The decision to patent, cumulative innovation, and optimal policy

https://doi.org/10.1016/j.ijindorg.2005.03.001Get rights and content

Abstract

This paper analyzes optimal policy in the context of cumulative innovation in a model that endogenizes patenting decisions of early innovators. Secrecy can significantly decrease investment in the second innovation. We show that as the effectiveness of secrecy as a protection mechanism increases, which may be the case if the government has a strong trade secret policy or innovators can monitor the flow of their technological information, it becomes optimal to have broad patent protection over a larger parameter space. In cases when patent policy is ineffective in achieving disclosure, it is socially desirable to have a lenient antitrust policy and allow collusive agreements.

Introduction

Recent developments in the economic theory of innovation have emphasized its cumulative nature. One of the main issues addressed in the cumulative innovation literature is how patent and antitrust policy can be designed to divide the profits between sequential innovators in a way that provides them with optimal incentives to invest. Since the cumulative nature of innovation implies that the social value of innovations should include the value of subsequent innovations they inspire, several papers have argued for broad patent protection in order to provide early innovators with sufficient incentives to invest.1

One of the common assumptions made in this literature is that all innovations are patented. This paper challenges this assumption by pointing out that innovators frequently rely on secrecy to protect their innovations (Cohen et al., 2000, Levin et al., 1987).2 When early innovators prefer secrecy over patenting, they can severely affect the investment incentives of rival firms in subsequent R&D races. The non-disclosure of early innovations also delays the benefit society gets from the use of those innovations. Hence, we argue that as the effectiveness of secrecy as a protection mechanism increases, the design of optimal policy should pay attention to encouraging disclosure and stimulating investment in subsequent innovations. This implies that it becomes optimal to have broad patent protection over a larger parameter space. If patent policy is ineffective in achieving disclosure, it is optimal to have a lenient antitrust policy.

We consider two consecutive R&D races as in Denicolo (2000). The firms race to achieve two symmetric and competing innovations, and the winner of the first race can also participate in the second race. At the end of the first race, the winner decides whether to patent the innovation or keep it secret. Under the patent law, an applicant is required to disclose sufficient information about the innovation to enable someone skilled in the art to make and use all the embodiments of the innovation claimed in the patent. Thus, the winner of the first race may be reluctant to patent, in order to get a head start in the second R&D race. After observing the patenting decision of the innovator, firms invest to develop an improved version in the second R&D race. If the first innovation is not disclosed, rival firms must spend resources to gain information about it or re-invent it, which reduces their competitiveness in the second R&D race.3

The two goals of the patent system are to encourage research and development, and to promote the disclosure of innovations. The analysis of optimal patent breadth has focused mainly on the first goal while our model aims to draw attention to the second. If innovators choose patenting, the effective life of a patent is determined by how the courts interpret the patent and antitrust laws.4 If innovators choose secrecy, the government's trade secret policy determines the amount of protection innovators get against theft and unauthorized disclosure of their trade secrets. In this paper, taking trade secret policy as given, we analyze optimal patent and antitrust policy.

Specifically, following Chang (1995), we compare the behavioral implications and welfare properties of three policy regimes. The first regime is broad patent protection, where the courts find that the second innovation infringes the patent of the first innovation. Such a policy regime allows the first innovator to collect some licensing fees from the second innovator if it loses the second race. Under narrow patent protection, two possibilities exist; the antitrust authorities can have a policy of collusion or no collusion. Thus, we ask whether it may be desirable to permit collusive licensing deals between holders of competing but non-infringing patents. Such collusive agreements increases the investment incentives of innovators by eliminating the competition between them.

