PLATFORMS AND THE RULE OF REASON: 

THE AMERICAN EXPRESS CASE 

Herbert Hovenkamp* 

I. Introduction: Antitrust on Two-Sided Platforms ........ 35 
II. The Amex Case in the Supreme Court......................... 46 
III. Observations: Applying the Rule of Reason to  

 Platform Exclusion ........................................................ 56 
A. The Significance of Burden Shifting ..................... 56 
B. Platform Market Delineation................................. 57 
C. Complements in Use or Production ....................... 64 
D. Free Rider Concerns ............................................... 65 
E. Use of the Record .................................................... 68 
F. Inattentiveness to Economic Analysis .................. 71 
G.  Marginal Harms and Benefits ................................ 75 
H. Implications for Market Definition ....................... 81 

IV. Conclusion ...................................................................... 88 
 

I. INTRODUCTION: ANTITRUST ON TWO-SIDED 
PLATFORMS 

In Ohio v. American Express Co. (“Amex”),1 the Supreme 

Court had its first explicit opportunity to apply antitrust’s 

rule of reason to an allegedly anticompetitive practice on a 

two-sided platform. The writ of certiorari petition asked the 

Court to consider “how Section 1 of the Sherman Act, which 

bans unreasonable restraints of trade, applies to ‘two-sided’ 

 

* James B. Dinan University Professor, Penn Law and the Wharton 

School, University of Pennsylvania. Thank you to Dennis Carlton, Harry 

First, Irving Scher, and Erik Hovenkamp for comments. A version of this 

paper was delivered at the William Howard Taft Lecture, September 14, 

2018, to the New York State Bar Association, Antitrust Law Section.  

© 2018. Herbert Hovenkamp. All rights reserved. 
1 Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018). 



  

36 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

platforms that unite distinct customer groups.”2 The 

challenge was to a vertical interbrand restraint,3 intended to 

prevent merchants from steering customers away from 

American Express (“Amex”) and toward lower cost credit 

cards.4 The suit was originally brought during the Obama 

administration by the Antitrust Division and seventeen 

states.5 The government won in the district court,6 but the 

decision was reversed by the United States Court of Appeals 

for the Second Circuit.7 The 2016 presidential election 

intervened, and the United States, under the new 

administration did not seek certiorari, but eleven states who 

had been co-plaintiffs did.8 After certiorari was granted, 

though, the United States filed a brief on behalf of the 

plaintiffs.9 

Careful fact-finding is essential to the rational 

administration of antitrust under the rule of reason. Under 

antitrust’s per se rule, once a practice is shown to fall within 

a certain classification, such as naked price fixing, little 

additional evidence of anticompetitive effects is relevant and 

defenses are limited.10 By contrast, proper application of the 

 

2 Petition for Writ of Certiorari at i, Am. Express Co., 138 S. Ct. 2274 

(No. 16-1454). 

3 A vertical interbrand restraint, such as exclusive dealing, tying, or 

most favored nation requirements, consists of limitations on the way that a 

dealer can promote or sell a brand other than the one owned by the firm 

imposing the restraint. By contrast, an intrabrand restraint, such as resale 

price maintenance or a territorial limitation, is a restraint on the disposition 

of the imposing firm’s own brand. 
4 See Am. Express Co., 138 S. Ct at 2280.  

5 See United States v. Am. Express Co., 88 F. Supp. 3d 143 (E.D.N.Y. 

2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express 

Co., 138 S. Ct. 2274 (2018). 
6 Id. at 238–39. 
7 United States v. Am. Express Co., 838 F.3d 179, 207 (2d Cir. 2016), 

aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018). 
8 See Petition for Writ of Certiorari, supra note 2. 

9 Brief for the United States as Respondent Supporting Petitioners, 

Am. Express Co., 138 S. Ct. 2274 (No. 16-1454). 
10 See 7 PHILLIP E. AREEDA & HERBERT HOVENKAMP, ANTITRUST LAW ¶ 

1509 (4th ed. 2017). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 37 

rule of reason requires a searching and detailed factual 

examination and careful development of a record, enabling the 

court to understand the structure and economic effects of the 

defendant’s activities.11 This in turn obliges appellate courts 

to review the record developed in the district court. The 

Supreme Court’s Amex opinion should be tested against this 

requirement. 

A significant portion of the debate among the various 

courts and the Supreme Court majority and dissenters 

concerned the way market power and anticompetitive effects 

should be measured. Therefore, it is essential to consider what 

it means that this dispute took place on a “two-sided” 

platform. Although some people speak of two-side markets, 

that term creates some confusion when used in juxtaposition 

with the term “relevant market” in antitrust. As both the 

majority and the dissent made clear, the fact that a platform 

is two-sided does not entail that it should be treated as a 

single relevant market for antitrust purposes.12 Indeed, 

because they lack market power, many platforms are not 

relevant markets even if both sides are considered. 

Many firms sell complementary products, and often to 

different groups of buyers, but that fact alone does not make 

them two-sided platforms. Rather, a two-sided platform is a 

business that depends on relationships between two different, 

noncompeting groups of transaction partners.13 A traditional 

example is the printed periodical, such as a newspaper, which 

earns revenue by selling both advertising and subscriptions to 

the paper itself. Depending on the chosen business model, 

 

11 See id. ¶ 1507 (tracking factual allegations and proof burdens in rule 

of reason cases). 
12 Am. Express Co., 138 S. Ct at 2285–86; id. at 2300–01 (Breyer, J., 

dissenting).  
13 The classic treatment of two-sided markets can be found in Jean-

Charles Rochet & Jean Tirole, Platform Competition in Two-Sided Markets, 

1 J. EUR. ECON. ASS’N 990 (2003). On definitional problems, see generally 

Michael Katz & Jonathan Sallet, Multisided Platforms and Antitrust 

Enforcement, 127 YALE L.J. 2142 (2018). Many of the issues were initially 

raised in William F. Baxter, Bank Interchange of Transactional Paper: 

Legal and Economic Perspectives, 26 J.L. & ECON. 541 (1983). 



  

38 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

such a periodical might obtain very different mixtures of 

advertising and subscriber revenue. At one extreme, 

Consumer Reports does not sell advertising but derives its 

revenue entirely from subscriptions and donations.14 At the 

other extreme, the local neighborhood shopping flier might be 

distributed free to customers, with its production and 

distribution supported entirely by advertising revenues. The 

manager of a two-sided platform maximizes profits by coming 

up with the optimal mixture of participation and revenue on 

the two sides. 

Two-sided platform sellers can be harmed by feedback 

effects if they make the wrong choice on one side of their 

platform. For example, a magazine might keep its user 

subscription price low by relying on relatively more 

advertising, but this may cause subscribers to cancel their 

subscriptions due to excessive advertising. As this happens, 

the magazine will become less attractive to advertisers, 

leading to a vicious cycle of revenue loss on both sides. The 

trick for the magazine is to find the “sweet spot” that 

optimizes revenue between paid subscribers and paid 

advertisers. Such an optimized allocation is a consequence not 

merely of the price level on the two sides of the platform, but 

also of the amount of participation on each – that is, of 

appropriate “participation balanc[ing].”15 This spot, once 

achieved, is also an equilibrium for that firm.16 That is, it has 

 

14 See About Us, CONSUMER REP., https://www.consumerreports.org/ 

cro/about-us/support-our-work/index.htm [https://perma.cc/L75W-HQ9R]. 
15 See Erik Hovenkamp, Platform Antitrust, 44 J. CORP. L. (forthcoming 

2019) (manuscript at 11–13), https://papers.ssrn.com/sol3/papers.cfm? 

abstract_id=3219396 [https://perma.cc/5UFG-92XT]; see also Dennis W. 

Carlton, The Anticompetitive Effects of Vertical Most-Favored-Nation 

Restraints and the Error of Amex, 2019 COLUM. BUS. L. REV. 93, 101 (2019) 

(“The insight of Rochet and Tirole is that a two-sided market has the 

property that the price to each side of the market matters separately. That 

is, it is not only the sum of the prices that matters but also the relative 

prices on each side of the market.”) (citing Jean-Charles Rochet & Jean 

Tirole, Two-Sided Markets: A Progress Report, 37 RAND J. ECON. 645 

(2006)). Identifying and reaching this sweet spot is sometimes called 

“participation balancing.” See Hovenkamp, supra at 12–14. 
16 See Hovenkamp, supra note 15, at 13. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 39 

no reason to change the balance as long as circumstances 

remain the same. Of course, if something changes that 

balance—such as a large postage rate increase for magazine 

subscribers—then the firm may have to seek out a new 

equilibrium. Significantly, not only the aggregate level of fees, 

but also their distribution determines the point that 

maximizes the platform operator’s profits.17 The Second 

Circuit seemed to ignore this attribute of platforms in Amex, 

when it spoke of the network’s profitability as determined by 

the “net price” of using the Amex card but said nothing about 

participation balancing.18 The same net price might yield very 

different levels of output and profits depending on how the 

price is distributed between the two sides. 

In the platform literature, the term “indirect” network 

effects describes situations in which the value of the platform 

to one side depends on either the revenue generated or the 

number of users on the other side.19 For example, ride-hailing 

platforms such as Uber can succeed only if they have a critical 

volume of drivers on one side and a critical number of 

passengers on the other side.20 If fares are set too high, the 

number of passengers will fall off. If they are set too low, the 

number of drivers will fall off. Ongoing antitrust litigation 

alleges that Uber is facilitating price fixing among drivers 

because its platform computes fares that are the same for 

similar rides.21 But platform economics suggests otherwise. 

The situation is more similar to the one in Broadcast Music, 

Inc. v. Columbia Broadcasting System, Inc. (“BMI”), a 1979 

 

17 See generally Rochet & Tirole, supra notes 13, 15. See also Dennis W. 

Carlton & Ralph A. Winter, Vertical Most-Favored-Nation Restraints and 

Credit Card No-Surcharge Rules, 61 J.L. & ECON. 215, 236 (2018); 

Hovenkamp, supra note 15, at 11–12. 
18 See United States v. Am. Express Co., 838 F.3d 179, 203 (2d Cir. 

2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018). 
19 See generally Rochet & Tirole, supra note 15 ; see also DAVID S. EVANS 

& RICHARD SCHMALENSEE, MATCHMAKERS: THE NEW ECONOMICS OF 

MULTISIDED PLATFORMS 25 (2016). 
20 See Hovenkamp, supra note 15, at 4, 9. 
21 See Meyer v. Kalanick, 174 F. Supp. 3d 817, 820–21 (S.D.N.Y. 2016); 

see also infra notes 173–76 and accompanying text. 



  

40 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

Supreme Court decision that also involved a two-sided 

platform, although the Court did not identify it as such.22 In 

BMI, the Court considered whether nonexclusive licenses, 

granted by owners of recorded music and assembled by the 

defendant into a “blanket license” granting nonexclusively to 

broadcasters, amounted to unlawful price fixing by the 

copyright owners.23 The Court held that this practice was not 

unlawful, because petitioners’ blanket licenses were a highly 

valuable product that could be assembled only by the 

cooperation of the participating artists.24 The parallels to the 

Uber case are quite strong. 

The fact that a platform has two sides does not necessarily 

mean that both sides are positive contributors of revenue. It 

is important to distinguish between the revenue level, which 

is the aggregate price, and the revenue distribution, which is 

how the price is divided up among participants on the two 

sides.25 Sometimes the price to users on one side of the 

platform is zero. For example, traditional “free” over-the-air 

television is supported entirely by paid advertising. Viewers 

pay nothing for program access. This is also the case for most 

consumer web search engines, such as Google Search, Bing, 

and Yahoo, and social networking sites such as Facebook. 

These services are generally free to users, but are supported 

by advertising revenue.26 Nevertheless, those advertising 

revenues still depend on the number of users or the number 

of page views. 

In some cases, as in Amex itself, the revenue from one side 

can be negative.27 Credit card companies routinely charge 

 

22 Broad. Music, Inc. v. CBS, Inc., 441 U.S. 1 (1979). 
23 Id. at 8–9 (assuming that the artists were “literally” fixing prices). 
24 Id. at 20–21, 24. 
25 See Hovenkamp, supra note 15, at 11–12. 
26 Some social networking sites, such as LinkedIn, also have a 

“premium” version for which users pay a monthly fee. See LinkedIn 

Premium, LINKEDIN, https://premium.linkedin.com/ 

[https://perma.cc/LD7E-VWSJ]. 
27 Ohio v. Am. Express Co., 138 S. Ct. at 2274, 2281 (2018); see also 

United States v. Am. Express Co., 88 F. Supp. 3d 143, 203 n.36 (E.D.N.Y. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 41 

merchants acceptance fees for the use of the cards, while the 

cost to customers can be zero or even negative depending on 

the terms of customer card ownership.28 A typical card might 

charge no annual fee to customers, and no usage fees other 

than interest on unpaid balances or penalties for late 

payments. In addition, the card may award “perks” or other 

inducements that make the cost of the credit card negative to 

the consumer. These can include favorable treatment such as 

airline travel miles, extended warranties on products 

purchased with the card, or increased insurance protection for 

vehicles rented on the card.29 That was largely the case with 

Amex’s card offerings: many of the company’s cardholders 

paid nothing for ownership of the card but received perks for 

each purchase made with it.30 As a result, it was actually 

cheaper to use the card than to pay cash. Significantly, as the 

district court noted but the Supreme Court majority 

overlooked, these perks are granted for card use, not simply 

for card ownership.31 For example, consumers do not receive 

product purchase protection simply because they happen to 

carry an Amex card in their wallet; they must actually use the 

card to make the qualifying purchase.32 

Some so-called “transactional” platforms, including credit 

card networks and ride-hailing apps, exhibit a very direct 

 

2015) rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express 

Co., 138 S. Ct. 2274 (2018). 
28 The process is described in Steven Semeraro, Settlement Without 

Consent: Assessing the Credit Card Merchant Fee Class Action, 2015 COLUM. 

BUS. L. REV. 186, 196 (2015). See generally Dany H. Assaf & Rebecca 

Moskowitz, Global Credit Card Wars: Litigation, Legislation, or Innovation 

as a Path to Peace, 29 ANTITRUST 42 (2015). 
29 See, e.g., infra note 30. 
30 Current perks for various classes of American Express credit cards 

are summarized at Retail and Travel Benefits, AM. EXPRESS, 

https://www.americanexpress.com/us/credit-cards/features-

benefits/policies/index.html [https://perma.cc/P7SJ-P59T]. 
31 See Am. Express Co., 88 F. Supp. 3d at 191. 

