theother the fact that data are available only at the aggregate, firm‐level. There are a fewstudiesthatovercometheselimitationsbyusinginvention‐leveldataincombinationwithexogenous differences in the legal protection of formal and informal IP over time andacrossjurisdictions.

40

Moser (2005)exploitssuchexogenousdifferences in theavailabilityofpatentprotectionacross countries in the 19th century.Moser looks at innovations presented at twoworldfairs(Londonin1851andPhiladelphiain1876).Someoftheseinnovationswerepatentedandsomewerenot,whichwaspartlyaresultofthefactthatnotallcountrieshadpatentlawsatthattime(SwitzerlandandDenmarkin1851andSwitzerlandandtheNetherlandsin1876).Herfindingssuggestthatpatentprotectionisnotcriticaltoinnovationbutitdoeshaveastrongeffectontheindustrialdistributionofinnovativeactivity.Countrieswithoutpatent protection tended to concentrate in industries where secrecy was effective (asreverse engineering was not so easy and law protected trade secrets). Textiles, foodprocessing,scientificinstrumentsandwatchmakingwereexamples;andcountriessuchasSwitzerland, which had no patent system, concentrated in these industries. In contrast,innovationsfromtheU.S.(whichhadarelativelylowcostandeffectivepatentprotectionandapatchywayofprotectingtradesecrets)concentratedinmachinery.TheNetherlandsabolisheditspatentlawsin1869andthisled,accordingtoMoser,toasubstantialincreaseininnovationscomingfromfoodprocessingwheresecrecywasimportant.Inotherwords,lack of an IP system (or aweak one) does not stop firms from innovating but can haveimplicationsforthedirectionofinnovativeactivity.

Inasimilaranalysis,Moser(2012)comparesBritishandU.S. inventionsexhibitedatfourglobalfairsinthenineteenthandtwentiethcentury.Shehasdataonover8,000inventionsexhibited at Crystal Palace World's Fair in London in 1851, Centennial Exhibition inPhiladelphiain1876,World'sColumbianExpositioninChicagoin1893,andthePanama‐PacificInternationalExpositioninSanFranciscoin1915.AlthoughboththeU.S.andBritainhaveapatentsysteminplacethroughouttheperiodstudied,patentinginBritainwasmoreexpensive(byafactorof60)thanintheU.S.Enforcementalsoappearstohavebeenmoredifficultanduncertain inBritainthantheU.S.Despitethesesubstantialdifferences inthepatentsystems,Moserfindsonlyamoderatedifferenceinpatentrates:11percentoftheexhibitedBritishand15percentof theU.S. inventionswerepatented.Mosershowsthatthe share ofBritish inventions thatwere rewarded a prize at theCrystal PalaceWorld'sFairislargeramongpatentedinventions.Thissuggeststhatmorevaluableinventionsaremore frequently patented, directly contradicting the theoretical predictionbyAnton andYao(2004).Thedataobtainedfromexhibitioncataloguesalsoreveallargecross‐sectionaldifferences across industries: amongU.S. inventions, the share of patented inventions inmanufacturingofmachineryis44percentwhereasnoneinchemicals.DuringthetimeoftheCrystalPalaceWorld'sFair, inventions inchemicals couldbeprotectedeffectivelybysecrecy whereas inventions in machinery would have been easy to copy and thereforereliedmoreheavilyonpatents.Perhapsmoreinterestingly,Mosershowsthatadecreaseinthe effectiveness of secrecy in the chemical industry (brought about by importantinventions in themid nineteenth century – themodel of benzene ring and the periodictable)resultedinasubstantialincreaseinpatenting.34Thisintra‐industryresultsupportstheexplanationthatinter‐industrydifferencesintherelianceonpatentsarelargelydrivenbytheireffectivenessinagivensector.

34Onpatentingandthe19thcenturyGermanchemicalindustry,seealsoMurmann(2003).

41

Moser’s work is noteworthy not only because of the marked differences in the type ofinventions exhibitedby inventors from jurisdictionswith andwithout patent protection,but also because she observes the combination of product‐level data and patents. Mostempiricalanalysisonlyhasfirm‐leveldatawhereusuallynodirectlinkbetweenproductsandthedifferentprotectionmechanismsexists.Thisisamajorshortcoming.Thinkonlyoflarger companies,which routinely comeupwith adiverse range of inventions.They arevery likely to use all types of protection mechanisms, although the type of protectionmechanismusedmayvary systematicallyby typeof invention.Moreover,productsoftenembody a range of inventions,whichmaybeprotectedbydifferentmechanisms.Hence,eventhecombinationofproductandpatentdatamaynotbeenoughifproductsembodyinventions protected by other means than patents, which remain unobserved to theresearcher.Thismakes the set‐up thatMoseruses forher study sovaluablebecause thechoiceofprotectionmechanismsislimitedforlargelyexogenousreasons.

Havingsaidthat,Moser(2013)emphasizesthattheevidenceobtainedfromhistoricaldatahas to be interpreted with care if the objective is to translate the findings into today’sworld.Thepatentinglandscapehasundoubtedlybecomemorecomplexinrecentyearsasthe range of technologies has expanded, especially in areas where the format of patentrightsappearstofitratherpoorly.Anexampleisthelargeincreaseinthenumberofpatentfilings inelectricalengineering,especiallytelecommunicationandcomputertechnologies,and the difficulty of delineating patents from each other, which has contributed to theemergenceofpatentthicketsinthesetechnologies(vonGraevenitzetal.,2012).

4.1.3. Evidencefromlegalchanges

Theliteratureoffersathirdapproachtotheproblemofestimatingempiricallytheimpactoftheeffectivenessofthedifferentappropriationmechanismsoncompanies’choicesandperformance.Png(2011)providesanexampleofanalysisoftheimpactofsecrecyonR&Dand the choice between patenting and secrecy without having survey information on afirm’sinnovativeactivitiesorself‐reporteduseofformalandinformalIP.HeusestheNBERCompustatPatentdataset,whichcontainsfirm‐leveldataforallpubliclytradedcompaniesintheU.S.manufacturingsector,toassesstheimpactofastrengtheningoflegalprotectionoftradesecretsthroughenactmentoftheUniformTradeSecretsAct(UTSA).Pngfindsthatenactmentof theUTSA isassociatedwithanaveragedropof2.4percent inR&Damongmanufacturingfirms.However,thefiguredisguisesimportantheterogeneityacrosssectors.Whereas the drop is evenmore pronounced for themedicinal chemicals and botanicalssector(‐4.2percent),aswellasthecomputerterminalsindustry(‐4.7percent),PngdoesnotfindanimpactinrelativelymoreR&Dintensiveindustriessuchaspharmaceuticalsandcomputercommunicationsequipment.TheresultsareinterpretedassuggestingthatownR&Dandknowledgespilloversarecomplements,i.e.,anincreaseintheuseofsecrecyleadstoadecreaseofspilloverswhichleadstoanetdecreaseinR&DgiventhecomplementarityofspilloverswithownR&Defforts.Pngalsoanalyzestheeffectofthestrengtheningoflegaltrade secret protection on firms’ patent filings, but finds overall no discernible impact.However,hefindssomeevidencethatitreducedpatentinginsectorsinwhichpatentsareeffective in protectingprocess innovations. Png interprets these findings as evidence for

42

firms filingpatentsmostly forstrategicreasons, ratherthan toappropriatereturns toaninnovation.

YoungeandMarx(2013)pursueasimilarapproachanalyzingtheeffectofastrengtheningofsecrecy, in theformofemployeenon‐competeagreements in thestateofMichigan,onfirms’marketvaluation.YoungeandMarxfindastrong,promptpositiveresponseonfirms’Tobin’sqtoalawthatmadenon‐competeagreementsenforceable.Thepositiveresponsetoastrengtheningoftradesecrecysuggeststhatcompaniesgainfromanimprovedabilitytomaintain knowledgewithin the firm.However, the effect patterns out over time. Thiseffectisstrongerinsectorsthatareknowntorelyrelativelymoreonsecrecy.YoungeandMarxinterpretevidencefromchangesinpatentingactivitiesofcompaniestosuggestthattheeventualdropinfirms’marketvaluationisduetofirms’R&Dbecomingnarrowerasaconsequenceofthelawchange.

