THE ANTICOMPETITIVE EFFECTS OF VERTICAL MOST-FAVORED-NATION RESTRAINTS AND THE ERROR OF AMEX Dennis W. Carlton* I. Introduction ................................................................... 93 II. Vertical Most-Favored-Nation Restraints ................... 95 III. Application to Credit Cards .......................................... 98 IV. Application to Amex .................................................... 100 I. INTRODUCTION This Essay explains why the Supreme Court’s economic reasoning in its recent Ohio v. American Express Co. (“Amex”) decision is wrong. The Amex case involved the use of what are called “antisteering” restraints in which a retailer is not allowed to use a variety of tactics to steer a consumer away from using an American Express (“Amex”) card and toward using another payment mechanism.1 The reason why a merchant might want to do this is because the cost that the merchant incurs when a customer uses an Amex card can be higher than the cost that the merchant incurs when the customer uses either another credit card, debit card, or cash.2 Although not challenged in the Amex case, the Amex contractual rules also prevent a retailer from imposing a surcharge on customers who use an Amex card to reflect the higher merchant cost.3 It is interesting to note that some * This paper is based on my William Howard Taft Lecture, presented September 14, 2018, delivered to the New York State Bar Association Antitrust Law Section. I thank Gustavo Bamberger, Daniel Feder, Patrick Gallagher, Allan Shampine, and Ralph Winter for helpful comments. 1 See Ohio v. Am. Express Co., 138 S. Ct. 2774, 2280 (2018). 2 See id. at 2282–83. 3 See United States v. Am. Express Co., 88 F. Supp. 3d 143, 163 (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). 94 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 countries—such as Australia4—have regulated certain credit card fees, others have forbidden credit card companies from telling merchants that they cannot surcharge,5 and some states in the United States—such as New York6—have forbidden merchants from surcharging. Restraints on surcharging or steering are examples of restraints that Ralph Winter and I call “vertical most-favored-nation restraints,” (“vMFN”) in which one supplier tells a retailer that the retailer cannot set the retail price of its product higher than that of a rival, even if its wholesale price is higher than that of its rival.7 Such restraints have been the subject of some litigation already, but I expect that with the increasing use of web based platforms where such restraints are often used, litigation regarding such restraints will increase. This Article illustrates the underlying economic logic behind the anticompetitive effect of vMFNs.8 I then apply the reasoning to credit cards9 and finally, using the economic framework developed, explain the economic errors in the Court’s Amex decision.10 For a more detailed discussion, please see the Carlton and Winter paper referenced herein.11 4 See, e.g., Competition and Consumer Amendment (Payment Surcharges) Act 2016 (Cth) sch 1 (Austl.). 5 See Regulation (EU) 2015/751 of the European Parliament and of the Council of 29 April 2015 on Interchange Fees for Card-based Transactions, art. 11, 2015 O.J. (L 123) 1, 13. 6 See N.Y. GEN. BUS. LAW § 518 (McKinney 2018). 7 Dennis W. Carlton & Ralph A. Winter, Vertical Most-Favored-Nation Restraints and Credit Card No-Surcharge Rules, 61 J.L. & ECON. 215 (2018). I refer the interested reader to that paper for a more detailed discussion of the issues in this Essay, as well as for the mathematical proofs of key propositions. Both authors of that article have appeared as experts adverse to credit card companies in litigation in the United States and in foreign countries. Id. at 215. 8 See infra Part II. 9 See infra Part III. 10 See infra Part IV. 11 See Carlton & Winter, supra note 7. No. 1:93] ANTICOMPETITIVE EFFECTS 95 II. VERTICAL MOST-FAVORED-NATION RESTRAINTS A vMFN is best illustrated by example. Suppose that there are two manufacturers. Manufacturer A produces product A while Manufacturer B produces product B. The two products compete with each other. Each manufacturer sells to a retailer at some wholesale price and the retailer then sets the retail price for each product. Manufacturer A tells the retailer that the retailer must abide by the following restriction: the retail price for product A can be no higher than that for product B, regardless of any difference in wholesale prices between the two products. Manufacturer B imposes the same restriction. What is the effect of these restrictions? Let’s use a simple example to illustrate the effect. Suppose