Friday, September 21, 2007
Posted by John Lopatka
In thinking about the decision of the Court of First Instance (CFI) in the European Microsoft case, it’s easy to confuse the economic theory of the European case with that of the United States DOJ case. After all, both have to do with the relationship between Microsoft’s dominant Windows PC operating system and related software products, both involve conduct by Microsoft that disadvantaged competing software producers, and both seem to have a lot to do with network effects. In the DOJ case, the principal theory was that Microsoft monopolized the PC operating system market by preventing Netscape and Java from developing products that would have rendered all operating systems fungible. The theory of the European case, however, was quite different. I focus here on the “interoperability” part of the case.
The European Commission claimed, and the CFI agreed, that Microsoft refused to provide and authorize the use of sufficient information about the interoperability of Windows PC operating systems and work group server operating systems to permit competing producers of server operating systems to remain in the market. A related claim was that Microsoft withheld needed information about interaction among servers. The behavior alleged constitutes a kind of tying arrangement: users of the Windows operating system on a network of client PCs effectively must use the Microsoft server operating system because alternative server operating systems are inferior. In this story, the Windows PC operating system is the tying product, the work group server operating system is the tied product, and the economic motivation for the tie is to leverage monopoly power in the PC operating system market into the server operating system market.
Network effects have little to do with this story. Indirect network effects are the source of monopoly power in the tying product market, but the story would not change materially if Microsoft instead dominated the PC operating system market because of patents, trade secrets, or ingenuity. This story does not assert that work group server operating systems are themselves subject to indirect (or direct) network effects. For example, it does not depend on the assumption that server applications will be written to the dominant server operating system. Contrast this with the DOJ case, in which the principal theory was that Microsoft monopolized the PC operating system market by artificially maintaining the network effects that protect its monopoly. The European case is based on a conventional theory of monopoly leverage.
The original Chicago School criticism of the leverage theory is that a monopolist of one product has no economic incentive to acquire a monopoly of a complementary product. It is not that tying is logically impossible, therefore, but that it is economically irrational. Later analysis demonstrates that in various settings tying for the purpose of leveraging is rational, and specifically, Dennis Carlton and Michael Waldman offer a sophisticated, dynamic model to explain the DOJ case against Microsoft: the monopolist squashes a superior complementary product in period one to prevent the supplier of that product from developing a competing tying product in period two that would eliminate the tying product monopoly. Bill Page and I argue in our book (pages 156-58) that the model does not fit the facts of Microsoft, but at least it is a coherent theory that could explain tying in the appropriate setting. By contrast, the European case lacks an economic foundation. There is no claim that excluded server operating system producers would eventually develop competing PC operating systems if not hampered by the tying arrangement. There is no claim that some users of server operating systems use them without any client operating systems, a condition which if it existed could explain leveraging under Michael Whinston’s classic model. On the surface, therefore, the monopolist would appear to benefit by facilitating the use of a competitor’s superior complementary product, because that conduct would increase the demand for the tying product, however marginally. And as applied to Microsoft, the evidence is compelling that Microsoft primarily strives to earn its revenue from the licensing of Windows. If the European authorities perceive a threat to economic welfare, it would seem to be that tying prevents rivals from developing superior server operating systems over time, but they offer no rigorous proof of just why that result should be anticipated.
An alternative explanation of Microsoft’s conduct, of course, is that it was not tying at all. Maybe all of the interoperability information that a competitor needed was available, and if users opted for Microsoft’s server operating system, the reason is that Microsoft offered a better product. Certainly a monopolist has an economic incentive to supply a better complementary product than its competitors do, again because that would increase the demand for the monopoly product. On this score, the dispute between the European authorities and Microsoft is not a matter of theory, but of fact – very technical fact – and the two sides seem to be talking past one another. Microsoft contends that the European Commission insists on the disclosure of enough information to allow competitors to clone its product. The CFI disagrees, drawing a critical distinction between specifications and implementations. In the CFI’s view, the Commission requires Microsoft to supply only information about the specifications related to interaction between work group server operating systems on the one hand and Windows PCs and other work group servers on the other. That information is not sufficient to permit rivals to clone the Microsoft server operating system, and indeed the implementation, which depends upon source code that need not be divulged, will vary across vendors. The CFI likens specifications to the vocabulary and syntax of a language, and the implementation in this analogy is the concrete expression of an idea in that language.
Certainly there is a distinction in software design between a specification and an implementation. But the issue is its significance in this context. Suppose a monopolist produces an unpatented toaster that cannot be reverse engineered. Specifications might explain how bread must be exposed to a heating element for a limited period of time, then ejected. The monopolist’s implementation has two slots and a handle on one side. The specifications may enable a competitor to design a toaster with four slots and handles on both sides. The rival’s implementation is different, but the disclosure of the specifications enabled the rival to appropriate the value of the invention – no specifications, no toaster. It is not a sufficient response to Microsoft’s argument, therefore, to say that implementations will differ if only specifications are disclosed. The question is whether disclosure of the specifications will enable competitors to free ride on Microsoft’s investment in developing its server operating system, and if it does, the incentives to innovate decline.
If we assume that Microsoft has an economic incentive to leverage, the difficulty in this case is that the quality of the toaster depends upon – really, is defined by – its interaction with another product. The whole purpose of a work group server operating system is to permit a network of client PCs and other servers to interact efficiently. The appropriate competitive contest is between work group server operating systems based on their intrinsic merit, and Microsoft claims that it gives competitors all of the information needed for competitors to engage in that contest. If customers choose Microsoft’s product, it is because its product is intrinsically superior. But suppose the performance of the toaster depends upon the amount of electricity entering the appliance, and the toaster manufacturer has no toaster monopoly but does have a monopoly over electrical outlets. The manufacturer discloses enough information to allow competitors to design toasters that use only 70% of the power available through an outlet, and as a result, its toaster toasts 30% faster. The fact that specifications permit rivals to make toasters that toast bread does not tell us whether the manufacturer exploited an advantage it held because of its outlet monopoly. On the face of it, we could not tell whether Microsoft succeeds in the server operating system market because of the intrinsic quality of its product – all toasters have access to 100% of the power coming from an outlet, and Microsoft’s toaster just has a better heating element – or because Microsoft withholds critical interoperability information from its competitors. That is a factual issue the resolution of which requires technical expertise. But an appropriate resolution is critical, because compulsory disclosure of information that pertains only to the intrinsic quality of a product will inevitably chill innovation.