idea patent

The economics of patent design: A selective survey Ryan Lampe and Anthony Niblett Intellectual Property Research Institute of Australia Working Paper No. 06/03 ISSN 1447-2317 March 2003 Intellectual Property Research Institute of Australia (IPRIA) Law School Building The University of Melbourne Victoria 3010 Australia Telephone (03) 8344 1127 Fax (03) 9348 2353 Email j.molloy@unimelb.edu.au WWW Address http://www.ipria.org 1 Introduction and background to patent rights 1.1 Introduction This paper selectively surveys the economic literature concerning patent design. We provide non-technical summaries of major papers, focussing on four key areas of research: (1) the optimal choice between patent length and breadth in a static environment; (2) the optimal choice between patent length and breadth in a dynamic environment; (3) the optimal structure of renewal fees; and (4) the optimality of granting patents for business methods and computer programs. In each of the above areas we have tried to select several key articles and discuss their findings and methodologies. research into patent design. Though the majority of articles in this literature are Hopefully this survey will provide an interesting theoretical, we have included a fifth section considering some of the recent empirical introduction to what has become a very rich field of economic research. 1.2 Background to patent rights Patents grant the patentee a temporary monopoly. This monopoly is conferred by law in order to generate monopoly profits that reward the inventor and provide incentive for research and development. Hence patents seek to overcome the inefficiencies associated with the under-production of knowledge.1 However, the conferral of a temporary monopoly may have significant social costs. Monopolists can overcharge for their product and, consequently, reduce the quantity of their product traded. Therefore patents can defer the arrival of improving innovations dependent on the patented innovation. An invention is patentable if it satisfies threshold tests required by the Patent Act 1990 including: 1 Because knowledge has public good characteristics, in the absence of a property rights such as the patent system, we would expect free-riding on the production of knowledge and too little knowledge produced from a social perspective. 1. the invention must be a ‘manner of new manufacture’;2 2. the invention must be ‘novel’ when compared to the prior art base;3 and 3. the invention must involve an ‘inventive step’ and not merely an advance that would be obvious to a person skilled in the field of the invention.4 Patent examiners will frequently reject an application at least once, usually because the scope of the claim is either unclear or too broad (usually because it is not ‘fairly based’5 on what is disclosed in the description or because it transgresses what is already known in the prior art.) 2 Patent length and breadth 2.1 Background The optimal patent length is modelled by equating the marginal social benefit, MSB(t), with the marginal social cost, MSC(t), of the patent over time. The social benefit of granting patents, as described above, is the increased level of inventive activity. The MSB(t) is the value of the incremental inventive activity generated by the patent in year t. It is assumed to slope downwards. This reflects the idea that the additional creative activity stimulated by the patent will be reduced as the length of the monopoly rises. As an example, extending the life of a patent from 10 years to 11 years will have a positive effect on the amount of investment activity. Extending the patent length from 20 years to 21 years will also have a positive effect, but the effect will be smaller. The social cost of patents, as described above, is the loss associated with the monopolistic behaviour of the patentee and the loss created by the reduction in improvements upon the patent (this dynamic effect is discussed in the next section). The MSC(t) is the value of 2 Patents Act 1990 (Cth), s 18(1)(a). See also, relevant case law, esp: National Research Development Corporation v Commissioner of Patents (1959) 102 CLR 252. 3 Patents Act 1990 (Cth), s 18(1)(b)(i). The assessment of novelty is legislatively provided for in s 7(1) of the same legislation. 4 Patents Act 1990 (Cth), s 18(1)(b)(ii). The assessment of inventive step is legislatively provided for in s 7(2). 1 the incremental harm to society in year t. The MSC(t) curve is assumed to slope upwards. This reflects the idea that the additional social cost incurred by extending patent length by one year increases over time due to an absence of substitutable products. The optimal patent length, as discussed above, is found by equating the MSB(t) and MSC(t). This year is denoted by t*. Extending the life of the patent beyond t* means that for each year beyond the optimal level, the incremental cost to society is greater than the additional benefit conferred by the patent. That is, the incremental value of the inventive activity will be outweighed by the additional costs of monopoly pricing and less improvement upon the patent. Conversely, if the length of the patent is shorter than the optimal length, then society will miss out on net benefits. Prior to 1995, the duration of a standard Australian patent was 16 years. However, this was changed as a result of Australia’s obligations under the Trade-Related Aspects of Intellectual Property Rights (TRIPs) in the Uruguay Round of Negotiations for the World Trade Organisation. Now, the length of a standard patent in Australia is 20 years.6 Is the optimal length of all patents in Australia 20 years? Under the simple model expounded above, almost certainly not. The benefits and costs deriving from any patent will inevitably depend on the particular invention that has been patented, and the nature and characteristics of the relevant industry. Ideally patent lives should be variable and reflect the optimal duration for each given invention. However, it is a requirement of the TRIPs agreement that patent term is a minimum of 20 years. Since the Intellectual Property and Competition Review Committee does not believe a case has been made for further extending the maximum patent term, it recommended in its 2000 Review of Intellectual Property Legislation under the Competition Principles Agreement that IP Australia consider implementing steeply rising renewal fees for patents to reduce the effective term of patents. In regard to patent breadth, a simple economic interpretation of breadth is that it reflects the extent to which innovations are immune to competition. Increasing patent breadth 5 6 Patents Act 1990 (Cth), s 40(3) Section 67 Patents Act 1990 (Cth). 2 therefore increases dead-weight loss because it lessens substitutability. However, reconciling this economic interpretation with a legal interpretation is difficult. It would seem that increasing patent breadth would mean increasing the difficulty in meeting the threshold for ‘novel’ and ‘non-obvious’. A patent with infinite breadth, for example, would mean that any product within that patent’s market (given an adequate definition of market) would be infringing. interpretations of patent breadth. The literature considered below offers varying 2.2 Recent economic literature Prior to 1990, research into patent design considered only the optimal lifetime of patents.7 However, more recent research has considered optimal patent design in terms of patent length as well as patent breadth. Gilbert & Shapiro (1990) and Klemperer (1990) were the first to model patent policy as a choice of these two instruments. A key result arising from these papers is that the optimal patent length may easily be infinite. Research following these papers by Gallini (1992), Denicolo (1996) and Wright (1999) have made only minor modifications to Gilbert & Shapiro and Klemperers' work. The seminal paper, Gilbert and Shapiro (1990), considers that the optimal mix between patent length and breadth rewards innovation. The authors do not consider how much to reward patentees. They simply consider how to structure each given patent reward. Their main result is that when patent policy is addressed as a choice between patent breadth and length, the optimal length may be infinite. Ultimately, their result reflects their definition of patent breadth. The authors define patent breadth to be the ability of the patentee to raise price. Hence increasing patent breadth increases deadweight loss (assuming a downward sloping demand curve). The authors define the second derivative of welfare with respect to profit as decreasing. So, not only is increasing breadth costly to welfare, it is increasingly costly to welfare. By contrast, increasing the length of the patent results in a constant trade-off between the 7 See, for example, Nordhaus (1969). In his seminal work, Nordhaus argued that the duration of the patent should maximise the discounted sum of consumers and producers surplus. It follows that the optimal patent length is longer if the demand curve is inelastic and the more responsive the level of invention is to an increase in R&D expenditure. 