1. the management of innovation 1994

Upload: dario-von-govedarovic

Post on 04-Jun-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 1. the Management of Innovation 1994

    1/27

    THE MANAGEMENT OF INNOVATION*PHILIPPEAGHION AND JEAN TIROLE

    The paper analyzes the organization of the R&D activity in an incompletecontract framework. It provides theoretical foundations: (a) to understand how theallocation of property rights on innovations may affect both the frequency and themagnitude of these innovations; (b) to rationalize commonly observed features inresearch employment contracts, such as shop rights, trailer clauses, and the hiredfor doctrine; (c) to discuss the robustness of the so-called Schumpeterian hypothe-ses to endogenizing the organization of R&D; and (d) to provide a rationale forcofintincing arrangements in research activities.

    I. INTRODUCTIONA feature common to the patent race and the endogenousgrowth literatures is their simplified representation of R&D activi-t ies which are assumed to be performed by an aggregate agentplaying simultaneously the roles offinancier creator owner and(often) user of the innovation. In practice, however, R&D takesplace either within firms where employees-inventors are subject to

    assignment contracts with their employers or through contractualagreements between independent research units and users of theirinnovations or fintmciers. In both cases, the contractual provisionson how to finance the research activities, how to allocate controlover the R&Dprocess bow to share property rights on innovations,and on how to structure the monetary compensations to theinventors^are far more complex than the current aggregated viewof the R&D process suggests.Using the incomplete contract framework introduced by Gross-man and Hart [1986] this paper analyzes tbe R&D activity from anWe thank Patrick Bolton, Fernando Branco, Drew Fudenberg, Oliver Hart,Rebecca Henderson, Gary Pisano, Diego Rodriguez, and Andrei Shieifer for theirhelp. This paper served as the basis for the 1993 Elisha Pazner lecture at theUniversity of Tel Aviv, which the second author had the honor to give. The secondauthor is also grateful to France Telecom (Centre National d'Etudes des Telecom-munications) and the Commissariat General du Plan for financial support.1. The sharing of return streams need not covary perfectly with propertyrights. For example, property rights may go to the employee in universities and tothe employer in aircraft manufacturers even though the sharing rules are similar

    (for instance, the four largest U. S. aircraft manufacturers offer their employeesshares of 10 percent to 30 percent on income collected from royalties [Neumeyer1971 Chapter 4]. Some universities give 15 percent of the royalties to their

  • 8/13/2019 1. the Management of Innovation 1994

    2/27

    1186 QU RTERLY JOURN L OF ECONOM ICSorganiza tional point of view. Firs t, it studies how t he allocation ofproperty rights on innovations can affect both tbe frequency andthe m agnitude of these innovations when their exact natu re cannotbe contracted upon ex ante. Second, it rationalizes a number ofcomm on con tractin g and legal features of the organization ofR&D.Th ird, it sheds light on th e findings ofth e empirical R&D lite ratu reand discusses th e rob ustn ess of tb e so-called Schu m peterian [1942]hypotheses on the effects of an increase in the scale, scope,monopoly power, or lon g-p urse enjoyed by tb e use r of aninnovation on R&D inputs and o utpu ts.

    The paper is structured as follows. Section II analyzes thebasic contractual relationsbip between a research unit and acus tom er. C usto m ers Eire tho se par ties who directly benefit fromthe innovation; namely, the m anufac turers who commercialize theinnovation, the users who will purchase the resulting product, andthe suppliers of complementary products or of inputs used by tbemanufacturer. (Which of these three kinds of customer financesthe innovation depends on the industry [von Hippel 1988], inparticular on who benefits most from the innovation. For thepurpose of this paper, we can content ourselves with a singleaggregated customer.) The research unit, wbicb may or may notbelong to tb e same firm as tbe custom er, performs th e creative tas kbut bas no independent resource to pay for salaries, equipment, orda ta. It m us t tberefo re look for outside financing. In a first step, weassum e tba t tb e financing is provided by tb e cu stomer.

    We posit that the exact nature of the innovation is ill-definedex ante and that the two parties cannot contract for delivery of aspecific innovation. Tbe contract (realistically) specifies a verifiableamount of customer investment, the allocation of property rightson any forthcom ing inn ovation, an d possibly, a shar ing rule on t heprofit (license fee) obtained by tbe research unit.In the integrated case the customer owns arid freely uses theinnovation. In tbe nonintegrated case the research unit owns theinnovation and, once the innovation is made, bargains witb thecustom er over th e license fee. The sh aring rule co ntracted u pon exante is shown to be irrelevant. The study then boils down to theclassic one of choosing property rights so as to best protect th e twoparties' specific investments in the relationship. Giving propertyrights to the research unit is optimal when it is more im porta nt to

  • 8/13/2019 1. the Management of Innovation 1994

    3/27

    THE MANAGEMENT OF INNOVATION 1187inefficiency of the allocation of property rights when the customerex ante h as sub stantial bargaining power.Is it optimal to bave a single provider of funds, namely thecustomer? Extending further the property rights model of Gross-m an an d H ar t by allowing for equity participation by tbe customeras well as by third parties (investors), we show in subsection II.4tbat it may be strictly optimal for tbe customer to give propertyrights to the research unit and to demand cofinancing by aninvestor (such as a ve ntu re capitalist, a ban k, or a pa ren t companynot in tbe customer's business). Furthermore, such a cofinancingcannot be duplicated by transferring the investor's share in theresearch unit to the customer, since the customer then facesconfiicting objectives wben bargaining with the research unit. Wethus obtain atheory ofthe existence of mu ltiple principals.