Inanotherstudythatlooksatmarketvaluationandthestrengthoflegalprotectionoftradesecrecy,CarrandGorman(2001) lookat theeffectof theEconomicEspionageActwhichwasenactedintheU.S.in1996.ItcriminalizedthetheftoftradesecretsintheU.S.(andbyU.S.citizensworldwide),whichofferedanadditionallegaltoolfortheenforcementoftradesecrecyoverexistingfederalandstatelaw.CarrandGormanlookatchangesinthestockmarket valuation of companies involved in 11 cases filed under EEA between 1996 and2000.CarrandGormangaugethevalueofthestolentradesecretsfromcourtrecordsandinterviewswithattorneysandcompanies.Theconservativeestimatesofthesevaluesrangewidely,between$0.04and$20million,withanaverageof$5million.CarrandGormanfindanegativeresponsebystockmarketstotherevelationoftradesecrecytheftinsevenoutofthe11cases.Themagnitudeofthestockmarketlossdwarfstheallegedvalueofthestolentradesecrets.

ThethreestudiesinconjunctionprovideinterestingevidenceontheeffectoftradesecrecyonperformanceandR&D.Inlightoftheempiricaldifficultyofobservingtheuseofsecrecyasanappropriationmechanism,relyingonexogenouschangesinthelegalframeworkthatgovernssecrecyoffersawindowtostudyingtheeffectofsecrecy.Theresultssuggestthatsecrecy is an important protection mechanism of valuable inventions for companies.However,strengtheninglegalsecrecyprotectionmaycomeatacostasitmaybecorrelatedwithfewerspilloversandevenchangethetypeofresearchthatfirmsconduct.

4.1.4. Evidencefromlitigation

There is a fourth approach to overcoming the “unobservability” of companies’ use ofsecrecy. Lerner (2006) offers an overview of the entire history of civil state and federalcourtscasesinCaliforniaandMassachusetts.However,statecasesarelimitedtocasesthatwereappealed.35Thisillustratesafundamentalproblemintheanalysisofcourtdata.Thecasesthatendupincourt,andespeciallythosethatendwithajudgmentonthemeritsinthefirstinstanceandarethenappealed,areunlikelytoberepresentativeofthepopulation

35Anotherproblemwiththeanalysisoflitigationoftradesecrecycasesisthattheyaredifficulttoidentifyincourtrecordsbecausetheyusuallyfigureundermuchbroadercategories,suchascontractlaw.

43

of court cases – let alone of all disputes. Thismeans that it is difficult to draw broaderlessonsontheuseofsecrecybycompaniesfromthesecourtdata.Thebreakdownofcasesby industry in Lerner’s study reveals a lot of dispersion of cases across industries.However, most cases are in industries where patents are either not available or firmstraditionally rely on secrecy and copyright, e.g. business services and computerprogrammingrespectively.Thedataalsoshowthatsecrecymaybedifficult toenforce incourt.Inslightlylessthan40percentofcasesthatendedwithjudgment,misappropriationof trade secrets was found. Damage awards average $1.5 million, which Lerner (2006)notesisa lot lowerthanthedamagestypicallyawardedinpatentcases.Thisagreeswiththeviewthatonaveragepatentsprotectmorevaluableinventionsthansecrecy.

Almeling et al. (2010a,b) collect data on trade secrecy cases in federal as well as statecourts across theU.S. Almeling et al. face the same challenges in their data collection asLerner(2006),whichmeanstheyonlyhavedataontradesecrecycasesinstateappellatecourts.Allinalltheycollect394tradesecrecycasesinfederaldistrictcourtsand358tradesecrecy cases in state appellate courts for the period 1950‐2008 and 1995‐2009respectively.Thedatashowthattherewasstronggrowthintradesecrecycasesoverthepast two decades. In 85 (93) per cent of federal (state) court cases, the allegedmisappropriator was known to the owner of the trade secret, as he/she was either a(former) employee or business partner. Most trade secrecy cases involve secrets ontechnical information and know‐how, customer lists, and internal business information.Therearehardlyanycasesonnegativesecrets(i.e.whatdoesnotwork).SimilartoLerner(2006), Almeling et al. find awin rate of 45 and 41 per cent for federal and state casesrespectively. Itappearsthatenforcingtradesecrecyvia‐à‐visanemployeeiseasier,mostlikely because discovery (which can be regarded as an important barrier to theenforcementofsecrecy)iseasier.

Courtdataofferaninterestingopportunitytoempiricallymeasuretheuseoftradesecrecybyfirms.Thisallowstostudythecharacteristicsof(anon‐randomsetof)companiesthatrely on secrecy and the information thatwas protected by secrecy. The limited data ondamagesoffersadditionalevidenceonthevalueoftradesecrets.Howeverinterestingthesedata are, it remains unclear what they tell us about the use of trade secrets, theircharacteristicsandvalueinthepopulationoffirms.

4.2. Combinedpatent‐secrecystrategy

Thereishardlyanybroadempiricalevidenceonwhetherfirmsinfactpursuethecombinedpatent‐secrecy strategy for a single invention, what its determinants are, and what itsconsequences on performance and innovation are. This is explained by the fact that theempirical analysis of theuseof patents in combinationwith secrecy is challenging.Withfirm‐leveldataitisimpossibletodetermineexactlywhatisprotectedbywhichprotectioninstrument.Thecombineduseofpatentsandsecrecyhastobeidentifiedattheinventionorproductlevel.Notonlyaredifferentelementsofaninventionoftenprotectedbypatentsorsecrecy,patentsandsecrecymayalsobeusedtoprotectthesameelementatdifferentstages of the innovative process. Anton et al. (2006: 9) state, “[b]ecause innovations are

44

rarely composed of a monolithic piece of knowledge, a combination of patenting andsecrecyiscommon.”

Theoreticalmodels tend to focuson the invention‐leveland tackle thequestionofwhichprotection tool is most suitable for this particular invention. It has proven difficult totranslatethesetheorymodelsintoempiricalresearchbecausetheavailablefirm‐leveldatacannot distinguish whether patents and secrecy are used for one or more particularinventions.Indeed,datasurveysthatarecommonlyusedforthistypeofanalysis(liketheCIS)areunabletotellwhetherfirmsusesecrecyandpatentsforthesameinventionsandwhether they apply to the different stages of the innovation process, where patents vs.secrecymightbedifferentiallyimportant.

GrahamandHegde (2012) offer some evidenceby looking at thepossibility tomaintainpatentsunpublisheduntilgrant.TheAmericaInventor'sProtectionAct(AIPA)of1999wasdesignedtoeliminatethispossibility.Itrequiredpublicationofanypatentapplicationafter18months,butofferedanopt‐out.Ifnopatentprotectionissoughtinotherjurisdictions,theapplicantcouldstillchoosetomaintaintheapplicationunpublisheduntilgrant.GrahamandHegdelookatallpatentgrants filedwiththeUSPTObetween1996and2005tofindthatpost‐AIPA(2001‐2005),7.5percentoffilingsoptout.Thelargestshareofassigneesoptingoutisfoundincomputersandcommunicationtechnologies(ICT).Theauthorsfindno evidence for small applicants more frequently opting‐out than large applicants. Forsmallapplicants,thereisalsoapositivecorrelationbetweenpatentvalueandnotchoosingtoopt‐out.Thiswouldsuggestthatthecombinationofpatentingandsecrecyintheformofdelayedpublication isnotchosen formorevaluable inventions.TheevidencebyGrahamandHegde(2012) looksataveryspecific typeof combiningpatentingwithsecrecy.Thestudy is also subject to considerable selectionbias, because foregoingpatenting in otherjurisdictionsisverycostlyinthecaseofanyvaluableinvention.Nodoubtthisexplainsthefact thatoptingout ismost frequent in ICTtechnologies,whichareusedbyasector thatengages in patent portfolio racing (a type of strategic patenting)wheremost individualpatentsarenotthatimportant,butquantitymatters.Nevertheless,itillustratesonewayinwhich these two appropriation mechanisms, which are in principle substitutes, can beemployedincombinationforthesameinvention.

4.3. Nopatent,nosecrecy–disclosure

Despiteawell‐developedbodyoftheoreticalworkontheuseofdisclosurebycompanies,thereisalackofempiricalanalysis.Merges(2004)discussestherolethatdisclosureplaysintwospecificexamples:theMerckGeneIndexandIBM’sinvestmentinLinux.In1995thepharmaceutical company Merck created a public database in which it discloses genesequencesandmakesthempubliclyavailable.Merck’smotivationwastopreventpatentinginthisarea,whichwouldaffectMerck’sbusinessbecauseitwidelyusesgenesequencesasan input. The logic is similar in the case of IBM’s investment in Linux. The open sourceoperatingsystemoffersapublicdomainalternativetoMicrosoft’sWindowsplatform.SinceMicrosoft’s dominance in the operating system market raises input costs for IBM, itsinvestmentinLinuxoffersanopportunitytobuildanIPfreesystemwhichwouldlowerthe

45

costsofanessentialinputforIBM.Theseexamplesillustratethestrategicuseofdisclosurebycompanies.

