Cooperation in R&D: The Case of Patent Infringement Agreements

Sugata Marjit, Arijit Mukherjee and Heling Shi
April, 1996

ABSTRACT: The paper attempts to provide a strategic rationale behind patent infringement agreements (PIA). PIA is to be viewed as a form of ¡§coopera-tion¡¨ in an otherwise non-cooperative environment. In this paper, we develop a symmetric duopolistic structure where firms initially hold certain patents and invest for further R&D. In the process of R&D, firms face a positive probabil-ity of infringing the patents held by the other. Once infringement takes place, the infringing firm has to pay a compensation to the patent holder. The PIA scheme generates a binding contract suggesting that such infringements would go unpunished. We prove that (a) If the infringement probability is not too ¡§low¡¨, firms would have incentives to sign a PIA and sacrifice the monopoly option. In case the infringement occurs with certainty, PIA dominates the non-cooperative outcome; (b) For relatively low and high probabilities of in-fringement, PIA dominates state-contingent PIA; and (c) The likelihood of signing PIA will also increase with the probability of failure in R&D.

JEL classification: L1

Key Words: Patent Infringement Agreement, Cross-licensing

Introduction

This paper is about patent infringement agreements (PIA hereafter), a term commonly used in the circuit of software specialists. Companies such as IBM and Apple have some PIA agree-ments in practice1. These agreements are drawn to make such that the software developers in course of developing new software programs can utilize the existing knowledge of the competing specialists. Interviews with self-employed programmers reveal that they are interested in accumulating a reasonable number of patents so that other big companies are willing to write such agreements with them. In fact this is perceived to be a mean of recogni-tion for the entrepreneurship in the business. This paper provides a theoretical rationale behind PIAs. We view PIA as a cooperative agreement in the sphere of R&D in the absence of explicit or implicit collusion in the product market. This immediately relates our work to the growing literature on cooperation in R&D2.

More Specifically, PIA is an arrangement that: i) is designed to deal with the acciden-tal infringement of other party's patents. An infringement is said to be accidental if the infringement or non-infringement does not significantly alter infringing party¡¦s investment costs in R&D, nor does it substantially influence the returns in marketing the final product3. Under such definition, infringement of patents is not a strategic choice; and ii) is usually signed before any R&D activities would be undertaken [like research joint ventures (RJV)].

Patents laws assign monopoly rights to the patent holder and it is usually expected that such rights will be exercised in the event of a patent violation. Apparently close competitors in a particular product market should be enthusiastic about enforcing their patent rights to protect their market share from their rivals. However, such aggressive behavior might be unsound when firms, in their bids to innovate a new process or product, are likely to overstep into each other's existing territory. In a world where innovations are extremely frequent and technological progress is intense, like information technology Industry, competing firms might find it costly to exercise their patent rights all the time.

Apart from such "transaction" costs, there is also a reciprocal infringement problem. Intuitively, if two firms operate in an industry where innovations are extremely frequent, though the protection of one's patents could enhance its market power by prohibiting rivals imitating the new technology and new product, it is equally possible that he would infringe other party's patents, be prohibited to use the new technology and new product, be forced to pay compensation, and therefore suffer the lose of market power. PIA works like risk-pooling: it scarifies the possibility of gaining market power to avoid the possibility of losing market power. success of these two operation systems are mainly lying on their own merits. In IT industry, technology innovation is converging and some level of overlapping seems inevitable. 4

Cross-licensing of patents is a well known phenomenon4 and recent papers by Fershtman and Kamien (1992) and Eswaran (1994) explain in detail the strategic motivation underlying such arrangements. Cross-licensing, as explained in these works differs from a PIA. Under cross-licensing firms can actually market the knowledge brand of the competitor. Fershtman and Kamien (1992) show how complementary technologies would be cross-licensed while Eswaran (1994) explains how cross-licensing of competing brands can facili-tate implicit collusion in the product market. With PIA firms are not supposed to market competing brands and in fact we would show that they do not have any such incentives. Even a absolutely non-cooperative product market can give enough incentive for the parties to write a PIA. A PIA is required simply because accidental patent-violation can be very costly. The right to utilise the existing technology of the purpose of R&D, however, does not allow the rivals to market each other¡¦s brand. This is a crucial difference from the pure cross-licensing phenomenon. It is assumed of generality that IBM packages can not be marketed by the Apple or vice verse. Under PIA, firms may not be observed to utilize the technologies of their rivals.

In fact in the ex post equilibrium it may or may not happen. But there will be an ex ante contract. This is not possible in the case with cross-licensing of complementary technologies; in which the firms have to actually use their technologies. But when the technologies are substitutable, we often observe that the firms are not willing to write cross-licensing contract. This is pointed out in Eswaran (1994) and our approach confirms this conjecture. However, Eswaran (1994) was interested in treating cross-licensing as a ¡§facilitating device¡¨ and highlights the case where firms would agree not to use the acquired technologies, ie. the other extreme of Fershtman and Kamien¡¦s (1994) model. As would be evident, our model accom-modates both such cross-licensing in a non-cooperative set up.

In what follows, we develop a symmetric duopolistic structure where firms initially hold certain patents and invest for further R&D. In the process of R&D, firms face a positive probability of accidentally infringing the patents held by the other. Without an agreement, once infringement takes place, the infringing firm has to pay a compensation to the patent holder, and the new technology can not be marketed. The PIA scheme, on the other hand, generates a binding contract suggesting that such infringement would go unpunished. PIA confers the benefit that if the R&D is succeeded the firms has utilized the rivals patented information, one does not have to worry about the legal problem in marketing the new technology. The implicit cost of writing a PIA is that one has to sacrifice the monopoly option which is there without the PIA, ie. non-infringing successful innovator can emerge as a monopolist when other player¡¦s success depends on the infringement. Trade-off between these incentives would generate a PIA provided the infringement probability is not too low. We show this in section II.

We introduce the problem of a self-enforcing PIA in section III, dropping the assump-tion that a PIA as developed in section II is enforceable in the court of law. We derive alternative contractual form called ¡§state-contingent PIA¡¨ or SPIA. We prove that SPIA will always dominate non-cooperative R&D. With monitoring costs, it is likely that PIA will dominate SPIA when the probability of infringement is relatively high or relatively low.

Both section II and III assume that there is no failure in R&D. Section IV extends the simple model of PIA into a model with a positive probability of failure and argue that chance are even higher that a PIA would be contracted relative to the ¡§no-failure¡¨ case in section II and III. Section V concludes the paper.

We can prove that: a) If the infringement probability is not too ¡§low¡¨, firms would have incentives to sign PIA. In case the infringement occurs with certainty, PIA dominates the non-cooperative outcome (In a simplified model with a Cournot-Nash example of cost-reducing drastic innovation, an infringement probability as low as 20% could induce firms to sign PIA); b) For relatively low and high probabilities of infringement, PIA dominates SPIA; for a moderate probability of infringement, SPIA dominates PIA; c) The likelihood of signing PIA will also increase with the probability of failure in R&D.

Download

Inframarginal Econimics Society www.inframarginal.com

 

© INFRAMARGINAL ECONOMICS --- ALL RIGHT RESERVED