W H E N A N A L Y Z I N G T H E intersection of antitrust and - - PDF document

▶

Jan 24, 2023 235 likes •343 views

A R T I C L E S A N D F E A T U R E S costs for Butch would be rather high, to say the least. But if Hatch-Waxman Harriman had paid Butch to leave his trains alone, he would have been shocked to learn that he faced exposure under the Patent

SLIDE 1

A R T I C L E S A N D F E A T U R E S

6 8 · A N T I T R U S T

Hatch-Waxman Patent Settlements and Antitrust: On “Probabilistic” Patent Rights and False Positives

B Y K E V I N D . M C D O N A L D

“[F]alse positives are much more harmful than false negatives. Market processes undercut monopolies wrongfully permitted, but no similar processes undercut judicial decisions that wrongly condemn efficient conduct.”

—Frank H. Easterbrook, Judge, United States Court of Appeals for the Seventh Circuit 1

W

H E N A N A L Y Z I N G T H E intersection of antitrust and patent law, I find it useful to consult the great economic thinkers, such as Judge Easterbrook (quoted above) and Butch Cassidy. The latter is, of course, the eponymous hero played by Paul Newman in the classic film, Butch Cassidy and the Sundance

Kid. Recall the scene in which Butch and Sundance return

from their long and harrowing escape after a failed train rob-

bery. Butch is shown a news article reporting that the Union

Pacific’s owner, E.H. Harriman, has hired and equipped a collection of famous lawmen to hunt Butch and Sundance

down. Butch’s look of puzzlement turns to disgust as he leans

forward, jabs his finger into the table top, and delivers an economic analysis worthy of Jevons and Walras:

That’s bad business . . . . If he’d just pay me what he’s spend- ing to make me stop robbing him, I’d stop robbing him.

It is unlikely, to be sure, that Mr. Harriman would ever have the opportunity to make such a payment. The transaction costs for Butch would be rather high, to say the least. But if Harriman had paid Butch to leave his trains alone, he would have been shocked to learn that he faced exposure under the recently-enacted Sherman Antitrust Law. After all, some apparently would argue, Butch could sell what he took on the black market at greatly reduced prices, and consumers would

benefit. Wait a minute, Mr. Harriman would sputter, it’s my
train. No sir, would come the confident reply, you mean it’s

probably your train. * * * * * * * The ongoing debate concerning the application of antitrust law to intellectual property has been hobbled by certain fal-

lacies. These fallacies have a common source: the perceived

need of some who seek to attack patent settlements to devel-

p a theory of liability that does not depend on the validity
f the patent. In other words, certain commentators and

members of the plaintiffs’ class action bar wish to argue that settlements of patent litigation are presumptively, if not per se, unlawful even if the patent in question is perfectly valid and thus would have precluded all competition for the patented good. The desire for such a theory is understand- able; proving that a patent is invalid is hard work. Making a verdict against a patent stick in the Federal Circuit is often harder work. Hence, the quest for an antitrust theory that can condemn a patent settlement while declaring the validity of the patent irrelevant. I have previously addressed one attempt at such a theory: the notion that settlement payments from a patent holder to the patent challenger flow the “wrong way” and thus are presumptively anticompetitive.2 In the “traditional” settlement, goes the argument, the patentee grants a license to the alleged infringer, who then pays a royalty flowing to the patentee. Payments going the other way (i.e., from the patentee) are therefore “reverse” payments. This argument has fared poor- ly, in my view, because it evades the fundamental question: No matter who paid what to whom, what lawful competition has been reduced by the settlement? That was the question from which the FTC’s administrative law judge could not be deflected in his recent decision in favor of Schering-Plough.3 There, the FTC staff learned to its chagrin that declaring the patent’s validity irrelevant was inconsistent with the minimum requirement of showing consumer harm, that is, showing that the alleged infringer had a right to be in the market at all:

Schering had the legal right to exclude Upsher-Smith from the market until Upsher-Smith either proved that the ’743 patent was invalid or that its product . . . did not infringe Schering’s patent. . . . [Thus,] Complaint Counsel has not proved that Upsher-Smith . . . could have even been on the market prior to the expiration of the ’743 patent.4

(If Mr. Harriman pays Butch, you can’t complain until you show that Butch had a right to be on the train in the first place.) In another thorough (and, in my view, unanswer- able) analysis, Daniel Crane has shown that reverse payments provide no useful evidence that a patent settlement is anti-

Kevin D. McDonald is a partner at Jones Day in Washington, D.C., and represents certain defendants in private antitrust litigation arising from the settlement of pharmaceutical patent litigation. No other person or entity is to blame for these views.

SLIDE 2

S P R I N G 2 3 · 6 9

competitive: “The ‘directional flow’ of the settlement payment does not, standing alone, provide a basis for evaluating the potential anticompetitive effects of a settlement agreement.” 5 To perform a proper antitrust analysis of such settlements, as one district court recently noted, “it appears beyond doubt that a court will have to determine the validity, enforceability or scope of [the] patent.” 6 Perhaps recognizing that a rhetorical flourish like “naked”

r “reverse” payments won’t render patent validity irrelevant,
thers have resorted to more extreme measures—such as

defining the problem out of existence. Which brings me to my subject here, a second theory claimed to render patent validity irrelevant. This is an argument, advanced by Professor Carl Shapiro in one paper, and by Professor Keith Leffler and co-author Cristofer Leffler in another, that patent rights are not like other property rights; they are merely “probabilistic”

r “partial” rights.7 Under this view, the patent grant confers
nly the right to go to court to stop perceived infringement.

