Weak Property Rights and Holdup in R&D - HBS … · Liebeskind, Felipe Zurita, two anonymous referees, the coeditor, and various colleagues for helpful suggestions, and to participants

Weak Property Rights and Holdup in RampD

Bharat N AnandHarvard Business School Boston MA 02163

banandhbsedu

Alexander GaletovicUniversidad de Chile Santiago Chile

agaletodiiuchilecl

We study how the sequence of nancing of RampD varies according to theease with which property rights over knowledge can be dened There aretwo nanciers a venture capitalist (VC) and a corporation The knowledgeacquired in costly research becomes embodied in the researcherrsquos human cap-ital and she may hold up the nancier and walk away with the project todevelop it elsewhere The main results are (1) When property rights arestrong research is always funded by the VC development is performed ef-ciently and breakaways from the VC to the corporation are observed in equi-librium (2) When property rights are weak projects may be nanced by theVC or the corporation or may remain unfunded (3) When property rightsare weak no breakaways occur in equilibrium local spillovers and strongproduct market competition increase the likelihood that research projects willget funding (4) The equilibrium sequence of RampD nance need not be rst-best efcient (5) In equilibrium and controlling for the strength of propertyrights VCs nance projects that are more protable on average

1 Introduction

When looking at the nancing of entrepreneurial ventures in theUnited States one confronts a few outstanding puzzles First despitethe prevalent view that venture capital is a better mode of governanceconsiderable RampD activity occurs within corporations Second thereis some evidence that venture-capital nancing has on average gen-erated superior returns and accounted for a disproportionate share of

We are grateful to Hank Chesbrough Rob Gertner Ulrich Hege Tom Hellmann JuliaLiebeskind Felipe Zurita two anonymous referees the coeditor and various colleaguesfor helpful suggestions and to participants at the 1998 American Economic Associa-tion meetings (Chicago) the Sloan Conference on Financing Innovation at ColumbiaUniversity Law School and the CEPR Conference on Venture Capital in Milano Partof this work was completed while Anand was at the Yale School of ManagementGaletovic gratefully acknowledges the nancial support of Fondecyt Fundacion Andesthe Mellon Foundation and the Hewlett Foundation

copy 2000 Massachusetts Institute of TechnologyJournal of Economics amp Management Strategy Volume 9 Number 4 Winter 2000 615ndash642

616 Journal of Economics amp Management Strategy

mega-successes Nevertheless persistent differences in performanceare surprising particularly since the robust venture-capital activityover the past two decades has attracted so much attention and thenature of venture-capital nance been dissected so thoroughly (seefor example Gompers and Lerner 1999) Indeed there is some evi-dence that when corporations mimic venture capitalists (VCs) theydo rather well (see Gompers and Lerner 1998) further amplifying thequestion why more corporate intrapreneurship is not organized simi-larly Third while the study of IPOs as an exit mechanism for venturecapitalists has attracted considerable attention more than one-third ofall VC exits are in the form of corporate acquisitions or trade salesThis raises the question when should one or the other form of exit beobserved Then a fourth question arises why is VC-backed researchfollowed by corporate acquisitions so common while the reversemdashcorporate-nanced research followed by independent developmentmdashis not Asset sales by corporations are certainly common so what isdifferent about RampD projects This paper uses an equilibriumapproach to study the sequence of nancing of RampD and analyzesthese puzzles within a unied framework

To explain these facts one must determine what is special aboutRampD We focus on two appropriability problems that are endemic toRampD1 First it is often the case that the knowledge acquired throughcostly research becomes embodied in the human capital of theresearcher not necessarily in the rm for which the research is per-formed2 In many cases property rights over this knowledge are weakbecause it is nonveriable Weak property rights enable the researcherto hold up the nancier by threatening to walk away with the project3

1 Theoretical papers that focus on appropriability problems include Anton and Yao(1994 1995) and Bhattacharya and Ritter (1983) Empirical papers include Cohen et al(1996) Levin et al (1987) and Anand and Khanna (2000)

2 As the The Economist notes ldquo the term lsquotechnology transferrsquo is something ofan oxymoron Real innovations do not move from laboratory to shopoor as patentsresearch reports or even working prototypes To stand any chance of success they haveto be transferred as concepts embedded in peoplersquos heads The one time an innovation(laser imaging) made it successfully out of (Xerox) PARC to become a multi-billiondollar business for Xerox it was because the person who championed it in the labora-tory Robert Adams moved with it and drove the innovation hard through engineeringmanufacturing marketing and salesrdquo (ldquoAdopting Orphansrdquo February 20 1999)

3 The importance of the threat of holdup can be inferred from the examples ofentrepreneurial breakaway For example ldquo[Xerox] PARC is famous for having pioneeredideas (including a superior personal computer the facsimile machine the Ethernetand the laser printer) that made fortunes for many of its Silicon Valley neighboursbut little for itselfrdquo (ldquoAdopting Orphansrdquo The Economist February 20 1999) see alsoLerner (1995a) and ldquoXerox Wonrsquot Duplicate Past Errorsrdquo Business Week September 291997) Elsewhere The Economist notes that ldquo people at Bell Labs have still not for-gotten how half a century ago William Shockley took the transistor idea which heand his colleagues had invented at Murray Hill to Palo Alto in California and started

Weak Property Rights and Holdup in RampD 617

Moreover the usual solution of allocating property rights over the out-put of research may not be feasible since knowledge itself isnonveriable and is embodied in the researcher

Second corporations and VCs are different nanciers speci-cally a corporation cannot replicate the control rights and cash-owallocations granted by a VC because many of their activities are jointor nonseparable Joint activities enable the corporation to inate costsand shift revenues across activities making surpluses nonveriableThis limits the ability of a corporation to commit ex ante to share prof-its with the researcher4 By contrast VCs nance standalone projectsand therefore can commit to share prots

Starting from these two premises we construct a model with twophases research and development The output of the research phaseis knowledge which becomes embodied in a researcher and is neces-sary to develop the the idea into a marketable innovation Each phasecan be nanced by either a VC or a corporation and the researchercan switch nanciers after research Both nanciers are equally ableto screen monitor and add value to a project in the research phaseNevertheless they differ on three accounts rst the VC can com-mit ex ante to any arbitrary surplus-sharing rule but the corporationcannot second the corporation can take advantage of local spilloversto learn about the project and develop the innovation without theresearcherrsquos participation but the VC cannot third development maybe more efcient with one or the other nancier We parametrizethe strength of property rights over knowledge as the probabilitythat the nancier can establish its rights over the innovationrsquos cashow should the researcher switch nanciers after the research phaseWe then study how the sequence of nancing of RampD endogenouslydepends on the strength of property rights over the knowledge cre-ated in research

We obtain the following results First when property rights arestrong research is always funded by the VC and development is per-formed efciently Second when property rights are weak either thecorporation or the VC nances both research and development or theproject gets no funding Thus researchers never break away whenproperty rights are weak breakaways from the VC to the corporationare observed in equilibrium but only when property rights are strongThird when property rights are weak the equilibrium sequence of

a company (Fairchild Semiconductor) that eventually became Intel As president ofBell Labsrsquo New Ventures Group Thomas Uhlman is out to see that no more Intels areallowed to escaperdquo (The Economist ibid February 20 1999)

4 Pakes and Nitzan (1983) also analyze the consequences of researcher mobilityNevertheless they assume that contracts are complete so that the corporation can com-mit ex ante to share the projectrsquos surplus


nancing of RampD need not be rst-best efcient Fourth when prop-erty rights are weak local spillovers (ie knowledge that spills over tothe corporation) and strong product-market competition by the corpo-ration in case of a spinoff increase the likelihood that research projectswill get funding This second-best result suggests that different appro-priability problems may neutralize each other instead of adding upLast we show that in equilibrium and controlling for the strengthof property rights VCs nance projects that are on average moreprotable

