The economics of patents: from natural rights to policy instruments. August 2003

Transcription

1 The economics of patents: from natural rights to policy instruments David ENCAOUA (EUREQua, CNRS & University Paris I) Dominique GUELLEC (OECD) Catalina MARTÍNEZ (OECD) Second Version 1 August 2003 Abstract This paper uses latest advances in economic research for examining recent changes in patent regimes aimed at strengthening patent protection, and beyond that, for rethinking the rationale of the patent system. Considering that economic theory does not regard patents as a natural right that should be systematically granted to inventors, but as a policy instrument aimed at fostering innovation and diffusion, three major implications can be drawn from economic theory regarding current policy debates. First, patents may not be the most effective means of protection for inventors to recover R&D investments when imitation is costly and first mover advantages are important. Moreover, they may do more bad than good to innovation if innovation is cumulative and first generation inventions are essential to develop further inventions, especially when patent protection is strong. Patents should not be seen as the solution by default, notably as regards new areas of patentability such as software, business methods and genetic inventions. Second, patentability requirements, such as novelty or non-obviousness, should be sufficiently stringent to avoid the grant of patents for inventions with low social value that increase the social cost of the patent system. Third, rather than the statutory patent life, what matters is the effective life of patents: the broader is a patent the longer is its effective life. Policy instruments affecting patent breadth (e.g. extra fee for independent claims above a certain threshold) and length (e.g. renewal fees) could be used to provide long effective lives to inventions with high social value. Beyond these currently debated issues, economic theory pleads for an in-depth reshuffling of the patent system. If the system were to be radically changed, an optimal patent policy could be based on a multidimensional menu of different degrees of patent protection associated to different patent fees, where stronger protection would correspond to higher fees. Patents could be transformed into self-selection mechanisms whereby patentees reveal the economic characteristics of their inventions, compensate society for the protection they are granted and obtain sufficient incentives to innovate. 1 A preliminary version of this paper was presented at the EPIP Conference "New Challenges to the Patent System", held in Munich at the EPO, on April 24/25, We are grateful to Dietmar Harhoff, Brian Kahin, Paul David and the participants to this conference for helpful comments. The opinions expressed in this paper are the sole responsibility of the authors and do not necessarily reflect any position or policy of the institutions to which the authors are affiliated.

2 1. Introduction There have been tremendous changes in patent regimes over the past two decades, all going in the same direction: expanding and strengthening protection. The patent community has been a major driving force behind this evolution, namely attorneys, judges, patent offices, as well as business intellectual property associations. In this context, the legal view of patents, a property right conferred to inventors in order to reward them, has been prevailing. Economic analysis has been essentially absent from the debate. Economists see patents as a policy instrument to foster innovation. They complement other instruments such as grants, subsidies and public research. Government intervention in the field of innovation is made necessary by the public good aspect of many inventions, and patents have certain advantages compared with these other policy instruments, insofar as their market friendly aspect does not involve a direct intervention by government. However, patents have drawbacks: like other policy instruments: they generate costs to society and should be implemented carefully. Looking at patents as a policy instrument both reinforces their status, as it highlights the benefits they bring to society, and leads to prescribe a careful design and implementation of the rights they confer to inventors. A reinforcement of patents might be a good thing in certain respects; it might be detrimental to society in others. The criterion for economists to evaluate a patent system is its impact on innovation and diffusion economy-wide. A patent regime that would encourage economic growth while reducing the share of income that accrues to inventors could be seen as positive by economists, while legal scholars would focus on preserving that share of income and the inventors rights. Whereas economists focus on ex-ante, incentive effects, legal scholars underline instead the ex-post, reward aspects of the patent system. It is the purpose of this paper to show the economic mechanisms at work in patenting, and consequently, what economists have to say about the strengthening of patent regimes. The first set of changes in patent regimes over the past decades has been the extension of the domain covered by patents, to include biotechnology (DNA, genes, living organisms), software and, in certain countries, methods of doing business. Part 2 of this paper will examine whether patents are necessary in all fields of knowledge, that is, whether the incentives provided by competitive market mechanisms need to be supplemented always by legal monopolies granted by governments to compensate inventors for their investment. As documented by empirical and theoretical studies, the answer is negative. Patents are necessary in certain but not all cases. A note of caution should then be attached to the expansion of the patent subject matter. A second type of change has been concerned with the design of patent rights: notably to a weakening of the criteria for granting patents in certain countries (with the view that more patents is better ), and a tendency to grant patents with broader scope in certain technology fields (as broader patents are more valuable, then better ). Part 3 of this paper will report that economics would not recommend weak criteria for granting patents, and that broad patents might in certain circumstances have detrimental effects on competition and follow-on invention. Part 4 of the paper will address more fundamental aspects of patent policy, such as to determine what type of patent-like mechanism would be more efficient as a policy instrument, in the sense of providing strong incentives to invent while minimising the social costs associated with a monopoly position. This is a rather new area for economists, and research is still highly theoretical. But it clearly emerges that an improved patent regime would be more diversified, escaping the current one size fits all system, and that it would encourage self-selection by patentees as much as possible, making the extension of their rights commensurable with the value of their invention to society. Economic research in the area of patents is relatively new; it has expanded and progressed considerably in recent years. It is not the purpose of this paper to present exhaustively research in this field. We will focus on a restricted number of key policy questions, leaving aside other important issues, such as the appropriate choice of patent regimes according to the state of development of a country, or the political economy of patent regimes. 2