Somewhat broader empirical evidence is provided by Henkel and Pangerl (2008), whoreportonthequalitativeevidencecollectedfrom56interviews(in2005and2006)with37largepublicly tradedmanufacturingcompanies inGermanyandpatentpractitioners.Theinterviews reveal that defensive publishing is indeed widely used. One of the mainmotivationsforpublishinginsteadofpatentingisthelowercostofpublishing.Obviously,thisargumentweighsmoreheavilyforlowervalueinventionsorforthoseinventionsthatarecomplementarytopatentedinventionsandcanthereforebeprotectedtosomeextentbythem.Still,patentpublicationsareoftenusedasvehiclefordefensivepublicationsandthe cost savings arise from not having to prosecute the patent to grant. Defensivepublicationsaremainlyseenasawaytomaintainfreedomtooperate.Toensurefreedomto operate, companies choose to disclose instead of maintaining the invention secret.Nevertheless, the interviewees indicate that there are ways of legally disclosing whilemaintaining de facto secrecy, for example by choosing obscure places (or foreignlanguages) forpublication insteadofstandardoutletssuchas IP.COM.Suchastrategy,ofcourse,contradicts the theoreticalmodels inwhichdefensivepublicationssignalcostsorraise thepriorartbar.According toHenkelandPangerl (2008), there isnoevidence fordisclosuretobeusedstrategically inpatent/R&Draces.Companiesclaimedthat theyareusuallynotsufficientlyawareoftheirpositioninaracetousedisclosurestrategically.Evenifcompaniesknewtheirpositioninsucharace,companiesstillindicatedthattheywouldnot rely on strategic disclosure for fear over encouraging entry of new competitors,contradictingtheassumptionsofsomeofthetheoreticalmodels.

4.4. TheChoiceofotherformsofIP

Unlike informalmethods of IP protection such as secrecy and lead time, the alternativeformal methods of IP protection such as trademarks and copyright are not necessarilysubstitutes for patents, but instead offer the ability to protect different aspects of aninnovation.Althoughthetheoreticalliteraturefocusesonthechoicebetweensecrecyandformal IP in the formofpatents, inreality theuseofother formsof formal IP,especiallytrademarks and copyright, is far more widespread. This section briefly reviews therelativelysparseempiricalevidenceontheuseofthesemethodsofprotection.

4.4.1. Trademarks

TrademarksareprobablythemostwidelyusedmethodofregisteredIPprotection,astheyare available to essentially any firm selling a good or service. In some cases they canrepresentanextremelyvaluableandlong‐livedbrandbutmostendupbeingoflittlevalueorhavingarelativelyshort life.Empiricalstudies into theeffectof trademarkingon firmperformancehavebeenscarce,althoughthisischangingwiththeadventofcomputerizeddataavailabilityatTrademarkOffices,notablytheOfficeforHarmonisationintheInternal

46

Market (OHIM) of the European Union.36 Most of the available studies look at theassociation of trademark ownership with firm value and typically find that the averagetrademark is valued positively but by less than the average patent, and also thattrademarksaremoreimportanttoservicesectorfirms.

Seethamraju (2003) analyzed the value of trademarks in 237 U.S. firms over 1993‐97,findingapositiverole for trademarkingonsalesandalsoonmarketvalue.Griffithsetal.(2011)usedasampleofslightlylessthan2,700largeAustralianfirmsover1989‐2002andfound that the stock of trademarks was a significant determinant of profits, but with asmallerimpactthaneitherpatentsorregistereddesigns.Theyalsofoundthatthevalueofatrademarkwasrisingovertheirdataperiod.TheirworkisconsistentwiththeearlierworkofBosworthandRogers(2001),whohadusedasampleof60Australianfirmsfrom1994‐96 and found a positive but insignificant coefficient for trademarks in themarket valueequation that also included R&D and patents. They also noted that trademarks weresomewhatmoreimportanttonon‐manufacturingfirmsthanmanufacturingfirms.

GreenhalghandRogers(2007b)analyzea largesampleofpubliclyquotedUKfirmsfrom1996‐2000, with both manufacturing and services firms being included. They look atwhether any trademark activity, and also the effects of increasing trademark intensity,impactonperformance,asmeasuredbyTobin’sq,ortheratioofmarketvaluetothebookvalue of the tangible assets. The results indicate that a firm’s stock market value ispositivelyassociatedwithtrademarkactivity(aswellaswithR&Dandpatents).Theyfindlargerdifferencesbetween firmswithandwithout trademarks in theservice sector thanformanufacturing.TheyalsofindbiggerdifferencesinTobin’sqwhentheservicesfirmisapplying for European Community trademarks, rather than just applying for UKmarks.Whenlookingatintensities(i.e.theratiooftrademarkstoassets),theyfindanincreaseinthe intensity of Community trademarks raisesmarket value for bothmanufacturing andservices,butthisrelationshipweakenedovertheirdataperiod.Sincetherewasanincreaseintrademarksduringthelate1990s,afallintheestimatedvalueofsuchactivitymightbeexpected.GreenhalghandRogers'interpretationoftheirfindingsisthat, ingeneral,trademark activity proxies for a range of other, unobservable, firm‐level characteristics,includinginnovationthatincreasebothproductivityandproductprices.

Greenhalgh and Rogers (2007) also analyze whether greater trademark intensity raisesproductivity growth. They find that higher trademark intensity has some positiveassociation with productivity growth in services, but the results are relatively weak formanufacturing firms. These results for the relationship between productivity andtrademarkswerebroadlyconsistentwiththosederivedfortheirquotedfirmsampleusingthemarket value approach, suggesting that stockmarkets are efficient in estimating thelikelybenefitsofnewintangibleassets,andthatmanagersarenotjustseekingtrademarksto followa ‘managementfad’.Evenso, themarginalreturnstoextratrademarksperfirmwere diminishing quite rapidly over the period, as indicated by exploration of the

36Veryrecently,theUSPTOhasreleasedacomprehensiveU.S.trademarkdatabase(Grahametal.2013),butitistooearlytoseeresearchusingthesedata.

47

interaction of time trends with trade mark intensity, suggesting decreasing returns tofurtherproliferationofproductvariety.

Empirical evidence on the determinants of trademarking is similarly sparse. Jensen andWebster (2004) consider the increase in trademarking in Australia from 1976 to 2002.Theyfindthattheincreasesareassociatedwitha)increasingglobalizationb)thegrowthofhousehold income c) an increase in service sector activity and d) that trademarkingappears linked to increases in product innovation and design. Rogers and Greenhalgh(2006) consider UK financial service sector firms (1996‐2000), finding thatwhile largerfirmsaccountformoretrademarks,thetrademarktoemploymentratioishigherforsmallfirms.Theyalsoinvestigatewhetherstockmarket listedfirmsandmorediversifiedfirmstrademarkmore,butfindnoroleforeitherfactor.

4.4.2. Copyright

Empirical analysis of the valueof copyright is difficult since there is (currently)no legalrequirementtoregistercreativework.37Nevertheless,thereareafewstudiesthatgeneratesome informationon theeconomic roleof copyright.Country‐level studiesprovidesomebackground. A study on the U.S. during the periodwhen copyright had to be registered(andrenewed),concludedthataround80percentofcopyrighthad littleeconomicvalue(see Landes and Posner, 2003, who looked at the 1910‐1991 period). This result isconsistentwiththegenerallyveryskewdistributionofvalueforawiderangeofinnovationmeasures(Scherer,1998).

BakerandCunningham(2009)lookataggregatequarterlycopyrightregistrationintheU.S.andCanadaduringthe1986‐2005periodandhowitrespondedtochangesinthecopyrightterm extensions and other changes to the law, finding a small positive impact of termextension.However,therearesomeproblemswiththeempiricalsetup,duetothelackofarequirement for registration unless legal enforcement is contemplated and the timing ofsuchregistration.PngandWang(2009)lookattheimpactofcopyrightextensionsontheproductionofmoviesin23OECDcountries,andfoundnostatisticallyrobustevidencethatcopyrighttermextensionwasassociatedwithhighermovieproduction.Thisresult isnotsurprising,sincethenetpresentvalueofsucha20‐yearincreaseisverylow(ifastandarddiscountrate isused).Theyalso lookedat the impactofEuropeanrevisionstocopyrightlaw in response to the EU’s Rental Directive (which arguably strengthened the rights ofmovieproducerstoreceivereturnsfromrentals)andfoundnoeffect.Lietal.(2012)usedata at the book‐level for the period 1790‐1840 to estimate the effect of copyrightextensionthroughtheBritishCopyrightActof1814onbookprices.Theauthorsshowthatthecopyrightacteffectivelyprolongedcopyrightprotectiononly fordeadauthors,whichallows themtocomparechangesdue to the termextension forbooksbydeadand livingauthors.The results indicate a largepositive effect onbookprices, that is, thepriceof abookbyadeadauthorsignificantlyincreasedinresponsetothetermextension.