    Our framework also provides a rigorous evaluation of theSchumpeterian hypotheses(subsectionII.5).It questions tbe robust-ness of the Schumpeterian conjectures mentioned earlier to varia-tions in the allocation of property rights on innovation. Once thisallocation is endogenized as pa rt of th e contractual arran gem entsbetween customers and research units, one should not expect aclear empirical aggregate relationship between R&D input orou tpu t and p ara m ete rs such a s scale, scope, or monopoly pow er.More complicated allocations of property rigbts commonlyobserved in researcb employment contracts include (see Neumeyer[1971]) a)property rights con tingent on the nature ofthe innovation (wbicb, for instance, confer ownersbip to the customer-employer for inventions th at are rela ted to the employer'sbusiness or make use of the employer's facilities or data, and tothe researcher-employee for inventions that pertain to otherbusinesses); (b) trailer clauses (which confer ownership to thecustomer-employer for innovations made by a breakaway re-searcher-employee sbortly after quitting tbe firm and to theresearcher-employee otherwise); (c) shop rights (which conferownership to the researcher-employee while at the same timeallocating a nonexclusive, nonassignable, and royalty-free licenseto use tbe innovation to the em ployer). Shop rights, property rightscontingent on the nature of the innovation, and rules governingbreakaway research are instances of mu ltiple,split property rights.Section III rationalizes these institutions by extending tbe basic

  • 8/13/2019 1. the Management of Innovation 1994

    4/27

    1188 QUARTERLY JOURNAL OF ECONOM ICSof research an d th esizeof the resu lting innovations. In p articular,it questions the vedidity of the common claim that independentresearch units have more incentives to pursue radical innovationstha n th e research division of a custom er.Last, Section V summarizes the main insights of the paper,and discusses some directions for res earch.

    II. TH E BASIC FRAMEWORKILL Research TechnologyA research unit {RU performs research for a customer (C).The value of innovation for the custom er is V > 0. Th e probabilityof discovery,/j(e,J), depe nds on tbe no nco ntrac tible effort cost byR U and on tbe inv estment Eby C The probability is increasing andstrictly concave in{e,E .We will also assum e thatpie,E < 1 in therelevant range and that the marginal productivities of effort andinvestment at their zero level are infinite so as to guaranteeinterior solutions (for strictly positive incentives). The minimumlevel of effort o fth e research un it, th at is , th e level of effort inducedby its res ear ch ers intellectual curiosity, ego, career concerns, andprospects of informal rewards, is normalized to be 0. So is tbeminimum level of investment by tbe customer. We will make twoopposite assum ption s concerning the custo m er s investm ent. Inthe first caseE is monetary and contractible. In the second case Estands for proprietary technological information freely supplied tothe research un it, or for interaction w ith the research un it to tailorthe innovation to the final demand;E will then be assumed to benoncontractible.We would of course expect a mixture of contract-ible and nonc ontractible investm ents in reality. The results for thetwo cases are most often identical, and so we will state the resultswith both cases in mind, unless they differ, in which case we willnote tbe points of dep arture .

    Without loss of insights, we posit a separable form for thetechnology:p{e,E = qie + r{E . Our theory c^n be straightfor-wardly extended to nonseparable tecbnologies. The new feature isthe n th at th e optimal specification ofth e custo m er s investm ent (ifit is con tractible) refiects its influence on tbe reseeirch un it s effortthrough complementarities or substitutabilities in the production

  • 8/13/2019 1. the Management of Innovation 1994

    5/27

    THE M N GEMENT OF INNOV TION 1189Let us define the socially optimal or first-best) effort and

    investment e*{V and.B*(V) bymax{pie,E)V- e - E],{e,E\

    orq ie*iV))V=r iE*{V)W= 1.

    11.2.ContractsA natu ral way of introducing property rights con siderationsintb e analysis ofR Dactivities is by postula ting th e incom pletenessof rese arch con trac ts. More specifically, and th is is ou r second basicassumption, we posit thattheexact n atu re ofthe innovation is illdefined ex ante , so th at the two pa rties cann ot contractfordeliveryof a specific inn ova tion. Th e co ntract only specifies the allocation oftbe property right on any forthcoming innovation,asharing ruleon the verifiable revenu e (license fee) obtained by the researchunit , and any verifiable amount of customer investment.^ Therealized valueofthe innovation for thecustomer is noncontract-ible;tba t is, eitberit is aprivate benefit itreducestbecustomer sproductive effort, say),or it is a monetary benefit that cannot berecovered amongthecustom er s many activities. (Assuming th atthis value is contractible would not affect the analysis in tbeabsence of thir d parties.)If property righ tsontheinnovation areallocated toC, the n Ccan freely usetheinnovation.In that case RU receivesnorewardfor innova ting.^ This we shall refer to as C -ownershipor integratedcase.If RU owns the innovation, C andRU bargain over tbelicensing fee once the innovation has been made. T his we call

    2. Allowing for announcement games between the two contracting partiesafter therealization of thestateof nature (i.e., after theinnovation haseitheroccurred or not occurred) need not destroy the property rights interpretationdevelopedinthe paper.Inparticula r, one can show th at, inthe context of this basictwo-party co ntracting m odel, the optimal complete contract can be implemented bya random allocationofproperty rights between RU and C, provided that we allowfor ex post renegotiation after the innovation h as occurred. This result, however,ingeneralis notrobustto theintroduction of third (cofinancing) parties, unlessweprecludeanynet transfer from theprincipalto such partiesin thestateofnaturewhere no innovation occurs.3 . To be certain, there exists some noncontrac tual orinformal sharin g intha t

  • 8/13/2019 1. the Management of Innovation 1994

    6/27

    1190 QU RTERLY JOURN L OF ECONOMICSRU-ownershipornonintegrated case.(So we assunie that C is theonly user of the innovation and is therefore indispensable for therealization ofthe value of innovation. This assumption is relaxed inSections III and IV.) For simplicity, we assume that the total pieVis then equally split ex post between the owner RU and thecustomer C,so that C pays a license fee equal toREMARK. By focusing on ownership of the innovation, we have

    ignored the possibility ofi?J7owningC.Suppose thus that C isa division owned byRU.We will then assume that ownershipofCdoes not imply thatRU can forceCto produce and that C sindispensability in tbe production process allows it to extractV/2 fromRUby threatening not to produce.^ RU then gainsnothing from owning C instead of only the innovation, and ourfocus on the ownership of innovation involves no loss ofgenerality. Alternatively, we could invoke RU s cash con-straint to rule out its owning C.

    11.3.Who Should Own the Innovation?This subsection assumes away tbe possibility of C s equity

    holding ini?U^(see subsection II.4 below). Whether C ori?C/shouldown the innovation hinges on two basic considerations: (a) themarginal efficiencyofRU s effort compared with the marginalefficiency of C s investment; (b) tbeex ante bargaining powerof tbetwo parties (who proposes tbe initial contract), which refiects theextent to which the research unit is the only candidate to performthe research. Tbe importance of ex ante bargaining assumptionsfollows from utility not jdways being ex ante transferable betweenRUand C. More specifically, tbe casb-constrainedRU isunable tocompensate C for a transfer of ownership toRU, even if such atransfer results in a higher total surplusipV - e E).