3 additional reward to the patentee and the increment to deadweight loss. Thus, the socially cost-effective way to achieve a given reward to innovators is to have infinitelylived patents with the minimum market power (breadth) necessary to achieve this reward. The authors cite a number of limitations with their result. First, it assumes the underlying environment is predictable. If there is uncertainty about the future, risk averse investors will prefer shorter and broader patents than risk neutral investors. Second, the result (not surprisingly) breaks down if the second derivative of welfare to profits is not negative. Klemperer (1990) provides another limitation of the Gilbert and Shapiro result, analysing a similar trade-off, finding that the optimal patent policy may also involve infinite breadth and minimum patent length. Like Gilbert and Shapiro, Klemperer chooses the patent breadth that minimises the deadweight loss per dollar of profit generated by the patent and then multiplies this by the lifetime required to generate the correct total profit. However, unlike Gilbert and Shapiro, Klemperer supposes that increasing the scope of the patent grant makes non-infringing substitute products less attractive to consumers. In the former, patent breadth has no effect on the set of substitute products that are offered to consumers; breadth only affects the price that the patentee can charge. In Klemperer's model, a patent of width (referred to as 'radius') w allows competing firms to produce product varieties a distance w from the patent holder's product. Each consumer has a substitution cost of t per unit distance per unit purchased of substituting toward alternative varieties for the patent holder's good. Hence consumers will prefer the patent holder's product only if tw is greater than or equal to price. Under Klemperer's model there are two types of social cost that depend on patent breadth. For sufficiently wide patents, the ratio of total social costs to profits depends only on the demand curve, while for sufficiently narrow patents the ratio depends only on the distribution of substitution costs. For wide patents the social costs that matter are the deadweight losses of a monopoly. For narrow patents, the only social costs that matter are the substitution costs incurred by consumers substituting away from the patent holder's product. 4 With this model, Klemperer considers two scenarios. First, he supposes that all consumers have identical substitution costs. This means that all consumers have the same preference between the patent holder's product and the competitively supplied variant. Hence the patent holder sets a price such that no-one substitutes toward the competitively supplied variant. With no substitution costs, the only social costs are the standard deadweight losses. Thus, Klemperer derives the Gilbert and Shapiro result that the narrowest possible patent is the most socially cost-effective way of awarding profits to an innovator. Klemperer does not find this surprising. In a homogeneous-product market (as in Gilbert and Shapiro), there is no possibility of consumers substituting related varieties of the product, so only standard deadweight losses are relevant. Under an alternative scenario, Klemperer supposes all consumers have identical reservation prices for the patented product. Here the patent holder sets price so that no consumers substitute out of the product class (the alternative is for the patent holder to sell no units). Therefore, the only losses are of consumers switching away from the patented product. Hence a patent with infinite breadth and minimum length is optimal. This result contrasts strongly with Gilbert and Shapiro’s finding and stems from the existence of substitute products. In Klemperer's model, patents that are wide in scope and short in duration can be preferred to patents that are narrow in scope and long in duration if wider patents discourage costly substitution away from the patented product to the inferior substitute product. Gallini (1992) finds that small patent lives are optimal in a different context. She considers the case where the innovation can be perfectly imitated at a cost whose size depends on the breadth of the patent. With a homogenous product and price competition, clearly no imitation would occur in equilibrium. Under different assumptions about the nature of competition prevailing in the product market (for instance, Cournot), however imitators would enter until their profits are driven to zero. Hence Gallini shows that broad patents may be optimal if they lower socially wasteful imitation costs. Denicolo (1996) extends the above theoretical framework by considering the case where many firms race for the patent. As in Gilbert & Shapiro and Klemperer, the social 5 problem involves minimising the ratio of the deadweight losses associated with the patent to the innovator's profits. In Denicolo's framework, the denominator of the ratio is an expression which measures the firms' incentive to innovate. It involves the profits earned by the race-winners, the non-innovators and the profits earned by every firm after the patent expires. This modification allows Denicolo to consider several examples that confirm a variety of possible results. The concept of breadth adopted in Denicolo is the degree of technical knowledge that the patent permits. If breadth is zero, then maximum protection against imitation is conferred. In his model, the equilibrium level of R&D is determined by the "profit incentive" (the difference between the patentee's profits and its pre-innovation profits) and the "competitive threat" (the difference between the profits to the winner and to the losers). Firms choose R&D effort to maximise expected profits. The probability of success is increasing (at a decreasing rate) in effort and their payoff is the present value of expected profits, net of R&D costs. It is assumed that the profit of the winner is higher than the profits losers receive and everyone receives an intermediate profit level after patent expiration. To derive a problem involving choice of patent breadth and length only, Denicolo supposes that the socially desired R&D effort is predetermined.8 of the patent while generating a given incentive to innovate. The optimal patent design depends on the derivative of the above ratio (between deadweight losses and incentive to innovate). If the second derivative of social welfare is decreasing in patent breadth and the second derivative of incentive is also decreasing in patent breadth (and at least one is a strict inequality) then the optimal patent has a maximum breadth and minimum length. If these second derivatives are reversed then the optimal patent has minimum breadth and maximum length.9 If social welfare and incentives are linear in breadth (second derivative is zero) then social welfare does not depend on the breadth-length mix. Denicolo notes that if the losers of the R&D race have 8 With a fixed number of firms in the R&D race, this means that the equilibrium R&D investment must equal a predetermined level. 9 It is assumed that social welfare is decreasing in patent breadth and profits are increasing in patent breadth. Patent breadth and length are chosen to minimise discounted deadweight loss over the lifetime 6 zero profits until the patent expires, the Gilbert and Shapiro result is derived. The author illustrates how these different results could occur with specific examples (such as the case of a vertically differentiated industry where consumers have utility functions based on quality). Despite the appeal of representing profit in terms of R&D effort, Denicolo's results do not extend greatly the results of Gilbert & Shapiro and Klemperer. Clearly, the patentbreadth optimal mix depends on the second derivatives of social welfare and postinnovation profits to patent breadth. Economic theory does not place any restriction on the concavity/convexity of these functions. As Denicolo notes, depending on how you specify these functions, either Gilbert & Shapiro or Klemperer’s results can be derived. If it is assumed that the additional competition brought about by narrowing the patent is on balance socially costly (e.g. consumers bear substitution costs), it is optimal to award patents of maximum breadth (Klemperer's result). If it assumed that a reduction in patent breadth increases social welfare more than it reduces the incentive to innovate of firms participating in the patent-race, Gilbert and Shapiro’s result can be derived. An alternate definition of patent breadth comes from Wright (1999) which defines it using the number of potential entrants that enter an industry. As before, the policy maker's problem is to choose patent breadth and length to minimise the ratio of deadweight loss to profit. However, in Wright's model both deadweight loss and profit are functions of the number of firms. Wright considers a Cournot market structure with linear, concave and convex demand curves. With a linear demand curve, entry reduces the deadweight loss and innovator profit in the same proportion so any patent length and breadth combination is optimal (the ratio is constant). With a concave demand curve, deadweight loss falls less than proportionately to innovator profits so the ratio is increasing in the number of entrants so the policy makers wants to make the patent as broad as possible. Supposing that the