37Historicallysomecountries,includingtheU.S.,requiredcopyrighttoberegisteredpriortoenforcementbutunderTRIPscountriescannotmakesucharequirement.

48

Waldfogel(2012)arguesthatitisimportanttolookatthequalityofmusiccreatedaswellasthequantitysold,asanindicatoroftheimpactonconsumersurplusfromtheriseoflowand free methods of distribution over the internet. He uses three indicators of quality:criticsratings,andtherelativetasteforrecordingsofvariousvintages(beforeandaftertheriseofNapster)asmeasuredbothbysalesandairplay.Allthreeindicatethatqualityhasslightlyincreasedratherthandecliningwiththeadventoftheinternetandthechangesindistribution ithas induced,suggestingthat increases in freeor lowercostaccesshavesofarnothadanegativeimpactontheincentivesforthecreationofmusicalrecordings.

Firm‐level studies on copyright aremore difficult. One approach is to use data on courtactions.BakerandCunningham(2006)lookattheeffectofU.S.federalcourtdecisionsthatbroadenedcopyrightonthemarketvalueoffirms.Theyfindthatanewcopyrightstatutecan raise return on equity by between0.4per cent and2.1 per cent;while a high courtdecisioncanraisereturnsby0.1percentto1.1percent.Inasimilartypeofstudy,MazehandRogers(2006)findthatplaintiffsincopyrightdisputeshavehighermarketvaluesthana peer group of similar firms. Overall, however, the empirical evidence on the value ofcopyright,especiallyatthefirm‐level,issparse.

It is probably worth emphasizing that for copyright as for other formal IP protectionmethods, there isagreatdifferencebetween itsroleasanexante incentiveandasanexpost profit generator. It is probably safe to say that there is very little evidence that theincentive toproduce creativeworks is impactedby termextensionsof thekindwehaveseen recently, but that does not mean that the firms holding the very small share ofcopyrightedworks that have a long lifetime (think of Disney Films)will not experiencemarketvalueeffectsinresponsetoextensionsoftheterm.

4.4.3. MultipleandoverlappingIPuse

Aspointedoutabove,firmstypicallyhavemorethanoneinventionand,furthermore,tendtobundledifferentIPprotectiontools(e.g.Levinetal.,1987).Infact,mostofthesurveysthat ask firms about their preferences for various IP protectionmethods find that theiranswers are correlated, implying that firms have a general taste for IPwhichmanifestsitselfasapreferenceforallthedifferentmethods.38

EvenwhenrestrictingattentiontoregisteredIPrights,theempiricalanalysisofIPbundlesremainschallengingbecausewithfirm‐leveldataitisimpossibletodetermineexactlywhatis protected by which IP protection instrument. The existing literature on the use of IPbundleshasfocusedonthequestionwhetherdifferentformsofIPactascomplementsorsubstitutes (i.e. whether the return to using one type of IP in/decreases in the use ofanother formof IP).GrahamandSomaya (2006) suggest that IPprotectionmethodsare38Inunpublishedwork,Hall(2013)showsthisforUKfirms.Aftercontrollingforfirmsize,age,exportstatus,ownership,R&D,andtwo‐digitindustry,thecorrelationsamongthepropensitytousefourdifferentkindsofformalIP(patents,trademarks,copyright,anddesignrights)rangefrom0.55to0.7,slightlyhigherthantheunconditionalcorrelationsof0.45to0.6.ThissuggestseitherthatfirmshaveheterogeneoustastesforIPingeneral,orthatthereisconsiderableheterogeneityintheimportanceofinnovationtodifferentfirms,evencontrollingforbasiccharacteristicslikesizeandindustry.

49

often complements rather than substitutes, and offer as an example computer software,where copyright, trademarks, and patents are often used together. The difficulty inexaminingtheuseofthesemethodsempiricallyisthatcopyrightisoftenunregistered,andtrademark data can be rather noisy. The authors address this problem by looking atchangesinlitigationratesforcopyrightandtrademarkswithinfirmsovertime.Theyshowthataftertheycontrol for firmsize,age,R&D, income,managerialattentionto IP, firmIPresources, and firm fixed effects, the residuals in the copyright and trademark litigationrates are correlated, suggesting complementary use of the two above and beyond theoverallIPprofileofthefirm.MillotandLlerena(2013)analyzepossiblecomplementaritiesbetweenpatentsand trademarks.They forward the theoretical idea that trademarks cancomplement patents by providing exclusivity for a product’s goodwill which was builtwhileapatentisinforce.Incontrast,patentsandtrademarksactassubstitutesaslongasapatent is in force because competition is precluded evenwithout trademark protection.The empirical evidence provides rather weak evidence for such complementarities,although thereare interestingdifferencesacross industrieswithpatentsand trademarksactingascomplements inpharmaceuticalsandsubstitutes in thecomputerandelectricalequipment industry. Nevertheless, these results have to be interpreted with cautionbecauseonlydataatthefirm‐levelisavailabletotesttheserelationshipsempirically.

5. Patentsvssecrecy:welfare

Theempiricalfactthatmanyfirmschoosetousesecrecyratherthanpatentshaspromptedvarioustheoreticalmodelsthatanalyzetheimpactofthischoiceonsocialwelfare.Inourcontext,oneoftheissuesatstakeconcernstheroleofdisclosure,whereherethisisdefinedasthefulldescriptionoftheinventioncontainedinthepatentdocument.Oneofthebasicrationales of the patent system is to encourage disclosure, since this prevents theduplicationofresearchand,oncethepatenthasexpired,allows‘thoseskilledintheart’toquickly replicate the invention. This rationale is referred to as the contract theory ofpatentsbylegalscholars,asopposedtotherewardtheory(whichfocusesonincentivestoinvent).Theroleofdisclosureincontracttheoryisveryspecific:preventduplicationandallow rapid diffusion once the patent has expired (for more detailed discussion seeAnderson,2011).

In several of the surveys mentioned previously, firms or inventors were asked for aqualitativeassessmentastohowimportantpatentswereasasourceofinformationforaparticularinvention.CockburnandHenderson’s(2003)surveydatashowsthatonlyathirdof respondents–whoaremostly IPcounsels ‐ conductapriorart searchbeforestartingnewR&Dorproductdevelopment.Thismayhavemultiplereasons.Forexample,Lemley(2008)isoneofmanywithexperienceaspractitionerstosuggestthatITandbiotechfirmsin the U.S. purposefully ignore existing patent documents in order to avoid charges ofwillful infringement and that researchers in these fields execute their research withoutconducting prior art searches. Holbrook (2006) suggests that also the 18‐month lagbetween application and publication (and possibilities to delay publication) renders the

50

information disclosed by patents largely obsolete in a range of fastmoving industries.39Also restrictions on the legitimate use of patented inventions for experimentation couldlimittheuseofpatentdisclosure.However,Ouellette(2011)providescontrastingevidencethat suggests that managers of nanotechnology firms find it useful to read patentdocuments. There is also regional variation: Cohen et al. (2002) find that JapaneseinventorsaretwiceaslikelyasU.S.inventorstoobtaininformationaboutfutureresearchdirections from patents. U.S. companies, in contrast, generally prefer other sources ofinformationoverpatents.WalshandNagaoka(2008)confirmtheselargedifferencesintheimportanceofpatentsbetweenU.S.andJapaneseinventors,withJapaneseinventorsabouttwice as likely to use patents to acquire information. In line with these studies,Gambardellaetal.(2011)findfortheirlargeinventorsurvey(around20,000inventorsinEurope,U.S.,andJapan)thatpatentsareparticularlyimportantsourcesofinformationinasmall number of technical areas, such as polymers, organic chemicals, pharmaceuticals,petrochemical and materials chemistry, and that Japanese inventors assign greaterimportance to information inpatents thaneitherEuropeanorU.S. inventors.Apart fromaskinginventorsabouttheimportanceofpatentsasasourceofinformation,Gambardellaet al. (2011)alsoask inventors toquantify the timesaved in an inventionprocesswhencomparedtoasituationinwhichtheinformationfrompatentshadnotbeenavailable.Thisoffersawayofquantifyingpotentialcost‐savingsincurredbyinventorsduetoknowledgeof thepatent literature.Time savings fromdisclosures followahighly skewdistribution,with estimated median values of 5.9 hours and mean values of 11.5 hours. There isconsiderableheterogeneityacross technical fields ‐medianvaluesrangebetween1hour(digitalcommunicationtechnology)and36hours(organicchemicals).Thusinfieldswherepatents have strong impact on appropriation such as chemicals and pharmaceuticals,disclosureeffectsalsoappeartomatterthemost.Still,eveninthosesectorscostssavingsfrompatentdisclosureimpliedbythesefiguresareatbestmodest.Itispossible,however,that disclosure through patents has other effects that are not captured by the surveyquestion, such as whether reading patent publications avoids duplication of research,eitherbynotevencommencinganewresearchprojectorbyleadingtheinventorstoabortanongoingprojectearly.But it ishard togaugehow important theseconsiderationsareempirically.Moreover,measuringtheeffectofdisclosureonlybasedonhowimportantthepatent literature is for innovation may be too narrow. If there is additional (informal)disclosure,whichisenabledbyapatent(iftheinventionhadtobekeptsecret,itcouldnotbe informally discussed), the disclosure effects could be substantially larger.Quantifyinganysuchinformaldisclosureeffectsischallenging,however.

