Less Restrictive Alternatives in Antitrust Law

C. Scott Hemphill* Antitrust courts often confront "mixed" conduct that has two contrasting effects, one harmful and the other beneficial. For example, a nationwide agreement not to pay college football players harms the players while benefiting fans of amateur sports. An important tool for analyzing mixed conduct is to compare the action to a hypothesized alternative and to ask whether the alternative action is "less restrictive" and hence less harmful. The less restrictive alternative (LRA) test is used widely, from the rule of reason to mergers to monopolization. The test often assumes a particular, narrow form, that the alternative must be dominant: not only less restrictive but also equally effective. In other words, could the benefits have been achieved equally well with less harm?

This Article offers a new account of the LRA test that draws inspiration from constitutional law and other fields. Dominant LRAs offer a shortcut that avoids the difficult tradeoff between increased benefit and increased harm. However, most LRAs are less effective rather than dominant. Such an alternative offers a basis for condemning conduct when the alternative is preferable on balance to the conduct. Balancing in antitrust is not a myth as many believe; instead, a tradeoff of incremental benefit and harm occurs in the assessment of LRAs. The LRA test serves the further function of "smoking out" an inference of anticompetitive effect.

As the Article shows, courts that restrict their analysis to dominant LRAs run a high risk of false negatives, particularly when they also

Antitrust courts often confront conduct that has two contrasting effects, one harmful and the other arguably beneficial. College football teams agree not to compete for players by paying them (harmful), a decision that also increases the popularity of the sport (beneficial). A performing-rights organization limits price competition among songwriters (harmful ) in the course of assembling the authors' rights into a blanket license that is valued by radio stations (beneficial). Two groups of doctors merge, allowing them to raise prices (harmful ) while also improving the quality of patient care (beneficial ).

To handle mixed conduct, courts have fashioned two common law tools. The first is to calculate the net effect of the conduct. In principle, a court totals up the benefits and harms to see which is larger. Courts and agencies frequently declare that identifying net effects is the heart of 2016] LESS RESTRICTIVE ALTERNATIVES 929 modern antitrust. In practice, however, courts seldom reach this balancing step. As a consequence, balancing in antitrust is regarded as rare, even a myth. 1 A second tool for handling mixed conduct is to compare the conduct to a hypothesized alternative and ask whether the alternative action is less harmful in the particular sense that it is "less restrictive." If so, then the defendant loses. Courts and agencies apply this less restrictive alternative (LRA) test widely, from agreements in restraint of trade to monopolization to mergers. 2 The LRA test employed in antitrust often takes a strikingly narrow form. The alternative must be dominant: not only less restrictive but also equally effective. In other words, could the good have been achieved equally well with less bad? 3 The LRA test is relevant wherever mixed conduct appears, from sports leagues to health care to search engines. Despite its importance, the test has received little systematic or sustained treatment across disparate doctrinal areas and industries. Nor has the distinctive role of dominant LRAs been recognized or examined. This Article aims to fill that gap by explaining the functions of the LRA test and identifying ways to improve its operation. Along the way, it situates the LRA test within the leading doctrinal frameworks for evaluating mixed conduct, such as ancillarity and the "no economic sense" test, and identifies errors in important recent applications of the LRA test.

In many instances, the test functions as a shortcut, allowing courts to avoid the balancing exercise that would otherwise be required in a search for net effects. That avoidance results from a widespread sense that balancing is hard. Courts have only a limited capacity to assess and compare the pros and cons. Quantification is difficult given that parties deny the existence of a tradeoff rather than providing guidance about making the tradeoff. Quantification is particularly complex when an effect on innovation is asserted. Thorny issues arise when the individuals harmed are different from the individuals (or firms) that benefit. The result is a judicial anxiety about balancing. 4 The LRA shortcut relieves that anxiety by permitting the court to answer a different question instead. When an LRA dominates the defendant's conduct along both dimensions of interest-that is, when the alternative is both less restrictive and equally effective-the conduct is worse without recourse to any tradeoff. In such cases, the court can condemn the conduct without further ado. 5 Usually, however, the only available alternative is less effective. Such an LRA may be superior to the defendant's conduct, but only on balance. Balanced LRAs are a valid basis for condemnation when the incremental positive effects, compared to the defendant's conduct, outweigh the incremental negative effects. Antitrust courts sometimes rely on balanced LRAs when they condemn the defendant's conduct. Balancing is not a myth after all, as courts trade off incremental benefit and incremental harm within the LRA test. 6 Courts that, in applying the LRA test, condemn in light of a balanced alternative, generally do not advertise that fact. They are engaged in balancing in disguise. Bringing balancing in the LRA into the open forces antitrust courts to confront these difficult tradeoffs.

Used as a shortcut or a locus of balancing, the LRA test serves as a benchmark that the defendant's conduct must clear. A third, alternative function for the LRA test is to serve as a diagnostic tool rather than as a benchmark. 7 The LRA test is used to "smoke out" anticompetitive effects. Here, another neglected distinction comes into view-the difference between LRAs that are equally procompetitive and LRAs that are equally profitable. Only LRAs that afford equal profits to the defendant, considering all legitimate sources of profit, provide a valid inference of anticompetitive effect. This inference is particularly useful when the court's analytical mode is to avoid balancing by choosing between two contrasting accounts of the conduct rather than calculating a net effect.

Many legal fields employ a version of the LRA test, 8 from torts to trade to deceptive advertising. The analysis herein draws particular inspiration from constitutional law, a field that shares with antitrust both the anxiety about balancing and the use of dominant LRAs as a defensive response. The analogy also reveals an important gap in the antitrust toolkit-the absence of an explicit inquiry into more beneficial alternatives.

The "narrow tailoring" analysis of constitutional law tests state action not only for an LRA, but also for underinclusiveness. The latter inquiry condemns conduct in light of an alternative that better serves the defendant's asserted goal. 9 Such a hypothetical is the flip side of the LRA test. The point is to smoke out pretext and thereby "diminish the credibility of the [defendant's] rationale." 10 A similar analysis is valuable in antitrust. For example, if a defendant defends a vertical contract on 6. See infra section II.C (describing and defending use of less effective LRAs within balancing analysis).

7. See infra section II.D (analyzing use of test as diagnostic tool). 8. See infra sections I.B-.C (discussing use of LRA test in tort, constitutional, and administrative law). 9. See, e.g., City of Cleburne v. Cleburne Living Ctr., 473 U.S. 432, 450 (1985) (rejecting justification for discriminatory zoning ordinance, namely "avoiding concentration of population . . . and lessening congestion," given extreme underinclusiveness of ordinance).

10. City of Ladue v. Gilleo, 512 U.S. 43, 52 (1994).

LESS RESTRICTIVE ALTERNATIVES 931 the ground that it confers desirable incentives on a dealer, a plaintiff might establish that a different contract would more powerfully serve the asserted goal. The existence of such an alternative raises the inference that the asserted justification is absent. This inference, a more powerful form of smoking out, is not available using the LRA test alone. An antitrust court that seeks to uncover pretext needs this additional tool. 11 Used as a benchmark, the LRA test interacts with the net-effects test in two important ways. First, the LRA test augments the net-effects analysis, condemning some conduct that a net-effects test would permit. The Article explains why this expanded range of liability is desirable, contrary to the concerns of some critics. 12 Second, the LRA test helps fill the gap in jurisdictions that omit any net-effects inquiry. Such a truncated analysis threatens to insulate conduct from effective antitrust review, particularly when the anticompetitive effect is large and the procompetitive effect is small. Here, the LRA test reduces the substantial risk of false negatives. Moreover, using balanced LRAs-not just dominant LRAs-does a better job in filling the gap. 13 An LRA test, particularly in its more expansive form, poses a risk of false positives. A final contribution of this Article is to propose best practices that reduce that risk. 14 First, plaintiffs properly bear the burden of persuasion in establishing an LRA, including for categories of conduct (such as mergers and tying) in which the burden traditionally has been borne by defendants. Second, an identified LRA must be profitable to the defendant; otherwise condemnation can have a perverse effect on corporate conduct. Third, the LRA must be practical and "standard," not speculative.

The Article proceeds in four parts. Part I presents the problem of mixed conduct and antitrust's reliance on LRAs, particularly its reliance on dominant alternatives. Part II examines the functions of the LRA test as a shortcut, a locus of balancing, and a method of smoking out anticompetitive effect. Part III assesses LRAs as a supplement to or substitute for net-effects balancing. Part IV outlines best practices in implementing the LRA test.

In 2009, Ed O'Bannon, a former UCLA basketball player, filed an antitrust suit against the National Collegiate Athletic Association 11. See infra section II.D.3 (assessing absence of explicit underinclusion analysis in antitrust).

12. See infra section III.A (defending augmentation). 13. See infra section III.B, which discusses the important recent example of O'Bannon v. NCAA, 802 F.3d 1049 (9th Cir. 2015).

14. See infra Part IV (describing best practices). [Vol. 116:927 (NCAA). 15 The suit challenged an NCAA rule that limits the amount of compensation offered to student-athletes. The rule capped scholarships at a level several thousand dollars below the full cost of attendance and prohibited additional cash compensation. 16 O'Bannon sought payment for the use of his likeness by videogame makers and other licensees. 17 Other suits have challenged the ban on payments to play the sport. 18 In September 2015, the Ninth Circuit affirmed-but also sharply narrowed-the district court's conclusion that the rule restrained trade. The court concluded that the rule eliminates price competition among the schools, depriving players of an opportunity for compensation. 19 At the same time, the court accepted the NCAA's proffered justification that denying payment preserves amateurism and thereby increases popular interest in college sports. 20 The court was therefore faced with mixed conduct that has both positive and negative effects and the challenge of reconciling the two.

Mixed conduct arises in the full range of antitrust cases. One common fact pattern is a joint venture of "horizontal" competitors, such as the NCAA, that limits competition among its members in the course of producing a new good that could not be produced by any member acting alone. 21 Horizontal mergers routinely package a threatened loss of competition together with certain efficiencies. 22 "Vertical" contracts 15. O'Bannon, 802 F.3d 1049. 16. Id. at 1054. 17. Id. at 1055 (noting prohibition on payment for athlete's name, image, or likeness).

18. See Complaint at 3-4, Hartman v. NCAA, No. 3:15-cv-00178 (N.D. Cal. Jan. 13, 2015) (challenging cap in women's basketball); Complaint & Jury Demand-Class Action Seeking Injunction and Individual Damages at 3, Jenkins v. NCAA, No. 14-1678 (D.N.J. Mar. 17, 2014), 2014 WL 1008526 (challenging cap in men's basketball and football). In practice, the rule may bind only as to football and basketball, the two most profitable college sports, and as to a subset of elite schools. See O'Bannon, 802 F.3d at 1056 n.4 (explaining financial basis for limitation of O'Bannon suit to football and men's basketball).

19. The court understood the agreement alternatively as a cartel of sellers of educational services or buyers of labor. See O'Bannon, 802 F.3d at 1071 n.14. 20. Id. at 1072-73 ("[T]he district court found, and the record supports that there is a concrete procompetitive effect in the NCAA's commitment to amateurism: namely, that the amateur nature of collegiate sports increases their appeal to consumers."). In the absence of "meaningful argument" on the issue by the NCAA, the court also accepted a second justification: that to some degree, the rule facilitates a scholar-athlete's successful integration within the academic community. Id. at 1072. The court also accepted the district court's findings that the rules fail to "promote competitive balance" or "increase output in the college education market." Id. at 1072. 21. See, e.g., id. at 1069 (noting, as to certain products, "' restraints on competition are essential if the product is to be available at all'" (quoting NCAA v. Bd. of Regents of Univ. of Okla., 468 U.S. 85, 101, 102 (1984))).

22. See, e.g., United States v. Phila. Nat'l Bank, 374 U.S. 321, 332-33 (1963) (noting, with respect to challenged bank merger, regulator's conclusion that "over-all effect upon between manufacturers and retailers can limit competition while also serving a desirable end, such as assuring quality or providing incentives to retailers. 23 Conduct by a monopolist, such as the assembly of a software system from multiple components, may impede a rival's competitive opportunities, while simultaneously improving the product. 24 In short, mixed conduct is everywhere in antitrust. Table 1 presents a set of illustrative examples drawn from antitrust litigation.

For conduct to have a mixed effect, the negative and positive elements must be bound together in some fashion. Sometimes the reduced competition gives rise to the benefit, as with the idea that limiting payments to players is needed to preserve the NCAA's stature with fans. 25 In other instances, reduced competition is a side effect of the benefit, as with the reduced price competition among songwriters that results from a performing-rights organization's blanket license. 26 Mixed conduct pairs together a wide variety of negative and positive economic effects. The negative effects can pertain to price, output, production costs, or innovation. The benefits cover the same broad territory. A typical benefit is a new or improved product, with improvement defined broadly to include service and product information. competition would not be unfavorable" due to certain benefits of transaction); id. at 370-71 (considering and rejecting asserted efficiencies).

23. See, e.g., Leegin Creative Leather Prods. v. PSKS, Inc., 551 U.S. 877, 889-90 (2007) (describing procompetitive justifications for such agreements); id. at 892-94 (describing anticompetitive effects); Victor P. Goldberg, The Free Rider Problem, Imperfect Pricing, and the Economics of Retailing Services, 79 Nw. U. L. Rev. 736, 738 (1984) (describing "rich array of plausible rationales for adopting vertical restrictions"); see also Cont'l T.V., Inc. v. GTE Sylvania Inc., 433 U.S. 36, 51-52 (1977) ("The market impact of vertical restrictions is complex because of their potential for a simultaneous reduction of intrabrand competition and stimulation of interbrand competition." (footnote omitted)).