that each product is so popular that every retail store wants to carry both products.12 In the absence of the vMFNs, Manufacturer A will set the wholesale price for its product and Manufacturer B will do the same. In deciding what wholesale price to set, Manufacturer A will recognize that if it lowers its wholesale price below that of product B, then that will typically lead the retailer to lower the retail price of product A below that of product B. By doing so, the retailer induces some consumers to switch from B to A in order to take advantage of the lower retail price of A. If Manufacturer A raises its wholesale price, it will recognize that it will lose some customers to B as the retailer responds to the higher wholesale price of A by raising the retail price of A relative to the retail price of B. Manufacturer B faces the same incentives. Depending on how strong the competitive forces are and the closeness of substitution of A and B, the wholesale price of A and B are driven down, perhaps all the way down to their own costs. That is how competition works. In contrast, with vMFNs, the incentives to lower price are dramatically changed. Now, if Manufacturer A lowers its wholesale price, the vMFN prevents the retailer from 12 The “must-carry” restriction allows the analysis to focus on the main source of competition: the consumer’s ability to substitute products. See id. at 217. 96 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 lowering the retail price of product A relative to that of B. Therefore, there is no gain in sales arising from the lowered wholesale price as a result of a lower retail price of A relative to the retail price of B. This makes lowering the wholesale price less desirable than without the vMFNs. If Manufacturer A raises its wholesale price, then it no longer has to worry that it will lose sales at retail because its retail price will rise relative to that of product B. Again, the vMFNs guarantee that that cannot happen. So, the incentive to lower wholesale price is reduced while the incentive to raise the wholesale price is increased. In equilibrium, the result is that the wholesale and retail prices of both products A and B are above the level that results when there are no vMFNs.13 In the absence of vMFNs, products A and B are substitutes in the precise sense that an increase in the wholesale price of product A, leads to a decline in demand for product A and an increase in demand for product B as customers respond to the lower retail price of product A. With vMFNs, if Manufacturer A raises his wholesale price, it creates an incentive for Manufacturer B to raise its wholesale price (assuming the price starts from below the monopoly level of Product B). In other words, instead of acting like substitutes, the products act like complements.14 Manufacturer A has an incentive to raise its price as does Manufacturer B, as long as each price is below the monopoly level.15 The vMFNs eliminate the competitive pressures between Manufacturer A and B and lead to higher prices than would occur without the vMFNs.16 There is an additional anticompetitive effect from the use of vMFNs. Suppose there is some firm, Firm X, that wishes to enter to compete against Manufacturers A and B. Suppose further that Firm X intends to enter by following the strategy of charging a very low wholesale price. That strategy might 13 See id. at 223. 14 See id. 15 In fact, there is an incentive to raise price even above the monopoly levels. See id. at 217. 16 See id. at 223 (“Proposition 1. The equilibrium price in the [vMFN] pricing subgame (i) exceeds the equilibrium price in the [non-vMFN] pricing subgame and (ii) exceeds the perfectly collusive price.”). No. 1:93] ANTICOMPETITIVE EFFECTS 97 work well in the world without a vMFN since the low wholesale price from Product X will be reflected in a low retail price for Product X, and that low retail price can induce large numbers of consumers to switch away from the higher priced Products A and B to Product X. In contrast, in a world with a vMFN, that incentive for retail customers to switch to Product X disappears as a result of the vMFN.17 Therefore, the vMFN creates an impediment to entry that does not exist in its absence. As a result, the vMFN restricts entry and thereby reduces the competitive pressures on Manufacturers A and B, leading to higher prices.18 There is an additional effect from the vMFN. It enables dominant firms (“must carry” products) to impose a tax on other products.19 To illustrate this effect in a simple example, consider the following. Suppose initially that there is only Manufacturer A. Now suppose that there is some Product C that is competitively