32 See, e.g., Purchase Protection, AM. EXPRESS, 

https://www.americanexpress.com/us/credit-cards/features-

benefits/policies/purchase-protection/faq.html#5 [https://perma.cc/P3TZ-

9PMW] (limiting purchase protection to goods purchased with Amex card). 



  

42 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

relationship between transactions on one side and those on 

the other.33 The Supreme Court emphasized this in its 

peculiar approach to market definition.34 For example, each 

time a customer uses an Amex card to make a purchase, the 

platform simultaneously logs one transaction on the customer 

side and an equal and offsetting transaction on the merchant 

side, less Amex’s merchant acceptance fee.35 The same thing 

is true of Uber.36 Each time a passenger hails a ride, the 

passenger pays the fare through the Uber application and the 

driver is compensated accordingly, after Uber subtracts its 

fee. This one-to-one transactional correspondence does not 

apply to all two-sided platforms. Health insurance networks, 

newspapers, search engines, and streaming sites, for example, 

exhibit a less direct relationship between transactions on the 

two sides of the market. In a market such as free television, 

advertising volume and rates might be based on Nielsen or 

other surveys that assess the size and composition of the 

audience.37 Advertising rates on a search engine such as 

Google are often based on clicks, which means that more 

heavily used search engines generate more advertising 

 

33 See Ohio v. Am. Express Co., 138 S. Ct. 2274, 2280 (2018) (citing 

Benjamin Klein, Andres V. Lerner, Kevin M. Murphy & Lacey L. Plache, 

Competition in Two-Sided Markets: The Antitrust Economics of Payment 

Card Interchange Fees, 73 ANTITRUST L.J. 571, 580, 583 (2006)). 
34 See infra notes 82–86 and accompanying text. 
35 See id. 
36 See Brief of the International Air Transport Association and Airlines 

for America as Amici Curiae in Support of Petitioners at 5–6, Am. Express 

Co., 138 S. Ct. 2274 (No. 16–1454). 
37 See Advertising Effectiveness, NIELSEN, https://www.nielsen.com/ 

us/en/solutions/advertising-effectiveness.html [https://perma.cc/TLL5-

63FX]; see also United States v. Gray Television, Inc., No. 1:15-CV-02232, 

2016 WL 1064377, at *10–11 (D.D.C. Mar. 3, 2016) (noting how television 

advertising rates are computed from Nielsen ratings based on audience size 

and demographic characteristics). Some pay television contains advertising 

as well, and Nielsen data may also be used to compute ad rates for such 

products. See, e.g., Dish Enlists Nielsen Digital Measurement to Power 

Advanced Advertising Across Sling TV, NIELSEN (Sept. 20, 2018), 

https://www.nielsen.com/us/en/press-room/2018/dish-enlists-nielsen-

digital-measurement-to-power-advanced-ads-across-sling.html 

[https://perma.cc/CM2W-CL93]. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 43 

revenue.38 The volume and price of advertising is certainly 

affected by these measures of traffic, but it is hardly true that 

a one-to-one correspondence exists between a viewer’s activity 

and the purchase of advertising.  

The antitrust challenge in Amex was to an “antisteering” 

rule that Amex imposed upon merchants accepting its credit 

cards.39 Amex charges merchants an acceptance fee, typically 

a percentage of the transaction price, that can run much 

higher than the fee charged by competing credit card issuers 

such as Visa, Mastercard, and Discover.40 Many customers 

who carry the Amex card likely also carry one or more other 

cards.41 This incentivizes the merchant to induce the 

customer to use a less costly card, which it might do by 

offering the customer a product price discount or other 

compensation if she agrees to switch. For example, if the 

merchant acceptance fee on a large purchase is thirty dollars 

with an Amex card, but only twenty dollars with a Visa card, 

the merchant might wish to offer the customer a price 

discount of six dollars for using the Visa card rather than 

Amex. Alternatively, it might offer free delivery or some other 

valuable good or service. The antisteering rule prevents the 

merchant from making this offer, or even from informing the 

customer that the Amex card was more costly to use.42 The 

rule does not apply to transactions that do not use a card at 

all, such as payment by cash or checks, and it does not apply 

 

38 For a summary of how these alternatives compute advertising rates, 

see generally Web Tracking Solutions, L.L.C. v. Google, Inc., No. 08-CV-

03139, 2011 WL 3418323 (E.D.N.Y. July 27, 2011) (providing the magistrate 

judge’s report and recommendation); see also Nathan Newman, Search, 

Antitrust, and the Economics of the Control of User Data, 31 YALE J. ON REG. 

401, 413–14 (2014). 
39 See Am. Express Co., 138 S. Ct at 2280. 
40 The district court found that Amex maintained higher merchant 

acceptance fees than rival cards, although the difference had been declining. 

United States v. Am. Express Co., 88 F. Supp. 3d 143, 200 (E.D.N.Y. 2015), 

rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express Co., 

138 S. Ct. 2274 (2018). The differences tended to be higher in purchases 

involving airlines, rental cars, and lodging. Id. 
41 See id. at 178. 
42 Id. at 165. 



  

44 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

to debit cards.43 The government alleged that the antisteering 

rule effectively forced customers to stay with the higher priced 

card, thus increasing not only merchant fees, but also product 

prices indirectly.44 Visa and Mastercard also used their own 

versions of such provisions, but signed consent judgments 

agreeing to abandon the practice.45 

“Steering” is fundamental to competition of any kind, 

including competition among platforms. It offers market 

participants an incentive to seek out lower cost alternatives. 

Some platforms are “single-homing,” which means that users 

typically engage with only one platform.46 For example, 

smartphones are costly, and managing two different phones 

would be inconvenient. For that reason, most smartphone 

users have only a single phone at a time. iPhone users 

purchase their apps on the App Store, and Android users 

purchase them on Google Play. Competition for a particular 

user exists for the platform—the smartphone itself—rather 

than among platform incumbents. Credit cards are different, 

however. They readily accommodate “multi-homing.”47 

Cardholders often own two or more general purpose credit 

cards and use whichever satisfies them most for a particular 

transaction. The same is true of ride hailing services, web 

browsers, and computer search engines. Competition among 

 

43 Id. 
44 See Am. Express Co., 138 S. Ct at 2288–89. As the Supreme Court 

described the antisteering rule, it prohibits: 

. . . merchants from implying a preference for non-Amex 

cards; dissuading customers from using Amex cards; 

persuading customers to use other cards; imposing any 

special restrictions, conditions, disadvantages, or fees on 

Amex cards; or promoting other cards more than Amex. The 

antisteering provisions do not, however, prevent merchants 

from steering customers toward debit cards, checks, or cash. 

Id. at 2283. 
45 See id. at 2293 (Breyer, J., dissenting); see also United States v. Am. 

Express Co., No. 10–CV–4496, 2011 WL 2974094, at *2 (E.D.N.Y. July 20, 

2011) (approving proposed consent judgment, 75 Fed. Reg. 62858–02 (Oct. 

13, 2010)). 
46 See Hovenkamp, supra note 15, at 18–19. 
47 See Hovenkamp, supra note 15, at 17. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 45 

platforms can help ensure competitive prices and high-quality 

service, but an antisteering rule, such as Amex’s, eliminates a 

consumer’s incentive to use the least costly alternative. 

The credit card steering problem presents some analogies 

to the existing economic literature on cartels. Limitations on 

price competition encourage firms to compete in ways other 

than price. For example, although cartel members may be 

forbidden from cutting the nominal price, they may compete 

with one another on nonprice terms, such as by offering perks 

that may be equivalent to those offered for credit card use.48 

At the margin, the cartel members may throw in nonprice 

perks right up to the point that their costs equal the price 

level. For example, if the competitive price is ten dollars but 

the cartel price is fourteen dollars, the cartel members may 

end up competing against one another by including nonprice 

perks costing them up to four dollars. Even apart from cartels, 

however, mixtures of price and nonprice competition are 

ubiquitous.  

Steering facilitates both price and nonprice competition by 

permitting merchants to reward cardholders for using less 

costly forms of payment. Card issuers can compete either by 

cutting their merchant acceptance fee or by increasing their 

perks. With steering, merchants can permit customers to 

choose between a lower product price obtained by using a card 

with a lower acceptance fee, or a higher-price product 

purchased with a card that rewards the customer with higher 

perks. An antisteering rule deprives customers of the 

opportunity to make this choice, at least among alternative 

credit cards. In the process, it serves to blunt both price and 

nonprice competition. 

Some Amex cardholders place greater value on the perks 

than others—a point that the Amex majority overlooked. It 

assumed that the higher transaction fees are justified by 

increased perks, as if they conferred the same value to 

 

48 See generally George J. Stigler, Price and Non-Price Competition, 76 

J. POL. ECON. 149 (1968). On application to transactions and merchant 

credit card fees, see Carlton & Winter, supra note 17, at 218, 230. See also 

In re Text Messaging Antitrust Litig., 782 F.3d 867 (7th Cir. 2015) (treating 

a rise in non-price competition as evidence of collusion). 



  

46 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

everyone. The very fact that steering worked, however, 

indicates that at least some Amex customers preferred the 

lower product price rather than the perks. Were that not true, 

a steering rule would not have been necessary. 

This Article considers how antitrust’s rule of reason should 

be applied to an exclusionary practice on a platform market. 

It considers the rule of reason’s basic burden-shifting 

framework, unique elements of market delineation on 

platform markets and the relevance of placing production 

complements into the same “market.” It also considers the 

Court’s regressive, antieconomic conclusion on a proposition 

that was never briefed—whether a market definition is 

necessary in an antitrust challenge to a vertical practice. Then 

it considers the Court’s odd treatment of free rider problems. 

It also faults the Court for paying so little attention to the 

record, its lack of economic analysis, and in particular its 

confusion of total with marginal harms and benefits. Finally, 

it looks at the implications of the Court’s decision for market 

delineation in cases involving platforms. 

II. THE AMEX CASE IN THE SUPREME COURT 

In Amex, the Supreme Court dismissed the government’s 

challenge to the Amex antisteering rule,49 affirming the 

Second Circuit’s decision reversing the district court. The 

Supreme Court acknowledged that antitrust’s rule of reason 

involves a “three-step, burden-shifting framework[.]”50 First, 

the plaintiff must show “that the challenged restraint has a 

substantial anticompetitive effect that harms consumers in 

the relevant market.”51 If the plaintiff carries this burden, 

“then the burden shifts to the defendant to show a 

procompetitive rationale[.]”52 If this showing is successful, 

“then the burden shifts back to the plaintiff to demonstrate 

that the procompetitive efficiencies could be reasonably 

 

49 Am. Express Co., 138 S. Ct. at 2290. 
50 Id. at 2284.  
51 Id. (citations omitted). 
52 Id. (citations omitted). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 47 

achieved through less anticompetitive means.”53 Because the 

dissenters agreed with this verbal formulation of the rule of 

reason, it appears to have the unanimous support of the 

Court.54 The differences lay mainly in how the prima facie 

case must be made out and how market power is to be 

established. 

The Court concluded that the government did not carry its 

burden at the first stage. It observed that anticompetitive 

effects could be shown in two ways—either “directly,” by 

“‘proof of actual detrimental effects on competition,’”55 which 

could include “reduced output, increased prices, or decreased 

quality in the relevant market[.]”56 It could also be shown 

“indirectly,” by “proof of market power plus some evidence 

that the challenged restraint harms competition.”57 As the 

dissent observed, both of these descriptions were inconsistent 

with traditionally accepted requirements under the rule of 

reason. First, “direct” proof of actual detrimental effects does 

not require a market definition;58 however, the majority spoke 

of direct evidence of “reduced output, increased prices, or 

decreased quality in the relevant market” as if it did.59 As 

Justice Stephen Breyer noted in his dissent, “[o]ne critical 

point that the majority’s argument ignores is that proof of 

 

53 Id. 
54 See id. at 2290 (Breyer, J., dissenting) (“I agree with the majority 

and the parties that this case is properly evaluated under the three-step 

‘rule of reason’[.]”). 
55 Id. at 2284 (quoting FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 460 

(1986)). 
56 Id. (internal citations omitted). 
57 Id. (citing Tops Mkts., Inc. v. Quality Mkts., Inc., 142 F.3d 90, 97 (2d 

Cir. 1998); Spanish Broad. Sys. of Fla. v. Clear Channel Commc’ns, Inc., 376 

F.3d 1065, 1073 (11th Cir. 2004)). 

58 E.g., 2B PHILLIP E. AREEDA, HERBERT HOVENKAMP & JOHN L. SOLOW, 

ANTITRUST LAW ¶521 (4th ed. 2014); Louis Kaplow, Why (Ever) Define 

Markets, 124 HARV. L. REV. 437 (2010). 
59 Am. Express Co., 138 S. Ct. at 2284. The Court cited FTC v. Ind. 

Fed’n of Dentists, 476 U.S. 447 (1986), but that decision never spoke of 

relevant market. 



  

48 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

actual adverse effects on competition is, a fortiori, proof of 

market power.”60 The dissent continued: 

The District Court’s findings of actual anticompetitive 

harm from the nondiscrimination provisions thus 

showed that, whatever the relevant market might be, 

American Express had enough power in that market 

to cause that harm. There is no reason to require a 

separate showing of market definition and market 

power under such circumstances. And so the 

majority’s extensive discussion of market definition is 

legally unnecessary.61 

Second, the Court’s formulation of “indirect” proof as 

requiring “proof of market power plus some evidence that the 

challenged restraint harms competition”62 resembles the 

general case for competitive harm under the rule of reason, 

but misses the point of “indirect” proof, which is that it draws 

inferences of market power from a market share and other 

features of a properly defined relevant market. 

In Amex, the plaintiff had relied on direct proof, which 

would not ordinarily require a market definition. As discussed 

below, however, the Court held that for vertical cases such as 

this one, even direct proof required a market definition.63 The 

Court then concluded that the relevant market consists of 

both sides of the platform as “the area of effective 

competition.”64 For example, firms such as the defendant earn 

a profit by maximizing revenue across both sides of the 

platform and can do so even if one side operates at a loss.65 

Further, “[p]rice increases on one side of the platform likewise 

do not suggest anticompetitive effects without some evidence 

that they have increased the overall cost of the platform’s 

services.”66 

 

60 Id. at 2297 (Breyer, J., dissenting). 
61 Id. (Breyer, J., dissenting). 
62 Id. at 2284. 

63 See infra text accompanying notes 76–77. 
64 Id. at 2285–86. 
65 See id. 
66 Id. at 2286. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 49 

The Court cited a great deal of literature on distinctive 

features of two-sided platforms. It concluded that “in two-

sided transaction markets, only one market should be 

defined.”67 Why these observations entailed that 

noncompeting goods should be grouped into the same relevant 

market is not clear. The price and output of complements 

certainly affect a firm’s profit-maximizing output and price,68 

but that hardly requires redefinition of its market.69 Further, 

the Court’s discussion about the “transactional”70 nature of 

Amex’s platform applies to conventional markets where 

sellers and buyers meet face to face. It is hardly unique to 

platforms. For example, if a gardener pays three dollars for a 

packet of spinach seeds in a hardware store there is a single 

simultaneous transaction, but we would never define a single 

market for gardeners and spinach seeds. 