    Consider first C-ownership. RU then receives no reward forinnovating and therefore supplies no effort: e = 0. On tbe other

    4. The equal split outcome will,forexample, result from a Rubinstein [1982]bargaining process with alternative offers by the twoparties and no time delaybetween two successive offers also see Stahl [1972]). Focu sing ontheequal splitcase involves no loss of insights in th is model where utility in the ex post barg aininggame is treuisferable in th e releva nt range where RU a income is positive). For theanecdote, the industry ruleofthumb is that the innovator receives between20percent an d 50 percen t ofth e pie [Cavesetal.1983;Bartonetal.1988].5. This is in thespirit ofHart and Moore [1994].Our assumption maybe

  • 8/13/2019 1. the Management of Innovation 1994

    7/27

    THE MANAGEMENT OF INNOVATION 1191band, C has appropriate incentives to invest. It maximizes[p{O,E)V - E]and therefore choosesE =E*(V).Utilities are thenpossibly up to a lump sum transfer from Cto RU)1) URU=0and

    Uc=p{O,E*iV))V - E*iV).Under RU-ownership,each receives V/2 in case of innovation.

    RU maximizes [pie,E)V/2 - e ] and so chooses e = e* V/2). Ifeither E is noncontractible or E is contractible and C bas thebargaining power ex ante,^ C chooses its investment so as tomaximize [p(e,E)V/2 - E], resulting in E =E*iV/2). Note thatunderinvestment by C occurs even when E is contractible, becauseRU cannot ex ante compensate C for an increase in investment.Utilities are then possibly up to a lump sum transfer from C toRU)

    and

    How is the property right determined? Clearly,Umj > URU-i?[7 prefers having the property right.URU s effort is importantenough thatUc > Uc,tben the property right is allocatedto RU.IfUc < Uc,the allocation of the property right depends on the exante relative bargaining strength.lfRU bas tbe bargaining power

    6. IfEis contractible andRUhas the bargaining power exante, RUwilldemandalevel of inve stme nt Etogether with acash tran sfer afrom C toRUsuchtha t C breaks even:a E=p{e*iV/2),E)V/2.

    Assuming thata > 0, i?f7 chooses Eso as to meiximize

  • 8/13/2019 1. the Management of Innovation 1994

    8/27

  • 8/13/2019 1. the Management of Innovation 1994

    9/27

    THEM N GEMENT OF INNOV TION 1 1 9 3Suppose th at C is given (1 - a) share s in RU. The real licensefee paid by C wb en b oth partie s agree on a nom inal license fee I is

    J I - {1 - a)l = al. T be equ ilibrium level of Zis, as before v dtb I,driven by tim e impatience only and not by th e sh aring ru le impliedby th e initial con tract. N othin g is tb u s affected by th e introdu ctionof an e quity participation of C inRU,^ as anticipated by Hart andMoore [1990, footnote 7]. (A similar reasoning shows that RU sbeing given a share in C has no effect on the net license fee and istherefore irrelevant.)This irrelevance result generalizes to nonlinear sharing rulesas long as the two parties can by m utu al consent tea r up th e initialcontract and renegotiate. The initial sharing rule cannot influencethe ex post bargaining game between R Uand C since in con trast tosome other incomplete contracts models with renegotiation,^ theobject (innovation) to be traded ex post is not ex an te contractible.In contrast with the above irrelevance result, introducingthir d partie s (outside investors) as co-owners ofi?{/ can help raiseC s profit un de r RU ownership and hence may occur if C hassubstan tial bargaining power ex ante .Suppose tberefore that C bas tbe ex ante bargaining power.Assume furtber that investment is contractible (this is not crucialfor the argument). C can then demand cofinancing Ei in exchangefor a claim of a fraction 1 - a) oiRU s profits. As before,RU and Cbargain over tbe licensing fee after the innovation occurs and Cmust still pay an observable license fee equal to V/2 to RU sowners. However tbe researcb ers themselves now receive a r etu rnof aV/2 for the innovation, whereas the outside investors receive(1 - a)V/2 . Note tha t in th e bargaining process the investors andthe research unit have congruent interests, namely to extract asmuch from the customer as is possible; therefore the investorshave no incentive to enter the bargaining process, to collude with

    8. We do not wish to imply that customers should never take equity in theindependent research units they sponsor. Equity participation here does not raisethe cu stom er s inve stme nt since it has no effect on the real transfe r price. Bu t itcould affect o ther m oral hazard com ponen ts of the cus tom er s activity. For instance,in the presence of alternative customers (see subsection III.3), if there wereappropriability problems so that the customer could resell the technology to othercustomers, an equity participation in the research unit would mitigate thecus tom er s incentive to expropriate t he reseeirch un it [Rodriguez 1992]. It mig htalso soften future competition between the research unit and the customer (seeAghion and Tirole [1993, section 6]).

  • 8/13/2019 1. the Management of Innovation 1994

    10/27

    1194 QUARTERLY JOURNAL OF ECONOM ICSth e research unit.^ Assum ing ex an te competition among outsideinvestors leads to the following free-entry condition, where Ej(respectively, .Ec) denotes th e outside inv esto r(s)' (respectively, thecustomer's) investment:

    \ \

    C's profit is thena

    where E = Ec -^Ej is total investment. The optimal total invest-ment for the customer is .B = E*{(1 - a/2)V). Cofinancing thusallows the customer to give the research unit any fraction of thevalue of the innovation between 0 (C-ownership) and V (purei?[/-ownership). W ith contractible, perfectly substitutable invest-men ts, co financing transforms a discrete choice of governancestructure into a continuous one.^^Note that our argument relies on the license fee beingverifiable. For, the customer and the research unit would have anincentive to collude aga inst th e thi rd p arty by specifying a sm all feeand making an additional transfer on tbe side. In view of tbe