Several recent studies examine on the one hand the range anddifculties of internal corporate venturing activities (see for exampleKanter 1989 Block and Macmillan 1995 Lerner 1995b) On the otherhand others have analyzed the structure and role of venture-capitalnancing (see for example Admati and Peiderer 1994 Amit et al1990 Barry 1990 Gompers 1995 Gompers and Lerner 1996 Lerner1994 Neher 1995 Sahlman 1990) We explicitly model the choice ofnancier and study it in an equilibrium framework

Several recent papers study the nancing of innovation theoreti-cally In an incomplete-contracts framework Aghion and Tirole (1994)assume that the output of research cannot be described ex ante Theyshow that incentives to make nonveriable investments and exerteffort can be provided by choosing the allocation of property rightsover the nal innovation Like them we assume that nancing con-tracts are incomplete and that knowledge becomes embodied in theresearcher Nevertheless in our model the strength of property rightsover knowledge is an attribute of the project and cannot be altered5

Since inefciencies cannot be mitigated by optimally choosing theallocation of property rights the question of whether the project willthen be nanced remains In addition we go beyond Aghion andTirole by allowing for an additional nancier (a VC) and explicitlydistinguishing between research and development This enables theresearcher to switch nanciers after acquiring knowledge and intro-duces a holdup problem forcing us to examine how the incentiveof the researcher not to cheat ex post is preserved The distinctionbetween research and development phases allows us to characterizethe equilibrium sequence of nancing of RampD

Our paper is also related to Hellmann (1997) who examines thechoice between VC and corporate nancing of new ventures In his

5 Intellectual-property rights may be weak when it is difcult to clearly specifythe boundaries of the knowledge being contracted upon Indeed there is evidencethat trailer clauses and other contractual devices aimed at preventing entrepreneurialholdup are remarkably weak in many situations Anand and Khanna (2000) discusswhy it may be useful to think of the strength of property rights as exogenous


model the corporation helps projects that complement existing oper-ations but hinders projects that are substitutes (eg a project thatwould cannibalize existing products) Hence the larger are the ben-ets to existing operations the greater are the advantages of corpo-rate nance Conversely VC nancing dominates when the projecthurts the corporationrsquos current activities We depart from Hellmannby assuming that the corporation cannot commit to share the surpluscreated by the innovation because it can manipulate the accounts ofthe new venture moreover this is independent of whether the projectis a complement or a substitute to existing operations Thus in ourmodel the researcher may prefer VC nancing even when the cor-poration can exploit complementarities Furthermore since Hellmannconsiders neither appropriability issues nor the possibility of switch-ing between nanciers both our results on the inuence of weak prop-erty rights and those concerning the nance cycle of RampD are novel

Finally Anton and Yao (1994 1995) have also studied the conse-quences of weak property rights In their 1995 paper they consider thecase of an employee-researcher who already has an idea that can beefciently developed with the help of her corporate employer Whilethe researcher may be expropriated after revealing the idea to heremployer they show that in equilibrium she will not break awayand will appropriate part of the projectrsquos surplus Like Anton andYao we assume that the researcherrsquos bargaining power stems fromthe possibility of leaving to develop the innovation elsewhere6 How-ever we add a research stage and study how the possibility of anex post breakaway by the researcher will affect the ex ante incentivesof the corporation to nance research7 We show that the possibilityof walking away with knowledge creates a reverse holdup problem(ie the researcher holds up the corporation) but need not preventnancing Indeed we nd conditions under which breakaway willnot occur in equilibrium even when property rights are weak andcorporate development is inefcient We also study the case whenproperty rights are strong and the nancier can establish propertyrights over knowledge Strong property rights reduce the value of theresearcherrsquos outside option and increase the corporationrsquos willingnessto nance but on the other hand they make the researcher vulnera-ble to an ex post holdup by the corporation We show that in this casethe researcher obtains funds from an outside nancier

6 Similarly Anton and Yao (1994) show that the inventor can prevent the nancierfrom stealing the innovation by threatening to reveal the idea to a competitor andincreasing product-market competition

7 Indeed this is the main (efciency) argument against limiting the scope of non-compete contracts for employees


FIGURE 1 TIMELINE

The rest of the paper proceeds as follows Section 2 presentsthe model and discusses why corporations and VCs are different n-anciers Section 3 solves the model and studies the sequence of nanc-ing of RampD Section 4 presents four applications of the model Sec-tion 5 concludes

2 The Model21 The Setup and the Main Questions

There are two periods a research phase and a development phaseand three dates date 0 before research date 1 in between researchand development and date 2 after development (see Fig 1 for thetimeline of the game) The output of the research phase is knowledgeThis knowledge is necessary in the development phase which yieldsa marketable innovation

There is a measure c of cash-constrained researchers and twonanciers a corporation and a VC All are risk-neutral One and onlyone researcher has the ability to turn an idea into useful knowledge atcost R and as in Hellmann (1997) there is asymmetric informationwhile each researcher knows her ability at date 0 nanciers do notAll uncertainty about researcher ability is revealed after the research

Weak Property Rights and Holdup in RampD 621TABLE I

Notation

R Cost of researchDr Dc Cost of developmentS Value of innovation (monopoly)Sr Sc Value of innovation (duopoly)v Icirc [0 1] Veriability of researchb Icirc [0 1] Strength of local spilloversAacuterampd Aacuter VC RampD contractAacuted VC development contractw rampd Corporate RampD contractw d Corporate development contract

phase at date 1 but ability is nonveriable and cannot be contractedupon8 Researchers cannot divert any funds during the research phaseand derive no intrinsic utility from being nanced in research Nev-ertheless a nancier will not make any cash payments beyond R atdate 0 because that would attract a lemon with probability one

In both the research and the development phase the researchercan obtain funding from either the VC or the corporation and shecan switch nanciers at date 1 The development cost is Dr if theresearcher partners with the VC (call this independent development)and Dc if the researcher partners with the corporation (call this jointdevelopment) with Dr Dc 9 (The notation is summarized in Table I)The assets required for development cannot be rented from the moreefcient party For example if Dc lt Dr the only way of exploitingthis cost advantage is to develop jointly Once brought to marketthe innovation yields a surplus worth S in present value if exploitedmonopolistically We further assume that S R maxDr Dc gt 0 netsurpluses are positive

8 This rules out a contract specifying ldquoI will pay you $X upon its being revealedthat you are a good researcherrdquo

9 The case of Dr gt Dc may result from a corporation being able to provide easieraccess to markets supplies and nance because of existing operations or because itcan leverage existing assets (for example its patent base) cheaply For example Kids RUs founded by Toys R Us was a new product line (childrenrsquos clothes) but in a familiarmarket so it could use its industry-specic skills to select styles and manufacturersbuy display and sell products process information and control inventories see Blockand Macmillan (1995) for other examples and Teece (1992) for a discussion of comple-mentary assets On the other hand there may be reasons why Dr lt Dc a VC by virtueof serving on multiple boards of portfolio companies and receiving and nancing alarge array of deals is probably more familiar with the common mistakes and sourcesof failure in a business than is a corporation which essentially has experience with asingle business model or process In addition venture capitalists are usually successfulformer entrepreneurs with industry-specic expertise and a unique network of personalrelationships For example at Kleiner Perkins each partner is required to have previousexperience running a company or working near the top According to Fortune (ldquoSiliconValley Machinerdquo October 26 1998) ldquoKleiner Perkinsrsquo pals include the founders ofNetscape Cisco AOL Intuit Intel TCI Sun Lotus and Comcastrdquo