Because the court may rule against the patent, argue the Lefflers, a settlement payment that prevents this possibility is per se illegal.8 Just like that. Under his more sophisticated approach, Professor Shapiro argues that the possibility of the patentee’s losing must be treated as diminishing the patent’s “strength.” Accordingly, if the settlement allows the patent lit- igants to share more than that reduced patent value, which he purports to derive mathematically, it is anticompetitive.9 Both conclusions depend upon the definition of patents as “probabilistic” rights. What’s more, the authors bristle at the suggestion that a court resolving a patent dispute might make a mistake: “For purposes of antitrust analysis, there are and can be no ‘wrong’ decisions reached by courts in patents litigation.” 10 (Honestly, it says that.) As I hope to show, this “theory” of probabilistic patent rights is simply an exercise in assuming one’s answer. Its premise is contrary to already existing principles of antitrust and patent law, and its conclusion would radically reform legal principles of causation that have been in place for cen-

turies. As a matter of competition policy, moreover, the argu-

ment proves vastly too much and is unworkable in practice. It would replace one inquiry that is difficult but feasible (i.e., proving a patent valid or not) with an inquiry that is candidly unknowable (i.e., how a particular patent lawsuit would have turned out) and beset by other imponderables, such as dif- fering assessments of the probability of winning (which can vary over time, between parties, and even within firms). Given that these theories expressly focus not on the patent’s

bjective validity, but on the potential result of a specific

patent case (the one being settled), the refusal of these authors to consider the possibility of an erroneous court decision may seem puzzling, to say the least. But the puzzle is solved when one realizes that introducing false positives into these theories causes them to crumple. Indeed, factoring in false positives reveals a hidden flaw in the theory of “probabilistic” patents: that adopting it would risk inflicting unambiguous and irrevocable consumer harm in certain cases—a risk that is not run if we simply face the issue that matters in most of these cases: patent validity. When Hatch Met Waxman While the explosion of interest in antitrust and intellectual property can be traced in large measure to cases like Microsoft and the revolution in information technology, the dominant arena for actual disputes and analysis has been pharmaceutical patent litigation under the so-called Hatch-Waxman Amendments.11 The scenario is this: A generic drug maker (G) applies to the FDA to manufacture a generic version of a branded drug. Under the Hatch-Waxman procedures, G can rely on the same proof of safety and efficacy used origi- nally by the patentee (P) of the branded, or pioneer, drug. Thus, G may file an Abbreviated New Drug Application (ANDA) showing only that its generic drug is “bioequiva- lent” to the pioneer drug. Hatch-Waxman also requires G to make one of four certifications as to the patent status of the pioneer drug: (I) that it never had a patent; (II) that it had a patent, but the patent has expired; (III) that it has a current patent, but the generic drug will be sold only after expiration;

r (IV) that it has a current patent, but the patent is either

invalid or will not be infringed by the generic drug.12 (No. IV is where the action is.) When P receives notice of the “ANDA(IV)” filing by G, P has forty-five days under the statute to sue G for patent infringement. If P does file suit, the FDA may not approve G’s ANDA for at least 30 months.13 Let us pause for a moment to consider the relative posi- tions of G and P as a result of this process. What we find is that the procedures created by Hatch-Waxman turn the “traditional” settlement scenario trumpeted by antitrust plaintiffs—where the patentee may grant a license and receive royalty payments—on its head. The traditional context pre- sumes an infringer who is actually, well, infringing the patent. This means that the infringer has made a capital investment to enter the market (particularly large in the case of drugs prior to Hatch-Waxman). This traditional infringer is therefore already in the market, taking sales away from the paten-

tee. Because it is selling at lower prices than the patentee,

moreover, every sale is producing less profit to the infringer than the patentee was making. Thus, the infringer’s profit on each sale is less than the damage the infringer will owe for that sale if the patent is upheld. The “traditional” infringer, therefore, is exposing itself to crushing liability for damages, which may also be trebled under the statute. Hatch-Waxman changed all that. It created what the Supreme Court has called a “highly artificial” act of infringement, i.e., the mere filing of an ANDA(IV) certification by the generic applicant.14 When the patentee sues under Hatch- Waxman, the generic applicant has, in effect, an opportunity to obtain a declaratory judgment that the patent is invalid

r not infringed.15 The statute expressly shields the generic

from damages unless it has actually marketed the drug. Thus, the only cost to G of obtaining a ruling that may strike down

SLIDE 3

A R T I C L E S A N D F E A T U R E S

7 · A N T I T R U S T

P’s patent is legal expense. If such a lawsuit is to be settled, there is only one way for payments to “flow”—from the party (P) with everything to lose (the patent) and nothing to gain (no prospect of damages) to the party (G) with everything to gain (free entry) and nothing to lose (no exposure to damages).16 Suppose, to keep our example straightforward, that P’s patent claims the drug’s active ingredient, meaning that no generic drug can be marketed without infringing it. (In other cases, a patent might claim only a certain formulation of the drug, meaning that some generic versions would infringe, while other differently formulated versions would not). Thus, the only issue in the patent suit will be whether the patent is valid.17 Now suppose that P and G agree to settle the case. G agrees to acknowledge the patent’s validity and not to infringe

it. In return, P agrees to make direct payments to G or, at P’s
ption, to provide a license.

The antitrust rub for some is that P appears to be “paying” G, another drug company, to stay out of the market. This means, they argue, that the agreement is preserving the monopoly “rent” of the patentee (by giving some of it to G to go away).18 Doesn’t that look a lot like a market division agreement? Perhaps, but there are people in Las Vegas who look a lot like Elvis. If you think about it a little longer, you will note (along with Professor Areeda) that, but for the existence of the patent, all patent licenses meet the definition of “per se unlawful naked horizontal market divisions.” 19 The reason that licenses usually present no antitrust problem is that the presence of the patent means they preclude no competition that would otherwise legally occur. In other words, when P shares its patent rights with its putative competitor, “we cannot say that it eliminates any competition that existed between [them].” 20 The same inquiry applies to the Hatch-Waxman settlement with “reverse” payments. Can we say that “it eliminated any competition that existed” between G and P? If the settlement only acts to bar infringing drugs, the answer is no—for precisely the same reason that applies to a license: “[T]he public [is] not entitled to profit by competition among infringers.” 21 Thus, calling settlement payments divided “rents” doesn’t advance the analysis unless we know—as a minimum start- ing point—whether those rents were generated by an invalid

patent. For it was a concern that settlements might shield

invalid patents that called for antitrust review in the first

place. As a former director of the FTC’s Bureau of Compe-

tition put it, the challenge for “antitrust enforcers lies in identifying those cases in which settlements enable parties to share economic rents through patents that the parties know to be invalid.” 22 Private plaintiffs and some commentators would like to read the last nine words out of that sentence, and condemn agreements that “share economic rents” through any patent, valid or not. But if you ignore the patent’s validity, you take away your ability to show consumer harm, for the reasons articulated with withering accu- racy by FTC Commissioner Thomas Leary: “If the patent is valid, the pioneer manufacturer is entitled to its monopoly profit, and a settlement that merely transfers a portion of that profit to a potential generic manufacturer causes no harm.” 23 (As long as the train belongs to Mr. Harriman, he can share it with whomever he pleases.) But there is more to the importance of patent validity than just the lack of competitive harm—and this is the truth from which so many avert their eyes: if the patent is valid, then consumers benefit from the very absence of rivalry and presence of monopoly rents that plaintiffs bemoan. As one prominent economist explained it:

