No. 324

DISMAL SCIENCE FICTIONSNetwork Effects, Microsoft, and Antitrust Speculation

by Stan Liebowitz and Stephen E. Margolis

Executive Summary

The Justice Department and other parties that havealigned against Microsoft have invoked novel economic theo-ries to justify new antitrust doctrines and to revive oldones. Those theories imply that in high-technology mar-kets, a product or technology with a head start or largemarket share may have an insurmountable advantage over itsrivals. The theories invoke factors called networkeffects, increasing returns, or path dependence. Any ofthose can allegedly create lock-in, leaving markets stuckwith inferior products or technologies. But these theoriesleave out important elements of real-world markets.Reexamination of the empirical evidence demonstrates thatthe claimed examples of lock-in are not market failures.

Scrutiny of the economic theories brought to bearagainst Microsoft show similar failings. Despite allega-tions that Microsoft uses lock-in to engage in exclusionaryand predatory business practices, exclusion and predationdo not explain its behavior. Furthermore, antitrustenforcers should not focus on who has the right to controlthe Windows desktop. Such a right matters little becauseconsumers can alter the desktop quite easily. At any rate,the market will eventually put desktop rights in the handsof those who value them most. Finally, progress in soft-ware inevitably involves increased functionality. A legalrule against adding functions to software products wouldimpede progress in the software industry.

____________________________________________________________

Stan Liebowitz is a professor of economics in theManagement School of the University of Texas at Dallas.Stephen E. Margolis is a professor of economics at NorthCarolina State University.

October 27, 1998

Introduction

Revolutions in science and technology, while benefit-ing large numbers of people, also bring various stresses.New technologies can alter the scale of business activi-ties, the geographic distribution of those activities, thetypes of firms that are involved in production and distri-bution, and the distribution of wealth. The benefits aremany: consumers may enjoy cheaper goods and new products;firms that implement the new technology may make very sub-stantial profits; and workers may enjoy higher wages, newtypes of careers, and generally expanded opportunities.At the same time, some businesses and workers will loseout as new skills and methods of commerce supplant oldones.

In such circumstances, interested parties have oftenenlisted legislation or regulation to preserve old inter-ests or defend new ones. The historical motivations forU.S. antitrust law have been at least in part an attemptby various parties to defend their stakes in the economy.The antitrust debates over new computer technologies ingeneral, and Microsoft in particular, are consistent withthat pattern. In particular, today, as in the past, thereare calls for restrictions on the leading firms in newtechnology industries. While Microsoft is the currentfocus of scrutiny, the effects are likely to reach muchfurther. As with past generations of antitrust law, theprecedent and enforcement practice reached in the currentdebate are likely to have a wide and long-lasting influ-ence.

In the policy debates surrounding the antitrust cam-paign against Microsoft, the Justice Department and otherparties that have aligned against Microsoft have invokednovel and incomplete economic theories to justify actionagainst a firm with a large market share. In markets forhigh-technology products, it is alleged, a company with ahead start or the largest market share will have what mayprove to be an insurmountable advantage over its rivals.Those new theories are associated with terminology such asincreasing returns, network effects, path dependence, andlock-in.

According to those theories, where industries exhibitincreasing returns, certain old-fashioned beliefs aboutmarket outcomes and market processes should be cast asidein favor of the following: First, the success of productsin the marketplace comes from size, good timing, aggres-sive strategies, or luck, rather than their inherent val-ues. Second, we should have no confidence that new prod-

Page 2

ucts, technologies, or standards should be able to dis-place their established counterparts, even if they offerimportant advantages. Finally, and perhaps most directlyof interest to the antitrust enforcers, any action takenby a market leader that might increase market share, suchas lowering price, should receive heightened scrutiny,since such actions have the likely consequence of lockingout superior products.

Widespread acceptance of such theories would necessi-tate a radical rethinking of antitrust policy. Further,such theories appear to hold considerable sway in today'santitrust debates. For example, Business Week reported:

Instead of basing his attack against Microsoft onoutdated economic theories that demonize bigness,Assistant Attorney General Joel I. Klein is rely-ing on a developing body of antitrust thinkingthat warns that the threat of anticompetitivebehavior could be even greater in high technologythan in traditional industries. This research on"network externalities" deserves to be takenseriously. . . . The Microsoft case is one ofthe first ever in which Justice has made use ofnetwork theory.1

Even a writer at the Wall Street Journal, a publica-tion not known for embracing radical expansions ofantitrust law, has fallen for lock-in theory. Alan Murrayrecently opined, on the paper's front page, that

high-tech industries might be more susceptible toantitrust problems than their low-tech brethren.That's because consumers often feel a need tobuy the product that everyone else is using,even if it isn't the best, so their equipment iscompatible. Economists call this "network exter-nalities."

It's why most people use a keyboard thatbegins clumsily with the letters QWERTY; why mostvideos are now available only in VHS format; andwhy Windows has become the dominant operatingsystem.2

The new theories provide a convenient solution forthose who would bring antitrust claims to bear againstmarket leaders such as Microsoft. Those "outdated econom-ic theories," dismissed so cavalierly in the Business Weekarticle, might fail to support antitrust enforcementagainst the current generation of market leaders in high-

Page 3

tech industries. Standard theories of monopoly, whichhave long provided what economic foundation there was forantitrust, hold that a monopoly restricts output in orderto elevate prices. Monopoly was bad only for those rea-sons. In contrast to that concern, what we seem to see inhigh-technology markets are falling prices and increasedquantities, even as market shares of the market leadersbecome extremely large.3 Absent an allegation of highprices, antitrust authorities have looked to the new lock-in theories to provide some economic support for theiractions against high-technology firms.

The problem is that the new economic theories arefundamentally flawed. Our writings, which have appearedin academic journals since 1990, show that the case forlock-in is an extraordinarily weak one.

While our work has criticized lock-in theories asbeing based on overly restrictive assumptions, the moretelling criticism has to do with the lack of empiricalsupport for those theories. Alleged examples of lock-inseem, when scrutinized, to be more the products of wishfulthinking than the fruits of serious study. This studyreviews the case against the economic theory of lock-inand analyzes the lock-in claims leveled against Microsoftin recent months. With regard to Microsoft, as elsewhere,neither theory nor fact supports the call for antitrustenforcement measures.

The Economics of Increasing Returns, Network Effects,and Path Dependence

A closely related group of ideas comes together underthe theory of lock-in. Increasing returns are said tooccur wherever the net benefits of an activity increasewith the scale of the activity. Within a firm, increasingreturns are present if the average cost of producing goodsdecreases as the output of the firm increases. All firmsare thought to have increasing returns at outputs that aresmall. Economists have also long considered cases inwhich increasing returns are more persistent, such thateven if a single firm were supplying the entire industrydemand, it would still experience decreases in averagecosts as output increased. In such a case, known as natu-ral monopoly, monopoly is the likely evolution of a freemarket. Further, in this instance, monopoly offers socie-ty the opportunity for the lowest possible production costbecause it takes full advantage of decreasing costs: inthat instance, monopoly is socially desirable. The prob-

Page 4

lem, however, is that even though such a monopoly wouldminimize costs, it would be expected to restrict outputand elevate prices.

Many industries regarded as experiencing such persist-ent economies of scale are treated as public utilities:electricity, telephone, natural gas, cable TV, and others.The policy response has been price (or rate of return)regulation. It is interesting, particularly in the con-text of the antitrust debate over operating systems, thatthe industries that we have traditionally regarded as nat-ural monopolies are now being widely deregulated.

Network effects, sometimes called network externali-ties, may be understood as a special case of increasingreturns. With a network effect, the benefit that someonegets from purchasing a product depends on the number ofother users of the product. For example, people who buyfax machines will find them to be more valuable as otherpeople buy compatible fax machines. The relationship toincreasing returns is straightforward. As a productbecomes more popular, it will become more valuable to con-sumers, giving it an increasing advantage over its smallerrivals. As a result, smaller rivals are likely to disap-pear. We may settle on a single format for videocassetterecorders or a single communication protocol for faxmachines. That need not constitute a monopoly in theusual sense of a single firm, depending on ownership ofthe standard, but in many cases it will.

As is the case for natural monopoly, a monopoly out-come may be socially desirable. That observation is crit-ical for consideration of antitrust policies. Of course,the potential for monopoly price elevation still applies,and if it occurred, such price elevation might result insocial losses. That has not been the concern of the net-work externality literature that apparently has influencedthe Justice Department and others, however.

The problem of path dependence, or lock-in, beginswith the observation that monopoly is a likely outcome ofnetwork effects. The concern then shifts to whether thebest technology or product is chosen. The literature thenalleges that we are likely to get the wrong monopolist,producing the wrong product, rather than a monopolistcharging the wrong price or producing the wrong quantity.

Page 5

Theories of Lock-In

A useful starting point for understanding the lock-intheory is an example presented by Brian Arthur, one of theleading figures in the literature of lock-in. Table 1presents his example.4 In Arthur's table, society has anopportunity to develop one of two technologies. For eachtechnology, the greater the number of adopters of thetechnology, the greater the payoffs to those adopters.Network effects, for example, could cause this relation-ship.

