DEPARTMENT OF JUSTICE
Thank you. It's a pleasure to be here today to speak to you about how we at the Department of Justice apply antitrust principles to so- called "high tech" markets.
This conference is certainly timely. We should focus, as this conference does, on the specifics of how antitrust policy should be applied to high-tech markets. In my experience I don't think its that unusual there is a disconnect in the worldviews of industry and the antitrust enforcement agencies on how high-tech markets (in particular, computer markets) function. I experienced this disconnect when I was out here in Silicon Valley in private practice on some software deals, and saw the utterly perplexed looks on the faces of business people from CEO's to marketing and product managers to programmers when they were confronted with the antitrust world, with the attempts of the enforcement agencies to analyze the likely competitive effects of their merger or other business arrangement. They thought the government was nuts, that the staff lawyers and economists had no idea how their markets worked. The government's theories on pricing, innovation and entry all seemed alien to them.
At the same time I knew (though my clients sometimes doubted) that the antitrust laws would not impede my clients' ability to compete. The principles I learned and that I knew the government was applying were meant to promote innovation, competition the free market. And I thought (I now know) that the enforcement agencies were, in good faith, grappling with the problems of determining whether market power could be exercised in these markets, notwithstanding the conventional wisdom.
This conventional wisdom was (perhaps it still is) that there couldn't be a competitive problem in Silicon Valley. The technology changes too fast. How could a firm act anticompetitively when its dominant product could be superseded in a heart beat by an upstart? Human capital is the key asset. There are a lot of smart, mobile people with nothing better to do than write programs. You don't even need a lot of them. Most of the time, small groups outproduce large development teams anyway, since big groups often don't work as well together as small ones do. And you can't keep track of them all; the big firm has no easy, reliable way to monitor the activities of all the people who pose a competitive threat. How could there be a competitive problem?
But despite this conventional wisdom, cases started popping up. Borland/Ashton Tate. Chipsoft/Mecca. Adobe/Aldus. Silicon Graphics/Wavefront/Alias. Microsoft/Intuit. And of course, the Microsoft con-sent decree. Were huge mistakes being made? I think it's more likely that we have a case of cognitive dissonance, as the conventional wisdom gives way to close analysis as it usually does, even in Washington.
In preparing my remarks, I kept thinking about the movie "Butch Cassidy and the Sundance Kid," which is basically about the end of the wild west. Throughout the movie, the Hole-in-the-Wall gang finds itself increasingly (and increasingly less successfully) doing battle with the forces of civilization. In an early scene, Butch and a mutinous member of the gang have a knife fight. As Butch stalls for time, trying to set up rules, the other guy perplexed says: "Rules? In a knife fight? There are no rules in a knife fight." These were the glory days of the frontier. Of course at the end of the movie, run out of America by the railroads, our heros meet their doom at the hands of about 2,000 soldiers of the Bolivian army. Civilization triumphs.
The reason this movie comes to mind isn't because civilization triumphs. It comes to mind because we like Butch and Sundance, just as we like those here in the Valley and elsewhere who have the vision and entrepreneurial spirit that drives innovation. At the same time, what do we do about knife fights?
As the theme of this conference makes plain, people know that high-tech markets, even where technology is rapidly evolving, are not automatically immune from antitrust scrutiny. Despite this disconnect in world view, the reality is that rapidly changing technology markets are not immune from anticompetitive behavior or market structures. For better or worse, the issue has become how to deal with the "civilizing" forces of antitrust enforcement. How do we establish and apply antitrust policy so that the world won't be a knife fight, and, at the same time, the antitrust laws won't be enforced with the sensitivity and subtlety of the Bolivian army.
I hope I have conveyed to you that it is the Division's policy to promote innovation and to permit innovators to appropriate returns for their efforts, while at the same time preventing undue restrictions on the opportunities of others to innovate. These are the principles underlying the 1995 Intellectual Property Guidelines.
Our basic approach to evaluating mergers in high technology industries follows the 1992 Horizontal Merger Guidelines. While the Merger Guidelines do not specifically address issues of intellectual property or innovation, they are flexible enough to permit us to investigate the impact of technology on the competitive effects of mergers.