demand curve is convex, the optimal patent is extremely narrow. In Wright's model the residual demand curve shifts in as patents are narrower. This contrasts with Gilbert and Shapiro’s model where no imitation could take place making 7 the patent holder's residual demand curve the industry demand curve (i.e. he/she is a monopolist). Assuming products are not perfect substitutes and using a linear demand structure, Wright shows that the policy maker's problem is to minimise a ratio that is strictly increasing in the number of firms. In-line with Klemperer’s results, extremely broad short-lived patents are optimal. The insight here is that market structure may be important in determining the optimal patent design. Clearly, the assumption that the policy maker can influence the number of entrants is unrealistic. Hence Wright supposes instead that the policy maker can influence the cost of imitation and then allow the number of entrants to be determined endogenously (i.e. smaller imitation costs allow more entrants to enter). Potential entrants are ordered according to increasing imitation costs and depending on the structure of imitation costs, patents can be very broad, very narrow or somewhere in-between. Note, imitation costs are modelled differently to Gallini (1992) who found that extremely broad patents were optimal. An obvious limitation of the above literature is that it assumes the underlying environment is static. However, inventions build on each other. So long patent grants may have negative effects on the incentives of other firms looking to engage in related research. The next section considers research that addresses this short-coming. 3 Patent design with sequential innovations 3.1 Background An important effect of temporary exclusive patent rights is that they may inhibit innovation based on the ideas involved. That is, while strong patent protection protects initial innovators, it discourages sequential innovators. Until recently, economic treatments of patents have assumed that innovations are isolated events with little impact on future innovations. However, it is obvious that not all industries are stationary. Biotechnology and computing hardware and software, for 8 example, have a high degree of ‘cumulativeness’. The following articles consider the patent design in a dynamic environment. 3.2 Recent economic literature Recent literature on patent design with sequential innovation has produced some interesting results. Green and Scotchmer (1995) analyses a game between two players, a first-generation innovator and a second-generation innovator, and show how the latter can appropriate profits from the former and therefore retard incentive to innovate. Matutes, Regibeau and Rockett (1996) considers how patents should be structured for innovations that will spawn second-generation applications. O’Donoghue, Scotchmer and Thisse (1998) consider an infinite series of innovations and show how some degree of protection from future improvements/innovations is socially optimal. O’Donoghue (1998) shows why this protection level might need to be broader rather than narrower. Conversely, Bessen and Maskin (1999) argue why patent protection in a dynamic setting might not be socially optimal. Denicolo and Zanchettin (2002) consider the optimal mix of the novelty requirement and leading breadth to provide protection from future innovations. In a related work, Hopenhayn and Mitchell (2001) show how the choice between patent breadth and length when innovations are sequential might reveal the true value of a firm’s innovation. As noted above, Green and Scotchmer (1995) considers the effect of sequential innovations on patent policy by modelling a game between a first generation innovator and a second generation innovator. Since first generation innovations are needed to facilitate second generation innovations, Green and Scotchmer consider not just the total profit earned by both the first and second innovators, but also the division of profit between them. Clearly, the first innovator will not invent if he/she cannot recover his/her innovation costs. The timing of Green and Scotchmer’s game is as follows. After the first product is patented, a second firm gets an "idea" for an improved product. This idea is described by a value and cost, each taken from separate distributions known to everyone prior to the first investment. Patent breadth is a value that determines whether the value of the 9 second product infringes the patent. There are two stages in the model at which an agreement between the sequential innovators can be reached. The first is when the second innovator gets his/her idea for the improved product but before he/she has sunk costs (an ex ante license or agreement). A second opportunity arises after the costs of the improved product have been sunk and the resulting product infringes (an ex post licence). Before considering the optimal patent breadth, Green and Scotchmer demonstrate that when ex ante agreements are possible, the second innovator has bargaining power so that the first innovator cannot collect all the profit. This is because the first innovator can be "threatened" with ex post competition if the second product does not infringe If the product infringes but its benefits outweigh its costs, the second innovator may find it profitable to invest even though he/she must negotiate licensing fees ex post (after his/her costs are sunk). He/she will not give away more profit in the ex ante agreement than is available by declining the ex ante agreement in favour of ex post licensing. Finally, if the second innovation would infringe but that ex post licensing would completely erode profits to the second innovator then he/she has a credible threat not to enter. Hence he/she will demand a positive fraction of the incremental profit. Green and Scotchmer then show that when there is uncertainty regarding the value of the idea and its cost, firm 2's profit in the ex ante agreement is minimised when the breadth of the patent is maximised. That is, no matter what the value of the idea or its cost, firm 2 will always make a smaller profit if it makes an ex ante agreement and its product infringes the patent. Interestingly, if there is uncertainty regarding the value of the idea only (i.e. the cost is known) then, under certain circumstances, a very wide patent can actually give firm 2 a credible threat not to enter, because it knows that with high probability its profit will be eroded through ex post licensing.10 However, if the first patent is narrower, there is some possibility that firm 2's product will turn out to be noninfringing, which raises the expected profit of entering without an ex ante license. This lessens firm 2's credible threat not to enter. 10 An added assumption is that the entire commercial value of the basic research resides in the secondgeneration products or that the two innovations are not close substitutes because they serve different markets. 10 Green and Scotchmer conclude their article by discussing the antitrust issues raised in their paper11 and what type of licensing (ex ante or ex post) should be legal. They first show that if ex ante licensing is legal, the first innovator earns greater profit if ex post licensing is permitted than if it is not. This is because the availability of ex post licensing takes away firm 2's credible threat not to enter which reduces his/her bargaining power in an ex ante agreement. Given that ex post licensing should be permitted if ex ante licensing is permitted, Green and Scotchmer then consider whether ex ante agreements should be permitted. If they were prohibited and the second product infringed then the two products would compete in the market and earn a smaller profit than if they combined. Consumer surplus would be larger but the policy maker would need to compensate firms by making patents longer. But this lengthening hurts consumers! Not surprisingly, Green and Scotchmer are unable to resolve this conflict with their model. However, they are able to offer one scenario where ex ante licensing improves social welfare. This occurs when the second-generation products are applications of the first technology which has no value as a stand-alone product. Matutes, Regibeau and Rockett (1996) find that patent ‘scope protection’ generates greater levels of social welfare than patent ‘length protection’. The authors concentrate on the protection for “basic innovations” - innovations that spawn a series of applications such as a new computer algorithm. Assuming that the basic innovation has already been produced, the authors argue that a patent providing protection against a subset of possible applications forever (‘scope protection’) is preferred to protecting it with a patent providing a period of exclusivity on all possible applications (‘length protection’). The authors do not consider combinations of length and scope since such combinations are dominated by either pure length protection or pure scope protection. There are two reasons why scope protection is preferred to length protection. First, if the innovator has a given list of applications it wishes to develop then ‘length’ precludes rival innovators from introducing alternate applications until after the incumbent’s term ends. Welfare is higher if scope protection is used since rivals can introduce their applications 11 For example, they have assumed that ex ante agreements are always legal. This might seem reasonable if firms are legitimately worried that ex post competition will significantly erode their profit. 