IDEAS FOR RENT: AN OVERVIEW OF MARKETS FOR TECHNOLOGY

Transcription

1 IDEAS FOR RENT: AN OVERVIEW OF MARKETS FOR TECHNOLOGY Ashish Arora Heinz School, Carnegie Mellon University Alfonso Gambardella Department of Management, Bocconi University, Milan June 2008 VERY PRELIMINARY 1. INTRODUCTION Markets for technology (henceforth MFT) i.e., the trade of technology disembodied from physical goods have grown considerably in recent years. The goal of this paper is to offer an overview of some relevant problems and issues regarding these markets. The discussion will provide some answers and it will highlight the many open questions on this topic. In this respect, the aim of the paper is to set a basis for further research in this field. We start by providing a definition of MFT (Section 2). We then discuss some evidence about the extent and the growth of these markets (Section 3). The core of the paper deals with three levels of analysis (Sections 4-6): 1) Transaction. How do the MFT functions? What factors facilitate markets for technology? What are the principal barriers to mutually beneficial trade in technology? 2) Firm. When do firms sell their technology? When they integrate the production of the technology with the commercialization of products? Are there differences between firms particularly according to whether the technology-holding firm owns the downstream assets? What factors condition the demand for external technology? 3) Industry. How do MFT affect industry structure and competition? How does industry structure affect MFT? 2. DEFINITION We define MFT in a broad sense. Strictly speaking, market transactions are arm s length, anonymous, and typically involve an exchange of a good for money. Many, if not most, transactions for technology would fail one or the other criterion. Often they involve detailed contracts, and they are often embedded in technological alliances of some sort. Thus, though we shall contrast market transactions with processes 1

2 inside a firm, this is not to dispute the existence of hybrid forms but to sharpen the exposition. In this respect, we will not linger on taxonomic distinctions. Some technology contracts may well involve more complicated forms of interactions than arm s length market transactions. The economics of such more elaborate agreements or co-development processes may have some additional elements compared to pure technology transactions. Yet, they maintain some of the key features that we would like to emphasize in our analysis of MFT, and some of the things that are different do not relate or impinge on our study of the latter. We will consider joint-ventures, which involve the formation of a new firm, or acquisitions, as more integrated organizational solutions. To some extent, we could take them as longer run commitments in the transactions, and hence we could include them in what we are trying to discuss here. Yet, in these arrangements the command of the relationships is most typically associated to administrative and bureaucratic forms of authority, routines or other typical organizational mechanisms. As a result, we will associate them to processes closer to organizational rather than market-based relationships. Acquisitions in particular, but also R&D joint-ventures to some extent, involve a whole set of issues that pertain more specifically to the market for firms, which are different from the MFT. 1 At any rate, here we are not interested in a detailed discussion of whether patterns at the border between markets and organizations fall in one or the other category. There are of course grey areas, but we think that most of the real cases fall into situations that clearly exhibit predominate features of the market or organizational models. 2 The way technology is traded is linked to the peculiar nature of technology as an economic asset, and as a potential object of exchange. From this point of view, technology comes in very different forms, and no general definition will fit. For instance, technology can take the form of intellectual property (patents) or intangibles (e.g. a software program, or a design), or it can be embodied in a product (e.g. a prototype, or a device like a chip designed to perform certain operations), or it can take the form of technical services. In this respect, the notion of MFT is broader than the market for patents per se. Much technology is not codified and embodied in patents. One reason is that innovations are typically associated with many patents. Some patents cover different aspects of the innovation and others cover variants that are not commercialized (a fence building strategy, see, e.g., Ziedonis, 2004). A second reason is that in many cases patents are merely used to block others and therefore may be offered for license. As such, they are permissions to infringe. Such licenses are included in the MFT. However, technologies may be licensed without patents, and even when there are patents, they include many non-patented components. There is nonetheless some reason to emphasize patents because patents can play a key role in facilitating such transactions, a point that we discuss in Section 3 below, and because the market for patents does overlap with the MFT. We will neither attempt to define technology, treating it instead as an imprecise term for useful knowledge rooted in engineering and scientific disciplines, which usually also draws from practical experience from 1 Indeed, it is widely accepted that the leading networking equipment firm, Cisco, acquires external technology principally by acquiring the firms that hold this technology. Cisco itself performs relatively little R&D, relying upon acquisitions to upgrade its technology. Many observers credit Cisco s ability to manage and integrate these acquisitions for its success. 2 Transactions in technology can also occur through the mobility of people. However, we shall ignore these cases here. 2