Without relying on an economically incoherent conception 

of a relevant market, the Court could simply have said that 

when power is sought to be proven by direct effects all relevant 

effects should be considered. It does not matter whether these 

effects occur in the same relevant market, because no relevant 

market need be defined in the first place. This is more 

consistent with Justice Breyer’s dissenting approach which (1) 

 

67 Id. at 2287 (internal quotations omitted) (quoting Lapo Filistrucchi, 

Damien Geradin, Eric van Damme & Pauline Affeldt, Market Definition in 

Two–Sided Markets: Theory and Practice, 10 J. COMPETITION L. & ECON. 

293, 302 (2014); see also David S. Evans & Michael Noel, Defining Antitrust 

Markets When Firms Operate Two–Sided Platforms, 2005 COLUM. BUS. L. 

REV. 667, 671 (2005)). 
68 In general, the presence of a substitute serves to increase a firm’s 

own price elasticity of demand, thus reducing its power; the presence of a 

complement serves to reduce a firm’s own price elasticity of demand, thus 

increasing its power. Both are considered in ordinary methodologies for 

computing residual demand. See generally Aviv Nevo, Mergers with 

Differentiated Products: The Case of the Ready-to-Eat Cereal Industry, 

31 RAND J. ECON. 395 (2000); Gregory J. Werden, Demand Elasticities in 

Antitrust Analysis, 66 ANTITRUST L.J. 363 (1998). On application to two-

sided platforms, see generally Andre Boik & Kenneth S. Corts, The Effects 

of Platform Most-Favored-Nation Clauses on Competition and Entry, 59 J. 

L. & ECON. 105 (2016). 
69 See infra notes 110–11 and accompanying text. 
70 See Am. Express Co., 138 S. Ct. at 2286–87. 



  

50 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

eschewed reliance on market definition;71 but (2) would 

consider all effects rather than just benefits on one side of the 

platform.72 The district court was also clear on that point, 

although the majority opinion ignored it. Indeed, one of the 

district court’s fact findings was that the antisteering rule 

resulted in higher product prices across the board for 

merchants who accepted the Amex card, whether or not the 

customer used that card.73 That finding alone was sufficient 

to establish the defendant’s power, as well as anticompetitive 

effects. 

In a footnote, the Court concluded that while direct proof 

of market power does not require proof of a relevant market 

in a horizontal case, it did in a vertical case such as this. The 

Court stated: 

The plaintiffs argue that we need not define the 

relevant market in this case because they have offered 

actual evidence of adverse effects on competition—

namely, increased merchant fees. We disagree. The 

cases that the plaintiffs cite for this proposition 

evaluated whether horizontal restraints had an 

adverse effect on competition. . . . Given that 

horizontal restraints involve agreements between 

competitors not to compete in some way, this Court 

concluded that it did not need to precisely define the 

relevant market to conclude that these agreements 

were anticompetitive. . . . But vertical restraints are 

different. Vertical restraints often pose no risk to 

competition unless the entity imposing them has 

market power, which cannot be evaluated unless the 

Court first defines the relevant market.74 

This confusing statement appears to do no more than 

assume the conclusion. The Court did not clarify why 

 

71 See id. at 2294–96. 
72 See id. at 2296–97. 
73 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 208 

(E.D.N.Y. 2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018); see also infra notes 146–53 and 

accompanying text. 
74 Am. Express Co., 138 S. Ct. at 2285 n.7 (internal citations omitted). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 51 

horizontal and vertical restraints should be treated differently 

in situations where both require proof of market power. One 

possibility, which Justice Breyer mentioned in his dissent, is 

that the majority believed that there was some category of 

anticompetitive effects that could be established without 

market power.75 The final sentence of the statement appears 

to conclude as a statement of law something that in reality 

presents a question of fact, and that in any event is incorrect. 

There is no obvious reason why power cannot be inferred from 

effects in both horizontal and vertical cases. For example, in 

exclusive dealing cases, which are vertical, evidence that a 

defendant was able to exclude a rival or suppress its sales 

even while keeping its own price high is certainly probative, 

as the Eleventh Circuit found in McWane, Inc. v. FTC.76 

Direct proof has its own limitations, of course. The facts must 

indicate that exclusion is a consequence of anticompetitive 

behavior rather than efficiency, but there is no obvious reason 

for thinking these things are fundamentally different in a 

vertical case. 

Further, the Court’s analysis is regressive, given the 

significant work in economics that both weakens the case for 

traditional market definition and improves upon econometric 

methodologies for measuring market power more directly.77 

This is particularly true when the immediate concern is the 

ability of a firm or group of firms to increase price above the 

competitive level through means other than collusion.78 When 

the issue is likelihood of collusion, on the other hand, then 

market definition may be an aid in identifying those in the 

 

75 See id. at 2297 (Breyer, J., dissenting) (“One critical point that the 

majority’s argument ignores is that proof of actual adverse effects on 

competition is, a fortiori, proof of market power.”). 
76 See McWane, Inc. v. FTC, 783 F.3d 814, 829–32 (11th Cir. 2015). 
77 See, e.g., Kaplow, supra note 58. 
78 See, e.g., Herbert Hovenkamp, Markets in Merger Analysis, 57 

ANTITRUST BULL. 887 (2012); Carl Shapiro, The 2010 Horizontal Merger 

Guidelines: From Hedgehog to Fox in Forty Years, 77 ANTITRUST L.J. 49, 68 

(2010) (noting the move away from traditional market definition in merger 

analysis); Gregory J. Werden, Market Delineation and the Justice 

Department’s Merger Guidelines, 1983 DUKE L.J. 514, 572–74 (1983). 



  

52 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

collusive group, and also their relative strengths as cartel 

contributors or enforcers.79 The alarming thing about the 

Court’s footnote is that it does not engage or even cite any of 

the extensive literature on the measurement of market power, 

providing not one single empirical rationale for the conclusion 

it draws. Indeed, the proposition that the Court asserted – 

namely, that a plaintiff in any vertical restraints case must 

define a relevant market no matter how power is sought to be 

established –was never briefed. Further, the Court stated this 

conclusion as a rule of law. Clearly, the permissible 

methodologies for proving power—a question of expert 

testimony—should be a question of fact.   

Market definition is even less reliable as an indicator of 

power in markets of significantly differentiated products.80 

Market definition is necessarily binary, putting products 

either inside or outside of the market. Placing differentiated 

products in the same market serves to exaggerate the degree 

of competition. Placing a differentiated product outside serves 

to understate the degree of competition. Amex concerned 

differentiated payment systems, as well as differentiation in 

costs and perks within the group of general purpose credit 

cards. Further, the fundamental concern was with high prices 

but not with collusion among issuers. In that case, measuring 

market power by reference to share of a defined market seems 

distinctly inferior. 

In any event, the Court then held that the proper relevant 

market for considering the restraint at issue was both sides of 

the platform.81 It also referred to “credit-card transactions as 

 

79 These are principal attributes of the Herfindahl-Hirschman Index, 

which is often used in merger enforcement and requires a market definition. 

See HERBERT HOVENKAMP, FEDERAL ANTITRUST POLICY: THE LAW OF 

COMPETITION AND ITS PRACTICE § 12.4 (5th ed. 2016); John Kwoka, Reviving 

Merger Control: A Comprehensive Plan for Reforming Policy and Practice 

7–8 (Oct. 9, 2018) (unpublished manuscript), https://www.antitrust 

institute.org/wp-content/uploads/2018/10/Kwoka-Reviving-Merger-Control-

October-2018.pdf [https://perma.cc/HX3Z-CKE2]. 
80 See 2B AREEDA ET AL., supra note 58, at ¶ 563; Herbert Hovenkamp, 

Response: Markets in IP and Antitrust, 100 GEO. L.J. 2133, 2146 (2012). 
81 See Ohio v. Am. Express Co., 138 S. Ct. 2274, 2286 (2018). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 53 

a whole” as the relevant market.82 That odd usage, of grouping 

both the buyer and seller into the same relevant market, 

would make any coherent economic analysis of the relevant 

market impossible, which is apparently why the opinion 

limited this definition to credit card transactions rather than 

transactions generally. Since every sale in a market involves 

a transaction, a broader conclusion would require holding that 

the spinach seeds and the gardener who purchases them are 

in the same market simply because they are simultaneously 

on the buy- and sell-side of the same transaction. 

Concluding that credit card transactions make up the 

relevant market should also have given the decision to the 

plaintiff. The record established unambiguously that the 

antisteering rule forced a specific buyer and seller to replace 

a lower price transaction that both preferred in favor of a 

higher cost transaction that injured both of them, as well as 

rival card issuers.83 In other words, it established exclusion, 

harm to the affected parties, and higher prices across the 

board. 

Importantly, the Court cabined this noneconomic market 

definition conclusion in other ways, adding this critical 

limitation: 

To be sure, it is not always necessary to consider both 

sides of a two-sided platform. A market should be 

treated as one sided when the impacts of indirect 

network effects and relative pricing in that market are 

minor. Newspapers that sell advertisements, for 

example, arguably operate a two-sided platform 

because the value of an advertisement increases as 

more people read the newspaper. But in the 

newspaper-advertisement market, the indirect 

networks effects operate in only one direction; 

newspaper readers are largely indifferent to the 

amount of advertising that a newspaper contains. 

Because of these weak indirect network effects, the 

market for newspaper advertising behaves much like 

a one-sided market and should be analyzed as such. 

 

82 Id. at 2287. 
83 See infra notes 180–91 and accompanying text. 



  

54 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

But two-sided transaction platforms, like the 

credit-card market, are different. These platforms 

facilitate a single, simultaneous transaction between 

participants. For credit cards, the network can sell its 

services only if a merchant and cardholder both 

simultaneously choose to use the network. Thus, 

whenever a credit-card network sells one transaction’s 

worth of card-acceptance services to a merchant it also 

must sell one transaction’s worth of card-payment 

services to a cardholder. It cannot sell transaction 

services to either cardholders or merchants 

individually. To optimize sales, the network must find 

the balance of pricing that encourages the greatest 

number of matches between cardholders and 

merchants.84 

Platforms that “facilitate a single, simultaneous 

transaction between participants”85 would include the credit 

card networks and very likely also ride-sharing platforms, 

such as Uber or Lyft, and perhaps eBay or Airbnb, which 

function mainly as brokers between buyers and sellers. The 

Court itself acknowledged that it would not include platforms 

such as newspapers, where there is no transaction-specific 

relationship between the two sides. A fortiori, it does not 

include television or radio stations that accept advertising 

revenue from one side. Nor would it include advertising-

supported computer search engines, music streaming, or other 

advertiser-supported computer applications in which 

advertising satisfies the company’s general revenue 

requirements, but there is no one-to-one transaction between 

the user and the advertiser. And it would exclude networks 

that sell things such as health insurance, where the buyer and 

seller do not engage in simultaneous transactions on a per-

service basis.86 

The Court concluded that assessing competitive effects of 

a two-sided transaction platform required the fact finder to 

 

84 Am. Express Co., 138 S. Ct. at 2286 (citations omitted). 
85 Id. 
86 See infra note 215 and accompanying text. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 55 

evaluate both sides.87 Here, the Court held, the plaintiffs had 

not carried their burden: 

Focusing on merchant fees alone misses the mark 

because the product that credit-card companies sell is 

transactions, not services to merchants, and the 

competitive effects of a restraint on transactions 

cannot be judged by looking at merchants alone. 

Evidence of a price increase on one side of a two-sided 

transaction platform cannot by itself demonstrate an 

anticompetitive exercise of market power. To 

demonstrate anticompetitive effects on the two-sided 

credit-card market as a whole, the plaintiffs must 

prove that Amex’s antisteering provisions increased 

the cost of credit-card transactions above a 

competitive level, reduced the number of credit-card 

transactions, or otherwise stifled competition in the 

credit-card market.88 

Further, the majority concluded, “[t]he plaintiffs did not 

offer any evidence that the price of credit-card transactions 

was higher than the price one would expect to find in a 

competitive market.”89 Rather, “Amex’s increased merchant 

fees reflect increases in the value of its services and the cost 

of its transactions, not an ability to charge above a competitive 

price.”90 

What the Court said could not possibly have been true of 

potential customers who would have agreed to a steering offer. 

Clearly, they did not value the perks by more than the 

merchant’s offered inducement, or else they would not have 

preferred to switch. That is, the relevant question concerned 

the marginal effect on consumers and merchants that 

resulted from the no steering rule.91 It is also worth noting 

that proof of any of these things does not rest on the premise 

that the merchant and customer sides of the platform were in 

the same relevant market. Further, the majority opinion 

 

87 Am. Express Co., 138 S. Ct. at 2287. 

88 Id. (citations omitted). 
89 Id. at 2288. 
90 Id. 
91 See infra notes 183–88 and accompanying text. 



  

56 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

ignored numerous explicit fact findings in the district court, 

all of which took effects on both sides into account but were 

based on an economically coherent market definition.92 

The Court also concluded that a dealer offering a discount 

to customers for purchasing with an alternative card was a 

form of free riding that “undermines the cardholder’s 

expectation of ‘welcome acceptance’—the promise of a 

frictionless transaction.”93 “A lack of welcome acceptance at 

one merchant makes a cardholder less likely to use Amex at 

all other merchants.”94 The Court described this “lack of 

welcome acceptance” as an “externality” that “endangers the 

viability of the entire Amex network,”95 and likened the 

situation to the use of resale price maintenance to prevent one 

seller from free riding on the efforts of a competing seller.96 

III. OBSERVATIONS: APPLYING THE RULE OF 
REASON TO PLATFORM EXCLUSION 

A. The Significance of Burden Shifting 

While all members of the Court nominally agreed with the 

rule of reason’s three-part burden-shifting analysis, the 

majority appeared to believe that the entire antitrust 

challenge depended on the plaintiff’s prima facie case. Some 

writers on two-sided platforms make the same mistake, 

 

92 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 174 

(E.D.N.Y. 2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018) (“American Express is correct that 

the court must account for the two-sided features of the credit card 

industry[.]”). The Second Circuit acknowledged this as well, but apparently 

concluded that benefits could be assessed only with a market definition that 

included both sides. United States v. Am. Express Co., 838 F.3d 179, 199–

200 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 

(2018); see also infra notes 149–51 and accompanying text. 
93 Am. Express Co., 138 S. Ct. at 2289 (citing Am. Express Co., 88 F. 