11 as soon as the innovation is patented (and the applications themselves have been developed). Second, scope protection allows the innovator greater flexibility in introducing its list of applications. The authors also consider the optimal scope of protection and find that it is increasing with the number of competing firms. Though their findings resemble those of Gilbert and Shapiro (1990) and Klemperer (1990), Matutes et al note that their analysis is quite different. Gilbert and Shapiro and Klemperer consider only how to structure a given reward. They do not consider the innovator’s incentive to reveal its basic innovation nor do they consider the dynamic effects of patent length and scope. O’Donoghue, Scotchmer and Thisse (1998) introduce the notion of effective patent life which is the expected time until a patented product is replaced in the market. It depends on patent breadth as well as on statutory patent life, since patent breadth determines which products can replace the patented product. The authors propose that there at least two types of patent breadth: lagging breadth and leading breadth. Lagging breadth provides protection against competition from products inferior to the patented product, while leading breadth provides protection from competition from superior quality products”. The authors’ model assumes there is an infinite sequence of innovations (quality improvements) where a Poisson process with parameter lambda determines the rate at which firms collectively receive ideas for improvements. Each idea is received by a single random firm (other than the original innovator) and quality improvement is distributed according to a stationary distribution. The policy tools available to the government are patent length, lagging breadth and leading breadth. The authors assume that lagging breadth protects the entire quality gap between the patent quality and the previous patent quality. Hence flow revenue to the patent holder is equal to the quality improvement and lasts until the next innovation. O’Donoghue et al show that without leading breadth, the rate of innovation is suboptimal since the patent effectively terminates when another firm invents a better product. The reasoning is as follows. It is optimal for all ideas to become innovations if the 12 improvement divided by the discount rate exceeds the costs of innovation. However, since a firm only receives the improvement divided by the sum of the discount rate plus the probability of a better product being produced, the rate of innovation is smaller than the optimal rate. Not surprisingly, leading breadth can stimulate R&D by increasing effective patent life. This is because some improvements will infringe a patent holder’s patent and therefore will not terminate the patent. The authors consider two types of patent with leading breadth: patents with finite leading breadth and infinite patent life, and patents with infinite leading breadth and finite patent life. Under the first type of policy, the effective patent life is determined endogenously by the time it takes for a sufficiently better product to be invented. Under the second type of policy, the effective patent life is the statutory patent life. Since the authors assume that infringing innovations are licensed to the market incumbent, profit surplus is made up of two components: (1) the direct addition to output market profits; and (2) the incremental claims on subsequent innovations that are patented before the patent expires. Not surprisingly, as the patent term becomes large, the rate of innovation also approaches optimality (i.e. increased length or leading breadth can increase effective patent life and stimulate R&D). However, they are not equivalent as they have different R&D costs. In particular, R&D costs would be higher under the policy of finite patent life. O’Donoghue et al also considers a scenario where consumers have heterogeneous tastes for quality and as a result do not all consume the highest quality product, even though that would be efficient. In fact, they assume consumers with lower willingness to pay for quality consume the lower quality product. Competition in the output market allows exactly two firms to earn positive profit at any point in time, a market leader with a higher quality product and a market follower with a lower-quality product. Hence the i-th innovator earns payoffs in two periods: as a market leader immediately following the innovation and as a market follower after the next innovation. innovations, the firm is displaced from the market. Without leading breadth, the fact that the innovator becomes a market follower creates a 'foot-in-the-door' effect. The profit an innovator will earn as a market follower will After subsequent 13 depend on the size of the next innovation, which could be large (meaning it is less likely a higher quality idea will arrive). As a result, firms may invest in small ideas simply as a means to secure a market position. This is inefficient since innovations of quality close to zero will be produced. But if the hit rate of ideas is high, lagging breadth is inadequate in the natural-oligopoly model despite the foot-in-the-door effect. With heterogeneous tastes, leading breadth can actually retard R&D relative to no leading breadth at all since the marginal innovation infringes the prior patent. The 'foot-in-thedoor' effect is absent. The marginal innovation is less profitable when it infringes the prior patent (and must be licensed). O'Donoghue (1998) also investigates patent design when there is a long (infinite) sequence of innovations. He shows how a patentability requirement, a minimum innovation size required to get a patent, can actually stimulate R&D investment in this environment. The basic problem, as discussed in O'Donoghue, Scotchmer and Thisse (1998) above, is that when there is a long sequence of innovations, firms tend to underinvest unless there is protection from future innovators (i.e. leading breadth). While the increase to flow social welfare last forever, any successful firm receives flow profit only until it is superseded. Hence the private reward is less than the social reward, so firms under-invest. However, a patentability requirement can increase the reward for R&D by delaying the next patentable innovation. This contrasts with the perception that if patents are easier to obtain then more R&D will take place. In O'Donoghue's model, innovations also occur according to a Poisson process. Firms select an innovation target and a desired level of R&D spending. The arrival rate is increasing at a decreasing rate with respect to R&D spending and decreasing at an increasing rate with respect to innovation size. O'Donoghue assumes the market leader is able to appropriate profits equal only to the quality gap between their innovation and the previous innovation. Hence firms always under-invest in the model because each consumer never pays more than the incremental value of consuming one unit of the leader's innovation. There is no "leadership-rent effect" in which followers have an extra incentive to capture leadership rent (which can lead to over-investment). 14 A patentability requirement stimulates R&D investment by extending the length of the market incumbency. When future firms target larger innovations, the expected length of market incumbency increases since larger innovations are more difficult to achieve. Hence the expected reward to any successful innovation increases, inducing firms to spend more. However, the patentability requirement introduces an inefficiency in terms of innovations size: without a patentability requirement firms target the first-best innovation size, and a patentability requirement induces firms to target sub-optimally large innovations. Therefore, social welfare is only increasing in the patentability requirement provided that it is not chosen too large. Conversely, Bessen and Maskin (1999) suggests that if innovation is both sequential and complementary, a firm's profit may actually be enhanced by competition, and a patent system may interfere with such competition and with innovation. The authors note that some of the most innovative industries of the last forty years - software, computers, and semi-conductors - have historically had weak patent protection. Imitation of a discovery may be socially desirable in a world of sequential and complementary innovation because it helps the imitator to develop further inventions. Since the imitator may have valuable ideas not available to the original discoverer, the overall pace of innovation may therefore be enhanced. Bessen and Maskin’s analysis differs from the above work primarily in their modelling of technological competition. Rather than producing direct 'knock offs', competitors produce differentiated products. The authors first construct a static model with two firms. If a firm discovers an innovation, the other firm can costlessly imitate the innovation provided there is no patent protection. In