Taking into account the patenting incentives of early innovators suggests new ways in which broad patent protection or collusive licensing agreements may be socially desirable. The results illustrate that broad patent protection gives early innovators the highest incentives to patent. Thus, the desirability of broad patent protection depends on how attractive secrecy is for innovators. We find that if innovators cannot rely on secrecy to protect their innovations, which may be the case if the government has a weak trade secret policy or if innovators cannot monitor the flow of their technological information, it may not be optimal to have strong patent protection. This is because as the rival firm's probability of success in case of no patenting increases, innovating firms have increased incentives to patent their innovations. In such cases, early innovations are likely to be disclosed through patenting whether or not there is broad patent protection. Therefore, the government may prefer to have narrow patent protection in order to encourage investment in the second R&D race. If, on the other hand, early innovations are likely to be kept secret in the absence of broad patent protection, the government may find it optimal to have broad patent protection in order to encourage patenting. In such cases, the rivals' probability of success in case of no patenting is low and innovators can rely on secrecy as a protection mechanism. Having broad patent protection increases social welfare by encouraging earlier disclosure of innovations.

If early innovators cannot be encouraged to disclose their innovations under any of the policy regimes, it is never optimal to have broad patent protection. In such cases, it becomes optimal to allow collusive licensing agreements between competing innovators in order to stimulate investment in subsequent R&D races. This is because non-disclosure of early innovations decreases the investment incentives of second-generation investors and second-generation investors have the strongest incentives to invest when the courts find no infringement and allow collusion. That is, while broad patent protection encourages investment in the first race, collusion encourages investment in the second race. This is because under broad patent protection, the innovators end up sharing the value of the second innovation while under no infringement and collusion, they end up sharing the value of the first innovation.

Thus, the paper makes a case for having a lenient antitrust policy in industries where innovators rely on secrecy. This result differs from Chang (1995), Priest (1977) and Kaplow (1984), who find little or no support for collusion, and has important policy implications. The general policy of the US Department of Justice and the US Federal Trade Commission is to allow cross-licensing agreements and patent pools that include complementary patents, but prohibit those that include substitute patents (Shapiro, 2001). Analyzing the optimality of collusive licensing agreements in a dynamic setting where secrecy is a possibility, this paper shows that having a lenient antitrust policy may also be desirable in case of substitute patents.

The paper proceeds as follows. Section 2 presents the model and discusses the assumptions made regarding the legal background. In 3 The innovation process, 4 The optimal policy, we analyze the private incentives of the firms and the optimal intellectual property policy respectively. Section 5 contains a detailed discussion of how our results differ from those in the literature. Section 6 concludes and indicates directions for future research. All proofs are in Appendix A.

Section snippets

Research environment and consumers

The research environment is adapted from Denicolo (2000). There are two sequential R&D races and free entry into each race. The size of each innovation is exogenously given and commonly known, but its timing is stochastic. For simplicity, we assume the two innovations are symmetric in terms of both their private and social values. The winner of the first race can participate in the second race, so the model allows for repeated innovation by the same firm. If the winner of the second race is

The innovation process

We are interested in finding the subgame perfect Nash equilibria of the game outlined in Section 2.2. Therefore, we start from the second race and work backwards. After analyzing the investment incentives of the firms under patenting and no patenting respectively, we determine the conditions under which patenting will take place. Finally, we move on to analyze the investment incentives in the first race.

The optimal policy

This section examines optimal policy by focusing on expected social welfare at the beginning of the first R&D race. Social welfare is defined as the sum of producer surplus, consumer surplus, and the non-appropriable values of the innovations. As pointed out in Arrow (1962), for a variety of reasons investors may not always be able to appropriate for themselves the entire social benefit of their innovations. Let s  0 stand for the non-appropriable value of the innovations. It represents the

Comparison with earlier literature

One of the main issues raised in the literature on cumulative innovation has been that if future innovations are improved versions of earlier innovations, the incentives to develop the earlier innovations may be too low. A natural solution is to have broad patent protection (Kitch, 1977). However, such broad protection may have the effect of stifling the investment incentives of subsequent innovators (Scotchmer, 1991). Thus, Merges and Nelson (1990) suggest that because of the hold-up problem,

Concluding remarks

This paper has examined optimal policy assuming innovators can try to protect their innovations by relying on either the patent system or secrecy. Although the patent system provides innovators with the right to exclude others from using their innovations for a fixed period of time, it also requires them to disclose their innovations. The literature on cumulative innovation has so far explored how broad patent protection and collusive agreements can be used to give early innovators sufficient