24. [Vol. 116:927 Another is reduced production cost, which may be passed along to purchasers. 27 In response to mixed conduct, courts often declare that their end goal is to identify the "net competitive effect" of the conduct. 28 The idea is to add up the beneficial effects of the conduct and subtract ("net") the harms from reduced competition. Canonical statements of doctrine for horizontal and vertical agreements, 29 mergers, 30 and monopolization 31 all culminate in a determination of net effects. This is a cost-benefit analysis that explicitly balances, or trades off, the incremental harms and incremental benefits of the conduct compared to a world without the challenged conduct. For example, the O'Bannon court concluded that the NCAA rule harms athletes but also yields the benefit of a more popular form of entertainment. 32 A cost-benefit approach would assess 27. For an extensive discussion of various benefits, see Rebecca Haw Allensworth, The Commensurability Myth in Antitrust, 69 Vand. L. Rev. 1, 18-22 (2016).

28. See, e.g., Cal. Dental Ass'n v. FTC, 526 U.S. 756, 771 (1999) (treating goal of analysis as identification of "net procompetitive effect").

29. See, e.g., id. (raising prospect that defendant's conduct had "net procompetitive effect"); id. at 774 (doubting conduct had "net anticompetitive effect"); id. at 782 (Breyer, J., dissenting) (emphasizing, for conduct with anticompetitive effect, analysis of "offsetting" justifications); Eastman Kodak Co. v. Image Tech. Servs., 504 U.S. 451, 486 (1992) (leaving as question on remand whether procompetitive effects "outweighed" anticompetitive effects); Atl. Richfield Co. v. USA Petroleum Co., 495 U.S. 328, 342 (1990) (framing core question of unreasonableness as "whether [restraint's] anticompetitive effects outweigh its procompetitive effects"); FTC v. Ind. Fed'n of Dentists, 476 U.S. 447, 459 (1986) (condemning conduct given absence of "countervailing procompetitive virtue"); Cont'l T.V., Inc. v. GTE Sylvania Inc., 433 U.S. 36, 57 n.27 (1977) (emphasizing "balancing intrabrand and interbrand competitive effects" is part of judicial function), remanded to 461 F. Supp. 1046, 1052 (N.D. Cal. 1978) (concluding vertical restraint had "overall procompetitive effect"); Interface Grp. v. Mass. Port Auth., 816 F.2d 9, 10 (1st Cir. 1987) (Breyer, J.) (finding conduct unreasonable when its "anticompetitive consequences . . . outweigh its legitimate business purposes"); U.S. Dep't of Justice & FTC, Antitrust Guidelines for Collaborations Among Competitors § 3.

which effect is more important or equivalently, whether the net effect is negative.

The net-effects test weighs the conduct against a baseline world without the conduct. If the outcome of interest (e.g., price) would be unchanged over time absent the conduct, then the test boils down to a simple before-and-after comparison. For example, did prices go up after the merger? In other cases, the baseline no-conduct world is itself changing. Supply or demand may shift over time, or competitive entry may be expected absent the conduct. If the baseline changes over time in important ways, then the before-and-after comparison must be adjusted or discarded in favor of a more sophisticated analysis.

Not all mixed conduct reaches this end stage of the analysis. One filter simplifies the calculation by removing some benefits from the calculus. For example, a benefit premised on the undesirability of competition itself is considered not cognizable. 33 Thus, engineers cannot refrain from price competition on the ground that competition will result in shoddy bridges. 34 However, the boundaries of this rule are porous in practice, as courts often take an expansive view of what counts as supporting competition. 35 Another source of simplification concerns benefits retained by producers, such as a reduction in production cost that is kept rather than shared with consumers. Benefits retained by producers are often considered not cognizable, but this conclusion is the subject of scholarly debate 36 and uncertainty in the case law. 37 Overall, a great deal of mixed 33. See Nat'l Soc'y of Prof'l Eng'rs v. United States, 435 U.S. 679, 696 (1978) ("[T]he Rule of Reason does not support a defense based on the assumption that competition itself is unreasonable."); see also Horizontal Merger Guidelines, supra note 30, § 10, at (excluding efficiencies that "arise from anticompetitive reductions in output or service"). A variant of this approach is to argue that innovation requires higher profits, either as a general matter (because it increases incentives or makes available cash for research and development), or as a condition for introducing a particular product. An example of the latter is Apple's assertion that it would not have entered the ebooks market but for certain conduct challenged by the Justice Department. A second filter is to lump together the harms and benefits as a single overall effect. For example, a merger of two beer producers that reduces competition between them will place upward pressure on prices. At the same time, if the merger yields production efficiencies, the resulting decrease in marginal costs would have the opposite tendency. These two forces can be conceptualized and measured as a single effect. A vertical contract that increases both price and quality similarly might be addressed by measuring the overall effect on output or quality-adjusted prices.

This filter, like the first, lets through a lot of mixed conduct. Often the winners and losers are different economic actors, as with an NCAA rule that harms players but benefits fans, or an airline merger that raises prices in some city pairs and lowers them in others. 38 If retained producer benefits count in the balance, a further example is a restraint that harms buyers but benefits producers. 39 In all these settings there are two distinct effects. Moreover, even when the benefits and harms are experienced by the same individual or firm, the overall effect is often calculated by comparing two components that are conceptually separable and separately estimated. The analysis of unilateral effects in merger cases offers an important illustration of this point. 39. This would include, for example, a merger to monopoly that reduced production costs without passing the savings on to consumers. 40. See, e.g., Carl Shapiro, The 2010 Horizontal Merger Guidelines: From Hedgehog to Fox in Forty Years, 77 Antitrust L.J. 49, 75-76 (2010) (identifying level of gross upward pricing pressure that exactly balances downward pressure from marginal cost savings). Doctrinally, the defendant's assertion that prices will not rise because of downward pricing pressure might be framed instead as a negation of the plaintiff's initial case, rather than as an assertion of mixed conduct.

To handle mixed conduct, courts employ a further tool beyond net effects and the filters discussed above. They ask whether an alternative exists that serves the same beneficial goal with less anticompetitive effect. For example, can the NCAA preserve the benefits of amateurism in college sports with less harm to competition for players?

LRAs take various forms. Mergers give way to more limited contractual relationships. Exclusive agreements are replaced by nonexclusive agreements. A vertical contract that indirectly supplies desirable incentives is swapped for an alternative that directly specifies the behavior. In O'Bannon, for example, the court condemned the NCAA rule in light of the LRA of a slightly larger scholarship that covers the full cost of attendance. 41 The LRA inquiry fits a common analytical pattern-that the restraint goes too far compared to its justification. The issue is sometimes described as overinclusiveness. More informally, courts speak of swinging a sledgehammer to crack a nut, firing a cannon to shoot a sparrow, 42 or "burn[ing] the house to roast the pig." 43 The alternative might be superior because it harms fewer consumers or because it harms them all to a lesser degree. Similar moves appear in other legal fields that confront mixed conduct, most famously in constitutional law, which scrutinizes a justification for state action in light of alternative means of achieving the goal. 44 The LRA test goes beyond the net-effects test by introducing an additional alternative to the analysis. The net-effects test evaluates the conduct in comparison to a world without the conduct at issue. By contrast, the LRA test considers a different possible action instead of the conduct. 45 An examination of unchosen alternatives-both their pros and cons-is a standard element of cost-benefit analysis. For example, a thorough evaluation of alternatives is required by official guidance for the development and review of agency regulations. 46 45. This separation of the net-effects and LRA analyses confines the net-effects inquiry to a comparison of the conduct with a no-conduct baseline. Another approach is to include and hence subsume an LRA inquiry within the net-effects analysis. One implementation of such an approach is considered infra note 166.

46. E.g., Exec. Order No. 12,866, § 1(a), 3 C.F.R. 638 (1994), reprinted as amended in 5 U.S.C. § 601 app. at 802 (2012) (requiring evaluation of alternatives available to agency in developing regulations); Office of Mgmt. & Budget, Exec. Office of the President, Circular A-4, Regulatory Analysis 1-2 (2003) http://www.whitehouse.gov/ [Vol. 116:927 plays an important role in tort law, where precautions are evaluated in light of alternative, untaken precautions that simultaneously offer incremental public benefit and impose an incremental burden on the alleged tortfeasor. 47 The LRA test appears in a wide range of antitrust cases. It is part of the rule of reason, antitrust's most important test for illegality, which is used to assess most horizontal and vertical agreements. 48 The rule of reason probes the unreasonableness (and hence illegality) of a challenged restraint. The LRA test is part of judicial statements of the rule of reason, 49 jury instructions, 50 and special verdict forms. 51 Federal agencies rely on the test in making enforcement determinations. 52 These applications are inspired in part by Addyston Pipe, a famous old case that insisted that horizontal restraints must be both "ancillary" to a desirable purpose and no more restrictive than necessary to achieve the purpose. 53 sites/default/files/omb/assets/regulatory_matters_pdf/a-4.pdf [http://perma.cc/KG6K -4KNA] (providing guidance on evaluating "various alternatives that should be considered in developing regulations").

47. This approach is adopted, for The LRA test is important outside the rule of reason as well. Courts and agencies apply the test to horizontal mergers by insisting that a claimed justification (a so-called "efficiency") must be "merger-specific." 54 Monopolization cases also employ an LRA test, 55 as one might expect given the close kinship between monopolization analysis and the rule of reason. 56 Finally, LRAs play a major role in assessing tying, wherein a firm conditions the purchase of one product or service on the purchase of another. 57 Courts test the tie's claimed benefit against the presumed availability of an LRA. 58 ecause a committed team can be assembled without" merger and therefore "is not a mergerspecific efficiency"); Horizontal Merger Guidelines, supra note 30, § 10, at 30 ("The Agencies credit only those efficiencies . . . unlikely to be accomplished in the absence of either the proposed merger or another means having comparable anticompetitive effects.").

55. See, e.g., LePage's Inc. v. 3M, 324 F.3d 141, 167 (3d Cir. 2003) (en banc) (asking, as part of jury charge, whether defendant "has impaired competition, in an unnecessarily restrictive way"); id. (defining "exclusionary conduct and predatory conduct" as, in relevant part, conduct that "either does not further competition on the merits, or does so in an unnecessarily restrictive way"); id. at 178 (Greenberg, J., dissenting) (concluding justification should be rejected "if a valid asserted purpose would be served fully by less restrictive means"); Trans Sport, Inc. v. Starter Sportswear, Inc., 964 F.2d 186, 188-89 (2d Cir. 1992) (prohibiting conduct that does "not further competition on the merits or does so in an unnecessarily restrictive way"); New York v. Actavis, PLC, No. The present focus is the use of LRA analyses to determine liability, but parts of the analysis apply in other contexts as well. LRA analyses also arise, for example, when courts consider the "but-for" world in establishing causation for damages; the but-for world may reflect adoption of an LRA. Moreover, courts may impose an identified LRA as an injunctive remedy. 59 The Supreme Court has given no sustained attention to the LRA test, despite its extensive use by lower courts. Most strikingly, the Court has never endorsed (or rejected) the test. Several horizontal agreement cases 60 and one merger review 61 assess restraints by reference to LRAs, though without explication or development. With respect to vertical agreements, a concurrence by Justice Brennan strongly endorses the test, 62 while dicta in a later opinion of the Court might offer a more skeptical take. 63 ,J., concurring) (suggesting "less restrictive alternatives" inquiry and describing its application to case at hand); see also infra section II.D (discussing Justice Brennan's concurrence).

63. In Continental T.V., Inc. v. GTE Sylvania Inc., the Court noted that the restriction at issue "was neither the least nor the most restrictive provision that [defendant] could have used . . . . We are unable to perceive significant social gain from channeling transactions into one form or another." 433 U.S. 36, 58 n. 29 (1977). This dictum might be read as a rejection of the least restrictive means. An alternative interpretation is that if a particular restraint is permissible, per se liability for a second restraint with similar effects is inapt. Phillip Areeda, The "Rule of Reason" in Antitrust Analysis: General Issues 9 (1981).

Skepticism has also been read into Justice O'Connor's concurrence in Jefferson Parish, which criticized the lower court's imposition of liability at the LRA stage. Jefferson Par., 466 U.S. at 44 n.13 (O'Connor, J., concurring) ("In the absence of an adequate basis to expect any harm to competition from the tie-in, this objection [that there is an LRA] is simply 2016] LESS RESTRICTIVE ALTERNATIVES 941 passing. 64 Tying cases are an exception, as several opinions have rejected claimed justifications in light of presumed LRAs. 65 Lower courts have filled the vacuum, elaborating a burden-shifting framework that combines the LRA test with a net-effects test. In one typical formulation, the plaintiff offers evidence establishing the presence of an anticompetitive effect-for example, higher prices to purchasers. 66 In response, the defendant offers evidence of a cognizable benefit. 67 Next, the court considers plaintiff's presentation of evidence about an LRA. 68 If there is no LRA, the final step is to determine whether there is a net anticompetitive effect. 69 The LRA and net-effects steps are alternative ways to establish liability.

This "typical" formulation masks important disagreement among the lower courts. For example, some courts (including the O'Bannon court) omit the net-effects inquiry, making the demonstration of an LRA necessary for liability. 70 A second difference is the burden of persuasion. Some courts assign this burden to defendants, particularly in tying and irrelevant."). In context, Justice O'Connor was pointing out the lack of mixed effect, rather than objecting to the LRA test.

64. Aspen Skiing Co. v. Aspen Highlands Skiing Co., 472 U.S. 585, 605 (1985) (asking, as relevant question for liability, whether conduct "impaired competition in an unnecessarily restrictive way").

65. See supra note 58 (identifying cases). 66. E.g., Geneva Pharms. Tech. Corp. v. Barr Labs. Inc., 386 F.3d 485, 506-07 (2d Cir. 2004) (noting plaintiff's initial burden to demonstrate anticompetitive effect). The plaintiff might show that prices actually increased or instead that an increase is likely given the usual tendency of such conduct, combined with a showing that the plaintiff has market power.

