THE LOGIC OF MARKET DEFINITION
For more than a half century, antitrust trials have usually begun with the
definition of a relevant market for the inquiry. Long experience has given this
exercise an air of familiarity, but closer examination reveals market definition
to be a confused exercise. Decades ago, Robert Pitofsky remarked that “no
aspect of antitrust enforcement has been handled nearly as badly as market
That sentiment remains frustratingly apt today.
Despite its long
tenure in antitrust analysis, and despite the crucial role it has played in many a
case and investigation, the process of defining relevant markets remains both
confused and uncertain.
Why do we define markets? How should we define them? One might think
that such fundamental questions would have long been settled. But the some-
times unclear rationale for the exercise, and its inconsistent evolution in the
* David Glasner is an economist at the Federal Trade Commission, Bureau of Economics.
Sean Sullivan is associate professor of law at the University of Iowa College of Law and an
Associate Editor of the Antitrust Law Journal. This article reflects the personal views of the
authors. It does not necessarily represent the views of the U.S. Government or Federal Trade
Commission, which have neither approved nor disapproved its content. The authors thank Chris-
tine Bartholomew, William Bishop, Aaron Edlin, Joshua Goodman, John Kirkwood, Christopher
Leslie, Matt Ralph, Daniel Sokol, Spencer Weber Waller, Gregory Werden, and conference par-
ticipants at the 2019 meeting of the American Law and Economics Association for thoughtful
comments on earlier drafts of this article. Alexander Asawa, Kassandra DiPietro, Reid Shepard,
and Madison Tallant provided superlative research assistance.
Robert Pitofsky, New Definitions of Relevant Market and the Assault on Antitrust, 90
. L. R
. 1805, 1807 (1990); see also Donald F. Turner, The Role of the “Market Con-
cept” in Antitrust Law, 49 A
L.J. 1145, 1150 (1980) (“Let me turn now to what some of
the current problems are with market definition. I have to say at the outset that as a general
matter this whole area is a bloody mess.”).
See, e.g., Louis Kaplow, Why (Ever) Define Markets?, 124 H
. L. R
. 437, 466 (2010)
(“[T]here is no canonical, operational statement of the standard for determining what constitutes
a relevant market and, a fortiori, no developed underlying rationalization for whatever the princi-
ple might be.”); William Blumenthal, Why Bother?: On Market Definition Under the Merger
Guidelines, Statement before the FTC/DOJ Merger Enforcement Workshop 2 (Feb. 17, 2004),
www.justice.gov/sites/default/files/atr/legacy/2007/08/30/202600.pdf (“[T]he meaning of ‘rele-
vant market’ today . . . probably is not understood by more than 500 people on the planet.”).
83 Antitrust Law Journal No. 2 (2020). Copyright 2020 American Bar Association. Reproduced
by permission. All rights reserved. This information or any portion thereof may not be copied
or disseminated in any form or by any means or downloaded or stored in an electronic
database or retrieval system without the express written consent of the American Bar
courts and scholarship, have not produced a straightforward set of criteria by
which to assess the validity of relevant markets in antitrust. Even the term,
market definition, is more ambiguous than it first appears. Does it refer to the
identification of popularly recognized lines of commerce or products with
similar characteristics? Does it refer to products with high cross-elasticity of
demand? Does it refer to things like the Hypothetical Monopolist Test (HMT)
and efforts to identify groups of producers with potential market power?
That, today, these are all potential answers, is both remarkable and unsettling.
In this article, we hope to cut through some of the ambiguity and confusion
surrounding market definition. Our goal is to trace the internal logic of the
exercise, identifying common errors and showing how the logic of market
definition can focus and guide antitrust inquiries. While we are mainly con-
cerned with how markets should be defined in antitrust, pragmatism requires
us to pause to say why we should aim for proper market definition as well.
The need for pause is the sometimes-popular claim that market definition is
unnecessary in antitrust law. While this argument is not new,
has advanced the thesis with a particularly pointed argument that (1) market
definition serves no role except to facilitate computing market shares, (2) mar-
ket shares are poor measures of market power, and (3) antitrust goals would
be better served by assessing market power from things like estimates of
residual-demand curves than by computing market shares.
This argument is
not without its strengths, and there are cases in which the traditional market
definition exercise can be skipped without adversely affecting the outcome of
the investigation or trial.
This list does not purport to exhaust the range of possibilities. See, e.g., Kenneth G. Elzinga
& Thomas F. Hogarty, The Problem of Geographical Market Delineation in Antimerger Suits, 18
. 45 (1973) (defining markets based on consumer flow information); Mario
Forni, Using Stationarity Tests in Antitrust Market Definition, 6 A
. L. & E
(2004) (defining markets based on price stationarity); Ira Horowitz, Market Definition in Anti-
trust Analysis: A Regression-Based Approach, 48 S. E
. J. 1 (1981) (defining markets based
on price movements); George J. Stigler & Robert A. Sherwin, The Extent of the Market, 28 J.L.
. 555 (1985) (defining markets based on empirical similarity of price movements).
E.g., Blumenthal, supra note 2, at 1 (“Worse than unnecessary, any effort formally to define
markets would [be] unduly costly, time-consuming, and invasive, and it probably would [yield]
less reliable outcomes than more streamlined techniques.”); Frank H. Easterbrook, The Limits of
Antitrust, 63 T
. L. R
. 1, 22 (1984) (“Market definition is just a tool in the investigation of
market power; it is an output of antitrust inquiry rather than an input into decisions, and it should
be avoided whenever possible.”); id. (“An inquiry into power does not entail the definition of a
‘market,’ a subject that has bedeviled the law of mergers.”).
See generally Kaplow, supra note 2; Louis Kaplow, Market Definition and the Merger
Guidelines, 39 R
. 107 (2011); Louis Kaplow, Market Definition Alchemy, 57
. 915 (2012) [hereinafter Kaplow, Alchemy]; Louis Kaplow, Market Definition:
Impossible and Counterproductive, 79 A
L.J. 361 (2013) [hereinafter Kaplow,
But even if traditional market definition is not necessary in every antitrust
we believe that courts and practitioners must still understand how to
properly define and interpret antitrust relevant markets in practice. There are
three reasons for this.
First, the claim that market definition can be entirely replaced by things like
econometric estimates of residual demand curves is doubtful, to say the least.
It is difficult, for example, to imagine courts and practitioners analyzing ease
of entry without a market concept. What exactly would firms be entering?
Similar difficulties beset efforts to assess the danger of anticompetitive coor-
dination without some idea of which firms would need to cooperate for their
coordinated action to be able to raise prices.
And while estimates of residual
demand elasticity may often suffice to establish current or historic market
power, they are not generally sufficient to predict future competitive effects—
as needed, for example, in cases involving unconsummated mergers or pro-
spective acts of exclusion.
In such situations, antitrust analysis is advanced
by defining relevant markets.
See infra Part IV.A (discussing cases where market definition is not necessary).
See generally Duncan Cameron, Mark Glick, & David Mangum, Good Riddance to Market
Definition?, 57 A
. 719 (2012); Malcolm B. Coate & Joseph J. Simons, In Defense
of Market Definition, 57 A
. 667 (2012); Gregory Werden, The Relevant Market:
Possible and Productive, A
(Apr. 2014) [hereinafter Werden, Possible],
online-pdf.pdf; Gregory Werden, Why (Ever) Define Markets? An Answer to Professor Kaplow,
L.J. 729 (2013) [hereinafter Werden, Answer].
See, e.g., Franklin M. Fisher, Economic Analysis and “Bright-Line” Tests, 4 J. C
L. & E
. 129, 131 (2008) (“Ease of entry must also be considered, and one might reason-
ably say that such a consideration requires one to know what it is that is being entered.”);
Werden, Answer, supra note 7, at 729 (“Even if antitrust analysis never used market shares, the
relevant market would remain essential for examining entry prospects and the durability of mar-
ket power.”). But see Louis Kaplow & Carl Shapiro, Antitrust, in 2 H
1073, 1185–86 (A. Mitchell Polinsky & Steven Shavell eds., 2007) (suggesting
some ways to analyze exclusionary conduct in terms of elasticities); Kaplow, Impossible, supra
note 5, at 363 n.3 (suggesting that potential entry analysis is similar to exclusionary conduct
See, e.g., Herbert Hovenkamp & Carl Shapiro, Horizontal Mergers, Market Structure, and
Burdens of Proof, 127 Y
L.J. 1996, 2000 (2018) (“In cases in which the government alleges
coordinated effects, the role of market definition and concentration measures such as the HHI is
much more fundamental.”); Werden, Answer, supra note 7, at 739 (“[Coordinated effects analy-
sis] uses the relevant market to determine how many, and which, competitors most likely would
be involved in the coordination.”).
See, e.g., 2B P
¶ 531d (4th ed.
2014) (commenting that assessment of current market power is insufficient to address concerns
about future market power); Phillip Areeda, Market Definition and Horizontal Restraints, 52
L.J. 553, 555 (1983) (“[Past] performance data cannot reveal unexercised power. . . .
Thus, performance data is not relevant for determining whether a new merger creates new
power.”); Gregory J. Werden, Market Delineation and the Justice Department’s Merger Guide-
lines, 1983 D
L.J. 514, 515 (1983) (“[A]pplication of [Section 7] requires predictions about
the effects on competition of changes in market structure.”). Kaplow’s best effort to extend his
approach to merger analysis rests on the assumption that the merger is between producers of a
Second, regardless of the academic debate, courts have long relied on mar-
ket definition in antitrust cases,
and the Supreme Court shows no indication
that it will abandon this practice soon. On the contrary, the Court has recently
reaffirmed its view that “courts usually cannot properly apply the rule of rea-
son without an accurate definition of the relevant market.”
So long as bind-
ing precedent continues to expect the definition of relevant markets in most
applications, lower courts and practitioners need to understand the logic and
proper execution of the market definition exercise.
Third, despite Kaplow’s insistence that market definition serves no purpose
other than to facilitate the calculation of market shares,
others perceive it to
play additional roles. During investigational stages—in the review of merger
notifications, for example—market definition is meant to clarify analysis by
imposing analytic discipline on investigators,
by providing a logical way to
homogeneous (undifferentiated) product. Kaplow, Impossible, supra note 5, at 370–71. Problem-
atically for this approach, it is unclear how courts and practitioners are supposed to identify a
homogeneous product; the matter normally arises within the context of market definition.
See, e.g., Jonathan Baker, Market Definition: An Analytical Overview, 74 A
129, 129 (2007) (“Market definition is often the most critical step in evaluating market power
and determining whether business conduct has or likely will have anticompetitive effects.”);
Fisher, supra note 8, at 130 (“Market definition has become a necessary part of every antitrust
case, and there is no avoiding discussing it.”); Dennis W. Carlton, Comment on Department of
Justice and Federal Trade Commission’s Proposed Horizontal Merger Guidelines 3 (June 4,
.pdf (“Any suggestion that the courts should abandon the use of market definition when analyz-
ing the competitive effects of mergers is unwise, as the failure to define markets would likely
increase the number of erroneous decisions reached by courts.”).
Ohio v. Am. Express Co., 138 S. Ct. 2274, 2285 (2018); see also id. (“Without a definition
of [the] market there is no way to measure [the defendant’s] ability to lessen or destroy competi-
tion.” (alterations in original)).
E.g., Kaplow, Impossible, supra note 5, at 363 (claiming the only point of market definition
is to “make market power inferences from market shares”); id. at n.3 (defending the prior claim
with the statement that “I am skeptical that market definition is useful for other purposes”).
See, e.g., Dennis W. Carlton, Revising the Horizontal Merger Guidelines, 6 J. C
L. & E
. 619, 626 (2010) (“The discipline of forcing decision-makers to have a reasonable
market definition in mind . . . is likely to be valuable in constraining agencies and especially
courts from making decisions based on arbitrary criteria.”); Turner, supra note 1, at 1145 (“[One
role of market definition is] to provide some sort of rational economic basis for assessing the
consequence of the particular kind of conduct that is involved in the antitrust case.”); Robert
Willig, Public Comments on the 2010 Draft Horizontal Merger Guidelines 2 (2010), www
purpose behind a requirement of market definition . . . is the imperative for disciplined consider-
ation of sources of competition beyond the parties’ own products, along with the need to gener-
ate a consistent calibration of the strength of that additional competition.”).
by helping to screen out implausible theories,
focusing the scope of competitive effects analysis.
stages—in court or before the agencies—market definition supports structural
inferences about competitive effects,
provides context for relevant eviden-
tiary considerations such as the possibility of entry or exit,
and again pro-
vides a conceptual framework to guide and discipline analysis.
discuss below, market definition may currently play different roles within the
agencies and the courts.
But in both contexts the exercise is meant to serve
the common goal of identifying conduct and structural conditions that raise
concerns about anticompetitive injury, and that therefore require scrutiny.
We do not challenge the consensus that market definition serves broad pur-
poses, but we suspect that this breadth of use may actually be a source of
some confusion. Common platitudes only reinforce the problem: the Supreme
Court does not mislead when it says that “the purpose of [market definition] is
to determine whether an arrangement has the potential for genuine adverse
effects on competition,”
but neither does it take anything off the table. One
See, e.g., L
ANDBOOK OF THE
(1977) (“[T]he only purpose for defining a market is to organize available data in a way which
facilitates judgment about the extent of that power.”); Fisher, supra note 8, at 130 (“Market
definition can be a useful tool, a way to begin organizing the material that must be studied.”).
See, e.g., Easterbrook, supra note 4, at 14–19.
See, e.g., Werden, Answer, supra note 7, at 739 (“The relevant market furthers the analysis
by separating the active forces of competition from forces properly treated as part of the back-
ground.”); see also R
85 (1962) (“[One purpose of defining a market in economic analysis is] that of delineat-
ing practical boundaries for any given inquiry, in order to narrow down to essentials the empiri-
cal points to be investigated.”).
See, e.g., Hovenkamp & Shapiro, supra note 9, at 2006 (“[E]conomic theory and a wide
range of economic evidence support the conclusion that horizontal mergers that significantly
increase market concentration are likely to lessen competition and harm consumers.”); Sean P.
Sullivan, What Structural Presumption?: Reuniting Evidence and Economics on the Role of Mar-
ket Concentration in Horizontal Merger Analysis, 42 J. C
. L. 403, 409–10, 426–28, 433
(2016) (discussing the probative value of market concentration evidence in predicting the com-
petitive effects of horizontal mergers).
See supra notes 7–10 and accompanying text.
See, e.g., Carlton, supra note 14, at 637 (“[E]ven though market definition may be a crude
tool to use, it does provide some structure to an antitrust analysis and its use likely prevents
courts from making egregious errors.”).
See infra note 227 and accompanying text (discussing differences in how merger review is
conducted in courts and before the federal agencies).
See, e.g., 2B A
, supra note 10, ¶ 531 (“Finding the relevant market
and its structure is typically not a goal in itself but a mechanism for considering the plausibility
of antitrust claims that the defendants’ business conduct will create, enlarge, or prolong market
power.”); Christine A. Varney, The 2010 Horizontal Merger Guidelines: Evolution, Not Revolu-
tion, 77 A
L.J. 651, 653 (2011) (“[Flexibility in market definition] flows from the pur-
pose of defining markets—helping to assess a merger’s potential to harm consumers.”).
FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460 (1986).
reason the logic of market definition remains obscure is that so little effort has
ever been devoted to specifying what should not factor into the exercise.
This article aims to fill that void. Our objective is to clarify the logic of
antitrust market definition; our strategy is to illustrate this logic by way of
exclusion. The explanation of what should factor into market definition is
hardened by the explanation of what should not. The article therefore focuses
on three common fallacies of antitrust market definition.
The first is what we call the natural market fallacy: the belief that relevant
markets should conform to intuition, convention, or observation. The reason
they should not is that markets are not real, observable, tangible things that
can be perceived and identified by simple observation. Markets are analytical
constructs without corporeal presence or tangible form. Recognition of this
fallacy leads to the conclusion that many traditional market-definition criteria
are inappropriate and also highlights that many common objections to market
definition—for example, that markets are unrealistic or gerrymandered unless
they conform to lay intuition or common description—are without merit.
The second is what we call the independent market fallacy: the belief that
relevant antitrust markets exist independent of underlying theories of harm.
The reason they do not lies in the prior fallacy. As purely analytical con-
structs, markets do not exist independent of a problem or inquiry but must be
defined in terms of a problem or inquiry. Specifically, antitrust markets are
defined in terms of specific theories of anticompetitive harm. Recognition of
this fallacy leads to the conclusion that a relevant market cannot be defined
without an anticompetitive theory of harm upon which the relevant market is
conditioned. Enduring confusion around the base price in HMT markets,
even the infamous Cellophane fallacy itself,
are illustrations of this error.
Cf. U.S. Dep’t of Justice & Fed. Trade Comm’n, Horizontal Merger Guidelines § 4.1.2
(2010) [hereinafter 2010 Merger Guidelines], www.justice.gov/sites/default/files/atr/legacy/
2010/08/19/hmg-2010.pdf (describing the typical baseline price against which the HMT markets
are constructed in horizontal merger cases).
See, e.g., R
150–51 (2d ed. 2001) (providing the mod-
ern textbook treatment of the Cellophane fallacy); George W. Stocking, Economic Tests of Mo-
nopoly and the Concept of the Relevant Market, 2 A
. 479 (1957) (providing
perhaps the earliest clear articulation of the Court’s error in the Cellophane case); see infra notes
113–124 and accompanying text (discussing the Cellophane fallacy).
Many aspects of this fallacy have been previously identified and articulated by Steven
Salop. See generally Steven C. Salop, The First Principles Approach to Antitrust, Kodak, and
Antitrust at the Millennium, 68 A
L.J. 187 (2000). Our approach adds to Salop’s first
principles approach in two respects. First, we extend the argument by showing not just the desir-
ability, but the logical necessity of defining markets by reference to specific theories of competi-
tive injury. Second, we show how the theory-dependence of market definition ties in with the
comprehensive logic of the market definition exercise.
The third is what we call the single market fallacy: the expectation that an
antitrust case must involve a unique relevant market (or set of markets if there
are multiple products). The reason not is implicit in the prior fallacies. Since a
relevant market must be conditioned on a specific theory of harm, and since
multiple theories of harm may be implicated by a given fact pattern, there may
be multiple relevant markets to delineate in analyzing the competitive effects
of the conduct in question. Recognition of this fallacy leads to the conclusion
that the number of relevant markets in a given case or investigation may be as
great as the number of theories of anticompetitive harm.