3 production. This is related to the point that collaborative agreements, alliances, co-development or cofunding processes are part of the phenomenon that we are dealing with. While pure forms of licenses (e.g. exchange of chip designs or specific patent rights) have become increasingly common, on many occasions technologies are exchanged by transferring both an object and some services associated with it (e.g. training with the technology), or the supplier-buyer relationship occurs ex-ante (e.g. co-development, co-funding). This is because the buyer may have to provide own assets in the process to make a technology better suited to its needs (co-development). Or a typical feature of the inventive business is that inventors are liquidity constrained (for various reason, e.g. because they are small), and this requires that they be provided with enough resources to carry out the activity (co-funding). The tremendous development in software technology further complicates attempts at strict definitions. Many technical ideas and designs, hitherto implemented in physical prototypes and paper designs, can now be more usefully expressed and implemented in software. A classic example is encryption. Although encryption algorithms can be expressed as mathematical formulae and implemented using physical devices 3, today, encryption algorithms are expressed in software and implemented in software. Another example is circuit design and design of various semiconductor devices, which are also implemented as software programs. The distinction between technology and product becomes difficult to make in such cases. These difficulties have manifest themselves in a variety of controversies in IP law, wherein sellers of software products use licensing like contracts to effect sales of their products. These difficulties have also led writers such as Boldrin and Levine (2008) to claim that there is, in essence, no difference between the two, and hence conclude that there is no need for intellectual property protection for technology. We believe that there important differences, even if it is difficult to draw clear distinctions in specific instances. Software products need to be produced to be compatible with other products, software and non software; they need to be maintained and upgraded, patched to fix bugs and so on. To conclude, technology transactions can take different forms, from pure licensing of well defined intellectual property, to complicated collaborative agreements which may well include the further development of the technology, or its realization from scratch. Figure 1 summarizes our definition of the MFT in the form of a simple typology, along with canonical examples for each case. The Figure emphasizes 2x2 dimensions. Technologies can be sold to firms in the same product-market (horizontal transactions) or to firms operating downstream (vertical markets). At the same time, they can refer to existing technologies that are licensed, or they can be markets for future technologies that could be more properly called markets for innovation. These include arrangements in which the parties agree to conduct activities, jointly or independently, leading to future developments of technologies that will be exchanged (or jointly owned) among them. This is typically the market for contract R&D and the various types of technological alliances. 3 The famous enigma machine used by the German armed forces during WWII is a case in point. 3

4 Figure 1 A simple typology of markets for technology Horizontal Market / Transactions with actual or potential rivals Vertical Market / Licensing to non rivals Existing Technology Union Carbide licensing Unipol polyethylene technology to Huntsman Chemicals Licensing of IP Core in Semiconductors Future Technology or component for future Sun licensing Java to IBM; R&D partnership between rivals (e.g., see Hagedoorn (2002) R&D agreements or other technological alliances; Affymax licensing combinatorial drug discovery technology to pharmaceutical companies In sum, MFT refer to transactions for the use, diffusion and creation of technology. This includes transactions involving full technology packages (patents and other intellectual property and know-how) and patent licensing. It also includes transactions involving knowledge that is not patentable or not patented (e.g., non-patented designs and innovations). It can involve parties in the same product markets or vertically related, and the transaction can take pure licensing forms as well as more elaborate collaborations and contractual agreements, which typically involve future or further developments of the innovation. An important aspect of MFT is that they are required to support a division of innovative labor. A division of labor may yield benefits from larger scale and learning, emphasized by Adam Smith, as well as a superior allocation of resources according to comparative advantage. A firm unsuited to commercialize inventions need not leave them on the shelf, nor must it tortuously change to become a commercializer. Instead, it can derive value by licensing them to others. Finally, in a world where commercialization is costly and slow, licensing can also lead to a more rapid diffusion and exploitation of technology. As well, markets for technology can lower entry barriers and increase competition in downstream product markets. Understanding the factors that determine when a division of innovative labor can emerge and be sustained will overlap with many of the issues discussed here. This, however, is not our present focus. 3. EVIDENCE In a long-term perspective there has been an acceleration of the technology licensing payments and receipts. By using International Monetary Fund (IMF) data at the country level, Athreye and Cantwell (2007) show that technology licensing payments and receipts have accelerated considerably since the 1980s, after being roughly constant between (Figure 2). From these data, one cannot distinguish whether the transactions were between different firms or affiliated firms, such as subsidiaries and parent companies. Evidence from the United States indicates that a substantial fraction of these flows involving American firms are between affiliated parties. Although this fraction has declined slightly over the 1990s, it remains high, around 75% (Figure 3). Thus, judging from Figure 3, if one considers only licensing to unaffiliated entities, 4