In the spirit of Grossman and Hart (1986) and like Aghion andTirole (1994) we assume that neither knowledge nor the innovationcan be described at date 0 Thus at date 0 parties cannot write acontract that xes a payment upon delivery of either knowledge atdate 1 or the innovation at date 2 In addition knowledge becomesembodied in the researcher and parties can dene property rightsover it only imperfectly We model this assuming that if the researcherswitches at date 1 after the research phase to pursue developmentelsewhere the nancier can establish its rights over the innovationrsquoscash ow as specied in the contract signed at date 0 with probabilityv with probability 1 v the nancier is unable to establish its rightsWeak property rights give rise to the holdup problem the researcherhas an incentive to switch nanciers at date 1 right after the researchphase The question is whether the holdup problem can be solvedsince the usual solution of allocating property rights over knowledgeis not feasible

The second question of interest concerns the sequence of nanc-ing of RampD Assuming that the holdup problem can be solved itwould seem that the nancier with the cost advantage in develop-ment should always be able to assure the researcher of getting at leastas much as she could by partnering with another nancier Conse-quently will breakaway ever be observed in equilibrium To studythis we must rst discuss the differences between the VC and thecorporation

22 What Is Different about Corporations

Since neither knowledge nor the innovation can be described at date0 contracts can at most specify how the projectrsquos surplus S will beshared As we discuss in this subsection however the next two as-sumptions imply that only the VC can commit to a sharing rule

Assumption 1 (Account Manipulation) Any costs incurred by the cor-poration in the nancing of research or development and any rev-enues generated in the subsequent marketing of the innovation arenonveriable By contrast the VC cannot manipulate the projectrsquosaccounts

Assumption 2 (Local Spillovers) If the corporation nances researchit learns the knowledge with probability b then it can develop theinnovation after date 1 even if the researcher leaves Local spilloversare nonveriable By contrast the VC cannot exploit local spillovers

These assumptions follow from the jointness or nonseparabilityof activities within corporations (Alchian and Demsetz 1972) whichmay become manifest in a variety of ways Assets or resourcesmdash


eg lab facilities centralized purchasing facilities and managerialtimemdashare often shared across activities and projects the transfer ofknow-how or best practices across projects may be common or thecommunication of information may be essential even inevitable (seeChandler 1966) Similarly processes and outputs may need to be com-patible in order to function together10 products may be cross-sold orbundled11 and new products may cannibalize existing ones

Of course a corporation should be able in principle to mimic theorganization of a stand alone project in which case the organizationof RampD within and outside corporations should look the same Never-theless the key point is that it may not be optimal for the corporationto do so since complementarities on both the cost side (arising fromshared ideas assets and time) and on the revenue side (arising fromcross-selling bundling etc) would be lost12 13

The consequence of joint activities is that the proper accountingof costs and revenues for each project may be difcult within the cor-poration because isolating the components of costs or the sources ofrevenue generation may be a genuinely hard task Moreover it maybe in the interest of the corporation to misrepresent costs and rev-enues to reduce payouts to the researcher In either case the net effectis that costs and revenues from projects are at least partially nonver-iable14 Holmstrom (1989) rationalizes similar assumptions arguingthat either ldquothe best intentioned rm does not know capital costsrdquoor even if it does it ldquohas control of many levers to make accountingmeasures less reliablerdquo Consequently ldquothe allocation of costs poses a

10 See for example Baldwin and Clarkrsquos (1999) study of evolution of the architec-ture of the PC

11 Shapiro and Varian (1998) give many examples to illustrate the benets ofbundling including its role in exploiting scale or scope economies and in reducingdispersion in consumersrsquo willingness to pay for products

12 Clearly the problem of cost and revenue accounting may be more or less severefor different projects Thus Assumption 1 is extreme because we do not allow for partialveriability which would mitigate accounting problems We do this because we want tofocus on the interaction between nonveriability and weak property rights In anotherpaper (Anand and Galetovic 1999) we assume that the corporation can choose amonga continuum of delegation levels and study the trade-off between complementaritiesand stronger incentives due to delegation

13 Companies often try to organize ventures to exploit existing corporate know-how relationships and operating processes most effectively For example Block andMacmillan (1995) argue that ldquo by far the most important factor to consider in ven-ture organization is the relatedness between the venturersquos business and the establishedbusinessrdquo

14 Several recent studies make a similar argument For example Feinstein (1995)examines how entrepreneurs can manipulate accounts because of superior informationAghion and Tirole (1994 p 1189) also assume that the value of the innovation to thecorporation (customer in their paper) is noncontractible because the benet is sharedacross many activities within its domain (although the revenue from the particularinnovation is veriable in their model)


dilemmardquo for rms trying to charge fees for using its assets Indeedhe argues

Difculties in identifying relevant costs and benets so asto make the innovator bear his marginal share are cen-tral Of course even as an individual entrepreneur mea-surement problems are substantial The entrepreneur doesnot know all the relevant gures either But the knowledgethat the money will ow into his own pockets that nothingwill be taken away still provides appropriate incentives Itis when nancial accounts are integrated that the difcul-ties of measurement become consequential and severe The key point is that veriability is an endogenous vari-able which depends on the incentives of those who collectthe information15

The rst implication of Assumption 1 is that the corporationcannot commit to a surplus-sharing rule at any date Moreover eventhough it can make a cash payment to the researcher at any date itwill not do so until date 1 otherwise it induces adverse selectionConsequently the most that the researcher can get from the corpora-tion is its outside option after research viz what it would get shouldit leave to pursue independent development with a VC any otherdivision of surplus agreed to at date 0 would be renegotiated How-ever the VC can commit to any surplus-sharing rule at date 0 becauseit cannot manipulate accounts16

The second implication of Assumption 1 regards the paymentsthe researcher makes in the event of switching nanciers Sinceresearch costs are nonveriable the corporation might claim they arearbitrarily high if the researcher leaves at date 1 This claim isbounded above by the project surplus S only because of limited lia-bility of the researcher so that the corporation cannot commit not to

15 The problem of nonveriability of costs and revenues is not restricted to RampDand may be endemic to corporations For example the revenue-accounting problemfaced by television networks is severe because lead-in effects and cross-promotionsare important determinants of viewing behavior and hence show ratings (see Shacharand Anand 1998) Similarly investment banks face a difcult problem when allocatingthe costs of their research divisions across other groups because many deals involve anumber of departments Nevertheless Eccles and Crane (1998) report that investmentbanks rarely resort to complicated fee-splitting algorithms since they believe that inldquofee splitting as in transfer pricing and cost allocations there are no right answersrdquoMoreover ldquoto attempt to do otherwise would ultimately be futile given the high levelsof interdependence created by the economic characteristics and production process ofthe industryrdquo (p 156)

16 It may be argued that the researcher could manipulate accounts to avoid sharingsurplus with the VC Nevertheless this is less likely because VCs typically nancestand-alone projects


appropriate the entire gross surplus from the project if the researcherswitches nanciers17 This implies that with probability 1 v theresearcher-VC partnership gets nothing thus this event can be inter-preted as the corporation having effectively blocked the developmentof the project In contrast since the VC cannot manipulate accountshe can commit at date 0 to limit his payoff in the event of a corporateacquisition

Finally Assumption 2 also stems from the fact that the inter-nal organization of the corporation is designed to exploit the com-plementarities that exist across projects or activities This will make itmore likely that the corporation learns the knowledge and is able todevelop the innovation even without the researcherrsquos participation18