The remainder of this article explores each of these fallacies separately and
then together in a discussion section that sketches some corollary implications
for antitrust practice. Many of our suggestions about market definition can be
extracted—to varying degrees of precision—from the text of the 2010 Hori-
zontal Merger Guidelines
and from various remarks, asides, and footnotes
scattered throughout the literature on market definition.
But the diffusion of
insights through subtle implication and easily overlooked margin notes is an
inefficient way to explain a concept as challenging as market definition. And
the enduring confusion surrounding this exercise suggests that further exposi-
tion may be warranted.
I. THE NATURAL MARKET FALLACY
The natural market fallacy is the mistaken belief that the boundaries of
relevant markets should conform to lay intuition, conventional language, or
direct observation. The error flows from a false assumption that markets are
tangible objects. They are not. Markets are analytical constructs with no nec-
essary relation to industry practices or popular conceptions of trade lines.
If this seems obvious, consider two themes that have long held sway in
discussions of market definition. The first is that market definition should
prevent courts and plaintiffs from delineating artificially narrow markets.
See generally 2010 Merger Guidelines, supra note 24.
See generally S
, supra note 15, at 41–74; T
, supra note 17, at 78–96; Baker,
supra note 11; Jonathan Baker, Stepping Out in an Old Brown Shoe: In Qualified Praise of
Submarkets, 68 A
L.J. 203 (2000); Hovenkamp & Shapiro, supra note 9; Werden, An-
swer, supra note 7; Gregory J. Werden, The 1982 Merger Guidelines and the Ascent of the
Hypothetical Monopolist Paradigm, 71 A
L.J. 253 (2003). Salop, supra note 26, dis-
cusses some of our points about the independent market fallacy in detail.
This concern has endured over a span of decades. Compare United States v. Mfrs. Hanover
Tr. Co., 240 F. Supp. 867, 918 (S.D.N.Y. 1965) (“[T]he government cannot gerrymander the
market any way it chooses.”), with Smalley & Co. v. Emerson & Cuming, Inc., 808 F. Supp.
1503, 1512 (D. Colo. 1992), aff’d, 13 F.3d 366 (10th Cir. 1993) (“Plaintiff cannot artificially
create antitrust claims by narrowly defining the relevant market to create the appearance of an
antitrust injury.”), and United States v. Anthem, Inc., 236 F. Supp. 3d 171, 202 (D.D.C. 2017)
(“The defense contends that [the plaintiff’s] proposed market is ‘gerrymander[ed]’ and ‘lacks
The second is that courts may fail to identify the correct market for analysis.
Each reflects a similar understanding of market definition. To say a market is
artificial is to say it is not a natural market; to say a market is defined incor-
rectly is to presume that the market could have been defined correctly. Both
notions betray the influence of the natural market fallacy on market definition
We now trace the natural market fallacy through case law and academic
commentary, showing how it hinders antitrust analysis. It will be helpful to
distinguish two types of natural-market concepts: those that define markets by
the observable characteristics of products and producers, and those that define
markets by observed substitutability. Both concepts are problematic. Markets
are nothing but mental constructs that facilitate the economic analysis of spe-
cific problems. Rejecting possible abstractions because they do not match an
imagined set of natural-market boundaries can only impede that analysis.
Of the many ways that the Supreme Court has tried to articulate standards
for defining markets, the most obviously naturalistic are those equating mar-
ket definition with the identification of observable product characteristics and
lay recognition of industry lines. At a high level, this approach seeks charac-
teristics of products and producers that indicate how a market should be cate-
gorized—almost as a biologist might try to identify an insect by comparing its
characteristics to those of known exemplars.
The leading authority for defining markets around the distinguishing char-
acteristics of products is du Pont-General Motors, a merger case in which the
Supreme Court distinguished a narrow relevant market of “automotive fin-
ishes and fabrics” from a broader class of similar industrial finishes and
fabrics. The Court based this market definition on its conclusion that “auto-
motive finishes and fabrics have sufficient peculiar characteristics and uses
See, e.g., Morris A. Adelman, The Antimerger Act, 1950–60, 51 A
. 236, 237
(1961) (“It is a pathetic illusion that the market is whatever the courts choose to call it. The
market, like the weather, is simply there, whether we only talk about it or do something: apply to
it the standards of Clayton, or of Sherman, or of any law, or none.”); Herbert Hovenkamp,
Markets in Merger Analysis, 50 A
. 887, 901 (2012) (“It is well known that the
relevant market estimates . . . are never ‘correct’ in product differentiated markets or in those that
have significant spatial dispersion and relatively high transportation costs.”).
This natural market fallacy might be viewed as a special case of the broader logical fallacy
of attributing tangible forms and properties to abstract concepts. See, e.g., A
CIENCE AND THE
51–52, 58 (1925) (describing the “Fallacy of
Misplaced Concreteness,” wherein abstract concepts are mistaken for concrete entities, and
therefor analyzed as though they were concrete facts); see also Rebecca Haw Allensworth, Law
and the Art of Modeling: Are Models Facts?, 103 G
. L.J. 825, 832–34 (2015) (describing
some properties of scientific models and their relationship to reality or optimal description).
[relative to the broader category] to make them [the relevant market for analy-
It is easy to imagine how differences in product characteristics and
uses could indicate the competitive closeness of products, but the opinion did
not explore this, resting instead on the mere observation of these differences.
Beyond providing an articulable basis for distinguishing one group of prod-
ucts from another, the Court did not even try to explain why such peculiar
characteristics and uses might affect the boundaries of the relevant market.
The leading authority for defining markets based on the public recognition
of specific trade and industry lines is another merger case, Brown Shoe.
district court, in this case, recited the standard litany of market definition con-
cepts before focusing almost exclusively on a simple reporting of how com-
mercial entities and the public viewed the boundaries of the market:
[A] “line of commerce” cannot be determined by any process of logic and
should be determined by the processes of observation. . . . Therefore, we
must go to the facts in the case and see what the testimony here reveals and
make a determination of the “line of commerce” from the practices in the
industry, the characteristics and uses of the products, their interchangeabil-
ity, price, quality and style. In other words, determine how the industry itself
and how the users, the public, treat the shoe product.
On appeal, the Supreme Court largely adopted this approach, articulating a
laundry list of practical indicia—observational factors—that might be used to
identify submarkets in which mergers could be assessed:
The boundaries of [a] submarket may be determined by examining such
practical indicia as industry or public recognition of the submarket as a sepa-
rate economic entity, the product’s peculiar characteristics and uses, unique
production facilities, distinct customers, distinct prices, sensitivity to price
changes, and specialized vendors.
We return briefly to the distinction between relevant markets and submarkets
below. For now, the interesting point is that, while some of the Brown Shoe
practical indicia could indeed reveal something about the competitive close-
ness of products and producers, the Court did not refer to this possibility in its
Instead, the opinion’s treatment of market definition was pre-
United States v. E. I. du Pont de Nemours & Co. (du Pont-General Motors), 353 U.S. 586,
See id. at n.12 (reciting trial testimony without further commentary).
Brown Shoe Co. v. United States, 370 U.S. 294 (1962).
United States v. Brown Shoe Co., 179 F. Supp. 721, 730 (E.D. Mo. 1959), aff’d, 370 U.S.
Brown Shoe, 370 U.S. at 325; see also id. at 336–37 (“Congress prescribed a pragmatic,
factual approach to the definition of the relevant market and not a formal, legalistic one. The
[market] must, therefore, . . . ‘correspond to the commercial realities’ of the industry.”).
See, e.g., id. at 326 (dismissing, without further explanation, the possibility of drawing
narrower markets around price/quality differences in shoes as “unrealistic”).
occupied with the question of the minimum size required for a potential sub-
market to warrant antitrust scrutiny.
Our aim here is not to criticize these opinions. Rather, we wish to empha-
size what is left unclear by modern reinterpretations of the tests they created.
It is not difficult to mold the peculiar-characteristics test into a rough approxi-
mation of how substitutable one product is for another. And courts and schol-
ars have similarly reinterpreted several of the practical indicia as factors
relevant to assessing closeness of competition.
Product characteristics and
some of the practical indicia may indeed be relevant to assessing the potential
for competitive injury in many cases.
But the discussion of market definition
in these opinions, and thus the tests themselves, were never meant to fit the
focus of modern antitrust analysis.
Antitrust law had not yet coalesced around the consumer welfare standard,
and both du Pont-General Motors and Brown Shoe were influenced by the
populist objectives that originally motivated the antitrust statutes. Congres-
sional motivations for Section 7 of the Clayton Act included the desire to
protect small, local competitors against larger rivals and to arrest and perhaps
reverse a perceived trend toward growing concentration in various indus-
Though less focused, Congressional concern with protecting small
competitors against larger rivals motivated the Sherman Act as well.
See, e.g., id. at 320 (reading legislative history to indicate that the “concern was with the
adverse effects of a given merger on competition only in an economically significant ‘section’ of
the country” (emphasis added)); id. at 325 (“[I]t is necessary to examine the effects of a merger
in each such economically significant submarket.” (emphasis added)); id. at 335 (“The 1950
amendments made plain Congress’ intent that the validity of such combinations was to be
gauged on a broader scale: their effect on competition generally in an economically significant
market.” (emphasis added)).
See, e.g., Rothery Storage & Van Co. v. Atlas Van Lines, Inc., 792 F.2d 210, 218–19 (D.C.
Cir. 1986) (reinterpreting the Court’s practical indicia as “evidentiary proxies for direct proof of
substitutability”); Baker, supra note 28, at 205 (providing a similar reinterpretation).
See, e.g., F
. R. E
. 401 (“Evidence is relevant if . . . it has any tendency to make a fact
[of consequence] more or less probable than it would be without the evidence.”).
See, e.g., Brown Shoe, 370 U.S. at 315–16 (“[C]onsiderations cited in support of [amend-
ments to Section 7] were the desirability of retaining ‘local control’ over industry and the protec-
tion of small businesses.”); id. at 344 (“[We] cannot fail to recognize Congress’ desire to
promote competition through the protection of viable, small, locally owned businesses.”);
Hovenkamp, supra note 30, at 894–95 (“[T]he rationale for market definition in Brown Shoe was
very different from, and is fundamentally at odds with, the rationale for market definition . . .
See, e.g., H
41 (2005) (“Although writers
heaped scorn upon Warren Court antitrust policy in the 1960s for its protection of small business,
that policy was probably the most faithful to Congress’s goals in passing the Sherman Act.”); 1
¶ 103 (4th ed. 2013) (“[W]hile
the framers of the Sherman Act were intent on condemning ‘monopoly,’ they saw the principal
injury of monopoly as reaching competitors rather than consumers.”); Carl Shapiro, Antitrust in a
Time of Populism, 61 I
. 714, 715 (2018) (“[T]he Sherman Act itself was
passed in 1890 in response to broad concerns about the political and economic power of large
Natural market concepts follow logically from these populist policy goals.
In common usage, the terms market, commodity, and industry are not clearly
related to substitutability and the closeness of competition between products
If Congress sought to protect small businesses and to prevent
further concentration in industries and markets as popularly conceived,
the naturalistic market-definition standards of these cases were reasonable ef-
forts to comply with legislative intent. The problem is not that these natural
market concepts were always inappropriate; the problem is that, having sur-
vived the evolution of antitrust policy, they have lost touch with what are now
the motivating principles of the antitrust framework. At best, these tests are
simply ignored as crude approximations of the modern approach, but they
may also confuse and detract from market definition analysis.
While characteristic-based markets present the clearest examples of natural
market reasoning, even in its older cases, the Supreme Court did not always
adhere strictly to lay concepts of industry when defining relevant markets,
and standards for defining relevant markets have always included various ap-
proximations to the economic idea of substitutability. Such standards include
the related notions of the cross-elasticity of demand and the reasonable inter-
changeability of use between products.
The idea is that somewhere in the
corporations in America.”); see also George J. Stigler, The Origin of the Sherman Act, 14 J.
. 1 (1985) (arguing that the Sherman Act was passed to protect competitors against
larger and more efficient rivals).
See 2B A
, supra note 10, ¶ 530a, at 235 n.5 (“In other contexts, of
course, ‘market’ means something else—for example, a trading center, as in ‘the stock market’
or ‘the town’s flea market.’ Data collections, including the Census, frequently lump together a
distribution level (‘retailing’) or a category of manufacture (‘motors and generators’) that covers
products that do not compete with each other.”); cf. T
, supra note 17, at 90 (“The term
‘commodity’ was one of those words which, for a long time, could be used without any question
being raised as to its exact meaning.”).
See Hovenkamp & Shapiro, supra note 9, at 2015 (commenting that the “line of commerce”
language of Section 7 was probably never intended to mean more than “a particular ‘line’ [of
products] that a seller might sell” as the term was used “by both businesspeople and courts” of
the time, and noting that Congress could have adopted the relevant market term of art, but did
not, when drafting revisions to this section of the statute); Hovenkamp, supra note 30, at 891
(“When it drafted the phrases ‘line of commerce’ and ‘section or community’ in 1914, and even
when it restated them as ‘section of the country’ in 1950, Congress almost certainly did not have
a technical definition of ‘relevant market’ in mind.”).
E.g., United States v. Cont’l Can Co., 378 U.S. 441, 457 (1964) (“Where the area of effec-
tive competition cuts across industry lines, so must the relevant [market]; otherwise an adequate
determination of the merger’s true impact cannot be made.”).
The cross-elasticity and reasonable interchangeability standards are not the only ways that
the Court has articulated this type of substitutability-based market concept. See, e.g., Tampa
Elec. Co. v. Nashville Coal Co., 365 U.S. 320, 327 (1961) (defining the area of effective compe-
tition as “the market area in which the seller operates, and to which the purchaser can practicably
turn for supplies.”).
field of increasingly distant competing products, substitutability becomes too
weak to warrant adding competitors to the relevant market—a generalization
of the intuition that two general stores in the same small town compete in a
common market but may not compete in a common market with stores in
neighboring towns and certainly do not compete in the same market as stores
hundreds of miles away. There are admirable qualities to this approach to
market definition, but—without additional details in the analysis—it too rests
on naturalistic market concepts.
A leading authority for defining markets using cross-elasticities of demand
is Times-Picayune, a tying case in which the Court undertook no detailed
market analysis but included a footnote commending market definition based
on degrees of substitutability:
For every product, substitutes exist. But a relevant market cannot meaning-
fully encompass that infinite range. The circle must be drawn narrowly to
exclude any other product to which, within reasonable variations in price,
only a limited number of buyers will turn; in technical terms, products
whose “cross-elasticities of demand” are small.
In using the term “cross-elasticities of demand” the Court probably overstated
the economic precision of its standard. Its own analysis of the “cross-elasticity
of demand” considered no more than “the trade’s own characterization of the
products involved,” observing that those in the trade “markedly differentiate”
between certain products.
Another leading authority for defining relevant markets by cross-elasticity of
demand is Cellophane.
Here, however, the Court confused the issue by
seeming to separate the cross-elasticity approach from another standard for
antitrust market definition based on the interchangeability of use between dif-
In considering what is the relevant market for determining the control of
price and competition, no more definite rule can be declared than that com-
modities reasonably interchangeable by consumers for the same purposes
make up that “part of the trade or commerce,” monopolization of which may
Times-Picayune Publ’g Co. v. United States, 345 U.S. 594, 612 n.31 (1953).
Id. (“Useful to [determining cross-elasticities of demand] is, among other things, the trade’s
own characterization of the products involved. The advertising industry and its customers, for
example, markedly differentiate between advertising in newspapers and in other mass media.”).
United States v. E.I. du Pont de Nemours & Co. (Cellophane), 351 U.S. 377, 380–81 (1956)
(“Every manufacturer is the sole producer of the particular commodity it makes but its control in
the . . . sense of the relevant market depends upon the availability of alternative commodities for
buyers: i.e., whether there is a cross-elasticity of demand [between products].”).
Id. at 395; see also id. at 404 (“[The relevant] market is composed of products that have
reasonable interchangeability for the purposes for which they are produced—price, use and qual-
Whether an intelligible distinction exists between the concepts of cross-elas-
ticity of demand and reasonable interchangeability of use is doubtful. Some
lower courts have understood reasonable interchangeability to mean the
court’s own assessment of the technical substitutability of products and cross-
elasticity of demand to mean customer willingness to substitute products at
But this bifurcated approach lacks economic rationale
and conflicts with the standard articulated in Cellophane, which encompassed
“reasonable interchangeability . . . price, use and qualities considered.”
It is easy to overstate the degree to which the substitutability standards dif-
fer from the characteristics-based standards discussed in the previous section.
The leading cases for the substitutability standards also relied upon evidence
of product characteristics and industry recognition,
and vice versa.
substitutability-based standards do, however, represent a conceptually distinct
approach to the market definition exercise.
The basic strategy of that approach is to draw market boundary lines where
the substitutability of products and producers becomes too attenuated. Unfor-
tunately, neither the reasonable-interchangeability-of-use test nor the cross-
elasticity-of-demand test even attempts to articulate where the cutoff lies.
How small must be the cross-elasticity of demand, and how poor must be the
interchangeability of use, before the edge of a relevant market has been
The omission of any attempt to answer these questions exposes the
naturalistic foundations of these standards.
First, and most importantly, these substitutability standards reflect the natu-
ralistic presumption that markets exist identifiably in the world, distinct from
any given inquiry or investigation. The error of ignoring the theory-depen-
E.g., FTC v. Arch Coal, Inc., 329 F. Supp. 2d 109, 119–20 (D.D.C. 2004), case dismissed,
No. 04-5291, 2004 WL 2066879 (D.C. Cir. Sept. 15, 2004); FTC v. Staples, Inc., 970 F. Supp.
1066, 1074–75 (D.D.C. 1997); United States v. Chas. Pfizer & Co., 246 F. Supp. 464, 468
Cellophane, 351 U.S. at 404; cf. United States v. Archer-Daniels-Midland Co., 866 F.2d
242, 246 (8th Cir. 1988) (“While sugar and HFCS are functionally interchangeable, they are not
reasonably interchangeable because of the price differential between the two products.”).