5 the acceleration in Athreye and Cantwell s (2007) data is probably less marked. However, licensing to unaffiliated entities also seems to have increased. Figure 2 Growth in non-us patents and worldwide licensing and royalty rates Source: Athreye and Cantwell (2007), p.217. Figure 3 US licensing flows, total and from unaffiliated entities Source: U.S. Bureau of Economic Analysis, Survey of Current Business 84(10):25 76 (2004) 5

6 As noted in the previous section, it is hard to make a tight distinction between alliances and pure arm-length market deals in technology. While there are pure technology deals or well-defined co-development agreements, many R&D partnerships involve technology exchange in addition to services or co-development work, and many licenses may also require follow-on services like training of personnel or further development of the applications. However, specific data for R&D partnerships offer a picture similar to Athreye and Cantwell. Hagedoorn (2002) shows that while between 1960 and 1980, R&D partnerships have been roughly constant worldwide, they have exhibited a significant acceleration between Moreover, he shows that there has been a parallel fall in the number of R&D joint-venture, which can be seen as a form of stronger integration between the partners. Thus, what makes our evidence clearer then is that both forms of technology transaction have increased in a similar fashion, and they have both accelerated after the In turn, R&D joint-ventures, i.e., more integrative and less contractual forms of partnership, have declined, which is suggestive of the trend towards greater market-based deals in technology exchange. The OECD provides additional evidence at the aggregate level. It has recently started collecting data on disembodied technology trade, in good part as a consequence of the recognized increase in technology transactions. According to OECD data, in the G8 countries, from 1980 to 2003, the technology royalty payments and receipts have increased by an average annual factor of 10.7%, reaching a volume of about USD 190 billion in 2003 (OECD 2006). Similarly, the U.S. Bureau of Economic Analysis indicate that royalties received by American corporations for industrial processes in 2002 may amount to as much as 25% of total U.S. R&D spending in that year (Robbins 2006). Since the 2002 US R&D was about 280 billion USD, this means that the royalty payments to American corporations were about 70 billion dollars. In turn, this is consistent with an estimate of the MFT worldwide well over 200 billion dollars, or in the same order of magnitude of the OECD figure of 190 billions. The increased importance of MFT is confirmed by more disaggregate data, both at the industry and firm level. First, many high-tech industries today feature specialized technology supplier firms e.g., biotechnology companies, fabless or chipless firms in semiconductors, software or medical devices start-ups. We will discuss the role and activities of these firms in the MFT in the next sections. Here we simply point out that the rise of these specialized technology suppliers since the 1980s is a manifestation of the growing importance of the underlying technology markets. Second, an OECD survey suggests that worldwide companies have increased their propensity to license in and out new technologies (Sheehan et al., 2004). The survey, which was administered in 2003, covered 105 firms in Europe (68 firms), North America (20), and Asia-Pacific (17, mostly from Japan). Most firms were large only 20% had fewer than 1000 employees. Almost 60% of the firms interviewed reported increased inward and outward licensing during the 1990s. Moreover, this was reported more frequently by North- American and Japanese than European firms. Also, almost 40% of the companies reported increased crosslicensing. Again, this was more marked in the case of North-American firms. 6

7 Across industries, Information and Communications Technology (ICT) shows the highest share of companies that reported an increase in the importance of outward licensing, almost 80%. This percentage is higher than the percentage of ICT firms that reported an increased importance of inward licensing (almost 60%). Pharmaceuticals is the sector with the highest share of companies reporting an increased importance of inward licensing: more than 80% compared to slightly less than 40% of the companies reporting an increase in the importance of outward licensing. The pharmaceutical firms in the sample are large, and the higher importance of inward licensing mirrors the increased opportunities to obtain licensing from specialized biotechnology companies. Thus, the OECD survey confirms that ICT and pharmaceuticals are the leading industries in which MFT have grown (Arora et al., 2001a). However, MFT are not confined to these two sectors. As noted above, almost 60% of the interviewed firms reported increased inward and outward licensing. Moreover, the share was higher for the large firms in the sample, which suggests that licensing is becoming more important for the established firms as well, and it is not confined to the technology specialist suppliers mentioned above. The survey also found that revenues from licensing was a very important reason for patenting for 39% of the ICT firms in the survey against fewer percentage points for the other sectors. Pharmaceuticals were again the next largest sector in terms of share of companies responding that licensing revenues was a very important reason for patenting (27%). Finally, the firms in the survey indicated that both inward and outward licensing will be increasingly important in the company strategies in the next five years, with ICT and pharmaceuticals reporting the highest share of firms responding the licensing will increase in importance. 4. TRANSACTION 4.1 Transaction costs MFT are bound by transaction costs (TC). TC are typical in the case of knowledge or intangibles (e.g., Teece, 1988). Though MFT have grown considerably in recent years, TC remain one of the main limitations to their expansion. We classified the TC faced by the MFT in three main types: a) Cognitive that is, a good deal of the knowledge associated to a certain technology is tacit, which makes it hard to define the object of exchange in a contract b) Ill-defined property rights which stem from the fact that property rights on intangibles are in general hard to define c) Uncertainty about c1) property rights; c2) value of the technology; c3) transaction process 7

