Property Rights, Firm Boundaries, and R&D Inputs Ashish Arora and Robert P. Merges Carnegie-Mellon University; U.C. Berkeley Contact: [email protected] Draft. Do not cite or quote without permission. ª 2001 AA & RPM Abstract This Article offers an explanation of the role of intellectual property rights (IPRs) in information-intensive vertical supply relationships. In particular, we explore the connection between stronger property rights and the enhanced viability of independent (versus vertically integrated) input supply firms when contracts are incomplete. We start by modeling a tradeoff between two types of information transfer in buyer-supplier relationships: “synergies,” in which joint efforts reveal new applications of existing technology; and “leakage,” or disclosure of existing information. We show that property rights in the hands of an independent input supplier can create the potential for greater inter-firm synergy, outweighing the risk of leakage. Greater synergies arise due to the supplier’s greater effort to adapt its generalized technology to the specific needs of the buyer. Property rights play a crucial role: they reduce the risk of buyer firm opportunism, in effect raising the cost of the buyer’s “outside option” in the event the supplier- buyer contract is terminated. The “residual” nature of property rights as described for example by Hart (1995) makes them more effective in this regard than contracts alone. We extend our basic results to analysis of buyouts and spinoffs, and assay an extensive body of empirical evidence. Broad support is found for our approach, pointing the way to future exploration of the relationship between property rights specifications and the opening up of new contracting horizons. 1 1.0 Introduction: The Story This Article offers an explanation of the role of intellectual property rights (IPRs) in information-intensive vertical supply relationships. We analyze how IPRs affect the tradeoff between high-powered incentives and information spillovers, and show that under plausible conditions, they favor the provision of information intensive inputs by an independent supplier. Suppose a manufacturing unit MU has a specialized production process utilizing Input Q. Suppose further that Q is a technologically sophisticated component of MU’s production process. In addition, Q is the unique specialty of scientists and engineers who comprise research unit RU. One way for MU to get the input is to employ RU – vertical integration. In the alternative, members of RU could found an independent firm to supply Q in an arm’s-length contract with MU. In making its decision, MU faces a well-recognized tradeoff pitting the possibility of opportunism against high-powered arm’s-length contracting (Coase, 1937; Williamson, 1996). The effort invested in customizing Q may be non-contractible so that RU may shirk. On the other hand, once expended, the effort is sunk, leaving RU vulnerable to reneging by MU. This is a familiar tradeoff in the literature of transaction cost economics. But the details of our story, drawn from the information-rich exchanges we describe, add some important theoretical twists, as formalized in section 2. In supplying Input Q, RU will have to interact extensively with the staff of MU. This entails considerable information exchange. In general, the information in this relationship is of two types: (1) leakage of pre-existing information held by each party; and (2) synergistic generation of new information. Leakage is straightforward: there is almost always information exchange in supplier-buyer transactions; this is particularly true where MU’s 2 production process and RU’s input are technologically complex.1 While it is difficult to quantify “typical” rates of information exchange in buyer-supplier contracting, legal disputes and practitioner guidance relating to these transactions provide some insight (MacLachlan, 1995). Theft of trade secret cases arising from this context are common. In addition, lawyers often advise clients to contemplate the degree of information exchange that may accompany a supply contract, and to take precautions to prevent undesired leakage (Pooley, 1999: 634). These are of course purely informal measures. But they do indicate that the issue is a real one. Synergistic generation of new information is also very common in the kinds of input transactions studied here. The empirical basis is presented in section 3, where we discuss the property rights allocations and contractual provisions real-world parties often craft to deal with leakages and synergies. In general, RU-MU interactions can generate useful information about potential new applications of Input Q, perhaps in the production of other products besides the one contemplated by the original contact. Or RU may learn about opportunities to add to the Input Q product line. For example, it might learn of a way to expand Input Q’s functionality to replace other inputs in MU’s production process.2 1 R&D Managers are well aware of this problem. See, e.g., Ragatz, Handfield and Scannell (1997: 199) (noting fear of information sharing in new product development outsourcing agreements); Stump and Heide (1996) (describing techniques for “controlling supplier opportunism.”). For a good overview, see Handfield, Krause, Scannell and Monczka (2000). 