Even the introduction of a product subject to monopoly power can represent a gain to society. That is the underlying logic of our patent system, in which monopoly profit expected from innovation creates an incentive to provide the gain to society. It has been estimated that the social return to invention significantly exceeds the private return.24

It is important to recognize that these consumer benefits are procompetitive benefits, and hence highly relevant to any antitrust analysis. Indeed, there is widespread agreement that the existence and protection of valid patent rights “driv[e] economic growth and increas[e] consumer welfare.” 25 Even the articles arguing for “probabilistic” patent rights pay lip service to the procompetitive benefits of patents and patent settlements—at least before adopting an analysis that necessarily assumes them away.26 The slide into fallacy begins when we ignore this essential dichotomy: that consumer interests potentially run in opposite directions when a patent is at issue. It is not that the patent laws and the antitrust laws are at cross-purposes; the tension lies within the antitrust laws. Depending on one circumstance, precisely the same outcome for consumers (exclusion of competitors, monopoly profits) is procompetitive and beneficial in one case, and is (or can be) anticompetitive and harmful in another.27 That circumstance is patent validity. The Birth of “Probabilistic” Patent Rights Despite the setbacks for the “wrong way” payment flow argument, the demand for a theory rendering patent validity irrelevant remains high. In a recent issue of ANTITRUST, for example, the authors of an article on patent settlements acknowledged that “any precise identification of the antitrust risk would require assessment of patent validity and scope,” but then promptly declared that “assessing probable validity and infringement in an antitrust proceeding fails to provide a tractable or predictive legal standard.” 28 At the risk of being inconvenient, I must ask: why? The inquiry can be grueling, to be sure, which is why patent litigation is long and expen- sive, but we do make the inquiry. The issue is presented in thousands of pending cases; there are established theories of invalidity and accepted ways of proving them. If, as Commis- sioner Leary acknowledges, “the ultimate competitive impact

f a pharmaceutical patent settlement is really dependent on

the merits of the underlying patent litigation,” 29 and if the

SLIDE 4

S P R I N G 2 3 · 7 1

source of our concern is that the settlement may be an attempt “to share monopoly profits from an invalid patent,” 30 how can we—and why should we—shrink from the crucial inquiry? I do not wish to appear naïve here. I know that antitrust enforcers are human. They work hard enough without adding the nightmare of proving a patent invalid. Nor (quite under- standably) are many of them qualified to do so. And the prospect of having the antitrust issues in these cases settled in appeals to the Federal Circuit, which tends to be highly respectful of patent rights, is no doubt enough to make any antitrust plaintiff hyperventilate. Hence the attempt to redefine the patent right so that the irksome issue of patent validity disappears. Hence the birth of “probabilistic” patent rights. The theory goes like this: Patents are not existing property rights in the conventional sense of land or chattels. They are merely a prediction of future rights, which are “uncertain

r imperfect until they have survived a challenge in court.” 31

Here is how Professor Carl Shapiro describes it:

A patent is best viewed as a partial or probabilistic property

right. What the patent grant actually gives the patent hold-

er is the right to sue to prevent others from infringing the

patent. Nothing in the patent grant guarantees that the

patent will be declared valid, or that the defendant in the patent suit will be found to have infringed. In other words, all real patents are less strong than the idealized patent grant usually imagined in economic theory.32

The Lefflers similarly declare that a patent is nothing more than “a right to ask a federal court to bar competitors under the procedures and rules dictated by Congress.” 33 Where, you may ask, does this definition come from? Nowhere, actually. Shapiro cites no source from law or eco-

nomics. 34 The Lefflers cite only Shapiro.35 Nonetheless, the

advantages of this definition for the authors are obvious: If a patent does not really exist until you go to court, then it is neither valid nor invalid. Before the court rules, the patent is unborn, so to speak—in a gestational period. Pesky rules about the presumption of patent validity or the failure of antitrust law to protect infringing entry need not detain us. The very possibility that the patent may be struck down becomes a consumer “benefit” that is lost when the parties

settle. That loss alone constitutes the present harm to con-

sumers when the settlement occurs. How liberating. How

tidy. How wrong.

“Probabilistic” Patent Rights and the Law The notion that property is not property until a court says so is novel. It will surely surprise all of those patent holders who routinely buy, sell, mortgage, and license patents with-

ut ever going to court. And the same semantic game can be

played to redefine any other form of property: a deed is only a “right to ask a court to bar” trespassers; the title to a car is

nly a “right to ask a court” for a judgment in conversion.

This would be a typical exchange from the land of probabilistic property: Ed: Hey Ralph, is that your new car? Ralph: Well, I paid for it and I have the tags, but no court has actually said it’s mine. There’s—you know—no, um, con- trolling legal authority. The authors cite no authority for their legal definition of patents because the precedent that exists contradicts them. One opinion made the point by carrying the metaphor of the “unborn” patent to a rather grisly extreme: “A patent is born

valid. It remains valid until a challenger proves it was stillborn
r had birth defects, or it is no longer viable as an enforce-

able right.” 36 The Administrative Law Judge in Schering-Plough also made it clear that the patents before him would be presumed valid until someone proved otherwise, citing Supreme Court precedent dating back to 1824: “[A] patent is presumed to be valid, until the contrary is shown.” 37 It is also worth mention- ing that Congress has passed a statute stating, with unchar- acteristic elegance: “A patent shall be presumed valid.” 35 U.S.C. § 282. The theory of “probabilistic” patent rights would render that statute largely meaningless. It would apply

nly after the issue of validity was resolved in court, when a

presumption is no longer needed. Their definition of patent rights also liberates the authors from other nettlesome rules of law. Based upon his definition, for example, Shapiro declares that, where an accused infringer obtains FDA approval and decides to enter the market before the patent case is over, consumers must be given the benefit of such competition whether it is infringing or not: “[C]ompetition that would take place under the shadow of patent litigation is considered entirely legitimate, even though it may wind up constituting infringement.” 38 This ignores, however, the fundamental rule that FDA approval does not give anyone the right to sell infringing products—any more than a license to operate gives a street vendor the right to sell “knock-off” clothing. The courts are therefore clear that FDA approval of a generic drug means

