Testimony of Richard Schmalensee(1)
On Antitrust Issues Related to Networks
before the Federal Trade Commission
December 1, 1995
It sometimes appears that in recent years network industries have given rise to a great variety of novel and difficult antitrust issues. In fact, however, most of the apparent novelty is illusory and stems in large measure from over-broad use of the term “network” to describe a range of economically distinct situations. After attempting to distinguish these situations, I discuss briefly some of the important antitrust issues raised by single-sponsor and multiple-sponsor networks. I conclude that, while network industries do give rise to difficult antitrust issues, new antitrust rules should not now be developed to deal with them -- both because these issues are already dealt with by existing rules and because the economics literature as yet provides no new, useful guidance for dealing with them in the network context.
What is a Network?
At the most general level, a network is a set of nodes connected, directly or indirectly, by a set of links. In the case of the telephone network, for instance, it is natural to think of the nodes as business and residential customers and of the links as the cables, wires, and other transmission facilities that connect them. On the other hand, an engineer might find it more useful to think of switches, not customers, as the primary nodes in this network, and he or she would certainly be aware that the telephone network is bound together by a set of standards and protocols as much as by copper and fiber.
Many other structures can be described in terms of nodes and links, and most of them seem to be regularly labeled networks. Airline route systems, for instance, in which airports are connected by scheduled flights, are often called networks, as are airlines’ computerized reservation systems. So are ATM and bank credit card systems, which are linked by contracts, use of a common brand name, and communications facilities and standards. The set of suppliers that are linked to a particular firm by ongoing business relationships is often described as that firm’s “network of suppliers.” The Internet plainly consists of nodes and links, and information providers such as America Online also provide links connecting their subscribers, among other services. New MBAs are regularly told to build and maintain “networks” of potentially useful friends and acquaintances. Finally, we sometimes hear that “network economies” effectively link together otherwise unconnected users of a common computer operating system or applications program. I hope it is clear that these “networks” and the many other situations that could be so labeled differ along a number of important dimensions. Indeed, their differences are often a good deal more important that their common features.
In light of the important differences among “networks,” why is there now so much interest in network-related antitrust issues? One obvious reason is that related advances in computation and communication have made possible much cheaper and faster information-related links than could have been imagined a few decades ago. As a result, information flows on some pre-existing networks, such as bank credit cards, have increased, and new networks, such as Prodigy, have been established. Thus networks with electronic links are more important and more visible than ever before -- and they are accordingly involved in more antitrust cases.
A second reason for increased antitrust interest in networks is that beginning in the mid-1980s, a literature on networks emerged in the economics journals.(2) The defining characteristic of “networks” in that literature is the presence of “network externalities.” Roughly, these are factors that cause the value of the network to increase more than proportionally as its size increases.(3) Thus when network externalities are present, the total value of one network is greater than the total value of two competing networks, and there is an element of increasing returns that tends in principle, all else equal, to drive the system toward monopoly. The classic example is the telephone network: the value to me of having telephone service rises as more people with whom I would like to talk join the network, so that doubling the number of subscribers is likely to more than double the total value of the network to them. Similarly, the value to me of using Microsoft’s Windows 95 operating system is higher the more others use it, both because I will have more opportunities for sharing information and because more applications software will be available for Windows 95 the larger its user base. Network externalities thus operate like economies of scale, except that network externalities arise on the demand side of the market and are conceptually distinct from any economies of scale, scope, or learning on the supply side that might affect the cost of production.
I believe that a good deal of confusion has flowed from a common tendency to assume that the economic literature on networks applies to every situation that can be described in terms of nodes and links. In some of these situations, which I refer to as broad-sense networks in what follows, there is simply no good reason to suppose that network economies exist or are of any importance. For the sake of clarity I use the label narrow-sense networks to refer to situations in which such economies are present, even if they are not so important as to make monopoly either inevitable or socially desirable.