8 4.2 Factors driving the growth of MFT Growing science base of industrial technology In the past few decades industrial knowledge has become more scientific. As suggested by Arora and Gambardella (1994) this stems both from a greater scientific intensity of engineering knowledge (as opposed to serenditipity or blind trial and error) and the reduced distance between some new sciences and commercial applications, as in the life sciences. Moreover, software is making the research process more codified e.g., through the development of software tools that automatize functions like design, engineering, or simulations. As technological outcomes become better defined e.g., a compound, an algorithm the cognitive limitation becomes less severe. This also facilitates patenting because the object of the patent can be identified more clearly. The property rights on the technology then face fewer ambiguities, which helps the patent owner exercise her right for instance, infringement can be singled out and sanctioned more easily. As already mentioned, the growth of software technology has greatly helped by allowing technology to be expressed usefully and, in many cases, more cheaply in software instead of physical prototypes. This is self evidently true for software technologies, as is the case with RSA, which provides a widely used encryption technology. It is also true for semi-conductors, with the rise of fabless chip firms, and now even chipless firms, such as Rambus, which specialize in specific components of microprocessor chips. Similarly, in telecom, firms such as Qualcom or Symbion, focus on developing technologies which are ultimately made available in the form of software programs. However, even outside the information technology and telecom sectors, the effects of a growing ability to express a design in software can be seen. For instance, in the area of industrial controls, a firm called Echelon has claimed to develop a general purpose control technology, which is implemented via a controller chip and interoperability standards implemented ultimately in software. We expect similar developments in many more important technical fields Intellectual Property Arora (1995 and 1996) shows that when part of the technology is codified and protected by patents, one can write simple contracts for its exchange, including the non codified parts. Specifically, this happens when the technology is made of a codified component, protected by patents and that can be transferred as blueprint, and complementary tacit know-how that can be transferred through training, technical support or other services. Examples include the transfer of a technology about the functioning of a plant, in which the supplier offers a blueprint license (the codified component) along with training or technical services; or the forms of technology trade discussed in Section 2 e.g., a licensing contract or an R&D alliance in which the technology supplier offers a patented technology together with the research services related to it. Arora s problem is that the provision of the service is costly for the supplier, who may then undersupply it. In turn, the buyer has an incentive to shirk because once the service is supplied, the knowledge outcome associated with it cannot be withdrawn. This is because in the case of intangibles, like training or learning, any tacit expertise, once transferred, cannot be taken back by the supplier. To solve this double-sided moral hazard problem, Arora s contract hinges on a two-stage payment. Apart from a first payment offered at the 8

9 start of the contract, the second payment, which is provided after the training is supplied, should be large enough to induce the supplier to transfer the optimal amount of training. Otherwise the buyer may not provide the full second payment. The hostage in the hands of the supplier is that he can withdraw the license to use the codified component if the buyer does not provide the specified second payment after the supplier offers the optimal amount of training. Key to this mechanism is that the codified and tacit components of the technology are complementary, viz. the learning is less valuable if it cannot be used together with the patented technology. For example, the supplier can withdraw the right to use the patented technology for the functioning of the plant or for the research outcome produced by the licensing agreement or the alliance. This reduces the ability of the buyer to use the technology even if the training or the learning is sunk. In this case, if the buyer uses the accumulated training or learning, she will run the risk of being liable for infringement, or has to bear the extra costs of inventing around the patented technology. Another important element of the model is that stronger intellectual property rights (IPR), in the sense of stronger enforcement or stronger ability of the technology owner to exercise his right, raise the odds of the buyer of being accused of infringement or of sanctioning an attempt to invent around. In other words, the supplier s threat is more effective because alternatives to the patented technology, if the right to use it is withdrawn, are costly. The supplier can use this threat to reduce the value of training even after it is supplied, in case the buyer does not abide by the contract. Ultimately, this means that stronger IPR can facilitate technology trade. With weak IPR even if the supplier withdraws the right to use the technology, the buyer can use the codified component with minimal risk of being sued for infringement, or he can circumvent the patent by inventing around it. If so, the buyer has an incentive to behave opportunistically about his second period payment, and the supplier does not have enough incentives to provide the optimal level of training. In turn, this lowers the incentives of the buyer to enter into the transaction in the first place. Some contributions in the literature address specifically the question whether and how IPR favor the MFT. The general point is that IPR better define the property rights on the invention or technology. This facilitates technology trade because the supplier is better protected against the opportunistic behavior of the buyer, or more generally against the loss of control of the rights on the technology. In particular, the literature suggests that this favors the supply of technology by smaller firms with no downstream assets or with no other means for protecting their technology. This is because the legal stake becomes their only tool, or the main one, to protect their intellectual property. For example, larger firms may take advantage of their control of the final markets (downstream assets). This prevents competitors to expand in these markets even if they use the same technology or if they can invent around it (e.g., Merges and Nelson, 1990). Arora and Merges (2004) provide specific examples of the advantages of holding IPR for firms of smaller size, which do not own the downstream assets. In particular, they highlight that IPR become the only means 9