Individuals make decisions based on their privateinterests and receive payoffs as shown in the table. Thefirst adopter of technology A would expect a payoff of 10.Similarly, the first adopter of technology B would expecta payoff of 4. Under those circumstances, Arthur notes,the very first adopter of any technology would certainlychoose A, thus receiving 10 instead of 4. A secondadopter would reach the same conclusion, and so on, sincethe advantage of technology A over technology B would onlyincrease with additional adoptions of A. But notice thatif the number of adopters eventually becomes large, tech-nology B offers greater payoffs. Thus for Arthur, thetable tells a story of lock-in to an undesirable outcome.

There are problems with the table and the lessonsthat are drawn from it. First, note that the increasingpayoffs in the table must be stronger for B than for A ifthe story is to unfold as presented. Yet there is no rea-son to think that among competing technologies the onewith the greatest payoffs with many users does not alsohave greater payoffs with smaller numbers of users. At aminimum, that restriction narrows the set of possiblelock-ins.

Table 1Adoption Payoffs ______________________________________________________________________________________

Number of Previous Adoptions 0 10 20 30 40 50 60 70 80 90 100 ______________________________________________________________________________________

Technology A 10 11 12 13 14 15 16 17 18 19 20Technology B 4 7 10 13 16 19 22 25 28 31 34______________________________________________________________________________________

Page 6

Also, the table does not allow adopters to anticipateor influence the outcome, although people clearly do both.If the first person faced with the opportunity to purchasea fax machine had assumed that he was going to be the onlyuser, we would still be waiting for the technology tocatch on. If a technology is owned, the owner of thetechnology may assure adopters that they will receive thehighest available payoffs by leasing applications with acancellation option, publicizing current and planned adop-tions, or simply bribing adopters with low prices or othercompensation. Of course, the owners of both of thosetechnologies can do this, but the owner of the technologythat creates more wealth can profitably invest more to winsuch a contest.5

Lock-in theory is often argued in the context of for-mal models with multiple equilibria. In such models, sev-eral different outcomes are equally likely, even thoughthey may not be equally desirable, and the choice amongthem is largely a matter of coincidence. Better products,it is shown, may not win. All of that comes to us dressedin the full rigor of mathematical proofs, supported bytheorems on stochastic processes.6 Those models, althoughmore complex than Arthur's table, again abstract from mostof the things that companies do to win technological racesand most of the things that consumers do to purchase well.7

Inclusion of those factors is difficult and cannot provethat markets always choose correctly. Therefore, whetherwe are considering the increasing returns story of Table 1or multiple equilibrium models, we are left with an empir-ical question: Have the models captured something importantabout the way that markets work? The next sectionaddresses that empirical question.

Evidence for Lock-In in the Economy:Do We Get the Wrong Monopolist?

Facts, or empirical evidence, must be the finalarbiters of those theories, as they are of all theories.Given the extensive publicity surrounding these theories,one might conclude that a large body of evidence supportsthem. Nothing, however, could be further from the truth.The little support that has been offered consists of a fewkey examples of markets that have supposedly settled onthe wrong system or standard and failed to change to apurportedly better system or standard. The following sub-sections address those examples.

Page 7

The Qwerty Keyboard

The most commonly cited example in the network-exter-nality, path-dependence literature is the prosaic type-writer keyboard.8 The importance of that example can begleaned from Paul Krugman's 1994 book, Peddling Pros-perity.9 In his book Krugman speaks glowingly of theentire literature in a chapter titled "The Economics ofQWERTY." He does, however, appear to have altered hisviews when made aware of the facts presented below.10

QWERTY refers to the letters in the upper left-handportion of the typewriter (and computer) keyboard. Thereceived story is that, by slowing typing speed, theQWERTY arrangement was able to minimize the problem ofjamming keys in the first typewriters. The story contin-ues that QWERTY's ascendance was due to a serendipitousassociation with the winner of a famous typing contest whoby happenstance used the then-innovative touch-typingmethod on a QWERTY keyboard.

The QWERTY design is reputed to be far inferior tothe "scientifically" designed Dvorak keyboard, whichallegedly offered a 40 percent increase in typing speed.Supposedly, the Navy conducted experiments during World WarII demonstrating that the costs of retraining typists onthe new keyboard could be fully recovered within 10 days.The story is claimed to validate path dependence: no typ-ists learn Dvorak because too many others use QWERTY,which increases the value of QWERTY all the more.

That is an ideal example because the number of dimen-sions of performance is small, and in those dimensions,the Dvorak keyboard appears overwhelmingly superior. Yetupon investigation, the story appears to be based on noth-ing more than wishful thinking and a shoddy reading of thehistory of the typewriter keyboard. The QWERTY keyboard,it turns out, is about as good a design as the Dvorak key-board and was better than most competing designs thatexisted in the late 1800s when there were many keyboarddesigns maneuvering for a place in the market.

Ignored in the stories of Dvorak's superiority is acarefully controlled experiment conducted under the aus-pices of the General Services Administration in the 1950scomparing QWERTY with Dvorak. That experiment contradictedthe claims made by advocates of Dvorak and concluded thatretraining typists on the Dvorak keyboard made no sense.Modern research in ergonomics also finds little advantagein the Dvorak keyboard layout, confirming the results ofthe GSA study.

Page 8

So on what bases were the claims of Dvorak's superi-ority made? Critical examination shows that most, if notall, of the claims of Dvorak's superiority can be tracedto the patent owner, August Dvorak. His book on the rela-tive merits of QWERTY and his own keyboard is about asobjective as a television infomercial.

The wartime Navy study turns out to have been con-ducted under the auspices of the Navy's chief expert intime-motion studies--Lt. Comdr. August Dvorak--and theresults of that study were clearly fudged. There is farmore to the story, but it all leads to the conclusion thatthe QWERTY story qualifies as no better than a convenientmyth.11

The acceptance of that story, wrong in almost everydetail, illustrates both the desire of path-dependence the-orists for empirical support and their reluctance to checkthe facts. The economic historian who wrote an influen-tial paper on the keyboard story and who cited the Navystudy to provide support for path dependence theoriesnever actually examined the Navy study.12

We published a very detailed account of that in thespring of 1990.13 Yet in spite of our paper, which has notbeen factually disputed, economist Garth Saloner, who iscertainly aware of our paper, used the keyboard example asrecently as last fall at Ralph Nader's anti-Microsoft con-ference.14 One could hardly find better evidence of thetheory's lack of empirical support than the continued useof a result that is known to be incorrect.

Beta-VHS

The second most popular example of how marketsallegedly get locked into poor standards is the Beta-VHSvideorecorder format struggle. It is sometimes claimedthat Beta was a better format and that VHS won the compe-tition between formats only because it fortuitously got alarge market share early in the competition with Beta.That story turns out to be just as inaccurate as the key-board story.15

In 1969 Sony developed a cartridge-based video-recorder, the U-matic, which it hoped to sell to house-holds. Because other companies were also working on suchproducts, Sony invited Matsushita and JVC to produce themachine jointly and to share technology and patents, whichthey did. Sony hoped by its behavior to achieve a stan-

Page 9

dard, which implies considerable foresight on the part ofthe market participants. The U-matic failed, however.

In the mid-1970s Sony developed the Betamax. Sonydemonstrated the machine to Matsushita and JVC and dis-closed technical details, hoping to establish a new set ofagreements. Months later, when JVC demonstrated its newmachine to Sony, Sony engineers concluded that JVC hadexpropriated their ideas. The resulting bitterness leftSony and Matsushita-JVC to go their separate ways.

The only real technical difference between Sony's Betaand Matsushita-JVC's VHS was the manner in which the tapewas threaded and, more important, the size of the cas-sette. Sony believed that a paperback-sized cassette,allowing easy transportability (although limiting recordingtime to one hour), was paramount to the consumer, whereasMatsushita believed that a two-hour recording time, allow-ing the taping of complete movies, was essential. Thelarger VHS tape meant that for any given state of record-ing technology, VHS machines could provide longer playingtime, or higher quality playback, or some combination ofthe two.

The behavior of the antagonists in that competitionis a wonderful example of forward-looking behavior. Theyused partnerships, advertising, pricing, and any other toolat their disposal. The behavior was nothing like the pas-sive adoption story that the theoretical models of lock-inpresent.

In an attempt to increase market share, Sony con-tracted to have its Beta machines sold under Zenith'sbrand name, a highly unusual move for Sony, and licensedthe format to Toshiba and Sanyo. To counter that move,Matsushita allowed RCA to put its name on VHS machines andbrought Hitachi, Sharp, and Mitsubishi into its camp.Sony slowed tape speed to increase its playing time to twohours; VHS did the same and increased playing time to fourhours. RCA radically lowered the price and came up with asimple but effective ad campaign: "Four hours. $1000.SelectaVision." Zenith responded by lowering the price ofits Beta machine to $996.

The market's referendum on playing time versus tapecompactness was decisive and rapid. Beta had an initialmonopoly for almost two years. But within six months ofVHS's introduction in the United States, VHS was out-selling Beta. Those results were repeated in Europe andJapan. By mid-1979 VHS was outselling Beta by more thantwo to one in the United States. By 1983 Beta's world

Page 10

share was down to 12 percent. By 1984 every VCR manufac-turer except Sony had adopted VHS.

