1. IntroductionSince Nordhaus’ early contribution (Nordhaus, 1969, 1972), economic theorists have undertakenextensive analysis of the problem of optimal patent design. Theorists have long recognized the tradeoffsinherent in the patent system – a system which creates static losses (by conferring upon innovatorstemporary monopoly power) in order to realize dynamic social gains by inducing greater innovative effort.Within policy circles, the debate surrounding intellectual property rights design and enforcementhas also increasingly come to the forefront. As has been widely noted, legal and procedural reforms in theU.S. in the early 1980s provided stronger protection to holders of existing patents, reflecting a shift in theU.S. policy consensus in favor of stronger intellectual property rights protection. This policy shift thentook on an international dimension as several consecutive U.S. administrations placed intellectual propertyrights at the center of the U.S. agenda in bilateral and multilateral trade negotiations.1 Since the early1980s, the U.S. has repeatedly sought to cajole or coerce its trading partners into significantlystrengthening their own intellectual property rights regimes, bringing them more in line with the new U.S.model.However, neither the expanding theoretical literature on patent design nor the international policydebate has been sufficiently guided by empirical work.2 Much recent theoretical work explicitly assumesthat R&D performing firms are quite responsive to subtle changes in patent design. The U.S. position inthe international fight over intellectual property rights is based on the proposition that the additionalinnovation induced by stronger patent systems is so substantial that the net social impact of a strengtheningof intellectual property rights abroad will be positive. These propositions are empirically testable, yet theempirical evidence underlining these propositions ranges from sketchy to nonexistent.Does an expansion of the scope of patent rights really induce more innovative effort? Our paperprovides badly needed empirical evidence on this question by analyzing the response of Japanese firms tothe Japanese patent reforms of 1987, enacted in January 1988 (hereafter the 1988 reforms or “patentreform”). For reasons which we document in the paper, this Japanese reform provides an unusually clean1

“natural experiment.”3 Furthermore, this policy reform was undertaken in a country with a large number ofR&D-intensive firms that had reached the global “technology frontier” by the time the reforms wereenacted.4 Thus, we have micro-level data on the R&D investment and patenting of a large number of firmsboth prior to and after the enactment of reforms. Finally, Japanese firms are not only aggressive aboutseeking patent protection at home, but they are also aggressive in patenting in the U.S., receivingapproximately 25% of all patents granted by the U.S. Patent and Trademark Office (U.S. PTO). Therewas no similar change in U.S. patent law at the time of the 1988 reforms. Because we possess two parallelpatent series for our sample firms, we may be able to identify more precisely just what the impact of thepatent reforms was on innovative output.We show that the 1988 patent reforms in Japan expanded the scope of patent protection, as definedby the theoretical literature on patent design, based on interviews with practitioners. Interview results alsosuggest that organizational setting has an effect on firms’ response to changes in patent systems.Empirically analyzing data on 307 Japanese firms during the period 1980 to 1994, we find little evidencethat this expansion of patent scope induced additional R&D effort by Japanese firms. We also examinewhether or not the Japanese patent reform resulted in additional innovative output by examining firmpatenting in the U.S. and Japan. We find some evidence consistent with the hypothesis that innovativeoutput increased. However, only a portion of this increase can be reasonably attributed to the effects ofpatent reform. Taken together, the empirical evidence suggests that firm responsiveness to even significantchanges in patent design is limited. This has important implications for both the current policy debate andthe theoretical literature, which we discuss in our conclusion.This article is organized as follows. Section 2 explains the pre-reform Japanese patent system.Section 3 describes the 1988 Japanese patent system reforms. This section also explains why these reformscan be considered an “exogenous” policy change. Section 4 discusses the perceptions of the effects of thereforms by practitioners and demonstrates that patent reform effectively expanded the scope of patentprotection. Section 5 establishes the linkage between the 1988 reforms and theoretical literature, and2