3 2. In what economic contexts are patents needed? The corner stone of the traditional argument in favour of patent protection is the non-rival character of knowledge, and more generally the idea that an invention can be imitated with no additional R&D cost. Economists have long challenged this idea, notably from the evolutionary school (Nelson and Winter, 1982) and arguments against the traditional view have also been recently formalised (Bester and Petrakis, 1998, Hellwig and Irmen, 2000, Boldrin and Levine, 2002, Quah, 2002). If imitation is as costly as invention, or if firms have economic means of protecting their inventions then, the argument goes, there is no need for further legal protection that is only a source of market distortions and associated rents Traditional arguments for and against patents Let us begin by recalling the usual argument in favour of intellectual property protection as it appears in the seminal works of Arrow (1962), Nordhaus (1969) and Romer (1990): the innovation process amounts essentially to the production of knowledge or informational assets, and whereas private goods belong to the class of rival goods, knowledge is inherently non-rival. Indeed, even if knowledge is generally embodied in new physical products or new technologies, the main point is that knowledge presents some features that distinguish it from private goods. The non-rival character of knowledge means that once it is produced, others can subsequently use it without its value being reduced. In other words, consumption of knowledge does not require any additional resources than those devoted to its initial production. This non-rivalry property, satisfied by public goods, is to be contrasted with the rivalry property satisfied by private goods for which the quantity produced must be at least equal to the total number of consumed units. Knowledge is also a non-exclusive good, in the sense that everyone can use it unless a specific property right legally protects it. Thus, insofar as the production of knowledge requires a fixed and indivisible cost in terms of R&D investment, and the goods and services in which the knowledge is embedded can be reproduced and distributed at low marginal cost, perfect competition in the product market does not allow the innovator to recover his initial investment. Whereas the production of knowledge is socially valuable, its non-rivalry and non-exclusivity features make it questionable for a competitive market mechanism to perform properly and public intervention is needed to re-establish private incentives to engage in R&D activities. Patents have been generally considered a valid policy instrument to overcome those problems, as an exante incentive mechanism giving the inventor the exclusive right to use or sell its invention. By imposing a legal exclusivity on the use of knowledge, society encourages ex-ante investment in R&D and thus the production of knowledge and innovation. However, whereas for rival goods strong property rights lead to efficient market outcomes, for non-rival goods a patent involves a trade-off. Weak rights may lead to under-provision of R&D, but strong rights may lead to a monopoly distortion (deadweight loss) and to a reduction in the pace of technical progress, since further innovations are confronted with the obstacles raised by previous innovators (Shapiro, 2001, Encaoua and Hollander, 2002). Patents appear to be a second best solution given that the first best, characterised by a socially desirable level of innovation without market power and with global diffusion, appears to be unreachable. Among the virtues of the patent system, the following properties must be emphasized. First, by giving some exclusionary rights to inventors, society delegates the R&D decision to the inventor and leaves to him the responsibility of recovering his R&D investment. Not only individual agents have better information on the costs and benefits of R&D than the government, but delegation has also the positive effect of avoiding the moral hazard problem on the part of the researchers which may be inherent to the implementation of other instruments such as ex-ante subsidies. Second, the assignment of costs is made to users rather than to tax payers. Third, the implementation of the patent system by the government does not require economic information that is only privately known (R&D cost and private value of the invention). The reward resulting from a patent is linked to the private value of the innovation, so that innovative firms weigh its cost and value when deciding whether to invest. Finally, the information disclosure requirement favours the diffusion of knowledge. 3

4 Among the drawbacks or weaknesses of patents, the following can be mentioned without the aim of being exhaustive. First, patents create static distortions corresponding to the classical deadweight loss that results from inefficient monopoly pricing. Not all consumers valuing goods above their marginal cost can buy them. Second, the market reward from a patented good is not directly linked to the R&D cost needed to develop it. Moreover, inventors cannot fully capture the social value of their invention, in terms of consumer surplus or positive spillovers of their ideas to other researchers, so that patents may provide insufficient incentives to develop socially valuable inventions. Third, a weakness of the patent system is that it does not avoid some duplication of resources, for instance through patent races. Fourth, patents may also distort the direction of research by creating too much incentive to develop substitutes for patented goods and too little to create complements, since by developing substitutes, firms can steal rents from existing patent holders (Eswaran and Gallini, 1996). Fifth, patent application and enforcement require a large amount of financial