10 by which these firms can prevent opportunistic behavior of the buyers. Hence, IPR encourage them to supply the technology. Similarly, Hall and Ziedonis (2001) show that in semiconductors IPR favor the entry of design firms that sell intellectual property and do not fabricate the chips. Gans et al. (2002) show that with stronger IPR the biotech firms find it profitable to sell the technology rather than integrate forward in testing the compounds. Arora and Ceccagnoli (2006) empirically show that the effect of IPR-protection on licensing is stronger when the firm does not own manufacturing assets. Indeed, for firms where R&D and manufacturing are complementary, increases in patent protection have no effect on licensing, whereas such increases markedly increase the propensity to license in firms without the complementarity between R&D and patenting. A similar difference can be seen between large and small firms, with the licensing propensity of small firms much more responsive to increases in patent effectiveness. 4.3 Persisting problems Transaction costs Cognitive issues or IPR have not fully solved the transaction cost (TC) problems involved in MFT. Razgaitis (2004) uses a sample of 229 US and Canadian companies and shows that out of 100 licensable technologies, in only 25 a potential licensee is found, in 6-7 the parties enter into a negotiation, and 3-4 deals are eventually concluded. Some of the most frequently cited reasons are search costs for finding a licensee, fear of opportunism in negotiations, lack of valid enforcement of intellectual property. (See also Cockburn, 2007.) A large scale survey on European patents, PatVal-EU, also finds that TC play a role in the market for European patents. PatVal-EU is based on a sample of more than 9000 patents applied for and eventually granted at the European Patent Office with priority date (Giuri e al., 2007.) The PatVal-EU estimate is that in 18% of the patents in the sample the applicant is willing to give out a license. However, only 11% are actually licensed. This is a lower attrition rate than the 3-4% suggested by Razgaitis. Yet, it suggests that a sizable share, about 40%, of the patents offered for licensing are not licensed. Moreover, after controlling for several factors, Gambardella et al. (2007) show that the patents offered for licensing that are not licensed are not different from the licensed patents. This suggests that the reason for the failure of the deals is not a lower demand for less valuable patents. TC is the residual explanation Uncertainty Economists have long identified uncertainty as a barrier to a MFT. However, their focus has been on asymmetric uncertainty coupled with the possibility of opportunistic behavior the lemons problem, where technologies offered for license are systematically inferior to those exploited in-house, much as lemons are offered in the second hand car market (cf. Zeckhauser, 1996; Pisano, 1997 and 2006). In many instances, one might even imagine that the asymmetries are reversed, with potential buyers being more knowledgeable about the market prospects of the technology. While such asymmetries are undoubtedly present, our sense from talking to participants is that this is overwhelmed by symmetric uncertainties about the future. 10

11 Moreover, it remains unclear why such asymmetries should significantly reduce the number of technology trades, as opposed to depressing the price at which such trades occur. A second potential culprit is risk aversion with uncertainty, risk aversion in general deters investment. This is correct but does not explain why risk aversion should favor internally developed technologies as opposed to technologies offered on the market. As yet, we lack a satisfactory understanding of why symmetric uncertainty creates difficulties. This is a major challenge and one that we do not take up here. Instead, we offer a brief discussion, leavened by anecdotal accounts, of the problem and possible explanations. As noted earlier, there are three levels of uncertainty that characterize technology deals: about the property rights; about the value of the technology; and about the transaction process. We examine them in turn. Uncertainty about property rights. Gans et al. (2008) show that when a US patent is licensed, the licensing takes place largely within a narrow window around the date of the grant of the patent by the US Patent Office. After controlling for several factors, they show that this stems from the reduced uncertainty and asymmetric information about the extent of the property right. Simply put, on the date of grant both parties know claims granted to the patent, or whether the patent is granted or not. It may be argued that the reason why licensing clusters around the date of grant is that before then the applicant has no property right to sell. However, the trade may well take place earlier at an appropriate discount for the probability that the patent is granted or for the expected number of claims. The problem is that because this information is asymmetric, the buyer may ask for a premium in the form of a discount. This induces the supplier to conclude the deal after the grant to avoid having to provide such a premium. Gans et al. (2008) also show that in areas where technology producers interact considerably with one another a good deal of the information that becomes available at the time of grant is available earlier as well. They show for instance that in Silicon Valley the concentration of licenses around the time of grant is less pronounced. This suggests that local interactions among the interested parties provide information about the technology that in less interactive environments only becomes available through formal institutional steps. For example, the exchange of information between the parties can reduce asymmetric information. As a result, the parties have the same expectation about whether the patent office will grant the patent or not or about the number of claims. This means that the buyer needs no premium in order to buy, or that the premium is lower, which facilitates the trade. More generally, these remarks are consistent with the perspective that licensing transactions are more likely to occur between parties that are geographically close, and thus that MFT are more local than managers or analysts sometimes claim (Alcacer et al., 2007). Uncertainty about the value of the technology. The economic value of patents is very skewed (Scherer, 1965; Scherer and Harhoff, 2000). While the firms know this distribution, they know much less about which patent is in the right-tail or even in the moderate high values of the support. Moreover, since many patents have in the end no or very small value, the difference between high and low value patents can be quite high. Thus, types 1 or 2 errors can be considerable. As shown in Gambardella et al. (2008) the gap between patents in the right and left tail of the distribution of patent values can be in the order of the ten million euros. 11

