Network Drama
Winners, Losers, and Microsoft: Competition and Antitrust in High Technology, by Stan J. Liebowitz and Stephen E. Margolis, Oakland, Calif.: The Independent Institute, 288 pages, $29.95
Any truthful history of the intersection between scholarship and public policy must resemble a tragicomedy. Usually, the scholars and their government disciples present the comic spectacle, while the public gets to play out the tragedy (see, for example, the misuse of systems analysis in the Vietnam War), although sometimes there is enough tragedy to go around, as in the fate of geneticists under Stalin. Good ideas, such as school choice, move tortuously through a hostile political process, while bad ideas, such as treating statistical representativeness as the sine qua non of nondiscrimination, are enshrined in law and practice.
Or look at the Justice Department's antitrust suit against Microsoft. The government complains that Microsoft has a monopoly on personal computer operating systems, a monopoly which it has abused to stifle innovation and to maintain its dominant position. The intellectual roots of these arguments are hard to disentangle. In many respects, they seem no different from the sort of predation and bundling complaints to which market leaders such as IBM have been subjected in the past.
But underlying the government's willingness to sue and vilify Microsoft is a belief that its current dominance is in some sense undeserved, that its products have succeeded through a combination of luck and deviousness, not because they gave buyers a better combination of attributes and price. This belief, or at least suspicion, has received considerable nourishment from an economic theory that burst into scholarly prominence in the mid-1980s. According to this theory, markets for high-technology products, where issues of technical compatibility matter a great deal to users, are particularly subject to the risk of choosing the wrong product from a set of mutually incompatible rivals.
Stan J. Liebowitz and Stephen E. Margolis' Winners, Losers, and Microsoft: Competition and Antitrust in High Technology is an all-out attack on the intellectual and empirical basis for this claim, with particular application to Microsoft's role in the software market. In a nutshell, Liebowitz, a professor of managerial economics at the University of Texas at Dallas, and Margolis, a professor of economics at North Carolina State, say that while it is logically possible for markets to pick inferior products, there should be a strong theoretical presumption that it doesn't happen; that often-cited examples of the phenomenon fall apart upon closer examination; and that a detailed study of the PC applications market shows that Microsoft has achieved dominance when and only when its products were rated superior by third-party observers. These conclusions position the authors as the leading academic defenders of Microsoft, a role in which they are almost certainly more effective than the company's paid experts.
The basic idea behind the claim that markets may get it wrong–that buyers may find themselves purchasing products that they themselves perceive as inferior to alternatives–is fairly simple: For some kinds of products, the value of the product to the user increases with the number of other people who own a compatible version. Such "network effects" are easiest to see with a product like the telephone, where no one would buy the service unless there were someone else to talk to. But they may also be important for products like computer operating systems, whose users may want to share files, answer each other's questions, and have access to a wide array of complementary products such as applications software.
So the comparative value of two incompatible products for a particular user depends both on what the user thinks about the intrinsic attributes of the two products compared to their prices and on how big the user expects the network size of each product to be. It is logically possible for those comparisons to point in opposite directions (i.e., the product that looks like it offers better quality for the money is expected to have a smaller network) and for all users to rationally choose the product that is intrinsically inferior but expected to be more popular. The problem is one of coordination among the users; if everyone thinks that a given product is going to be the winner, even if they're all unhappy about it, that product can indeed become the winner.
But why would people expect an intrinsically inferior product-price combination to win? Here is where a second idea, "path dependence," comes in. Path dependence means that history matters; in this case, if the inferior product gets out on the market first, or acquires a sizable user base first, then it may get a leg up on its superior rival that "locks in" its dominance.
The archetypal example, cited by nearly all adherents of the lock-in hypothesis, is the standard typewriter keyboard, whose QWERTY layout is supposedly vastly inferior to alternatives that were developed later. In a famous 1985 article in The American Economic Review, Stanford University economist Paul David argued that the Dvorak alternative keyboard had been proven to be superior to QWERTY, that the gains from switching keyboards were large, but that the ubiquity of the QWERTY format deterred people from learning the Dvorak format. Liebowitz and Margolis' convincing refutation of these claims, first published in 1990 in the Journal of Law and Economics, is reprinted in Winners, Losers, and Microsoft.
