Two-Sided Red Herrings: Q&A with Richard Schmalensee
Professor Richard Schmalensee served as the John C Head III Dean of the MIT Sloan School of Management from 1998 through 2007. He was a member of the President’s Council of Economic Advisers from 1989 through 1991 and served for 12 years as Director of the MIT Center for Energy and Environmental Policy Research. Professor Schmalensee is the author or co-author of 11 books and more than 120 published articles. He has served as a consultant to the U.S. Federal Trade Commission, the U.S. Department of Justice, and numerous private corporations.
In DisCo’s Q&A Session with Professor Schmalensee we recap some of the arguments in his article Two-Sided Red Herrings, which was featured in the October 2018 issue of CPI’s Antitrust Chronicle.
DisCo: The Amex decision raised a lot of questions on how to deal with platform businesses that do not follow Amex’s credit-card platform. You mention in your article that there have been two decades of economic literature on two-sided platforms, which have been heavily cited by the District Court, the Appeals Court and the Supreme Court. However, you make the argument that there are five ‘red herrings’ that have been used to marginalize the extensive economics learning on two-sided platforms in antitrust analysis. What is the first red herring, and why does it have no merit?
Schmalensee: The first red herring is the assertion that there is nothing really different about multi-sided platforms. If a platform provides services to members of groups A and B, and each group values the participation of members of the other group, a price increase to A will reduce A’s participation, and this will make the platform less attractive to B and thus reduce B’s participation. In his Amex dissent, Justice Breyer argued that this is just like the interdependent demand for complements like gasoline and tires. An amicus brief in Amex argued that tennis racquets and tennis balls were an example of such an interdependent demand.
Even though there is a formal similarity between the pricing problem facing a seller of complements and the pricing problem faced by some platforms, there are fundamental differences between the two cases. First, complements are usually sold to the same customers, and individual preferences make the demands interdependent, while platform services are sold to members of distinct groups linked by indirect network effects. Second, a filling station can sell gasoline without selling tires, but Amex needs to have both merchants and consumers to be in business at all. Finally, platforms facilitate links between members of the groups they serve, but I can buy gasoline without interacting with anyone who has bought tires.
DisCo: The second red herring is the idea that if any business is two-sided, so are almost all businesses. Is the definition of two-sided platforms too broad, or just ill-defined, and how could that be problematic in antitrust analysis?
Schmalensee: The valid criticism is that the definition of a two-sided platform in the Amex decision did not focus on platforms’ distinctive features and was thus overbroad. If the definition is that the business deals with two or more distinct groups, for instance, then even brick-and-mortar retailers are platforms, since they deal with both suppliers and customers. If all businesses are two-sided platforms, we might as well continue to use only the one-sided analyses that have served antitrust well across a wide range of markets.
Most, if not all economists who have studied them would agree that two-sided platforms are firms that facilitate interactions between members of two different groups, between which the indirect network effects are strong enough to affect firm behavior. There are no debates in the literature about the definition of multi-sided platforms, and the same businesses are identified as multi-sided platforms in many papers. While many businesses, particularly businesses that rely on the Internet, satisfy this definition, most do not. And the economic analysis of two-sided platforms is different from that of ordinary businesses. There is agreement among economists that payment cards need to be modeled differently than drug stores. Jean Tirole got his Nobel Prize in part by showing that the differences matter.
DisCo: You argue that the third red herring is the notion that two-sidedness is irrelevant in mature markets. What is the problem with this red herring?
Schmalensee: The problem is that this notion, advanced as far as I know only in one short paragraph in a single published article, lacks both theoretical and empirical support. It is at least plausible, though unsupported, that when all possible market participants are on one or more platforms, indirect network effects vanish at the industry level.
But even accepting for the sake of argument that this is true, it does not follow that indirect network effects vanish at the firm level, and firm behavior is the focus of antitrust, not industry-level welfare analyses. Suppose, for example, that all consumers in some region go to one or more shopping malls, that all merchants are present in at least one of the malls, and this will continue almost no matter what. Even though by assumption indirect network effects don’t exist at the industry level, if one mall charged for parking and consumers stopped coming, stores would surely leave that mall for its competitors. Firm-level demands would still be interdependent, and indirect network effects would still be present at the firm level.
DisCo: What is the fourth red herring, and why is this red herring so dangerous?
Schmalensee: It is generally true that goods or services that are not reasonably interchangeable cannot be in the same market. The red herring in the Amex case is the assertion that since the services Amex provides to consumers and to merchants fail this test, they can’t be in the same market, so that services to merchants and services to consumers must be treated as separate markets.
This is a very bright red herring. The engine, brakes, and steering wheel on a new car are not reasonably interchangeable, yet nobody argues it is fundamentally wrong to think about a market for new cars. In fact, it’s even worse. For example, GM can sell engines separately from other parts, and they do, but Amex cannot sell merchant services without also dealing with consumers.
This red herring is used to support the argument that it is always appropriate to focus on only one side of a multi-sided platform and to ignore the other side. Sometimes that’s appropriate, as it was in the issue in the Times-Picayune case that reached the Supreme Court in the early 1950s, but whether it is appropriate can’t be decided a priori without looking at both sides.
DisCo: The last red herring you mention is one that has been propelled by Professor Tim Wu, which is that considering two-sidedness explicitly will devastate antitrust. Why does this red herring distract from substance?
Schmalensee: Professor Wu and others argue that trying to deal with the real complexities of two-sided businesses will be such a burden for plaintiffs that antitrust will be dramatically weakened. I don’t think that’s right.
In the first place, one-sided analyses of two-sided platforms can lead to both false positive and false negatives as David Evans and I have demonstrated. Moreover, accounting for business realities, even complex ones, and using the best available economic tools has generally helped courts and enforcement agencies make better decisions over the years. I see no reason to think that this will not be the case as courts and agencies rely on the large and growing literature on multi-sided platforms as they deal with these increasingly important business.
DisCo: What is your own personal view on the Amex decision? How will it affect future antitrust analysis?
Schmalensee: Though I certainly do not agree with everything in the Amex decision, I think it is fundamentally right and that its key lessons will be durable. The first and more novel lesson is that where two-sidedness is important, it must be considered explicitly. The second lesson is not novel at all: the effects of vertical restrictions under the rule of reason must be analyzed at the market level, not at the firm level. In Amex the government focused on merchant services and ignored the consumer side of the platform. They failed to show that but for Amex’s policies market-level payment card transactions volume would be higher or some defensible measure of average per-transaction price would be lower. Since Amex had only around a quarter of payment card transactions volume, even if debit cards are excluded, and given that even in countries in which merchants are free to tell consumers who present an Amex card that they would rather the consumers use another card such steering is rare, it is not surprising that the government did not show that Amex’s anti-steering contracts had an adverse market-level impact.