12 The problem is that the value of a technology is gradually revealed as time elapses. For example, after some time a dominant design or standard may emerge, and this reduces the value of alternative technologies; some technologies can eventually prove to be more effective than others, while others become obsolete. Yet, if the transaction takes place too late, when the uncertainty about the value of the technology is reduced considerably, a good technology will no longer be sold under competitive conditions because the alternative technologies have shown not to be valuable, or they have not emerged as the standard or the dominant design, or for other reasons. By contrast, if the transaction takes place early, when there is still uncertainty about the exact value of a set of alternative technologies, there will be competition among them. Take the example of a standard (e.g., the CDMA technology by Qualcomm). If the winner technology is sold when the battle of standards is over, the supplier can enjoy significant rents. If they are sold before the standard is set, competing standards may exercise some moderating effects on the technology price. A similar problem arises with patent infringement. Today many research terrains are so widely protected by patents that any new project in the area runs the risks of infringing some existing patent. The problem is that once the infringement takes place, the infringed party knows that the infringer may have committed resources to carry out the research, and hence her exit costs may be high. Since they are sunk costs for the potential infringer, in a classical hold-up situation the infringed party can ask for a compensation that does not enable the infringer to cover the sunk costs. The potential infringer may then be discouraged to start the research in the first place, to avoid being caught in the hold-up. The alternative would be to go to court. But the litigation costs can be equally severe, especially for small research-intensive firms. A potential solution to the infringement problem is the so-called soft patent or licenses of rights system. Essentially, a group of patent holders agree on committing ex-ante to a certain compensation if others infringe the patent. The ex-ante bargaining can take place before the uncertainty about the values of these patents is resolved. As a result, once the infringement takes place, the patent-holder will simply ask for the compensation to which he committed, without taking advantage of the additional information that he may have about the value of his patent or about the fact that he now observes the infringement and the infringer has exit costs. To summarize, this discussion suggests that some uncertainty in the MFT arises because ex-post the returns from holding a technology, or from undertaking a research projects are far more skewed than ex-ante. This is a natural consequence of the winner-takes-all nature of many investments in technology. Ex-ante, many technologies may have equal odds of being successful, and hence the proprietor may be content with receiving the expected value of it. Ex-post, once the value is revealed, the winner may gain extraordinarily high returns, while all the competing technologies receive zero or negative returns. The ex-post situation is clearly more risky than the ex-ante one. Risk-averse parties are then discouraged from investing in the creation of new technologies. Ultimately, this calls for finding ways to commit the parties early enough, which in turn means finding institutions or other mechanisms for pre-commitment and coordination. Some authors emphasize the role of clearinghouses, which pool licensors and licensees, collect the rights ex-ante, and provide solutions through 12