    10. The lack of incentive to collude comes from the linear sharing rule. Thecorporate finance literatu re has dem onstrated that, in the absence of renegotiationor collusion, a nonlinear contract between an investor and an agent, such as a debtcontract, can strengthen the agent's bargaining position with a third party.Nonlinear contracts, on the other hand, are sensitive to the possibility of secretrenegotiation between the investor and the agent. Indeed if the research unit andthe custom er barga in toge ther, it is optirnal for the resea rch un it and the inves tor tosecretly r eneg otiate tow ard a lineeir snarin g rule so as to obtedn con gruence.We should also point out that our analysis carries over to the situation wherethe investor is drawn into the bargaining process (indeed in a generalization of thealternative-move model with sho rt periods, the research u nit's payoff aV/2, is notaffected by this possibility: see our discussion paper) .11. An altern ative way of achieving the sam e continu um [O,' ] of governancestructures on the innovation would be through the random allocation of theproperty right between RU and C. However, there are several reasons whyiJ [/-ownership with cofinancing hy a th ird peirty cannot be d uplicated by a ran domallocation of authority between RU an d C.One reason m ight be tha t C-ownershipdiscourages RU s initiative excessively in situations where RU s current effortaffects not only the occurrence ofthe current innovation but also the occurrence orvalue of future innovations. (See Aghion and Tirole [1993, section 6] for aformalization of this argument in the context of a comparison between researchjoint ventures and fully vertically integrated structures.) Another, more straightfor-

  • 8/13/2019 1. the Management of Innovation 1994

    11/27

    THE M N GEMENT OF INNOV TION 1195prevalence of shareholdings by outside shareholders or headquar-ter s in th e real world, we do not find the verifiability assum ptionunr ealistic, althoug h tb e collusion argu m en t indicates one possiblelimit in cases where the customer and the research unit have theoppo rtunity to make large hidden side tran sfers.Th e choice of a is based on two con siderations. F irst, a lower aallows cofinancing and thus reduces the customer's investmentburden. Second, the dilution of RU's share reduces its incentivesand therefore the probability of discovery. UsingRU's first-ordercondition (q''(e*(aV/2))aV/2 = 1), the derivative of C's profit withrespect to a is equal to [ pV/2 + [(2 - a)/am-q'/q )]. Th e firstterm in this derivative is therent extractioneffect and the secondthe incentive effect. That a = 0 is never optimal implies thatC-ow nership is no t optimEil either. ^ Pu re cus tom er financing(a = 1) may or may not be optimal depending on tbe size of tbeincentive effect. If effort is quite sensitive to dilution (tbat is, ifq lq' is small), th en a = 1 is indeed optim al. But if effort isrelatively inelastic, tbe rent extraction effect dominates and co-financing occurs. Tb e am ou nt of investor financing for the optimala* is th en given by^^

    V

    II.5. Schumpeterian HypothesesTh e second most tested set of byp otheses in ind ustria l organi-zation (after the cross-sectional analysis of tbe structure-conduct-performance paradigm) relates R&D input (R&D expenditures orpersonnel engaged in R&D) or outpu t (as measured by the num berof significant innovations) to variables th at presum ably alter th eincentives forR&D.Thescale effectstates som ewhat vaguely th at a l ar ge r firm has m ore incentives for R&D. A first inte rpr etatio nof this scale effect is that a larger market for a good that benefitsfrom a process innovation raises the value of the innovation. The

    12. However, C-ownership can again become optimal when the custom er sinksnoncontractible investment such as advice on design or release of proprietarytechnological information. Th en C-ownership cannot be duplicated by i? [/-ownershipwith cofinancing at a = 0 if.E;a ndE c are not perfect substitutes in the productionfunction. A move to i? [/-ownership reduces th e cu stom er's inve stme nt, and thisinvestm ent c annot be perfectly offset by an increase in the in vesto r's financing.

  • 8/13/2019 1. the Management of Innovation 1994

    12/27

    1196 QUARTERLY JOURNAL OF ECONOM ICSrelevant explanatory variable is then the size of the business unitratber than the size of the firm. An alternative view takes the sizeof th e firm a s the relevan t em pirical variable. Relatedly, the scopeeffect (Nelson [1959]) posits that a more diversified firm exploitsinnovations more easily than a specialized firm and "therefore"has mo re incentives to innovate. The market powereffectpresumesthat firms with market power gain more from an innovation. Howdo these hypotheses fit in our framework? Let us assume that thescale, scope, and m arke t pow er effect all boil down to a change in V.(See Aghion and Tirole [1993] for a brief discussion of tbis.)

    First, one can study how R&D inputs and output react to anincrease in th e value of innovation, taking the organizational formas given (in the case of cofinancing, the investor's share in theresearch unit is kept fixed). It is easily shown that effort, invest-m en t, and prob ability of discovery all increase with V. *However, these monotonicity results may not hold any longeronce tbe organizational form or ownersbip structure of R&Dactivities is endogenized. For, as th e value of innovation increases,C may either insist on keeping or acquiring property rights (itbecomes even more imp orta nt for th e custome r to fully ca pture th ewhole value of innovation) or instead want to voluntarily relin-quish property rights to RU (it also becomes more important toraise the researcb unit 's incentives through i?[/-ownership). TheAppendix illustrates this trade-off with two simple examples, onewhereRU s effort co ntrib utes proportionally more to th e probabil-ity of discovery th an C's inv estm ent as V increases, and a (polar)example where C's investment becomes relatively more importantas V increases.

    Consider now a switcb from C-ownership to i?[7-ownersbip.The standard R&D input measure, the customer 's monetaryinvestme nt, declines. The o utp ut m easure, here t he probability ofdiscovery, on the other hand, may well increase due to improvedincentives of the research unit. Input and output measures thenmove in opposite directions because part of the input cannot bemea sured. 1 Fu rtbe rm ore , tbere is no clear relationship betweeneither measure and the value of the innovation and therefore theSchumpeterian parameters.14. Indeed, under C-ownership,E = E*(.V), andp =riE*iV)) withdE*/dV >

  • 8/13/2019 1. the Management of Innovation 1994

    13/27

    THE M N GEMENT OF INNOV TION 1197II. 6. The Long-Purse Effect

    O ur analysis so far ha s assum ed away the existence of cash orcredit constraints on the customer's side. Introducing such con-straints into our framework has obvious implications for theoptimal ownership structure.Consider, for instance, the situation where C-ownership isoptimal provided that the customer can finance E*{V) withoutgoing on the capital m arke t and assum e now tha t the customer hasless tha nE*i{V)an d m ust borrow in a m arke t th at is imperfect forinformational reasons.^ External financing is then more costlyth an inte rna l financing a nd a m ove from C-ownership to i?C/-ownership may become attractive since it reduces tbe customer'sbenchmark (monetary) investment from E*(V) to E*{V/2), andthu s reduces the am oun t of cash C needs to borrow on the capitalmarket. Financial constraints therefore bias tbe organizationalform toward the use of creative inputs and away from capitalexpenditures. An interesting implication of this analysis is theprediction that new firms or firms which have experienced hardtimes will tend to feirm out their research activities more thanestablished, healthy firms. This is nothing but a refinement of tbewell-known long -purse hypothesis enunciated by Schum peter,according to which a firm's R&D investment should be positivelycorrelated witb its assets.