The implicit assumption that all broad-sense networks are also narrow-sense networks, which I believe is often made in this context, tends to carry one a good deal of the way to the conclusion that the industry involved is a natural monopoly. It may be useful for some purposes, for instance, to label the set of users of Intuit’s Quicken personal finance software package a network, but it is unjustified and potentially dangerous to infer from that label alone that network externalities drive the market for such packages toward monopoly. (I believe that conclusion is also probably wrong as a factual matter.) In the analysis of single-sponsor networks the apparent result of assuming that network industries generally have important demand-side economies of scale is winner-bashing, while in multiple-sponsor networks the problem appears to be excessive application of the essential facility doctrine.
Single-Sponsor Networks
Perhaps the most interesting cases involving single-sponsor networks turn on actual or potential competition between broad-sense networks, which often amounts to competition between standards. These cases involve antitrust analysis of the difficult terrain dividing vigorous competition from monopolization. Competition between standards was at issue in the Commission’s and Division’s investigations of Microsoft’s operating system licensing practices. In this case the networks involved were the sets of users of competing operating system software for microcomputers. I believe there was evidence of some network externalities in this case, though the persistent survival of the Apple operating system and continuing investments in operating systems by Microsoft’s rivals would seem to indicate the absence of a natural monopoly.
Though some plainly feel otherwise, under the conditions just described there is simply no consistent economic argument for applying unusually strict standards to evaluate the conduct of the leading vendor. This would be true even if network economies were so strong as to point toward natural monopoly. Those who feel otherwise usually observe that the economic literature suggests that in competition between narrow-sense networks, the less efficient network can prevail and obtain a monopoly as a consequence of small first-mover advantages -- or small advantages gained through aggressive or restrictive conduct.(4) Since small offenses can have large adverse consequences, it is argued, very strict standards should be used evaluate leading firm conduct in narrow-sense network industries. To guard against the loss of kingdoms, we have to watch the nails in horseshoes very carefully.
In evaluating this argument, it is important to recognize that the theoretical possibility of large effects small causes is not confined only to narrow-sense network industries. Small advantages can also have large consequences in industries marked by important scale or learning economies. Thus the applicable theory suggests that it is possible that because auto production is characterized by such economies, Henry Ford’s Model T dominated the US auto market as a result of historical accident or (purely hypothetical) small antitrust offenses committed by Ford. A better product may have been available but unable to overcome Ford’s head start or other advantages. Of course, any attempt to handicap Ford so as to give his competitors a “level playing field” would have had real costs and only hypothetical benefits. Even if Ford had committed antitrust violations, for instance, the Model T may still have been far and away the best product of its kind available.
It simply does not follow from theoretical models implying that the wrong standard can win that the winning standard in any particular market, even if its sponsor has been unusually aggressive, is generally wrong. All else equal, even if scale economies of various sorts give random events greater influence on market outcomes, it would be surprising indeed if the better standard didn’t win more often than not. More generally, it is important to recognize that the existing theory of narrow-sense networks is about what can happen, not about what must happen. It simply provides no useful, general rules for antitrust policy. In particular, it does not provide a rationale for a policy of bashing winners who have engaged in conduct that would be acceptable in non-network industries, particularly when it is considered that such a policy would tend to discourage vigorous competition across the board.
Moreover, it is unclear how seriously to take models in which market competition yields inefficient outcomes because of demand-side or supply-side economies of scale. Most inefficiencies of these sorts correspond exactly to unexploited profit opportunities, and the strategies available to actors in theoretical models to exploit those opportunities are typically much less rich than the strategies available to real firms. It is thus perhaps not surprising that there appear to be no clear-cut examples of inefficient outcomes of competition between standards.(5) In addition, it is important to recognize that a monopoly based on network externalities is much more fragile than one based on scale economies in production or on learning economies. Over the last decade numerous software products that had dominated their categories, including WordStar, VisiCalc, and dBase, saw their sales disappear rapidly in the face of innovative competition. Being popular mainly because you are popular is a precarious business.