67. See id. at 507. 68. See id. 69. See, e.g., id. ("Ultimately, the factfinder must engage in a careful weighing of the competitive effects of the agreement-both pro and con-to determine if the effects of the challenged restraint tend to promote or destroy competition."); County of Tuolumne v. Sonora Cmty. Hosp., 236 F.3d 1148, 1160 (9th Cir. 2001) (progressing to "balancing stage" after plaintiffs failed to establish LRA); Model Jury Instructions, supra note 50, at A-12 (Instruction 3D) (instructing, when no LRA exists, to "balance . . . competitive benefits against the competitive harm") Collaboration Guidelines, supra note 29, § 3.37, at 24-25 (if agreement is "reasonably necessary to achieve cognizable efficiencies," agencies proceed to analysis of "likelihood and magnitude" of harms and efficiencies to determine "overall actual or likely effect on competition"); see also Law v. NCAA, 134 F.3d 1010, 1019 (10th Cir. 1998) (describing steps discussed above, including LRA, but then proceeding to balancing "if these steps are met," raising puzzling prospect that plaintiff is forced to balance even if LRA exists); Einer Elhauge, United States Antitrust Law and Economics 141-42 (2d ed. 2011) [hereinafter Elhauge, Antitrust Law] (proposing more complex set of procedural steps distinguishing arguments made at dismissal and summary judgment stages). 70. See infra section III.B (examining role of LRA test if net-effects analysis is omitted). merger cases. 71 Other courts give the burden to plaintiffs. These disagreements are taken up in Parts III and IV.

Perhaps echoing its neglect by the Supreme Court, the LRA test has received limited critical attention. Several scholars have criticized the LRA analysis as too demanding in rule of reason 72 or monopolization 73 cases. A larger literature considers the test in a more limited or scattered fashion, often in the course of analyzing a particular type of antitrust claim 74 or specific matter. 75 71. See infra section IV.A (arguing plaintiffs properly bear burden of persuasion in establishing LRA in rule of reason, merger, and tying cases).

To say that an alternative is less restrictive is only half of the LRA analysis. A further question is effectiveness-how well the alternative serves the procompetitive end. Some alternatives are dominant in the sense indicated in the Introduction: They are not only less restrictive, but also equally (or more) effective. Other alternatives are not only less restrictive, but also less effective. They may or may not be superior, on balance.

This distinction is illustrated by Figure 1, which depicts the conduct (point A) and an alternative action (point Z) along the dimensions of justification and restriction. Here, A has equal or lesser benefit and greater restriction compared to Z. Thus, the existence of the dominant alternative Z provides a basis for condemning A. The same is true for any conduct to the southwest of Z. By contrast, consider Z'. Z' is less effective than A and hence fails as a dominant alternative. 76 FIGURE 1: DOMINANT ALTERNATIVES Condemning conduct in light of a dominant alternative is a form of cost-benefit analysis just as much as the net-effects inquiry. Although costbenefit comparisons often require courts to confront a tradeoff, this is not always the case. If one action, compared to another, has greater or 76. Z' may or may not be superior to A on balance, depending upon the degree to which it is less restrictive. equal benefit and also imposes a lesser burden on competition, it is decisively better. Such alternatives offer a free lunch that we may choose without regret. These are the easy cases, in which cost-benefit analysis can be performed without needing to explore any tradeoff.

Antitrust courts and agencies often limit the LRA analysis to dominant alternatives. In O'Bannon, the equal-effectiveness limitation had determinative force. 77 The court accepted the slightly larger aid package as an equally effective LRA, while rejecting a second LRA as less effective. 78 Equal effectiveness is an explicit limitation in cases, 79 jury 77. See O'Bannon v. NCAA, 802 F.3d 1049, 1064 (9th Cir. 2015) ("[A] restraint that serves a procompetitive purpose can still be invalid under the Rule of Reason if a substantially less restrictive rule would further the same objectives equally well." (emphasis added)); id. at 1074, 1076-78 (requiring "virtually as effective" LRA); id. at 1075 (limiting analysis to following proposition: "where, as here, a restraint is patently and inexplicably stricter than is necessary to accomplish all of its procompetitive objectives, an antitrust court can and should invalidate it and order it replaced with a less restrictive alternative" (emphasis omitted)); id. at 1076 ("We cannot agree that a rule permitting schools to pay students pure cash compensation and a rule forbidding them from paying . . . compensation are both equally effective in promoting amateurism and preserving consumer demand.").

78. Id. at 1076 (rejecting deferred-compensation LRA as less effective). Similarly, the dissent accepted this limitation, arguing that the deferred-compensation LRA was in fact equally effective. Id. at 1081 (Thomas, C.J., concurring in part and dissenting in part) (framing relevant question as whether LRA is "' virtually as effective' in preserving popular demand for college sports," as opposed to amateurism per se (quoting id. at 1076 (majority opinion))).

79. See, e.g., United States v. Apple, Inc., 791 F.3d 290, 350 (2d Cir. 2015) (Jacobs, J., dissenting) (crediting justification of procompetitive entry given that "[n]obody has proposed . . . any 'less restrictive means' by which Apple could have achieved the same competitive benefits"); Geneva Pharms. Tech. Corp. v. Barr Labs. Inc., 386 F.3d 485, 507 (2d Cir. 2004) (noting at LRA step, plaintiffs must prove "any legitimate competitive benefits . . . could have been achieved" through LRA); Virgin Atl. Airways Ltd. v. British Airways PLC, 257 F.3d 256, 265 (2d Cir. 2001) (requiring, at LRA step, plaintiff must show "same procompetitive effect could be achieved through alternative means"); County of Tuolumne v. Sonora Cmty. Hosp., 236 F.3d 1148, 1159 (9th Cir. 2001) (requiring plaintiffs show alternative is "virtually as effective" in serving defendant's objective, concluding proposed LRAs were less effective, and ruling in favor of defendants at final net-effects step); Sullivan v. NFL, 34 F.3d 1091, 1103 (1st Cir. 1994) (stating "restriction is not reasonable" if "less restrictive alternative . . . exists that would provide the same benefits"). Some cases impose liability in light of a dominant alternative, without specifying whether a less effective alternative would have sufficed. See

LESS RESTRICTIVE ALTERNATIVES 945 instructions, 80 and commentary. 81 The limitation to dominant alternatives is not an instruction of the Supreme Court, which is hardly surprising given the Court's inattention to the LRA test altogether. Nor is the limitation uniformly applied by lower courts. A somewhat broader view recognizes alternatives that confer "similar" or "comparable" benefits. 82 As a logical matter, there is a further, more expansive possibility, which is to recognize alternatives that are less effective in serving the claimed aim. As developed in section II.C, less effective alternatives might be preferred to the challenged conduct, provided that the lesser restrictiveness outweighs the lesser effectiveness.

Many cases are effectively silent on the degree of effectiveness needed to establish an LRA. 83 In particular, some courts employ the vague formulation that the restraint must be "reasonably necessary" to achieve the benefit. 84 That phrase can be read to implement any of the positions above. Perhaps a wide range of dominant and nondominant LRAs may be used to demonstrate that the restraint was not reasonably necessary to achieve the benefit. 85 Or perhaps only LRAs with similar or comparable benefits may be used. 86 Or more narrowly, perhaps only 80. See, e.g., Model Jury Instructions, supra note 50, at A-10 (Instruction 3C) ("If the plaintiff proves that the same benefits could have been readily achieved by other, reasonably available alternative means . . . then they cannot be used to justify the restraint.").

81. See, e.g., Elhauge, Antitrust Law, supra note 69, at 142 (framing analysis as restricted to LRAs that "equally achieve . . . procompetitive effects"); Herbert Hovenkamp, Federal Antitrust Policy: The Law of Competition and Its Practice 280 (4th ed. 2011) (limiting LRA analysis to alternatives that achieve "same efficiencies").

82. See, e.g., United States v. Brown Univ., 5 F.3d 658, 679 (3d Cir. 1993) (considering, at LRA step, whether "comparable benefits could be achieved through a substantially less restrictive alternative"); 11 Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law ¶ 1912i, at 371-72 (3d ed. 2011) (asking, at LRA step, whether "same (or nearly the same) procompetitive benefits could be achieved by" LRA); Lawrence A. Sullivan & Warren S. Grimes, The Law of Antitrust: An Integrated Handbook 223 (1st ed. 2000) (criticizing opinion in which court failed to consider whether LRA would have been "nearly as effective" in achieving claimed end); Collaboration Guidelines, supra note 29, § 3.36(b), at 24 (considering, at LRA step, whether "participants . . . could achieve similar efficiencies by practical, significantly less restrictive means"); see also Meese, Price Theory, supra note 74, at 111 n.172 (collecting commentary).

83. For example, in Law v. NCAA, the court considered whether the defendant's "objectives can be achieved in a substantially less restrictive manner." 134 F.3d 1010, 1019 (10th Cir. 1998). That formulation leaves open whether the objectives must be achieved to the same degree or may be achieved to a lesser degree.

84. See, e.g., Brown Univ., 5 F.3d at 669 (formulating standard as "reasonably necessary to achieve the stated objective").

85. See, e.g., 7 Areeda & Hovenkamp, supra note 82, ¶ 1505b, at 419 (equating "reasonably necessary" test with absence of LRA); id. (reading test to require "discriminating judgment about the allegedly less restrictive alternative: how much worse for the parties or how much better for society," apparently accepting liability in light of LRA that is "worse for the parties").

86. See, e.g., Brown Univ., 5 F.3d at 669 (formulating standard as "reasonably necessary to achieve the stated objective"); id. at 679 (requiring "comparable benefits"). [Vol. 116:927 dominant LRAs count. 87 Most lenient of all, the phrase might mean that the restraint is permitted provided it makes an incremental contribution to the claimed benefit, even if there exists an LRA. 88 Dominant LRAs are used to varying degrees in areas of law other than antitrust. For example, constitutional scrutiny of state action is often limited to equally effective alternatives. 89 By contrast, as noted above, administrative law and tort law evaluate both the benefits and the costs of an alternative. 90 It is remarkable and even surprising that in this respect, antitrust bears a stronger resemblance to constitutional law. That resemblance is a clue to one function of the LRA test in antitrust, the subject of the next Part.

87. On this view, reasonable necessity includes achievement of the objective to a particular degree.

88. This appears to be the approach taken by Carrier, Real Rule, supra note 74, at 1341-46, in discussing a reasonable-necessity requirement that is less searching than condemnation in light of an LRA. The existence of a separate, more limited test is suggested by cases that treat the absence of reasonable necessity and the existence of an LRA as distinct inquiries. See, e.g., United States v. Visa U.S.A., Inc., 344 F.3d 229, 238 (2d Cir. 2003) ("[T]he government must prove either that the challenged restraint is not reasonably necessary to achieve the defendants' procompetitive justifications, or that those objectives may be achieved in a manner less restrictive of free competition."); Law, 134 F.3d at 1019 ("[T]he plaintiff . . . must prove that the challenged conduct is not reasonably necessary to achieve [defendant's] legitimate objectives or that those objectives can be achieved in a substantially less restrictive manner."). Of course, Visa and Law both include an explicit LRA test, so they are of limited help in determining whether a case that invokes reasonable necessity, without separately mentioning LRAs, really means to dispense with an LRA analysis.

The LRA test serves multiple functions in antitrust law. To varying degrees, each of them serves to ease an anxiety discussed in section II.A-that courts resist and avoid calculation and integration of the tradeoff between positive and negative effects that inheres in mixed conduct.

This Part identifies three functions of the LRA test-as a shortcut, a locus of balancing, and a tool of smoking out. Section II.B demonstrates that sometimes courts can legitimately avoid net-effects balancing by the shortcut of identifying a dominant LRA. In other cases, the only available alternative is less effective. Section II.C argues that condemning in light of a less effective alternative is best understood and defended as a form of balancing. A third approach is to employ the LRA not as a benchmark that the conduct must clear but as a diagnostic tool. Section II.D makes the limited case for a diagnostic to smoke out the anticompetitive effect of ambiguous conduct.

Although modern antitrust law has economics at its core, applying cost-benefit analysis to real-world conduct is not a straightforward task. Assessing the net effect of mixed conduct is hard. It is challenging to measure the effects of, say, a loss in competition for players arising from limits on compensation. (How much would the players receive? Which ones? What would be the change in output, if any?) Measuring the benefits to competition among sports leagues is also hard. (Would fans otherwise watch less college sports? How much worse is the next-best use of fans' time? What are the effects on advertising markets?) A full analysis requires an assessment of size, probability, and error costs. Such quantification, at least in a rough sense, is necessary in order to compare the two effects.

Implementation is a further, potentially difficult issue. Judges are generalists, see antitrust cases only rarely, are generally unfamiliar with the practice and industry at issue, and are therefore not well equipped to evaluate these effects. The limited fact-finding capacity of courts is a familiar refrain, 91 as is courts' limited capacity in antitrust cases, 92 particularly if the fact-finding is performed by a jury. The adversarial nature of antitrust litigation deepens these difficulties. Party economists frequently reach the conclusion that there is no mixed conduct to begin with. The plaintiff's expert concludes that there is an anticompetitive effect and that the justification is illogical or unsupported. 93 The defendant's expert concludes, to the contrary, that there is no anticompetitive effect and strong evidence of a justification. 94 Neither economist is likely to offer much insight about the integration of the contrasting effects. Doing so requires the party expert to embrace hypotheticals-"if defendant's evidence of justification is correct, how does that compare to your evidence of anticompetitive effect?"-that are poorly suited to the conduct of adversarial litigation. Federal courts have substantial authority to appoint neutral economists to support the court's work, 95 and neutrals would help, but they remain a rarity in antitrust litigation. 96 Another issue is more fundamental-that the court might view the values at stake to be incommensurable. 97 In the context of constitutional scrutiny of mixed state action, Justice Scalia memorably described this problem as the futile attempt to "judg[e] whether a particular line is longer than a particular rock is heavy." 98 The quotation comes from a dormant commerce clause case; the values at stake in an antitrust case economy against promotion of competition in another sector is one important reason we have formulated per se rules." (footnote omitted)); see also Herbert Hovenkamp, The Antitrust Enterprise 47 (2006) [hereinafter Hovenkamp, Antitrust Enterprise] ("[T]here is relatively little disagreement about the basic proposition that often our general judicial system is not competent to apply the economic theory necessary for identifying strategic behavior as anticompetitive."); Christopher R. Leslie, Rationality Analysis in Antitrust, 158 U. Pa. L. Rev. 261, 264 (2010) (arguing "courts are generally not effective arbiters of whether alleged business conduct is implausible").