resources which are diverted from the innovation process itself. Finally, the patent system does not lead to a useful aggregation for efficient decision-making of private information on individual research efficiency disseminated among heterogeneous firms. In summary, patents correspond to an ex-post decentralised solution with many virtues, but also many drawbacks Market-based means of protection for inventions The view that no profit can be made from an invention if it is not protected by legal means, in other words, that market forces are not sufficient to compensate inventors, has been challenged on different grounds for the past two decades in a series of empirical works (Levin et al. 1987; Cohen et al. 2000). Broad empirical evidence supports the view that knowledge used in economic processes is to a large extent difficult or costly to imitate, which contrasts with the traditional view that knowledge is a public good. If knowledge was like any other ordinary rival asset, there would be no need for special protection and competitive rents would provide sufficient incentives to innovate. Companies would be able to charge customers for the additional cost incurred in R&D for the production of new knowledge embedded in their products and processes, without being undercut by lower cost imitators. The market power enjoyed by firms in this context would thus be justified by the nature of technology itself, not by any specific economic or legal device Empirical evidence on the effectiveness of patents across sectors Industrial surveys show that first mover advantages, secrecy and the existence of complementary assets are effectively used by firms to protect their inventions without necessarily relying on legal means of protection such as patents. The 1983 Yale Survey and the 1994 Carnegie Mellon Survey indicated that US manufacturing firms tend to use private appropriation mechanisms, such as exploitation of lead time and the use of complementary sales, manufacturing and service capabilities, in addition to secrecy and patents to capture and protect the competitive advantage provided by innovations (Levin et al., 1987; Cohen et al., 2000). Replicating an existing invention is costly and time consuming because knowledge is to a large extent embodied in individuals, in firms, and in physical goods and equipment. The characterisation of knowledge as blueprint is to a large extent irrelevant in many technology areas, although it might apply to science where many discoveries can be summarised in formulas. In certain industries knowledge is more science-based and codified and hence can be more easily imitated than in others (e.g. pharmaceuticals). However, when knowledge is not codified, imitation is delayed and limited and gives rise to first mover advantages that could allow inventors to recover their cost. Providing protection when imitation is costly may be unjustified, not only because the innovator has enough incentives to introduce new products in the market but also because protection often creates barriers to entry. Lead-time in the market can be sustained if there is further learning, which guarantees the initial inventor permanent advantage over followers. Moreover, secrecy can lengthen the duration of the first mover advantage and reduce the ability of competitors to improve upon initial inventions, and it is to a certain extent protected by law (e.g. trade secrets). Finally, the need to have access to complementary assets in order to obtain profits from a particular invention may also give an advantage to the innovator with respect to imitators. Such complementary assets include marketing (e.g. distribution networks, brand name and reputation, etc.), 4

5 manufacturing facilities and specific competencies. Firms not endowed with these assets face barriers to enter the market and compete on unequal arms length with the innovator. Results from the Yale and Carnegie Mellon surveys also show that firms commonly employ a variety of mechanisms to protect the same invention, and that the frequency of reliance on a certain mechanism differs across industries. Patents are unambiguously reported to be the least central of the major appropriability mechanisms in all industries, significantly below complementary sales and services, but relatively more effective in industries such as medical equipment and drugs, special purpose machinery, computers and auto parts (Cohen et al. 2000). Moreover, patents do not seem to be essential prerequisite for innovation in certain sectors. Mansfield (1986) asked US manufacturing firms what fraction of innovations they would not have developed in the absence of patents in , finding out that it was generally very low, except for pharmaceuticals (60%) and fine chemicals (40%). 2 Using data from the 1994 Carnegie Mellon Survey, Arora et al. (2003) show that the additional payoff obtained from a patented innovation relative to an unpatented innovation (patent premium) differs largely across industries and is positive only in a few manufacturing industries: drugs, biotech, medical instruments, machinery, computers and industrial chemicals Theoretical arguments in favour of market-based mechanisms The question of whether market-based mechanisms (i.e. relying on standard competitive forces, with no special rights interfering) are sufficient for inventors to recover innovation costs is at the heart of the debate opened by recent research. Boldrin and Levine (2002) define the right of first sale as the property right allowing an inventor to use his intellectual asset for a productive purpose or to sell the first unit or prototype to a third party. As for any other good, this property right is perfectly legitimate and commanded by economic efficiency. However, patent protection goes beyond this right of first sale since it also involves the right to control and limit the usage of the intellectual property after sale. This second right is the one conferred by patent protection and corresponds to what they call downstream protection or intellectual monopoly. Boldrin and Levine argue that the right of first sale may be sufficient to obtain revenues reflecting the full market value of the invention (i.e. the net discounted value of the future stream of consumption services generated by the first unit produced) that allows the inventor to recover his R&D investment. In their model, consumers are confronted with the following choice: at each date either they buy and consume the product; or they buy and replicate the product, withdrawing revenues from the inventor. By assumption, subsequent replication involves non-increasing returns to scale, because it takes time to replicate the original prototype