    III . SPLIT PROPERTY RIGHTSWe have analyzed the allocation of a single property right. Inpractice, the innovation may have more than one customer, orthere may be more that one innovation. There is then more thanone prope rty righ t to allocate, and p roperty righ ts canbe,and oftenare,split.

    III.l. Multiple Innovations: Contingent Property Rightsand the Hired for DoctrineWe observed in tbe introduction that botb emplojonent con-tracts and the law allocate property rights on the basis of bow

    much customer investment was used by tbe research unit and ofwhether tbe researcb unit bad been hired for the innovation and

  • 8/13/2019 1. the Management of Innovation 1994

    14/27

    1198 QU RTERLY JOURN L OF ECONOMICSmade it during normal working hours. We argue that thesecontingent property rights stem from incentive considerations.Coming back to tbe single-customer case, suppose that theeffort e and the investmentEcan yield one in a subset T C + oftypes of innovation. With probabilityXt,with S^^T Xt= 1, innova-tiont,with value V,, is the relevant one. Some types of innovationare consumed by tbe customer, some others are purchased at priceVtby alternative customers (so, tbe customer can be both acustomer and an investor in our terminology). Besides tbeir value,types of innovation differ in the extent they make use of thecustomer's investment. Namely, we assume tbat the probability ofdiscovery conditional on innovation t being the relevant one isp(e,E,t) = q(e) + tr(E). We also assume that e andE arechosenbefore the parties known which innovation is relevant, i Let adenote i?t7's share ofthe value ofthe innovation of typet.Note thatthe nature of innovation is contractible. Because both parties neverhave an incentive to overinvest (their individual stakes, neverexceed 1.tXt Vt ,an optimal contract must maximize tbe stake ofeither party in the innovation given tbe otber party'sstake. ^Wetherefore maximize C's incentive subject toRU sincentive exceed-ing some level:

    max2 ^(^(1~() V VRUteTand

    0 < a

  • 8/13/2019 1. the Management of Innovation 1994

    15/27

    THE MANAGEMENT OF INNOVATION 1 1 9 9in th e Na sh barg ainin g solution equa l to balf of C /s valuation.) Weth u s focus only on th e allocation of incen tives; the optimjdV uatisdeterminedbyth e sam e considerations (incentives, ex an te barg ain-ing powers) as in Section IL Th e solution to this pro gram satisfiesfor some oGT,

    t tf a=* a=> 0 t = (

  • 8/13/2019 1. the Management of Innovation 1994

    16/27

    1200 QUARTERLY JOURNAL OF ECONOM ICSbe obtained by following the lines of the previous subsection,replacing the n atu re of tbe inno vation by its date.More precisely, suppose th at th e effort e and th e inve stm ent Eproduce a tim e sequence of innovations with p rese nt value V( inwhich the customer's marginal efficiency decreases over time:p{e,E,t) =qie) + h{t)r{E), h {t) < 0. Let a, deno te 72 t/'s sha re of theinnovation discovered at date t .Because both p arties never have anincentive to overinvest,^ an optimal contract maximizes th e stakeof one party (say, the customer) for a given value of the otherparty's stake:

    subject to (1 - at)h{t)Vt

    The solution to this program satisfies, for some^o,t at = 0 (C-ownership)t > ta > at = 1 (7?[/-ownership).

    The optimal policy is thus very similar to the trailer clausesobserved in practice.777.3.M ultiple Users: A Rationale for Shop Rights

    This subsection derives foundations for the observed practiceof employers allocating ownership to employees while keeping aroyalty-free, nonassignable license for themselves. Consider thefollowing situation . An innovation designed and m anaged thro ug ha contract between a research unit RU and a customer Ci can expost be sold to a second, yet unidentified custom erC as well as toCj.Ci andC do not compete on tbe product m arket. Let Vi andVdenote tbeir valuations of the innovation, where Vj may refiect

    inventions made or acquired during employment or for one year aftertermination of employment. The agreement is also equipped with a trailerclause valid for two years after the termination of employment. . . . Theemployee will not con tribu te his knowledge to . . . any corporation orperson engaged in competition with P olaroid.At Gulf Oil, [employees in techn ical or scientific work] agree . . . for oneyear after employment not to engage in the same type of work for acompetitor within the same geographical area or territory. This secondpa rt o fth e [trailer] clause covers practically all employees who are expectedto make patentable inventions and improvements.

  • 8/13/2019 1. the Management of Innovation 1994

    17/27

    THE M N GEMENT OF INNOV TION 1 2 0 1some noncontractible investment E by Ci, that increases Ci'swillingness to pay for the innovation (e.g., Vi = VQ-I-E,whereVQisa con stant). We adopt th e convention t h at C2 purch ases theinnovation at priceV2.LetV R U andVc denote R U s mdCi's stakes in the innovation,th at is, the e xtra payoffs they ob tain wben inn ovation o ccurs. Onehas

    V R U + V C ^ = V,-^V2 an d V^^; < V i/2 + V .Let ai and a2denote7?IT Ssb are ofViandV2.W hen there is no

    custom er investm ent in raising the value of innovation {E = 0),a noptimal contract m us t in particula r solvemax{( l -

    subject toaiVi +

    and0 < tti < i , 0 < a2 :< 1.In th is program ai is inde term inate, as botb incentives depend onlyon R U s total sbare (aiVi + a2V2). However, wben Ci's inv estm entaffects not only tbe probability of discovery, but also its ownvalua tion Vi for the innov ation (E > 0), the optimal contract mustgive maximal incentives to Ci on its own use of the innovation. Inparticuleir, ai = 0 if02< 1. In wo rds, ifCican affect V ibut notV2,its relative sbare sbould be tilted as much as possible toward thefirst use. Indeed, if the optimal stake for the research unit, V R U ,

    does no t exceedV2,th e optimal co ntract gives a shop right toC ia ndallocates the licensing fees in proportionsa =V u/V2 toR U and1 - a2 to Ci. We th u s ob tain a rationa le for shop rights.22We sum m arize Section III with th e following proposition.PROPOSITION 2. In the presence of multiple users or multipleinnovations, the property rights may be split between thecustomer and tbe research unit. Each should get propertyrights on those activities for which it has a comparative

    advantage in creating value. This principle gives rise to shoprights, property rights based on the nature or the date of the