As Hellmann (1997) points out the fact that a corporation will gener-ally have a strategic in addition to a purely nancial motive to investis in fact ldquocentral to the identity of a corporationrdquo This feature is alsocentral in Aghion and Tirole (1994) who assume that the corporationhas a distinctive position relative to an outside nancier (eg a bank)because it is also a potential customer of the innovation Note that ifboth the corporation and the researcher develop the innovation theywill compete in the product market We assume that in that case theyearn respectively surpluses Sc and Sr with Sc 1 Sr pound S

23 Timeline

We now describe the timing of actions also summarized in Figure 1

1 Contract offers and research At date 0 the corporation and the VCsimultaneously offer contracts to the researcher19

The VC offers a nancing contract Aacuterampd Aacuter at date 0 HereAacuterampd is the amount to be paid to the researcher at date 2 afterthe innovation is brought to market if the VC nanced both

17 Even if the corporation writes a contract claiming that it contributes only R tothe research project it can claim later that the project used assets and employee ormanagerial time from other projects This is consistent with the broad claims made bycorporations ling for breach of noncompetes (eg Hertz vs Avis and Wal Mart vsAmazon) and the attempts of courts to severely restrict these (eg Augat Inc vs AegisInc)

18 An example is the recent attempt by Microsoft to develop technology for trans-mitting video and audio over the Internet to reduce the value of this technology to itsbreakaway inventor Rob Glaser who left in 1993 to start RealNetworks Inc (see ldquoAFormer Ally Joins the War On Microsoftrdquo Wall Street Journal July 24 1998)

19 We can restrict attention to high-quality researchers without loss of generalitySince researchers get paid from project surplus when nanced by a VC and low-qualityresearchers generate projects with no surplus the latter would not approach a VC inthe rst place since they derive no intrinsic utility from being nanced and cannotdivert research funds Although the corporation can make a cash payment right afterresearch by then all uncertainty about researcher quality is revealed so a corporationcan refuse to pay a low-quality researcher


research and development (then the VC gets S R Dr Aacuterampd) and Aacuter is the amount to be paid to the VC if the researcherobtains development nancing from the corporation at date 1and the VC can prove that he seed-nanced the project Limitedliability implies that Aacuter pound SThe corporation offers a nancing contract w rampd Here w rampd

is the amount to be paid to the researcher at date 1 if jointdevelopment occurs20 21

2 Research nanced by the VC If the researcher accepts VC nancethen

The VC pays R Then knowledge is acquired by the researcherAt date 1 (after the research phase) either of the followinghappens

1 VC nances development The VC pays the devel-opment cost Dr 22 Surplus S is then shared accord-ing to the nancing contract Aacuterampd Aacuter and thegame ends

2 Corporate acquisition or trade sale The researcheraccepts a take-it-or-leave-it offer from the corpo-ration to pursue development nancing the con-tract species a payment w d to the researcher withresidual prots going to the corporation Anytimeafter development costs have been incurred theVC is paid Aacuter according to its initial nancing con-tract with the researcher23 If no payment is madeto the VC the VC sues and with probability v suc-cessfully proves that he seed-nanced the projectthereby establishing (senior) claims in the amount

20 Exactly like a VC the corporation can also offer a nancing contract to theresearcher that species a sharing rule over project surplus However as discussedabove because costs and surplus are nonveriable the corporation cannot crediblycommit to share any surplus from the innovation with the researcher therefore w rampdcannot be greater than the researcherrsquos ex post outside option

21 Again recall that since costs are nonveriable the corporation cannot committo limit damages in the event of contract breach Hence without loss of generality wecan restrict attention to cases where the corporation does not specify such damages inthe nancing contract

22 In our model neither the VC nor the entrepreneur can commit upfront to con-tinuing with the other party after the research phase the former because of uncertainresearcher quality the latter because of weak property rights over knowledge Theeffects of entrepreneurial holdup on the optimal VC nancing contract (timing of pay-ments and staging of nancing respectively) have been studied by Hart and Moore(1994) and Neher (1995)

23 The reason for this restriction is that sinking development costs is the onlyveriable action that indicates that the researcher has partnered with another agent


Aacuter on the projectrsquos surplus24 With probability 1 vthe VC is unsuccessful and gets nothing SurplusS is then shared at date 2 between the researcherand corporation according to the acquisition con-tract and the game ends

3 Research with the corporation If the researcher accepts the corpora-tionrsquos offer then

The corporation pays R Then knowledge is acquired by theresearcherAt date 1 (after the research phase) one of the followinghappens

1 Corporate development The researcher continueswith the corporation who pays Dc to develop theinnovation and bring it to market At date 2 thecorporation pays w rampd and appropriates surplus Sand the game ends

2 Spinoff The researcher accepts a take-it-or-leave-itoffer from the VC to pursue development nanc-ing the contract species a payment Aacuted to theresearcher with the remaining share of the sur-plus going to the VC After the researcher leavesthe corporation proves with probability v that itseed-nanced the project in which case develop-ment is blocked25 With probability 1 v the cor-poration is unsuccessful the innovation is broughtto market surpluses are shared at date 2 and thegame ends

3 The corporation decides whether to pursue devel-opment on its own if it does it spends Dc and issuccessful in development with probability b

3 The Sequence of RampD Finance

In this section we study the determination of the equilibrium nanc-ing of RampD We rst look at variation in the strength of propertyrights ignoring local spillovers (Sections 31 and 32) We then exam-ine the consequences of local spillovers

24 This is a simplifying assumption the effects of other forms of damage paymentsare qualitatively similar

25 Equivalently one can assume that the project goes ahead and with probabilityv the corporation receives damages up to the entire project surplus S in either casethe researcher and VC get nothing This assumption follows from the fact that sinceresearch costs are nonveriable when nanced by the corporation the latter cannotcommit to limiting damages in the event of contract breach


31 Strong Property Rights

What does the sequence of nancing of RampD look like when propertyrights are strong Proposition 31 shows that the VC always nancesresearch but that the innovation is developed efciently

Proposition 31 (RampD Financing When Property Rights Are Strong)Let v sup3 R S Then (i) the VC always nances research (ii) if Dr lt Dc theVC nances development (iii) if Dr sup3 Dc the innovation is developed bythe corporation

Proof To prove part (i) we show rst that the VC is always willing tonance research If the researcher breaks away at date 1 the VC willget paid with probability v Since v sup3 R S it follows that the projectrsquosgross surplus S is at least as large as R v the amount required for theVC to recoup R in expected value Next we show that the corporationcannot offer a research contract that lures the researcher At date 1 allthe bargaining power of the researcher stems from the possibility ofleaving to develop independently In that case with probability v thecorporation successfully prevents development with probability 1 vthe corporation cannot prevent development and the researcher getsat most S Dr (or even less if bSc sup3 Dc and the corporation developsindependently) Hence if the researcher partners with the corporationat date 0 she gets at most

max(1 v)S Dr 0 (31)

Since by assumption 1 v pound (S R) S this is at most equal to thesurplus that is created if the VC nances both research and develop-ment S R Dr and generally less Thus the VC can offer a contractthat matches and improves anything that the corporation may offer

(ii) Suppose that Dr lt Dc Then the surplus from staying withthe VC at date 1 is greater than the surplus when leaving and the VCcan offer a contract that makes the researcher stay

(iii) Let Dr sup3 Dc Then the surplus created if the researcherleaves to develop with the corporation S R Dc is always atleast as large as the surplus that is created if the researcher stayswith the VC S R Dr Thus whatever the share of the surplusthe VC would appropriate if the researcher stays it can get at leastas much in expected value from the breakaway fee Thus the VCand the researcher will always sign a contract that allows efcientbreakaway u

Figure 2 summarizes the sequence of nancing of RampD Whenproperty rights are strong (ie to the right of v 5 R S) the VC always