13 royalty distribution schemes (Aoki and Schiff, 2008a). Yet, Aoki and Schiff (2008b) show that while the royalty distribution schemes of the clearinghouses can increase the average profits from licensing, and hence restore the incentives to perform R&D, they may also reduce the incentives of suppliers who have the ability to produce crucial components, and hence expect higher returns. In short, the welfare implications are ambiguous. However, the natural intuition here is that the clearinghouses are more likely to be welfareenhancing when the R&D costs are high and the probability of success of the individual inventors is low. However, uncertainty can have the other effects as well. For instance, Erkal and Scotchmer (2007) point out that in choosing to develop an idea, one may have to give the option of developing a possibly better idea in the future. This option is valuable for society, but not to the firm that has come up with the idea, whose probability of being the one to come up with the improved idea in the future is small. Erkal and Scotchmer use this set up to analyze patent policy. In the MFT context, this may explain why licensing ideas can sometimes be frustrating: Potential licensees, with an eye to the future, may be willing to pay less than the expected value of the technology itself. Further, the greater the uncertainty, the more valuable is the option. This implies that conditions of uncertainty would reduce the reward that idea developers unable to commercialize their innovations can expect. This effect of uncertainty is not linked to risk aversion as such. Rather, it is linked to the uncertainty about future developments in the technology and the inability to write contracts today with those who will come up with those developments. These contracts are infeasible simply because no one (including the developers themselves) knows who will have the improved idea in the future. Uncertainty about the transaction process. We recently interviewed a licensing executive responsible for the in-licensing of a leading European pharmaceutical firm. 4 He suggested the following characteristics of the transaction process. The buyer starts by looking around for a compound from a biotech firm, or sometimes the biotech company approaches the buyer firm with a compound to offer. If the buyer is interested, he makes an initial bid, and asks the biotech company to provide more information. Once the biotech firm sends the information, the buyer evaluates it and if he is still interested, he will produce a more detailed bid, and so on. The buyer firm is often in competition with other buyers, but this is private information of the biotech company. It is like an auction in which the buyer does not know who are the other buyers and the offers that they make. Our executive notes that there are the conditions for a classical winner curse problem. Because of the uncertainty surrounding the value and the potential of the technology, the buyers may propose different bids on the technology for reasons that can be highly subjective. As the theory of common value auctions had demonstrated (Milgrom and Weber, 1982), and as a growing body of empirical evidence, and results of lab experiments indicate, the winner s curse is real. Faced with the prospect, prospective buyers bid less than their expected value of the technology, and in some cases (e.g., Kagel and Levin, 2002) may shade their bids by as much as 65%. There is less clarity on what this means for the expected revenue that the seller (in this case, the licensor) can get. 4 Unfortunately we cannot disclose the name of the company at this stage. 13

14 Note that the key uncertainty here is the private information about the distributions of the bid by the buyers. If the buyers knew the other bids, they could adjust their offers as they could figure out whether they are in the right- or left-tail of the distribution. They would then move their offers closer to the mean. But they do not know it, and they are only revealed that they are in the right-tail when they win and if they win. The uncertainty about the transaction process is reinforced by the vagaries about the value of the technology. Unlike physical goods, the price of a technology cannot be hanged to some anchor measure of value, like a mark-up on costs. As a result, the predictions of the two parties of the reservation price of the other are characterized by severe uncertainty. This stems from two reasons. The first one is that the prediction about the future streams is affected by considerable errors by the very party that is making the prediction because of the underlying uncertainty about the future value of the technology. The second one is that both calculations are based on private information (use of the technology, other costs of the firm, etc.). Thus, the prediction of each party about the reservation price of the other can be quite volatile, leading to impasses when profitable trade could have taken place. The pharmaceutical executives that we interviewed confirmed these uncertainties associated to the price of the technology in the negotiation. Prices can fluctuate a lot in the life-span of a transaction. He pointed out that because the prices are not anchored to anything, the first price that is made in a bid (viz. when the uncertainty and asymmetric information is the highest) will determine the order of magnitude of the transaction if the deal is concluded. For example, if the buyer offers 500 millions, the final price could range between 50 millions and 1 billion. If for the same transaction he started with 100 million, it could end up between 10 and 200 millions. Interestingly, we also interviewed an executive working for a biotech company and responsible for the licensing out of her firm. She confirms such a volatility and vagary of the price, including the possibility that the price ends up being very different according to the starting point. Yet, she was not concerned. She saw it as a natural feature of the way the price is formed in this market. 5 Two other factors add to the volatility of the price according to our executive in the pharmaceutical firm. First, most technologies that his firm seeks to buy are unique, in the sense that there are few alternatives. This provides the supplier with bargaining power. However, if a competing technology comes up, the seller s power drops dramatically e.g., from an order of magnitude of hundred millions, to simply millions. Second, our executive pointed out that uncertainty in the transactions is also induced by the parties themselves to gain bargaining power. In particular, when the buyer is an established firm with repeated transactions with biotech producers, it tries not to establish a reputation. That is, the bidder will not start systematically with low prices in the bargaining, eventually accepting high prices when the deal ends up. This would suggest to any counterpart that has observed the behavior of the buyer in the past to wait till this particular buyer starts raising the bid-price. Similarly, the buyer cannot establish a reputation of starting high and then moving to low prices. Thus, the buyers start either high or low trying not establishing patterns that the suppliers can recognize and adjust their bargaining strategies accordingly. 5 At this stage we cannot disclose the name of this company either. 14