available to the retailer at some constant wholesale cost. The reader can think of Product C as a store brand. Let’s see what happens as a result of the vMFN. Without the vMFN, the retail price of Product A would be higher than the retail price of Product C. With the vMFN, those retail prices must be the same. The price of Product C will rise in order to equal the price of Product A. The new price of A can be either higher or lower than in the case without the vMFN. The retailer will now earn a margin on Product C (its retail price has risen but costs are unchanged). Manufacturer A, if she is able, will try to get that extra profit through some sort of payment from the retailer. It is as if Manufacturer A can place a tax on consumers of Product C and collect the proceeds of that tax. 17 Discover attempted to implement this business model in the late 1990s, but was unsuccessful due to the nondiscrimination provisions used by Amex and other card companies. See Ohio v. Am. Express Co., 138 S. Ct. 2774, 2293–94 (2018) (Breyer, J., dissenting). 18 Carlton & Winter, supra note 7, at 231. 19 See id. 98 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 III. APPLICATION TO CREDIT CARDS The application of the above economic reasoning to credit cards is straightforward. Visa, Mastercard, and American Express all compete to attract merchants to accept their cards and customers to use their cards.20 The credit card firm influences the fee that a merchant must pay every time a customer uses one of its cards.21 If surcharging or steering were allowed at the point of sale, then a merchant could surcharge the most expensive card or otherwise discourage its use in order to induce the customer to use a cheaper card. Just as in the example involving Manufacturers A and B, the use of the vMFN prevents the forces of competition from acting, distorts them, and winds up increasing the fee that the merchant pays for each card compared to what it would have been in the absence of the no surcharge or no steering rule.22 Just as in the example involving Manufacturers A and B, the use of the vMFN makes entry more difficult for any new credit card firm.23 This too causes the merchant fees to be elevated relative to what they would have been in the absence of the vMFN.24 Debit cards and cash often are less costly for the merchant as a payment mechanism than credit cards. But no steering and no surcharge rules prevent a merchant from using certain incentives to induce retail customers to use these lower cost payment mechanisms.25 Just as in the example involving Manufacturer A and Product C, the use of a vMFN raises the 20 See Am. Express Co., 138 S. Ct. at 2282. For simplicity, assume that each card belonging to the same payment network charges the same fees. This simplification is not necessary but makes the exposition easier. 21 See id. at 2281. 22 See supra Part II. 23 See supra note 17 and accompanying text. 24 See supra Part II. 25 A merchant is not allowed to indicate to the consumer its preference for the means of payment (i.e., is not allowed to “steer” customers). See Am. Express Co., 138 S. Ct. at 2283. Amex rules do not allow surcharging of Amex cards unless other credit and debit cards are also surcharged the same amount. See id. However, discounts for cash and all debit are allowed. Id. No. 1:93] ANTICOMPETITIVE EFFECTS 99 price to customers who wish to use debit and cash.26 That is an unambiguous harm to them. Finally, the reader may notice that I have not discussed rewards to cardholders. Indeed, the Court in Amex placed special emphasis on the fact that consumers receive rewards27 as a unique feature of the industry that called for different legal treatment than usually occurs when there is a claim of an anticompetitive vertical restriction.28 I will discuss this a bit more in the next Part (and discuss it extensively in the Carlton & Winter article29). Here I make just one point: rewards used by credit card companies to attract consumers are similar in their function to promotions.30 Firms often engage in advertising, sales efforts, or discount coupons to select customers. In fact, analytically, one can show that there is no difference between the economic incentives to engage in promotion—a feature that has long been understood and studied—and the incentive to engage in rewards. Rewards are just another name for promotion. In Carlton & Winter, we prove this exact equivalence.31 No new economic principles are at work!32 As we will see in the next Part, the Court in Amex seems to think that there is something very special in credit cards flowing from the fact that high merchant fees can be used to 26 Carlton & Winter, supra note 7, at 234. 