Supp. 3d at 156). 
94 Id. 

95 Id. 
96 Id. at 2289–90 (citing Leegin Creative Leather Prods., Inc. v. PSKS, 

Inc., 551 U.S. 877, 890–91. (2007)); see also infra notes 134–38 and 

accompanying text. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 57 

faulting analyses such as the district court’s for looking only 

at one side of the market.97 But that clearly is not what the 

district court did in this case, and it is not the proper way to 

think of the burden shifting rule-of-reason framework. The 

prima facie case considers whether the plaintiff has presented 

enough evidence of competitive harm to require the defendant 

to offer an explanation.98 Because the defendant is the creator 

of its restraint and presumably knows what its motives were, 

it is in a far better position to provide proof of its rationale and 

effects. If it had a procompetitive justification, such as cost 

reduction or product improvement, that must have been a 

motivating factor in its creation of the restraint.99 In any 

event, the district court made clear that its analysis 

considered both sides of the market even under the burden 

shifting approach.100 

B. Platform Market Delineation 

The majority never explained why assessing effects on both 

sides of a platform required jettisoning economically coherent 

conceptions of the relevant market as a group of substitute 

goods or services. That is, a relevant market is a “collusive 

group.”101 Putting production complements into the same 

market simply because making a deal requires both 

introduces economic nonsense into the law and economics of 

market power. Superior techniques exist for evaluating the 

pricing relationship among substitutes and complements, and 

 

97 See, e.g., Filistrucchi et al., supra note 67, at 301. 
98 See supra text accompanying notes 50–51.  
99 See 7 AREEDA & HOVENKAMP, supra note 10, at ¶ 1505a; Herbert 

Hovenkamp, The Rule of Reason, 70 FLA. L. REV. 81, 107 (2018). 
100 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 171 

(E.D.N.Y. 2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018). 

101 See, e.g., Kenneth D. Boyer, Industry Boundaries, in ECONOMIC 

ANALYSIS & ANTITRUST LAW 70, 73–74 (Terry Calvani & John Siegfried eds., 

2d ed. 1988); Gregory J. Werden, The History of Antitrust Market 

Delineation, 76 MARQ. L. REV. 123, 188–89 (1992). 



  

58 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

their effects on market power.102 These techniques do not 

require abandonment of sensible economics. In some cases, 

however, they may serve to strengthen or weaken the 

inference of power that can be drawn from computation of a 

market share. Unfortunately, parties will likely waste many 

hours of litigation resources disputing whether the “relevant 

market” in their particular case should include complements 

as well as substitutes. 

As Justice Breyer observed in his dissent, “[t]he phrase 

‘two-sided transaction platform’ is not one of antitrust 

art[.]”103 The important points, as he observed, are that such 

platforms “(1) offer different products or services, (2) to 

different groups of customers, (3) whom the ‘platform’ 

connects, (4) in simultaneous transactions.”104 But the 

majority made no attempt to explain why this set of facts 

required the Court to develop an economically incoherent 

conception of the relevant market. 

The dissent’s logic here is inescapable. Market definition 

and market share are only the starting points in the analysis 

of market power by indirect measures. At that point, a court 

must consider a number of things that give meaning to market 

shares.105 The Second Circuit lost sight of this when it held 

that the district court could not have accounted for the two-

sided features of the credit-card industry without a market 

definition that included both.106 When relying on proof from 

market share, a far better approach is to start out with a 

 

102 For an introduction, see generally Jonathan B. Baker, Unilateral 

Competitive Effects Theories in Merger Analysis, 11 ANTITRUST 21 (1997). 

See also Jonathan B. Baker & Timothy F. Bresnahan, Empirical Methods 

of Identifying and Measuring Market Power, 61 ANTITRUST L.J. 3 (1992). For 

an updated and somewhat more critical overview, see Daniel A. Crane, 

Market Power Without Market Definition, 90 NOTRE DAME L. REV. 31 (2014). 
103 Ohio v. Am. Express Co., 138 S. Ct. 2274, 2298 (2018) (Breyer, J., 

dissenting). 
104 Id. 
105 See 2B AREEDA ET AL., supra note 58, at ¶¶ 532–33. 

106 See United States v. Am. Express Co., 838 F.3d 179, 196, 200 (2d 

Cir. 2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018). 

The Supreme Court did not discuss the fact finding. See generally Am. 

Express Co., 138 S. Ct. 2274. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 59 

properly defined group of substitutes and then consider other 

factors that might strengthen or weaken any inference of 

power drawn from market share. In this case, the alleged 

relevant market and the one the district court focused on was 

economically sensible—namely a “network services market” 

for general purpose charge cards, in which the purchasers 

were merchants.107 To this, the Second Circuit required the 

addition of cardholders.108 Cardholder transactions were not 

competitors with network services, but rather were 

complements in production.109 

It is difficult to see any added value coming from a 

linguistic requirement that both sides of the platform be 

placed within the same market.110 It certainly cannot be to 

assess collusion possibilities because the two sides do not 

compete with each other. The presence of a second side may 

affect the ability of the first side to exercise power, but no part 

of that determination requires a conclusion that the second 

side is in the same market. The availability or price of 

complements can certainly limit a firm’s ability to increase its 

price. For example, the limited supply and high price of 

gasoline might limit a firm’s ability to charge a higher price 

for its automobiles. This is another way of saying that 

substitutes and complements pull in opposite directions when 

one is estimating a firm’s market power: high prices for 

substitutes tend to increase it, while high prices for 

complements tend to decrease it.111 Economists have been 

 

107 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 171 

(E.D.N.Y. 2015), rev’d., 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018). 
108 See Am. Express Co., 838 F.3d at 197. 
109 See infra at notes 132–133 and accompanying text. 
110 See Carlton, supra note 15. 
111 See supra notes 68–69 and accompanying text; see also Werden, 

supra note 68, at 398. Conceptually, a firm’s market power is determined 

by its own price elasticity of demand, which is the weighted sum of the cross 

elasticities of all other products with respect to the first product’s price. See 

Kaplow, supra note 58, at 485–86. In practice, direct measurement of a 

firm’s own price elasticity is easier than, and preferable to, attempts to 

measure cross-elasticities with potentially competing products. Id. at 490. 



  

60 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

making such calculations for decades,112 and without doing 

anything as irrational as grouping substitutes and 

complements into a single market. 

What the Supreme Court majority was apparently trying 

to do is force the plaintiff to consider burdens and benefits on 

both sides of the platform as part of its prima facie case. The 

district court seems to have done that quite adequately,113 but 

it did not add the verbal flourish that the two sides were in a 

single relevant market. How its analysis would have been any 

different if it had done so is not clear. 

The Second Circuit expressed concern that limiting the 

market to substitutes would ignore “feedback effects”—

namely that a “reduction in cardholders’ demand for cards (or 

card transactions) . . . would accompany any degree of 

merchant attrition.”114 Of course, that would be true of any 

firm that sold complementary products. For example, a grocer 

that sells both milk and Cheerios would have to consider that 

a price increase in milk might reduce the demand for 

Cheerios. This would not warrant creating a single market for 

milk and Cheerios, although it might require the fact finder 

to consider what impact reduced demand for Cheerios might 

have on the profitability of the higher-priced milk. That is 

fundamentally a demand elasticity problem, not a problem in 

market delineation, and further illustrates why the Court’s 

insistence on a traditional market definition in a vertical case 

makes so little sense. In an extreme case, an increase in the 

price of milk might impact Cheerios so severely as to make it 

an unsustainable product. In all events, the “feedback” 

equilibrium depends on all market conditions, including price, 

that operate on both sides of the platform. A judicial order 

permitting steering would simply move this equilibrium to a 

different place. In general, the more competitive the two sides 

of the market are, the less profitable this equilibrium would 

 

112 See Werden, supra note 68, at 398 (citing JOE S. BAIN, PRICE THEORY 

25–26, 50–53 (1952); FRITZ MACHLUP, THE ECONOMICS OF SELLERS 

COMPETITION 213–14 (1952)). 
113 See supra notes 99–100 and accompanying text. 
114 United States v. Am. Express Co., 838 F.3d 179, 200 (2d Cir. 2016), 

aff’d sub nom. United States v. Am. Express Co., 138 S. Ct. 2274 (2018). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 61 

be. In any event, a lost transaction to Amex would be a gained 

transaction to a lower priced network, affecting the latter’s 

feedback effect as well. 

Indeed, the district court also used a version of this 

feedback argument to make the opposite point, and its 

analysis was more persuasive. It concluded that the impact of 

the antisteering rule was to create “derived demand” for the 

Amex card by deterring customers from switching away, 

which in turn supported increased merchant fees.115 This was 

more persuasive because it considered marginal—rather than 

average—behavior. The reason this rationale is more 

persuasive is that each and every customer who would have 

switched away from Amex as a result of steering in fact 

experienced a loss in value, as did each and every merchant 

forced to make the transaction using the Amex card.116 There 

were no gains on either side, but rather losses on both. Profits 

accrued only to the network operator, not to merchants or card 

users. The only remaining question was whether the 

antisteering rule was necessary to provide the minimum 

volume necessary for the network operator’s viability. 

Assuming that is a viable defense, one does not need a special 

theory of platforms in order to answer that question.  

Similar situations can arise in exclusive dealing cases, 

where viability of production facilities is sometimes raised as 

a defense. Suppose, for example, that Uber should impose 

exclusive dealing117 on its drivers, preventing them from 

driving for any competing company. To the extent Uber had 

the power to do so, this would prevent the drivers from selling 

their services to rivals such as Lyft or traditional taxicabs. 

One standard defense, as presented in Amex, would be that 

exclusive dealing is necessary to increase volume in order to 

make the Uber platform viable. That is a testable fact. 

 

115 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 188 

(E.D.N.Y. 2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018). 

116 See infra notes 180–91 and accompanying text. 
117 More specifically, this would be an output contract, in which 

exclusive dealing is imposed on sellers rather than buyers. See 11 HERBERT 

HOVENKAMP, ANTITRUST LAW ¶ 1803 (4th ed. 2018). 



  

62 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

However, it could also be raised as a defense to any exclusive 

dealing claim, whether or not it involved a transactional 

network or any network at all. For example, in McWane, a 

more conventional exclusive dealing case, the defendant 

argued that exclusive dealing was necessary for it to maintain 

minimum viable scale in one of its plants.118 The Eleventh 

Circuit concluded that McWane did not show it needed 

exclusive dealing in order to get its volume up.119 Given its 

very high profit margins, it should simply have cut its 

prices.120 

By the same token, the evidence in Amex pertained to 

whether the antisteering rule was needed to protect Amex’s 

scale of operations necessary to keep its platform viable. Here, 

the record was clear. Amex offered high-cost cards and 

continuously raised its merchant prices under the 

antisteering rule.121 None of that depended on the fact that 

Amex was operating on a transaction platform. Even 

assuming that Amex had to increase card usage in order to 

make its platform viable—a fact that was never established—

it might have achieved viability by cutting its price. Further, 

the opinion says nothing about what Amex’s viability 

requirements were. As the district court observed and Justice 

Breyer noted, at trial Amex “presented no expert testimony, 

financial analysis, or other direct evidence establishing that 

without its [nondiscrimination provisions] it will, in fact, be 

unable to adapt its business to a more competitive market.”122 

Finally, the viability question is not one of networks, but 

rather of plain, old economies of scale or perhaps of scope. 

 

118 See McWane, Inc. v. FTC, 783 F.3d 814, 841 (11th Cir. 2015). 
119 See id. 
120 See id. 
121 See Ohio v. Am. Express Co., 138 S. Ct. 2274, 2293 (2018) (Breyer, 

J., dissenting) (noting that “American Express raised the prices it charged 

merchants on 20 separate occasions.”). 

122 Am. Express Co., 138 S. Ct. at 2304 (Breyer, J., dissenting) (quoting 

United States v. Am. Express Co., 88 F. Supp. 3d 143, 231 (E.D.N.Y. 2015), 

rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express Co., 

138 S. Ct. 2274 (2018)). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 63 

Properly conducted antitrust law already requires analysis 

of two sides in cases of vertical interbrand restraints, 

including the tying of complementary products, although it 

does not proceed by anything as economically incoherent as 

putting them into the same relevant market.123 The law of 

tying and exclusive dealing both assess competitive effects by 

examining power in a primary market and foreclosure in a 

secondary market. Typically, as in Amex, the two products are 

complements. The principal exception is tying law’s 

inappropriate but nevertheless tenacious per se rule, which 

permits an anticompetitive tie to be inferred without 

reference to foreclosure.124 

To illustrate, in Jefferson Parish v. Hyde, the Supreme 

Court dismissed a challenge to a hospital-anesthesiology 

services tie under the rule of reason after finding that the 

defendant hospital’s admission of thirty percent of the 

surgical patients in the relevant market was inadequate.125 

Importantly, the Court did not define a single market for 

anesthesiologists and surgical patients. Hospitals, it should 

be noted, are two-sided platforms servicing providers on one 

side and patients on the other. 

Anticompetitive harm in vertical interbrand foreclosure 

cases such as Jefferson Parish requires a showing of some 

degree of presence in the primary market and some minimum 

amount of foreclosure, exclusion, or perhaps a price increase 

in the complementary market.126 For example, in the well-

known Tampa Electric Co. v. Nashville Coal Co. exclusive 

dealing case, the utility controlled a dominant position in the 

market for electric power in its service area and consumed 

eighteen percent of the coal consumed in Florida and 

 

123 Cf. Am. Express Co., 88 F. Supp. 3d at 171 (including effects on the 

other side of the market expressly). 
124 See 9 PHILLIP E. AREEDA & HERBERT HOVENKAMP, ANTITRUST LAW ¶ 

1720 (4th ed. 2018). 
125 Jefferson Par. Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2, 7, 26–27 (1984) 

(noting that defendant lacked power because seventy percent of surgical 

patients in the area went to other hospitals). 
126 On market definition for purposes of assessing vertical foreclosure, 

see 2B AREEDA ET AL., supra note 58, at ¶ 570. 