93. For example, in United States v. American Express Co., the court cited testimony by plaintiff's expert that the challenged conduct had an anticompetitive effect and that a claimed justification (avoiding free-riding) was inapt in light of an LRA. 88 F. Supp. 3d 143, 209-10, 236-37 (E.D.N.Y. 2015).

94. The expert for American Express opined in part that the challenged conduct lacked anticompetitive effect because the form of competition suppressed is undesirable and that the conduct had the procompetitive effect of avoiding free-riding. See are comparably commercial. Not all antitrust cases raise deep issues, 99 but relatively difficult comparisons arise when the harms and benefits fall on different economic actors, requiring an interpersonal (or intergenerational) comparison 100 or a summing of contrasting effects on individuals and firms.

These difficulties tend to be particularly pronounced in cases alleging anticompetitive exclusion of a rival. 101 The harm to consumers is an indirect consequence of limiting a rival's production. Often, the plaintiff asserts a harm to innovation, 102 the welfare effects of which can be larger than with high prices or reduced output 103 while being particularly difficult to measure. 104 For its part, the defendant often asserts a benefit to innovation thanks to the conduct. 105 The resulting anxiety about balancing has led courts and commentators to shy away from an analysis of net effects. Addyston Pipe worried about the "sea of doubt" that accompanies a search for net effects. Judge Robert Bork dismissed the possibility of "weigh[ing] procompetitive effects against anticompetitive effects" thusly: "[W]e do not think that a usable formula if it implies an ability to quantify the two effects and compare the values found." 107 The leading treatise, while accepting the need for both an LRA test and a net-effects balancing step, regards the prospect with trepidation: If the court reaches that step, it "must somehow weigh and balance the harm against the benefit." 108 This anxiety is shared with constitutional law. There the anxiety is rooted in analogous concerns about quantification and incommensurability. 109 Constitutional scrutiny raises a further concern that is absent from antitrust law-that courts might be stepping beyond their proper bounds when they second guess the work of the legislature. 110 The response in constitutional law has two elements of relevance to antitrust-first, to suppress any explicit calculation of net effects. Although net-effects balancing is an acknowledged part of constitutional analysis in many democracies, 111 courts generally avoid it in the United States. 112 The second response is to emphasize an analysis of intent in place of an analysis of effects.

To a degree, antitrust law has embraced both responses. Some courts have dispensed with net-effects balancing. For example, the O'Bannon court and other courts have simply dropped the net-effects test (while retaining the LRA test), an approach endorsed by some commentators. 113 Other commentators have proposed replacing balancing with an inquiry into intent grounded in an evaluation of profitability. If the mixed conduct makes "no economic sense" for a profit-maximizing firm apart from its anticompetitive tendency, the conduct is subject to 109. See, e.g., T. Alexander Aleinikoff, Constitutional Law in the Age of Balancing, 96 Yale L.J. 943, 984 (1987) (noting and criticizing "opinions that have openly explored the 'costs' and 'benefits' of constitutional rules and appealed to empirical evidence of the effect of constitutional doctrine on societal interests"); Frederick Schauer, Commensurability and Its Constitutional Consequences, 45 Hastings L.J. 785, 809 (1994) (noting and "deflat[ing] some of the moral rhetoric surrounding the commensurability debate in legal and constitutional context"). 111. Barak, supra note 42, at 340, 348 (discussing "balancing in the strict sense" as final stage of proportionality review).

112. Commentators often remark upon this contrast in comparisons of the regimes. See, e.g., id. at 518; Jud Mathews & Alec Stone Sweet, All Things in Proportion? American Rights Review and the Problem of Balancing, 60 Emory L.J. 797, 833-36 (2011) (contrasting "all-purpose" German proportionality analysis with "rigid" U.S. approach).

113. See, e.g., Bork, Antitrust Paradox, supra note 36, at 279 (proposing categorical approach to assessing legality of horizontal restraints); Arthur, supra note 74, at 367 (characterizing balancing test as impractical); see also Carstensen, supra note 74, at 65-66 (advocating Addyston Pipe approach). But see id. at 67 (acknowledging insufficiency of approach in identifying unreasonable restraints). For further discussion, see infra section III.B.

LESS RESTRICTIVE ALTERNATIVES condemnation, while conduct that makes economic sense is permitted. Notwithstanding these developments, however, net-effects balancing remains an important part of the law on the books for the Supreme Court and many lower courts. 115 Beyond omission and a shift toward intent, avoidance of net effects in antitrust takes a further, more subtle form. Courts may avoid balancing in practice, while acknowledging balancing in theory, by deciding the case on other grounds that arise earlier in the analysis. Some evidence of this avoidance appears in rule-of-reason cases. Although it is commonplace to understand the rule of reason as a fact-intensive search for net effects, cases are seldom decided on that explicit basis. An extensive survey of rule of reason final judgments concluded that very few are decided on net-effect balancing grounds. 116 A similar avoidance appears in monopolization cases such as the famous D.C. Circuit opinion in United States v. Microsoft Corp., which reached balancing in just two instances, despite repeated assertions of mixed conduct. 117 Careful observers have gone so far as to declare that explicit balancing is a "myth." 118

The LRA test is a response to the anxiety about balancing. It replaces a hard question-the calculation of net effects-with an analysis that requires no tradeoff between justification and restriction. The shortcut works, but only if the LRA is a dominant alternative. The analysis thus shifts to an inquiry into whether the proposed alternative is less restrictive and equally (or more) effective. 119 This section examines the circumstances under which that inquiry offers a valid and effective shortcut.

As an example, consider the public performance licenses issued by the American Society of Composers, Authors and Publishers (ASCAP). In the early years, songwriters granted exclusive public performance licenses to ASCAP, which were repackaged into a blanket license and made available to radio stations and other public performers of musical works. 120 The exclusive licenses eliminated price competition among the songwriters, essentially creating a market-wide cartel, to the detriment of some consumers. But at the same time, they made possible the creation of a valuable new good, a blanket license offering immediate and easy access to a large repertoire of songs. 121 This mixed conduct, however, was subject to an obvious LRA-to make the licenses nonexclusive. Nonexclusive licenses fully retained the benefit of the blanket license while permitting individually negotiated licenses. 122 Put in the terms of Addyston Pipe, the licenses were ancillary to a desirable transaction but more restrictive than necessary to achieve the benefit. Under pressure from the Justice Department, ASCAP shifted to nonexclusive licenses. 123 Taking the LRA shortcut-nonexclusive licenses in the case of ASCAP-has several benefits. One benefit, already emphasized, is to avoid the calculation of harm and benefit and thereby relieve the anxiety 119. Although there is no "balancing" in the sense of a tradeoff of incremental benefit and incremental harm, there may be balancing in other senses of the term. For example, finding a dominant alternative might be viewed as identifying a different balance, compared to that selected by the defendant. 120. See 17 U.S.C. § 106(4) (2012) (granting, to owner of musical work copyright, right to authorize public performances). ASCAP does not handle all compositions; Broadcast Music, Inc. and the Society of European Stage Authors and Composers also assemble blanket licenses. For digital audio transmissions, a second license to the sound recording is also required. See id. § 106(6) (granting, to owner of sound recording copyright, right to authorize public performances by means of digital audio transmissions).

121. Walter L. Pforzheimer, Comment, Copyright Reform and the Duffy Bill, 47 Yale L.J. 433, 443 (1938) (noting benefit to music consumers from access to "extensive repertoire required by modern entertainment demands").

122. A different question would be presented if nonexclusivity caused some songwriters to withdraw their music from the blanket license. Courts would then be entering the territory of balanced LRAs.

LESS RESTRICTIVE ALTERNATIVES 953 about balancing. The court also avoids deciding whether a particular benefit is cognizable. 124 Even if the mixed conduct can be conceptualized as a single effect, the shortcut may be useful if measurement of the overall effect is difficult.

As shortcuts go, however, the LRA test is unusual. Many shortcuts, in antitrust as elsewhere, offer simplicity at the expense of accuracy. For example, courts have adopted an explicitly lenient rule for predatory pricing. 125 Price fixing is per se illegal even though sometimes it is welfare enhancing. 126 Many antitrust claims require a finding of horizontal agreement, which serves as an imperfect proxy for socially harmful activity that can occur even without agreement. 127 By contrast, the LRA-as-shortcut approach is accurate, provided its demanding conditions are met, but not necessarily simple. The main demand pertains to the benefit: to have enough information to determine that the alternative has (at least) equal benefits compared to the conduct. This informational demand may be similar in nature to the avoided costs of calculating net effects. As for restrictiveness, showing that the LRA is indeed less restrictive is comparatively easy. Often it is not restrictive at all. The alternative to an exclusive license may be a nonexclusive contract with no anticompetitive tendency. The contractual alternative to merger may offer no opportunity for raising price.

In meeting this informational demand, an easy special case arises when the action makes no incremental contribution to the claimed justification. Such cases might be said to fail ancillarity. The LRA is simply to refrain from the conduct. Without the conduct, the level of benefit is unchanged. The conduct is unnecessary, not merely in the sense that switching to an alternative achieves the same benefit with less harm but in the stronger sense that halting the conduct would do so.

This scenario, a special case of Figure 1, is depicted in Figure 2. The alternative of cessation is located at the origin. That LRA suffices to condemn actions (such as point A) to the west of the alternative (point Z). A glance at the figure reveals an overlap between the LRA and neteffects inquiries. Where the LRA is mere cessation, rather than alternative affirmative conduct, the two inquiries essentially merge.

124. See supra section I.A (discussing benefits premised on avoiding competition and benefits retained by defendant); cf. Ross, supra note 75, at 492-93 (identifying avoidance of costly inquiry into market power as further benefit).

125. A restraint fails ancillarity when the harm lacks a causal relationship to the claimed benefit. One common fact pattern arises when a defendant takes several conceptually separable actions at the same time, and the LRA test is used to exclude from the analysis a contemporaneous but irrelevant beneficial action. 128 For example, in considering Apple's conspiracy to raise prices on ebooks, the benefits from Apple's introduction of the iPad, which coincided with the conspiracy, should not count in Apple's favor, because the iPad would have been introduced in any event. 129 A second fact pattern is illustrated by another case about college sports, NCAA v. Board of Regents of the University of Oklahoma, 130 decided more than thirty years before O'Bannon. Plaintiffs challenged an NCAA rule that limited the number of games broadcast on television, thereby increasing the price the networks paid. 131 The NCAA contended that its rule benefited consumers by helping to maintain competitive balance

LESS RESTRICTIVE ALTERNATIVES 955 among the teams. 132 The Supreme Court rejected this argument because competitive balance was already promoted equally well by other existing NCAA rules that had no restrictive effect. 133 In other words, the benefit was already being achieved to some degree, and the restraint provided no incremental benefit. Here, once again, cessation served as an LRA.

An LRA only works as a shortcut if it is dominant. Most LRAs, however, flunk this test because they are less effective than the conduct in serving a desirable end. Section II.C.1 provides an account and typology of less effective alternatives. Section II.C.2 offers a qualified defense of such alternatives as a basis for condemning anticompetitive conduct.

1. Less Effective Alternatives. -Many, and perhaps most, LRAs are less effective than the challenged conduct, a point that appears to be well recognized. 134 The judicial treatment of less effective alternatives can be divided into four basic patterns.

a. No Liability. -Some cases deny liability upon a determination that the LRA is less effective, as one would expect from a rule limited to dominant alternatives. For example, in O'Bannon, the court considered a second proposed LRA, deferred compensation to be paid once the player leaves college. 135 This alternative only very imperfectly serves the NCAA's goal of promoting amateurism. 136 The players are still being paid a substantial amount, just delayed. That hardly preserves college athletics as a money-free zone. Thus, the LRA was accurately labeled as a less 132. This was one of several arguments considered (and rejected) by the Court. See id. at 117-20.

133. Id. at 119 ("[T]he NCAA imposes . . . other restrictions designed to preserve amateurism . . . much better tailored to the goal of competitive balance than . . . the television plan, and which are clearly sufficient to preserve competitive balance to the extent it is within the NCAA's power to do so." (internal quotation marks omitted) ( 136. Id. (explaining court's view that permitting payment is less effective in "promoting amateurism and preserving consumer demand"). The court did not consider whether this LRA might nevertheless serve the other credited justification, ensuring student integration. Here, the court might have considered whether players who anticipate deferred compensation may shift spending forward, in possible derogation of the integration goal. effective alternative, and as a consequence, the plaintiff's claim was rejected. 137 b. Liability: Lesser Effectiveness Misidentified as Equally Effective. -In other cases, a less effective alternative is a basis for condemnation, an outcome reached with varying degrees of candor. For example, consider the St. Luke's case, challenging a transaction that brought eighty percent of primary care physicians in an Idaho town under the same roof. 138 After a bench trial, the district court concluded that the merger was likely to raise prices. 139 At the same time, the court recognized an important efficiency resulting from the deal: increased integration of doctors with other medical providers. Integrated providers could better coordinate care, and by negotiating payment based on outcomes rather than procedures performed, they had an incentive to economize on utilization over time. 140 The court "complimented" the acquirer on its "foresight and vision" in acquiring practices to advance its integration goal. 141 Nevertheless, the court condemned the merger on the ground that the "same effect" could be accomplished instead through the LRA, a contract-based network of providers. 142 In reaching this conclusion, the court largely relied on the general proposition that a network would achieve the same benefits. 143 The court's unsupported confidence that contracts are equally effective is doubtful, given previous, failed efforts to accomplish integration without merger 144 and the literature indicating that integrated care requires a high degree of formal integration. 145 Here, the court embraced a less effective LRA. 146 137. Id. at 1079 (vacating district court's judgment regarding deferred compensation). 138. Saint Alphonsus Med. Ctr.-Nampa, Inc. v. St. Luke's Health Sys., Ltd., Nos. 1:12-CV-00560-BLW, 1:13-CV-00116-BLW, 2014 WL 407446 (D. Idaho Jan. 24, 2014).