nly that, “aside from potential patent issues, the product is an

approved generic drug,” 39 and that even after such approval “the patent holder is still entitled to the benefits of its right to exclude until the patent expires.” 40 The generic entrant who reads the Shapiro essay, enters the market, and is later found to have infringed the patent is certain to feel put upon. If any infringing competition during the lawsuit is “considered entirely legitimate,” G may ask, why do I have to pay trebled damages? If the professor’s response is, we don’t care if G is held liable as long as consumers benefit (which strikes me as a bit cold), I have more bad news. The patent law makes anyone who uses the infringing drug an infringer as well, which includes every consumer who swallows a pill. “Contrary to widely held belief, the statute does not immunize or exempt personal or noncom- mercial use.” 41 In Shapiro’s view, therefore, a patentee would be liable to consumers for a settlement that precluded this interim, infringing competition at the same time that the consumers in his “but for” world would be liable to the

SLIDE 5

A R T I C L E S A N D F E A T U R E S

7 2 · A N T I T R U S T

patentee, in the same amount, for infringement of the patent. Hmmm. These legal obstacles to the probabilistic theory of patents, while insurmountable absent several acts of Congress, only hint at the radical nature of the argument. The conclusion, you will recall, is that the mere possibility of the patent being found invalid is a consumer benefit that a settlement may not

eliminate. If it does so, the settlement is illegal. In Shapiro’s

view, the settlement must result in a price to consumers suf- ficiently lower than the full value of the patent to reflect the present value of the possibility that the patent might be struck down. This theory, if implemented by the courts, would be the first to find that a settlement caused competitive harm where that harm is derived exclusively from the failure of the parties to carry the litigation to fruition. As Shapiro puts it, “consumers have a ‘property right’ to the level of competition that would have prevailed, on average, had the two parties litigated.” 42 This “right” exists, and limits the parties’ settlement options, even where it is far more likely that the patent would be upheld. If there is a 10 percent likelihood of an invalidity finding, with a 90 percent likelihood

f a validity finding, the settlement has caused competitive

harm unless it conveys to consumers the expected value of that 10 percent chance. For these authors, this possibility of consumer harm, even if tiny, justifies a conclusion that the settlement has, in fact, reduced competition. In our legal system, however, a 10 percent chance of rain is not rain. No area of the law—antitrust or otherwise— permits a plaintiff to substitute the possibility of harm for actual harm. The Supreme Court in Associated General Contractors made it clear that the elements of causation and injury in antitrust law are properly derived from common law principles employed in tort law dating back to Holmes and further.43 These principles are unambiguous: “A mere possibility of such causation is not enough; and when the matter remains one of pure speculation and conjecture, or the probabilities are at best evenly balanced, it becomes the duty of the court to direct a verdict for the defendant.” 44 This principle that “[p]ossibility alone cannot serve as the basis for recov- ery” 45 applies with special force where the injury claimed arises from the uncertain outcome of a contest. As one court put it, “[d]eprivation of the chance of winning a horserace or any sporting event does not present a basis for . . . liability.” 46 I do not mean to suggest here that litigation is as fair as

horseracing. But I do agree with the ALJ in Schering-Plough

that “there is no recognized methodology for handicapping trials or for testing the reliability of predictions of litigation

utcome.” 47

Probabilistic Patent Rights and Policy Even if we could change all of these legal principles, or have Congress carve an exception for Hatch-Waxman settlements, would it make economic sense to replace an objective inquiry into patent validity with a theory of consumer harm based on the predicted value of a possible ruling against the patent? Not much, I’m afraid. First, there is no limiting principle to these arguments. All patent settlements preclude a judicial resolution by definition, and all settlements therefore cause the consumer harm these authors posit. The Lefflers are particularly vulnerable here, for they concede that various settlements with precisely the same consumer outcome (i.e., no rivalry and monop-

ly prices) are perfectly legal while settlements with reverse

payments are per se illegal.48 But why, one may ask, do the payments matter? Under the “probabilistic” theory, consumers are hurt by losing even a miniscule chance that the patent will be struck down. That precise harm still occurs, however, when a patent challenger simply gives up and drops the challenge, or when the challenger who has already entered and faces damages decides to exit in exchange for the patentee’s agreement to drop (or reduce) the damage claim. In both cases, the irreducible “possibility” that the generic could prevail remains. How can this consumer harm simply evap-

rate when the “reverse” payments disappear?

The difference, according to the Lefflers, does not come from any economic theory; they emphasize that, whatever may be “the optimal balance between dynamic and static efficiency as it relates to the reward given to inventors,” that balance has been struck by Congress and may not be second- guessed.49 Instead, the difference lies in the bundle of “sub- stantive and procedural rights” Congress has seen fit to grant patent holders. For the Lefflers, a settlement where G gives up, or where P grants G an exclusive license at the maximum royalty, is somehow within the “bundle of rights granted by Congress” to patent holders, while the settlement with reverse payments is not.50 They do not pretend, however, that this distinction can be found in any word of the statute

r piece of legislative history. Indeed, no citation to any

statute or other Congressional statement appears in the Lefflers’ entire article. And little wonder, for there is no hint that Congress considered limiting settlements in the leg- islative history of Hatch-Waxman, and there are several admonitions that the Act was “not intended to modify existing patent law.” 51 In the end, this curiously specific “bundle”

f rights granted by Congress is whatever these authors

declare it to be: ipse dixit. The more interesting analysis of Shapiro has its own difficulty with breadth, because it applies not just to every settlement, but to every patent. If we define patents as having diminished “strength,” why should protection of that alleged “consumer surplus” be limited to settlements? Consider

licenses. As Shapiro perceptively observes, “[v]irtually every

patent license can be viewed as a settlement of a patent dispute.” 52 When a patentee grants any license, therefore, should the government be scrutinizing the royalty rate to ensure that consumers face a price low enough to preserve their “property right” in the possibility that the patent is invalid? Perhaps there should be a federal office for filing a “pre- licensing notification” or “PLN,” with a modest filing fee. Nor need we stop there. Who says that a patentee that choos-