2 Management literature shows an awareness of these opportunities for synergy. Stuart and McCutcheon (2000: 35) (“They [suppliers] are in on the engineering meetings. They can drop in on the research guys. They know more about our requirements than some of our own people do and are instrumental for concurrent engineering of new products.”); McCutcheon, Grant and Hartley (1997: 275) (empirical study of 79 cases of outsourcing involving new component or product design: “Increasing the role of the supplier in design enables the buyer to tap more effectively into the ideas of the supplier for product improvements.”); Ragatz, Handfield and Scannell (1997: 200) (summarizing industry experience integrating suppliers into new product development, and finding that the greater the sharing of “intellectual assets” among the partners, the greater the degree of success of the product. Sen and Rubinstein (1989: 130) find, in a study of 31 technology outsourcing contracts, a “high level of R&D involvement by the buyer firm,” which includes “new uses, new applications, and new products”; and (1989: 125) note that “flowback [i.e., grantback] clauses may be necessary in outsourcing contracts because of the likelihood that the parties may “improve the acquired technology” during an outsourcing agreement. 3 These two types of information spillovers – leakage and synergies – are at the heart of the tradeoff modeled in this Article. By choosing to integrate (i.e., by owning RU), MU prevents leakage of information about its products and processes. Greater control over disclosure of internal information is a well-recognized feature of the employment relationship, as compared to independent contractor status (Masten, 1996). In addition, the law by default vests a firm with ownership of employee inventions, thus allocating to the employer the residual that accompanies ownership of property rights (Merges, 1999). Integration also internalizes the benefits of synergistic information. In the absence of integration, both RU and MU may base future products on the information generated in the supply relationship. Or they may compete in the market for this information per se, as rival licensors. In either case, (some of) the rents made possible from the new information will be competed away.3 As the literature on technology pioneers and improvers shows, integration prevents rent dissipation (Scotchmer, 1991). On the other side of the ledger, an independent RU has certain information-related advantages. One is obvious: the canonical “high-powered incentives” that flow from arm’s- length contracting. This leads in turn to a second, more particular to our context: RU’s increased efforts create more synergistic information. The combination of independence and property rights over synergistic information induces RU to work hard to expand the frontiers of Input Q. New applications and extensions of the technology become more likely. Because an independent RU can directly appropriate the value of the new applications, RU team members will work harder to uncover them.4 Further, though we do not model it, an independent RU can aggregate 3 The management literature shows a sophisticated awareness of these issues. See, e.g., Leavy (1994, 1996). For example, Leavy (1996: 50) states: “Even in the closest of outsourcing relationships, the partners will always remain potential future competitors.”) 4 Practitioners recognize this. Michael A. Corbett Associates (2000): When organizations are not changing they favor internal sourcing. Similarly, when they view the internal operation as highly innovative they also favor internal sourcing. However, when there is a 4 information across supply relationships, thus gaining a “multiplier effect” for each unit of synergistic information. The emphasis on information in this Article represents a departure from prior treatments of vertical integration. Transaction cost economics (TCE) first called attention to the high- powered incentives of arm’s-length contracting. TCE incentives take the form of a more direct relationship between effort and profit, with no mention of informational synergies. On the other hand, the discussion of integration in this Article is closer to TCE. Preventing information leakage is closely related to the core TCE concern of integrating to reduce opportunism. It might also be argued that competing for rents from synergistic information is a form of opportunism. To the extent it is not, this Article extends current thinking by introducing rent dissipation as an explicit factor influencing vertical integration. The ideas in this Article are close to the spirit of the property rights-firm boundaries literature (Grossman and Hart, 1986; Hart and Moore, 1990; Hart, 1995) (“GHM”), but not identical. While the model here is concerned with the allocation of property rights in facilitating more efficient sequential investment, fewer restrictive assumptions are employed. For example, we avoid GHM’s exclusive emphasis on the strictly marginal effects of property rights allocations (Holmstrom and Roberts, 1998: 79). In addition, we explicitly consider “post- contractual” benefits arising from supply relationships – i.e., learning that has value in periods beyond those in simple two-stage models of investment and asset transfer. Most importantly, in this Article property rights specifications are not assumed to be fixed; patent strength is an explicit variable in the model. This permits more robust exploration of the relationship between features of the property rights regime and the “make versus buy” decision. recognized need for change and when the internal group cannot or is not viewed as being capable of bringing forward the needed changes, then external sourcing becomes the preferred option. 