  • 8/13/2019 1. the Management of Innovation 1994

    18/27

    1202 QU RTERLY JOURN L OF ECONOMICSIV. PROPERTY RIGHTS AND THE SIZE OF INNOVATIONS

    This section investigates the relationship between propertyrigbts and tbe nature of the innovation: does an independentresearch unit have more incentives to pursue radical (drastic)versus incremental (nondrastic) innovations compared with thesame research unit within an integrated firm?

    A main motivation for analyzing this question is derived fromthe recent neo-Schumpeterian literature on endogenous techni-cal change and the process of economic growth. This literature, andparticularly the models with vertical innovations or qualityladders, expresses the long-run rate of productivity growth as afunction of both the frequency and the size of innovations.^^Another motivation is Henderson [1991]'s tests ofthe hypothesisthat establisbed firms would pursue incremental rather thanradical innovations (tests wbich yielded mixed results).

    In the previous sections we concentrated our analysis on therelationship between the organizational form of research and thefrequencyof innovations. Our emphasis in this section is on therelationship between the organization of research and thesize(ordrasticity) of technological or quality improvement brought aboutby the innovation.

    More precisely, we consider a single innovation and assumethat the probability of discovery decreases with the size of theinnovation. To focus on the choice of technology, we ignore in afirst step the inputs e and E in the notation and denote theprobability of discoveryhy piy). The parameter 7 > 1 indexes the size ofthe innovation or research line, with 7 = 1 correspond-ing to the existing technology. What was previously said aboutinputs still applies and of course infiuences the distribution ofproperty rights.

    Following tbe patent race literature, we refer to the currentcustomer C as the incumbent and to other potential customersas entrants. Let a j^iy) denote tbe incumbent's profit when beobtains an exclusive license for the innovation, and let TT = ^^(l)denote the incumbent's (monopoly) profit in the absence of innova-tion. Finally,171(7)is tbe profit of an entrant who has purchased anexclusive license. One has'TTI(7)

  • 8/13/2019 1. the Management of Innovation 1994

    19/27

    THE MANAGEMENT OF INNOVATION 1 2 0 3technology is not constrained by the competitive pressure of theincumbent. Since 7 indexes the size of the innovation, -ni^ -) andn-i(-) are bo th inc reasing .We now compare the sizes 7c and JRU of the innovation thatobtain when C (respectively, R U ) has property rights on theinno vation and chooses its size.^*UnderC-own ership we have

    7c = argmaxp (7)[ n-^(7) - TT^].Note that 7c is the efficient research line for the industry. By arevealed preference argument, the larger n , the more drastic theinnovation.^^ T his is a version of A rrow s [1962] celebrated rep lace-ment effect. The incumbent prefers a lower probabilitybigberpayoff research technology if its profit in tbe absence of innovationis high.Consider now R U -ownership (with or witbout cofinancing).Note that in the absence of potential entrants, the research unitgets a fraction of the value of the innovation in either case andtherefore chooses the technology preferred by C: yR u = 7c- Tbiscongruence between tbe researcb un it s and the incum bent spreferences disappears when the research unit can sell to anen tra nt. Although in equilibrium tbe research un it optimally sellsan exclusive license to the incumbent, an entrant acts as a threatand creates an appropriation effect.More precisely, if the innovation is licensed hy R U to apotential en tra nt (instead of being licensed to the incu m bent), tb een tra nt can obtain a license fee equal to -niiy). In othe r words, theresearch unit has the outside option to sell an exclusive license toan en tr an t a t price 171(7). A well-known re su lt in B inm ore et al.[1986] then shows that, under alternating-offers bargaining witbthis o utside option, tb e researc b u nit can obtain a license fee equa lto max ( (< (7) - iT ^)/2, 771(7)) from C.Suppose first that tbe optimal cboiceyRuby the research u nitis such th at

    24. In the case where property rights are split from control rights andtherefore C can impose the research line even though the ow nership of the

  • 8/13/2019 1. the Management of Innovation 1994

    20/27

    1204 QUARTERLY JOURNAL OFECONOMICThen, the outside option of selling to an entrant is irrelevant, andtherefore, as in the ahsence of a potential entrant, the research unitchooses the same technology (or technologies if there are multipleoptima) as the incumbent customer: -^inj= 7c- i

    Ontheother hand, ifTri(7Ri7) > ('Trf(7R(;)-T7y)/2, the researchunit's choice~iRuwill generally differ from the incunihent's 7c sincenow the outside option of selling to an entrant hecomes credibleand the incumbent must pay T^iiyRu to obtain; the license. Tocompare the customer's and the research unit's preferred researchlines, note that, by definition of in uand 7c,

    and

    Multipl3dng these two inequalities, we obtain

    Reintroducing (ownership-contingent) inputs, the latter inequalitystill holds as long as the discovery function is ''multiplicative,namely p(e,,7) =g{y h{e,E .We have established the followingproposition:PROPOSITION 3. Interests may diverge as to the choice of the

    research line when there are potential entrants. Let r{y =Tri(7)/(TT (7) - 17 ) denote the appropriability ratio or relativewillingness to pay of an entrant with respect to the incumbentfor the innovation. Then, under the multiplicative discoveryfunction, yRu> ycif'' is strictly increasing, andy/m < ycif r isstrictly decreasing.

    In other words, if the relative willingness to pay of a potentialentrant increases with the size of the innovation, then the researchunit tends to choose a more drastic innovation than the incumbent,since by doing so it is able to appropriate a higher fraction of thesurplus of the incumbent. Similarly, if the relative willingness topay of a potential entrant decreases with the size of the innovation,the research unit appropriates more of the surplus by choosing a

  • 8/13/2019 1. the Management of Innovation 1994

    21/27

    THE MANAGEMENT OF INNOVATION 1205and strictly decreasing over another, then it is possible to find tworesearch technologies p( ) andj5() such that (with obvious nota-tion)-ynu .