SLIDE 6

S P R I N G 2 3 · 7 3

es to produce and market the invention by itself is allowed to charge whatever the market will bear, i.e., the full monopoly price? Well, a lot of caselaw and economic literature says exactly that, but they will have to be re-thought. According to Shapiro, that patentee is reaping more rent than the strength of its “probabilistic” patent will justify, and thus appropriating the consumers’ property right. The case for reg- ulating the prices charged for patented products is equally as strong as the case for policing settlements based on the diminished “strength” of every patent. And equally as weak. Another problem with Shapiro’s approach is that it is unworkable, if not unknowable, in practice. As noted, the assessment of a patent’s validity may be occasionally arduous, but it is done all the time. By contrast, we have neither the experience nor the science to predict the outcome of litigation, especially a specific lawsuit that is being settled. How will we make what Shapiro calls an “informed judgment as to the strength of various patents that are at issue?” 53 Will we take the personal views of the parties? Should we care more about P’s prediction, or more about G’s? Which views of the corporate officers and scientists within P and G really count? And what if they were wrong? If we care about consumers, and thus we want to expose invalid patents and protect valid patents, there is no compelling reason to let the issue turn on the selfish interests of P and G. Shapiro also concedes that his mathematical model has not yet worked in fundamental fac- tors that will affect any measurement of the parties’ predic-

tions. He notes that his model “assumes away asymmetric

beliefs,” 54 meaning that both P and G will make precisely the same assessment of the patent’s strength. (Unlikely.) Nor has he yet included the parties’ relative degrees of risk aversion, a crucial determinant in any settlement negotiation.55 In the end, the conclusion is almost irresistible that any measurement of the probability that a patent will be set aside will devolve to the very inquiry that these theories were designed to avoid: whether the patent is actually valid, based upon the kinds of objective evidence that would be used in a patent case. Indeed, Shapiro appears to reach the same conclusion, as he drily notes, that “antitrust authorities may have difficulty determining the probabilities of the various states of the world, particularly inasmuch as this requires a technical assessment of one or more patents and their various claims.” 56 Probabilistic Patent Rights and False Positives If we are going to do this wrong, let’s at least do it right. If we are not going to make the rational and straightforward inquiry into the issue of patent validity that we should, let’s not show bad form by cutting corners in our attempt to predict the outcome of litigation. When dealing with court decisions, as Judge Easterbrook’s opening quotation indicates, it is impossible to avoid the question of false positives. Courts are institutions whose function is to accept or reject the hypothesis of a given claimant (e.g., “this patent is invalid”). These institutions are run by humans, who make mistakes. It is quite a surprise to have an economist announce that the potential for such mistakes must be ignored “for purposes of economic analysis.” What do they think a court decision striking down a patent is, but a ruling that the PTO’s issuance

f the patent was a false positive? What do they call a deci-

sion that is reversed or overruled—or a subsequent decision in a second, indistinguishable case that reaches the opposite conclusion about the patent? For years, the antitrust econo- mists have been the ones to point out the opportunities for “Type I” and “Type II” errors. Why these tools of analysis should be abandoned here is left unexplained. As a general scientific proposition, “[a] Type I error occurs when a false hypothesis is accepted, and a Type II error occurs when a true hypothesis is rejected.” 57 For this discussion, the hypothesis is G’s claim that the patent is invalid. If the patent is valid, the hypothesis is false; if G wins anyway, therefore, a Type I error (or false positive) has occurred. If the patent is invalid, but P still prevails, a Type II error (or false negative) has occurred (G’s true hypothesis was rejected). Since the hypothesis may also be accepted or rejected with-

ut any error, there are four separate outcomes to be consid-

ered if we are to measure patent settlements against the “possible” outcomes of patent litigation. We can sympathize with the refusal of these authors to consider false positives. The Lefflers, after all, are arguing for a per se rule condemning settlements with “reverse” payments in every case. It is essential, therefore, to ignore the procompetitive benefits to consumers that flow from an agreement not to infringe a valid patent. For the Lefflers, the possibility (however small) that the patent will be struck down must be treated as an unambiguous consumer bene-

fit. “The patent rules result in consumer benefits flowing

directly from the risk that the patent will be found invalid.” 58 If that result is not always beneficial to consumers, as when it represents a false positive, the argument gets com- plicated, even messy. That is not supposed to happen to per se theories. But they can’t have it both ways. Remember that the Lefflers’ argument applies to all patents (validity is irrelevant) because, they say, there is always some possibility that G will prevail. Well, why is that? Because no matter how strong your patent is, there is the possibility that the court will make, well, a mistake. As others have long recognized, it is always a “gamble . . . [to place] a technology case in the hands of a lay judge or jury.” 59 Some experts estimate that the “chances of prevailing in [patent] litigation rarely exceed sev- enty percent . . . even in that rare case with great prospects.” 60 When the Lefflers themselves describe the reasons a party may have for settling, they do not refer to the objective merit

f the patent, but to drawing a “pro-patent judge” or having

“a key witness” die.61 It is at best unseemly to rely upon a “risk” of patent invalidity that derives in large measure from institutional failure, and then assume that failure away by announcing that “there are no ‘wrong’ results generated by the patent litigation process.” 62 Allowing for false positives

SLIDE 7

A R T I C L E S A N D F E A T U R E S

7 4 · A N T I T R U S T

would require the recognition that the “risk” of patent invalidity may augur consumer harm as well as benefit, and the per se conclusion collapses. Nor will Shapiro’s analysis brook the introduction of false

positives. He uses the possibility of the patent’s defeat in lit-

igation to conclude that predicted consumer surplus result- ing from continued litigation is higher than the consumer surplus that would result from a so-called “ironclad” patent.63 If we determine the actual likelihood that the patent will win

r lose and take an “average” of their expected values, he

says that we can derive the proper amount of consumer surplus that a settlement must produce. If a settlement produces less, it is anticompetitive. There is much to be said about this approach but I will limit myself for now to this: There is no fault in constructing a model that produces predicted values of consumer surplus and a price that generates that surplus as long as you do not pretend that they describe real world phenomena. In other words, one can “average” the expected value of two uncertain outcomes in litigation—

ne can do the arithmetic—but one cannot make the answer
meaningful. One cannot conclude that such a number rep-