introduces a model which summarizes the expected effect of the 1988 reforms on innovation. Section 6examines the impact of patent reform on Japanese firms at the aggregated level. Section 7 empiricallyanalyzes the impact of patent reform on R&D spending. Section 8 examines the effect of the patentreforms on Japanese firms’ patenting in Japan and the U.S., and section 9 concludes.2. The Japanese patent system before the 1988 reformsJapanese and American patent examiners followed similar standards of what constituted apatentable invention. A major difference between the two patent systems concerned the scope of a patent.5A claim defines the subject matter which an applicant regards as his or her invention. Japanese patentscovered a single, independent claim - meaning that one novel advance was permitted per patent. Incontrast, American and European patents often listed multiple, independently valid claims. The Japanesesingle-claim system (or the “head claim system”) meant that, compared with other nations, many moreJapanese patents had to be filed to cover the same technology. This difference led to a much greaternumber of patent applications in Japan than in other countries.It was also argued that the scope of each claim tended to be narrower in Japan than in the U.S. Apatent in the U.S. could claim protection for broad classes of a product, whereas in Japan only specificcases of the product that had been proven in practice could be patented. For example, if a new materialcould conceivably contain 10% to 50% of a certain ingredient, the U.S. patent could specify this range; inJapan the patent covered only the specific percentage that the inventor had used.6 As each patent was sothinly defined under the Japanese patent system,7 some critics even called it the sashimi system, after theJapanese sliced fish delicacy.83. The 1988 reformsFrom the single-claim system to the “improved” multiple-claim system. Before 1976, Japanesepatent law allowed only one independent, single claim to be included in an invention. A 1976 amendment3

to the patent law allowed the inclusion of multiple dependent claims, which defined the technical ways toimplement an independent claim, in the same patent application. This amendment, however, did notsubstantially change the number of claims included in a patent.In contrast, the 1988 reforms significantly expanded the extent to which multiple claims could beincluded in one patent. Patent applicants can now define the coverage of an invention with multiple claims,and those claims can be either independent of or dependent on other claims. In addition, the extent to whichrelated inventions could be included in a single patent was dramatically expanded by the new law. Forexample, in patenting chemical compounds, both the final chemical compound and the intermediateproducts generated in its manufacture can be covered by one patent (Niihara ed., 1987). Patent expertscontend that, after the 1988 reforms, the scope of invention covered by a single patent applicationequaled or even exceeded that conferred by the U.S. and European patent systems (Okamoto et. al.,1996). The consequences of expansion of patent scope under these reforms will be our primary empiricalfocus in this paper.Patent term restoration for pharmaceuticals. The 1988 reforms also provided patent termrestoration of up to five years for the period necessary for drug safety and efficacy examination if thepatented invention could not be implemented for more than two years due to delays in the examinationprocess. This amounted to an effective extension of the length of patent rights in the pharmaceuticalindustry. Our data set includes all of the important participants in the Japanese pharmaceutical industry,allowing us, in principle, to explore the innovation-inducing effects of increased patent length as well asincreased patent scope.9Motives of the series of reforms. When a patent system reform comes about primarily throughthe lobbying efforts of the system’s primary beneficiaries, it is not clear that such a reform can be taken asexogenous. In the case of the 1988 reforms, much of the political impetus for change came, not from heavydomestic users of the patent system, such as Sony or NEC, but from Japan’s trading partners – chiefly theU.S. The purpose of those reforms was to harmonize the Japanese patent system with that of other4