139. See id. at *7-8 (noting likely price rise and other bad effects). The court noted that some patients in Nampa, Idaho might go to neighboring Boise, Idaho in response to a price rise but concluded that Boise was an ineffective constraint on pricing. See id. at *7 (noting, prior to transaction, "[o]nly 15% of Nampa residents obtain[ed] their primary care in Boise"). 140. See id. at *1 (arguing transition to integrated care "require[s] a major shift away from our fragmented delivery system"). 141. Id.; see also id. at *2 ("The Acquisition was intended . . . primarily to improve patient outcomes. The Court believes that it would have that effect if left intact, and St. Luke's is to be applauded for its efforts to improve the delivery of health care . . . .").

142. Id. at *2. Defendants failed to carry their burden of persuasion in establishing the absence of an (equally effective) LRA. See id. at *14-19 (rejecting several efficiency defenses). For a discussion of the allocation of burdens, see infra section IV.A.

143. See id. at *17 ("Because a committed team can be assembled without employing physicians, a committed team is not a merger-specific efficiency of the Acquisition."); id. at *23 (rejecting "committed team of physicians" as merger-specific efficiency). The court provided a detailed rejection of just one component of the claimed merger-specific efficiencies, namely, that shared electronic records required a merger. Id. at *19.

144. See id. at *4 (noting failure of earlier collaborations). 145. See, e.g., Berkeley Forum, A New Vision for California's Healthcare System: Integrated Care with Aligned Financial Incentives 6 (2013), http://berkeleyhealth careforum.berkeley.edu/wp-content/uploads/A-New-Vision-for-California's-Healthcare-c. Liability: Lesser Effectiveness Ignored. -Yet other courts condemn conduct in light of an LRA without considering the effectiveness of the alternative. Often the LRA is in fact less effective. For example, in United States v. Philadelphia National Bank, the Supreme Court enjoined a bank merger defended on the ground (among others) that "only through mergers can banks follow their customers to the suburbs and retain their business." 147 The Court rejected this asserted benefit in favor of the LRA of internal growth. 148 Internal growth avoids the anticompetitive concern, but there is no basis for concluding that it would provide a certain or effective means of geographic expansion.

Tying cases furnish additional examples. One early case tested IBM's claim that a tie between its tabulating machines and punch cards was necessary to avoid machine damage from inferior non-IBM punch cards. 149 The Supreme Court rejected this justification in favor of the LRAs of customer outreach and detailed contractual specification. 150 However, if the goal were to preserve goodwill, an outright ban would have been more effective than a public relations campaign.

d. Liability: Lesser Effectiveness Acknowledged. -Finally, we can occasionally observe condemnation in light of an alternative that is acknowledged to be less effective. In Arizona v. Maricopa County Medical Society, the Court condemned a maximum fee schedule set by doctors. 151 The Court entertained the doctors' justification-that a fee schedule lowers premi-System.pdf [http://perma.cc/6T9P-9PZQ] (endorsing increase in care via "fully-or highly-integrated care systems"); Alain C. Enthoven & Laura A. Tollen, Competition in Health Care: It Takes Systems to Pursue Quality and Efficiency, Health Aff. 420 (2005), http: //content.healthaffairs.org/content/early/2005/09/07/hlthaff.w5.420.full.pdf (on file with the Columbia Law Review) (identifying healthcare markets "that encourage the formation of high-quality, efficient, integrated delivery systems" as cure for deficiencies in current system); Kaiser Comm'n on Medicaid & the Uninsured, Integrating Physical and Behavioral Health Care: Promising Medicaid Models 1-3 (2014), https://kaiserfamily foundation.files.wordpress.com/2014/02/8553-integrating-physical-and-behavioral-healthcare-promising-medicaid-models.pdf [http://perma.cc/8EWD-5V3S] (discussing issues resulting from fragmented care and describing methods to increase integration).

146 as means "to preserve . . . the good will of its patrons by preventing the use of unsuitable cards which would interfere with . . . performance"). 150. Id. at 139-40 ("Appellant is not prevented from proclaiming the virtues of its own cards or warning against the danger of using, in its machines, cards which do not conform to the necessary specifications").

151. 457 U.S. 332, 336 (1982). The case is styled as a per se condemnation of the conduct, but the Court considered the defendants' proffered justification at length. ums by reducing the cost of providing insurance 152 -but rejected it in light of the LRA that insurers could set the schedule instead of doctors. The Court's analysis conceded quietly that such an LRA would be less effective. 153 In all but the first category, the court condemns conduct in light of a less effective LRA. Such decisions might be dismissed as simple error. 154 An alternative interpretation, however, presents itself-that courts are engaged in balancing.

This scenario is depicted in Figure 3, which revisits the setting of Figure 1. Although the alternative Z' is somewhat less effective than A, it is much less restrictive. There is a net loss from choosing A instead of Z', and hence Z' is a basis for condemning A. The set of conduct condemned on the basis of Z' includes any conduct to the west of Z' that is below the forty-five-degree line, where incremental effectiveness and incremental restrictiveness exactly balance. FIGURE 3: BALANCED ALTERNATIVES 152. Id. at 352 (noting resulting "potential for lower insurance premiums"); see also id. at 352 n.25 (noting fee schedule could make it "easier" and "less expensive" to calculate risks).

153. Id. at 353-54 (acknowledging "doctors may be able to do it [i.e. set maximum prices] more efficiently than insurers").

154. See, e.g., Meese, Price Theory, supra note 74, at 168 (arguing "[LRA] test is plainly flawed" because LRAs are less effective in practice).

Balancing within the LRA test is a rough cost-benefit analysis between the conduct and the alternative. The analysis is constrained in important respects. First, an alternative that is more effective and more restrictive-an alternative to the northwest of A-does not count. Alternatives that are more restrictive are not LRAs, by definition. They are also out of place in antitrust analysis, which condemns conduct for its restrictiveness, not merely because one can imagine a superior alternative. 155 Second, the analysis measures the action and the alternative only along the dimensions of the restriction and the cognizable benefit. It omits other axes along which conduct might be desirable or harmful. 156 Net-effects balancing is frequently recited as the law on the books but seldom observed in practice. 157 Meanwhile, LRA balancing is roundly ignored. Like a balloon squeezed on one end, balancing has merely been shifted to a different part of the doctrine. One consequence is that, contrary to conventional wisdom, balancing turns out not to be a myth after all. Instead, balancing is happening out of sight, in the LRA step. 158 Often, this is balancing in disguise. Antitrust courts usually do not announce that they are conducting balancing with respect to the alternative. They are either silent about the nature of the alternative, or assert falsely that the alternative is equally effective.

Balancing in disguise has undesirable consequences. Opacity encourages inadequate analysis. The court never engages publicly with the tradeoff at the heart of the case. There is no full and open resolution about how the interest in player compensation should be traded off against the popularity of the sport or how low prices are integrated with clinical improvement. Where balancing is not explicit, important information may never even be presented by the parties. Moreover, when courts do not show their work, review by the appellate court is much less effective. 159 With the analysis obscured from view, it is unclear whether LRA-based balancing is currently achieving optimal results. Conceivably courts are getting the balance right. Possibly they are routinely imposing false positives. It is impossible to say where balancing is disguised. Acknowledging LRA-based balancing is therefore crucial.

155. See Verizon Commc'ns Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398, 415-16 (2004) (noting antitrust law "does not give judges carte blanche to insist that a monopolist alter its way of doing business whenever some other approach might yield greater competition").

156. See supra section I.A (discussing limits to cognizable justifications). 157. See supra notes 116-118 and accompanying text (reviewing commentary concluding net-effects balancing is rare).

158. The LRA test is not the only mechanism for balancing in disguise. See Greene, supra note 105, at 78, 86, 89 (arguing, in context of product design, courts may engage in "stealth balancing" by "aggressively dismissing innovation as pretextual").

159. Cf. Bert I. Huang, Concurrent Damages, 100 Va. L. Rev. 711, 766-67 (2014) (discussing procedural benefits that result when appellate court need not "reverse engineer" lower court's damages calculation).

Alternatives. -Bringing LRA-asbalancing out into the open raises the obvious question: Is this a good idea? LRA-as-balancing has several points in its favor. At the outset, it is important to note that balancing against alternatives would hardly be a development unique to antitrust. LRAs are a locus of balancing in other fields, with varying degrees of openness, ranging from tort 160 to trade 161 to constitutional law. 162 Considering balanced alternatives would thus bring antitrust into line with other areas of law that openly trade off the pros and cons of an alternative.

Balancing within the LRA test also complements and mirrors balancing at the final net-effects step. The LRA test and the inactionfocused net-effects test are doing parallel work, just as in a cost-benefit analysis that incorporates multiple alternatives. 163 Fully implementing a net-effects approach implies consideration of less effective alternatives. Such an approach has a further consequence: It opens up a new set of easy cases, in which the alternative is somewhat less effective but much less restrictive. These cases would tend to arise where the anticompetitive effect is large and clear. The larger the anticompetitive effect of the conduct, the more leeway there is for a less effective alternative to furnish a basis for condemnation without fitting the demanding requirements of a dominant LRA.

Balanced LRAs are largely consistent with current doctrine. The Supreme Court has not limited the inquiry to dominant alternatives, and its general endorsement of a net-effects approach can be understood to encompass a broad examination of alternatives. Some lower courts, in characterizing the inquiry as a matter of "similar" rather than "same" 160. See supra note 47 and accompanying text (discussing full evaluation of alternatives in tort law context). In fact, the Restatement of Torts devotes a comment to the separate analysis of dominant alternatives and balanced alternatives:

If the actor can advance or protect his interest as adequately by other conduct which involves less risk of harm to others, the risk contained in his conduct is clearly unreasonable. If any other practicable course of conduct is clearly likely to give his interest a less adequate advancement or protection the question whether the risk is or is not unreasonable depends upon whether the additional risk involved in the particular course of conduct outweighs the additional advancement or protection which it is likely to secure. Restatement (Second) of Torts § 292(c) cmt. ( 163. Indeed, the net-effects baseline can be conceived as a balanced LRA. This is an extension of the point made in section II.B that cessation is an LRA. There, cessation was recognized to be a dominant LRA compared to harmful conduct that offered no incremental benefit. If the conduct does make a contribution, then cessation can once again be conceived as an alternative-albeit a less effective alternative-that furnishes a potential basis for condemnation depending on the balance of benefits and harms.

LESS RESTRICTIVE ALTERNATIVES effectiveness, have conceded the need for a degree of balancing. Others have taken no view 165 and hence are free to adopt a balanced approach. That leaves those lower courts that have rejected a balanced approach in favor of equal effectiveness. To adopt balanced LRAs, such a court would need to either change its approach or else conceptualize the net-effects step-if it has one-to include a comparison of alternatives beyond inaction. A model for such an approach is constitutional scrutiny outside the United States under one version of "proportionality." 166 Notwithstanding these favorable points, LRA-as-balancing runs headlong into the anxiety about balancing discussed above. LRA-asbalancing has coexisted with an anxiety about balancing in part because the LRA test does not announce itself as a balancing test. Pushed into the open, courts are forced to confront the very difficulties they had initially hoped to avoid. These difficulties are substantial, for reasons extensively discussed by others. 167 Identifying a balanced LRA has many of the same pitfalls that afflict net-effects balancing between the conduct and inaction. The balancing performed by courts in the LRA test is necessarily crude. A crude comparison is adequate for some cases but not others.

In a subset of cases, the difficulty of the LRA comparison is greater than with net effects because the LRA is hypothetical rather than actually implemented. Evaluating both its incremental benefits and incremental costs may entail some guesswork. 168 However, the same is often true of net effects, which also requires estimation of projected incremental benefits and incremental costs, whether because the conduct has not yet taken hold or because the unobserved baseline no-conduct world is itself changing. Thus, this is an extra criticism of LRAs in the subset of cases 164. See supra note 82 and accompanying text (discussing scholarship and lower-court decisions addressing "similar" effectiveness).

165. See supra notes 83-86 and accompanying text (identifying courts taking no clear view as to effectiveness). 166. See Barak, supra note 42, at 350-51 (noting usual comparison in last step is to inaction but other alternatives should be considered as well). Under this approach, the court first checks whether there is a dominant LRA compared to the conduct. See id. at 321 (arguing LRA must "achieve the proper purpose to the same extent"). Next, it asks whether the conduct is superior to both inaction and alternatives, trading off both costs and benefits. Such an approach renders the LRA step, thus limited to dominant LRAs, a lesser-included analysis compared to an evaluation of net effects that has been expanded to include balanced LRAs.

167. See, e.g., Frank H. Easterbrook, The Limits of Antitrust, 63 Tex. L. Rev. 1, (1984) ("[I]t is impossible to determine the difference in efficiency between a known practice and some hypothetical alternative."); Louis Kaplow, The Patent-Antitrust Intersection: A Reappraisal, 97 Harv. L. Rev. 1813, 1833-34 (1984) (summarizing difficulties in doing cost-benefit analysis for patent licenses).