Two other issues deserve brief mention in this context. First, while the economic literature makes clear that single-firm decisions regarding standards and compatibility may have anticompetitive effects, it does not provide operational rules that can reliably predict the likelihood of such effects. In the absence of such rules, litigation is unlikely to be either efficient or effective at evaluating the social or private merits of particular technical decisions. Since decisions regarding standards and compatibility are typically made in the process of innovation, the social costs of regulating those decisions badly is potentially very high. Accordingly, in the absence of reliable rules, I would not advocate stricter antitrust scrutiny of single-firm standard-setting.
A similar situation arises in connection with mergers between sponsors of complementary standards -- involving, for instance, components of a system. These can be usefully viewed as vertical mergers.(6) Recent work in economics has shown that under some conditions, vertical mergers can have anticompetitive effects.(7) It is thus difficult to maintain the extreme Chicago view that such mergers are always benign. But the existing theory deals with special cases and much of it rests on controversial assumptions; it does not have any real empirical support or offer any workable rules for predicting what is likely to happen as a consequence of any particular merger. Given that there are costs of challenging desirable vertical mergers and that the relevant theory seems to hold out the possibility of anticompetitive effects only in rather extreme cases, I would advocate at most a slight tightening of vertical merger policy.(8)
Multiple-Sponsor Networks
The most interesting issues raised by networks involving more than one sponsor involve some measure of cooperation among actual or potential competitors, and I limit my attention to such situations. The most familiar examples of multiple-sponsor networks are joint ventures or functionally equivalent organizations such as ATM networks or bank credit card systems. One would also include under this heading the public switched telecommunications network, in light of the importance of interconnection and common standards, as well as other sets of sellers engaged in (explicit or implicit) collective standard-setting. The sellers of fax machines form an interesting set of this sort.
As above, there is no guarantee that any particular broad-sense multiple-sponsor network is characterized by important network economies. Even in the case of narrow-sense networks, there is no guarantee that network economies are so important as to make monopoly the likely or socially preferred outcome. Moreover, even if a network is a natural monopoly, its sponsors often compete vigorously with each other along a number of dimensions. Competition can thus occur both between and within multiple-sponsor networks. Multiple-sponsor networks raise two basic sorts of antitrust issues: those relating to operations and those relating to membership.
In an economic evaluation of the operations of a multiple-sponsor network, one must weigh the benefits of various sorts of closer cooperation between rivals against the costs of diminished competition. One must also consider the general benefits and costs of encouraging partial integration via joint ventures, as opposed to permitting only complete integration via merger. These are familiar, though difficult, issues; they are basic to the antitrust analysis of joint ventures and other collective action (including standard-setting) and have nothing in particular to do with networks.(9) The economic literature does not indicate that broad-sense or narrow-sense networks are easier to analyze in this context than other joint ventures, and it provides no useful rules specific to network contexts.
Judgments about these likely economic effects of a joint venture’s operating policies should clearly be informed by the facts of the industry being considered, and, as many observers have noted, the level and nature of innovative activity are likely to be important facts in this setting. Perhaps even more basically, as many observers have noted, if a multiple-sponsor network (or any other joint venture) produces a product that the individual sponsors could not have produced by themselves, then (whether or not new technology is employed) the venture involved has social benefits.
Economic evaluation of a joint venture’s membership rules clearly must consider potential threats to competition from both exclusion and inclusion decisions. Exclusion of a set of firms from a joint venture may reduce competition by either excluding those firms completely from a relevant market or rendering them ineffective competitors. (Of course, excluding a few firms without special advantages from an already competitive market cannot have this effect.) On the other hand, inclusion of a large fraction of actual or potential competitors may reduce or eliminate competition from other networks, either by robbing them of scale economies (if the included firms leave those other networks) or by affecting a partial merger with them (if the included firms remain sponsors of those other networks). Moreover, in a world of imperfect monitoring, collective decision-making by most or all sellers in an industry must be a cause of unease, at least. Finally, requiring members of a joint venture that has created tangible or intangible property to admit non-members immediately raises the question of the appropriate price of membership. This tends to require the sort of detailed economic regulation to which courts are traditionally and properly averse as well as adding a level of uncertainty (at least) likely broadly to reduce incentives to create property through joint ventures.
