TABLE OF CONTENTS I. INTRODUCTION 418 II. PATENT SETTLEMENTS BETWEEN RIVALS 424 A. Horizontal Restraints: Examples 424 B. What's the Problem? 426 1. Contrast with Settlements in Other Contexts 427 C. The Scope of the Patent Doctrine 429 D. Actavis and Proportional Effects 431 1. Pay-for-Delay Settlements 433 2. The Actavis Decision 435 E. The Administrability Problem 436 III. THE ECONOMICS OF SETTLEMENT DESIGN: AN OVERVIEW 438 A. Relevant Aspects of Settlement Design 439 B. Licensing Restraints 440 1. Applications 445 C. Reverse Payments and Counter-Restraints 450 1. Reverse Payments 451 2. Counter-Restraints 454 3. The Pareto Problem 456 IV. MULTIPLE ENTRY, HATCH-WAXMAN, AND OTHER CONSIDERATIONS 457 A. Clarifying the Hatch-Waxman Problem 460 B. Litigation Costs 462 C. Cross-Licensing Settlements 463 V. REFRAMING THE ANTITRUST QUESTION 465 A. Administration 465 B. Demystifying Causation 471 C. Evaluating Potential Objections 473 1. What if the Patentee is Risk-Averse? 473 2. Appeals to Non-Pareto Hypotheticals 475 VI. CONCLUSION 477 I. INTRODUCTION
Antitrust usually prohibits rival firms from striking agreements that forestall competition. Patent settlements provide an exception, however, because a patent on a significant technology may provide a lawful basis for excluding competitors from the marketplace. Problematically, firms always prefer to restrain competition into monopoly (and share in the proceeds), even if they privately believe that the patent is very likely invalid or noninfringed. (1) This might not be such a challenging problem if it were easy to discern whether a given patent is valid and infringed. But in practice this is almost never the case. (2)
The settlement problem is emblematic of the analytical difficulties that emerge at the intersection of antitrust and patent law, which has long been a source of widespread confusion and debate. (3) In some cases, the settlement restrains inter-party competition in a way that is not even facially authorized by patent law, even if the relevant patent is definitively valid and infringed. (4) But in the more challenging cases--which are the focus of this paper--the firms' settlement restrains competition in a manner that potentially falls within the patentee's exclusionary entitlement. Specifically, the settlement forecloses competition by no more than a permanent injunction would, which is the most restrictive remedy available for patent infringement. (5) For example, if a settlement limits the monthly sales of a patented product by a rival licensee, it clearly diminishes competition. Indeed, if not for the patent, it would be illegal per se. But it is less restrictive than an injunction, which would not permit the rival to make any sales at all.
As outlined below, antitrust analysis of these settlements has become increasingly concerned with how patent litigation would have played out but for the settlement. This is now widely-regarded as the appropriate benchmark for evaluating the settlement's competitive effects. (6) To that end, scholars and courts have framed the antitrust inquiry to require evidence or "signals" bearing on the likelihood that the relevant patent is invalid or noninfringed. This gives the antitrust claim a "case-within-a-case" structure, as liability then hinges on speculation about the expected outcome of counterfactual patent litigation. Notwithstanding the challenges this presents in practice, such speculation is thought to be logically necessary to answer the antitrust question.
But this is not so. In fact, it is the design of the settlement agreement--the way it restrains competition or otherwise influences the distribution of profits--that determines how the parties' ultimate agreement will compare with their expectations about litigation. (7) As I argue below, this supports a wholesale shift in how the antitrust inquiry is conceptualized and administered. In short: evaluate the agreement, not the patent.
Consider an example illustrating the underlying problem. A patent holder and its only rival are very confident that the rival's product does not infringe the operative patent--so confident that the rival will not agree to pay even a moderate royalty, as it gives little weight to the litigation threat. But the patentee proposes an alternative settlement, which restrains the rival in a different way. It offers to license the rival to be the exclusive seller of the patented product in one half of the country, so long as it does not make sales elsewhere; (8) the patentee will then be the exclusive seller in the other half of the country. (9) This generates a monopoly, albeit a divided one, since there is no competition in any given territory. It is thus easy to see why the rival accepted this proposal, as the settlement essentially creates a cartel. (10) Still, the arrangement is no more restrictive than an injunction would have been.