resents what would have occurred “had the two parties litigated.” 64 This is especially true when the outcomes being modeled are, as in our example, binary (either P wins and there is no generic entry, or G wins and the patent is gone). Simply put, you can take an average of the probability that a coin will land “heads” and the probability that it will land “tails,” but you cannot make the coin land on its side. Any conclusion that the result of his model produces real-world values that reliably measure anticompetitive effects awaits an economic justification that has not yet been attempted. Leaving these difficulties to one side, however, false positives render the model itself useless. Professor Shapiro must be able to conclude in all cases that more of this static consumer surplus is preferred to less, and that lower prices are preferred to higher. That is obviously a fair assumption in many antitrust cases—unless, and just as obviously, a valid patent is involved. In that case, consumers benefit from maximum incentives to innovation, which occur when the patentee charges the highest obtainable price, which produces the lowest obtainable consumer surplus. Thus, if we also attempt to model the probability that a valid patent is incorrectly

verturned (Type I error), or an invalid patent incorrectly

preserved (Type II error), the model will dutifully generate numbers, but what do they mean? We now have “good” high prices with “bad” high prices, and “bad” low prices with “good” low prices. Average that. You may believe that this talk of erroneous patent decisions is much ado about little. Perhaps the Type I and Type II errors just cancel each other out, and we can ignore them. That may well be a testable hypothesis, but to show that it cannot be accepted on its face, let us consider briefly how consumer interests may be affected by the four potential results of carrying the litigation to conclusion (set out in the accompanying chart).

“RESULTS” OF LITIGATION PATENT VALID PATENT INVALID

PATENTEE WINS

A. NO ERROR
B. TYPE II ERROR

GENERIC WINS

C. TYPE I ERROR
D. NO ERROR
A. Patent Valid/P Wins/No Error. If the patent is valid,

the patentee should win. The settlement removes this possibility, but also emulates its result if G concedes patent validity and infringement. Thus, the settlement makes consumers no worse off. One can argue that consumers would have been even better off if the patent had been upheld in court, because that would be a more effective deterrent to other

challengers. (Consumers do not benefit from meritless chal-

lenges to valid patents.) But an arguably better point is that the consumer interest in maximizing incentives to innovation is promoted by letting the innovator (P) decide what best maximizes the value of its valid patent, and P has chosen the settlement.

B. Patent Invalid/P Wins/Type II Error. Here, the patent

is invalid, but P wins anyway. As a result, the settlement causes no additional consumer harm by preserving the patent because the consumers would have been harmed in any event by the Type II error if the litigation had continued. One undeniable consumer benefit from the settlement preventing this Type II error is that it avoids placing a judicial impri- matur on a bad patent. Rather than be deterred by a judicial precedent that wrongly declares this patent valid, other challengers will be enticed by the settlement to see whether a challenge is feasible. If, as this example assumes, the patent is invalid, they will keep coming.

C. Patent Valid/G Wins/Type I Error. Here is the false
positive. A valid patent has been set aside. Forever. The exclu-

sive rights and monopoly profits deemed essential to spur innovation in the case of valid patents are eliminated. The rivalry that Congress has sought to exclude as harmful to consumers is now ensured. This may be especially harmful in the pharmaceutical industry, where the profits from a patented drug are quite literally paying for the development of the next innovative drug. Brand name drug companies are in the business of inventing new drugs, and the investments are

enormous. Where, as posited here, a valid patent is struck

down, the procompetitive incentive to innovate sought by Congress has been reduced.65 Even worse, this consumer harm is irrevocable because the current law of collateral estoppel will prevent the patentee from enforcing the patent against others.66 Unlike Type II error, therefore, Type I error in the patent system cannot be

corrected. While Judge Easterbrook was speaking of erro-

neous antitrust decisions when he said that “false positives are much more harmful than false negatives,” 67 the same obser- vation applies to patent decisions—and for the same reason. False negatives have the potential for self-correction; false positives do not. Accordingly, to the extent that a Hatch- Waxman settlement prevents a false positive, it provides an

SLIDE 8

S P R I N G 2 3 · 7 5

unambiguous consumer benefit; the settlement removes the

ne possibility of consumer harm that cannot be fixed.
D. Patent Invalid/G Wins/No Error. This is the key

result for those who attack Hatch-Waxman settlements. By preventing an invalid patent from being struck down, the settlement prevents additional rivalry that can benefit con-

sumers. This potential for consumer harm is tempered by two
factors. First, if the patent was obtained and prosecuted in

good faith, consumer interest in the integrity of the patent system may, in a given case, outweigh the loss of rivalry during the unexpired term of a specific patent. This is why the second Justice Harlan and others have urged that antitrust liability for agreements concerning invalid patents should be confined to cases in which patents were procured or prosecuted in bad faith.68 Second, as in the outcome involving Type II error, there is the opportunity for self-correction

here. The settlement will attract other applicants, who will see

(as we posit) that the patent is invalid. A patent is always subject to being invalidated by new challengers, no matter how

ften it has been upheld in other cases.

How do these four potential litigation results add up? By preventing results A and B, the effect of the settlement is somewhat procompetitive. By preventing result C (the false positive), the effect is unambiguously and significantly pro-

competitive. By preventing result D (exposing an invalid

patent), the effect is probably—assuming away the cavils mentioned above—anticompetitive. These are all only “possible” results, of course, and hence legally inadequate to base a conclusion of competitive harm. But even if we were to take seriously the notion of basing harm on the predicted outcome

f litigation, who could argue that the potential anticom-

petitive effect of preventing result D always outweighs the procompetitive effects of preventing results A, B, and (especially) C? That is what one must conclude to apply the per se rule under the theory of probabilistic patent rights, and the conclusion will not follow. Conclusion Let us learn from Butch and Sundance. They thought that a quick trip to South America was an easy solution to their immediate problem (Mr. Harriman’s posse), and so it seemed. Similarly, here, some see the theory of “probabilistic” patent rights as an easy evasion of the presumption of patent validity, as well as the venerated legal rules that refuse to equate the possibility of harm with harm itself. But the consequence of basing consumer harm on the potential outcome of a specific lawsuit is a thicket of analytical problems that defy evalu- ation—from false positives to risk aversion to the age and fal- libility of a given federal judge. And these problems are gratuitous. For in a proper antitrust analysis of Hatch-Waxman settlements, we should not attempt to measure the possible outcomes of a single lawsuit (which cannot rationally exclude false positives). We should conduct instead an orderly rule of reason analysis, which begins by measuring the patent. If the patent is valid and the exclusion of competition no broader than that inherent in the patent, then in the words of the Federal Circuit “that ends the inquiry.” 69 If there is reason to believe that the patent is invalid, then the antitrust analysis may begin, considering all relevant circumstances. This may seem like hard work, even dull. But if our goal is sensible antitrust policy (not to mention analytical rigor), we must avoid specious detours like “probabilistic” patent

rights. Otherwise, we will find, like Butch and Sundance, that

we have traded a mere posse for the entire Bolivian Army. As Butch would say: “That’s bad business.”