this case, both firms earn duopoly profits that are less than the monopoly profits. Complementarity is modelled by assuming that increasing the number of firms in pursuit of an innovation raises the probability that one of them will succeed.12 In this static model, the authors show that, if it is socially desirable for a second high-cost firm to invest, patent protection encourages would-be imitators to innovate themselves. However, in a dynamic setting with a sequence of potential innovations, a second firm is 12 Assuming that the probability of not succeeding is (1-p), then the total probability of successful 15 more likely to engage in R&D without patents. This is because its R&D investment raises the probability, not only of the next innovation, but of subsequent innovations, and this is advantageous to the firm, even if subsequently it merely imitates the first firm. Hopenhayn and Mitchell (2001) is concerned with optimal patent design in the presence of innovation heterogeneity. The problem is modelled as an optimal mechanism design problem where innovators have private information about the nature of their innovations. Their article ties together the two strands of research covered in this discussion paper since it involves the trade-off between patent length and breadth and also involves mechanism design to overcome information asymmetries (as used in the patent renewal literature). In Hopenhayn and Mitchell’s model, the patent authority seeks to maximise a social welfare function that is decreasing in the breadth and length of the patent assigned to an innovator. Individual rationality and incentive compatibility constraints are included in the mechanism as is a constraint ensuring fees are non-negative. Breadth and length are functions of the innovator's "idea" which is unobservable to the patent authority. The authors introduce a set of sorting conditions that are sufficient for the optimal level of fees to be zero.13 Under the sorting conditions, the optimal contract has a specific form. Projects with higher value ideas get more breadth and less length, since that is the instrument they favour most. Fees are not necessary for information revelation. To illustrate their model with a more structural form, Hopenhayn and Mitchell build a model that satisfies the sorting condition and incorporates zero fees. In the model, the inventor arrives with an idea of unobservable value. The value of the idea also determines the expected arrival of the second innovation (i.e. the more valuable the idea, the sooner improvements arrive). The faster are arrivals (higher value idea), the more valuable breadth is, since it is more likely to be useful (to block improvements) and likely to be useful sooner. Hence the model satisfies the sorting condition introduced above: fast arrival means that breadth is more valuable and length is less valuable. innovation is 1-(1-p)^2=2p-p^2. 13 The sorting conditions are that the profit function of the innovator is monotonic in the idea and strictly increasing at a decreasing rate in the idea (i.e. profit function is concave). 16 The authors offer a second model to illustrate their hypothesis. The model is reminiscent of Klemperer's model (discussed above) in that consumers have a substitution cost per unit in the product space away from the patent good. Here the value of the idea determines the cost of imitation by competitors such that a high value indicates low-cost substitutes are available. Hopenhayn and Mitchell show that the optimal patent menu in this model has zero fees provided the patentee's profit function is concave. Interestingly, they also show that if only one patent is offered it is not necessarily infinitely lived (as Klemperer argues). This relates to the unobserved heterogeneity of the innovations.14 Denicolo (2000) re-examines the patent race by analysing the value of ‘forward patent protection’ under various regimes. The model presented in this paper differs from that of the large majority of literature by providing for a patent race at each stage of the game. It differs from the Green and Scotchmer model by allowing for competition in R&D and, by assuming that the first innovator can compete as well as any outsider in the second race, repeated innovation is expressly modelled. Forward protection is at its strongest when the second innovation infringes the patent on the first innovation and cannot be patented. Forward protection is at its weakest when the second innovation can be patented and does not infringe. The author demonstrates that as the second innovation ‘moves’ from unpatentable to patentable, and from infringing to non-infringing, the returns to the first innovation decrease while the returns to the second innovation increase. Denicolo shows that forward protection in a regime where second innovations are unpatentable and infringing always lead to underinvestment in subsequent improvements upon the first innovation. This is a result flowing from the winner-takes-all nature of innovation from the first period, which leads to overinvestment because of inefficient competitive externalities. The author also argues that if the innovations are symmetric, 14 Suppose there are two value types: high and low. Starting with an infinitely-lived patent that is of sufficient breadth to cover the high-value types, it is clear that the low-value types will enjoy monopoly power since their product is more costly to imitate (breadth is less important and patent life is more important). Thus, it is welfare enhancing to reduce the patent life and compensate high-value innovators by increased breadth. High-value innovators are not affected but low-value innovators lose monopoly power. 17 then social welfare is at its greatest if the innovations are patentable and non-infringing, and social welfare is at its lowest if the innovations are unpatentable and do infringe. Denicolo relaxes the symmetry assumption and notes that the appeal of strong forward protection diminishes. If the first innovation becomes relatively more profitable, then there is greater incentive to invest in the first innovation. This may lead to inefficiencies generated by over-investment. Also, strong forward protection becomes less desirable the more obvious the first innovation. Denicolo and Zanchettin (2002) builds upon this model and considers the optimal combination of the novelty requirement and breadth to provide forward patent protection in a sequential two-stage model of innovation. The authors stress that the two types of forward protection are in fact different. If a second-generation innovation must satisfy a novelty requirement then it may be blocked. If a second-generation innovation is found to infringe the original innovation it may still be introduced if the two parties can bargain over profit shares. Denicolo and Zanchettin begin their analysis by (once again) demonstrating that there must be some form of forward protection since there is under-investment in the first stage without it. The question is then how it should be provided. With a basic model, the authors find that the optimal policy involves no novelty requirement. When the level of forward protection necessary is low, leading breadth and the novelty requirement are perfect substitutes. This is because leading breadth, like a novelty requirement, blocks some second-generation patents which, after profit sharing is negotiated, leave the imitator insufficient reward to innovate. However, as forward protection rises, leading breadth should be employed since it promotes profit sharing which is preferable to blocking. The authors also offer scenarios where the novelty requirement might play a role. Supposing that the original innovator can compete for a second innovation then a novelty requirement will make the second innovation unpatentable to competitors. The original innovator will be able to capture the entire (social) value of the second innovation which is optimal. If there is uncertainty surrounding outcomes then it may also be necessary to 18 incorporate a novelty requirement should the original innovator have insufficient incentive to invest even with maximum leading breadth. Hence the authors advocate a novelty requirement only after leading breadth is set at its maximum feasible level. 4 Patent renewals 4.1 Background Section 143(a) of the Patents Act 1990 stipulates that a standard patent will cease to exist if the patentee does not pay the renewal fee for the patent. Renewal fees for standard patents become payable upon the fifth anniversary of registration of the patent. To renew the patent for the year following the fifth anniversary the patentee must pay a fee of $180. From the sixth anniversary it increases to $200 and then in fifty dollar increments until the sixteenth anniversary ($700) when it increases in one-hundred increments until the nineteenth anniversary ($1,000). If an extension of the patent is granted under s 76, such that the life of the patent is extended beyond twenty years, then the renewal fee for the twentieth year and each subsequent year is $1,200.15 This policy has been introduced (in part) to prevent inefficient use of the patent system. In order to ensure optimality, the patent renewal fee at time t should equal the marginal social cost at time t. Such a policy would ensure that the social cost of patent monopolies is reduced to zero. As the marginal social cost increases over time, the renewal fee should also increase over time. The system will thus operate such that the inventor will continue to renew the patent if the marginal private benefit of maintaining the patent exceeds the renewal fee. As the marginal private benefit falls over time and the renewal fee increases, there may come a time where the marginal cost exceeds the marginal benefit, and the rational decision for the inventor is to not renew the patent. However, this assumes that the returns flowing to the inventor from maintaining the patent for an extra year (the marginal private benefit) are equal to the incremental value of inventive activity (marginal social benefit). 