5 The approach here thus has some similarity to Zingales and Rajan (1998), whose concept of “access” to assets introduces a more nuanced interpretation of property rights. This Article pays particular attention to R&D-intensive inputs that do not often meet the conditions of their model, however. Finally, the bargaining between pioneers and improvers (Scotchmer, 1991) has some similarity to that between input suppliers and manufacturers (e.g., RU and MU). In both cases, multiple contributors together generate new, complementary information. And in this Article, as in Scotchmer (1991: 35), integration has the advantage of preventing rent dissipation. In the pioneer-improver scenario, however, ex ante bargaining is often impossible. Many improvement patents are owned by firms that had no opportunity to negotiate with the pioneer, often because lags in the patent system make it difficult to determine which of two parties will end up the pioneer, and which (if any) the improver (Merges, 1994). By contrast, suppliers of information-intensive inputs must negotiate with manufacturers ex ante. Indeed, in our model, synergistic information cannot be generated in the absence of RU-MU interaction. Thus our model analyzes the effects of rent dissipation identified in the pioneer-improver literature in a setting where ex ante bargaining is a necessary part of the situation. 2.0 The Model Let V(X) be the benefit to the MU from the purchase of the specialized input, where X represents the stage 1 effort by RU to customize the input, at a cost C(X, Z), where Z is the level of openness chosen by MU. We assume that X e [Xmin , Xmax], where Xmin is a baseline level of effort which can be verified. Similarly, we assume Z e [Zmin , Zmax]. We further assume efficient bargaining. Thus, whether MU chooses to integrate RU or not depends on which form yields the greatest joint surplus. We assume that V(X) and C(X, Z) increasing and concave in X, and C(X, Z) and C (X, Z) are both decreasing in Z, where C (X, Z) represents the marginal cost of x x 6 customization. In other words, openness by MU reduces the marginal cost of customization by increasing the flow of information to RU. For simplicity, we assume that openness has no direct costs for MU, though a more realistic situation would be to assume that controlling information flow is costly for MU.5 The process of customizing an input will require information flows between RU and MU. As noted earlier, such flows can have two types of consequences. They can reveal valuable information about RU to MU and vice versa. If both MU and RU are part of the same firm, leakage of proprietary information is of no consequence. However, if they are independent firms, RU may be able to use the information in ways that reduce MU’s rents. For instance, it may reveal this information to MU’s rivals, or embody it in services it provides to MU’s rivals. In sum, this leakage may lead to a partial dissipation of rents. In principal, the situation is symmetric. MU may likewise use what it learns from RU. It is less likely, albeit not impossible, that MU would use what it learns to compete, directly or indirectly, with RU, thereby also dissipating rents.6 Insofar as this usage does not result in rents being dissipated, one can simply think of this as an additional cost to RU, subsumed under C(X, Z). We only analyze the case where RU may use what it learns to compete with MU. Information exchange can also lead to the synergistic creation of new information. Because only an independent RU has an incentive to work hard to reveal new information, dis- integrating RU from MU can unlock significant potential value. New applications and extensions 5 Trade secret law protects information only upon a showing of “reasonable precautions” against disclosure, such as costly monitoring and sequestering. See, e.g., Rockwell Graphic Systems, Inc. v. DEV Industries, Inc., 925 F.2d 174 (7th Cir. 1991) (Posner, J.). For a descriptive account of such “fencing costs,” see Merges et al. (2000: 49). 6 Handfield, (2000: 40): One of the biggest challenges in supplier development is cultivating mutual trust. Suppliers may be reluctant to share information on costs and processes; the need to release sensitive and confidential information may compound this hesitation. Ambiguous or intimidating legal issues and ineffective lines of communication also may inhibit the trust building necessary for a successful supplier-development effort. 7 of the technology may be revealed.7 We model this as follows. After the customization is completed, with probability P(X, Z), MU and RU receive payoffs of P and P , and with M R probability 1-P(X,Z) MU receives a payoff of W and RU a payoff of zero. Here W represents the rents that MU would earn if information exchange did not result in leakage of information or the synergistic creation of new information. In the leakage case, we have W > P with synergy, P M; M > W. If both MU and RU are part of the same firm, then the combined entity receives a payoff of P with probability P(X, Z). Rents are dissipated if P > P + P . Under synergy, P > W, but in M R the leakage case, P = W.8 Clearly, these are not mutually exclusive possibilities. Information exchange is likely to reveal proprietary information that is the source of existing rents, as well as lead to the creation of new and useful information, which in turn may be the source of additional rents. Extending our model to accommodate both leakage and synergy is straightforward. The timing and structure of the game is as follows: RU begins the game with a property right (i.e., patent) over the general design of its input. If RU