1 Frank H. Easterbrook, On Identifying Exclusionary Conduct, 61 NOTRE DAME

L. REV. 972, 977 (1986).

2 Kevin D. McDonald, Patent Settlements and Payments That Flow the “Wrong”

Way: The Early History of a Bad Idea, 15 ABA Section of Antitrust Law Healthcare Chronicle No. 4, Winter 2002, at 2 (Winter 2002) [hereinafter Wrong Way Payments].

3 Schering-Plough, FTC Docket No. 9297, (Initial Decision) (June 27, 2002). 4 Schering-Plough, slip op. at 104. Id. at 110–11 (rejecting the argument that

payments were unlawful because payments by patentee gave “incentive” to stay out of the market).

5 Daniel A. Crane, Exit Payments in Settlement of Patent Infringement Lawsuits:

Antitrust Rules and Economic Implications, 54 FLA. L. REV. 747, 775 (2002).

6 In re Tamoxifen Citrate Antitrust Litig., MDL No. 1408 (ILG), Memorandum

and Order at 11 (E.D.N.Y. Aug. 20, 2002).

7 Carl Shapiro, Antitrust Limits to Patent Settlements, original version (May

2001); updated version (August 2002), RAND J. ECON. (forthcoming), available at http://faculty.haas.berkeley.edu/shapiro/settle.pdf. My citations are to the original version of May 2001, which was presented at two con- ferences in 2001, and is the version cited by the Lefflers in their article. Keith

B. Leffler & Cristofer I. Leffler, Want to Pay a Competitor to Exit the Market?

Settle a Patent Infringement Case, ABA Section of Antitrust Law Economics Committee Newsletter, Spring 2002, Vol. 2, No. 1 [hereinafter Lefflers]. One or more of the Lefflers have worked for private plaintiffs claiming that patent settlements are anticompetitive.

8 Lefflers, supra note 7, at 31. 9 Shapiro, supra note 7, at 6–7 & 9. 10 Lefflers, supra note 7, at 31. 11 Officially known as “The Drug Price Competition and Patent Term Restoration

Act of 1984,” Pub. L. 98-417, 98 Stat. 1585 (1984), 21 U.S.C. § 355 (1994).

12 21 U.S.C. § 355(j)(2)(A)(vii)(I–IV). 13 Id. § 355(j)(4)(B)(iii). 14 Eli Lilly & Co. v. Medtronic, Inc., 496 U.S. 661, 676–78 (1990). 15 See Wrong Way Payments, supra note 2, at 8–9 (explaining how prior law

would not permit such a cost-free declaratory judgment on the validity of a patent).

16 Indeed, in the course of his article, Carl Shapiro demonstrates mathemati-

cally that, under one set of assumptions, no settlement is possible without a payment from P to G. Shapiro, supra note 7, at 19.

17 As in my prior article (note 2, supra), the scenario examined here has sev-

eral assumptions designed to test the logic of those who argue the irrele- vance of patent validity. This scenario assumes that the patent is broad enough to preclude all generic competition and that the settlement’s preclu- sive effect is no broader than the patent itself. That is to say, the settlement does not attempt to “extend” the patent by barring drugs that would not infringe a patent, as well as those drugs that do.

18 E.g., David A. Balto, Pharmaceutical Patent Settlements: The Antitrust Risks,

55 FOOD & DRUG L.J. 321, 335 (2000).

SLIDE 9

A R T I C L E S A N D F E A T U R E S

7 6 · A N T I T R U S T 19 12 HERBERT HOVENKAMP, ANTITRUST LAW ¶ 2040b at 199 (1999) (footnote

mitted).

20 Id. at 200. 21 Rubber Tire Wheel Co. v. Milwaukee Rubber Works Co., 154 F. 358, 364 (7th

Cir. 1907). See Monarch Marking Sys., Inc. v. Duncan Parking Motor Maint.

Co., 1988 WL 5038 at *5 (N.D. Ill. Jan. 19, 1988) (“Neither [antitrust plaintiffs] nor consumers have a right to the sale of labels which infringe Monarch’s patents”), partially vacated on other grounds, 1988 WL 23830 (N.D. Ill. Mar. 8, 1988).

22 Kevin J. Arquit, Patent Abuse and the Antitrust Laws, 59 ANTITRUST L.J. 739,

744 (1991) (emphasis added).

23 Thomas B. Leary, Antitrust Issues in Settlement of Pharmaceutical Patent

Disputes, ABA Section of Antitrust Law Healthcare Chronicle, Vol. 14, No. 4, Winter 2000/2001, at 6 (emphasis added) [hereafter Leary I].

24 Dennis W. Carlton, A General Analysis of Exclusionary Conduct and Refusal

to Deal—Why Aspen and Kodak Are Misguided, 68 ANTITRUST L.J. 659, 674 (2001).

25 Charles F. Rule, Patent-Antitrust Policy: Looking Back and Ahead, 59 ANTITRUST

L.J. 729, 730 (1991). Accord Atari Games Corp. v. Nintendo of Am., Inc., 897 F.2d 1572, 1576 (Fed. Cir. 1990).