  

64 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

Georgia.127 Under existing exclusive dealing standards, that 

percentage would have been sufficient for illegality.128 The 

Supreme Court dismissed the complaint, however, because 

the coal that it purchased represented “less than 1% of the 

total marketed production” of coal that was available for 

Tampa Electric to purchase.129 The Court did not attempt to 

define a single market for coal and electric power. Indeed, 

doing so would have undermined the competitive analysis. 

Looking at the Tampa utility’s side of the market alone 

seriously exaggerated the competitive harm. In Tampa 

Electric, the Court got to the correct result by limiting both 

the upstream and downstream markets to substitutes and 

then assessing the competitive harm. As Justice Breyer 

queried in Amex, “What is it about the economic relationship 

between merchant-related and shopper-related services that 

would justify the majority’s novel approach to market 

definition?”130 Economically speaking, the analysis of market 

pairs for purposes of evaluating vertical interbrand restraints 

is no different from the analysis of platforms.131 

C. Complements in Use or Production 

On the question of complements, the Amex majority 

concluded that the two sides of the transaction were not 

complements because they were not purchased by the same 

buyers.132 The dissent countered that they were, and that 

putting complements and substitutes into the same market 

 

127 See Tampa Electric Co. v. Nashville Coal Co., 365 U.S. 320, 331 

(1961). 
128 See Standard Oil Co. of Cal. v. United States, 337 U.S. 293, 295, 305 

(1949) (condemning exclusive dealing contracts covering 6.7% of total 

retailed gasoline). 
129 Tampa Electric, 365 U.S. at 331. 
130 Ohio v. Am. Express Co., 138 S. Ct. 2274, 2299 (2018) (Breyer, J., 

dissenting). 

131 On this point, see generally Carlton & Winter, supra note 17. 
132 Am. Express Co., 138 S. Ct. at 2286 n.8 (asserting that the two were 

not complements “in which both products are bought by the same buyers” 

(quoting Filistrucchi et al., supra note 67, at 297)). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 65 

was “economic nonsense.”133 What neither side stated clearly 

was that the two sides were clearly complements, but they 

were complements in production rather than use. The 

majority was thinking of complements in use, such as toast 

and jam, or gardeners and spinach seed. Complements in 

production are goods or services that are produced together, 

such as beef and cowhide, oil and natural gas, lumber and 

sawdust, or voice services and messaging services. The 

majority’s acknowledgement that each cardholder transaction 

is necessarily offset by an equal merchant transaction 

acknowledged as much.134 Complements in production behave 

in ways similar to complements in use. For example, strong 

demand for oil leads to higher oil prices. Necessarily, however, 

increasing oil production will increase gas production because 

gas is a natural byproduct. If the demand for gas remains 

constant, its price will fall just as oil prices rise. 

D. Free Rider Concerns 

The majority was confused about the existence and nature 

of free riding. It spoke of the other card companies as free 

riding on Amex’s “investments in rewards[,]” likening the 

antisteering provision to the use of resale price maintenance 

to protect against dealer free riding.135 But as the dissent 

properly noted, the Amex rewards attach to specific 

transactions, not to mere possession of the card.136 If a 

cardholder earned its perks simply by owning an Amex card, 

then, of course, free riding would be possible. A customer 

might acquire the card in order to obtain the perks but then 

make its actual transactions with a lower priced card. The 

majority paid little attention to the record, but the district 

 

133 Id. at 2295–96 (Breyer, J., dissenting) (quoting 2B AREEDA ET AL., 

supra note 58, at ¶ 565a, at 431). 

134 See id. at 2286. 
135 See id. at 2289–90 (citing Leegin Creative Leather Prods., Inc. v. 

PSKS, Inc., 551 U.S. 877, 890–91 (2007)). 
136 Id. at 2304 (Breyer, J., dissenting). 



  

66 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

court was clear on the point137 and it seems incontrovertible. 

The policy is clear on Amex’s own website.138 

The all-important ingredient in the free-rider explanation 

of resale price maintenance or other dealer restraints is that 

the manufacturer is unable to price out services distinctly 

from the product itself. For example, it cannot charge 

customers separately for the well-trained sales staff but must 

include it in the price of the basic product. That makes it 

possible for the customer to segregate the two—obtaining 

product education from the full-service dealer, but then 

purchasing the product from the discounter.139 However, if 

the issuer attaches the perks strictly to payment for the 

product, then “the ride is not free,” in Judge Frank 

Easterbrook’s words.140 

The majority’s free rider argument was attached to a 

perspective on consumer behavior that can only be described 

as economically bizarre, and certainly contrary to our usual 

assumption that consumers are rational maximizers of their 

own utility. The majority spoke of the antisteering provision 

as promoting “welcome acceptance” of its card.141 “Welcome 

acceptance” in this case apparently meant that the buyer 

 

137 United States v. Am. Express Co, 88 F. Supp. 3d 143, 237 (E.D.N.Y. 

2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express 

Co., 138 S. Ct. 2274 (2018). (“Plainly . . . investments tied to card use (such 

as Membership Rewards points, purchase protection, and the like) are not 

subject to free-riding, since the network does not incur any cost if the 

cardholder is successfully steered away from using his or her American 

Express card.”). 
138 See, e.g., Extended Warranty Description of Coverage, AM. EXPRESS 

https://www.americanexpress.com/content/dam/amex/us/credit-

cards/features-benefits/policies/pdf/EW%20Benefit%20Guide_Tier% 

201%20Rev%2007-18.pdf [https://perma.cc/8TEZ-HNNL] (describing 

extended warranty protection offered with Amex Platinum Card, provided 

that the product in question was purchased with the card). 
139 See Lester G. Telser, Why Should Manufacturers Want Fair Trade?, 

3 J.L. & ECON. 86, 91–92 (1960); see also 8 PHILLIP E. AREEDA & HERBERT 

HOVENKAMP, ANTITRUST LAW ¶ 1613 (4th ed. 2017). 

140 Chi. Prof’l Sports Ltd. P’ship v. NBA, 961 F.2d 667, 675 (7th Cir. 

1992) (“When payment is possible, free-riding is not a problem because the 

‘ride’ is not free.”). 
141 Am. Express Co., 138 S. Ct. at 2289. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 67 

should be prevented from being offered or even told about the 

availability of a cheaper alternative. Amex made this 

argument in its brief, offered as a Hail Mary pass, which the 

majority caught.142 The Second Circuit had decided that 

“welcome acceptance” was a viable defense because loss of a 

sale via steering could have a negative impact on both sides of 

the market.143 Factually, of course, it could be true that loss 

of “welcome acceptance” on one product could impair a firm’s 

earnings on a complementary product. For example, 

condemnation of a cartel’s boycott of a price cutter in its 

primary market could also have a negative impact on the sales 

of a complement, or impairing a grocer’s milk sales might also 

harm that grocer’s sale of Cheerios. By contrast, the district 

court took a much more economically rational view of the 

situation, namely that “[a]llowing merchants to actively 

participate in their customers’ point-of-sale decisions would 

remove the artificial barrier that now segregates merchant 

demand from the price of network services[.]”144 

Putting the most sensible gloss on the argument, Amex 

had invested in a business model that depended on high 

merchant acceptance fees. When a merchant offered a 

customer a lower price to use a different card, that offer 

undermined Amex’s model. Factually, of course, that is true. 

Anytime a merchant tells a buyer that a better deal is 

available than the buyer’s initially chosen one it serves to 

diminish “welcome acceptance” of that initial offer. Indeed, 

competition of any sort does that. One wonders if that 

argument, which now has the majority’s imprimatur, will also 

 

142 See Brief for Respondents Am. Express Co. & Am. Express Travel 

Related Servs. Co., Inc. at 9, Am. Express Co., 138 S. Ct. 2274 (No. 16-1454). 
143 See United States v. Am. Express Co., 838 F.3d 179, 191 (2d Cir. 

2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 (“Although 

merchants across various industries regularly try to ‘steer’ their customers 

toward certain purchasing decisions via strategic product placement, 

discounts, and other deals, steering within the credit-card industry can be 

harmful insofar as it interferes with a network’s ability to balance its two-

sided net price.”). 
144 United States v. Am. Express Co., 88 F. Supp. 3d 143, 220–21 

(E.D.N.Y. 2015), rev’d, 838 F.3d 179 (2d Cir. 2016, aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018). 



  

68 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

appear as a defense to price fixing. After all, permitting rivals 

to offer a lower price than the cartel’s offer will undermine 

“welcome acceptance” of the cartel’s product.145 

Further, the majority failed to observe the one instance of 

free riding that the facts indicated. The antisteering rule 

made the Amex cardholder indifferent as to which card he or 

she used, because the increased cost was borne entirely by the 

merchant. The merchant for its part had to absorb these 

higher costs, which it did via higher product prices, as the 

district court also found.146 Effectively, all purchasers, 

whether they used an Amex card or any card at all, were 

forced to subsidize Amex’s higher merchant acceptance fees. 

That was a true case of free riding. 

The Second Circuit did not disagree with this fact finding, 

but it held that higher product prices could be justified by the 

greater value of Amex’s perks to its own cardholders, 

particularly for merchants where Amex cardholders insisted 

on paying with the Amex card and nothing else.147 What the 

Second Circuit apparently did not see is that this finding 

conceded Amex’s market power: that is, it had the power to 

compel higher merchant product prices across the board in 

order to subsidize Amex’s perks to its own cardholders. A firm 

that lacked power would not be able to compel higher prices 

market wide, harming everyone else for the benefit of its own 

customers. The Supreme Court did not discuss the Second 

Circuit’s treatment of the issue. 

E. Use of the Record 

One of the rule of reason’s most essential features is 

development and analysis of a record. The rule of reason 

requires detailed factual analysis because the restraint is not 

of a type that can be disposed of categorically, as under per se 

rules of illegality or legality.148 The Brooke Group, Ltd. v. 

Brown & Williamson Tobacco Corp. predatory pricing case 

 

145 See Hovenkamp, supra note 15, at 31. 
146 See infra note 154 and accompanying text. 
147 See Am. Express Co., 838 F.3d at 202. 
148 See 7 AREEDA & HOVENKAMP, supra note 10, at ¶¶ 1507–08. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 69 

provides an instructive example, where the Supreme Court 

carefully reviewed the record, citing it nearly twenty times 

and ultimately agreeing with the district court that the 

plaintiff’s legal case had failed.149 

By contrast, the Supreme Court’s Amex opinion contains 

only a single mention of the record, for a proposition unrelated 

to the challenged restraint, and virtually no analysis.150 The 

Second Circuit had discussed the record multiple times, 

mainly concerning the district court’s conclusion that the 

relevant market must be limited to reasonably 

interchangeable goods. The Second Circuit also concluded 

that harms to merchants needed to be offset by benefits to 

cardholders, and that only a market definition that grouped 

the two together could do that.151 It did not acknowledge that 

for cardholders affected by the antisteering rule, the benefits 

were necessarily less than the higher fees.152 The Supreme 

Court did not discuss these findings either. On the district 

court’s finding of higher retail prices, the Second Circuit held 

that this conclusion was error because it “fail[ed] to take into 

account offsetting benefits to cardholders in the form of 

rewards and other services.”153 That was tantamount to a 

conclusion that Amex was entitled to raise merchant prices 

across the board for the benefit of its own cardholders.154 

While the majority did not disagree with or repudiate the 

district court’s detailed fact findings, it made almost no use of 

them—a point that the dissenters noted.155 Indeed, the only 

way that the Court could reach its conclusions was by ignoring 

the record. While the majority opinion did cite significant 

 

149 See generally Brooke Grp., Ltd. v. Brown & Williamson Tobacco 

Corp., 509 U.S. 209 (1993). 
150 See Ohio v. Am. Express Co., 138 S. Ct. 2274, 2282 (2018) (citing 

record for proposition that Amex made some banking and payment services 

available to low-income individuals). 
151 See Am. Express Co., 838 F.3d at 205. 
152 See infra notes 188–92 and accompanying text. 

153 See Am. Express Co., 838 F.3d at 204 n.52. 
154 On the free rider issue, see supra notes 94–96, 134–45 and 

accompanying text. 
155 See Am. Express Co., 138 S. Ct. at 2293 (Breyer, J., dissenting). 



  

70 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

academic economics literature on the issue of platforms, the 

Court made very little use of it other than setting up some 

basic definitions. In, fact, the Court never bothered to analyze 

the particular transactions at issue and how the antisteering 

rule affected consumer behavior and welfare. 

By contrast, the dissent summarized the district court’s 

conclusions: Under the antisteering rule, Amex was able to 

increase merchant acceptance fees approximately twenty 

times during a five-year period, specifically because it did not 

have to worry about merchants shifting their customers to a 

less costly card.156 It found that in the absence of the no-

steering rule, merchant acceptance fees “would likely have 

been lower.”157 It also found that the antisteering rule had 

successfully deterred an attempt by Discover, a competitor, to 

switch merchants by offering them lower acceptance fees.158 

The district court also found that for many merchants the 

costs of credit card acceptance were “among many merchants’ 

highest,” giving them “a strong economic incentive to take 

steps to reduce” them.159 The court concluded that the 

“Plaintiffs additionally are able to show harm to those same 

merchants’ customers on the other side of the GPCC platform, 

as inflated merchant discount rates are passed on to all 

customers—AmEx cardholders and non-cardholders alike—in 

the form of higher retail prices.”160 

The key target of the Sherman Act’s restraint of trade 

standard is reduced output and, consequently, higher 

prices.161 The particular fact finding by the district court hit 

 

156 Id. 

157 Id. 
158 Id. at 2293–94. On the use of vertical most-favored-nation clauses 

or similar practices to deter entry, see Carlton, supra note 15. 
159 United States v. Am. Express Co., 88 F. Supp. 3d 143, 221 (E.D.N.Y. 

2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express 

Co., 138 S. Ct. 2274 (2018). 
160 Id. at 208. 

161 See Herbert Hovenkamp, Is Antitrust’s Consumer Welfare Principle 

Imperiled?, 44 J. CORP. L. (forthcoming 2019) (manuscript at 3), 

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3197329 

[https://perma.cc/EG3A-6N3C]. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 71 

that target’s bullseye. On the question of output, the record 

showed everything that the Court needed to know. The 

antisteering rule shifted deals to a higher cost transaction 

that resulted in lost value to both the affected cardholder and 

merchant. Further, it resulted in higher product prices across 

the board, even to those who purchased without the use of the 

Amex card. 