15 5. F IRM 5.1 Demand for external technology Many organizations are believed to underinvest in external technology, inefficiently preferring internally developed technology, sometimes dubbed the Not Invented Here syndrome. This behavior is simply taken to mark large organizations but its causes have not been seriously investigated, especially by economists. The recent calls towards open innovation certainly target the NIH syndrome, and the CEOs of a number of large firms have taken to almost ritual affirmation of the importance of being open. To a considerable extent, as large corporations have systematically reduced in-house R&D, particularly more long range and basic R&D, this may be little more than making a virtue out of a necessity. Nonetheless, it is a worthwhile question to ask how it is that firms came to be afflicted with the NIH syndrome and the possible remedies. Once again, this is a difficult challenge and once again we content ourselves with pointing to possible components of an answer. The need to provide incentives to internal innovation may be an important reason for an organization to be less receptive to external technology. Simply put, the NIH may be a commitment device to overcome incomplete contracting within a firm. As Rotemberg and Saloner (1994) argue in a slightly different context, firms wish to reward their own employees (in this case, likely those involved in R&D) to innovate. However, when external ideas can also be readily adopted and a firm may only reward an employee if the internal idea is adopted, then the competition posed by the external idea may deter internal innovation. However, one could turn this argument on its head by noting that competition from external ideas may spur in-house teams to greater effort, as Fosfuri and Roende (2007) argue, once again in a different context. Absorptive capacity provides a link to a second possible answer. As is well known, buyers in the MFT can do better if they themselves have significant in-house R&D capability. This helps them identify and absorb external knowledge (Cohen and Levinthal, 1989; Arora and Gambardella, 1994). It also provides them with credible threat point in bargaining over the price of a license. But it is plausible that the in-house R&D group may feel threatened by external technology and may use its political power inside the organization to block such access. Of course, recognizing this, firms could try to align the incentives to mitigate this possibility. The practice and theory of such alignment remains a topic for future research. Simple synergies, perhaps due to close communication between the in-house R&D function and marketing and manufacturing may be an efficiency based reason for NIH. Acquiring external technology may put such synergies at risk. However, although widely accepted, the extent and importance of such synergies has not been systematically documented. 5.2 Profiting from innovation Exclusive licensing In this section we look at the strategic decision of the firms that hold a technology whether they want to license it or not, thereby feeding the supply of the MFT. In this context, the natural question to start with is 15

16 the classical one raised by Teece (1986) viz. how do firms profit from innovation? Here one has to distinguish between exclusive and non-exclusive licensing. Exclusive licensing implies a complete loss of control by the licensor, whereas non-exclusive licensing may not have the same implication. Moreover, with non-exclusive licensing, the nature of the product market competition, between different licensees, and between licensees and the licensor also become salient. Teece s classical approach is that a firm holding a technology or an innovation can profit from it two ways: i) by embedding the technology in a final product ii) or by selling it Teece (1986) then argues that two factors affect the decision of firms between i) and ii): a) the appropriability regime and b) the control of the downstream assets The first factor was discussed in the previous section. A well functioning system of IPR reduces the TC because the property rights are well defined. As noted by Gans et al. (2002) it also encourages the firms to trade their technologies as opposed to integrating forward. Weak IPR imply that the main alternative by which the firms can secure some returns from the innovation is to embody it in products. By contrast, well defined IPR imply that they can profit from the innovation by selling the technology or the idea. Note that this raises the incentives to innovate at the system-level, and therefore the rate of innovation and exploration in the economy. Many smaller firms or even individuals may be discouraged to invest in innovation because they know that in order to profit from it, they have to own the downstream assets that are necessary to embody it in final products and sell the products in final markets. Technology development costs or manufacturing and commercialization costs can be severe. Since these firms lack the resources to undertake these investments, they do not invest in the innovation in the first place. By contrast, IP and technology markets encourage them to invest in innovation because they can draw rents from this activity even if they cannot invest in the downstream assets. As firms try more innovations, the degree of exploration of the economy increases. The other dimension viz. the control of downstream assets is a key factor in the decision to license. Firms with downstream assets need to feed them in order to utilize the capacity, which is a primary reason for embodying the technology in the downstream operations of the company. For this very reason, Teece also argues that the technology and the downstream assets are made to fit one another. Simply put, the firm will produce a technology that best fits its use with the type of downstream assets that the firm owns and vice versa. In short, the upstream and downstream assets of the firm are co-specialized. Thus, a firm with downstream assets will use the technology internally both because in this way it feeds the use of its downstream resources and capabilities, and because the technology is deliberately developed to best fit these resources and capabilities. 16