See supra note 48 and accompanying text; Cellophane, 351 U.S. at 380–81 (commenting
that “interchangeability is largely gauged by the purchase of competing products for similar uses
considering the price, characteristics and adaptability of the competing commodities.”).
E.g., Brown Shoe Co. v. United States, 370 U.S. 294, 325 (1962) (“The outer boundaries of
a product market are determined by the reasonable interchangeability of use or the cross-elastic-
ity of demand between the product itself and substitutes for it.”).
213 (1952) (“If it is un-
derstood that the products of different firms are generally not identical but different, what degree
of similarity or dissimilarity or, more concisely, what degree of substitutability would justify us
in speaking of the ‘same’ industry or of ‘different’ industries?”); Patrick Massey, Market Defini-
tion and Market Power in Competition Analysis: Some Practical Issues, 31 E
. & S
309, 314 (“It is unclear how high the cross price elasticity of demand needs to be before goods
can be considered to be part of the same market.”).
dence of market definition is addressed in Part II. Here, we are concerned
with the naturalistic character of these standards. The Court’s old sub-
stitutability standards conceive of markets as observable groups of products or
producers, at least when all relevant elasticities are known. Such a situation
would arise only if there were natural or physical boundaries to competition: a
gap in the chain of substitutes so vast that the products, producers, and cus-
tomers on one side of the gap were not competitively relevant to those on the
Such markets would be naturally identifiable in the world, obviat-
ing the need to identify any particular cutoff as part of the test itself.
Second, the expectation that substitutability standards should reveal natural
markets is revealed in the failure of this ideal. Perfect gaps in competition are
rare; a given product usually faces competition from products of varying de-
grees of substitutability at different price points.
Precisely because the sub-
stitutability standards offer no clear guidance on what to do in these cases,
courts and scholars have come to describe relevant markets as imperfect and
Often, but not always, these apologies are founded on the notion
that a correct market exists, but has not been identified. Worse yet, the exis-
tence of plausible alternative markets is somehow thought to undermine the
validity of any given choice of market.
Our point is not that market definition is unimportant, unprincipled, or im-
possible. As discussed further below, there are logical and helpful ways to
approach the exercise. An appropriate use of the HMT, for example, may
even consider some of the same evidence as the tests discussed above, but in
the context of an analytical exercise keyed to modern competitive effects
analysis. To reiterate, the point is not that market definition is not helpful, is
not evidence based, or is not a factual question. The point is simply that mar-
17 (2d ed. 1969) (“The
correspondence of [the classical economic idea of industry] to the industries of the real world is
not perhaps very close. But in some cases, where a commodity in the real world is bounded on
all sides by a marked gap between itself and its closest substitutes [the real-world industry will
approximate the theoretic ideal.]”).
, supra note 55, at 213 (“The use of the expression ‘entry into the industry’
presupposes that there are borderlines of some sort between one industry and another. Yet we
know that often in reality there are no such borderlines of any sort.”); Nicholas Kaldor, Mrs.
Robinson’s “Economics of Imperfect Competition,” 1 E
335, 335 (1934) (“Different
producers are not selling either ‘identical’ or ‘different’ products, but ‘more or less different’
products—the demand confronting them being neither completely sensitive nor completely un-
sensitive to the prices charged by other producers.”); Stocking, supra note 25, at 483 (“All prod-
ucts compete with each other for the consumer’s dollar, and in this sense each product is a
substitute for any other.”).
See United States v. Phila. Nat’l Bank, 374 U.S. 321, 361 (1963) (describing market defini-
tion as a “workable compromise” that merely avoids the indefensible extremes of overly narrow
and overly broad markets); 2B A
, supra note 10, ¶ 530d (“The Supreme
Court has wisely recognized there is ‘some artificiality’ in any boundaries, but that ‘such fuzzi-
ness’ is inherent in bounding any market.”).
ket definition has often been understood as aiming to identify natural or pre-
existing markets. That is a problematic goal, because in the general run of
cases, natural markets simply do not exist.
Unsurprisingly, the natural market bias of early antitrust opinions paralleled
early economic thinking on this topic. Starting from an uncritical identifica-
tion of markets with commodity concepts in textbook price-theory models,
economists contemplated the classification of industries around observable
the definition of markets by close substitutability,
termination of market boundaries at discrete gaps in the chain of substitutes.
The unsatisfying properties of all these approaches were foreshadowed in the
The turning point in economic thinking occurred in the 1930s and 1940s
with the development of the theory of monopolistic competition by Edward
Chamberlin. In 1950, Chamberlin surveyed extant market concepts with this
“Industry” or “commodity” boundaries are a snare and a delusion—in the
highest degree arbitrarily drawn, and, wherever drawn, establishing at once
wholly false implications both as to competition of substitutes within their
limits, which supposedly stops at their borders, and as to the possibility of
ruling on the presence or absence of oligopolistic forces by the simple de-
vice of counting the number of producers included. As for the conventional
categories of industries, it seems increasingly evident to me that they have
their origin, not primarily in substitution at all, but in similarity of raw
materials or other inputs or of technical methods used.
85 (1966) (“A market, according to the
masters, is the area within which the price of a commodity tends to uniformity, allowance being
made for transportation costs.”); T
,supra note 17, at 78–79 (noting that the equation of
industry and commodity made sense in the classical perfect competition model because “Under
pure competition, a number of sellers were supposed to compete for the sale of a homogeneous,
identical commodity: these sellers constituted a group, or an industry.”); Andreas G. Papandreou,
Market Structure and Monopoly Power, 39 A
. 883, 885 (1949) (“Before the advent
of the theory of monopolistic or imperfect competition, the concept of a ‘group’ of firms compet-
ing in the sale of a ‘commodity’ was considered self-explanatory.”).
See, e.g., Edward S. Mason, Price and Production Policies of Large-Scale Enterprise, 29
. 61, 69 (1939) (proposing to classify market structures by observation of “simi-
lar objective conditions” including “the economic characteristics of the product” and “the cost
and production characteristics of the firm’s operation.”).
Cf. Abba P. Lerner, The Concept of Monopoly and the Measurement of Monopoly Power, 1
. 157, 167 (1934) (“In calling the same thing at different places different com-
modities, we have rejected the criterion of physical similarity as a basis for [identifying markets]
and have put in its place the principle of substitutability at the margin.”).
, supra note 56.
Edward H. Chamberlin, Product Heterogeneity and Public Policy, 40 A
Yet Chamberlin did not reject the idea of the market itself. Instead, his writing
suggested a different way to conceptualize markets: not as “definite economic
[entities], the existence of which has merely to be recognized by the investiga-
tor,” but as analytical tools which “may and should be used with all degrees of
inclusiveness” in the process of studying a problem.
Under this approach,
markets are not concrete arenas or free-standing entities. They are lenses for
focusing analysis on a given problem. Such markets can—in fact, must—be
molded and shaped to fit the specifics of each problem.
Put another way,
market definition remains a factual inquiry upon which evidence is needed,
but that evidence is mixed with an understanding of the purpose of the exer-
cise to define a market fit for the specific role it is to serve. In this approach,
there are no right or wrong markets, only markets of varying utility in study-
ing the particular economic question at hand.
The novelty of Chamberlin’s approach is taken for granted today.
ample, though the underlying logic often goes unstated, this analytical under-
standing of the market concept is foundational in much of modern equilibrium
Deeply integrated as modern antitrust is with economic analysis, this quick
history of the economic concept of markets suggests two fundamental proper-
ties of antitrust market definition. First, there is no such thing as an economi-
cally interesting natural antitrust market. No real markets are waiting to be
A market is not true (false), correct (incorrect), or real (unreal), but
merely appropriate (inappropriate) for a specific purpose. Appropriateness is a
contextual quality. And, in antitrust, the context is the theory of harm. Thus,
, supra note 17, at 84 (explaining the “Chamberlinian ‘group’” concept).
See Papandreou, supra note 59, at 886 (“For Professor Chamberlin the ‘group’ concept is
merely an analytical tool which derives its content from the problem at hand.”).
See, e.g., M
, supra note 55, at 217 (defining industry as “merely a short expression
which stands [for] all firms whose operations affect one another’s selling opportunities and sales
revenues so definitely that we must not neglect taking account of them.”); id. at 213–14 (explain-
ing that “[i]t saves time and effort in analysis to assume certain variables as constant or, what
often comes to the same thing, to disregard them; and it is quite legitimate to do so if changes of
these variables are negligible for the particular problem or if the direction of the relationship is
, supra note 17 (explaining that such a market “abstracts those firms that are
more tightly linked with the enterprise under consideration and which, as a consequence, cannot
be ignored in a discussion of its problems”).
See, e.g., Kaplow, supra note 2, at 507 (“[T]here is no way to see (or feel or otherwise
directly sense) the magnitude of a firm’s market power . . . . No aspect of the analysis is sensory;
‘markets’ as the term is used in this context are pure abstractions.”).
Werden, Answer, supra note 7, at 746 (“Separating active and passive competitive forces is
part of economic analysis because economic models distinguish the strategic action of competi-
tors from background influences on them.”); Werden, Possible, supra note 7, at 2 (“[S]eparating
active from passive competitive forces is the defining feature of the [most] ubiquitous modeling
technique in the field—partial equilibrium analysis.”).
, supra note 15, at 41 (“There is not for any product a single, real ‘market’ wait-
ing to be discovered.”).
the second proposition: as a purely analytical construct, the definition of a
relevant market must always depend upon the nature of the problem for which
it is to be used. The second point is developed at length in Part II of this
article, but the first point deserves brief attention as well.
While it would be unfair to criticize the Supreme Court for its early focus
on natural-market concepts, the same cannot be said of the continued adher-
ence to these concepts today. The problem is that the policy and framework of
antitrust law have shifted over time to the economic analysis of effects on
consumer welfare while the precedential language of these old cases has not
changed accordingly. The old naturalistic standards have thus become a bar-
rier to proper market definition and a lure toward naturalistic thinking.
The barrier is that even when the factors implicated by these standards are
relevant to market definition, nothing in the Court’s market-definition stan-
dards says what to do with this information. The substitutability standards do
not say what degree of substitution defines a market;
the characteristics stan-
dards do not say how to interpret any given combination of characteristics.
As discussed in Part II, the modern HMT and related tests of market definition
provide answers to these questions. But if courts and practitioners continue to
treat natural-market concepts as controlling standards—deserving equal or
greater weight than the HMT—then market definition will continue to be
The lure of the Court’s old market-definition standards is what keeps natu-
ral-market thinking entrenched in market-definition practice. One example is
the continued focus of courts and advocates on Brown Shoe’s practical indicia
in merger cases.
While some of the underlying factors in this test may well
be relevant in HMT analysis,
treating them as having any separate relevance
outside of the HMT framework can only muddle market definition.
Another example is the occasional argument, based on the reasonable inter-
changeability standard in du Pont-General Motors, that a judge’s own impres-
sion of substitutability should outweigh evidence of actual consumer
See supra notes 55, 58, and accompanying text.
See Turner, supra note 1, at 1151 (“The problem is, you see, the courts really have not
gotten around to trying to spell out the necessary analysis, and what the consequences are of
certain facts.”); id. (“[The Brown Shoe factors are] a laundry list, not a mode of analysis.”).
E.g., United States v. Anthem, Inc., 236 F. Supp. 3d 171, 193–207 (D.D.C. 2017) (treating
the Brown Shoe indicia and the HMT as complementary means of defining markets); FTC v.
Sysco Corp., 113 F. Supp. 3d 1, 25–38 (D.D.C. 2015) (same); FTC v. CCC Holdings Inc., 605 F.
Supp. 2d 26, 38–45 (D.D.C. 2009) (same).
See supra note 72 and sources cited therein.
See supra note 39.
preferences in market definition.
The incoherence of this idea reflects its
naturalistic roots. Only if markets were observable entities in the world could
it possibly make sense to elevate a judge’s impression of product characteris-
tics above otherwise reliable evidence of consumer preferences in market
A final example is the misuse of Brown Shoe’s admonition that relevant
markets must “correspond to . . . commercial realities”
to discredit proposed
markets as “unrealistic” or “gerrymandered.”
If there is no such thing as a
natural or realistic market, then it makes no sense to reject a putative antitrust
market for being artificial or unrealistic.
Defendants are always free to argue
that a market is invalid for failing to meet a proper test—not satisfying the
HMT, for example. But to reject any valid market as unrealistic or gerry-
mandered, simply because it fails to conform to lay observation or intuition, is
to commit the naturalistic fallacy in its most extreme form.
United States v. Oracle Corp., 331 F. Supp. 2d 1098, 1131 (N.D. Cal. 2004) (“The test of
market definition turns on reasonable substitutability. This requires the court to determine
whether or not products have ‘reasonable interchangeability’ based upon ‘price, use and quali-
ties.’ What, instead, these witnesses testified to was, largely, their preferences. Customer prefer-
ences towards one product over another do not negate interchangeability.” (citations omitted));
James A. Keyte & Kenneth B. Schwartz, “Tally-Ho!”: UPP and the 2010 Horizontal Merger
Guidelines, 77 A
L.J. 587, 607 (2011) (“While the case law remains somewhat murky
on the role of cross-elasticity, it is now well-established precedent that consumer preferences are,
at most, a component of reasonable interchangeability and should not provide a separate basis for
defining a relevant market.”).
This is not to say that evidence of consumer preferences is entitled to uncritical deference.
But, if product characteristics are relevant at all, it must be because they help to evaluate con-
sumer preferences—not the other way around. Cf. supra note 75 and sources cited therein.
Brown Shoe Co. v. United States, 370 U.S. 294, 336 (1962). Strictly, the court referred only
to geographic markets in this assertion. Use of the idea has not remained so contained.
E.g., U.S. Healthcare, Inc. v. Healthsource, Inc., No. CIV. 91-113-D, 1992 WL 59713, at *5
(D.N.H. Jan. 30, 1992), aff’d, 986 F.2d 589 (1st Cir. 1993) (“I find that the concept of a geo-
graphic market being the southern tier of New Hampshire is an unrealistic form of gerrymander-
ing in light of the parties’ recruiting, marketing and sales efforts.”); United States v. Gen.
Dynamics Corp., 341 F. Supp. 534, 560 (N.D. Ill. 1972), aff’d, 415 U.S. 486 (1974) (“even were
this court to accept the Government’s unrealistic product and geographic market definitions
. . . .”).
See Baker, supra note 11, at 139 (“[T]here is no reason to expect that the concept of market
employed by business executives when discussing issues of business strategy or marketing . . .
would be the same as the concept of an ‘antitrust market’ or ‘relevant market’ defined for the
purpose of antitrust analysis.”); 2010 Merger Guidelines, supra note 24, § 4, ¶ 8 (“Relevant
antitrust markets defined according to the hypothetical monopolist test are not always intuitive
and may not align with how industry members use the term ‘market.’”).
There is no shortage of relevant markets about which claims of unrealistic gerrymandering
still hang. See, e.g., FTC v. Whole Foods Mkt., Inc., 548 F.3d 1028, 1035 (D.C. Cir. 2008)
(involving the FTC’s allegation of a “premium natural and organic supermarkets” relevant mar-
ket); United States v. Oracle Corp., 331 F. Supp. 2d 1098, 1130 (N.D. Cal. 2004) (involving the
DOJ’s allegation of a “high function FMS and HRM [software]” relevant market); Hill Wellford
& Gregory Wells, The “Litigation Mulligan” in the 2010 Merger Guidelines: Better Economics
but Not (Necessarily) More Clarity Before the Agencies and the Courts, CPI A
2010, at 1, 11 (“The Agencies’ attempts to define as relevant markets certain ‘high function’
One response to our argument, up to this point, is that it shows how natural-
market concepts can derail market definition but does not show that these
concepts have actually led to errors in recent cases. We provide one case
study in Part IV.C below, but must admit that we cannot easily prove how the
attention courts and advocates pay to these old tests affects outcomes in run-
of-the-mill cases. The modern convention is to enumerate essentially all the
different tests in reaching a market-definition conclusion. Perhaps no court or
analyst is ever swayed by the old tests, which are simply announced as lip
service to obviously decrepit precedent. We cannot disprove this.
But why continue to invoke the old standards at all? At worst, dropping
these tests from the market definition exercise changes nothing. That would
be the case, for example, if indeed no court or advocate was ever swayed by
the old tests. Perhaps that is the case. But suppose the world is less extreme.
Suppose some fact-finders do assign weight to these old tests and that some
advocates do foment confusion with natural-market arguments.
In a world
where such things are possible, dropping the old tests would increase the clar-
ity and accuracy of the market-definition exercise. It would eliminate an un-
necessary distraction and would focus analysis on the critical task of defining
markets to identify the scope of potential competitive harm.
There is no legal or rational defense for continuing to perpetuate the natural
market fallacy. The Supreme Court’s early efforts at market definition never
purported to exhaust the ways to define a market.
And in more recently
stating that the purpose of market definition “is to determine whether an ar-
rangement has the potential for genuine adverse effects on competition,”
Court has set an objective at odds with its earlier natural-market standards.
enterprise software . . . and ‘premium natural and organic supermarkets’ . . .—neither being a
term used in the relevant industry—are now recognized as litigation mistakes.”).
See, e.g., 3A K
§ 150:66 (6th ed. 2012) (proposing model instructions in which ju-
rors are told that “[I]n determining the product market, the basic idea is that the products within
it are interchangeable as a practical matter from the buyer’s point of view. . . . What you are
being asked to do is to decide which products compete with each other.”). But see A
[hereinafter ABA M
] (more clearly limiting the use of observational
evidence to an application of the HMT in market definition).
See FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460 (1986).
Cf. Brown Shoe Co. v. United States, 370 U.S. 294, 320 (1962) (“Congress neither adopted
nor rejected specifically any particular tests for measuring the relevant markets, either as defined
in terms of product or in terms of geographic locus of competition.”).
Indiana Federation, 476 U.S. at 460; see also B
NALYSIS OF THE
59 (4th ed. 1965) (“Be-
cause of the relative flexibility with which the Court has dealt with the problems of market
boundaries, the term ‘relevant market’ . . . does not, and cannot, refer to a ‘market’ in any simple
economic or trade sense, but refers rather to the . . . locale where [competitors, suppliers, or
customers] may be affected by an acquisition—and nothing more.”).