The district court had also concluded that “the customer 

neither sees nor pays the additional cost when networks 

increase the price of network services to merchants (other 

than in the form of higher retail prices, which are paid by all 

consumers); thus, the customer cannot be expected to initiate 

substitution in the first instance.”162 As Justice Breyer 

described this evidence, it showed that “[c]onsumers 

throughout the economy paid higher retail prices as a 

result[.]”163 The district court also observed that Amex 

“sharply dispute[d]” these fact findings but resolved them in 

favor of the government.164 The Supreme Court did not upset 

that conclusion. 

F. Inattentiveness to Economic Analysis 

Another point missing from the majority opinion was 

analysis of how the antisteering rule affected market 

participants. The Court apparently reasoned that because 

transactions on the two sides of the platform balanced out, 

this meant that “costs” on one side of a platform are offset by 

“benefits” on the other side.165 This assumption, which is 

 

162 Am. Express Co. 88 F. Supp. 3d at 177. The district court concluded 

that “[i]n the longer term, the court expects that merchants will pass along 

some amount of the savings associated with declining swipe fees to their 

customers in the form of lower retail prices.” Id. at 221. 
163 See Am. Express Co., 138 S. Ct. at 2294 (Breyer, J., dissenting). 
164 See Am. Express Co., 88 F. Supp. 3d at 222. 
165 See Am. Express Co., 138 S. Ct. at 2288. The Second Circuit made 

the same assumption. See United States v. Am. Express Co., 838 F.3d 179, 

203 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 

(2018). Like the Supreme Court, it did not look at individual transactions 

but rather made general observations about increased aggregate value to 

one or the other side of the platform. 



  

72 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

frequently made in bird’s-eye views of networks, drove much 

of the Court’s analysis. Thus in the Court’s mind it became 

essential to compare burdens and benefits on the two sides in 

making out a prima facie case.166 

Some cases involving two-sided platforms do involve 

offsetting costs and benefits on the two sides. This is 

particularly likely to be true when the only issue is source and 

amount of revenue. A good example is Wallace v. IBM.167 IBM 

used an open source program as the operating system for one 

of its computer lines.168 As part of the open source licensing 

obligations, the program had to be given away for free.169 This 

is a common practice. For example, a large variety of smart 

phones are sold with the Google Android operating system 

included at a price of zero. IBM, of course, is not in the 

business of giving away software, but of selling computers, to 

which the software was distributed as a complement. Daniel 

Wallace had developed a competing, proprietary operating 

system that he wanted to sell to users of these computers,170 

but competing with a price of zero is difficult. Wallace accused 

IBM of predatory pricing.171 Judge Easterbrook dismissed the 

complaint, holding mainly that the facts did not fall within 

predatory pricing rules requiring a likelihood of 

recoupment.172 In any event, one cannot evaluate a claim such 

as predatory pricing without looking at all sources of revenue. 

The price of the computer-plus-operating system was never 

 

166 See, e.g., Am. Express Co., 138 S. Ct. at 2288 (“Amex’s higher 

merchant fees are based on a careful study of how much additional value its 

cardholders offer merchants. On the other side of the market, Amex uses its 

higher merchant fees to offer its cardholders a more robust rewards 

program, which is necessary to maintain cardholder loyalty and encourage 

the level of spending that makes Amex valuable to merchants. That Amex 

allocates prices between merchants and cardholders differently from Visa 

and MasterCard is simply not evidence that it wields market power to 

achieve anticompetitive ends.”) (citations omitted). 
167 Wallace v. IBM Corp., 467 F.3d 1104 (7th Cir. 2006). 
168 Id. at 1107. 

169 Id. at 1105. 
170 Id. at 1106. 
171 See id. 
172 See id. at 1106, 1108. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 73 

alleged to be below cost, so the court rejected the predatory 

pricing claim. 

Another good example arose in Meyer v. Kalanick, which 

alleged that Uber was engaged in price fixing by setting a 

common price among its drivers.173 Price fixing is 

anticompetitive because it is an output reducing practice that 

results in higher consumer prices. But in order to determine 

output effects on a two-sided platform, one must see how the 

effects on one side play out on the other side, just as in 

Wallace. For that purpose, it is not necessary to do anything 

as irrational as define a single market for riders and drivers. 

Nevertheless, one must examine the economic rationales for 

Uber’s price in order to determine its effects on market output. 

As the operator of a platform, Uber needs to seek out the 

equilibrium spot that will bring in the optimal number of 

drivers and riders. Setting fares too high discourages riders, 

while setting them too low discourages drivers.174 

This process does not necessarily mean that Uber’s conduct 

is lawful, although it does indicate that it should not be 

addressed under the per se rule for ordinary price fixing. 

Indeed, the very fact that this price fixing occurs in the context 

of an elaborate joint venture should be a sufficient trigger for 

the rule of reason.175 Under this analysis, Uber’s conduct may 

still be unlawful. For example, perhaps Uber is controlled by 

a local cartel of drivers that has market power in some area. 

They set prices that are too high to maximize overall profits, 

but instead try to maximize the profits of the colluding 

drivers.176 The complaint challenges Uber’s “peak load” 

 

173 See Meyer v. Kalanick, 174 F. Supp. 3d 817, 819–20 (S.D.N.Y. 2016) 

(discussing antitrust issues in the context of a class action). 
174 See supra notes 20–21 and accompanying text. 
175 Cf. NCAA v. Bd. of Regents of Univ. of Okla., 468 U.S. 85, 101–03 

(1984) (stating that the rule of reason must be applied if a restraint is 

essential to making the product available at all); Broad. Music, Inc. v. CBS, 

Inc., 441 U.S. 1, 22–25 (1979) (rejecting claim of per se unlawful price fixing 

and applying the rule of reason to conduct on a two-sided platform that the 

Supreme Court described as price fixing). 
176 See First Amended Complaint at ¶ 4, Meyer, 174 F. Supp. 3d 817 

(No. 1:15-CV-9796), 2016 WL 950376 (alleging that Uber’s CEO conspired 

with drivers to increase prices at expense of riders); but see id. ¶ 47 (noting 



  

74 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

pricing model that increases prices during busy periods, when 

demand is large in relation to driver supply.177 But that is 

precisely how one would expect an efficient two-sided platform 

to work in order to maintain an equilibrium between the two 

sides of the market. When rider side demand increases, a fare 

increase is necessary to balance that demand with the supply 

of drivers. Proof of antitrust harm would require additional 

evidence, such as proof of driver control and of an entry 

restriction that prevents additional drivers in such an area. 

In sum, anticompetitive harm is an unlikely scenario, but one 

that cannot be dismissed out of hand. 

Whether or not a firm operates a platform, evaluating its 

revenue requires examination of all relevant sources. For a 

non-platform illustration, Coca-Cola provides free coke 

machines to employers who agree to stock the machines 

exclusively with Coca-Cola products.178 But that does not 

necessarily mean that Coca-Cola is engaging in predatory 

pricing of dispensing machines. In order to determine the 

profitability of this enterprise one must look at the revenues 

obtained from both the dispensing machine and the products. 

Notably, one does not need to define a single relevant market 

for Coca-Cola and dispensing machines in order to answer 

that question. 

In Amex, however, the charge was not predatory pricing or 

price fixing, but rather an exclusionary practice more akin to 

exclusive dealing or a most-favored-nation (MFN) clause.179 

 

that “[f]ares are calculated based on an Uber-generated algorithm” which 

increases fares as demand increases in relation to supply (surge pricing)). 

Similar peak load pricing is common in intermediate sales of electric power, 

which also occurs on a two-sided platform between generators and users. 

See, e.g., RAFAL WERON, MODELING AND FORECASTING ELECTRICITY LOADS 

AND PRICES: A STATISTICAL APPROACH 4–5 (2006). 
177 See First Amended Complaint, supra note 176, at ¶¶ 2, 4, 47. 
178 Coke Vending Machine, VENDINGSOLUTIONS, 

http://www.vendingsolutions.com/coke-vending-machines 

[https://perma.cc/MVR6-4EJP]; see also Erik Hovenkamp & Herbert 

Hovenkamp, Tying Arrangements and Antitrust Harm, 52 ARIZ. L. REV. 925, 

942, 943 n.78 (2010). 
179 On antisteering rules as most-favored-nation clauses, see Carlton, 

supra note 15; see also Carlton & Winter, supra note 17, at 215–16. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 75 

Not only did the Court pay scant attention to the record, but 

it also never analyzed the transaction to determine how the 

harms and benefits balance out on the two sides. As Ronald 

Coase taught, when one wants to understand a practice, there 

is no good substitute for examining the incentives to make 

each individual transaction, small as they might be, and 

considering how they affect the whole.180 However, neither 

the Second Circuit nor the Supreme Court did this. 

G.  Marginal Harms and Benefits 

Competition always exists at the margin. It requires firms 

to make incremental changes to their business methods, 

continuously tracking and adjusting to reflect their successes 

or failures. These changes may or may not be anticompetitive, 

quite apart from the overall structure of the firm or 

organization that implements them. To the extent that a 

chosen rule or decision might harm competition, the rule of 

reason is designed to manage this process. For example, in 

Amex the government was not trying to tear down Amex’s 

entire business model, but only to enjoin its antisteering rule. 

Such a case requires an assessment of the marginal costs and 

benefits of the antisteering rule itself. One pervasive error in 

the Supreme Court majority’s analysis was that it failed to 

distinguish the challenged rule’s marginal effects from the 

overall impact of the defendant’s business model. It simply 

assumed that harms on one side were offset by benefits on the 

other, or else it spoke of evidence about the defendant’s overall 

business model. The record was clear, however, that at the 

margin each merchant affected by the steering rule was worse 

off, and each cardholder was worse off as well. Competitive 

harm was clear. 

One cannot evaluate the competitive effects of a particular 

restraint by considering whether the overall costs of a 

defendant’s business practices exceed the benefits. For 

example, in NCAA v. Board of Regents of University of 

Oklahoma, the challenge was not to the existence or 

 

180 See R. H. Coase, The Nature of the Firm, 4 ECONOMICA 386, 397–98 

(1937). 



  

76 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

legitimacy of the NCAA as an association; rather it was to the 

effect of a limitation on the televising of games.181 The issue 

was whether the incremental harm to competition caused by 

the challenged restriction on the number of televised games 

was justified by offsetting incremental benefits.182 By the 

same token, the question in Amex was not whether Amex’s 

general business model of charging higher merchant fees for 

large perks produced overall benefits. Rather it was whether 

the incremental harm caused by the antisteering rule 

produced incremental competitive harms greater than 

incremental benefits. That would require an assessment of the 

competitive effects of that particular rule.  

The Second Circuit also confused total and marginal effects 

by stating that “because the NDPs183 affect competition for 

cardholders as well as merchants, the Plaintiffs initial burden 

was to show that the NDPs made all Amex consumers on both 

sides of the platform—i.e., both merchants and cardholders—

worse off overall.”184 

But that was not the question. As with any restraint, many 

customers were not affected at all. For example, the restraint 

on game televising in NCAA did not affect those who did not 

watch televised games at all.185 Similarly, rules imposing 

resale price maintenance affect only discounters who would 

otherwise charge a lower price.186 Standard setting and other 

boycott rules affect only producers at risk of violating a 

 

181 NCAA v. Bd. of Regents of Univ. of Okla., 468 U.S. 85, 88 (1984); see 

also 7 AREEDA & HOVENKAMP, supra note 10, at ¶¶ 1502–03, 1511. 
182 See id. at 94. 

183 The Second Circuit used the term “nondiscrimination provisions,” 

or NDPs to describe Amex’s policies “barring merchants from (1) offering 

customers any discounts or nonmonetary incentives to use credit cards less 

costly for merchants to accept, (2) expressing preferences for any card, or (3) 

disclosing information about the costs of different cards to merchants who 

accept them. United States v. Am. Express Co., 838 F.3d 179, 184 (2d Cir. 

2016), aff’d sub nom. Ohio v. Am. Express Co., 138 S. Ct. 2274 (2018). 

184 Id. at 205. 
185 See NCAA, 468 U.S. 85. 
186 See 8 AREEDA & HOVENKAMP, supra note 139, at ¶¶ 1624d, 1625, 

1627. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 77 

standard.187 One might continue with such illustrations. But 

it should be clear that when the government was seeking only 

an injunction against the rule rather than complete 

destruction of the defendant’s business method, then it is the 

effect of that particular rule that must be examined. Here, 

those customers affected by the antisteering rule were those 

that would have switched to a less costly card but for the rule. 

By implication, the value they placed on the defendant’s perks 

was less than the incremental price to merchants of using the 

Amex card. 

Consider the following example, which can readily be 

generalized.188 On a typical transaction, the Amex merchant 

acceptance fee may be fifty percent greater than the fee 

charged by competing cards. Suppose that on a particular 

purchase Amex’s merchant fee was $30, but $20 for Visa. This 

$10 difference creates bargaining room—a “surplus,” in 

Coasean terms189—for the merchant and the cardholder to 

strike a mutually beneficial deal. Suppose that the merchant 

offers the customer a $6 discount for using a Visa card instead, 

which would make the customer $6 better off for that 

particular transaction and the merchant $4 better off. The 

customer would agree if the value it placed on Amex’s perks 

was less than the $6 price discount. 

The antisteering provision prevents this transaction from 

occurring, however. As a result, the customer stays with the 

Amex card and experiences a $6 loss. The merchant loses $4 

as well. So, far from being a situation where value goes up on 

one side and down on the other, it actually goes down on both 

sides. At the margin, both the cardholder side and the 

merchant side of the platform are losers. In addition, the 

competing platform, Visa, is also worse off because it was 

denied the opportunity to offer a lower cost substitute 

transaction. The only entity that is better off is Amex—the 

owner of the platform itself, but not the dealing parties on one 

 

187 See 13 HERBERT HOVENKAMP, ANTITRUST LAW, ¶ 2231 (3d ed. 2012). 

188 The analysis here relies on Hovenkamp, supra note 15, at 27. 
189 See, e.g., Robert Cooter, The Cost of Coase, 11 J. LEGAL STUD. 1, 4 

(1982) (discussing R. H. Coase, The Problem of Social Cost, 3 J.L. & ECON. 

1 (1960)).  



  

78 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

or the other side of the platform. It is better off because the 

Amex cardholder who would have switched did not place much 

value on the Amex perks, suggesting that margins on those 

sales would have been particularly high. The Amex 

cardholders most likely to switch are those that would benefit 

most from using a different card. 