and only a limited number of units can be replicated per period. The inventor internalises this loss in the expected revenues he can capture. A paradoxical feature of the model is that a decrease in the cost of imitation generates an increase in the rent of the inventor: the easier it is to imitate an invention, the less protection it needs. Since the technology of production is characterised by nonincreasing returns, it generates infra-marginal or competitive rents that could be sufficient to recover the R&D sunk cost. Thus, even without intellectual monopoly, innovators have incentives to undertake R&D investments in this setting. However, their results depend strongly on certain critical assumptions of the model. First, for competitive rents to be sufficient, demand has to be elastic. When demand is elastic, if the cost of copying the new product decreases over time, due to the introduction of a more competitive technology of reproducing and diffusing the copies of the prototype, output increases more than prices decrease at each date, so that the revenue appropriated by the innovator increases. Second, the inventor is supposed to anticipate the residual demand for its product at each date, without any informational limitation, which is a rather stringent assumption. In addition, the market value of the prototype may be too low to cover the indivisible R&D costs. Given the indivisibility of the initial R&D investment, competitive markets would 2 Less than 10% for firms in electrical equipment, primary metals, instruments, office equipment, motor vehicles and others. 5

6 not provide enough rents to inventors when initial R&D costs are too high. Moreover, although Quah (2002) obtains similar results under certain conditions, notably when the process occurs in a discrete time framework, he obtains a different result when actions are undertaken in a continuous time framework. Quah shows that in a 24/7 Internet-like scenario competitive markets are not sufficient to recover R&D costs. 3 The mechanism proposed by Boldrin and Levine may nevertheless be quite common, as it may amount to protecting inventions with secrecy. When inventors do not diffuse their inventions and protect them by trade secret law, customers have the right to reverse engineer the product and duplicate the invention, which indicates that this could be the advocated diffusion technology in the model developed by Boldrin and Levine. The key question is whether patents are preferable to secrecy. On the one hand patents favour the diffusion of knowledge because of disclosure requirements, but they also have drawbacks. On the other, secrecy prevents disclosure but allows the working of competitive markets. Empirical evidence shows that the balance might be different across sectors. In what follows the special case of financial innovations, traditionally protected by secrecy and recently included in the realm of patentability, is analysed Informational advantages: the case of financial innovations Research undertaken to date indicates that patents are not essential to promote financial innovation. Banks and other financial institutions have developed new products and processes over the past four centuries in the absence of legal protection mechanisms, and financial methods only started to be extensively patented in the US after the State Street Bank decision in The question is how innovators recover costs in a sector where imitation occurs relatively rapidly. An empirical study on the incentives to innovate in financial markets shows that underwriting spreads on first offerings were not much larger than on late offerings for a sample of financial services introduced by investment banks between 1974 and It also shows that although lead-time was relatively short for innovators, they usually enjoyed higher market shares than followers (Tufano, 1989). 5 Consistent with these empirical findings and additional information from interviews with credit derivatives dealers and developers in investment banks, a theoretical model developed by Herrera and Schroth (2000) concludes that the incentive of investment banks to launch new financial products comes from the informational advantage provided by being the first in the market. Lead-time is important because it provides an informational advantage in a setting where learning by doing is crucial, rather than because of the short-term monopoly profits it may confer to the innovator until its competitors enter the market. In fact, Herrera and Schroth argue that the threat of imitation prevents the innovator from charging monopoly profits in the learning stage, when he is the only supplier. The innovator is able to recover investment costs because his informational advantage and expertise enables him to charge higher prices and enjoy larger market shares than his competitors at the competitive stage. Information is essential in a market where profit depends on risk management, so although the design of a financial product is fairly easy to imitate, its optimal exploitation can only be imperfectly imitated because it requires some expertise that is carefully protected by the innovator with secrecy. 6 3 Bester and Petrakis (1998) obtain similar results to Boldrin and Levine in a partial equilibrium framework, and Hellwig and Irmen (2000) in a general equilibrium framework. 4 Since then USPTO has issued patents on financial inventions on asset valuation, debt management, education finance, mortgages, privatisation, risk assessment, stock picking and working capital finance, among others (Thomas, 2001). 5 Recent empirical findings by Carrow (1999) confirm that leaders tend to have larger market shares than followers but show that prices decrease as the number of competitors increases when the reputation of sellers and specific characteristics of the products compared is introduced into the analysis. 6 One of the persons interviewed by Herrera and Schroth, a J.P. Morgan credit derivatives trader, explained that the economic value of a financial product lies in information about its performance, rather than in its design: Everybody can see the laid-out contract but what I am very careful not to disclose are the positions in my book. With this information you could track down the logic and see where I make money. Without it you could not price correctly the 6