The onus, here, is on courts and scholars to normalize the omission of the old
tests in market definition. When advocates are finally free from the need to
pay obeisance to natural-market concepts, the market-definition exercise will
be clarified and rationalized.
II. THE INDEPENDENT MARKET FALLACY
The independent market fallacy is the common misconception that relevant
markets exist independent of a theory of anticompetitive injury. The error in
this assumption rests in the prior fallacy. If markets were indeed freestanding
entities, then the process of defining a relevant market would be a matter of
observation independent of competitive effects analysis. But antitrust markets
are not observable entities; they are mental constructs designed to help assess
specific competitive concerns. As a result, market definition can never be sep-
arated from the act of hypothesizing a specific competitive concern.
If this seems obvious, note that market definition has often been treated as
an independent step in rule-of-reason analysis. Early Supreme Court cases
were consistent with this view, articulating market-definition tests that were
independent of theories of anticompetitive harm and treating market definition
as a step preceding the start of competitive effects analysis.
A similar idea
long applied in merger review. The understanding of market definition as a
discrete, initial step in merger review could be inferred from the text and
structure of the 1982 and 1992 Merger Guidelines,
and while the 2010 Hori-
zontal Merger Guidelines now deny this rote ordering of analysis, advocates
of the old way remain.
The following shows how the economic understanding of markets as ana-
lytic tools necessitates conditioning market definition on specific theories of
harm. Standards like the HMT broadly align market definition with theories of
anticompetitive harm. But even these standards generally require further tai-
loring to fit the specific circumstances of a given case and theory.
E.g., United States v. E.I. du Pont de Nemours & Co. (du Pont-General Motors), 353 U.S.
586, 593 (1957) (“Determination of the relevant market is a necessary predicate to a finding of a
See U.S. Dep’t of Justice, Merger Guidelines § 2 (1982), www.justice.gov/sites/default/
files/atr/legacy/2007/07/11/11248.pdf (appearing to make market definition a discrete and initial
step in merger analysis); U.S. Dep’t of Justice & Fed. Trade Comm’n, Merger Guidelines § 0.2,
¶ 1 (1992) [hereinafter 1992 Merger Guidelines], www.justice.gov/?sites/?default/?files/?atr/
Cf. Varney, supra note 22, at 655 (responding to criticism of the 2010 Merger Guidelines’
position that merger review does not need to begin with market definition).
EPENDENCE OF THE
Antitrust scholars have long suspected that mainstream economics has little
to say about market definition.
It has become popular to claim that econo-
mists never define markets.
Both ideas are curious, given the substantial
effort that economists have devoted to studying the theory of market concepts
and the identification of markets.
In microeconomic theory markets are typi-
cally taken as a primitive concept, leaving it to the applied microeconomist to
define a market appropriate to a given empirical study. Perhaps the point is
that mainstream microeconomists rarely define antitrust markets in their re-
search. That is correct, but unsurprising.
Because antitrust markets are constructed for the specific purpose of study-
ing antitrust problems, it would be remarkable indeed if they were of interest
to economists studying non-antitrust problems. The point is not that antitrust
market concepts differ in any fundamental way from the market concepts used
other places in economics.
What really differs from one context to the other
is the problem at hand.
E.g., Horowitz, supra note 3, at 1–2 (“Curiously enough, economists have had compara-
tively little to say about how to delineate markets.”); George J. Stigler, The Economists and the
Problem of Monopoly, 72 A
. 1, 9 (1982) (“My lament is that this battle on market
definitions, which is fought thousands of times what with all the private antitrust suits, has re-
ceived virtually no attention from us economists.”); Stigler & Sherwin, supra note 3, at 555
(“The infrequency with which one encounters actual market size determinations outside the anti-
trust area is surprising and perhaps disquieting.”).
Kaplow, Impossible, supra note 5, at 364 (“[T]he notion of a relevant market does not exist
[in industrial organization economics].”); Kaplow, Alchemy, supra note 5, at 927 n.16 (“It is the
market definition approach that is unsubstantiated; . . . in the economic theory of industrial
organization it does not even exist.”); Hovenkamp, supra note 30, at 910 (“Indeed, as Kaplow
observes, the concept of market definition has virtually no presence in the theoretical or empiri-
cal literature of industrial organization today.”); Fisher, supra note 8, at 132 (“What, then, does
economic analysis have to say about market definition? In one sense, the answer is ‘Nothing at
all.’ The question of what is ‘the’ relevant market never arises in economics outside of
See supra Part I.C (citing and discussing just some of this work).
Cf. Massey, supra note 55, at 317 (“An important development in the literature on market
definition . . . is the distinction between the concept of a relevant market used in competition
analysis, and traditional economic definitions of a market.”); Turner, supra note 1, at 1147
(“[T]here is bound to be, it seems to me, a difference between the economic and legal concepts
of the market.”); Werden, supra note 10, at 515–16 (“[T]he concept of market delineation as it is
used in the antitrust context is quite foreign to economic theorists, and it is only this context that
gives meaning to the market delineation question.”); Papandreou, supra note 59, at 883 (“One
important reason for this gap between the legal and economic concepts of monopoly is their
difference in emphasis. Whereas the lawyer deals with competitive relationships, the economist
is primarily interested in the allocation mechanism and welfare economics.”). Note that we are
referring to antitrust relevant markets here; the term “relevant market” may deviate from the
economic concept in other areas of law. See generally C
In antitrust trials and investigations, attention is typically devoted to one of
two questions: (1) has suspect conduct caused any anticompetitive injury? or
(2) might it, or some future conduct, cause future anticompetitive injury?
One market concept helpful in answering either question is an outer bound on
the entities sufficient to bring about the potential injury: a market defined as a
class of transactions in which the suspected injury could occur, at least under
assumptions favorable to that theory of harm.
Sometimes defining such a
market will answer the question whether competitive injury did, or is likely to,
occur. In most cases, this type of market definition cannot itself provide an
answer but will facilitate further analysis by identifying questions that must be
answered to dispose of the ultimate question of the past or future likelihood of
In the right circumstance, the HMT exemplifies this type of market con-
The basic idea is to take a putative market (e.g., a set of products) and
ask whether a hypothetical monopolist over this market would profitably raise
the price of some of the products by a “small but significant amount” (e.g., by
5 percent) over an appropriate baseline (e.g., the competitive price) for an
appreciable period of time (e.g., a year). If the answer is yes, then the candi-
date market satisfies the test and is a relevant antitrust market; if the answer is
no, then the candidate market is expanded to include additional products and
the test is repeated until a relevant antitrust market is found. Less elaborate
articulations of the HMT,
and comparable standards based on the identifica-
This does not describe all antitrust applications. Proof of anticompetitive injury is not re-
quired in hard-core collusion cases under Section 1 of the Sherman Act, for example. Tellingly,
this is one situation in which courts have long dispensed with the need to define relevant
See, e.g., P
, supra note 25, at 148–49 (“[A group of sellers] is thus a market in the
sense, which is the only one relevant to an economic analysis of competition and monopoly, of a
group of sellers who have the power to increase the market price by merging or colluding.”);
Werden, Answer, supra note 7, at 741 (“When the [HMT] is used, the allegation of the relevant
market certifies at least the possibility of harm the antitrust laws were designed to prevent.”).
See generally 2010 Merger Guidelines, supra note 24, § 4.1.1 (providing the modern ex-
pression of the HMT as applied to horizontal mergers).
See, e.g., 2B A
, supra note 10, ¶ 533, at 267 (“A ‘market’ is any
grouping of sales whose sellers, if unified by a hypothetical cartel or merger, could profitably
raise prices significantly above the competitive level.”); S
, supra note 15, at 41 (“To
define a market . . . is to say that if prices were appreciably raised or volume appreciably cur-
tailed . . . while demand held constant, supply from other sources could not be expected to enter
promptly enough and in large enough amounts to restore the old price or volume.”); Hovenkamp
& Shapiro, supra note 9, at 1999 (“Any candidate market for which the court concludes that a
perfectly functioning cartel would lead to a significant price increase qualifies as a relevant
tion of possible collusive groups
or producers that would otherwise constrain
a given exercise of market power,
amount to roughly the same idea.
Tests of this form lead to the delineation of appropriate antitrust markets if
they correspond to the theory of harm at issue. To illustrate, suppose a pro-
posed merger of two rival producers raises concern that coordination among
the remaining producers of similar products will cause modest, near-term
price elevation. The typical articulation of the HMT (focusing on a non-transi-
tory 5 percent price increase) defines a relevant market corresponding to just
such a theory of harm. Proper implementation of this version of the HMT will
delineate only those markets in which coordination among the set of produc-
ers could result in something like a 5 percent price increase in the near future.
Whether the merger would bring about this result is a question requiring fur-
ther analysis of market structure and the feasibility of coordination. Either
way, the relevant market identifies suspect participants, helping to contextual-
ize the analysis needed to assess the likelihood of a post-merger price
But the usual articulation of the HMT fails to identify a proper market un-
less it corresponds to the relevant theory of harm. Suppose, for example, that a
proposed merger instead raised concern about coordination causing a large
price increase, not taking effect until the next bidding cycle three years in the
future. The usual articulation of the HMT is ill-suited to identifying a relevant
market for this theory of harm. The typical HMT parameters target moderate
price elevation in the near term, not large price elevation in the longer term.
The groups of producers with sufficient power and incentive to impose large
price increases may well differ from those with the power and incentive to
target moderate price increases. And the ability of customers to avoid price
increases in the more distant future by substituting products or suppliers may
well differ from the ability of customers to cover against near-term price in-
creases. The point is not that the HMT cannot be used in this setting, just that
it must be adjusted to fit the relevant theory of harm.
E.g., Kenneth D. Boyer, Is There a Principle for Defining Industries?, 50 S. E
. J. 761,
763 (1984) (defining markets around possible collusive groups).
See, e.g., M
(“[T]he relevant market should . . . [contain] the set of products (and geographical areas) that
exercise some competitive constraint on each other.”); Fisher, supra note 8, at 133 (“[A] useful
market definition should include in the market all of the firms and products or services that
constrain the exploitation of monopoly power by the firm.”); see also Boyer, supra note 96, at
763 (“A firm’s competitors . . . are those sellers who would cause significant losses if that firm
took independent action.”).
USTOMIZATION OF THE
Decades ago, Phillip Areeda complained, “I am repeatedly disappointed
that my students leap into market definition without first specifying the partic-
ular legal question that the tribunal hopes to answer through market defini-
We doubt Areeda would be better disposed toward current practices.
Even in a framework as sophisticated as the HMT, market definition is often
treated as if it were a theory-independent step in the analysis. This is clearly
wrong. A meaningful relevant market must be conditioned on a specific the-
ory of anticompetitive harm;
otherwise, the relevant market cannot properly
inform the analysis of competitive effects.
This simple proposition clarifies the important role of market definition in
antitrust. Consider the 2010 Horizontal Merger Guidelines. While conceding
that the base price and hypothesized price increase may differ from one appli-
cation of the HMT to another,
the Guidelines offer no rationale for how or
why these parameters should be chosen in particular applications. Confusion
would have been avoided if the Guidelines had stated that the base price and
hypothesized price increase must correspond to the specific theory of an-
ticompetitive harm being investigated. We now discuss how this correspon-
dence can be achieved in practice.
1. Customizing the Price Increase
The gravest omission in the 2010 Horizontal Merger Guidelines may be its
failure to explain how to choose the size of the hypothesized price increase in
Areeda, supra note 10, at 553; see also Fishman v. Estate of Wirtz, 807 F.2d 520, 568–69
(7th Cir. 1986) (Easterbrook, J., dissenting in part) (“The market definition in this case shows
why you can’t pick a market without knowing the purpose of the choice. The court has defined a
market of professional basketball in Chicago. This is a plausible market, if the question is
whether anything injured consumers. . . . If, instead, we seek to learn whether CPSC harmed
competition for a sports franchise, we must define a market that looks at the demand and supply
possibilities facing Rich and IBI.”); Fisher, supra note 8, at 130 (“The first thing to understand
about market definition is that how it is done depends on the purpose for which it is used.”).
Areeda, supra note 10, at 553 (“[A] subordinate question needs to be focused before market
definition can be attempted: namely, what particular impairment of competition is to be
See, e.g., Salop, supra note 26, at 191 (“Market definition and market power should be
evaluated in the context of the alleged anticompetitive conduct and effect, not as a flawed filter
carried out in a vacuum divorced from these factors.”); Gregory J. Werden, Four Suggestions on
Market Delineation, 37 A
. 107, 108 (1992) (“Assuring that markets are suitable
for the purposes to which they are put requires that a preliminary step be taken before market
delineation. This step is the identification of who might exercise market power, against whom it
might be exercised, and how it might be exercised.”).
E.g., 2010 Merger Guidelines, supra note 24, § 4.1.1, ¶ 2 n.4 (considering circumstances in
which a hypothetical cartel should be used in place of a hypothetical monopolist); id. § 4.1.2, ¶ 1
(defining the HMT benchmark price as that which “would likely prevail absent the merger”); id.
§ 4.1.2, ¶ 3 (allowing the size of the hypothesized price increase to depend on “the nature of the
industry and the merging firms’ positions in it”).
applying the HMT.
After noting that the size of the increase is a method-
ological question, not a policy choice,
the Guidelines say only that the size
of a hypothetical price increase depends on “the nature of the industry and the
merging firms’ positions in it.”
This guidance verges on uselessness, and
may even be read as reverting to natural market concepts—the “nature” of
some metaphysical “industry”—even while articulating an analytical market
concept. The clearer and more useful guidance would have been to say that
the size of the price increase should reflect the specifics of the potential injury
To put this point in perspective, it helps to consider the typical and flawed
policy argument that relevant markets should be defined based on price in-
creases of no less than, say, 5 percent because only substantial price increases
constitute antitrust violations in warrant of enforcement actions.
two problems with this argument. First, apart from some loose qualifiers (un-
due restraint of trade; substantial lessening of competition)
there is no quan-
titative threshold for violating any antitrust statute,
so it would be odd to
contemplate implementing any particular choice of threshold at the market
definition stage. Second, even if there were something like a minimum 5 per-
cent price-increase requirement to prove a violation, the requirement would
We recognize that omitting this detail could serve important ends. It might preserve some
flexibility in the process or increase accessibility to lay readers. Our focus is on the Guidelines’
role as a description of the market definition process, and we hope our comment will be under-
stood as limited to this scope.
2010 Merger Guidelines, supra note 24, § 4.1.1, ¶ 2 (“The SSNIP is . . . a methodological
tool for performing the hypothetical monopolist test; it is not a tolerance level for price increases
resulting from a merger.”).
Id. § 4.1.2, ¶ 3.
See, e.g., 2B A
, supra note 10, ¶ 530a, at 238 (“[T]he extent of the
market for legal purposes depends on the magnitude and duration of power that antitrust law
deems critical. . . . [S]electing the relevant degree and duration are questions of legal policy.”);
id. ¶ 530e, at 241 (“Because the market power that concerns antitrust law must be ‘substantial,’ a
product that can be profitably priced only a few percentage points above the perfectly competi-
tive level . . . should not be deemed a ‘market’ for antitrust purposes.”); Werden, supra note 10,
at 538–39 (“The Guidelines also require price increases to be ‘significant and non-transitory’
because collusion that increased price only slightly or for a very short time would not have a
significant adverse effect on the economic welfare of the nation, and therefore would not justify
governmental intervention in the marketplace.”).
E.g., Standard Oil Co. of N.J. v. United States, 221 U.S. 1, 58–60 (1911) (interpreting
Sherman Act § 1 to reach only “undue” restraints of trade); Clayton Act, ch. 323, § 7, 38 Stat.
730, 731–32 (1914) (current version at 15 U.S.C. § 18 (2012)) (prohibiting mergers, the effects
of which “may be substantially to lessen competition, or to tend to create a monopoly”).
See Brown Shoe Co. v. United States, 370 U.S. 294, 321 (1962) (“[Congress did not] adopt
a definition of the word ‘substantially,’ whether in quantitative terms . . . or in designated quali-
tative terms.”). Note that we are referring to U.S. antitrust law here. Cf. Eur. Comm’n, Notice on
Agreements of Minor Importance Which Do Not Appreciably Restrict Competition Under Arti-
cle 101(1) of the Treaty on the Functioning of the European Union (De Minimis Notice), 2014
O.J. (C 291) 1.
logically apply to the competitive effect of some challenged conduct, not the
relevant market in which that conduct is assessed.
Suppose that a hypothetical monopolist over a given set of products would
raise the price of every product in the set by 10 percent. This set of products
constitutes a valid relevant market when tested by the HMT with a 5 percent
hypothesized price increase, but the actual price effect resulting from
whatever conduct is being explored is unclear. If the conduct in question is a
merger to monopoly, then the price effect would be double the assumed 5
percent threshold. If the conduct in question is the merger of two out of ten
firms in the relevant market, then the price effect could be much smaller than
the assumed 5 percent threshold.
In either case, implementing the substanti-
ality threshold at the market definition stage is both clumsy and misleading.
So, on what basis should the hypothetical price increase be chosen when
applying the HMT? The answer is on the basis of the theory of anticompeti-
tive injury in question. At a minimum, then, we ought never to choose a hypo-
thetical price increase larger than what the actual anticompetitive effect is
expected to be. Suppose, for example, that the anticompetitive concern around
some challenged conduct is a price increase of only 3 percent—because an
elastic supply of an alternative product is known to be available at a price 3
percent above the current price.
As we discuss below, a 3 percent price
increase on a large mass of sales may amount to substantial competitive harm
and thus warrant antitrust scrutiny.
How should we define the market for
this inquiry? A relevant market defined by the HMT with the usual 5 percent
price increase is inappropriate. Of what importance are the suppliers that are
“in” the market for a 4–5 percent price increase when our actual concern is a
1–3 percent price increase? If these suppliers would not constrain the hypoth-
esized 3 percent price increase, then their inclusion in the market could only
confuse analysis, overstating the competitive forces constraining the hypothe-
sized injury. The maximum price increase for the HMT in this example would
be at most 3 percent—the upper bound on the price increase posited by the
anticompetitive theory at issue.
See Gregory J. Werden, The History of Antitrust Market Delineation, 76 M
. L. R
123, 204 (1992) (offering a similar observation); Dennis W. Carlton, Market Definition: Use and
Abuse, 3 C
3, 6 n.3 (2007) (same).