As the district court observed, other Amex cardholders 

would decline the merchant’s offer to switch, because for them 

the value of the perks might be as high as the merchant’s 

acceptance fee, or at least as high as that portion of the fee 

that the merchant offered them for switching.190 Cardholders 

whose behavior was actually changed by the antisteering rule 

were worse off as a result of the rule, thus creating lost value 

on both sides of the platform.191 That is why, even if the 

market were irrationally defined as including both sides, the 

way the majority defined it,192 competitive harm was 

apparent: at the margin, cardholders, merchants, and rival 

platforms were all injured by an output-reducing restraint. 

Whom does the antisteering rule benefit in this case? Not 

the traders on either side of the platform, but only Amex itself, 

who is able to retain that transaction at margins in excess of 

 

190 United States v. Am. Express Co., 88 F. Supp. 3d 143, 220 (E.D.N.Y. 

2015), rev’d, 838 F.3d 179 (2d Cir. 2016), aff’d sub nom. Ohio v. Am. Express 

Co., 138 S. Ct. 2274 (2018) (“[E]ven if a merchant is inclined to steer away 

from American Express, the cardholder would still have the freedom to use 

an Amex card if the cardholder decides the rewards offered by American 

Express are of greater value than the discount, in-kind perk, or other benefit 

offered by the merchant.”). 
191 The Second Circuit stated the requirement as whether the 

antisteering rule “made all Amex consumers on both sides of the platform—

i.e., both merchants and cardholders—worse off over all.” United States v. 

Am. Express Co., 838 F.3d 179, 205 (2d Cir. 2016), aff’d sub nom. Ohio v. 

Am. Express Co., 138 S. Ct. 2274 (2018). The phrase “over all” is ambiguous, 

but it may imply the proper question, and the one that the government 

actually answered, which is that it made every affected merchant and 

cardholder worse off. For example, in an exclusive dealing case courts do 

not ask if every customer would have switched but for exclusive dealing, 

thus giving the defendant a market share of zero. Instead, the question is 

whether enough would have switched to create an inference of competitive 

harm. 
192 See supra notes 64–65 and accompanying text. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 79 

the value that these cardholders placed on use of the card. 

Amex benefits because its volume increases by that one 

transaction, but Visa loses that same transaction.193 One 

might be tempted to describe that shift as a wash, but in fact 

it is not. The shift to Visa, the preferred platform for both the 

cardholder and the merchant, would produce both higher 

cardholder and merchant value as well as higher output in the 

product market. To be sure, the loss of these transactions 

would cost Amex revenue and there would be feedback effects 

that would generate a new, and likely less profitable 

equilibrium for Amex, but that is what competition is all 

about. 

This is the point at which a possible defense would come 

in. One factor worth examining is whether preserving the 

transaction to Amex was necessary to the viability of its 

business model. And if so, does that provide a benefit to 

competition in excess of costs? For example, Amex might 

argue that it needs a certain minimum transaction volume 

coupled with higher prices in order to be profitable. First of 

all, this query does not depend on whether there is one market 

or two.194 Indeed, it does not even depend on the existence of 

a platform, but only on the existence of scale economies or 

other attributes relating profitability to scale. These are core 

issues in industrial organization. Second, the need to 

maintain viability while charging higher prices hardly sounds 

like a meritorious antitrust defense. 

Steering would permit the bargaining parties—i.e., 

merchants and customers—to negotiate to the joint 

maximizing position. Consumers who place a small value on 

Amex’s perks could use a different form of payment and would 

be better off. For their part, merchants could bargain by 

discounting the price, or offering collateral services such as 

free delivery, to reflect the merchant costs of a particular 

payment form. The important thing is that in the absence of 

transaction costs and under good information, everything 

 

193 See Hovenkamp, supra note 15, at 30–31. 
194 See id. at 25–27. 



  

80 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

would get discounted into the purchase price.195 This becomes 

an important efficiency principle: payment systems should be 

“neutral” and transparent, permitting the parties to negotiate 

to a mutually beneficial maximum.196 The Amex antisteering 

rule was a bargaining impediment that prevented the 

payment platform’s own participants from reaching a joint-

maximizing deal. Moreover, in the process of injuring its own 

participants, it also excluded rival card platforms who were 

ready, willing, and able to offer better terms. Consequently, it 

resulted in higher product prices across the board.  

Looking at the situation as a whole, it seems clear that the 

district court and Justice Breyer got the issue right through a 

careful examination of the record. Stated in rule of reason 

terms, the question was whether the plaintiff had presented 

enough evidence of competitive harm to require the defendant 

to offer a defense. The harms were clear: cardholders were 

denied an opportunity to obtain a lower price, the merchants 

were denied the opportunity for a less costly transaction, and 

rival cards that were less costly lost sales. From a consumer 

welfare perspective, the directly affected consumers were 

worse off, as well as other consumers who were forced to pay 

higher product prices regardless of the form of payment they 

 

195 See R. H. Coase, The Problem of Social Cost, 3 J.L. & ECON. 1, 1–16 

(1960). 
196 See Rochet & Tirole, supra note 15, at 648 (“Neutrality in payment 

systems. The choice of an interchange fee paid by the merchant’s bank, the 

acquirer, to the cardholder’s bank, the issuer, is irrelevant if the following 

conditions are jointly satisfied: First, issuers and acquirers pass through the 

corresponding charge (or benefit) to the cardholder and the merchant. 

Second, the merchant can charge two different prices for goods or services 

depending on whether the consumer pays by cash or by card; in other words, 

the payment system does not impose a no-surcharge rule as a condition for 

the merchant to be affiliated with the system. Third, the merchant and the 

consumer incur no transaction cost associated with a dual-price system.”). 

As Rochet and Tirole observe, the Coase Theorem indicates that in a well-

functioning market, merchants and customers would move to a wealth 

maximizing equilibrium. Id. at 649. But the minimum conditions are that 

the parties are free to bargain (i.e., no prohibition on steering) and that they 

have adequate information about the gains that would be available from 

trading. See id. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 81 

chose.197 While Amex itself was benefitted by preserving the 

transaction to its own system, this was at best a wealth 

transfer from whatever platform lost the transaction.198 In 

fact, the antisteering rule was not a neutral wealth transfer 

at all, but rather a transfer from a competing platform that 

generated higher value to one that generated lower value. As 

Justice Breyer noted in dissent, looking at the prima facie 

case, “there is little more that need be said.”199 

H. Implications for Market Definition 

Amex raises important issues concerning market definition 

in antitrust cases. As discussed above, grouping both sides of 

a platform into a single relevant market was economic 

nonsense and, in any event, unnecessary to the analysis that 

the Court undertook.200 Nevertheless, the law is what it is 

and, that leaves important questions about the scope of the 

decision’s holding. 

The Court held that not every two-sided platform qualified 

for its unique approach, but noted that transactional 

platforms in which there is a simultaneous one-to-one 

correspondence between the transactions on one side of the 

platform and those on the other side are “different.”201 Even 

this definition was too broad to be supported by some of the 

literature the majority cited.202 Further, it was apparently 

 

197 See supra notes 180–92 and accompanying text. 
198 See Hovenkamp, supra note 15, at 30. 
199 Ohio v. Am. Express Co., 138 S. Ct. 2274, 2294 (2018) (Breyer, J., 

dissenting). 

200 See supra notes 101–12 and accompanying text. 
201 See Am. Express Co., 138 S. Ct. at 2286 (requiring “a single, 

simultaneous transaction between participants”). 
202 As Justice Breyer noted in his dissent, see id. at 2300, the Court’s 

definition was broader than the definition given even in the economic 

literature that the majority opinion cited. For example, Rochet & Tirole, 

whom the majority relies on, say this: 

“Getting the two sides on board” is a useful characterization, 

but it is not restrictive enough. Indeed, if the analysis just 

stopped there, pretty much any market would be two-sided, 

since buyers and sellers need to be brought together for 



  

82 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

driven by the Court’s mistaken view that in a transactional 

platform each gain on one side generates an equal-and-

offsetting benefit on the other side.203 

Nevertheless, under the majority’s approach the two sides 

of the Amex platform constitute a single relevant market 

because a $50 transaction on the consumer side of the Amex 

platform is simultaneously offset by a $50 transaction on the 

merchant side, less Amex’s acceptance fee. 

Upon first glance, Uber would appear to be another 

instance of a pure transaction platform under this approach. 

Each ride that a passenger purchases generates a 

corresponding and simultaneous payment to the driver, less 

the fee. Interbank ATM platforms are very likely in the same 

category. The same may also be true of airline and hotel 

reservation websites, such as Orbitz and Expedia, and 

perhaps of venue websites such as Ticketmaster.204 All of 

these cases, however, will depend on a careful evaluation of 

the facts. 

The Court also observed that other types of platforms 

exhibit a looser relationship between transactions on one side 

and those on the other, and these would not fall within the 

Court’s “single market” rule. It noted sales of newspapers and 

 

markets to exist and gains from trade to be realized. We 

define a two-sided market as one in which the volume of 

transactions between end-users depends on the structure 

and not only on the overall level of the fees charged by the 

platform. A platform’s usage or variable charges impact the 

two sides’ willingness to trade once on the platform and, 

thereby, their net surpluses from potential interactions; the 

platforms’ membership or fixed charges in turn condition 

the end-users’ presence on the platform. The platforms’ fine 

design of the structure of variable and fixed charges is 

relevant only if the two sides do not negotiate away the 

corresponding usage and membership externalities. 

Rochet & Tirole, supra note 15, at 646. 
203 See supra notes 188–93 and accompanying text. 
204 See, e.g., Brief of US Airways at 2, US Airways, Inc. v. Sabre 

Holdings Corp., No. 17-960(L) & 17-983(XAP), 2018 WL 3456163, (2d Cir. 

July 16, 2018) (considering whether airline Global Distribution System 

(GDS) for travel agents, which facilitates ticket transactions, is a qualifying 

single market under Amex). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 83 

advertising, where the relationship is more attenuated.205 

That would also be the case for most magazines. It is also true 

of computer search engines, advertiser-supported music 

streaming, and other advertiser-supported services. For 

these, there is no balanced one-to-one transaction between the 

two sides. Other pay websites that do not exhibit one-to-one 

relationships include services such as Netflix, Amazon Prime 

Video, and Hulu. For these, subscribers typically pay a 

regular monthly fee after which the incremental cost of 

content is free. Therefore, there is neither simultaneity nor a 

very close correspondence between the size of the fee and the 

volume of content that the viewer can access.206 For example, 

a Netflix subscriber pays $8.99 a month for a basic 

subscription whether she watches one movie during that time 

period or a dozen.207 Amazon Prime Video is a little more 

complex because it concurrently offers “Prime” movies at an 

incremental price of zero and also pay movies, for which 

Amazon collects an individual fee, say $3.99, from 

customers.208 

There is an additional problem with movie and music 

streaming, however, which is that the platform operator itself 

is often the seller or licensor. Both Netflix and Amazon, as 

well as the music streamers, typically obtain nonexclusive 

licenses to the content that they stream, often in fixed-cost 

license agreements.209 As a result, they are not acting as a 

 

205 See Am. Express Co., 138 S. Ct. at 2286. 
206 For example, all Netflix plans offer unlimited viewing of Netflix 

content for a single monthly fee. See Choose Your Plan, NETFLIX, 

https://www.netflix.com/signup/planform (on file with the Columbia 

Business Law Review). 
207 See Choose a Plan That’s Right for You, NETFLIX, 

https://www.netflix.com/signup/planform (on file with the Columbia 

Business Law Review). 
208 See Andy Beatman, What Is Prime Video? – Amazon Prime Insider, 

AMAZON, https://www.amazon.com/primeinsider/video/prime-video-qa.html 

[https://perma.cc/Y5B9-LYDF] (describing both “Prime” content and rental 

content). 
209 See Erik Hovenkamp & Neel U. Sukhatme, Vertical Mergers and 

the MFN Thicket in Television, CPI ANTITRUST CHRON., Aug. 2018, at 4, 

https://www.competitionpolicyinternational.com/wp-content/uploads 



  

84 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

platform intermediary between the movie’s owner and the 

customer. Having acquired a nonexclusive license at a fixed 

fee that permits relicensing, Netflix or Amazon are 

themselves the sellers, so there are not two sides of the market 

to define but only a seller on one side and a buyer on the other 

side, as with any market. This is a critical distinction. The 

premise in a case such as Amex is that the merchant is selling 

its own product and the customer is buying it. Amex is only 

the platform intermediary facilitating the transaction. The 

merchant does not sell the product to Amex, who then in turn 

sells it to the customer. The same is true of Uber: Drivers do 

not sell rides to Uber, which in turn sells them to passengers. 

The Amex majority’s approach does not apply when the 

merchandise is actually sold or licensed to the operator of the 

platform. 

By the same token, to the extent that Amazon, Walmart, 

Target, or numerous other retailers purchase goods from 

manufacturers and then sell them on their websites, the 

platform operator is not merely a transaction facilitator. 

Therefore, the established economics of market definition 

should apply in those cases. If websites such as Orbitz and 

Expedia purchase blocks of rooms for resale, or if 

Ticketmaster purchases a block of tickets for a particular 

performance, those transactions do not qualify for Amex’s 

market definition approach either. Likewise, in a blanket 

license case such as BMI, the artist provides BMI with a non-

exclusive license and subsequently BMI sells a blanket license 

to a radio station.210 There is neither simultaneity nor a one-

to-one transactional correspondence. 

The status of app stores on smartphones may become 

relevant in consideration of Apple, Inc. v. Pepper, which the 

 

/2018/08/CPI-Hovenkamp-Sukhatme.pdf [https://perma.cc/4SBK-RTH7]. 