15 This fee structure is found in Patent Regulations 1991, Schedule 7: Fees, Item 211. 19 As mentioned above, the Intellectual Property and Competition Review Committee recommended in its 2000 Review that IP Australia consider implementing steeply rising renewal fees for patents to reduce the effective term of patents. 4.2 Recent economic literature With the exception of Hopenhayn and Mitchell (2001), the research into patent design discussed above has been narrowly focussed on such issues as optimal uniform patent length and patent breadth. Only two other articles, Scotchmer (1999) and Cornelli and Schankerman (1999), consider the optimality of differentiated patent protection. Scotchmer considers the issue with a model of asymmetric information on costs and benefits of innovation and finds that, under certain circumstances, patents should have uniform lives. Conversely, Cornelli and Schankerman argue that such uniformity is suboptimal. Their model incorporates moral hazard and asymmetric information. One further article by Crampes and Langinier (1998) considers a different aspect of patents renewals, namely their signalling value. Using a static model with asymmetric information but no moral hazard, Scotchmer (1999) shows that the only feasible incentive mechanisms are equivalent to patent renewal systems. In her model, firms have private information on the cost and the value of their innovations. The patent authority cannot observe these variables ex ante or ex post. Central to her analysis is the notion of a "cutoff function" that determines the maximum R&D cost that the patent mechanism will support for various innovation values. Scotchmer first shows that for the mechanism to be both individually rational and incentive compatible the "cutoff function" must be convex and that its slope represents the patent life.16 Two implications quickly follow. First, higher-value innovations receive longer patents.17 Second, firms must pay back money to the patent office to prolong its patent. This is because the firm's total compensation for innovating is determined by the "cutoff function" which comprises the value of the patent and a transfer from the patent authority. Since the "cutoff function" (and hence the value of the 16 The mechanism is individually rational if firms earn positive profits from any innovation and incentive compatible if these profits are maximized by reporting the 'true' value of the innovation. 17 Actually, the patent length could be uniform since a linear "cutoff function" is also convex. The paper 20 patent) is nondecreasing, the transfer from the patent authority must be nonincreasing. Combined, the above findings allow Scotchmer to demonstrate equivalence with renewal mechanisms. Having shown that in an incentive-compatible mechanism, the "cutoff function" must be convex, Scotchmer then investigates what shape the function has. This is important since the shape (and hence, slope) determine the renewal fee structure. Here, she considers two different research environments: one in which innovation costs and value are independently distributed, and another in which costs are a deterministic function of value. In the former case, the "cutoff function" that maximizes social welfare is found to be linear. Hence it is optimal to have a trivial renewal system of uniform patent lengths. In the latter case, it is optimal to have a nontrivial renewable system with varying patent lengths. As no evidence on the joint distribution of the value and costs of innovations was available at the time of publication, Scotchmer is unable to favour either of these two systems. Using a model with moral hazard and asymmetric information, Cornelli and Schankerman (1998) show that it can be welfare improving to differentiate patent lives when firms have different R&D productivities. In the model, firms have a certain capacity to generate an innovation which is unobservable by the government but which is known to be drawn from some known distribution of production capacities. Cornelli and Schankerman begin their analysis by solving the government's maximization problem: choose patent length based on capacity to maximize social welfare. This section assumes full information. They then construct a mechanism to solve for the optimal differentiated patent schedule and optimal fee schedule (both of which are functions of firm capacity) while relaxing the assumption of full information. The solution to the mechanism yields three key results. First, the conditions for welfare maximization are the same as those for the full information problem. Second, the optimal patent schedule does not depend on the distribution of firm capacities. In contrast, the mechanism with uniform patent lives does depend on this distribution. Third, patent fees are set equal to a level that extracts as much rent as possible while ensuring firms still have an incentive to reveal their true does not seem to emphasize this. 21 capacity. Using these results in conjunction with particular specifications for firm profit, social welfare and deadweight loss, they show that a differentiated patent policy is both optimal and can be implemented using renewals or upfront fees.18 Using simulation analysis with the above specification for profit, Cornelli and Schankerman conclude by demonstrating that optimum mechanism raises welfare by 27% as compared to the optimal uniform patent length. Here they also show that optimal renewal fees should rise much more with patent length than existing (French, German and UK) fee schedules. One further article, Crampes and Langinier (1998), considers the signalling value of patents. The authors present a sequential two-period game with two agents: an informed patent holder who knows the characteristics of the demand function and an uninformed potential entrant who does not. The patent holder must pay a renewal fee each year in order to maintain her patent. The challenger only knows the probability distribution associated with demand but can observe the patentee’s renewal decision. Paying the renewal fee is interpreted as a positive signal to the challenger because the incumbent would never pay the fee if the market was unprofitable. Hence paying the renewal fee increases the probability that the incumbent will face competition. Crampes and Langinier analyse several Perfect Bayesian Equilibria and find equilibria where the incumbent prefers not to pay the renewal fee. This is when the challenger has low beliefs about the profitability of the market and when the incumbent knows that conditions are favourable. Here the incumbent will choose not to pay the renewal fee hoping that it will be interpreted by the challenger as a signal of low market profitability. In mixed strategies though, the patent holder decides to not pay the renewal fee only when market decisions are poor. When conditions are good the decision is randomised. Consequently the challenger will enter when the fee is paid and will decide to enter on the basis of revised probabilities if the fee is left unpaid. 18 The particular specification employed has profit equal to the innovation's value multiplied by the firm's capacity. 22 5 Patenting of business methods and computer programs Over the past two decades patent protection in the US has been extended to new subject matter including business methods and computer programs. Examples of business methods that have been patented in the past few years include Amazon.com’s one-click Internet ordering process and Priceline.com’s reverse auction method for booking airline tickets and other products on the Internet. Prior to 1998, the US Patent Office had not considered business method patents to be within the proper scope of the law.19 Whether it is socially optimal to patent business methods is unclear since there are both positive and negative effects. In economic models where innovation is sequential, extending patent protection to less significant innovations can either raise or lower the level of innovation depending on the characteristics of the industry. Several articles, all discussed above, highlight this point. Bessen and Maskin (1999) contend that the extension of patent protection to software has not been a spur to innovation in the computer industry. They argue that weaker protection furthers innovation by permitting the development of complementary innovations.20 O’Donoghue (1998) argues along similar lines by advocating a ‘patent requirement’ which precludes patenting of low-value innovations. However, O'Donoghue, Scotchmer and Thisse (1998) show that unless first-generation innovators are afforded protection for second-generation innovators, they will have insufficient incentive to innovate in the first place. Interestingly, a recent literature has developed on how excess patenting actually deters innovation (see Heller and Eisenberg, 1998; and Shapiro, 2001). The metaphor used is the “tragedy of the anticommons” whereby a resource is prone to underuse when multiple owners each have a right to exclude others from a scarce resource and no-one has an 19 The Patent and Trademark Office (1994), Manual of Patent Examining Procedure, US Department of Commerce, available at ftp://ftp.uspto.gov/pub/mpep/. 