See, e.g., United States v. Archer-Daniels-Midland Co., 866 F.2d 242, 246 (8th Cir. 1988)
(finding that a monopolist of high fructose corn syrup could raise the price of this product “to
just below the . . . price of sugar before being constrained by the competitive forces of sugar”).
We provide a more detailed discussion of how a proper understanding of market definition
informs the substantiality inquiry in Part IV.B.
Whether a price increase of at most 3% warrants legal relief is a conceptually and logically
separate question from how to define the relevant market for this injury. The logic of market
definition does inform this question, however, and we return to this subject in Part IV.B.
This point can be generalized. If the hypothesized anticompetitive injury is
a large price increase, something like 10–20 percent, then a relevant market
defined by the HMT with a 5 percent hypothesized price increase is similarly
inapt to the inquiry. If a large price increase is the concern, then the absence
of adequate substitutes in response to a small price increase is at most of
partial interest. Here, it is critical to know which producers are “in” the mar-
ket for a large price increase on the order of the theorized concern. An appro-
priate choice of price increase for the HMT in this example is thus something
like 10–20 percent.
We recognize that in many cases it may be impossible to state the anticom-
petitive concern with such precision. Early in the review of merger notifica-
tions, for example, precise predictions of the potential anticompetitive
concern cannot be expected. In other situations, such as those involving a
range of potential coordination strategies, it may never be possible to specify
competitive concerns in fine detail. Where nothing about the conduct or con-
text suggests otherwise, the longstanding default 5 percent price increase of
the Merger Guidelines seems reasonable.
But it should not be treated as the
presumptive standard that it is today. If the relevant market is to be helpful in
analyzing questions, then as the nature of anticompetitive concern evolves and
comes into focus, so should the relevant market.
2. Customizing the Base Price
Another HMT parameter in need of customization is the base price. Though
rarely described in quite these terms,
the poster child for failing to heed this
requirement is the Supreme Court’s infamous blunder in Cellophane.
translate the Court’s mistake into HMT terms, in attempting to define the
relevant market for assessing a monopolization claim under Section 2 of the
Sherman Act, the Court asked whether DuPont would have found it profitable
to increase the price of its cellophane product by 5 percent. Finding that Du-
Pont would not have profited from such a price increase—to the surprise of no
one operating under the assumption that firms set prices to maximize prof-
its—the Court went on to expand the relevant market and misleadingly mask
DuPont’s apparent ability to exercise market power. If the base price in the
HMT had not been the current price, but instead some estimate of Cello-
2010 Merger Guidelines, supra note 24, § 4.1.2, ¶ 3 (“The Agencies most often use a
SSNIP of five percent of the price paid by customers.”).
An exception is Salop’s treatment of this subject. Salop, supra note 26, at 194, 197 (provid-
ing a description of the “Cellophane trap” analogous to what we describe here).
United States v. E.I. du Pont de Nemours & Co. (Cellophane), 351 U.S. 377 (1956).
phane’s competitive price, then the relevant market would have been nar-
rowed and DuPont’s market power would have been evident.
Few antitrust decisions have been pilloried so deservedly as Cellophane.
But just what made the Court’s reasoning fallacious has never fully crystal-
ized in the literature. Comments on the topic often draw a distinction between
market definition in monopolization cases versus merger cases, wrongly sug-
gesting that the statutory standard is of more than derivative importance in
dictating how markets should be defined.
Similar ambiguity envelops the
2010 Horizontal Merger Guidelines’ approach to selecting the baseline price
in the HMT. The Guidelines expressly contemplate alternatives to the current-
price default baseline but describe the situations justifying alternative base-
lines by vague and confusing references to likely future prices absent the
While none of these commentaries is entirely off the mark, none
answers the root question: what dictates the proper choice of baseline price?
The answer is, once again, the theory of anticompetitive injury. As purely
conceptual tools, relevant markets are defined to address specific economic
questions. And like any tool, the appropriate market definition depends on the
purpose for which it is to be used.
The appropriate market for one question
may be inappropriate for another. Cellophane illustrates just this.
As noted above, Cellophane presented the Court with a theory of monopoli-
Possession of monopoly power is an element of this claim, and the
purpose of defining a market in Cellophane was to help determine whether
DuPont already possessed monopoly power.
Whether market definition was
See generally 2B A
, supra note 10, ¶ 539 (providing an extended
treatment of the Cellophane fallacy); P
, supra note 25, at 150–51 (same).
E.g., Massey, supra note 55, at 323 (“Applying the SSNIP test ignores the fact that a firm
may already have market power. However, such considerations are not relevant for defining a
market in merger cases. . . . The cellophane trap means that a different approach is required in
abuse of dominance cases.”); Werden, supra note 10, at 526 (asserting non-applicability of the
Cellophane fallacy in merger cases); cf. P
, supra note 25, at 151 (stating that because the
Cellophane fallacy “may seem not to be a problem in a merger case,” “the criteria for defining
the market should be different in monopolization and merger cases,” but noting paradoxes with
See 2010 Merger Guidelines, supra note 24, § 4.1.2, ¶ 1 (“The Agencies apply the SSNIP
starting from prices that would likely prevail absent the merger. . . . If prices are likely to change
absent the merger . . . the Agencies may use anticipated future prices as the benchmark for the
See Werden, supra note 10, at 516 (“Markets are an analytical tool, and in economics and
law as well as in carpentry and auto mechanics the most useful tools are those designed for a
The government had originally made claims of attempted monopolization and conspiracy to
monopolize as well, but these theories were not before the Court on appeal. Cellophane, 351
U.S. at 379.
Cf. id. at 380 (“Market delimitation is necessary . . . to determine whether an alleged mo-
nopolist violates [Section 2]. The ultimate consideration in such a determination is whether the
really needed to address this element is debatable,
but insofar as market
definition was used, the appropriate market concept would have been obtained
by asking whether DuPont had sufficient market power to raise its price above
the competitive level, at least under assumptions favorable to the claim. The
HMT with base price equal to an estimate of Cellophane’s competitive price
would have validated markets responsive to this question.
This is not to say
that estimating that competitive price would have been easy.
It is simply to
observe that the competitive price is the baseline demanded by the substantive
law if market definition is to be used in assessing the monopolization theory
presented in cases like Cellophane.
To make that point another way, what the Court did in Cellophane was not
to define the wrong market but to define the right market for the wrong ques-
tion. The HMT with a base price equal to the current price—roughly the test
used in the case—validates markets responsive to the following question:
what group of competitors would a monopolist need to control or collude with
in order to further raise the price of its product? This is an interesting ques-
tion, and this market concept may well have been appropriate for assessing a
claim of attempted monopolization or conspiracy to monopolize.
lem is not, therefore, that the current price is the wrong baseline for every
market definition exercise in a Section 2 claim. The problem is simply that
this base price was inappropriate for the specific claim of monopolization that
the Court was trying to address in Cellophane.
The same logic applies when selecting the base price for defining relevant
markets in merger cases. An opportunity for illustration is the long-standing
unease attaching to whether the Cellophane fallacy applies in the merger con-
text. Is the proper HMT base price in merger cases the current price—even if
it reflects the ongoing exercise of market power—or is it an estimate of the
defendants control the price and competition in the market for such part of trade or commerce as
they are charged with monopolizing.”); S
, supra note 15, at 56 (“The purpose for market
definition in a monopoly case is to see whether the alleged monopolist has power to maintain a
price substantially higher than costs (or, by lowering price . . . to drive others out).”).
See generally supra notes 2, 5, and sources cited therein (proposing the use of estimates
derived from residual demand curves in place of market definition).
See Werden, supra note 108, at 139 (“The relevant question for assessing the firm’s market
power is whether the cross-elasticities of demand were so great near competitive price levels as
to prevent a significant elevation of prices above the competitive level in the first instance.”).
See generally Lawrence J. White, Market Power and Market Definition in Monopolization
Cases: A Paradigm Is Missing, in 2 ABA S
913 (W. Dale Collins ed., 2008) (discussing this and related challenges in the
Section 2 context); Carlton, supra note 108, at 19–20 (discussing issues with market definition in
the Section 2 context).
Cf. 2B A
, supra note 10, ¶ 539a, at 321 (commenting that relevant
markets defined around price increases may be appropriate in attempted monopolization cases).
The 2010 Horizontal Merger Guidelines continue a tradi-
tion of not really answering this question, instead defining the HMT baseline
as the price “that would likely prevail absent the merger.”
aspires to have a current-price default baseline but to allow for deviations
from this rule in vaguely identified special cases.
The theory-dependence of market definition provides a simple answer to
the choice-of-baseline question. The answer is—once again—that the proper
choice of base price depends on the theory of harm. If the concern is that a
merger will allow the remaining firms to elevate price above the current level,
then the current price is the appropriate baseline against which to define the
market. If the concern is that a merger could stabilize or entrench already
existing price elevation, then some measure of competitive pricing but for the
ongoing cause of price elevation is the appropriate baseline. The reason for
the difference is the fundamental difference in the question posed. In the case
of already ongoing coordination, the economic question in the entrenchment
theory is not “what firms would need to coordinate to further raise the price?”
Rather, the question is “what firms would need to deviate from coordination
to effectively depress the price?”
Again, an analytically helpful relevant
market must be customized to fit a particular theory of anticompetitive harm.
The theory-dependence of market definition also highlights deficiencies in
the Guidelines’ likely-future-price paradigm. While the Guidelines prescribe
the right result in the case of a theory of harm based on price elevation, they
mask the reason for this choice of baseline. Use of a current or likely-future-
price baseline is only appropriate when the theory of harm is an increase in
price above this current or likely-future-price level. If the theory of harm is
entrenchment of existing market power, then the appropriate choice of base-
line price is something like a competitive price—a result reached awkwardly,
if it is reached at all, by the likely-future-price paradigm.
Compare Gene C. Schaerr, The Cellophane Fallacy and the Justice Department’s Guide-
lines for Horizontal Mergers, 94 Y
L.J. 670 (1985) (suggesting that failure to adopt a com-
petitive price baseline benefits mergers between firms already exercising market power and may
result in overly broad relevant markets), with Werden, supra note 10, at 525–26 (suggesting that,
because “the ultimate question is whether a merger would create or enhance market power,” the
current price is the appropriate baseline “[even if it is already] well above competitive levels
because of collusion or monopoly”).
2010 Merger Guidelines, supra note 24, § 4.1.2, ¶ 1.
Cf. Werden, supra note 10, at 526 (“The only possible exception [to using current price as
the baseline] would be when a merger would strengthen a shaky cartel and prevent price from
falling. In this case, a price significantly below the prevailing price could be considered to be a
‘likely future’ price.”).
See generally Sean P. Sullivan, Anticompetitive Entrenchment, 68 U. K
. L. R
(2020) (discussing entrenchment theories of merger enforcement).
To illustrate, suppose that several firms have settled into a pattern of stable
coordination on elevated prices. If two of these firms propose to merge, a
possible theory of harm is that the merger may reduce the likelihood that a
future shock would disrupt this pattern of coordination—random disruption
being more likely the more independent concerns there are in the coordinated
As this theory of harm centers on disruption of coordination,
the proper base price for defining the relevant market is some measure of the
competitive price but for coordination. The Guidelines appear to recognize
this answer, but only through a vague and unexplained exception, rather than
by applying a general approach to market definition.
As another illustration, consider the proposed acquisition of a fringe com-
petitor, already in the market, by a dominant firm that has raised its price to
the limit that would trigger entry by higher-cost firms. An entrenchment the-
ory of harm is that this merger will allow the dominant firm to forestall possi-
ble future price depression by the fringe competitor, not that the merger will
allow the dominant firm to further raise its prices.
In defining the relevant
market for this theory of harm, the appropriate base price is some estimate of
the competitive price. It is debatable whether the Guidelines anticipate market
definition addressing this type of maintenance of monopoly situation at all.
Even if this situation is meant to be addressed, the likely-future-price para-
digm excuses reliance on a competitive base price only when that is the price
We are grateful to Robert Tovsky for suggesting this particular articulation of the theory of
anticompetitive entrenchment, which exists in various forms in caselaw and scholarship. See id.
at 1151–57 (identifying tacit collusion as a target for the entrenchment approach to merger en-
forcement); see also Am. Needle, Inc. v. Nat’l Football League, 560 U.S. 183, 195 (2010) (fo-
cusing on the existence of “separate economic actors pursuing separate economic interests,”
“independent centers of decisionmaking,” and “diversity of entrepreneurial interests” as the cata-
lysts of the “actual or potential competition” that is protected by the antitrust laws); Jonathan B.
Baker, Mavericks, Mergers, and Exclusion: Proving Coordinated Competitive Effects under the
Antitrust Laws, 77 N.Y.U. L. R
. 135, 163–73 (2002) (discussing the tensions in partial coordi-
nation and the role of disruptive “maverick” firms); Christopher R. Leslie, Trust, Distrust, and
Antitrust, 82 T
. L. R
. 515, 564–95 (2004) (discussing tensions in trust among cartel
2010 Merger Guidelines, supra note 24, § 4.1.2, ¶ 1 (“If prices might fall absent the merger
due to the breakdown of pre-merger coordination, the Agencies may use those lower prices as
the benchmark for the test.”).
See, e.g., Stanley Works, 78 F.T.C. 1023, 1066–68 (1971) (noting minimal increase in con-
centration resulting from a merger but focusing on the elimination of prospective competition
and foreclosure of future price reductions as anticompetitive concern), aff’d, Stanley Works v.
FTC, 469 F.2d 498, 505–08 (2d Cir. 1972).
Cf. Werden, supra note 10, at 526–27 (describing merger cases as focused on the question
whether a merger would “create or enhance market power” and thus interpreting deviation from
a current-price baseline as limited to a merger that would “strengthen a shaky cartel”).
that “would likely prevail absent the merger,”
circularly conditioning the
ability to analyze this theory of harm on its own conclusion.
The likely-future-price paradigm is doubly problematic. As shown above, it
obscures the logic of selecting a proper HMT base price in merger cases. But
its greater fault may be that its framing of market definition seems inadver-
tently to bias merger analysis against theories of anticompetitive harm alleg-
ing entrenchment of market power.
Indeed, these cases are rarely brought
today, despite both case law and statutory authority for mergers to be blocked
if they entrench rather than augment existing market power.
If antitrust au-
thorities are unwilling to rely on entrenchment theories in challenging merg-
ers, that unwillingness ought to follow from an explicit policy decision and
not from the adoption of market definition practices that just happen to pre-
clude challenging mergers on such grounds.
The need to condition market definition on a theory of harm has implica-
tions beyond those discussed above.
We will not discuss every implication
in detail, but four deserve highlighting. First, besides clarifying the level of
the HMT baseline, the need to customize market definition explains the selec-
tion of price concepts in the HMT. Whether a hypothesized price increase
should be relative to the immediate price of an intermediate product, to its
value-added price, or to the final price of the end product
should be guided
by the theory of harm. Second, whereas the 2010 Horizontal Merger Guide-
lines seem to treat price discrimination as a special case of market defini-
price discrimination markets are actually just applications of the
principle that market definition should be conditioned on a theory of harm. If
2010 Merger Guidelines, supra note 24, § 4.1.2, ¶ 1; see also id. (limiting deviations from
the current-price baseline to situations in which “prices are likely to change absent the merger,
e.g., because of innovation or entry”).
See Sullivan, supra note 128, at 1142–51 (describing entrenchment theories of harm in
horizontal merger enforcement).
See United States v. Phila. Nat’l Bank, 374 U.S. 321, 365 n.42 (“[I]f concentration is al-
ready great, the importance of preventing even slight increases in concentration and so preserv-
ing the possibility of eventual deconcentration is correspondingly great.”); Areeda, supra note
10, at 564 (“Merger precedents have been concerned not only with combinations creating new
power but also with those reinforcing present power. One need not endorse all the cases making
or misusing that point to accept the proposition that Clayton Act Section 7’s prophylactic man-
date is violated by a merger which reinforces pre-existing monopoly or oligopoly pricing.”);
2010 Merger Guidelines, supra note 24 (“The unifying theme of these Guidelines is that mergers
should not be permitted to create, enhance, or entrench market power or to facilitate its exer-
cise.” (emphasis added)).
See Salop, supra note 26, at 194–95 (outlining five analytic traps that can be avoided by
explicitly recognizing the theory-dependence of market definition).
Cf. Werden, supra note 10, at 534–38 (discussing these types of alternatives).
See 2010 Merger Guidelines, supra note 24, §§ 3, 4.1.4, 4.2.2.
the theory of harm is market-wide price elevation, it is unnecessary to specify
the customer component of the market. If the theory of harm is price elevation
to a subset of customers, then this should be reflected in the relevant market.
Third, when considering the appropriateness of including captive capacity or
excluding presently uncontested share from the definition of a market,
fixed answer can ever be given in the abstract. Like all aspects of market
definition, the appropriateness of including or excluding any given facet of
competition in a relevant market depends on the theory of harm and the ana-
lytical purpose for which the market is being defined. Fourth, the theory-de-
pendence of market definition helps to clarify market delineation in tying
cases. Without embarking on too detailed an aside on tying doctrine,
note that the value of defining a relevant market for the tying product differs
fundamentally from the value of defining a relevant market for the tied prod-
The former helps assess ability to leverage (a question of current mar-
ket power); the latter helps assess whether leveraging could harm consumers
(a question of the acquisition of market power). Approaching market defini-
tion from a theory-dependent perspective clarifies how and why the exercise
differs between the two separate relevant markets in this analysis.
To summarize, a relevant antitrust market must always be conditioned on a
specific theory of competitive injury, and courts and advocates should resist
the urge to see market definition as a rote, theory-independent process. Nor
need they submit to the independent-market fallacy. While the Court’s com-
mentary on market definition has not always been clear, the principle that
relevant markets should correspond to specific theories of harm is evident.
This was clear in Philadelphia National Bank, which defined the market by
looking to where competitive effects would be direct and immediate.
reaffirmed in Indiana Federation of Dentists, which explained that the point
of market definition is to determine “whether an arrangement has the potential
Cf. Fiona M. Scott Morton & Zachary Abrahamson, A Unifying Analytical Framework for
Loyalty Rebates, 81 A
L.J. 777 (2017) (discussing the complicated interplay between
contested and uncontested market shares in the context of contracts referencing rivals).