FIGURE 2 THE SEQUENCE OF RampD FINANCE (NO LOCALSPILLOVERS)

nances research Nevertheless when Dr gt Dc the corporation devel-ops the innovation (area I) examples of these include ex post licensesmanufacturing joint ventures and acquisitions by the corporation Bycontrast when Dc gt Dr both research and development are nancedby the VC (area II) Thus the VC always nances research In fact onecan obtain an even stronger result namely that when property rightsare strong the VC will nance research even when corporate researchis more efcient To see this assume for a moment that research costsdiffer with nanciers and let Rr be the cost of independent researchNote that condition (31) implies that the VC can always match thecorporation when v sup3 Rr S since

max(1 v)S Dr 0 pound S Rr Dr (32)


Now condition (32) is independent of Rc the cost of corporateresearch because that is already sunk when the holdup occurs Hencewe have established the following result

Result 32 Let v sup3 Rr S Then the VC will nance research even whenthe corporation is more efcient in both research and development

The unattractiveness of corporate-nanced research is funda-mental and stems from the interaction of the corporationrsquos inabil-ity to commit to a sharing rule and strong property rights (whichenables the corporation to expropriate the researcher) The corpora-tionrsquos inability to commit implies that it can only pay the researcherrsquosoutside option which as seen from condition (31) is of little valuewhen property rights are strong By contrast the VC can commit toany sharing rule to match the corporation

Result 32 contrasts with Aghion and Tirole (1994) In theirmodel the corporation (or ldquocustomerrdquo in their terminology) will ownthe research project whenever it is (second-best) efcient26 Howeverif contractible outside nancing were available with no capital con-straints then the rst best would always be attained27 In our modelwhile allocating ownership to the corporation may give the rightinvestment incentives it also transfers to it most of the bargainingpower Thus the researcher may prefer to get funding from an out-side nancier to appropriate part of the projectrsquos surplus even if thatis inefcient28 Result 32 also contrasts with Hellmann (1997) whonds that the corporation will nance research whenever it can takeadvantage of complementarities His result hinges on the assumptionthat surplus is veriable so the corporation can commit to matchwhatever the researcher would get by partnering with a VC

A key characteristic of the nancing of RampD when propertyrights are strong is

Result 33 In equilibrium efcient breakaway from the VC is observed

Strong property rights imply that the researcher can crediblycommit to compensate the VC in the event of a breakaway despitethe fact that knowledge is embodied in her Thus a prediction of themodel is that corporate acquisitions should be common when prop-erty rights are strong and that when they occur they are efcient Note

26 In their model corporate ownership of the research project is (second-best) ef-cient when returns are very sensitive to the corporationrsquos investment

27 Corporate ownership blunts the researcherrsquos effort incentives in Aghion andTirole (1994) Contractible outside nancing would allow the corporation to choose theoptimal investment level and transfer ownership to the researcher thus achieving therst best

28 Of course this depends on the fact that the researcher cannot sell the project tothe corporation at date 0


that at date 1 the adverse-selection problem is no longer relevant sothat the corporation will be willing to pay cash for the invention

32 Weak Property Rights

When property rights are weak the VC will not recoup R in expectedvalue if the researcher breaks away since v lt R S This reduces thescope of VC nancing Proposition 34 characterizes the sequence ofnancing when local spillovers do not matter (that is bSc lt Dc )

Proposition 34 (RampD Financing When Property Rights Are Weak)Let v lt R S and bSc lt Dc Then (i) the corporation will nance research ifand only if

Dr Dc sup3 R vS (33)

(ii) the VC will nance research if and only if

Dc Dr sup3 R vS (34)

(iii) a project will not be nanced if and only if

R vS gt Dr Dc gt (R vS) (35)

Proof (i) Note rst that since bSc lt Dc the corporation will neverdevelop independently Thus if the researcher breaks away from thecorporation at date 1 she gets (1 v)S Dr gt 0 since v lt R S andS R Dr gt 0 by assumption Hence the corporation will be willingto nance research if and only if S R Dc sup3 (1 v)S Dr whichleads to the incentive-compatibility condition (33)

(ii) When v lt R S the projectrsquos gross surplus S is not sufcientto compensate the VC in expected value since vS lt R Thus whenproperty rights are weak the VC nances research only if a holdupis not protable Assuming that Aacuter 5 S the surplus to be shared withthe corporation in the event of a breakaway is (1 v)S Dc gt 0 Hencethe VC will nance the project if and only if S R Dr sup3 (1 v)S Dc which leads to the incentive-compatibility condition (34)

(iii) Note that inequalities (33) and (34) imply (35) u

Figure 2 summarizes Proposition 34 To the left of v 5 R S thecorporation nances research and development but only if its costadvantage in development is sufciently large (area V) examples ofthis case include research joint ventures ex ante licenses or the hiringof the researcher as an employee of the corporation Moreover no


breakaways are observed in equilibrium Mutatis mutandis exactly thesame holds for the VC (area III) Thus a striking feature of Proposition34 which is apparent from Figure 2 is the symmetry of corporate andVC nance Thus

Result 35 When property rights are weak the corporationrsquos inability tocommit to a sharing rule does not affect the sequence of nancing of RampD

When property rights are strong the inability of the corporationto commit to a sharing rule is key to determining the nancing ofRampD Weak property rights change two things First the VCrsquos abilityto contractually x the value of its compensation when a breakawayoccurs Aacuter is no longer relevant to its nancing decision because it cannever recover its research costs in the event of breakaway (v lt R Sso that vAacuter lt R) Thus the VCrsquos decision to nance depends solely onwhether he can match the value of the researcherrsquos outside optionmdashjust as is always the case with the corporation Second while thecompensation that the corporation can credibly offer is still limitedby the researcherrsquos outside option the condition Dr Dc 5 R vSidenties points where the researcher gets all the projectrsquos net surplusThus the corporationrsquos inability to commit is no longer relevant Bothfacts explain symmetry Next it is also apparent from Figure 2 andProposition 34 that weak property rights may prevent the nancingof research altogether (area IV)

Result 36 When development costs are similar across the differentnanciers protable projects may not be nanced

To see the intuition behind this result consider the case ofcorporate-nanced research The VC will be willing to nance inde-pendent development as long as (1 v)S Dr sup3 0 The key thingto note is that by breaking away the researcher can avoid payingthe research cost R Since v is small when property rights are weak(1 v)S Dr S Dr To prevent a breakaway the corporationwould need to have a signicant advantage in development since itshould not only match this outside option but also cover the researchcost (recall that the researcher appropriates all surplus of marginalprojects) Hence the incentives for a breakaway exist even when inde-pendent development is more costly By symmetry exactly the sameargument explains why the VC may not be willing to nance In thesecases projects will be executed only if the researcher can nance theresearch cost on her own29 Result 36 also shows how the requirement

29 Note that there is no a priori reason to think that the government will be moreable than private nanciers to solve the holdup problem and thus nance the area IVprojects


that ex ante nancing be feasible imposes restrictions on the conditionsunder which ex post development can occur

The preceding discussion suggests that nanciers will anticipatebreakaways and then will not nance Therefore

Result 37 When property rights are weak no breakaways occur inequilibrium

Putting together Results 33 and 37 leads to the conclusion thatbreakaways occur only when property rights are strong The explana-tion for this somewhat surprising result is that strong property rightsenable the nancier to recoup research costs with high probabilityBy contrast when property rights are weak nanciers cannot recoupresearch costs in the event of a breakaway and they will lend only ifthey have a large cost advantage in development This result also con-trasts with Anton and Yao (1995) who nd conditions under whichbreakaways from a corporation occur when property rights are weakThe reason for this difference is that in their model the corporationdoes not incur any costs prior to the breakaway hence the questionof what preserves ex ante incentives to nance research is moot

Note that since only the VC nances when property rights arestrong it follows that