Probably the second-most trotted out example of lock-in by path dependence is the videocassette recorder war between the VHS and Betamax formats. "Everybody knows" that Betamax was superior but VHS wound up dominating the market due to some early good luck in the marketplace. Liebowitz and Margolis' analysis of this episode is the polemical equivalent of intercepting a pass and returning it for a touchdown. Most devastating is the simple fact that Sony had Betamax out on the market for two years before VHS even got there, so path dependence seems an unlikely explanation for VHS's success. In addition, VHS cassettes' significantly longer recording time provides a simple explanation for the format's success, while Beta's supposed advantage in picture quality turns out to be another myth, if independent technical experts and Consumer Reports are to be given any credibility.
Liebowitz and Margolis do more than simply explode these just-so stories of market failure. They explain why the mathematical models are too pessimistic about the behavior of markets for network technologies: The models do not allow actors the same degree of freedom they have in the real world.
To make the mathematics tractable, the "strategy spaces" allowed the firms and users in these models are usually fairly restrictive: There is typically no scope for offering guarantees, disseminating messages through various media, price discriminating between new and replacement purchasers, and so on. But a firm with a superior product has a strong incentive to take such steps if they will shift user expectations of market success toward its own standard and away from its rival's. Moreover, the greater the harm to consumers if they coordinate on buying the wrong product, the greater the potential profit for an entrepreneur who can fix the problem. The models' simplifying assumptions about seller behavior, which rule out the possibility of sellers' taking corrective actions, therefore bias the models' predictions toward finding more market failure than should really occur.
The part of the book most likely to stir controversy is its examination of PC software markets. Looking at word processors, spreadsheets, personal finance software, desktop publishing, browsers, and online services, Liebowitz and Margolis compare products' market shares to the quality ratings these products received from computer magazines. In each case, they find that Microsoft's products captured the largest sales only after those products received superior quality ratings; when Microsoft's ratings lagged its rivals' (as in personal finance and online services), so did its market share. (This pattern holds true for other firms' products as well.) Furthermore, prices were lower when Microsoft was dominant than when other firms led the market.
In all product markets, there was a pronounced tendency for the highest-rated product to grab the vast majority of the market very quickly. Liebowitz and Margolis refer to this phenomenon as "serial monopoly" and suggest that the software market is, if anything, even more frictionless in responding to quality differences than other markets, because a firm with a better offering can almost instantly ramp up production to satisfy the whole market (a feature they call "instant scalability"). Firms, in this view, are held back only when they can't come up with a competitive product. Hence, VisiCalc did a poor job of adapting its product to the DOS operating system, allowing Lotus 1-2-3 to dominate; Lotus lagged in adjusting its product to the Macintosh operating system and to Windows, allowing Microsoft Excel to take over.
There is an alternative interpretation of this pattern, however. Perhaps each new operating system presents a window of opportunity for new products, with the first firm to come up with a suitably adapted piece of software able to seize an advantage it never relinquishes. In this view, the leader's product is never challenged again, in either market share or the quality ratings of magazines, because the lagging firms recognize that a locked-in market will not switch programs even if they invest in product improvements. Microsoft critics would go on to argue that the company deliberately deceived its rivals into working on products adapted to OS/2, its joint operating system project with IBM, before abandoning OS/2 and instead pushing Windows, so that its "inside information" enabled it to get to market first.
More basic than this debate between the "frictionless serial monopoly" and "lock-in" interpretations is the question of whether the relevant markets should be expected to display network effects in the first place. Most word processors, for example, can read files created by other programs. The cost of learning to use a new word processor is not that great, and most users are familiar enough with these products not to need that much help. Given these conditions, it is not clear that this market presents a very good test of network lock-in theory.