Under fixed-cost license agreements the licensee, such as Netflix, pays a 

one-time fee for a nonexclusive license to a film for a given time period, but 

does not pay a per use fee. See id. 
210 See Broad. Music, Inc. v. CBS, Inc., 441 U.S. 1, 4 (1979); see also 

supra notes 22–24 and accompanying text. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 85 

Supreme Court will soon address.211 The issue is whether 

customers who purchase apps for their iPhones or other Apple 

devices through Apple’s App Store, are direct purchasers from 

Apple. Assuming that the app producers are the violators, the 

underlying legal question is whether the app platform is a 

mere broker between the app producer and the consumer, or 

a purchaser-reseller. If the latter, the indirect purchaser rule 

barring damages actions to indirect purchasers applies.212 A 

question pertinent to Amex that is not necessarily governed 

by Illinois Brick is whether the transactions were 

“simultaneous,” as Amex requires. For instance, the app 

producer may have licensed and delivered its app to Apple in 

advance, which held them in its own cloud or storage devices, 

delivering them to a customer upon order. Under that 

scenario, the developer sold (licensed) the app to Apple, which 

held it until a later time when a customer bought (licensed) a 

copy.213 One important principle is that Amex does not provide 

a basis for turning ordinary vertical distribution into a single 

market at both the upstream and downstream levels. Many of 

the economic effects commonly attributed to platforms are 

similar to those that result from ordinary vertical 

distribution.214 

There are also intermediate platforms where the relation 

between transactions on one side of the platform and those on 

the other side is not one-to-one and, in most cases, not 

simultaneous either. A prominent example is health 

 

211 See In re Apple iPhone Antitrust Litig., 846 F.3d 313 (9th Cir. 2017), 

cert. granted sub. nom. Apple, Inc. v. Pepper, 138 S. Ct. 2647 (2018); see also 

Salveson v. JP Morgan Chase & Co., 663 F. App’x 71 (2d Cir. 2016); In re 

Payment Card Interchange Fee & Merch. Disc. Antitrust Litig., 05-MD-

1720, 2018 WL 4158290, at *3 (E.D.N.Y. Aug. 30, 2018) (noting possible 

relevance of Amex to decision considering whether credit cardholders were 

indirect purchasers from issuing banks). 
212 See Ill. Brick v. Illinois, 431 U.S. 720, 730–31 (1977); see also 2A 

PHILLIP E. AREEDA, HERBERT HOVENKAMP, ROGER D. BLAIR & CHRISTINE 

PIETTE DURRANCE, ANTITRUST LAW ¶ 346 (4th ed. 2014). 

213 Apple’s position on the issue is articulated in a brief to the Supreme 

Court. See Brief of Petitioner at 2–3, In re Apple iPhone Antitrust Litig., 138 

S. Ct. 2647 (No. 17-204). 
214 On this point, see Carlton & Winter, supra note 17, at 215. 



  

86 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

insurance networks. On one side is the insured patient, who 

receives a covered medical procedure. On the other side is a 

health care provider. In the middle is the platform, which is 

an insurer who collects premiums from the insured or her 

employer and pays the provider’s claim.215 Here, however, the 

sales to insured are neither simultaneous nor are they made 

on a matching fee-for-service basis. 

Actually, the arrangement that the Supreme Court 

condemned under the per se rule in its Arizona v. Maricopa 

County Medical Society216 decision looks a little more like a 

two-sided platform as the Amex opinion defined it. In that 

case, physician participants agreed to be compensated at 

stipulated fees per service, which were paid by the insured or, 

more likely, an employer.217 Even this arrangement would not 

meet the Amex definition unless payment and receipt were 

simultaneous. In any event, since Maricopa condemned such 

arrangements, insurers have taken a more actuarial approach 

that requires providers to share a certain amount of risk.218 

The Maricopa decision itself contemplated that result, 

suggesting that firms that “pool their capital and share the 

risks of loss as well as the opportunities for profit” be treated 

more like an integrated single entity.219 

In the modern health insurance network, characterized by 

risk sharing and actuarial pricing, it clearly is not the case 

that the platform “facilitate[s] a single, simultaneous 

 

215 One decision that discussed the then-pending Amex decision 

involved a horizontal territorial division and price-fixing agreement among 

Blue Cross affiliates is In re Blue Cross Blue Shield Antitrust Litig., 308 F. 

Supp. 3d 1241, 1276 n.20 (N.D. Ala. 2018); see also Lifewatch Servs. Inc. v. 

Highmark Inc., 902 F.3d 323, 337 (3d Cir. 2018), which mentioned the 

possible relevance of Amex to a market definition question in a health 

insurance market, but did not decide the case on that basis. 
216 See Arizona v. Maricopa Cty. Med. Soc’y, 457 U.S. 332 (1982). 
217 Id. at 339–40. 
218 See Scott D. Danzis, Revising the Revised Guidelines: Incentives, 

Clinically Integrated Physician Networks, and the Antitrust Laws, 87 VA. L. 

REV. 531, 531–44 (2001); Thomas L. Greaney, Managed Competition, 

Integrated Delivery Systems and Antitrust, 79 CORNELL L. REV. 1507, 1530 

(1994). 
219 Maricopa, 457 U.S. at 356. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 87 

transaction between participants[,]” as the Supreme Court 

required.220 The insurer might receive a premium of, say, 

$500 per month from the insured and pay the insured’s 

medical expenses for that month, less any deductibles, copays, 

and the like. The insurer’s payment could be greater or 

smaller than $500 depending on the insured’s needs. To be 

sure, there is often a co-payment, such as a flat fee per office 

visit. However, this co-payment is often not made to the 

insurer at all, but rather directly to the health care provider. 

In that case the network is not acting as an intermediary. 

Once again, there is no substitute for careful examination of a 

record. 

Suppose that an insurer network enters an exclusive 

agreement with a provider, effectively denying that provider 

the ability to service other networks or health payment 

systems. Ordinary exclusive dealing principles might require 

foreclosure on the order of, say, thirty percent for a prima facie 

case.221 This might require that both a provider market and a 

consumer market be defined, but it would not require 

anything as economically incoherent as putting them into the 

same market.222 In response, the defense might be raised that 

the network provider needs exclusive dealing with, say, 

anesthesiologists in order to make its network economically 

viable, but that is a question that can largely be answered 

independently of network considerations. 

Other platforms not included in Amex’s “single market” 

definition include intermediaries that bring buyers and 

sellers together but have little to do with the resulting 

transaction. For example, real estate websites such as 

realtor.com and Zillow.com identify real properties that are 

for sale or rent. Having settled on a property, a prospective 

purchaser then contacts the broker by email or telephone and, 

after subsequent negotiations, there may be a sale. But none 

of this comes close to the kind of simultaneous one-to-one 

 

220 Ohio v. Am. Express Co., 138 S. Ct. 2274, 2286 (2018). 
221 See 11 HOVENKAMP, supra note 117 ¶ 1821c (discussing foreclosure 

percentages and suggesting a minimum in the range of thirty percent). 
222 See supra notes 122–26 and accompanying text. 



  

88 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

transaction that was present in Amex. The same thing is 

largely true of dating websites such as Match.com or 

OkCupid.com. Typically, members on both sides pay a 

monthly or annual fee, although some sites also offer free 

versions. As in the case of real estate sites, however, the 

website does little more than introduce two people to each 

other. What, if anything, happens later occurs largely off the 

site. At first glance, Craigslist appears to resemble these sites 

more than, say, eBay, which actually completes transactions 

on the site. On Craigslist, offerors of merchandise or services 

essentially post an advertisement with contact information, 

but prospective purchasers typically make their contact and 

any subsequent transaction off the site. 

The same thing appears to be true of the NCAA, which may 

be a multi-sided platform that, according to one expert, brings 

together students, student athletes, alumni, coaches and 

athletic staff.223 Factually, this may be true, but that will not 

satisfy the definition of a single two-sided relevant market for 

antitrust purposes unless someone can show the requisite 

simultaneous one-to-one transaction between both sides. 

There are certainly other examples, but the important 

point is that only a relatively small subset of two-sided 

platforms fall within the Court’s requirements for treating the 

two sides as a single market. On this question, maintaining a 

coherent economic approach to antitrust policy requires that 

Amex be limited to its facts. In any event, the Supreme Court 

was clear that the two sides should not be treated as a single 

market unless they were characterized by “transactions” that 

were both “simultaneous” and one-to-one.224 

IV. CONCLUSION 

One danger of the Amex decision is its signal that neither 

close economic analysis nor careful examination of the record 

is necessary to apply antitrust law under the rule of reason. 

That, of course, flies in the face of a century-long history of 

 

223 See In re NCAA Athletic Grant-in-Aid Cap Antitrust Litig., No. 14-

md-02541 CW, 2018 WL 4241981, at *4 n.3 (N.D. Cal. Sep. 3, 2018). 
224 Am. Express Co., 138 S. Ct. at 2286. 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 89 

rule of reason analysis in the federal courts, which has always 

emphasized careful examination of a well-developed record on 

issues pertaining to both power and conduct.225 Judge William 

Howard Taft himself distinguished ancillary from naked 

restraints only by careful examination of the facts.226 

The Amex majority never concluded that the district court’s 

fact findings were an abuse of discretion or otherwise 

improper. Rather, it simply ignored them. Nor did it declare 

as a matter of law that close examination of a factual record 

is unimportant in antitrust cases under the rule of reason. As 

a result, the lower federal courts should not feel precluded 

from engaging in the kind of close transactional analysis that 

the rule of reason traditionally requires if decisions are to be 

economically coherent. 

The economic literature on two-sided platforms has made 

major contributions to price and industrial organization 

theory in a wide variety of markets. It deserves an important 

position in both industrial economics and competition policy. 

At the same time, however, its influence should not be 

exaggerated. It is, essentially, a tool of neoclassical economics, 

not a discovery that realistically threatens to alter the 

foundations of economics. 

The two-sided-platform literature is strongly reminiscent 

of another development in the theory of industrial 

organization thirty years ago. That theory, termed 

“contestable markets,” grew out of the imminently reasonable 

observation that where a market contains only one seller, 

competition “for the market” can yield competitive outcomes 

just as much as competition by multiple incumbents “in the 

market.”227 The theory of contestable markets was introduced 

 

225 See Hovenkamp, supra note 99, at 101–02. 
226 See United States v. Addyston Pipe & Steel Co., 85 F. 271, 280–83 

(6th Cir. 1898), modified and aff’d, 175 U.S. 211 (1899). 
227 On the initial debate, see generally Harold Demsetz, Why Regulate 

Utilities?, 11 J.L. & ECON. 55 (1968) and Oliver E. Williamson, Franchise 

Bidding for Natural Monopolies—In General and With Respect to CATV, 7 

BELL J. ECON. 73 (1976). For a reprise, see Herbert Hovenkamp, Regulation 

and the Marginalist Revolution, FLA. L. REV (forthcoming 2019), 



  

90 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

by prominent economists with great fanfare, producing a 

spate of articles and at least one important book.228 The late 

William J. Baumol, a past president of the American 

Economic Association, proclaimed it to be an “uprising” in the 

theory of industry structure.229 It promised to eliminate the 

need for such things as public utility or airline regulation 

because even a natural monopolist incumbent knew that the 

instant it attempted to charge too high a price a potential rival 

would swoop in and steal the business. 

But the theory never lived up to anything remotely 

resembling its expectations, although it did provide some 

valuable lessons. Even in the airline industry, thought to be a 

prime target for contestability, competition among incumbent 

carriers remains an important determinant of price and 

output. The theory of platform markets will pursue much the 

same course. After a brief period of exaggeration, industrial 

organization theory will be enriched, but will remain 

fundamentally the same. The Amex majority opinion serves to 

highlight what happens when a Court abandons fundamental 

economics in its haste to encounter something new. 

The decision that seems to come closest to Amex as an 

economic “misfire” is the Supreme Court’s 1992 ruling in 

Eastman Kodak Co. v. Image Technical Services, in which the 

Court held that sufficient power to condemn a tie of parts and 

service by a nondominant firm could be inferred from 

consumer “lock in.”230 Kodak was a six to three decision, but 

the reaction to Kodak was so strongly critical that subsequent 

lower court decisions went to great lengths to limit it.231 It has 

 

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3181852 

[https://perma.cc/6ZWD-MAKB]. 
228 See WILLIAM J. BAUMOL, JOHN C. PANZAR & ROBERT D. WILLIG, 

CONTESTABLE MARKETS AND THE THEORY OF INDUSTRY STRUCTURE (1988). 
229 William J. Baumol, Contestable Markets: An Uprising in the Theory 

of Industry Structure, 72 AM. ECON. REV. 1, 1 (1982). 
230 See Eastman Kodak Co. v. Image Tech. Servs., 504 U.S. 451, 496 

(1992). 
231 On the case law limiting Kodak, see 10 PHILLIP E. AREEDA & 

HERBERT HOVENKAMP, ANTITRUST LAW ¶ 1740 (4th ed. 2018) (discussing 

tying arrangements and the Kodak decision). 



  

No. 1:35] PLATFORMS AND THE RULE OF REASON 91 

had little impact on antitrust outcomes even though lock-in is 

more prevalent today in our modern networked world than it 

was in 1992. 

Other consequences could be on the horizon. This decision 

will encourage more legislation and regulation as more 

decision makers lose confidence in judge-made antitrust rules 

to promote competition. As Justice Breyer noted in his dissent, 

several jurisdictions around the world have acted against high 

interchange fees and antisteering rules, mostly by statute or 

agency rule.232 The United States legal system has 

historically relied less on regulation and more on antitrust 

law, which can be much less intrusive. But what this decision 

describes as “steering” is actually among the most ordinary 

and essential of competitive functions: encouraging people to 

acquire information and giving them the option to choose. 

This process protects the competitive process, both improving 

product quality and driving prices to the competitive level. For 

example, a common concern about healthcare costs is that 

they are so high because patients are indifferent to prices. 

First, medical bills are paid indirectly by insurers. Second, 

most patients do not even pay the insurance premium; rather, 

it is paid by either an employer or a government agency. As a 

result, the patient bears only a small portion of the cost and is 

inclined to spend too much. The antisteering rule operates in 

much the same way: it makes the cardholder indifferent to 

merchant costs and thus diminishes the consumer incentive 

to reduce them. 

Today, the consumer welfare principle in antitrust is under 

attack from people who argue for abandonment of economic 

approaches to antitrust in favor of populism, political theory, 

or some other source.233 Decisions like Amex add fuel to their 

 

232 See Ohio v. Am. Express Co., 138 S. Ct. 2274, 2290 (2018) (Breyer, 

J., dissenting) (citing U.S. GOV’T ACCOUNTABILITY OFF., GAO-08-558, CREDIT 

AND DEBIT CARDS: FEDERAL ENTITIES ARE TAKING ACTIONS TO LIMIT THEIR 

INTERCHANGE FEES, BUT ADDITIONAL REVENUE COLLECTION COST SAVINGS 

MAY EXIST 31–35 (2008)). 
233 See generally Barak Orbach, Antitrust Populism, 14 N.Y.U. J.L. & 

BUS. 1 (2017); Carl Shapiro, Antitrust in a Time of Populism, 61 INT’L J. 

INDUS. ORG. 714 (2018); see also Hovenkamp, supra note 161, at 1–4. 



  

92 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 

cause. The success of antitrust as an enterprise driven by 

economic policy depends on the ability and willingness of 

judges to use economics effectively, bringing monopoly prices 

and output restrictions under control while protecting 

provable efficiencies. The rule of reason cannot be simply an 

excuse for judges to ignore well developed records and sound 

economic theory in order to reach a conclusion that they find 

pleasing on noneconomic grounds.