Result 38 The researcher may break away from the VC but not fromthe corporation

A nal observation is appropriate Note that knowing the exactdivision of surplus between the researcher and the nancier was notnecessary to determine the equilibrium sequence of nancing of RampDThis is so because along the boundaries dened by the incentive com-patibility conditions (33) and (34) the researcherrsquos outside optionequals project surplus thus whether the nancier can commit to asharing rule is irrelevant Nevertheless the ability to commit matterswithin each of the regions I II III and V and determines surplussharing

33 Local Spillovers

Local spillovers in research may give the corporation enough infor-mation to attempt independent development and become a product-market competitor In turn competition from the corporation in theproduct market reduces Sr and the value of the researcherrsquos outsideoption The next proposition studies the effects of local spillovers onthe sequence of nancing of RampD

Proposition 39 (The Effect of Local Spillovers) Let v lt R SbSc sup3 Dc and b ordm b(1 Sr S) Then (i) if v gt 1 (Dr S) (1 b ) the


corporation will always nance research (ii) if v pound 1 (Dr S) (1 b ) thecorporation will nance research if and only if

Dr Dc sup3 R [v 1 (1 v)b ]S (36)

Proof Since bSc sup3 Dc at date 2 the corporation will compete in theproduct market with probability b If the researcher breaks away fromthe corporation at date 1 the expected surplus (gross of developmentcosts) is equal to (1 v) [(1 b)S 1 bSr ] ordm (1 v)(1 b )S To provepart (i) note that if this does not cover the development cost Dr thenno VC will be willing to nance independent development Thus theresearcher will be unable to break away if (1 v) (1 b )S lt Dr whichyields v gt 1 (Dr S) (1 b ) after straightforward manipulation

To prove part (ii) note that if (1 v)(1 b )S sup3 Dr then indepen-dent development is feasible hence the corporation will be willing tonance research only if it can match the outside option that is ifS R Dc sup3 (1 v)(1 b )S Dr or Dr Dc sup3 R [v 1 (1 v)b ]S u

Figure 3 shows that local spillovers enlarge area V [we assumethat 1 (Dr S) (1 b ) lt R S] case (i) obtains to the right of the kinkin the line separating areas IV and V case (ii) to the left30 Thus

Result 310 When property rights are weak local spillovers enlarge thescope for corporate research because appropriability problems cancel outinstead of adding up

When v pound 1 (Dr S) (1 b ) an additional term appears in thecorporationrsquos individual rationality constraint (36) namely (1 v)b ordm(1 v)b(1 Sr S) The stronger are local spillovers (the larger is b) orthe tougher is competition from the corporation in the product mar-ket (the smaller is Sr) the less protable independent developmentbecomes and consequently the more attractive is corporate RampDIn the extreme when v gt 1 (Dr S) (1 b ) competition from thecorporation in the product market becomes so strong that the VCis no longer willing to nance independent research The result thatproduct-market competition can help the victim of holdup to appro-priate part of the projectrsquos surplus has also been recognized by Antonand Yao (1994) Our result shows that such competition also may giveincentives to nance research and overcome the holdup problem

The second part of Result 310 is a standard application of thesecond-best principle Local spillovers and weak property rights over

30 Note that 1 (Dr S) (1 b ) lt R S simplies to the condition (S R)(1 b )Dr lt 0 it can be easily shown that the kink must occur in the region where Dr Dc lt 0


FIGURE 3 THE EFFECT OF LOCAL SPILLOVERS

knowledge are usually thought to reduce the incentives for innova-tion but here they work to increase the scope of nancing The rea-son is that weak property rights create a preexisting distortion that isneutralized by local spillovers Since different sorts of appropriabil-ity problems are common in RampD markets this hints at a more gen-eral implication namely that imperfections in RampD markets shouldnot be analyzed separately rather their interactions must be carefullyassessed The next result is also apparent from Figure 3

Result 311 When property rights are weak the corporation may nanceRampD even when independent development is more efcient

Even when corporate development is not efcient (that is Dc gtDr ) the corporation may still provide research funding in equilibriumOn the one hand although it is efcient for the VC to nance research


and development this is not feasible because the researcher wouldbreak the VC contract at date 1 The corporation on the other handis protected from a holdup by its ability to exploit local spillovers andbecome a product-market competitor

What is the effect of local spillovers when property rights arestrong The last result indicates that under such conditions localspillovers do not affect the nancing of RampD

Result 312 When property rights are strong local spillovers may affectthe division of surplus between the researcher and the nancier but not thesequence of nancing of RampD

We know from the proof of Proposition 31 that local spilloversmake corporate research even less attractive for the researcher How-ever this means that the VC can offer her less and still get the contractThus when property rights are strong local spillovers may reduce thepayoff to the researcher but affect neither the feasibility nor the loca-tion of RampD

4 Applications

41 Some Performance Implications

There is some evidence that VC-nanced projects obtain higher ret-urns on average It is often concluded that VCs are inherently betterat selecting projects and corporations should imitate their governancestructure Our model however suggests caution in interpreting thisevidence

Consider the consequences of a smaller research cost R cent lt Rkeeping all else equal (ie the project has a higher return) Figure 4shows that high-return projects are more likely to obtain researchnancing when property rights are weak either by the corporationor by the VC (local spillovers are assumed irrelevant in the gure)When property rights are strong however the predominant effect isthat VC nancing substitutes for corporate nancing (a given grosssurplus S is more useful for compensating the VC in the event of abreakaway thus enlarging the region where knowledge is veriable)Therefore the average return of a VC-nanced project will be higherthan that of a corporate-nanced project ceteris paribus It can be easilyshown that a similar result obtains when S Sr and Sc increase propor-tionately This suggests that VCs may obtain higher returns becauseresearchers with better projects self-select due to the inability of cor-porations to commit to a sharing rule Thus project quality may drivenancier selection not the other way round


FIGURE 4 A DECREASE IN RESEARCH COSTS

42 Strategic Local Spillovers

In the model we treat local spillovers as a primitive This is sen-sible when little unique knowledge is revealed to others no mat-ter how much interaction or communication is facilitated betweenkey researchers and other employees In other cases however localspillovers may be endogenously affected by the internal organizationof research activities For example requiring frequent updates andreports organizing researchers in teams and conducting research incommon labs may all enhance local spillovers In such cases howshould research be organized internally A common line of reasoningsuggests that corporations would do well to structure research activi-ties as close to a VC model as possible to provide the sharpest incen-tives to researchers [see for example Sahlman (1990)] Neverthelessthere may be an important benet to doing the opposite Integrating


the researcher into the rest of the rmrsquos activities is likely to enhancelocal spillovers and by making the corporation a likely competitordeter holdup Again this is an application of the second-best Result310

The model also has an empirical prediction When the corpora-tion has a large cost advantage in development local spillovers willnot be important for preventing a holdup Thus corporate researchfacilities that are separated from each other or from the rest of therm are more likely when these cost advantages are large By contrastwhen these cost advantages are small local spillovers will becomeimportant and the locus of innovation within rms is likely to bemore concentrated

43 Alliances or Venture Capital

A vast literature studying interrm alliances has focused on two areasthe tradeoff between complementarities between partners (in RampDproduction or marketing) which enhance the scope for cooperationand mutual learning (and other local spillovers) which enhances thescope for conict While there are clearly gains to such learning (forexample learning to manage alliances to deal with conict resolutionetc) learning or acquiring knowledge from an alliance partner mayalso increase the incentive to break away from the partner and pursuethe project on onersquos own31 Our approach may help to explain howthese forces interact First contrary to common belief holdup prob-lems that arise in research activities may not be solved by contractualprovisions that attempt to optimally allocate control and cash-owrights over the resulting output of the partnership because owner-ship rights over the output of the partnership (knowledge) may notbe well dened Second enhanced local spillovers or learning betweenthe partners can serve as credible commitments not to cheat on eachother so the usual conict thought to arise from such local spilloversmay also conceal a benet that serves to increase alliance stabilityFinally Result 36 states that holdup is likely to be more severe formore generic capabilities of the partner a corollary to this is that part-nerships that create greater total surplus may not be pursued if thecost differences between competing partners in development is small