Therearerelativelyfewtheoreticalpapersthatexaminethewelfareimplicationsofpatentsystem design in the presence of the patent‐secrecy choice. Earlier we discussed thepioneeringworkofScotchmerandGreen(1990),whofocusonthewelfareconsequencesfor subsequent invention induced by the choice of first‐to‐file priority vs. first‐to‐invent,showingthatneithersystemcanachievefirstbestdisclosureandinventionlevels.Denicolo

39Holbrook(2006)notesthat intheU.S. incontrast tograntedpatents, to infringeonapatentafter itwaspublishedbutbeforeitwasgranted,theinfringerhadtobeawareofthepublication.Thiscreatesincentivesnottoreviewpatentpublicationsbeforetheyaregranted.

51

andFranzoni(2004)studythewelfareconsequencesofprioruserrights.Usingamodelofan inventor and a followerwhomay try to imitate, they first show thatwhen there areprioruserrights(therightofthefirstinventortouseaninventionwhichithadkeptsecretevenifitislaterpatentedbyathirdparty),secondinventorswillnotpatentinequilibrium.When a patent system is optimized (that is, when patent life is optimal for welfare),introducingprioruserrightscannotimprovewelfare.Inthemorerealisticsituationwherepatentlifeisseldomoptimal,theincentivetoinnovateisincreasedbytheintroductionofprior user rights, but there is more duplication of research, and the welfare effects areunclear.TheyconcludethatprioruserrightsmaybeoptimalinhighlycompetitivesettingswithunderinvestmentinR&D,andalsothattradesecrecyisseldompreferabletopatentingfromawelfareperspective.

CugnoandOttoz(2006)presentamodel inwhich inventions thatarekeptsecretcanbelicensed.Theyassumethatinnovatorshavealreadyobtainedtheinnovationandonlyfacethedecisionwhethertopatentorkeepitsecret.CugnoandOttozalsoassumethatpatentedinventionscannotbe imitatedwithout infringement, that is,patentsprotectan inventionperfectly. In this case, the only relevant variable determining profits from patenting ispatent lifetime. In the case of secrecy, competitors can legally duplicate the invention atsomepositive cost. It is this threat of duplication that leads the innovator to license thesecret innovation to competitors to keep them from imitating. A simple comparison ofwelfareundertheassumptionthatlicensingdoesnotentailtransactioncostsrevealsthattheexpostwelfarelossassociatedwithpatentingislargerthanthatforsecrecy.Thisresultis based on a number of rather unrealistic assumptions such as that patents perfectlyprevent imitationor that thedecision topatentormaintainan inventionsecretdoesnotaffectR&Dincentives.Soitsapplicabilityisdoubtful.

6. Conclusions

Althoughnotideal,duetothelevelofaggregation(firmlevelratherthanproductlevel)andthe qualitative nature of the questions, the survey evidencewe reviewed gives a ratherconsistentoverallimpressionofIPuseanditsimportanceforfirms.Anumberofvaluablestylized facts about invention, company and industry characteristics that influence thechoice between patenting, secrecy and other formal as well as informal mechanismsemergefromthesesurveys.Certainlythemostrobustfindingisheterogeneityintheuseofpatents across industries. Patents play an important role in the pharmaceutical andchemical industry, and sometimes in the medical instrument industry and parts of themachinerysector.Wealso learnthatmost firmsconsiderpatentsasrelatively ineffectivemeans to protecting their inventions. Instead they favor a range of different informalprotectionmechanisms, above all lead time. But companies also regard secrecy asmoreeffective thanpatents.Finally, from the surveyevidence it is apparent that firms tend totreatvariouskindsofIPprotectionascomplements,inthesensethatuseofonemakesithighlyprobablytheywillusetheothers.

52

From our review of the theoretical literature, the heterogeneity of firm behavior withrespect to IP protection should come as no surprise: theory suggests that the nature ofinnovation (product vs. process and discrete vs. complex) along with the degree ofcompetitionamonginnovatorsandintheproductmarketarethekeyfactorsthatshapeafirm’spropensity touse secrecy rather thanpatents. Since these factorsalsovaryacrosstime,andacrosscountries,weshouldalsoexpecttoseethepropensitytopatentvarying.Even in patent intensive industries, secrecy can be important in protecting processinnovations. This apparent inconsistencywith theory is due to the fact thatmany firmshaveabundleofinventionsandinnovations,andtheirdifferentcharacteristicsmaycallfordifferentprotectionmechanisms.

Ourreviewalsohighlightsthatmostofthetheoreticalworkconcentratesonthechoiceofpatentsvs.secrecy.Thebinarychoicebetweentwosubstitutesisaconvenientassumptionthathelpsidentifyarangeoffactorsthatinfluencethedecisiontousepatentsorsecrecy.One of the key determinants of the choice is the disclosure required by patents and thestrategic options (the possibility to influence competitors’ behavior) that disclosurethrough patent publications confers to companies. The available survey data, however,suggest thatpatentsandsecrecyareoftenusedas complements.Moreover, the focusonpatents is too narrow as trademarks and copyright are farmorewidely used formal IPrightsthanpatents.Similarly,secrecyisnotthemosteffectiveandfrequentlyusedinformalappropriationmethod.Thenarrowfocusofthetheoreticalliteraturecreatesagapbetweenthetheoreticalmodelsandtheempiricalworkinthisarea.

Animportantlimitationinmostoftheexistingempiricalworkrelativetothetheoryistheabsence of data at the invention level. Theorymodels the choice betweenpatenting andsecrecy at the invention‐level; most empirical studies rely on firm‐level data. Sincecompanies pursue amultitude of activities, thismakes it difficult to infer from the firm‐leveldatainformationonchoicesmadeattheinvention‐level.

Fromapolicypointofview,thelackofabettertheoreticalunderstandingofthetrade‐offthat companies face when choosing between the large range formal and informal IPmethods represents a challenge. The impact of changes to the patent system or the lawgoverning trade secrecy depends on their effect on the trade‐off between the use ofdifferent appropriation mechanisms. Our review suggests that a simplistic view of thistrade‐off, with companies patenting by default patentable inventions, is misplaced. Theavailableempiricalevidencestronglysuggeststhatinmostindustriespatentsarenotseenas an effective tool to appropriate returns to innovation ‐‐ especially not in isolation.Instead, companies appear to use a combination of different appropriationmechanismseven for the same invention. For policy, an improved understanding of possibleinteractionsandoverlapsbetweenthedifferentmechanismswouldbeuseful.

Our review shows that there is ample room for further research in this area. However,furtherresearchinformofadditionalcross‐sectionalsurveyevidencemaynotbethebestway forward.Asdiscussedabove, surveydata in thisareahave somebuilt‐in limitationswhichrestricttheinsightsthatcanbederivedfromtheanalysisofthesedata.Despitethelimitations, the various robust stylized facts generated by the survey data are useful in

53

guiding further theoretical and empiricalwork. One such stylized fact is the joint use ofdifferentformsofIP,includingthejointuseofpatentsandsecrecy.Itmightbeworthwhiletorelaxtheassumptionthatpatentsandsecrecyaremutuallyexclusiveandtoconsideramorecomplexandrealisticscenarioinwhichcompaniesemploydifferentmechanismstoprotectthesameinvention.Fromanempiricalpointofview,thesmall literaturethathasusedthecombinationofinvention‐leveldatawithexogenouschangesinthelegalregimesgoverning the patent‐secrecy trade‐off offers the most insightful findings. If patentprotection is not available, there is still innovation, albeit innovation that can be moreeasily protected by informal mechanisms including secrecy. This literature, however, isbasedonthesametypeofdataandsetting:inventionsexhibitedatfairsbetweenthemid‐nineteenth and mid‐twentieth centuries. It would be useful to have these findingsconfirmedbydatafromdifferentsettings.