and MU are part of the same firm, the property rights belong to the firm as a whole. After the initial contract details are settled, RU and MU choose levels of X and Z respectively. We assume that neither X nor Z are contractible, but both MU and RU can observe the levels of X and Z. MU may also make a transfer payment to RU. The role of this first stage payment, T , is essentially to divide the total surplus between 1 the two. This concludes stage 1 of the game. In stage 2, RU and MU negotiate second stage payments, T . At this point, both X and Z are “sunk,” thus opening the door to potential holdup 2 problems. In stage 3, which is unique to our model, the information flows result in spillovers 7 To the extent that only RU becomes aware of the new application, this is formally equivalent to a reduction in cost, C(X, Z), and likewise, if only MU becomes aware of the new application, this is equivalent to an increase in V(X). 8 Note that P > W defines synergy whereas P = W defines leakages. It is clear that these cases are M mutually exclusive. However, they are not exhaustive i.e., we are ruling out by assumption the case where P > W and P < W, where information exchange always leads to a loss for MU but a net social gain. M 8 with probability P(X, Z). This may be thought of as the post-contractual period: the “out years” when learning gained during the supply relationship is applied to the economic activities of RU and MU. We assume that an independent RU and MU may not contract not to compete with each other in stage 3, implying rent dissipation in stage 3. To highlight the role played by patents, we first analyze the special case when stage 3 is absent. We demonstrate that patents can make possible contracts where an independent RU invests in customization. 2.1 Special Case: No spillovers Specialization (RU is independent) Once the investment is sunk, the parties bargain over the payment MU must make to RU. We assume the bargaining results in an equal split of the surplus defined by the “threat points” of the two parties. MU can threaten to end the relationship. Should it do so, RU will withdraw its input. After termination, MU would be able to duplicate RU’s design of the input, or transfer the RU design to a third party supplier, getting a net benefit of L(X). Here we assume that the ability of MU to produce the input for itself may benefit from the effort RU makes in customizing the input, and the disclosure of information by RU in the process of customizing the input.9 The joint profit maximizing level of effort is given by XOPT = argmax {V(X) – C(X, Z)}, so that even with L(X) = 0, the RU’s investment in customization in stage 1 is sub-optimal. If T is the second stage payment, then we have 2 T =argmax (V(X) - T - L(X))1/2(T )1/2 (1) 2 2 2 9 MU learns from RU in several ways: directly, through sharing of blueprints and the like; indirectly, e.g., by closely inspecting the physical embodiment of the input; or through some combination of the two. In this sense it is not particularly important what form the input takes. If RU is a software supplier, for example, it could supply MU either with finished computer code to be directly incorporated into MU’s own end-user software product, or 9 so that T = 1 (V(X) - L(X)) (2) 2 2 Knowing this, in stage 1, RU chooses X to maximize T – C(X, Z). MU chooses Z to 2 maximize V(X) - T . Since we assume Z has no direct cost, the choice of Z is indeterminate. We 2 assume that MU will choose Z = Zmax.10 The joint surplus is V(XS) – C(XS, Zmax) where XS is the effort level chosen by RU. It is easy to see that T is given by 1 T = 1 (V(X) +C(X, Z)) - 1 (V(X) - L(X)) = 1 (L(X) +C(X,Z)) (3) 1 2 2 2 This is the point where RU’s patents play a role. RU’s patent on the general design of its input implies that if MU had to “invent around” RU’s patents, L(X) would be lower than V(X). In this sense, the level of L(X) is inversely related to the effectiveness of intellectual property protection. This formulation is similar to the one used in Gallini (1985) and Arora (1995; 1996). For analytical convenience, we assume that L(X) = kV(X), where k e [0, kmax], kmax £ 1. A decrease in k corresponds to an increase in the “strength” of patent protection. Thus, one can write the choice of X as Xs =argmax 1 (1-k)V(X)- C(X, Z) (4) 2 We assume throughout that argmax {½(1- kmin) V(X) - C(X, Zmax)} > Xmin , so that strong enough patent protection will induce customization effort beyond the baseline level.11 with “high level” design information on how to achieve a particular software objective. We assume only that the input supplied by RU has a high degree of information content. 10 This would be true if, for instance, MU could move earlier or could commit. This would also be true if we added a small component to V(X) that was increasing in Z. 11 Note further that if T the second stage payment, can be contracted for in advance and MU can commit 2, not to renegotiate, it will be set so that MU is indifferent between ending the contract and making the payment, i.e., T = L(X) = kV(X). In this case, for k small enough, X = XOPT. This is formally shown in Arora (1996) and is 2, similar to the result in Noldeke and Schmidt (1998). The outline of the proof is simple enough. Set T = (1- 2 k)V(XOPT). Now for any choice of X < XOPT, MU will end the contract, giving RU a payoff of – C(X, Z). For X ‡ XOPT, MU will make the payment, providing RU with a payoff of (1-k)V(XOPT)- C(X,Z). Since C(X,Z) is increasing X, the RU either chooses X = XOPT or Xmin. For k small enough, (1-k)V(XOPT) - C(XOPT, Z) > - C(Xmin,Z). 10