countries, and so was driven by motives largely exogenous to the wishes of Japanese firms.Many of the reasons for American dissatisfaction with the pre-reform Japanese patent system arewell illustrated by the case of Fusion Systems, which was made famous by a Harvard Business School casestudy. Fusion Systems was a small U.S. high-tech company which invented a microwave lamp in 1974.Fusion Systems sought patent protection in Japan for its invention, filing 20 patents in Japan by 1983.While 20 U.S.-style multi-claim patents might have been enough to secure Fusion Systems’ intellectualproperty rights, the 20 single-claim patents the company filed in Japan left substantial “holes” in theintellectual product space it had sought to protect. Before Fusion Systems could correct this error, its chiefcompetitor in the Japanese market, Mitsubishi Electric, exploited its mistake by filing 139 patents in closelyrelated technologies – an example of a practice known as “patent flooding.” After finding that MitsubishiElectric’s 139 patent applications had effectively prevented it from plugging the “holes” in its intellectualproperty protection, Fusion Systems appealed to USTR.10 Other U.S. firms with similar complaints putadditional political pressure on the U.S. government to pressure the Japanese government for a strongerintellectual property rights regime. The major outcome was the introduction of the multi-claim system.At the same time, the U.S. and other countries requested that Japan shorten the patent examinationperiod. Due to a severe backlog, the average examination period (from the request of the examination11 tothe end of the examination) reached 36 months in 1985. One cause of the delay was the very large numberof patent applications filed by Japanese firms. The broadening of patent scope was expected to reduce thenumber of patent applications. Parallel to the 1988 reforms, the Japanese Patent Office asked the top 100patenting companies, which account for 50% of the total patenting, to reduce the number of unnecessaryand trivial patent applications in 1988.124. Expected results of the reform and the perception by practitionersThe significance of the multi-claim system. In order to evaluate the expected and perceivedoutcomes of the 1988 reforms, we conducted a series of interviews in Japan in March - September, 1998.5

Six high-ranking officials at the Japanese Patent Office, general managers of intellectual propertydepartments at 10 major Japanese companies in electronics, office equipment, pharmaceuticals, chemicals,and food industries, and three officers at the Japan Patent Information Organization (JAPIO), a quasigovernment organization which provides a Japanese patent electronic database, were -------Figure 1 about igure 1 illustrates the effect the new multi-claim system, as perceived by company managers andJapanese Patent Office officers. The Xs in the center are inventions by a firm and each circle represents aclaim. Let us assume this company invented a chip which is used for image processing. The Xs are thechip itself, and the processes used to make the chip. One circle represents a camera which contains thischip. Another circle represents a personal computer which includes this chip. The third circle might be asystem which combines this chip, a camera, and a personal computer. The multi-claim system allows onepatent to contain all of these claims. Thus, a firm tries to maximize the territory covered by these circleswithout tapping into claims associated with inventions by other firms. Under the single-claim system, acompany can draw only one circle, resulting in much narrower scope of intellectual property protection.Does the multi-claim system really matter? One might argue that before the 1988 reforms,innovators could duplicate the protection afforded by patents of greater scope by simply taking out multiplepatents.13 This argument would be incorrect if (1) the multi-claim system substantially reduced the cost ofa given level of patent protection, or (2) if the multi-claim system substantially improved the breadth ofpatent protection (meaning innovations that could not be completely protected under the old system couldnow receive full protection). Based on our interviews and on more objective data, we will argue that patentreform did indeed produce effects (1) and (2). The extent and relative strength of these two effects, ofcourse, is difficult to calculate with precision, but we believe the effect of (2) is more substantial forreasons discussed below.Before addressing these issues, however, we feel that it is important at the outset to dispel the6