20 Open sourcing of software-providing code that individuals can use freely to develop new code or collaborate on new products-demonstrates the productive effect that complementary information sharing can have on innovation. 23 effective privilege of use.21 The result is a very lengthy and expensive licensing process where firms can hold up other inventions from entering the market. For example, Gallini (2002) notes that in 1999 Amazon.com sued Barnes and Noble to prevent it implementing a similar ‘one-click’ online ordering process. In 1998, Priceline.com used its patent rights on its ‘name your own price’ reverse auction business method against Microsoft’s Expedia travel service. Since patent protection on business methods and software is unlikely to be revoked, several academics have considered what policy amendments can alleviate the problem. Not surprisingly, the focus has often been on reducing barriers to future progress for second-stage researchers without reducing the initial innovator’s incentives. For example, Maurer and Scotchmer (1998) advocate an “independent invention defence” in which subsequent researchers could develop a patented invention, as long as they can prove that it was done independently. However, Gallini notes such remedies “do not directly reduce the error and accompanying administrative and litigation costs of granting low-value patents in the first place”.22 Hunt (2001) suggests that more resources be put into patent offices and more emphasis should be placed on the development of computerised databases to make searches more effective. 6 Empirical literature on patent design This section examines some of the recent empirical literature on patent design. To date the majority of research has used patent data to estimate either innovation intensity or the private value of patent rights.23 However, the articles below have used patent data to analyse more diverse areas such as the effect of patent renewal fees and the effect of increasing patent scope. 21 A ‘patent thicket’ is said to form in this situation. Shapiro (2001) defines this as an overlapping set of patent rights requiring those seeking to commercialize new technology obtain license from multiple patentees. 22 Nancy Gallini, ‘The Economics of Patents: Lessons from Recent U.S. Patent Reform’, Journal of Economic Perspectives, 16(2), p. 147. 23 See, for example, Schankerman (1998). 24 6.1 Patent renewals Though the focus of their studies is not in assessing the effect of varying renewal fees, Lanjouw (1998) and Lanjouw, Pakes and Putnam (1998) use patent renewal data to investigate how renewal fees affect innovation.24 Combining patent application and renewal data with models of application and renewal behaviour, Lanjouw et al. find that decreasing renewal fees would have little impact on patent values. For example, they find that introducing a renewal fee schedule as suggested by Cornelli and Schankerman (1998) with lower fees in early stages which rise sharply thereafter, would only raise the average private value of a computer patent by 5 per cent. Abandoning renewal fees altogether would result in a maximum average increase of 15 per cent. These results are based on data on 1,172 German computer patents issued prior to 1975. The authors note that the benefit to patentees from a decrease in renewal fees results from two effects. First, they pay less for each year of protection. Second, because it is less costly, they also tend to take advantage of more years of protection. Thomas (1999) focuses on the relationship between the technological impact of patents and the renewal decision of the patentee. The technological impact of a patent is estimated by the number of times a patent is cited by subsequent patent applications. The intuition is that significant patents will influence a large number of second-generation patents and will therefore by more heavily cited. The data set used by Thomas contains 189,359 US patents, applied for and issued between December 1980 and December 1985. The results suggest that, if citations are a valid proxy for technological importance, the decision to renew a patent is influenced by its technological impact. 6.2 Effect of increased patent protection Lerner (1994) examines the impact of patent scope on firm value. As a proxy for patent scope, he counts the number of International Patent Classification (IPC) classes to which the patent examiners assign each patent. Lerner validates the relevance of this measure by showing that patent assigned to more classes are more likely to be cited and litigated. 24 Lanjouw focuses on estimating the private value of patent protection while Lanjouw et al. discusses how patent renewal data can be used to improve on innovation indicators. 25 Using the valuations placed on 173 biotechnology firms during the venture capital investment process between 1978 and 1992, increasing scope is found to be associated with higher valuations. Lerner also explores whether, as Klemperer’s model suggests, the marginal value of broader patent scope is higher when consumers find it easier to switch within the same product class (smaller value of t above). Lerner's proxy for patent uniqueness is the ratio of the number of patents held by a firm in an International Patent Classification (IPC) subclass to the total number of patents held in that subclass. His hypothesis is that firms whose research is unique face less substitution. Thus, their value should be less sensitive to patent scope (as measured by the number of IPC classes to which patent examiners assign each patent).25 Lerner's results support this hypothesis. Sakakibara and Branstetter (2001) analyses the effects of 1987 Japanese patent reforms which expanded patent scope. The central issue is whether these reforms induced more innovative effort. The authors’ data set comprises 307 publicly traded Japanese manufacturing firms, drawn from various industries. It is found that the firms have been unresponsive to reforms expanding patent scope. Though they have adjusted the number of claims per patent, there is no evidence of an increase in innovative effort. This suggests that strengthening patent protection is socially inefficient and merely leads to the formation of patent thickets (as discussed in the previous section). Such a suggestion is echoed in Hall and Ziedonis (2001) which examines the patenting behaviour of 95 US semiconductor firms during 1979-1995. The period is of interest since it spans the “pro-patent” shift in the US legal environment, strengthening patent protection. In determining the factors underlying the surge in patenting during this period, the authors conducted interviews and sampled approximately 100 publicly traded US firms. They find that the surge in patenting was driven by firms engaging in “patent portfolio races” hoping to mitigate the need to negotiate with rival firms holding critical patents. The authors note however that this is not an inevitable outcome of increasing patent protection. For if patent rights are awarded to inventors of ‘novel’ inventions then 26 patent thickets will not develop so easily. Lanjouw and Schankerman (2001) analyses what characteristics of patents increase the likelihood of litigation. The authors’ data includes 5,452 patent cases between 1975 and 1991 involving 3,887 US patents. The characteristics of the patents included in the data include two measures of patent scope: the number of claims26 and the number of IPC assignments. Interestingly, while Lanjouw and Schankerman find that the probability of litigation rises strongly with the number of claims, they find no evidence of a positive relationship between litigation and the number of IPC assignments. This result contradict Lerner’s findings above. The authors also find that the probability of litigation is strongly correlated with the number of forward citations (i.e. the number of times a given patent is cited by subsequent patent applications). This suggests that patentees looking to appropriate rents from improvements innovation need to exercise control over the initial innovation. Lanjouw and Schankerman note that this finding is consistent with Green and Scotchmer (1995) which analyses bargaining between first and second-generation innovators. 