Acknowledgements

This paper is based on the second chapter of my dissertation (University of Maryland, College Park, 2002). I am grateful to Rachel Kranton, Dan Vincent and Larry Ausubel for their valuable comments. I also thank David Brennan, John Creedy, Stephen King, Stephen Martin, two anonymous referees, and seminar participants at Australian Economic Theory Workshop (2004), European Economic Association Annual Congress (2003), the North American Summer Meeting of the Econometric Society (2003), the

References (33)

  • A. Arundel

    The relative effectiveness of patents and secrecy for appropriation

    Research Policy

    (2001)
  • V. Denicolo et al.

    How should forward patent protection be provided?

    International Journal of Industrial Organization

    (2002)
  • Aoki, R., Spiegel, Y., 2003. Pre-grant patent publication, R&D, and welfare....
  • K.J. Arrow

    Economic welfare and the allocation of resources for invention

  • H.F. Chang

    Patent scope, antitrust policy, and cumulative innovation

    Rand Journal of Economics

    (1995)
  • Cohen, W.M., Nelson, R.R., Walsh, J.P., 2000. Protecting their intellectual assets: appropriability conditions and why...
  • V. Denicolo

    Two-stage patent races and patent policy

    Rand Journal of Economics

    (2000)
  • V. Denicolo

    Sequential innovation and the patent–antitrust conflict

    Oxford Economic Papers

    (2002)
  • V. Denicolo et al.

    Patents, Secrets, and the First Inventor Defense

    (2002)
  • N.T. Gallini

    Patent policy and costly imitation

    Rand Journal of Economics

    (1992)
  • R.J. Gilbert et al.

    Preemptive patenting and persistence of monopoly

    American Economic Review

    (1982)
  • J. Green et al.

    On the division of profit in sequential innovation

    Rand Journal of Economics

    (1995)
  • I. Horstmann et al.

    Patents as information transfer mechanisms: to patent or (maybe) not to patent

    Journal of Political Economy

    (1985)
  • C.I. Jones et al.

    Measuring the social return to R&D

    Quarterly Journal of Economics

    (1998)
  • L. Kaplow

    The patent–antitrust intersection: a reappraisal

    Harvard Law Review

    (1984)
  • E.W. Kitch

    The nature and function of the patent system

    Journal of Law & Economics

    (1977)
  • Cited by (18)

    • Optimal use of patents and trade secrets for complex innovations

      2021, International Journal of Industrial Organization
      Citation Excerpt :

      Erkal (2005) and Bhattacharya and Guriev (2006) consider a cumulative innovation framework. More specifically, Erkal (2005) considers two consecutive R&D races between two firms in which the winner of the first race has an option to protect their innovation either by patenting or through secrecy, and compares several patent policy regimes. Bhattacharya and Guriev (2006) build a model in which the upstream inventor has two potential licensees in the downstream market.

    • Do firms face a trade-off between the quantity and the quality of their inventions?

      2013, Research Policy
      Citation Excerpt :

      As explained in Section 2, the endogeneity of the propensity to patent in Eq. (6) needs to be addressed. The economic literature often models the decision to patent as a trade-off between the positive protective effect and the negative disclosure effect (e.g. Anton and Yao, 2004; Erkal, 2005; Zaby, 2010). Simply said, a firm applies for a patent when the benefit arising from increased protection against imitation outweighs the cost of disclosing the invention.

    • Patenting in the shadow of independent discoveries by rivals

      2012, International Journal of Industrial Organization
      Citation Excerpt :

      This paper has a different focus, examining innovators' choice between patenting and secrecy decisions in a particular legal environment in which the patent system grants no prior user rights. Choi (1990) and Erkal (2005) investigate decisions to patent from another interesting and important angle. In a framework of cumulative innovation, they examine two options for an innovator: to patent (and commercialize) the basic version of a product or to keep it secret and work on developing an improved version.

    • Pre-grant patent publication and cumulative innovation

      2009, International Journal of Industrial Organization
    View all citing articles on Scopus
    View full text