widely held misconception that patents are neither important nor effective as instruments of appropriationof innovation in Japan. Strong evidence to the contrary is presented in the recent paper by Cohen et. al.(1998). The authors conduct a comprehensive survey of the R&D managers of U.S. and Japanese R&Dintensive firms. Firms’ responses to these surveys suggest that Japanese firms regard patents as among themost effective instruments of appropriation of innovation.14 In striking contrast, U.S. firms’ responsessuggest that American firms find patents to be a much less effective means of protection than, for instance,trade secrecy. Likewise, when asked their most important reasons for patenting innovations, Japanesefirms listed prevention of imitation, blocking similar innovation by rivals, and the prevention of lawsuits asthe most important reasons for filing patents for both product innovations and process innovations.15Financial implications of patent reform. The most obvious implications of patent reform, and theeasiest to demonstrate quantitatively, are the reductions in private patenting costs to the firms. To put it simply,taking out multiple patents to protect a single idea under the “sashimi” system could prove quite costly.Examination fees and agent fees, which are several times higher than the actual examination fees, had to beincurred for each patent if multiple patents were filed to protect a single innovation. Multiple patenting perinvention would also increase the cost of enforcing patent rights if patent infringement took place, since a lawsuitwould have to be filed for each patent infringed. Many companies suggested in interviews that the multiple-claim system brought an opportunity to save patent related fees through the combination of multiple claims(and therefore multiple inventions) in one patent application, because a substantial portion of patent feesare incurred per patent, not per claim.16As an illustration of the potential cost savings, we use the patent fee schedule as of 1996. Officialfees including the filing fee and the examination fees are 105,300 yen per patent and 2,700 yen per claim.Okamoto (1996) estimates that the agent fees incurred in the processing of these patent documents (asexplained later, some interviewed firms rely on agents to process as much as 70% of their patentapplications) were 604,000 yen per patent. Finally, maintenance fees for the life of the patent, assumingthe patent is registered at the fourth year from the filing, include 1,301,000 yen per patent and 132,300 yen7

per claim. Under this fee schedule, the total costs of protection for a patent is 2,010,300 yen per patent(denoted F, for the “fixed cost” of a patent) and 135,000 yen per claim (denoted V, for the “variable cost”of a claim).In 1993, for example, the average number of claims per patent (denoted N) is 3.77. Under the1996 fee schedule, the total cost for the life of a typical patent is NV F 2,519,250 yen. Assuming thatunder the old system each claim has to be filed as an independent patent17, the total cost for the life of apatent to cover the same invention was N(V F) 8,087,781 yen. The cost saving from the multi-claimsystem is C N(V F) - (NV F) 5,568,531 yen, or a 68.9% cost saving. The number of the totalpatent applications in Japan in 1993 was 366,486, and so the total cost saving would be 2,041 billion yenor 17 billion (using 1 120 yen). This illustration suggests that the total cost savings could be quitesubstantial.18Admittedly, this sum is an upper bound. It overstates our case, as not all patents which areexamined are granted, nor are maintenance fees paid on all patents. Nevertheless, given the fee structure,the potential costs savings are so large that they remain substantial even after more realistic adjustmentsare made. Assuming only 10% of filed patents are examined and maintained, for example, the total costsaving will still be 224 billion yen or 1.8 billion.19While the private benefits of cost reduction for a given amount of patent protection are clear, thesocial benefits are less so. A substantial reduction in the cost of patenting could lead firms to seek toprotect lower quality ideas. Certainly, we might expect the “quality” of the marginal protected idea todecline with a large enough decline in cost. In fact, if the only real effect of patent reform was to reducethe private cost of patent protection, we would expect to observe firms “repackaging” the same number ofideas into fewer patents, thereby realizing cost savings, but we would not necessarily expect such a reformto induce additional innovative output.Real expansion of intellectual property rights under patent reform.8An alternative