7 2.3. Sequential innovation: the case of software When innovation is cumulative, the exclusive rights provided by patents may impede access to the knowledge embedded in previous inventions and slow down technological progress. Some authors argue that this is the case for software innovations, which have been nevertheless increasingly patented since the 1980s. Prior to patents, copyright and market-based mechanisms such as the offering of complementary assets and sales proved to be an efficient way to recover investment costs in the software industry. Even if the initial rents earned by an innovator in the absence of patents may be lower than with patents, the benefits that accrue to him when he is allowed, in his turn, to build around the next innovation made by a competitor may outweigh the current loss. Bessen and Maskin (2000) argue that in a dynamic setting, firms are better off when they are imitated and have competitors in the market because that increases the probability of further innovation. They state that although patents preserve innovation incentives in a static world; in a dynamic world, firms may have plenty of incentives to innovate without patents and patents may constrict complementary innovation. 7 They use evidence on the lack of increase in R&D intensity of top US software patentees after the expansion of patent subject matter in the US to software in the 1980s to illustrate that patents are not per se a good instrument to spur innovation in the software industry. They argue that where innovation is sequential and complementary imitation becomes a spur to innovation, while strong patents become an impediment. 8 In addition, using a regression model of the cost of patenting, Bessen and Hunt (2003) argue that the growth of patents together with a decrease in R&D intensity in the software industry is consistent with strategic patent portfolio behaviour. They claim that manufacturing firms are patenting relatively more software inventions than software publishers because they have imported patent portfolio strategies from their traditional sectors of activity to the software sector. They also argue that this behaviour is one of the main reasons behind the increase in software patents, as it has also been noted by Hall and Ziedonis (2001) as regards the increase in the number of patents in the semiconductor industry Essential research inputs: the case of biotechnology A particular innovation lacking economic value on its own (i.e. having no direct industrial application) may nevertheless be essential as a research input for highly valuable innovations in the future. Basic discoveries, genetic material and research tools in general, which can be considered as essential facilities for follow-on innovations, come under this category. Patent protection was effectively expanded to biotechnological and genetic inventions at the beginning of the 1980s in the US. Also at that time, the Bayh-Dole Act reinforced patenting and licensing incentives for federally funded research in the US, which to a large extent consists of basic discoveries made at universities and other public research organisations (Mazzoleni and Nelson, 1998). The underlying argument for the introduction of the Act was that although federal funds are necessary to support fundamental research, its development into commercial applications needed additional R&D investment that would only be funded privately if patent protection allowed firms to appropriate the economic return of their exploitation. A recent paper by Walsh, Arora and Cohen (2002) provides empirical evidence on the impact of patenting and licensing research tools in the field of biomedical innovations. The authors note that advances in molecular biology, automated sequencing techniques and bio-informatics have led biomedical research to increasingly depend on prior scientific discoveries or research tools. They also note that the role of product, break down the risks involved, and understand what the components are. New ideas are not easily imitated: the developing process is a set of complex skills that are not easy to acquire. 7 Bessen and Maskin, 2000, p By sequential we mean that each successive invention builds on t he preceding one in the way that Windows built on DOS. And by complementary, we mean that each potential innovator takes a somewhat different research line and thereby enhances the overall probability that a particular goal is reached within a given time (Bessen and Maskin, 2000). 7

8 universities has become more important in the past two decades, as sources of both patented biomedical inventions and start-up firms that are often founded on the strength of university-origin patents. Despite the obstacles posed by patents in an area where scientific discoveries were traditionally placed in the public domain until the 1980s, they note that firms seem to have developed working solutions that combine taking licenses, i nventing around patents, infringement (often informally invoking a research exemption), developing and using public tools and challenging patents in court). 9 The question is whether the benefits of patenting outweigh the cost in terms of resources needed to find such a working solution Conclusion: patents should not be the solution by default Patents are not needed as an incentive mechanism for all types of inventions. A series of characteristics of technologies and markets such as the importance of market-lead, the arrival rate of innovative ideas, the ease of imitation and the more or less marked sequential character of innovation, determine whether market-based means of protection provide sufficient incentives for innovation and the impact of patents on the pace of innovation. The optimal level of patent protection may differ across fields, with different solutions applying to industries as diverse as pharmaceuticals, software and finance. However, patents serve also other purposes than excluding competitors from the market, and these purposes should be taken into account when evaluating the relevance of patent as a policy instrument. Patents are used as an argument in negotiations for cross-licensing agreements, as a signalling mechanism for shareholders, banks, venture capitalists, competitors or customers. They also contribute to social welfare by facilitating the diffusion of knowledge through information disclosure requests and allowing the development of markets for technology. When the invention can easily be kept secret, as in the case on process innovation, patent might be socially beneficial for diffusion of knowledge even it is not necessary as an incentive. These factors seem to differ across technology fields and to a large extent economic theory falls short of addressing these other private and social motives for patenting that may lead to social welfare gains. 