    Example: Process innovation in a homogeneous good industry.The incumbent monopolist produces at marginal cost Co- Thepotential innovation is a process innovation that reduces themarginal cost to c < Co. So 7 = Co/c . If an entrant purchases theexclusivelicense,the two firms wage Bertrand competition. LettingDip) denote the demand curve and p' {c) denote the monopolyprice for costc one has (abusing notation slightly)dx

    -n-i(c) = min {I>(co)(co - c),I>(p' (c))(p' (c) - c)).As is well-known, an entrant has more incentive to innovate thanan incumbent due to the replacement effect, so r > 1. A fortiori,iri(c) > (TT7(C)- 'n^)/2, so that the outside option is bindingregardless of the size of the process innovation. Furthermore, rdecreases from Z)(co)/i)(p' (co) to 1 as c decreases(yincreases .In ahomogeneous good industry, an independent research unit willpursue less drastic innovations than an integrated one. On theother hand, our working paper [1993] analyzes two other standardmodels of industry behavior, respectively, with horizontal andvertical product differentiation, that yield ambiguous conclusionsas to the monotonicity of the appropriability ratio. There istherefore no general conclusion about the impact of the ownershipstructure on the size ofinnovations.

    V. SUMMARY AND DIRECTIONS FOR FUTURE RESEARCHManaging innovation properly is one of the most importantchallenges faced by developed economies. This exploratory paperargues that property rights analysis offers a conceptual frameworkto understand the organizational aspects of R&D activities and

    the innovation is minor, ind conversely when the innovation is drastic. On the onehand, the replacement effect implies that the iricumbent is not in a hurry toinnovate. On the other hand, competition destroys industry profit and therefore theincumbent gains more from remaining a monopoly than an entrant from becominga duopolist. The latter efficiency effect is absent here, since the research uriit sells to

  • 8/13/2019 1. the Management of Innovation 1994

    22/27

    1206 QU RTERLY JOURN L OF ECONOMICStheir implications in terms of the frequency and size of innova-tions. Let us summarize our main insights. First, research willmore likely be conducted in an integrated structure if (a) capitalinputs are substantial relative to intellectual inputsin contrast,when intellectual inputs dominate as for software and biotechnol-ogy, research will often be performed by independent units; (b) thecustomer has more bargaining power ex ante,; say because ofintense competition among potential research teams; (c) the cus-tomer has a deep pocket. Otherwise, research activities are morelikely to be performed by nonintegrated research units. In thatcase,cofinancingby an outside investor (venture capitalist or bank)may benefit the customer of the innovation. Second, when thereare multiple innovations or multiple customers, property rightsmust be split on the basis of comparative advantage in creatingvalue. This principle gives rise to shop rights, the hired fordoctrine, and trailer clauses. Third, the drasticity or size ofinnovations is also affected by the organization of research. How-ever, to predict whether an independent research unit wants topursue more drastic innovations than an integrated one requiresdetailed knowledge of the industry, and in particular of theappropriability ratio.

    The potential of the property rights analysis for studyinginnovation management is broader than is suggested by the limitedscope of this paper.

    First, except in our brief discussion of breakaway research inSection III, we have focused on thestaticmanagement of innova-tion. However, the dynamic aspects of the organization of researchactivities are also important, and in particular they underlie the(business economics) debate on the relative merits and drawbacksof vertical Research Joint Ventures (RJV) ' vis-a-vis (perma-nently) integrated structures (or direct equity participations).RJVs have specific objectives and are generallylimited in scope orin time The short-term horizon of the RJV matters when theresearch unit cannot protect its intellectual property. By releasingits technological knowledge to its partner, the research unit raises

    27, Such RJVs involve independ ent researcher s (or inventors) and ma nufactu r-ing firms (developers), where the former contribute their knowledge and the lattercontribute capitalorother productive inputs. Theseare to bedistinguished from

  • 8/13/2019 1. the Management of Innovation 1994

    23/27

    THE MANAG EMENT OF INNOVATION 1 2 0 7the probability of success of the joint venture, but also creates itsown competition in the fu ture. RJVs are th u s no t very conducive totechnology transfers, although future competition can be softenedby letting the custom er take an equity participation in the researchunit. Vertical integration reduces the research unit 's incentive tohold back the transfer of knowledge by reducing its payoff fromhaving exclusive knowledge of the technology tomorrow. One canshow, however, that integration imperils future technologicalprogres s (see Aghion and Tirole [1993]).Second, government promotion of R&D is one of the mostimportant areas of public policy. Analyzing the government as acustomer, an investor, or a benefactor (depending on the circum-stances) ought to shed light on efficient ways of channelinggovernment m oney into R&D.Third, we have not considered competition among researchteam s. It would be fruitful to merge the prop erty righ ts approach ofthis paper and of the literature on strategic vertical integrationtogethe r with the tradition al paten t race analysis.Las t, we believe th at ou r andysis provides some microfounda-tions for extending the new growth litera ture in several in terestingdirections. In particular, it may help to introduce financial andorganizational considerations into the neo -Schu m peterian frame-work and also enrich o ur cu rre nt views on technological innovationand diffusion within sectors and ind ustries as major de term inan tsof productivity growth.

    APPENDIX: THE AMBIGUOUS EFFECT OF A CHANGE IN THE VALUEOF INNOVATION ON THE ORGANIZATIONAL FORM

    Example 1 an Increase in VM akes RU Ow nershipMore D esirable)Suppose that the probability of discovery depends only on theresearch unit's effort: riE) = ro > 0for allE. ThenERU = EC = O.Suppose further thatq{e = 2e/Vo for 0 < e < 1, andq{e = 2/Vofor e > 1. For V < Vo, RU chooses e = 0 even under purei?[/-ow nersh ip, since (2e/Vo)(W2) - e < 0 for alle > 0. The refore,

    C prefers C-ownership to pure i?[/-ownership since roV > roV/2,and more generally to cofinancing. For V > VQ , C prefers i?[/-

  • 8/13/2019 1. the Management of Innovation 1994

    24/27

    1 2 0 8 QUARTERLY JOURNAL OF ECONOM ICSto Vo (any nonnegligible a m ou nt of equity ta ken by an investorthen destroys RU s incentives). We th us conclude tha t a t Vo th eorganization al form jum ps from C-ownership to i? [/-ownership.Example 2 (an Increase in VM akes C-Ow nership More Desirable)