168. See Feldman, Misuse, supra note 72, at 603-04 (describing difficulties).

where the net-effects test is able to use the observed preconduct status quo as a baseline. 169 Despite these substantial difficulties, the importance of net effects as a tool for identifying anticompetitive conduct, including the Supreme Court's emphasis on a net-effects approach, supports incorporating that same incrementalist analysis into the LRA test. This conclusion is strengthened by the best practices discussed in Part IV, which limit the scope of the test and reduce the likelihood and costliness of error in identifying balanced LRAs. The case for incorporation is particularly strong as to those LRAs that are nearly dominant and much less restrictive. Recognizing near-dominance as a second-best approach to balancing would generalize the approach of those courts that have embraced LRAs with similar (albeit not identical) benefits as the challenged conduct. 170 It bears emphasis that under the balancing approach, it is not enough to simply identify an LRA without any inquiry into effectiveness. That might seem to be an obvious implication of any analysis of incremental harms and benefits. However, some courts have required that the defendant select the "least" restrictive alternative. 171 Logically, this formulation might seem to be a distinction without a difference. An obligation to pick the "least" bad alternative is equivalent to the obligation to choose an available "less" bad alternative. 172 However, the emphasis on "least" might embolden a court to abandon the inquiry into effectiveness, perhaps echoing a distinction made in First Amendment cases. There, "least" has been taken to suggest a particularly searching analysis, 173 though this distinction is inconsistently applied. 174 Due attention to the effectiveness of the alternative is essential under the balancing test, whatever its verbal formulation.

169. Another potential difference arises when the effect of particular conduct can be measured in some other way, for example, by comparing prices and output between territories that have the restraint and others that do not. For a discussion of the availability of such tools in evaluating an LRA, see infra section IV.C. 170.

The shortcut and balancing approaches to the LRA test share the same basic approach, which is to identify an alternative with a superior mix of benefits and costs. Both approaches reflect the cost-benefit technique that is central to the self-conception of modern antitrust law and familiar from administrative law and tort law, among other areas. A different approach is to eschew the investigation of overall effect in favor of uncovering or "smoking out" a single, decisive aspect of defendant's conduct. If the single feature is made decisive, there is no need to balance and hence no anxiety about balancing.

This section considers the LRA test as a tool of smoking out three distinct aspects of the defendant's behavior: bad intent, anticompetitive effect, and a pretextual procompetitive justification. Section II.D.1 considers the idea that just as the LRA test smokes out the bad intent of state actors in constitutional law, it might have a similar role in antitrust law. As shown below, however, the kind of intent thereby uncovered in an antitrust case is usually uninformative. Section II.D.2 offers an account of the LRA test as a means to strengthen an inference of anticompetitive effect. Section II.D.3 explains why an LRA test is, at least without modification, incapable of demonstrating that an asserted procompetitive justification is merely pretextual.

1. Bad Intent. -In constitutional law, the smoking out inquiry aims to smoke out the true intent of the state actor. The court assesses the legislature's conduct for an impermissible motive rather than the potentially mixed effects of its actions. The focus on intent avoids a judicial judgment about the desirability of effects and associated infringement of the legislature's prerogative in selecting appropriate ends and the best means of achieving them. 175 Smoking out intent has assumed an important role in constitutional doctrine. 176 For example, in R.A.V. v. City of St. Paul, the Court decided a First Amendment challenge to a city ordinance prohibiting "fighting words that insult, or provoke violence, on the basis of race, color, creed, religion, or gender." 177 The city defended the ordinance as a means to protect the rights of individuals "that have historically been subjected to discrimination." 178 The Court struck down the ordinance in light of an LRA with "precisely the same beneficial effect," a prohibition not limited to the enumerated topics. 179 The Court was open about its search for impermissible motive. 180 Comparing the city's conduct to the LRA revealed the legislators' "special hostility towards the particular biases singled out," a hostility impermissible under the First Amendment. 181 One might imagine a similar role for the LRA test in antitrust law. In fact, Justice Brennan suggested this more than fifty years ago. In White Motor Co. v. United States, the Justice Department challenged vertical restraints that granted exclusive territories to dealers reselling trucks and parts. 182 The Court held that per se condemnation was inappropriate and remanded for trial. 183 Justice Brennan's concurrence urged an examination of LRAs, the first appearance of the term "less restrictive alternatives" in the U.S. Reports. 184 The explicit goal was to smoke out the defendant's intent: "If the restraint is shown to be excessive for the manufacturer's needs, then its presence invites suspicion . . . that the real purpose of its adoption was to restrict price competition." 185 Paul, 505 U.S. 377, 386 (1992) (noting strict scrutiny prevents regulation "based on hostility-or favoritism-towards the underlying message expressed"); see also Richard H. Fallon, Jr., The Supreme Court, 1996 Term-Foreword: Implementing the Constitution, 111 Harv. L. Rev. 54, 96 (1997) (discussing uses of smoking out with respect to affirmative action and First Amendment); Elena Kagan, Private Speech, Public Purpose: The Role of Governmental Motive in First Amendment Doctrine, 63 U. Chi. L. Rev. 413, 414, 451 (1996) (arguing First Amendment doctrine "comprises a series of tools to flush out illicit [governmental] motives and to invalidate actions infected with them"); Regan, supra note 110 at 1209-33 (perceiving similar role in evaluating claims of state protectionism under dormant commerce clause).

177. 505 U.S. at 391 (internal quotation marks omitted). 178. Id. at 395. 179. Id. at 396. This is an example of conduct that, though labeled a less "restrictive" alternative, in fact regulates more speech. 180. See, e.g., id. at 390 (stating goal of identifying "realistic possibility that official suppression of ideas [was] afoot"); id. at 394 (expressing concern that government "seek[s] to handicap the expression of particular ideas").

181. Id. at 396. 182. 372 U.S. 253, 255-56 (1963). The restraint also restricted sales as to particular customers. See id.

183. Id. at 264. 184. Id. at 271 (Brennan, J., concurring); see also id. at 270 (Brennan, J., concurring) (arguing relevant question "is not simply whether some justification can be found, but whether the restraint so justified is more restrictive than necessary, or excessively anticompetitive, when viewed in light of the extenuating interests").

185. Id. at 270 n.9 (Brennan, J., concurring). That said, Justice Brennan also called for "a full inquiry into the pros and cons" of the restraint, which might be read as consistent with a net-effects approach. Id. at 270 (Brennan, J., concurring).

Justice Brennan's approach was likely influenced by contemporaneous developments in constitutional law. In the same Term that Justice Brennan wrote his White Motor concurrence and Philadelphia National Bank, Sherbert v. Verner (another Justice Brennan majority opinion) 186 and other cases introduced a modern conception of strict scrutiny that includes an LRA test. 187 Justice Brennan was open about his proposed inquiry into a firm's motives, though more guarded about questioning the bad motives of the state in constitutional cases decided the same Term. 188 The facts of the case illustrate how this inquiry might work. The defendant claimed that exclusivity encouraged distributors to invest in fully developing their territories. 189 That incentive would be undercut if dealers from other areas could sweep in and cherry-pick customers. The upshot, according to White Motor, was to encourage interbrand competition with other manufacturers. 190 The government's proposed LRA was a detailed contract that directly specified the desired activities. 191 If the detailed contract also elicited services without restricting dealer competition, Justice Brennan was suggesting, this would show that the true purpose was to restrict intrabrand competition among its dealers.

Such an inquiry in antitrust cases faces the initial hurdle that intent is often assigned only a minor role in antitrust analysis. 192 After all, factfinders can be misled by vivid language that merely reflects hard competition. 193 Moreover, even when firms lack any clear provable intent to behave anticompetitively, there remains an interest in halting and deterring that conduct if it has adverse effects.

Nevertheless, intent surely still plays some role in examinations of mixed conduct. 194 The Supreme Court's test for monopolization refers to "willful acquisition or maintenance" of monopoly power. 195 Other cases indicate the relevance of an "inference . . . that [conduct was] intended to restrain trade and enhance prices" 196 or more colorfully, a defendant's "dreams of monopoly." 197 Another prominent example is Microsoft, which tests a procompetitive justification for "pretext." 198 A more important problem is that the question of interest in a constitutional case is seldom informative in antitrust law. In a constitutional case, what is uncovered is the unknown, hidden type of the decisionmaker. The question asked is: Is this the type of decisionmaker that harbors an improper motive and thereby benefits (improperly) from its action?

In constitutional cases such as R.A.V., the comparison to an LRA reveals a gratuitous harm, which is a likely result if the state actor desired the harm and an unlikely result otherwise. The court is making a Bayesian inference about the probability that the state had an ill motive. 199 The idea is to begin with a "prior" probability about the issue in question, such as the probability that the city of St. Paul had a "special hostility." The court then adjusts this probability upward or downward in light of a new fact. Suppose the new fact is likely to exist if the city has special hostility and is unlikely to arise if the city had no special hostility. The court then increases its perceived probability that the city has special hostility. 200 Here is a numerical illustration of the inference, loosely based on the facts of R.A.V. Suppose that, without taking account of the ordinance, the court believes that the city is twenty-five percent likely to harbor special hostility. The court examines the ordinance and identifies a gratuitous harm in light of the LRA, a prohibition on fighting words not limited to particular topics. Suppose further that a gratuitous harm is forty percent likely to result if a state actor is motivated by special hostility and ten percent likely if a state actor is indifferent. Once the court observes a gratuitous harm, it updates its assessment of a likely bad motive. Applying Bayes' Rule, instead of twenty-five percent, the court's new ("posterior") probability that the state has a bad motive is fifty-seven percent. 201 Importantly, a dominant LRA ("same beneficial effect") is necessary for the analysis; otherwise there is no basis for updating our probability because either type would reject an LRA that was less effective in achieving its claimed unobjectionable end. 202 This approach, useful in constitutional cases, does not do much work in antitrust. In antitrust, the decisionmaker's type is seldom in doubt. It is typically clear that the firm is the kind that benefits from the 199. The R.A.V. Court's stated interest in assessing the probability that the state action reflects an impermissible motive, and how that probability varies depending on the facts, invites such an analysis. R.A.V. v. City harmful effect. For example, if postmerger prices are higher for medical procedures, there is no question of whether these providers are the type that benefits from higher prices. The harmful effect normally implies a transfer that benefits the providers. Thus, if an anticompetitive effect exists, normally the conclusion readily follows that the firm knowingly benefitted from that effect. This idea is captured by Judge Learned Hand's encomium that "no monopolist monopolizes unconscious of what he is doing." 203 It is a rare antitrust case in which the connection between the anticompetitive harm and the firm's benefit is unclear, and hence there is little to be gained from using an LRA to smoke out a hidden type. Perhaps for this reason, Justice Brennan's suggestion has received little attention in later case law. 204 2. Anticompetitive Effect. -Justice Brennan's suggested approach can be adapted instead to an examination of effect. This inquiry makes use of the idea, frequently expressed in case law, that intent sheds light on effect. 205 The existence of an LRA raises the question of why the defendant chose an action that harms consumers rather than the alternative. 206 The likely reason is that the defendant expected to gain from the resulting consumer harm. The defendant's choice is consequently a basis for inferring that there is such an anticompetitive effect in fact. 207 This inference is valuable when the evidence of anticompetitive effect is ambiguous. For example, in a typical merger challenge, the merger has not yet occurred, and plaintiff's evidence is limited to a controverted prediction of higher prices in the future. 208 Even after the conduct, the evidence of a price rise is often ambiguous. In the face of such uncertainties, it is desirable to have greater confidence of anticompetitive effect before condemning the conduct. The limit of this inference is illustrated by cases like O'Bannon and St. Luke's, in which the fact-finder perceived a clear anticompetitive effect. 209 In such cases this inquiry is of little importance.

An LRA provides a useful inference when the alternative is equally (or more) profitable compared to the defendant's conduct, considering all sources of legitimate profits. Given an inquiry into the firm's decisionmaking, the focus on profit is crucial. The point is to isolate and identify a "bad" source of profits by pointing to an alternative that confers the same level of legitimate benefit. If the alternative is less profitable, then the defendant has an independent, sufficient explanation for its conduct. For example, White Motor would avoid detailed contracts if they were less profitable, even apart from any anticompetitive effect. 210 This would be equally true, whether or not the conduct also had an anticompetitive effect. 211 If the LRA is more profitable than the defendant's conduct, then the inference of anticompetitive effect is strengthened. For a rational, wellinformed firm, forgoing legitimate profits makes no sense unless the firm is making up the difference by earning extra profits through anticompetitive conduct. The existence of a more profitable LRA thus raises a strong inference of an anticompetitive effect. 212 Antitrust cases do not discuss "more profitable LRAs" in those terms, but one prominent proposal for addressing mixed conduct implements the search for more profitable LRAs. As introduced above, the "no economic sense" test singles out competition-restricting (usually exclusionary) conduct that is unprofitable, considering only the legitimate profits. 213 The two tests are identical: If the conduct is unprofitable and more restrictive, that is equivalent to the alternative being more profitable and less restrictive.

The usual interpretation of the no-economic-sense test is that it offers a conservative approach to liability for mixed conduct, at the acknowledged cost of some false negatives-a kind of no-regret shortcut. 211. The focus on profits and the implication that only some LRAs produce a reliable inference are absent in Professor Feldman's account. See Feldman, Misuse, supra note 72, at 624 (crediting LRA when defendant failed to use "most efficient" method).

212. This is subject to the familiar caveat that what looks like a sacrifice of profits might in fact be an innocent investment in future profits without any anticompetitive tendency.

213. See supra note 114 and accompanying text (discussing "no economic sense" test).

However, the test is not entirely conservative. It carries a possibility of false positives due to the distinction between profitability and welfare discussed above. Just because the alternative is more profitable does not guarantee that it is more procompetitive. The false positive arises when the conduct has lower legitimate profits-the basis for condemnationbut also higher procompetitive benefits thanks to benefits that are not captured by the firm. 214 The equal-profits limitation contains a subtle but important difference from the shortcut approach. As with the shortcut approach, not all LRAs count. For the shortcut to the LRA test, the alternative must be equally effective in serving a cognizable justification. By contrast, for smoking out, it is legitimate profits, not welfare effects, that matter. Legitimate profits might accrue from a procompetitive justification or instead result from a "neutral" business reason that is not cognizable. 215 One upshot is that an LRA that suffices for one purpose might fail for another. For example, an LRA that dominates a defendant's conduct, for purposes of a shortcut analysis, might fail to satisfy equal profitability for purposes of smoking out. 216 The smoking-out approach offers an alternative interpretation of the no-economic-sense test. Where the evidence of restrictiveness is ambiguous, the no-economic-sense test furnishes an inference that the conduct indeed has an anticompetitive effect. If the court observes conduct with a potential anticompetitive effect but the evidence is uncertain and it is possible to discern that the firm has sacrificed legitimate profits, then the defendant's choice suggests that there is indeed an anticompetitive effect.