None of the general points raised in the preceding paragraph depend in any essential way on whether the joint venture being analyzed is a broad-sense network, a narrow-sense network, or not in any important way a network at all. Nor should the analysis in any particular case depend importantly on whether or not a joint venture is labeled a network. As a general matter, I believe that advances in communications and computation have increased the scope for efficiency-enhancing joint ventures, but whether any particular joint venture in fact produces efficiencies must be determined primarily on the basis of the facts of the case. It happens that the MountainWest case that has drawn attention to these membership issues involved VISA, a broad-sense network in which, at least up to a certain scale, network externalities are likely to have been important, but this is more or less coincidental.(10)
On balance, given the considerations discussed above, I believe that an efficiency-enhancing joint venture should be required to admit a new member only when it can be shown that doing so is essential for effective competition in some market.(11) That is, I believe that the balance of policy considerations points toward a presumption of legality for refusal to broaden a joint venture’s scope -- just as a refusal to merge is presumptively legal.
Other observers would apparently condemn any decision to exclude an applicant for membership in a joint venture if exclusion would reduce the applicant’s effectiveness as a competitor, unless that exclusion (or, more generally, the rule that compels it) could be shown to be reasonably necessary to achieve efficiencies.(12) This alternative rule, which would place a heavy burden of proof on joint ventures to justify refusals to admit new members, seems to rest on two presumptions. The first is that refusal to admit a new member to a joint venture generally reduces competition on balance. In light of the danger of reducing competition between networks when competitors are admitted, I do not believe that this presumption is generally justified. Moreover, it is certainly incorrect if competition is already vigorous and the excluded firm or firms would bring no special advantages to the market. The second necessary presumption is that there is generally no net cost to broadening the scope of a joint venture. In light of the competitive risks of over-broad joint ventures, this presumption seems justifiable only if the venture is a natural monopoly -- and thus an essential facility. Even though, as I noted above, some authors tend to assume implicitly that all networks are natural monopolies, it seems clear that most in fact are not. Those that are should generally be required to admit new members on “reasonable” terms.
Conclusions
The negative tone of much of this testimony may suggest that I believe that network industries should receive less vigorous antitrust scrutiny than other industries. To be clear, I do not hold that view. I do believe that industries in which innovation is an important form of rivalry should be viewed through a somewhat different set of presumptions than technologically stagnant industries, but not all innovative industries involve networks, and not all networks occur in innovative industries. I have tried to show that the economic literature on network externalities does not apply to all structures we commonly call networks, so that policy toward such structures should not be based on an explicit or implicit assumption of natural monopoly. I have also argued that existing antitrust rules deal adequately (though inevitably imperfectly) with the main difficult antitrust issues that arise in network industries and that the economic literature does not provide workable, theoretically superior rules for such industries.
ENDNOTES:
(1) Gordon Y Billard Professor of Economics and Management, Massachusetts Institute of Technology and Special Consultant, National Economic Research Associates. I have been a consultant to VISA U.S.A. and Microsoft on matters involving issues touched on in this testimony, but this testimony has neither been supported nor reviewed, directly or indirectly, by either organization. I alone am responsible for the views expressed herein.
(2) A good general introduction to this literature and some ongoing controversies within it is provided by the three essays in “Symposium on Network Externalities,” Journal of Economic Perspectives, 8 (Spring 1984): 93-150; see also W.E. Cohen, “Competition and Foreclosure in the Context of Installed Base and Compatibility Effects,” mimeo, Federal Trade Commission, November 16, 1995.
(3) S.J. Liebowitz and S.E. Margolis, “Network Externalities: An Uncommon Tragedy,” Journal of Economic Perspectives, 8 (Spring 1994): 133-150 argue that some effects that have been labeled “network externalities” are pecuniary rather than technical and thus do not tend to produce inefficiency. Perusal of the literature suggests that the first part of this argument is largely correct, but the second part does not follow. In the presence of increasing returns, fixed costs, imperfect competition, and other departures from classical conditions, pecuniary externalities can have (positive or negative) efficiency effects.