54

Bibliography

Allen,R.C.(1983).CollectiveInvention.JournalofEconomicBehaviorandOrganization4(1):1‐24.

Almeling,D.S.,D.W.Snyder,M.Sapoznikow,W.E.McCollum,andJ.Weader(2010a).AStatisticalAnalysisofTradeSecretLitigationinFederalCourts.GonzagaLawReview,45(2),291‐334.

Almeling,D.S.,D.W.Snyder,M.Sapoznikow,W.E.McCollum,andJ.Weader(2010b).AStatisticalAnalysisofTradeSecretLitigationinStateCourts.GonzagaLawReview,46(1),57‐101.

Anderson,J.(2011).SecretInventions.BerkeleyTechnologyLawJournal,26:917‐978.

Anton,J.,H.Greene,andD.Yao(2006).PolicyImplicationsofWeakPatentRights.InA.Jaffe,J.Lerner,andS.Stern(eds.)InnovationPolicyandtheEconomy,MITPress.

Anton,J.andD.Yao(1994).ExpropriationandInventions:AppropriableRentsintheAbsenceofPropertyRights.AmericanEconomicReview84(1):190‐209.

Anton,J.andD.Yao(2004).LittlePatentsandBigSecrets:ManagingIntellectualProperty.RANDJournalofEconomics35(1):1‐22.

Anton,J.andD.Yao(2008).AttractingSkepticalBuyers:NegotiatingforIntellectualPropertyRights,InternationalEconomicReview49(1):319‐348.

Arora,A.(1997).Patents,licensing,andmarketstructureinthechemicalindustry.ResearchPolicy,26(4‐5):391‐403.

Arrow,K.J.(1962).EconomicWelfareandtheAllocationofResourcesforInventionInR.R.Nelson(ed.),TheRateandDirectionofInventiveActivity:EconomicandSocialFactors.Princeton,NJ:PrincetonUniversityPress,609‐625.

Arundel,A.,G.vandePaal,andL.Soete(1995).InnovationStrategiesofEurope’sLargestIndustrialFirms:ResultsofthePACESurvey.DirectorateGeneralXIII,EuropeanCommission,EIMSPublication23.

Arundel,A.(2001).Therelativeeffectivenessofpatentsandsecrecyforappropriation.ResearchPolicy,30(4):611‐624.

Arundel,A.andI.Kabla(1998).Whatpercentageofinnovationsarepatented?EmpiricalestimatesforEuropeanfirms.ResearchPolicy,27:127‐141.

Baker,S.andC.Mezzetti(2005).DisclosureasaStrategyinthePatentRace.JournalofLawandEconomics,48(1):173–194.

55

Baker,M.J.andB.M.Cunningham(2006).CourtDecisionsandEquityMarkets:EstimatingtheValueofCopyrightProtection.JournalofLawandEconomics,49(2):567‐596.

Baker,M.J.andB.M.Cunningham(2009).LawandInnovationinCopyrightIndustries.ReviewofEconomicResearchonCopyrightIssues,6(1):61‐82.AvailableatSSRN:http://ssrn.com/abstract=1436160

Bar, T. (2006). Defensive Publications in an R&D Race. Journal of Economics andManagementStrategy,15(1),229‐254.

Bessen,J.andM.J.Meurer(2008).PatentFailure:HowJudges,Bureaucrats,andLawyersPutInnovatorsatRisk.Princeton,NJ:PrincetonUniversityPress.

Bhattarchaya,S.andS.Guriev(2006).Patentsvs.TradeSecrets:KnowledgeLicensingandSpillover,JournaloftheEuropeanEconomicAssociation4:1112‐1147.

Blind,K.,J.Edler,R.FrietschandU.Schmoch(2006).Motivestopatent:EmpiricalevidencefromGermany.ResearchPolicy,35(3):655‐672.

Bosworth,D.,andM.Rogers(2001).MarketValue,R&DandIntellectualProperty:AnEmpiricalAnalysisofLargeAustralianFirms,EconomicRecord77:323‐337.

Boldrin,M.andD.Levine(2012).Thecaseagainstpatents.JournalofEconomicPerspectives,27:3‐22.

Bound,J.,C.Cummins,Z.Griliches,B.H.Hall,andA.B.Jaffe(1984).WhoDoesR&DandWhoPatents?InR&D,Patents,andProductivity,Z.Griliches(ed.).Chicago:UniversityofChicagoPress,pp.21‐54.

Brouwer,E.andA.Kleinknecht(1999).InnovativeOutput,andaFirm’sPropensitytoPatent.AnexplorationofCISmicrodata.ResearchPolicy,28(6):615‐624.

Carr,C.andL.Gorman(2001).TheRevictimizationofCompaniesbytheStockMarketWhoReportTradeSecretTheftundertheEconomicEspionageAct,TheBusinessLawyer,57(1):25‐53.

Cockburn,I.andR.Henderson(2003).SurveyResultsfromthe2003IntellectualPropertyOwnersAssociationonStrategicManagementofIntellectualProperty.Availableathttp://www.ipo.org/AM/Template.cfm?Section=Past_Meetings_and_Events&Template=/CM/HTMLDisplay.cfm&ContentID=13628

Cohen,W.M.,A.Goto,A.Nagata,R.Nelson,andJ.P.Walsh(2002).R&Dspillovers,patentsandtheincentivestoinnovateinJapanandtheUnitedStates.ResearchPolicy,31:1349–1367

56

Cohen,W.M.,R.R.Nelson,andJ.Walsh(2000).ProtectingTheirIntellectualAssets:AppropriabilityConditionsandWhyU.S.ManufacturingFirmsPatent(orNot).Cambridge,MA:NBERWorkingPaperNo.7552.

ComanorW.andF.M.Scherer(1969).PatentStatisticsasaMeasureofTechnicalChange.JournalofPoliticalEconomy,77:392‐398.

Cordes,J.J.,H.R.Hertzfeld,andN.S.Vonortas(1999).Asurveyofhightechnologyfirms.Availableathttp://archive.sba.gov/advo/research/rs189tot.pdf

Cosh,A.andJ.J.Zhang(2011).OpenInnovationChoices‐WhatisBritishEnterprisedoing?ImperialCollegeandCBR,UniversityofCambridge:UKInnovationResearchCentreReport.

Cugno,F.andE.Ottoz(2006).TradeSecretvs.BroadPatent:TheRoleofLicensing.ReviewofLaw&Economics,2(2).BerkeleyElectronicPress.

Denicolò,V.andL.Franzoni(2004).Patents,Secrets,andtheFirst‐InventorDefense.JournalofEconomics&ManagementStrategy,13(3):517–538.

Encaoua,D.,andY.Lefouili(2005).ChoosingIntellectualProtection:Imitation,PatentStrengthandLicensing.InJ.Mairesse,andM.Trajtenberg(eds),Annalesd'ÉconomieetdeStatistique(SpecialissueofContributionsinMemoryofZviGriliches),79/80:241‐271.

Encaoua,D.,D.,Guellec,andC.Martinez(2006).Patentsystemsforencouraginginnovation:lessonsfromeconomicanalysis,ResearchPolicy,35:1423‐1440.

Erkal,N.(2005).Thedecisiontopatent,cumulativeinnovation,andoptimalpolicy,InternationalJournalofIndustrialOrganization23:535–562.

Foray,D.,andL.Hilaire‐Perez(2001).TheEconomicsofOpenTechnology:CollectiveOrganizationandIndividualClaimsinthe"FabriqueLyonnaise"DuringtheOldRegime.InC.Antonelli,D.Foray,B.H.HallandW.E.Steinmueller(eds.),EssaysinHonorofPaulA.David.Cheltenham,UK:EdwardElgar.

Friedman,D.,W.Landes,R.A.Posner(1991).SomeEconomicsofTradeSecretLaw,JournalofEconomicPerspectives5(1):61‐72.

Gambardella,A.,andB.H.Hall(2006).ProprietaryVersusPublicDomainLicensingOfSoftwareAndResearchProducts,ResearchPolicy35(6):875‐892.

GambardellaA.,D.Harhoff,andS.Nagaoka.(2011).TheSocialValueofPatentDisclosure.Munich,Germany:LMUMunich,manuscript.

57

Gilbert,R.(1987).CommentsandDiscussionof“AppropriatingtheReturnsfromIndustrialResearchandDevelopment”(Levin,R.C.,A.K.Klevorick,R.R.Nelson,andS.G.Winter).BrookingsPapersonEconomicActivity3:821‐824.