26 Shapiro, supra note 7, at 10; Lefflers, supra note 7, at 27. 27 I say that settlements involving invalid patents “can be” anticompetitive to

make clear that invalidity alone is insufficient to conclude the antitrust

analysis. Justice Harlan made clear in his Walker Process concurrence that

the requirement of actual fraud by the patentee in that case was crucial because the antitrust laws do not and should not outlaw all “monopolies practiced under patents that for one reason or another may turn out to be voidable under one or more of the numerous technicalities attending the issuance of a patent.” Walker Process Equip., Inc. v. Food Mach. & Chem. Corp., 382 U.S. 172, 179 (1965) (Harlan, J. concurring). Others agree that invalidity alone is insufficient to establish liability without a showing that the patent was procured or prosecuted in bad faith: “[P]atent fraud is required . . . [to render unlawful] the use of an invalid patent to monopolize . . . a seg- ment of the market.” SSP Agricultural Equip., Inc. v. Orschand-Rite, Ltd., 592 F.2d 1096, 1103–04 (9th Cir. 1979) (emphasis added). See Nobelpharma AB v. Implant Innovations, Inc., 141 F.3d 1059, 1069 (Fed. Cir. 1998) (antitrust laws apply differently to “patents procured by ‘deliberate fraud’ [than to] those rendered invalid or unenforceable for other reasons”). My point is that, for a settlement that precludes no more competition than the patent itself, a showing of invalidity is essential to permit the antitrust analysis to commence.

28 Joseph F. Brodley & Maureen A. O’Rourke, Patent Settlement Agreements,

ANTITRUST, Summer 2002, at 53.

29 Thomas B. Leary, Antitrust Issues in the Settlement of Pharmaceutical

Patent Disputes, Part II at IV, ¶ 5, ABA Section of Antitrust Law Healthcare Program (May 17, 2001), available at http://www.ftc.gov/speeches/ leary.htm at 1.

30 Leary I, supra note 23, at 2. 31 Shapiro, supra note 7, at 6. 32 Id. at 9. 33 Lefflers, supra note 7, at 29. 34 Shapiro, supra note 7, at 9. 35 Lefflers, supra note 7, at 29. 36 Roper Corp. v. Litton Sys., Inc., 757 F.2d 1266, 1270 (Fed. Cir. 1985)

(emphasis added).

37 Schering-Plough, slip op. at 103 (citing Doddridge v. Thompson, 22 U.S. 469,

483 (1824)). Id. at 104.

38 Shapiro, supra note 7, at 11. 39 Biovail Corp. Int’l v. Hoechst AG, 49 F. Supp. 2d 750, 766 (D.N.J. 1999)

(emphasis added).

40 Zeneca Ltd. v. Mylan Pharms., Inc., 173 F.3d 829, 832–33 (Fed. Cir. 1999).

See cases cited supra note 21.

41 3 JOHN G. MILLS III ET AL., PATENT LAW FUNDAMENTALS § 18:4 at 18–29 (2d

ed. 2002); 3 PETER D. ROSENBERG, PATENT LAW

AND FUNDAMENTALS,

§ 17.02 at 17–29 (2d ed. 2001).

42 Shapiro, supra note 7, at 10. 43 Associated Gen. Contractors v. Cal. State Council of Carpenters, 459 U.S.

519, 534 n.30 (1983).

44 RESTATEMENT (SECOND) OF TORTS § 433B cmt. 1 (1965). 45 Gideon v. Johns-Manville Sales Corp., 761 F.2d 1129, 1137 (5th Cir. 1985). 46 Youst v. Longo, 729 P.2d 728, 741 (Cal. 1987). 47 Schering-Plough, slip op. at 74–75 (¶ 393). See In re Tamoxifen Citrate

Antitrust Litig., supra note 6, Slip. op. at 9.

48 Lefflers, supra note 7, at 30. 49 Id. at 33; see id. at 29. 50 Id. at 30 & 33. 51 H.R. REP. No. 98-857 pt. 1, at 28 (1984), reprinted in 1984 U.S.C.C.A.N.

2647, 2661; see id. (patents will have “same statutory presumption of validity as is afforded under current law”).

52 Shapiro, supra note 7, at 3. 53 Id. at 12. 54 Id. at 13. 55 Id. at 40. In the proceedings before the ALJ in Schering Plough, the FTC

staff’s expert conceded that his model did not account for risk aversion and that risk aversion could produce settlements unrelated to his calculation of expected litigation outcomes. See Appeal Brief of Respondent Schering- Plough Corporation at 62, In the Matter of Schering-Plough Corp. et al. (Sept. 30, 2002), FTC Docket No. 9297 (Public Version).

56 Shapiro, supra note 7, at 15. 57 JOHN ALLEN PAULOS, INNUMERACY: MATHEMATICAL ILLITERACY

AND ITS

CONSEQUENCES 108 (1988).

58 Lefflers, supra note 7, at 32. 59 Seymour E. Hollander, Why ADR May Be Superior in Patent Disputes, 2 No.

1 Intell. Prop. Strategist 1, at *6 (1995).

60 Steven Z. Szczepanski, Licensing or Settlement: Deferring the Fight to Another

Day, 15 AIPLA Q.J. 298, 300–01 (1987).

61 Lefflers, supra note 7, at 30. 62 Id. at 31. 63 Shapiro, supra note 7, at 10. 64 Id. 65 Some might argue that the false positive did not really hurt consumers,

because this invention, covered by this patent, was in fact produced and will be enjoyed by all. The false positive, in this view, is only a “gotcha” for the patentee; consumers get a windfall. There are two responses. First, putting aside the effect on incentives for future innovation, the argument does not necessarily work even for this invention. The loss of a patent causes the patent holder to pull back its distribution and marketing efforts severely because the new competitors will simply get a free ride. The former patentee may invest its resources in other areas, and output can actually fall. See, e.g., Edmund W. Kitch, The Nature and Function of the Patent System, 20 J.

OF LAW & ECON. 277 (1977). Second, and of far greater importance, this

argument depends on an entirely static view of the incentive to innovate. Yesterday’s patented invention provides the profit and incentive for tomor- row’s patented invention—often, as in the drug industry, by the same inven-

tor. We cannot rationally embrace the promotion of innovation in our patent

system—as both Shapiro and the Lefflers purport to do—and then suggest that consumers are unhurt when a patent is erroneously struck down. Before we applaud Type I error, therefore, we should recall—as Justice Black

nce reminded us—“the possible relevancy of this ancient remark: ‘Another

such victory and I am undone.’” Beauharnais v. Illinois, 343 U.S. 250, 275 (1952).

66 See, e.g., Blonder-Tongue Labs., Inc., v. University of Illinois Found., 402 U.S.

313 (1971); Stevenson v. Sears, Roebuck & Co., 713 F.2d 705 (Fed. Cir. 1983).

67 See supra note 1. 68 See supra note 27. 69 Mallincrodt, Inc. v. Medipart, Inc., 976 F.2d 700, 708 (Fed. Cir. 1992).