27 See, e.g., Am. Express Co., 138 S. Ct. at 2280 (“Cardholders also can receive rewards based on the amount of money they spend, such as airline miles, points for travel, or cash back.”). 28 See id. at 2286 (“[C]ourts must include both sides of the platform— merchants and cardholders—when defining the credit-card market.”). 29 See generally Carlton & Winter, supra note 7. 30 See Carlton & Winter, supra note 7, at 240. 31 Id. at 237–38. 32 In no way does this statement diminish the importance of understanding the two-sided nature of a credit card platform—or other platforms—or the intellectual contribution of Rochet and Tirole. See generally Jean-Charles Rochet & Jean Tirole, Two-Sided Markets: A Progress Report, 37 RAND J. ECON. 645 (2006). Quite the reverse. Failure to understand that contribution led the Court astray, as I explain below. 100 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 finance high reward payments to cardholders.33 There is no doubt that as merchant fees rise, more rewards can be financed and that the high merchant fees create an incentive for credit card firms to use rewards to get customers to use their card.34 But that same incentive exists if that high fee resulted from a cartel of the card companies to set the merchant fee at a high level. A cartel price creates incentives for cartel members to use promotion to get customers and earn the (inflated) cartel price.35 IV. APPLICATION TO AMEX The Court in Amex made two key points. First, the Court said that the credit card market is “two-sided.”36 As a result, any economic analysis must take account of both sides of the market.37 Second, the Court said that because the credit card market is two-sided, different legal rules from those ordinarily used are needed for evaluating the effect of any vertical restriction in terms of whether the plaintiff or defendant bears the burden of proof of certain elements of the case. The second point is the more important one but we need to understand the first point before we can discuss the second. It is true, as the Court said, that the “market” for credit cards is two-sided. I use quotation marks around the word “market” because an antitrust market differs from how the term market has often been used in the economic literature. It would be more accurate to say that a credit card firm competes with other credit card firms and that each firm is a “platform” that has two sides.38 By two sides, the economic literature means that a credit card company must attract 33 Am. Express Co., 138 S. Ct. at 2288 (“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.”). 34 See Carlton & Winter, supra note 7, at 230. 35 See id.; see also George J. Stigler, Price and Non-Price Competition, 76 J. POL. ECON. 149, 150–51 (1968). 36 Am. Express Co., 138 S. Ct. at 2280. 37 See id. at 2286. 38 See Rochet & Tirole, supra note 32, at 664. No. 1:93] ANTICOMPETITIVE EFFECTS 101 merchants to make using the card desirable to consumers and it must attract merchants to make the card desirable to customers.39 The more merchants that accept the card, the more customers will be interested in having the card. Similarly, the more customers who have the card, the more merchants will be interested in accepting the card. There is a positive feedback between the two sides of the market. There is no question that this understanding of credit cards is an accepted view as reflected in a large body of literature, to which I and my various co-authors have contributed.40 The economic literature sometimes describes a two-sided platform as one where a firm must get both sides “on board.”41 A successful credit card firm needs both merchants and customers. But such a definition is vague and could apply to many markets since every transaction requires sellers and customers. As Rochet and Tirole explain, with such a definition “pretty much any market would be two-sided[.]”42 But Rochet and Tirole explain that there is a way to distinguish one-sided from two-sided markets. 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.43 That is, it is not only the sum of the prices that matters but also the relative prices on each side of the market.44 So, for example, consider a standard example of a two-sided market, a newspaper. A newspaper sells ads to merchants and sells newspapers to readers. The newspaper sets two prices, one to its readers and the other to its merchants placing ads. The more readers the newspaper gets, the higher the rate it can charge for ads. The price to each side of the market matters. Similarly, in a transaction platform in which a transaction results in a simultaneous exchange between a consumer and a seller, the price to the seller and the price to the buyer 39 See Carlton & Winter, supra note 7, at 217. 