3. Pretextual Justification. -The previous section examined an inference about anticompetitive effect that arises from some (but not all) LRAs. This section considers to what extent the LRA test sheds light on the second component of mixed conduct, the asserted procompetitive effect. In particular, can the test indicate whether an asserted justification is pretextual, as the Microsoft court put it, 217 in the sense of being absent? 218 If the procompetitive effect can be ruled out or minimized, the problem of assessing a tradeoff goes away. The main result, developed 214. For 218. Even if a procompetitive effect is present, one might argue that the justification is pretextual in the distinct sense (not pursued here) that it did not motivate the conduct. below, is that the firm's choices can again shed light as to asserted procompetitive conduct that is also profitable to the firm. However, the relevant alternative is a more profitable alternative rather than an LRA.

As an example, consider an early round in the FCC's ongoing effort to implement "network neutrality" regulation. 219 In 2008, the Commission reviewed allegations that Comcast had anticompetitively interfered with its Internet customers' efforts to use BitTorrent and other file-sharing applications. 220 In its order, the Commission concluded that Comcast had discriminated against, and thereby impeded, the applications. 221 Comcast justified its actions as "reasonable network management" that prevented congestion by large bandwidth users. 222 In rejecting these claims, the Commission noted a superior alternative that Comcast had failed to implement, which was to also curb other dataintensive applications in addition to file-sharing. 223 The superior alternative identified by the Commission was more profitable but equally restrictive. If Comcast imposed limits on all heavy users, BitTorrent would still be suppressed. The inference offered by a broader restriction is the inference of underinclusiveness. It suggests that the claimed benefit is an illusion, for if it were real, the defendant would have pursued the beneficial goal in a more thorough way.

The search for underinclusiveness plays an important role in constitutional scrutiny. Narrow tailoring, frequently invoked in a variety of constitutional contexts, is a combination of LRA and underinclusiveness tests. 224 Within narrow tailoring, underinclusiveness is used to reveal that a claimed justification is pretextual. 225 The court hypothesizes steps that the state could be expected to take to effectively serve its asserted purpose. The state's failure to take those steps supports the conclusion that the asserted purpose is not genuine. 226 This inference is valuable when it is the existence of the claimed benefit that is in doubt.

In antitrust, as in constitutional law, it is a more-beneficialalternatives analysis that smokes out a pretextual justification, not the LRA test. For example, in Comcast, the alternative is not less restrictive. It is equally restrictive. The key point of comparison is profitability, not restrictiveness. 227 And in fact, a less restrictive alternative usually fails to produce a similar inference of pretext. Even if the LRA is more profitable, the inference fails because the failure to pursue the benefit to the fullest extent has another explanation-that the defendant was sacrificing those benefits to pursue an anticompetitive effect. Such conduct is troubling, but not on the ground that the claimed benefit is an illusion. The possibility remains open that there is an anticompetitive effect but also a benefit, coexisting side by side. Hence the LRA test is ineffective in casting doubt on an asserted justification. 228 Smoking out pretext in antitrust law instead requires the recognition of a more-profitable-alternatives test. Today, there is no explicit doctrine that performs that function, just a few scattered hints. For example, in light of the foregoing analysis, Microsoft's reference to pretext might be read as an oblique call for such an analysis. Moreover, antitrust liability for selective refusals to deal is consistent with this approach. 229 A full examination would result in the adoption of a narrow tailoring doctrine for antitrust, both the LRA test and underinclusiveness. This approach is exemplified by the Comcast order, which explicitly adopted a narrow tailoring approach, insisting that the conduct "be narrowly or carefully tailored to serve that interest," 230 with both LRA and underinclusiveness components to the tailoring analysis. 231 Even without a full narrow tailoring inquiry, there is a second way to analyze pretext. Doing so requires a reconceptualization of the antitrust inquiry. In the cases we have examined so far, there are two offsetting effects that arise simultaneously. For example, the conduct both reduces competition for players' services and increases the value of the product 227. In principle, a more restrictive alternative would also serve this function, were it not for antitrust law. A decision not to choose a more restrictive alternative could be explained as an effort to avoid antitrust liability.

228. This is a further important difference from Professor Feldman's account. See Feldman, Misuse, supra note 72, at 624 (concluding failure to use most efficient method demonstrates "pretext"). As developed below, an LRA test can shed light on pretext under a particular additional condition-that the presence of an anticompetitive effect necessarily implies the absence of a procompetitive effect.

229. See, e.g., Otter Tail Power Co. v. United States, 410 U.S. 366, 370-71, 377-78 (1973) (concluding electric power company's selective refusal to serve some cities violated antitrust law).

230. Comcast Corp., 23 FCC Rcd. 13,028, 13,056 (2008); see also id. at 13,055 (requiring "tight fit between [Comcast's] chosen practices and a significant goal"); id. at 13,091 (McDowell, Comm'r, dissenting) (criticizing adoption of "' strict scrutiny' type standard").

231. See id. at 13,057-58 (opinion of the Commission) (identifying LRAs for managing network traffic); id. at 13,056-57 (finding underinclusiveness in light of more effective alternatives). sold to fans. The presence of both a bad effect and an offsetting good effect is clearly stated. The end of the analysis, if it is reached, is to sum up the two effects to determine the net effect.

In some cases, however, courts seem to approach mixed conduct in a different fashion-not by adding up the two effects but by choosing between the two parties' stories. Essentially, the court sees the stories as mutually exclusive. This move rests upon an additional assumption: a negative correlation between the probability of anticompetitive effect and probability of procompetitive effect. Thus, even though the LRA test does not directly shed light on pretext, it may indirectly do so by making use of this negative correlation. By reinforcing the evidence of anticompetitive effect, the test simultaneously undermines the procompetitive effect.

For example, in an exclusive-dealing case, there are often conflicting narratives about the contract. One anticompetitive story is that the contract denies a rival an outlet for its goods, restricting its opportunities with resulting harm to consumer welfare. One procompetitive story is that the contract protects the producer's investments in a retailer from interbrand free-riding-that is, from rival producers who would otherwise take advantage of the producer's investments. In practice, courts may choose between these stories rather than adding them up.

Beyond exclusive dealing, choosing rather than summing may characterize some courts' approach to resale price maintenance (RPM). The Supreme Court's leading RPM decision appears to take an either/or approach, 232 and some commentators approach RPM in either/or terms, 233 although in principle both effects could arise simultaneously. This perspective may also help explain the rarity of balancing in rule of reason and monopolization cases. In short, when an either/or frame is employed, the LRA test assumes heightened importance as a tool for smoking out.

The LRA test augments-and in some cases substitutes for-the work of the net-effects test. The test condemns some conduct even when the conduct has a positive net effect compared to inaction. Section III.A explains and defends this result, which is particularly pronounced if balanced LRAs are recognized as a baseline for condemning conduct. Section III.B considers the role of the LRA test if the net-effects analysis is omitted. Here the LRA test picks up some but not all of the slack.

A benchmark approach to the LRA test expands the set of conduct condemned by an antitrust inquiry. The LRA test imposes a second benchmark. Conduct may be condemned either because it is worse than the LRA, or because it is worse than inaction. Figure 4 Use of the dual benchmark expands the range of condemnation compared to an analysis that merely compares the conduct to inaction. To illustrate, consider point D. D would not be condemned by comparison to inaction; it is above the forty-five-degree line from O, which means that it is more effective than restrictive. However, it is still condemned by comparison to Z, as it is more restrictive and less effective than Z.

Such augmentation of the range of condemnation is a possibility whenever the LRA is used as a benchmark. 234 The augmentation effect is 234. The issue does not arise when the LRA is used as a diagnostic tool to smoke out an anticompetitive effect rather than as a benchmark for comparing costs and benefits.

to some degree obscured by the usual formulation of the antitrust inquiry, which first compares the conduct to an alternative, and then if necessary proceeds to a comparison between the conduct and inaction. 235 When a court performs an LRA analysis first, it does not yet know how the conduct compares to inaction and hence does not know whether the conduct would be condemned absent the LRA test.

The degree of augmentation varies depending on the range of accepted LRAs. A dominant LRA is sufficient to handle the condemnation of D in Figure 4. Recognizing balanced LRAs increases the degree of augmentation. If balanced LRAs are recognized, one would condemn conduct, such as point C, that is somewhat more effective and much more restrictive than Z. Point C would be permitted if the analysis is limited to dominant LRAs.

Critics have objected to augmentation. 236 They argue that antitrust merely requires that welfare must not fall overall by virtue of the conduct. Beyond that, firms are free to organize their activities as they see fit. They bristle at the proposition that a court might condemn conduct in light of the superiority of an unchosen alternative. They point to influential dicta in a refusal-to-deal case, in which the Court concluded that "[t]he Sherman Act is indeed the 'Magna Carta of free enterprise,' but it does not give judges carte blanche to insist that a monopolist alter its way of doing business whenever some other approach might yield greater competition." 237 This and other language in the opinion 238 have been read to imply a watered-down scrutiny of alternatives. 239 Critics of augmentation have not distinguished between dominant and balanced LRAs. Such critics would surely be even more opposed to the use of balanced LRAs given the expanded range of condemnation.

Augmentation, however, makes eminent sense. First, to insist on mere equipoise between the conduct and inaction ignores marginal 239. It bears noting that Trinko is not an analysis of mixed conduct, as the Court concluded that the defendant had not engaged in any cognizable anticompetitive conduct to begin with. Id. at 410 (concluding allegations of "insufficient assistance" fell short of any "recognized antitrust claim").

analysis. By way of analogy, suppose surgery is necessary to save a patient's life but is performed negligently, leaving the patient maimed. It is hardly a defense to note that the patient lived and that this outcome is better, all things considered, than if no surgery had been performed. We hold the defendant to a higher standard. If a small, inexpensive increase in care would yield a large improvement in outcomes, the defendant is properly held liable.

Second, if antitrust decisionmaking is ultimately bottomed on a costbenefit analysis of the restraint, then it is natural to examine a potentially more ambitious alternative in addition to the prerestraint status quo. An examination of such alternatives, after all, is a logical and standard part of cost-benefit analysis. 240 Third, public evaluation of private restraints is different from review of agency regulations in a way that strengthens the case for evaluating alternatives. There is no reason to expect a firm to choose the means that maximize net welfare. Far from it: The firm will choose the privately optimal course subject to whatever constraint is imposed by antitrust law. If antitrust law only insists upon equipoise, compared to inaction, then there is a powerful incentive to game the system. Private parties have substantial leeway to design their restraints in anticipation of antitrust scrutiny. If they merely need to ensure that consumers break even, they will take a minimalist approach if it is profitable to do so, draining consumer welfare down to the level that just meets the inaction baseline. 241 A dual-benchmark approach avoids this problem.

The previous section considered the LRA test as a second benchmark for anticompetitive conduct, in addition to a net-effects analysis. As noted above, other courts scrutinize mixed conduct by means of the LRA test alone, omitting any overall evaluation of the conduct. 242 240. See supra note 46 and accompanying text (discussing weighing of alternatives in cost-benefit analysis).

241. See Picker, supra note 75, at 1, 5-6 (arguing more lenient standard allows parties to avoid scrutiny provided overall effect is "just a bit better than the status quo").

If the defendant presents evidence of a justification, the plaintiff must establish an LRA or lose the case. This approach was apparently embraced by the O'Bannon court. The court's three-step analysis identified an anticompetitive effect (step one), then a procompetitive effect (step two). Upon plaintiff's failure to demonstrate an equally effective LRA (step three), the court found for defendants. 243 The NCAA did not escape liability entirely because the court recognized the slight increase in scholarship aid as an equally effective LRA. 244 Thus, the practical consequence was to sharply narrow the injunctive relief to the increased scholarship. In other cases (or a future NCAA case), the truncated analysis eliminates relief entirely.

Dropping the net-effects step opens up a significant gap in enforcement, particularly when combined with a limitation to dominant LRAs. To see this in the context of O'Bannon, consider a future challenge to the NCAA rule. The challenge seeks compensation for playing, rather than for use of a player's likeness. The NCAA has now implemented the slight increase in scholarship aid. Under the O'Bannon rule, the NCAA rule survives-even if the harm to the players is very large and the benefit for the fans is small-unless plaintiffs can demonstrate an equally effective LRA. 245 The court does not even try to approximate the relative size of the anticompetitive and procompetitive effects.

This enforcement gap can be understood by reference to Figure 4. If there is no LRA that substantially advances the justification, then mixed conduct gets no scrutiny at all. 246 Points A and B are permitted, even though they impose a net loss compared to inaction. 247 Moreover, a court that rejects net-effects balancing will surely insist that the LRA test be limited to dominant LRAs, since balanced LRAs present the same sorts of concerns as net-effects balancing. The limitation to dominant LRAs means that even when a dominant LRA is identified points B and C will be erroneously permitted. 248 Though a full defense of balancing is beyond the scope of this Article, it appears unlikely that the acknowledged difficulties justify the resulting enforcement gap. Even when the restraint is essential to the achievement of some procompetitive end, further analysis is appropriate. Reading between the lines, this appears to be the view embraced by the Supreme Court in horizontal agreement cases. 249 A closer question is presented by exclusion cases in which, for the reasons given in Part II, the difficulties with balancing are greatest. 250 A second form of truncation is to drop the LRA test, rather than the net-effects test. Some commentators have argued that the LRA test can be omitted in exclusion cases because a dominant firm's unilateral conduct is less suspicious. 251 That conclusion seems unwarranted, however, given the work of the previous steps in the analysis. Even if it is true in general that most conduct by a dominant firm is not suspicious, by the time a court reaches the LRA inquiry, it has already satisfied itself that there is an actual or likely anticompetitive effect. Establishing that first step is a difficult and indeed often insuperable hurdle. But once that threshold is crossed, the defendant no longer enjoys the benefit of the doubt.