Not only did the market not get stuck on the Betapath; it was able to make the switch to the slightly bet-ter VHS path. Although Beta was first, VHS was able toovertake Beta very quickly. That, of course, is the exactopposite of what path-dependence theory predicted: that thefirst product to reach the market is likely to win therace even if it is inferior to later rivals.

Now listen to the version of that story found inBrian Arthur's work:

Both systems were introduced at about the sametime and so began with roughly equal marketshares. . . . Increasing returns on early gainseventually tilted the competition toward VHS:. . . if the claim that Beta was technicallysuperior is true, then the market's choice didnot represent the best outcome.16

The story is little more than an inaccurate anecdote.The elevation of poorly researched anecdotes to the cate-gory of "proof" for narrowly constructed theories reappearsin the current discussions surrounding Microsoft, as shownbelow.

Other Purported Examples, Including the Macintosh

Path-dependence advocates have sometimes claimed thatthe continued use of FORTRAN by academics and scientistsis an example of getting stuck on a wrong standard. Onedoesn't have to browse through too many computer magazinesto realize that FORTRAN was long ago superseded by lan-guages such as Pascal, C, C++, and now, perhaps, Java.Individuals continue to use Fortran, not because they wantto be like everyone else, but because the cost of switch-ing is too high. Network effects, as normally modeled,should have induced them to switch years ago. That is astory of ordinary sunk costs, not of network "externality"or other market failure.

Path-dependence proponents have also sometimes claimedthat the gasoline-powered engine might have been a mis-take, and that steam or electricity might have been asuperior choice for vehicle propulsion. That is in spiteof the fact that in the century since automobiles becamecommon, with all of the applications of motors and batter-ies in other endeavors, and with all the advantages of

Page 11

digital electronic power-management systems, today's mostadvanced electric automobiles do not yet equal the bestinternal combustion automobiles of the late 1920s.

The most captivating of those other stories, however,is the success of the PC over the Macintosh. Mac usersnaturally favor the claim that they chose operating sys-tems wisely whereas the rest of the world ignorantly optedfor Microsoft's products. The presence of that large andsomewhat embittered audience probably explains why the ideaof getting stuck with an inferior product resonates sostrongly in the Microsoft case, playing as it does in thearena of personal computer aficionados.17

Yet even here the facts do not support the lock-inthesis.18 Yes, Macintosh owners were forward looking whenthey made their purchases in the early and mid 1980s. Atthat time, the advantages of a graphical interface wereclearly understood, if not fully implemented. The lack ofimplementation, however, had to do with the high price, interms of speed and cost, that owners of graphical comput-ers had to pay at a time when processors were slow andmemory was expensive.

A graphical user interface requires considerably morepower to get the job done, significantly increasing costswhen that power is hard to come by. Macintosh owners hadto wait for their screen displays to change, whereas PCowners had almost instantaneous updates.19 True, Macintoshowners could see italics and bold onscreen, but to printthe screen as they saw it required a postscript printer,and such printers cost about $1,000 more than ordinarylaser printers. The graphical user interface allowedusers to learn programs much more easily, but in manybusiness settings, a computer tended to be used for only asingle application. In that environment, the operator hadvery little interaction with the operating system inter-face, and once the operator had learned the application,the advantages of the graphical user interface were dimin-ished.

The case for DOS, therefore, was stronger thanappears from the vantage of the 1990s with our multi-megabyte memories and multigigabyte hard drives. Now thatwe routinely use computers that can run 30 times as fastas those old DOS machines and have 50 times the memory and100 times the hard drive capacity, the requirements of agraphical operating system seem rather puny. They wereenormous, however, in the days of DOS.

Page 12

As processors became faster, memory cheaper, and harddrives larger, the advantages of a graphical user inter-face should have overcome any command-based, or text-based,system such as DOS. If we were still using DOS, thatwould certainly be an example of being stuck with an infe-rior product. We are not using DOS, however.

Instead we are using a Mac-like graphical user inter-face. If someone went to sleep in 1983 and awoke in 1995to see a modern PC, he most likely would think that theMacintosh graphical user interface had been colorized andupdated, with a second button added to the mouse.20 Ourmodern Rip Van Winkle might be surprised to learn, howev-er, that the owner of the graphical user interface was notApple but Microsoft.

The movement from DOS to Windows was costly, yet itoccurred quite rapidly. As in the other examples, whatthe evidence shows is quite the opposite of what the path-dependence pundits predict: not markets getting stuck inruts, but markets that make changes when there is a clearadvantage in doing so.

Microsoft's Dispute with the Justice Department

Historically, new antitrust doctrines have developedin connection with big cases, which most often involvedthe biggest and most successful companies. That patternis being repeated today. Various antitrust actionsagainst Microsoft have been the main venue for discussionsand actions that propose and explore new economic founda-tions for antitrust and new interpretations of oldantitrust doctrines.

Microsoft's antitrust problems began with a governmentinvestigation of its pricing of software sold to originalequipment manufacturers (OEMs). Microsoft agreed to endthose practices in a highly publicized 1994 consent decreewith the Department of Justice. Whether or not thosepractices were anticompetitive, they doubtless had littleto do with Microsoft's successes in the market.21

The consent decree did little, however, to endMicrosoft's legal problems with the DOJ. When Microsoftattempted to purchase Intuit, a maker of financial soft-ware, the DOJ opposed the deal. In a highly publicizeddecision, the consent decree itself was temporarily over-turned by Judge Stanley Sporkin's (later overturned) deci-sion, which appears to be the first instance of path-dependence theory's having a serious influence on policy.

Page 13

There were other skirmishes as well. The DOJ exam-ined Microsoft's inclusion of the Microsoft Network iconon the Windows 95 desktop. It claimed that consumerswould be unwittingly forced into acceptance of that prod-uct to the detriment of competition in the online serviceindustry.

Most recently, the DOJ has accused Microsoft ofattempting to divide the browser market with Netscape and,when Netscape refused its offer, of employing a variety ofexclusionary tactics to unlawfully drive Netscape out ofbusiness.22 In large part, the DOJ's claims revolve aroundMicrosoft's inclusion of its Web browser in the Windowsoperating system23 and Microsoft's insistence that OEMs nottamper with consumers' access to its browser.24 The nextsection discusses those issues.

The DOJ's most recent lawsuit against Microsoft bearsthe clear imprint of path-dependence theory. The com-plaint argues that Microsoft's allegedly anti-competitiveagreements "threaten to 'tip' the market permanently toInternet Explorer"25 and that "because of the market's net-work effects, the significant increase in Microsoft's shareof the browser market . . . will tip the market inMicrosoft's favor and accelerate its dominance and competi-tion's demise."26 Ultimately, the DOJ regards networkeffects as problematic whether or not it can relate themto traditional antitrust violations: "[B]arriers that existto the entry of new competitors or the expansion of small-er existing competitors, including network effects, meanthat dominance once achieved cannot readily be reversed."27

A consortium of 21 attorneys general from variousstates and the District of Columbia joined the DOJ both insuing Microsoft and in embracing path-dependence theory."Microsoft's operating system software enjoyed greater con-sumer acceptance, familiarity, and loyalty than its latercompetitors by virtue of the fact that it was first in themarketplace,"28 their complaint argued, denying Microsoftany credit for offering quality goods at low prices. Theattorneys general had a more trendy explanation ofMicrosoft's success: "Through 'network effects,' Microsoftwas able to build upon this initial advantage."29

Newspaper accounts and public statements by other par-ticipants confirm that path-dependence theory has, in wholeor in part, driven the DOJ's antitrust attack onMicrosoft. The most famous, and perhaps most influential,attempt to connect these theories to antitrust is a studyprepared by Gary Reback, a lawyer working for several ofMicrosoft's competitors, with the assistance of two econo-

Page 14

mists who have played prominent roles in promoting path-dependence theory, Brian Arthur and Garth Saloner.30

Their "White Paper" on Microsoft goes far beyond theeconomics literature. They do not stop with the tradi-tional path-dependence claim that a market-based economy islikely to choose all sorts of wrong products. Nor do theystop with the claim that innovation might be eliminated inthe computing industry. Reback, Arthur, and Salonerinstead portray Microsoft as an evil empire intent onnothing less than world domination. To hear them tell it,the American way of life will be imperiled if the govern-ment does not rein in Microsoft: "It is difficult to imag-ine that in an open society such as this one with multipleinformation sources, a single company could seize suffi-cient control of information transmission so as to consti-tute a threat to the underpinnings of a free society. Butsuch a scenario is a realistic (and perhaps probable) out-come."31

Those are fantastic claims, indeed, and they wererepeated at a recent conference on Microsoft, held byRalph Nader, at which Reback, Arthur, and Saloner madepresentations.32 To judge from the DOJ's recent claimsagainst Microsoft, all the talk about path dependence hasbegun to influence public policy. As the next sectionexplains, however, path-dependence theory provides no jus-tification for expanding antitrust enforcement.