The problem is that litigation would have produced a much more competitive result--not with certainty, but in expectation. (11) The patentee would likely have lost, in which case the rival would have an unambiguous right to compete freely. Framing the antitrust inquiry to focus on the expected result of litigation is relatively new. Considerations of counterfactual litigation were rarely raised, let alone dispositive, in cases prior to the 21st century. (12) In a 2003 article, Carl Shapiro proposed that all horizontal settlements should be evaluated based on how their competitive effects compare to the expected result of litigation. (13) This approach quickly came to dominate the antitrust literature (14) and was recently embraced by the Supreme Court in FTC v. Actavis. (15) Under this standard, which I call "the proportional-effects rule," anti-trust's goal is for settlements to be proportional: to restrain competition to an extent commensurate with the expected result of patent litigation. (16)
At the outset, this standard would appear to suffer from a major administrability problem. (17) How do you enforce a standard that hinges on the likelihood that a patent would have been held valid and infringed in counterfactual litigation? This would ostensibly require a casewithin-a-case analysis of the relevant patent. However, in one well-known settlement format--"pay-for-delay," also known as a "reverse payment" settlement--many antitrust commentators advocate a workaround. In such settlements, the patentee makes a lump sum "reverse payment" to its prospective rival in exchange for the latter's agreement to delay its entry until later in the patent term, often soon before expiration. (18) A large reverse payment is thought to be a suitable proxy for the patent's likely invalidity. (19) The Supreme Court's Actavis decision endorsed this proxy-based approach, (20) which has since been dubbed the "Actavis Inference." (21)
However, there remain major obstacles to enforcement of the proportional-effects rule. First, the Actavis Inference is applicable only in settlements that involve reverse payments. Many settlement arrangements--including essentially all types that received antitrust scrutiny prior to the 21st century--do not contain such payments. Second, even in post-Actavis pay-for-delay cases, some courts have essentially reverted to the requirement that a private plaintiff show (directly, not by proxy) that the patent is invalid. (22) In practice, this may be an insurmountable barrier to enforcement. (23)
These problems could be avoided entirely if the patent issues were wholly removed from the antitrust analysis. In a recent economics paper, Jorge Lemus and I demonstrate that a settlement's proportionality (or lack thereof) is ultimately driven by its design--the particular way it restrains competition, along with any other provisions influencing the distribution of profits. (24) That paper provides a formal economic theory of the relationship between settlement design and proportionality. The theory relies on only the most general properties that all restraints exhibit and can be applied within any economic model of oligopoly competition.
The present article applies those economic insights to antitrust law and practice. Among other things, I emphasize that: (a) this approach can be administered much more practicably, accurately, and broadly than the prevailing patent-focused approach; (25) (b) it is consistent with the Supreme Court's Actavis decision, notwithstanding that it removes the patent issues from the antitrust claim; (26) (c) it simplifies the antitrust analysis by disentangling the relevant antitrust violation from the extent of the resulting harm; (27) and (d) it clarifies a number of critical errors in arguments advocating against antitrust intervention in patent settlements. (28)
In broad outline, a settlement's design determines what the parties can agree on in relation to their litigation expectations. That is precisely the comparison that drives antitrust liability under the proportional-effects rule. This leads to a counterintuitive result: the "design-focused" approach makes it possible to administer the proportional-effects rule--a standard intimately concerned with probability--without having to assess the probability that any particular patent is valid and infringed. This brings the antitrust analysis back into familiar territory, for it can focus entirely on the nature of the agreement; no more case-within-a-case. And, significantly, this analysis can be applied to all kinds of settlements, including those involving multiple rivals or multiple patentees. (29)
In fact, one example of this design-focused approach in action is already well-known. In a "pure delay" agreement, the patentee's rival agrees that its entry will be delayed for some negotiated period of time, but it does not receive any reverse payment. As numerous commentators have noted, (30) this settlement forces the firms to settle on terms that emulate their litigation expectations no matter...