44 The Changing Structure ofthe Pharmaceutical Industry

Up until 25 years ago pharmaceutical research was based largely onbrute-force trial-and-error methods performed in large-scale facilitieswhere researchers would synthesize an immense array of chemical

31 Adarkar et al (1997) provide numerous examples that highlight this possibility


entities to nd a useful one32 Most research was done by big rmsthat also developed and marketed them In recent years the industryhas changed signicantly First fundamental advances in molecularbiology and genetic engineering have led to discovery by design33 Thusbrute-force trial-and-error-methods are less common today Secondnow researchers can better describe innovations not only after usefulchemical compounds have been found but also ex ante before they arediscovered Third advances in experimentation techniques (eg com-puter simulations) have reduced the scale of experimentation Fourthwhile advances in science have had a notable inuence on the dis-covery process they have not changed the type of assets needed fordrug development and marketing Last the industry has seen a largeincrease in the division of labor a signicant part of pharmaceuticalresearch is performed nowadays by small or medium-size research-intensive biotechnology companies that specialize in research a largefraction of those companiesrsquo revenues comes from research contractswith larger companies

Our model offers a simple explanation for the changes in thenance cycle of RampD in the pharmaceutical industry Both discov-ery by design and advances in experimentation techniques lead toa smaller research cost R per discovered drug ceteris paribus Fur-thermore the fact that innovations can be better described ex ante(not ex post see the discussion below) should lead to a larger veri-ability parameter v As our model predicts both a smaller R and alarger v lead to independent research nanced by the VC Moreoversince the assets needed for development and commercialization havenot changed it is not surprising that large companies still dominatethat stage Transfers of knowledge are feasible because patents can beclearly dened

It could be argued that the only driving force behind thesechanges is reduced economies of scale (a smaller R) which haveshifted research to smaller rms This illustrates the importance of dis-tinguishing between the ability to describe innovations ex ante (beforethey are made) and ex post Pharmaceutical rms have always reliedon patents to establish property rights ex post over their discover-ies (see Levin et al 1987) Nevertheless patents help only after dis-covery By contrast when the corporation has a large advantage in

32 Particular features of the market described in this section have received carefultreatment in Henderson and Cockburn (1996) Pisano and Mang (1993) and Zuckeret al (1998) A broader overview of the industry is provided in Gambardella (1995)

33 As Gambardella (1995 pp 23ndash24) points out ldquoWith discovery by design sci-entists use knowledge about the causes of human disorders the properties of drugcompounds and their action in the human organism to conceptualize the structure ofan ideal molecule that is expected to restore the altered equilibrium The ideal moleculeis then given to the laboratory chemists which search for substances whose molecularstructures match as closely as possible the theoretical modelrdquo


development a VCrsquos priority is whether it is possible to describe theproduct of research ex ante With brute-force methods of discoveryVC nancing was very difcult because it was not possible to specifyex ante what would be discovered much less its characteristics anduses thus it would have been far easier for a researcher to switchnanciers after learning which compound in a million was the rightone Thus despite the fact that drugs have always been patentablev was very small (In fact note that patents made VC nancing evenmore difcult once the discovery has been made patents make iteasier to sell the knowledge and reduce the chances of holdup by theacquiring corporation34) Corporations in contrast took advantage oftheir huge cost advantages in development to prevent holdups

5 Concluding Remarks

This paper has examined how the strength or weakness of propertyrights affects the sequence of nancing of RampD We conclude by point-ing out some directions for further research

While we assumed that VC- and corporate-nanced researchwere equally efcient our results suggest that self-selection ofresearchers may lead VCs to nance higher-quality projects In ourmodel more protable projects attract VCs because they can com-mit to a surplus-sharing rule while the corporation cannot The lit-erature has repeatedly pointed out a second reason why VCs showbetter performance they are able to provide stronger incentives Onemay argue that stronger incentives can be traced to the same factorthat makes researchers self-select the ability of VCs to commit to asurplus-sharing rule Thus joint activities limit not only a corpora-tionrsquos ability to transfer surplus to the researcher but also the powerof its incentive schemes Examining the link between the internal orga-nization of RampD incentives and the mode of nancing is a fruitfularea for research

Despite stronger performance VCs still account for a small frac-tion of all RampD nance and most corporations do not try to mimic theVC mode of governance Consequently it is unlikely that corporateRampD is simply an off-equilibrium phenomenon bound to disappearWe have shown here that joint activities may also facilitate corporateRampD nance For example local spillovers and strong product-marketcompetition from the corporation enlarge the scope of corporate RampDnance when property rights are weak More generally corporationsexist because doing many activities facilitates the exploitation of econ-omies of scope and complementarities which are absent in standalone

34 Lamoreaux and Sokoloff (1999) have shown that large corporations have regu-larly bought patents since the second half of the nineteenth century


projects This suggests that corporations may optimally sacrice incen-tives to exploit complementarities

REFERENCES

Adarkar A A Adil D Ernst and P Vaish 1997 ldquoEmerging Market Alliances MustThey Be Win-Loserdquo McKinsey Quarterly 4 120ndash137

Admati A and P Peiderer 1994 ldquoRobust Financial Contracting and the Role of Ven-ture Capitalistsrdquo Journal of Finance 49 371ndash402

Aghion P and J Tirole 1994 ldquoThe Management of Innovationrdquo Quarterly Journal ofEconomics 109 1185ndash1209

Alchian A and H Demsetz 1972 ldquoProduction Information Costs and Economic Orga-nizationrdquo American Economic Review 62 777ndash795

Amit R L Glosten and E Muller 1990 ldquoEntrepreneurial Ability Venture Investmentsand Risk-Sharingrdquo Management Science 36 1232ndash1245

Anand B and A Galetovic 1999 ldquoDecentralization and Incentives in RampDrdquoworking paper

and T Khanna 2000 ldquoThe Structure of Licensing Contractsrdquo Journal of IndustrialEconomics 47 103ndash135

Anton J and D Yao 1994 ldquoExpropriation and Inventions Appropriable Rents in theAbsence of Property Rightsrdquo American Economic Review 84 190ndash209

and 1995 ldquoStart-ups Spin-offs and Internal Projectsrdquo Journal of Law Eco-nomics and Organization 11 362ndash378

Baldwin C and K Clark 1999 Design Rules Volume I The Power of ModularityCambridge MA MIT Press

Barry CB 1990 ldquoThe Role of Venture Capital in the Creation of Public CompaniesEvidence from the Going Public Processrdquo Journal of Financial Economics 27 447ndash471

Bhattacharya S and J Ritter 1983 ldquoInnovation and Communication Signalling withPartial Disclosurerdquo Review of Economic Studies 50 331ndash346

Block Z and I MacMillan 1995 Corporate Venturing Boston Harvard Business SchoolPress

Chandler A 1966 Strategy and Structure New York DoubledayCohen W R Nelson and J Walsh 1996 ldquoAppropriability Conditions and Why Firms

Patent and Why They Do Not in the American Manufacturing Sectorrdquo paper pre-sented at the OECD Conference on New Indicators for the Knowledge-BasedEconomy

Eccles R and D Crane 1988 Doing Deals Investment Banks at Work Boston HarvardBusiness School Press