3. Instruments of patent policy design In what follows we consider some policy issues arising in the current patent debate. Although they do not exhaust all the problems raised by the evolution of the patent system, they are nevertheless of prime importance. Recent trends in the patent system may indicate a weakening of patentability requirements and an increase of patent breadth, at least in certain jurisdictions and technology fields. The purpose of this section is to inquire whether these trends are justified from an economic perspective. 3.1 Patentability requirement: non-obviousness Patent offices grant patents to innovations which comply with the patentability criteria of utility, novelty, and non-obviousness. 10 Non-obviousness is a technical concept meaning that the invention should not be obvious in itself or a simple combination of existing techniques that is obvious to somebody skilled in the relevant art. Some observers argue that the non-obviousness requirement has been quite low in the US in recent years, especially in new areas such as software and computer-implemented business methods, where experts argue that a large number of US patents have been granted on inventions that were neither obvious nor novel. 11 The effective grant rate at USPTO, including continuing applications, has been very high in recent years, and was between 87% and 97% in the mid 1990s (Quillen and Webster, 2001). Moreover, new intellectual property rights with lower granting requirements that provide weaker protection than patents have been introduced in the past few years in the US, such as the 1984 US Semiconductor Chips 9 See research exemptions in the section on access to patented knowledge by subsequent inventors below. 10 Utility is called industrial application, and non-obviousness is called inventive step in Europe

9 Protection Act, which may indicate a generalised trend towards a reduction in the threshold of nonobviousness to qualify for some form of intellectual property protection. 12 In order to analyse the economic impact of lowering the patentability requirement, it should be noted that there are some differences between practice at patent offices and the approach taken in the economic literature. Whereas patent examiners define non-obviousness according to the technical character of the invention, by reference to a person with ordinary skill in the art, economists interpret it as a threshold under which the reduction in costs for process innovations or the degree of quality improvement for product innovations would be insufficient for the patent to be granted. For economists, the scheme is a one-dimensional scale on which all techniques of the relevant domain can be measured, according to some performance feature (e.g. cost, quality). Whereas the technical characteristics of an invention are usually known ex-ante, economic features are only known ex-post, after the technique has been actually implemented, giving rise to products sold on the market. For instance, a new technical device, which does not improve performance with respect to a currently used technique, may satisfy the patentability requirement on the examiners scale, but it would not be acceptable prima facie on the economists' scale because it provides no direct or immediate gain to society. Traditional tests of non-obviousness are primarily based on the following technical features: i) scope and contents of prior art; ii) differences between prior art and patent claims; iii) level of ordinary skills in the relevant art. However, in the US, a set of secondary economic factors have been recently emphasised in some decisions of the US Court of Appeals for the Federal Circuit, such as commercial success, failure of others and long felt need (Hunt, 1999). EPO also considers commercial success as relevant information, but subordinate to technical criteria: Commercial success alone is not to be regarded as indicative of inventive step, but evidence of immediate commercial success when coupled with evidence of a long-felt want is of relevance provided that the examiner is satisfied that the success derived from the technical features of the invention and not from other influences (e.g. selling techniques or advertising). 13 Merges (1992) notes that the use of commercial success as a test of non-obviousness may have the drawback of rewarding firms with better marketing and distribution systems rather than to those introducing more innovative products in the market. What is the advantage of imposing non-obviousness requirements? In other words, why not grant patents to all inventions? The answer is that strong patentability requirements provide higher incentives to innovate by extending the effective life of patents, that is, the length of market incumbency for the inventor (O Donoghue, 1998, Hunt, 1999, O'Donnoghue and Zwimüüller, 2000). To analyse the effect of the patentability requirement on the pace of innovation, Hunt (1999) considers a situation where successive quality improvements occur and research activity has uncertain outcomes so that the size of innovations is exogenous and stochastic. At each point in time, only the highest quality product is profitable ( winner takes all ) and for an improvement to be patented, it must exceed some threshold value, which corresponds to the patentability requirement. In addition, unless it is patented, an invention is in the public domain, so that a non-patented product would be immediately copied by all firms and therefore would have a zero return. A patented technology is fully disclosed and serves as a basis for further improvements. A number of firms are racing to improve existing products and it is assumed that decreasing returns on R&D prevail for each firm. In this setting, the net effect of increasing the patentability threshold on the pace of innovation depends on the magnitude of two opposing effects. The short-term effect is negative. Indeed, increasing the 12 Another example is the new system of innovation patents introduced in Australia in 2001 to replace the previous system of petty patents, which is intended to provide protection to incremental and lower level inventions that would not be sufficiently inventive to qualify for standard patent protection EPO guidelines, Part C, Chapter IV, page 72. 9