7 Conclusion The economic literature on patent design presents a wide variety of robust results regarding optimal patent design and policy. On the question of trading off patent breadth for patent length, the literature provides circumstances where both infinite length and infinite breadth are optimal. Given a static framework, where future innovations do not build on prior patents, these results suggest that Australia’s current regime is too rigid. A patent term of twenty years might be optimal in some cases, but it will not be optimal in most. Hence the literature suggests patent offices should have greater flexibility in 25 'Value' is derived from valuations placed on firms during the venture capital investment process. 26 A patent includes a set of claims that delineate the boundaries of the property rights provided by the patent. 27 structuring patent rewards. However, Australia’s obligation under the TRIPS agreement make such reform difficult. The less-developed literature on Patent Renewals suggests that greater flexibility in patent policy can be achieved with rising renewal fees. As noted above, Australia’s fee structure is already increasing. However, given the infancy of the literature on this topic, there is scope for further research investigating the optimal schedule of renewal fees for Australia and other countries. On the question of optimal patent policy in a dynamic environment, the literature is less conclusive. Depending on the characteristics of the particular industry, weakening the requirements for patent protection can be shown to either increase or decrease innovation. Though a weaker policy of patenting would seem to increase innovation, the recent literature on patent thickets suggests exactly the opposite. If anything, this divergence in findings should stress the importance of any changes to Australia’s policy, particularly in following the US in granting patents to new subject matter such as business methods. The vast majority of economic literature discussed in this survey has considered perfect protection against imitation. However, it should be noted that patent protection is ultimately meaningless unless it can be enforced. Crampes and Langinier (2002) suggest that patents merely grant a right to sue imitators. They note that if the patentee is unable to observe or identify an infringing imitator then the patent offers little protection. So the value of a patent lies in the competency of the legal advisors of both the patentee and the imitator.27 It is also worth mentioning that no matter how well they are designed, patents are not always the ideal tool for stimulating innovation. Boldrin and Levine (2002) show that in a competitive environment with no downstream licensing, IP rental possibilities may be sufficient to compensate producers for the sunk costs of production. Boldrin and Levine 27 A modest literature on patent litigation exists, see also Choi (1998), Meurer (1989) and Waterson (1990). Some interesting results have come out of this literature. For example, Choi (1998) considers the information externalities arising from litigation. He finds that in certain circumstances the patent holder may choose not to litigate against an entrant if the negative information effect (e.g. of alerting 28 also show that as new technology (e.g. Napster) increases the number of copies that can be made, the reward to innovators might increase. entrants to a potentially invalid patent) dominates the positive effect of entry deterrence. 29 References Bessen, James and Maskin, Eric (1999), 'Sequential Innovation, Patents and Imitation', Working Paper, available at www.researchoninnovation.org/patent.pdf. Boldrin, Michele and Levine, David K., ‘The Case Against Intellectual Property’, American Economic Review, 92(2): 209-212. Chang, Howard (1995), 'Patent Scope, Antitrust Policy, and Cumulative Innovation', Rand Journal of Economics, 26: 34-57. Choi, Jay Pil, ‘Patent Litigation as an Information-Transmission Mechanism’, American Economic Review, 88(5): 1249-1263. Cooter, Robert and Ulen, Thomas (1988), ‘Law and Economics’, HarperCollinsPublishers. Cornelli, Frances and Mark Schankerman (1999), 'Patent Renewals and R&D Incentives', Rand Journal of Economics, 30: 197-213. Crampes, Claude and Corinne Langinier (1998), 'Information Disclosure in the Renewal of Patents', Annales d'Economie et de Statistique 49-50, 265-288. Crampes, Claude and Corinne Langinier (2002), ‘Litigation and Settlement in Patent Infringement Cases’, Rand Journal of Economics, 33(2): 258-274. Denicolo, Vincenzo (1996), 'Patent Races and Optimal Patent Breadth and length', Journal of Industrial Economics, 44: 249-265. Denicolo, Vincenzo (2000), ‘Two-Stage Patent Races and Patent Policy’, RAND Journal of Economics, 31(2): 488-501. Denicolo, Vincenzo and Zanchettin, Piercarlo (2002), ‘How Should Forward Patent Protection Be Provided?’, International Journal of Industrial Organization, 20: 801-827. Gallini, Nancy (1992), ‘Patent Policy and Costly Imitation’, Rand Journal of Economics, 23: 52-63. Gallini, Nancy (2002), ‘The Economics of Patents: Lessons from Recent U.S. Patent Reform’, Journal of Economic Perspectives, 16:131-154. Gilbert, Richard and Shapiro, Carl (1990), 'Optimal patent length and breadth', Rand Journal of Economics, 21: 106-112. Green, Jerry and Scotchmer, Suzanne (1995), 'On the Division of Profit in Sequential Innovation', Rand Journal of Economics, 26: 20-33. 30 Hall, Bronwyn and Ziedonis, Rosemarie Ham (2001), ‘The Patent Paradox Revisited: An Empirical Study of Patenting in the U.S. Semiconductor Industry, 1979-1995’, Rand Journal of Economics, 32(1): 101-128. Heller, M.A. and R. S. Eisenberg, (1998), “Can Patents Deter Innovation? The Anticommons in Biomedical Research,” Science, 280: 698-701. Hopenhayn, Hugo and Mitchell, Matthew (2001), 'Innovation Variety and Patent Breadth', Rand Journal of Economics, 32: 152-166. Hunt, Robert (2001), ‘You Can Patent That? Are Patents on Computer Programs and Business Methods Good for the Economy’, Federal Reserve Bank of Philadelphia Business Review. First Quarter, pp.5-15. Klemperer, Paul (1990), 'How Broad Should the Scope of Patent Protection Be?'. Rand Journal of Economics, 21: 113-130. Kortum, Samuel and Lerner, Joshua (1998), ‘Stronger Protection or Technological Revolution: What Is Behind the Recent Surge in Patenting?’, Carnegie-Rochester Conference Series on Public Policy, 48: 247-304. Lanjouw, Jean (1998), 'Patent Protection in the Shadow of Infringement: Simulation Estimations of Patent Value', Review of Economic Studies, 65(4): 671-710. Lanjouw, Jean, Ariel Pakes and Jonathan Putnam (1998), 'How to Count Patents and Value Intellectual Property: The Uses of Patent Renewal and Application Data', Journal of Industrial Economics, 46(4): 405-432. Lanjouw, Jean and Schankerman, Mark (2001), ‘Characteristics of Patent Litigation: A Window on Competition’, Rand Journal of Economics, 32(1): 129-151. Lerner, Joshua (1994), 'The Importance of Patent Scope: An Empirical Analysis', Rand Journal of Economics, 25: 319-333. Lerner, Joshua (2002), ‘Where does State Street Lead? A First Look at Finance Patents, 1971-2000’, Journal of Finance, 57: 901-30. Matutes, Carmen, Regibeau, Pierre and Rockett, Katharine (1996), ‘Optimal Patent Design and the Diffusion of Innovations’, Rand Journal of Economics, 27(1): 60-83. Maurer, Stephan and Scotchmer, Suzanne (1998), ‘The Independent Invention Defense in Intellectual Property’, John M. Olin Working Paper 98-11, Boalt School of Law, University of California at Berkeley. Meurer, Michael J., ‘The Settlement of Patent Litigation’, Rand Journal of Economics, 31 20(1): 77-91. Nordhaus, W. (1969), ‘Invention, Growth and Welfare: A Theoretical Treatment of Technological Change’, Cambridge Mass, MIT Press. O'Donoghue, Ted (1998), 'A Patentability Requirement for Sequential Innovation', Rand Journal of Economics, 29: 654-79. O'Donoghue, Ted, Scotchmer, Suzanne and Thisse, Jacques-Francois (1998), 'Patent Breadth, Patent Life, and the Pace of Technological Progress', Journal of Economics and Management Strategy, 7: 1-32. Pakes, Ariel (1986), 'Patents as Options: Some Estimates of the Value of Holding European Patent Stocks', Econometrica, 54: 755-784. Sakakibara, Mariko and Branstetter, Lee (2001), ‘Do Stronger Patents Induce More Innovation? Evidence from the 1988 Japanese Patent Law Reforms’, Rand Journal of Economics, 32(1): 77-100. Schankerman, Mark (1998), 'How Valuable is Patent Protection? Estimates by Technology Field', Rand Journal of Economics, 29: 77-107. Schankerman, Mark and Ariel Pakes (1986), 'Estimates of the Value of Patent Rights in European Countries During the Post-1950 Period', Economic Journal, 96: 1052-1076. Scotchmer, Suzanne (1996), Protecting Early Innovators: Should Second Generation Products by Patentable?', Rand Journal of Economics, 27: 322-331. Scotchmer, Suzanne (1999), 'On the Optimality of the Patent Renewal System', Rand Journal of Economics, 30: 181-196. Shapiro, Carl (2001), “Navigating the Patent Thicket: Cross Licenses, Patent Pools, and Standard-Setting”, http://faculty.haas.berkeley.edu/shapiro/thicket.pdf. Thomas, Patrick (1999), 'The Effect of Technological Impact upon Patent Renewal Decisions', Technology Analysis and Strategic Management, 11(2): 181-197. Waterson, Michael (1990), ‘The Economics of Product Patents’, American Economic Review, 80(4): 860-869. Wright, Donald (1999), 'Optimal Patent Breadth and Length with Costly Imitation', International Journal of Industrial Organisation, 17: 419-436. 32