Feinstein J 1995 ldquoAsymmetric Information Accounting Manipulations and Partner-shipsrdquo Journal of Economic Behavior and Organization 26 49ndash73

Gambardella A 1995 Science and Innovation The US Pharmaceutical Industry During the1980s New York Cambridge University Press

Gompers P 1995 ldquoOptimal Investment Monitoring and the Staging of Venture Capi-talrdquo Journal of Finance 50 1461ndash1489

and J Lerner 1996 ldquoThe Use of Covenants An Analysis of Venture PartnershipAgreementsrdquo Journal of Law and Economics 39 463ndash498

and 1998 ldquoThe Determinants of Corporate Venture Capital Successes Orga-nizational Structure Incentives and Complementaritiesrdquo NBER Working PaperNo W6725


and 1999 The Venture Capital Cycle Cambridge MIT PressGrossman S and O Hart 1986 ldquoThe Costs and Benets of Ownership A Theory of

Vertical and Lateral Integrationrdquo Journal of Political Economy 94 691ndash719Hart O and J Moore 1994 ldquoA Theory of Debt Based on the Inalienability of Human

Capitalrdquo Quarterly Journal of Economics 109 841ndash879Hellmann T 1997 ldquoA Theory of Corporate Venture Investingrdquo Stanford Business

School Working Paper No 1452Henderson R and I Cockburn 1996 ldquoScale Scope and Spillovers The Determinants

of Research Productivity in Drug Discoveryrdquo Rand Journal of Economics 27 32ndash59Holmstrom B 1989 ldquoAgency Costs and Innovationrdquo Journal of Economic Behavior and

Organization 12 305ndash327Kanter RM 1989 When Giants Learn to Dance New York Simon and SchusterLamoreaux N and K Sokoloff 1999 ldquoInventive Activity and the Market for Technol-

ogy in the US 1840ndash1920rdquo NBER Working Paper No W7107Lerner J 1994 ldquoVenture Capitalists and the Decision to Go Publicrdquo Journal of Financial

Economics 35 293ndash316 1995a ldquoXerox Technology Ventures March 1995rdquo Harvard Business School Case

Study 9-295ndash127 1995b ldquoVenture Capitalists and the Oversight of Private Firmsrdquo Journal of Finance

50 301ndash318Levin R A Klevorick R Nelson and S Winter 1987 ldquoAppropriating the Returns

from Industrial Research and Developmentrdquo Brookings Papers on Economic Activity3 783ndash820

Neher D 1995 ldquoStage Financing An Agency Perspectiverdquo Mimeo Boston UniversityPakes A and S Nitzan 1983 ldquoOptimum Contracts for Research Personnel Research

Employment and the Establishment of lsquoRivalrsquo Enterprisesrdquo Journal of Labor Eco-nomics 1 345ndash365

Pisano G and P Mang 1993 ldquoCollaborative Product Development and the Market forKnow-How Strategies and Structures in the Biotechnology Industryrdquo inR Burgelman R Rosenbloom and S Richard eds Research on Technological Innova-tion Management and Policy 5 109ndash136

Sahlman W 1990 ldquoThe Structure and Governance of Venture Capital OrganizationsrdquoJournal of Financial Economics 27 473ndash521

Shachar R and B Anand 1998 ldquoThe Effectiveness and Targeting of Television Adver-tisingrdquo Journal of Economics and Management Strategy 7 363ndash396

Shapiro C and H Varian 1998 Information Rules Boston Harvard Business SchoolPress

Teece D 1992 ldquoCompetition Cooperation and Innovation Organizational Arrange-ments for Regimes of Rapid Technological Progressrdquo Journal of Economic Behaviorand Organization 18 1ndash25

Zucker L M Darby and M Brewer 1998 ldquoIntellectual Human Capital and the Birthof US Biotechnology Enterprisesrdquo Americal Economic Review 88 290ndash306


mega-successes Nevertheless persistent differences in performanceare surprising particularly since the robust venture-capital activityover the past two decades has attracted so much attention and thenature of venture-capital nance been dissected so thoroughly (seefor example Gompers and Lerner 1999) Indeed there is some evi-dence that when corporations mimic venture capitalists (VCs) theydo rather well (see Gompers and Lerner 1998) further amplifying thequestion why more corporate intrapreneurship is not organized simi-larly Third while the study of IPOs as an exit mechanism for venturecapitalists has attracted considerable attention more than one-third ofall VC exits are in the form of corporate acquisitions or trade salesThis raises the question when should one or the other form of exit beobserved Then a fourth question arises why is VC-backed researchfollowed by corporate acquisitions so common while the reversemdashcorporate-nanced research followed by independent developmentmdashis not Asset sales by corporations are certainly common so what isdifferent about RampD projects This paper uses an equilibriumapproach to study the sequence of nancing of RampD and analyzesthese puzzles within a unied framework

To explain these facts one must determine what is special aboutRampD We focus on two appropriability problems that are endemic toRampD1 First it is often the case that the knowledge acquired throughcostly research becomes embodied in the human capital of theresearcher not necessarily in the rm for which the research is per-formed2 In many cases property rights over this knowledge are weakbecause it is nonveriable Weak property rights enable the researcherto hold up the nancier by threatening to walk away with the project3

1 Theoretical papers that focus on appropriability problems include Anton and Yao(1994 1995) and Bhattacharya and Ritter (1983) Empirical papers include Cohen et al(1996) Levin et al (1987) and Anand and Khanna (2000)

2 As the The Economist notes ldquo the term lsquotechnology transferrsquo is something ofan oxymoron Real innovations do not move from laboratory to shopoor as patentsresearch reports or even working prototypes To stand any chance of success they haveto be transferred as concepts embedded in peoplersquos heads The one time an innovation(laser imaging) made it successfully out of (Xerox) PARC to become a multi-billiondollar business for Xerox it was because the person who championed it in the labora-tory Robert Adams moved with it and drove the innovation hard through engineeringmanufacturing marketing and salesrdquo (ldquoAdopting Orphansrdquo February 20 1999)

3 The importance of the threat of holdup can be inferred from the examples ofentrepreneurial breakaway For example ldquo[Xerox] PARC is famous for having pioneeredideas (including a superior personal computer the facsimile machine the Ethernetand the laser printer) that made fortunes for many of its Silicon Valley neighboursbut little for itselfrdquo (ldquoAdopting Orphansrdquo The Economist February 20 1999) see alsoLerner (1995a) and ldquoXerox Wonrsquot Duplicate Past Errorsrdquo Business Week September 291997) Elsewhere The Economist notes that ldquo people at Bell Labs have still not for-gotten how half a century ago William Shockley took the transistor idea which heand his colleagues had invented at Murray Hill to Palo Alto in California and started





















FIGURE 1 TIMELINE






Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES






























































FIGURE 1 TIMELINE






Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES























































FIGURE 1 TIMELINE






Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES
















































FIGURE 1 TIMELINE






Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES











































FIGURE 1 TIMELINE






Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES











































Notation


































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES






































































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES






























































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES





















































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES














































23 Timeline


































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES





































































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES


























































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES















































max(1 v)S Dr 0 (31)

















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES






















































































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES


















































































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES







































































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES

























































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES


















































33 Local Spillovers





Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES












































Dr Dc sup3 R [v 1 (1 v)b ]S (36)


















4 Applications































REFERENCES




















































4 Applications































REFERENCES















































4 Applications































REFERENCES





































































REFERENCES

































































REFERENCES

























































REFERENCES



















































REFERENCES












































REFERENCES






























































Documents

Weak Property Rights and Holdup in R&D - HBS … · Liebeskind, Felipe Zurita, two anonymous referees, the coeditor, and various colleagues for helpful suggestions, and to participants