(4) This is the apparent thrust of the argument in the famous “Reback Brief”: Memorandum of Amici Curiae In Opposition to Proposed Final Judgment, submitted to the Court on January 10, 1994 in U.S. v. Microsoft Corporation, Civil Action No. 94-1564 (SS), (D. Columbia 1994).
(5) See S.J. Liebowitz and S.E. Margolis, “The Fable of the Keys,” Journal of Law and Economics, 33 (April 1990): 1-26, and “Network Externality: An Uncommon Tragedy,” Journal of Economic Perspectives, 8 (Spring 1994), 133-150.
(6) Cohen, supra note 2 frames this issue nicely in the present context.
(7) This is a large literature, a good bit of which is cited by Cohen, supra note 2. Two important contributions are J.A. Ordover, G.A. Saloner, and S.C. Salop, “Equilibrium Vertical Foreclosure,” American Economic Review, 80 (March 1990): 127-142 and O. Hart and J. Tirole, “Vertical Integration and Market Foreclosure,” Brookings Papers on Economic Activity, Microeconomics, (1990): 205-286.
(8) In particular, I believe that the proposal of M.H. Riordan and S.C. Salop, “Evaluating Vertical Mergers: A Post-Chicago Approach,” Antitrust Law Journal, 63 (1995): 513, tilts too strongly against vertical mergers and imposes burdens of proof on defendants that they would have extreme difficulty using available economic tools to meet in benign cases. See also D. Reiffen and M. Vita, “Comment: Is There New Thinking on Vertical Mergers” Antitrust Law Journal, 63 (Spring 1995): 917-941 and M.H. Riordan and S.C. Salop, “Evaluating Vertical Mergers: Reply to Reiffen and Vita Comment,” Antitrust Law Journal, 63 (Spring 1995): 943-950.
(9) See, for instance, the discussion in R. Schmalensee, “Agreements Between Competitors,” in T.M. Jorde and D.J. Teece, eds., Antitrust, Innovation, and Competitiveness, Oxford: Oxford University Press, 1992. A word about collective (as opposed to single-firm) standard-setting may be in order. Because, as I noted in the text, I do not believe that litigation is likely to be an effective or efficient method for judging the social costs and benefits of particular technical standards, I would be inclined to allow procedural propriety (perhaps with some variant of the current ANSI process as a minimum standard) to be a complete defense. Recognizing that procedural niceties can easily be overlooked when pressures to innovate are intense, however, I would also permit use of something like the “substantive reasonable basis” standard of J. Anton and D. Yao, “Standard-Setting Consortia, Antitrust, and High-Technology Industries,” Antitrust Law Journal, 64 (Fall 1995): 247-265 as a defense.
(10) SCFC ILC, Inc. v. VISA U.S.A., Inc., 819 F. Supp. 956 (D. Utah 1993), 36 F.3d 958 (10th Cir. 1994), cert. denied, 115 S.Ct. 2600 (1995).
(11) Since no such showing was made in the MountainWest case, I believe that case was correctly decided for defendant VISA. See D.S. Evans and R. Schmalensee, “Economic Aspects of Payment Card Systems and Antitrust Policy Toward Joint Ventures,” Antitrust Law Journal, 63 (Spring 1995): 861-901. See also D.I Baker, “Compulsory Access to Network Joint Ventures Under the Sherman Act: Rules or Roulette?” Utah Law Review, 1993: 999.
(12) See D.W. Carlton and A.S. Frankel, “The Antitrust Economics of Credit Card Networks,” Antitrust Law Journal, 63 (1995): 643 and “The Antitrust Economics of Credit Card Networks: Reply to Evans and Schmalensee,” Antitrust Law Journal, 63 (Spring 1995): 903-915. See also D.W. Carlton and S.C. Salop, “You Keep On Knocking But You Can’t Come In: Evaluating Restrictions on Access to Input Joint Ventures,” Working Paper 111, Center for the Study of the Economy and the State, University of Chicago, April 1995 and H. Hovenkamp, “Exclusive Joint Ventures and Antitrust Policy,” Columbia Business Law Review, (1995):1.