In other words, if Liebowitz and Margolis' interpretation of the data is correct, does that mean that network theory is not applicable to these markets, or does it mean that the theory is applicable, but its unrealistic restrictions on entrepreneurial behavior cause it to make false predictions? If the latter, what business strategy not included in the network models leads to the observed result? For polemical purposes–get your stinking Justice Department off my spreadsheets!–the answers might not matter, but from the standpoint of understanding high-tech markets they seem crucial. The same concern applies to the Betamax-VHS battle: Was Sony's failure to lock in the market a sign that network effects are unimportant in the case of VCRs, or proof that they can be overcome by entrepreneurial creativity? Does Liebowitz tell his managerial economics students that network effects are mostly a myth and should be ignored, or does he coach them on how to overcome them if they have a superior product starting from an inferior market position?
An important lesson to be learned from episodes of standards competition is that the outcome depends not simply on management's ability to create superior products but also on its skill at planning production capacity, setting prices, choosing marketing techniques, and so on. A strong argument can be made, for instance, that the Macintosh operating system did not fail to become the dominant standard because of its intrinsic technical drawbacks, as Liebowitz and Margolis claim in a surprisingly casual treatment. Rather, Apple's management team didn't understand the importance of network effects and therefore overpriced the product, underpromoted it to business users, and failed to expand hardware production capacity sufficiently.
Jim Carlton's meticulously reported 1997 book Apple: The Inside Story of Intrigue, Egomania, and Business Blunders includes a 1985 memo to Apple CEO John Sculley from none other than Bill Gates, at that time dependent on selling Macintosh applications software for a large chunk of his profits. The memo recommends that Apple allow other manufacturers to produce the Mac, so that Apple would "have the independent support required to gain momentum and establish a standard." Gates' advice was rejected, with the results for Apple that we see today: It has about 3 percent of the worldwide PC market. A world where network effects are pervasive may be a world of men, not laws, as far as technology choices are concerned.
One of the bizarre aspects of Liebowitz and Margolis' book's reception was a September 17 New York Times report that the publisher, The Independent Institute, had received more than $200,000 from Microsoft without trumpeting the fact. Even though the Times reported that the authors had no idea of their publisher's funding arrangements, the article strained to leave the impression that the "secret" funding somehow called into question the validity of their research. REASON's readers will find the suggestion particularly amusing, since the authors presented their views on the economic issues relevant to the Microsoft case in these pages several years ago (see "Typing Errors," June 1996). In fact, a little research by the Times would have shown that the authors first published their views back in 1990.
If the book's publisher is guilty of any-thing, it is failing to provide better editorial guidance. Three of the first five chapters, which are primarily theoretical, are put together from a number of the authors' previous writings and don't relate very well to the empirical analysis of software markets. And rather than present a coherent point of view about the importance of network effects and the appropriate assumptions to make in modeling different sorts of markets, Liebowitz and Margolis engage in what debaters call "straight refutation," meaning a strategy of objecting to every assertion made by an opponent without regard to the internal consistency of one's own position.
Thus, the book repeatedly criticizes the lock-in models for assuming that consumers behave myopically in choosing technologies to buy, picking the current leader rather than anticipating that everyone might be better off with a newer, currently less popular alternative. But many of these models feature consumers who are not only forward-looking but correctly predict the market equilibrium; it is Liebowitz and Margolis whose formal models assume that consumers are completely myopic, looking only at current market share in deciding which network to join. Similarly, the book equivocates about whether software markets constitute natural monopolies, questioning the idea in the first chapter and proclaiming it in the last. The theoretical sections make a point of distinguishing between "fixed" and "flexible" standards, with the latter evolving and not being subject to lock-in, but the authors never apply these ideas in their empirical analysis of software, even when those programs feature quite a bit of backward compatibility between new and old versions.
These flaws aside, Winners, Losers, and Microsoft serves a useful purpose not merely in throwing cold water on the idea that markets routinely pick inferior technologies and in providing the first thorough summary of what really happens in software markets, but also in forcing the reader to confront the difficulties of applying scholarship to public policy. The authors make a compelling Hayekian argument that if scholarly analysis of voluntarily adopted institutions and practices seems to reveal opportunities to improve matters through regulation, the odds are that something important was left out of the analysis. Too bad that the antitrust tragicomedy seems to be proceeding heedless of this wisdom. The show, it seems, must go on.