10 patentability requirement lowers the probability that an innovation is qualified for a patent, reducing therefore the short-term incentive to innovate. In contrast, the long-term effect is positive for two reasons. First, an increase of the patentability requirement leads to a longer effective patent life of an innovation, since it will be replaced by improved techniques later. Second, it also leads to a higher average profit flow from a patented discovery. Hunt (1999) shows that the balance of these two opposing effects is in favour of the positive effect, insofar as the patentability threshold is not too low. He shows that there exists an inverted U relationship between the increase of the patentability requirement and the rate of innovation. Innovation is first positively affected by the height of the patentability requirement but beyond a certain level the relationship changes and innovation decreases. O Donoghue (1998) obtains a similar result considering that the size of innovations is endogenous, and a higher patentability requirement extends the effective life of patents by inducing firms to pursue more ambitious R&D projects. Moreover, Hunt (1999) shows that the optimal patentability requirement increases with the arrival rate of innovative ideas. He argues that in rapidly innovating industries, a smaller proportion of inventions can be protected by patents without causing the rate of innovation to decline. Similarly, reductions in the patentability requirement are more likely to encourage innovation in industries that innovate slowly. Increasing the patentability requirement may encourage larger and riskier inventions, something that is socially desirable when markets tend to favour smaller inventions with more certain rewards. Merges (1992) argues that patents should be used to encourage firms to engage projects with low certainty of commercial success, as inventions with more certain gains would be also implemented in the absence of patent protection. Optimal patentability requirements are higher when technical change is more rapid or innovative ideas arrive more frequently because in that case the length of the incumbency period is shorter, and thus the opportunity cost of not winning the patent race decreases. An implication of this finding is that the factors affecting the optimal level of the patentability requirement are technology specific, whereas the current patent system is characterised by uniform rules, according to the one size fits all principle. The reduction in the patentability requirement experienced in certain countries over the past two decades might have favoured industries where technical change is slow at the expense of more dynamic areas. This is all the more important since patentability requirements seem to have particularly decreased in new areas of patentability such as biotechnological or software-related inventions, generally characterised by high frequency of arrival of innovative ideas where the optimal patentability requirements should be higher. Moreover, if imitation were assumed to be costly even for a fully disclosed invention, the positive effect of increasing patentability requirements on innovation would be reinforced and the optimal level of the patentability requirement would be higher. Indeed, if the incumbent cannot freely capture a non-patented technology, an innovator would need more incentives to undertake an R&D project. Increasing the patentability requirement would give such an incentive by extending the effective life of the incumbent s patent Patent breadth and the effective life of patents A patent holder has the right to prevent others from making, selling or using the invention protected by the patent, so that the breadth of a patent is measured by the set of products that courts would find to infringe the patent. In principle, patent breadth is determined by the claims accorded by the patent examiners to the patentee, defining the boundaries between what is protected and what is not, and by the courts interpretation of these claims during litigation procedures. Policy instruments affecting patent breadth may involve additional fees on patent applications in which the number of claims exceeds a certain threshold. In addition, courts may rely on doctrines that either expand patent breadth, such as the doctrine of equivalents, or restrict it, such as the enablement doctrine. Whereas the doctrine of equivalents states that essentially equivalent products not explicitly included in the patent claims could be judged to infringe the patent, the enablement doctrine states that only what is disclosed is protected by the patent. The existence of discrepancies between the technical scope of discoveries and the claims granted by patent protection is a currently debated issue, notably in biotechnology (Bar-Shalom and Cook-Deegan, 2002). 10

11 The surveyed literature on patent breadth shows that economic models introduce more policy instruments than those available in the real world. While the economic literature emphasises the trade-off between patent length and breadth, in practice one of these instruments is fixed since the statutory patent length is uniform across industries and countries. The fine-tuning of patent policy that can be achieved in economic models is more difficult to reach in reality, and under the constraint on the uniform statutory patent life, broad patents may do more bad than good to innovation. Nevertheless, the main conclusion to be drawn from the economic literature is that broad patents may distort incentives and allocation of research funds. While broad patents increase the rents accruing to inventors, they might generate rent-seeking behaviours if the patent office is not entirely successful in screening the legitimate claims of each patent application. In other words, broad patents increase the social cost of imperfections in the management of the patent system. In addition, broad patents tend to skew the reward distribution associated with research. Large breadth makes research resemble more a "winner takes all" game, as modelled in patent races. One drawback of such a scheme is that it may lead to duplication and concentration of R&D efforts in some areas at the expense of insufficient investment in other areas where the return is lower. Another conclusion emerging from this literature is that the trade-off between patent breadth and length depends on industrial characteristics such as the level of technological opportunities or the rate of arrival of innovative ideas. Both patent breadth and length determine the extent of patent protection and affect the expected revenues from patenting an innovation, but they work in different ways, with different effects on the economic behaviour of the patent holder and his potential competitors. Whereas broad patents allow the patent holder to set a higher market price for the patented product, patents with longer lives allow patent holders to obtain revenues for longer periods of time. Larger breadth makes it more difficult to imitate or improve upon the protected invention, whereas increasing the duration of patent protection enhances the incentives to imitate or to improve the invention (Gallini 1992). The effects of patent breadth and length may differ according to whether an