Gilbert,R.andD.Newbery(1982).PreemptivePatentingandthePersistenceofMonopoly,AmericanEconomicReview72:514‐526.

Gill,D.(2008).StrategicDisclosureofIntermediateResearchResults.JournalofEconomicsandManagementStrategy,17(3),733‐758.

Graham,S.J.H.(2004).HidinginthePatent’sShadow:Firms’UsesofSecrecytoCaptureValuefromNewDiscoveries.GeorgiaInstituteofTechnologyTI:GERWorkingPaperSeries.Availableathttp://smartech.gatech.edu/xmlui/handle/1853/10725

Graham,S.J.H.,G.Hancock,A.Marco,andA.F.Myers(2013).TheUSPTOTrademarkCaseFilesDataset:Descriptions,Lessons,andInsights.Alexandria,VA:U.S.PatentandTrademarkOffice.Availableathttp://ssrn.com/abstract=2188621

Graham,S.J.H.andT.Sichelman(2008).WhyDoStart‐UpsPatent?BerkeleyTechnologyLawJournal,23(3):1063‐1097.

Graham,S.J.H.,andD.Somaya(2006).VermeersandRembrandtsinthesameattic:Complementaritybetweencopyrightandtrademarkleveragingstrategiesinsoftware.Availableathttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=887484

Graham,S.J.H.,andD.Hedge(2012).Doinventorsvaluesecrecyinpatenting?EvidencefromtheAmericanInventor’sProtectionActof1999.Availableathttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=2170555

Graham,S.J.H.,R.Merges,P.Samuelson,andT.Sichelman(2010):HighTechnologyEntrepreneursandthePatentSystem:Resultsofthe2008BerkeleyPatentSurvey.BerkeleyTechnologyLawJournal,24(4):1255‐1328.

Green,J.andS.Scotchmer(1995).OntheDivisionofProfitinSequentialInnovation.RANDJournalofEconomics,26:20‐33.

Greenhalgh,C.A.,andM.Rogers(2007a).TheValueofIntellectualPropertytoFirmsandSociety.OxfordReviewofEconomicPolicy23(4):541‐67.

Greenhalgh,C.A.andM.Rogers(2007b).TradeMarksandPerformanceinUKFirms:EvidenceofSchumpeterianCompetitionthroughInnovation.OxfordUniversity,EconomicsDept.Availableathttp://www.economics.ox.ac.uk/Research/wp/pdf/paper300.pdf

Griffiths,W.,P.Jensen,andE.Webster.(2011).WhatCreatesAbnormalProfits?ScottishJournalofPoliticalEconomy,58(3):323‐346.

58

Griliches,Zvi(1990).PatentStatisticsasEconomicIndicators:ASurvey.JournalofEconomicLiterature28:1661‐707.

Hagedoorn,J.andA.‐K.Ridder(2012).Openinnovation,contracts,andintellectualpropertyrights:anexploratoryempiricalstudy.Maastricht,NL:UNU‐MERITWorkingPaperNo.2012‐025.

Hall,B.H.(2013).UnpublishedworkfortheUKDepartmentofBusiness,Innovation,andSkills,ManufacturingForesightProject.

Hall,B.H.,C.Helmers,M.Rogers,andV.Sena(2013).Theimportance(ornot)ofpatentstoUKfirms,OxfordEconomicPapers,forthcoming.Availableathttp://emlab.berkeley.edu/~bhhall/papers/HHRS12_UK_patenting.pdf

Hanel,P.(2008).TheuseofintellectualpropertyrightsandinnovationbymanufacturingfirmsinCanada.EconomicsofInnovationandNewTechnology,17(4):285‐309.

Harabi,N.(1995).Appropriabilityoftechnicalinnovations:anempiricalanalysis,ResearchPolicy24:981‐992.

Hausman,J.A.,B.H.HallandZ.Griliches(1984).EconometricModelsforCountDatawithanApplicationtothePatents‐R&DRelationship,Econometrica52:909‐937.

Heger,D.andA.Zaby(2013).TheHeterogeneousCostsofDisclosureandthePropensitytoPatent,OxfordEconomicPapers,forthcoming.

Hegde,D.,D.C.Mowery,andS.Graham(2009).Pioneeringinventorsorthicket‐builders:WhichU.S.firmsusecontinuationsinpatenting?ManagementScience,55(7):1214‐1226.

Helmers,C.andL.McDonagh(2012).PatentLitigationintheUK.LSELegalStudiesWorkingPaperNo.12/2012.AvailableatSSRN:http://ssrn.com/abstract=2154939

Henkel,J.,andS.M.Lernbecher(neéPangerl)(2008).DefensivePublishing‐AnEmpiricalStudy.Availableathttp://ssrn.com/paper=981444

Henry,E.,andF.Ruiz‐Aliseda(2012).InnovationBeyondPatents:TechnologicalComplexityasaProtectionagainstImitation,CEPRDiscussionPapers8870.

Henry,E.,andC.J.Ponce(2011).WaitingtoImitate:OntheDynamicPricingofKnowledge.JournalofPoliticalEconomy119(5):959‐981.

Holbrook,T.R.(2006).PossessioninPatentLaw,59SMULawReview123.

Horstmann,I.,G.M.MacDonald,andA.Slivinski(1985).PatentsasInformationTransferMechanisms:ToPatentor(Maybe)NottoPatent.JournalofPoliticalEconomy,93(5):837‐858.

59

Hounshell,D.A.(1984).FromAmericanSystemtoMassProduction1800‐1932:TheDevelopmentofManufacturingTechnologyintheUnitedStates.Baltimore:JohnsHopkinsUniversityPress.

HsuD.H.,andR.H.Ziedonis.(2008).PatentsasQualitySignalsforEntrepreneurialVentures.AcademyofManagementBestPaperProceedings.Availableathttp://www.management.wharton.upenn.edu/hsu/inc/doc/papers/david‐hsu‐signals.pdf

Hussinger,K.(2006).Issilencegolden?Patentsversussecrecyatthefirmlevel.EconomicsofInnovationandNewTechnology,15(8):735‐752.

Hyytinen,A.,andM.Pajarinen(2002).Financingoftechnology‐intensivesmallbusinesses:someevidenceontheuniquenessoftheICTsector.InformationEconomicsandPolicy,17(1):115‐132.

Hyytinen,A.,andM.Pajarinen(2003).ExternalFinance,FirmGrowthandtheBenefitsofInformationDisclosure:EvidencefromFinland(Revised),ResearchInstituteoftheFinnishEconomyDiscussionPaper805.

Jensen,P.H.,andE.Webster(2004).PatternsofTrademarkinginAustralia,IPRIAWorkingPaperNo.03/04.(UniversityofMelbourne,Melbourne,Australia).AustralianJournalofIntellectualProperty,15(2):112‐126.

Jensen,P.H.,andE.Webster(2009).KnowledgeManagement:DoesCaptureImpedeCreation?IndustrialandCorporateChange,18:701‐727

Jovanovic,B.andG.M.MacDonald(1994).TheLifeCycleofaCompetitiveIndustry.JournalofPoliticalEconomy,102:322‐347.

Jung,T.andJ.P.Walsh(2010).WhatDrivesStrategicPatenting?EvidencefromtheGeorgiaTechInventorSurvey.ConferencepresentationtotheAOM,MontrealandKoreanInsituteofScienceandTechnology.

Klepper,S.(1996).Entry,Exit,Growth,andInnovationovertheProductLifeCycle.AmericanEconomicReview,86:562‐583.

Kultti,K.,T.Takalo,andJ.Toikka(2006).SimultaneousModelofInnovation,Secrecy,andPatentPolicy.AmericanEconomicReview,96(2):82‐86.

Kultti,K.,T.Takalo,andJ.Toikka(2007).SecrecyversusPatenting.RANDJournalofEconomics,38(1):22‐42.

Landes,W.andR.Posner(2003).TheEconomicStructureofIntellectualPropertyLaw.Boston,MA:Belknap/Harvard.

60

Lanjouw,J.O.,A.Pakes,andJ.Putnam(1998).HowtoCountPatentsandValueIntellectualProperty:UsesofPatentRenewalandApplicationData.JournalofIndustrialEconomicsXLVI(4):405‐33.

Leiponen,A.andJ.Byma(2009).Ifyoucannotblock,youbetterrun:Smallfirms,cooperativeinnovation,andappropriationstrategies.ResearchPolicy,38:1478–1488.

Lemley,M.(2008).IgnoringPatents.MichiganStateLawReview,19‐34.