So, beyond taking refuge in the Guidelines, what can I say specifically about antitrust and high-tech industries? I don't have enough time to set out a complete, systematic discussion of the topic. Actually, I don't think such a discussion is possible at this point: The issues are too complicated and the answers are too tentative. Instead, allow me to make a couple of observations on what informs the analysis.
In many cases, high-tech, rapidly changing markets involve networks. Examples of networks include telephone, railroad and electricity networks, as well as networks of compatible technology (say a computer platform or software product or a video tape format). It's important to keep in mind some key points about network industries that will inform our analysis. First, networks tend to involve demand-side scale economies, which influence consumer expectations and choices: Large networks are more attractive to consumers, and so consumers are more likely to buy into a network they expect will succeed. Consumers are reluctant to buy a technology they believe will fail. Thus, attractive networks tend to get larger. Second, compatibility with other products in the network affects that product's attractiveness to consumers. Incompatibility can constitute a significant entry barrier. Third, cooperation among producers is necessary to facilitate the ability of consumers to attach to the network. For example, vertical cooperation between hardware and software producers is essential to the computer industry. Even cooperation between direct rivals is often beneficial, say, where standards are needed to achieve compatibility. These points all suggests that rapid technological progress does not necessarily equate to low entry barriers in circumstances where users find it costly to switch to new brands that are incompatible in some way with established technology.
Thus, the modalities of competition in network industries involve practices that may seem problematic from a traditional antitrust perspective. Our job is to see through the form and analyze the competitive effect. But, we want to take care that innovation is not hampered and that prices do not go up.
In merger analysis, the role of networks most clearly manifests itself in respect to entry issues. In particular, the existence of an installed base, which is typical of software markets, makes entry more difficult when the new firm is trying to compete against an incumbent that already has a substantial installed base. This is in part because customer switching costs can be high. For example, switching from one data base program to another may be unfeasible if the consumer's historical data cannot be transferred in a suitable format from the old program to the new program. Where customer switching costs are high, it will take more than writing lines of code to be successful.
The need for an installed base also affects the decisions of possible new entrants. For example, an independent software programmer is relatively less likely to focus his or her efforts on developing a program that offers only incremental advances to existing products. The upfront fixed costs of writing a new software program can be relatively high. Only firms with investment in the product (that is, the incumbent firms) will want to devote the resources to such a project. Moreover, it is not necessarily easier or cheaper to write an application to compete with an existing program than it is to write a new application that will face little or no competition from an incumbent. If an installed base is hard to amass, the return necessary to cover the fixed cost will be much more difficult to achieve if the developer chooses to go head to head with an existing product. His or her efforts would be better spent elsewhere. That being so, the incentives for new entry likely to be created by anticompetitive behavior of incumbents may well not be sufficient to spur de novo entry, even where ample human capital is available. In short, the incentive to enter created by the merger must be measured against the benefits to be obtained by actually entering before an entry argument will be persuasive.
Another area where networks affect an entry analysis is in evaluating the probability that firms would reposition their products in response to a merger of two differentiated products, such as software applications. Products are differentiated when their respective attributes are perceived by consumers to differ. Cement is not often a differentiated product; as I said earlier, computer programs usually are. In a differentiated product market, a product competes with some products in the market to a greater degree than with others. To illustrate: If the price of product A goes up, and we see customers switching in greater number to product B than to product C, then product A is a closer competitor to product B than to product C. Mergers of close competitors in a differentiated product market can give rise to a unilateral anticompetitive effect. In other words, by eliminating (or rather, internalizing in one firm via merger) the competition between close substitutes, the merged firm may find it profitable unilaterally to raise price, since lost sales in large part will go to the other product now under common ownership.
One way such a unilateral price increase would be defeated is through the firms in the market repositioning (that is, changing the attributes) of their products so that the merged firm will face new close substitutes. The newly repositioned products would capture a larger portion of lost sales following a price increase of one or both of the merged products.
In a network industry, repositioning can be difficult. Intellectual capital is not likely to be devoted to a me-to or slightly different version of an existing product. Also, consumer expectations and installed-base compatibility issues weigh against easy repositioning by a more distant competitor. Thus, the reward for repositioning may well not be great enough to justify the cost and effort needed to reposition the product.