interpretation of these reforms is that, not only did they provide cost savings, but the advent of the multiclaim system really did generate an increase in the breadth of intellectual property protection. Whether thisreally happened or not is quite difficult to verify without detailed knowledge of changes in patent practicethat occurred after reform. When we went to those individuals most knowledgeable about these changes –Japanese Patent Office officials and the heads of intellectual property departments at major corporations –they claimed that this increase in effective scope did occur.In practice, the total protection afforded by multiple narrow patents fell short of the protectionconferred by the new broader patents. All possible modifications to the original innovation were often notregarded by patent examiners as bona fide independent innovations. Interviewees suggested that, inpractice, the “burden of proof” in establishing novelty was less stringent for a claim than for a patent.Some firm managers also said that the multi-claim system allowed firms to describe an invention inmultiple dimensions, which increased the opportunities to demonstrate the novelty of the invention and,thus, actually increase the chance that a patent is granted.20The multi-claim system allowed the patent applicant to define the invention in ways which madeinfringement easier to prove and prosecute under Japanese legal practice. This is an extremely importantbut rather complicated point, which is best illustrated by example. Fortunately, Hiraoka et. al. (1988) andJapan Patent Association (1988) provide many such examples. In the following case, a firm found that ifchemical compounds A and B (both publicly known materials) were combined at a specific ratio withpolyethylene, they significantly prevent the deterioration of polyethylene products. In filing a patentapplication for this invention, there were several alternative ways to describe it.(1) An additive which consists of materials A and B, that prevents the deterioration of polyethylene(which emphasizes the product)(2)A strengthened shaped polyethylene product which consists of materials A and B andpolyethylene (which emphasizes the product)(3)A method to produce strengthened shaped polyethylene products which is characterized by9

adding materials A and B to polyethylene (which emphasizes the production process)After the 1988 reforms, (1), (2), and (3) could be included as separate claims in one patent. Underthe old system, however, the inventor had to choose one claim from the above three to include in the patentapplication, since all three claims point to the same invention. If a patent including one of above claimswas granted, subsequent patent applications by the same inventor which contained any of the remainingclaims would be rejected on the grounds of not having met the “one invention, one patent” principle or the“unity” requirement. Subsequent patent applications by other inventors which contained any of theseremaining claims would be rejected on the grounds of not having met the novelty requirement, as theseoverlapping claims point to essentially the same invention. This does not mean, however, that a patent withone claim protects the intellectual property space covered by the other closely related claims, becauseclaims are narrowly interpreted under historical Japanese legal practice. If the inventor is granted a patentwhich includes (1), then the inventor would not be able to sue someone selling a shaped polyethyleneproduct for infringement unless he/she could prove that the produce is the result of adding the protectedadditive to polyethylene. Similarly, if the inventor chooses to include (2) in a patent filing, then he/shecould only sue a rival firm selling the protected additive for infringement if it can be proven that theadditive will be used only for the production of the strengthened shaped product. Finally, in case theinventor chose to include (3), he/she would not be able to sue someone selling the additive for infringementunless he/she could prove that the additive would only be used for the patented method. In practice, theseindirect infringement cases were very difficult to prove in court. The old system illustrated here mightsound very strange to Western readers, but the philosophy behind it is that a claim should briefly andclearly define an invention, and so a claim (and a patent) can include only the essential features of aninvention.To sum up, the narrow interpretation of claims for purposes of determining infringement created a“commons” of intellectual product space around existing patent grants which no patent holder couldeffectively appropriate and within which any firm was free to operate. This situation changed with patent10

reform. Firms were now allowed to use claims “defensively.” In addition to describing independentinnovations, the claims could be used to “fill in” the traditional “commons,” substantially strengthening thefirms’ scope of intellectual property rights.21 It is this sort of effect that the theoretical literature highlightsin models of patent scope. To the extent that this effect was present in the wake of Japanese patent reform,examining the Japanese case will allow us to get some sense of how firms respond to such an expansion ofpatent scope.Note, however, that this “defensive” use of overlapping claims is a very different use of themulticlaim system from that outlined in the preceding subsection. Here firms are not using the multiclaimsystem to repackage several independent ideas into one patent document so much as they are usingoverlapping claims to fill in the “no man’s land” around a single invention. Interview evidence andquantitative firm-level evidence suggests that firms used the multi-claim system in both ways in differentpatents, so that both effects had an actual impact on Japanese firms’ patenting practices.Managerial implications of patent reform. Interview results of general managers of intellectualproperty departments showed interesting differences in the effects of the multi-claim system across firms.They indicated that the degree of change in firms’ patenting behavior and the utilization of the new multiclaim system was strongly related to the degree to which patent-related tasks are conducted within the firm,and to the number of people in a company involved in patent processing. In one electronics company, forexample, over the course of a single year 30 people in the intellectual property department engaged inproducing 20,000 domestic patent applications as well as thousands of overseas patents. However, thegeneration of more than 70% of the actual patent documents, both for Japan and overseas filings, wascontracted out to patent agents. For another company, the degree of outsourcing is more than 90%. Insituations such as these, where there is such a profound organizational disconnect between those in the firm(or outside the firm) dealing directly with the patent system, and those in the firm actually conductingR&D, it is perhaps not surprising that the “feedback” from changes in the patent system to the conduct ofR&D was muted and slowed. On the other hand, at one consumer electronics company, 70% of the patent11