Antitrust Doctrines and Network Technologies

Both the DOJ and some of Microsoft's private competi-tors have used theories of lock-in to support a call forheightened antitrust scrutiny of Microsoft. By itself,lock-in would seem not to constitute an antitrust offense.There is nothing in the law that makes it a crime to havetechnologies that are less than the best available or lessthan the best imaginable.33 Instead, lock-in theoriesoffer an alternative way to claim harm in the absence ofthe usual monopoly problem of elevated prices andrestricted outputs. Lock-in stories also offer new lifeto and a contemporary spin on old antitrust doctrines.The following two subsections consider some of theantitrust issues that have been raised in the context ofthe software industry. The first subsection describes whymonopoly leverage in theory requires special conditionsthat virtually guarantee it will not occur in practice.The second describes why no smart monopolist would trypredatory bundling.

Page 15

Monopoly Leverage, Tie-Ins, and Bundling

In theory, monopoly leverage occurs when a firm usesits monopoly in one industry to win a monopoly in anotherindustry. Tie-in sales and bundling are contractual prac-tices that are sometimes alleged to facilitate monopolyleverage, but tie-ins and bundling do not have to create anew monopoly to be profitable. Nor do tie-ins necessarilyharm consumers.34 In fact, as this subsection explains,the theory of monopoly leverage requires so many specialconditions that monopoly leverage seems certain to remaina theoretical problem.

Economists have long been skeptical that monopolyleverage is either feasible or profitable. In most cir-cumstances, forcing consumers to purchase some other prod-uct so as to create a second monopoly will not add to afirm's profits. A monopolist can instead simply extractthe value of its monopoly through the pricing of the goodin the market where it has its first monopoly.

Suppose, for example, that a firm held a monopoly onoil furnaces. Such a monopoly might be quite profitable;oil furnaces are useful and offer some advantages overother kinds of furnaces. The monopolist's ability to max-imize profits would face some limits, of course, such asthe availability of substitutes like propane and electricheating.35 Still, the monopolist could devise a pricingsystem that captured the extra value of using an oil fur-nace rather than a competing source of heat. The lowerthe price of heating oil relative to the price of propaneor electricity, the greater that value would be. If thefurnace monopolist were to become the oil monopolist too,he might raise the price of heating oil, but that wouldonly reduce what he could extract through the furnaceprice.

Consider this analogy: regardless of whether it wor-ried you that someone had a key to the front door of yourhouse, it would not worry you more if that person also hada key to your back door. Nevertheless, the idea that thesecond monopoly could be used for something has intuitiveappeal. Even if the monopoly in furnaces could be used toextract everything that can be extracted from the furnaceusers, could not a monopoly in heating oil be used toextract something from people who use heating oil foranother purpose? It turns out that, yes, there is a cir-cumstance in which a second monopoly is worth something.That circumstance is a very limited one, however. If the

Page 16

furnace monopolist could also monopolize the heating oilindustry, he could extract additional monopoly rents fromheating oil users who were not also his furnace customers.

The question then arises of whether one monopolycould ever be extended to capture customers of anothermarket. The answer again is yes, it is possible--but,again, only under very special circumstances. If therewere economies of scale in the heating oil industry and iftoo few customers bought heating oil for nonfurnace usesto support a separate supply of heating oil, then the fur-nace seller could lever his monopoly in furnaces into amonopoly in heating oil by preventing furnace customersfrom buying heating oil from other sources. By assump-tion, the nonfurnace customers would not offer a largeenough market to support any independent oil supplier andthe furnace monopolist could then extract new monopolyrents in that other market. That explanation of leverageis sometimes referred to as market foreclosure.36 Iron-ically, the larger the furnace monopolist relative to theheating oil industry, the less likely it will benefit frommonopolizing heating oil since it will already have virtu-ally all the potential customers.37

That explanation shows that there is a theoreticalpossibility of harmful monopoly leverage but that itrequires very special conditions. The leveraged marketmust be big enough to matter, but not so big as to allowcompetitive independent suppliers to survive. There mustbe some economies of scale in the leveraged market, butnot enough to have caused prior monopolization of the mar-ket. The leveraged market must have many of the same cus-tomers as the initial monopoly, so as to provide controlof the new market, but not too many, or there will be nonew rents to extract by establishing the second monopoly.In short, leveraging can be viewed as the Goldilocks theo-ry of monopoly extension--everything has to be just theright size.

Do the facts of the Microsoft case fit within theleverage story at all? If Microsoft requires each cus-tomer to buy one copy of some other Microsoft product,that, in and of itself, adds nothing to its profits. Thatsort of tie-in sale with fixed proportions has long beenunderstood to offer no particular advantage to the monopo-list.38 So the issue becomes whether Microsoft could crowdout any rivals that sell to customers who do not useMicrosoft's own operating system.

Here the application to Microsoft of the market fore-closure theory runs into trouble. If the products

Page 17

allegedly crowded out by Microsoft's bundling are productsthat run only under the Windows operating system, thenmonopoly leverage offers Microsoft no advantage.

To illustrate that point, consider a hypotheticalexample of successful tying-foreclosure using personalsoftware products, such as Quicken and Microsoft Money.Both are sold in both the Macintosh market and the Windowsmarket. If Microsoft were to build Microsoft Money intothe Windows operating system, and if that eliminatedQuicken in the Windows market, and if the Macintosh marketwere too small to allow a product like Quicken to be pro-duced at reasonable average cost for that market alone,and if Microsoft continued to sell the product separatelyin the Macintosh market (now at a monopoly price), and ifthere were few additional costs for Microsoft in creatinga Macintosh version, then, and only then, would Microsoftbenefit from leveraging monopoly.

Has that occurred? Does Microsoft sell in theMacintosh market disk compression, backup, fax software, orany other program that is included in the Windows operat-ing system? The only product that comes to mind is aMacintosh version of Internet Explorer. But Microsoftgives away that product in the Macintosh market and prom-ises a permanent price of zero. If Microsoft sticks toits promise, it cannot profit from including the browserin the operating system. Even then, the other requiredconditions for market foreclosure (that the Macintosh mar-ket be too small to support Navigator and the costs ofMicrosoft's creating a Macintosh version not be too large)probably would not be satisfied.

A simple rule that would prevent that type of fore-closure would be to prevent Microsoft from including inits operating system any program that it sells separatelyin another market. But to bar Microsoft from selling sep-arate versions in other markets would exclude customers inthose markets from the benefits of those programs, evenwhen Microsoft did not contemplate market leverage. Givenall the special conditions required for successful leverag-ing, it would be unwise to implement such a rule withoutfurther investigation of the potential harm of denyingMicrosoft products to consumers in tied markets. It bearsrepeating, moreover, that intervention would remain unjus-tified absent a showing that Microsoft has a sustainablemonopoly in PC operating systems and that the costs ofintervention, including its stifling effect on innovation,outweigh its putative benefits.

Page 18

Predatory Bundling

The most recent allegations against Microsoft concernpredatory use of its ownership of the Windows operatingsystem. The specific allegation is that Microsoft's inte-gration of its browser into the operating system is large-ly predatory in intent, aimed at forcing other firms outof the browser market. The implications of that issue,however, reach well beyond the browser market and extendto the very issues of what an operating system can be andthe nature of progress in the software industry.

Antitrust law defines as predatory actions that areinconsistent with profit-maximizing behavior except whenthey succeed in driving a competitor out of business. Inpredatory pricing, for example, a would-be monopolistallegedly charges a price that is so low that other firmscannot sell their outputs at prices that will cover eventheir variable costs. The other firms are then forcedeither into bankruptcy or to exit the industry becausethey have become unprofitable. Upon completing the preda-tory phase, the predator then enjoys the benefits ofmonopoly pricing. Note that during the predatory phaseconsumers benefit greatly from the low prices; only thelater monopoly pricing harms consumers.

Economists are generally skeptical of claims thatprice cuts or other actions have predatory intent becausethey have determined, both in theory and in practice, thatpredatory campaigns are unlikely to have profitable end-ings. First, the predatory action is likely to be moreexpensive for the predator than for the prey. The preda-tor cannot just cut price; it must also meet market demandat the lower price. Otherwise, customers will be forcedto patronize the prey, even if at higher prices. If thepredator is a large firm, it stands to lose money at afaster rate than the prey. Second, even if the predationsucceeds in bankrupting the prey, there is no guaranteethat a reestablished firm will not just reenter the indus-try once the predator has established monopoly pricing.If there are fixed investments in the industry, such asdurable specialized equipment, the predator cannot estab-lish monopoly prices as long as the durable assets canreturn to the market. If there are no durable assets,then the prey can cheaply exit the industry and reenterwhen monopoly prices return. Either way, predation drainsthe predator while imposing uncertain burdens on the prey.

Another problem with predation is that almost anyaction that a firm takes to become more attractive to con-sumers can be alleged to be predatory. If customers like

Page 19

something a firm is doing, its competitors will not. Inthe most elementary case, a price cut or product improve-ment will damage the prospects for some competitor. Itbears noting that most of the alleged cases of predationhave been demonstrated to be false.39

Predatory bundling, like predatory pricing, is a sim-ple idea that ultimately proves too simple. If a firmwith a controlling share of one product bundles in someother product, competitors who sell the bundled-in productwill have to compete with a product that, to the consumer,has a zero cost. If Microsoft includes in its operatingsystem a piece of software that competes with other ven-dors in what had been a separate market, Microsoft ensuresthat virtually all purchasers of computers then have acopy of the new software.