See generally 9 P
2011) (providing exhaustive treatment of the antitrust analysis of tying and related practices); 10
(3d ed. 2011) (same).
See Jefferson Parish Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2, 21 (1984) (“The requirement
that two distinguishable product markets be involved follows from the underlying rationale of the
rule against tying.”); Allen-Myland, Inc. v. IBM Corp., 33 F.3d 194, 200–01 (3d Cir. 1994)
(“The first inquiry in any § 1 tying case is whether the defendant has sufficient market power
over the tying product, which requires a finding that two separate product markets exist and a
determination of precisely what the tying and tied product markets are.”); H
§ 3.3.a1, at 127 (5th ed. 2016) (noting that market
power in the primary market may or may not exist independently of market power in an
aftermarket, such that “each of the markets must be evaluated separately); id. at 524–25
United States v. Phila. Nat’l Bank, 374 U.S. 321, 357 (1963).
for genuine adverse effects on competition.”
Where a theory of harm has
been identified, and to as much specificity as that theory allows, the delinea-
tion of relevant markets should always aim to identify market concepts tai-
lored to the specific injuries implied by the anticompetitive theory at issue.
III. THE SINGLE MARKET FALLACY
The single market fallacy is the common mistake of assuming each antitrust
case involves a single relevant market. Most markets encompass a variety of
products, and we are not suggesting that courts have failed to recognize this or
to define multiple product markets where different sets of products are impli-
cated by the facts. Rather, our concern is the assumption that a single relevant
market (or markets) applies in common to every aspect of an antitrust case.
The error in this assumption stems from both of the prior fallacies. The exis-
tence of a single market is congruent with the assumption of natural, free-
standing markets. And the mistake is amplified by a failure to grasp the
theory-dependence of relevant markets. When these foundational errors are
recognized, the single-market fallacy becomes obvious. The number of poten-
tial theories of harm implicated by a given fact pattern marks the upper limit
on the number of potentially helpful relevant markets to define.
If this seems obvious, note that it has not deterred legions of courts, practi-
tioners, and scholars from battling over selection of the relevant market for a
given case or investigation. Since the inception of antitrust litigation, trial
courts have treated market definition as a fact question to be answered defini-
tively by the fact-finder, taking advocacy under advisement but ultimately
finding by themselves what the relevant market is to be.
have proven no less fixated on trying to identify unique relevant markets,
at least on trying to distill principles for choosing a single market from among
FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460 (1986).
See, e.g., JBL Enters., Inc. v. Jhirmack Enters., Inc, 698 F.2d 1011, 1016 (9th Cir. 1983)
(“In determining what the field of competition is, courts are not free to accept whatever market is
suggested by the plaintiff, but must examine the commercial realities within the industry in
question.” (citations omitted) (internal quotation marks omitted)); Gough v. Rossmoor Corp.,
585 F.2d 381, 389 (9th Cir. 1978) (“Thus in determining the relevant market the courts are not
free to accept whatever market is suggested by the plaintiff as fitting most persuasively with his
contention that his power to compete effectively has suffered injury.”).
See, e.g., Areeda, supra note 10, at 584 (“[F]or each such product and region, there can be
only a single legally relevant market and not a multiplicity of legal relevant submarkets.”);
Horowitz, supra note 3, at 5 (“[T]he platitude that the geographic market is ‘the area of effective
competition’ fails to provide a comprehensive guide for delineating the geographic market that is
uniquely relevant for the antitrust issue in question.”); Werden, supra note 108, at 194–95 (stat-
ing that the smallest market principle means “there is a unique relevant market for every initial
In each of these examples, the focus of attention makes
sense only if one starts from the premise that there is and must be a unique
relevant market for each inquiry.
An accurate understanding of markets compels the conclusion that multiple
relevant markets can—and often should—be defined within a single case or
This multiple-market paradigm highlights aspects of current
practice in need of change. Specifically, the multiple-market paradigm sug-
gests that a distinct relevant market should be defined for each theory of harm
in a trial or investigation. It also suggests a subordinate role for courts and
defendants in market definition: the choice of relevant markets should be left
to the plaintiff, with fact-finders and defendants limited to the task of testing
the validity of whatever market the plaintiff proposes by applying the HMT or
a related concept. To be clear, the plaintiff’s choice is not unconstrained in
this approach: any relevant market must satisfy something like the HMT, and
any proposal failing such a test refutes the underlying theory of harm. But
where multiple alternative market delineations would satisfy the requirements
of the HMT or a comparable market concept, the choice among them is the
plaintiff’s. These may seem like tectonic changes, but they are all logical out-
growths of a principled approach to market definition, and they are all within
the bounds of what courts are already free to do today.
It is often considered beyond dispute that any antitrust case or investigation
should give rise to a single relevant market, or, where different products are
involved, a single set of relevant markets. A single definition of the relevant
market is usually assumed to apply to every theory or alternative theory in a
case. This is not to say that courts have never defined alternative markets in
antitrust cases. Indeed, some older merger cases multiplied markets with al-
most comedic abandon.
But the modern view is that market definition
should produce a single market concept germane to all aspects of the case or
Cf. Werden, supra note 100, at 117 (“Under the Guidelines, there are many markets but
generally only one relevant market, and it is determined by the smallest market principle, which
holds that the smallest market generally is the relevant market.”).
This should not be taken as a suggestion that market definition is always necessary. But if
market definition would be helpful to analysis, then a multiple-market paradigm is appropriate.
E.g., United States v. Bethlehem Steel Corp., 168 F. Supp. 576, 603 (S.D.N.Y. 1958) (find-
ing relevant markets to include “(1) the iron and steel industry, (2) hot rolled sheets, (3) cold
rolled sheets, and (4) hot rolled bars, in (a) the United States as a whole, (b) the northeast
quadrant of the United States, (c) Michigan, Ohio, Pennsylvania, and New York, (d) Michigan
and Ohio, (e) Michigan, and (f) Ohio, (5) buttweld pipe, (6) electric weld pipe, (7) seamless pipe,
(8) oil field equipment, (9) oil field equipment and supplies, (10) tin plate, and (11) track spikes,
in (a) the United States as a whole”).
with at most an allowance for nested price discrimination
markets in special circumstances.
This single-market philosophy manifests
in several observable practices.
One such practice is for courts to treat market definition as their own re-
sponsibility. Not always, but often, antitrust trials play out with the plaintiff
and defendant putting forth competing arguments for what the relevant market
and with the judge or jury deciding which of these definitions
Whatever market the court announces as the outcome of this tour-
nament becomes the relevant market for the case.
Closely related to the previous practice is scholarly discussion of market
definition as an exercise in selecting the best relevant market from among the
E.g., FTC v. Whole Foods Mkt., Inc., 548 F.3d 1028, 1036 (D.C. Cir. 2008) (“[T]he FTC
may have alternate theories of the merger’s anticompetitive harm, depending on inconsistent
market definitions . . . [but] on the merits, the FTC would have to proceed with only one of those
theories.”); 2B A
, supra note 10, ¶ 533, at 266 (“[D]egrees of constraint
do in fact vary [but] the ‘market’ for antitrust purposes is the one relevant to the particular legal
issue at hand.”); Werden, supra note 100, at 117 (“Under the Guidelines, there are many markets
but generally only one relevant market” (emphasis added)).
See, e.g., FTC v. Sysco Corp., 113 F. Supp. 3d 1, 40–48 (D.D.C. 2015) (finding both
“broadline foodservice distribution” and “broadline foodservice distribution to national custom-
ers” to be relevant markets for the case); 2010 Merger Guidelines, supra note 24, § 4.1.4, ¶ 1 (“If
a hypothetical monopolist could profitably target a subset of customers for price increases, the
Agencies may identify relevant markets defined around those targeted customers.”).
See, e.g., Sysco, 113 F. Supp. 3d at 24–25 (describing the competing market definition
proposals of the government and defendant); United States v. H & R Block, Inc., 833 F. Supp. 2d
36, 50 (D.D.C. 2011) (same); see also Turner, supra note 1, at 1150 (describing the typical trial
as follows: “you have two protagonists, one on each side, plaintiff and defendant, both seeking to
establish the market definition most favorable to them”); O’M
, supra note
81, § 150:132 (providing model jury instructions in which the plaintiff and defendant both offer
competing market definition proposals); ABA M
, supra note 81, at 109
(providing model jury instructions that include “defendant’s contention, if any, about the scope
of the relevant . . . market”).
See, e.g., supra note 144 and sources cited therein; FTC v. Arch Coal, Inc., 329 F. Supp. 2d
109, 123 (D.D.C. 2004), case dismissed, No. 04-5291, 2004 WL 2066879 (D.C. Cir. Sept. 15,
2004) (rejecting portions of both plaintiff’s and defendant’s proposed market definitions); see
also Turner, supra note 1, at 1152 (commenting that if the defendant’s market concept would not
lead to illegality, and the plaintiff “has not really shown enough to indicate that his market is
better, then [the plaintiff] loses.”). This is not a new practice. See R
211 (2000) (“The first opinion to consider
explicitly alternative market definitions and to give reasons for choosing one over the other was
Learned Hand’s opinion in [Alcoa]. The Supreme Court followed Hand’s example in
Cf. H & R Block, 833 F. Supp. 2d at 64 (“While some inappropriate proposed relevant
markets would be ruled out by the critical loss test, the fact that the test could still confirm
multiple relevant markets means that the Court must rely on additional evidence in reaching the
single, appropriate market definition.” (emphasis added)); Areeda, supra note 10, at 583 (com-
menting that submarkets allow courts to avoid choosing the relevant market from among alterna-
tives—implying that a choice is necessary in the first place).
The best market, in this approach, is typically under-
stood to mean the market concept that most accurately reflects the market
power of the relevant parties.
Again, the best or most representative market
is taken as the relevant market for a given application.
Finally, and related to both of the previous practices, the disfavored status
of Brown Shoe submarkets appears largely driven by the single-market idea.
This is not to say that submarket concepts were sound as originally conceived,
nor that we endorse their occasional abuse in antitrust practice.
It is simply
to say that the frequent claim that relevant markets and submarkets cannot
simultaneously coexist appears to rest on the assumption that different defini-
tions of the markets are always mutually exclusive.
What the foregoing examples show is that these futile practices and argu-
ments could be avoided if practitioners and courts understood that identifying
a unique or best relevant market is not a factual question that needs to be
resolved in antitrust cases and investigations. The irony of all the above exam-
ples is that the need for these practices and arguments would rarely arise if
unique or best markets really did exist in most antitrust applications. That they
do not might have been taken as a sign that economic reality is inconsistent
with the expectation of a single best market. This warning unheeded, the exis-
E.g., Kaplow, supra note 2, at 442 (assuming market definition to encompass the rule that
“the best market is that which yields the most accurate inference about market power”); Kaplow,
Alchemy, supra note 5, at 941 (“[I]t is well understood that, for any [market structure] statements
to be meaningful, one must look at the market shares in the relevant (best) market.”).
E.g., Areeda, supra note 10, at 583–84 (describing useful market definition as “identifying
the one product and geographic market that best gives the tribunal insight into the defendant’s
power with respect to each of his products or regions”); Kaplow, supra note 2, at 439 (“[Defin-
ing a relevant market] involves choosing from among candidate markets that which most accu-
rately depicts the extent of market power.”); see also S
, supra note 15, at 44 (“Courts in
monopolization cases usually begin by defining a single geographic and product market. In most
cases the effort is to identify what seems to be . . . the one market which is most meaningful
See Baker, supra note 28, at 206 (discussing some of the errors that have resulted when
practical indicia factors have been “applied blindly, without reference to the goals of identifying
buyer and seller substitution possibilities”).
E.g., 2B A
, supra note 10, ¶ 533c, at 269–70 (“The mischief of
submarket talk is the frequent supposition that a shoe market and an HQMS submarket can both
be simultaneously relevant to appraising the merger of two HQMS producers. Although that is
not possible . . .”); G.E. Hale & Rosemary D. Hale, A Line of Commerce: Market Definition in
Anti-Merger Cases, 52 I
. 406, 426 (1966) (“[T]he notion of a submarket is an odd
one: either there is or there is not a market in which competition may be affected. . . . If the line
of commerce is men’s shoes, it should not also be men’s golf shoes: if one boundary is right, the
other must be wrong.”); Turner, supra note 1, at 1151 (“If you have applied proper analysis in
trying to decide what the market is and, for example, you have concluded that price responsive-
ness among this group of products is so high that they really belong in the same market, that is
the end. . . . Once you have said these products are so closely substitutable they are in the same
market, there are no meaningful submarkets.”).
tence of alternative plausible market concepts has instead been seen as a chal-
lenge to be overcome—with confusion the only result of the effort.
The simultaneous existence of multiple relevant markets flows immediately
from economic fundamentals. Since economically meaningful markets are
mental constructs, and since in antitrust analysis these constructs are condi-
tioned on specific theories of anticompetitive harm, there is potentially a dif-
ferent relevant market for every theory of harm. There will often be additional
or alternative theories of harm in a single case or investigation and, if so, there
will be multiple relevant markets to be drawn. Whether these markets overlap,
nest, or intersect is of no consequence; each of the relevant markets serves a
To see the reason for this, note that to search for the best relevant market
gives rise to an obvious question: best for what? Best in the sense of whatever
market “most accurately depicts the extent of market power” is no answer.
The best market concept for gauging the possibility of unilateral harm from a
merger need not correspond to the best market for gauging the possibility of
coordinated price elevation as the result of the same merger. The best market
concept for gauging the possibility for modest coordinated price elevation
need not be the best for gauging the possibility of more substantial coordi-
nated price elevation. And the best market for studying the anticompetitive
potential of one type of exclusionary conduct need not be the best for studying
another type of conduct. As discussed in Part II, the appropriate definition of a
relevant market depends on the specific features of a theory of harm. For each
theory of harm implicated by a given fact pattern, a different relevant market
may be needed to guide analysis.
Logical clarity in antitrust requires a mul-
Because there may be many theories of harm in a given trial or investiga-
tion, there may be multiple helpful relevant markets in the trial or investiga-
tion. This simple proposition has remarkable potential to streamline and
clarify antitrust practice. We now discuss a few important implications.
Kaplow, supra note 2, at 439.
Cf. Pitofsky, supra note 1, at 1812–13 (“The tendency to see relevant market definition as
an all-or-nothing proposition rather than as an array of estimates with no market description
being exactly right has led to the most serious errors in antitrust enforcement.”); Easterbrook,
supra note 4, at 22 (“Usually the search for the ‘right’ market is a fool’s errand. . . . [T]here may
be tens of possible markets, each offering a little insight into conditions of competition.”).
1. Multiple Markets by Type of Harm
The most obvious reason for defining multiple markets is to tailor relevant
markets to distinct theories of harm. This is not an alien concept. The Su-
preme Court adopted just such a siloed approach to market definition in
Brown Shoe, conducting separate relevant market analyses for the vertical and
horizontal theories of harm at issue in that case.
The D.C. Circuit contem-
plated a similar separation of market definition by theory of harm in
But the principle applies to lower-level differences in the theory
of harm as well.
An example is the definition of relevant markets in horizontal merger cases
in which the plaintiff alleges both unilateral and coordinated theories of an-
ticompetitive injury. Despite drawing a sharp distinction between the analysis
of unilateral and coordinated effects, the 2010 Horizontal Merger Guidelines
can be read to imply that the same relevant market should be used to assess
both of these different theories of harm.
Recent merger cases have tended to
follow this suggestion.
This is a strange practice, as it is actually unclear when it would ever make
sense to use the same market concept to analyze both unilateral effects and
coordinated theories of harm. In unilateral effects analysis, there is a strong
argument that separate market definition analysis is unnecessary—the rele-
vant market being an output of the model rather than an input. But in coordi-
nated effects analysis, market definition can never be dispensed with, since it
plays the central role of identifying in the first instance the groups of produc-
ers that could achieve anticompetitive outcomes through coordination, and
whose coordination incentives must therefore be considered.
It is natural
Brown Shoe Co. v. United States, 370 U.S. 294, 325–28 (1962) (defining the relevant mar-
ket for the vertical aspects of the merger); id. at 336–39 (defining the relevant market for the
horizontal aspects of the merger).
United States v. Microsoft Corp., 253 F.3d 34, 81 (D.C. Cir. 2001) (noting that “[d]efining a
market for an attempted monopolization claim involves the same steps as defining a market for a
monopoly maintenance claim” before observing that “[t]he District Court never engaged in such
an analysis nor entered detailed findings defining what a browser is or what products might
constitute substitutes [in its analysis of the government’s attempted monopolization claim]”).
The Guidelines do not specifically oppose the definition of alternative relevant markets for
alternative theories of harm. They simply do not discuss this possibility at all. See generally 2010
Merger Guidelines, supra note 24, §§ 4, 6–7. This omission could be read to suggest the (im-
proper) norm that a common relevant market should apply across all alternative theories of harm.
See infra note 166 and accompanying text (discussing one way that this single-market fallacy
could hinder sound antitrust analysis).
E.g., United States v. H & R Block, Inc., 833 F. Supp. 2d 36, 50–71 (D.D.C. 2011)
(mechanically defining the same relevant market for both unilateral and coordinated effects anal-
ysis); FTC v. CCC Holdings Inc., 605 F. Supp. 2d 26, 38–44 (D.D.C. 2009) (same).
Compare Joseph Farrell & Carl Shapiro, Antitrust Evaluation of Horizontal Mergers: An
Economic Alternative to Market Definition, 10 B.E. J. T
., Vol. 10, No. 1, Art.
9, at 1 (2010) (describing definition of a relevant market as “clumsy and inaccurate in industries
and intuitive to define different relevant markets when assessing different the-
ories of anticompetitive harm.
The contrary practice, though common, is unnatural and unintuitive. There
is nothing to be gained, and much to be lost, from trying to force a compro-
mise between a market focused on the loss of competition between merging
firms and a market focused on the potential for coordination among the re-
maining firms in the market.
Worse yet is the possibility that a viable theory
of harm could be marginalized or omitted in order to protect the market con-
cept favorable to an alternative theory of harm. Nothing could be more flawed
than forcing an antitrust plaintiff to decline to allege a coordinated-effects
theory, for example, because the relevant market corresponding to that theory
of harm is broader than the relevant market implied by an additional unilat-
Multiple relevant markets can and should be defined to
focus analysis and avoid such unnecessary and self-defeating compromises.