40 For further reading, readers should look to the references in Carlton & Winter, supra note 7, at 250–51. 41 See Rochet & Tirole, supra note 32, at 646. 42 Id. 43 See id. at 664–65. 44 See id. 102 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 matters separately if the platform is two-sided. That is, the price a merchant pays for accepting a credit card and the rewards that a consumer receives for using a card matter separately. Otherwise, it is incorrect to label the platform as two-sided. The legal issue the Court opined on has to do with the legal procedure for evaluating vertical restrictions, of which no steering is an example.45 In the typical vertical restriction case, the three step procedure—which both sides in Amex adopted—is: (1) the plaintiff establishes that there is a harm to the competitive process; (2) the defendant can rebut by showing that there is a procompetitive justification for the restrictions, and (3) that the plaintiff can rebut, if necessary, by showing that even if there were a procompetitive justification, that there is a less restrictive alternative.46 So, for example, in an exclusive dealing case: the plaintiff shows that its inability to obtain distribution is a harm to the competitive process. The defendant then shows that the exclusive dealing generates so much promotion that output expands. The plaintiff can then rebut, if necessary, by showing there is a less restrictive alternative to exclusive dealing. From an economic view, it does seem sensible to place the burden of explaining the procompetitive rationale for the challenged conduct on the defendant, rather than the plaintiff, since that is the party with the most information about the practice.47 In Amex, though, the Court ruled that because credit cards are a two-sided market, the plaintiff has the burden of doing both step one and step two in order to show that there is a harm to the competitive process.48 The Court said that the plaintiff had the burden of defining an antitrust market that includes both sides and showing a harm in that market.49 The Court indicated that the plaintiff needed to show that the net price rose, presumably meaning 45 See Ohio v. Am. Express Co., 138 S. Ct. 2774, 2284 (2018). 46 Id. 47 Carlton & Winter, supra note 7, at 239. 48 See Am. Express Co., 138 S. Ct. at 2287. 49 Id. No. 1:93] ANTICOMPETITIVE EFFECTS 103 the merchant fee minus rewards.50 Because the Department of Justice had failed to do that, the Court ruled in favor of Amex.51 By relabeling promotion as “rewards,” the Court turned an ordinary vertical restriction case into one that it felt broke new ground economically and legally and, therefore, had to be treated differently than the ordinary way vertical restriction cases have been handled. The Court’s economic reasoning is muddled. Breyer’s dissent is a clear enumeration of several of its salient flaws.52 I discuss four errors here. The first error is its misunderstanding of two-sided markets. The Court’s use of the term and understanding of two-sided markets is confused. The Court claimed that because credit cards are a two-sided market, the plaintiff had the burden of showing that the net price was adversely affected by the restrictions.53 By “net price,” the Court appears to mean the price to merchants minus rewards to consumers.54 But the key insight of Rochet and Tirole is that in a two-sided platform, the price on each side of the platform matters separately.55 That is, it is not just the sum of prices to each side, but the relative prices on each side, that matters in a two-sided platform.56 Therefore, any interference with the setting of these relative prices is a distortion of the competitive process. That means that the antisteering rules, by their very nature, are a distortion of the competitive process since they alter the relative prices.57 This distortion of 50 See id. The majority states that one must examine both sides of the market in order to determine the cost of credit card services “as a whole.” Id. I take that to mean that the Court focused on the “net price,” which I define as merchant fee minus customer reward. 51 Id. at 2290. 52 See id. at 2297–301 (Breyer, J., dissenting). 53 Id. at 2287. 54 The Court appears to take the view that there are only linear fees on each side of the market, so that “net price” is well defined. It pays no attention to the fact that there may be annual fees or other nonlinear fees, in which case one cannot even define “net price” unambiguously. 55 See Rochet & Tirole, supra note 32, at 664–65. 56 See id. 57 See Carlton & Winter, supra note 7, at 231. 