Suppose Microsoft bundles a fax program intoWindows98. If Microsoft's fax program, relative to itscost, is better than other fax products, then the bundlingis not predatory. The Microsoft product would win in themarketplace anyway; adding it to the operating systemcosts less than its value to consumers. If the product isworth more to consumers than the costs of creating it,then bundling will also be profitable without any exclu-sionary consequences. In contrast, if Microsoft's faxprogram, again considering its cost, is inferior to alter-natives or provides less value than its cost, thenMicrosoft would profit only if bundling caused other firmsto exit the market and if Microsoft were able to raise theoperating system's price by the now-higher implicit monop-oly price of its fax product.

As a strategy, however, predatory bundling has thesame liabilities as predatory pricing. As in predatorypricing, Microsoft stands to lose money (relative to notincluding the fax software) faster than its rivals do ifits fax program costs more to produce than its value toconsumers. Moreover, a rival with a superior fax programcould keep the product on the market at a price thatreflects the advantages that it offers over the bundledproduct. The rival could not charge more than thatbecause the Windows consumer would already have the infe-rior fax program. The rival could still capture the extravalue that its own intellectual property contributes, how-ever, especially since it would enjoy low marginal costsof "producing" (i.e., copying) its software and the mar-keting edge of an installed customer base. While it maylose profits or market share, therefore, the rival willretire its fax program only if it is inferior toMicrosoft's.

Page 20

From a social or consumer welfare perspective, then,Microsoft's bundling action would do no harm. The rivalsoftware is a fixed asset in the industry; it does notwear out. In the extreme case, a bankrupt producer mightput its fax program on the Web, making it freely availableto all. That would limit what consumers would be willingto pay for the program bundled into Windows98 to its extravalue, which is zero. Thus Microsoft would be unable tocharge a higher price for the bundled software despitehaving incurred the costs of creating the fax program.Microsoft would lose money and fail to monopolize the mar-ket. Furthermore, the talents that created the rival faxprogram still exist, ready for some other firm to hireshould Microsoft ever achieve a monopoly price on the faxprogram.

Of course, an antitrust enforcer might reply that theoperating system producer has distribution or coordinationadvantages that an independent rival lacks. If thoseadvantages outweigh any quality advantages of the rival,however, it is efficient for the operating system producerto bundle its fax program.

That suggests that bundling, as a predatory action,is unlikely to succeed. Furthermore, the software indus-try has very important nonpredatory reasons to bundlefunctions into operating systems and other software prod-ucts. As explained below, new sales of software willrequire continual additions to functionality.

In the Netscape case, antitrust enforcers might allegethat Microsoft is not interested in defeating the Netscapebrowser as much as in destroying Netscape as a company.Industry pundits have often theorized that Web browsersmight constitute a means of establishing a new operatingsystem. Netscape, they allege, constitutes a threat toMicrosoft's position in the operating system market.Regardless of the technical reasonableness of that claim,however, it runs into the same problems as other allega-tions of predation.

Here, as elsewhere, predation would not destroy thedurable assets of the prey. Netscape's software willhardly disappear if Microsoft bundles a browser intoWindows. Indeed, Netscape has already made the sourcecode for its Navigator program publicly available. Evenif Microsoft still tried to destroy Netscape in order toprotect Windows' market share, it would ultimately fail.Any of Microsoft's several large and vigorous competitors,such as IBM or Sun, would happily purchase Netscape, or

Page 21

hire its engineers, if they thought that by doing so theycould share some of Microsoft's enviable profits.

The Rate of Innovation

Putative Dangers

One concern that has been raised by the JusticeDepartment, in the Judiciary Committee hearings, by somejournalists, and by several path-dependence theorists isthat Microsoft's dominant position in the market willsomehow inhibit innovation. The suggestion is thatMicrosoft will be able to so dominate the software marketthat no small firms will dare compete with it. Firms willbe unwilling to create new products in any market that islikely to attract Microsoft's attention, especially inproducts that are possible additions to the operating sys-tem. It is not clear that current antitrust law addressessuch concerns. If valid, however, and if not addressed byantitrust law, they might encourage new legislation. Ofcourse, the impact of such legislation would probablyreach beyond the computer industry.

Concerns about lock-in drive the accusations againstMicrosoft. Consumers are viewed as being so locked intoMicrosoft's products that, even if the Wintel platformfell far behind the cutting edge of computer technology,no other combination of an operating system, applications,and support could displace it. Obviously, no one canempirically disprove the claim that products might havebeen created that would have been better than currentlyexisting products. Instead, the analysis here focuses onwhether lock-in theory correctly concludes that Microsoftwill stifle innovation in the computer industry.

Certainly there are instances in which Microsoft hasincorporated programs into the operating system and theformer providers of such programs have gone on to otherthings. Disk compression and memory management programsare two examples. Fax programs, backup programs, anddisk-defragmenting programs are counterexamples: the inclu-sion of such programs in the operating system has noteliminated the separate market. The key factor appears tobe whether the programs Microsoft includes in its operat-ing system are as good as the separate programs or not.When Microsoft produces a product as good as or betterthan the competition, the separate market usually doesdisappear. It is difficult, however, to conceive of con-sumer harm in that case.

Page 22

The general claim that innovation will suffer ifMicrosoft is allowed to grow and add programs to the oper-ating system has several shortcomings. For one thing, itwrongly assumes that programmers hired by Microsoft lackor lose creativity. It proves nothing to observe thatsmall startup companies generate creative ideas more fre-quently than Microsoft does. There are 15 times as manyoutside programmers developing programs for Windows asthere are programmers working for Microsoft. ThatMicrosoft does not lead the industry in creative program-ming thus hardly shows that Microsoft could not, or wouldnot, hire programmers to produce software just as creativeas that produced by smaller companies.

It is, of course, conceivable that large firms pro-duce less innovation than small firms do (adjusting forsize). That has been investigated at length in the eco-nomics literature with no clear consensus.40 Every firm,no matter what its size, benefits from profitable newideas. Microsoft itself recently announced plans to dou-ble its already huge research and development budget.41

The argument that large firms might not innovate asmuch as small firms usually relies on some variation ofthe view that large firms are fat and lazy; that is, thatthey do not innovate because they do not have to. Still,a dollar is a dollar. Most investors are just as eagerfor their large-firm stocks to maximize returns as theyare for their small-firm stocks to do so. For the fat-and-lazy condition to hold, it must be that large firmswith dispersed ownership of their stock do not have thesame incentives to maximize shareholder value and profitsas do small firms, which are usually closely held. Thatreal possibility is known as the problem of separation ofownership and control.

With regard to Microsoft and many other successfulhigh-technology firms, however, that argument would seem tohave little force. The ownership of Microsoft and mostother high-tech firms is not widely dispersed. For exam-ple, Bill Gates owns almost 25 percent of Microsoft andseveral other early Microsoft investors also own very sub-stantial stakes. That may in fact explain why Microsoftis still considered such an intense competitor.

Alternatively, Microsoft's critics vaguely suggestthat it stifles innovation because it copies softwareideas from others, leaving the other firms no reward fortheir efforts. If that claim were true, the problem wouldappear to lie in intellectual property law, notMicrosoft's purported monopoly power. After all, if

Page 23

Microsoft were allowed to copy the ideas of its rivals, sowould be a host of other large (or small) firms in theindustry, in each instance lowering the profits of theinnovator, large or small.42

It would be a serious problem if innovators in soft-ware were not being properly rewarded for their efforts.The purpose of intellectual property laws is to allowinnovators to collect economic rewards for their efforts.Without such laws, imitators could free ride on theefforts of innovators and produce similar products atlower cost, driving true innovators out of business. So,while deserving of investigation, those problems do notseem to pertain to Microsoft, or its ownership of theoperating system. Perhaps a reevaluation of intellectualproperty laws is in order. That claim seems to have lit-tle to do with antitrust.

Some factual matters do not seem consistent with theclaim that Microsoft reduces innovation. Microsoft'sbehavior toward its developers, for example, does not seemto square with the claim that it is intent on driving outindependent software producers. It launched a programlast fall to "help out startups that are developingMicrosoft-compatible software by giving them space in tradeshow booths, discounted software and passes to tradeshows."43 Unsurprisingly, some startups regard Microsoftwarily and refuse its help. But "the companies thatMicrosoft invests in have few scare stories to report."44

More broadly, there seems to be little evidence thatthe pace of innovation is insufficiently rapid. The paceof innovation in the computer industry is generallyregarded with some awe. Certainly, the Windows marketdoes not appear to have suffered from stifled developmentof applications.45

Finally, there appear to be tremendous rewards tothose who do innovate in this industry. Even the foundersof Netscape, a supposed victim of Microsoft's power,walked away with hundreds of millions of dollars. Doesthat discourage others from taking the same path? Unlessand until careful research answers those sorts of ques-tions, any antitrust action would be premature and poten-tially dangerous to the software industry and economy.