    Assume that the customer 's and the investor 's investmentsare not substitu tes. Indeed, th e custom er's investment is nonmone-tary and noncontractible w hile investor's investm ent is useless. Weposit that r{E) = 2M,V/Eandqie =2e/Vo fore < Vo/4 and ^(e) = V2fore > Vo/4.We assum e th at V > Vo.Under C-ownership, the customer chooses Eso as to maxim ize2[L\JEV- E,jdelding payoff |x^V^. Under pure i?[/-ownership, RUchooses e = Vo/4. C chooses E so as to maximize [V2 +2JJLV/E]V / 2 - E, yielding payoff V /4 -I- (i,2y2/4 Suppose tha t thes etwo payoffs ar e equ al:

    V

    Then

    Thus,an increa se in ma kes C-ownership optimal at = V*.Until now, we have ruled out cofinancing. As in example 1,having an investor take equity in RU (and give a lump sumpaym ent to C) is optima l for C bu t cann ot really be distinguishedfrom pure i2[/ ownership if V is close to Vo. Any nonnegligibleequity of an investor in RU then destroysRU s incentives.Last, note that at = V*,p^^ > p^. Therefore, an increase in Vat V* discontinuously reduces the probability of discovery, whilediscontinuously raising the inpu t E.N U F F I E L D C O L L E G E , O X F O R D , A ND E U R O P E A N B A N K F O R R EC O N S T R U C T I O N AN D D E V E L -O P M E N T , L O N D O NI N S T I T U T D ' E C O N O M I E I N D U S T R I E L L E , T O U L O U S E , M A S S A C H U S E T T S I N S T I T U T E O F T E C H -NOLOGY, CAMBRIDGE AND CERAS, PARIS

    REFERENCES

  • 8/13/2019 1. the Management of Innovation 1994

    25/27

    THE MANAGEMENT OF INNOVATION 1 2 0 9Aghion, P., and J. Tirole, O n the Managem ent of Innov ation, working paper, MITand Nuffield College, Oxford, 1993.Aghion, P., M. Dew atripont, and P. Rey, Reneg otiation Design with UnverifiableInformation, Econometrica,LXII (1994), 257-82 .Arrow, K., Econo mic Welfare and the Allocation of Resou rces for Inventio ns , inTheRcite and Direction, of Inventive Activity, Richard Nelson, ed. (Princeton,NJ: Princeton University Press, 1962).Barto n, J., R. Dellenbach, and P. Ku ruk, Tow ard a Theory of TechnologyLicensing, inStanford Journ al o f International L aw, Symp osium on the Lawand Economics of Intern ation al Technology Licensing, XXV (1988), 195-229.Binmore, K., A. Rubinstein, an d A. Wolinsky, Th e N ash Ba rgaining Solution inEconomic M odeling, Rand Journal of Economics,XVII (1986), 176-88.Brodley, J., Jo int Ven tures and A ntitr ust Policy, Harvard Law Review,LXXXXV(1982),1523-90.Caves, R., H. Crookell, and J. P. Killing, Th e Imperfect M arke t for TechnologyLicenses, Oxford Bulletin of Economics and Statistics, XXXXV (1983),223-48.Fisher, F., and P. Temin, R etu rns to Scale in Research and Development: W hatDoes the Schumpeteritin Hypothesis Imply? Journal of Political Econo my,LXXXI (1973), 56-70 .Gilbert, R., and D. Newbery, Preem ptive P aten ting and the Persistence ofMonopoly, American Economic Review,LXXII (1982), 514 -26.Grossm an, G., and E. Helpm an, Qu ality Ladders in the Theory of Gr ow th,Review of Economic Studies, LVIII (1991),43-61.Grossm an, S., and O. Ha rt, Th e C osts emd Benefits of Ow nership: A Theory ofLateral and Vertical Integration, Journal of Political Economy, LXXXXFV(1986),691-719.Ha rt, O., and J. Moore, Incom plete Contracts and Ren egotiation, Econometrica,LVI (1988), 755-86 .Ha rt, O., and J. Moore, Prop erty Rights and the Nature of the Firm , Journal ofPolitical Econom y,XCVIII (1990), 1119-50.H art, O., and J. Moore, A Theory of Debt Based on the Inalienability of Hu manCapital, Quarterly Journal of'Econom ics,CIX (1994), 841-7 9.Hend erson, R., Un derinv estm ent and Incompetence as Responses to RadicalInnovation: Evidence from t he Photolithog raphic Alignment Equipm ent Ind us-try , mimeo, Sloan School, MIT, 1991.Holm strom, B., and P. Milgrom, M ultitask Principal-Agent Anedyses: IncentiveContracts, Asset Ownership, and Job Design, Journal of Law, Economics andOrganization,VII (1991), 24 -52 (special issue).Laffont, J. J., and J. Tirole,A Theory of Incentives in P rocurement and Regulation(Cambridge, MA: MIT P ress , 1993).Nelson, R., Th e Simple Economics of Basic Scientific Res earch, Journal of

    Political E conomy,LXVII (1959), 297-306.Neumeyer, F.,The Em ployed Inventor in the United States: R Dpo licies. Law andPractice(Cam bridge, MA: M IT Press , 1971).ReingEuium, J., Un certa in Innovation and the Persisten ce of Monopoly, Ameri-can Economic Review,LXXIII (1983), 741 -48 .Rodriguez, D., The Gro wth and Diffusion of Knowledge and the Th eory of th eFirm , mimeo, MIT, 1992.Rubinstein, A., Perfeci; Equilibrium in a Bargaining Model, Econometrica, L(1982), 97-1 10.Schumpeter, J.,Capitalism, Socialism and emocracy(NewYork Harper, 1942).Segestrom, P., T. An ant, and E. Dinopoulos, A Schum peterian M odel of theProduct Life Cycle, American Economic Review,LXXX (1990), 107 7-92 .Stflhl, I.,Bargaining Theory(Stockholm: Stock holm School of Economics, 1972).von Hippel, E.,The S ources of Innovation(Oxford: Oxford University Press, 1988).

  • 8/13/2019 1. the Management of Innovation 1994

    26/27

  • 8/13/2019 1. the Management of Innovation 1994

    27/27

    1994 President & Fellows of Harvard University