104 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 the competitive process arises from the elimination of the competition among credit card firms at the point of sale, competition that would occur if merchants could surcharge or otherwise steer consumers away from more expensive credit cards. Second, the Court seems to understand that there can be justifications for vertical restrictions based on some well- known free riding arguments.58 For instance, in the above example of exclusive dealing, a standard justification that defendants provide is that the exclusive dealing creates incentives to promote the product.59 But, as Breyer’s dissent makes clear, the free riding justification that the Court provides for Amex makes no economic sense.60 The free riding the Court talks about is the rewards that a consumer receives if she uses the Amex card.61 But if the merchant steers the consumer to use another card, the consumer does not receive the Amex rewards.62 There is no free riding off of the Amex rewards in the Court’s exposition. Third, the Court ignores what would seem like relevant economic evidence on several of the issues it raises. It ignores the evidence that the district court found that the antisteering rule harmed the entry of Discover into credit cards.63 It ignored the district court finding that there was no procompetitive justification for the antisteering rule.64 Even if one adopted the legal rules the Court espoused, this evidence would seem to lead to the conclusion that the plaintiff had established its case at step one. The fourth error relates to antitrust market definition. Market definition is always at most a crude first step in any 58 See Ohio v. Am. Express Co., 138 S. Ct. 2774, 2289–90 (2018). See generally DENNIS W. CARLTON & JEFFREY M. PERLOFF, MODERN INDUSTRIAL ORGANIZATION 414–25 (4th ed. 2005). 59 See CARLTON & PERLOFF, supra note 58, at 421–22. 60 Ohio v. Am. Express Co., 138 S. Ct. 2274, 2304 (2018) (Breyer, J., dissenting). 61 Id. 62 Id. 63 Id. at 2293–94 (Breyer, J., dissenting). 64 Id. at 2294 (Breyer, J., dissenting). No. 1:93] ANTICOMPETITIVE EFFECTS 105 antitrust analysis and many complications regarding its usefulness arise when dealing with two-sided platforms. The Court claimed that although merchants obtain one service from a credit card firm and customers another, because the result is one transaction, the two services must be considered as one antitrust market.65 That approach can obscure the underlying economic forces. The credit card performs two different functions and those two go into making a transaction. Steel and rubber are used to make a golf club, but it would make no sense to claim that steel and rubber are in one market. Complements are not in the same antitrust market, even though the price of one complement can affect the demand for the other.66 An alternative procedure is to define antitrust markets in the usual way, from the point of view of a demander (either the merchant or consumer) and if they interact as they do in this case, to figure out the overall effect of any restriction. That is what would have happened from following the typical three-step procedure. Instead, the Court conflated steps one and two. What will be the effect of the Amex decision? From the viewpoint of economic analysis, which by its nature, would have considered all effects (i.e., on both sides of the market) from the vertical restriction even in the usual three step procedure, I don’t think much will change, at least conceptually. From the viewpoint of burden shifting, a lot could change. I suspect that placing the burden on the plaintiff in the way the Court proposes will make it more difficult for plaintiffs to prevail, even in cases where there is a clear interference in the process of competition with no offsetting justification. The beauty of the common law is that a bad decision can be either overturned or so confined to its unique facts that the effect of bad decisions can be mitigated. I hope that is what happens here. One thing I can predict is that, given the vagueness with which the Court has defined a two-sided market, a firm that is charged with using vertical restrictions in violation of the 65 Id. at 2285–87. 66 See id. at 2295 (Breyer, J., dissenting). 106 COLUMBIA BUSINESS LAW REVIEW [Vol. 2019 antitrust laws will have an incentive to claim that it is operating in a two-sided, not one-sided, market in order to take advantage of the Amex decision, which I suspect will make it harder for plaintiffs to win. This illustrates why having different legal rules for promotional activity depending on whether the market is one-sided or two-sided markets is a mistake.