A Real Danger to Innovation

The nature of software markets requires that softwareproducers continually add functionality to their products.

Page 24

In contrast, McDonald's can keep selling Big Macs becausefood is not durable and consumers want to purchase BigMacs that taste just like those they ate the day before.Because software lasts forever with no diminution in qual-ity, however, consumers have no reason to repurchase aword processor or operating system unless an improved ver-sion comes to market. Undoubtedly, improving softwarewill mean adding functionality.

Consider what it means to improve software. Whilesoftware could be made faster, real speed improvementsoccur when consumers replace their old computers withfaster ones. And while intuitive interfaces are useful,consumers can with practice become adept even with poorlydesigned software. Consumers will naturally tend to stickwith a familiar version of a program (or operating system)unless a newer version performs some useful tasks notavailable in the old version. That requires adding func-tionality not found in previous versions.

Added functionality can be seen in every category ofsoftware. Newer word processors have far more functional-ity than earlier versions. Spell and grammar checkers,mail-merge programs, and thesauruses are among the manyfeatures now standard in virtually all word processors.Spreadsheets, database programs, and virtually every othercategory of program have far more functionality thanbefore.46 That is one reason why new software seems tofill our ever-expanding hard drives, which have hundredsor thousands of times the storage capacity of earliermachines.

The consumer benefits in many ways from added func-tionality. Large programs almost always cost far lessthan the sum of the prices that the individual componentproducts used to command. The various components alsotend to work together far better than separate componentsbecause they are made for each other. If that were notthe case, consumers would not purchase new generations ofsoftware products.

As programs become more functional, the number ofsmall companies specializing in add-ons may shrink. Butthat is no reason to prevent creators of word processorsfrom including grammar checkers and thesauruses. Shouldthe producers of dominant programs be forbidden to addfunctionality while producers of less successful programsare allowed to add new functions? That hardly seems arecipe for success. Do we really believe that innovationwas retarded because add-on companies feared that theymight be put out of business? Do we even know if they

Page 25

have been put out of business, or that those programmersare no longer working on new ideas? Again, the weak logicand scant evidence supporting them make such claims sus-pect.

Yet some Microsoft critics, including those within thegovernment, have proposed freezing the operating system,putting an end to new functionality. If that proposalwere accepted for the operating system, it might logicallyapply to other categories of software. The results wouldbe disastrous, for software producers, who would have nonew sales except to new computer users, for computer manu-facturers, who would find little demand for more capablehardware, and, most important, for users, who would beforced to use seriously crippled software. Few proposedantitrust policies are as dangerous as that one.

Who Should Get to Assign Desktop Icons?The Irrelevance of the "Browser Wars"

At the Senate hearings and in the media, considerableattention has been given to the claim that Microsoft'sdesire to prevent OEMs from removing the Internet Explorericon from the desktop was somehow inimical to competition.This section explains why Microsoft and OEMs might eachwant to control the placement of desktop icons and pro-vides an economic framework for deciding who should beallowed to control the desktop icons--if it really mat-ters. Ultimately, icon placement should probably not mat-ter even to the computer and software industry, much lessto antitrust enforcers.

Control of the desktop might be valuable since, as apractical matter, all computer users see the desktop. Inprinciple, desktop placements of "advertisements," whetherprograms or messages, could be sold to companies interest-ed in such exposure. For example, assume that an icon forAmerica Online appears on the desktop. Consumers inter-ested in an online service might just click on the iconand begin the process of becoming an America Online cus-tomer. Knowing that, a company such as America Onlinemight be willing to pay the controller of the desktop fora good placement of its icon.47

Assume, then, that icon placements are indeed valu-able. The next subsection explains why, nonetheless, reg-ulators should not care whether Microsoft or OEMs controlicon placement. Following that, the discussion criticallyreexamines the assumption that control of icons shouldmatter even to the computer industry.

Page 26

A Simple Theory of "Desktop Rights"

If placing icons on the desktop can generate revenue,it should not be surprising that OEMs and the owner of theoperating system (Microsoft) will both claim the right toplace the icons. Economic analysis allows us to examinewhether it makes any difference who has that right. Italso may provide some guidance as to who should get it.

The Coase theorem helps to explain the tradeoffsinvolved.48 If the rights to place desktop icons were welldefined, and if there were no transactions costs or wealtheffects,49 the Coase theorem tells us that regardless ofwho initially had the rights, they would end up where theyhad the greatest value.

Consider the following example. If the rights tosell desktop placement were worth $5 to Microsoft and $10to OEMs, then OEMs would wind up controlling the desktopicons regardless of who initially did so. If Microsoftinitially controlled the desktop, OEMs would be willing topay up to $10 to Microsoft for the rights, and Microsoftwould be better off selling the rights to OEMs--perhaps byraising the price of the operating system by more than $5but less than $10.

If, on the other hand, OEMs initially controlleddesktop placements, Microsoft would be willing to lowerthe price of the operating system by up to $5 in exchangefor the right to control icon placements. OEMs would pre-fer to keep the rights themselves, however, since they cangenerate more than $5 in revenues by maintaining control.In either case, OEMs wind up with the rights, and bothparties share the $10 in extra revenue brought about byicon placement sales. Although the two parties might beexpected to fight over the rights, it makes no differenceto the rest of us who prevails. By analogy, as virtuallyall microeconomics textbooks explain, if the governmentsubsidizes gasoline purchases, it makes no differencewhether automobile drivers or service stations receive thesubsidy, because in either case the subsidy is shared inexactly the same way.

Sometimes the assumptions of the Coase theorem arenot met. For example, if negotiations between OEMs andMicrosoft were not feasible, efficiency considerationswould require that the property rights be assigned to theparty who could generate the highest value for desktopplacements.50 Since Microsoft and OEMs are already negoti-ating over other aspects of the desktop (e.g., price),

Page 27

however, there is little reason to believe that the marketwill not work efficiently.

The current anxiety regarding desktop placements ismisplaced. As long as the parties freely enter into newcontracts, neither party will benefit from a legal stipu-lation of who initially controls the desktop. It shouldnot matter at all to the government who has the right.

The reader may naturally ask, "If it makes no differ-ence, why is there fighting over who places the icons?"There are two answers. First, there is little evidencethat Microsoft and OEMs are fighting. It is Microsoft'scompetitors who are complaining. Second, it is not unusu-al in such circumstances for there to be contract disputesor strategic behavior. Two parties can negotiate a con-tract, then subsequently dispute their understanding of theterms of that contract. If, for example, OEMs are receiv-ing a lower price from Microsoft because Microsoft thoughtit controlled desktop placement, but now OEMs have achance to sell icon placement while remaining under afixed contractual price for Windows, it would not be sur-prising that a dispute would arise.

Is Icon Placement Valuable?

For icon placement to be valuable, it must generatefuture revenues. America Online benefits in the previousexample because consumers could not use its services with-out paying a monthly fee. Having its icon on the desktopincreased the chances that consumers would sign up for theservice.

For a typical software product to be on the desktop,however, the software is usually already installed on thecomputer, and thus already purchased. The icon placementonly increases its likelihood of use. The only additionalbenefits to the software producer from having the consumeruse the software after purchasing it is that the consumermight purchase upgrades or ancillary products.

For the Netscape and Microsoft browsers there areseveral reasons why the icon placement might be important.(This analysis ignores any future revenues from upgradessince both companies have agreed not to charge forbrowsers or upgrades.) First, Netscape and Microsoftmight be able to trade off the success of their browsersto sell software specializing in serving up Web pages(known as servers) because of their large presence among

Page 28

the base of users and the (presumably) assured compatibil-ity with these browsers.

Second, when a browser is first put into use, it goesto a default starting location on the Internet.51 If manyWeb users (surfers) view a particular location, advertisingrevenues can be generated, as some popular locations onthe Internet, such as Yahoo, have discovered. Yahoo, infact, paid Netscape many millions of dollars to provideNetscape users an easy way to reach the Yahoo page.Netscape and Microsoft, although somewhat late to thatgame, are both working on new start pages (to which thebrowsers will be preprogrammed to go) in the hopes ofenticing users to stay at their Web sites. By leadingconsumers to particular pages, browsers might become apotent revenue-generating force.

Some industry analysts are skeptical that browsericons provide valuable control of the start page, however.First, users can easily alter the start page. To cite afamiliar example, a radio station would not pay automobiledealers to have car radios preset to its frequency. Theradio station would not benefit from such a deal becausecar buyers can easily change the preset stations. Simi-larly, how can icon owners benefit when consumers find iteasy to change the icons on the desktop? The parallelwill become still more compelling as consumers become moreaccustomed to customizing their operating systems.

There is, however, a more fundamental impediment tothe claim that desktop placement is important forbrowsers. Just having the icon on the desktop is insuffi-cient to gain access to the Internet. Clicking on thaticon will not connect users to the Internet. For thatthey will have to use one of many Internet serviceproviders. The Internet service provider will almost cer-tainly provide its own browser, independent of what iconis on the desktop.52 Therefore, the icon on the desktop atthe time of sale does not provide much value.