2. Multiple Markets Within a Type of Harm
By the same logic, multiple relevant markets can be defined within a given
type of anticompetitive harm. This point was foreshadowed in our earlier dis-
cussion of how market definition depends on the implied magnitude of the
competitive effect corresponding to a theory of harm.
To illustrate, suppose
that six producers of differentiated products could, if united in a cartel, maxi-
mize profits by raising prices 10 percent above prevailing levels; suppose fur-
ther that three of these producers make products of closer similarity to each
other than to the others, and that these three could, if alone united in a cartel,
maximize profits by raising prices 5 percent above prevailing levels.
is the appropriate definition of the relevant market for assessing possible coor-
dination resulting from a merger of two of the three close competitors?
with differentiated products where the theory of harm is related to unilateral (rather than coordi-
nated) effects”), with J. Thomas Rosch, Litigating Merger Challenges: Lessons Learned, Re-
marks Presented at the Bates White Fifth Annual Antitrust Conference 2 (June 2, 2008),
reproduced at 5 H
App. E 165 (2018) (“A coordinated effects
challenge requires an assessment of who is ‘in’ and ‘out’ of a market. Only once the market
participants have been identified, can one assess the likelihood that a merger will facilitate the
coordination of pricing or output decisions and thus substantially lessen competition.”).
See Baker, supra note 28, at 216 (“A market definition analyzing the loss of localized
competition may well be unduly narrow for analyzing the likelihood of post-merger coordina-
tion, even though the same economic force, buyer substitution, is at stake in each.”).
Cf. Carlton, supra note 14, at 621 (discussing the different but related practice of the agen-
cies concentrating on unilateral effects analysis “when standard ‘coordinated effects’ analysis
based on market definition implies a very narrow market that might make agencies or courts
uncomfortable for advocacy purposes”).
See Part II.B.1.
This hypothetical is a variation on a puzzle described by Areeda and Hovenkamp, itself
based on prior discussion by Baker, Bresnahan, and others. See 2B A
supra note 10, ¶ 537d, at 311–12 and sources cited therein.
There are at least two possible theories of harm: (1) modest potential price
elevation resulting from coordination among the two remaining close compet-
itors, and (2) greater potential price elevation resulting from coordination
among the five remaining competitors overall. While a finding of likely harm
on the latter theory might obviate consideration of the former, it does not
follow that the latter theory is better or the only antitrust concern implicated
by these facts. Proper analysis of the competitive effects of this merger re-
quires that both theories of harm be considered.
Specific economic conditions are needed to bring about this type of tiered
system of price constraints, but such conditions are neither impossible,
even necessarily rare.
Insofar as these conditions arise, to insist, as modern
market definition practice seems to do, that one market concept or another
should be chosen as the relevant market only obscures the analysis. Harm in a
broad market deserves no less consideration than harm in a narrow market.
If both theories of harm are possible, then both warrant scrutiny.
market definition is to aid in this consideration, then each theory should be
assessed within its own relevant market.
A question that follows from the previous discussion is how antitrust viola-
tions should be defined in the absence of a singular-relevant-market concept.
What should happen if analysis suggests that anticompetitive injury has oc-
curred or is likely to occur in some relevant markets but not in others? Special
cases like out-of-market efficiencies aside,
the answer is that injury in any
See supra note 109 (discussing the tiered constraints that evidently beset sugar producers).
, supra note 141, § 3.2c, at 118 (“The existence of a relatively large rele-
vant market does not preclude the existence of smaller relevant markets within it.”); id. at
118–19 (discussing how market concepts can be diagramed as concentric or overlapping circles,
each representing a potential relevant market); S
, supra note 15, at 72 (“The position of
any seller can be [diagrammatically] represented within a series of concentric circles, each repre-
senting groups of other sellers which affect the subject seller less and less directly.”).
See Baker supra note 28, at 207 (“To the extent this slogan [there are no submarkets, only
markets] suggests that when a broad aggregation of products constitutes a market, a narrower
collection cannot also do so, it misleads.”); Werden, supra note 10, at 532 (“[I]t need not be the
case that the smallest market is a better basis for predicting the likelihood of collusion than a
slightly larger market.”).
See Baker, supra note 11, at 148 (“Recognizing the possibility of multiple markets in which
the competitive effects of firm conduct could be evaluated allows for more accurate targeting of
the competitive effects analysis in each case. It is appropriate to analyze firm conduct in any or
all relevant markets in which harm to competition may be found.”).
See 2010 Merger Guidelines, supra note 24, § 10, ¶ 6 n.14 (discussing efficiencies arising
outside of the relevant market); see also Ohio v. Am. Express Co., 138 S. Ct. 2274, 2282–83
(2018) (discussing related complications in two-sided markets).
properly defined relevant market is sufficient to establish a violation of the
This principle has further implications for market definition.
First, an immediate corollary is that market definition should largely be left
to the antitrust plaintiff. Since proof of anticompetitive injury in any relevant
market is sufficient, it is no answer for a court or defendant to point to other
alternative market definitions in which harm is unlikely to occur. If injury is
sufficiently proved in the plaintiff’s chosen market, that ends the inquiry.
This relieves antitrust plaintiffs of the burden of proving more than potential
injury in a chosen, valid relevant market.
Second, and closely related to the previous comment, the “smallest-market
principle” should be dropped in a multiple-market approach. As a reminder,
this principle instructs that, when several different markets might satisfy a test
like the HMT, the “smallest” or “narrowest” of these markets should usually
A charitable reading of the smallest-market principle is that it
approximates the previous comment in a single-market paradigm—suggesting
that if anticompetitive harm is possible in a narrow market, then that possibil-
ity should not be ignored simply because harm would not be possible in some
broader market. A less charitable reading is that the smallest-market principle
represents a crude heuristic for reducing the set of possible relevant markets to
a single choice that is hoped to be more likely than others to reflect common
theories of anticompetitive harm.
Either way, the principle is redundant in a
See supra note 172; 2010 Horizontal Merger Guidelines, supra note 24, § 10, ¶ 6 n.14
(“The Agencies normally assess competition in each relevant market affected by a merger inde-
pendently and normally will challenge the merger if it is likely to be anticompetitive in any
relevant market.”); see also Hovenkamp & Shapiro, supra note 9, at 1999 (“[T]he government
should be entitled to the structural presumption if the merger causes the requisite increase in
concentration in any properly defined relevant market. Even if the defense can identify an alter-
native relevant market (whether broader or narrower) in which the level or increase in concentra-
tion is insufficient to trigger the structural presumption, that showing does not negate or rebut the
This is obviously true at the pleading stage as a point of civil procedure. A plaintiff does not
fail to state a claim for relief, for example, because some alternative definition of the market may
be possible. See F
. R. C
. P. 12(b)(6). Our point is that the same is true at the merits stage. If
the plaintiff’s proposal of a cellophane market would be valid under a proper application of the
HMT, it is immaterial that a broader all-flexible-wrapping-materials market might also state a
valid relevant market.
, supra note 15, at 64 (“Economic theory, sensitively utilized, often suggests
that there is no one ‘right’ market, but congeries of interlinked ‘markets’ . . . . Thus, the party
asserting monopoly should have no burden other than that of showing a market which is plausi-
ble in the sense that those included within it have a clear and substantial commercial advantage
over those who are excluded from it in selling to a designated class of customers.”).
See, e.g., 2010 Horizontal Merger Guidelines, supra note 24, § 4.1.1, ¶ 5 (“[W]hen the
Agencies rely on market shares and concentration, they usually do so in the smallest relevant
market satisfying the hypothetical monopolist test.”).
See id. (justifying this approach on the basis that “the relative competitive significance of
more distant substitutes is apt to be overstated by their share of sales”); Werden, Answer, supra
note 7, at 742 (“Standardization is accomplished with the help of the ‘smallest market principle,’
proper understanding of market definition. The irrelevance of alternative
broader markets is explained above; the smallest-market principle is not
needed to reach this result. And leaving the choice of relevant markets up to
the plaintiff obviates the need to try to guess which market concepts would
best fit the theories of harm brought forth by plaintiffs in future cases. Provid-
ing plaintiffs with flexibility over the choice of relevant markets focuses anal-
ysis on the market concepts that are relevant to the actual theories of harm at
Third, allowing antitrust plaintiffs to select any relevant market satisfying
the HMT test does not allow the plaintiff to gerrymander an artificial and
As already explained above (Part I.C), antitrust markets
are not real or observable entities. Every market is in some sense artificial, so
to criticize any market as artificial or gerrymandered is a category mistake.
As explained in Part II.A, a relevant market—validated by the HMT or an-
other comparable standard—has demonstrated the potential for competitive
injury, at least under assumptions favorable to the theory of harm. This leaves
no room for objection to any valid relevant market chosen by the plaintiff. So
long as an antitrust injury is possible within the proposed market, there is no
basis or excuse for refusing to consider it.
Nor is there reason to waste time or attention on claims of gerrymandering,
better alternative markets, or any other facet of the single market fallacy. As
already noted, there is precedent for defining multiple relevant markets where
and all that we argue here is that multiple market con-
cepts are often helpful. Courts have never been prohibited from accepting any
holding that, of the markets that could be delineated around some starting product or location, the
relevant market is, roughly, the narrowest one.”).
Cf. Brown Shoe Co. v. United States, 370 U.S. 294, 368 n.3 (1962) (Harlan, J., dissenting in
part and concurring in part) (“If the Government were permitted to choose its ‘line of commerce’
it could presumably draw the market narrowly in a case that turns on the existence vel non of
monopoly power and draw it broadly when the question is whether both parties to a merger are
within the same competitive market.”); P
, supra note 25, at 145 (“Given enough flexibility
in market definition, high concentration becomes ubiquitous and a surprising number of innocu-
ous mergers can be made to appear dangerously monopolistic.”); Werden, supra note 10, at 532
(“[I]f the Guidelines permitted the exercise of considerable discretion in selecting the relevant
market, there would be considerable potential for gerrymandering.”).
See supra notes 78–79 and accompanying text.
See United States v. Pabst Brewing Co., 384 U.S. 546, 549 (1966) (“[W]hen the Govern-
ment brings an action under [Section 7] it must, according to the language of the statute, prove
no more than that . . . [the] effect of the merger may be substantially to lessen competition . . . in
any line of commerce ‘in any section of the country.’ . . . The language of this section requires
merely that the Government prove the merger may have a substantial anticompetitive effect
somewhere in the United States.”); Carlton, supra note 14, at 638 (“While I sense that enforce-
ment agencies may be reluctant to define [narrow markets]—for fear a court will think the defi-
nition is artificial—my view is that one should use and defend a narrow market if it is indeed
See supra note 160 and accompanying text.
valid market proposed by the plaintiff.
Nor have courts ever been bound by
the smallest-market principle or any other rote heuristic for defining relevant
On the contrary, in determining whether an arrangement has the
“potential for genuine adverse effects on competition,”
a court must con-
sider any market in which anticompetitive harm could arise. A multiple-mar-
ket paradigm is not only permissible; it is required.
IV. FOLLOWING THE LOGIC OF MARKET DEFINITION
Our objective has been to explain the logic of market definition and to
expose some troubling errors in market-definition practice. Having addressed
that objective, we now explore how the logic of market definition informs
market definition practice and some related aspects of antitrust law.
A peculiarly persistent artifact of the Supreme Court’s early work on mar-
ket definition is the vestigial notion that market definition must be performed
in every rule-of-reason case, or at least every merger case.
for requiring market definition is that this exercise weeds out insubstantial
antitrust injuries, a subject addressed in Part IV.B below. Another justification
conceives of market definition as the necessary starting point of all antitrust
Another views it as a necessary element in any claim for relief.
Indeed, modern jury instructions typically include proof of a relevant market
among the factual elements of a plaintiff’s case,
and failure to allege a plau-
See United States v. Aetna Inc., 240 F. Supp. 3d 1, 39–40 (D.D.C. 2017) (“The Guidelines
make clear that the hypothetical monopolist test does not aim to identify a ‘single relevant mar-
ket.’ . . . [T]he government ‘may evaluate a merger in any relevant market satisfying the [hypo-
thetical monopolist] test,’ and will ‘usually do so in the smallest’ market that qualifies. . . . The
government has operated within those parameters here.”).
See Baker, supra note 28, at 207 (“Although a court might often focus its concern and
analysis on the smallest such market, as the Merger Guidelines ‘generally’ recommend, a court is
entitled to identify a violation of the antitrust laws based on harmful effects in any market, even
one that is not the smallest.”).
See FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460 (1986).
See, e.g., Walker Process Equip., Inc. v. Food Mach. & Chem. Corp., 382 U.S. 172, 177
(1965) (“Without [market definition] there is no way to measure [an attempted monopolist’s]
ability to lessen or destroy competition.”); FTC v. Whole Foods Mkt., Inc., 548 F.3d 1028, 1036
(D.C. Cir. 2008) (Brown, C.J.) (“Inexplicably, the FTC now asserts a market definition is not
necessary in a § 7 case . . . in contravention of the statute itself.”).
E.g., FTC v. Cardinal Health, Inc., 12 F. Supp. 2d 34, 45 (D.D.C. 1998) (“Defining the
relevant market is the starting point for any merger analysis.”).
E.g., Keyte & Schwartz, supra note 75, at 589 (“[M]arket definition unquestionably remains
a statutory predicate to finding a Section 7 violation.”).
, supra note 81, § 150:66 (proposing model instructions in
which “proof of monopolization requires determination of [a relevant market]”); id. § 150:132
(“Determination of the relevant product market is a necessary predicate to the determination of
the legality of an acquisition under Section 7 of the Clayton Act.”); ABA M
sible relevant market may be taken as grounds for dismissal.
All of these
ideas are flawed because all are based on confused notions of what relevant
markets are and of what role they play in antitrust analysis.
As we have explained, relevant markets are analytical tools for evaluating
specific theories of anticompetitive harm.
The theory-dependence of rele-
vant markets invalidates the notion that market definition must be the starting
point of antitrust analysis. Even if it were possible to define a relevant market
without first considering potential theories of harm—and it is not—such rote
prioritization would only confuse analysis
and distract attention from the
ultimate question whether injury has occurred or is likely to occur.
The logic of market definition also invalidates the idea that a claim for
antitrust relief requires alleging a relevant market. What a relevant market
does is identify a group of transactions in which competitive injury could
occur, at least under conditions favorable to the theory of harm. Thus, at the
pleading stage, allegation of a relevant market is no more than explanatory
context for the allegation of competitive injury itself.
If injury has otherwise
been adequately plead, nothing but hollow formalism is advanced by insisting
on separate allegation of a relevant market. The same holds at the fact-finding
stage. Often the definition of a relevant market provides important context for
the fact-finder’s assessment of evidence of an alleged injury. And, as dis-
, supra note 81, at 106 (“Defining the relevant market is essential because you are required
to make a judgment about whether defendant has monopoly power in a properly defined eco-
E.g., City of New York v. Grp. Health Inc., 649 F.3d 151, 155 (2d Cir. 2011) (“To state a
claim under § 7 of the Clayton Act, §§ 1 or 2 of the Sherman Act, or New York’s Donnelly Act,
a plaintiff must allege a plausible relevant market in which competition will be impaired.”);
Queen City Pizza, Inc. v. Domino’s Pizza, Inc., 124 F.3d 430, 436 (3d Cir. 1997) (“Where the
plaintiff fails to define its proposed relevant market . . . a motion to dismiss may be granted.”).
See Part II.A (explaining this point in greater detail).
See Salop, supra note 26, at 189 (explaining that the “threshold test approach is fraught with
potential for error” because it is generally “impossible to evaluate market power accurately with-
out understanding the conduct and effect claims at issue and analyzing market power in the
context of those claims”); id. at 198 (observing that the use of market definition as a threshold
test can lead to a confused conclusion that a firm lacks market power when the very conduct at
question is targeted at allowing the firm to obtain such market power).
See Rosch, supra note 164, at 1 (“Judges have also often focused on market definition as a
‘threshold issue’ in merger litigation. I would suggest this is a mistake. A focus on market defini-
tion risks obscuring the ultimate question under Section 7 of the Clayton Act, which is whether
the transaction is likely to substantially lessen competition.”).
Cf. Carlton, supra note 14, at 626 (“[A] finding that a merger will have an anticompetitive
effect implies that competition in a particular economic market would be harmed. Viewed in this
way, an analysis that identifies an anticompetitive effect should be viewed as defining a market
in which a merger harms consumers.”); Kaplow, Alchemy, supra note 5, at 930 (“[I]f one insists
on market definition, one can satisfy such a formal doctrinal requirement by working back-
wards—which it appears courts and enforcement agencies already sometimes do.”).
where the plaintiff proposes a relevant market, validity
of the market concept remains a fact question to be tested by the trier of
But in those cases where an antitrust injury can be sufficiently proven
without explicit market definition, nothing is gained from objecting to the
omission of this inherently supportive and ancillary step in the analysis.
In sum, the logic of market definition clarifies that market definition can be
a useful step in antitrust analysis, but also clarifies that it is not a necessary
step. The Supreme Court recognized as much when it said that “Proof of the
section of the country where the anticompetitive effect exists is entirely sub-
sidiary to the crucial question . . . whether a merger may substantially lessen
competition anywhere in the United States.”
The per se rules under Section
1 of the Sherman Act exemplify this logic, and cases like Staples I show how
something akin to direct proof of anticompetitive injury can make separate
market definition redundant.
The ability of appropriate proof of competitive
effects to substitute for the definition of a relevant market is a proposition
with academic support,
and doctrinal support.
put, there is no need for courts or litigants to waste time and resources insist-
See supra notes 179–181 and accompanying text (discussing the limited role of defendants
and fact-finders in testing the validity of proposed relevant markets).
Strictly speaking, market definition under something like the HMT is a conjectural fact
question, as it is based on the assumption of a hypothetical monopolist (or cartel) that doesn’t
actually exist. A near analog is the mixed question of fact and law. See, e.g., Francis H. Bohlen,
Mixed Questions of Law and Fact, 72 U. P
. L. R
. 111 (1924); J.L. Clark, A Mixed Question
of Law and Fact, 18 Y
L.J. 404 (1909); Frederick Green, Mixed Questions of Law and Fact,
. L. R
. 271 (1901). Another analog is counterfactual reasoning. See generally Barbara
A. Spellman & Alexandra Kincannon, The Relation Between Counterfactual (“But For”) and
Causal Reasoning: Experimental Findings and Implications for Jurors’ Decisions, 64 L. & C
. 241 (2001).