Lemley,M.(2012).TheMythoftheSoleInventor.MichiganStateLawReview,709‐760.

Lemley,M.andC.Shapiro(2005).ProbabilisticPatents.JournalofEconomicPerspectives,19(2):75‐98.

Lerner,J.(2006).Usinglitigationtounderstandtradesecrets:apreliminaryexploration.Availableathttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=922520

Levin,R.C.,A.K.Klevorick,R.R.Nelson,andS.G.Winter(1987).AppropriatingtheReturnsfromIndustrialResearchandDevelopment.BrookingsPapersonEconomicActivity,3:783‐831.

Li,X.,M.MacGarvie,andP.Moser(2012).DeadPoets'Property‐TheCopyrightActof1814andthePriceofBooksintheRomanticPeriod.Availableathttp://dx.doi.org/10.2139/ssrn.2170447

Lichtman,D.,S.Baker,andK.Kraus(2000).StrategicDisclosureinthePatentSystem.VanderbiltLawReview,53,2175–2217.

Love,B.,andC.B.Seaman(2012).BestModeTradeSecrets.YaleJournalofLaw&Technology,15,1‐17.

Mansfield,E.(1986).Patentsandinnovation:Anempiricalstudy.ManagementScience,32(2):173‐181.

Mansfield,E.,M.Schwartz,andS.Wagner(1981).ImitationCostsandPatents:AnEmpiricalStudy.EconomicJournal,91:907‐918.

Mazeh,Y.andM.Rogers(2006).Theeconomicsignificanceandextentofcopyrightcases:ananalysisoflargeUKfirms.IntellectualPropertyQuarterly,4:404‐420.

Merges,R.(2004).ANewDynamisminthePublicDomain,UniversityofChicagoLawReview,71:183‐203.

Millot,V.andP.Llerena(2013).AreTrademarksandPatentsComplementaryorSubstituteProtectionsforInnovation.BETAWorkingPaper2013‐01.

61

Mosel,M.(2011).Bigpatents,smallsecrets:howfirmsprotectinventionswhenR&Doutcomeisheterogeneous.BavarianGraduatePrograminEconomics(BGPE)WorkingPaperNo.105.

Moser,P.(2005).HowDoPatentLawsInfluenceInnovation?EvidencefromNineteenth‐CenturyWorld’sFairs.AmericanEconomicReview,95(4):1214‐1236.

MoserP.(2012).InnovationWithoutPatents‐EvidencefromtheWorldFairs.ForthcomingintheJournalofLawandEconomics,55(1):43‐74.

MoserP.(2013).PatentsandInnovation:EvidencefromEconomicHistory.JournalofEconomicPerspectives,27(1):23‐44.

Mowery,D.C.,R.R.Nelson,B.N.Sampat,andA.A.Ziedonis(2004).IvoryTowerandIndustrialInnovation:University‐IndustryTechnologyTransferbeforeandaftertheBayh‐DoleAct.Stanford,CA:StanfordUniversityPress.

MurmannJ.P.(2003).KnowledgeandCompetitiveAdvantage–TheCoevolutionofFirms,TechnologiesandNationalInstitutions.Cambridge,UK:CambridgeUniversityPress

NagaokaS.,andJ.P.Walsh.(2009).CommercializationandotherusesofpatentsinJapanandtheUS:MajorfindingsfromtheRIETI‐GeorgiaTechinventorsurvey.Tokyo,Japan:RIETIDiscussionPaper.

Nuvolari,A.(2004).CollectiveinventionduringtheBritishIndustrialRevolution:thecaseoftheCornishpumpingengine.CambridgeJournalofEconomics,28(3):347‐363.

Olson,M.,(1971)[1965].TheLogicofCollectiveAction:PublicGoodsandtheTheoryofGroups(Revisededition).Cambridge,MA:HarvardUniversityPress.

Ouellette,L.L.(2011).DoPatentsDiscloseUsefulInformation?NewHaven,CT:YaleLawSchool.

Pajak,S.(2010).DoFirmsRelyonBigSecrets?AnAnalysisofIPProtectionStrategieswiththeCIS4Survey.Availableathttp://papers.ssrn.com/sol3/Papers.cfm?abstract_id=1538980

Pitkethly,R.H.(2001).IntellectualpropertystrategyinJapaneseandUKcompanes:patentlicensingdecisionsandlearningopportunities.ResearchPolicy,30:425‐442.

Png,I.P.L.(2011).LawandInnovation:EvidencefromtheUniformTradeSecretsAct.NationalUniversityofSingapore.Availableathttp://www.comp.nus.edu.sg/~ipng/research/2011_secrecy.pdf

Png,I.P.L.andQ.Wang(2006).CopyrightLawandtheSupplyofCreativeWork:EvidencefromtheMovies.NationalUniversityofSingapore.Availableathttp://www.comp.nus.edu.sg/~ipng/research/copyrt.pdf

62

Ponce,C.(2007).MoreSecrecy...MoreKnowledgeDisclosure?OnDisclosureOutsideofPatents.UniversidadCarlosIIIdeMadrid,DepartamentodeEconomía,EconomicSeries41,WorkingPaper07‐72.

Ponce,C.(2011).KnowledgeDisclosureasIntellectualPropertyRightsProtection.JournalofEconomicBehaviorandOrganization,80(3)418‐434.

Rogers,M.andC.Greenhalgh(2006).UseofIPbytheUKFinancialServicesSector.InThemanagementofintellectualproperty(D.BosworthandE.Webster,eds.).Cheltenham,UK:EdwardElgar.

Rogers,M.,C.Greenhalgh,andC.Helmers(2007).TheassociationbetweentheuseofIPbyUKSMEsin2001andsubsequentperformancein2002to2004.ReportforUKIntellectualPropertyOffice,http://www.ipo.gov.uk/ipresearch‐association‐200710.pdf

Romer,P.M.(1990).EndogenousTechnologicalChange.JournalofPoliticalEconomy,98:S71‐S102.

Salop,S.(1979).Monopolisticcompetitionwithoutsidegoods,TheBellJournalofEconomics,10(1):141–156.

Samuelson,P.,andS.Scotchmer(2002).Thelawandeconomicsofreverseengineering.TheYaleLawJournal,111(7):1575‐1663.

SchererF.M.(1998).Thesizedistributionofprofitsfrominnovation.Annalesd'EconomieetdeStatistique,49/50:495‐516

Schmookler,J.(1966).InventionandEconomicGrowth.Cambridge,Mass.:HarvardUniversityPress.

Schneider,C.(2008).Fencesandcompetitioninpatentraces.InternationalJournalofIndustrialOrganization,26(6):1348‐1364.

Scotchmer,S.andJ.Green(1990).NoveltyandDisclosureinPatentLaw.RANDJournalofEconomics,21(1):131‐146.

Seethamraju,C.(2003).TheValueRelevanceofTrademarks.InIntangibleAssets:Values,Measures,andRisks,J.HandandB.Lev(eds.).Oxford,UK:OxfordUniversityPress.

U.S.NationalScienceFoundation(2013).BusinessUseofIntellectualPropertyProtectionDocumentedinNSFSurvey.NSFInfobrief12‐307,NationalCenterforScienceandEngineeringStatistics.

VonGraevenitz,G.,S.Wagner,andD.Harhoff(2011).Howtomeasurepatentthickets:Anovelapproach.EconomicsLetters,111(1):6‐9.

63

Waldfogel,J.(2012).CopyrightProtection,TechnologicalChange,andtheQualityofNewProducts:EvidencefromRecordedMusicsinceNapster.JournalofLawandEconomics,55(4):715‐740.

Walsh,J.andS.Nagaoka(2008).How“open”isinnovationintheUSandJapan?:EvidencefromtheRIETI‐GeorgiaTechinventorsurvey.Tokyo,Japan:RIETIDiscussionPapers.

Wright,B.(1983).TheEconomicsofInventionIncentives:Patents,PrizesandResearchContracts.AmericanEconomicReview,73(4):691‐707.

Younge,K.,andM.Marx(2013).Themarketvalueofsecrecy:evidencefromanaturalexperiement.Availableathttp://www.funginstitute.berkeley.edu/sites/default/files/Market%20Value%20of%20Secrecy.pdf

Zaby,A.(2010).Losingthelead:thepatentingdecisioninthelightofthedisclosurerequirement.EconomicsofInnovationandNewTechnology,19(2):147–164.

Zhao,M.(2006).ConductingR&DinCountrieswithWeakIntellectualPropertyRightsProtection.ManagementScience,52(8):1185‐1199.

the choice between formal and informal intellectual property: a literature review · 2013-05-30 ·...

Documents