innovation is considered in isolation or as part of a sequence of innovations building on each other. The following two sections describe separately economic research related to these two scenarios Isolated innovations The literature on patent design examines isolated innovations by emphasising the traditional trade-off between creating incentives to conduct R&D and reducing the deadweight loss resulting from strong patents. It identifies patent breadth with lagging breadth, understood as the protection accorded against imitators. For instance, for a quality product innovation, lagging breadth defines the minimal product differentiation that must exist between the patented product and an inferior product so that the latter does not infringe the patent of the former. For a process innovation leading to a cost reduction, lagging breadth defines the maximal cost reduction allowed to a non-infringing inferior process. In this setting, optimal patent breadth is defined by minimising the discounted value of the deadweight loss created by the patent under the constraint that the discounted profit provides enough incentives to invest. A number of authors have studied the optimal mix of patent breadth and length in the context of isolated innovations, but no consensus has been reached and results depend heavily on the different assumptions of their models as regards the relationship between patent breadth and deadweight loss (Nordhaus, 1969, Scherer, 1972, Tandon, 1982, Gilbert and Shapiro, 1990, Klemperer, 1990, Waterson, 1990, Gallini, 1992, Denicolò, 1996). Gilbert and Shapiro (1990) define patent breadth as the minimum price that provides the patentee a prespecified reward allowing him to recover the R&D cost (participation constraint). In this case, minimising the deadweight loss under the participation constraint calls for a long and narrow patent, where patent breadth is such that it just allows the specified reward. This result crucially depends on the fact that the deadweight loss increases more than proportionally with breadth (i.e. it is a convex function of breadth). Tandon (1982) obtains the same qualitative result in the context of compulsory licensing. 11

12 Klemperer (1990) obtains different results by developing a model with horizontal product differentiation, linear disutility costs and price competition. In his model, patent breadth is defined as the set of characteristics in the product space that infringe the patented product, so that consumers may be deprived of their ideal products if these products happen to infringe on an existing patented product. In this context, Klemperer shows that each one of the two combinations of patent length and breadth can be optimal, either long and narrow patents or short and broad patents. Short and broad patents are also found to be optimal in Gallini (1992). If an invention that reduces unit cost by a certain amount is patented, breadth is defined as the maximal cost reduction that an imitator can achieve to avoid infringement. Larger breadth is associated with higher imitation cost. With imitation, the social cost of patents has now two components: the deadweight loss and the duplicative waste resulting from imitation. Gallini shows that short and broad patents are optimal because they deter socially wasteful imitation. However, this result has been criticised by Maurer and Scotchmer (2002) because it depends on an assumption about the absence of licensing. They note that licensing agreements between the innovator and potential entrants could also avoid the duplicative waste induced by imitation and reverse the optimal design. Even if no conclusive results stem from the literature on the optimal mix between length and breadth for isolated innovations, there is a strong presumption that narrow and long protection is a preferable combination, when more realistic features such as patent races and richer descriptions of competition in product markets are introduced in the model (Denicolò, 1996) Sequential innovations The key question in a sequential innovation setting is how to increase the rate of innovation by designing the patent system in terms of length and breadth. Suppose that ideas for improving a current technology occur randomly and the level of improvement is itself random. Not only the initial and subsequentinnovators must receive sufficient rewards to recover their costs but also these rewards must take into account the fact that each innovation opens the way to further improvements. By deciding whether an improvement infringes or not a previous patent, the courts determine a patent regime that has an important effect on the rate of innovation. The economic analysis of this effect investigated in some recent papers, as set out below (O'Donoghue, Scotchmer and Thisse, 1998, O'Donoghue, 1998, O'Donoghue and Zweimuller, 2000). O'Donnoghue et al. (1998) show that in a cumulative setting, lagging breadth alone (even when combined with an infinite statutory patent length) does not provide sufficient incentives for R&D. Indeed, without leading breadth, firms under-invest in R&D. Lagging breadth offers protection only against imitators and not against future innovations; therefore, no future innovation will infringe a previous patent in a regime where forward protection is absent. As a consequence, the concept of leading breadth is introduced in the setting, understood as forward protection that limits the behaviour of subsequent innovators. 14 O'Donoghue et al. (1998) also analyse the problem of the optimal mix of leading patent breadth and patent length for sequential innovations, starting from the observation that a specified rate of innovation can be achieved by one of two different patent regimes. The first one involves a long statutory patent life and a weak leading breadth; the second, a short statutory patent life and a broad leading breadth. Under the first regime, the effective patent life is determined by the time elapsed until a sufficiently better product is invented. Under the second regime, the effective patent life coincides with it statutory length and no new patentable innovation occurs before the expiration date of the previous patent. One of the main conclusions obtained by comparing the two regimes is that the regime combining a long patent life and a narrow leading breadth is superior in the sense that it reduces total R&D costs, while a regime with short statutory patent life and broad forward protection leads to a smaller patent life than the expected patent life induced by the arrival rate of innovations. 14 Note that leading breadth elicits investment in infringing innovations, insofar as they can be licensed to the market incumbent. 12