Moreover, incomplete scrutiny of exclusionary conduct comes at a high cost. Exclusion of new entrants restricts the introduction of new products and services. Such innovation, and the accompanying creative destruction, is crucial to economic growth. 252 Conduct that slows or blocks innovation is ultimately more worrying than high prices from reduced competition among insiders and more resistant to selfcorrection by the market. 253 As a consequence, exclusion, particularly in 248. The LRA test still augments the range of condemned conduct. To see this, note that, in addition to A, D is condemned.

249. For example, in American Needle, Inc. v. NFL, the Court considered a sports league joint venture, opining in dicta that even if the restraint is "essential" to the benefit, further analysis would be required. 560 U.S. 183, 203 (2010) (stating such restraints would "likely" survive subsequent application of rule of reason (citing BMI v. CBS, Inc., 441 U.S. innovative industries, raises a particularly high cost of false negatives. For both reasons, a more thorough evaluation of the justification is justified.

This Part identifies best practices that reduce the risk of false positives in applying the LRA test. First, plaintiffs properly bear the burden of persuasion in establishing an LRA, not only in rule-of-reason cases but also in horizontal mergers and tying cases (section IV.A). Second, the LRA must be preferred by the defendant to inaction (section IV.B). Third, the LRA must be practical and plausibly chosen by the defendant (section IV.C).

These best practices each take on heightened importance if the LRA test includes balanced alternatives. With only a little "imagination," one can usually identify an LRA. 254 If LRAs are not restricted to equal effectiveness, the scope for imaginative alternatives is substantially increased. These best practices place significant real-world limits on the scope of the test.

The burden of persuasion is a crucial issue in antitrust cases. Imposing upon defendants a burden to establish the absence of an LRA places a potentially heavy thumb on the scales in favor of plaintiffs. Indeed, the allocation to defendants in constitutional cases 255 may be a decisive disadvantage. In antitrust cases, the burden of persuasion is an inconsistent and poorly theorized mix that varies across jurisdictions and types of case. The majority of courts applying the rule of reason allocate the burden to plaintiffs, 256 though some allocate the burden to [Vol. 116:927 defendants. 257 By contrast, in horizontal merger 258 and tying 259 cases, defendants typically have the burden.

When the LRA test is used as a benchmark for evaluating conduct, plaintiffs properly bear the burden of persuasion. At heart, the plaintiff is arguing that the net benefit of the alternative is positive compared to the conduct, as a means of establishing the anticompetitive effect of the conduct. This claim is conceptually analogous to the comparison between the conduct and inaction. Whether premised on one benchmark or the other, the plaintiff is attempting to establish an unreasonable restraint of trade. A difference in treatment between the two benchmarks is hard to justify.

Moreover, to the extent the test recognizes balanced alternatives, the case for a burden on plaintiffs is strengthened. Given that most LRAs are less effective, their superiority must be established rather than presumed. Moreover, the fuller comparison of both benefits and costs strengthens the identity between the LRA inquiry and net-effects balancing.

A further reason to assign plaintiffs this burden is that otherwise defendants are forced to prove a negative. This poses a practical difficulty where a potentially wide range of LRAs might be posed in response to the defendant's conduct. The difficulty of proving a negative strengthens the case for placing the burden on plaintiffs. That is especially true if the LRA is a balanced LRA, the asserted net benefit of which may be harder ("An asserted business justification cannot salvage a tying arrangement that is otherwise per se unlawful without proof that means less restrictive than the tie-in were not feasible to achieve the desired protection.").

LESS RESTRICTIVE ALTERNATIVES to disprove. The strength of this argument, however, must not be overstated. Even if defendant had the burden of persuasion, the plaintiff could be assigned a burden of pleading and limited to one or several "standard" alternatives, as discussed in more detail below.

The plaintiff's burden properly extends to a horizontal merger, which raises some similar concerns as horizontal agreement, such as coordinated pricing and, like other forms of agreement, is a "contract, combination, . . . or conspiracy" covered by section 1 of the Sherman Act. 260 Here, the traditional allocation to defendants has had an important effect. For example, the St. Luke's court insisted that the merging medical providers prove the absence of an LRA. 261 The allocation of burden here was likely decisive to the result. Given the limited state of current knowledge about what is needed for successful integration, 262 it was very difficult for defendants to show that their merger was a superior vehicle for delivering the beneficial effects.

The foregoing analysis suggests that the St. Luke's court applied the wrong rule and that plaintiffs ought to have borne the burden of establishing the comparison between the conduct and the LRA benchmark. This conclusion holds even if the governing statute, the Clayton Act, establishes a prophylactic goal and even if it weighs false negatives especially highly. Such a goal does not require that defendants bear the burden of persuasion as to an LRA. Sensitivity to false negatives is already reflected in other elements of doctrine, particularly the presumption of anticompetitive effect from increased concentration and the skepticism with which courts evaluate claims of efficiencies. 264 Tying, like horizontal mergers, places a thumb on the scales for plaintiffs. The reason is different: a judicial belief that most tying is anticompetitive and that usually there exist LRAs. 265 Even if the first belief were correct-a conclusion that is sharply contested 266 -the [Vol. 116:927 playing field is already tilted in favor of plaintiffs by dispensing with the need to show substantial foreclosure as to the tied product. 267 As to the second belief, the LRAs are generally less effective. 268 Thus, once again, overall superiority is something to be established, rather than presumed, suggesting that plaintiffs ought to have the burden of persuasion here, too.

If plaintiffs have the burden of persuasion, defendants ought to bear a burden of production. 269 Defendants have better access to information about their reasons for adopting a particular practice. 270 Giving defendants a burden of production is not unduly onerous. The burden is lightened by requiring the plaintiff to propose the set of alternatives to which the defendant responds, and it is further constrained by the "practicality" requirement discussed below. 271 Giving defendants a limited justificatory burden has additional benefits. A firm that may later be required to explain its actions is more likely to create a suitable record at the time that it implements the conduct. A burden of production is therefore likely to induce more careful deliberation by the firm before taking its action. 272 At least some anticompetitive conduct is the product of inattention by the firm. Deliberation is frequently in short supply, and often firms may have no considered reason for having taken an action with anticompetitive consequences. Encouraging a culture of justification and reason giving, a familiar part of constitutional discussion, is valuable in antitrust as a source of deterrence. 273 The plaintiff's burden of persuasion is appropriate when the LRA test is used as a benchmark, whether as a shortcut or as a locus of balancing. By contrast, it does not arise for LRA-as-smoking-out, for in that instance it is an input to an evaluation of effects, rather than a benchmark. Instead, plaintiffs bear a burden of persuasion on anticompetitive effect, and the LRA test is one piece of evidence among many that bears on that effect.

A further exception arises when the LRA test is used to reveal a lack of ancillarity. As discussed in Part II, when conduct makes no incremental contribution to a claimed beneficial end, cessation is an LRA. Another way to say this is that there is no benefit in the first place, a point on which the defendant properly bears the burden of persuasion.

Section III.A defended the augmentation that arises from use of the LRA test. However, augmentation can produce perverse outcomes if courts are not careful. This issue potentially arises where the conduct is already superior to inaction. In such cases, the LRA condemns the conduct with a view to further welfare improvement.

For condemnation to have the desired effect, the LRA must be preferable to the firm compared to inaction. If not, as Professor Richard Craswell has noted in the context of deceptive advertising, a perverse outcome may result. 274 A firm that anticipates condemnation for its conduct will just stick with inaction, rather than choosing the LRA. Condemning the conduct imposes a social loss, the difference between the conduct and the inferior results of inaction. 275 This perversity problem has received little attention.

One important setting for the perversity problem arises when the defendant depends on the anticompetitive effect to make the conduct profitable. Suppose the firm would not enter, but for the opportunity to raise prices through collusive and otherwise illegal conduct, or through conduct that raised prices though it was not the point of the conduct. For example, in the Apple case, the Justice Department accused Apple of taking part in a conspiracy to raise ebook retail prices. 276 Apple defended on the ground (among others) that its entry into ebook retailing brought important consumer benefits that would otherwise not have been legitimacy of that objective; it may also suggest that alternative, less injurious means are available.").

274. See Richard Craswell, "Compared to What?" The Use of Control Ads in Deceptive Advertising Litigation, 65 Antitrust L.J. 757, 779 (1997) ("[I]f the advertiser were barred from using the original ad . . . and was instead limited to using either [the ideal solution from the FTC's standpoint] or no ad at all, the advertiser might choose no ad at all.").

275. Courts cannot simply order the LRA as preferable to inaction. Among other problems, there is no conduct on which to premise the antitrust violation.

276. United States v. Apple, Inc., 791 F.3d 290, 296 (2d Cir. 2015). [Vol. 116:927 available and that it would not have entered but for the conduct that resulted in higher prices. 277 Here, the benefit is premised on the harm itself. The LRA test is not easily applied in this instance. It is not enough to simply say there exists an LRA with higher welfare, namely entry without collusion. If the firm would not have entered on those terms, the LRA is off the table. That is not to say the conduct is permissible. The conduct may still be condemned, either because a benefit premised on collusion is not cognizablefoot_2 or because the cognizable benefit is smaller than the loss. But courts cannot use the LRA test to avoid that analysis.

Crediting balanced LRAs exacerbates concerns about the perversity problem. It does so, in part, simply by expanding the range of credited LRAs. Beyond that, less effective alternatives-the set of alternatives added by expanding the range of LRAs-are particularly apt to raise the perversity issue. That is because less effective alternatives are often (though not always, it must be stressed) less profitable. Thus a disproportionate share of balanced LRAs are likely ruled out by the zeroprofit restriction. Consequently, this restriction sets a potentially important limit on the set of balanced LRAs available to condemn mixed conduct.

The solution is to credit only those LRAs that meet the defendant's participation constraint. For example, we might insist that the plaintiff furnish evidence, including expert testimony based on economic theory, that the LRA would likely be chosen if the conduct is condemned. The defendant's own asserted justification may tend to show that the LRA is profitable.

It is worth pausing to consider whether this additional requirement is overkill. Perhaps the perversity problem is so rare, small when it occurs, or costly and error prone to calculate that courts should ignore the complication. To answer that question, courts need better information, currently lacking, about two questions-first, how often scrutinized conduct is superior to inaction. To the extent it is inferior, there is no undue deterrence concern; deterring the conduct is desirable. The second question is how often identified LRAs are unprofitable compared to inaction. Certainly the LRA is (weakly) less profitable compared to the conduct; otherwise the defendant would have chosen it. The lesser restrictiveness will ordinarily reduce or remove the illicit profits arising from an anticompetitive effect. The legitimate profits may be lower as well. Even a dominant LRA may deliver lower legitimate profits to the defendant, as explained in section II.D. These considerations suggest that the perversity problem is not merely theoretical, though further work is required to establish its size.

An idea developed in the previous section-that the LRA test is properly limited to alternatives a defendant might actually pursue-has a further implication. The LRA ought to be practically rooted in real commercial experience. Some courts have operationalized this idea by requiring that the posited alternative must be "based on actual experience" or else "fairly obvious." 279 A rigorous test of practicality is needed both to avoid the perversity problem discussed above and to avoid an overestimate of the benefit from conduct whose speculative basis makes it hard to evaluate.

One useful indicia of practicality is that the alternative has been implemented by this or other firms in similar circumstances. For example, a professional sports league's internal restrictions might be evaluated in light of the alternatives implemented by other sports leagues. 280 Similarly, certain types of standard contractual provisions have clear real-world alternatives. For example, if a retailer implements resale price maintenance, asserting the avoidance of intrabrand free-riding as a justification for resale price maintenance, the alternative of a more detailed contract may be ready at hand.

Practicality is harder to establish when the conduct itself is more unusual and less standardized. Where the alternative has never been implemented in a related context, there is less of a basis for confidence in its effectiveness. Judged from the perspective of practicality, the O'Bannon district court erred in embracing a highly nonstandard alternative to the NCAA's rule. The deferred compensation alternative was an ingenious response to the justifications credited by the court. However, this LRA lacked the indicia of reliability based on previous experience. As a consequence, it was less clear how well the alternative would perform, raising the expected error costs of its evaluation.

Practicality also poses a challenge where alleged exclusion is accomplished through product design. In some cases alleging predatory product design, the proposed alternative is a redesign of the product.

The LRA test plays several roles in antitrust cases. Sometimes a dominant LRA offers a valid shortcut that avoids the need to weigh incremental restriction against incremental effectiveness. In other cases, a less effective alternative is a locus for balancing. And sometimes it may serve not as a benchmark but as a diagnostic tool that smokes out anticompetitive effect.

While the LRA test offers a searching inquiry of mixed conduct, it is subject to significant limitations. The posited alternative must be profitable to the defendant, compared to inaction, and practical rather than speculative. The plaintiff properly has the burden of persuasion, even for conduct such as mergers and tying, where the burden has traditionally been assigned to defendants.

Several features of the LRA test apply broadly beyond antitrust. For example, expansion of the range of condemnation results whenever the LRA test offers a second baseline beyond inaction, whether in antitrust or constitutional law. 287 Moreover, balancing against alternatives is inevitable wherever LRAs arise. For example, the limitation to dominant LRAs in constitutional scrutiny is breached there, just as in antitrust law. 288 At the same time, a few courts have openly acknowledged the role of balanced LRAs in constitutional scrutiny, making possible a fuller comparison of the alternative's incremental benefits and costs. 289 Such frankness about balancing-and its inevitably-might profitably be brought to both constitutional law and antitrust law.