application documents are drafted by the corporate R&D personnel actually producing the innovations. Insuch a context, the “feedback” from patenting changes to the conduct of R&D is, not surprisingly, muchstronger.Some companies also admitted that the internal incentive structures worked, in some cases, toprevent R&D departments from responding to patent law changes. At one large diversified company,business units shared costs related to patent administration through the intellectual property division of thecompany’s headquarters, with the cost allocation determined by the number of patent applications inprevious years. Business units were not allowed to retain any “patent administration cost” savings if theyfiled fewer patents than in previous years, which deterred these business units from responding to patentlaw changes which were designed to reduce the number of patent filings. On the other hand, in somecompanies, the internal reward structure was changed, with greater emphasis placed on maximizing thenumber of claims rather than simply the number of patents.Interviewees suggested that company-specific factors were likely to influence the degree of firmresponsiveness to patent reforms, in addition to the aforementioned organizational factors. The degree oflearning and the utilization of the new patent system in R&D efforts also depended on the efforts ofintellectual property departments. In many companies, commentaries and examples of the multiple-claimsystem were documented and circulated among R&D departments, which stimulated the thinking ofengineers concerning how to utilize an invention in multiple ways. In some companies, a series of seminarswere held to communicate the change in the patent system to researchers.In the cases where the “feedback” from patent changes to R&D conduct was strong, somecorporate R&D managers reported that it had an impact on the manner in which researchers approachedinnovation. Some R&D managers in the office equipment industry noted, for example, that R&D personnelbegan to look for new/multiple applications of an innovation due to the multi-claim system. However, thisresponse was far from unanimous. Many respondents claimed that neither the multi-claim system nor theeffective extension of patent rights in the pharmaceutical industry had a significant impact on the nature or12

level of R&D activity.22Theorists in this literature typically assume that the “innovating unit,” be it an individual or a firm,directly confronts and internalizes the incentives in the patent system. Our interviews drove home thereality that corporate R&D departments are embedded in larger organizations, and the feedback fromchanges in the patent system to their activity is anything but automatic. Organizational settings affect howinnovations are turned to patents, and organizational settings vary across firm. We need to keep this realityin mind not only in interpreting our empirical results, but also in determining the lessons our study has forthe next generation of theoretical models.5. Theoretical linkage between the 1988 reforms and R&DThe multi-claim system and the theoretical literature on patent scope. In this section, weestablish a link between these institutional changes in Japan and the theoretical literature on patent scope.This requires that we briefly review this literature. However, due to space constraints, we will necessarilybe selective and incomplete in our review.In Nordhaus’ original work, and in the famous exchange with Scherer that followed, emphasis wasplaced on the optimal length of a patent as a policy variable.23 In this setup with the assumption that patentprotection allows patent recipients to fully command monopoly rents, the optimal length of a patent wasfound to be finite. Gilbert and Shapiro (1990) and Klemperer (1990) used more sophisticated modelingframeworks to consider both the optimal length and breadth of a patent. They found conditions underwhich a narrow but infinite

Japanese and American patent examiners followed similar standards of what constituted a . example, in patenting chemical compounds, both the final chemical compound and the intermediate . government organization which provides a Japanese patent el