Finally, the concept of detailed governmental controlover desktop placement leads to other endless and absurdquestions. What about the Start button in Windows? Theorder of programs tends to be alphabetical. Should thegovernment be concerned about the ordering of those pro-grams, and who gets the rights to order them? Has anyoneinvestigated whether the various color schemes found inWindows benefit Microsoft's icons over the alternatives?Does the Windows screen saver that animates a Microsoft

Page 29

Windows icon have anti-competitive subliminal effects? Inconclusion, and in all seriousness, we should ask, Shouldthe government really be involved in those decisions?

Implications

The theories of path dependence and lock-in are rela-tively new to the economic literature. They have not wonover the economics profession after years of debate; norhave they made their way into many economics textbooks.They do not draw on first principles in obvious and secureways. That does not make theories of path dependence andlock-in bad economics, or wrong economics, or inappropriatetopics for academic research. On the contrary, it makesthe academic debate that much more important. Those theo-ries, however, are a poor foundation for public policiesthat could affect the progressiveness of the Americaneconomy.

If we were treating a dying industry, even specula-tive economic medicine might be worth a try. But the com-puter and software industries continue to astound mostpeople--both with the rates at which products improve andthe pace at which prices decline. It makes no sense tosubmit such a robust patient to the risks of economicquackery.

In our academic writings, summarized above, we showthat there is a poor connection between theories of pathdependence and the real-world behaviors of entrepreneursand consumers. Moreover, no connection exists between thealleged empirical support for those theories and realevents. Contrary to the lock-in claim, and contrary tosome popular stories of markets gone awry, good productsdo seem to displace bad ones. Since there is no real sup-port for the supposition that markets fail in an environ-ment of increasing returns to scale, there is no morebasis for antitrust in high-tech markets than in any oth-ers.

There might even be less reason to apply antitrust tosuch markets. Our most basic theory of increasing returnsimplies that monopoly or near-monopoly equilibria are like-ly. Where people do value compatibility, or whereincreases in the scale of a firm really do lower itscosts, dominant formats or single producers will probablyresult at particular moments.

Furthermore, consumers want it that way. Anythingelse will frustrate the urge for compatibility or unneces-

Page 30

sarily raise costs, or both. Monopoly outcomes thus neednot imply that anything has gone wrong or has been donewrong. Monopolies that are undone by the government maylead only to monopolies that are redone in the market.The faces may change, but market structure may not. If weinsist that natural monopolies be populated by severalfirms kept at inefficiently small size, we are likely tofind those markets taken over by foreign companies withoutsuch restrictions.

In such markets, firms compete to win a monopoly.Thanks to this competition, firms selling products thatcreate more value for consumers prevail against those thatcreate less. The somewhat paradoxical result: the veryacts of competition that bring us superior standards--thestrategies that save us from things like inadequate video-tapes and expensive operating systems--will look likemonopolizing acts. In some sense, after all, they are.Competitive strategies, when successful, award temporarymonopolies. Eventually, however, better products promptnew campaigns to unseat existing monopolies in marketswith increasing returns.

Many of the other claims that surround the newantitrust debate are disconnected, not only from realworld observations, but also from any real theoreticalsupport. One such claim is that Microsoft would like tocrush any would-be direct competitor. It probably would.Theory and history, however, do not tell us how predationcould ever work in a world in which assets are perfectlydurable. Further, Microsoft has been demonstrably unsuc-cessful in crushing anything except where its products arebetter than those of the opposition. They had to resortto buying uncrushed Intuit; they have barely dentedAmerica Online with the much ballyhooed Microsoft Network;and they only began to erode Netscape's near monopoly whentheir own browser came up to snuff. Microsoft's productsthat dominate in the Windows environment are the very onesthat have dominated elsewhere.53

There is, finally, the vaguely posed claim thatMicrosoft stifles innovation--another disconnect. Theclaim fails to conform to several prominent features ofthe PC landscape. First, Microsoft courts and supportsits many software developers, who now number in the hun-dreds of thousands. Second, the personal computing indus-try, by any practical standard of comparison, is astonish-ingly innovative.

Finally and most important, antitrust doctrinesbrought to bear against Microsoft cannot be construed to

Page 31

apply to Microsoft alone. If regulators can put a shortleash on the biggest operating system, then why not alsoon the biggest producer of database software, which setsthe standards for that activity, or the biggest producerof printers, or scanners, or modems, or microprocessors,and so on? If those new technologies do exhibit increas-ing returns, or important reliance on standards, or net-work effects, then we are likely to see high concentrationin many of those areas. Unless regulators embark on arelentless attack on whatever succeeds most, they mustacknowledge that firms in high-tech markets hurt competi-tors even as they benefit consumers.

Notes

1. Susan Garland, "Commentary: Justice vs. Microsoft: WhyIt Has a Case," Business Week, November 17, 1997, p. 147.

2. Alan Murray, "The Outlook: Antitrust Isn't Obsolete inan Era of High-Tech," Wall Street Journal, November 10,1997, p. A1.

3. We are not, however, aware of any formal studies thatexamine market shares and prices in high-tech markets.

4. See W. Brian Arthur, "Competing Technologies,Increasing Returns, and Lock-In by Historical Events,"Economic Journal 97 (1989): 642-65.

5. For a more complete discussion of this table, seeS. J. Liebowitz and S. E. Margolis, "Path Dependence,Lock-in, and History," Journal of Law, Economics andOrganization 11 (1995): 205-26.

6. Stochastic processes are characterized by randomincremental change on a fine scale but probabilisticallypredictable change on a gross scale. For example, a sin-gle coin flip has a random outcome whereas a series ofcoin flips will tend toward an even distribution of headsand tails.

7. See S. J. Liebowitz and S. E. Margolis, "ShouldTechnology Choice Be a Concern of Antitrust Policy?"Harvard Journal of Law and Technology 9 (1996): 283-318,for an example of a multiple equilibria model.

8. See, for example, Paul A. David, "Clio and theEconomics of QWERTY," American Economic Review 75 (1985):332-37; Joseph Farrell and Garth Saloner, "Standardization,Compatibility, and Innovation," RAND Journal of Economics

Page 32

(1985): 70-83; and Michael L. Katz and Carl C. Shapiro,"Network Externalities, Competition, and Compatibility,"American Economic Review 75 (1985): 424-40.

9. Paul R. Krugman, Peddling Prosperity: Economic Senseand Nonsense in the Age of Diminishing Expectations (NewYork: W.W. Norton, 1994).

10. Lee Gomes, "QWERTY Spells a Saga of Market Economics,"Wall Street Journal, February 25, 1998, pp. B1 and B6.

11. For a full debunking of the QWERTY myth, see S. J.Liebowitz and S. E. Margolis, "Fable of the Keys," Journalof Law and Economics 33 (1990): 1-25.

12. Gomes reports that David, author of "Clio and theEconomics of QWERTY," the paper that introduced economiststo this story, never saw the Navy study. See Gomes, p.B6.

13. Liebowitz and Margolis, "Fable of the Keys."

14. "Appraising Microsoft and Its Global Strategy,"Washington, November 13-14, 1997. RealAudio files of theconference are available at http://www.appraising-microsoft.org/day1rm.html.

15. For a fuller account of the facts summarized here, seeLiebowitz and Margolis, "Path Dependence, Lock-in, andHistory," pp. 218-22. See also Liebowitz and Margolis,"Should Technology Choice Be a Concern of AntitrustPolicy?" and S. J. Liebowitz and S. E. Margolis, "NetworkExternality: An Uncommon Tragedy," Journal of EconomicPerspectives 8 (1994): 147-49.

16. W. Brian Arthur, "Positive Feedbacks in the Economy,"Scientific American, February 1990, p. 92.

17. Pick up Newsweek in 1996 and there is Steve Wozniak,the engineering wunderkind largely responsible for Apple'searly success, explaining that Apple's recent failures werejust another example of a better product losing out to aninferior alternative. "Like the Dvorak keyboard, Apple'ssuperior operating system lost the market-share war," hesaid. Steve Wozniak, "How We Failed Apple," Newsweek,February 19, 1996, p. 48.

18. For a fuller account of the facts summarized here, seeLiebowitz and Margolis, "Should Technology Choice Be aConcern of Antitrust Policy?" pp. 316-17.

Page 33

19. The screen display on the PC required only 5 or 10percent of the computer memory that was required by theMacintosh screen.

20. The original graphical user interface, developed atthe Xerox PARC research center, had three buttons. TheMicrosoft community takes great pride in its use of but asingle button and often defends it as optimal.

21. See "Declaration of Kenneth J. Arrow," in Memorandumof the United States of America in Support of Motion toEnter Final Judgment and in Opposition to the Positions ofIDE Corporation and Amici, January 17, 1995.

22. See Complaint, U.S. v. Microsoft Corp., No. __ (U.S.Dist. Ct., D.C. filed May 18, 1998), http://www.usdoj.gov/atr/cases3/micros/1763.htm.