United States v. Pabst Brewing Co., 384 U.S. 546, 549–50 (1966).
FTC v. Staples, Inc., 970 F. Supp. 1066, 1075–76 (D.D.C. 1997) (noting substantial price
differences depending on which office supply superstores competed in a given locality in con-
cluding that a merger of Staples and Office Depot would likely harm consumers).
See 2B A
, supra note 10, ¶ 521 (commenting that market definition is
unnecessary if anticompetitive effects are directly observable); see also John B. Kirkwood, Mar-
ket Power and Antitrust Enforcement, 98 B.U. L. R
. 1169 (2108) (suggesting how market
power may be inferred from the likely competitive effects of challenged conduct).
See 2010 Merger Guidelines, supra note 24, § 4, ¶ 2 (noting that “the Agencies’ analysis
need not start with market definition” and that some tools of competitive effects analysis do not
require formal market definition at all).
E.g., FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460–61 (1986) (“Since the purpose of the
inquiries into market definition and market power is to determine whether an arrangement has
the potential for genuine adverse effects on competition, proof of actual detrimental effects, such
as a reduction of output, can obviate the need for an inquiry into market power, which is but a
surrogate for detrimental effects.” (citations omitted) (internal quotation marks omitted)); see
also Allen-Myland, Inc. v. IBM Corp., 33 F.3d 194, 209 (3d Cir. 1994) (“Market share is just a
way of estimating market power, which is the ultimate consideration. When there are better ways
to estimate market power, the court should use them.”); Ball Mem’l Hosp., Inc. v. Mut. Hosp.
Ins., Inc., 784 F.2d 1325, 1336 (7th Cir. 1986) (same).
ing on traditional market definition when other evidence is sufficient to meet
the relevant standard of proof in demonstrating anticompetitive effects.
Closely related to arguments that market definition is necessary is the no-
tion that market definition ensures an injury is substantial enough to warrant
relief under the antitrust laws. This thinking appears to be derived from a bald
proposition in du Pont-General Motors: “Determination of the relevant mar-
ket is a necessary predicate to a finding of a violation of the Clayton Act
because the threatened monopoly must be one which will substantially lessen
competition ‘within the area of effective competition.’”
The vitality of this language in modern opinions is disproportionate to its
threadbare rationale. In the protectionist context of merger review in the
1950s and 1960s, a lay concept of the market might have helped to determine
which competitors required antitrust protection.
But under the modern con-
sumer-welfare standard, defining a relevant market has negligible value in
assessing or assuring the substantiality of antitrust injury.
To illustrate, consider a common justification for relying on a hypothetical
price increase of at least 5 percent when defining relevant markets under the
HMT. While the size of this price increase is not intended to signify a toler-
ance level for antitrust enforcement,
the usual argument for using a signifi-
cant price increase to test a relevant market under the HMT is transparently
just that. The Areeda & Hovenkamp treatise explains the typical thinking:
A “significant” price increase for market definition purposes must be large
enough to suggest that antitrust enforcement will be worth its cost while
minimizing interference with private activity that is generally desirable or
unavoidable though it creates small amounts of market power.
Whether a de minimis standard is necessary is a debatable proposition,
even assuming that it is, there is little defense for implementing it as a mini-
mum percentage price increase in market definition. That approach has two
United States v. E.I. du Pont de Nemours & Co. (du Pont-General Motors), 353 U.S. 586,
See supra notes 39–44 and accompanying text.
See supra note 103 and accompanying text.
, supra note 10, ¶ 537a, at 306.
One wonders, for example, why private plaintiffs or enforcement agencies need the gui-
dance of a de minimis standard to aid them in efficiently allocating their own resources. Perhaps
antitrust defendants deserve the protection of a de minimis standard as an assurance that small
anticompetitive injuries will not result in investigations or litigation. But even if this is accepted,
it is not obvious how any given choice of threshold would adequately protect defendants’ inter-
ests without also creating a vehicle for the accretion of market power through series of individu-
ally modest anticompetitive acts.
flaws. First, as noted earlier, the ultimate concern of a de minimis requirement
is the size of the competitive effect—and this has little correspondence with
the size of hypothesized price increase in something like the HMT.
if indeed some fixed quantity of harm were needed to justify relief, that
should be tested by a measure of total harm, not a percentage price increase. A
one percent increase in the price of $100,000,000 in transactions would seem
a far greater social concern than a 100 percent increase in the price of $1,000
in transactions. That a percentage price increase, alone, says so little about the
social significance of potential harm should arrest any suggestion that relevant
markets defined around less than 5 percent hypothetical price increases are
somehow categorically unworthy of antitrust scrutiny or attention.
In carpentry, a drill is a poor substitute for a hammer. So, too, with market
definition and the substantiality of antitrust injury. Relevant markets are tools,
but they are not tools for assessing the substantiality of injury, and they should
not be distorted to try to serve this purpose.
With courts increasingly required to apply antitrust law to two-sided plat-
the question naturally arises whether our approach has anything to
say about the appropriate definition of relevant markets in this context. The
salience of the question was underscored by the Supreme Court’s decision in
Ohio v. American Express,
delineating a controversial relevant market in a
platform context, and by related scholarly commentary asserting that standard
market definition approaches are inappropriate where platform competition is
The answer is that our approach to market definition is entirely
applicable in the two-sided platform context. The unique features of platform
competition must surely be accounted for, but the basic logic of market defini-
tion remains unchanged.
Indeed, American Express illustrates the problems that arise when the logic
of market definition is wrongly assumed to change between different contexts
and applications. The plaintiffs in this case alleged that “anti-steering” provi-
sions in American Express (Amex) merchant contracts constituted anticompe-
See supra notes 108, 107, and accompanying text.
See, e.g., Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018) (involving an antitrust claim
concerning American Express’s credit payment platform); Meyer v. Kalanick, 174 F. Supp. 3d
817, 819 (S.D.N.Y. 2016) (involving an antitrust claim involving Uber’s ride-sharing platform).
American Express, 138 S. Ct. at 2286 (holding that “courts must include both sides of the
platform . . . when defining the credit-card market”).
See, e.g., David S. Evans, Two-Sided Market Definition, in M
ch. XI § D ¶ 1 (2012) (“The standard approaches to market
definition do not apply to two-sided markets without significant modification. Even the two-
sided versions of these approaches should be used with caution.”).
titive vertical restraints under Section 1 of the Sherman Act.
provisions prohibited merchants from “steering” customers toward the use of
competing credit cards, even when those competing cards charged lower
merchant fees that might have been shared with the customer in the form of a
lower purchase price.
Focusing on the two-sided character of credit-card-
network platforms, the majority held that “courts must include both sides of
the platform—merchants and cardholders—when defining the credit-card
The majority opinion in American Express is peculiar in many respects, and
is marred by both substantive and procedural errors.
For present purposes,
our only interest is the confusion displayed by the majority opinion regarding
market definition in the two-sided platform context. Perhaps distracted by the
apparent novelty of the setting, the opinion deviated from appropriate market
definition principles in several important respects.
First, the majority opinion exhibits the natural market fallacy. After reciting
the naturalistic platitude that markets must “reflect[ ]” and “correspond to the
commercial realities of the industry,”
the opinion goes on to hold that, at
least where credit card networks are implicated, relevant markets must include
both the merchant and cardholder sides of the market.
The statement betrays
its naturalistic underpinnings. Categorical statements about how relevant mar-
kets should be defined in a given “industry” epitomize the confusion of lay
concepts of industry with antitrust relevant markets. As already explained,
relevant markets cannot be defined except by reference to a given theory of
anticompetitive harm. To define a market on the basis of industry or competi-
tive classifications (credit card networks or two-sided transaction platforms),
without regard to the theory of harm, is to commit a category error.
Second, the opinion illustrates the effects of the independent market fallacy.
The theory of anticompetitive harm at issue was that Amex’s anti-steering
Ohio v. Am. Express Co., 138 S. Ct. 2274, 2282–83 (2018).
Id. at 2282–83, 2288 (describing Amex’s anti-steering provision); Herbert Hovenkamp,
Platforms and the Rule of Reason: The American Express Case, 2019 C
. L. R
53, 77–78 (2019) (providing a clear explanation of how both customers and merchants could
have shared the benefits of having customers “steered” to lower-merchant-fee cards).
American Express, 138 S. Ct. at 2286.
See generally Hovenkamp, supra note 212 (noting that the majority opinion contained only
a single reference to the record, ignored without disrupting findings of fact that directly contra-
dicted its conclusions, and cited a complex academic literature while providing almost no analy-
sis of how that literature applied to the particular facts of the case); John B. Kirkwood, Antitrust
and Two-Sided Platforms: The Failure of American Express, 41 C
. 1805 (2020)
(noting these and other substantive problems with the majority opinion).
American Express, 138 S. Ct. at 2285 (citations omitted) (internal quotation marks omitted).
See supra note 213 and accompanying text; see also id. at 2286 (ostensibly expanding this
holding to cover a class of “two-sided transaction platforms”).
provisions were raising merchant fees,
which, in turn, were raising the retail
prices paid by consumers.
The undisturbed factual record suggests no rea-
son why a relevant market narrowly defined around the merchant side of the
credit-card transaction would have failed something like the HMT.
contrary, the record reveals direct evidence of the alleged injury, validating a
merchant-side relevant market by implication.
While the majority is cer-
tainly correct that network effects must be accounted for in the antitrust analy-
sis of this case, this does not require both the merchant and cardholder sides
of the platform to be included in the relevant market—at least, not under the
specific theory of harm at issue in this case.
Third, the majority opinion displays the single market fallacy. While the
plaintiff’s proposed market concept would evidently have satisfied something
like the HMT and was thus a valid relevant market,
the majority rejected
this valid choice of relevant market in favor of a different market more to the
majority’s liking. In fairness, the point was evidently to account for the possi-
bility that injury to one group of consumers might be offset by a benefit to
But even assuming that a benefit to one group would justify
anticompetitive injury to another, the majority was not compelled to conflate
both sides of the credit card platform into a single relevant market. Nothing
about the plaintiff’s proposed market precluded the consideration of offsetting
On the contrary, considering the alleged injury in light of the
United States v. Am. Express Co., 88 F. Supp. 3d 143, 215 (E.D.N.Y. 2015) (finding as fact
that the anti-steering provisions “allowed all four networks to raise their swipe fees” and “re-
sulted in higher all-in merchant prices across the network service market”).
Id. at 216 (finding as fact that “[m]erchants facing increased [fees] will pass most, if not all,
of their additional costs along to their customers in the form of higher retail prices”).
See American Express, 138 S. Ct. at 2293–94 (Breyer J., dissenting) (noting the majority’s
lack of discussion of the factual record and summarizing the district court’s findings).
See Hovenkamp, supra note 212, at 53 (noting that record evidence and basic economic
inferences on the support of it “established unambiguously” that the anti-steering provisions led
to “higher prices across the board”).
See Dennis W. Carlton & Ralph A. Winter, Vertical Most-Favored-Nation Restraints and
Credit Card No-Surcharge Rules, 61 J.L. & E
. 215, 217–18 (2018) (commenting on the false
belief that explicit accounting for two-sidedness is essential to competitive effects analysis and
stating that “[d]espite its two-sided property, one can analyze a credit card market using the
vertical structure of a one-sided market”); id. at 242 (“The two-sidedness of credit card markets
does not require a new set of economic principles for assessing competition policy because the
difference between the credit card setting and a conventional one-sided market is essentially a
matter of labeling.”); Hovenkamp, supra note 212, at 49 (noting that a court can always consider
network effects as part of its competitive effects analysis “[w]ithout relying on an economically
incoherent conception of a relevant market”).
See supra note 176 and accompanying text.
See American Express, 138 S. Ct. at 2286 (“Price increases on one side of the platform . . .
do not suggest anticompetitive effects without some evidence that they have increased the overall
cost of the platform’s services.”).
See Kirkwood, supra note 214, at 1838–46 (noting that it is possible to define the relevant
market as one side of a transaction while still considering effects on the other side).
plaintiff’s proposed market makes clear that the effects of Amex’s anti-steer-
ing provisions necessarily accrued to the detriment of both merchants and
marginal cardholders in this setting.
The antitrust implications of two-sided platforms are still being worked out
in a rapidly growing scholarly literature, and it may be some time before con-
sensus forms around the appropriate frameworks for analysis in this setting.
While the errors of American Express may seem like obstacles to the future
maturation of this area of antitrust practice, perhaps the opinion will be lim-
ited to its narrow facts or treated as sui generis in the arc of the Court’s recent
and more soundly reasoned decisions. Either way, American Express illus-
trates the dangers of straying from fundamentals in market definition, espe-
cially in the context of new and complex competitive environments.
Finally, we wish to recognize a reasonable critique of market definition as
we describe it in this article: that it is too complicated and too demanding to
be used in many practical applications. A common version of this critique is
that concepts like the HMT are simply too sophisticated for the minds of
generalist judges and juries, leading to unpredictable results in practice.
related idea is that sophisticated market definition is possible within the fed-
eral antitrust agencies, but that different techniques are needed when disputes
enter the generalist legal system.
Another version of the critique amounts to
Intuitively, if a cardholder valued the perks of using the Amex card more than the lower
effective price that a steering practice allowed, that cardholder would simply use the Amex card
despite the steering effort. The actual effect of the anti-steering provision is to stop some card-
holders from switching to other cards when their revealed preference is for a lower-price transac-
tion over the recoupment of Amex perks. See Hovenkamp, supra note 212 (providing a more
detailed explanation of this point).
See, e.g., Gopal Das Varma, Market Definition, Upward Pricing Pressure, and the Role of
Courts: A Response to Carlton and Israel, A
1 (Dec. 2010), www.americanbar.
courts—presided over by generalist judges—lack the economic sophistication that is required to
evaluate the merits of competing econometric analyses of market definition that are submitted by
opposing experts.”); J. Douglas Richards, Is Market Definition Necessary in Sherman Act Cases
When Anticompetitive Effects Can Be Shown with Direct Evidence, A
, Summer 2012, at
53, 57 (“[I]n retrospective analysis of the effects of past conduct, direct evidence of the actual
effects of such conduct is often more probative than comparatively confusing and misleading
market definition and market share analysis.”); Rosch, supra note 164, at 3 (“A case focused on
market definition risks getting bogged down in esoteric fights over the SSNIP test. Asking a
customer witness whether they would have switched to an alternative in the face of a 5% price
increase is arguably not a persuasive line of questioning.”); see also S
, supra note 15, at
63 (discussing concerns about the adjudicatory institution’s competence to make the kinds of
judgements that sophisticated economic theory demands).
Cf. D. Daniel Sokol, Antitrust, Institutions, and Merger Control, 17 G
1055, 1100–04 (2010) (discussing how merger review before the federal antitrust agencies dif-
fers from purely adjudicatory antitrust practice); Daniel A. Crane, Technocracy and Antitrust, 86
a pragmatic claim that data limitations often constrain and dictate the scope of
market definition more than any consideration of economic theory.
knowledge the challenges raised in these critiques, and we do not endeavor to
refute these essentially empirical claims about what judges and juries under-
stand and what available data allow.
Nor need we. Because even if all these claims were true, practicing market
definition without an understanding of its logical underpinnings would still be
inappropriate and counterproductive. Thus, the challenge of coping with data
limitations may well affect the plaintiff’s choice of relevant market, but a
proper understanding of market definition would allow the plaintiff to propose
whatever valid market the imperfect data support, and would foreclose com-
plaints that such a market is unrealistic or defective because it does not con-
form to an alternative for which there is insufficient data. The claim that
market definition practice before the agencies differs from market definition
practice before the courts may well be true today, but this only underscores
the need for a common and coherent rationale to guide market definition in
both contexts. Similarly, the challenge of guiding generalist judges and juries
through the intricacies of market definition may continue to weigh on liti-
gants, but that challenge only highlights the importance of exposing common
fallacies and misperceptions to help triers of fact discharge their market defi-
nition duties. The errors of American Express illustrate this need quite vividly.
Market definition is challenging. Nothing in our approach changes that. The
difficulty of the task makes clarity all the more important. Where the logic of
market definition reveals the irrelevance of a previously contentious consider-
ation, it simplifies the exercise. Where it calls for additional work—defining
different relevant markets for different theories of harm, for example—it com-
plicates the exercise. But in either case, the outcome is improved by focusing
attention on a coherent and economically meaningful market concept. Regard-
less of the constraints and limitations that beset the market definition exercise
in practice, the economic analysis of antitrust issues can only be strengthened
by clarifying the logic of market definition.
. L. R
. 1159, 1194–96 (2008) (similar); Wellford & Wells, supra note 80, at 2, 11
(describing growing divergence between merger advocacy before the courts and Agencies).
See, e.g., 2B A
, supra note 10, ¶ 530a, at 236 (“As a matter of
practicality . . . the only data we ever have is historical. . . . To at least some extent, future
behavior must be inferred from historical observations.”); S
, supra note 15, at 61 (“An-
other pragmatic factor is the availability of data. One can only count things for which there are
numbers. Unless exhaustive statistical surveys are to be done the parties must utilize either the
data gathered by the census taker, or the business records of firms or trade associations, or both.
Markets, then, will tend to be defined the way the Bureau of the Census has defined them, or the
way firms have perceived them, despite imperfections.”); Werden, Answer, supra note 7, at 742
n.59 (“In practice . . . relevant markets tend to be delineated on the basis of natural market
boundaries and hence are broader than absolutely necessary to satisfy the test.”).
As stated at the outset, our objective in this article has been to clarify the
logic of market definition. We have illustrated this logic in part by pointing
out several prominent mistakes in market definition thinking: the natural mar-
ket fallacy, the independent market fallacy, and the single market fallacy.
These fallacies are a vehicle for understanding the internal logic of market
definition. Because there is no meaningful natural market, relevant markets
are just analytic constructs. Because analytic constructs are tied to the subject
of analysis, relevant markets can be defined only with respect to particular
theories of anticompetitive harm. And because theories of anticompetitive
harm may be numerous in common applications, multiple helpful relevant
markets can and should be defined if and when doing so aids antitrust analy-
sis. The whole of this article, we hope, is a useful guide to the logic of market