In sum, in network industries such as software, entry is not always easy, even if some indicia of low entry barriers are present (for example, the code can be easily written and the time and upfront investment appear small). Careful evaluation of the impact of an installed base must be done before it can be concluded that a merger poses no problem on entry grounds.
On the other hand, some features of network industries suggest that rivalry between incumbents may not be the sole determinant of innovation or even pricing. For example, incumbents cannot innovate as freely because they must maintain a link with installed base. Changing a program too dramatically will alienate current users. Firms with a large installed base may not be the source of the "next generation" product, since they would be reluctant to damage their lucrative installed base in the existing technology. They are thus more susceptible to leap-frog innovation.
Also, firms with installed bases do not necessarily depend on rivalry as the sole reason to continue to innovate or keep prices low. Software programs are like durable goods. Once they're installed, they can reside on a computer indefinitely, subject to incompatibilities caused by hardware and other system developments. Software companies make money from the installed base by selling upgrades. To get its customers to buy the upgrade, the software firm must convince the customer that the upgrade has something useful beyond what's in the original program. They don't get used up. Without innovating and, in effect, competing against its own prior version, this lucrative income stream would dry-up; each decision not to upgrade is a lost sale (and a profitable one at that). Thus, regardless of the existence of other competitors, a software company has an incentive to continue to innovate and to "compete" against itself in selling upgrades. Moreover, a firm's general reputation for delivering value meaningful upgrades at good prices can be an important factor in persuading a new customer to buy the program in the first instance. This is not to say that rivalry is secondary. Indeed, competitive upgrades, side-grades other schemes evidence fierce competition for customers from a rival's installed base. My point is that other dimensions are also relevant to the analysis.
The analytical points I've made this morning I believe can be found in the analysis of the transactions I mentioned above. For example, the Division's challenge of the Microsoft/Intuit transaction rested substantially on the network aspects of the market. Intuit and Microsoft were the top two competitors in personal finance software, with market shares of 70 percent and 20 percent respectively. There was a fix proposed, but the Division concluded that it would not have solved the competitive problems. In any event, beyond the basic horizontal issues, the Division was concerned that the merger would threaten competition in the emerging business of home banking. Microsoft and Intuit were the two most likely competitors to develop software that could serve as the "front-end" of the network. Microsoft and Intuit's front-end products would connect through a bridge (for example, an on line service) to the "back-end" banks and banking services.
The Division believed the deal ran the risk of creating a bottleneck at the front-end, because, owing to its installed base, Intuit had a tremendous advantage in providing the personal financial software part of the system. At the same time, Microsoft had the next best personal finance program, as well as a monopoly position in the operating system, plus plans for Microsoft Network. The competitive problems posed by the deal were confirmed by the CEO of Intuit in a memo to his board analyzing the transaction, where he claimed the deal would leave financial institutions with one clear option, eliminate a bloody share war and enrich the terms of trade. No antitrust disconnect here.
The Division's enforcement actions in the BT/MCI and Sprint/FT/DT transactions also rest on the impact of networks. These deals were done to provide enhanced global business communications services. Obviously, the Division approved of the deals' objectives. However, these transactions raised concerns because they could limit access to the UK (the BT/MCI deal), France and Germany (the Sprint/FT/DT deal) by requiring competitors to deal with the respective ventures to provide services into these countries. The connections needed to accomplish this access also would have resulted in competitively sensitive information being given to BT/MCI and Sprint/FT/DT. Because these ventures could appropriate some of the rewards of the competitors' innovations and delay implementation of others, the access aspects of the network contributed to the competitive concerns. The consent decrees entered in these transactions ameliorated these concerns by spelling out some key terms governing access and prohibiting confidential information from being improperly disseminated and used by the parties.
In focusing on enforcement actions, it may be easy to get the idea that there is a predilection to find a competitive problem. This is not so. In most cases, a rapidly changing technological landscape will strongly suggest that the exercise of market power is unlikely. The conventional wisdom still applies. It is important, however, in planning a transaction or contemplating a business arrangement carefully to consider whether the particular market dynamics and structure would give rise to a competitive problem.