23. Ibid. at 35-46, ¶¶ 103-23.

24. Ibid. at 31-34, ¶¶ 93-102.

25. Ibid. at 12, ¶ 35.

26. Ibid. at 46, ¶ 127.

27. Ibid. at 47, ¶ 128. See also Memorandum of the UnitedStates in Support of Motion for Preliminary Injunction,U.S. v. Microsoft Corp., No. __ (U.S. Dist. Ct., D.C.filed May 18, 1998), http://www.usdoj.gov/atr/cases3/micros/1762.htm, p. 11: "Microsoft's conduct threatens ineconomic terms irreversibly to 'tip' the market in itsfavor"; and ibid. at 17: "[N]etwork effects give Microsofta tremendous advantage."

28. Complaint, New York, et al. v. Microsoft Corp., No. __(U.S. Dist. Ct., D.C. filed May 18, 1998), http://www.oag.state.ny.us/press/may98/microsoft. html, ¶15.

29. Ibid. at ¶ 16. See also ibid. at ¶ 26: "Substantialbarriers prevent entry and establishment in the PC operat-ing system market. These include large 'sunk' costs, net-work effects, the 'lock-in' effect, and high switchingcosts."

30. Gary L. Reback, Brian Arthur, and Garth Saloner,"Microsoft White Paper: Technological, Economic and LegalPerspectives Regarding Microsoft's Business Strategy inLight of the Proposed Acquisition of Intuit, Inc.,"February 1, 1995, http://www.upside.com/texis/mvm/story?id=34712c0e38. See also Revised Joint Brief of

Page 34

Amici Curiae on Common Issues, U.S. v. Microsoft Corp.,Nos. 95-5037, 95-5039 (U.S. Dist. Ct., D.C. filed April 6,1995), http://wsgrgate.wsgr.com/Resources/IntProp/Briefs/ms_brief.pdf (citing works of Arthur and Saloner).

31. Reback, Arthur, and Soloner.

32. See RealAudio files of Nader's conference, held inWashington, November 13-14, 1997, "Appraising Microsoft andIts Global Strategy," http://www. appraising-microsoft.org/day1rm.html.

33. That is not to say, however, that antitrust laws aswritten are ideal. If there really were serious problemswith lock-ins to inferior technology, we might want torewrite the antitrust laws. (Even then, serious questionsabout whether antitrust violates rights to property andfree exchange would remain unanswered.) For the reasonsset forth here, however, there is no evidence that lock-inpresents a serious problem.

34. Tie-in sales may allow the monopolist to capture moreof the surplus created by the monopolized good, may spreadrisk, may contribute to quality control, or may provide acheap means for monitoring intellectual property infringe-ment. Such effects of tie-ins do not require monopoliza-tion of a second market. Furthermore, where tie-ins areprofitable for any of these reasons, they may contributeto economic efficiency. See S. J. Liebowitz, "Tie-inSales, Risk Reduction and Price Discrimination," EconomicInquiry 21 (1983): 387-99. Bundling is very common forall kinds of goods. People buy season tickets, cars withtires and transmissions, and houses with microwave ovensand furnaces. Bundling can be explained by efficienciesof either production or purchase and commonly occurs inhighly competitive markets.

35. That analysis ignores natural gas, which in fact isusually cheaper than oil where it can be had.

36. While that explanation has been around in antitrusteconomics for some time, it is formalized in MichaelWhinston, "Tying, Foreclosure and Exclusion," AmericanEconomic Review 80 (1990): 837-59.

37. Furthermore, the existence of economies of scale inheating oil makes it likely that someone else has alreadymonopolized the industry, in which case extending themonopoly from furnaces to oil would cause no economicharm; it would merely change the identity of the monopo-list. The furnace monopolist is likely to benefit if it

Page 35

can avoid dealing with an oil monopolist who can share inthe furnace monopolist's profits, or lead to lower jointprofits if the two monopolists each try to take largershares of the profit.

38. That does not hold true if a different firm monopo-lizes the tied market, as discussed in the previous note.However, consumers suffer no harm at all if Microsoftreplaces the other monopolist; thus there is no reasonthat antitrust should care about that outcome.

39. The most famous case is John D. Rockefeller's StandardOil. John McGee, however, demonstrates that there werevirtually no instances of predatory pricing by Rockefeller.See John McGee, "Predatory Price Cutting: The Standard Oil(N.J.) Case," Journal of Law and Economics 1 (1958):137-69.

40. See F. M. Scherer, Industrial Market Structure andEconomic Performance, 2d ed. (Chicago: Rand McNally, 1980),p. 418: "One conclusion relevant to public policy followsimmediately. No single firm size is uniquely conducive totechnological progress. There is room for firms of allsizes." See also William Shughart, The Organization ofIndustry (Homewood, Ill.: Richard D. Irwin, 1990), p. 356:"Industrial creativity does seem to depend on somethingother than pure size."

41. "Microsoft now boasts one of the corporate world'sbiggest annual research and development budgets at $2.6billion. . . . almost a quarter of its 1996-97 sales rev-enue." Gates nonetheless recently announced plans to dou-ble Microsoft's R&D budget. Reuters, "Gates ExpectsMicrosoft's Research Budget to Double," Minneapolis StarTribune, March 18, 1998, p. 5D.

42. Note that many small start-ups have in fact gainedaccess to enormous amounts of capital in the equities mar-ket when they have been able to convince investors of thepotential value of their ideas. Those would includeNetscape, Yahoo, and many other Internet companies.

43. Julia Angwin, "Stealth Financiers; Microsoft'sVentures; Company Has $3 Billion in Startups," SanFrancisco Chronicle, November 13, 1997, p. B1.

44. Ibid.

45. Surely no one would want to argue that the softwaremarket has been insufficiently innovative in recent years.

Page 36

Nonetheless, it remains very difficult to prove that theindustry has been optimally innovative.

46. For example, word processors now contain draw pack-ages, paint packages, graphing packages, dictionaries, the-sauruses, grammar checkers, equation editors, outliners,and so forth. Spreadsheets now routinely include spellcheckers, graphics packages, statistics programs, financialprograms, programming languages, linear and nonlinear pro-gramming routines, and so forth. Even fax programs nowcontain optical character recognition software (to convertfaxes to text), draw packages to create cover pages, andso forth.

47. That was supposedly the main ingredient in a well-pub-licized deal whereby America Online agreed to includeInternet Explorer on its installation disks (although userscould use Netscape's browser if they so wished). SeeJulie Pitta, "The Cutting Edge; Showdown on the Web;Microsoft, Netscape Battle for Net Dominance," Los AngelesTimes, July 8, 1996, p. D1.

48. See Ronald Coase, "The Problem of Social Cost,"Journal of Law and Economics 3 (1960): 1-44.

49. "Transactions costs" include such things as the costsof finding parties willing to conduct business, the costsof negotiating deals, and the costs of arranging for pay-ment and delivery. "Wealth effects" refers to the factthat parties enriched by an initial entitlement to a goodmight skew final outcomes in a market because they willconsume in accord with their peculiar preferences.Whoever has the initial entitlement to a glass of water ina desert will, for example, probably end up drinking it.

50. Who is most likely to maximize the value of desktopplacement? The ability to generate value in desktopplacement depends largely on the costs of searching, mar-keting, and negotiating desktop placement with both usersand placement purchasers. In that case, it might appearthat Microsoft would be able to transact at lower costswith placement purchasers than could OEMs, arguing forgiving Microsoft the property rights. First, OEMs are notincluded in the upgrade market, and thus Microsoft alreadywould be negotiating for the desktop placements. Theadditional costs for Microsoft's controlling OEM placementswould seem trivial. Second, each OEM would likely dupli-cate the marketing, search, and negotiation costs of otherOEMs. On the other hand, OEMs often sell other softwareto customers of whom Microsoft has no knowledge. Althoughthe placement of the icons could be preordered in a par-

Page 37

ticular way, that might impose its own inefficiencies andcould conceivably tilt the result toward giving OEMs prop-erty rights. The bottom line is that at this time itremains unclear who can maximize desktop value.

51. Early in the history of the Internet, it was possibleto have a browser and not know what to do to get started.Start pages cropped up to help consumers learn to maneuvertheir way around the Web.

52. For example, both America Online and AT&T's WorldcomInternet services use a version of Internet Explorer thatis specially set up to go to America OnLine's and AT&T'shomepage, respectively. Note that control of the startpage is a Coasian problem analogous to the icon placementproblem.

53. The market share for Microsoft applications (such asWord and Excel) in the Macintosh market, for example, hashistorically been far larger than in the Windows market.

Page 38

Published by the Cato Institute, Policy Analysis is a regular series evaluating government policies and offering proposals for reform. Nothing in Policy Analysis should be construed as necessarily reflecting the views of the Cato Institute or as an attempt toaid or hinder the passage of any bill before congress. Contact the Cato Institute for reprint permission. Additional copies of PolicyAnalysis are $6.00 each ($3.00 each for five or more). To order, or for a complete listing of available studies, write the CatoInstitute, 1000 Massachusetts Ave., N.W., Washington, D.C. 20001 or call toll free 1-800-767-1241 (noon - 9 p.m. eastern time) orfax (202) 842-3490 or visit our website at www.cato.org.